diff options
Diffstat (limited to 'llvm/test/CodeGen')
124 files changed, 17045 insertions, 3634 deletions
diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/combine-freeze.mir b/llvm/test/CodeGen/AArch64/GlobalISel/combine-freeze.mir index 6b84a84..1950e60 100644 --- a/llvm/test/CodeGen/AArch64/GlobalISel/combine-freeze.mir +++ b/llvm/test/CodeGen/AArch64/GlobalISel/combine-freeze.mir @@ -1440,3 +1440,50 @@ body: | %freeze:_(<4 x s32>) = G_FREEZE %extract $q0 = COPY %freeze(<4 x s32>) RET_ReallyLR implicit $x0 +... +--- +name: ubfx_does_not_generate_poison +body: | + bb.1: + liveins: $w0 + + ; CHECK-LABEL: name: ubfx_does_not_generate_poison + ; CHECK: liveins: $w0 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $x0 + ; CHECK-NEXT: %c1:_(s64) = G_CONSTANT i64 1 + ; CHECK-NEXT: [[FREEZE:%[0-9]+]]:_(s64) = G_FREEZE [[COPY]] + ; CHECK-NEXT: [[UBFX:%[0-9]+]]:_(s64) = G_UBFX [[FREEZE]], %c1(s64), %c1 + ; CHECK-NEXT: $x0 = COPY [[UBFX]](s64) + ; CHECK-NEXT: RET_ReallyLR implicit $x0 + %0:_(s64) = COPY $x0 + %c1:_(s64) = G_CONSTANT i64 1 + %1:_(s64) = G_UBFX %0, %c1, %c1 + %2:_(s64) = G_FREEZE %1 + $x0 = COPY %2(s64) + RET_ReallyLR implicit $x0 + +... +--- +name: sbfx_does_not_generate_poison +body: | + bb.1: + liveins: $w0 + + ; CHECK-LABEL: name: sbfx_does_not_generate_poison + ; CHECK: liveins: $w0 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $x0 + ; CHECK-NEXT: %c1:_(s64) = G_CONSTANT i64 1 + ; CHECK-NEXT: [[FREEZE:%[0-9]+]]:_(s64) = G_FREEZE [[COPY]] + ; CHECK-NEXT: [[SBFX:%[0-9]+]]:_(s64) = G_SBFX [[FREEZE]], %c1(s64), %c1 + ; CHECK-NEXT: $x0 = COPY [[SBFX]](s64) + ; CHECK-NEXT: RET_ReallyLR implicit $x0 + %0:_(s64) = COPY $x0 + %c1:_(s64) = G_CONSTANT i64 1 + %1:_(s64) = G_SBFX %0, %c1, %c1 + %2:_(s64) = G_FREEZE %1 + $x0 = COPY %2(s64) + RET_ReallyLR implicit $x0 + +... diff --git a/llvm/test/CodeGen/AArch64/arm64-srl-and.ll b/llvm/test/CodeGen/AArch64/arm64-srl-and.ll index b58f6ba..330f27b 100644 --- a/llvm/test/CodeGen/AArch64/arm64-srl-and.ll +++ b/llvm/test/CodeGen/AArch64/arm64-srl-and.ll @@ -1,22 +1,38 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc -mtriple=aarch64-linux-gnu -O3 < %s | FileCheck %s +; RUN: llc -mtriple=aarch64-linux-gnu -O3 < %s | FileCheck %s --check-prefixes=CHECK,CHECK-SD +; RUN: llc -mtriple=aarch64-linux-gnu -O3 -global-isel < %s | FileCheck %s --check-prefixes=CHECK,CHECK-GI ; This used to miscompile: ; The 16-bit -1 should not become 32-bit -1 (sub w8, w8, #1). @g = global i16 0, align 4 define i32 @srl_and() { -; CHECK-LABEL: srl_and: -; CHECK: // %bb.0: // %entry -; CHECK-NEXT: adrp x8, :got:g -; CHECK-NEXT: mov w9, #50 -; CHECK-NEXT: ldr x8, [x8, :got_lo12:g] -; CHECK-NEXT: ldrh w8, [x8] -; CHECK-NEXT: eor w8, w8, w9 -; CHECK-NEXT: mov w9, #65535 -; CHECK-NEXT: add w8, w8, w9 -; CHECK-NEXT: and w0, w8, w8, lsr #16 -; CHECK-NEXT: ret +; CHECK-SD-LABEL: srl_and: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: adrp x8, :got:g +; CHECK-SD-NEXT: mov w9, #50 // =0x32 +; CHECK-SD-NEXT: ldr x8, [x8, :got_lo12:g] +; CHECK-SD-NEXT: ldrh w8, [x8] +; CHECK-SD-NEXT: eor w8, w8, w9 +; CHECK-SD-NEXT: mov w9, #65535 // =0xffff +; CHECK-SD-NEXT: add w8, w8, w9 +; CHECK-SD-NEXT: and w0, w8, w8, lsr #16 +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: srl_and: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: adrp x8, :got:g +; CHECK-GI-NEXT: mov w9, #50 // =0x32 +; CHECK-GI-NEXT: ldr x8, [x8, :got_lo12:g] +; CHECK-GI-NEXT: ldrh w8, [x8] +; CHECK-GI-NEXT: eor w8, w8, w9 +; CHECK-GI-NEXT: mov w9, #65535 // =0xffff +; CHECK-GI-NEXT: add w8, w9, w8, uxth +; CHECK-GI-NEXT: and w9, w8, #0xffff +; CHECK-GI-NEXT: cmp w8, w9 +; CHECK-GI-NEXT: cset w8, ne +; CHECK-GI-NEXT: and w0, w9, w8 +; CHECK-GI-NEXT: ret entry: %0 = load i16, ptr @g, align 4 %1 = xor i16 %0, 50 @@ -29,3 +45,5 @@ entry: ret i32 %and } +;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line: +; CHECK: {{.*}} diff --git a/llvm/test/CodeGen/AArch64/eor3.ll b/llvm/test/CodeGen/AArch64/eor3.ll index eccd091..594a73f 100644 --- a/llvm/test/CodeGen/AArch64/eor3.ll +++ b/llvm/test/CodeGen/AArch64/eor3.ll @@ -277,3 +277,154 @@ define <2 x i64> @eor3_vnot(<2 x i64> %0, <2 x i64> %1) { ret <2 x i64> %4 } +define <1 x i64> @eor3_1x64(<1 x i64> %0, <1 x i64> %1, <1 x i64> %2) { +; SHA3-LABEL: eor3_1x64: +; SHA3: // %bb.0: +; SHA3-NEXT: // kill: def $d0 killed $d0 def $q0 +; SHA3-NEXT: // kill: def $d2 killed $d2 def $q2 +; SHA3-NEXT: // kill: def $d1 killed $d1 def $q1 +; SHA3-NEXT: eor3 v0.16b, v1.16b, v2.16b, v0.16b +; SHA3-NEXT: // kill: def $d0 killed $d0 killed $q0 +; SHA3-NEXT: ret +; +; NOSHA3-LABEL: eor3_1x64: +; NOSHA3: // %bb.0: +; NOSHA3-NEXT: eor v1.8b, v1.8b, v2.8b +; NOSHA3-NEXT: eor v0.8b, v1.8b, v0.8b +; NOSHA3-NEXT: ret +; +; SVE2-LABEL: eor3_1x64: +; SVE2: // %bb.0: +; SVE2-NEXT: // kill: def $d1 killed $d1 def $z1 +; SVE2-NEXT: // kill: def $d2 killed $d2 def $z2 +; SVE2-NEXT: // kill: def $d0 killed $d0 def $z0 +; SVE2-NEXT: eor3 z1.d, z1.d, z2.d, z0.d +; SVE2-NEXT: fmov d0, d1 +; SVE2-NEXT: ret +; +; SHA3-SVE2-LABEL: eor3_1x64: +; SHA3-SVE2: // %bb.0: +; SHA3-SVE2-NEXT: // kill: def $d0 killed $d0 def $q0 +; SHA3-SVE2-NEXT: // kill: def $d2 killed $d2 def $q2 +; SHA3-SVE2-NEXT: // kill: def $d1 killed $d1 def $q1 +; SHA3-SVE2-NEXT: eor3 v0.16b, v1.16b, v2.16b, v0.16b +; SHA3-SVE2-NEXT: // kill: def $d0 killed $d0 killed $q0 +; SHA3-SVE2-NEXT: ret + %4 = xor <1 x i64> %1, %2 + %5 = xor <1 x i64> %4, %0 + ret <1 x i64> %5 +} + +define <2 x i32> @eor3_2x32(<2 x i32> %0, <2 x i32> %1, <2 x i32> %2) { +; SHA3-LABEL: eor3_2x32: +; SHA3: // %bb.0: +; SHA3-NEXT: // kill: def $d0 killed $d0 def $q0 +; SHA3-NEXT: // kill: def $d2 killed $d2 def $q2 +; SHA3-NEXT: // kill: def $d1 killed $d1 def $q1 +; SHA3-NEXT: eor3 v0.16b, v1.16b, v2.16b, v0.16b +; SHA3-NEXT: // kill: def $d0 killed $d0 killed $q0 +; SHA3-NEXT: ret +; +; NOSHA3-LABEL: eor3_2x32: +; NOSHA3: // %bb.0: +; NOSHA3-NEXT: eor v1.8b, v1.8b, v2.8b +; NOSHA3-NEXT: eor v0.8b, v1.8b, v0.8b +; NOSHA3-NEXT: ret +; +; SVE2-LABEL: eor3_2x32: +; SVE2: // %bb.0: +; SVE2-NEXT: // kill: def $d1 killed $d1 def $z1 +; SVE2-NEXT: // kill: def $d2 killed $d2 def $z2 +; SVE2-NEXT: // kill: def $d0 killed $d0 def $z0 +; SVE2-NEXT: eor3 z1.d, z1.d, z2.d, z0.d +; SVE2-NEXT: fmov d0, d1 +; SVE2-NEXT: ret +; +; SHA3-SVE2-LABEL: eor3_2x32: +; SHA3-SVE2: // %bb.0: +; SHA3-SVE2-NEXT: // kill: def $d0 killed $d0 def $q0 +; SHA3-SVE2-NEXT: // kill: def $d2 killed $d2 def $q2 +; SHA3-SVE2-NEXT: // kill: def $d1 killed $d1 def $q1 +; SHA3-SVE2-NEXT: eor3 v0.16b, v1.16b, v2.16b, v0.16b +; SHA3-SVE2-NEXT: // kill: def $d0 killed $d0 killed $q0 +; SHA3-SVE2-NEXT: ret + %4 = xor <2 x i32> %1, %2 + %5 = xor <2 x i32> %4, %0 + ret <2 x i32> %5 +} + +define <4 x i16> @eor3_4x16(<4 x i16> %0, <4 x i16> %1, <4 x i16> %2) { +; SHA3-LABEL: eor3_4x16: +; SHA3: // %bb.0: +; SHA3-NEXT: // kill: def $d0 killed $d0 def $q0 +; SHA3-NEXT: // kill: def $d2 killed $d2 def $q2 +; SHA3-NEXT: // kill: def $d1 killed $d1 def $q1 +; SHA3-NEXT: eor3 v0.16b, v1.16b, v2.16b, v0.16b +; SHA3-NEXT: // kill: def $d0 killed $d0 killed $q0 +; SHA3-NEXT: ret +; +; NOSHA3-LABEL: eor3_4x16: +; NOSHA3: // %bb.0: +; NOSHA3-NEXT: eor v1.8b, v1.8b, v2.8b +; NOSHA3-NEXT: eor v0.8b, v1.8b, v0.8b +; NOSHA3-NEXT: ret +; +; SVE2-LABEL: eor3_4x16: +; SVE2: // %bb.0: +; SVE2-NEXT: // kill: def $d1 killed $d1 def $z1 +; SVE2-NEXT: // kill: def $d2 killed $d2 def $z2 +; SVE2-NEXT: // kill: def $d0 killed $d0 def $z0 +; SVE2-NEXT: eor3 z1.d, z1.d, z2.d, z0.d +; SVE2-NEXT: fmov d0, d1 +; SVE2-NEXT: ret +; +; SHA3-SVE2-LABEL: eor3_4x16: +; SHA3-SVE2: // %bb.0: +; SHA3-SVE2-NEXT: // kill: def $d0 killed $d0 def $q0 +; SHA3-SVE2-NEXT: // kill: def $d2 killed $d2 def $q2 +; SHA3-SVE2-NEXT: // kill: def $d1 killed $d1 def $q1 +; SHA3-SVE2-NEXT: eor3 v0.16b, v1.16b, v2.16b, v0.16b +; SHA3-SVE2-NEXT: // kill: def $d0 killed $d0 killed $q0 +; SHA3-SVE2-NEXT: ret + %4 = xor <4 x i16> %1, %2 + %5 = xor <4 x i16> %4, %0 + ret <4 x i16> %5 +} + +define <8 x i8> @eor3_8x8(<8 x i8> %0, <8 x i8> %1, <8 x i8> %2) { +; SHA3-LABEL: eor3_8x8: +; SHA3: // %bb.0: +; SHA3-NEXT: // kill: def $d0 killed $d0 def $q0 +; SHA3-NEXT: // kill: def $d2 killed $d2 def $q2 +; SHA3-NEXT: // kill: def $d1 killed $d1 def $q1 +; SHA3-NEXT: eor3 v0.16b, v1.16b, v2.16b, v0.16b +; SHA3-NEXT: // kill: def $d0 killed $d0 killed $q0 +; SHA3-NEXT: ret +; +; NOSHA3-LABEL: eor3_8x8: +; NOSHA3: // %bb.0: +; NOSHA3-NEXT: eor v1.8b, v1.8b, v2.8b +; NOSHA3-NEXT: eor v0.8b, v1.8b, v0.8b +; NOSHA3-NEXT: ret +; +; SVE2-LABEL: eor3_8x8: +; SVE2: // %bb.0: +; SVE2-NEXT: // kill: def $d1 killed $d1 def $z1 +; SVE2-NEXT: // kill: def $d2 killed $d2 def $z2 +; SVE2-NEXT: // kill: def $d0 killed $d0 def $z0 +; SVE2-NEXT: eor3 z1.d, z1.d, z2.d, z0.d +; SVE2-NEXT: fmov d0, d1 +; SVE2-NEXT: ret +; +; SHA3-SVE2-LABEL: eor3_8x8: +; SHA3-SVE2: // %bb.0: +; SHA3-SVE2-NEXT: // kill: def $d0 killed $d0 def $q0 +; SHA3-SVE2-NEXT: // kill: def $d2 killed $d2 def $q2 +; SHA3-SVE2-NEXT: // kill: def $d1 killed $d1 def $q1 +; SHA3-SVE2-NEXT: eor3 v0.16b, v1.16b, v2.16b, v0.16b +; SHA3-SVE2-NEXT: // kill: def $d0 killed $d0 killed $q0 +; SHA3-SVE2-NEXT: ret + %4 = xor <8 x i8> %1, %2 + %5 = xor <8 x i8> %4, %0 + ret <8 x i8> %5 +} diff --git a/llvm/test/CodeGen/AArch64/hoist-and-by-const-from-lshr-in-eqcmp-zero.ll b/llvm/test/CodeGen/AArch64/hoist-and-by-const-from-lshr-in-eqcmp-zero.ll index c3fdc7d..8438f0b0 100644 --- a/llvm/test/CodeGen/AArch64/hoist-and-by-const-from-lshr-in-eqcmp-zero.ll +++ b/llvm/test/CodeGen/AArch64/hoist-and-by-const-from-lshr-in-eqcmp-zero.ll @@ -1,5 +1,6 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc -mtriple=aarch64-unknown-unknown < %s | FileCheck %s --check-prefix=CHECK +; RUN: llc -mtriple=aarch64-unknown-unknown < %s | FileCheck %s --check-prefixes=CHECK,CHECK-SD +; RUN: llc -mtriple=aarch64-unknown-unknown -global-isel < %s | FileCheck %s --check-prefixes=CHECK,CHECK-GI ; We are looking for the following pattern here: ; (X & (C l>> Y)) ==/!= 0 @@ -13,12 +14,21 @@ ; i8 scalar define i1 @scalar_i8_signbit_eq(i8 %x, i8 %y) nounwind { -; CHECK-LABEL: scalar_i8_signbit_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: lsl w8, w0, w1 -; CHECK-NEXT: tst w8, #0x80 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i8_signbit_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: lsl w8, w0, w1 +; CHECK-SD-NEXT: tst w8, #0x80 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i8_signbit_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #128 // =0x80 +; CHECK-GI-NEXT: and w9, w1, #0xff +; CHECK-GI-NEXT: lsr w8, w8, w9 +; CHECK-GI-NEXT: tst w8, w0 +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = lshr i8 128, %y %t1 = and i8 %t0, %x %res = icmp eq i8 %t1, 0 @@ -26,12 +36,21 @@ define i1 @scalar_i8_signbit_eq(i8 %x, i8 %y) nounwind { } define i1 @scalar_i8_lowestbit_eq(i8 %x, i8 %y) nounwind { -; CHECK-LABEL: scalar_i8_lowestbit_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: lsl w8, w0, w1 -; CHECK-NEXT: tst w8, #0x1 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i8_lowestbit_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: lsl w8, w0, w1 +; CHECK-SD-NEXT: tst w8, #0x1 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i8_lowestbit_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #1 // =0x1 +; CHECK-GI-NEXT: and w9, w1, #0xff +; CHECK-GI-NEXT: lsr w8, w8, w9 +; CHECK-GI-NEXT: tst w8, w0 +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = lshr i8 1, %y %t1 = and i8 %t0, %x %res = icmp eq i8 %t1, 0 @@ -39,12 +58,21 @@ define i1 @scalar_i8_lowestbit_eq(i8 %x, i8 %y) nounwind { } define i1 @scalar_i8_bitsinmiddle_eq(i8 %x, i8 %y) nounwind { -; CHECK-LABEL: scalar_i8_bitsinmiddle_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: lsl w8, w0, w1 -; CHECK-NEXT: tst w8, #0x18 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i8_bitsinmiddle_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: lsl w8, w0, w1 +; CHECK-SD-NEXT: tst w8, #0x18 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i8_bitsinmiddle_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #24 // =0x18 +; CHECK-GI-NEXT: and w9, w1, #0xff +; CHECK-GI-NEXT: lsr w8, w8, w9 +; CHECK-GI-NEXT: tst w8, w0 +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = lshr i8 24, %y %t1 = and i8 %t0, %x %res = icmp eq i8 %t1, 0 @@ -54,12 +82,21 @@ define i1 @scalar_i8_bitsinmiddle_eq(i8 %x, i8 %y) nounwind { ; i16 scalar define i1 @scalar_i16_signbit_eq(i16 %x, i16 %y) nounwind { -; CHECK-LABEL: scalar_i16_signbit_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: lsl w8, w0, w1 -; CHECK-NEXT: tst w8, #0x8000 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i16_signbit_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: lsl w8, w0, w1 +; CHECK-SD-NEXT: tst w8, #0x8000 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i16_signbit_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #32768 // =0x8000 +; CHECK-GI-NEXT: and w9, w1, #0xffff +; CHECK-GI-NEXT: lsr w8, w8, w9 +; CHECK-GI-NEXT: tst w8, w0 +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = lshr i16 32768, %y %t1 = and i16 %t0, %x %res = icmp eq i16 %t1, 0 @@ -67,12 +104,21 @@ define i1 @scalar_i16_signbit_eq(i16 %x, i16 %y) nounwind { } define i1 @scalar_i16_lowestbit_eq(i16 %x, i16 %y) nounwind { -; CHECK-LABEL: scalar_i16_lowestbit_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: lsl w8, w0, w1 -; CHECK-NEXT: tst w8, #0x1 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i16_lowestbit_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: lsl w8, w0, w1 +; CHECK-SD-NEXT: tst w8, #0x1 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i16_lowestbit_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #1 // =0x1 +; CHECK-GI-NEXT: and w9, w1, #0xffff +; CHECK-GI-NEXT: lsr w8, w8, w9 +; CHECK-GI-NEXT: tst w8, w0 +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = lshr i16 1, %y %t1 = and i16 %t0, %x %res = icmp eq i16 %t1, 0 @@ -80,12 +126,21 @@ define i1 @scalar_i16_lowestbit_eq(i16 %x, i16 %y) nounwind { } define i1 @scalar_i16_bitsinmiddle_eq(i16 %x, i16 %y) nounwind { -; CHECK-LABEL: scalar_i16_bitsinmiddle_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: lsl w8, w0, w1 -; CHECK-NEXT: tst w8, #0xff0 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i16_bitsinmiddle_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: lsl w8, w0, w1 +; CHECK-SD-NEXT: tst w8, #0xff0 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i16_bitsinmiddle_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #4080 // =0xff0 +; CHECK-GI-NEXT: and w9, w1, #0xffff +; CHECK-GI-NEXT: lsr w8, w8, w9 +; CHECK-GI-NEXT: tst w8, w0 +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = lshr i16 4080, %y %t1 = and i16 %t0, %x %res = icmp eq i16 %t1, 0 @@ -95,12 +150,20 @@ define i1 @scalar_i16_bitsinmiddle_eq(i16 %x, i16 %y) nounwind { ; i32 scalar define i1 @scalar_i32_signbit_eq(i32 %x, i32 %y) nounwind { -; CHECK-LABEL: scalar_i32_signbit_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: lsl w8, w0, w1 -; CHECK-NEXT: tst w8, #0x80000000 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i32_signbit_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: lsl w8, w0, w1 +; CHECK-SD-NEXT: tst w8, #0x80000000 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i32_signbit_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #-2147483648 // =0x80000000 +; CHECK-GI-NEXT: lsr w8, w8, w1 +; CHECK-GI-NEXT: tst w8, w0 +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = lshr i32 2147483648, %y %t1 = and i32 %t0, %x %res = icmp eq i32 %t1, 0 @@ -108,12 +171,20 @@ define i1 @scalar_i32_signbit_eq(i32 %x, i32 %y) nounwind { } define i1 @scalar_i32_lowestbit_eq(i32 %x, i32 %y) nounwind { -; CHECK-LABEL: scalar_i32_lowestbit_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: lsl w8, w0, w1 -; CHECK-NEXT: tst w8, #0x1 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i32_lowestbit_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: lsl w8, w0, w1 +; CHECK-SD-NEXT: tst w8, #0x1 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i32_lowestbit_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #1 // =0x1 +; CHECK-GI-NEXT: lsr w8, w8, w1 +; CHECK-GI-NEXT: tst w8, w0 +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = lshr i32 1, %y %t1 = and i32 %t0, %x %res = icmp eq i32 %t1, 0 @@ -121,12 +192,20 @@ define i1 @scalar_i32_lowestbit_eq(i32 %x, i32 %y) nounwind { } define i1 @scalar_i32_bitsinmiddle_eq(i32 %x, i32 %y) nounwind { -; CHECK-LABEL: scalar_i32_bitsinmiddle_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: lsl w8, w0, w1 -; CHECK-NEXT: tst w8, #0xffff00 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i32_bitsinmiddle_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: lsl w8, w0, w1 +; CHECK-SD-NEXT: tst w8, #0xffff00 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i32_bitsinmiddle_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #16776960 // =0xffff00 +; CHECK-GI-NEXT: lsr w8, w8, w1 +; CHECK-GI-NEXT: tst w8, w0 +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = lshr i32 16776960, %y %t1 = and i32 %t0, %x %res = icmp eq i32 %t1, 0 @@ -136,12 +215,20 @@ define i1 @scalar_i32_bitsinmiddle_eq(i32 %x, i32 %y) nounwind { ; i64 scalar define i1 @scalar_i64_signbit_eq(i64 %x, i64 %y) nounwind { -; CHECK-LABEL: scalar_i64_signbit_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: lsl x8, x0, x1 -; CHECK-NEXT: tst x8, #0x8000000000000000 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i64_signbit_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: lsl x8, x0, x1 +; CHECK-SD-NEXT: tst x8, #0x8000000000000000 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i64_signbit_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov x8, #-9223372036854775808 // =0x8000000000000000 +; CHECK-GI-NEXT: lsr x8, x8, x1 +; CHECK-GI-NEXT: tst x8, x0 +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = lshr i64 9223372036854775808, %y %t1 = and i64 %t0, %x %res = icmp eq i64 %t1, 0 @@ -149,12 +236,20 @@ define i1 @scalar_i64_signbit_eq(i64 %x, i64 %y) nounwind { } define i1 @scalar_i64_lowestbit_eq(i64 %x, i64 %y) nounwind { -; CHECK-LABEL: scalar_i64_lowestbit_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: lsl x8, x0, x1 -; CHECK-NEXT: tst x8, #0x1 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i64_lowestbit_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: lsl x8, x0, x1 +; CHECK-SD-NEXT: tst x8, #0x1 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i64_lowestbit_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #1 // =0x1 +; CHECK-GI-NEXT: lsr x8, x8, x1 +; CHECK-GI-NEXT: tst x8, x0 +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = lshr i64 1, %y %t1 = and i64 %t0, %x %res = icmp eq i64 %t1, 0 @@ -162,12 +257,20 @@ define i1 @scalar_i64_lowestbit_eq(i64 %x, i64 %y) nounwind { } define i1 @scalar_i64_bitsinmiddle_eq(i64 %x, i64 %y) nounwind { -; CHECK-LABEL: scalar_i64_bitsinmiddle_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: lsl x8, x0, x1 -; CHECK-NEXT: tst x8, #0xffffffff0000 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i64_bitsinmiddle_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: lsl x8, x0, x1 +; CHECK-SD-NEXT: tst x8, #0xffffffff0000 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i64_bitsinmiddle_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov x8, #281474976645120 // =0xffffffff0000 +; CHECK-GI-NEXT: lsr x8, x8, x1 +; CHECK-GI-NEXT: tst x8, x0 +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = lshr i64 281474976645120, %y %t1 = and i64 %t0, %x %res = icmp eq i64 %t1, 0 @@ -179,14 +282,24 @@ define i1 @scalar_i64_bitsinmiddle_eq(i64 %x, i64 %y) nounwind { ;------------------------------------------------------------------------------; define <4 x i1> @vec_4xi32_splat_eq(<4 x i32> %x, <4 x i32> %y) nounwind { -; CHECK-LABEL: vec_4xi32_splat_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: movi v2.4s, #1 -; CHECK-NEXT: ushl v0.4s, v0.4s, v1.4s -; CHECK-NEXT: and v0.16b, v0.16b, v2.16b -; CHECK-NEXT: cmeq v0.4s, v0.4s, #0 -; CHECK-NEXT: xtn v0.4h, v0.4s -; CHECK-NEXT: ret +; CHECK-SD-LABEL: vec_4xi32_splat_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: movi v2.4s, #1 +; CHECK-SD-NEXT: ushl v0.4s, v0.4s, v1.4s +; CHECK-SD-NEXT: and v0.16b, v0.16b, v2.16b +; CHECK-SD-NEXT: cmeq v0.4s, v0.4s, #0 +; CHECK-SD-NEXT: xtn v0.4h, v0.4s +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: vec_4xi32_splat_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: movi v2.4s, #1 +; CHECK-GI-NEXT: neg v1.4s, v1.4s +; CHECK-GI-NEXT: ushl v1.4s, v2.4s, v1.4s +; CHECK-GI-NEXT: and v0.16b, v1.16b, v0.16b +; CHECK-GI-NEXT: cmeq v0.4s, v0.4s, #0 +; CHECK-GI-NEXT: xtn v0.4h, v0.4s +; CHECK-GI-NEXT: ret %t0 = lshr <4 x i32> <i32 1, i32 1, i32 1, i32 1>, %y %t1 = and <4 x i32> %t0, %x %res = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 0, i32 0> @@ -211,44 +324,86 @@ define <4 x i1> @vec_4xi32_nonsplat_eq(<4 x i32> %x, <4 x i32> %y) nounwind { } define <4 x i1> @vec_4xi32_nonsplat_undef0_eq(<4 x i32> %x, <4 x i32> %y) nounwind { -; CHECK-LABEL: vec_4xi32_nonsplat_undef0_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: movi v2.4s, #1 -; CHECK-NEXT: ushl v0.4s, v0.4s, v1.4s -; CHECK-NEXT: and v0.16b, v0.16b, v2.16b -; CHECK-NEXT: cmeq v0.4s, v0.4s, #0 -; CHECK-NEXT: xtn v0.4h, v0.4s -; CHECK-NEXT: ret +; CHECK-SD-LABEL: vec_4xi32_nonsplat_undef0_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: movi v2.4s, #1 +; CHECK-SD-NEXT: ushl v0.4s, v0.4s, v1.4s +; CHECK-SD-NEXT: and v0.16b, v0.16b, v2.16b +; CHECK-SD-NEXT: cmeq v0.4s, v0.4s, #0 +; CHECK-SD-NEXT: xtn v0.4h, v0.4s +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: vec_4xi32_nonsplat_undef0_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #1 // =0x1 +; CHECK-GI-NEXT: neg v1.4s, v1.4s +; CHECK-GI-NEXT: fmov s2, w8 +; CHECK-GI-NEXT: mov v2.s[1], w8 +; CHECK-GI-NEXT: mov v2.s[3], w8 +; CHECK-GI-NEXT: ushl v1.4s, v2.4s, v1.4s +; CHECK-GI-NEXT: and v0.16b, v1.16b, v0.16b +; CHECK-GI-NEXT: cmeq v0.4s, v0.4s, #0 +; CHECK-GI-NEXT: xtn v0.4h, v0.4s +; CHECK-GI-NEXT: ret %t0 = lshr <4 x i32> <i32 1, i32 1, i32 undef, i32 1>, %y %t1 = and <4 x i32> %t0, %x %res = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 0, i32 0> ret <4 x i1> %res } define <4 x i1> @vec_4xi32_nonsplat_undef1_eq(<4 x i32> %x, <4 x i32> %y) nounwind { -; CHECK-LABEL: vec_4xi32_nonsplat_undef1_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: movi v2.4s, #1 -; CHECK-NEXT: neg v1.4s, v1.4s -; CHECK-NEXT: ushl v1.4s, v2.4s, v1.4s -; CHECK-NEXT: and v0.16b, v1.16b, v0.16b -; CHECK-NEXT: cmeq v0.4s, v0.4s, #0 -; CHECK-NEXT: xtn v0.4h, v0.4s -; CHECK-NEXT: ret +; CHECK-SD-LABEL: vec_4xi32_nonsplat_undef1_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: movi v2.4s, #1 +; CHECK-SD-NEXT: neg v1.4s, v1.4s +; CHECK-SD-NEXT: ushl v1.4s, v2.4s, v1.4s +; CHECK-SD-NEXT: and v0.16b, v1.16b, v0.16b +; CHECK-SD-NEXT: cmeq v0.4s, v0.4s, #0 +; CHECK-SD-NEXT: xtn v0.4h, v0.4s +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: vec_4xi32_nonsplat_undef1_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: movi d2, #0000000000000000 +; CHECK-GI-NEXT: movi v3.4s, #1 +; CHECK-GI-NEXT: neg v1.4s, v1.4s +; CHECK-GI-NEXT: mov v2.s[1], wzr +; CHECK-GI-NEXT: ushl v1.4s, v3.4s, v1.4s +; CHECK-GI-NEXT: and v0.16b, v1.16b, v0.16b +; CHECK-GI-NEXT: mov v2.s[3], wzr +; CHECK-GI-NEXT: cmeq v0.4s, v0.4s, v2.4s +; CHECK-GI-NEXT: xtn v0.4h, v0.4s +; CHECK-GI-NEXT: ret %t0 = lshr <4 x i32> <i32 1, i32 1, i32 1, i32 1>, %y %t1 = and <4 x i32> %t0, %x %res = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 undef, i32 0> ret <4 x i1> %res } define <4 x i1> @vec_4xi32_nonsplat_undef2_eq(<4 x i32> %x, <4 x i32> %y) nounwind { -; CHECK-LABEL: vec_4xi32_nonsplat_undef2_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: movi v2.4s, #1 -; CHECK-NEXT: neg v1.4s, v1.4s -; CHECK-NEXT: ushl v1.4s, v2.4s, v1.4s -; CHECK-NEXT: and v0.16b, v1.16b, v0.16b -; CHECK-NEXT: cmeq v0.4s, v0.4s, #0 -; CHECK-NEXT: xtn v0.4h, v0.4s -; CHECK-NEXT: ret +; CHECK-SD-LABEL: vec_4xi32_nonsplat_undef2_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: movi v2.4s, #1 +; CHECK-SD-NEXT: neg v1.4s, v1.4s +; CHECK-SD-NEXT: ushl v1.4s, v2.4s, v1.4s +; CHECK-SD-NEXT: and v0.16b, v1.16b, v0.16b +; CHECK-SD-NEXT: cmeq v0.4s, v0.4s, #0 +; CHECK-SD-NEXT: xtn v0.4h, v0.4s +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: vec_4xi32_nonsplat_undef2_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #1 // =0x1 +; CHECK-GI-NEXT: movi d2, #0000000000000000 +; CHECK-GI-NEXT: neg v1.4s, v1.4s +; CHECK-GI-NEXT: fmov s3, w8 +; CHECK-GI-NEXT: mov v3.s[1], w8 +; CHECK-GI-NEXT: mov v2.s[1], wzr +; CHECK-GI-NEXT: mov v3.s[3], w8 +; CHECK-GI-NEXT: mov v2.s[3], wzr +; CHECK-GI-NEXT: ushl v1.4s, v3.4s, v1.4s +; CHECK-GI-NEXT: and v0.16b, v1.16b, v0.16b +; CHECK-GI-NEXT: cmeq v0.4s, v0.4s, v2.4s +; CHECK-GI-NEXT: xtn v0.4h, v0.4s +; CHECK-GI-NEXT: ret %t0 = lshr <4 x i32> <i32 1, i32 1, i32 undef, i32 1>, %y %t1 = and <4 x i32> %t0, %x %res = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 undef, i32 0> @@ -260,11 +415,20 @@ define <4 x i1> @vec_4xi32_nonsplat_undef2_eq(<4 x i32> %x, <4 x i32> %y) nounwi ;------------------------------------------------------------------------------; define i1 @scalar_i8_signbit_ne(i8 %x, i8 %y) nounwind { -; CHECK-LABEL: scalar_i8_signbit_ne: -; CHECK: // %bb.0: -; CHECK-NEXT: lsl w8, w0, w1 -; CHECK-NEXT: ubfx w0, w8, #7, #1 -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i8_signbit_ne: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: lsl w8, w0, w1 +; CHECK-SD-NEXT: ubfx w0, w8, #7, #1 +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i8_signbit_ne: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #128 // =0x80 +; CHECK-GI-NEXT: and w9, w1, #0xff +; CHECK-GI-NEXT: lsr w8, w8, w9 +; CHECK-GI-NEXT: tst w8, w0 +; CHECK-GI-NEXT: cset w0, ne +; CHECK-GI-NEXT: ret %t0 = lshr i8 128, %y %t1 = and i8 %t0, %x %res = icmp ne i8 %t1, 0 ; we are perfectly happy with 'ne' predicate @@ -315,14 +479,24 @@ define i1 @scalar_i8_bitsinmiddle_slt(i8 %x, i8 %y) nounwind { } define i1 @scalar_i8_signbit_eq_with_nonzero(i8 %x, i8 %y) nounwind { -; CHECK-LABEL: scalar_i8_signbit_eq_with_nonzero: -; CHECK: // %bb.0: -; CHECK-NEXT: mov w8, #128 // =0x80 -; CHECK-NEXT: lsr w8, w8, w1 -; CHECK-NEXT: and w8, w8, w0 -; CHECK-NEXT: cmp w8, #1 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i8_signbit_eq_with_nonzero: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: mov w8, #128 // =0x80 +; CHECK-SD-NEXT: lsr w8, w8, w1 +; CHECK-SD-NEXT: and w8, w8, w0 +; CHECK-SD-NEXT: cmp w8, #1 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i8_signbit_eq_with_nonzero: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #128 // =0x80 +; CHECK-GI-NEXT: and w9, w1, #0xff +; CHECK-GI-NEXT: lsr w8, w8, w9 +; CHECK-GI-NEXT: and w8, w8, w0 +; CHECK-GI-NEXT: cmp w8, #1 +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = lshr i8 128, %y %t1 = and i8 %t0, %x %res = icmp eq i8 %t1, 1 ; should be comparing with 0 diff --git a/llvm/test/CodeGen/AArch64/hoist-and-by-const-from-shl-in-eqcmp-zero.ll b/llvm/test/CodeGen/AArch64/hoist-and-by-const-from-shl-in-eqcmp-zero.ll index 4a73b10..cc1bf27 100644 --- a/llvm/test/CodeGen/AArch64/hoist-and-by-const-from-shl-in-eqcmp-zero.ll +++ b/llvm/test/CodeGen/AArch64/hoist-and-by-const-from-shl-in-eqcmp-zero.ll @@ -1,5 +1,6 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc -mtriple=aarch64-unknown-unknown < %s | FileCheck %s --check-prefix=CHECK +; RUN: llc -mtriple=aarch64-unknown-unknown < %s | FileCheck %s --check-prefixes=CHECK,CHECK-SD +; RUN: llc -mtriple=aarch64-unknown-unknown -global-isel < %s | FileCheck %s --check-prefixes=CHECK,CHECK-GI ; We are looking for the following pattern here: ; (X & (C << Y)) ==/!= 0 @@ -13,13 +14,23 @@ ; i8 scalar define i1 @scalar_i8_signbit_eq(i8 %x, i8 %y) nounwind { -; CHECK-LABEL: scalar_i8_signbit_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: and w8, w0, #0xff -; CHECK-NEXT: lsr w8, w8, w1 -; CHECK-NEXT: tst w8, #0x80 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i8_signbit_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: and w8, w0, #0xff +; CHECK-SD-NEXT: lsr w8, w8, w1 +; CHECK-SD-NEXT: tst w8, #0x80 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i8_signbit_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #-128 // =0xffffff80 +; CHECK-GI-NEXT: and w9, w1, #0xff +; CHECK-GI-NEXT: lsl w8, w8, w9 +; CHECK-GI-NEXT: and w8, w8, w0 +; CHECK-GI-NEXT: tst w8, #0xff +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = shl i8 128, %y %t1 = and i8 %t0, %x %res = icmp eq i8 %t1, 0 @@ -27,13 +38,23 @@ define i1 @scalar_i8_signbit_eq(i8 %x, i8 %y) nounwind { } define i1 @scalar_i8_lowestbit_eq(i8 %x, i8 %y) nounwind { -; CHECK-LABEL: scalar_i8_lowestbit_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: and w8, w0, #0xff -; CHECK-NEXT: lsr w8, w8, w1 -; CHECK-NEXT: tst w8, #0x1 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i8_lowestbit_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: and w8, w0, #0xff +; CHECK-SD-NEXT: lsr w8, w8, w1 +; CHECK-SD-NEXT: tst w8, #0x1 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i8_lowestbit_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #1 // =0x1 +; CHECK-GI-NEXT: and w9, w1, #0xff +; CHECK-GI-NEXT: lsl w8, w8, w9 +; CHECK-GI-NEXT: and w8, w8, w0 +; CHECK-GI-NEXT: tst w8, #0xff +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = shl i8 1, %y %t1 = and i8 %t0, %x %res = icmp eq i8 %t1, 0 @@ -41,13 +62,23 @@ define i1 @scalar_i8_lowestbit_eq(i8 %x, i8 %y) nounwind { } define i1 @scalar_i8_bitsinmiddle_eq(i8 %x, i8 %y) nounwind { -; CHECK-LABEL: scalar_i8_bitsinmiddle_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: and w8, w0, #0xff -; CHECK-NEXT: lsr w8, w8, w1 -; CHECK-NEXT: tst w8, #0x18 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i8_bitsinmiddle_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: and w8, w0, #0xff +; CHECK-SD-NEXT: lsr w8, w8, w1 +; CHECK-SD-NEXT: tst w8, #0x18 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i8_bitsinmiddle_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #24 // =0x18 +; CHECK-GI-NEXT: and w9, w1, #0xff +; CHECK-GI-NEXT: lsl w8, w8, w9 +; CHECK-GI-NEXT: and w8, w8, w0 +; CHECK-GI-NEXT: tst w8, #0xff +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = shl i8 24, %y %t1 = and i8 %t0, %x %res = icmp eq i8 %t1, 0 @@ -57,13 +88,23 @@ define i1 @scalar_i8_bitsinmiddle_eq(i8 %x, i8 %y) nounwind { ; i16 scalar define i1 @scalar_i16_signbit_eq(i16 %x, i16 %y) nounwind { -; CHECK-LABEL: scalar_i16_signbit_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: and w8, w0, #0xffff -; CHECK-NEXT: lsr w8, w8, w1 -; CHECK-NEXT: tst w8, #0x8000 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i16_signbit_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: and w8, w0, #0xffff +; CHECK-SD-NEXT: lsr w8, w8, w1 +; CHECK-SD-NEXT: tst w8, #0x8000 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i16_signbit_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #-32768 // =0xffff8000 +; CHECK-GI-NEXT: and w9, w1, #0xffff +; CHECK-GI-NEXT: lsl w8, w8, w9 +; CHECK-GI-NEXT: and w8, w8, w0 +; CHECK-GI-NEXT: tst w8, #0xffff +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = shl i16 32768, %y %t1 = and i16 %t0, %x %res = icmp eq i16 %t1, 0 @@ -71,13 +112,23 @@ define i1 @scalar_i16_signbit_eq(i16 %x, i16 %y) nounwind { } define i1 @scalar_i16_lowestbit_eq(i16 %x, i16 %y) nounwind { -; CHECK-LABEL: scalar_i16_lowestbit_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: and w8, w0, #0xffff -; CHECK-NEXT: lsr w8, w8, w1 -; CHECK-NEXT: tst w8, #0x1 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i16_lowestbit_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: and w8, w0, #0xffff +; CHECK-SD-NEXT: lsr w8, w8, w1 +; CHECK-SD-NEXT: tst w8, #0x1 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i16_lowestbit_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #1 // =0x1 +; CHECK-GI-NEXT: and w9, w1, #0xffff +; CHECK-GI-NEXT: lsl w8, w8, w9 +; CHECK-GI-NEXT: and w8, w8, w0 +; CHECK-GI-NEXT: tst w8, #0xffff +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = shl i16 1, %y %t1 = and i16 %t0, %x %res = icmp eq i16 %t1, 0 @@ -85,13 +136,23 @@ define i1 @scalar_i16_lowestbit_eq(i16 %x, i16 %y) nounwind { } define i1 @scalar_i16_bitsinmiddle_eq(i16 %x, i16 %y) nounwind { -; CHECK-LABEL: scalar_i16_bitsinmiddle_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: and w8, w0, #0xffff -; CHECK-NEXT: lsr w8, w8, w1 -; CHECK-NEXT: tst w8, #0xff0 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i16_bitsinmiddle_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: and w8, w0, #0xffff +; CHECK-SD-NEXT: lsr w8, w8, w1 +; CHECK-SD-NEXT: tst w8, #0xff0 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i16_bitsinmiddle_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #4080 // =0xff0 +; CHECK-GI-NEXT: and w9, w1, #0xffff +; CHECK-GI-NEXT: lsl w8, w8, w9 +; CHECK-GI-NEXT: and w8, w8, w0 +; CHECK-GI-NEXT: tst w8, #0xffff +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = shl i16 4080, %y %t1 = and i16 %t0, %x %res = icmp eq i16 %t1, 0 @@ -101,12 +162,20 @@ define i1 @scalar_i16_bitsinmiddle_eq(i16 %x, i16 %y) nounwind { ; i32 scalar define i1 @scalar_i32_signbit_eq(i32 %x, i32 %y) nounwind { -; CHECK-LABEL: scalar_i32_signbit_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: lsr w8, w0, w1 -; CHECK-NEXT: tst w8, #0x80000000 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i32_signbit_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: lsr w8, w0, w1 +; CHECK-SD-NEXT: tst w8, #0x80000000 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i32_signbit_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #-2147483648 // =0x80000000 +; CHECK-GI-NEXT: lsl w8, w8, w1 +; CHECK-GI-NEXT: tst w8, w0 +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = shl i32 2147483648, %y %t1 = and i32 %t0, %x %res = icmp eq i32 %t1, 0 @@ -114,12 +183,20 @@ define i1 @scalar_i32_signbit_eq(i32 %x, i32 %y) nounwind { } define i1 @scalar_i32_lowestbit_eq(i32 %x, i32 %y) nounwind { -; CHECK-LABEL: scalar_i32_lowestbit_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: lsr w8, w0, w1 -; CHECK-NEXT: tst w8, #0x1 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i32_lowestbit_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: lsr w8, w0, w1 +; CHECK-SD-NEXT: tst w8, #0x1 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i32_lowestbit_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #1 // =0x1 +; CHECK-GI-NEXT: lsl w8, w8, w1 +; CHECK-GI-NEXT: tst w8, w0 +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = shl i32 1, %y %t1 = and i32 %t0, %x %res = icmp eq i32 %t1, 0 @@ -127,12 +204,20 @@ define i1 @scalar_i32_lowestbit_eq(i32 %x, i32 %y) nounwind { } define i1 @scalar_i32_bitsinmiddle_eq(i32 %x, i32 %y) nounwind { -; CHECK-LABEL: scalar_i32_bitsinmiddle_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: lsr w8, w0, w1 -; CHECK-NEXT: tst w8, #0xffff00 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i32_bitsinmiddle_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: lsr w8, w0, w1 +; CHECK-SD-NEXT: tst w8, #0xffff00 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i32_bitsinmiddle_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #16776960 // =0xffff00 +; CHECK-GI-NEXT: lsl w8, w8, w1 +; CHECK-GI-NEXT: tst w8, w0 +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = shl i32 16776960, %y %t1 = and i32 %t0, %x %res = icmp eq i32 %t1, 0 @@ -142,12 +227,20 @@ define i1 @scalar_i32_bitsinmiddle_eq(i32 %x, i32 %y) nounwind { ; i64 scalar define i1 @scalar_i64_signbit_eq(i64 %x, i64 %y) nounwind { -; CHECK-LABEL: scalar_i64_signbit_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: lsr x8, x0, x1 -; CHECK-NEXT: tst x8, #0x8000000000000000 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i64_signbit_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: lsr x8, x0, x1 +; CHECK-SD-NEXT: tst x8, #0x8000000000000000 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i64_signbit_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov x8, #-9223372036854775808 // =0x8000000000000000 +; CHECK-GI-NEXT: lsl x8, x8, x1 +; CHECK-GI-NEXT: tst x8, x0 +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = shl i64 9223372036854775808, %y %t1 = and i64 %t0, %x %res = icmp eq i64 %t1, 0 @@ -155,12 +248,20 @@ define i1 @scalar_i64_signbit_eq(i64 %x, i64 %y) nounwind { } define i1 @scalar_i64_lowestbit_eq(i64 %x, i64 %y) nounwind { -; CHECK-LABEL: scalar_i64_lowestbit_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: lsr x8, x0, x1 -; CHECK-NEXT: tst x8, #0x1 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i64_lowestbit_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: lsr x8, x0, x1 +; CHECK-SD-NEXT: tst x8, #0x1 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i64_lowestbit_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #1 // =0x1 +; CHECK-GI-NEXT: lsl x8, x8, x1 +; CHECK-GI-NEXT: tst x8, x0 +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = shl i64 1, %y %t1 = and i64 %t0, %x %res = icmp eq i64 %t1, 0 @@ -168,12 +269,20 @@ define i1 @scalar_i64_lowestbit_eq(i64 %x, i64 %y) nounwind { } define i1 @scalar_i64_bitsinmiddle_eq(i64 %x, i64 %y) nounwind { -; CHECK-LABEL: scalar_i64_bitsinmiddle_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: lsr x8, x0, x1 -; CHECK-NEXT: tst x8, #0xffffffff0000 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i64_bitsinmiddle_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: lsr x8, x0, x1 +; CHECK-SD-NEXT: tst x8, #0xffffffff0000 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i64_bitsinmiddle_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov x8, #281474976645120 // =0xffffffff0000 +; CHECK-GI-NEXT: lsl x8, x8, x1 +; CHECK-GI-NEXT: tst x8, x0 +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %t0 = shl i64 281474976645120, %y %t1 = and i64 %t0, %x %res = icmp eq i64 %t1, 0 @@ -216,42 +325,81 @@ define <4 x i1> @vec_4xi32_nonsplat_eq(<4 x i32> %x, <4 x i32> %y) nounwind { } define <4 x i1> @vec_4xi32_nonsplat_undef0_eq(<4 x i32> %x, <4 x i32> %y) nounwind { -; CHECK-LABEL: vec_4xi32_nonsplat_undef0_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: movi v2.4s, #1 -; CHECK-NEXT: ushl v1.4s, v2.4s, v1.4s -; CHECK-NEXT: and v0.16b, v1.16b, v0.16b -; CHECK-NEXT: cmeq v0.4s, v0.4s, #0 -; CHECK-NEXT: xtn v0.4h, v0.4s -; CHECK-NEXT: ret +; CHECK-SD-LABEL: vec_4xi32_nonsplat_undef0_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: movi v2.4s, #1 +; CHECK-SD-NEXT: ushl v1.4s, v2.4s, v1.4s +; CHECK-SD-NEXT: and v0.16b, v1.16b, v0.16b +; CHECK-SD-NEXT: cmeq v0.4s, v0.4s, #0 +; CHECK-SD-NEXT: xtn v0.4h, v0.4s +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: vec_4xi32_nonsplat_undef0_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #1 // =0x1 +; CHECK-GI-NEXT: fmov s2, w8 +; CHECK-GI-NEXT: mov v2.s[1], w8 +; CHECK-GI-NEXT: mov v2.s[3], w8 +; CHECK-GI-NEXT: ushl v1.4s, v2.4s, v1.4s +; CHECK-GI-NEXT: and v0.16b, v1.16b, v0.16b +; CHECK-GI-NEXT: cmeq v0.4s, v0.4s, #0 +; CHECK-GI-NEXT: xtn v0.4h, v0.4s +; CHECK-GI-NEXT: ret %t0 = shl <4 x i32> <i32 1, i32 1, i32 undef, i32 1>, %y %t1 = and <4 x i32> %t0, %x %res = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 0, i32 0> ret <4 x i1> %res } define <4 x i1> @vec_4xi32_nonsplat_undef1_eq(<4 x i32> %x, <4 x i32> %y) nounwind { -; CHECK-LABEL: vec_4xi32_nonsplat_undef1_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: movi v2.4s, #1 -; CHECK-NEXT: ushl v1.4s, v2.4s, v1.4s -; CHECK-NEXT: and v0.16b, v1.16b, v0.16b -; CHECK-NEXT: cmeq v0.4s, v0.4s, #0 -; CHECK-NEXT: xtn v0.4h, v0.4s -; CHECK-NEXT: ret +; CHECK-SD-LABEL: vec_4xi32_nonsplat_undef1_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: movi v2.4s, #1 +; CHECK-SD-NEXT: ushl v1.4s, v2.4s, v1.4s +; CHECK-SD-NEXT: and v0.16b, v1.16b, v0.16b +; CHECK-SD-NEXT: cmeq v0.4s, v0.4s, #0 +; CHECK-SD-NEXT: xtn v0.4h, v0.4s +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: vec_4xi32_nonsplat_undef1_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: movi d3, #0000000000000000 +; CHECK-GI-NEXT: movi v2.4s, #1 +; CHECK-GI-NEXT: mov v3.s[1], wzr +; CHECK-GI-NEXT: ushl v1.4s, v2.4s, v1.4s +; CHECK-GI-NEXT: and v0.16b, v1.16b, v0.16b +; CHECK-GI-NEXT: mov v3.s[3], wzr +; CHECK-GI-NEXT: cmeq v0.4s, v0.4s, v3.4s +; CHECK-GI-NEXT: xtn v0.4h, v0.4s +; CHECK-GI-NEXT: ret %t0 = shl <4 x i32> <i32 1, i32 1, i32 1, i32 1>, %y %t1 = and <4 x i32> %t0, %x %res = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 undef, i32 0> ret <4 x i1> %res } define <4 x i1> @vec_4xi32_nonsplat_undef2_eq(<4 x i32> %x, <4 x i32> %y) nounwind { -; CHECK-LABEL: vec_4xi32_nonsplat_undef2_eq: -; CHECK: // %bb.0: -; CHECK-NEXT: movi v2.4s, #1 -; CHECK-NEXT: ushl v1.4s, v2.4s, v1.4s -; CHECK-NEXT: and v0.16b, v1.16b, v0.16b -; CHECK-NEXT: cmeq v0.4s, v0.4s, #0 -; CHECK-NEXT: xtn v0.4h, v0.4s -; CHECK-NEXT: ret +; CHECK-SD-LABEL: vec_4xi32_nonsplat_undef2_eq: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: movi v2.4s, #1 +; CHECK-SD-NEXT: ushl v1.4s, v2.4s, v1.4s +; CHECK-SD-NEXT: and v0.16b, v1.16b, v0.16b +; CHECK-SD-NEXT: cmeq v0.4s, v0.4s, #0 +; CHECK-SD-NEXT: xtn v0.4h, v0.4s +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: vec_4xi32_nonsplat_undef2_eq: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #1 // =0x1 +; CHECK-GI-NEXT: movi d3, #0000000000000000 +; CHECK-GI-NEXT: fmov s2, w8 +; CHECK-GI-NEXT: mov v2.s[1], w8 +; CHECK-GI-NEXT: mov v3.s[1], wzr +; CHECK-GI-NEXT: mov v2.s[3], w8 +; CHECK-GI-NEXT: mov v3.s[3], wzr +; CHECK-GI-NEXT: ushl v1.4s, v2.4s, v1.4s +; CHECK-GI-NEXT: and v0.16b, v1.16b, v0.16b +; CHECK-GI-NEXT: cmeq v0.4s, v0.4s, v3.4s +; CHECK-GI-NEXT: xtn v0.4h, v0.4s +; CHECK-GI-NEXT: ret %t0 = shl <4 x i32> <i32 1, i32 1, i32 undef, i32 1>, %y %t1 = and <4 x i32> %t0, %x %res = icmp eq <4 x i32> %t1, <i32 0, i32 0, i32 undef, i32 0> @@ -263,12 +411,22 @@ define <4 x i1> @vec_4xi32_nonsplat_undef2_eq(<4 x i32> %x, <4 x i32> %y) nounwi ;------------------------------------------------------------------------------; define i1 @scalar_i8_signbit_ne(i8 %x, i8 %y) nounwind { -; CHECK-LABEL: scalar_i8_signbit_ne: -; CHECK: // %bb.0: -; CHECK-NEXT: and w8, w0, #0xff -; CHECK-NEXT: lsr w8, w8, w1 -; CHECK-NEXT: lsr w0, w8, #7 -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i8_signbit_ne: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: and w8, w0, #0xff +; CHECK-SD-NEXT: lsr w8, w8, w1 +; CHECK-SD-NEXT: lsr w0, w8, #7 +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i8_signbit_ne: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #-128 // =0xffffff80 +; CHECK-GI-NEXT: and w9, w1, #0xff +; CHECK-GI-NEXT: lsl w8, w8, w9 +; CHECK-GI-NEXT: and w8, w8, w0 +; CHECK-GI-NEXT: tst w8, #0xff +; CHECK-GI-NEXT: cset w0, ne +; CHECK-GI-NEXT: ret %t0 = shl i8 128, %y %t1 = and i8 %t0, %x %res = icmp ne i8 %t1, 0 ; we are perfectly happy with 'ne' predicate @@ -310,13 +468,24 @@ define i1 @scalar_i32_x_is_const2_eq(i32 %y) nounwind { } define i1 @scalar_i8_bitsinmiddle_slt(i8 %x, i8 %y) nounwind { -; CHECK-LABEL: scalar_i8_bitsinmiddle_slt: -; CHECK: // %bb.0: -; CHECK-NEXT: mov w8, #24 // =0x18 -; CHECK-NEXT: lsl w8, w8, w1 -; CHECK-NEXT: and w8, w8, w0 -; CHECK-NEXT: ubfx w0, w8, #7, #1 -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i8_bitsinmiddle_slt: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: mov w8, #24 // =0x18 +; CHECK-SD-NEXT: lsl w8, w8, w1 +; CHECK-SD-NEXT: and w8, w8, w0 +; CHECK-SD-NEXT: ubfx w0, w8, #7, #1 +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i8_bitsinmiddle_slt: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #24 // =0x18 +; CHECK-GI-NEXT: and w9, w1, #0xff +; CHECK-GI-NEXT: lsl w8, w8, w9 +; CHECK-GI-NEXT: and w8, w8, w0 +; CHECK-GI-NEXT: sxtb w8, w8 +; CHECK-GI-NEXT: cmp w8, #0 +; CHECK-GI-NEXT: cset w0, mi +; CHECK-GI-NEXT: ret %t0 = shl i8 24, %y %t1 = and i8 %t0, %x %res = icmp slt i8 %t1, 0 @@ -324,15 +493,20 @@ define i1 @scalar_i8_bitsinmiddle_slt(i8 %x, i8 %y) nounwind { } define i1 @scalar_i8_signbit_eq_with_nonzero(i8 %x, i8 %y) nounwind { -; CHECK-LABEL: scalar_i8_signbit_eq_with_nonzero: -; CHECK: // %bb.0: -; CHECK-NEXT: mov w8, #-128 // =0xffffff80 -; CHECK-NEXT: lsl w8, w8, w1 -; CHECK-NEXT: and w8, w8, w0 -; CHECK-NEXT: and w8, w8, #0x80 -; CHECK-NEXT: cmp w8, #1 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: scalar_i8_signbit_eq_with_nonzero: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: mov w8, #-128 // =0xffffff80 +; CHECK-SD-NEXT: lsl w8, w8, w1 +; CHECK-SD-NEXT: and w8, w8, w0 +; CHECK-SD-NEXT: and w8, w8, #0x80 +; CHECK-SD-NEXT: cmp w8, #1 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: scalar_i8_signbit_eq_with_nonzero: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w0, wzr +; CHECK-GI-NEXT: ret %t0 = shl i8 128, %y %t1 = and i8 %t0, %x %res = icmp eq i8 %t1, 1 ; should be comparing with 0 diff --git a/llvm/test/CodeGen/AArch64/llvm.sincos.ll b/llvm/test/CodeGen/AArch64/llvm.sincos.ll index f1dcb2a..21da864 100644 --- a/llvm/test/CodeGen/AArch64/llvm.sincos.ll +++ b/llvm/test/CodeGen/AArch64/llvm.sincos.ll @@ -215,6 +215,133 @@ define { <2 x half>, <2 x half> } @test_sincos_v2f16(<2 x half> %a) nounwind { ret { <2 x half>, <2 x half> } %result } +define { <3 x half>, <3 x half> } @test_sincos_v3f16(<3 x half> %a) nounwind { +; CHECK-LABEL: test_sincos_v3f16: +; CHECK: // %bb.0: +; CHECK-NEXT: sub sp, sp, #64 +; CHECK-NEXT: // kill: def $d0 killed $d0 def $q0 +; CHECK-NEXT: mov h1, v0.h[1] +; CHECK-NEXT: str q0, [sp] // 16-byte Folded Spill +; CHECK-NEXT: add x0, sp, #36 +; CHECK-NEXT: add x1, sp, #32 +; CHECK-NEXT: str x30, [sp, #48] // 8-byte Folded Spill +; CHECK-NEXT: fcvt s0, h1 +; CHECK-NEXT: bl sincosf +; CHECK-NEXT: ldr q0, [sp] // 16-byte Folded Reload +; CHECK-NEXT: add x0, sp, #28 +; CHECK-NEXT: add x1, sp, #24 +; CHECK-NEXT: fcvt s0, h0 +; CHECK-NEXT: bl sincosf +; CHECK-NEXT: ldr q0, [sp] // 16-byte Folded Reload +; CHECK-NEXT: add x0, sp, #44 +; CHECK-NEXT: add x1, sp, #40 +; CHECK-NEXT: mov h0, v0.h[2] +; CHECK-NEXT: fcvt s0, h0 +; CHECK-NEXT: bl sincosf +; CHECK-NEXT: ldr q0, [sp] // 16-byte Folded Reload +; CHECK-NEXT: add x0, sp, #60 +; CHECK-NEXT: add x1, sp, #56 +; CHECK-NEXT: mov h0, v0.h[3] +; CHECK-NEXT: fcvt s0, h0 +; CHECK-NEXT: bl sincosf +; CHECK-NEXT: ldp s2, s0, [sp, #32] +; CHECK-NEXT: ldr x30, [sp, #48] // 8-byte Folded Reload +; CHECK-NEXT: ldp s3, s1, [sp, #24] +; CHECK-NEXT: fcvt h4, s0 +; CHECK-NEXT: fcvt h2, s2 +; CHECK-NEXT: fcvt h0, s1 +; CHECK-NEXT: fcvt h1, s3 +; CHECK-NEXT: ldp s5, s3, [sp, #40] +; CHECK-NEXT: fcvt h3, s3 +; CHECK-NEXT: mov v0.h[1], v4.h[0] +; CHECK-NEXT: fcvt h4, s5 +; CHECK-NEXT: mov v1.h[1], v2.h[0] +; CHECK-NEXT: ldp s5, s2, [sp, #56] +; CHECK-NEXT: mov v0.h[2], v3.h[0] +; CHECK-NEXT: fcvt h2, s2 +; CHECK-NEXT: fcvt h3, s5 +; CHECK-NEXT: mov v1.h[2], v4.h[0] +; CHECK-NEXT: mov v0.h[3], v2.h[0] +; CHECK-NEXT: mov v1.h[3], v3.h[0] +; CHECK-NEXT: // kill: def $d0 killed $d0 killed $q0 +; CHECK-NEXT: // kill: def $d1 killed $d1 killed $q1 +; CHECK-NEXT: add sp, sp, #64 +; CHECK-NEXT: ret +; +; NO-LIBCALL-LABEL: test_sincos_v3f16: +; NO-LIBCALL: // %bb.0: +; NO-LIBCALL-NEXT: sub sp, sp, #80 +; NO-LIBCALL-NEXT: // kill: def $d0 killed $d0 def $q0 +; NO-LIBCALL-NEXT: mov h1, v0.h[1] +; NO-LIBCALL-NEXT: stp d9, d8, [sp, #48] // 16-byte Folded Spill +; NO-LIBCALL-NEXT: str q0, [sp] // 16-byte Folded Spill +; NO-LIBCALL-NEXT: stp d11, d10, [sp, #32] // 16-byte Folded Spill +; NO-LIBCALL-NEXT: str x30, [sp, #64] // 8-byte Folded Spill +; NO-LIBCALL-NEXT: fcvt s8, h1 +; NO-LIBCALL-NEXT: fmov s0, s8 +; NO-LIBCALL-NEXT: bl sinf +; NO-LIBCALL-NEXT: ldr q1, [sp] // 16-byte Folded Reload +; NO-LIBCALL-NEXT: fcvt h0, s0 +; NO-LIBCALL-NEXT: fcvt s9, h1 +; NO-LIBCALL-NEXT: str q0, [sp, #16] // 16-byte Folded Spill +; NO-LIBCALL-NEXT: fmov s0, s9 +; NO-LIBCALL-NEXT: bl sinf +; NO-LIBCALL-NEXT: ldr q1, [sp] // 16-byte Folded Reload +; NO-LIBCALL-NEXT: fcvt h0, s0 +; NO-LIBCALL-NEXT: mov h1, v1.h[2] +; NO-LIBCALL-NEXT: fcvt s10, h1 +; NO-LIBCALL-NEXT: ldr q1, [sp, #16] // 16-byte Folded Reload +; NO-LIBCALL-NEXT: mov v0.h[1], v1.h[0] +; NO-LIBCALL-NEXT: str q0, [sp, #16] // 16-byte Folded Spill +; NO-LIBCALL-NEXT: fmov s0, s10 +; NO-LIBCALL-NEXT: bl sinf +; NO-LIBCALL-NEXT: ldr q1, [sp] // 16-byte Folded Reload +; NO-LIBCALL-NEXT: fcvt h0, s0 +; NO-LIBCALL-NEXT: mov h1, v1.h[3] +; NO-LIBCALL-NEXT: fcvt s11, h1 +; NO-LIBCALL-NEXT: ldr q1, [sp, #16] // 16-byte Folded Reload +; NO-LIBCALL-NEXT: mov v1.h[2], v0.h[0] +; NO-LIBCALL-NEXT: fmov s0, s11 +; NO-LIBCALL-NEXT: str q1, [sp, #16] // 16-byte Folded Spill +; NO-LIBCALL-NEXT: bl sinf +; NO-LIBCALL-NEXT: fcvt h0, s0 +; NO-LIBCALL-NEXT: ldr q1, [sp, #16] // 16-byte Folded Reload +; NO-LIBCALL-NEXT: mov v1.h[3], v0.h[0] +; NO-LIBCALL-NEXT: fmov s0, s8 +; NO-LIBCALL-NEXT: str q1, [sp, #16] // 16-byte Folded Spill +; NO-LIBCALL-NEXT: bl cosf +; NO-LIBCALL-NEXT: fcvt h0, s0 +; NO-LIBCALL-NEXT: str q0, [sp] // 16-byte Folded Spill +; NO-LIBCALL-NEXT: fmov s0, s9 +; NO-LIBCALL-NEXT: bl cosf +; NO-LIBCALL-NEXT: fcvt h0, s0 +; NO-LIBCALL-NEXT: ldr q1, [sp] // 16-byte Folded Reload +; NO-LIBCALL-NEXT: mov v0.h[1], v1.h[0] +; NO-LIBCALL-NEXT: str q0, [sp] // 16-byte Folded Spill +; NO-LIBCALL-NEXT: fmov s0, s10 +; NO-LIBCALL-NEXT: bl cosf +; NO-LIBCALL-NEXT: fcvt h0, s0 +; NO-LIBCALL-NEXT: ldr q1, [sp] // 16-byte Folded Reload +; NO-LIBCALL-NEXT: mov v1.h[2], v0.h[0] +; NO-LIBCALL-NEXT: fmov s0, s11 +; NO-LIBCALL-NEXT: str q1, [sp] // 16-byte Folded Spill +; NO-LIBCALL-NEXT: bl cosf +; NO-LIBCALL-NEXT: fmov s1, s0 +; NO-LIBCALL-NEXT: ldp d9, d8, [sp, #48] // 16-byte Folded Reload +; NO-LIBCALL-NEXT: ldp d11, d10, [sp, #32] // 16-byte Folded Reload +; NO-LIBCALL-NEXT: ldr q0, [sp, #16] // 16-byte Folded Reload +; NO-LIBCALL-NEXT: ldr x30, [sp, #64] // 8-byte Folded Reload +; NO-LIBCALL-NEXT: // kill: def $d0 killed $d0 killed $q0 +; NO-LIBCALL-NEXT: fcvt h2, s1 +; NO-LIBCALL-NEXT: ldr q1, [sp] // 16-byte Folded Reload +; NO-LIBCALL-NEXT: mov v1.h[3], v2.h[0] +; NO-LIBCALL-NEXT: // kill: def $d1 killed $d1 killed $q1 +; NO-LIBCALL-NEXT: add sp, sp, #80 +; NO-LIBCALL-NEXT: ret + %result = call { <3 x half>, <3 x half> } @llvm.sincos.v3f16(<3 x half> %a) + ret { <3 x half>, <3 x half> } %result +} + define { float, float } @test_sincos_f32(float %a) nounwind { ; CHECK-LABEL: test_sincos_f32: ; CHECK: // %bb.0: @@ -493,3 +620,71 @@ define { <2 x double>, <2 x double> } @test_sincos_v2f64(<2 x double> %a) nounwi %result = call { <2 x double>, <2 x double> } @llvm.sincos.v2f64(<2 x double> %a) ret { <2 x double>, <2 x double> } %result } + +define { <3 x double>, <3 x double> } @test_sincos_v3f64(<3 x double> %a) nounwind { +; CHECK-LABEL: test_sincos_v3f64: +; CHECK: // %bb.0: +; CHECK-NEXT: sub sp, sp, #80 +; CHECK-NEXT: add x0, sp, #16 +; CHECK-NEXT: add x1, sp, #8 +; CHECK-NEXT: stp d9, d8, [sp, #48] // 16-byte Folded Spill +; CHECK-NEXT: str x30, [sp, #64] // 8-byte Folded Spill +; CHECK-NEXT: fmov d8, d2 +; CHECK-NEXT: fmov d9, d1 +; CHECK-NEXT: bl sincos +; CHECK-NEXT: fmov d0, d9 +; CHECK-NEXT: add x0, sp, #32 +; CHECK-NEXT: add x1, sp, #24 +; CHECK-NEXT: bl sincos +; CHECK-NEXT: fmov d0, d8 +; CHECK-NEXT: add x0, sp, #72 +; CHECK-NEXT: add x1, sp, #40 +; CHECK-NEXT: bl sincos +; CHECK-NEXT: ldp d3, d0, [sp, #8] +; CHECK-NEXT: ldr d2, [sp, #72] +; CHECK-NEXT: ldp d4, d1, [sp, #24] +; CHECK-NEXT: ldr d5, [sp, #40] +; CHECK-NEXT: ldr x30, [sp, #64] // 8-byte Folded Reload +; CHECK-NEXT: ldp d9, d8, [sp, #48] // 16-byte Folded Reload +; CHECK-NEXT: add sp, sp, #80 +; CHECK-NEXT: ret +; +; NO-LIBCALL-LABEL: test_sincos_v3f64: +; NO-LIBCALL: // %bb.0: +; NO-LIBCALL-NEXT: stp d13, d12, [sp, #-64]! // 16-byte Folded Spill +; NO-LIBCALL-NEXT: stp d11, d10, [sp, #16] // 16-byte Folded Spill +; NO-LIBCALL-NEXT: fmov d10, d0 +; NO-LIBCALL-NEXT: stp d9, d8, [sp, #32] // 16-byte Folded Spill +; NO-LIBCALL-NEXT: fmov d8, d2 +; NO-LIBCALL-NEXT: fmov d9, d1 +; NO-LIBCALL-NEXT: str x30, [sp, #48] // 8-byte Folded Spill +; NO-LIBCALL-NEXT: bl sin +; NO-LIBCALL-NEXT: fmov d11, d0 +; NO-LIBCALL-NEXT: fmov d0, d9 +; NO-LIBCALL-NEXT: bl sin +; NO-LIBCALL-NEXT: fmov d12, d0 +; NO-LIBCALL-NEXT: fmov d0, d8 +; NO-LIBCALL-NEXT: bl sin +; NO-LIBCALL-NEXT: fmov d13, d0 +; NO-LIBCALL-NEXT: fmov d0, d10 +; NO-LIBCALL-NEXT: bl cos +; NO-LIBCALL-NEXT: fmov d10, d0 +; NO-LIBCALL-NEXT: fmov d0, d9 +; NO-LIBCALL-NEXT: bl cos +; NO-LIBCALL-NEXT: fmov d9, d0 +; NO-LIBCALL-NEXT: fmov d0, d8 +; NO-LIBCALL-NEXT: bl cos +; NO-LIBCALL-NEXT: fmov d5, d0 +; NO-LIBCALL-NEXT: fmov d0, d11 +; NO-LIBCALL-NEXT: ldr x30, [sp, #48] // 8-byte Folded Reload +; NO-LIBCALL-NEXT: fmov d3, d10 +; NO-LIBCALL-NEXT: fmov d4, d9 +; NO-LIBCALL-NEXT: fmov d1, d12 +; NO-LIBCALL-NEXT: ldp d9, d8, [sp, #32] // 16-byte Folded Reload +; NO-LIBCALL-NEXT: fmov d2, d13 +; NO-LIBCALL-NEXT: ldp d11, d10, [sp, #16] // 16-byte Folded Reload +; NO-LIBCALL-NEXT: ldp d13, d12, [sp], #64 // 16-byte Folded Reload +; NO-LIBCALL-NEXT: ret + %result = call { <3 x double>, <3 x double> } @llvm.sincos.v3f64(<3 x double> %a) + ret { <3 x double>, <3 x double> } %result +} diff --git a/llvm/test/CodeGen/AArch64/signbit-test.ll b/llvm/test/CodeGen/AArch64/signbit-test.ll index c74a934..298495b 100644 --- a/llvm/test/CodeGen/AArch64/signbit-test.ll +++ b/llvm/test/CodeGen/AArch64/signbit-test.ll @@ -1,13 +1,21 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc < %s -mtriple=aarch64-- | FileCheck %s +; RUN: llc -mtriple=aarch64-- < %s | FileCheck %s --check-prefixes=CHECK,CHECK-SD +; RUN: llc -mtriple=aarch64-- -global-isel < %s | FileCheck %s --check-prefixes=CHECK,CHECK-GI define i64 @test_clear_mask_i64_i32(i64 %x) nounwind { -; CHECK-LABEL: test_clear_mask_i64_i32: -; CHECK: // %bb.0: // %entry -; CHECK-NEXT: mov w8, #42 // =0x2a -; CHECK-NEXT: cmn w0, #1 -; CHECK-NEXT: csel x0, x8, x0, gt -; CHECK-NEXT: ret +; CHECK-SD-LABEL: test_clear_mask_i64_i32: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: mov w8, #42 // =0x2a +; CHECK-SD-NEXT: cmn w0, #1 +; CHECK-SD-NEXT: csel x0, x8, x0, gt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: test_clear_mask_i64_i32: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: mov w8, #42 // =0x2a +; CHECK-GI-NEXT: tst x0, #0x80000000 +; CHECK-GI-NEXT: csel x0, x8, x0, eq +; CHECK-GI-NEXT: ret entry: %a = and i64 %x, 2147483648 %r = icmp eq i64 %a, 0 diff --git a/llvm/test/CodeGen/AArch64/signed-truncation-check.ll b/llvm/test/CodeGen/AArch64/signed-truncation-check.ll index 7c80f93..fc01c6b 100644 --- a/llvm/test/CodeGen/AArch64/signed-truncation-check.ll +++ b/llvm/test/CodeGen/AArch64/signed-truncation-check.ll @@ -1,5 +1,6 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc -mtriple=aarch64-unknown-linux-gnu < %s | FileCheck %s +; RUN: llc -mtriple=aarch64-unknown-linux-gnu < %s | FileCheck %s --check-prefixes=CHECK,CHECK-SD +; RUN: llc -mtriple=aarch64-unknown-linux-gnu -global-isel < %s | FileCheck %s --check-prefixes=CHECK,CHECK-GI ; https://bugs.llvm.org/show_bug.cgi?id=38149 @@ -19,13 +20,22 @@ ; ---------------------------------------------------------------------------- ; define i1 @shifts_eqcmp_i16_i8(i16 %x) nounwind { -; CHECK-LABEL: shifts_eqcmp_i16_i8: -; CHECK: // %bb.0: -; CHECK-NEXT: sxtb w8, w0 -; CHECK-NEXT: and w8, w8, #0xffff -; CHECK-NEXT: cmp w8, w0, uxth -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: shifts_eqcmp_i16_i8: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: sxtb w8, w0 +; CHECK-SD-NEXT: and w8, w8, #0xffff +; CHECK-SD-NEXT: cmp w8, w0, uxth +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: shifts_eqcmp_i16_i8: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: lsl w8, w0, #8 +; CHECK-GI-NEXT: sbfx w8, w8, #8, #8 +; CHECK-GI-NEXT: and w8, w8, #0xffff +; CHECK-GI-NEXT: cmp w8, w0, uxth +; CHECK-GI-NEXT: cset w0, eq +; CHECK-GI-NEXT: ret %tmp0 = shl i16 %x, 8 ; 16-8 %tmp1 = ashr exact i16 %tmp0, 8 ; 16-8 %tmp2 = icmp eq i16 %tmp1, %x @@ -97,26 +107,43 @@ define i1 @shifts_eqcmp_i64_i8(i64 %x) nounwind { ; ---------------------------------------------------------------------------- ; define i1 @add_ugecmp_i16_i8(i16 %x) nounwind { -; CHECK-LABEL: add_ugecmp_i16_i8: -; CHECK: // %bb.0: -; CHECK-NEXT: and w8, w0, #0xffff -; CHECK-NEXT: sub w8, w8, #128 -; CHECK-NEXT: lsr w8, w8, #8 -; CHECK-NEXT: cmp w8, #254 -; CHECK-NEXT: cset w0, hi -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ugecmp_i16_i8: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: and w8, w0, #0xffff +; CHECK-SD-NEXT: sub w8, w8, #128 +; CHECK-SD-NEXT: lsr w8, w8, #8 +; CHECK-SD-NEXT: cmp w8, #254 +; CHECK-SD-NEXT: cset w0, hi +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ugecmp_i16_i8: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #-128 // =0xffffff80 +; CHECK-GI-NEXT: mov w9, #65280 // =0xff00 +; CHECK-GI-NEXT: add w8, w8, w0, uxth +; CHECK-GI-NEXT: cmp w8, w9 +; CHECK-GI-NEXT: cset w0, hs +; CHECK-GI-NEXT: ret %tmp0 = add i16 %x, -128 ; ~0U << (8-1) %tmp1 = icmp uge i16 %tmp0, -256 ; ~0U << 8 ret i1 %tmp1 } define i1 @add_ugecmp_i32_i16_i8(i16 %xx) nounwind { -; CHECK-LABEL: add_ugecmp_i32_i16_i8: -; CHECK: // %bb.0: -; CHECK-NEXT: and w8, w0, #0xffff -; CHECK-NEXT: cmp w8, w8, sxtb -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ugecmp_i32_i16_i8: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: and w8, w0, #0xffff +; CHECK-SD-NEXT: cmp w8, w8, sxtb +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ugecmp_i32_i16_i8: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #-128 // =0xffffff80 +; CHECK-GI-NEXT: add w8, w8, w0, uxth +; CHECK-GI-NEXT: cmn w8, #256 +; CHECK-GI-NEXT: cset w0, hs +; CHECK-GI-NEXT: ret %x = zext i16 %xx to i32 %tmp0 = add i32 %x, -128 ; ~0U << (8-1) %tmp1 = icmp uge i32 %tmp0, -256 ; ~0U << 8 @@ -124,55 +151,92 @@ define i1 @add_ugecmp_i32_i16_i8(i16 %xx) nounwind { } define i1 @add_ugecmp_i32_i16(i32 %x) nounwind { -; CHECK-LABEL: add_ugecmp_i32_i16: -; CHECK: // %bb.0: -; CHECK-NEXT: cmp w0, w0, sxth -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ugecmp_i32_i16: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: cmp w0, w0, sxth +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ugecmp_i32_i16: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: sub w8, w0, #8, lsl #12 // =32768 +; CHECK-GI-NEXT: cmn w8, #16, lsl #12 // =65536 +; CHECK-GI-NEXT: cset w0, hs +; CHECK-GI-NEXT: ret %tmp0 = add i32 %x, -32768 ; ~0U << (16-1) %tmp1 = icmp uge i32 %tmp0, -65536 ; ~0U << 16 ret i1 %tmp1 } define i1 @add_ugecmp_i32_i8(i32 %x) nounwind { -; CHECK-LABEL: add_ugecmp_i32_i8: -; CHECK: // %bb.0: -; CHECK-NEXT: cmp w0, w0, sxtb -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ugecmp_i32_i8: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: cmp w0, w0, sxtb +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ugecmp_i32_i8: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: sub w8, w0, #128 +; CHECK-GI-NEXT: cmn w8, #256 +; CHECK-GI-NEXT: cset w0, hs +; CHECK-GI-NEXT: ret %tmp0 = add i32 %x, -128 ; ~0U << (8-1) %tmp1 = icmp uge i32 %tmp0, -256 ; ~0U << 8 ret i1 %tmp1 } define i1 @add_ugecmp_i64_i32(i64 %x) nounwind { -; CHECK-LABEL: add_ugecmp_i64_i32: -; CHECK: // %bb.0: -; CHECK-NEXT: cmp x0, w0, sxtw -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ugecmp_i64_i32: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: cmp x0, w0, sxtw +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ugecmp_i64_i32: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov x8, #-2147483648 // =0xffffffff80000000 +; CHECK-GI-NEXT: mov x9, #-4294967296 // =0xffffffff00000000 +; CHECK-GI-NEXT: add x8, x0, x8 +; CHECK-GI-NEXT: cmp x8, x9 +; CHECK-GI-NEXT: cset w0, hs +; CHECK-GI-NEXT: ret %tmp0 = add i64 %x, -2147483648 ; ~0U << (32-1) %tmp1 = icmp uge i64 %tmp0, -4294967296 ; ~0U << 32 ret i1 %tmp1 } define i1 @add_ugecmp_i64_i16(i64 %x) nounwind { -; CHECK-LABEL: add_ugecmp_i64_i16: -; CHECK: // %bb.0: -; CHECK-NEXT: cmp x0, w0, sxth -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ugecmp_i64_i16: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: cmp x0, w0, sxth +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ugecmp_i64_i16: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: sub x8, x0, #8, lsl #12 // =32768 +; CHECK-GI-NEXT: cmn x8, #16, lsl #12 // =65536 +; CHECK-GI-NEXT: cset w0, hs +; CHECK-GI-NEXT: ret %tmp0 = add i64 %x, -32768 ; ~0U << (16-1) %tmp1 = icmp uge i64 %tmp0, -65536 ; ~0U << 16 ret i1 %tmp1 } define i1 @add_ugecmp_i64_i8(i64 %x) nounwind { -; CHECK-LABEL: add_ugecmp_i64_i8: -; CHECK: // %bb.0: -; CHECK-NEXT: cmp x0, w0, sxtb -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ugecmp_i64_i8: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: cmp x0, w0, sxtb +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ugecmp_i64_i8: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: sub x8, x0, #128 +; CHECK-GI-NEXT: cmn x8, #256 +; CHECK-GI-NEXT: cset w0, hs +; CHECK-GI-NEXT: ret %tmp0 = add i64 %x, -128 ; ~0U << (8-1) %tmp1 = icmp uge i64 %tmp0, -256 ; ~0U << 8 ret i1 %tmp1 @@ -180,14 +244,23 @@ define i1 @add_ugecmp_i64_i8(i64 %x) nounwind { ; Slightly more canonical variant define i1 @add_ugtcmp_i16_i8(i16 %x) nounwind { -; CHECK-LABEL: add_ugtcmp_i16_i8: -; CHECK: // %bb.0: -; CHECK-NEXT: and w8, w0, #0xffff -; CHECK-NEXT: sub w8, w8, #128 -; CHECK-NEXT: lsr w8, w8, #8 -; CHECK-NEXT: cmp w8, #254 -; CHECK-NEXT: cset w0, hi -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ugtcmp_i16_i8: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: and w8, w0, #0xffff +; CHECK-SD-NEXT: sub w8, w8, #128 +; CHECK-SD-NEXT: lsr w8, w8, #8 +; CHECK-SD-NEXT: cmp w8, #254 +; CHECK-SD-NEXT: cset w0, hi +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ugtcmp_i16_i8: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #-128 // =0xffffff80 +; CHECK-GI-NEXT: mov w9, #65279 // =0xfeff +; CHECK-GI-NEXT: add w8, w8, w0, uxth +; CHECK-GI-NEXT: cmp w8, w9 +; CHECK-GI-NEXT: cset w0, hi +; CHECK-GI-NEXT: ret %tmp0 = add i16 %x, -128 ; ~0U << (8-1) %tmp1 = icmp ugt i16 %tmp0, -257 ; ~0U << 8 - 1 ret i1 %tmp1 @@ -198,68 +271,113 @@ define i1 @add_ugtcmp_i16_i8(i16 %x) nounwind { ; ---------------------------------------------------------------------------- ; define i1 @add_ultcmp_i16_i8(i16 %x) nounwind { -; CHECK-LABEL: add_ultcmp_i16_i8: -; CHECK: // %bb.0: -; CHECK-NEXT: sxtb w8, w0 -; CHECK-NEXT: and w8, w8, #0xffff -; CHECK-NEXT: cmp w8, w0, uxth -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ultcmp_i16_i8: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: sxtb w8, w0 +; CHECK-SD-NEXT: and w8, w8, #0xffff +; CHECK-SD-NEXT: cmp w8, w0, uxth +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ultcmp_i16_i8: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: add w8, w0, #128 +; CHECK-GI-NEXT: and w8, w8, #0xffff +; CHECK-GI-NEXT: cmp w8, #256 +; CHECK-GI-NEXT: cset w0, lo +; CHECK-GI-NEXT: ret %tmp0 = add i16 %x, 128 ; 1U << (8-1) %tmp1 = icmp ult i16 %tmp0, 256 ; 1U << 8 ret i1 %tmp1 } define i1 @add_ultcmp_i32_i16(i32 %x) nounwind { -; CHECK-LABEL: add_ultcmp_i32_i16: -; CHECK: // %bb.0: -; CHECK-NEXT: cmp w0, w0, sxth -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ultcmp_i32_i16: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: cmp w0, w0, sxth +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ultcmp_i32_i16: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: add w8, w0, #8, lsl #12 // =32768 +; CHECK-GI-NEXT: cmp w8, #16, lsl #12 // =65536 +; CHECK-GI-NEXT: cset w0, lo +; CHECK-GI-NEXT: ret %tmp0 = add i32 %x, 32768 ; 1U << (16-1) %tmp1 = icmp ult i32 %tmp0, 65536 ; 1U << 16 ret i1 %tmp1 } define i1 @add_ultcmp_i32_i8(i32 %x) nounwind { -; CHECK-LABEL: add_ultcmp_i32_i8: -; CHECK: // %bb.0: -; CHECK-NEXT: cmp w0, w0, sxtb -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ultcmp_i32_i8: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: cmp w0, w0, sxtb +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ultcmp_i32_i8: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: add w8, w0, #128 +; CHECK-GI-NEXT: cmp w8, #256 +; CHECK-GI-NEXT: cset w0, lo +; CHECK-GI-NEXT: ret %tmp0 = add i32 %x, 128 ; 1U << (8-1) %tmp1 = icmp ult i32 %tmp0, 256 ; 1U << 8 ret i1 %tmp1 } define i1 @add_ultcmp_i64_i32(i64 %x) nounwind { -; CHECK-LABEL: add_ultcmp_i64_i32: -; CHECK: // %bb.0: -; CHECK-NEXT: cmp x0, w0, sxtw -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ultcmp_i64_i32: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: cmp x0, w0, sxtw +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ultcmp_i64_i32: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #-2147483648 // =0x80000000 +; CHECK-GI-NEXT: mov x9, #4294967296 // =0x100000000 +; CHECK-GI-NEXT: add x8, x0, x8 +; CHECK-GI-NEXT: cmp x8, x9 +; CHECK-GI-NEXT: cset w0, lo +; CHECK-GI-NEXT: ret %tmp0 = add i64 %x, 2147483648 ; 1U << (32-1) %tmp1 = icmp ult i64 %tmp0, 4294967296 ; 1U << 32 ret i1 %tmp1 } define i1 @add_ultcmp_i64_i16(i64 %x) nounwind { -; CHECK-LABEL: add_ultcmp_i64_i16: -; CHECK: // %bb.0: -; CHECK-NEXT: cmp x0, w0, sxth -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ultcmp_i64_i16: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: cmp x0, w0, sxth +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ultcmp_i64_i16: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: add x8, x0, #8, lsl #12 // =32768 +; CHECK-GI-NEXT: cmp x8, #16, lsl #12 // =65536 +; CHECK-GI-NEXT: cset w0, lo +; CHECK-GI-NEXT: ret %tmp0 = add i64 %x, 32768 ; 1U << (16-1) %tmp1 = icmp ult i64 %tmp0, 65536 ; 1U << 16 ret i1 %tmp1 } define i1 @add_ultcmp_i64_i8(i64 %x) nounwind { -; CHECK-LABEL: add_ultcmp_i64_i8: -; CHECK: // %bb.0: -; CHECK-NEXT: cmp x0, w0, sxtb -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ultcmp_i64_i8: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: cmp x0, w0, sxtb +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ultcmp_i64_i8: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: add x8, x0, #128 +; CHECK-GI-NEXT: cmp x8, #256 +; CHECK-GI-NEXT: cset w0, lo +; CHECK-GI-NEXT: ret %tmp0 = add i64 %x, 128 ; 1U << (8-1) %tmp1 = icmp ult i64 %tmp0, 256 ; 1U << 8 ret i1 %tmp1 @@ -267,13 +385,21 @@ define i1 @add_ultcmp_i64_i8(i64 %x) nounwind { ; Slightly more canonical variant define i1 @add_ulecmp_i16_i8(i16 %x) nounwind { -; CHECK-LABEL: add_ulecmp_i16_i8: -; CHECK: // %bb.0: -; CHECK-NEXT: sxtb w8, w0 -; CHECK-NEXT: and w8, w8, #0xffff -; CHECK-NEXT: cmp w8, w0, uxth -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ulecmp_i16_i8: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: sxtb w8, w0 +; CHECK-SD-NEXT: and w8, w8, #0xffff +; CHECK-SD-NEXT: cmp w8, w0, uxth +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ulecmp_i16_i8: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: add w8, w0, #128 +; CHECK-GI-NEXT: and w8, w8, #0xffff +; CHECK-GI-NEXT: cmp w8, #255 +; CHECK-GI-NEXT: cset w0, ls +; CHECK-GI-NEXT: ret %tmp0 = add i16 %x, 128 ; 1U << (8-1) %tmp1 = icmp ule i16 %tmp0, 255 ; (1U << 8) - 1 ret i1 %tmp1 @@ -284,12 +410,20 @@ define i1 @add_ulecmp_i16_i8(i16 %x) nounwind { ; Adding not a constant define i1 @add_ultcmp_bad_i16_i8_add(i16 %x, i16 %y) nounwind { -; CHECK-LABEL: add_ultcmp_bad_i16_i8_add: -; CHECK: // %bb.0: -; CHECK-NEXT: add w8, w0, w1 -; CHECK-NEXT: tst w8, #0xff00 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ultcmp_bad_i16_i8_add: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: add w8, w0, w1 +; CHECK-SD-NEXT: tst w8, #0xff00 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ultcmp_bad_i16_i8_add: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: add w8, w0, w1 +; CHECK-GI-NEXT: and w8, w8, #0xffff +; CHECK-GI-NEXT: cmp w8, #256 +; CHECK-GI-NEXT: cset w0, lo +; CHECK-GI-NEXT: ret %tmp0 = add i16 %x, %y %tmp1 = icmp ult i16 %tmp0, 256 ; 1U << 8 ret i1 %tmp1 @@ -311,12 +445,20 @@ define i1 @add_ultcmp_bad_i16_i8_cmp(i16 %x, i16 %y) nounwind { ; Second constant is not larger than the first one define i1 @add_ultcmp_bad_i8_i16(i16 %x) nounwind { -; CHECK-LABEL: add_ultcmp_bad_i8_i16: -; CHECK: // %bb.0: -; CHECK-NEXT: and w8, w0, #0xffff -; CHECK-NEXT: add w8, w8, #128 -; CHECK-NEXT: lsr w0, w8, #16 -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ultcmp_bad_i8_i16: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: and w8, w0, #0xffff +; CHECK-SD-NEXT: add w8, w8, #128 +; CHECK-SD-NEXT: lsr w0, w8, #16 +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ultcmp_bad_i8_i16: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: and w8, w0, #0xffff +; CHECK-GI-NEXT: add w8, w8, #128 +; CHECK-GI-NEXT: cmp w8, w8, uxth +; CHECK-GI-NEXT: cset w0, ne +; CHECK-GI-NEXT: ret %tmp0 = add i16 %x, 128 ; 1U << (8-1) %tmp1 = icmp ult i16 %tmp0, 128 ; 1U << (8-1) ret i1 %tmp1 @@ -324,12 +466,20 @@ define i1 @add_ultcmp_bad_i8_i16(i16 %x) nounwind { ; First constant is not power of two define i1 @add_ultcmp_bad_i16_i8_c0notpoweroftwo(i16 %x) nounwind { -; CHECK-LABEL: add_ultcmp_bad_i16_i8_c0notpoweroftwo: -; CHECK: // %bb.0: -; CHECK-NEXT: add w8, w0, #192 -; CHECK-NEXT: tst w8, #0xff00 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ultcmp_bad_i16_i8_c0notpoweroftwo: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: add w8, w0, #192 +; CHECK-SD-NEXT: tst w8, #0xff00 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ultcmp_bad_i16_i8_c0notpoweroftwo: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: add w8, w0, #192 +; CHECK-GI-NEXT: and w8, w8, #0xffff +; CHECK-GI-NEXT: cmp w8, #256 +; CHECK-GI-NEXT: cset w0, lo +; CHECK-GI-NEXT: ret %tmp0 = add i16 %x, 192 ; (1U << (8-1)) + (1U << (8-1-1)) %tmp1 = icmp ult i16 %tmp0, 256 ; 1U << 8 ret i1 %tmp1 @@ -351,12 +501,20 @@ define i1 @add_ultcmp_bad_i16_i8_c1notpoweroftwo(i16 %x) nounwind { ; Magic check fails, 64 << 1 != 256 define i1 @add_ultcmp_bad_i16_i8_magic(i16 %x) nounwind { -; CHECK-LABEL: add_ultcmp_bad_i16_i8_magic: -; CHECK: // %bb.0: -; CHECK-NEXT: add w8, w0, #64 -; CHECK-NEXT: tst w8, #0xff00 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ultcmp_bad_i16_i8_magic: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: add w8, w0, #64 +; CHECK-SD-NEXT: tst w8, #0xff00 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ultcmp_bad_i16_i8_magic: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: add w8, w0, #64 +; CHECK-GI-NEXT: and w8, w8, #0xffff +; CHECK-GI-NEXT: cmp w8, #256 +; CHECK-GI-NEXT: cset w0, lo +; CHECK-GI-NEXT: ret %tmp0 = add i16 %x, 64 ; 1U << (8-1-1) %tmp1 = icmp ult i16 %tmp0, 256 ; 1U << 8 ret i1 %tmp1 @@ -364,12 +522,20 @@ define i1 @add_ultcmp_bad_i16_i8_magic(i16 %x) nounwind { ; Bad 'destination type' define i1 @add_ultcmp_bad_i16_i4(i16 %x) nounwind { -; CHECK-LABEL: add_ultcmp_bad_i16_i4: -; CHECK: // %bb.0: -; CHECK-NEXT: add w8, w0, #8 -; CHECK-NEXT: tst w8, #0xfff0 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ultcmp_bad_i16_i4: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: add w8, w0, #8 +; CHECK-SD-NEXT: tst w8, #0xfff0 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ultcmp_bad_i16_i4: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: add w8, w0, #8 +; CHECK-GI-NEXT: and w8, w8, #0xffff +; CHECK-GI-NEXT: cmp w8, #16 +; CHECK-GI-NEXT: cset w0, lo +; CHECK-GI-NEXT: ret %tmp0 = add i16 %x, 8 ; 1U << (4-1) %tmp1 = icmp ult i16 %tmp0, 16 ; 1U << 4 ret i1 %tmp1 @@ -377,12 +543,20 @@ define i1 @add_ultcmp_bad_i16_i4(i16 %x) nounwind { ; Bad storage type define i1 @add_ultcmp_bad_i24_i8(i24 %x) nounwind { -; CHECK-LABEL: add_ultcmp_bad_i24_i8: -; CHECK: // %bb.0: -; CHECK-NEXT: add w8, w0, #128 -; CHECK-NEXT: tst w8, #0xffff00 -; CHECK-NEXT: cset w0, eq -; CHECK-NEXT: ret +; CHECK-SD-LABEL: add_ultcmp_bad_i24_i8: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: add w8, w0, #128 +; CHECK-SD-NEXT: tst w8, #0xffff00 +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: add_ultcmp_bad_i24_i8: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: add w8, w0, #128 +; CHECK-GI-NEXT: and w8, w8, #0xffffff +; CHECK-GI-NEXT: cmp w8, #256 +; CHECK-GI-NEXT: cset w0, lo +; CHECK-GI-NEXT: ret %tmp0 = add i24 %x, 128 ; 1U << (8-1) %tmp1 = icmp ult i24 %tmp0, 256 ; 1U << 8 ret i1 %tmp1 diff --git a/llvm/test/CodeGen/AArch64/stackmap.ll b/llvm/test/CodeGen/AArch64/stackmap.ll index 995d254..26221d0 100644 --- a/llvm/test/CodeGen/AArch64/stackmap.ll +++ b/llvm/test/CodeGen/AArch64/stackmap.ll @@ -81,14 +81,14 @@ ; CHECK-NEXT: .hword 8 ; CHECK-NEXT: .hword 0 ; CHECK-NEXT: .hword 0 -; CHECK-NEXT: .word 65535 +; CHECK-NEXT: .word -1 ; SmallConstant ; CHECK-NEXT: .byte 4 ; CHECK-NEXT: .byte 0 ; CHECK-NEXT: .hword 8 ; CHECK-NEXT: .hword 0 ; CHECK-NEXT: .hword 0 -; CHECK-NEXT: .word 65535 +; CHECK-NEXT: .word -1 ; SmallConstant ; CHECK-NEXT: .byte 4 ; CHECK-NEXT: .byte 0 diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/add.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/add.ll new file mode 100644 index 0000000..e117200 --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/add.ll @@ -0,0 +1,612 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=hawaii < %s | FileCheck -check-prefix=GFX7 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=gfx900 < %s | FileCheck -check-prefix=GFX9 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=fiji < %s | FileCheck -check-prefix=GFX8 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=gfx1010 < %s | FileCheck -check-prefix=GFX10 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 < %s | FileCheck -check-prefix=GFX11 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=gfx1200 -amdgpu-enable-delay-alu=0 < %s | FileCheck -check-prefix=GFX12 %s + +define i16 @s_add_i16(i16 inreg %a, i16 inreg %b) { +; GFX7-LABEL: s_add_i16: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: s_add_i32 s16, s16, s17 +; GFX7-NEXT: v_mov_b32_e32 v0, s16 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: s_add_i16: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: s_add_i32 s16, s16, s17 +; GFX9-NEXT: v_mov_b32_e32 v0, s16 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: s_add_i16: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: s_add_i32 s16, s16, s17 +; GFX8-NEXT: v_mov_b32_e32 v0, s16 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: s_add_i16: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_add_i32 s16, s16, s17 +; GFX10-NEXT: v_mov_b32_e32 v0, s16 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: s_add_i16: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_add_i32 s0, s0, s1 +; GFX11-NEXT: v_mov_b32_e32 v0, s0 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: s_add_i16: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: s_add_co_i32 s0, s0, s1 +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: v_mov_b32_e32 v0, s0 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = add i16 %a, %b + ret i16 %c +} + +define i16 @v_add_i16(i16 %a, i16 %b) { +; GFX7-LABEL: v_add_i16: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: v_add_i32_e32 v0, vcc, v0, v1 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: v_add_i16: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_add_u16_e32 v0, v0, v1 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: v_add_i16: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: v_add_u16_e32 v0, v0, v1 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: v_add_i16: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: v_add_nc_u16 v0, v0, v1 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_add_i16: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_add_nc_u16 v0.l, v0.l, v1.l +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: v_add_i16: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: v_add_nc_u16 v0, v0, v1 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = add i16 %a, %b + ret i16 %c +} + +define i32 @s_add_i32(i32 inreg %a, i32 inreg %b) { +; GFX7-LABEL: s_add_i32: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: s_add_i32 s16, s16, s17 +; GFX7-NEXT: v_mov_b32_e32 v0, s16 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: s_add_i32: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: s_add_i32 s16, s16, s17 +; GFX9-NEXT: v_mov_b32_e32 v0, s16 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: s_add_i32: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: s_add_i32 s16, s16, s17 +; GFX8-NEXT: v_mov_b32_e32 v0, s16 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: s_add_i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_add_i32 s16, s16, s17 +; GFX10-NEXT: v_mov_b32_e32 v0, s16 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: s_add_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_add_i32 s0, s0, s1 +; GFX11-NEXT: v_mov_b32_e32 v0, s0 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: s_add_i32: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: s_add_co_i32 s0, s0, s1 +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: v_mov_b32_e32 v0, s0 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = add i32 %a, %b + ret i32 %c +} + +define i32 @v_add_i32(i32 %a, i32 %b) { +; GFX7-LABEL: v_add_i32: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: v_add_i32_e32 v0, vcc, v0, v1 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: v_add_i32: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_add_u32_e32 v0, v0, v1 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: v_add_i32: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: v_add_u32_e32 v0, vcc, v0, v1 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: v_add_i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: v_add_nc_u32_e32 v0, v0, v1 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_add_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_add_nc_u32_e32 v0, v0, v1 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: v_add_i32: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: v_add_nc_u32_e32 v0, v0, v1 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = add i32 %a, %b + ret i32 %c +} + +define <2 x i16> @s_add_v2i16(<2 x i16> inreg %a, <2 x i16> inreg %b) { +; GFX7-LABEL: s_add_v2i16: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: s_add_i32 s16, s16, s18 +; GFX7-NEXT: s_add_i32 s17, s17, s19 +; GFX7-NEXT: v_mov_b32_e32 v0, s16 +; GFX7-NEXT: v_mov_b32_e32 v1, s17 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: s_add_v2i16: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: s_lshr_b32 s4, s16, 16 +; GFX9-NEXT: s_lshr_b32 s5, s17, 16 +; GFX9-NEXT: s_add_i32 s16, s16, s17 +; GFX9-NEXT: s_add_i32 s4, s4, s5 +; GFX9-NEXT: s_pack_ll_b32_b16 s4, s16, s4 +; GFX9-NEXT: v_mov_b32_e32 v0, s4 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: s_add_v2i16: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: s_lshr_b32 s4, s16, 16 +; GFX8-NEXT: s_lshr_b32 s5, s17, 16 +; GFX8-NEXT: s_add_i32 s4, s4, s5 +; GFX8-NEXT: s_add_i32 s16, s16, s17 +; GFX8-NEXT: s_and_b32 s4, 0xffff, s4 +; GFX8-NEXT: s_and_b32 s5, 0xffff, s16 +; GFX8-NEXT: s_lshl_b32 s4, s4, 16 +; GFX8-NEXT: s_or_b32 s4, s5, s4 +; GFX8-NEXT: v_mov_b32_e32 v0, s4 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: s_add_v2i16: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_lshr_b32 s4, s16, 16 +; GFX10-NEXT: s_lshr_b32 s5, s17, 16 +; GFX10-NEXT: s_add_i32 s16, s16, s17 +; GFX10-NEXT: s_add_i32 s4, s4, s5 +; GFX10-NEXT: s_pack_ll_b32_b16 s4, s16, s4 +; GFX10-NEXT: v_mov_b32_e32 v0, s4 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: s_add_v2i16: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_lshr_b32 s2, s0, 16 +; GFX11-NEXT: s_lshr_b32 s3, s1, 16 +; GFX11-NEXT: s_add_i32 s0, s0, s1 +; GFX11-NEXT: s_add_i32 s2, s2, s3 +; GFX11-NEXT: s_pack_ll_b32_b16 s0, s0, s2 +; GFX11-NEXT: v_mov_b32_e32 v0, s0 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: s_add_v2i16: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: s_lshr_b32 s2, s0, 16 +; GFX12-NEXT: s_lshr_b32 s3, s1, 16 +; GFX12-NEXT: s_add_co_i32 s0, s0, s1 +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: s_add_co_i32 s2, s2, s3 +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: s_pack_ll_b32_b16 s0, s0, s2 +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: v_mov_b32_e32 v0, s0 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = add <2 x i16> %a, %b + ret <2 x i16> %c +} + +define <2 x i16> @v_add_v2i16(<2 x i16> %a, <2 x i16> %b) { +; GFX7-LABEL: v_add_v2i16: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: v_add_i32_e32 v0, vcc, v0, v2 +; GFX7-NEXT: v_add_i32_e32 v1, vcc, v1, v3 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: v_add_v2i16: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_pk_add_u16 v0, v0, v1 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: v_add_v2i16: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: v_add_u16_e32 v2, v0, v1 +; GFX8-NEXT: v_add_u16_sdwa v0, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_or_b32_e32 v0, v2, v0 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: v_add_v2i16: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: v_pk_add_u16 v0, v0, v1 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_add_v2i16: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_pk_add_u16 v0, v0, v1 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: v_add_v2i16: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: v_pk_add_u16 v0, v0, v1 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = add <2 x i16> %a, %b + ret <2 x i16> %c +} + +define i64 @s_add_i64(i64 inreg %a, i64 inreg %b) { +; GFX7-LABEL: s_add_i64: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: s_add_u32 s4, s16, s18 +; GFX7-NEXT: s_addc_u32 s5, s17, s19 +; GFX7-NEXT: v_mov_b32_e32 v0, s4 +; GFX7-NEXT: v_mov_b32_e32 v1, s5 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: s_add_i64: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: s_add_u32 s4, s16, s18 +; GFX9-NEXT: s_addc_u32 s5, s17, s19 +; GFX9-NEXT: v_mov_b32_e32 v0, s4 +; GFX9-NEXT: v_mov_b32_e32 v1, s5 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: s_add_i64: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: s_add_u32 s4, s16, s18 +; GFX8-NEXT: s_addc_u32 s5, s17, s19 +; GFX8-NEXT: v_mov_b32_e32 v0, s4 +; GFX8-NEXT: v_mov_b32_e32 v1, s5 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: s_add_i64: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_add_u32 s4, s16, s18 +; GFX10-NEXT: s_addc_u32 s5, s17, s19 +; GFX10-NEXT: v_mov_b32_e32 v0, s4 +; GFX10-NEXT: v_mov_b32_e32 v1, s5 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: s_add_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_add_u32 s0, s0, s2 +; GFX11-NEXT: s_addc_u32 s1, s1, s3 +; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: s_add_i64: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: s_add_nc_u64 s[0:1], s[0:1], s[2:3] +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = add i64 %a, %b + ret i64 %c +} + +define i64 @v_add_i64(i64 %a, i64 %b) { +; GFX7-LABEL: v_add_i64: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: v_add_i32_e32 v0, vcc, v0, v2 +; GFX7-NEXT: v_addc_u32_e32 v1, vcc, v1, v3, vcc +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: v_add_i64: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_add_co_u32_e32 v0, vcc, v0, v2 +; GFX9-NEXT: v_addc_co_u32_e32 v1, vcc, v1, v3, vcc +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: v_add_i64: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: v_add_u32_e32 v0, vcc, v0, v2 +; GFX8-NEXT: v_addc_u32_e32 v1, vcc, v1, v3, vcc +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: v_add_i64: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2 +; GFX10-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_add_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2 +; GFX11-NEXT: v_add_co_ci_u32_e64 v1, null, v1, v3, vcc_lo +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: v_add_i64: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2 +; GFX12-NEXT: s_wait_alu 0xfffd +; GFX12-NEXT: v_add_co_ci_u32_e64 v1, null, v1, v3, vcc_lo +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = add i64 %a, %b + ret i64 %c +} + +define void @s_uaddo_uadde(i64 inreg %a, i64 inreg %b, ptr addrspace(1) %res, ptr addrspace(1) %carry) { +; GFX7-LABEL: s_uaddo_uadde: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: s_add_u32 s4, s16, s18 +; GFX7-NEXT: s_addc_u32 s5, s17, s19 +; GFX7-NEXT: v_mov_b32_e32 v4, s4 +; GFX7-NEXT: s_mov_b32 s6, 0 +; GFX7-NEXT: s_cselect_b32 s8, 1, 0 +; GFX7-NEXT: v_mov_b32_e32 v5, s5 +; GFX7-NEXT: s_mov_b32 s7, 0xf000 +; GFX7-NEXT: s_mov_b64 s[4:5], 0 +; GFX7-NEXT: buffer_store_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 +; GFX7-NEXT: v_mov_b32_e32 v0, s8 +; GFX7-NEXT: buffer_store_dword v0, v[2:3], s[4:7], 0 addr64 +; GFX7-NEXT: s_waitcnt vmcnt(0) +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: s_uaddo_uadde: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: s_add_u32 s4, s16, s18 +; GFX9-NEXT: s_addc_u32 s5, s17, s19 +; GFX9-NEXT: v_mov_b32_e32 v4, s4 +; GFX9-NEXT: s_cselect_b32 s6, 1, 0 +; GFX9-NEXT: v_mov_b32_e32 v5, s5 +; GFX9-NEXT: global_store_dwordx2 v[0:1], v[4:5], off +; GFX9-NEXT: v_mov_b32_e32 v0, s6 +; GFX9-NEXT: global_store_dword v[2:3], v0, off +; GFX9-NEXT: s_waitcnt vmcnt(0) +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: s_uaddo_uadde: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: s_add_u32 s4, s16, s18 +; GFX8-NEXT: s_addc_u32 s5, s17, s19 +; GFX8-NEXT: v_mov_b32_e32 v4, s4 +; GFX8-NEXT: s_cselect_b32 s6, 1, 0 +; GFX8-NEXT: v_mov_b32_e32 v5, s5 +; GFX8-NEXT: flat_store_dwordx2 v[0:1], v[4:5] +; GFX8-NEXT: v_mov_b32_e32 v0, s6 +; GFX8-NEXT: flat_store_dword v[2:3], v0 +; GFX8-NEXT: s_waitcnt vmcnt(0) +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: s_uaddo_uadde: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_add_u32 s4, s16, s18 +; GFX10-NEXT: s_addc_u32 s5, s17, s19 +; GFX10-NEXT: s_cselect_b32 s6, 1, 0 +; GFX10-NEXT: v_mov_b32_e32 v4, s4 +; GFX10-NEXT: v_mov_b32_e32 v5, s5 +; GFX10-NEXT: v_mov_b32_e32 v6, s6 +; GFX10-NEXT: global_store_dwordx2 v[0:1], v[4:5], off +; GFX10-NEXT: global_store_dword v[2:3], v6, off +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: s_uaddo_uadde: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_add_u32 s0, s0, s2 +; GFX11-NEXT: s_addc_u32 s1, s1, s3 +; GFX11-NEXT: s_cselect_b32 s2, 1, 0 +; GFX11-NEXT: v_dual_mov_b32 v5, s1 :: v_dual_mov_b32 v4, s0 +; GFX11-NEXT: v_mov_b32_e32 v6, s2 +; GFX11-NEXT: global_store_b64 v[0:1], v[4:5], off +; GFX11-NEXT: global_store_b32 v[2:3], v6, off +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: s_uaddo_uadde: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: s_add_co_u32 s0, s0, s2 +; GFX12-NEXT: s_add_co_ci_u32 s1, s1, s3 +; GFX12-NEXT: s_cselect_b32 s2, 1, 0 +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: v_dual_mov_b32 v5, s1 :: v_dual_mov_b32 v4, s0 +; GFX12-NEXT: v_mov_b32_e32 v6, s2 +; GFX12-NEXT: global_store_b64 v[0:1], v[4:5], off +; GFX12-NEXT: global_store_b32 v[2:3], v6, off +; GFX12-NEXT: s_setpc_b64 s[30:31] + %uaddo = call {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b) + %add = extractvalue {i64, i1} %uaddo, 0 + %of = extractvalue {i64, i1} %uaddo, 1 + %of32 = select i1 %of, i32 1, i32 0 + store i64 %add, ptr addrspace(1) %res + store i32 %of32, ptr addrspace(1) %carry + ret void +} + +define void @v_uaddo_uadde(i64 %a, i64 %b, ptr addrspace(1) %res, ptr addrspace(1) %carry) { +; GFX7-LABEL: v_uaddo_uadde: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: v_add_i32_e32 v0, vcc, v0, v2 +; GFX7-NEXT: v_addc_u32_e32 v1, vcc, v1, v3, vcc +; GFX7-NEXT: s_mov_b32 s6, 0 +; GFX7-NEXT: s_mov_b32 s7, 0xf000 +; GFX7-NEXT: s_mov_b64 s[4:5], 0 +; GFX7-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc +; GFX7-NEXT: buffer_store_dwordx2 v[0:1], v[4:5], s[4:7], 0 addr64 +; GFX7-NEXT: buffer_store_dword v2, v[6:7], s[4:7], 0 addr64 +; GFX7-NEXT: s_waitcnt vmcnt(0) +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: v_uaddo_uadde: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_add_co_u32_e32 v0, vcc, v0, v2 +; GFX9-NEXT: v_addc_co_u32_e32 v1, vcc, v1, v3, vcc +; GFX9-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc +; GFX9-NEXT: global_store_dwordx2 v[4:5], v[0:1], off +; GFX9-NEXT: global_store_dword v[6:7], v2, off +; GFX9-NEXT: s_waitcnt vmcnt(0) +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: v_uaddo_uadde: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: v_add_u32_e32 v0, vcc, v0, v2 +; GFX8-NEXT: v_addc_u32_e32 v1, vcc, v1, v3, vcc +; GFX8-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc +; GFX8-NEXT: flat_store_dwordx2 v[4:5], v[0:1] +; GFX8-NEXT: flat_store_dword v[6:7], v2 +; GFX8-NEXT: s_waitcnt vmcnt(0) +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: v_uaddo_uadde: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2 +; GFX10-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo +; GFX10-NEXT: global_store_dwordx2 v[4:5], v[0:1], off +; GFX10-NEXT: global_store_dword v[6:7], v2, off +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_uaddo_uadde: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2 +; GFX11-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo +; GFX11-NEXT: global_store_b64 v[4:5], v[0:1], off +; GFX11-NEXT: global_store_b32 v[6:7], v2, off +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: v_uaddo_uadde: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2 +; GFX12-NEXT: s_wait_alu 0xfffd +; GFX12-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX12-NEXT: s_wait_alu 0xfffd +; GFX12-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo +; GFX12-NEXT: global_store_b64 v[4:5], v[0:1], off +; GFX12-NEXT: global_store_b32 v[6:7], v2, off +; GFX12-NEXT: s_setpc_b64 s[30:31] + %uaddo = call {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b) + %add = extractvalue {i64, i1} %uaddo, 0 + %of = extractvalue {i64, i1} %uaddo, 1 + %of32 = select i1 %of, i32 1, i32 0 + store i64 %add, ptr addrspace(1) %res + store i32 %of32, ptr addrspace(1) %carry + ret void +} + +declare {i64, i1} @llvm.uadd.with.overflow.i64(i64 %a, i64 %b) diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/fadd.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/fadd.ll new file mode 100644 index 0000000..e440bee --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/fadd.ll @@ -0,0 +1,165 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mattr=-real-true16 -mcpu=gfx1100 -o - %s | FileCheck -check-prefixes=GCN,GFX11,GFX11-FAKE16 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mattr=+real-true16 -mcpu=gfx1100 -o - %s | FileCheck -check-prefixes=GCN,GFX11,GFX11-TRUE16 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mattr=-real-true16 -mcpu=gfx1200 -o - %s | FileCheck -check-prefixes=GCN,GFX12,GFX12-FAKE16 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mattr=+real-true16 -mcpu=gfx1200 -o - %s | FileCheck -check-prefixes=GCN,GFX12,GFX12-TRUE16 %s + +define amdgpu_ps half @fadd_s16_uniform(half inreg %a, half inreg %b) { +; GFX11-FAKE16-LABEL: fadd_s16_uniform: +; GFX11-FAKE16: ; %bb.0: +; GFX11-FAKE16-NEXT: v_add_f16_e64 v0, s0, s1 +; GFX11-FAKE16-NEXT: ; return to shader part epilog +; +; GFX11-TRUE16-LABEL: fadd_s16_uniform: +; GFX11-TRUE16: ; %bb.0: +; GFX11-TRUE16-NEXT: v_add_f16_e64 v0.l, s0, s1 +; GFX11-TRUE16-NEXT: ; return to shader part epilog +; +; GFX12-LABEL: fadd_s16_uniform: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_add_f16 s0, s0, s1 +; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_3) +; GFX12-NEXT: v_mov_b32_e32 v0, s0 +; GFX12-NEXT: ; return to shader part epilog + %fadd = fadd half %a, %b + ret half %fadd +} + +define amdgpu_ps half @fadd_s16_div(half %a, half %b) { +; GFX11-FAKE16-LABEL: fadd_s16_div: +; GFX11-FAKE16: ; %bb.0: +; GFX11-FAKE16-NEXT: v_add_f16_e32 v0, v0, v1 +; GFX11-FAKE16-NEXT: ; return to shader part epilog +; +; GFX11-TRUE16-LABEL: fadd_s16_div: +; GFX11-TRUE16: ; %bb.0: +; GFX11-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v1.l +; GFX11-TRUE16-NEXT: ; return to shader part epilog +; +; GFX12-FAKE16-LABEL: fadd_s16_div: +; GFX12-FAKE16: ; %bb.0: +; GFX12-FAKE16-NEXT: v_add_f16_e32 v0, v0, v1 +; GFX12-FAKE16-NEXT: ; return to shader part epilog +; +; GFX12-TRUE16-LABEL: fadd_s16_div: +; GFX12-TRUE16: ; %bb.0: +; GFX12-TRUE16-NEXT: v_add_f16_e32 v0.l, v0.l, v1.l +; GFX12-TRUE16-NEXT: ; return to shader part epilog + %fadd = fadd half %a, %b + ret half %fadd +} + +define amdgpu_ps float @fadd_s32_uniform(float inreg %a, float inreg %b) { +; GFX11-LABEL: fadd_s32_uniform: +; GFX11: ; %bb.0: +; GFX11-NEXT: v_add_f32_e64 v0, s0, s1 +; GFX11-NEXT: ; return to shader part epilog +; +; GFX12-LABEL: fadd_s32_uniform: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_add_f32 s0, s0, s1 +; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_3) +; GFX12-NEXT: v_mov_b32_e32 v0, s0 +; GFX12-NEXT: ; return to shader part epilog + %fadd = fadd float %a, %b + ret float %fadd +} + +define amdgpu_ps float @fadd_s32_div(float %a, float %b) { +; GCN-LABEL: fadd_s32_div: +; GCN: ; %bb.0: +; GCN-NEXT: v_add_f32_e32 v0, v0, v1 +; GCN-NEXT: ; return to shader part epilog + %fadd = fadd float %a, %b + ret float %fadd +} + +define amdgpu_ps void @fadd_s64_uniform(double inreg %a, double inreg %b, ptr addrspace(1) %ptr) { +; GFX11-LABEL: fadd_s64_uniform: +; GFX11: ; %bb.0: +; GFX11-NEXT: v_add_f64 v[2:3], s[0:1], s[2:3] +; GFX11-NEXT: global_store_b64 v[0:1], v[2:3], off +; GFX11-NEXT: s_endpgm +; +; GFX12-LABEL: fadd_s64_uniform: +; GFX12: ; %bb.0: +; GFX12-NEXT: v_add_f64_e64 v[2:3], s[0:1], s[2:3] +; GFX12-NEXT: global_store_b64 v[0:1], v[2:3], off +; GFX12-NEXT: s_endpgm + %fadd = fadd double %a, %b + store double %fadd, ptr addrspace(1) %ptr + ret void +} + +define amdgpu_ps void @fadd_s64_div(double %a, double %b, ptr addrspace(1) %ptr) { +; GFX11-LABEL: fadd_s64_div: +; GFX11: ; %bb.0: +; GFX11-NEXT: v_add_f64 v[0:1], v[0:1], v[2:3] +; GFX11-NEXT: global_store_b64 v[4:5], v[0:1], off +; GFX11-NEXT: s_endpgm +; +; GFX12-LABEL: fadd_s64_div: +; GFX12: ; %bb.0: +; GFX12-NEXT: v_add_f64_e32 v[0:1], v[0:1], v[2:3] +; GFX12-NEXT: global_store_b64 v[4:5], v[0:1], off +; GFX12-NEXT: s_endpgm + %fadd = fadd double %a, %b + store double %fadd, ptr addrspace(1) %ptr + ret void +} + +define amdgpu_ps <2 x half> @fadd_v2s16_uniform(<2 x half> inreg %a, <2 x half> inreg %b) { +; GFX11-LABEL: fadd_v2s16_uniform: +; GFX11: ; %bb.0: +; GFX11-NEXT: v_pk_add_f16 v0, s0, s1 +; GFX11-NEXT: ; return to shader part epilog +; +; GFX12-LABEL: fadd_v2s16_uniform: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_lshr_b32 s2, s0, 16 +; GFX12-NEXT: s_lshr_b32 s3, s1, 16 +; GFX12-NEXT: s_add_f16 s0, s0, s1 +; GFX12-NEXT: s_add_f16 s1, s2, s3 +; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_3) | instskip(NEXT) | instid1(SALU_CYCLE_1) +; GFX12-NEXT: s_pack_ll_b32_b16 s0, s0, s1 +; GFX12-NEXT: v_mov_b32_e32 v0, s0 +; GFX12-NEXT: ; return to shader part epilog + %fadd = fadd <2 x half> %a, %b + ret <2 x half> %fadd +} + +define amdgpu_ps <2 x half> @fadd_v2s16_div(<2 x half> %a, <2 x half> %b) { +; GCN-LABEL: fadd_v2s16_div: +; GCN: ; %bb.0: +; GCN-NEXT: v_pk_add_f16 v0, v0, v1 +; GCN-NEXT: ; return to shader part epilog + %fadd = fadd <2 x half> %a, %b + ret <2 x half> %fadd +} + +define amdgpu_ps <2 x float> @fadd_v2s32_uniform(<2 x float> inreg %a, <2 x float> inreg %b) { +; GFX11-LABEL: fadd_v2s32_uniform: +; GFX11: ; %bb.0: +; GFX11-NEXT: v_add_f32_e64 v0, s0, s2 +; GFX11-NEXT: v_add_f32_e64 v1, s1, s3 +; GFX11-NEXT: ; return to shader part epilog +; +; GFX12-LABEL: fadd_v2s32_uniform: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_add_f32 s0, s0, s2 +; GFX12-NEXT: s_add_f32 s1, s1, s3 +; GFX12-NEXT: s_delay_alu instid0(SALU_CYCLE_3) +; GFX12-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1 +; GFX12-NEXT: ; return to shader part epilog + %fadd = fadd <2 x float> %a, %b + ret <2 x float> %fadd +} + +define amdgpu_ps <2 x float> @fadd_v2s32_div(<2 x float> %a, <2 x float> %b) { +; GCN-LABEL: fadd_v2s32_div: +; GCN: ; %bb.0: +; GCN-NEXT: v_dual_add_f32 v0, v0, v2 :: v_dual_add_f32 v1, v1, v3 +; GCN-NEXT: ; return to shader part epilog + %fadd = fadd <2 x float> %a, %b + ret <2 x float> %fadd +} diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-copy-scc-vcc.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-copy-scc-vcc.ll new file mode 100644 index 0000000..588802c --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-copy-scc-vcc.ll @@ -0,0 +1,66 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx700 < %s | FileCheck -check-prefixes=GFX7 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx803 < %s | FileCheck -check-prefixes=GFX8 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1100 < %s | FileCheck -check-prefixes=GFX11 %s + +define amdgpu_kernel void @fcmp_uniform_select(float %a, i32 %b, i32 %c, ptr addrspace(1) %out) { +; GFX7-LABEL: fcmp_uniform_select: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x9 +; GFX7-NEXT: s_load_dword s3, s[4:5], 0xb +; GFX7-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xd +; GFX7-NEXT: s_mov_b32 s2, -1 +; GFX7-NEXT: s_waitcnt lgkmcnt(0) +; GFX7-NEXT: v_cmp_eq_f32_e64 s[4:5], s6, 0 +; GFX7-NEXT: s_or_b64 s[4:5], s[4:5], s[4:5] +; GFX7-NEXT: s_cselect_b32 s4, 1, 0 +; GFX7-NEXT: s_and_b32 s4, s4, 1 +; GFX7-NEXT: s_cmp_lg_u32 s4, 0 +; GFX7-NEXT: s_cselect_b32 s3, s7, s3 +; GFX7-NEXT: v_mov_b32_e32 v0, s3 +; GFX7-NEXT: s_mov_b32 s3, 0xf000 +; GFX7-NEXT: buffer_store_dword v0, off, s[0:3], 0 +; GFX7-NEXT: s_endpgm +; +; GFX8-LABEL: fcmp_uniform_select: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24 +; GFX8-NEXT: s_load_dword s6, s[4:5], 0x2c +; GFX8-NEXT: s_load_dwordx2 s[2:3], s[4:5], 0x34 +; GFX8-NEXT: s_waitcnt lgkmcnt(0) +; GFX8-NEXT: v_cmp_eq_f32_e64 s[4:5], s0, 0 +; GFX8-NEXT: s_cmp_lg_u64 s[4:5], 0 +; GFX8-NEXT: s_cselect_b32 s0, 1, 0 +; GFX8-NEXT: s_and_b32 s0, s0, 1 +; GFX8-NEXT: s_cmp_lg_u32 s0, 0 +; GFX8-NEXT: s_cselect_b32 s0, s1, s6 +; GFX8-NEXT: v_mov_b32_e32 v0, s2 +; GFX8-NEXT: v_mov_b32_e32 v2, s0 +; GFX8-NEXT: v_mov_b32_e32 v1, s3 +; GFX8-NEXT: flat_store_dword v[0:1], v2 +; GFX8-NEXT: s_endpgm +; +; GFX11-LABEL: fcmp_uniform_select: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_clause 0x2 +; GFX11-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 +; GFX11-NEXT: s_load_b32 s6, s[4:5], 0x2c +; GFX11-NEXT: s_load_b64 s[2:3], s[4:5], 0x34 +; GFX11-NEXT: v_mov_b32_e32 v1, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: v_cmp_eq_f32_e64 s0, s0, 0 +; GFX11-NEXT: s_cmp_lg_u32 s0, 0 +; GFX11-NEXT: s_cselect_b32 s0, 1, 0 +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1) +; GFX11-NEXT: s_and_b32 s0, s0, 1 +; GFX11-NEXT: s_cmp_lg_u32 s0, 0 +; GFX11-NEXT: s_cselect_b32 s0, s1, s6 +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) +; GFX11-NEXT: v_mov_b32_e32 v0, s0 +; GFX11-NEXT: global_store_b32 v1, v0, s[2:3] +; GFX11-NEXT: s_endpgm + %cmp = fcmp oeq float %a, 0.0 + %sel = select i1 %cmp, i32 %b, i32 %c + store i32 %sel, ptr addrspace(1) %out + ret void +} diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-copy-scc-vcc.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-copy-scc-vcc.mir new file mode 100644 index 0000000..b6652f6 --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-copy-scc-vcc.mir @@ -0,0 +1,37 @@ +# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py +# RUN: llc -mtriple=amdgcn -mcpu=gfx700 -run-pass=instruction-select %s -o - | FileCheck -check-prefixes=GFX7 %s +# RUN: llc -mtriple=amdgcn -mcpu=gfx803 -run-pass=instruction-select %s -o - | FileCheck -check-prefixes=GF8 %s +# RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -run-pass=instruction-select %s -o - | FileCheck -check-prefixes=GFX11 %s + +--- +name: test_copy_scc_vcc +legalized: true +regBankSelected: true +tracksRegLiveness: true +body: | + bb.0: + ; GFX7-LABEL: name: test_copy_scc_vcc + ; GFX7: [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF + ; GFX7-NEXT: [[S_OR_B64_:%[0-9]+]]:sreg_64 = S_OR_B64 [[DEF]], [[DEF]], implicit-def $scc + ; GFX7-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $scc + ; GFX7-NEXT: $sgpr0 = COPY [[COPY]] + ; GFX7-NEXT: S_ENDPGM 0, implicit $sgpr0 + ; + ; GF8-LABEL: name: test_copy_scc_vcc + ; GF8: [[DEF:%[0-9]+]]:sreg_64_xexec = IMPLICIT_DEF + ; GF8-NEXT: S_CMP_LG_U64 [[DEF]], 0, implicit-def $scc + ; GF8-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $scc + ; GF8-NEXT: $sgpr0 = COPY [[COPY]] + ; GF8-NEXT: S_ENDPGM 0, implicit $sgpr0 + ; + ; GFX11-LABEL: name: test_copy_scc_vcc + ; GFX11: [[DEF:%[0-9]+]]:sreg_32_xm0_xexec = IMPLICIT_DEF + ; GFX11-NEXT: S_CMP_LG_U32 [[DEF]], 0, implicit-def $scc + ; GFX11-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY $scc + ; GFX11-NEXT: $sgpr0 = COPY [[COPY]] + ; GFX11-NEXT: S_ENDPGM 0, implicit $sgpr0 + %0:vcc(s1) = G_IMPLICIT_DEF + %1:sgpr(s32) = G_AMDGPU_COPY_SCC_VCC %0 + $sgpr0 = COPY %1 + S_ENDPGM 0, implicit $sgpr0 +... diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.abs.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.abs.ll index 02d0e52..6facdfd 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.abs.ll +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.abs.ll @@ -104,109 +104,110 @@ define amdgpu_cs <4 x i32> @abs_sgpr_v4i32(<4 x i32> inreg %arg) { ret <4 x i32> %res } -define amdgpu_cs i16 @abs_vgpr_i16(i16 %arg) { +define i16 @abs_vgpr_i16(i16 %arg) { ; GFX6-LABEL: abs_vgpr_i16: ; GFX6: ; %bb.0: +; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX6-NEXT: v_bfe_i32 v0, v0, 0, 16 ; GFX6-NEXT: v_sub_i32_e32 v1, vcc, 0, v0 ; GFX6-NEXT: v_max_i32_e32 v0, v0, v1 -; GFX6-NEXT: v_readfirstlane_b32 s0, v0 -; GFX6-NEXT: ; return to shader part epilog +; GFX6-NEXT: s_setpc_b64 s[30:31] ; ; GFX8-LABEL: abs_vgpr_i16: ; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX8-NEXT: v_sub_u16_e32 v1, 0, v0 ; GFX8-NEXT: v_max_i16_e32 v0, v0, v1 -; GFX8-NEXT: v_readfirstlane_b32 s0, v0 -; GFX8-NEXT: ; return to shader part epilog +; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX10-LABEL: abs_vgpr_i16: ; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: v_sub_nc_u16 v1, 0, v0 ; GFX10-NEXT: v_max_i16 v0, v0, v1 -; GFX10-NEXT: v_readfirstlane_b32 s0, v0 -; GFX10-NEXT: ; return to shader part epilog +; GFX10-NEXT: s_setpc_b64 s[30:31] ; ; GFX1250-LABEL: abs_vgpr_i16: ; GFX1250: ; %bb.0: +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 ; GFX1250-NEXT: v_sub_nc_u16 v1, 0, v0 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) ; GFX1250-NEXT: v_max_i16 v0, v0, v1 -; GFX1250-NEXT: v_readfirstlane_b32 s0, v0 -; GFX1250-NEXT: ; return to shader part epilog +; GFX1250-NEXT: s_set_pc_i64 s[30:31] %res = call i16 @llvm.abs.i16(i16 %arg, i1 false) ret i16 %res } -define amdgpu_cs i32 @abs_vgpr_i32(i32 %arg) { +define i32 @abs_vgpr_i32(i32 %arg) { ; GFX6-LABEL: abs_vgpr_i32: ; GFX6: ; %bb.0: +; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX6-NEXT: v_sub_i32_e32 v1, vcc, 0, v0 ; GFX6-NEXT: v_max_i32_e32 v0, v0, v1 -; GFX6-NEXT: v_readfirstlane_b32 s0, v0 -; GFX6-NEXT: ; return to shader part epilog +; GFX6-NEXT: s_setpc_b64 s[30:31] ; ; GFX8-LABEL: abs_vgpr_i32: ; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX8-NEXT: v_sub_u32_e32 v1, vcc, 0, v0 ; GFX8-NEXT: v_max_i32_e32 v0, v0, v1 -; GFX8-NEXT: v_readfirstlane_b32 s0, v0 -; GFX8-NEXT: ; return to shader part epilog +; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX10-LABEL: abs_vgpr_i32: ; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: v_sub_nc_u32_e32 v1, 0, v0 ; GFX10-NEXT: v_max_i32_e32 v0, v0, v1 -; GFX10-NEXT: v_readfirstlane_b32 s0, v0 -; GFX10-NEXT: ; return to shader part epilog +; GFX10-NEXT: s_setpc_b64 s[30:31] ; ; GFX1250-LABEL: abs_vgpr_i32: ; GFX1250: ; %bb.0: +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 ; GFX1250-NEXT: v_sub_nc_u32_e32 v1, 0, v0 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) ; GFX1250-NEXT: v_max_i32_e32 v0, v0, v1 -; GFX1250-NEXT: v_readfirstlane_b32 s0, v0 -; GFX1250-NEXT: ; return to shader part epilog +; GFX1250-NEXT: s_set_pc_i64 s[30:31] %res = call i32 @llvm.abs.i32(i32 %arg, i1 false) ret i32 %res } -define amdgpu_cs i64 @abs_vgpr_i64(i64 %arg) { +define i64 @abs_vgpr_i64(i64 %arg) { ; GFX6-LABEL: abs_vgpr_i64: ; GFX6: ; %bb.0: +; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX6-NEXT: v_ashrrev_i32_e32 v2, 31, v1 ; GFX6-NEXT: v_add_i32_e32 v0, vcc, v0, v2 ; GFX6-NEXT: v_addc_u32_e32 v1, vcc, v1, v2, vcc ; GFX6-NEXT: v_xor_b32_e32 v0, v0, v2 ; GFX6-NEXT: v_xor_b32_e32 v1, v1, v2 -; GFX6-NEXT: v_readfirstlane_b32 s0, v0 -; GFX6-NEXT: v_readfirstlane_b32 s1, v1 -; GFX6-NEXT: ; return to shader part epilog +; GFX6-NEXT: s_setpc_b64 s[30:31] ; ; GFX8-LABEL: abs_vgpr_i64: ; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX8-NEXT: v_ashrrev_i32_e32 v2, 31, v1 ; GFX8-NEXT: v_add_u32_e32 v0, vcc, v0, v2 ; GFX8-NEXT: v_addc_u32_e32 v1, vcc, v1, v2, vcc ; GFX8-NEXT: v_xor_b32_e32 v0, v0, v2 ; GFX8-NEXT: v_xor_b32_e32 v1, v1, v2 -; GFX8-NEXT: v_readfirstlane_b32 s0, v0 -; GFX8-NEXT: v_readfirstlane_b32 s1, v1 -; GFX8-NEXT: ; return to shader part epilog +; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX10-LABEL: abs_vgpr_i64: ; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: v_ashrrev_i32_e32 v2, 31, v1 ; GFX10-NEXT: v_add_co_u32 v0, vcc_lo, v0, v2 ; GFX10-NEXT: v_add_co_ci_u32_e32 v1, vcc_lo, v1, v2, vcc_lo ; GFX10-NEXT: v_xor_b32_e32 v0, v0, v2 ; GFX10-NEXT: v_xor_b32_e32 v1, v1, v2 -; GFX10-NEXT: v_readfirstlane_b32 s0, v0 -; GFX10-NEXT: v_readfirstlane_b32 s1, v1 -; GFX10-NEXT: ; return to shader part epilog +; GFX10-NEXT: s_setpc_b64 s[30:31] ; ; GFX1250-LABEL: abs_vgpr_i64: ; GFX1250: ; %bb.0: +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 ; GFX1250-NEXT: v_ashrrev_i32_e32 v2, 31, v1 ; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) ; GFX1250-NEXT: v_mov_b32_e32 v3, v2 @@ -214,17 +215,15 @@ define amdgpu_cs i64 @abs_vgpr_i64(i64 %arg) { ; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2) ; GFX1250-NEXT: v_xor_b32_e32 v0, v0, v2 ; GFX1250-NEXT: v_xor_b32_e32 v1, v1, v2 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX1250-NEXT: v_readfirstlane_b32 s0, v0 -; GFX1250-NEXT: v_readfirstlane_b32 s1, v1 -; GFX1250-NEXT: ; return to shader part epilog +; GFX1250-NEXT: s_set_pc_i64 s[30:31] %res = call i64 @llvm.abs.i64(i64 %arg, i1 false) ret i64 %res } -define amdgpu_cs <4 x i32> @abs_vgpr_v4i32(<4 x i32> %arg) { +define <4 x i32> @abs_vgpr_v4i32(<4 x i32> %arg) { ; GFX6-LABEL: abs_vgpr_v4i32: ; GFX6: ; %bb.0: +; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX6-NEXT: v_sub_i32_e32 v4, vcc, 0, v0 ; GFX6-NEXT: v_max_i32_e32 v0, v0, v4 ; GFX6-NEXT: v_sub_i32_e32 v4, vcc, 0, v1 @@ -233,14 +232,11 @@ define amdgpu_cs <4 x i32> @abs_vgpr_v4i32(<4 x i32> %arg) { ; GFX6-NEXT: v_max_i32_e32 v2, v2, v4 ; GFX6-NEXT: v_sub_i32_e32 v4, vcc, 0, v3 ; GFX6-NEXT: v_max_i32_e32 v3, v3, v4 -; GFX6-NEXT: v_readfirstlane_b32 s0, v0 -; GFX6-NEXT: v_readfirstlane_b32 s1, v1 -; GFX6-NEXT: v_readfirstlane_b32 s2, v2 -; GFX6-NEXT: v_readfirstlane_b32 s3, v3 -; GFX6-NEXT: ; return to shader part epilog +; GFX6-NEXT: s_setpc_b64 s[30:31] ; ; GFX8-LABEL: abs_vgpr_v4i32: ; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX8-NEXT: v_sub_u32_e32 v4, vcc, 0, v0 ; GFX8-NEXT: v_max_i32_e32 v0, v0, v4 ; GFX8-NEXT: v_sub_u32_e32 v4, vcc, 0, v1 @@ -249,14 +245,11 @@ define amdgpu_cs <4 x i32> @abs_vgpr_v4i32(<4 x i32> %arg) { ; GFX8-NEXT: v_max_i32_e32 v2, v2, v4 ; GFX8-NEXT: v_sub_u32_e32 v4, vcc, 0, v3 ; GFX8-NEXT: v_max_i32_e32 v3, v3, v4 -; GFX8-NEXT: v_readfirstlane_b32 s0, v0 -; GFX8-NEXT: v_readfirstlane_b32 s1, v1 -; GFX8-NEXT: v_readfirstlane_b32 s2, v2 -; GFX8-NEXT: v_readfirstlane_b32 s3, v3 -; GFX8-NEXT: ; return to shader part epilog +; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX10-LABEL: abs_vgpr_v4i32: ; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: v_sub_nc_u32_e32 v4, 0, v0 ; GFX10-NEXT: v_sub_nc_u32_e32 v5, 0, v1 ; GFX10-NEXT: v_sub_nc_u32_e32 v6, 0, v2 @@ -265,14 +258,12 @@ define amdgpu_cs <4 x i32> @abs_vgpr_v4i32(<4 x i32> %arg) { ; GFX10-NEXT: v_max_i32_e32 v1, v1, v5 ; GFX10-NEXT: v_max_i32_e32 v2, v2, v6 ; GFX10-NEXT: v_max_i32_e32 v3, v3, v7 -; GFX10-NEXT: v_readfirstlane_b32 s0, v0 -; GFX10-NEXT: v_readfirstlane_b32 s1, v1 -; GFX10-NEXT: v_readfirstlane_b32 s2, v2 -; GFX10-NEXT: v_readfirstlane_b32 s3, v3 -; GFX10-NEXT: ; return to shader part epilog +; GFX10-NEXT: s_setpc_b64 s[30:31] ; ; GFX1250-LABEL: abs_vgpr_v4i32: ; GFX1250: ; %bb.0: +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 ; GFX1250-NEXT: v_dual_sub_nc_u32 v4, 0, v0 :: v_dual_sub_nc_u32 v5, 0, v1 ; GFX1250-NEXT: v_dual_sub_nc_u32 v6, 0, v2 :: v_dual_sub_nc_u32 v7, 0, v3 ; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_3) @@ -281,13 +272,7 @@ define amdgpu_cs <4 x i32> @abs_vgpr_v4i32(<4 x i32> %arg) { ; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_4) ; GFX1250-NEXT: v_max_i32_e32 v2, v2, v6 ; GFX1250-NEXT: v_max_i32_e32 v3, v3, v7 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4) -; GFX1250-NEXT: v_readfirstlane_b32 s0, v0 -; GFX1250-NEXT: v_readfirstlane_b32 s1, v1 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4) -; GFX1250-NEXT: v_readfirstlane_b32 s2, v2 -; GFX1250-NEXT: v_readfirstlane_b32 s3, v3 -; GFX1250-NEXT: ; return to shader part epilog +; GFX1250-NEXT: s_set_pc_i64 s[30:31] %res = call <4 x i32> @llvm.abs.v4i32(<4 x i32> %arg, i1 false) ret <4 x i32> %res } @@ -304,44 +289,43 @@ define amdgpu_cs <2 x i8> @abs_sgpr_v2i8(<2 x i8> inreg %arg) { ret <2 x i8> %res } -define amdgpu_cs <2 x i8> @abs_vgpr_v2i8(<2 x i8> %arg) { +define <2 x i8> @abs_vgpr_v2i8(<2 x i8> %arg) { ; GFX6-LABEL: abs_vgpr_v2i8: ; GFX6: ; %bb.0: +; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX6-NEXT: v_bfe_i32 v0, v0, 0, 8 ; GFX6-NEXT: v_sub_i32_e32 v2, vcc, 0, v0 ; GFX6-NEXT: v_bfe_i32 v1, v1, 0, 8 ; GFX6-NEXT: v_max_i32_e32 v0, v0, v2 ; GFX6-NEXT: v_sub_i32_e32 v2, vcc, 0, v1 ; GFX6-NEXT: v_max_i32_e32 v1, v1, v2 -; GFX6-NEXT: v_readfirstlane_b32 s0, v0 -; GFX6-NEXT: v_readfirstlane_b32 s1, v1 -; GFX6-NEXT: ; return to shader part epilog +; GFX6-NEXT: s_setpc_b64 s[30:31] ; ; GFX8-LABEL: abs_vgpr_v2i8: ; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX8-NEXT: v_mov_b32_e32 v2, 0 ; GFX8-NEXT: v_sub_u16_sdwa v3, v2, sext(v0) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX8-NEXT: v_sub_u16_sdwa v2, v2, sext(v1) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX8-NEXT: v_max_i16_sdwa v0, sext(v0), v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; GFX8-NEXT: v_max_i16_sdwa v1, sext(v1), v2 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD -; GFX8-NEXT: v_readfirstlane_b32 s0, v0 -; GFX8-NEXT: v_readfirstlane_b32 s1, v1 -; GFX8-NEXT: ; return to shader part epilog +; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX10-LABEL: abs_vgpr_v2i8: ; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: v_bfe_i32 v0, v0, 0, 8 ; GFX10-NEXT: v_bfe_i32 v1, v1, 0, 8 ; GFX10-NEXT: v_sub_nc_u16 v2, 0, v0 ; GFX10-NEXT: v_sub_nc_u16 v3, 0, v1 ; GFX10-NEXT: v_max_i16 v0, v0, v2 ; GFX10-NEXT: v_max_i16 v1, v1, v3 -; GFX10-NEXT: v_readfirstlane_b32 s0, v0 -; GFX10-NEXT: v_readfirstlane_b32 s1, v1 -; GFX10-NEXT: ; return to shader part epilog +; GFX10-NEXT: s_setpc_b64 s[30:31] ; ; GFX1250-LABEL: abs_vgpr_v2i8: ; GFX1250: ; %bb.0: +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 ; GFX1250-NEXT: v_bfe_i32 v0, v0, 0, 8 ; GFX1250-NEXT: v_bfe_i32 v1, v1, 0, 8 ; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) @@ -350,10 +334,7 @@ define amdgpu_cs <2 x i8> @abs_vgpr_v2i8(<2 x i8> %arg) { ; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) ; GFX1250-NEXT: v_max_i16 v0, v0, v2 ; GFX1250-NEXT: v_max_i16 v1, v1, v3 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX1250-NEXT: v_readfirstlane_b32 s0, v0 -; GFX1250-NEXT: v_readfirstlane_b32 s1, v1 -; GFX1250-NEXT: ; return to shader part epilog +; GFX1250-NEXT: s_set_pc_i64 s[30:31] %res = call <2 x i8> @llvm.abs.v2i8(<2 x i8> %arg, i1 false) ret <2 x i8> %res } @@ -372,9 +353,10 @@ define amdgpu_cs <3 x i8> @abs_sgpr_v3i8(<3 x i8> inreg %arg) { ret <3 x i8> %res } -define amdgpu_cs <3 x i8> @abs_vgpr_v3i8(<3 x i8> %arg) { +define <3 x i8> @abs_vgpr_v3i8(<3 x i8> %arg) { ; GFX6-LABEL: abs_vgpr_v3i8: ; GFX6: ; %bb.0: +; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX6-NEXT: v_bfe_i32 v0, v0, 0, 8 ; GFX6-NEXT: v_sub_i32_e32 v3, vcc, 0, v0 ; GFX6-NEXT: v_bfe_i32 v1, v1, 0, 8 @@ -384,13 +366,11 @@ define amdgpu_cs <3 x i8> @abs_vgpr_v3i8(<3 x i8> %arg) { ; GFX6-NEXT: v_max_i32_e32 v1, v1, v3 ; GFX6-NEXT: v_sub_i32_e32 v3, vcc, 0, v2 ; GFX6-NEXT: v_max_i32_e32 v2, v2, v3 -; GFX6-NEXT: v_readfirstlane_b32 s0, v0 -; GFX6-NEXT: v_readfirstlane_b32 s1, v1 -; GFX6-NEXT: v_readfirstlane_b32 s2, v2 -; GFX6-NEXT: ; return to shader part epilog +; GFX6-NEXT: s_setpc_b64 s[30:31] ; ; GFX8-LABEL: abs_vgpr_v3i8: ; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX8-NEXT: v_mov_b32_e32 v3, 0 ; GFX8-NEXT: v_sub_u16_sdwa v4, v3, sext(v0) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX8-NEXT: v_max_i16_sdwa v0, sext(v0), v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD @@ -398,13 +378,11 @@ define amdgpu_cs <3 x i8> @abs_vgpr_v3i8(<3 x i8> %arg) { ; GFX8-NEXT: v_sub_u16_sdwa v3, v3, sext(v2) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0 ; GFX8-NEXT: v_max_i16_sdwa v1, sext(v1), v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD ; GFX8-NEXT: v_max_i16_sdwa v2, sext(v2), v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD -; GFX8-NEXT: v_readfirstlane_b32 s0, v0 -; GFX8-NEXT: v_readfirstlane_b32 s1, v1 -; GFX8-NEXT: v_readfirstlane_b32 s2, v2 -; GFX8-NEXT: ; return to shader part epilog +; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX10-LABEL: abs_vgpr_v3i8: ; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: v_bfe_i32 v0, v0, 0, 8 ; GFX10-NEXT: v_bfe_i32 v1, v1, 0, 8 ; GFX10-NEXT: v_bfe_i32 v2, v2, 0, 8 @@ -414,13 +392,12 @@ define amdgpu_cs <3 x i8> @abs_vgpr_v3i8(<3 x i8> %arg) { ; GFX10-NEXT: v_max_i16 v0, v0, v3 ; GFX10-NEXT: v_max_i16 v1, v1, v4 ; GFX10-NEXT: v_max_i16 v2, v2, v5 -; GFX10-NEXT: v_readfirstlane_b32 s0, v0 -; GFX10-NEXT: v_readfirstlane_b32 s1, v1 -; GFX10-NEXT: v_readfirstlane_b32 s2, v2 -; GFX10-NEXT: ; return to shader part epilog +; GFX10-NEXT: s_setpc_b64 s[30:31] ; ; GFX1250-LABEL: abs_vgpr_v3i8: ; GFX1250: ; %bb.0: +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 ; GFX1250-NEXT: v_bfe_i32 v0, v0, 0, 8 ; GFX1250-NEXT: v_bfe_i32 v1, v1, 0, 8 ; GFX1250-NEXT: v_bfe_i32 v2, v2, 0, 8 @@ -433,12 +410,7 @@ define amdgpu_cs <3 x i8> @abs_vgpr_v3i8(<3 x i8> %arg) { ; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3) ; GFX1250-NEXT: v_max_i16 v1, v1, v4 ; GFX1250-NEXT: v_max_i16 v2, v2, v5 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3) -; GFX1250-NEXT: v_readfirstlane_b32 s0, v0 -; GFX1250-NEXT: v_readfirstlane_b32 s1, v1 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_3) -; GFX1250-NEXT: v_readfirstlane_b32 s2, v2 -; GFX1250-NEXT: ; return to shader part epilog +; GFX1250-NEXT: s_set_pc_i64 s[30:31] %res = call <3 x i8> @llvm.abs.v3i8(<3 x i8> %arg, i1 false) ret <3 x i8> %res } @@ -485,44 +457,44 @@ define amdgpu_cs <2 x i16> @abs_sgpr_v2i16(<2 x i16> inreg %arg) { ret <2 x i16> %res } -define amdgpu_cs <2 x i16> @abs_vgpr_v2i16(<2 x i16> %arg) { +define <2 x i16> @abs_vgpr_v2i16(<2 x i16> %arg) { ; GFX6-LABEL: abs_vgpr_v2i16: ; GFX6: ; %bb.0: +; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX6-NEXT: v_bfe_i32 v0, v0, 0, 16 ; GFX6-NEXT: v_sub_i32_e32 v2, vcc, 0, v0 ; GFX6-NEXT: v_bfe_i32 v1, v1, 0, 16 ; GFX6-NEXT: v_max_i32_e32 v0, v0, v2 ; GFX6-NEXT: v_sub_i32_e32 v2, vcc, 0, v1 ; GFX6-NEXT: v_max_i32_e32 v1, v1, v2 -; GFX6-NEXT: v_readfirstlane_b32 s0, v0 -; GFX6-NEXT: v_readfirstlane_b32 s1, v1 -; GFX6-NEXT: ; return to shader part epilog +; GFX6-NEXT: s_setpc_b64 s[30:31] ; ; GFX8-LABEL: abs_vgpr_v2i16: ; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX8-NEXT: v_mov_b32_e32 v2, 0 ; GFX8-NEXT: v_sub_u16_e32 v1, 0, v0 ; GFX8-NEXT: v_sub_u16_sdwa v2, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1 ; GFX8-NEXT: v_max_i16_e32 v1, v0, v1 ; GFX8-NEXT: v_max_i16_sdwa v0, v0, v2 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX8-NEXT: v_or_b32_e32 v0, v1, v0 -; GFX8-NEXT: v_readfirstlane_b32 s0, v0 -; GFX8-NEXT: ; return to shader part epilog +; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX10-LABEL: abs_vgpr_v2i16: ; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: v_pk_sub_i16 v1, 0, v0 ; GFX10-NEXT: v_pk_max_i16 v0, v0, v1 -; GFX10-NEXT: v_readfirstlane_b32 s0, v0 -; GFX10-NEXT: ; return to shader part epilog +; GFX10-NEXT: s_setpc_b64 s[30:31] ; ; GFX1250-LABEL: abs_vgpr_v2i16: ; GFX1250: ; %bb.0: +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 ; GFX1250-NEXT: v_pk_sub_i16 v1, 0, v0 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_1) ; GFX1250-NEXT: v_pk_max_i16 v0, v0, v1 -; GFX1250-NEXT: v_readfirstlane_b32 s0, v0 -; GFX1250-NEXT: ; return to shader part epilog +; GFX1250-NEXT: s_set_pc_i64 s[30:31] %res = call <2 x i16> @llvm.abs.v2i16(<2 x i16> %arg, i1 false) ret <2 x i16> %res } @@ -576,9 +548,10 @@ define amdgpu_cs <3 x i16> @abs_sgpr_v3i16(<3 x i16> inreg %arg) { ret <3 x i16> %res } -define amdgpu_cs <3 x i16> @abs_vgpr_v3i16(<3 x i16> %arg) { +define <3 x i16> @abs_vgpr_v3i16(<3 x i16> %arg) { ; GFX6-LABEL: abs_vgpr_v3i16: ; GFX6: ; %bb.0: +; GFX6-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX6-NEXT: v_bfe_i32 v0, v0, 0, 16 ; GFX6-NEXT: v_sub_i32_e32 v3, vcc, 0, v0 ; GFX6-NEXT: v_bfe_i32 v1, v1, 0, 16 @@ -588,13 +561,11 @@ define amdgpu_cs <3 x i16> @abs_vgpr_v3i16(<3 x i16> %arg) { ; GFX6-NEXT: v_max_i32_e32 v1, v1, v3 ; GFX6-NEXT: v_sub_i32_e32 v3, vcc, 0, v2 ; GFX6-NEXT: v_max_i32_e32 v2, v2, v3 -; GFX6-NEXT: v_readfirstlane_b32 s0, v0 -; GFX6-NEXT: v_readfirstlane_b32 s1, v1 -; GFX6-NEXT: v_readfirstlane_b32 s2, v2 -; GFX6-NEXT: ; return to shader part epilog +; GFX6-NEXT: s_setpc_b64 s[30:31] ; ; GFX8-LABEL: abs_vgpr_v3i16: ; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX8-NEXT: v_mov_b32_e32 v3, 0 ; GFX8-NEXT: v_sub_u16_e32 v2, 0, v0 ; GFX8-NEXT: v_sub_u16_sdwa v3, v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1 @@ -603,31 +574,27 @@ define amdgpu_cs <3 x i16> @abs_vgpr_v3i16(<3 x i16> %arg) { ; GFX8-NEXT: v_max_i16_sdwa v0, v0, v3 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD ; GFX8-NEXT: v_or_b32_e32 v0, v2, v0 ; GFX8-NEXT: v_max_i16_e32 v1, v1, v4 -; GFX8-NEXT: v_readfirstlane_b32 s0, v0 -; GFX8-NEXT: v_readfirstlane_b32 s1, v1 -; GFX8-NEXT: ; return to shader part epilog +; GFX8-NEXT: s_setpc_b64 s[30:31] ; ; GFX10-LABEL: abs_vgpr_v3i16: ; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) ; GFX10-NEXT: v_pk_sub_i16 v2, 0, v0 ; GFX10-NEXT: v_sub_nc_u16 v3, 0, v1 ; GFX10-NEXT: v_pk_max_i16 v0, v0, v2 ; GFX10-NEXT: v_max_i16 v1, v1, v3 -; GFX10-NEXT: v_readfirstlane_b32 s0, v0 -; GFX10-NEXT: v_readfirstlane_b32 s1, v1 -; GFX10-NEXT: ; return to shader part epilog +; GFX10-NEXT: s_setpc_b64 s[30:31] ; ; GFX1250-LABEL: abs_vgpr_v3i16: ; GFX1250: ; %bb.0: +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 ; GFX1250-NEXT: v_pk_sub_i16 v2, 0, v0 ; GFX1250-NEXT: v_sub_nc_u16 v3, 0, v1 ; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) ; GFX1250-NEXT: v_pk_max_i16 v0, v0, v2 ; GFX1250-NEXT: v_max_i16 v1, v1, v3 -; GFX1250-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX1250-NEXT: v_readfirstlane_b32 s0, v0 -; GFX1250-NEXT: v_readfirstlane_b32 s1, v1 -; GFX1250-NEXT: ; return to shader part epilog +; GFX1250-NEXT: s_set_pc_i64 s[30:31] %res = call <3 x i16> @llvm.abs.v3i16(<3 x i16> %arg, i1 false) ret <3 x i16> %res } diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.ballot.i32.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.ballot.i32.ll index 7714c03..d3e2118 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.ballot.i32.ll +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.ballot.i32.ll @@ -113,9 +113,9 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_non_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_non_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s0, 0, s0 -; CHECK-NEXT: s_cmp_eq_u32 s0, 0 +; CHECK-NEXT: s_xor_b32 s0, s0, 1 +; CHECK-NEXT: s_and_b32 s0, s0, 1 +; CHECK-NEXT: s_cmp_lg_u32 s0, 0 ; CHECK-NEXT: s_cbranch_scc1 .LBB8_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 @@ -161,16 +161,17 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_non_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_eq_zero_non_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s0, 0, s0 +; CHECK-NEXT: s_xor_b32 s0, s0, 1 +; CHECK-NEXT: s_xor_b32 s0, s0, 1 +; CHECK-NEXT: s_and_b32 s0, s0, 1 ; CHECK-NEXT: s_cmp_lg_u32 s0, 0 -; CHECK-NEXT: s_cbranch_scc0 .LBB10_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB10_3 -; CHECK-NEXT: .LBB10_2: ; %true +; CHECK-NEXT: s_cbranch_scc1 .LBB10_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB10_3 +; CHECK-NEXT: .LBB10_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB10_3 ; CHECK-NEXT: .LBB10_3: %c = trunc i32 %v to i1 %ballot = call i32 @llvm.amdgcn.ballot.i32(i1 %c) @@ -208,11 +209,7 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_cmp_lt_u32 s0, 12 -; CHECK-NEXT: s_cselect_b32 s0, 1, 0 -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s0, 0, s0 -; CHECK-NEXT: s_cmp_eq_u32 s0, 0 +; CHECK-NEXT: s_cmp_ge_u32 s0, 12 ; CHECK-NEXT: s_cbranch_scc1 .LBB12_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 @@ -258,17 +255,13 @@ define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_eq_zero_compare: ; CHECK: ; %bb.0: ; CHECK-NEXT: s_cmp_lt_u32 s0, 12 -; CHECK-NEXT: s_cselect_b32 s0, 1, 0 -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s0, 0, s0 -; CHECK-NEXT: s_cmp_lg_u32 s0, 0 -; CHECK-NEXT: s_cbranch_scc0 .LBB14_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB14_3 -; CHECK-NEXT: .LBB14_2: ; %true +; CHECK-NEXT: s_cbranch_scc1 .LBB14_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB14_3 +; CHECK-NEXT: .LBB14_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB14_3 ; CHECK-NEXT: .LBB14_3: %c = icmp ult i32 %v, 12 %ballot = call i32 @llvm.amdgcn.ballot.i32(i1 %c) @@ -310,14 +303,12 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_and(i32 inreg %v1, i32 inreg %v2) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_and: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_cmp_lt_u32 s0, 12 +; CHECK-NEXT: s_cmp_ge_u32 s0, 12 ; CHECK-NEXT: s_cselect_b32 s0, 1, 0 -; CHECK-NEXT: s_cmp_gt_u32 s1, 34 +; CHECK-NEXT: s_cmp_le_u32 s1, 34 ; CHECK-NEXT: s_cselect_b32 s1, 1, 0 -; CHECK-NEXT: s_and_b32 s0, s0, s1 -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s0, 0, s0 -; CHECK-NEXT: s_cmp_eq_u32 s0, 0 +; CHECK-NEXT: s_or_b32 s0, s0, s1 +; CHECK-NEXT: s_cmp_lg_u32 s0, 0 ; CHECK-NEXT: s_cbranch_scc1 .LBB16_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 @@ -372,16 +363,14 @@ define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_and(i32 inreg %v1, i32 inreg ; CHECK-NEXT: s_cmp_gt_u32 s1, 34 ; CHECK-NEXT: s_cselect_b32 s1, 1, 0 ; CHECK-NEXT: s_and_b32 s0, s0, s1 -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s0, 0, s0 ; CHECK-NEXT: s_cmp_lg_u32 s0, 0 -; CHECK-NEXT: s_cbranch_scc0 .LBB18_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB18_3 -; CHECK-NEXT: .LBB18_2: ; %true +; CHECK-NEXT: s_cbranch_scc1 .LBB18_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB18_3 +; CHECK-NEXT: .LBB18_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB18_3 ; CHECK-NEXT: .LBB18_3: %v1c = icmp ult i32 %v1, 12 %v2c = icmp ugt i32 %v2, 34 diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.ballot.i64.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.ballot.i64.ll index 7b81669..250fbc7 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.ballot.i64.ll +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.ballot.i64.ll @@ -116,9 +116,9 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_non_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_non_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s[0:1], 0, s0 -; CHECK-NEXT: s_cmp_eq_u64 s[0:1], 0 +; CHECK-NEXT: s_xor_b32 s0, s0, 1 +; CHECK-NEXT: s_and_b32 s0, s0, 1 +; CHECK-NEXT: s_cmp_lg_u32 s0, 0 ; CHECK-NEXT: s_cbranch_scc1 .LBB8_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 @@ -164,16 +164,17 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_non_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_eq_zero_non_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s[0:1], 0, s0 -; CHECK-NEXT: s_cmp_lg_u64 s[0:1], 0 -; CHECK-NEXT: s_cbranch_scc0 .LBB10_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB10_3 -; CHECK-NEXT: .LBB10_2: ; %true +; CHECK-NEXT: s_xor_b32 s0, s0, 1 +; CHECK-NEXT: s_xor_b32 s0, s0, 1 +; CHECK-NEXT: s_and_b32 s0, s0, 1 +; CHECK-NEXT: s_cmp_lg_u32 s0, 0 +; CHECK-NEXT: s_cbranch_scc1 .LBB10_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB10_3 +; CHECK-NEXT: .LBB10_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB10_3 ; CHECK-NEXT: .LBB10_3: %c = trunc i32 %v to i1 %ballot = call i64 @llvm.amdgcn.ballot.i64(i1 %c) @@ -211,11 +212,7 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_cmp_lt_u32 s0, 12 -; CHECK-NEXT: s_cselect_b32 s0, 1, 0 -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s[0:1], 0, s0 -; CHECK-NEXT: s_cmp_eq_u64 s[0:1], 0 +; CHECK-NEXT: s_cmp_ge_u32 s0, 12 ; CHECK-NEXT: s_cbranch_scc1 .LBB12_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 @@ -261,17 +258,13 @@ define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_eq_zero_compare: ; CHECK: ; %bb.0: ; CHECK-NEXT: s_cmp_lt_u32 s0, 12 -; CHECK-NEXT: s_cselect_b32 s0, 1, 0 -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s[0:1], 0, s0 -; CHECK-NEXT: s_cmp_lg_u64 s[0:1], 0 -; CHECK-NEXT: s_cbranch_scc0 .LBB14_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB14_3 -; CHECK-NEXT: .LBB14_2: ; %true +; CHECK-NEXT: s_cbranch_scc1 .LBB14_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB14_3 +; CHECK-NEXT: .LBB14_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB14_3 ; CHECK-NEXT: .LBB14_3: %c = icmp ult i32 %v, 12 %ballot = call i64 @llvm.amdgcn.ballot.i64(i1 %c) @@ -313,14 +306,12 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_and(i32 inreg %v1, i32 inreg %v2) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_and: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_cmp_lt_u32 s0, 12 +; CHECK-NEXT: s_cmp_ge_u32 s0, 12 ; CHECK-NEXT: s_cselect_b32 s0, 1, 0 -; CHECK-NEXT: s_cmp_gt_u32 s1, 34 +; CHECK-NEXT: s_cmp_le_u32 s1, 34 ; CHECK-NEXT: s_cselect_b32 s1, 1, 0 -; CHECK-NEXT: s_and_b32 s0, s0, s1 -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s[0:1], 0, s0 -; CHECK-NEXT: s_cmp_eq_u64 s[0:1], 0 +; CHECK-NEXT: s_or_b32 s0, s0, s1 +; CHECK-NEXT: s_cmp_lg_u32 s0, 0 ; CHECK-NEXT: s_cbranch_scc1 .LBB16_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 @@ -375,16 +366,14 @@ define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_and(i32 inreg %v1, i32 inreg ; CHECK-NEXT: s_cmp_gt_u32 s1, 34 ; CHECK-NEXT: s_cselect_b32 s1, 1, 0 ; CHECK-NEXT: s_and_b32 s0, s0, s1 -; CHECK-NEXT: s_and_b32 s0, 1, s0 -; CHECK-NEXT: v_cmp_ne_u32_e64 s[0:1], 0, s0 -; CHECK-NEXT: s_cmp_lg_u64 s[0:1], 0 -; CHECK-NEXT: s_cbranch_scc0 .LBB18_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB18_3 -; CHECK-NEXT: .LBB18_2: ; %true +; CHECK-NEXT: s_cmp_lg_u32 s0, 0 +; CHECK-NEXT: s_cbranch_scc1 .LBB18_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB18_3 +; CHECK-NEXT: .LBB18_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB18_3 ; CHECK-NEXT: .LBB18_3: %v1c = icmp ult i32 %v1, 12 %v2c = icmp ugt i32 %v2, 34 diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.mir new file mode 100644 index 0000000..097372a --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.mir @@ -0,0 +1,524 @@ +# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py +# RUN: llc -mtriple=amdgcn -mcpu=fiji -run-pass=amdgpu-regbankselect,amdgpu-regbanklegalize %s -verify-machineinstrs -o - | FileCheck %s +--- +name: add_s16_ss +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $sgpr1 + ; CHECK-LABEL: name: add_s16_ss + ; CHECK: liveins: $sgpr0, $sgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[COPY]](s32) + ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:sgpr(s16) = G_TRUNC [[COPY1]](s32) + ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:sgpr(s32) = G_ANYEXT [[TRUNC]](s16) + ; CHECK-NEXT: [[ANYEXT1:%[0-9]+]]:sgpr(s32) = G_ANYEXT [[TRUNC1]](s16) + ; CHECK-NEXT: [[ADD:%[0-9]+]]:sgpr(s32) = G_ADD [[ANYEXT]], [[ANYEXT1]] + ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:sgpr(s16) = G_TRUNC [[ADD]](s32) + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s16) = G_AND [[TRUNC2]], [[TRUNC2]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s16) = G_TRUNC %0 + %3:_(s16) = G_TRUNC %1 + %4:_(s16) = G_ADD %2, %3 + %5:_(s16) = G_AND %4, %4 +... + +--- +name: add_s16_sv +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr0 + ; CHECK-LABEL: name: add_s16_sv + ; CHECK: liveins: $sgpr0, $vgpr0 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[COPY]](s32) + ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY1]](s32) + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s16) = COPY [[TRUNC]](s16) + ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s16) = G_ADD [[COPY2]], [[TRUNC1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s16) = G_AND [[ADD]], [[ADD]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $vgpr0 + %2:_(s16) = G_TRUNC %0 + %3:_(s16) = G_TRUNC %1 + %4:_(s16) = G_ADD %2, %3 + %5:_(s16) = G_AND %4, %4 +... + +--- +name: add_s16_vs +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr0 + ; CHECK-LABEL: name: add_s16_vs + ; CHECK: liveins: $sgpr0, $vgpr0 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY]](s32) + ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:sgpr(s16) = G_TRUNC [[COPY1]](s32) + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s16) = COPY [[TRUNC1]](s16) + ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s16) = G_ADD [[TRUNC]], [[COPY2]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s16) = G_AND [[ADD]], [[ADD]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $sgpr0 + %2:_(s16) = G_TRUNC %0 + %3:_(s16) = G_TRUNC %1 + %4:_(s16) = G_ADD %2, %3 + %5:_(s16) = G_AND %4, %4 +... + +--- +name: add_s16_vv +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $vgpr1 + ; CHECK-LABEL: name: add_s16_vv + ; CHECK: liveins: $vgpr0, $vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY]](s32) + ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY1]](s32) + ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s16) = G_ADD [[TRUNC]], [[TRUNC1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s16) = G_AND [[ADD]], [[ADD]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s16) = G_TRUNC %0 + %3:_(s16) = G_TRUNC %1 + %4:_(s16) = G_ADD %2, %3 + %5:_(s16) = G_AND %4, %4 +... + +--- +name: add_s32_ss +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $sgpr1 + ; CHECK-LABEL: name: add_s32_ss + ; CHECK: liveins: $sgpr0, $sgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[ADD:%[0-9]+]]:sgpr(s32) = G_ADD [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s32) = G_AND [[ADD]], [[ADD]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s32) = G_ADD %0, %1 + %3:_(s32) = G_AND %2, %2 +... + +--- +name: add_s32_sv +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr0 + ; CHECK-LABEL: name: add_s32_sv + ; CHECK: liveins: $sgpr0, $vgpr0 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY [[COPY]](s32) + ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s32) = G_ADD [[COPY2]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[ADD]], [[ADD]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $vgpr0 + %2:_(s32) = G_ADD %0, %1 + %3:_(s32) = G_AND %2, %2 +... + +--- +name: add_s32_vs +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr0 + ; CHECK-LABEL: name: add_s32_vs + ; CHECK: liveins: $sgpr0, $vgpr0 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32) + ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s32) = G_ADD [[COPY]], [[COPY2]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[ADD]], [[ADD]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $sgpr0 + %2:_(s32) = G_ADD %0, %1 + %3:_(s32) = G_AND %2, %2 +... + +--- +name: add_s32_vv +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $vgpr1 + ; CHECK-LABEL: name: add_s32_vv + ; CHECK: liveins: $vgpr0, $vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s32) = G_ADD [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[ADD]], [[ADD]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s32) = G_ADD %0, %1 + %3:_(s32) = G_AND %2, %2 +... + +--- +name: add_s64_ss +legalized: true + +body: | + bb.0: + liveins: $sgpr0_sgpr1, $sgpr2_sgpr3 + ; CHECK-LABEL: name: add_s64_ss + ; CHECK: liveins: $sgpr0_sgpr1, $sgpr2_sgpr3 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s64) = COPY $sgpr0_sgpr1 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s64) = COPY $sgpr2_sgpr3 + ; CHECK-NEXT: [[ADD:%[0-9]+]]:sgpr(s64) = G_ADD [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s64) = G_CONSTANT i64 255 + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s64) = G_AND [[ADD]], [[ADD]] + %0:_(s64) = COPY $sgpr0_sgpr1 + %1:_(s64) = COPY $sgpr2_sgpr3 + %2:_(s64) = G_ADD %0, %1 + %3:_(s64) = G_CONSTANT i64 255 + %4:_(s64) = G_AND %2, %2 +... + +--- +name: add_s64_sv +legalized: true + +body: | + bb.0: + liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + ; CHECK-LABEL: name: add_s64_sv + ; CHECK: liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s64) = COPY $sgpr0_sgpr1 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s64) = COPY $vgpr0_vgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s64) = COPY [[COPY]](s64) + ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s64) = G_ADD [[COPY2]], [[COPY1]] + ; CHECK-NEXT: [[UV:%[0-9]+]]:vgpr(s32), [[UV1:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[ADD]](s64) + ; CHECK-NEXT: [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[ADD]](s64) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UV]], [[UV2]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:vgpr(s32) = G_AND [[UV1]], [[UV3]] + ; CHECK-NEXT: [[MV:%[0-9]+]]:vgpr(s64) = G_MERGE_VALUES [[AND]](s32), [[AND1]](s32) + %0:_(s64) = COPY $sgpr0_sgpr1 + %1:_(s64) = COPY $vgpr0_vgpr1 + %2:_(s64) = G_ADD %0, %1 + %3:_(s64) = G_AND %2, %2 +... + +--- +name: add_s64_vs +legalized: true + +body: | + bb.0: + liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + ; CHECK-LABEL: name: add_s64_vs + ; CHECK: liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s64) = COPY $vgpr0_vgpr1 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s64) = COPY $sgpr0_sgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s64) = COPY [[COPY1]](s64) + ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s64) = G_ADD [[COPY]], [[COPY2]] + ; CHECK-NEXT: [[UV:%[0-9]+]]:vgpr(s32), [[UV1:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[ADD]](s64) + ; CHECK-NEXT: [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[ADD]](s64) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UV]], [[UV2]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:vgpr(s32) = G_AND [[UV1]], [[UV3]] + ; CHECK-NEXT: [[MV:%[0-9]+]]:vgpr(s64) = G_MERGE_VALUES [[AND]](s32), [[AND1]](s32) + %0:_(s64) = COPY $vgpr0_vgpr1 + %1:_(s64) = COPY $sgpr0_sgpr1 + %2:_(s64) = G_ADD %0, %1 + %3:_(s64) = G_AND %2, %2 +... + +--- +name: add_s64_vv +legalized: true + +body: | + bb.0: + liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 + ; CHECK-LABEL: name: add_s64_vv + ; CHECK: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s64) = COPY $vgpr0_vgpr1 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s64) = COPY $vgpr2_vgpr3 + ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s64) = G_ADD [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[UV:%[0-9]+]]:vgpr(s32), [[UV1:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[ADD]](s64) + ; CHECK-NEXT: [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[ADD]](s64) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UV]], [[UV2]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:vgpr(s32) = G_AND [[UV1]], [[UV3]] + ; CHECK-NEXT: [[MV:%[0-9]+]]:vgpr(s64) = G_MERGE_VALUES [[AND]](s32), [[AND1]](s32) + %0:_(s64) = COPY $vgpr0_vgpr1 + %1:_(s64) = COPY $vgpr2_vgpr3 + %2:_(s64) = G_ADD %0, %1 + %3:_(s64) = G_AND %2, %2 +... + +--- +name: uaddo_s32_ss +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $sgpr1 + ; CHECK-LABEL: name: uaddo_s32_ss + ; CHECK: liveins: $sgpr0, $sgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[UADDO:%[0-9]+]]:sgpr(s32), [[UADDO1:%[0-9]+]]:sgpr(s32) = G_UADDO [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s32) = G_AND [[UADDO1]], [[C]] + ; CHECK-NEXT: [[C1:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0 + ; CHECK-NEXT: [[SELECT:%[0-9]+]]:sgpr(s32) = G_SELECT [[AND]](s32), [[C]], [[C1]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:sgpr(s32) = G_AND [[SELECT]], [[UADDO]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s32), %3:_(s1) = G_UADDO %0, %1 + %4:_(s32) = G_ZEXT %3 + %5:_(s32) = G_AND %4, %2 +... + +--- +name: uaddo_s32_sv +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr1 + ; CHECK-LABEL: name: uaddo_s32_sv + ; CHECK: liveins: $sgpr0, $vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY [[COPY]](s32) + ; CHECK-NEXT: [[UADDO:%[0-9]+]]:vgpr(s32), [[UADDO1:%[0-9]+]]:vcc(s1) = G_UADDO [[COPY2]], [[COPY1]] + ; CHECK-NEXT: [[C:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0 + ; CHECK-NEXT: [[SELECT:%[0-9]+]]:vgpr(s32) = G_SELECT [[UADDO1]](s1), [[C]], [[C1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UADDO]], [[SELECT]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s32), %3:_(s1) = G_UADDO %0, %1 + %4:_(s32) = G_ZEXT %3 + %5:_(s32) = G_AND %2, %4 +... + +--- +name: uaddo_s32_vs +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $sgpr1 + ; CHECK-LABEL: name: uaddo_s32_vs + ; CHECK: liveins: $vgpr0, $sgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32) + ; CHECK-NEXT: [[UADDO:%[0-9]+]]:vgpr(s32), [[UADDO1:%[0-9]+]]:vcc(s1) = G_UADDO [[COPY]], [[COPY2]] + ; CHECK-NEXT: [[C:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0 + ; CHECK-NEXT: [[SELECT:%[0-9]+]]:vgpr(s32) = G_SELECT [[UADDO1]](s1), [[C]], [[C1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UADDO]], [[SELECT]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s32), %3:_(s1) = G_UADDO %0, %1 + %4:_(s32) = G_ZEXT %3 + %5:_(s32) = G_AND %2, %4 +... + +--- +name: uaddo_s32_vv +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $vgpr1 + ; CHECK-LABEL: name: uaddo_s32_vv + ; CHECK: liveins: $vgpr0, $vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[UADDO:%[0-9]+]]:vgpr(s32), [[UADDO1:%[0-9]+]]:vcc(s1) = G_UADDO [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[C:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0 + ; CHECK-NEXT: [[SELECT:%[0-9]+]]:vgpr(s32) = G_SELECT [[UADDO1]](s1), [[C]], [[C1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UADDO]], [[SELECT]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s32), %3:_(s1) = G_UADDO %0, %1 + %4:_(s32) = G_ZEXT %3 + %5:_(s32) = G_AND %2, %4 +... + +--- +name: uadde_s32_ss +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $sgpr1, $sgpr2 + ; CHECK-LABEL: name: uadde_s32_ss + ; CHECK: liveins: $sgpr0, $sgpr1, $sgpr2 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr2 + ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s32) = G_AND [[COPY2]], [[C]] + ; CHECK-NEXT: [[UADDE:%[0-9]+]]:sgpr(s32), [[UADDE1:%[0-9]+]]:sgpr(s32) = G_UADDE [[COPY]], [[COPY1]], [[AND]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:sgpr(s32) = G_AND [[UADDE1]], [[C]] + ; CHECK-NEXT: [[C1:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0 + ; CHECK-NEXT: [[SELECT:%[0-9]+]]:sgpr(s32) = G_SELECT [[AND1]](s32), [[C]], [[C1]] + ; CHECK-NEXT: [[AND2:%[0-9]+]]:sgpr(s32) = G_AND [[UADDE]], [[SELECT]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s32) = COPY $sgpr2 + %3:_(s1) = G_TRUNC %2 + %4:_(s32), %5:_(s1) = G_UADDE %0, %1, %3 + %6:_(s32) = G_ZEXT %5 + %7:_(s32) = G_AND %4, %6 +... + +--- +name: uadde_s32_sv +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr1, $sgpr2 + ; CHECK-LABEL: name: uadde_s32_sv + ; CHECK: liveins: $sgpr0, $vgpr1, $sgpr2 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr2 + ; CHECK-NEXT: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY]](s32) + ; CHECK-NEXT: [[AMDGPU_COPY_VCC_SCC:%[0-9]+]]:vcc(s1) = G_AMDGPU_COPY_VCC_SCC [[COPY2]](s32) + ; CHECK-NEXT: [[UADDE:%[0-9]+]]:vgpr(s32), [[UADDE1:%[0-9]+]]:vcc(s1) = G_UADDE [[COPY3]], [[COPY1]], [[AMDGPU_COPY_VCC_SCC]] + ; CHECK-NEXT: [[C:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0 + ; CHECK-NEXT: [[SELECT:%[0-9]+]]:vgpr(s32) = G_SELECT [[UADDE1]](s1), [[C]], [[C1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UADDE]], [[SELECT]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s32) = COPY $sgpr2 + %3:_(s1) = G_TRUNC %2 + %4:_(s32), %5:_(s1) = G_UADDE %0, %1, %3 + %6:_(s32) = G_ZEXT %5 + %7:_(s32) = G_AND %4, %6 +... + +--- +name: uadde_s32_vs +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $sgpr1, $sgpr2 + ; CHECK-LABEL: name: uadde_s32_vs + ; CHECK: liveins: $vgpr0, $sgpr1, $sgpr2 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr2 + ; CHECK-NEXT: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32) + ; CHECK-NEXT: [[AMDGPU_COPY_VCC_SCC:%[0-9]+]]:vcc(s1) = G_AMDGPU_COPY_VCC_SCC [[COPY2]](s32) + ; CHECK-NEXT: [[UADDE:%[0-9]+]]:vgpr(s32), [[UADDE1:%[0-9]+]]:vcc(s1) = G_UADDE [[COPY]], [[COPY3]], [[AMDGPU_COPY_VCC_SCC]] + ; CHECK-NEXT: [[C:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0 + ; CHECK-NEXT: [[SELECT:%[0-9]+]]:vgpr(s32) = G_SELECT [[UADDE1]](s1), [[C]], [[C1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UADDE]], [[SELECT]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s32) = COPY $sgpr2 + %3:_(s1) = G_TRUNC %2 + %4:_(s32), %5:_(s1) = G_UADDE %0, %1, %3 + %6:_(s32) = G_ZEXT %5 + %7:_(s32) = G_AND %4, %6 +... + +--- +name: uadde_s32_vv +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $vgpr1, $vgpr2 + ; CHECK-LABEL: name: uadde_s32_vv + ; CHECK: liveins: $vgpr0, $vgpr1, $vgpr2 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr2 + ; CHECK-NEXT: [[C:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[COPY2]], [[C]] + ; CHECK-NEXT: [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0 + ; CHECK-NEXT: [[ICMP:%[0-9]+]]:vcc(s1) = G_ICMP intpred(ne), [[AND]](s32), [[C1]] + ; CHECK-NEXT: [[UADDE:%[0-9]+]]:vgpr(s32), [[UADDE1:%[0-9]+]]:vcc(s1) = G_UADDE [[COPY]], [[COPY1]], [[ICMP]] + ; CHECK-NEXT: [[SELECT:%[0-9]+]]:vgpr(s32) = G_SELECT [[UADDE1]](s1), [[C]], [[C1]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:vgpr(s32) = G_AND [[UADDE]], [[SELECT]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s32) = COPY $vgpr2 + %3:_(s1) = G_TRUNC %2 + %4:_(s32), %5:_(s1) = G_UADDE %0, %1, %3 + %6:_(s32) = G_ZEXT %5 + %7:_(s32) = G_AND %4, %6 +... + +--- +name: uadde_s32_ss_scc_use +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $sgpr1, $sgpr2 + ; CHECK-LABEL: name: uadde_s32_ss_scc_use + ; CHECK: liveins: $sgpr0, $sgpr1, $sgpr2 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr2 + ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s32) = G_AND [[COPY2]], [[C]] + ; CHECK-NEXT: [[UADDE:%[0-9]+]]:sgpr(s32), [[UADDE1:%[0-9]+]]:sgpr(s32) = G_UADDE [[COPY]], [[COPY1]], [[AND]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:sgpr(s32) = G_AND [[UADDE1]], [[C]] + ; CHECK-NEXT: [[C1:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0 + ; CHECK-NEXT: [[SELECT:%[0-9]+]]:sgpr(s32) = G_SELECT [[AND1]](s32), [[C]], [[C1]] + ; CHECK-NEXT: [[AND2:%[0-9]+]]:sgpr(s32) = G_AND [[UADDE]], [[SELECT]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s32) = COPY $sgpr2 + %3:_(s1) = G_TRUNC %2 + %4:_(s32), %5:_(s1) = G_UADDE %0, %1, %3 + %6:_(s32) = G_ZEXT %5 + %8:_(s32) = G_AND %4, %6 +... diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.s16.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.s16.mir index 54ee69f..30c958f 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.s16.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.s16.mir @@ -1,6 +1,5 @@ # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py -# RUN: llc -mtriple=amdgcn -mcpu=fiji -run-pass=regbankselect %s -verify-machineinstrs -o - -regbankselect-fast | FileCheck %s -# RUN: llc -mtriple=amdgcn -mcpu=fiji -run-pass=regbankselect %s -verify-machineinstrs -o - -regbankselect-greedy | FileCheck %s +# RUN: llc -mtriple=amdgcn -mcpu=fiji -run-pass=amdgpu-regbankselect,amdgpu-regbanklegalize %s -verify-machineinstrs -o - | FileCheck %s --- name: add_s16_ss legalized: true @@ -19,13 +18,13 @@ body: | ; CHECK-NEXT: [[ANYEXT1:%[0-9]+]]:sgpr(s32) = G_ANYEXT [[TRUNC1]](s16) ; CHECK-NEXT: [[ADD:%[0-9]+]]:sgpr(s32) = G_ADD [[ANYEXT]], [[ANYEXT1]] ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:sgpr(s16) = G_TRUNC [[ADD]](s32) - ; CHECK-NEXT: S_ENDPGM 0, implicit [[TRUNC2]](s16) + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s16) = G_AND [[TRUNC2]], [[TRUNC2]] %0:_(s32) = COPY $sgpr0 %1:_(s32) = COPY $sgpr1 %2:_(s16) = G_TRUNC %0 %3:_(s16) = G_TRUNC %1 %4:_(s16) = G_ADD %2, %3 - S_ENDPGM 0, implicit %4 + %5:_(s16) = G_AND %4, %4 ... --- @@ -44,13 +43,13 @@ body: | ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY1]](s32) ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s16) = COPY [[TRUNC]](s16) ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s16) = G_ADD [[COPY2]], [[TRUNC1]] - ; CHECK-NEXT: S_ENDPGM 0, implicit [[ADD]](s16) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s16) = G_AND [[ADD]], [[ADD]] %0:_(s32) = COPY $sgpr0 %1:_(s32) = COPY $vgpr0 %2:_(s16) = G_TRUNC %0 %3:_(s16) = G_TRUNC %1 %4:_(s16) = G_ADD %2, %3 - S_ENDPGM 0, implicit %4 + %5:_(s16) = G_AND %4, %4 ... --- @@ -69,13 +68,13 @@ body: | ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:sgpr(s16) = G_TRUNC [[COPY1]](s32) ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s16) = COPY [[TRUNC1]](s16) ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s16) = G_ADD [[TRUNC]], [[COPY2]] - ; CHECK-NEXT: S_ENDPGM 0, implicit [[ADD]](s16) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s16) = G_AND [[ADD]], [[ADD]] %0:_(s32) = COPY $vgpr0 %1:_(s32) = COPY $sgpr0 %2:_(s16) = G_TRUNC %0 %3:_(s16) = G_TRUNC %1 %4:_(s16) = G_ADD %2, %3 - S_ENDPGM 0, implicit %4 + %5:_(s16) = G_AND %4, %4 ... --- @@ -93,11 +92,11 @@ body: | ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY]](s32) ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY1]](s32) ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(s16) = G_ADD [[TRUNC]], [[TRUNC1]] - ; CHECK-NEXT: S_ENDPGM 0, implicit [[ADD]](s16) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s16) = G_AND [[ADD]], [[ADD]] %0:_(s32) = COPY $vgpr0 %1:_(s32) = COPY $vgpr1 %2:_(s16) = G_TRUNC %0 %3:_(s16) = G_TRUNC %1 %4:_(s16) = G_ADD %2, %3 - S_ENDPGM 0, implicit %4 + %5:_(s16) = G_AND %4, %4 ... diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.v2s16.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.v2s16.mir index 97018fa..01eb391 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.v2s16.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-add.v2s16.mir @@ -1,6 +1,5 @@ # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py -# RUN: llc -mtriple=amdgcn -mcpu=gfx900 -run-pass=regbankselect %s -verify-machineinstrs -o - -regbankselect-fast | FileCheck %s -# RUN: llc -mtriple=amdgcn -mcpu=gfx900 -run-pass=regbankselect %s -verify-machineinstrs -o - -regbankselect-greedy | FileCheck %s +# RUN: llc -mtriple=amdgcn -mcpu=gfx900 -run-pass=amdgpu-regbankselect,amdgpu-regbanklegalize %s -verify-machineinstrs -o - | FileCheck %s --- name: add_v2s16_ss @@ -18,16 +17,19 @@ body: | ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 16 ; CHECK-NEXT: [[LSHR:%[0-9]+]]:sgpr(s32) = G_LSHR [[BITCAST]], [[C]](s32) ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:sgpr(s32) = G_BITCAST [[COPY1]](<2 x s16>) - ; CHECK-NEXT: [[C1:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 16 - ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:sgpr(s32) = G_LSHR [[BITCAST1]], [[C1]](s32) + ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:sgpr(s32) = G_LSHR [[BITCAST1]], [[C]](s32) ; CHECK-NEXT: [[ADD:%[0-9]+]]:sgpr(s32) = G_ADD [[BITCAST]], [[BITCAST1]] ; CHECK-NEXT: [[ADD1:%[0-9]+]]:sgpr(s32) = G_ADD [[LSHR]], [[LSHR1]] ; CHECK-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:sgpr(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[ADD]](s32), [[ADD1]](s32) - ; CHECK-NEXT: S_ENDPGM 0, implicit [[BUILD_VECTOR_TRUNC]](<2 x s16>) + ; CHECK-NEXT: [[C1:%[0-9]+]]:sgpr(s16) = G_CONSTANT i16 255 + ; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:sgpr(<2 x s16>) = G_BUILD_VECTOR [[C1]](s16), [[C1]](s16) + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(<2 x s16>) = G_AND [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR]] %0:_(<2 x s16>) = COPY $sgpr0 %1:_(<2 x s16>) = COPY $sgpr1 %2:_(<2 x s16>) = G_ADD %0, %1 - S_ENDPGM 0, implicit %2 + %3:_(s16) = G_CONSTANT i16 255 + %4:_(<2 x s16>) = G_BUILD_VECTOR %3, %3 + %5:_(<2 x s16>) = G_AND %2, %4 ... --- @@ -44,11 +46,11 @@ body: | ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(<2 x s16>) = COPY $vgpr0 ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(<2 x s16>) = COPY [[COPY]](<2 x s16>) ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(<2 x s16>) = G_ADD [[COPY2]], [[COPY1]] - ; CHECK-NEXT: S_ENDPGM 0, implicit [[ADD]](<2 x s16>) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(<2 x s16>) = G_AND [[ADD]], [[ADD]] %0:_(<2 x s16>) = COPY $sgpr0 %1:_(<2 x s16>) = COPY $vgpr0 %2:_(<2 x s16>) = G_ADD %0, %1 - S_ENDPGM 0, implicit %2 + %3:_(<2 x s16>) = G_AND %2, %2 ... --- @@ -65,9 +67,11 @@ body: | ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(<2 x s16>) = COPY $sgpr0 ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(<2 x s16>) = COPY [[COPY1]](<2 x s16>) ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(<2 x s16>) = G_ADD [[COPY]], [[COPY2]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(<2 x s16>) = G_AND [[ADD]], [[ADD]] %0:_(<2 x s16>) = COPY $vgpr0 %1:_(<2 x s16>) = COPY $sgpr0 %2:_(<2 x s16>) = G_ADD %0, %1 + %3:_(<2 x s16>) = G_AND %2, %2 ... --- @@ -83,9 +87,9 @@ body: | ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(<2 x s16>) = COPY $vgpr0 ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(<2 x s16>) = COPY $vgpr1 ; CHECK-NEXT: [[ADD:%[0-9]+]]:vgpr(<2 x s16>) = G_ADD [[COPY]], [[COPY1]] - ; CHECK-NEXT: S_ENDPGM 0, implicit [[ADD]](<2 x s16>) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(<2 x s16>) = G_AND [[ADD]], [[ADD]] %0:_(<2 x s16>) = COPY $vgpr0 %1:_(<2 x s16>) = COPY $vgpr1 %2:_(<2 x s16>) = G_ADD %0, %1 - S_ENDPGM 0, implicit %2 + %3:_(<2 x s16>) = G_AND %2, %2 ... diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sext.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sext.mir index 7378c93..e0e783e 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sext.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sext.mir @@ -77,10 +77,14 @@ body: | ; CHECK-NEXT: [[C2:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0 ; CHECK-NEXT: [[SELECT:%[0-9]+]]:sgpr(s32) = G_SELECT [[AND]](s32), [[C1]], [[C2]] ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[SELECT]](s32) + ; CHECK-NEXT: [[C3:%[0-9]+]]:sgpr(s16) = G_CONSTANT i16 255 + ; CHECK-NEXT: [[AND1:%[0-9]+]]:sgpr(s16) = G_AND [[TRUNC]], [[C3]] %0:_(s32) = COPY $sgpr0 %1:_(s32) = COPY $sgpr1 %2:_(s1) = G_ICMP intpred(eq), %0, %1 %3:_(s16) = G_SEXT %2 + %4:_(s16) = G_CONSTANT i16 255 + %5:_(s16) = G_AND %3, %4 ... --- @@ -215,9 +219,13 @@ body: | ; CHECK-NEXT: [[C2:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0 ; CHECK-NEXT: [[SELECT:%[0-9]+]]:sgpr(s32) = G_SELECT [[AND]](s32), [[C1]], [[C2]] ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[SELECT]](s32) + ; CHECK-NEXT: [[C3:%[0-9]+]]:sgpr(s16) = G_CONSTANT i16 255 + ; CHECK-NEXT: [[AND1:%[0-9]+]]:sgpr(s16) = G_AND [[TRUNC]], [[C3]] %0:_(s32) = COPY $sgpr0 %1:_(s1) = G_TRUNC %0 %2:_(s16) = G_SEXT %1 + %3:_(s16) = G_CONSTANT i16 255 + %4:_(s16) = G_AND %2, %3 ... --- diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sub.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sub.mir index b0199d3..e3c01c0 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sub.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-sub.mir @@ -1,5 +1,107 @@ # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py -# RUN: llc -mtriple=amdgcn -mcpu=fiji -run-pass="amdgpu-regbankselect,amdgpu-regbanklegalize" %s -o - | FileCheck %s +# RUN: llc -mtriple=amdgcn -mcpu=gfx900 -run-pass=amdgpu-regbankselect,amdgpu-regbanklegalize %s -verify-machineinstrs -o - -regbankselect-fast | FileCheck %s +# RUN: llc -mtriple=amdgcn -mcpu=gfx900 -run-pass=amdgpu-regbankselect,amdgpu-regbanklegalize %s -verify-machineinstrs -o - -regbankselect-greedy | FileCheck %s + +--- +name: sub_s16_ss +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $sgpr1 + ; CHECK-LABEL: name: sub_s16_ss + ; CHECK: liveins: $sgpr0, $sgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[COPY]](s32) + ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:sgpr(s16) = G_TRUNC [[COPY1]](s32) + ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:sgpr(s32) = G_ANYEXT [[TRUNC]](s16) + ; CHECK-NEXT: [[ANYEXT1:%[0-9]+]]:sgpr(s32) = G_ANYEXT [[TRUNC1]](s16) + ; CHECK-NEXT: [[SUB:%[0-9]+]]:sgpr(s32) = G_SUB [[ANYEXT]], [[ANYEXT1]] + ; CHECK-NEXT: [[TRUNC2:%[0-9]+]]:sgpr(s16) = G_TRUNC [[SUB]](s32) + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s16) = G_AND [[TRUNC2]], [[TRUNC2]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s16) = G_TRUNC %0 + %3:_(s16) = G_TRUNC %1 + %4:_(s16) = G_SUB %2, %3 + %6:_(s16) = G_AND %4, %4 +... + +--- +name: sub_s16_sv +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr0 + ; CHECK-LABEL: name: sub_s16_sv + ; CHECK: liveins: $sgpr0, $vgpr0 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[COPY]](s32) + ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY1]](s32) + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s16) = COPY [[TRUNC]](s16) + ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(s16) = G_SUB [[COPY2]], [[TRUNC1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s16) = G_AND [[SUB]], [[SUB]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $vgpr0 + %2:_(s16) = G_TRUNC %0 + %3:_(s16) = G_TRUNC %1 + %4:_(s16) = G_SUB %2, %3 + %6:_(s16) = G_AND %4, %4 +... + +--- +name: sub_s16_vs +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr0 + ; CHECK-LABEL: name: sub_s16_vs + ; CHECK: liveins: $sgpr0, $vgpr0 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY]](s32) + ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:sgpr(s16) = G_TRUNC [[COPY1]](s32) + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s16) = COPY [[TRUNC1]](s16) + ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(s16) = G_SUB [[TRUNC]], [[COPY2]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s16) = G_AND [[SUB]], [[SUB]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $sgpr0 + %2:_(s16) = G_TRUNC %0 + %3:_(s16) = G_TRUNC %1 + %4:_(s16) = G_SUB %2, %3 + %6:_(s16) = G_AND %4, %4 +... + +--- +name: sub_s16_vv +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $vgpr1 + ; CHECK-LABEL: name: sub_s16_vv + ; CHECK: liveins: $vgpr0, $vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY]](s32) + ; CHECK-NEXT: [[TRUNC1:%[0-9]+]]:vgpr(s16) = G_TRUNC [[COPY1]](s32) + ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(s16) = G_SUB [[TRUNC]], [[TRUNC1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s16) = G_AND [[SUB]], [[SUB]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s16) = G_TRUNC %0 + %3:_(s16) = G_TRUNC %1 + %4:_(s16) = G_SUB %2, %3 + %6:_(s16) = G_AND %4, %4 +... --- name: sub_s32_ss @@ -14,9 +116,11 @@ body: | ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 ; CHECK-NEXT: [[SUB:%[0-9]+]]:sgpr(s32) = G_SUB [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s32) = G_AND [[SUB]], [[SUB]] %0:_(s32) = COPY $sgpr0 %1:_(s32) = COPY $sgpr1 %2:_(s32) = G_SUB %0, %1 + %4:_(s32) = G_AND %2, %2 ... --- @@ -33,9 +137,11 @@ body: | ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY [[COPY]](s32) ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(s32) = G_SUB [[COPY2]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[SUB]], [[SUB]] %0:_(s32) = COPY $sgpr0 %1:_(s32) = COPY $vgpr0 %2:_(s32) = G_SUB %0, %1 + %4:_(s32) = G_AND %2, %2 ... --- @@ -52,9 +158,11 @@ body: | ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32) ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(s32) = G_SUB [[COPY]], [[COPY2]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[SUB]], [[SUB]] %0:_(s32) = COPY $vgpr0 %1:_(s32) = COPY $sgpr0 %2:_(s32) = G_SUB %0, %1 + %4:_(s32) = G_AND %2, %2 ... --- @@ -70,7 +178,376 @@ body: | ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(s32) = G_SUB [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[SUB]], [[SUB]] %0:_(s32) = COPY $vgpr0 %1:_(s32) = COPY $vgpr1 %2:_(s32) = G_SUB %0, %1 + %4:_(s32) = G_AND %2, %2 +... + +--- +name: sub_v2s16_ss +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $sgpr1 + ; CHECK-LABEL: name: sub_v2s16_ss + ; CHECK: liveins: $sgpr0, $sgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(<2 x s16>) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(<2 x s16>) = COPY $sgpr1 + ; CHECK-NEXT: [[BITCAST:%[0-9]+]]:sgpr(s32) = G_BITCAST [[COPY]](<2 x s16>) + ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 16 + ; CHECK-NEXT: [[LSHR:%[0-9]+]]:sgpr(s32) = G_LSHR [[BITCAST]], [[C]](s32) + ; CHECK-NEXT: [[BITCAST1:%[0-9]+]]:sgpr(s32) = G_BITCAST [[COPY1]](<2 x s16>) + ; CHECK-NEXT: [[LSHR1:%[0-9]+]]:sgpr(s32) = G_LSHR [[BITCAST1]], [[C]](s32) + ; CHECK-NEXT: [[SUB:%[0-9]+]]:sgpr(s32) = G_SUB [[BITCAST]], [[BITCAST1]] + ; CHECK-NEXT: [[SUB1:%[0-9]+]]:sgpr(s32) = G_SUB [[LSHR]], [[LSHR1]] + ; CHECK-NEXT: [[BUILD_VECTOR_TRUNC:%[0-9]+]]:sgpr(<2 x s16>) = G_BUILD_VECTOR_TRUNC [[SUB]](s32), [[SUB1]](s32) + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(<2 x s16>) = G_AND [[BUILD_VECTOR_TRUNC]], [[BUILD_VECTOR_TRUNC]] + %0:_(<2 x s16>) = COPY $sgpr0 + %1:_(<2 x s16>) = COPY $sgpr1 + %2:_(<2 x s16>) = G_SUB %0, %1 + %5:_(<2 x s16>) = G_AND %2, %2 +... + +--- +name: sub_v2s16_sv +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr0 + ; CHECK-LABEL: name: sub_v2s16_sv + ; CHECK: liveins: $sgpr0, $vgpr0 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(<2 x s16>) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(<2 x s16>) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(<2 x s16>) = COPY [[COPY]](<2 x s16>) + ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(<2 x s16>) = G_SUB [[COPY2]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(<2 x s16>) = G_AND [[SUB]], [[SUB]] + %0:_(<2 x s16>) = COPY $sgpr0 + %1:_(<2 x s16>) = COPY $vgpr0 + %2:_(<2 x s16>) = G_SUB %0, %1 + %5:_(<2 x s16>) = G_AND %2, %2 +... + +--- +name: sub_v2s16_vs +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr0 + ; CHECK-LABEL: name: sub_v2s16_vs + ; CHECK: liveins: $sgpr0, $vgpr0 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(<2 x s16>) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(<2 x s16>) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(<2 x s16>) = COPY [[COPY1]](<2 x s16>) + ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(<2 x s16>) = G_SUB [[COPY]], [[COPY2]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(<2 x s16>) = G_AND [[SUB]], [[SUB]] + %0:_(<2 x s16>) = COPY $vgpr0 + %1:_(<2 x s16>) = COPY $sgpr0 + %2:_(<2 x s16>) = G_SUB %0, %1 + %5:_(<2 x s16>) = G_AND %2, %2 +... + +--- +name: sub_v2s16_vv +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $vgpr1 + ; CHECK-LABEL: name: sub_v2s16_vv + ; CHECK: liveins: $vgpr0, $vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(<2 x s16>) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(<2 x s16>) = COPY $vgpr1 + ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(<2 x s16>) = G_SUB [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(<2 x s16>) = G_AND [[SUB]], [[SUB]] + %0:_(<2 x s16>) = COPY $vgpr0 + %1:_(<2 x s16>) = COPY $vgpr1 + %2:_(<2 x s16>) = G_SUB %0, %1 + %5:_(<2 x s16>) = G_AND %2, %2 +... + +--- +name: sub_s64_ss +legalized: true + +body: | + bb.0: + liveins: $sgpr0_sgpr1, $sgpr0_sgpr1 + ; CHECK-LABEL: name: sub_s64_ss + ; CHECK: liveins: $sgpr0_sgpr1, $sgpr0_sgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s64) = COPY $sgpr0_sgpr1 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s64) = COPY $sgpr0_sgpr1 + ; CHECK-NEXT: [[SUB:%[0-9]+]]:sgpr(s64) = G_SUB [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s64) = G_AND [[SUB]], [[SUB]] + %0:_(s64) = COPY $sgpr0_sgpr1 + %1:_(s64) = COPY $sgpr0_sgpr1 + %2:_(s64) = G_SUB %0, %1 + %4:_(s64) = G_AND %2, %2 +... + +--- +name: sub_s64_sv +legalized: true + +body: | + bb.0: + liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + ; CHECK-LABEL: name: sub_s64_sv + ; CHECK: liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s64) = COPY $sgpr0_sgpr1 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s64) = COPY $vgpr0_vgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s64) = COPY [[COPY]](s64) + ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(s64) = G_SUB [[COPY2]], [[COPY1]] + ; CHECK-NEXT: [[UV:%[0-9]+]]:vgpr(s32), [[UV1:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[SUB]](s64) + ; CHECK-NEXT: [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[SUB]](s64) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UV]], [[UV2]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:vgpr(s32) = G_AND [[UV1]], [[UV3]] + ; CHECK-NEXT: [[MV:%[0-9]+]]:vgpr(s64) = G_MERGE_VALUES [[AND]](s32), [[AND1]](s32) + %0:_(s64) = COPY $sgpr0_sgpr1 + %1:_(s64) = COPY $vgpr0_vgpr1 + %2:_(s64) = G_SUB %0, %1 + %4:_(s64) = G_AND %2, %2 +... + +--- +name: sub_s64_vs +legalized: true + +body: | + bb.0: + liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + ; CHECK-LABEL: name: sub_s64_vs + ; CHECK: liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s64) = COPY $vgpr0_vgpr1 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s64) = COPY $sgpr0_sgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s64) = COPY [[COPY1]](s64) + ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(s64) = G_SUB [[COPY]], [[COPY2]] + ; CHECK-NEXT: [[UV:%[0-9]+]]:vgpr(s32), [[UV1:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[SUB]](s64) + ; CHECK-NEXT: [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[SUB]](s64) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UV]], [[UV2]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:vgpr(s32) = G_AND [[UV1]], [[UV3]] + ; CHECK-NEXT: [[MV:%[0-9]+]]:vgpr(s64) = G_MERGE_VALUES [[AND]](s32), [[AND1]](s32) + %0:_(s64) = COPY $vgpr0_vgpr1 + %1:_(s64) = COPY $sgpr0_sgpr1 + %2:_(s64) = G_SUB %0, %1 + %4:_(s64) = G_AND %2, %2 +... + +--- +name: sub_s64_vv +legalized: true + +body: | + bb.0: + liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 + ; CHECK-LABEL: name: sub_s64_vv + ; CHECK: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s64) = COPY $vgpr0_vgpr1 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s64) = COPY $vgpr2_vgpr3 + ; CHECK-NEXT: [[SUB:%[0-9]+]]:vgpr(s64) = G_SUB [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[UV:%[0-9]+]]:vgpr(s32), [[UV1:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[SUB]](s64) + ; CHECK-NEXT: [[UV2:%[0-9]+]]:vgpr(s32), [[UV3:%[0-9]+]]:vgpr(s32) = G_UNMERGE_VALUES [[SUB]](s64) + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[UV]], [[UV2]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:vgpr(s32) = G_AND [[UV1]], [[UV3]] + ; CHECK-NEXT: [[MV:%[0-9]+]]:vgpr(s64) = G_MERGE_VALUES [[AND]](s32), [[AND1]](s32) + %0:_(s64) = COPY $vgpr0_vgpr1 + %1:_(s64) = COPY $vgpr2_vgpr3 + %2:_(s64) = G_SUB %0, %1 + %4:_(s64) = G_AND %2, %2 +... + +--- +name: usubo_s32_ss +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $sgpr1 + ; CHECK-LABEL: name: usubo_s32_ss + ; CHECK: liveins: $sgpr0, $sgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[USUBO:%[0-9]+]]:sgpr(s32), [[USUBO1:%[0-9]+]]:sgpr(s32) = G_USUBO [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s32) = G_AND [[USUBO]], [[USUBO]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s32), %3:_(s1) = G_USUBO %0, %1 + %5:_(s32) = G_AND %2, %2 +... + +--- +name: usubo_s32_sv +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr1 + ; CHECK-LABEL: name: usubo_s32_sv + ; CHECK: liveins: $sgpr0, $vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY [[COPY]](s32) + ; CHECK-NEXT: [[USUBO:%[0-9]+]]:vgpr(s32), [[USUBO1:%[0-9]+]]:vcc(s1) = G_USUBO [[COPY2]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[USUBO]], [[USUBO]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s32), %3:_(s1) = G_USUBO %0, %1 + %5:_(s32) = G_AND %2, %2 +... + +--- +name: usubo_s32_vs +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $sgpr1 + ; CHECK-LABEL: name: usubo_s32_vs + ; CHECK: liveins: $vgpr0, $sgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32) + ; CHECK-NEXT: [[USUBO:%[0-9]+]]:vgpr(s32), [[USUBO1:%[0-9]+]]:vcc(s1) = G_USUBO [[COPY]], [[COPY2]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[USUBO]], [[USUBO]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s32), %3:_(s1) = G_USUBO %0, %1 + %5:_(s32) = G_AND %2, %2 +... + +--- +name: usubo_s32_vv +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $vgpr1 + ; CHECK-LABEL: name: usubo_s32_vv + ; CHECK: liveins: $vgpr0, $vgpr1 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[USUBO:%[0-9]+]]:vgpr(s32), [[USUBO1:%[0-9]+]]:vcc(s1) = G_USUBO [[COPY]], [[COPY1]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[USUBO]], [[USUBO]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s32), %3:_(s1) = G_USUBO %0, %1 + %5:_(s32) = G_AND %2, %2 +... + +--- +name: usube_s32_ss +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $sgpr1, $sgpr2 + ; CHECK-LABEL: name: usube_s32_ss + ; CHECK: liveins: $sgpr0, $sgpr1, $sgpr2 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr2 + ; CHECK-NEXT: [[C:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[AND:%[0-9]+]]:sgpr(s32) = G_AND [[COPY2]], [[C]] + ; CHECK-NEXT: [[USUBE:%[0-9]+]]:sgpr(s32), [[USUBE1:%[0-9]+]]:sgpr(s32) = G_USUBE [[COPY]], [[COPY1]], [[AND]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:sgpr(s32) = G_AND [[USUBE]], [[USUBE]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s32) = COPY $sgpr2 + %3:_(s1) = G_TRUNC %2 + %4:_(s32), %5:_(s1) = G_USUBE %0, %1, %3 + %7:_(s32) = G_AND %4, %4 +... + +--- +name: usube_s32_sv +legalized: true + +body: | + bb.0: + liveins: $sgpr0, $vgpr1, $sgpr2 + ; CHECK-LABEL: name: usube_s32_sv + ; CHECK: liveins: $sgpr0, $vgpr1, $sgpr2 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:sgpr(s32) = COPY $sgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr2 + ; CHECK-NEXT: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY]](s32) + ; CHECK-NEXT: [[AMDGPU_COPY_VCC_SCC:%[0-9]+]]:vcc(s1) = G_AMDGPU_COPY_VCC_SCC [[COPY2]](s32) + ; CHECK-NEXT: [[USUBE:%[0-9]+]]:vgpr(s32), [[USUBE1:%[0-9]+]]:vcc(s1) = G_USUBE [[COPY3]], [[COPY1]], [[AMDGPU_COPY_VCC_SCC]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[USUBE]], [[USUBE]] + %0:_(s32) = COPY $sgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s32) = COPY $sgpr2 + %3:_(s1) = G_TRUNC %2 + %4:_(s32), %5:_(s1) = G_USUBE %0, %1, %3 + %7:_(s32) = G_AND %4, %4 +... + +--- +name: usube_s32_vs +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $sgpr1, $sgpr2 + ; CHECK-LABEL: name: usube_s32_vs + ; CHECK: liveins: $vgpr0, $sgpr1, $sgpr2 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:sgpr(s32) = COPY $sgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sgpr(s32) = COPY $sgpr2 + ; CHECK-NEXT: [[COPY3:%[0-9]+]]:vgpr(s32) = COPY [[COPY1]](s32) + ; CHECK-NEXT: [[AMDGPU_COPY_VCC_SCC:%[0-9]+]]:vcc(s1) = G_AMDGPU_COPY_VCC_SCC [[COPY2]](s32) + ; CHECK-NEXT: [[USUBE:%[0-9]+]]:vgpr(s32), [[USUBE1:%[0-9]+]]:vcc(s1) = G_USUBE [[COPY]], [[COPY3]], [[AMDGPU_COPY_VCC_SCC]] + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[USUBE]], [[USUBE]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $sgpr1 + %2:_(s32) = COPY $sgpr2 + %3:_(s1) = G_TRUNC %2 + %4:_(s32), %5:_(s1) = G_USUBE %0, %1, %3 + %7:_(s32) = G_AND %4, %4 +... + +--- +name: usube_s32_vv +legalized: true + +body: | + bb.0: + liveins: $vgpr0, $vgpr1, $vgpr2 + ; CHECK-LABEL: name: usube_s32_vv + ; CHECK: liveins: $vgpr0, $vgpr1, $vgpr2 + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr(s32) = COPY $vgpr0 + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr(s32) = COPY $vgpr1 + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:vgpr(s32) = COPY $vgpr2 + ; CHECK-NEXT: [[C:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 1 + ; CHECK-NEXT: [[AND:%[0-9]+]]:vgpr(s32) = G_AND [[COPY2]], [[C]] + ; CHECK-NEXT: [[C1:%[0-9]+]]:vgpr(s32) = G_CONSTANT i32 0 + ; CHECK-NEXT: [[ICMP:%[0-9]+]]:vcc(s1) = G_ICMP intpred(ne), [[AND]](s32), [[C1]] + ; CHECK-NEXT: [[USUBE:%[0-9]+]]:vgpr(s32), [[USUBE1:%[0-9]+]]:vcc(s1) = G_USUBE [[COPY]], [[COPY1]], [[ICMP]] + ; CHECK-NEXT: [[AND1:%[0-9]+]]:vgpr(s32) = G_AND [[USUBE]], [[USUBE]] + %0:_(s32) = COPY $vgpr0 + %1:_(s32) = COPY $vgpr1 + %2:_(s32) = COPY $vgpr2 + %3:_(s1) = G_TRUNC %2 + %4:_(s32), %5:_(s1) = G_USUBE %0, %1, %3 + %7:_(s32) = G_AND %4, %4 ... diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-zext.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-zext.mir index 088c20a3..d4baa5f 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-zext.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/regbankselect-zext.mir @@ -73,10 +73,14 @@ body: | ; CHECK-NEXT: [[C1:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0 ; CHECK-NEXT: [[SELECT:%[0-9]+]]:sgpr(s32) = G_SELECT [[AND]](s32), [[C]], [[C1]] ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[SELECT]](s32) + ; CHECK-NEXT: [[C2:%[0-9]+]]:sgpr(s16) = G_CONSTANT i16 255 + ; CHECK-NEXT: [[AND1:%[0-9]+]]:sgpr(s16) = G_AND [[TRUNC]], [[C2]] %0:_(s32) = COPY $sgpr0 %1:_(s32) = COPY $sgpr1 %2:_(s1) = G_ICMP intpred(eq), %0, %1 %3:_(s16) = G_ZEXT %2 + %4:_(s16) = G_CONSTANT i16 255 + %5:_(s16) = G_AND %3, %4 ... --- @@ -209,9 +213,13 @@ body: | ; CHECK-NEXT: [[C1:%[0-9]+]]:sgpr(s32) = G_CONSTANT i32 0 ; CHECK-NEXT: [[SELECT:%[0-9]+]]:sgpr(s32) = G_SELECT [[AND]](s32), [[C]], [[C1]] ; CHECK-NEXT: [[TRUNC:%[0-9]+]]:sgpr(s16) = G_TRUNC [[SELECT]](s32) + ; CHECK-NEXT: [[C2:%[0-9]+]]:sgpr(s16) = G_CONSTANT i16 255 + ; CHECK-NEXT: [[AND1:%[0-9]+]]:sgpr(s16) = G_AND [[TRUNC]], [[C2]] %0:_(s32) = COPY $sgpr0 %1:_(s1) = G_TRUNC %0 %2:_(s16) = G_ZEXT %1 + %3:_(s16) = G_CONSTANT i16 255 + %4:_(s16) = G_AND %2, %3 ... --- diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/sub.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/sub.ll new file mode 100644 index 0000000..8b5958d --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/sub.ll @@ -0,0 +1,535 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=hawaii < %s | FileCheck -check-prefix=GFX7 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=gfx900 < %s | FileCheck -check-prefix=GFX9 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=fiji < %s | FileCheck -check-prefix=GFX8 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=gfx1010 < %s | FileCheck -check-prefix=GFX10 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 < %s | FileCheck -check-prefix=GFX11 %s +; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn-amd-amdpal -mcpu=gfx1200 -amdgpu-enable-delay-alu=0 < %s | FileCheck -check-prefix=GFX12 %s + +define i16 @s_sub_i16(i16 inreg %a, i16 inreg %b) { +; GFX7-LABEL: s_sub_i16: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: s_sub_i32 s4, s16, s17 +; GFX7-NEXT: v_mov_b32_e32 v0, s4 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: s_sub_i16: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: s_sub_i32 s4, s16, s17 +; GFX9-NEXT: v_mov_b32_e32 v0, s4 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: s_sub_i16: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: s_sub_i32 s4, s16, s17 +; GFX8-NEXT: v_mov_b32_e32 v0, s4 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: s_sub_i16: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_sub_i32 s4, s16, s17 +; GFX10-NEXT: v_mov_b32_e32 v0, s4 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: s_sub_i16: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_sub_i32 s0, s0, s1 +; GFX11-NEXT: v_mov_b32_e32 v0, s0 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: s_sub_i16: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: s_sub_co_i32 s0, s0, s1 +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: v_mov_b32_e32 v0, s0 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = sub i16 %a, %b + ret i16 %c +} + +define i16 @v_sub_i16(i16 %a, i16 %b) { +; GFX7-LABEL: v_sub_i16: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: v_sub_i32_e32 v0, vcc, v0, v1 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: v_sub_i16: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_u16_e32 v0, v0, v1 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: v_sub_i16: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: v_sub_u16_e32 v0, v0, v1 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: v_sub_i16: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: v_sub_nc_u16 v0, v0, v1 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_sub_i16: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_nc_u16 v0.l, v0.l, v1.l +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: v_sub_i16: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: v_sub_nc_u16 v0, v0, v1 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = sub i16 %a, %b + ret i16 %c +} + +define i32 @s_sub_i32(i32 inreg %a, i32 inreg %b) { +; GFX7-LABEL: s_sub_i32: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: s_sub_i32 s4, s16, s17 +; GFX7-NEXT: v_mov_b32_e32 v0, s4 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: s_sub_i32: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: s_sub_i32 s4, s16, s17 +; GFX9-NEXT: v_mov_b32_e32 v0, s4 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: s_sub_i32: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: s_sub_i32 s4, s16, s17 +; GFX8-NEXT: v_mov_b32_e32 v0, s4 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: s_sub_i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_sub_i32 s4, s16, s17 +; GFX10-NEXT: v_mov_b32_e32 v0, s4 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: s_sub_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_sub_i32 s0, s0, s1 +; GFX11-NEXT: v_mov_b32_e32 v0, s0 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: s_sub_i32: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: s_sub_co_i32 s0, s0, s1 +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: v_mov_b32_e32 v0, s0 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = sub i32 %a, %b + ret i32 %c +} + +define i32 @v_sub_i32(i32 %a, i32 %b) { +; GFX7-LABEL: v_sub_i32: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: v_sub_i32_e32 v0, vcc, v0, v1 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: v_sub_i32: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_u32_e32 v0, v0, v1 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: v_sub_i32: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: v_sub_u32_e32 v0, vcc, v0, v1 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: v_sub_i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: v_sub_nc_u32_e32 v0, v0, v1 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_sub_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_nc_u32_e32 v0, v0, v1 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: v_sub_i32: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: v_sub_nc_u32_e32 v0, v0, v1 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = sub i32 %a, %b + ret i32 %c +} + +; TODO: Add test for s_sub_v2i16. Instruction selector currently fails +; to handle G_UNMERGE_VALUES. + +define <2 x i16> @v_sub_v2i16(<2 x i16> %a, <2 x i16> %b) { +; GFX7-LABEL: v_sub_v2i16: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: v_sub_i32_e32 v0, vcc, v0, v2 +; GFX7-NEXT: v_sub_i32_e32 v1, vcc, v1, v3 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: v_sub_v2i16: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_pk_sub_i16 v0, v0, v1 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: v_sub_v2i16: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: v_sub_u16_e32 v2, v0, v1 +; GFX8-NEXT: v_sub_u16_sdwa v0, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:WORD_1 +; GFX8-NEXT: v_or_b32_e32 v0, v2, v0 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: v_sub_v2i16: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: v_pk_sub_i16 v0, v0, v1 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_sub_v2i16: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_pk_sub_i16 v0, v0, v1 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: v_sub_v2i16: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: v_pk_sub_i16 v0, v0, v1 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = sub <2 x i16> %a, %b + ret <2 x i16> %c +} + +define i64 @s_sub_i64(i64 inreg %a, i64 inreg %b) { +; GFX7-LABEL: s_sub_i64: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: s_sub_u32 s4, s16, s18 +; GFX7-NEXT: s_subb_u32 s5, s17, s19 +; GFX7-NEXT: v_mov_b32_e32 v0, s4 +; GFX7-NEXT: v_mov_b32_e32 v1, s5 +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: s_sub_i64: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: s_sub_u32 s4, s16, s18 +; GFX9-NEXT: s_subb_u32 s5, s17, s19 +; GFX9-NEXT: v_mov_b32_e32 v0, s4 +; GFX9-NEXT: v_mov_b32_e32 v1, s5 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: s_sub_i64: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: s_sub_u32 s4, s16, s18 +; GFX8-NEXT: s_subb_u32 s5, s17, s19 +; GFX8-NEXT: v_mov_b32_e32 v0, s4 +; GFX8-NEXT: v_mov_b32_e32 v1, s5 +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: s_sub_i64: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_sub_u32 s4, s16, s18 +; GFX10-NEXT: s_subb_u32 s5, s17, s19 +; GFX10-NEXT: v_mov_b32_e32 v0, s4 +; GFX10-NEXT: v_mov_b32_e32 v1, s5 +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: s_sub_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_sub_u32 s0, s0, s2 +; GFX11-NEXT: s_subb_u32 s1, s1, s3 +; GFX11-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1 +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: s_sub_i64: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: s_sub_nc_u64 s[0:1], s[0:1], s[2:3] +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1 +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = sub i64 %a, %b + ret i64 %c +} + +define i64 @v_sub_i64(i64 %a, i64 %b) { +; GFX7-LABEL: v_sub_i64: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: v_sub_i32_e32 v0, vcc, v0, v2 +; GFX7-NEXT: v_subb_u32_e32 v1, vcc, v1, v3, vcc +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: v_sub_i64: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_co_u32_e32 v0, vcc, v0, v2 +; GFX9-NEXT: v_subb_co_u32_e32 v1, vcc, v1, v3, vcc +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: v_sub_i64: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: v_sub_u32_e32 v0, vcc, v0, v2 +; GFX8-NEXT: v_subb_u32_e32 v1, vcc, v1, v3, vcc +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: v_sub_i64: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: v_sub_co_u32 v0, vcc_lo, v0, v2 +; GFX10-NEXT: v_sub_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_sub_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_co_u32 v0, vcc_lo, v0, v2 +; GFX11-NEXT: v_sub_co_ci_u32_e64 v1, null, v1, v3, vcc_lo +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: v_sub_i64: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: v_sub_co_u32 v0, vcc_lo, v0, v2 +; GFX12-NEXT: s_wait_alu 0xfffd +; GFX12-NEXT: v_sub_co_ci_u32_e64 v1, null, v1, v3, vcc_lo +; GFX12-NEXT: s_setpc_b64 s[30:31] + %c = sub i64 %a, %b + ret i64 %c +} + +define void @s_usubo_usube(i64 inreg %a, i64 inreg %b, ptr addrspace(1) %res, ptr addrspace(1) %carry) { +; GFX7-LABEL: s_usubo_usube: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: s_sub_u32 s4, s16, s18 +; GFX7-NEXT: s_subb_u32 s5, s17, s19 +; GFX7-NEXT: v_mov_b32_e32 v4, s4 +; GFX7-NEXT: s_mov_b32 s6, 0 +; GFX7-NEXT: s_cselect_b32 s8, 1, 0 +; GFX7-NEXT: v_mov_b32_e32 v5, s5 +; GFX7-NEXT: s_mov_b32 s7, 0xf000 +; GFX7-NEXT: s_mov_b64 s[4:5], 0 +; GFX7-NEXT: buffer_store_dwordx2 v[4:5], v[0:1], s[4:7], 0 addr64 +; GFX7-NEXT: v_mov_b32_e32 v0, s8 +; GFX7-NEXT: buffer_store_dword v0, v[2:3], s[4:7], 0 addr64 +; GFX7-NEXT: s_waitcnt vmcnt(0) +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: s_usubo_usube: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: s_sub_u32 s4, s16, s18 +; GFX9-NEXT: s_subb_u32 s5, s17, s19 +; GFX9-NEXT: v_mov_b32_e32 v4, s4 +; GFX9-NEXT: s_cselect_b32 s6, 1, 0 +; GFX9-NEXT: v_mov_b32_e32 v5, s5 +; GFX9-NEXT: global_store_dwordx2 v[0:1], v[4:5], off +; GFX9-NEXT: v_mov_b32_e32 v0, s6 +; GFX9-NEXT: global_store_dword v[2:3], v0, off +; GFX9-NEXT: s_waitcnt vmcnt(0) +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: s_usubo_usube: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: s_sub_u32 s4, s16, s18 +; GFX8-NEXT: s_subb_u32 s5, s17, s19 +; GFX8-NEXT: v_mov_b32_e32 v4, s4 +; GFX8-NEXT: s_cselect_b32 s6, 1, 0 +; GFX8-NEXT: v_mov_b32_e32 v5, s5 +; GFX8-NEXT: flat_store_dwordx2 v[0:1], v[4:5] +; GFX8-NEXT: v_mov_b32_e32 v0, s6 +; GFX8-NEXT: flat_store_dword v[2:3], v0 +; GFX8-NEXT: s_waitcnt vmcnt(0) +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: s_usubo_usube: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: s_sub_u32 s4, s16, s18 +; GFX10-NEXT: s_subb_u32 s5, s17, s19 +; GFX10-NEXT: s_cselect_b32 s6, 1, 0 +; GFX10-NEXT: v_mov_b32_e32 v4, s4 +; GFX10-NEXT: v_mov_b32_e32 v5, s5 +; GFX10-NEXT: v_mov_b32_e32 v6, s6 +; GFX10-NEXT: global_store_dwordx2 v[0:1], v[4:5], off +; GFX10-NEXT: global_store_dword v[2:3], v6, off +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: s_usubo_usube: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_sub_u32 s0, s0, s2 +; GFX11-NEXT: s_subb_u32 s1, s1, s3 +; GFX11-NEXT: s_cselect_b32 s2, 1, 0 +; GFX11-NEXT: v_dual_mov_b32 v5, s1 :: v_dual_mov_b32 v4, s0 +; GFX11-NEXT: v_mov_b32_e32 v6, s2 +; GFX11-NEXT: global_store_b64 v[0:1], v[4:5], off +; GFX11-NEXT: global_store_b32 v[2:3], v6, off +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: s_usubo_usube: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: s_sub_co_u32 s0, s0, s2 +; GFX12-NEXT: s_sub_co_ci_u32 s1, s1, s3 +; GFX12-NEXT: s_cselect_b32 s2, 1, 0 +; GFX12-NEXT: s_wait_alu 0xfffe +; GFX12-NEXT: v_dual_mov_b32 v5, s1 :: v_dual_mov_b32 v4, s0 +; GFX12-NEXT: v_mov_b32_e32 v6, s2 +; GFX12-NEXT: global_store_b64 v[0:1], v[4:5], off +; GFX12-NEXT: global_store_b32 v[2:3], v6, off +; GFX12-NEXT: s_setpc_b64 s[30:31] + %usubo = call {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b) + %sub = extractvalue {i64, i1} %usubo, 0 + %of = extractvalue {i64, i1} %usubo, 1 + %of32 = select i1 %of, i32 1, i32 0 + store i64 %sub, ptr addrspace(1) %res + store i32 %of32, ptr addrspace(1) %carry + ret void +} + +define void @v_usubo_usube(i64 %a, i64 %b, ptr addrspace(1) %res, ptr addrspace(1) %carry) { +; GFX7-LABEL: v_usubo_usube: +; GFX7: ; %bb.0: +; GFX7-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX7-NEXT: v_sub_i32_e32 v0, vcc, v0, v2 +; GFX7-NEXT: v_subb_u32_e32 v1, vcc, v1, v3, vcc +; GFX7-NEXT: s_mov_b32 s6, 0 +; GFX7-NEXT: s_mov_b32 s7, 0xf000 +; GFX7-NEXT: s_mov_b64 s[4:5], 0 +; GFX7-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc +; GFX7-NEXT: buffer_store_dwordx2 v[0:1], v[4:5], s[4:7], 0 addr64 +; GFX7-NEXT: buffer_store_dword v2, v[6:7], s[4:7], 0 addr64 +; GFX7-NEXT: s_waitcnt vmcnt(0) +; GFX7-NEXT: s_setpc_b64 s[30:31] +; +; GFX9-LABEL: v_usubo_usube: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_co_u32_e32 v0, vcc, v0, v2 +; GFX9-NEXT: v_subb_co_u32_e32 v1, vcc, v1, v3, vcc +; GFX9-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc +; GFX9-NEXT: global_store_dwordx2 v[4:5], v[0:1], off +; GFX9-NEXT: global_store_dword v[6:7], v2, off +; GFX9-NEXT: s_waitcnt vmcnt(0) +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-LABEL: v_usubo_usube: +; GFX8: ; %bb.0: +; GFX8-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-NEXT: v_sub_u32_e32 v0, vcc, v0, v2 +; GFX8-NEXT: v_subb_u32_e32 v1, vcc, v1, v3, vcc +; GFX8-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc +; GFX8-NEXT: flat_store_dwordx2 v[4:5], v[0:1] +; GFX8-NEXT: flat_store_dword v[6:7], v2 +; GFX8-NEXT: s_waitcnt vmcnt(0) +; GFX8-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-LABEL: v_usubo_usube: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-NEXT: v_sub_co_u32 v0, vcc_lo, v0, v2 +; GFX10-NEXT: v_sub_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo +; GFX10-NEXT: global_store_dwordx2 v[4:5], v[0:1], off +; GFX10-NEXT: global_store_dword v[6:7], v2, off +; GFX10-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_usubo_usube: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_co_u32 v0, vcc_lo, v0, v2 +; GFX11-NEXT: v_sub_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo +; GFX11-NEXT: global_store_b64 v[4:5], v[0:1], off +; GFX11-NEXT: global_store_b32 v[6:7], v2, off +; GFX11-NEXT: s_setpc_b64 s[30:31] +; +; GFX12-LABEL: v_usubo_usube: +; GFX12: ; %bb.0: +; GFX12-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX12-NEXT: s_wait_expcnt 0x0 +; GFX12-NEXT: s_wait_samplecnt 0x0 +; GFX12-NEXT: s_wait_bvhcnt 0x0 +; GFX12-NEXT: s_wait_kmcnt 0x0 +; GFX12-NEXT: v_sub_co_u32 v0, vcc_lo, v0, v2 +; GFX12-NEXT: s_wait_alu 0xfffd +; GFX12-NEXT: v_sub_co_ci_u32_e32 v1, vcc_lo, v1, v3, vcc_lo +; GFX12-NEXT: s_wait_alu 0xfffd +; GFX12-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo +; GFX12-NEXT: global_store_b64 v[4:5], v[0:1], off +; GFX12-NEXT: global_store_b32 v[6:7], v2, off +; GFX12-NEXT: s_setpc_b64 s[30:31] + %usubo = call {i64, i1} @llvm.usub.with.overflow.i64(i64 %a, i64 %b) + %sub = extractvalue {i64, i1} %usubo, 0 + %of = extractvalue {i64, i1} %usubo, 1 + %of32 = select i1 %of, i32 1, i32 0 + store i64 %sub, ptr addrspace(1) %res + store i32 %of32, ptr addrspace(1) %carry + ret void +} diff --git a/llvm/test/CodeGen/AMDGPU/amdgpu-miscellaneous-uniform-intrinsic.ll b/llvm/test/CodeGen/AMDGPU/amdgpu-miscellaneous-uniform-intrinsic.ll new file mode 100644 index 0000000..34d4c51 --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/amdgpu-miscellaneous-uniform-intrinsic.ll @@ -0,0 +1,173 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 +; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -o - %s | FileCheck %s +define amdgpu_kernel void @readfirstlane_with_readfirstlane(ptr addrspace(1) %out) { +; CHECK-LABEL: readfirstlane_with_readfirstlane: +; CHECK: ; %bb.0: +; CHECK-NEXT: s_load_b64 s[0:1], s[4:5], 0x0 +; CHECK-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, 5 +; CHECK-NEXT: s_waitcnt lgkmcnt(0) +; CHECK-NEXT: global_store_b32 v0, v1, s[0:1] +; CHECK-NEXT: s_endpgm + %v1 = call i32 @llvm.amdgcn.readfirstlane(i32 5) + %v2 = call i32 @llvm.amdgcn.readfirstlane(i32 %v1) + store i32 %v2, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @readfirstlane_with_readlane(ptr addrspace(1) %out) { +; CHECK-LABEL: readfirstlane_with_readlane: +; CHECK: ; %bb.0: +; CHECK-NEXT: s_load_b64 s[0:1], s[4:5], 0x0 +; CHECK-NEXT: v_bfe_u32 v1, v0, 10, 10 +; CHECK-NEXT: v_and_b32_e32 v0, 0x3ff, v0 +; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) +; CHECK-NEXT: v_readfirstlane_b32 s2, v1 +; CHECK-NEXT: v_readlane_b32 s2, v0, s2 +; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_1) +; CHECK-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, s2 +; CHECK-NEXT: s_waitcnt lgkmcnt(0) +; CHECK-NEXT: global_store_b32 v0, v1, s[0:1] +; CHECK-NEXT: s_endpgm + %tidx = call i32 @llvm.amdgcn.workitem.id.x() + %tidy = call i32 @llvm.amdgcn.workitem.id.y() + %v1 = call i32 @llvm.amdgcn.readlane(i32 %tidx, i32 %tidy) + %v2 = call i32 @llvm.amdgcn.readfirstlane(i32 %v1) + store i32 %v2, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @readlane_with_firstlane(ptr addrspace(1) %out) { +; CHECK-LABEL: readlane_with_firstlane: +; CHECK: ; %bb.0: +; CHECK-NEXT: s_load_b64 s[0:1], s[4:5], 0x0 +; CHECK-NEXT: v_and_b32_e32 v0, 0x3ff, v0 +; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; CHECK-NEXT: v_readfirstlane_b32 s2, v0 +; CHECK-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, s2 +; CHECK-NEXT: s_waitcnt lgkmcnt(0) +; CHECK-NEXT: global_store_b32 v0, v1, s[0:1] +; CHECK-NEXT: s_endpgm + %tidx = call i32 @llvm.amdgcn.workitem.id.x() + %v1 = call i32 @llvm.amdgcn.readfirstlane(i32 %tidx) + %v2 = call i32 @llvm.amdgcn.readlane(i32 %v1, i32 3) + store i32 %v2, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @readlane_readlane(ptr addrspace(1) %out) { +; CHECK-LABEL: readlane_readlane: +; CHECK: ; %bb.0: +; CHECK-NEXT: s_load_b64 s[0:1], s[4:5], 0x0 +; CHECK-NEXT: v_bfe_u32 v1, v0, 10, 10 +; CHECK-NEXT: v_and_b32_e32 v0, 0x3ff, v0 +; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) +; CHECK-NEXT: v_readfirstlane_b32 s2, v1 +; CHECK-NEXT: v_readlane_b32 s2, v0, s2 +; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_1) +; CHECK-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, s2 +; CHECK-NEXT: s_waitcnt lgkmcnt(0) +; CHECK-NEXT: global_store_b32 v0, v1, s[0:1] +; CHECK-NEXT: s_endpgm + %tidx = call i32 @llvm.amdgcn.workitem.id.x() + %tidy = call i32 @llvm.amdgcn.workitem.id.y() + %v1 = call i32 @llvm.amdgcn.readlane(i32 %tidx, i32 %tidy) + %v2 = call i32 @llvm.amdgcn.readlane(i32 %v1, i32 2) + store i32 %v2, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @permlane64_uniform(ptr addrspace(1) %out, i32 %src) { +; CHECK-LABEL: permlane64_uniform: +; CHECK: ; %bb.0: +; CHECK-NEXT: s_clause 0x1 +; CHECK-NEXT: s_load_b32 s2, s[4:5], 0x8 +; CHECK-NEXT: s_load_b64 s[0:1], s[4:5], 0x0 +; CHECK-NEXT: s_waitcnt lgkmcnt(0) +; CHECK-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, s2 +; CHECK-NEXT: global_store_b32 v0, v1, s[0:1] +; CHECK-NEXT: s_endpgm + %v = call i32 @llvm.amdgcn.permlane64(i32 %src) + store i32 %v, ptr addrspace(1) %out + ret void +} + +define amdgpu_kernel void @permlane64_nonuniform(i32 addrspace(1)* %out) { +; CHECK-LABEL: permlane64_nonuniform: +; CHECK: ; %bb.0: +; CHECK-NEXT: s_load_b64 s[0:1], s[4:5], 0x0 +; CHECK-NEXT: v_and_b32_e32 v0, 0x3ff, v0 +; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_1) +; CHECK-NEXT: v_permlane64_b32 v1, v0 +; CHECK-NEXT: v_lshlrev_b32_e32 v0, 2, v0 +; CHECK-NEXT: s_waitcnt lgkmcnt(0) +; CHECK-NEXT: global_store_b32 v0, v1, s[0:1] +; CHECK-NEXT: s_endpgm + %tid = call i32 @llvm.amdgcn.workitem.id.x() + %v = call i32 @llvm.amdgcn.permlane64(i32 %tid) + %out_ptr = getelementptr i32, i32 addrspace(1)* %out, i32 %tid + store i32 %v, i32 addrspace(1)* %out_ptr + ret void +} + +define amdgpu_kernel void @permlane64_nonuniform_expression(i32 addrspace(1)* %out) { +; CHECK-LABEL: permlane64_nonuniform_expression: +; CHECK: ; %bb.0: +; CHECK-NEXT: s_load_b64 s[0:1], s[4:5], 0x0 +; CHECK-NEXT: v_and_b32_e32 v0, 0x3ff, v0 +; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2) +; CHECK-NEXT: v_add_nc_u32_e32 v1, 1, v0 +; CHECK-NEXT: v_lshlrev_b32_e32 v0, 2, v0 +; CHECK-NEXT: v_permlane64_b32 v1, v1 +; CHECK-NEXT: s_waitcnt lgkmcnt(0) +; CHECK-NEXT: global_store_b32 v0, v1, s[0:1] +; CHECK-NEXT: s_endpgm + %tid = call i32 @llvm.amdgcn.workitem.id.x() + %tid2 = add i32 %tid, 1 + %v = call i32 @llvm.amdgcn.permlane64(i32 %tid2) + %out_ptr = getelementptr i32, i32 addrspace(1)* %out, i32 %tid + store i32 %v, i32 addrspace(1)* %out_ptr + ret void +} + +define protected amdgpu_kernel void @trivial_waterfall_eq_zero(ptr addrspace(1) %out) { +; CHECK-LABEL: trivial_waterfall_eq_zero: +; CHECK: ; %bb.0: ; %entry +; CHECK-NEXT: s_load_b64 s[0:1], s[4:5], 0x0 +; CHECK-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, 5 +; CHECK-NEXT: s_mov_b32 s2, 0 +; CHECK-NEXT: s_branch .LBB7_2 +; CHECK-NEXT: .LBB7_1: ; %Flow +; CHECK-NEXT: ; in Loop: Header=BB7_2 Depth=1 +; CHECK-NEXT: s_and_not1_b32 vcc_lo, exec_lo, s2 +; CHECK-NEXT: s_mov_b32 s2, -1 +; CHECK-NEXT: s_cbranch_vccz .LBB7_4 +; CHECK-NEXT: .LBB7_2: ; %while +; CHECK-NEXT: ; =>This Inner Loop Header: Depth=1 +; CHECK-NEXT: s_and_b32 vcc_lo, exec_lo, s2 +; CHECK-NEXT: s_mov_b32 s2, -1 +; CHECK-NEXT: s_cbranch_vccnz .LBB7_1 +; CHECK-NEXT: ; %bb.3: ; %if +; CHECK-NEXT: ; in Loop: Header=BB7_2 Depth=1 +; CHECK-NEXT: s_mov_b32 s2, 0 +; CHECK-NEXT: s_waitcnt lgkmcnt(0) +; CHECK-NEXT: global_store_b32 v0, v1, s[0:1] +; CHECK-NEXT: s_branch .LBB7_1 +; CHECK-NEXT: .LBB7_4: ; %exit +; CHECK-NEXT: s_endpgm +entry: + br label %while + +while: + %done = phi i1 [ 0, %entry ], [ 1, %if ] + %not_done = xor i1 %done, true + %ballot = tail call i64 @llvm.amdgcn.ballot.i64(i1 %not_done) + %is_done = icmp eq i64 %ballot, 0 ; in this case is_done = !not_done + br i1 %is_done, label %exit, label %if + +if: + store i32 5, ptr addrspace(1) %out + br label %while + +exit: + ret void +} diff --git a/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-uniform-waterfall.ll b/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-uniform-waterfall.ll index 33ce278..c962c05 100644 --- a/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-uniform-waterfall.ll +++ b/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-uniform-waterfall.ll @@ -1,6 +1,7 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6 ; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -amdgpu-enable-uniform-intrinsic-combine=0 -O3 -S < %s | FileCheck %s -check-prefix=CURRENT-CHECK ; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -passes=amdgpu-uniform-intrinsic-combine -S < %s | FileCheck %s -check-prefix=PASS-CHECK +; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -amdgpu-uniform-intrinsic-combine -S < %s | FileCheck %s -check-prefix=PASS-CHECK ; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -O3 -S < %s | FileCheck %s -check-prefix=O3-CHECK define protected amdgpu_kernel void @trivial_waterfall_eq_zero(ptr addrspace(1) %out) { diff --git a/llvm/test/CodeGen/AMDGPU/amdgpu-uniform-intrinsic-combine.ll b/llvm/test/CodeGen/AMDGPU/amdgpu-uniform-intrinsic-combine.ll index a3e42e5..a7e828c 100644 --- a/llvm/test/CodeGen/AMDGPU/amdgpu-uniform-intrinsic-combine.ll +++ b/llvm/test/CodeGen/AMDGPU/amdgpu-uniform-intrinsic-combine.ll @@ -1,5 +1,6 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6 ; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -amdgpu-enable-uniform-intrinsic-combine=0 -O3 -S < %s | FileCheck %s -check-prefix=CURRENT-CHECK +; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -amdgpu-uniform-intrinsic-combine -S < %s | FileCheck %s -check-prefix=PASS-CHECK ; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -passes=amdgpu-uniform-intrinsic-combine -S < %s | FileCheck %s -check-prefix=PASS-CHECK ; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -passes=amdgpu-uniform-intrinsic-combine,dce -S < %s | FileCheck %s -check-prefix=DCE-CHECK diff --git a/llvm/test/CodeGen/AMDGPU/amdgpu-uniform-temporal-divergence.ll b/llvm/test/CodeGen/AMDGPU/amdgpu-uniform-temporal-divergence.ll index 2fde3e3..7929261 100644 --- a/llvm/test/CodeGen/AMDGPU/amdgpu-uniform-temporal-divergence.ll +++ b/llvm/test/CodeGen/AMDGPU/amdgpu-uniform-temporal-divergence.ll @@ -1,5 +1,6 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5 ; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -passes=amdgpu-uniform-intrinsic-combine -S < %s | FileCheck %s -check-prefix=PASS-CHECK +; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -amdgpu-uniform-intrinsic-combine -S < %s | FileCheck %s -check-prefix=PASS-CHECK ; RUN: opt -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1010 -passes=amdgpu-uniform-intrinsic-combine,instcombine,early-cse,simplifycfg -S < %s | FileCheck %s -check-prefix=COMB-CHECK ; This should not be optimized diff --git a/llvm/test/CodeGen/AMDGPU/fix-sgpr-copies-wwm.ll b/llvm/test/CodeGen/AMDGPU/fix-sgpr-copies-wwm.ll index db32135..b8f084d 100644 --- a/llvm/test/CodeGen/AMDGPU/fix-sgpr-copies-wwm.ll +++ b/llvm/test/CodeGen/AMDGPU/fix-sgpr-copies-wwm.ll @@ -4,24 +4,14 @@ define amdgpu_gs i32 @main() { ; CHECK-LABEL: main: ; CHECK: ; %bb.0: ; %bb -; CHECK-NEXT: s_bitcmp1_b32 0, 0 ; CHECK-NEXT: s_mov_b32 s0, 0 -; CHECK-NEXT: s_cselect_b32 s1, -1, 0 -; CHECK-NEXT: s_or_saveexec_b32 s2, -1 -; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, 1, s1 -; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_1) -; CHECK-NEXT: v_readfirstlane_b32 s1, v0 -; CHECK-NEXT: s_mov_b32 exec_lo, s2 -; CHECK-NEXT: s_or_b32 s0, s0, s1 -; CHECK-NEXT: s_wait_alu 0xfffe +; CHECK-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1) ; CHECK-NEXT: s_bitcmp1_b32 s0, 0 ; CHECK-NEXT: s_cselect_b32 s0, -1, 0 -; CHECK-NEXT: s_wait_alu 0xfffe ; CHECK-NEXT: s_xor_b32 s0, s0, -1 -; CHECK-NEXT: s_wait_alu 0xfffe -; CHECK-NEXT: v_cndmask_b32_e64 v1, 0, 1, s0 -; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_1) -; CHECK-NEXT: v_readfirstlane_b32 s0, v1 +; CHECK-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, 1, s0 +; CHECK-NEXT: v_readfirstlane_b32 s0, v0 ; CHECK-NEXT: s_wait_alu 0xf1ff ; CHECK-NEXT: ; return to shader part epilog bb: diff --git a/llvm/test/CodeGen/AMDGPU/llc-pipeline-npm.ll b/llvm/test/CodeGen/AMDGPU/llc-pipeline-npm.ll index 3aa3663..704ea37 100644 --- a/llvm/test/CodeGen/AMDGPU/llc-pipeline-npm.ll +++ b/llvm/test/CodeGen/AMDGPU/llc-pipeline-npm.ll @@ -9,11 +9,11 @@ ; RUN: | FileCheck -check-prefix=GCN-O3 %s -; GCN-O0: require<MachineModuleAnalysis>,require<profile-summary>,require<collector-metadata>,pre-isel-intrinsic-lowering,function(expand-large-div-rem,expand-fp<O0>),amdgpu-remove-incompatible-functions,amdgpu-printf-runtime-binding,amdgpu-lower-ctor-dtor,expand-variadics,amdgpu-always-inline,always-inline,amdgpu-export-kernel-runtime-handles,amdgpu-sw-lower-lds,amdgpu-lower-module-lds,function(atomic-expand,verify,gc-lowering,lower-constant-intrinsics,unreachableblockelim,ee-instrument<post-inline>,scalarize-masked-mem-intrin,expand-reductions,amdgpu-lower-kernel-arguments),amdgpu-lower-buffer-fat-pointers,amdgpu-lower-intrinsics,cgscc(function(lower-switch,lower-invoke,unreachableblockelim,amdgpu-unify-divergent-exit-nodes,fix-irreducible,unify-loop-exits,StructurizeCFGPass,amdgpu-annotate-uniform,si-annotate-control-flow,amdgpu-rewrite-undef-for-phi,lcssa,require<uniformity>,callbr-prepare,safe-stack,stack-protector,verify)),cgscc(function(machine-function(amdgpu-isel,si-fix-sgpr-copies,si-i1-copies,finalize-isel,localstackalloc))),require<reg-usage>,cgscc(function(machine-function(reg-usage-propagation,phi-node-elimination,two-address-instruction,regallocfast,si-fix-vgpr-copies,remove-redundant-debug-values,fixup-statepoint-caller-saved,prolog-epilog,post-ra-pseudos,si-post-ra-bundler,fentry-insert,xray-instrumentation,patchable-function,si-memory-legalizer,si-insert-waitcnts,si-late-branch-lowering,post-RA-hazard-rec,amdgpu-wait-sgpr-hazards,amdgpu-lower-vgpr-encoding,branch-relaxation,reg-usage-collector,remove-loads-into-fake-uses,live-debug-values,machine-sanmd,stack-frame-layout,verify),free-machine-function)) +; GCN-O0: require<MachineModuleAnalysis>,require<profile-summary>,require<collector-metadata>,pre-isel-intrinsic-lowering,function(expand-large-div-rem,expand-fp<O0>),amdgpu-remove-incompatible-functions,amdgpu-printf-runtime-binding,amdgpu-lower-ctor-dtor,function(amdgpu-uniform-intrinsic-combine),expand-variadics,amdgpu-always-inline,always-inline,amdgpu-export-kernel-runtime-handles,amdgpu-sw-lower-lds,amdgpu-lower-module-lds,function(atomic-expand,verify,gc-lowering,lower-constant-intrinsics,unreachableblockelim,ee-instrument<post-inline>,scalarize-masked-mem-intrin,expand-reductions,amdgpu-lower-kernel-arguments),amdgpu-lower-buffer-fat-pointers,amdgpu-lower-intrinsics,cgscc(function(lower-switch,lower-invoke,unreachableblockelim,amdgpu-unify-divergent-exit-nodes,fix-irreducible,unify-loop-exits,StructurizeCFGPass,amdgpu-annotate-uniform,si-annotate-control-flow,amdgpu-rewrite-undef-for-phi,lcssa,require<uniformity>,callbr-prepare,safe-stack,stack-protector,verify)),cgscc(function(machine-function(amdgpu-isel,si-fix-sgpr-copies,si-i1-copies,finalize-isel,localstackalloc))),require<reg-usage>,cgscc(function(machine-function(reg-usage-propagation,phi-node-elimination,two-address-instruction,regallocfast,si-fix-vgpr-copies,remove-redundant-debug-values,fixup-statepoint-caller-saved,prolog-epilog,post-ra-pseudos,si-post-ra-bundler,fentry-insert,xray-instrumentation,patchable-function,si-memory-legalizer,si-insert-waitcnts,si-late-branch-lowering,post-RA-hazard-rec,amdgpu-wait-sgpr-hazards,amdgpu-lower-vgpr-encoding,branch-relaxation,reg-usage-collector,remove-loads-into-fake-uses,live-debug-values,machine-sanmd,stack-frame-layout,verify),free-machine-function)) -; GCN-O2: require<MachineModuleAnalysis>,require<profile-summary>,require<collector-metadata>,pre-isel-intrinsic-lowering,function(expand-large-div-rem,expand-fp<O2>),amdgpu-remove-incompatible-functions,amdgpu-printf-runtime-binding,amdgpu-lower-ctor-dtor,function(amdgpu-image-intrinsic-opt),expand-variadics,amdgpu-always-inline,always-inline,amdgpu-export-kernel-runtime-handles,amdgpu-sw-lower-lds,amdgpu-lower-module-lds,function(amdgpu-atomic-optimizer,atomic-expand,amdgpu-promote-alloca,separate-const-offset-from-gep<>,slsr,early-cse<>,nary-reassociate,early-cse<>,amdgpu-codegenprepare,loop-mssa(licm<allowspeculation>),verify,loop-mssa(canon-freeze,loop-reduce),mergeicmps,expand-memcmp,gc-lowering,lower-constant-intrinsics,unreachableblockelim,consthoist,replace-with-veclib,partially-inline-libcalls,ee-instrument<post-inline>,scalarize-masked-mem-intrin,expand-reductions,early-cse<>),amdgpu-preload-kernel-arguments,function(amdgpu-lower-kernel-arguments,codegenprepare,load-store-vectorizer),amdgpu-lower-buffer-fat-pointers,amdgpu-lower-intrinsics,cgscc(function(lower-switch,lower-invoke,unreachableblockelim,flatten-cfg,sink,amdgpu-late-codegenprepare,amdgpu-unify-divergent-exit-nodes,fix-irreducible,unify-loop-exits,StructurizeCFGPass,amdgpu-annotate-uniform,si-annotate-control-flow,amdgpu-rewrite-undef-for-phi,lcssa)),amdgpu-perf-hint,cgscc(function(require<uniformity>,objc-arc-contract,callbr-prepare,safe-stack,stack-protector,verify)),cgscc(function(machine-function(amdgpu-isel,si-fix-sgpr-copies,si-i1-copies,finalize-isel,early-tailduplication,opt-phis,stack-coloring,localstackalloc,dead-mi-elimination,early-machinelicm,machine-cse,machine-sink,peephole-opt,dead-mi-elimination,si-fold-operands,gcn-dpp-combine,si-load-store-opt,si-peephole-sdwa,early-machinelicm,machine-cse,si-fold-operands,dead-mi-elimination,si-shrink-instructions))),require<reg-usage>,cgscc(function(machine-function(reg-usage-propagation,amdgpu-prepare-agpr-alloc,detect-dead-lanes,dead-mi-elimination,init-undef,process-imp-defs,unreachable-mbb-elimination,require<live-vars>,si-opt-vgpr-liverange,require<machine-loops>,phi-node-elimination,si-lower-control-flow,two-address-instruction,register-coalescer,rename-independent-subregs,amdgpu-rewrite-partial-reg-uses,machine-scheduler,amdgpu-pre-ra-optimizations,si-wqm,si-optimize-exec-masking-pre-ra,si-form-memory-clauses,amdgpu-pre-ra-long-branch-reg,greedy<sgpr>,virt-reg-rewriter<no-clear-vregs>,stack-slot-coloring,si-lower-sgpr-spills,si-pre-allocate-wwm-regs,greedy<wwm>,si-lower-wwm-copies,virt-reg-rewriter<no-clear-vregs>,amdgpu-reserve-wwm-regs,greedy<vgpr>,amdgpu-nsa-reassign,virt-reg-rewriter,amdgpu-mark-last-scratch-load,machine-cp,machinelicm,si-fix-vgpr-copies,si-optimize-exec-masking,remove-redundant-debug-values,fixup-statepoint-caller-saved,postra-machine-sink,shrink-wrap,prolog-epilog,branch-folder,tailduplication,machine-latecleanup,machine-cp,post-ra-pseudos,si-shrink-instructions,si-post-ra-bundler,postmisched,block-placement,fentry-insert,xray-instrumentation,patchable-function,gcn-create-vopd,si-memory-legalizer,si-insert-waitcnts,si-late-branch-lowering,si-pre-emit-peephole,post-RA-hazard-rec,amdgpu-wait-sgpr-hazards,amdgpu-lower-vgpr-encoding,amdgpu-insert-delay-alu,branch-relaxation,reg-usage-collector,remove-loads-into-fake-uses,live-debug-values,machine-sanmd,stack-frame-layout,verify),free-machine-function)) +; GCN-O2: require<MachineModuleAnalysis>,require<profile-summary>,require<collector-metadata>,pre-isel-intrinsic-lowering,function(expand-large-div-rem,expand-fp<O2>),amdgpu-remove-incompatible-functions,amdgpu-printf-runtime-binding,amdgpu-lower-ctor-dtor,function(amdgpu-image-intrinsic-opt,amdgpu-uniform-intrinsic-combine),expand-variadics,amdgpu-always-inline,always-inline,amdgpu-export-kernel-runtime-handles,amdgpu-sw-lower-lds,amdgpu-lower-module-lds,function(amdgpu-atomic-optimizer,atomic-expand,amdgpu-promote-alloca,separate-const-offset-from-gep<>,slsr,early-cse<>,nary-reassociate,early-cse<>,amdgpu-codegenprepare,loop-mssa(licm<allowspeculation>),verify,loop-mssa(canon-freeze,loop-reduce),mergeicmps,expand-memcmp,gc-lowering,lower-constant-intrinsics,unreachableblockelim,consthoist,replace-with-veclib,partially-inline-libcalls,ee-instrument<post-inline>,scalarize-masked-mem-intrin,expand-reductions,early-cse<>),amdgpu-preload-kernel-arguments,function(amdgpu-lower-kernel-arguments,codegenprepare,load-store-vectorizer),amdgpu-lower-buffer-fat-pointers,amdgpu-lower-intrinsics,cgscc(function(lower-switch,lower-invoke,unreachableblockelim,flatten-cfg,sink,amdgpu-late-codegenprepare,amdgpu-unify-divergent-exit-nodes,fix-irreducible,unify-loop-exits,StructurizeCFGPass,amdgpu-annotate-uniform,si-annotate-control-flow,amdgpu-rewrite-undef-for-phi,lcssa)),amdgpu-perf-hint,cgscc(function(require<uniformity>,objc-arc-contract,callbr-prepare,safe-stack,stack-protector,verify)),cgscc(function(machine-function(amdgpu-isel,si-fix-sgpr-copies,si-i1-copies,finalize-isel,early-tailduplication,opt-phis,stack-coloring,localstackalloc,dead-mi-elimination,early-machinelicm,machine-cse,machine-sink,peephole-opt,dead-mi-elimination,si-fold-operands,gcn-dpp-combine,si-load-store-opt,si-peephole-sdwa,early-machinelicm,machine-cse,si-fold-operands,dead-mi-elimination,si-shrink-instructions))),require<reg-usage>,cgscc(function(machine-function(reg-usage-propagation,amdgpu-prepare-agpr-alloc,detect-dead-lanes,dead-mi-elimination,init-undef,process-imp-defs,unreachable-mbb-elimination,require<live-vars>,si-opt-vgpr-liverange,require<machine-loops>,phi-node-elimination,si-lower-control-flow,two-address-instruction,register-coalescer,rename-independent-subregs,amdgpu-rewrite-partial-reg-uses,machine-scheduler,amdgpu-pre-ra-optimizations,si-wqm,si-optimize-exec-masking-pre-ra,si-form-memory-clauses,amdgpu-pre-ra-long-branch-reg,greedy<sgpr>,virt-reg-rewriter<no-clear-vregs>,stack-slot-coloring,si-lower-sgpr-spills,si-pre-allocate-wwm-regs,greedy<wwm>,si-lower-wwm-copies,virt-reg-rewriter<no-clear-vregs>,amdgpu-reserve-wwm-regs,greedy<vgpr>,amdgpu-nsa-reassign,virt-reg-rewriter,amdgpu-mark-last-scratch-load,machine-cp,machinelicm,si-fix-vgpr-copies,si-optimize-exec-masking,remove-redundant-debug-values,fixup-statepoint-caller-saved,postra-machine-sink,shrink-wrap,prolog-epilog,branch-folder,tailduplication,machine-latecleanup,machine-cp,post-ra-pseudos,si-shrink-instructions,si-post-ra-bundler,postmisched,block-placement,fentry-insert,xray-instrumentation,patchable-function,gcn-create-vopd,si-memory-legalizer,si-insert-waitcnts,si-late-branch-lowering,si-pre-emit-peephole,post-RA-hazard-rec,amdgpu-wait-sgpr-hazards,amdgpu-lower-vgpr-encoding,amdgpu-insert-delay-alu,branch-relaxation,reg-usage-collector,remove-loads-into-fake-uses,live-debug-values,machine-sanmd,stack-frame-layout,verify),free-machine-function)) -; GCN-O3: require<MachineModuleAnalysis>,require<profile-summary>,require<collector-metadata>,pre-isel-intrinsic-lowering,function(expand-large-div-rem,expand-fp<O3>),amdgpu-remove-incompatible-functions,amdgpu-printf-runtime-binding,amdgpu-lower-ctor-dtor,function(amdgpu-image-intrinsic-opt),expand-variadics,amdgpu-always-inline,always-inline,amdgpu-export-kernel-runtime-handles,amdgpu-sw-lower-lds,amdgpu-lower-module-lds,function(amdgpu-atomic-optimizer,atomic-expand,amdgpu-promote-alloca,separate-const-offset-from-gep<>,slsr,gvn<>,nary-reassociate,early-cse<>,amdgpu-codegenprepare,loop-mssa(licm<allowspeculation>),verify,loop-mssa(canon-freeze,loop-reduce),mergeicmps,expand-memcmp,gc-lowering,lower-constant-intrinsics,unreachableblockelim,consthoist,replace-with-veclib,partially-inline-libcalls,ee-instrument<post-inline>,scalarize-masked-mem-intrin,expand-reductions,gvn<>),amdgpu-preload-kernel-arguments,function(amdgpu-lower-kernel-arguments,codegenprepare,load-store-vectorizer),amdgpu-lower-buffer-fat-pointers,amdgpu-lower-intrinsics,cgscc(function(lower-switch,lower-invoke,unreachableblockelim,flatten-cfg,sink,amdgpu-late-codegenprepare,amdgpu-unify-divergent-exit-nodes,fix-irreducible,unify-loop-exits,StructurizeCFGPass,amdgpu-annotate-uniform,si-annotate-control-flow,amdgpu-rewrite-undef-for-phi,lcssa)),amdgpu-perf-hint,cgscc(function(require<uniformity>,objc-arc-contract,callbr-prepare,safe-stack,stack-protector,verify)),cgscc(function(machine-function(amdgpu-isel,si-fix-sgpr-copies,si-i1-copies,finalize-isel,early-tailduplication,opt-phis,stack-coloring,localstackalloc,dead-mi-elimination,early-machinelicm,machine-cse,machine-sink,peephole-opt,dead-mi-elimination,si-fold-operands,gcn-dpp-combine,si-load-store-opt,si-peephole-sdwa,early-machinelicm,machine-cse,si-fold-operands,dead-mi-elimination,si-shrink-instructions))),require<reg-usage>,cgscc(function(machine-function(reg-usage-propagation,amdgpu-prepare-agpr-alloc,detect-dead-lanes,dead-mi-elimination,init-undef,process-imp-defs,unreachable-mbb-elimination,require<live-vars>,si-opt-vgpr-liverange,require<machine-loops>,phi-node-elimination,si-lower-control-flow,two-address-instruction,register-coalescer,rename-independent-subregs,amdgpu-rewrite-partial-reg-uses,machine-scheduler,amdgpu-pre-ra-optimizations,si-wqm,si-optimize-exec-masking-pre-ra,si-form-memory-clauses,amdgpu-pre-ra-long-branch-reg,greedy<sgpr>,virt-reg-rewriter<no-clear-vregs>,stack-slot-coloring,si-lower-sgpr-spills,si-pre-allocate-wwm-regs,greedy<wwm>,si-lower-wwm-copies,virt-reg-rewriter<no-clear-vregs>,amdgpu-reserve-wwm-regs,greedy<vgpr>,amdgpu-nsa-reassign,virt-reg-rewriter,amdgpu-mark-last-scratch-load,machine-cp,machinelicm,si-fix-vgpr-copies,si-optimize-exec-masking,remove-redundant-debug-values,fixup-statepoint-caller-saved,postra-machine-sink,shrink-wrap,prolog-epilog,branch-folder,tailduplication,machine-latecleanup,machine-cp,post-ra-pseudos,si-shrink-instructions,si-post-ra-bundler,postmisched,block-placement,fentry-insert,xray-instrumentation,patchable-function,gcn-create-vopd,si-memory-legalizer,si-insert-waitcnts,si-late-branch-lowering,si-pre-emit-peephole,post-RA-hazard-rec,amdgpu-wait-sgpr-hazards,amdgpu-lower-vgpr-encoding,amdgpu-insert-delay-alu,branch-relaxation,reg-usage-collector,remove-loads-into-fake-uses,live-debug-values,machine-sanmd,stack-frame-layout,verify),free-machine-function)) +; GCN-O3: require<MachineModuleAnalysis>,require<profile-summary>,require<collector-metadata>,pre-isel-intrinsic-lowering,function(expand-large-div-rem,expand-fp<O3>),amdgpu-remove-incompatible-functions,amdgpu-printf-runtime-binding,amdgpu-lower-ctor-dtor,function(amdgpu-image-intrinsic-opt,amdgpu-uniform-intrinsic-combine),expand-variadics,amdgpu-always-inline,always-inline,amdgpu-export-kernel-runtime-handles,amdgpu-sw-lower-lds,amdgpu-lower-module-lds,function(amdgpu-atomic-optimizer,atomic-expand,amdgpu-promote-alloca,separate-const-offset-from-gep<>,slsr,gvn<>,nary-reassociate,early-cse<>,amdgpu-codegenprepare,loop-mssa(licm<allowspeculation>),verify,loop-mssa(canon-freeze,loop-reduce),mergeicmps,expand-memcmp,gc-lowering,lower-constant-intrinsics,unreachableblockelim,consthoist,replace-with-veclib,partially-inline-libcalls,ee-instrument<post-inline>,scalarize-masked-mem-intrin,expand-reductions,gvn<>),amdgpu-preload-kernel-arguments,function(amdgpu-lower-kernel-arguments,codegenprepare,load-store-vectorizer),amdgpu-lower-buffer-fat-pointers,amdgpu-lower-intrinsics,cgscc(function(lower-switch,lower-invoke,unreachableblockelim,flatten-cfg,sink,amdgpu-late-codegenprepare,amdgpu-unify-divergent-exit-nodes,fix-irreducible,unify-loop-exits,StructurizeCFGPass,amdgpu-annotate-uniform,si-annotate-control-flow,amdgpu-rewrite-undef-for-phi,lcssa)),amdgpu-perf-hint,cgscc(function(require<uniformity>,objc-arc-contract,callbr-prepare,safe-stack,stack-protector,verify)),cgscc(function(machine-function(amdgpu-isel,si-fix-sgpr-copies,si-i1-copies,finalize-isel,early-tailduplication,opt-phis,stack-coloring,localstackalloc,dead-mi-elimination,early-machinelicm,machine-cse,machine-sink,peephole-opt,dead-mi-elimination,si-fold-operands,gcn-dpp-combine,si-load-store-opt,si-peephole-sdwa,early-machinelicm,machine-cse,si-fold-operands,dead-mi-elimination,si-shrink-instructions))),require<reg-usage>,cgscc(function(machine-function(reg-usage-propagation,amdgpu-prepare-agpr-alloc,detect-dead-lanes,dead-mi-elimination,init-undef,process-imp-defs,unreachable-mbb-elimination,require<live-vars>,si-opt-vgpr-liverange,require<machine-loops>,phi-node-elimination,si-lower-control-flow,two-address-instruction,register-coalescer,rename-independent-subregs,amdgpu-rewrite-partial-reg-uses,machine-scheduler,amdgpu-pre-ra-optimizations,si-wqm,si-optimize-exec-masking-pre-ra,si-form-memory-clauses,amdgpu-pre-ra-long-branch-reg,greedy<sgpr>,virt-reg-rewriter<no-clear-vregs>,stack-slot-coloring,si-lower-sgpr-spills,si-pre-allocate-wwm-regs,greedy<wwm>,si-lower-wwm-copies,virt-reg-rewriter<no-clear-vregs>,amdgpu-reserve-wwm-regs,greedy<vgpr>,amdgpu-nsa-reassign,virt-reg-rewriter,amdgpu-mark-last-scratch-load,machine-cp,machinelicm,si-fix-vgpr-copies,si-optimize-exec-masking,remove-redundant-debug-values,fixup-statepoint-caller-saved,postra-machine-sink,shrink-wrap,prolog-epilog,branch-folder,tailduplication,machine-latecleanup,machine-cp,post-ra-pseudos,si-shrink-instructions,si-post-ra-bundler,postmisched,block-placement,fentry-insert,xray-instrumentation,patchable-function,gcn-create-vopd,si-memory-legalizer,si-insert-waitcnts,si-late-branch-lowering,si-pre-emit-peephole,post-RA-hazard-rec,amdgpu-wait-sgpr-hazards,amdgpu-lower-vgpr-encoding,amdgpu-insert-delay-alu,branch-relaxation,reg-usage-collector,remove-loads-into-fake-uses,live-debug-values,machine-sanmd,stack-frame-layout,verify),free-machine-function)) define void @empty() { ret void diff --git a/llvm/test/CodeGen/AMDGPU/llc-pipeline.ll b/llvm/test/CodeGen/AMDGPU/llc-pipeline.ll index 6e52125..ee6caab 100644 --- a/llvm/test/CodeGen/AMDGPU/llc-pipeline.ll +++ b/llvm/test/CodeGen/AMDGPU/llc-pipeline.ll @@ -31,6 +31,11 @@ ; GCN-O0-NEXT: AMDGPU Remove Incompatible Functions ; GCN-O0-NEXT: AMDGPU Printf lowering ; GCN-O0-NEXT: Lower ctors and dtors for AMDGPU +; GCN-O0-NEXT: FunctionPass Manager +; GCN-O0-NEXT: Dominator Tree Construction +; GCN-O0-NEXT: Cycle Info Analysis +; GCN-O0-NEXT: Uniformity Analysis +; GCN-O0-NEXT: AMDGPU Uniform Intrinsic Combine ; GCN-O0-NEXT: Expand variadic functions ; GCN-O0-NEXT: AMDGPU Inline All Functions ; GCN-O0-NEXT: Inliner for always_inline functions @@ -179,6 +184,11 @@ ; GCN-O1-NEXT: AMDGPU Remove Incompatible Functions ; GCN-O1-NEXT: AMDGPU Printf lowering ; GCN-O1-NEXT: Lower ctors and dtors for AMDGPU +; GCN-O1-NEXT: FunctionPass Manager +; GCN-O1-NEXT: Dominator Tree Construction +; GCN-O1-NEXT: Cycle Info Analysis +; GCN-O1-NEXT: Uniformity Analysis +; GCN-O1-NEXT: AMDGPU Uniform Intrinsic Combine ; GCN-O1-NEXT: Expand variadic functions ; GCN-O1-NEXT: AMDGPU Inline All Functions ; GCN-O1-NEXT: Inliner for always_inline functions @@ -466,6 +476,11 @@ ; GCN-O1-OPTS-NEXT: AMDGPU Remove Incompatible Functions ; GCN-O1-OPTS-NEXT: AMDGPU Printf lowering ; GCN-O1-OPTS-NEXT: Lower ctors and dtors for AMDGPU +; GCN-O1-OPTS-NEXT: FunctionPass Manager +; GCN-O1-OPTS-NEXT: Dominator Tree Construction +; GCN-O1-OPTS-NEXT: Cycle Info Analysis +; GCN-O1-OPTS-NEXT: Uniformity Analysis +; GCN-O1-OPTS-NEXT: AMDGPU Uniform Intrinsic Combine ; GCN-O1-OPTS-NEXT: Expand variadic functions ; GCN-O1-OPTS-NEXT: AMDGPU Inline All Functions ; GCN-O1-OPTS-NEXT: Inliner for always_inline functions @@ -783,6 +798,10 @@ ; GCN-O2-NEXT: Lower ctors and dtors for AMDGPU ; GCN-O2-NEXT: FunctionPass Manager ; GCN-O2-NEXT: AMDGPU Image Intrinsic Optimizer +; GCN-O2-NEXT: Dominator Tree Construction +; GCN-O2-NEXT: Cycle Info Analysis +; GCN-O2-NEXT: Uniformity Analysis +; GCN-O2-NEXT: AMDGPU Uniform Intrinsic Combine ; GCN-O2-NEXT: Expand variadic functions ; GCN-O2-NEXT: AMDGPU Inline All Functions ; GCN-O2-NEXT: Inliner for always_inline functions @@ -1104,6 +1123,10 @@ ; GCN-O3-NEXT: Lower ctors and dtors for AMDGPU ; GCN-O3-NEXT: FunctionPass Manager ; GCN-O3-NEXT: AMDGPU Image Intrinsic Optimizer +; GCN-O3-NEXT: Dominator Tree Construction +; GCN-O3-NEXT: Cycle Info Analysis +; GCN-O3-NEXT: Uniformity Analysis +; GCN-O3-NEXT: AMDGPU Uniform Intrinsic Combine ; GCN-O3-NEXT: Expand variadic functions ; GCN-O3-NEXT: AMDGPU Inline All Functions ; GCN-O3-NEXT: Inliner for always_inline functions diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ballot.i32.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ballot.i32.ll index e00e1f1..c1f3a12 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ballot.i32.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ballot.i32.ll @@ -110,9 +110,8 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_non_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_non_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_and_b32 s0, s0, 1 -; CHECK-NEXT: v_cmp_ne_u32_e64 vcc_lo, s0, 0 -; CHECK-NEXT: s_cbranch_vccz .LBB8_2 +; CHECK-NEXT: s_bitcmp0_b32 s0, 0 +; CHECK-NEXT: s_cbranch_scc1 .LBB8_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB8_3 @@ -156,15 +155,16 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_non_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_eq_zero_non_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_and_b32 s0, s0, 1 -; CHECK-NEXT: v_cmp_ne_u32_e64 vcc_lo, s0, 0 -; CHECK-NEXT: s_cbranch_vccz .LBB10_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB10_3 -; CHECK-NEXT: .LBB10_2: ; %true +; CHECK-NEXT: s_bitcmp1_b32 s0, 0 +; CHECK-NEXT: s_cselect_b32 s0, -1, 0 +; CHECK-NEXT: s_and_b32 vcc_lo, exec_lo, s0 +; CHECK-NEXT: s_cbranch_vccnz .LBB10_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB10_3 +; CHECK-NEXT: .LBB10_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB10_3 ; CHECK-NEXT: .LBB10_3: %c = trunc i32 %v to i1 %ballot = call i32 @llvm.amdgcn.ballot.i32(i1 %c) @@ -201,8 +201,8 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: v_cmp_lt_u32_e64 vcc_lo, s0, 12 -; CHECK-NEXT: s_cbranch_vccz .LBB12_2 +; CHECK-NEXT: s_cmp_gt_u32 s0, 11 +; CHECK-NEXT: s_cbranch_scc1 .LBB12_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB12_3 @@ -245,14 +245,14 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_eq_zero_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: v_cmp_lt_u32_e64 vcc_lo, s0, 12 -; CHECK-NEXT: s_cbranch_vccz .LBB14_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB14_3 -; CHECK-NEXT: .LBB14_2: ; %true +; CHECK-NEXT: s_cmp_lt_u32 s0, 12 +; CHECK-NEXT: s_cbranch_scc1 .LBB14_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB14_3 +; CHECK-NEXT: .LBB14_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB14_3 ; CHECK-NEXT: .LBB14_3: %c = icmp ult i32 %v, 12 %ballot = call i32 @llvm.amdgcn.ballot.i32(i1 %c) @@ -293,13 +293,13 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_and(i32 inreg %v1, i32 inreg %v2) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_and: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_cmp_lt_u32 s0, 12 +; CHECK-NEXT: s_cmp_gt_u32 s0, 11 ; CHECK-NEXT: s_cselect_b32 s0, -1, 0 -; CHECK-NEXT: s_cmp_gt_u32 s1, 34 +; CHECK-NEXT: s_cmp_lt_u32 s1, 35 ; CHECK-NEXT: s_cselect_b32 s1, -1, 0 -; CHECK-NEXT: s_and_b32 s0, s0, s1 -; CHECK-NEXT: s_and_b32 s0, s0, exec_lo -; CHECK-NEXT: s_cbranch_scc0 .LBB16_2 +; CHECK-NEXT: s_or_b32 s0, s0, s1 +; CHECK-NEXT: s_and_b32 vcc_lo, exec_lo, s0 +; CHECK-NEXT: s_cbranch_vccnz .LBB16_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB16_3 @@ -353,14 +353,14 @@ define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_and(i32 inreg %v1, i32 inreg ; CHECK-NEXT: s_cmp_gt_u32 s1, 34 ; CHECK-NEXT: s_cselect_b32 s1, -1, 0 ; CHECK-NEXT: s_and_b32 s0, s0, s1 -; CHECK-NEXT: s_and_b32 s0, s0, exec_lo -; CHECK-NEXT: s_cbranch_scc0 .LBB18_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB18_3 -; CHECK-NEXT: .LBB18_2: ; %true +; CHECK-NEXT: s_and_b32 vcc_lo, exec_lo, s0 +; CHECK-NEXT: s_cbranch_vccnz .LBB18_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB18_3 +; CHECK-NEXT: .LBB18_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB18_3 ; CHECK-NEXT: .LBB18_3: %v1c = icmp ult i32 %v1, 12 %v2c = icmp ugt i32 %v2, 34 @@ -591,3 +591,24 @@ exit: store i32 %ballot, ptr addrspace(1) %out ret void } + +define amdgpu_cs i32 @compare_bfloats(bfloat %x, bfloat %y) { +; GFX10-LABEL: compare_bfloats: +; GFX10: ; %bb.0: +; GFX10-NEXT: v_lshlrev_b32_e32 v1, 16, v1 +; GFX10-NEXT: v_lshlrev_b32_e32 v0, 16, v0 +; GFX10-NEXT: v_cmp_gt_f32_e64 s0, v0, v1 +; GFX10-NEXT: ; return to shader part epilog +; +; GFX11-LABEL: compare_bfloats: +; GFX11: ; %bb.0: +; GFX11-NEXT: v_mov_b16_e32 v2.l, 0 +; GFX11-NEXT: v_mov_b16_e32 v2.h, v1.l +; GFX11-NEXT: v_mov_b16_e32 v1.h, v0.l +; GFX11-NEXT: v_mov_b16_e32 v1.l, v2.l +; GFX11-NEXT: v_cmp_gt_f32_e64 s0, v1, v2 +; GFX11-NEXT: ; return to shader part epilog + %cmp = fcmp ogt bfloat %x, %y + %ballot = call i32 @llvm.amdgcn.ballot.i32(i1 %cmp) + ret i32 %ballot +} diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ballot.i64.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ballot.i64.ll index b4adf7f..827a01f 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ballot.i64.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ballot.i64.ll @@ -113,9 +113,8 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_non_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_non_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_and_b32 s0, s0, 1 -; CHECK-NEXT: v_cmp_ne_u32_e64 vcc, s0, 0 -; CHECK-NEXT: s_cbranch_vccz .LBB8_2 +; CHECK-NEXT: s_bitcmp0_b32 s0, 0 +; CHECK-NEXT: s_cbranch_scc1 .LBB8_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB8_3 @@ -159,15 +158,16 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_non_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_eq_zero_non_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_and_b32 s0, s0, 1 -; CHECK-NEXT: v_cmp_ne_u32_e64 vcc, s0, 0 -; CHECK-NEXT: s_cbranch_vccz .LBB10_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB10_3 -; CHECK-NEXT: .LBB10_2: ; %true +; CHECK-NEXT: s_bitcmp1_b32 s0, 0 +; CHECK-NEXT: s_cselect_b64 s[0:1], -1, 0 +; CHECK-NEXT: s_and_b64 vcc, exec, s[0:1] +; CHECK-NEXT: s_cbranch_vccnz .LBB10_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB10_3 +; CHECK-NEXT: .LBB10_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB10_3 ; CHECK-NEXT: .LBB10_3: %c = trunc i32 %v to i1 %ballot = call i64 @llvm.amdgcn.ballot.i64(i1 %c) @@ -204,8 +204,8 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: v_cmp_lt_u32_e64 vcc, s0, 12 -; CHECK-NEXT: s_cbranch_vccz .LBB12_2 +; CHECK-NEXT: s_cmp_gt_u32 s0, 11 +; CHECK-NEXT: s_cbranch_scc1 .LBB12_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB12_3 @@ -248,14 +248,14 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_compare(i32 inreg %v) { ; CHECK-LABEL: branch_uniform_ballot_eq_zero_compare: ; CHECK: ; %bb.0: -; CHECK-NEXT: v_cmp_lt_u32_e64 vcc, s0, 12 -; CHECK-NEXT: s_cbranch_vccz .LBB14_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB14_3 -; CHECK-NEXT: .LBB14_2: ; %true +; CHECK-NEXT: s_cmp_lt_u32 s0, 12 +; CHECK-NEXT: s_cbranch_scc1 .LBB14_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB14_3 +; CHECK-NEXT: .LBB14_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB14_3 ; CHECK-NEXT: .LBB14_3: %c = icmp ult i32 %v, 12 %ballot = call i64 @llvm.amdgcn.ballot.i64(i1 %c) @@ -296,13 +296,13 @@ false: define amdgpu_cs i32 @branch_uniform_ballot_ne_zero_and(i32 inreg %v1, i32 inreg %v2) { ; CHECK-LABEL: branch_uniform_ballot_ne_zero_and: ; CHECK: ; %bb.0: -; CHECK-NEXT: s_cmp_lt_u32 s0, 12 +; CHECK-NEXT: s_cmp_gt_u32 s0, 11 ; CHECK-NEXT: s_cselect_b64 s[2:3], -1, 0 -; CHECK-NEXT: s_cmp_gt_u32 s1, 34 +; CHECK-NEXT: s_cmp_lt_u32 s1, 35 ; CHECK-NEXT: s_cselect_b64 s[0:1], -1, 0 -; CHECK-NEXT: s_and_b64 s[0:1], s[2:3], s[0:1] -; CHECK-NEXT: s_and_b64 s[0:1], s[0:1], exec -; CHECK-NEXT: s_cbranch_scc0 .LBB16_2 +; CHECK-NEXT: s_or_b64 s[0:1], s[2:3], s[0:1] +; CHECK-NEXT: s_and_b64 vcc, exec, s[0:1] +; CHECK-NEXT: s_cbranch_vccnz .LBB16_2 ; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB16_3 @@ -356,14 +356,14 @@ define amdgpu_cs i32 @branch_uniform_ballot_eq_zero_and(i32 inreg %v1, i32 inreg ; CHECK-NEXT: s_cmp_gt_u32 s1, 34 ; CHECK-NEXT: s_cselect_b64 s[0:1], -1, 0 ; CHECK-NEXT: s_and_b64 s[0:1], s[2:3], s[0:1] -; CHECK-NEXT: s_and_b64 s[0:1], s[0:1], exec -; CHECK-NEXT: s_cbranch_scc0 .LBB18_2 -; CHECK-NEXT: ; %bb.1: ; %false -; CHECK-NEXT: s_mov_b32 s0, 33 -; CHECK-NEXT: s_branch .LBB18_3 -; CHECK-NEXT: .LBB18_2: ; %true +; CHECK-NEXT: s_and_b64 vcc, exec, s[0:1] +; CHECK-NEXT: s_cbranch_vccnz .LBB18_2 +; CHECK-NEXT: ; %bb.1: ; %true ; CHECK-NEXT: s_mov_b32 s0, 42 ; CHECK-NEXT: s_branch .LBB18_3 +; CHECK-NEXT: .LBB18_2: ; %false +; CHECK-NEXT: s_mov_b32 s0, 33 +; CHECK-NEXT: s_branch .LBB18_3 ; CHECK-NEXT: .LBB18_3: %v1c = icmp ult i32 %v1, 12 %v2c = icmp ugt i32 %v2, 34 @@ -557,3 +557,15 @@ exit: store i64 %ballot, ptr addrspace(1) %out ret void } + +define amdgpu_cs i64 @compare_bfloats(bfloat %x, bfloat %y) { +; CHECK-LABEL: compare_bfloats: +; CHECK: ; %bb.0: +; CHECK-NEXT: v_lshlrev_b32_e32 v1, 16, v1 +; CHECK-NEXT: v_lshlrev_b32_e32 v0, 16, v0 +; CHECK-NEXT: v_cmp_gt_f32_e64 s[0:1], v0, v1 +; CHECK-NEXT: ; return to shader part epilog + %cmp = fcmp ogt bfloat %x, %y + %ballot = call i64 @llvm.amdgcn.ballot.i64(i1 %cmp) + ret i64 %ballot +} diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ll index 6dd2258..39191d2 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ll @@ -23,10 +23,8 @@ define amdgpu_kernel void @test_s_i32(ptr addrspace(1) %out, i32 %src0) { ; GFX11-SDAG-NEXT: s_load_b32 s2, s[4:5], 0x2c ; GFX11-SDAG-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v0 -; GFX11-SDAG-NEXT: global_store_b32 v1, v0, s[0:1] +; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, s2 +; GFX11-SDAG-NEXT: global_store_b32 v0, v1, s[0:1] ; GFX11-SDAG-NEXT: s_endpgm ; ; GFX11-GISEL-LABEL: test_s_i32: @@ -36,8 +34,6 @@ define amdgpu_kernel void @test_s_i32(ptr addrspace(1) %out, i32 %src0) { ; GFX11-GISEL-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 ; GFX11-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; GFX11-GISEL-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2 -; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-GISEL-NEXT: v_permlane64_b32 v0, v0 ; GFX11-GISEL-NEXT: global_store_b32 v1, v0, s[0:1] ; GFX11-GISEL-NEXT: s_endpgm %v = call i32 @llvm.amdgcn.permlane64.i32(i32 %src0) @@ -50,12 +46,9 @@ define amdgpu_kernel void @test_s_i64(ptr addrspace(1) %out, i64 %src0) { ; GFX11-SDAG: ; %bb.0: ; GFX11-SDAG-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_mov_b32 v0, s3 -; GFX11-SDAG-NEXT: v_mov_b32_e32 v2, s2 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-SDAG-NEXT: v_permlane64_b32 v1, v0 -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v2 -; GFX11-SDAG-NEXT: global_store_b64 v3, v[0:1], s[0:1] +; GFX11-SDAG-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3 +; GFX11-SDAG-NEXT: v_mov_b32_e32 v0, s2 +; GFX11-SDAG-NEXT: global_store_b64 v2, v[0:1], s[0:1] ; GFX11-SDAG-NEXT: s_endpgm ; ; GFX11-GISEL-LABEL: test_s_i64: @@ -64,9 +57,6 @@ define amdgpu_kernel void @test_s_i64(ptr addrspace(1) %out, i64 %src0) { ; GFX11-GISEL-NEXT: v_mov_b32_e32 v2, 0 ; GFX11-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; GFX11-GISEL-NEXT: v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3 -; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-GISEL-NEXT: v_permlane64_b32 v0, v0 -; GFX11-GISEL-NEXT: v_permlane64_b32 v1, v1 ; GFX11-GISEL-NEXT: global_store_b64 v2, v[0:1], s[0:1] ; GFX11-GISEL-NEXT: s_endpgm %v = call i64 @llvm.amdgcn.permlane64.i64(i64 %src0) @@ -79,12 +69,9 @@ define amdgpu_kernel void @test_s_f64(ptr addrspace(1) %out, double %src0) { ; GFX11-SDAG: ; %bb.0: ; GFX11-SDAG-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_mov_b32 v0, s3 -; GFX11-SDAG-NEXT: v_mov_b32_e32 v2, s2 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-SDAG-NEXT: v_permlane64_b32 v1, v0 -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v2 -; GFX11-SDAG-NEXT: global_store_b64 v3, v[0:1], s[0:1] +; GFX11-SDAG-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3 +; GFX11-SDAG-NEXT: v_mov_b32_e32 v0, s2 +; GFX11-SDAG-NEXT: global_store_b64 v2, v[0:1], s[0:1] ; GFX11-SDAG-NEXT: s_endpgm ; ; GFX11-GISEL-LABEL: test_s_f64: @@ -93,9 +80,6 @@ define amdgpu_kernel void @test_s_f64(ptr addrspace(1) %out, double %src0) { ; GFX11-GISEL-NEXT: v_mov_b32_e32 v2, 0 ; GFX11-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; GFX11-GISEL-NEXT: v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3 -; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-GISEL-NEXT: v_permlane64_b32 v0, v0 -; GFX11-GISEL-NEXT: v_permlane64_b32 v1, v1 ; GFX11-GISEL-NEXT: global_store_b64 v2, v[0:1], s[0:1] ; GFX11-GISEL-NEXT: s_endpgm %v = call double @llvm.amdgcn.permlane64.f64(double %src0) @@ -116,19 +100,15 @@ define amdgpu_kernel void @test_i_i32(ptr addrspace(1) %out) { ; GFX11-SDAG-LABEL: test_i_i32: ; GFX11-SDAG: ; %bb.0: ; GFX11-SDAG-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 -; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, 0x63 :: v_dual_mov_b32 v1, 0 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v0 +; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, 0x63 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: global_store_b32 v1, v0, s[0:1] +; GFX11-SDAG-NEXT: global_store_b32 v0, v1, s[0:1] ; GFX11-SDAG-NEXT: s_endpgm ; ; GFX11-GISEL-LABEL: test_i_i32: ; GFX11-GISEL: ; %bb.0: ; GFX11-GISEL-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 ; GFX11-GISEL-NEXT: v_dual_mov_b32 v0, 0x63 :: v_dual_mov_b32 v1, 0 -; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-GISEL-NEXT: v_permlane64_b32 v0, v0 ; GFX11-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; GFX11-GISEL-NEXT: global_store_b32 v1, v0, s[0:1] ; GFX11-GISEL-NEXT: s_endpgm @@ -141,19 +121,15 @@ define amdgpu_kernel void @test_i_f32(ptr addrspace(1) %out) { ; GFX11-SDAG-LABEL: test_i_f32: ; GFX11-SDAG: ; %bb.0: ; GFX11-SDAG-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 -; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, 0x449a5000 :: v_dual_mov_b32 v1, 0 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v0 +; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, 0x449a5000 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: global_store_b32 v1, v0, s[0:1] +; GFX11-SDAG-NEXT: global_store_b32 v0, v1, s[0:1] ; GFX11-SDAG-NEXT: s_endpgm ; ; GFX11-GISEL-LABEL: test_i_f32: ; GFX11-GISEL: ; %bb.0: ; GFX11-GISEL-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 ; GFX11-GISEL-NEXT: v_dual_mov_b32 v0, 0x449a5000 :: v_dual_mov_b32 v1, 0 -; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-GISEL-NEXT: v_permlane64_b32 v0, v0 ; GFX11-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; GFX11-GISEL-NEXT: global_store_b32 v1, v0, s[0:1] ; GFX11-GISEL-NEXT: s_endpgm @@ -166,23 +142,16 @@ define amdgpu_kernel void @test_i_i64(ptr addrspace(1) %out) { ; GFX11-SDAG-LABEL: test_i_i64: ; GFX11-SDAG: ; %bb.0: ; GFX11-SDAG-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 -; GFX11-SDAG-NEXT: v_mov_b32_e32 v2, 0 -; GFX11-SDAG-NEXT: v_mov_b32_e32 v0, 0x63 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-SDAG-NEXT: v_permlane64_b32 v1, v2 -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v0 +; GFX11-SDAG-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, 0x63 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: global_store_b64 v2, v[0:1], s[0:1] +; GFX11-SDAG-NEXT: global_store_b64 v1, v[0:1], s[0:1] ; GFX11-SDAG-NEXT: s_endpgm ; ; GFX11-GISEL-LABEL: test_i_i64: ; GFX11-GISEL: ; %bb.0: ; GFX11-GISEL-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 ; GFX11-GISEL-NEXT: v_mov_b32_e32 v0, 0x63 -; GFX11-GISEL-NEXT: v_mov_b32_e32 v2, 0 -; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-GISEL-NEXT: v_permlane64_b32 v0, v0 -; GFX11-GISEL-NEXT: v_permlane64_b32 v1, v2 +; GFX11-GISEL-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, 0 ; GFX11-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; GFX11-GISEL-NEXT: global_store_b64 v2, v[0:1], s[0:1] ; GFX11-GISEL-NEXT: s_endpgm @@ -195,22 +164,16 @@ define amdgpu_kernel void @test_i_f64(ptr addrspace(1) %out) { ; GFX11-SDAG-LABEL: test_i_f64: ; GFX11-SDAG: ; %bb.0: ; GFX11-SDAG-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 -; GFX11-SDAG-NEXT: v_mov_b32_e32 v0, 0x40934a00 -; GFX11-SDAG-NEXT: v_mov_b32_e32 v2, 0 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-SDAG-NEXT: v_permlane64_b32 v1, v0 -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v2 +; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, 0x40934a00 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: global_store_b64 v2, v[0:1], s[0:1] +; GFX11-SDAG-NEXT: global_store_b64 v0, v[0:1], s[0:1] ; GFX11-SDAG-NEXT: s_endpgm ; ; GFX11-GISEL-LABEL: test_i_f64: ; GFX11-GISEL: ; %bb.0: ; GFX11-GISEL-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 -; GFX11-GISEL-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, 0x40934a00 -; GFX11-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-GISEL-NEXT: v_permlane64_b32 v0, v2 -; GFX11-GISEL-NEXT: v_permlane64_b32 v1, v1 +; GFX11-GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GFX11-GISEL-NEXT: v_dual_mov_b32 v1, 0x40934a00 :: v_dual_mov_b32 v2, 0 ; GFX11-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; GFX11-GISEL-NEXT: global_store_b64 v2, v[0:1], s[0:1] ; GFX11-GISEL-NEXT: s_endpgm diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ptr.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ptr.ll index b0149f7..672b658 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ptr.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ptr.ll @@ -6,12 +6,9 @@ define amdgpu_kernel void @test_p0(ptr addrspace(1) %out, ptr %src0) { ; GFX11-SDAG: ; %bb.0: ; GFX11-SDAG-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_mov_b32 v0, s3 -; GFX11-SDAG-NEXT: v_mov_b32_e32 v2, s2 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-SDAG-NEXT: v_permlane64_b32 v1, v0 -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v2 -; GFX11-SDAG-NEXT: global_store_b64 v3, v[0:1], s[0:1] +; GFX11-SDAG-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s3 +; GFX11-SDAG-NEXT: v_mov_b32_e32 v0, s2 +; GFX11-SDAG-NEXT: global_store_b64 v2, v[0:1], s[0:1] ; GFX11-SDAG-NEXT: s_endpgm %v = call ptr @llvm.amdgcn.permlane64.p0(ptr %src0) store ptr %v, ptr addrspace(1) %out @@ -22,21 +19,14 @@ define amdgpu_kernel void @test_v3p0(ptr addrspace(1) %out, <3 x ptr> %src0) { ; GFX11-SDAG-LABEL: test_v3p0: ; GFX11-SDAG: ; %bb.0: ; GFX11-SDAG-NEXT: s_clause 0x2 -; GFX11-SDAG-NEXT: s_load_b128 s[0:3], s[4:5], 0x44 ; GFX11-SDAG-NEXT: s_load_b64 s[6:7], s[4:5], 0x54 +; GFX11-SDAG-NEXT: s_load_b128 s[0:3], s[4:5], 0x44 ; GFX11-SDAG-NEXT: s_load_b64 s[4:5], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v1, s2 -; GFX11-SDAG-NEXT: v_dual_mov_b32 v4, s1 :: v_dual_mov_b32 v5, s7 -; GFX11-SDAG-NEXT: v_mov_b32_e32 v8, s6 -; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, s3 :: v_dual_mov_b32 v7, s0 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_4) -; GFX11-SDAG-NEXT: v_permlane64_b32 v2, v1 -; GFX11-SDAG-NEXT: v_permlane64_b32 v1, v4 -; GFX11-SDAG-NEXT: v_permlane64_b32 v5, v5 -; GFX11-SDAG-NEXT: v_permlane64_b32 v4, v8 -; GFX11-SDAG-NEXT: v_permlane64_b32 v3, v0 -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v7 +; GFX11-SDAG-NEXT: v_dual_mov_b32 v6, 0 :: v_dual_mov_b32 v5, s7 +; GFX11-SDAG-NEXT: v_dual_mov_b32 v4, s6 :: v_dual_mov_b32 v1, s1 +; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v3, s3 +; GFX11-SDAG-NEXT: v_mov_b32_e32 v2, s2 ; GFX11-SDAG-NEXT: s_clause 0x1 ; GFX11-SDAG-NEXT: global_store_b64 v6, v[4:5], s[4:5] offset:16 ; GFX11-SDAG-NEXT: global_store_b128 v6, v[0:3], s[4:5] @@ -53,10 +43,8 @@ define amdgpu_kernel void @test_p3(ptr addrspace(1) %out, ptr addrspace(3) %src0 ; GFX11-SDAG-NEXT: s_load_b32 s2, s[4:5], 0x2c ; GFX11-SDAG-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v0 -; GFX11-SDAG-NEXT: global_store_b32 v1, v0, s[0:1] +; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, s2 +; GFX11-SDAG-NEXT: global_store_b32 v0, v1, s[0:1] ; GFX11-SDAG-NEXT: s_endpgm %v = call ptr addrspace(3) @llvm.amdgcn.permlane64.v3p0(ptr addrspace(3) %src0) store ptr addrspace(3) %v, ptr addrspace(1) %out @@ -70,14 +58,9 @@ define amdgpu_kernel void @test_v3p3(ptr addrspace(1) %out, <3 x ptr addrspace(3 ; GFX11-SDAG-NEXT: s_load_b128 s[0:3], s[4:5], 0x34 ; GFX11-SDAG-NEXT: s_load_b64 s[4:5], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v3, s0 -; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s1 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-SDAG-NEXT: v_permlane64_b32 v2, v0 -; GFX11-SDAG-NEXT: v_permlane64_b32 v1, v1 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_4) -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v3 -; GFX11-SDAG-NEXT: global_store_b96 v4, v[0:2], s[4:5] +; GFX11-SDAG-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_mov_b32 v0, s0 +; GFX11-SDAG-NEXT: v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v2, s2 +; GFX11-SDAG-NEXT: global_store_b96 v3, v[0:2], s[4:5] ; GFX11-SDAG-NEXT: s_endpgm %v = call <3 x ptr addrspace(3)> @llvm.amdgcn.permlane64.v3p3(<3 x ptr addrspace(3)> %src0) store <3 x ptr addrspace(3)> %v, ptr addrspace(1) %out @@ -91,10 +74,8 @@ define amdgpu_kernel void @test_p5(ptr addrspace(1) %out, ptr addrspace(5) %src0 ; GFX11-SDAG-NEXT: s_load_b32 s2, s[4:5], 0x2c ; GFX11-SDAG-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v0 -; GFX11-SDAG-NEXT: global_store_b32 v1, v0, s[0:1] +; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, s2 +; GFX11-SDAG-NEXT: global_store_b32 v0, v1, s[0:1] ; GFX11-SDAG-NEXT: s_endpgm %v = call ptr addrspace(5) @llvm.amdgcn.permlane64.p5(ptr addrspace(5) %src0) store ptr addrspace(5) %v, ptr addrspace(1) %out @@ -108,14 +89,9 @@ define amdgpu_kernel void @test_v3p5(ptr addrspace(1) %out, <3 x ptr addrspace(5 ; GFX11-SDAG-NEXT: s_load_b128 s[0:3], s[4:5], 0x34 ; GFX11-SDAG-NEXT: s_load_b64 s[4:5], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v3, s0 -; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s1 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-SDAG-NEXT: v_permlane64_b32 v2, v0 -; GFX11-SDAG-NEXT: v_permlane64_b32 v1, v1 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_4) -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v3 -; GFX11-SDAG-NEXT: global_store_b96 v4, v[0:2], s[4:5] +; GFX11-SDAG-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_mov_b32 v0, s0 +; GFX11-SDAG-NEXT: v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v2, s2 +; GFX11-SDAG-NEXT: global_store_b96 v3, v[0:2], s[4:5] ; GFX11-SDAG-NEXT: s_endpgm %v = call <3 x ptr addrspace(5)> @llvm.amdgcn.permlane64.v3p5(<3 x ptr addrspace(5)> %src0) store <3 x ptr addrspace(5)> %v, ptr addrspace(1) %out @@ -129,10 +105,8 @@ define amdgpu_kernel void @test_p6(ptr addrspace(1) %out, ptr addrspace(6) %src0 ; GFX11-SDAG-NEXT: s_load_b32 s2, s[4:5], 0x2c ; GFX11-SDAG-NEXT: s_load_b64 s[0:1], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v0, s2 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v0 -; GFX11-SDAG-NEXT: global_store_b32 v1, v0, s[0:1] +; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v1, s2 +; GFX11-SDAG-NEXT: global_store_b32 v0, v1, s[0:1] ; GFX11-SDAG-NEXT: s_endpgm %v = call ptr addrspace(6) @llvm.amdgcn.permlane64.p6(ptr addrspace(6) %src0) store ptr addrspace(6) %v, ptr addrspace(1) %out @@ -146,14 +120,9 @@ define amdgpu_kernel void @test_v3p6(ptr addrspace(1) %out, <3 x ptr addrspace(6 ; GFX11-SDAG-NEXT: s_load_b128 s[0:3], s[4:5], 0x34 ; GFX11-SDAG-NEXT: s_load_b64 s[4:5], s[4:5], 0x24 ; GFX11-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; GFX11-SDAG-NEXT: v_dual_mov_b32 v4, 0 :: v_dual_mov_b32 v3, s0 -; GFX11-SDAG-NEXT: v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s1 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2) -; GFX11-SDAG-NEXT: v_permlane64_b32 v2, v0 -; GFX11-SDAG-NEXT: v_permlane64_b32 v1, v1 -; GFX11-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_4) -; GFX11-SDAG-NEXT: v_permlane64_b32 v0, v3 -; GFX11-SDAG-NEXT: global_store_b96 v4, v[0:2], s[4:5] +; GFX11-SDAG-NEXT: v_dual_mov_b32 v3, 0 :: v_dual_mov_b32 v0, s0 +; GFX11-SDAG-NEXT: v_dual_mov_b32 v1, s1 :: v_dual_mov_b32 v2, s2 +; GFX11-SDAG-NEXT: global_store_b96 v3, v[0:2], s[4:5] ; GFX11-SDAG-NEXT: s_endpgm %v = call <3 x ptr addrspace(6)> @llvm.amdgcn.permlane64.v3p6(<3 x ptr addrspace(6)> %src0) store <3 x ptr addrspace(6)> %v, ptr addrspace(1) %out diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll index d1ba892..02d2990 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll @@ -396,8 +396,7 @@ define amdgpu_kernel void @test_readfirstlane_imm_f64(ptr addrspace(1) %out) { ; ; CHECK-GISEL-LABEL: test_readfirstlane_imm_f64: ; CHECK-GISEL: ; %bb.0: -; CHECK-GISEL-NEXT: s_mov_b32 s0, 0 -; CHECK-GISEL-NEXT: s_mov_b32 s1, 0x40400000 +; CHECK-GISEL-NEXT: s_mov_b64 s[0:1], 0x4040000000000000 ; CHECK-GISEL-NEXT: ;;#ASMSTART ; CHECK-GISEL-NEXT: ; use s[0:1] ; CHECK-GISEL-NEXT: ;;#ASMEND @@ -456,14 +455,13 @@ define amdgpu_kernel void @test_readfirstlane_imm_fold_i64(ptr addrspace(1) %out ; CHECK-GISEL-LABEL: test_readfirstlane_imm_fold_i64: ; CHECK-GISEL: ; %bb.0: ; CHECK-GISEL-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 -; CHECK-GISEL-NEXT: s_mov_b64 s[2:3], 32 ; CHECK-GISEL-NEXT: s_add_i32 s12, s12, s17 -; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, s2 +; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, 32 ; CHECK-GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13 +; CHECK-GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 ; CHECK-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; CHECK-GISEL-NEXT: v_mov_b32_e32 v3, s1 -; CHECK-GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 -; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, s3 +; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, 0 ; CHECK-GISEL-NEXT: v_mov_b32_e32 v2, s0 ; CHECK-GISEL-NEXT: flat_store_dwordx2 v[2:3], v[0:1] ; CHECK-GISEL-NEXT: s_endpgm @@ -490,15 +488,13 @@ define amdgpu_kernel void @test_readfirstlane_imm_fold_f64(ptr addrspace(1) %out ; CHECK-GISEL-LABEL: test_readfirstlane_imm_fold_f64: ; CHECK-GISEL: ; %bb.0: ; CHECK-GISEL-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 -; CHECK-GISEL-NEXT: s_mov_b32 s2, 0 ; CHECK-GISEL-NEXT: s_add_i32 s12, s12, s17 -; CHECK-GISEL-NEXT: s_mov_b32 s3, 0x40400000 -; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, s2 -; CHECK-GISEL-NEXT: s_waitcnt lgkmcnt(0) -; CHECK-GISEL-NEXT: v_mov_b32_e32 v3, s1 +; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, 0 ; CHECK-GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13 ; CHECK-GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 -; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, s3 +; CHECK-GISEL-NEXT: s_waitcnt lgkmcnt(0) +; CHECK-GISEL-NEXT: v_mov_b32_e32 v3, s1 +; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, 0x40400000 ; CHECK-GISEL-NEXT: v_mov_b32_e32 v2, s0 ; CHECK-GISEL-NEXT: flat_store_dwordx2 v[2:3], v[0:1] ; CHECK-GISEL-NEXT: s_endpgm @@ -588,17 +584,17 @@ define amdgpu_kernel void @test_readfirstlane_copy_from_sgpr_i64(ptr addrspace(1 ; CHECK-SDAG: ; %bb.0: ; CHECK-SDAG-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 ; CHECK-SDAG-NEXT: s_add_i32 s12, s12, s17 -; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 -; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 ; CHECK-SDAG-NEXT: ;;#ASMSTART ; CHECK-SDAG-NEXT: s_mov_b64 s[2:3], 0 ; CHECK-SDAG-NEXT: ;;#ASMEND +; CHECK-SDAG-NEXT: v_mov_b32_e32 v2, s2 +; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 ; CHECK-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; CHECK-SDAG-NEXT: v_mov_b32_e32 v3, s1 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s2 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s3 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v2, s0 -; CHECK-SDAG-NEXT: flat_store_dwordx2 v[2:3], v[0:1] +; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s0 +; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s1 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v3, s3 +; CHECK-SDAG-NEXT: flat_store_dwordx2 v[0:1], v[2:3] ; CHECK-SDAG-NEXT: s_endpgm ; ; CHECK-GISEL-LABEL: test_readfirstlane_copy_from_sgpr_i64: @@ -628,17 +624,17 @@ define amdgpu_kernel void @test_readfirstlane_copy_from_sgpr_f64(ptr addrspace(1 ; CHECK-SDAG: ; %bb.0: ; CHECK-SDAG-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 ; CHECK-SDAG-NEXT: s_add_i32 s12, s12, s17 -; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 -; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 ; CHECK-SDAG-NEXT: ;;#ASMSTART ; CHECK-SDAG-NEXT: s_mov_b64 s[2:3], 0 ; CHECK-SDAG-NEXT: ;;#ASMEND +; CHECK-SDAG-NEXT: v_mov_b32_e32 v2, s2 +; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 ; CHECK-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; CHECK-SDAG-NEXT: v_mov_b32_e32 v3, s1 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s2 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s3 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v2, s0 -; CHECK-SDAG-NEXT: flat_store_dwordx2 v[2:3], v[0:1] +; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s0 +; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s1 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v3, s3 +; CHECK-SDAG-NEXT: flat_store_dwordx2 v[0:1], v[2:3] ; CHECK-SDAG-NEXT: s_endpgm ; ; CHECK-GISEL-LABEL: test_readfirstlane_copy_from_sgpr_f64: diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll index 7ff5eb4..0795f40 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readlane.ll @@ -9,7 +9,7 @@ declare double @llvm.amdgcn.readlane.f64(double, i32) #0 define amdgpu_kernel void @test_readlane_sreg_sreg_i32(i32 %src0, i32 %src1) #1 { ; CHECK-SDAG-LABEL: test_readlane_sreg_sreg_i32: ; CHECK-SDAG: ; %bb.0: -; CHECK-SDAG-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 +; CHECK-SDAG-NEXT: s_load_dword s0, s[8:9], 0x0 ; CHECK-SDAG-NEXT: s_waitcnt lgkmcnt(0) ; CHECK-SDAG-NEXT: ;;#ASMSTART ; CHECK-SDAG-NEXT: ; use s0 @@ -18,7 +18,7 @@ define amdgpu_kernel void @test_readlane_sreg_sreg_i32(i32 %src0, i32 %src1) #1 ; ; CHECK-GISEL-LABEL: test_readlane_sreg_sreg_i32: ; CHECK-GISEL: ; %bb.0: -; CHECK-GISEL-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 +; CHECK-GISEL-NEXT: s_load_dword s0, s[8:9], 0x0 ; CHECK-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; CHECK-GISEL-NEXT: ;;#ASMSTART ; CHECK-GISEL-NEXT: ; use s0 @@ -224,14 +224,13 @@ define amdgpu_kernel void @test_readlane_imm_sreg_i64(ptr addrspace(1) %out, i32 ; CHECK-GISEL-LABEL: test_readlane_imm_sreg_i64: ; CHECK-GISEL: ; %bb.0: ; CHECK-GISEL-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 -; CHECK-GISEL-NEXT: s_mov_b64 s[2:3], 32 ; CHECK-GISEL-NEXT: s_add_i32 s12, s12, s17 -; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, s2 +; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, 32 ; CHECK-GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13 +; CHECK-GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 ; CHECK-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; CHECK-GISEL-NEXT: v_mov_b32_e32 v3, s1 -; CHECK-GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 -; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, s3 +; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, 0 ; CHECK-GISEL-NEXT: v_mov_b32_e32 v2, s0 ; CHECK-GISEL-NEXT: flat_store_dwordx2 v[2:3], v[0:1] ; CHECK-GISEL-NEXT: s_endpgm @@ -258,15 +257,13 @@ define amdgpu_kernel void @test_readlane_imm_sreg_f64(ptr addrspace(1) %out, i32 ; CHECK-GISEL-LABEL: test_readlane_imm_sreg_f64: ; CHECK-GISEL: ; %bb.0: ; CHECK-GISEL-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 -; CHECK-GISEL-NEXT: s_mov_b32 s2, 0 ; CHECK-GISEL-NEXT: s_add_i32 s12, s12, s17 -; CHECK-GISEL-NEXT: s_mov_b32 s3, 0x40400000 -; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, s2 -; CHECK-GISEL-NEXT: s_waitcnt lgkmcnt(0) -; CHECK-GISEL-NEXT: v_mov_b32_e32 v3, s1 +; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, 0 ; CHECK-GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13 ; CHECK-GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 -; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, s3 +; CHECK-GISEL-NEXT: s_waitcnt lgkmcnt(0) +; CHECK-GISEL-NEXT: v_mov_b32_e32 v3, s1 +; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, 0x40400000 ; CHECK-GISEL-NEXT: v_mov_b32_e32 v2, s0 ; CHECK-GISEL-NEXT: flat_store_dwordx2 v[2:3], v[0:1] ; CHECK-GISEL-NEXT: s_endpgm @@ -660,17 +657,17 @@ define amdgpu_kernel void @test_readlane_copy_from_sgpr_i64(ptr addrspace(1) %ou ; CHECK-SDAG: ; %bb.0: ; CHECK-SDAG-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 ; CHECK-SDAG-NEXT: s_add_i32 s12, s12, s17 -; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 -; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 ; CHECK-SDAG-NEXT: ;;#ASMSTART ; CHECK-SDAG-NEXT: s_mov_b64 s[2:3], 0 ; CHECK-SDAG-NEXT: ;;#ASMEND +; CHECK-SDAG-NEXT: v_mov_b32_e32 v2, s2 +; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 ; CHECK-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; CHECK-SDAG-NEXT: v_mov_b32_e32 v3, s1 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s2 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s3 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v2, s0 -; CHECK-SDAG-NEXT: flat_store_dwordx2 v[2:3], v[0:1] +; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s0 +; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s1 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v3, s3 +; CHECK-SDAG-NEXT: flat_store_dwordx2 v[0:1], v[2:3] ; CHECK-SDAG-NEXT: s_endpgm ; ; CHECK-GISEL-LABEL: test_readlane_copy_from_sgpr_i64: @@ -700,17 +697,17 @@ define amdgpu_kernel void @test_readlane_copy_from_sgpr_f64(ptr addrspace(1) %ou ; CHECK-SDAG: ; %bb.0: ; CHECK-SDAG-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 ; CHECK-SDAG-NEXT: s_add_i32 s12, s12, s17 -; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 -; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 ; CHECK-SDAG-NEXT: ;;#ASMSTART ; CHECK-SDAG-NEXT: s_mov_b64 s[2:3], 0 ; CHECK-SDAG-NEXT: ;;#ASMEND +; CHECK-SDAG-NEXT: v_mov_b32_e32 v2, s2 +; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 ; CHECK-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; CHECK-SDAG-NEXT: v_mov_b32_e32 v3, s1 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s2 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s3 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v2, s0 -; CHECK-SDAG-NEXT: flat_store_dwordx2 v[2:3], v[0:1] +; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s0 +; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s1 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v3, s3 +; CHECK-SDAG-NEXT: flat_store_dwordx2 v[0:1], v[2:3] ; CHECK-SDAG-NEXT: s_endpgm ; ; CHECK-GISEL-LABEL: test_readlane_copy_from_sgpr_f64: diff --git a/llvm/test/CodeGen/AMDGPU/promote-alloca-array-to-vector.ll b/llvm/test/CodeGen/AMDGPU/promote-alloca-array-to-vector.ll new file mode 100644 index 0000000..05a0e39 --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/promote-alloca-array-to-vector.ll @@ -0,0 +1,325 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6 +; RUN: opt -S -mtriple=amdgcn-- -mcpu=gfx1100 -passes=amdgpu-promote-alloca < %s | FileCheck -check-prefix=OPT %s + +define amdgpu_kernel void @large_array_vectors_small_users(<16 x i8> %in, <16 x i8> %add, ptr addrspace(3) %out) #0 { +; OPT-LABEL: define amdgpu_kernel void @large_array_vectors_small_users( +; OPT-SAME: <16 x i8> [[IN:%.*]], <16 x i8> [[ADD:%.*]], ptr addrspace(3) [[OUT:%.*]]) #[[ATTR0:[0-9]+]] { +; OPT-NEXT: [[ENTRY:.*:]] +; OPT-NEXT: [[ALLOCA:%.*]] = freeze <128 x i8> poison +; OPT-NEXT: [[TMP0:%.*]] = extractelement <16 x i8> [[IN]], i64 0 +; OPT-NEXT: [[TMP1:%.*]] = insertelement <128 x i8> [[ALLOCA]], i8 [[TMP0]], i32 0 +; OPT-NEXT: [[TMP2:%.*]] = extractelement <16 x i8> [[IN]], i64 1 +; OPT-NEXT: [[TMP3:%.*]] = insertelement <128 x i8> [[TMP1]], i8 [[TMP2]], i32 1 +; OPT-NEXT: [[TMP4:%.*]] = extractelement <16 x i8> [[IN]], i64 2 +; OPT-NEXT: [[TMP5:%.*]] = insertelement <128 x i8> [[TMP3]], i8 [[TMP4]], i32 2 +; OPT-NEXT: [[TMP6:%.*]] = extractelement <16 x i8> [[IN]], i64 3 +; OPT-NEXT: [[TMP7:%.*]] = insertelement <128 x i8> [[TMP5]], i8 [[TMP6]], i32 3 +; OPT-NEXT: [[TMP8:%.*]] = extractelement <16 x i8> [[IN]], i64 4 +; OPT-NEXT: [[TMP9:%.*]] = insertelement <128 x i8> [[TMP7]], i8 [[TMP8]], i32 4 +; OPT-NEXT: [[TMP10:%.*]] = extractelement <16 x i8> [[IN]], i64 5 +; OPT-NEXT: [[TMP11:%.*]] = insertelement <128 x i8> [[TMP9]], i8 [[TMP10]], i32 5 +; OPT-NEXT: [[TMP12:%.*]] = extractelement <16 x i8> [[IN]], i64 6 +; OPT-NEXT: [[TMP13:%.*]] = insertelement <128 x i8> [[TMP11]], i8 [[TMP12]], i32 6 +; OPT-NEXT: [[TMP14:%.*]] = extractelement <16 x i8> [[IN]], i64 7 +; OPT-NEXT: [[TMP15:%.*]] = insertelement <128 x i8> [[TMP13]], i8 [[TMP14]], i32 7 +; OPT-NEXT: [[TMP16:%.*]] = extractelement <16 x i8> [[IN]], i64 8 +; OPT-NEXT: [[TMP17:%.*]] = insertelement <128 x i8> [[TMP15]], i8 [[TMP16]], i32 8 +; OPT-NEXT: [[TMP18:%.*]] = extractelement <16 x i8> [[IN]], i64 9 +; OPT-NEXT: [[TMP19:%.*]] = insertelement <128 x i8> [[TMP17]], i8 [[TMP18]], i32 9 +; OPT-NEXT: [[TMP20:%.*]] = extractelement <16 x i8> [[IN]], i64 10 +; OPT-NEXT: [[TMP21:%.*]] = insertelement <128 x i8> [[TMP19]], i8 [[TMP20]], i32 10 +; OPT-NEXT: [[TMP22:%.*]] = extractelement <16 x i8> [[IN]], i64 11 +; OPT-NEXT: [[TMP23:%.*]] = insertelement <128 x i8> [[TMP21]], i8 [[TMP22]], i32 11 +; OPT-NEXT: [[TMP24:%.*]] = extractelement <16 x i8> [[IN]], i64 12 +; OPT-NEXT: [[TMP25:%.*]] = insertelement <128 x i8> [[TMP23]], i8 [[TMP24]], i32 12 +; OPT-NEXT: [[TMP26:%.*]] = extractelement <16 x i8> [[IN]], i64 13 +; OPT-NEXT: [[TMP27:%.*]] = insertelement <128 x i8> [[TMP25]], i8 [[TMP26]], i32 13 +; OPT-NEXT: [[TMP28:%.*]] = extractelement <16 x i8> [[IN]], i64 14 +; OPT-NEXT: [[TMP29:%.*]] = insertelement <128 x i8> [[TMP27]], i8 [[TMP28]], i32 14 +; OPT-NEXT: [[TMP30:%.*]] = extractelement <16 x i8> [[IN]], i64 15 +; OPT-NEXT: [[TMP31:%.*]] = insertelement <128 x i8> [[TMP29]], i8 [[TMP30]], i32 15 +; OPT-NEXT: [[TMP32:%.*]] = extractelement <16 x i8> [[IN]], i64 0 +; OPT-NEXT: [[TMP33:%.*]] = insertelement <128 x i8> [[TMP31]], i8 [[TMP32]], i32 0 +; OPT-NEXT: [[TMP34:%.*]] = extractelement <16 x i8> [[IN]], i64 1 +; OPT-NEXT: [[TMP35:%.*]] = insertelement <128 x i8> [[TMP33]], i8 [[TMP34]], i32 1 +; OPT-NEXT: [[TMP36:%.*]] = extractelement <16 x i8> [[IN]], i64 2 +; OPT-NEXT: [[TMP37:%.*]] = insertelement <128 x i8> [[TMP35]], i8 [[TMP36]], i32 2 +; OPT-NEXT: [[TMP38:%.*]] = extractelement <16 x i8> [[IN]], i64 3 +; OPT-NEXT: [[TMP39:%.*]] = insertelement <128 x i8> [[TMP37]], i8 [[TMP38]], i32 3 +; OPT-NEXT: [[TMP40:%.*]] = extractelement <16 x i8> [[IN]], i64 4 +; OPT-NEXT: [[TMP41:%.*]] = insertelement <128 x i8> [[TMP39]], i8 [[TMP40]], i32 4 +; OPT-NEXT: [[TMP42:%.*]] = extractelement <16 x i8> [[IN]], i64 5 +; OPT-NEXT: [[TMP43:%.*]] = insertelement <128 x i8> [[TMP41]], i8 [[TMP42]], i32 5 +; OPT-NEXT: [[TMP44:%.*]] = extractelement <16 x i8> [[IN]], i64 6 +; OPT-NEXT: [[TMP45:%.*]] = insertelement <128 x i8> [[TMP43]], i8 [[TMP44]], i32 6 +; OPT-NEXT: [[TMP46:%.*]] = extractelement <16 x i8> [[IN]], i64 7 +; OPT-NEXT: [[TMP47:%.*]] = insertelement <128 x i8> [[TMP45]], i8 [[TMP46]], i32 7 +; OPT-NEXT: [[TMP48:%.*]] = extractelement <16 x i8> [[IN]], i64 8 +; OPT-NEXT: [[TMP49:%.*]] = insertelement <128 x i8> [[TMP47]], i8 [[TMP48]], i32 8 +; OPT-NEXT: [[TMP50:%.*]] = extractelement <16 x i8> [[IN]], i64 9 +; OPT-NEXT: [[TMP51:%.*]] = insertelement <128 x i8> [[TMP49]], i8 [[TMP50]], i32 9 +; OPT-NEXT: [[TMP52:%.*]] = extractelement <16 x i8> [[IN]], i64 10 +; OPT-NEXT: [[TMP53:%.*]] = insertelement <128 x i8> [[TMP51]], i8 [[TMP52]], i32 10 +; OPT-NEXT: [[TMP54:%.*]] = extractelement <16 x i8> [[IN]], i64 11 +; OPT-NEXT: [[TMP55:%.*]] = insertelement <128 x i8> [[TMP53]], i8 [[TMP54]], i32 11 +; OPT-NEXT: [[TMP56:%.*]] = extractelement <16 x i8> [[IN]], i64 12 +; OPT-NEXT: [[TMP57:%.*]] = insertelement <128 x i8> [[TMP55]], i8 [[TMP56]], i32 12 +; OPT-NEXT: [[TMP58:%.*]] = extractelement <16 x i8> [[IN]], i64 13 +; OPT-NEXT: [[TMP59:%.*]] = insertelement <128 x i8> [[TMP57]], i8 [[TMP58]], i32 13 +; OPT-NEXT: [[TMP60:%.*]] = extractelement <16 x i8> [[IN]], i64 14 +; OPT-NEXT: [[TMP61:%.*]] = insertelement <128 x i8> [[TMP59]], i8 [[TMP60]], i32 14 +; OPT-NEXT: [[TMP62:%.*]] = extractelement <16 x i8> [[IN]], i64 15 +; OPT-NEXT: [[TMP63:%.*]] = insertelement <128 x i8> [[TMP61]], i8 [[TMP62]], i32 15 +; OPT-NEXT: [[TMP64:%.*]] = extractelement <16 x i8> [[IN]], i64 0 +; OPT-NEXT: [[TMP65:%.*]] = insertelement <128 x i8> [[TMP63]], i8 [[TMP64]], i32 0 +; OPT-NEXT: [[TMP66:%.*]] = extractelement <16 x i8> [[IN]], i64 1 +; OPT-NEXT: [[TMP67:%.*]] = insertelement <128 x i8> [[TMP65]], i8 [[TMP66]], i32 1 +; OPT-NEXT: [[TMP68:%.*]] = extractelement <16 x i8> [[IN]], i64 2 +; OPT-NEXT: [[TMP69:%.*]] = insertelement <128 x i8> [[TMP67]], i8 [[TMP68]], i32 2 +; OPT-NEXT: [[TMP70:%.*]] = extractelement <16 x i8> [[IN]], i64 3 +; OPT-NEXT: [[TMP71:%.*]] = insertelement <128 x i8> [[TMP69]], i8 [[TMP70]], i32 3 +; OPT-NEXT: [[TMP72:%.*]] = extractelement <16 x i8> [[IN]], i64 4 +; OPT-NEXT: [[TMP73:%.*]] = insertelement <128 x i8> [[TMP71]], i8 [[TMP72]], i32 4 +; OPT-NEXT: [[TMP74:%.*]] = extractelement <16 x i8> [[IN]], i64 5 +; OPT-NEXT: [[TMP75:%.*]] = insertelement <128 x i8> [[TMP73]], i8 [[TMP74]], i32 5 +; OPT-NEXT: [[TMP76:%.*]] = extractelement <16 x i8> [[IN]], i64 6 +; OPT-NEXT: [[TMP77:%.*]] = insertelement <128 x i8> [[TMP75]], i8 [[TMP76]], i32 6 +; OPT-NEXT: [[TMP78:%.*]] = extractelement <16 x i8> [[IN]], i64 7 +; OPT-NEXT: [[TMP79:%.*]] = insertelement <128 x i8> [[TMP77]], i8 [[TMP78]], i32 7 +; OPT-NEXT: [[TMP80:%.*]] = extractelement <16 x i8> [[IN]], i64 8 +; OPT-NEXT: [[TMP81:%.*]] = insertelement <128 x i8> [[TMP79]], i8 [[TMP80]], i32 8 +; OPT-NEXT: [[TMP82:%.*]] = extractelement <16 x i8> [[IN]], i64 9 +; OPT-NEXT: [[TMP83:%.*]] = insertelement <128 x i8> [[TMP81]], i8 [[TMP82]], i32 9 +; OPT-NEXT: [[TMP84:%.*]] = extractelement <16 x i8> [[IN]], i64 10 +; OPT-NEXT: [[TMP85:%.*]] = insertelement <128 x i8> [[TMP83]], i8 [[TMP84]], i32 10 +; OPT-NEXT: [[TMP86:%.*]] = extractelement <16 x i8> [[IN]], i64 11 +; OPT-NEXT: [[TMP87:%.*]] = insertelement <128 x i8> [[TMP85]], i8 [[TMP86]], i32 11 +; OPT-NEXT: [[TMP88:%.*]] = extractelement <16 x i8> [[IN]], i64 12 +; OPT-NEXT: [[TMP89:%.*]] = insertelement <128 x i8> [[TMP87]], i8 [[TMP88]], i32 12 +; OPT-NEXT: [[TMP90:%.*]] = extractelement <16 x i8> [[IN]], i64 13 +; OPT-NEXT: [[TMP91:%.*]] = insertelement <128 x i8> [[TMP89]], i8 [[TMP90]], i32 13 +; OPT-NEXT: [[TMP92:%.*]] = extractelement <16 x i8> [[IN]], i64 14 +; OPT-NEXT: [[TMP93:%.*]] = insertelement <128 x i8> [[TMP91]], i8 [[TMP92]], i32 14 +; OPT-NEXT: [[TMP94:%.*]] = extractelement <16 x i8> [[IN]], i64 15 +; OPT-NEXT: [[TMP95:%.*]] = insertelement <128 x i8> [[TMP93]], i8 [[TMP94]], i32 15 +; OPT-NEXT: [[TMP96:%.*]] = extractelement <16 x i8> [[IN]], i64 0 +; OPT-NEXT: [[TMP97:%.*]] = insertelement <128 x i8> [[TMP95]], i8 [[TMP96]], i32 0 +; OPT-NEXT: [[TMP98:%.*]] = extractelement <16 x i8> [[IN]], i64 1 +; OPT-NEXT: [[TMP99:%.*]] = insertelement <128 x i8> [[TMP97]], i8 [[TMP98]], i32 1 +; OPT-NEXT: [[TMP100:%.*]] = extractelement <16 x i8> [[IN]], i64 2 +; OPT-NEXT: [[TMP101:%.*]] = insertelement <128 x i8> [[TMP99]], i8 [[TMP100]], i32 2 +; OPT-NEXT: [[TMP102:%.*]] = extractelement <16 x i8> [[IN]], i64 3 +; OPT-NEXT: [[TMP103:%.*]] = insertelement <128 x i8> [[TMP101]], i8 [[TMP102]], i32 3 +; OPT-NEXT: [[TMP104:%.*]] = extractelement <16 x i8> [[IN]], i64 4 +; OPT-NEXT: [[TMP105:%.*]] = insertelement <128 x i8> [[TMP103]], i8 [[TMP104]], i32 4 +; OPT-NEXT: [[TMP106:%.*]] = extractelement <16 x i8> [[IN]], i64 5 +; OPT-NEXT: [[TMP107:%.*]] = insertelement <128 x i8> [[TMP105]], i8 [[TMP106]], i32 5 +; OPT-NEXT: [[TMP108:%.*]] = extractelement <16 x i8> [[IN]], i64 6 +; OPT-NEXT: [[TMP109:%.*]] = insertelement <128 x i8> [[TMP107]], i8 [[TMP108]], i32 6 +; OPT-NEXT: [[TMP110:%.*]] = extractelement <16 x i8> [[IN]], i64 7 +; OPT-NEXT: [[TMP111:%.*]] = insertelement <128 x i8> [[TMP109]], i8 [[TMP110]], i32 7 +; OPT-NEXT: [[TMP112:%.*]] = extractelement <16 x i8> [[IN]], i64 8 +; OPT-NEXT: [[TMP113:%.*]] = insertelement <128 x i8> [[TMP111]], i8 [[TMP112]], i32 8 +; OPT-NEXT: [[TMP114:%.*]] = extractelement <16 x i8> [[IN]], i64 9 +; OPT-NEXT: [[TMP115:%.*]] = insertelement <128 x i8> [[TMP113]], i8 [[TMP114]], i32 9 +; OPT-NEXT: [[TMP116:%.*]] = extractelement <16 x i8> [[IN]], i64 10 +; OPT-NEXT: [[TMP117:%.*]] = insertelement <128 x i8> [[TMP115]], i8 [[TMP116]], i32 10 +; OPT-NEXT: [[TMP118:%.*]] = extractelement <16 x i8> [[IN]], i64 11 +; OPT-NEXT: [[TMP119:%.*]] = insertelement <128 x i8> [[TMP117]], i8 [[TMP118]], i32 11 +; OPT-NEXT: [[TMP120:%.*]] = extractelement <16 x i8> [[IN]], i64 12 +; OPT-NEXT: [[TMP121:%.*]] = insertelement <128 x i8> [[TMP119]], i8 [[TMP120]], i32 12 +; OPT-NEXT: [[TMP122:%.*]] = extractelement <16 x i8> [[IN]], i64 13 +; OPT-NEXT: [[TMP123:%.*]] = insertelement <128 x i8> [[TMP121]], i8 [[TMP122]], i32 13 +; OPT-NEXT: [[TMP124:%.*]] = extractelement <16 x i8> [[IN]], i64 14 +; OPT-NEXT: [[TMP125:%.*]] = insertelement <128 x i8> [[TMP123]], i8 [[TMP124]], i32 14 +; OPT-NEXT: [[TMP126:%.*]] = extractelement <16 x i8> [[IN]], i64 15 +; OPT-NEXT: [[TMP127:%.*]] = insertelement <128 x i8> [[TMP125]], i8 [[TMP126]], i32 15 +; OPT-NEXT: [[TMP128:%.*]] = extractelement <16 x i8> [[IN]], i64 0 +; OPT-NEXT: [[TMP129:%.*]] = insertelement <128 x i8> [[TMP127]], i8 [[TMP128]], i32 0 +; OPT-NEXT: [[TMP130:%.*]] = extractelement <16 x i8> [[IN]], i64 1 +; OPT-NEXT: [[TMP131:%.*]] = insertelement <128 x i8> [[TMP129]], i8 [[TMP130]], i32 1 +; OPT-NEXT: [[TMP132:%.*]] = extractelement <16 x i8> [[IN]], i64 2 +; OPT-NEXT: [[TMP133:%.*]] = insertelement <128 x i8> [[TMP131]], i8 [[TMP132]], i32 2 +; OPT-NEXT: [[TMP134:%.*]] = extractelement <16 x i8> [[IN]], i64 3 +; OPT-NEXT: [[TMP135:%.*]] = insertelement <128 x i8> [[TMP133]], i8 [[TMP134]], i32 3 +; OPT-NEXT: [[TMP136:%.*]] = extractelement <16 x i8> [[IN]], i64 4 +; OPT-NEXT: [[TMP137:%.*]] = insertelement <128 x i8> [[TMP135]], i8 [[TMP136]], i32 4 +; OPT-NEXT: [[TMP138:%.*]] = extractelement <16 x i8> [[IN]], i64 5 +; OPT-NEXT: [[TMP139:%.*]] = insertelement <128 x i8> [[TMP137]], i8 [[TMP138]], i32 5 +; OPT-NEXT: [[TMP140:%.*]] = extractelement <16 x i8> [[IN]], i64 6 +; OPT-NEXT: [[TMP141:%.*]] = insertelement <128 x i8> [[TMP139]], i8 [[TMP140]], i32 6 +; OPT-NEXT: [[TMP142:%.*]] = extractelement <16 x i8> [[IN]], i64 7 +; OPT-NEXT: [[TMP143:%.*]] = insertelement <128 x i8> [[TMP141]], i8 [[TMP142]], i32 7 +; OPT-NEXT: [[TMP144:%.*]] = extractelement <16 x i8> [[IN]], i64 8 +; OPT-NEXT: [[TMP145:%.*]] = insertelement <128 x i8> [[TMP143]], i8 [[TMP144]], i32 8 +; OPT-NEXT: [[TMP146:%.*]] = extractelement <16 x i8> [[IN]], i64 9 +; OPT-NEXT: [[TMP147:%.*]] = insertelement <128 x i8> [[TMP145]], i8 [[TMP146]], i32 9 +; OPT-NEXT: [[TMP148:%.*]] = extractelement <16 x i8> [[IN]], i64 10 +; OPT-NEXT: [[TMP149:%.*]] = insertelement <128 x i8> [[TMP147]], i8 [[TMP148]], i32 10 +; OPT-NEXT: [[TMP150:%.*]] = extractelement <16 x i8> [[IN]], i64 11 +; OPT-NEXT: [[TMP151:%.*]] = insertelement <128 x i8> [[TMP149]], i8 [[TMP150]], i32 11 +; OPT-NEXT: [[TMP152:%.*]] = extractelement <16 x i8> [[IN]], i64 12 +; OPT-NEXT: [[TMP153:%.*]] = insertelement <128 x i8> [[TMP151]], i8 [[TMP152]], i32 12 +; OPT-NEXT: [[TMP154:%.*]] = extractelement <16 x i8> [[IN]], i64 13 +; OPT-NEXT: [[TMP155:%.*]] = insertelement <128 x i8> [[TMP153]], i8 [[TMP154]], i32 13 +; OPT-NEXT: [[TMP156:%.*]] = extractelement <16 x i8> [[IN]], i64 14 +; OPT-NEXT: [[TMP157:%.*]] = insertelement <128 x i8> [[TMP155]], i8 [[TMP156]], i32 14 +; OPT-NEXT: [[TMP158:%.*]] = extractelement <16 x i8> [[IN]], i64 15 +; OPT-NEXT: [[TMP159:%.*]] = insertelement <128 x i8> [[TMP157]], i8 [[TMP158]], i32 15 +; OPT-NEXT: [[TMP160:%.*]] = extractelement <16 x i8> [[IN]], i64 0 +; OPT-NEXT: [[TMP161:%.*]] = insertelement <128 x i8> [[TMP159]], i8 [[TMP160]], i32 0 +; OPT-NEXT: [[TMP162:%.*]] = extractelement <16 x i8> [[IN]], i64 1 +; OPT-NEXT: [[TMP163:%.*]] = insertelement <128 x i8> [[TMP161]], i8 [[TMP162]], i32 1 +; OPT-NEXT: [[TMP164:%.*]] = extractelement <16 x i8> [[IN]], i64 2 +; OPT-NEXT: [[TMP165:%.*]] = insertelement <128 x i8> [[TMP163]], i8 [[TMP164]], i32 2 +; OPT-NEXT: [[TMP166:%.*]] = extractelement <16 x i8> [[IN]], i64 3 +; OPT-NEXT: [[TMP167:%.*]] = insertelement <128 x i8> [[TMP165]], i8 [[TMP166]], i32 3 +; OPT-NEXT: [[TMP168:%.*]] = extractelement <16 x i8> [[IN]], i64 4 +; OPT-NEXT: [[TMP169:%.*]] = insertelement <128 x i8> [[TMP167]], i8 [[TMP168]], i32 4 +; OPT-NEXT: [[TMP170:%.*]] = extractelement <16 x i8> [[IN]], i64 5 +; OPT-NEXT: [[TMP171:%.*]] = insertelement <128 x i8> [[TMP169]], i8 [[TMP170]], i32 5 +; OPT-NEXT: [[TMP172:%.*]] = extractelement <16 x i8> [[IN]], i64 6 +; OPT-NEXT: [[TMP173:%.*]] = insertelement <128 x i8> [[TMP171]], i8 [[TMP172]], i32 6 +; OPT-NEXT: [[TMP174:%.*]] = extractelement <16 x i8> [[IN]], i64 7 +; OPT-NEXT: [[TMP175:%.*]] = insertelement <128 x i8> [[TMP173]], i8 [[TMP174]], i32 7 +; OPT-NEXT: [[TMP176:%.*]] = extractelement <16 x i8> [[IN]], i64 8 +; OPT-NEXT: [[TMP177:%.*]] = insertelement <128 x i8> [[TMP175]], i8 [[TMP176]], i32 8 +; OPT-NEXT: [[TMP178:%.*]] = extractelement <16 x i8> [[IN]], i64 9 +; OPT-NEXT: [[TMP179:%.*]] = insertelement <128 x i8> [[TMP177]], i8 [[TMP178]], i32 9 +; OPT-NEXT: [[TMP180:%.*]] = extractelement <16 x i8> [[IN]], i64 10 +; OPT-NEXT: [[TMP181:%.*]] = insertelement <128 x i8> [[TMP179]], i8 [[TMP180]], i32 10 +; OPT-NEXT: [[TMP182:%.*]] = extractelement <16 x i8> [[IN]], i64 11 +; OPT-NEXT: [[TMP183:%.*]] = insertelement <128 x i8> [[TMP181]], i8 [[TMP182]], i32 11 +; OPT-NEXT: [[TMP184:%.*]] = extractelement <16 x i8> [[IN]], i64 12 +; OPT-NEXT: [[TMP185:%.*]] = insertelement <128 x i8> [[TMP183]], i8 [[TMP184]], i32 12 +; OPT-NEXT: [[TMP186:%.*]] = extractelement <16 x i8> [[IN]], i64 13 +; OPT-NEXT: [[TMP187:%.*]] = insertelement <128 x i8> [[TMP185]], i8 [[TMP186]], i32 13 +; OPT-NEXT: [[TMP188:%.*]] = extractelement <16 x i8> [[IN]], i64 14 +; OPT-NEXT: [[TMP189:%.*]] = insertelement <128 x i8> [[TMP187]], i8 [[TMP188]], i32 14 +; OPT-NEXT: [[TMP190:%.*]] = extractelement <16 x i8> [[IN]], i64 15 +; OPT-NEXT: [[TMP191:%.*]] = insertelement <128 x i8> [[TMP189]], i8 [[TMP190]], i32 15 +; OPT-NEXT: [[TMP192:%.*]] = extractelement <16 x i8> [[IN]], i64 0 +; OPT-NEXT: [[TMP193:%.*]] = insertelement <128 x i8> [[TMP191]], i8 [[TMP192]], i32 0 +; OPT-NEXT: [[TMP194:%.*]] = extractelement <16 x i8> [[IN]], i64 1 +; OPT-NEXT: [[TMP195:%.*]] = insertelement <128 x i8> [[TMP193]], i8 [[TMP194]], i32 1 +; OPT-NEXT: [[TMP196:%.*]] = extractelement <16 x i8> [[IN]], i64 2 +; OPT-NEXT: [[TMP197:%.*]] = insertelement <128 x i8> [[TMP195]], i8 [[TMP196]], i32 2 +; OPT-NEXT: [[TMP198:%.*]] = extractelement <16 x i8> [[IN]], i64 3 +; OPT-NEXT: [[TMP199:%.*]] = insertelement <128 x i8> [[TMP197]], i8 [[TMP198]], i32 3 +; OPT-NEXT: [[TMP200:%.*]] = extractelement <16 x i8> [[IN]], i64 4 +; OPT-NEXT: [[TMP201:%.*]] = insertelement <128 x i8> [[TMP199]], i8 [[TMP200]], i32 4 +; OPT-NEXT: [[TMP202:%.*]] = extractelement <16 x i8> [[IN]], i64 5 +; OPT-NEXT: [[TMP203:%.*]] = insertelement <128 x i8> [[TMP201]], i8 [[TMP202]], i32 5 +; OPT-NEXT: [[TMP204:%.*]] = extractelement <16 x i8> [[IN]], i64 6 +; OPT-NEXT: [[TMP205:%.*]] = insertelement <128 x i8> [[TMP203]], i8 [[TMP204]], i32 6 +; OPT-NEXT: [[TMP206:%.*]] = extractelement <16 x i8> [[IN]], i64 7 +; OPT-NEXT: [[TMP207:%.*]] = insertelement <128 x i8> [[TMP205]], i8 [[TMP206]], i32 7 +; OPT-NEXT: [[TMP208:%.*]] = extractelement <16 x i8> [[IN]], i64 8 +; OPT-NEXT: [[TMP209:%.*]] = insertelement <128 x i8> [[TMP207]], i8 [[TMP208]], i32 8 +; OPT-NEXT: [[TMP210:%.*]] = extractelement <16 x i8> [[IN]], i64 9 +; OPT-NEXT: [[TMP211:%.*]] = insertelement <128 x i8> [[TMP209]], i8 [[TMP210]], i32 9 +; OPT-NEXT: [[TMP212:%.*]] = extractelement <16 x i8> [[IN]], i64 10 +; OPT-NEXT: [[TMP213:%.*]] = insertelement <128 x i8> [[TMP211]], i8 [[TMP212]], i32 10 +; OPT-NEXT: [[TMP214:%.*]] = extractelement <16 x i8> [[IN]], i64 11 +; OPT-NEXT: [[TMP215:%.*]] = insertelement <128 x i8> [[TMP213]], i8 [[TMP214]], i32 11 +; OPT-NEXT: [[TMP216:%.*]] = extractelement <16 x i8> [[IN]], i64 12 +; OPT-NEXT: [[TMP217:%.*]] = insertelement <128 x i8> [[TMP215]], i8 [[TMP216]], i32 12 +; OPT-NEXT: [[TMP218:%.*]] = extractelement <16 x i8> [[IN]], i64 13 +; OPT-NEXT: [[TMP219:%.*]] = insertelement <128 x i8> [[TMP217]], i8 [[TMP218]], i32 13 +; OPT-NEXT: [[TMP220:%.*]] = extractelement <16 x i8> [[IN]], i64 14 +; OPT-NEXT: [[TMP221:%.*]] = insertelement <128 x i8> [[TMP219]], i8 [[TMP220]], i32 14 +; OPT-NEXT: [[TMP222:%.*]] = extractelement <16 x i8> [[IN]], i64 15 +; OPT-NEXT: [[TMP223:%.*]] = insertelement <128 x i8> [[TMP221]], i8 [[TMP222]], i32 15 +; OPT-NEXT: [[TMP224:%.*]] = extractelement <16 x i8> [[IN]], i64 0 +; OPT-NEXT: [[TMP225:%.*]] = insertelement <128 x i8> [[TMP223]], i8 [[TMP224]], i32 0 +; OPT-NEXT: [[TMP226:%.*]] = extractelement <16 x i8> [[IN]], i64 1 +; OPT-NEXT: [[TMP227:%.*]] = insertelement <128 x i8> [[TMP225]], i8 [[TMP226]], i32 1 +; OPT-NEXT: [[TMP228:%.*]] = extractelement <16 x i8> [[IN]], i64 2 +; OPT-NEXT: [[TMP229:%.*]] = insertelement <128 x i8> [[TMP227]], i8 [[TMP228]], i32 2 +; OPT-NEXT: [[TMP230:%.*]] = extractelement <16 x i8> [[IN]], i64 3 +; OPT-NEXT: [[TMP231:%.*]] = insertelement <128 x i8> [[TMP229]], i8 [[TMP230]], i32 3 +; OPT-NEXT: [[TMP232:%.*]] = extractelement <16 x i8> [[IN]], i64 4 +; OPT-NEXT: [[TMP233:%.*]] = insertelement <128 x i8> [[TMP231]], i8 [[TMP232]], i32 4 +; OPT-NEXT: [[TMP234:%.*]] = extractelement <16 x i8> [[IN]], i64 5 +; OPT-NEXT: [[TMP235:%.*]] = insertelement <128 x i8> [[TMP233]], i8 [[TMP234]], i32 5 +; OPT-NEXT: [[TMP236:%.*]] = extractelement <16 x i8> [[IN]], i64 6 +; OPT-NEXT: [[TMP237:%.*]] = insertelement <128 x i8> [[TMP235]], i8 [[TMP236]], i32 6 +; OPT-NEXT: [[TMP238:%.*]] = extractelement <16 x i8> [[IN]], i64 7 +; OPT-NEXT: [[TMP239:%.*]] = insertelement <128 x i8> [[TMP237]], i8 [[TMP238]], i32 7 +; OPT-NEXT: [[TMP240:%.*]] = extractelement <16 x i8> [[IN]], i64 8 +; OPT-NEXT: [[TMP241:%.*]] = insertelement <128 x i8> [[TMP239]], i8 [[TMP240]], i32 8 +; OPT-NEXT: [[TMP242:%.*]] = extractelement <16 x i8> [[IN]], i64 9 +; OPT-NEXT: [[TMP243:%.*]] = insertelement <128 x i8> [[TMP241]], i8 [[TMP242]], i32 9 +; OPT-NEXT: [[TMP244:%.*]] = extractelement <16 x i8> [[IN]], i64 10 +; OPT-NEXT: [[TMP245:%.*]] = insertelement <128 x i8> [[TMP243]], i8 [[TMP244]], i32 10 +; OPT-NEXT: [[TMP246:%.*]] = extractelement <16 x i8> [[IN]], i64 11 +; OPT-NEXT: [[TMP247:%.*]] = insertelement <128 x i8> [[TMP245]], i8 [[TMP246]], i32 11 +; OPT-NEXT: [[TMP248:%.*]] = extractelement <16 x i8> [[IN]], i64 12 +; OPT-NEXT: [[TMP249:%.*]] = insertelement <128 x i8> [[TMP247]], i8 [[TMP248]], i32 12 +; OPT-NEXT: [[TMP250:%.*]] = extractelement <16 x i8> [[IN]], i64 13 +; OPT-NEXT: [[TMP251:%.*]] = insertelement <128 x i8> [[TMP249]], i8 [[TMP250]], i32 13 +; OPT-NEXT: [[TMP252:%.*]] = extractelement <16 x i8> [[IN]], i64 14 +; OPT-NEXT: [[TMP253:%.*]] = insertelement <128 x i8> [[TMP251]], i8 [[TMP252]], i32 14 +; OPT-NEXT: [[TMP254:%.*]] = extractelement <16 x i8> [[IN]], i64 15 +; OPT-NEXT: [[TMP255:%.*]] = insertelement <128 x i8> [[TMP253]], i8 [[TMP254]], i32 15 +; OPT-NEXT: [[TMP256:%.*]] = extractelement <128 x i8> [[TMP255]], i32 80 +; OPT-NEXT: [[TMP257:%.*]] = insertelement <16 x i8> poison, i8 [[TMP256]], i64 0 +; OPT-NEXT: [[TMP258:%.*]] = extractelement <128 x i8> [[TMP255]], i32 81 +; OPT-NEXT: [[TMP259:%.*]] = insertelement <16 x i8> [[TMP257]], i8 [[TMP258]], i64 1 +; OPT-NEXT: [[TMP260:%.*]] = extractelement <128 x i8> [[TMP255]], i32 82 +; OPT-NEXT: [[TMP261:%.*]] = insertelement <16 x i8> [[TMP259]], i8 [[TMP260]], i64 2 +; OPT-NEXT: [[TMP262:%.*]] = extractelement <128 x i8> [[TMP255]], i32 83 +; OPT-NEXT: [[TMP263:%.*]] = insertelement <16 x i8> [[TMP261]], i8 [[TMP262]], i64 3 +; OPT-NEXT: [[TMP264:%.*]] = extractelement <128 x i8> [[TMP255]], i32 84 +; OPT-NEXT: [[TMP265:%.*]] = insertelement <16 x i8> [[TMP263]], i8 [[TMP264]], i64 4 +; OPT-NEXT: [[TMP266:%.*]] = extractelement <128 x i8> [[TMP255]], i32 85 +; OPT-NEXT: [[TMP267:%.*]] = insertelement <16 x i8> [[TMP265]], i8 [[TMP266]], i64 5 +; OPT-NEXT: [[TMP268:%.*]] = extractelement <128 x i8> [[TMP255]], i32 86 +; OPT-NEXT: [[TMP269:%.*]] = insertelement <16 x i8> [[TMP267]], i8 [[TMP268]], i64 6 +; OPT-NEXT: [[TMP270:%.*]] = extractelement <128 x i8> [[TMP255]], i32 87 +; OPT-NEXT: [[TMP271:%.*]] = insertelement <16 x i8> [[TMP269]], i8 [[TMP270]], i64 7 +; OPT-NEXT: [[TMP272:%.*]] = extractelement <128 x i8> [[TMP255]], i32 88 +; OPT-NEXT: [[TMP273:%.*]] = insertelement <16 x i8> [[TMP271]], i8 [[TMP272]], i64 8 +; OPT-NEXT: [[TMP274:%.*]] = extractelement <128 x i8> [[TMP255]], i32 89 +; OPT-NEXT: [[TMP275:%.*]] = insertelement <16 x i8> [[TMP273]], i8 [[TMP274]], i64 9 +; OPT-NEXT: [[TMP276:%.*]] = extractelement <128 x i8> [[TMP255]], i32 90 +; OPT-NEXT: [[TMP277:%.*]] = insertelement <16 x i8> [[TMP275]], i8 [[TMP276]], i64 10 +; OPT-NEXT: [[TMP278:%.*]] = extractelement <128 x i8> [[TMP255]], i32 91 +; OPT-NEXT: [[TMP279:%.*]] = insertelement <16 x i8> [[TMP277]], i8 [[TMP278]], i64 11 +; OPT-NEXT: [[TMP280:%.*]] = extractelement <128 x i8> [[TMP255]], i32 92 +; OPT-NEXT: [[TMP281:%.*]] = insertelement <16 x i8> [[TMP279]], i8 [[TMP280]], i64 12 +; OPT-NEXT: [[TMP282:%.*]] = extractelement <128 x i8> [[TMP255]], i32 93 +; OPT-NEXT: [[TMP283:%.*]] = insertelement <16 x i8> [[TMP281]], i8 [[TMP282]], i64 13 +; OPT-NEXT: [[TMP284:%.*]] = extractelement <128 x i8> [[TMP255]], i32 94 +; OPT-NEXT: [[TMP285:%.*]] = insertelement <16 x i8> [[TMP283]], i8 [[TMP284]], i64 14 +; OPT-NEXT: [[TMP286:%.*]] = extractelement <128 x i8> [[TMP255]], i32 95 +; OPT-NEXT: [[TMP287:%.*]] = insertelement <16 x i8> [[TMP285]], i8 [[TMP286]], i64 15 +; OPT-NEXT: [[SUM:%.*]] = add <16 x i8> [[TMP287]], [[ADD]] +; OPT-NEXT: store <16 x i8> [[SUM]], ptr addrspace(3) [[OUT]], align 16 +; OPT-NEXT: ret void +; +entry: + %alloca = alloca [8 x <16 x i8>], align 16, addrspace(5) + %gep0 = getelementptr [8 x <16 x i8>], ptr addrspace(5) %alloca, i64 0, i64 0 + store <16 x i8> %in, ptr addrspace(5) %gep0, align 16 + %gep1 = getelementptr [8 x <16 x i8>], ptr addrspace(5) %alloca, i64 0, i64 1 + store <16 x i8> %in, ptr addrspace(5) %gep0, align 16 + %gep2 = getelementptr [8 x <16 x i8>], ptr addrspace(5) %alloca, i64 0, i64 2 + store <16 x i8> %in, ptr addrspace(5) %gep0, align 16 + %gep3 = getelementptr [8 x <16 x i8>], ptr addrspace(5) %alloca, i64 0, i64 3 + store <16 x i8> %in, ptr addrspace(5) %gep0, align 16 + %gep4 = getelementptr [8 x <16 x i8>], ptr addrspace(5) %alloca, i64 0, i64 4 + store <16 x i8> %in, ptr addrspace(5) %gep0, align 16 + %gep5 = getelementptr [8 x <16 x i8>], ptr addrspace(5) %alloca, i64 0, i64 5 + store <16 x i8> %in, ptr addrspace(5) %gep0, align 16 + %gep6 = getelementptr [8 x <16 x i8>], ptr addrspace(5) %alloca, i64 0, i64 6 + store <16 x i8> %in, ptr addrspace(5) %gep0, align 16 + %gep7 = getelementptr [8 x <16 x i8>], ptr addrspace(5) %alloca, i64 0, i64 7 + store <16 x i8> %in, ptr addrspace(5) %gep0, align 16 + %load = load <16 x i8>, ptr addrspace(5) %gep5, align 16 + %sum = add <16 x i8> %load, %add + store <16 x i8> %sum, ptr addrspace(3) %out, align 16 + ret void +} + +attributes #0 = {"amdgpu-waves-per-eu"="2,2"} diff --git a/llvm/test/CodeGen/AMDGPU/readcyclecounter.ll b/llvm/test/CodeGen/AMDGPU/readcyclecounter.ll index f67cbe3..ddb522a8 100644 --- a/llvm/test/CodeGen/AMDGPU/readcyclecounter.ll +++ b/llvm/test/CodeGen/AMDGPU/readcyclecounter.ll @@ -1,17 +1,17 @@ ; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=tahiti < %s | FileCheck -check-prefix=MEMTIME -check-prefix=SIVI -check-prefix=GCN %s ; -global-isel=1 SI run line skipped since store not yet implemented. ; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=MEMTIME -check-prefix=SIVI -check-prefix=GCN %s -; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=MEMTIME -check-prefix=SIVI -check-prefix=GCN %s +; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=MEMTIME -check-prefix=SIVI -check-prefix=GCN %s ; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1010 < %s | FileCheck -check-prefix=MEMTIME -check-prefix=GCN %s -; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1010 < %s | FileCheck -check-prefix=MEMTIME -check-prefix=GCN %s +; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1010 < %s | FileCheck -check-prefix=MEMTIME -check-prefix=GCN %s ; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1030 < %s | FileCheck -check-prefixes=MEMTIME -check-prefix=GCN %s -; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1030 < %s | FileCheck -check-prefixes=MEMTIME -check-prefix=GCN %s +; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1030 < %s | FileCheck -check-prefixes=MEMTIME -check-prefix=GCN %s ; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1100 -amdgpu-enable-vopd=0 < %s | FileCheck -check-prefixes=GETREG,GETREG-SDAG -check-prefix=GCN %s -; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1100 -amdgpu-enable-vopd=0 < %s | FileCheck -check-prefixes=GETREG,GETREG-GISEL -check-prefix=GCN %s +; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1100 -amdgpu-enable-vopd=0 < %s | FileCheck -check-prefixes=GETREG,GETREG-GISEL -check-prefix=GCN %s ; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1200 < %s | FileCheck -check-prefixes=GCN,GFX12 %s -; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1200 < %s | FileCheck -check-prefixes=GCN,GFX12 %s +; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1200 < %s | FileCheck -check-prefixes=GCN,GFX12 %s ; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck -check-prefixes=GCN,GFX1250 %s -; RUN: llc -global-isel=1 -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck -check-prefixes=GCN,GFX1250 %s +; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1250 < %s | FileCheck -check-prefixes=GCN,GFX1250 %s declare i64 @llvm.readcyclecounter() #0 diff --git a/llvm/test/CodeGen/AMDGPU/schedule-amdgpu-trackers.ll b/llvm/test/CodeGen/AMDGPU/schedule-amdgpu-trackers.ll index c573253..48ed5c4 100644 --- a/llvm/test/CodeGen/AMDGPU/schedule-amdgpu-trackers.ll +++ b/llvm/test/CodeGen/AMDGPU/schedule-amdgpu-trackers.ll @@ -73,10 +73,10 @@ define amdgpu_kernel void @constant_zextload_v64i16_to_v64i32(ptr addrspace(1) % } ; CHECK-LABEL: {{^}}excess_soft_clause_reg_pressure: -; GFX908: NumSgprs: 64 -; GFX908-GCNTRACKERS: NumSgprs: 64 +; GFX908: NumSgprs: 56 +; GFX908-GCNTRACKERS: NumSgprs: 56 ; GFX908: NumVgprs: 43 -; GFX908-GCNTRACKERS: NumVgprs: 39 +; GFX908-GCNTRACKERS: NumVgprs: 40 ; GFX908: Occupancy: 5 ; GFX908-GCNTRACKERS: Occupancy: 6 diff --git a/llvm/test/CodeGen/AMDGPU/spill-vgpr-to-agpr-update-regscavenger.ll b/llvm/test/CodeGen/AMDGPU/spill-vgpr-to-agpr-update-regscavenger.ll index 586579f..ef96944 100644 --- a/llvm/test/CodeGen/AMDGPU/spill-vgpr-to-agpr-update-regscavenger.ll +++ b/llvm/test/CodeGen/AMDGPU/spill-vgpr-to-agpr-update-regscavenger.ll @@ -20,38 +20,33 @@ define void @test() { ; CHECK-NEXT: ; in Loop: Header=BB0_1 Depth=1 ; CHECK-NEXT: .LBB0_3: ; %bb.3 ; CHECK-NEXT: ; in Loop: Header=BB0_1 Depth=1 -; CHECK-NEXT: ; implicit-def: $sgpr4 -; CHECK-NEXT: v_mov_b32_e32 v0, s4 -; CHECK-NEXT: v_readfirstlane_b32 s6, v0 ; CHECK-NEXT: s_mov_b64 s[4:5], -1 -; CHECK-NEXT: s_mov_b32 s7, 0 -; CHECK-NEXT: s_cmp_eq_u32 s6, s7 ; CHECK-NEXT: ; implicit-def: $vgpr1 : SGPR spill to VGPR lane ; CHECK-NEXT: v_writelane_b32 v1, s4, 0 ; CHECK-NEXT: v_writelane_b32 v1, s5, 1 -; CHECK-NEXT: s_mov_b64 s[10:11], exec -; CHECK-NEXT: s_mov_b64 exec, -1 +; CHECK-NEXT: s_or_saveexec_b64 s[8:9], -1 +; CHECK-NEXT: s_nop 0 ; CHECK-NEXT: v_accvgpr_write_b32 a0, v1 ; Reload Reuse -; CHECK-NEXT: s_mov_b64 exec, s[10:11] +; CHECK-NEXT: s_mov_b64 exec, s[8:9] ; CHECK-NEXT: s_cbranch_scc1 .LBB0_5 ; CHECK-NEXT: ; %bb.4: ; %bb.4 ; CHECK-NEXT: ; in Loop: Header=BB0_1 Depth=1 -; CHECK-NEXT: s_or_saveexec_b64 s[10:11], -1 +; CHECK-NEXT: s_or_saveexec_b64 s[8:9], -1 ; CHECK-NEXT: v_accvgpr_read_b32 v1, a0 ; Reload Reuse -; CHECK-NEXT: s_mov_b64 exec, s[10:11] +; CHECK-NEXT: s_mov_b64 exec, s[8:9] ; CHECK-NEXT: s_mov_b64 s[4:5], 0 ; CHECK-NEXT: v_writelane_b32 v1, s4, 0 ; CHECK-NEXT: v_writelane_b32 v1, s5, 1 -; CHECK-NEXT: s_or_saveexec_b64 s[10:11], -1 +; CHECK-NEXT: s_or_saveexec_b64 s[8:9], -1 ; CHECK-NEXT: s_nop 0 ; CHECK-NEXT: v_accvgpr_write_b32 a0, v1 ; Reload Reuse -; CHECK-NEXT: s_mov_b64 exec, s[10:11] +; CHECK-NEXT: s_mov_b64 exec, s[8:9] ; CHECK-NEXT: .LBB0_5: ; %Flow ; CHECK-NEXT: ; in Loop: Header=BB0_1 Depth=1 -; CHECK-NEXT: s_or_saveexec_b64 s[10:11], -1 +; CHECK-NEXT: s_or_saveexec_b64 s[8:9], -1 ; CHECK-NEXT: s_nop 0 ; CHECK-NEXT: v_accvgpr_read_b32 v1, a0 ; Reload Reuse -; CHECK-NEXT: s_mov_b64 exec, s[10:11] +; CHECK-NEXT: s_mov_b64 exec, s[8:9] ; CHECK-NEXT: v_readlane_b32 s4, v1, 0 ; CHECK-NEXT: v_readlane_b32 s5, v1, 1 ; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[4:5] diff --git a/llvm/test/CodeGen/AMDGPU/spill_kill_v16.mir b/llvm/test/CodeGen/AMDGPU/spill_kill_v16.mir index 0c694d9..6989583 100644 --- a/llvm/test/CodeGen/AMDGPU/spill_kill_v16.mir +++ b/llvm/test/CodeGen/AMDGPU/spill_kill_v16.mir @@ -1,5 +1,6 @@ # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py # RUN: llc -march=amdgcn -verify-machineinstrs -mcpu=gfx1100 -mattr=+real-true16 -run-pass=prologepilog -o - %s | FileCheck -check-prefix=EXPANDED %s +# RUN: llc -march=amdgcn -verify-machineinstrs -mcpu=gfx1250 -mattr=+real-true16 -run-pass=prologepilog -o - %s | FileCheck -check-prefix=SRAMECC-EXPANDED %s --- name: spill_restore_vgpr16 @@ -31,6 +32,28 @@ body: | ; EXPANDED-NEXT: $vgpr0_lo16 = SCRATCH_LOAD_SHORT_D16_SADDR_t16 $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.0, align 4, addrspace 5) ; EXPANDED-NEXT: $vgpr0_hi16 = SCRATCH_LOAD_SHORT_D16_SADDR_t16 $sgpr32, 4, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.1, align 4, addrspace 5) ; EXPANDED-NEXT: S_NOP 0, implicit killed renamable $vgpr0_lo16, implicit killed renamable $vgpr0_hi16 + ; + ; SRAMECC-EXPANDED-LABEL: name: spill_restore_vgpr16 + ; SRAMECC-EXPANDED: bb.0: + ; SRAMECC-EXPANDED-NEXT: successors: %bb.1(0x80000000) + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: S_NOP 0, implicit-def renamable $vgpr0_lo16, implicit-def renamable $vgpr0_hi16 + ; SRAMECC-EXPANDED-NEXT: SCRATCH_STORE_SHORT_SADDR_t16 killed $vgpr0_hi16, $sgpr32, 4, 0, implicit $exec, implicit $flat_scr :: (store (s16) into %stack.1, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: S_NOP 0, implicit renamable $vgpr0_lo16 + ; SRAMECC-EXPANDED-NEXT: SCRATCH_STORE_SHORT_SADDR_t16 killed $vgpr0_lo16, $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (store (s16) into %stack.0, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: S_CBRANCH_SCC1 %bb.1, implicit undef $scc + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: bb.1: + ; SRAMECC-EXPANDED-NEXT: successors: %bb.2(0x80000000) + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: S_NOP 1 + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: bb.2: + ; SRAMECC-EXPANDED-NEXT: $vgpr1 = SCRATCH_LOAD_USHORT_SADDR $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.0, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: $vgpr0_lo16 = V_MOV_B16_t16_e64 0, killed $vgpr1_lo16, 0, implicit $exec + ; SRAMECC-EXPANDED-NEXT: $vgpr1 = SCRATCH_LOAD_USHORT_SADDR $sgpr32, 4, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.1, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: $vgpr0_hi16 = V_MOV_B16_t16_e64 0, killed $vgpr1_lo16, 0, implicit $exec + ; SRAMECC-EXPANDED-NEXT: S_NOP 0, implicit killed renamable $vgpr0_lo16, implicit killed renamable $vgpr0_hi16 bb.0: successors: %bb.1(0x80000000) S_NOP 0, implicit-def renamable $vgpr0_lo16, implicit-def renamable $vgpr0_hi16 @@ -78,6 +101,29 @@ body: | ; EXPANDED-NEXT: $vgpr0_lo16 = SCRATCH_LOAD_SHORT_D16_SADDR_t16 $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.0, align 4, addrspace 5) ; EXPANDED-NEXT: $vgpr0_hi16 = SCRATCH_LOAD_SHORT_D16_SADDR_t16 $sgpr32, 4, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.1, align 4, addrspace 5) ; EXPANDED-NEXT: S_NOP 0, implicit killed renamable $vgpr0_lo16, implicit killed renamable $vgpr0_hi16 + ; + ; SRAMECC-EXPANDED-LABEL: name: spill_restore_vgpr16_middle_of_block + ; SRAMECC-EXPANDED: bb.0: + ; SRAMECC-EXPANDED-NEXT: successors: %bb.1(0x80000000) + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: S_NOP 0, implicit-def renamable $vgpr0_lo16, implicit-def renamable $vgpr0_hi16 + ; SRAMECC-EXPANDED-NEXT: SCRATCH_STORE_SHORT_SADDR_t16 killed $vgpr0_hi16, $sgpr32, 4, 0, implicit $exec, implicit $flat_scr :: (store (s16) into %stack.1, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: S_NOP 0, implicit renamable $vgpr0_lo16 + ; SRAMECC-EXPANDED-NEXT: SCRATCH_STORE_SHORT_SADDR_t16 killed $vgpr0_lo16, $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (store (s16) into %stack.0, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: S_CBRANCH_SCC1 %bb.1, implicit undef $scc + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: bb.1: + ; SRAMECC-EXPANDED-NEXT: successors: %bb.2(0x80000000) + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: S_NOP 1 + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: bb.2: + ; SRAMECC-EXPANDED-NEXT: S_NOP 1 + ; SRAMECC-EXPANDED-NEXT: $vgpr1 = SCRATCH_LOAD_USHORT_SADDR $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.0, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: $vgpr0_lo16 = V_MOV_B16_t16_e64 0, killed $vgpr1_lo16, 0, implicit $exec + ; SRAMECC-EXPANDED-NEXT: $vgpr1 = SCRATCH_LOAD_USHORT_SADDR $sgpr32, 4, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.1, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: $vgpr0_hi16 = V_MOV_B16_t16_e64 0, killed $vgpr1_lo16, 0, implicit $exec + ; SRAMECC-EXPANDED-NEXT: S_NOP 0, implicit killed renamable $vgpr0_lo16, implicit killed renamable $vgpr0_hi16 bb.0: successors: %bb.1(0x80000000) S_NOP 0, implicit-def renamable $vgpr0_lo16, implicit-def renamable $vgpr0_hi16 @@ -124,6 +170,27 @@ body: | ; EXPANDED-NEXT: bb.2: ; EXPANDED-NEXT: $vgpr0_lo16 = SCRATCH_LOAD_SHORT_D16_SADDR_t16 $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.0, align 4, addrspace 5) ; EXPANDED-NEXT: $vgpr0_hi16 = SCRATCH_LOAD_SHORT_D16_SADDR_t16 $sgpr32, 4, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.1, align 4, addrspace 5) + ; + ; SRAMECC-EXPANDED-LABEL: name: spill_restore_vgpr16_end_of_block + ; SRAMECC-EXPANDED: bb.0: + ; SRAMECC-EXPANDED-NEXT: successors: %bb.1(0x80000000) + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: S_NOP 0, implicit-def renamable $vgpr0_lo16, implicit-def renamable $vgpr0_hi16 + ; SRAMECC-EXPANDED-NEXT: SCRATCH_STORE_SHORT_SADDR_t16 killed $vgpr0_hi16, $sgpr32, 4, 0, implicit $exec, implicit $flat_scr :: (store (s16) into %stack.1, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: S_NOP 0, implicit renamable $vgpr0_lo16 + ; SRAMECC-EXPANDED-NEXT: SCRATCH_STORE_SHORT_SADDR_t16 killed $vgpr0_lo16, $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (store (s16) into %stack.0, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: S_CBRANCH_SCC1 %bb.1, implicit undef $scc + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: bb.1: + ; SRAMECC-EXPANDED-NEXT: successors: %bb.2(0x80000000) + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: S_NOP 1 + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: bb.2: + ; SRAMECC-EXPANDED-NEXT: $vgpr1 = SCRATCH_LOAD_USHORT_SADDR $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.0, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: $vgpr0_lo16 = V_MOV_B16_t16_e64 0, killed $vgpr1_lo16, 0, implicit $exec + ; SRAMECC-EXPANDED-NEXT: $vgpr1 = SCRATCH_LOAD_USHORT_SADDR $sgpr32, 4, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.1, align 4, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: $vgpr0_hi16 = V_MOV_B16_t16_e64 0, killed $vgpr1_lo16, 0, implicit $exec bb.0: successors: %bb.1(0x80000000) S_NOP 0, implicit-def renamable $vgpr0_lo16, implicit-def renamable $vgpr0_hi16 diff --git a/llvm/test/CodeGen/AMDGPU/spillv16.ll b/llvm/test/CodeGen/AMDGPU/spillv16.ll index 0e45df2..2d54ac8 100644 --- a/llvm/test/CodeGen/AMDGPU/spillv16.ll +++ b/llvm/test/CodeGen/AMDGPU/spillv16.ll @@ -1,6 +1,8 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=+real-true16 -enable-misched=0 -post-RA-scheduler=0 -stress-regalloc=8 < %s | FileCheck %s -check-prefixes=GCN,GCN-TRUE16 ; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-real-true16 -enable-misched=0 -post-RA-scheduler=0 -stress-regalloc=8 < %s | FileCheck %s -check-prefixes=GCN,GCN-FAKE16 +; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1250 -mattr=+real-true16 -enable-misched=0 -post-RA-scheduler=0 -stress-regalloc=8 < %s | FileCheck %s -check-prefixes=GFX1250,GFX1250-TRUE16 +; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1250 -mattr=-real-true16 -enable-misched=0 -post-RA-scheduler=0 -stress-regalloc=8 < %s | FileCheck %s -check-prefixes=GFX1250,GFX1250-FAKE16 define void @spill_i16_alu() { ; GCN-TRUE16-LABEL: spill_i16_alu: @@ -32,6 +34,41 @@ define void @spill_i16_alu() { ; GCN-FAKE16-NEXT: scratch_store_b16 off, v0, s32 dlc ; GCN-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0 ; GCN-FAKE16-NEXT: s_setpc_b64 s[30:31] +; +; GFX1250-TRUE16-LABEL: spill_i16_alu: +; GFX1250-TRUE16: ; %bb.0: ; %entry +; GFX1250-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-TRUE16-NEXT: s_wait_kmcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_load_u16 v0, off, s32 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x7b, v0.l +; GFX1250-TRUE16-NEXT: scratch_store_b16 off, v0, s32 offset:2 ; 2-byte Folded Spill +; GFX1250-TRUE16-NEXT: s_wait_xcnt 0x0 +; GFX1250-TRUE16-NEXT: ;;#ASMSTART +; GFX1250-TRUE16-NEXT: ;;#ASMEND +; GFX1250-TRUE16-NEXT: scratch_load_u16 v1, off, s32 offset:2 th:TH_LOAD_LU ; 2-byte Folded Reload +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: v_mov_b16_e32 v0.l, v1.l +; GFX1250-TRUE16-NEXT: scratch_store_b16 off, v0, s32 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-TRUE16-NEXT: s_set_pc_i64 s[30:31] +; +; GFX1250-FAKE16-LABEL: spill_i16_alu: +; GFX1250-FAKE16: ; %bb.0: ; %entry +; GFX1250-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-FAKE16-NEXT: s_wait_kmcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_load_u16 v0, off, s32 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: v_add_nc_u16 v0, 0x7b, v0 +; GFX1250-FAKE16-NEXT: scratch_store_b32 off, v0, s32 offset:4 ; 4-byte Folded Spill +; GFX1250-FAKE16-NEXT: s_wait_xcnt 0x0 +; GFX1250-FAKE16-NEXT: ;;#ASMSTART +; GFX1250-FAKE16-NEXT: ;;#ASMEND +; GFX1250-FAKE16-NEXT: scratch_load_b32 v0, off, s32 offset:4 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_store_b16 off, v0, s32 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-FAKE16-NEXT: s_set_pc_i64 s[30:31] entry: %alloca = alloca i16, i32 1, align 4, addrspace(5) @@ -88,6 +125,51 @@ define void @spill_i16_alu_two_vals() { ; GCN-FAKE16-NEXT: scratch_store_b16 off, v0, s32 offset:4 dlc ; GCN-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0 ; GCN-FAKE16-NEXT: s_setpc_b64 s[30:31] +; +; GFX1250-TRUE16-LABEL: spill_i16_alu_two_vals: +; GFX1250-TRUE16: ; %bb.0: ; %entry +; GFX1250-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-TRUE16-NEXT: s_wait_kmcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_load_u16 v0, off, s32 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x7b, v0.l +; GFX1250-TRUE16-NEXT: scratch_store_b16 off, v0, s32 offset:6 ; 2-byte Folded Spill +; GFX1250-TRUE16-NEXT: s_wait_xcnt 0x0 +; GFX1250-TRUE16-NEXT: ;;#ASMSTART +; GFX1250-TRUE16-NEXT: ;;#ASMEND +; GFX1250-TRUE16-NEXT: scratch_load_u16 v0, off, s32 offset:4 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_load_u16 v1, off, s32 offset:6 th:TH_LOAD_LU ; 2-byte Folded Reload +; GFX1250-TRUE16-NEXT: v_add_nc_u16 v0.l, 0x7b, v0.l +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: v_mov_b16_e32 v0.h, v1.l +; GFX1250-TRUE16-NEXT: scratch_store_d16_hi_b16 off, v0, s32 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_store_b16 off, v0, s32 offset:4 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-TRUE16-NEXT: s_set_pc_i64 s[30:31] +; +; GFX1250-FAKE16-LABEL: spill_i16_alu_two_vals: +; GFX1250-FAKE16: ; %bb.0: ; %entry +; GFX1250-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-FAKE16-NEXT: s_wait_kmcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_load_u16 v0, off, s32 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: v_add_nc_u16 v0, 0x7b, v0 +; GFX1250-FAKE16-NEXT: scratch_store_b32 off, v0, s32 offset:8 ; 4-byte Folded Spill +; GFX1250-FAKE16-NEXT: s_wait_xcnt 0x0 +; GFX1250-FAKE16-NEXT: ;;#ASMSTART +; GFX1250-FAKE16-NEXT: ;;#ASMEND +; GFX1250-FAKE16-NEXT: scratch_load_u16 v0, off, s32 offset:4 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_load_b32 v1, off, s32 offset:8 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-FAKE16-NEXT: v_add_nc_u16 v0, 0x7b, v0 +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_store_b16 off, v1, s32 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_store_b16 off, v0, s32 offset:4 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-FAKE16-NEXT: s_set_pc_i64 s[30:31] entry: %alloca = alloca i16, i32 1, align 4, addrspace(5) %alloca2 = alloca i16, i32 1, align 4, addrspace(5) @@ -140,6 +222,22 @@ define void @spill_i16() { ; GCN-FAKE16-NEXT: scratch_store_b16 off, v0, s32 dlc ; GCN-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0 ; GCN-FAKE16-NEXT: s_setpc_b64 s[30:31] +; +; GFX1250-LABEL: spill_i16: +; GFX1250: ; %bb.0: ; %entry +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 +; GFX1250-NEXT: scratch_load_u16 v0, off, s32 scope:SCOPE_SYS +; GFX1250-NEXT: s_wait_loadcnt 0x0 +; GFX1250-NEXT: scratch_store_b32 off, v0, s32 offset:4 ; 4-byte Folded Spill +; GFX1250-NEXT: s_wait_xcnt 0x0 +; GFX1250-NEXT: ;;#ASMSTART +; GFX1250-NEXT: ;;#ASMEND +; GFX1250-NEXT: scratch_load_b32 v0, off, s32 offset:4 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-NEXT: s_wait_loadcnt 0x0 +; GFX1250-NEXT: scratch_store_b16 off, v0, s32 scope:SCOPE_SYS +; GFX1250-NEXT: s_wait_storecnt 0x0 +; GFX1250-NEXT: s_set_pc_i64 s[30:31] entry: %alloca = alloca i16, i32 1, align 4, addrspace(5) @@ -183,6 +281,22 @@ define void @spill_half() { ; GCN-FAKE16-NEXT: scratch_store_b16 off, v0, s32 dlc ; GCN-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0 ; GCN-FAKE16-NEXT: s_setpc_b64 s[30:31] +; +; GFX1250-LABEL: spill_half: +; GFX1250: ; %bb.0: ; %entry +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 +; GFX1250-NEXT: scratch_load_u16 v0, off, s32 scope:SCOPE_SYS +; GFX1250-NEXT: s_wait_loadcnt 0x0 +; GFX1250-NEXT: scratch_store_b32 off, v0, s32 offset:4 ; 4-byte Folded Spill +; GFX1250-NEXT: s_wait_xcnt 0x0 +; GFX1250-NEXT: ;;#ASMSTART +; GFX1250-NEXT: ;;#ASMEND +; GFX1250-NEXT: scratch_load_b32 v0, off, s32 offset:4 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-NEXT: s_wait_loadcnt 0x0 +; GFX1250-NEXT: scratch_store_b16 off, v0, s32 scope:SCOPE_SYS +; GFX1250-NEXT: s_wait_storecnt 0x0 +; GFX1250-NEXT: s_set_pc_i64 s[30:31] entry: %alloca = alloca half, i32 1, align 4, addrspace(5) @@ -226,6 +340,22 @@ define void @spill_i16_from_v2i16() { ; GCN-FAKE16-NEXT: scratch_store_b16 off, v0, s32 offset:2 dlc ; GCN-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0 ; GCN-FAKE16-NEXT: s_setpc_b64 s[30:31] +; +; GFX1250-LABEL: spill_i16_from_v2i16: +; GFX1250: ; %bb.0: ; %entry +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 +; GFX1250-NEXT: scratch_load_u16 v0, off, s32 offset:2 scope:SCOPE_SYS +; GFX1250-NEXT: s_wait_loadcnt 0x0 +; GFX1250-NEXT: scratch_store_b32 off, v0, s32 offset:8 ; 4-byte Folded Spill +; GFX1250-NEXT: s_wait_xcnt 0x0 +; GFX1250-NEXT: ;;#ASMSTART +; GFX1250-NEXT: ;;#ASMEND +; GFX1250-NEXT: scratch_load_b32 v0, off, s32 offset:8 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-NEXT: s_wait_loadcnt 0x0 +; GFX1250-NEXT: scratch_store_b16 off, v0, s32 offset:2 scope:SCOPE_SYS +; GFX1250-NEXT: s_wait_storecnt 0x0 +; GFX1250-NEXT: s_set_pc_i64 s[30:31] entry: %alloca = alloca <2 x i16>, i32 2, align 1, addrspace(5) @@ -283,6 +413,54 @@ define void @spill_2xi16_from_v2i16() { ; GCN-FAKE16-NEXT: scratch_store_b16 off, v0, s32 dlc ; GCN-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0 ; GCN-FAKE16-NEXT: s_setpc_b64 s[30:31] +; +; GFX1250-TRUE16-LABEL: spill_2xi16_from_v2i16: +; GFX1250-TRUE16: ; %bb.0: ; %entry +; GFX1250-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-TRUE16-NEXT: s_wait_kmcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_load_u16 v0, off, s32 offset:2 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: s_clause 0x1 +; GFX1250-TRUE16-NEXT: scratch_store_b32 off, v0, s32 offset:12 +; GFX1250-TRUE16-NEXT: scratch_load_u16 v0, off, s32 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_store_b32 off, v0, s32 offset:8 ; 4-byte Folded Spill +; GFX1250-TRUE16-NEXT: s_wait_xcnt 0x0 +; GFX1250-TRUE16-NEXT: ;;#ASMSTART +; GFX1250-TRUE16-NEXT: ;;#ASMEND +; GFX1250-TRUE16-NEXT: scratch_load_b32 v0, off, s32 offset:12 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_store_b16 off, v0, s32 offset:2 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_load_b32 v0, off, s32 offset:8 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_store_b16 off, v0, s32 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-TRUE16-NEXT: s_set_pc_i64 s[30:31] +; +; GFX1250-FAKE16-LABEL: spill_2xi16_from_v2i16: +; GFX1250-FAKE16: ; %bb.0: ; %entry +; GFX1250-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-FAKE16-NEXT: s_wait_kmcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_load_u16 v0, off, s32 offset:2 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: s_clause 0x1 +; GFX1250-FAKE16-NEXT: scratch_store_b32 off, v0, s32 offset:8 +; GFX1250-FAKE16-NEXT: scratch_load_u16 v0, off, s32 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_store_b32 off, v0, s32 offset:12 ; 4-byte Folded Spill +; GFX1250-FAKE16-NEXT: s_wait_xcnt 0x0 +; GFX1250-FAKE16-NEXT: ;;#ASMSTART +; GFX1250-FAKE16-NEXT: ;;#ASMEND +; GFX1250-FAKE16-NEXT: scratch_load_b32 v0, off, s32 offset:8 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_store_b16 off, v0, s32 offset:2 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_load_b32 v0, off, s32 offset:12 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_store_b16 off, v0, s32 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-FAKE16-NEXT: s_set_pc_i64 s[30:31] entry: %alloca = alloca <2 x i16>, i32 2, align 1, addrspace(5) @@ -341,6 +519,47 @@ define void @spill_2xi16_from_v2i16_one_free_reg() { ; GCN-FAKE16-NEXT: scratch_store_b16 off, v0, s32 dlc ; GCN-FAKE16-NEXT: s_waitcnt_vscnt null, 0x0 ; GCN-FAKE16-NEXT: s_setpc_b64 s[30:31] +; +; GFX1250-TRUE16-LABEL: spill_2xi16_from_v2i16_one_free_reg: +; GFX1250-TRUE16: ; %bb.0: ; %entry +; GFX1250-TRUE16-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-TRUE16-NEXT: s_wait_kmcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_load_u16 v7, off, s32 offset:2 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_load_u16 v0, off, s32 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_store_b32 off, v0, s32 offset:8 ; 4-byte Folded Spill +; GFX1250-TRUE16-NEXT: s_wait_xcnt 0x0 +; GFX1250-TRUE16-NEXT: ;;#ASMSTART +; GFX1250-TRUE16-NEXT: ;;#ASMEND +; GFX1250-TRUE16-NEXT: v_mov_b16_e32 v0.l, v7.l +; GFX1250-TRUE16-NEXT: scratch_store_b16 off, v0, s32 offset:2 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_load_b32 v0, off, s32 offset:8 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-TRUE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-TRUE16-NEXT: scratch_store_b16 off, v0, s32 scope:SCOPE_SYS +; GFX1250-TRUE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-TRUE16-NEXT: s_set_pc_i64 s[30:31] +; +; GFX1250-FAKE16-LABEL: spill_2xi16_from_v2i16_one_free_reg: +; GFX1250-FAKE16: ; %bb.0: ; %entry +; GFX1250-FAKE16-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-FAKE16-NEXT: s_wait_kmcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_load_u16 v7, off, s32 offset:2 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_load_u16 v0, off, s32 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_store_b32 off, v0, s32 offset:8 ; 4-byte Folded Spill +; GFX1250-FAKE16-NEXT: s_wait_xcnt 0x0 +; GFX1250-FAKE16-NEXT: ;;#ASMSTART +; GFX1250-FAKE16-NEXT: ;;#ASMEND +; GFX1250-FAKE16-NEXT: scratch_store_b16 off, v7, s32 offset:2 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_load_b32 v0, off, s32 offset:8 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-FAKE16-NEXT: s_wait_loadcnt 0x0 +; GFX1250-FAKE16-NEXT: scratch_store_b16 off, v0, s32 scope:SCOPE_SYS +; GFX1250-FAKE16-NEXT: s_wait_storecnt 0x0 +; GFX1250-FAKE16-NEXT: s_set_pc_i64 s[30:31] entry: %alloca = alloca <2 x i16>, i32 2, align 1, addrspace(5) @@ -375,6 +594,22 @@ define void @spill_v2i16() { ; GCN-NEXT: scratch_store_b32 off, v0, s32 offset:4 dlc ; GCN-NEXT: s_waitcnt_vscnt null, 0x0 ; GCN-NEXT: s_setpc_b64 s[30:31] +; +; GFX1250-LABEL: spill_v2i16: +; GFX1250: ; %bb.0: ; %entry +; GFX1250-NEXT: s_wait_loadcnt_dscnt 0x0 +; GFX1250-NEXT: s_wait_kmcnt 0x0 +; GFX1250-NEXT: scratch_load_b32 v0, off, s32 offset:4 scope:SCOPE_SYS +; GFX1250-NEXT: s_wait_loadcnt 0x0 +; GFX1250-NEXT: scratch_store_b32 off, v0, s32 offset:8 ; 4-byte Folded Spill +; GFX1250-NEXT: s_wait_xcnt 0x0 +; GFX1250-NEXT: ;;#ASMSTART +; GFX1250-NEXT: ;;#ASMEND +; GFX1250-NEXT: scratch_load_b32 v0, off, s32 offset:8 th:TH_LOAD_LU ; 4-byte Folded Reload +; GFX1250-NEXT: s_wait_loadcnt 0x0 +; GFX1250-NEXT: scratch_store_b32 off, v0, s32 offset:4 scope:SCOPE_SYS +; GFX1250-NEXT: s_wait_storecnt 0x0 +; GFX1250-NEXT: s_set_pc_i64 s[30:31] entry: %alloca = alloca <2 x i16>, i32 2, align 1, addrspace(5) diff --git a/llvm/test/CodeGen/AMDGPU/spillv16.mir b/llvm/test/CodeGen/AMDGPU/spillv16.mir index 05569bf..ba2d926 100644 --- a/llvm/test/CodeGen/AMDGPU/spillv16.mir +++ b/llvm/test/CodeGen/AMDGPU/spillv16.mir @@ -1,6 +1,7 @@ # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py # RUN: llc -march=amdgcn -verify-machineinstrs -mcpu=gfx1100 -mattr=+real-true16 -run-pass=regallocfast -o - %s | FileCheck -check-prefix=SPILLED %s # RUN: llc -march=amdgcn -verify-machineinstrs -mcpu=gfx1100 -mattr=+real-true16 -run-pass=regallocfast,prologepilog -o - %s | FileCheck -check-prefix=EXPANDED %s +# RUN: llc -march=amdgcn -verify-machineinstrs -mcpu=gfx1250 -mattr=+real-true16 -run-pass=regallocfast,prologepilog -o - %s | FileCheck -check-prefix=SRAMECC-EXPANDED %s --- name: spill_restore_vgpr16 @@ -46,6 +47,27 @@ body: | ; EXPANDED-NEXT: $vgpr0_lo16 = SCRATCH_LOAD_SHORT_D16_SADDR_t16 $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.0, addrspace 5) ; EXPANDED-NEXT: $vgpr0_hi16 = SCRATCH_LOAD_SHORT_D16_SADDR_t16 $sgpr32, 2, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.1, addrspace 5) ; EXPANDED-NEXT: S_NOP 0, implicit killed renamable $vgpr0_lo16, implicit killed renamable $vgpr0_hi16 + ; + ; SRAMECC-EXPANDED-LABEL: name: spill_restore_vgpr16 + ; SRAMECC-EXPANDED: bb.0: + ; SRAMECC-EXPANDED-NEXT: successors: %bb.1(0x80000000) + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: S_NOP 0, implicit-def renamable $vgpr0_lo16, implicit-def renamable $vgpr0_hi16 + ; SRAMECC-EXPANDED-NEXT: SCRATCH_STORE_SHORT_SADDR_t16 killed $vgpr0_hi16, $sgpr32, 2, 0, implicit $exec, implicit $flat_scr :: (store (s16) into %stack.1, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: SCRATCH_STORE_SHORT_SADDR_t16 killed $vgpr0_lo16, $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (store (s16) into %stack.0, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: S_CBRANCH_SCC1 %bb.1, implicit undef $scc + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: bb.1: + ; SRAMECC-EXPANDED-NEXT: successors: %bb.2(0x80000000) + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: S_NOP 1 + ; SRAMECC-EXPANDED-NEXT: {{ $}} + ; SRAMECC-EXPANDED-NEXT: bb.2: + ; SRAMECC-EXPANDED-NEXT: $vgpr1 = SCRATCH_LOAD_USHORT_SADDR $sgpr32, 0, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.0, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: $vgpr0_lo16 = V_MOV_B16_t16_e64 0, killed $vgpr1_lo16, 0, implicit $exec + ; SRAMECC-EXPANDED-NEXT: $vgpr1 = SCRATCH_LOAD_USHORT_SADDR $sgpr32, 2, 0, implicit $exec, implicit $flat_scr :: (load (s16) from %stack.1, addrspace 5) + ; SRAMECC-EXPANDED-NEXT: $vgpr0_hi16 = V_MOV_B16_t16_e64 0, killed $vgpr1_lo16, 0, implicit $exec + ; SRAMECC-EXPANDED-NEXT: S_NOP 0, implicit killed renamable $vgpr0_lo16, implicit killed renamable $vgpr0_hi16 bb.0: S_NOP 0, implicit-def %0:vgpr_16, implicit-def %1:vgpr_16 S_CBRANCH_SCC1 implicit undef $scc, %bb.1 diff --git a/llvm/test/CodeGen/AMDGPU/splitkit-getsubrangeformask.ll b/llvm/test/CodeGen/AMDGPU/splitkit-getsubrangeformask.ll index 5aafb0f..364598f 100644 --- a/llvm/test/CodeGen/AMDGPU/splitkit-getsubrangeformask.ll +++ b/llvm/test/CodeGen/AMDGPU/splitkit-getsubrangeformask.ll @@ -31,8 +31,8 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x ; CHECK-NEXT: [[COPY13:%[0-9]+]]:sgpr_32 = COPY $sgpr10 ; CHECK-NEXT: [[COPY14:%[0-9]+]]:sgpr_32 = COPY $sgpr8 ; CHECK-NEXT: undef [[S_LOAD_DWORDX2_IMM:%[0-9]+]].sub0_sub1:sgpr_128 = S_LOAD_DWORDX2_IMM [[COPY]], 232, 0 :: (invariant load (s64) from %ir.39, addrspace 4) - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM undef %125:sgpr_128, 0, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: KILL undef %125:sgpr_128 + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM undef %117:sgpr_128, 0, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: KILL undef %117:sgpr_128 ; CHECK-NEXT: [[S_LSHL_B32_:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY5]], 4, implicit-def dead $scc ; CHECK-NEXT: [[S_LSHL_B32_1:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY4]], 4, implicit-def dead $scc ; CHECK-NEXT: [[S_LSHL_B32_2:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY3]], 4, implicit-def dead $scc @@ -44,87 +44,85 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x ; CHECK-NEXT: [[S_SUB_I32_1:%[0-9]+]]:sreg_32 = S_SUB_I32 [[S_BUFFER_LOAD_DWORD_IMM]], 30, implicit-def dead $scc ; CHECK-NEXT: undef [[S_ADD_U32_:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], [[S_LSHL_B32_2]], implicit-def $scc ; CHECK-NEXT: [[S_ADD_U32_:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_]], 16, 0 :: (invariant load (s128) from %ir.81, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_]], 16, 0 :: (invariant load (s128) from %ir.71, addrspace 4) ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM1:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM undef %74:sreg_64, 0, 0 :: (invariant load (s128) from `ptr addrspace(4) poison`, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM2:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_]], 64, 0 :: (invariant load (s128) from %ir.88, addrspace 4) ; CHECK-NEXT: KILL undef %74:sreg_64 ; CHECK-NEXT: KILL [[S_ADD_U32_]].sub0, [[S_ADD_U32_]].sub1 ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM1:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM [[S_LOAD_DWORDX4_IMM]], 0, 0 :: (dereferenceable invariant load (s32)) ; CHECK-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec ; CHECK-NEXT: undef [[S_MOV_B32_:%[0-9]+]].sub1:sgpr_128 = S_MOV_B32 0 - ; CHECK-NEXT: [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET undef %118:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], undef %89:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) + ; CHECK-NEXT: [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET undef %112:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) + ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], undef %87:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN1:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: KILL undef %89:sgpr_128 - ; CHECK-NEXT: KILL undef %118:sgpr_128 + ; CHECK-NEXT: KILL undef %112:sgpr_128 + ; CHECK-NEXT: KILL undef %87:sgpr_128 ; CHECK-NEXT: [[S_SUB_I32_2:%[0-9]+]]:sreg_32 = S_SUB_I32 [[S_BUFFER_LOAD_DWORD_IMM1]], 31, implicit-def dead $scc ; CHECK-NEXT: undef [[S_ADD_U32_1:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], [[S_LSHL_B32_]], implicit-def $scc ; CHECK-NEXT: [[S_ADD_U32_1:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc ; CHECK-NEXT: undef [[S_ADD_U32_2:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], [[S_LSHL_B32_1]], implicit-def $scc ; CHECK-NEXT: [[S_ADD_U32_2:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_3:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], [[S_LSHL_B32_2]], implicit-def $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM2:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_1]], 64, 0 :: (invariant load (s128) from %ir.87, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM3:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_2]], 64, 0 :: (invariant load (s128) from %ir.93, addrspace 4) - ; CHECK-NEXT: KILL [[S_ADD_U32_1]].sub0, [[S_ADD_U32_1]].sub1 + ; CHECK-NEXT: [[S_ASHR_I32_3:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 undef %148:sreg_32, 31, implicit-def dead $scc + ; CHECK-NEXT: undef [[S_ADD_U32_3:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], undef %148:sreg_32, implicit-def $scc + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM3:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_1]], 64, 0 :: (invariant load (s128) from %ir.77, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM4:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_2]], 64, 0 :: (invariant load (s128) from %ir.83, addrspace 4) ; CHECK-NEXT: KILL [[S_ADD_U32_2]].sub0, [[S_ADD_U32_2]].sub1 - ; CHECK-NEXT: [[S_ADD_U32_3:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_ASHR_I32_3:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 undef %169:sreg_32, 31, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_4:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], undef %169:sreg_32, implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_4:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_3]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_5:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_5:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_6:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_1]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_6:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_7:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, undef %169:sreg_32, implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_7:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_3]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_8:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_2]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_8:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_9:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY8]], [[S_LSHL_B32_]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_9:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %48:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_10:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY9]], [[S_LSHL_B32_1]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_10:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %45:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_11:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY9]], [[S_LSHL_B32_2]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_11:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %45:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: KILL [[S_ADD_U32_1]].sub0, [[S_ADD_U32_1]].sub1 + ; CHECK-NEXT: [[S_ADD_U32_3:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_3]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_4:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_4:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_5:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_1]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_5:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_6:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, undef %148:sreg_32, implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_6:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_3]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_7:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_2]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_7:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_8:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY8]], [[S_LSHL_B32_]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_8:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %48:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_9:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY9]], [[S_LSHL_B32_1]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_9:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %45:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_10:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY9]], [[S_LSHL_B32_2]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_10:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %45:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc ; CHECK-NEXT: [[S_ADD_I32_:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_LSHL_B32_]], 16, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_1:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_LSHL_B32_2]], 16, implicit-def dead $scc ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM [[S_MOV_B32_]], [[S_ADD_I32_]], 0, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM1:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM [[S_MOV_B32_]], undef %302:sreg_32, 0, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM1:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM [[S_MOV_B32_]], undef %279:sreg_32, 0, 0 :: (dereferenceable invariant load (s32)) ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM2:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM [[S_MOV_B32_]], [[S_ADD_I32_1]], 0, 0 :: (dereferenceable invariant load (s32)) ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM2:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM [[S_MOV_B32_]], 16, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM3:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %357:sgpr_128, undef %358:sreg_32, 0, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM3:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM undef %368:sgpr_128, 16, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM4:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_3]], 64, 0 :: (invariant load (s128) from %ir.99, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM5:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_4]], 64, 0 :: (invariant load (s128) from %ir.107, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM6:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_5]], 0, 0 :: (invariant load (s128) from %ir.112, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM7:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_6]], 0, 0 :: (invariant load (s128) from %ir.117, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM8:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_7]], 0, 0 :: (invariant load (s128) from %ir.124, addrspace 4) - ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN2:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM2]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM4:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %352:sgpr_128, [[S_ADD_I32_]], 0, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM5:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %363:sgpr_128, [[S_ADD_I32_1]], 0, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN3:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM3]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM3:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %334:sgpr_128, undef %335:sreg_32, 0, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM3:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM undef %345:sgpr_128, 16, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM5:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_3]], 64, 0 :: (invariant load (s128) from %ir.95, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM6:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_4]], 0, 0 :: (invariant load (s128) from %ir.100, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM7:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_5]], 0, 0 :: (invariant load (s128) from %ir.105, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM8:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_6]], 0, 0 :: (invariant load (s128) from %ir.112, addrspace 4) + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM4:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %329:sgpr_128, [[S_ADD_I32_]], 0, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM5:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %340:sgpr_128, [[S_ADD_I32_1]], 0, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN2:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM3]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) + ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN3:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ADD_I32_2:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM]], -98, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_3:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM1]], -114, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_4:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM2]], -130, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_5:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM2]], -178, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_12:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY10]], [[S_LSHL_B32_]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_12:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %42:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_13:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_13:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_14:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_1]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_14:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_15:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_2]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_15:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_11:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY10]], [[S_LSHL_B32_]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_11:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %42:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_12:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_12:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_13:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_1]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_13:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_14:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_2]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_14:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc ; CHECK-NEXT: [[S_LSHL_B32_3:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY12]], 4, implicit-def dead $scc - ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN4:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) + ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN4:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM2]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ADD_I32_6:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_LSHL_B32_3]], 16, implicit-def dead $scc - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM6:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %384:sgpr_128, [[S_ADD_I32_6]], 0, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM6:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %361:sgpr_128, [[S_ADD_I32_6]], 0, 0 :: (dereferenceable invariant load (s32)) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN5:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM5]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM9:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_5]], 224, 0 :: (invariant load (s128) from %ir.129, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM10:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[COPY7]], 224, 0 :: (invariant load (s128) from %ir.145, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM11:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_5]], 576, 0 :: (invariant load (s128) from %ir.150, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM9:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_4]], 224, 0 :: (invariant load (s128) from %ir.117, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM10:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[COPY7]], 224, 0 :: (invariant load (s128) from %ir.133, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM11:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_4]], 576, 0 :: (invariant load (s128) from %ir.138, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN6:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM6]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM12:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_6]], 224, 0 :: (invariant load (s128) from %ir.134, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM13:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_7]], 576, 0 :: (invariant load (s128) from %ir.162, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM14:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_8]], 224, 0 :: (invariant load (s128) from %ir.140, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM12:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_5]], 224, 0 :: (invariant load (s128) from %ir.122, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM13:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_6]], 576, 0 :: (invariant load (s128) from %ir.150, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM14:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_7]], 224, 0 :: (invariant load (s128) from %ir.128, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN7:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM7]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN8:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM8]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ADD_I32_7:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM4]], -217, implicit-def dead $scc @@ -135,49 +133,49 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x ; CHECK-NEXT: [[S_ADD_I32_12:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM3]], -329, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_13:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM3]], -345, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_14:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM6]], -441, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_16:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY2]], [[S_LSHL_B32_2]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_16:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %36:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_15:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY2]], [[S_LSHL_B32_2]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_15:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %36:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc ; CHECK-NEXT: [[S_LSHL_B32_4:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY13]], 4, implicit-def dead $scc ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN9:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM9]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ASHR_I32_4:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_4]], 31, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_17:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY2]], [[S_LSHL_B32_4]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_17:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %36:sreg_32, [[S_ASHR_I32_4]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_16:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY2]], [[S_LSHL_B32_4]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_16:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %36:sreg_32, [[S_ASHR_I32_4]], implicit-def dead $scc, implicit $scc ; CHECK-NEXT: [[S_LSHL_B32_5:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY5]], 3, implicit-def dead $scc ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN10:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM12]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ASHR_I32_5:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_5]], 31, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_18:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_5]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_18:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_5]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORD_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM [[S_ADD_U32_18]], 168, 0 :: (invariant load (s32) from %ir.273, align 8, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM15:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_8]], 576, 0 :: (invariant load (s128) from %ir.157, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_17:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_5]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_17:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_5]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORD_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM [[S_ADD_U32_17]], 168, 0 :: (invariant load (s32) from %ir.260, align 8, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM15:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_7]], 576, 0 :: (invariant load (s128) from %ir.145, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN11:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM14]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN12:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM10]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN13:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM11]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_LOAD_DWORDX2_IMM:%[0-9]+]].sub3:sgpr_128 = S_MOV_B32 553734060 ; CHECK-NEXT: [[S_LOAD_DWORDX2_IMM:%[0-9]+]].sub2:sgpr_128 = S_MOV_B32 -1 ; CHECK-NEXT: [[COPY15:%[0-9]+]]:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM]] - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM16:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_9]], 0, 0 :: (invariant load (s128) from %ir.170, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM16:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_8]], 0, 0 :: (invariant load (s128) from %ir.158, addrspace 4) ; CHECK-NEXT: [[COPY15:%[0-9]+]].sub1:sgpr_128 = COPY [[S_MOV_B32_]].sub1 ; CHECK-NEXT: [[COPY15:%[0-9]+]].sub0:sgpr_128 = COPY [[S_LOAD_DWORD_IMM]] ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM4:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM [[COPY15]], 0, 0 :: (dereferenceable invariant load (s32)) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN14:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM15]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN15:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM13]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM17:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_10]], 0, 0 :: (invariant load (s128) from %ir.178, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM18:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_11]], 0, 0 :: (invariant load (s128) from %ir.183, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM17:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_9]], 0, 0 :: (invariant load (s128) from %ir.166, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM18:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_10]], 0, 0 :: (invariant load (s128) from %ir.171, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN16:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM16]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_LSHL_B32_6:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY4]], 3, implicit-def dead $scc ; CHECK-NEXT: [[BUFFER_LOAD_DWORD_OFFSET1:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[S_LOAD_DWORDX4_IMM1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ASHR_I32_6:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_6]], 31, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_15:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM4]], -467, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_19:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_6]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_19:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_6]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX2_IMM1:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[S_ADD_U32_19]], 168, 0 :: (invariant load (s64) from %ir.282, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_18:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_6]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_18:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_6]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORDX2_IMM1:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[S_ADD_U32_18]], 168, 0 :: (invariant load (s64) from %ir.269, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_DWORD_OFFSET2:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[S_LOAD_DWORDX4_IMM17]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_DWORD_OFFSET3:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[S_LOAD_DWORDX4_IMM18]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM19:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_12]], 0, 0 :: (invariant load (s128) from %ir.205, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM20:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_13]], 0, 0 :: (invariant load (s128) from %ir.211, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM19:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_11]], 0, 0 :: (invariant load (s128) from %ir.193, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM20:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_12]], 0, 0 :: (invariant load (s128) from %ir.199, addrspace 4) ; CHECK-NEXT: [[COPY16:%[0-9]+]]:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM]] - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM21:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_14]], 0, 0 :: (invariant load (s128) from %ir.216, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM22:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_15]], 0, 0 :: (invariant load (s128) from %ir.221, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM21:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_13]], 0, 0 :: (invariant load (s128) from %ir.204, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM22:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_14]], 0, 0 :: (invariant load (s128) from %ir.209, addrspace 4) ; CHECK-NEXT: [[S_AND_B32_:%[0-9]+]]:sreg_32 = S_AND_B32 [[S_LOAD_DWORDX2_IMM1]].sub1, 65535, implicit-def dead $scc ; CHECK-NEXT: [[COPY16:%[0-9]+]].sub0:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM1]].sub0 ; CHECK-NEXT: [[COPY16:%[0-9]+]].sub1:sgpr_128 = COPY [[S_AND_B32_]] @@ -189,30 +187,30 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN20:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM22]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ASHR_I32_7:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_7]], 31, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_16:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM5]], -468, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_20:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_7]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_20:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_7]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX2_IMM2:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[S_ADD_U32_20]], 168, 0 :: (invariant load (s64) from %ir.293, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_19:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_7]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_19:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_7]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORDX2_IMM2:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[S_ADD_U32_19]], 168, 0 :: (invariant load (s64) from %ir.280, addrspace 4) ; CHECK-NEXT: [[COPY17:%[0-9]+]]:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM]] ; CHECK-NEXT: [[S_AND_B32_1:%[0-9]+]]:sreg_32 = S_AND_B32 [[S_LOAD_DWORDX2_IMM2]].sub1, 65535, implicit-def dead $scc ; CHECK-NEXT: [[COPY17:%[0-9]+]].sub0:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM2]].sub0 ; CHECK-NEXT: [[COPY17:%[0-9]+]].sub1:sgpr_128 = COPY [[S_AND_B32_1]] ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM6:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM [[COPY17]], 0, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM23:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_16]], 160, 0 :: (invariant load (s128) from %ir.256, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORD_IMM1:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM undef %470:sreg_64, 0, 0 :: (invariant load (s32) from `ptr addrspace(4) poison`, addrspace 4) - ; CHECK-NEXT: KILL [[S_ADD_U32_16]].sub0, [[S_ADD_U32_16]].sub1 - ; CHECK-NEXT: KILL undef %470:sreg_64 + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM23:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_15]], 160, 0 :: (invariant load (s128) from %ir.244, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORD_IMM1:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM undef %443:sreg_64, 0, 0 :: (invariant load (s32) from `ptr addrspace(4) poison`, addrspace 4) + ; CHECK-NEXT: KILL [[S_ADD_U32_15]].sub0, [[S_ADD_U32_15]].sub1 ; CHECK-NEXT: KILL [[COPY17]].sub0_sub1_sub2, [[COPY17]].sub3 + ; CHECK-NEXT: KILL undef %443:sreg_64 ; CHECK-NEXT: [[S_LSHL_B32_8:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY14]], 3, implicit-def dead $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM24:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_17]], 160, 0 :: (invariant load (s128) from %ir.265, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM24:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_16]], 160, 0 :: (invariant load (s128) from %ir.252, addrspace 4) ; CHECK-NEXT: [[S_ASHR_I32_8:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_8]], 31, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_17:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM6]], -469, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_21:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_8]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_21:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_8]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORD_IMM2:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM [[S_ADD_U32_21]], 168, 0 :: (invariant load (s32) from %ir.305, align 8, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_20:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_8]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_20:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_8]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORD_IMM2:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM [[S_ADD_U32_20]], 168, 0 :: (invariant load (s32) from %ir.291, align 8, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN21:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM23]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN22:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM24]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM24]] ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM23]] + ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM24]] ; CHECK-NEXT: [[S_AND_B32_2:%[0-9]+]]:sreg_32 = S_AND_B32 [[S_LOAD_DWORD_IMM1]], 65535, implicit-def dead $scc ; CHECK-NEXT: [[COPY18:%[0-9]+]]:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM]] ; CHECK-NEXT: [[COPY18:%[0-9]+]].sub1:sgpr_128 = COPY [[S_AND_B32_2]] @@ -224,22 +222,22 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x ; CHECK-NEXT: [[S_ADD_I32_21:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM3]], -507, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_22:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM3]], -539, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_23:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM7]], -473, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_22:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_22:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM25:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_22]], 96, 0 :: (invariant load (s128) from %ir.323, addrspace 4) - ; CHECK-NEXT: undef [[S_ADD_U32_23:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_1]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_23:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM26:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_23]], 96, 0 :: (invariant load (s128) from %ir.329, addrspace 4) - ; CHECK-NEXT: undef [[S_ADD_U32_24:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_2]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_24:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM27:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_24]], 96, 0 :: (invariant load (s128) from %ir.335, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_21:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_21:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM25:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_21]], 96, 0 :: (invariant load (s128) from %ir.309, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_22:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_1]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_22:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM26:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_22]], 96, 0 :: (invariant load (s128) from %ir.315, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_23:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_2]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_23:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM27:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_23]], 96, 0 :: (invariant load (s128) from %ir.321, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN23:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM25]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN24:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM26]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN25:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM27]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM27]] ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM25]] - ; CHECK-NEXT: KILL [[V_MOV_B32_e32_]] ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM26]] + ; CHECK-NEXT: KILL [[V_MOV_B32_e32_]] + ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM27]] ; CHECK-NEXT: [[V_ADD_U32_e64_:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 -2, [[BUFFER_LOAD_FORMAT_X_IDXEN]], 0, implicit $exec ; CHECK-NEXT: [[V_ADD_U32_e64_1:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 -1, [[BUFFER_LOAD_FORMAT_X_IDXEN1]], 0, implicit $exec ; CHECK-NEXT: [[V_ADD_U32_e64_2:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 -3, [[BUFFER_LOAD_FORMAT_X_IDXEN]], 0, implicit $exec @@ -351,13 +349,13 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x ; CHECK-NEXT: [[V_OR_B32_e64_64:%[0-9]+]]:vgpr_32 = V_OR_B32_e64 [[V_OR_B32_e64_63]], [[V_ADD_U32_e64_28]], implicit $exec ; CHECK-NEXT: [[V_ADD_U32_e64_30:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 -593, [[BUFFER_LOAD_FORMAT_X_IDXEN]], 0, implicit $exec ; CHECK-NEXT: [[V_OR_B32_e64_65:%[0-9]+]]:vgpr_32 = V_OR_B32_e64 [[V_OR_B32_e64_64]], [[V_ADD_U32_e64_29]], implicit $exec - ; CHECK-NEXT: [[S_LOAD_DWORDX8_IMM:%[0-9]+]]:sgpr_256 = S_LOAD_DWORDX8_IMM undef %543:sreg_64, 0, 0 :: (invariant load (s256) from `ptr addrspace(4) poison`, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX8_IMM:%[0-9]+]]:sgpr_256 = S_LOAD_DWORDX8_IMM undef %516:sreg_64, 0, 0 :: (invariant load (s256) from `ptr addrspace(4) poison`, addrspace 4) ; CHECK-NEXT: [[V_OR_B32_e64_66:%[0-9]+]]:vgpr_32 = V_OR_B32_e64 [[V_OR_B32_e64_65]], [[V_ADD_U32_e64_30]], implicit $exec ; CHECK-NEXT: [[S_ADD_I32_24:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM8]], -594, implicit-def dead $scc ; CHECK-NEXT: [[V_OR_B32_e64_67:%[0-9]+]]:vgpr_32 = V_OR_B32_e64 [[S_ADD_I32_24]], [[V_OR_B32_e64_66]], implicit $exec ; CHECK-NEXT: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U32_e64 0, [[V_OR_B32_e64_67]], implicit $exec ; CHECK-NEXT: undef [[V_CNDMASK_B32_e64_:%[0-9]+]].sub3:vreg_128 = V_CNDMASK_B32_e64 0, 0, 0, 1, [[V_CMP_EQ_U32_e64_]], implicit $exec - ; CHECK-NEXT: IMAGE_STORE_V4_V2_nsa_gfx10 [[V_CNDMASK_B32_e64_]], undef %557:vgpr_32, undef %559:vgpr_32, [[S_LOAD_DWORDX8_IMM]], 15, 1, -1, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), addrspace 8) + ; CHECK-NEXT: IMAGE_STORE_V4_V2_nsa_gfx10 [[V_CNDMASK_B32_e64_]], undef %530:vgpr_32, undef %532:vgpr_32, [[S_LOAD_DWORDX8_IMM]], 15, 1, -1, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), addrspace 8) ; CHECK-NEXT: S_ENDPGM 0 .expVert: %0 = extractelement <31 x i32> %userData, i64 2 diff --git a/llvm/test/CodeGen/AMDGPU/umin-sub-to-usubo-select-combine.ll b/llvm/test/CodeGen/AMDGPU/umin-sub-to-usubo-select-combine.ll new file mode 100644 index 0000000..22e4a24 --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/umin-sub-to-usubo-select-combine.ll @@ -0,0 +1,236 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx900 < %s | FileCheck -check-prefix=GFX9 %s +; RUN: llc -mtriple=amdgcn-amd-amdpal -mcpu=gfx1100 < %s | FileCheck -check-prefix=GFX11 %s + +define i16 @v_underflow_compare_fold_i16(i16 %a, i16 %b) #0 { +; GFX9-LABEL: v_underflow_compare_fold_i16: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_u16_e32 v1, v0, v1 +; GFX9-NEXT: v_min_u16_e32 v0, v1, v0 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_underflow_compare_fold_i16: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_nc_u16 v0.h, v0.l, v1.l +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_min_u16 v0.l, v0.h, v0.l +; GFX11-NEXT: s_setpc_b64 s[30:31] + %sub = sub i16 %a, %b + %cond = call i16 @llvm.umin.i16(i16 %sub, i16 %a) + ret i16 %cond +} + +define i32 @v_underflow_compare_fold_i32(i32 %a, i32 %b) #0 { +; GFX9-LABEL: v_underflow_compare_fold_i32: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_u32_e32 v1, v0, v1 +; GFX9-NEXT: v_min_u32_e32 v0, v1, v0 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_underflow_compare_fold_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_nc_u32_e32 v1, v0, v1 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_min_u32_e32 v0, v1, v0 +; GFX11-NEXT: s_setpc_b64 s[30:31] + %sub = sub i32 %a, %b + %cond = call i32 @llvm.umin.i32(i32 %sub, i32 %a) + ret i32 %cond +} + +define i32 @v_underflow_compare_fold_i32_commute(i32 %a, i32 %b) #0 { +; GFX9-LABEL: v_underflow_compare_fold_i32_commute: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_u32_e32 v1, v0, v1 +; GFX9-NEXT: v_min_u32_e32 v0, v0, v1 +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_underflow_compare_fold_i32_commute: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_nc_u32_e32 v1, v0, v1 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_min_u32_e32 v0, v0, v1 +; GFX11-NEXT: s_setpc_b64 s[30:31] + %sub = sub i32 %a, %b + %cond = call i32 @llvm.umin.i32(i32 %a, i32 %sub) + ret i32 %cond +} + +define i32 @v_underflow_compare_fold_i32_multi_use(i32 %a, i32 %b, ptr addrspace(1) %ptr) #0 { +; GFX9-LABEL: v_underflow_compare_fold_i32_multi_use: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_u32_e32 v1, v0, v1 +; GFX9-NEXT: v_min_u32_e32 v0, v1, v0 +; GFX9-NEXT: global_store_dword v[2:3], v1, off +; GFX9-NEXT: s_waitcnt vmcnt(0) +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_underflow_compare_fold_i32_multi_use: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_nc_u32_e32 v1, v0, v1 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_min_u32_e32 v0, v1, v0 +; GFX11-NEXT: global_store_b32 v[2:3], v1, off +; GFX11-NEXT: s_setpc_b64 s[30:31] + %sub = sub i32 %a, %b + store i32 %sub, ptr addrspace(1) %ptr + %cond = call i32 @llvm.umin.i32(i32 %sub, i32 %a) + ret i32 %cond +} + +define i64 @v_underflow_compare_fold_i64(i64 %a, i64 %b) #0 { +; GFX9-LABEL: v_underflow_compare_fold_i64: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_co_u32_e32 v2, vcc, v0, v2 +; GFX9-NEXT: v_subb_co_u32_e32 v3, vcc, v1, v3, vcc +; GFX9-NEXT: v_cmp_lt_u64_e32 vcc, v[2:3], v[0:1] +; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc +; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_underflow_compare_fold_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_co_u32 v2, vcc_lo, v0, v2 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_sub_co_ci_u32_e64 v3, null, v1, v3, vcc_lo +; GFX11-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[2:3], v[0:1] +; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v2 :: v_dual_cndmask_b32 v1, v1, v3 +; GFX11-NEXT: s_setpc_b64 s[30:31] + %sub = sub i64 %a, %b + %cond = call i64 @llvm.umin.i64(i64 %sub, i64 %a) + ret i64 %cond +} + +define i64 @v_underflow_compare_fold_i64_commute(i64 %a, i64 %b) #0 { +; GFX9-LABEL: v_underflow_compare_fold_i64_commute: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_co_u32_e32 v2, vcc, v0, v2 +; GFX9-NEXT: v_subb_co_u32_e32 v3, vcc, v1, v3, vcc +; GFX9-NEXT: v_cmp_lt_u64_e32 vcc, v[0:1], v[2:3] +; GFX9-NEXT: v_cndmask_b32_e32 v0, v2, v0, vcc +; GFX9-NEXT: v_cndmask_b32_e32 v1, v3, v1, vcc +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_underflow_compare_fold_i64_commute: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_co_u32 v2, vcc_lo, v0, v2 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_sub_co_ci_u32_e64 v3, null, v1, v3, vcc_lo +; GFX11-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[0:1], v[2:3] +; GFX11-NEXT: v_dual_cndmask_b32 v0, v2, v0 :: v_dual_cndmask_b32 v1, v3, v1 +; GFX11-NEXT: s_setpc_b64 s[30:31] + %sub = sub i64 %a, %b + %cond = call i64 @llvm.umin.i64(i64 %a, i64 %sub) + ret i64 %cond +} + +define i64 @v_underflow_compare_fold_i64_multi_use(i64 %a, i64 %b, ptr addrspace(1) %ptr) #0 { +; GFX9-LABEL: v_underflow_compare_fold_i64_multi_use: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: v_sub_co_u32_e32 v2, vcc, v0, v2 +; GFX9-NEXT: v_subb_co_u32_e32 v3, vcc, v1, v3, vcc +; GFX9-NEXT: v_cmp_lt_u64_e32 vcc, v[2:3], v[0:1] +; GFX9-NEXT: global_store_dwordx2 v[4:5], v[2:3], off +; GFX9-NEXT: v_cndmask_b32_e32 v0, v0, v2, vcc +; GFX9-NEXT: v_cndmask_b32_e32 v1, v1, v3, vcc +; GFX9-NEXT: s_waitcnt vmcnt(0) +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: v_underflow_compare_fold_i64_multi_use: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: v_sub_co_u32 v2, vcc_lo, v0, v2 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_sub_co_ci_u32_e64 v3, null, v1, v3, vcc_lo +; GFX11-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[2:3], v[0:1] +; GFX11-NEXT: global_store_b64 v[4:5], v[2:3], off +; GFX11-NEXT: v_dual_cndmask_b32 v0, v0, v2 :: v_dual_cndmask_b32 v1, v1, v3 +; GFX11-NEXT: s_setpc_b64 s[30:31] + %sub = sub i64 %a, %b + store i64 %sub, ptr addrspace(1) %ptr + %cond = call i64 @llvm.umin.i64(i64 %sub, i64 %a) + ret i64 %cond +} + +define amdgpu_ps i16 @s_underflow_compare_fold_i16(i16 inreg %a, i16 inreg %b) #0 { +; GFX9-LABEL: s_underflow_compare_fold_i16: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_sub_i32 s1, s0, s1 +; GFX9-NEXT: s_and_b32 s0, 0xffff, s0 +; GFX9-NEXT: s_and_b32 s1, s1, 0xffff +; GFX9-NEXT: s_min_u32 s0, s1, s0 +; GFX9-NEXT: ; return to shader part epilog +; +; GFX11-LABEL: s_underflow_compare_fold_i16: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_sub_i32 s1, s0, s1 +; GFX11-NEXT: s_and_b32 s0, 0xffff, s0 +; GFX11-NEXT: s_and_b32 s1, s1, 0xffff +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) +; GFX11-NEXT: s_min_u32 s0, s1, s0 +; GFX11-NEXT: ; return to shader part epilog + %sub = sub i16 %a, %b + %cond = call i16 @llvm.umin.i16(i16 %sub, i16 %a) + ret i16 %cond +} + +define amdgpu_ps i32 @s_underflow_compare_fold_i32(i32 inreg %a, i32 inreg %b) #0 { +; GFX9-LABEL: s_underflow_compare_fold_i32: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_sub_i32 s1, s0, s1 +; GFX9-NEXT: s_min_u32 s0, s1, s0 +; GFX9-NEXT: ; return to shader part epilog +; +; GFX11-LABEL: s_underflow_compare_fold_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_sub_i32 s1, s0, s1 +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) +; GFX11-NEXT: s_min_u32 s0, s1, s0 +; GFX11-NEXT: ; return to shader part epilog + %sub = sub i32 %a, %b + %cond = call i32 @llvm.umin.i32(i32 %sub, i32 %a) + ret i32 %cond +} + +define amdgpu_ps i64 @s_underflow_compare_fold_i64(i64 inreg %a, i64 inreg %b) #0 { +; GFX9-LABEL: s_underflow_compare_fold_i64: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_sub_u32 s2, s0, s2 +; GFX9-NEXT: v_mov_b32_e32 v0, s0 +; GFX9-NEXT: s_subb_u32 s3, s1, s3 +; GFX9-NEXT: v_mov_b32_e32 v1, s1 +; GFX9-NEXT: v_cmp_lt_u64_e32 vcc, s[2:3], v[0:1] +; GFX9-NEXT: s_and_b64 s[4:5], vcc, exec +; GFX9-NEXT: s_cselect_b32 s1, s3, s1 +; GFX9-NEXT: s_cselect_b32 s0, s2, s0 +; GFX9-NEXT: ; return to shader part epilog +; +; GFX11-LABEL: s_underflow_compare_fold_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_sub_u32 s2, s0, s2 +; GFX11-NEXT: s_subb_u32 s3, s1, s3 +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) +; GFX11-NEXT: v_cmp_lt_u64_e64 s4, s[2:3], s[0:1] +; GFX11-NEXT: s_and_b32 s4, s4, exec_lo +; GFX11-NEXT: s_cselect_b32 s0, s2, s0 +; GFX11-NEXT: s_cselect_b32 s1, s3, s1 +; GFX11-NEXT: ; return to shader part epilog + %sub = sub i64 %a, %b + %cond = call i64 @llvm.umin.i64(i64 %sub, i64 %a) + ret i64 %cond +} + +attributes #0 = { nounwind } diff --git a/llvm/test/CodeGen/AMDGPU/vgpr-lowering-gfx1250-t16.mir b/llvm/test/CodeGen/AMDGPU/vgpr-lowering-gfx1250-t16.mir index 8a70a8a..32cc398 100644 --- a/llvm/test/CodeGen/AMDGPU/vgpr-lowering-gfx1250-t16.mir +++ b/llvm/test/CodeGen/AMDGPU/vgpr-lowering-gfx1250-t16.mir @@ -36,7 +36,7 @@ body: | ; GCN-NEXT: v_add_f16_e64 v128.l /*v384.l*/, v129.l /*v385.l*/, v130.l /*v386.l*/ $vgpr384_lo16 = V_ADD_F16_t16_e64 0, undef $vgpr385_lo16, 0, undef $vgpr386_lo16, 0, 0, 0, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0x8a + ; GCN-NEXT: s_set_vgpr_msb 0x458a ; ASM-SAME: ; msbs: dst=2 src0=2 src1=2 src2=0 ; GCN-NEXT: v_add_f16_e64 v0.h /*v512.h*/, v1.h /*v513.h*/, v2.h /*v514.h*/ $vgpr512_hi16 = V_ADD_F16_t16_e64 0, undef $vgpr513_hi16, 0, undef $vgpr514_hi16, 0, 0, 0, implicit $exec, implicit $mode @@ -50,7 +50,7 @@ body: | ; GCN-NEXT: v_add_f16_e64 v128.l /*v640.l*/, v129.l /*v641.l*/, v130.l /*v642.l*/ $vgpr640_lo16 = V_ADD_F16_t16_e64 0, undef $vgpr641_lo16, 0, undef $vgpr642_lo16, 0, 0, 0, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0xcf + ; GCN-NEXT: s_set_vgpr_msb 0x8acf ; ASM-SAME: ; msbs: dst=3 src0=3 src1=3 src2=0 ; GCN-NEXT: v_add_f16_e64 v0.h /*v768.h*/, v1.h /*v769.h*/, v2.h /*v770.h*/ $vgpr768_hi16 = V_ADD_F16_t16_e64 0, undef $vgpr769_hi16, 0, undef $vgpr770_hi16, 0, 0, 0, implicit $exec, implicit $mode diff --git a/llvm/test/CodeGen/AMDGPU/vgpr-lowering-gfx1250.mir b/llvm/test/CodeGen/AMDGPU/vgpr-lowering-gfx1250.mir index f508df2..7e1c28f 100644 --- a/llvm/test/CodeGen/AMDGPU/vgpr-lowering-gfx1250.mir +++ b/llvm/test/CodeGen/AMDGPU/vgpr-lowering-gfx1250.mir @@ -22,13 +22,13 @@ body: | $vgpr257 = V_MOV_B32_e32 undef $vgpr510, implicit $exec ; Single bit change - ; GCN-NEXT: s_set_vgpr_msb 1 + ; GCN-NEXT: s_set_vgpr_msb 0x4101 ; ASM-SAME: ; msbs: dst=0 src0=1 src1=0 src2=0 ; GCN-NEXT: v_rcp_f32_e64 v255, v2 /*v258*/ $vgpr255 = V_RCP_F32_e64 0, undef $vgpr258, 0, 0, implicit $exec, implicit $mode ; Reset - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x100 ; ASM-SAME: ; msbs: dst=0 src0=0 src1=0 src2=0 ; GCN-NEXT: v_rcp_f32_e64 v255, v1 $vgpr255 = V_RCP_F32_e64 0, undef $vgpr1, 0, 0, implicit $exec, implicit $mode @@ -40,7 +40,7 @@ body: | ; GCN-NEXT: v_add_nc_u32_e32 v0, v253 /*v509*/, v252 /*v508*/ $vgpr0 = V_ADD_U32_e32 undef $vgpr509, undef $vgpr508, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0x44 + ; GCN-NEXT: s_set_vgpr_msb 0x544 ; ASM-SAME: ; msbs: dst=1 src0=0 src1=1 src2=0 ; GCN-NEXT: s_delay_alu instid0(VALU_DEP_1) ; GCN-NEXT: v_add_f32_e64 v2 /*v258*/, v0, v251 /*v507*/ @@ -48,7 +48,7 @@ body: | ; VOP3 - ; GCN-NEXT: s_set_vgpr_msb 0x55 + ; GCN-NEXT: s_set_vgpr_msb 0x4455 ; ASM-SAME: ; msbs: dst=1 src0=1 src1=1 src2=1 ; GCN-NEXT: v_fma_f32 v3 /*v259*/, v4 /*v260*/, v5 /*v261*/, v6 /*v262*/ $vgpr259 = V_FMA_F32_e64 0, undef $vgpr260, 0, undef $vgpr261, 0, undef $vgpr262, 0, 0, implicit $exec, implicit $mode @@ -58,32 +58,32 @@ body: | $vgpr259 = V_FMA_F32_e64 0, undef $vgpr260, 0, undef $vgpr261, 0, undef $vgpr262, 0, 0, implicit $exec, implicit $mode ; Tuple crossing the 256 boundary - ; GCN-NEXT: s_set_vgpr_msb 17 + ; GCN-NEXT: s_set_vgpr_msb 0x5511 ; ASM-SAME: ; msbs: dst=0 src0=1 src1=0 src2=1 ; GCN-NEXT: v_mqsad_u32_u8 v[254:257], v[2:3] /*v[258:259]*/, v0, v[244:247] /*v[500:503]*/ $vgpr254_vgpr255_vgpr256_vgpr257 = V_MQSAD_U32_U8_e64 $vgpr258_vgpr259, $vgpr0, undef $vgpr500_vgpr501_vgpr502_vgpr503, 0, implicit $exec ; DPP/tied operand - ; GCN-NEXT: s_set_vgpr_msb 0x45 + ; GCN-NEXT: s_set_vgpr_msb 0x1145 ; ASM-SAME: ; msbs: dst=1 src0=1 src1=1 src2=0 ; GCN-NEXT: s_delay_alu instid0(VALU_DEP_1) ; GCN-NEXT: v_add_nc_u16_e64_dpp v0 /*v256*/, v1 /*v257*/, v2 /*v258*/ quad_perm:[1,0,0,0] row_mask:0xf bank_mask:0xf bound_ctrl:1 $vgpr256 = V_ADD_NC_U16_fake16_e64_dpp $vgpr256, 0, $vgpr257, 0, undef $vgpr258, 0, 0, 1, 15, 15, 1, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 17 + ; GCN-NEXT: s_set_vgpr_msb 0x4511 ; ASM-SAME: ; msbs: dst=0 src0=1 src1=0 src2=1 ; GCN-NEXT: v_add3_u32_e64_dpp v0, v1 /*v257*/, v0, v2 /*v258*/ quad_perm:[1,0,0,0] row_mask:0xf bank_mask:0xf bound_ctrl:1 $vgpr0 = V_ADD3_U32_e64_dpp $vgpr0, $vgpr257, $vgpr0, undef $vgpr258, 1, 15, 15, 1, implicit $exec ; DS (addr, data0, and data1 operands) - ; GCN-NEXT: s_set_vgpr_msb 20 + ; GCN-NEXT: s_set_vgpr_msb 0x1114 ; ASM-SAME: ; msbs: dst=0 src0=0 src1=1 src2=1 ; GCN-NEXT: ds_store_2addr_b32 v0, v248 /*v504*/, v249 /*v505*/ offset1:1 DS_WRITE2_B32_gfx9 $vgpr0, undef $vgpr504, undef $vgpr505, 0, 1, 0, implicit $exec ; Reset - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x1400 ; ASM-SAME: ; msbs: dst=0 src0=0 src1=0 src2=0 ; GCN-NEXT: ds_store_2addr_b32 v0, v248, v249 offset1:1 DS_WRITE2_B32_gfx9 $vgpr0, undef $vgpr248, undef $vgpr249, 0, 1, 0, implicit $exec @@ -93,13 +93,13 @@ body: | ; GCN-NEXT: ds_load_b32 v0, v255 /*v511*/ $vgpr0 = DS_READ_B32_gfx9 $vgpr511, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0x44 + ; GCN-NEXT: s_set_vgpr_msb 0x144 ; ASM-SAME: ; msbs: dst=1 src0=0 src1=1 src2=0 ; GCN-NEXT: ds_add_rtn_u32 v255 /*v511*/, v0, v248 /*v504*/ $vgpr511 = DS_ADD_RTN_U32_gfx9 $vgpr0, undef $vgpr504, 0, 0, implicit $exec ; Reset - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4400 ; ASM-SAME: ; msbs: dst=0 src0=0 src1=0 src2=0 ; GCN-NEXT: ds_add_rtn_u32 v0, v0, v0 $vgpr0 = DS_ADD_RTN_U32_gfx9 $vgpr0, $vgpr0, 0, 0, implicit $exec @@ -111,17 +111,17 @@ body: | ; GCN-NEXT: global_load_b32 v2, v[2:3] /*v[258:259]*/, off $vgpr2 = GLOBAL_LOAD_DWORD undef $vgpr258_vgpr259, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 64 + ; GCN-NEXT: s_set_vgpr_msb 0x140 ; ASM-SAME: ; msbs: dst=1 src0=0 src1=0 src2=0 ; GCN-NEXT: global_load_b32 v255 /*v511*/, v0, s[0:1] $vgpr511 = GLOBAL_LOAD_DWORD_SADDR undef $sgpr0_sgpr1, $vgpr0, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 1 + ; GCN-NEXT: s_set_vgpr_msb 0x4001 ; ASM-SAME: ; msbs: dst=0 src0=1 src1=0 src2=0 ; GCN-NEXT: scratch_load_u8 v0, v255 /*v511*/, s0 $vgpr0 = SCRATCH_LOAD_UBYTE_SVS $vgpr511, undef $sgpr0, 0, 0, implicit $exec, implicit $flat_scr - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x100 ; ASM-SAME: ; msbs: dst=0 src0=0 src1=0 src2=0 ; GCN-NEXT: global_store_b32 v[0:1], v2, off GLOBAL_STORE_DWORD $vgpr0_vgpr1, $vgpr2, 0, 0, implicit $exec @@ -135,13 +135,13 @@ body: | ; GCN-NEXT: global_store_b96 v[0:1] /*v[256:257]*/, v[244:246] /*v[500:502]*/, off GLOBAL_STORE_DWORDX3 $vgpr256_vgpr257, $vgpr500_vgpr501_vgpr502, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0x44 + ; GCN-NEXT: s_set_vgpr_msb 0x544 ; ASM-SAME: ; msbs: dst=1 src0=0 src1=1 src2=0 ; GCN-NEXT: flat_atomic_add_u32 v254 /*v510*/, v[0:1], v255 /*v511*/ th:TH_ATOMIC_RETURN $vgpr510 = FLAT_ATOMIC_ADD_RTN $vgpr0_vgpr1, $vgpr511, 0, 1, implicit $exec, implicit $flat_scr ; Reset - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4400 ; ASM-SAME: ; msbs: dst=0 src0=0 src1=0 src2=0 ; GCN-NEXT: flat_atomic_add_u32 v0, v[0:1], v255 th:TH_ATOMIC_RETURN $vgpr0 = FLAT_ATOMIC_ADD_RTN $vgpr0_vgpr1, $vgpr255, 0, 1, implicit $exec, implicit $flat_scr @@ -156,12 +156,12 @@ body: | ; GCN-NEXT: buffer_load_b32 v1 /*v257*/, v0, s[8:11], s3 offen $vgpr257 = BUFFER_LOAD_DWORD_VBUFFER_OFFEN $vgpr0, undef $sgpr8_sgpr9_sgpr10_sgpr11, undef $sgpr3, 0, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0x41 + ; GCN-NEXT: s_set_vgpr_msb 0x4041 ; ASM-SAME: ; msbs: dst=1 src0=1 src1=0 src2=0 ; GCN-NEXT: buffer_load_b32 v1 /*v257*/, v0 /*v256*/, s[8:11], s3 offen $vgpr257 = BUFFER_LOAD_DWORD_VBUFFER_OFFEN $vgpr256, undef $sgpr8_sgpr9_sgpr10_sgpr11, undef $sgpr3, 0, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4100 ; ASM-SAME: ; msbs: dst=0 src0=0 src1=0 src2=0 ; GCN-NEXT: buffer_store_b32 v0, v1, s[0:3], s3 offen BUFFER_STORE_DWORD_VBUFFER_OFFEN $vgpr0, $vgpr1, undef $sgpr0_sgpr1_sgpr2_sgpr3, undef $sgpr3, 0, 0, 0, implicit $exec @@ -171,7 +171,7 @@ body: | ; GCN-NEXT: buffer_store_b32 v0 /*v256*/, v1 /*v257*/, s[0:3], s3 offen BUFFER_STORE_DWORD_VBUFFER_OFFEN $vgpr256, $vgpr257, undef $sgpr0_sgpr1_sgpr2_sgpr3, undef $sgpr3, 0, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4100 ; ASM-SAME: ; msbs: dst=0 src0=0 src1=0 src2=0 ; GCN-NEXT: buffer_atomic_add_f32 v0, v1, s[0:3], s3 offen BUFFER_ATOMIC_ADD_F32_VBUFFER_OFFEN $vgpr0, $vgpr1, undef $sgpr0_sgpr1_sgpr2_sgpr3, undef $sgpr3, 0, 0, implicit $exec @@ -183,44 +183,44 @@ body: | ; VGPRs above 512 - ; GCN-NEXT: s_set_vgpr_msb 0xaa + ; GCN-NEXT: s_set_vgpr_msb 0x41aa ; ASM-SAME: ; msbs: dst=2 src0=2 src1=2 src2=2 ; GCN-NEXT: v_fma_f32 v0 /*v512*/, v1 /*v513*/, v2 /*v514*/, v3 /*v515*/ $vgpr512 = V_FMA_F32_e64 0, undef $vgpr513, 0, undef $vgpr514, 0, undef $vgpr515, 0, 0, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0xab + ; GCN-NEXT: s_set_vgpr_msb 0xaaab ; ASM-SAME: ; msbs: dst=2 src0=3 src1=2 src2=2 ; GCN-NEXT: v_fma_f32 v0 /*v512*/, v0 /*v768*/, v2 /*v514*/, v3 /*v515*/ $vgpr512 = V_FMA_F32_e64 0, undef $vgpr768, 0, undef $vgpr514, 0, undef $vgpr515, 0, 0, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0xae + ; GCN-NEXT: s_set_vgpr_msb 0xabae ; ASM-SAME: ; msbs: dst=2 src0=2 src1=3 src2=2 ; GCN-NEXT: v_fma_f32 v0 /*v512*/, v1 /*v513*/, v2 /*v770*/, v3 /*v515*/ $vgpr512 = V_FMA_F32_e64 0, undef $vgpr513, 0, undef $vgpr770, 0, undef $vgpr515, 0, 0, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0xba + ; GCN-NEXT: s_set_vgpr_msb 0xaeba ; ASM-SAME: ; msbs: dst=2 src0=2 src1=2 src2=3 ; GCN-NEXT: v_fma_f32 v0 /*v512*/, v1 /*v513*/, v2 /*v514*/, v3 /*v771*/ $vgpr512 = V_FMA_F32_e64 0, undef $vgpr513, 0, undef $vgpr514, 0, undef $vgpr771, 0, 0, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0xea + ; GCN-NEXT: s_set_vgpr_msb 0xbaea ; ASM-SAME: ; msbs: dst=3 src0=2 src1=2 src2=2 ; GCN-NEXT: v_fma_f32 v255 /*v1023*/, v1 /*v513*/, v2 /*v514*/, v3 /*v515*/ $vgpr1023 = V_FMA_F32_e64 0, undef $vgpr513, 0, undef $vgpr514, 0, undef $vgpr515, 0, 0, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0xff + ; GCN-NEXT: s_set_vgpr_msb 0xeaff ; ASM-SAME: ; msbs: dst=3 src0=3 src1=3 src2=3 ; GCN-NEXT: v_fma_f32 v0 /*v768*/, v1 /*v769*/, v2 /*v770*/, v3 /*v771*/ $vgpr768 = V_FMA_F32_e64 0, undef $vgpr769, 0, undef $vgpr770, 0, undef $vgpr771, 0, 0, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0x42 + ; GCN-NEXT: s_set_vgpr_msb 0xff42 ; ASM-SAME: ; msbs: dst=1 src0=2 src1=0 src2=0 ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v0 /*v512*/ $vgpr256 = V_MOV_B32_e32 undef $vgpr512, implicit $exec ; Reset - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4200 ; ASM-SAME: ; msbs: dst=0 src0=0 src1=0 src2=0 ; GCN-NEXT: v_fma_f32 v0, v1, v2, v3 $vgpr0 = V_FMA_F32_e64 0, undef $vgpr1, 0, undef $vgpr2, 0, undef $vgpr3, 0, 0, implicit $exec, implicit $mode @@ -232,12 +232,12 @@ body: | ; GCN-NEXT: global_store_b96 v[0:1] /*v[512:513]*/, v[0:2] /*v[512:514]*/, off GLOBAL_STORE_DWORDX3 $vgpr512_vgpr513, $vgpr512_vgpr513_vgpr514, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 11 + ; GCN-NEXT: s_set_vgpr_msb 0xa0b ; ASM-SAME: ; msbs: dst=0 src0=3 src1=2 src2=0 ; GCN-NEXT: global_store_b64 v[254:255] /*v[1022:1023]*/, v[254:255] /*v[766:767]*/, off GLOBAL_STORE_DWORDX2 $vgpr1022_vgpr1023, $vgpr766_vgpr767, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0x55 + ; GCN-NEXT: s_set_vgpr_msb 0xb55 ; ASM-SAME: ; msbs: dst=1 src0=1 src1=1 src2=1 ; GCN-NEXT: v_wmma_f32_16x16x32_bf16 v[14:21] /*v[270:277]*/, v[26:33] /*v[282:289]*/, v[34:41] /*v[290:297]*/, v[14:21] /*v[270:277]*/ early-clobber $vgpr270_vgpr271_vgpr272_vgpr273_vgpr274_vgpr275_vgpr276_vgpr277 = V_WMMA_F32_16X16X32_BF16_w32_twoaddr 8, undef $vgpr282_vgpr283_vgpr284_vgpr285_vgpr286_vgpr287_vgpr288_vgpr289, 8, undef $vgpr290_vgpr291_vgpr292_vgpr293_vgpr294_vgpr295_vgpr296_vgpr297, 8, killed undef $vgpr270_vgpr271_vgpr272_vgpr273_vgpr274_vgpr275_vgpr276_vgpr277, 0, 0, 0, 0, implicit $exec @@ -247,6 +247,7 @@ body: | ... # ASM-LABEL: {{^}}vopd: + # DIS-LABEL: <vopd>: --- name: vopd @@ -262,35 +263,35 @@ body: | ; GCN-NEXT: v_dual_sub_f32 v244 /*v500*/, v1, v2 :: v_dual_mul_f32 v0 /*v256*/, v3, v4 $vgpr500, $vgpr256 = V_DUAL_SUB_F32_e32_X_MUL_F32_e32_gfx1250 undef $vgpr1, undef $vgpr2, undef $vgpr3, undef $vgpr4, implicit $mode, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0x41 + ; GCN-NEXT: s_set_vgpr_msb 0x4041 ; GCN-NEXT: v_dual_sub_f32 v244 /*v500*/, s1, v2 :: v_dual_mul_f32 v0 /*v256*/, v44 /*v300*/, v4 $vgpr500, $vgpr256 = V_DUAL_SUB_F32_e32_X_MUL_F32_e32_gfx1250 undef $sgpr1, undef $vgpr2, undef $vgpr300, undef $vgpr4, implicit $mode, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 4 + ; GCN-NEXT: s_set_vgpr_msb 0x4104 ; GCN-NEXT: v_dual_sub_f32 v255, v1, v44 /*v300*/ :: v_dual_mul_f32 v6, v0, v1 /*v257*/ $vgpr255, $vgpr6 = V_DUAL_SUB_F32_e32_X_MUL_F32_e32_gfx1250 undef $vgpr1, undef $vgpr300, undef $vgpr0, $vgpr257, implicit $mode, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 1 + ; GCN-NEXT: s_set_vgpr_msb 0x401 ; GCN-NEXT: v_dual_sub_f32 v255, 0, v1 :: v_dual_mul_f32 v6, v44 /*v300*/, v3 $vgpr255, $vgpr6 = V_DUAL_SUB_F32_e32_X_MUL_F32_e32_gfx1250 0, undef $vgpr1, undef $vgpr300, undef $vgpr3, implicit $mode, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 64 + ; GCN-NEXT: s_set_vgpr_msb 0x140 ; GCN-NEXT: v_dual_fmamk_f32 v243 /*v499*/, v0, 0xa, v3 :: v_dual_fmac_f32 v0 /*v256*/, v1, v1 $vgpr499, $vgpr256 = V_DUAL_FMAMK_F32_X_FMAC_F32_e32_gfx1250 undef $vgpr0, 10, undef $vgpr3, undef $vgpr1, undef $vgpr1, $vgpr256, implicit $mode, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 5 + ; GCN-NEXT: s_set_vgpr_msb 0x4005 ; GCN-NEXT: v_dual_mov_b32 v2, v3 /*v259*/ :: v_dual_add_f32 v3, v1 /*v257*/, v2 /*v258*/ $vgpr2, $vgpr3 = V_DUAL_MOV_B32_e32_X_ADD_F32_e32_gfx1250 undef $vgpr259, undef $vgpr257, undef $vgpr258, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0x44 + ; GCN-NEXT: s_set_vgpr_msb 0x544 ; GCN-NEXT: v_dual_fmamk_f32 v244 /*v500*/, v0, 0xa, v44 /*v300*/ :: v_dual_fmac_f32 v3 /*v259*/, v1, v1 /*v257*/ $vgpr500, $vgpr259 = V_DUAL_FMAMK_F32_X_FMAC_F32_e32_gfx1250 undef $vgpr0, 10, undef $vgpr300, undef $vgpr1, undef $vgpr257, $vgpr259, implicit $mode, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 16 + ; GCN-NEXT: s_set_vgpr_msb 0x4410 ; GCN-NEXT: v_dual_fma_f32 v0, v6, v6, v44 /*v300*/ :: v_dual_fma_f32 v1, v4, v5, v45 /*v301*/ $vgpr0, $vgpr1 = V_DUAL_FMA_F32_e64_X_FMA_F32_e64_e96_gfx1250 0, undef $vgpr6, 0, undef $vgpr6, 0, undef $vgpr300, 0, undef $vgpr4, 0, undef $vgpr5, 0, undef $vgpr301, implicit $mode, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x1000 ; GCN-NEXT: v_dual_fmac_f32 v2, v6, v6 :: v_dual_fma_f32 v3, v4, v5, v3 $vgpr2, $vgpr3 = V_DUAL_FMAC_F32_e32_X_FMA_F32_e64_e96_gfx1250 0, undef $vgpr6, 0, undef $vgpr6, undef $vgpr2, 0, undef $vgpr4, 0, undef $vgpr5, 0, $vgpr3, implicit $mode, implicit $exec @@ -298,7 +299,7 @@ body: | ; GCN-NEXT: v_dual_fma_f32 v244 /*v500*/, v6, v7, v8 :: v_dual_add_f32 v3 /*v259*/, v4, v5 $vgpr500, $vgpr259 = V_DUAL_FMA_F32_e64_X_ADD_F32_e32_e96_gfx1250 0, undef $vgpr6, 0, undef $vgpr7, 0, undef $vgpr8, 0, undef $vgpr4, 0, undef $vgpr5, implicit $mode, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0xae + ; GCN-NEXT: s_set_vgpr_msb 0x40ae ; GCN-NEXT: v_dual_fmac_f32 v2 /*v514*/, v6 /*v518*/, v8 /*v776*/ :: v_dual_fma_f32 v3 /*v515*/, v4 /*v516*/, v7 /*v775*/, v3 /*v515*/ $vgpr514, $vgpr515 = V_DUAL_FMAC_F32_e32_X_FMA_F32_e64_e96_gfx1250 0, undef $vgpr518, 0, undef $vgpr776, undef $vgpr514, 0, undef $vgpr516, 0, undef $vgpr775, 0, $vgpr515, implicit $mode, implicit $exec @@ -319,31 +320,31 @@ body: | ; GCN-NEXT: v_fmaak_f32 v0 /*v256*/, v1 /*v257*/, v2 /*v258*/, 0x1 $vgpr256 = V_FMAAK_F32 undef $vgpr257, undef $vgpr258, 1, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 5 + ; GCN-NEXT: s_set_vgpr_msb 0x4505 ; GCN-NEXT: v_fmaak_f32 v0, v1 /*v257*/, v2 /*v258*/, 0x1 $vgpr0 = V_FMAAK_F32 undef $vgpr257, undef $vgpr258, 1, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0x41 + ; GCN-NEXT: s_set_vgpr_msb 0x541 ; GCN-NEXT: v_fmaak_f32 v0 /*v256*/, v1 /*v257*/, v2, 0x1 $vgpr256 = V_FMAAK_F32 undef $vgpr257, undef $vgpr2, 1, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0x44 + ; GCN-NEXT: s_set_vgpr_msb 0x4144 ; GCN-NEXT: v_fmaak_f32 v0 /*v256*/, v1, v2 /*v258*/, 0x1 $vgpr256 = V_FMAAK_F32 undef $vgpr1, undef $vgpr258, 1, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0x45 + ; GCN-NEXT: s_set_vgpr_msb 0x4445 ; GCN-NEXT: v_fmamk_f32 v0 /*v256*/, v1 /*v257*/, 0x1, v2 /*v258*/ $vgpr256 = V_FMAMK_F32 undef $vgpr257, 1, undef $vgpr258, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 5 + ; GCN-NEXT: s_set_vgpr_msb 0x4505 ; GCN-NEXT: v_fmamk_f32 v0, v1 /*v257*/, 0x1, v2 /*v258*/ $vgpr0 = V_FMAMK_F32 undef $vgpr257, 1, undef $vgpr258, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0x41 + ; GCN-NEXT: s_set_vgpr_msb 0x541 ; GCN-NEXT: v_fmamk_f32 v0 /*v256*/, v1 /*v257*/, 0x1, v2 $vgpr256 = V_FMAMK_F32 undef $vgpr257, 1, undef $vgpr2, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0x44 + ; GCN-NEXT: s_set_vgpr_msb 0x4144 ; GCN-NEXT: v_fmamk_f32 v0 /*v256*/, v1, 0x1, v2 /*v258*/ $vgpr256 = V_FMAMK_F32 undef $vgpr1, 1, undef $vgpr258, implicit $exec, implicit $mode @@ -389,15 +390,15 @@ body: | ; GCN-NEXT: v_lshlrev_b32_e64 v0, v0 /*v256*/, v2 $vgpr0 = V_LSHLREV_B32_e64 undef $vgpr256, undef $vgpr2, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 4 + ; GCN-NEXT: s_set_vgpr_msb 0x104 ; GCN-NEXT: v_lshlrev_b32_e64 v0, v1, v0 /*v256*/ $vgpr0 = V_LSHLREV_B32_e64 undef $vgpr1, undef $vgpr256, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 1 + ; GCN-NEXT: s_set_vgpr_msb 0x401 ; GCN-NEXT: v_subrev_nc_u32_e32 v0, v0 /*v256*/, v2 $vgpr0 = V_SUBREV_U32_e32 undef $vgpr256, undef $vgpr2, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 4 + ; GCN-NEXT: s_set_vgpr_msb 0x104 ; GCN-NEXT: v_subrev_nc_u32_e32 v0, v1, v0 /*v256*/ $vgpr0 = V_SUBREV_U32_e32 undef $vgpr1, undef $vgpr256, implicit $exec @@ -417,7 +418,7 @@ body: | ; GCN-NEXT: v_fma_f32 v3 /*v259*/, v4 /*v260*/, v5 /*v261*/, v6 /*v262*/ $vgpr259 = V_FMA_F32_e64 0, undef $vgpr260, 0, undef $vgpr261, 0, undef $vgpr262, 0, 0, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x5500 ; GCN-NEXT: v_add_nc_u32_e32 v0, v1, v2 $vgpr0 = V_ADD_U32_e32 undef $vgpr1, undef $vgpr2, implicit $exec @@ -431,7 +432,7 @@ body: | ; GCN-NEXT: v_add_nc_u32_e32 v0 /*v256*/, v1, v2 $vgpr256 = V_ADD_U32_e32 undef $vgpr1, undef $vgpr2, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 ; GCN-NEXT: v_fma_f32 v3, v4, v5, s2 $vgpr3 = V_FMA_F32_e64 0, undef $vgpr4, 0, undef $vgpr5, 0, undef $sgpr2, 0, 0, implicit $exec, implicit $mode @@ -439,17 +440,17 @@ body: | ; GCN-NEXT: v_fma_f32 v3, v4 /*v260*/, v5, 1 $vgpr3 = V_FMA_F32_e64 0, undef $vgpr260, 0, undef $vgpr5, 0, 1, 0, 0, implicit $exec, implicit $mode - ; GCN-NEXT: s_set_vgpr_msb 4 + ; GCN-NEXT: s_set_vgpr_msb 0x104 ; GCN-NEXT: v_mov_b32_e32 v0, v1 $vgpr0 = V_MOV_B32_e32 undef $vgpr1, implicit $exec ; GCN-NEXT: v_add_nc_u32_e32 v2, v1, v3 /*v259*/ $vgpr2 = V_ADD_U32_e32 undef $vgpr1, undef $vgpr259, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 1 + ; GCN-NEXT: s_set_vgpr_msb 0x401 ; GCN-NEXT: v_mov_b32_e32 v0, v0 /*v256*/ ; GCN-NEXT: v_add_nc_u32_e32 v1, v1 /*v257*/, v1 - ; GCN-NEXT: s_set_vgpr_msb 5 + ; GCN-NEXT: s_set_vgpr_msb 0x105 ; GCN-NEXT: v_add_nc_u32_e32 v2, v2 /*v258*/, v2 /*v258*/ $vgpr0 = V_MOV_B32_e32 undef $vgpr256, implicit $exec $vgpr1 = V_ADD_U32_e32 undef $vgpr257, undef $vgpr1, implicit $exec @@ -478,16 +479,18 @@ body: | ; ASM: .LBB{{.*_1}}: ; GCN-NEXT: s_set_vgpr_msb 64 ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v1 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 $vgpr256 = V_MOV_B32_e32 undef $vgpr1, implicit $exec - ; No mode switch on fall through + ; Reset on fallthrough block end bb.2: ; ASM-NEXT: %bb.2: - ; GCN-NEXT: s_nop 0 - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 64 + ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v1 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 ; GCN-NEXT: s_branch - S_NOP 0 + $vgpr256 = V_MOV_B32_e32 undef $vgpr1, implicit $exec S_BRANCH %bb.3 ; Reset mode on terminator @@ -496,7 +499,7 @@ body: | ; ASM: .LBB{{.*_3}}: ; GCN-NEXT: s_set_vgpr_msb 64 ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v1 - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 ; GCN-NEXT: s_swap_pc_i64 $vgpr256 = V_MOV_B32_e32 undef $vgpr1, implicit $exec $exec = S_SWAPPC_B64 undef $sgpr0_sgpr1 @@ -518,7 +521,7 @@ body: | ; GCN-NEXT: v_mov_b32_e32 v0, v1 ; GCN-NEXT: s_set_vgpr_msb 64 ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v1 - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 ; GCN-NEXT: s_set_pc_i64 $vgpr0 = V_MOV_B32_e32 undef $vgpr1, implicit $exec $vgpr256 = V_MOV_B32_e32 undef $vgpr1, implicit $exec @@ -538,7 +541,7 @@ body: | ; ASM-NEXT: %bb.7: ; GCN-NEXT: s_set_vgpr_msb 64 ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v1 - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 ; ASM-NEXT: ; return to shader part epilog $vgpr256 = V_MOV_B32_e32 undef $vgpr1, implicit $exec SI_RETURN_TO_EPILOG undef $vgpr0, implicit-def $exec @@ -556,7 +559,7 @@ body: | ; ASM-NEXT: %bb.9: ; GCN-NEXT: s_set_vgpr_msb 64 ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v1 - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 ; GCN-NEXT: s_set_pc_i64 $vgpr256 = V_MOV_B32_e32 undef $vgpr1, implicit $exec S_SETPC_B64_return undef $sgpr0_sgpr1, implicit-def $exec @@ -574,13 +577,14 @@ body: | ; ASM: %bb.0: ; GCN-NEXT: s_set_vgpr_msb 64 ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v0 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 $vgpr256 = V_MOV_B32_e32 undef $vgpr0, implicit $exec bb.1: ; ASM: .LBB{{[0-9]+}}_1: ; GCN-NEXT: s_set_vgpr_msb 64 ; GCN-NEXT: v_mov_b32_e32 v1 /*v257*/, v1 - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 ; GCN-NEXT: s_cbranch_scc0 $vgpr257 = V_MOV_B32_e32 undef $vgpr1, implicit $exec S_CBRANCH_SCC0 %bb.1, undef implicit $scc @@ -604,7 +608,7 @@ body: | ; ASM: %bb.0: ; GCN-NEXT: s_set_vgpr_msb 64 ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v1 - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 ; ASM: def v0 ; GCN-NOT: s_set_vgpr_msb ; ASM: use v0 @@ -638,7 +642,7 @@ body: | ; GCN-NEXT: s_set_vgpr_msb 64 ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v1 ; GCN-NEXT: s_nop 0 - ; GCN-NEXT: s_set_vgpr_msb 1 + ; GCN-NEXT: s_set_vgpr_msb 0x4001 ; GCN-NEXT: v_mov_b32_e32 v1, v0 /*v256*/ BUNDLE implicit-def $vgpr256 { $vgpr256 = V_MOV_B32_e32 undef $vgpr1, implicit $exec @@ -680,7 +684,7 @@ body: | ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v1 ; GCN-NEXT: v_mov_b32_e32 v1 /*v257*/, v1 - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 ; GCN-NEXT: v_mov_b32_e32 v2, v1 ; GCN-NEXT: v_mov_b32_e32 v3, v1 BUNDLE implicit-def $vgpr256, implicit-def $vgpr257, implicit-def $vgpr2, implicit-def $vgpr3, implicit undef $vgpr1 { @@ -709,7 +713,7 @@ body: | ; GCN-NEXT: s_clause 0x3e ; GCN-NEXT: v_mov_b32_e32 v0 /*v256*/, v1 - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x4000 ; GCN-NEXT: v_mov_b32_e32 v1, v1 ; GCN-NEXT: v_mov_b32_e32 v2, v1 ; GCN-COUNT-60: v_mov_b32_e32 v1, v1 @@ -823,7 +827,7 @@ body: | ; GCN-NEXT: v_wmma_scale_f32_16x16x128_f8f6f4 v[210:217], v[244:259] /*v[500:515]*/, v[244:259] /*v[500:515]*/, v[10:17], v1, v2 $vgpr210_vgpr211_vgpr212_vgpr213_vgpr214_vgpr215_vgpr216_vgpr217 = V_WMMA_SCALE_F32_16X16X128_F8F6F4_f8_f8_w32_threeaddr undef $vgpr500_vgpr501_vgpr502_vgpr503_vgpr504_vgpr505_vgpr506_vgpr507_vgpr508_vgpr509_vgpr510_vgpr511_vgpr512_vgpr513_vgpr514_vgpr515, undef $vgpr500_vgpr501_vgpr502_vgpr503_vgpr504_vgpr505_vgpr506_vgpr507_vgpr508_vgpr509_vgpr510_vgpr511_vgpr512_vgpr513_vgpr514_vgpr515, 0, undef $vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17, undef $vgpr1, undef $vgpr2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x500 ; GCN-NEXT: v_wmma_scale_f32_16x16x128_f8f6f4 v[210:217], v[100:115], v[100:115], v[10:17], v1, v2 $vgpr210_vgpr211_vgpr212_vgpr213_vgpr214_vgpr215_vgpr216_vgpr217 = V_WMMA_SCALE_F32_16X16X128_F8F6F4_f8_f8_w32_threeaddr undef $vgpr100_vgpr101_vgpr102_vgpr103_vgpr104_vgpr105_vgpr106_vgpr107_vgpr108_vgpr109_vgpr110_vgpr111_vgpr112_vgpr113_vgpr114_vgpr115, undef $vgpr100_vgpr101_vgpr102_vgpr103_vgpr104_vgpr105_vgpr106_vgpr107_vgpr108_vgpr109_vgpr110_vgpr111_vgpr112_vgpr113_vgpr114_vgpr115, 0, undef $vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17, undef $vgpr1, undef $vgpr2, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, implicit $exec @@ -835,11 +839,11 @@ body: | ; GCN-NEXT: v_wmma_ld_scale16_paired_b64 v[0:1], v[2:3] V_WMMA_LD_SCALE16_PAIRED_B64 undef $vgpr0_vgpr1, undef $vgpr2_vgpr3, 0, 0, 0, 0, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 5 + ; GCN-NEXT: s_set_vgpr_msb 0x105 ; GCN-NEXT: v_wmma_scale16_f32_16x16x128_f8f6f4 v[210:217], v[244:259] /*v[500:515]*/, v[244:259] /*v[500:515]*/, v[10:17], v[0:1], v[2:3] $vgpr210_vgpr211_vgpr212_vgpr213_vgpr214_vgpr215_vgpr216_vgpr217 = V_WMMA_SCALE16_F32_16X16X128_F8F6F4_f8_f8_w32_threeaddr undef $vgpr500_vgpr501_vgpr502_vgpr503_vgpr504_vgpr505_vgpr506_vgpr507_vgpr508_vgpr509_vgpr510_vgpr511_vgpr512_vgpr513_vgpr514_vgpr515, undef $vgpr500_vgpr501_vgpr502_vgpr503_vgpr504_vgpr505_vgpr506_vgpr507_vgpr508_vgpr509_vgpr510_vgpr511_vgpr512_vgpr513_vgpr514_vgpr515, 0, undef $vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17, undef $vgpr0_vgpr1, undef $vgpr2_vgpr3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, implicit $exec - ; GCN-NEXT: s_set_vgpr_msb 0 + ; GCN-NEXT: s_set_vgpr_msb 0x500 ; GCN-NEXT: v_wmma_scale16_f32_16x16x128_f8f6f4 v[210:217], v[100:115], v[100:115], v[10:17], v[0:1], v[2:3] $vgpr210_vgpr211_vgpr212_vgpr213_vgpr214_vgpr215_vgpr216_vgpr217 = V_WMMA_SCALE16_F32_16X16X128_F8F6F4_f8_f8_w32_threeaddr undef $vgpr100_vgpr101_vgpr102_vgpr103_vgpr104_vgpr105_vgpr106_vgpr107_vgpr108_vgpr109_vgpr110_vgpr111_vgpr112_vgpr113_vgpr114_vgpr115, undef $vgpr100_vgpr101_vgpr102_vgpr103_vgpr104_vgpr105_vgpr106_vgpr107_vgpr108_vgpr109_vgpr110_vgpr111_vgpr112_vgpr113_vgpr114_vgpr115, 0, undef $vgpr10_vgpr11_vgpr12_vgpr13_vgpr14_vgpr15_vgpr16_vgpr17, undef $vgpr0_vgpr1, undef $vgpr2_vgpr3, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, implicit $exec diff --git a/llvm/test/CodeGen/AMDGPU/vgpr-spill-emergency-stack-slot.ll b/llvm/test/CodeGen/AMDGPU/vgpr-spill-emergency-stack-slot.ll index db49339..9c16b3c 100644 --- a/llvm/test/CodeGen/AMDGPU/vgpr-spill-emergency-stack-slot.ll +++ b/llvm/test/CodeGen/AMDGPU/vgpr-spill-emergency-stack-slot.ll @@ -22,8 +22,6 @@ ; GFX9-DAG: s_mov_b32 s[[DESC3:[0-9]+]], 0xe00000 ; OFFREG is offset system SGPR -; GCN: buffer_store_dword {{v[0-9]+}}, off, s[[[DESC0]]:[[DESC3]]], 0 offset:{{[0-9]+}} ; 4-byte Folded Spill -; GCN: buffer_load_dword v{{[0-9]+}}, off, s[[[DESC0]]:[[DESC3]]], 0 offset:{{[0-9]+}} ; 4-byte Folded Reload ; GCN: NumVgprs: 256 ; GCN: ScratchSize: 640 diff --git a/llvm/test/CodeGen/AMDGPU/wait-xcnt.mir b/llvm/test/CodeGen/AMDGPU/wait-xcnt.mir index 1b8e126..a1381ec 100644 --- a/llvm/test/CodeGen/AMDGPU/wait-xcnt.mir +++ b/llvm/test/CodeGen/AMDGPU/wait-xcnt.mir @@ -945,7 +945,6 @@ body: | $vgpr0 = V_MOV_B32_e32 0, implicit $exec ... -# FIXME: Missing S_WAIT_XCNT before overwriting vgpr0. --- name: wait_kmcnt_with_outstanding_vmem_2 tracksRegLiveness: true @@ -971,6 +970,7 @@ body: | ; GCN-NEXT: {{ $}} ; GCN-NEXT: S_WAIT_KMCNT 0 ; GCN-NEXT: $sgpr2 = S_MOV_B32 $sgpr2 + ; GCN-NEXT: S_WAIT_XCNT 0 ; GCN-NEXT: $vgpr0 = V_MOV_B32_e32 0, implicit $exec bb.0: liveins: $vgpr0_vgpr1, $sgpr0_sgpr1, $scc @@ -986,6 +986,180 @@ body: | ... --- +name: wait_kmcnt_and_wait_loadcnt +tracksRegLiveness: true +machineFunctionInfo: + isEntryFunction: true +body: | + ; GCN-LABEL: name: wait_kmcnt_and_wait_loadcnt + ; GCN: bb.0: + ; GCN-NEXT: successors: %bb.2(0x40000000), %bb.1(0x40000000) + ; GCN-NEXT: liveins: $vgpr0_vgpr1, $sgpr0_sgpr1, $scc + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: $sgpr2 = S_LOAD_DWORD_IMM $sgpr0_sgpr1, 0, 0 + ; GCN-NEXT: S_CBRANCH_SCC1 %bb.2, implicit $scc + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: bb.1: + ; GCN-NEXT: successors: %bb.2(0x80000000) + ; GCN-NEXT: liveins: $vgpr0_vgpr1, $sgpr2 + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: $vgpr2 = GLOBAL_LOAD_DWORD $vgpr0_vgpr1, 0, 0, implicit $exec + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: bb.2: + ; GCN-NEXT: liveins: $sgpr2 + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: S_WAIT_KMCNT 0 + ; GCN-NEXT: $sgpr2 = S_MOV_B32 $sgpr2 + ; GCN-NEXT: S_WAIT_LOADCNT 0 + ; GCN-NEXT: $vgpr2 = V_MOV_B32_e32 0, implicit $exec + bb.0: + liveins: $vgpr0_vgpr1, $sgpr0_sgpr1, $scc + $sgpr2 = S_LOAD_DWORD_IMM $sgpr0_sgpr1, 0, 0 + S_CBRANCH_SCC1 %bb.2, implicit $scc + bb.1: + liveins: $vgpr0_vgpr1, $sgpr2 + $vgpr2 = GLOBAL_LOAD_DWORD $vgpr0_vgpr1, 0, 0, implicit $exec + bb.2: + liveins: $sgpr2 + $sgpr2 = S_MOV_B32 $sgpr2 + $vgpr2 = V_MOV_B32_e32 0, implicit $exec +... + +--- +name: implicit_handling_of_pending_vmem_group +tracksRegLiveness: true +machineFunctionInfo: + isEntryFunction: true +body: | + ; GCN-LABEL: name: implicit_handling_of_pending_vmem_group + ; GCN: bb.0: + ; GCN-NEXT: successors: %bb.2(0x40000000), %bb.1(0x40000000) + ; GCN-NEXT: liveins: $vgpr0_vgpr1, $sgpr0_sgpr1, $scc + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: $sgpr2 = S_LOAD_DWORD_IMM $sgpr0_sgpr1, 0, 0 + ; GCN-NEXT: S_CBRANCH_SCC1 %bb.2, implicit $scc + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: bb.1: + ; GCN-NEXT: successors: %bb.2(0x80000000) + ; GCN-NEXT: liveins: $vgpr0_vgpr1, $sgpr2 + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: $vgpr2 = GLOBAL_LOAD_DWORD $vgpr0_vgpr1, 0, 0, implicit $exec + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: bb.2: + ; GCN-NEXT: liveins: $sgpr0_sgpr1, $sgpr2 + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: S_WAIT_KMCNT 0 + ; GCN-NEXT: $sgpr2 = S_MOV_B32 $sgpr2 + ; GCN-NEXT: $sgpr2 = S_LOAD_DWORD_IMM $sgpr0_sgpr1, 0, 0 + ; GCN-NEXT: $vgpr0 = V_MOV_B32_e32 0, implicit $exec + ; GCN-NEXT: S_WAIT_XCNT 0 + ; GCN-NEXT: $sgpr0 = S_MOV_B32 $sgpr0 + bb.0: + liveins: $vgpr0_vgpr1, $sgpr0_sgpr1, $scc + $sgpr2 = S_LOAD_DWORD_IMM $sgpr0_sgpr1, 0, 0 + S_CBRANCH_SCC1 %bb.2, implicit $scc + bb.1: + liveins: $vgpr0_vgpr1, $sgpr2 + $vgpr2 = GLOBAL_LOAD_DWORD $vgpr0_vgpr1, 0, 0, implicit $exec + bb.2: + liveins: $sgpr0_sgpr1, $sgpr2 + $sgpr2 = S_MOV_B32 $sgpr2 + $sgpr2 = S_LOAD_DWORD_IMM $sgpr0_sgpr1, 0, 0 + $vgpr0 = V_MOV_B32_e32 0, implicit $exec + $sgpr0 = S_MOV_B32 $sgpr0 +... + +--- +name: pending_vmem_event_between_block +tracksRegLiveness: true +machineFunctionInfo: + isEntryFunction: true +body: | + ; GCN-LABEL: name: pending_vmem_event_between_block + ; GCN: bb.0: + ; GCN-NEXT: successors: %bb.2(0x40000000), %bb.1(0x40000000) + ; GCN-NEXT: liveins: $vgpr0_vgpr1, $sgpr0_sgpr1, $scc + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: $sgpr2 = S_LOAD_DWORD_IMM $sgpr0_sgpr1, 0, 0 + ; GCN-NEXT: S_CBRANCH_SCC1 %bb.2, implicit $scc + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: bb.1: + ; GCN-NEXT: successors: %bb.2(0x80000000) + ; GCN-NEXT: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3, $sgpr2 + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: $vgpr4 = GLOBAL_LOAD_DWORD $vgpr0_vgpr1, 0, 0, implicit $exec + ; GCN-NEXT: $vgpr5 = GLOBAL_LOAD_DWORD $vgpr2_vgpr3, 0, 0, implicit $exec + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: bb.2: + ; GCN-NEXT: liveins: $sgpr0_sgpr1, $sgpr2, $vgpr2 + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: S_WAIT_KMCNT 0 + ; GCN-NEXT: $sgpr2 = S_MOV_B32 $sgpr2 + ; GCN-NEXT: S_WAIT_XCNT 1 + ; GCN-NEXT: $vgpr1 = V_MOV_B32_e32 0, implicit $exec + ; GCN-NEXT: S_WAIT_XCNT 0 + ; GCN-NEXT: $vgpr2 = V_MOV_B32_e32 0, implicit $exec + ; GCN-NEXT: $sgpr0 = S_MOV_B32 $sgpr0 + bb.0: + liveins: $vgpr0_vgpr1, $sgpr0_sgpr1, $scc + $sgpr2 = S_LOAD_DWORD_IMM $sgpr0_sgpr1, 0, 0 + S_CBRANCH_SCC1 %bb.2, implicit $scc + bb.1: + liveins: $vgpr0_vgpr1, $vgpr2_vgpr3, $sgpr2 + $vgpr4 = GLOBAL_LOAD_DWORD $vgpr0_vgpr1, 0, 0, implicit $exec + $vgpr5 = GLOBAL_LOAD_DWORD $vgpr2_vgpr3, 0, 0, implicit $exec + bb.2: + liveins: $sgpr0_sgpr1, $sgpr2, $vgpr2 + $sgpr2 = S_MOV_B32 $sgpr2 + $vgpr1 = V_MOV_B32_e32 0, implicit $exec + $vgpr2 = V_MOV_B32_e32 0, implicit $exec + $sgpr0 = S_MOV_B32 $sgpr0 +... + +--- +name: flushing_vmem_cnt_on_block_entry +tracksRegLiveness: true +machineFunctionInfo: + isEntryFunction: true +body: | + ; GCN-LABEL: name: flushing_vmem_cnt_on_block_entry + ; GCN: bb.0: + ; GCN-NEXT: successors: %bb.2(0x40000000), %bb.1(0x40000000) + ; GCN-NEXT: liveins: $vgpr0_vgpr1, $sgpr0_sgpr1, $scc + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: $sgpr2 = S_LOAD_DWORD_IMM $sgpr0_sgpr1, 0, 0 + ; GCN-NEXT: S_CBRANCH_SCC1 %bb.2, implicit $scc + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: bb.1: + ; GCN-NEXT: successors: %bb.2(0x80000000) + ; GCN-NEXT: liveins: $vgpr0_vgpr1, $vgpr2_vgpr3, $sgpr2 + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: $vgpr4 = GLOBAL_LOAD_DWORD $vgpr0_vgpr1, 0, 0, implicit $exec + ; GCN-NEXT: $vgpr5 = GLOBAL_LOAD_DWORD $vgpr2_vgpr3, 0, 0, implicit $exec + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: bb.2: + ; GCN-NEXT: liveins: $sgpr0_sgpr1, $sgpr2, $vgpr2 + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: S_WAIT_XCNT 0 + ; GCN-NEXT: $vgpr1 = V_MOV_B32_e32 0, implicit $exec + ; GCN-NEXT: $vgpr2 = V_MOV_B32_e32 0, implicit $exec + ; GCN-NEXT: $sgpr0 = S_MOV_B32 $sgpr0 + bb.0: + liveins: $vgpr0_vgpr1, $sgpr0_sgpr1, $scc + $sgpr2 = S_LOAD_DWORD_IMM $sgpr0_sgpr1, 0, 0 + S_CBRANCH_SCC1 %bb.2, implicit $scc + bb.1: + liveins: $vgpr0_vgpr1, $vgpr2_vgpr3, $sgpr2 + $vgpr4 = GLOBAL_LOAD_DWORD $vgpr0_vgpr1, 0, 0, implicit $exec + $vgpr5 = GLOBAL_LOAD_DWORD $vgpr2_vgpr3, 0, 0, implicit $exec + bb.2: + liveins: $sgpr0_sgpr1, $sgpr2, $vgpr2 + $vgpr1 = V_MOV_B32_e32 0, implicit $exec + $vgpr2 = V_MOV_B32_e32 0, implicit $exec + $sgpr0 = S_MOV_B32 $sgpr0 +... + +--- name: wait_loadcnt_with_outstanding_smem tracksRegLiveness: true machineFunctionInfo: diff --git a/llvm/test/CodeGen/AMDGPU/whole-wave-functions.ll b/llvm/test/CodeGen/AMDGPU/whole-wave-functions.ll index a42c8ac7..7581710 100644 --- a/llvm/test/CodeGen/AMDGPU/whole-wave-functions.ll +++ b/llvm/test/CodeGen/AMDGPU/whole-wave-functions.ll @@ -3182,7 +3182,7 @@ define amdgpu_gfx_whole_wave <2 x half> @call_gfx_from_whole_wave(i1 %active, <2 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v253 /*v509*/, s33 offset:1592 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v254 /*v510*/, s33 offset:1596 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v255 /*v511*/, s33 offset:1600 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 8 ; msbs: dst=0 src0=0 src1=2 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x408 ; msbs: dst=0 src0=0 src1=2 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v0 /*v512*/, s33 offset:1604 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v1 /*v513*/, s33 offset:1608 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v2 /*v514*/, s33 offset:1612 @@ -3443,7 +3443,7 @@ define amdgpu_gfx_whole_wave <2 x half> @call_gfx_from_whole_wave(i1 %active, <2 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v253 /*v765*/, s33 offset:2616 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v254 /*v766*/, s33 offset:2620 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v255 /*v767*/, s33 offset:2624 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 12 ; msbs: dst=0 src0=0 src1=3 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x80c ; msbs: dst=0 src0=0 src1=3 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v0 /*v768*/, s33 offset:2628 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v1 /*v769*/, s33 offset:2632 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v2 /*v770*/, s33 offset:2636 @@ -3706,7 +3706,7 @@ define amdgpu_gfx_whole_wave <2 x half> @call_gfx_from_whole_wave(i1 %active, <2 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v255 /*v1023*/, s33 offset:3648 ; GFX1250-DAGISEL-NEXT: s_wait_xcnt 0x0 ; GFX1250-DAGISEL-NEXT: s_mov_b32 exec_lo, -1 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0 ; msbs: dst=0 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0xc00 ; msbs: dst=0 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v40, s33 ; 4-byte Folded Spill ; GFX1250-DAGISEL-NEXT: s_wait_xcnt 0x0 ; GFX1250-DAGISEL-NEXT: v_writelane_b32 v40, s0, 3 @@ -4135,7 +4135,7 @@ define amdgpu_gfx_whole_wave <2 x half> @call_gfx_from_whole_wave(i1 %active, <2 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v253 /*v509*/, off, s33 offset:1592 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v254 /*v510*/, off, s33 offset:1596 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v255 /*v511*/, off, s33 offset:1600 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x80 ; msbs: dst=2 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x4080 ; msbs: dst=2 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v0 /*v512*/, off, s33 offset:1604 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v1 /*v513*/, off, s33 offset:1608 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v2 /*v514*/, off, s33 offset:1612 @@ -4396,7 +4396,7 @@ define amdgpu_gfx_whole_wave <2 x half> @call_gfx_from_whole_wave(i1 %active, <2 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v253 /*v765*/, off, s33 offset:2616 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v254 /*v766*/, off, s33 offset:2620 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v255 /*v767*/, off, s33 offset:2624 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0xc0 ; msbs: dst=3 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x80c0 ; msbs: dst=3 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v0 /*v768*/, off, s33 offset:2628 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v1 /*v769*/, off, s33 offset:2632 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v2 /*v770*/, off, s33 offset:2636 @@ -4661,7 +4661,7 @@ define amdgpu_gfx_whole_wave <2 x half> @call_gfx_from_whole_wave(i1 %active, <2 ; GFX1250-DAGISEL-NEXT: s_mov_b32 exec_lo, s4 ; GFX1250-DAGISEL-NEXT: s_mov_b32 s33, s0 ; GFX1250-DAGISEL-NEXT: s_wait_loadcnt 0x0 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0 ; msbs: dst=0 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0xc000 ; msbs: dst=0 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: s_set_pc_i64 s[30:31] %ret = call amdgpu_gfx <2 x half>(<2 x half>, <2 x half>) @gfx_callee(<2 x half> %y, <2 x half> %x) convergent ret <2 x half> %ret @@ -6346,7 +6346,7 @@ define amdgpu_gfx_whole_wave <2 x half> @tail_call_gfx_from_whole_wave(i1 %activ ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v253 /*v509*/, s32 offset:1588 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v254 /*v510*/, s32 offset:1592 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v255 /*v511*/, s32 offset:1596 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 8 ; msbs: dst=0 src0=0 src1=2 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x408 ; msbs: dst=0 src0=0 src1=2 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v0 /*v512*/, s32 offset:1600 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v1 /*v513*/, s32 offset:1604 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v2 /*v514*/, s32 offset:1608 @@ -6607,7 +6607,7 @@ define amdgpu_gfx_whole_wave <2 x half> @tail_call_gfx_from_whole_wave(i1 %activ ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v253 /*v765*/, s32 offset:2612 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v254 /*v766*/, s32 offset:2616 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v255 /*v767*/, s32 offset:2620 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 12 ; msbs: dst=0 src0=0 src1=3 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x80c ; msbs: dst=0 src0=0 src1=3 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v0 /*v768*/, s32 offset:2624 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v1 /*v769*/, s32 offset:2628 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v2 /*v770*/, s32 offset:2632 @@ -6872,7 +6872,7 @@ define amdgpu_gfx_whole_wave <2 x half> @tail_call_gfx_from_whole_wave(i1 %activ ; GFX1250-DAGISEL-NEXT: s_mov_b32 exec_lo, -1 ; GFX1250-DAGISEL-NEXT: v_mov_b32_e32 v2, v0 ; GFX1250-DAGISEL-NEXT: s_mov_b64 s[36:37], gfx_callee@abs64 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0 ; msbs: dst=0 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0xc00 ; msbs: dst=0 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: v_swap_b32 v0, v1 ; GFX1250-DAGISEL-NEXT: s_xor_b32 exec_lo, s0, -1 ; GFX1250-DAGISEL-NEXT: s_clause 0x3e @@ -7283,7 +7283,7 @@ define amdgpu_gfx_whole_wave <2 x half> @tail_call_gfx_from_whole_wave(i1 %activ ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v253 /*v509*/, off, s32 offset:1588 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v254 /*v510*/, off, s32 offset:1592 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v255 /*v511*/, off, s32 offset:1596 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x80 ; msbs: dst=2 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x4080 ; msbs: dst=2 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v0 /*v512*/, off, s32 offset:1600 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v1 /*v513*/, off, s32 offset:1604 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v2 /*v514*/, off, s32 offset:1608 @@ -7544,7 +7544,7 @@ define amdgpu_gfx_whole_wave <2 x half> @tail_call_gfx_from_whole_wave(i1 %activ ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v253 /*v765*/, off, s32 offset:2612 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v254 /*v766*/, off, s32 offset:2616 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v255 /*v767*/, off, s32 offset:2620 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0xc0 ; msbs: dst=3 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x80c0 ; msbs: dst=3 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v0 /*v768*/, off, s32 offset:2624 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v1 /*v769*/, off, s32 offset:2628 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v2 /*v770*/, off, s32 offset:2632 @@ -7807,7 +7807,7 @@ define amdgpu_gfx_whole_wave <2 x half> @tail_call_gfx_from_whole_wave(i1 %activ ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v255 /*v1023*/, off, s32 offset:3644 ; GFX1250-DAGISEL-NEXT: s_wait_xcnt 0x0 ; GFX1250-DAGISEL-NEXT: s_mov_b32 exec_lo, s0 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0 ; msbs: dst=0 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0xc000 ; msbs: dst=0 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: s_set_pc_i64 s[36:37] %ret = tail call amdgpu_gfx <2 x half>(<2 x half>, <2 x half>) @gfx_callee(<2 x half> %y, <2 x half> %x) convergent ret <2 x half> %ret @@ -9657,7 +9657,7 @@ define amdgpu_gfx_whole_wave void @call_from_whole_wave(i1 %unused, <8 x float> ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v253 /*v509*/, s33 offset:1600 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v254 /*v510*/, s33 offset:1604 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v255 /*v511*/, s33 offset:1608 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 8 ; msbs: dst=0 src0=0 src1=2 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x408 ; msbs: dst=0 src0=0 src1=2 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v0 /*v512*/, s33 offset:1612 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v1 /*v513*/, s33 offset:1616 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v2 /*v514*/, s33 offset:1620 @@ -9918,7 +9918,7 @@ define amdgpu_gfx_whole_wave void @call_from_whole_wave(i1 %unused, <8 x float> ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v253 /*v765*/, s33 offset:2624 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v254 /*v766*/, s33 offset:2628 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v255 /*v767*/, s33 offset:2632 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 12 ; msbs: dst=0 src0=0 src1=3 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x80c ; msbs: dst=0 src0=0 src1=3 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v0 /*v768*/, s33 offset:2636 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v1 /*v769*/, s33 offset:2640 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v2 /*v770*/, s33 offset:2644 @@ -10181,7 +10181,7 @@ define amdgpu_gfx_whole_wave void @call_from_whole_wave(i1 %unused, <8 x float> ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v255 /*v1023*/, s33 offset:3656 ; GFX1250-DAGISEL-NEXT: s_wait_xcnt 0x0 ; GFX1250-DAGISEL-NEXT: s_mov_b32 exec_lo, -1 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0 ; msbs: dst=0 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0xc00 ; msbs: dst=0 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: s_clause 0x2 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v42, s33 ; GFX1250-DAGISEL-NEXT: scratch_store_b32 off, v40, s33 offset:164 @@ -10616,7 +10616,7 @@ define amdgpu_gfx_whole_wave void @call_from_whole_wave(i1 %unused, <8 x float> ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v253 /*v509*/, off, s33 offset:1600 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v254 /*v510*/, off, s33 offset:1604 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v255 /*v511*/, off, s33 offset:1608 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x80 ; msbs: dst=2 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x4080 ; msbs: dst=2 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v0 /*v512*/, off, s33 offset:1612 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v1 /*v513*/, off, s33 offset:1616 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v2 /*v514*/, off, s33 offset:1620 @@ -10877,7 +10877,7 @@ define amdgpu_gfx_whole_wave void @call_from_whole_wave(i1 %unused, <8 x float> ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v253 /*v765*/, off, s33 offset:2624 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v254 /*v766*/, off, s33 offset:2628 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v255 /*v767*/, off, s33 offset:2632 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0xc0 ; msbs: dst=3 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0x80c0 ; msbs: dst=3 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v0 /*v768*/, off, s33 offset:2636 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v1 /*v769*/, off, s33 offset:2640 ; GFX1250-DAGISEL-NEXT: scratch_load_b32 v2 /*v770*/, off, s33 offset:2644 @@ -11142,7 +11142,7 @@ define amdgpu_gfx_whole_wave void @call_from_whole_wave(i1 %unused, <8 x float> ; GFX1250-DAGISEL-NEXT: s_mov_b32 exec_lo, s4 ; GFX1250-DAGISEL-NEXT: s_mov_b32 s33, s0 ; GFX1250-DAGISEL-NEXT: s_wait_loadcnt_dscnt 0x0 -; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0 ; msbs: dst=0 src0=0 src1=0 src2=0 +; GFX1250-DAGISEL-NEXT: s_set_vgpr_msb 0xc000 ; msbs: dst=0 src0=0 src1=0 src2=0 ; GFX1250-DAGISEL-NEXT: s_set_pc_i64 s[30:31] %ret = call float(ptr, ...) @llvm.amdgcn.call.whole.wave(ptr @callee, <8 x float> %x) convergent store float %ret, ptr %p diff --git a/llvm/test/CodeGen/AMDGPU/wqm.ll b/llvm/test/CodeGen/AMDGPU/wqm.ll index ad8dcd3..21f0c00 100644 --- a/llvm/test/CodeGen/AMDGPU/wqm.ll +++ b/llvm/test/CodeGen/AMDGPU/wqm.ll @@ -3477,13 +3477,10 @@ define amdgpu_gs void @wqm_init_exec_wwm() { ; GFX9-W64-NEXT: s_mov_b64 exec, 0 ; GFX9-W64-NEXT: s_mov_b32 s1, 0 ; GFX9-W64-NEXT: s_mov_b32 s0, s1 -; GFX9-W64-NEXT: s_cmp_lg_u64 exec, 0 -; GFX9-W64-NEXT: s_cselect_b64 s[2:3], -1, 0 -; GFX9-W64-NEXT: s_cmp_lg_u64 s[0:1], 0 +; GFX9-W64-NEXT: s_cmp_eq_u64 s[0:1], 0 ; GFX9-W64-NEXT: s_cselect_b64 s[0:1], -1, 0 -; GFX9-W64-NEXT: s_xor_b64 s[0:1], s[2:3], s[0:1] -; GFX9-W64-NEXT: v_cndmask_b32_e64 v0, 0, 1.0, s[0:1] -; GFX9-W64-NEXT: v_mov_b32_e32 v1, 0 +; GFX9-W64-NEXT: v_mov_b32_e32 v0, 0 +; GFX9-W64-NEXT: v_cndmask_b32_e64 v1, 0, 1.0, s[0:1] ; GFX9-W64-NEXT: exp mrt0 off, off, off, off ; GFX9-W64-NEXT: s_endpgm ; @@ -3491,14 +3488,11 @@ define amdgpu_gs void @wqm_init_exec_wwm() { ; GFX10-W32: ; %bb.0: ; GFX10-W32-NEXT: s_mov_b32 exec_lo, 0 ; GFX10-W32-NEXT: s_mov_b32 s1, 0 -; GFX10-W32-NEXT: s_cmp_lg_u64 exec, 0 +; GFX10-W32-NEXT: v_mov_b32_e32 v0, 0 ; GFX10-W32-NEXT: s_mov_b32 s0, s1 -; GFX10-W32-NEXT: s_cselect_b32 s2, -1, 0 -; GFX10-W32-NEXT: s_cmp_lg_u64 s[0:1], 0 -; GFX10-W32-NEXT: v_mov_b32_e32 v1, 0 +; GFX10-W32-NEXT: s_cmp_eq_u64 s[0:1], 0 ; GFX10-W32-NEXT: s_cselect_b32 s0, -1, 0 -; GFX10-W32-NEXT: s_xor_b32 s0, s2, s0 -; GFX10-W32-NEXT: v_cndmask_b32_e64 v0, 0, 1.0, s0 +; GFX10-W32-NEXT: v_cndmask_b32_e64 v1, 0, 1.0, s0 ; GFX10-W32-NEXT: exp mrt0 off, off, off, off ; GFX10-W32-NEXT: s_endpgm call void @llvm.amdgcn.init.exec(i64 0) diff --git a/llvm/test/CodeGen/ARM/fp-intrinsics.ll b/llvm/test/CodeGen/ARM/fp-intrinsics.ll index 93b6a58..cb87508 100644 --- a/llvm/test/CodeGen/ARM/fp-intrinsics.ll +++ b/llvm/test/CodeGen/ARM/fp-intrinsics.ll @@ -76,7 +76,6 @@ define i32 @fptosi_f32(float %x) #0 { ; CHECK-NOSP: bl __aeabi_f2iz ; CHECK-NOSP: bl __aeabi_f2iz ; CHECK-SP: vcvt.s32.f32 -; FIXME-CHECK-SP: vcvt.s32.f32 define void @fptosi_f32_twice(float %arg, ptr %ptr) #0 { entry: %conv = call i32 @llvm.experimental.constrained.fptosi.i32.f32(float %arg, metadata !"fpexcept.strict") #0 @@ -146,6 +145,80 @@ define float @tan_f32(float %x) #0 { ret float %val } +; CHECK-LABEL: acos_f32: +; CHECK: bl acosf +define float @acos_f32(float %x, float %y) #0 { + %val = call float @llvm.experimental.constrained.acos.f32(float %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret float %val +} + +; CHECK-LABEL: asin_f32: +; CHECK: bl asinf +define float @asin_f32(float %x, float %y) #0 { + %val = call float @llvm.experimental.constrained.asin.f32(float %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret float %val +} + +; CHECK-LABEL: atan_f32: +; CHECK: bl atanf +define float @atan_f32(float %x, float %y) #0 { + %val = call float @llvm.experimental.constrained.atan.f32(float %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret float %val +} + +; CHECK-LABEL: cosh_f32: +; CHECK: bl coshf +define float @cosh_f32(float %x, float %y) #0 { + %val = call float @llvm.experimental.constrained.cosh.f32(float %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret float %val +} + +; CHECK-LABEL: sinh_f32: +; CHECK: bl sinhf +define float @sinh_f32(float %x, float %y) #0 { + %val = call float @llvm.experimental.constrained.sinh.f32(float %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret float %val +} + +; CHECK-LABEL: tanh_f32: +; CHECK: bl tanhf +define float @tanh_f32(float %x, float %y) #0 { + %val = call float @llvm.experimental.constrained.tanh.f32(float %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret float %val +} + +; CHECK-LABEL: fmuladd_f32: +; CHECK-SP: vfma.f32 +; CHECK-NOSP: bl __aeabi_fmul +; CHECK-NOSP: bl __aeabi_fadd +define float @fmuladd_f32(float %x, float %y, float %z) #0 { + %val = call float @llvm.experimental.constrained.fmuladd.f32(float %x, float %y, float %z, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret float %val +} + +; CHECK-LABEL: ldexp_f32: +; CHECK: bl ldexpf +define float @ldexp_f32(float %x, i32 %y) #0 { + %val = call float @llvm.experimental.constrained.ldexp.f32.i32(float %x, i32 %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret float %val +} + +; CHECK-LABEL: roundeven_f32: +; CHECK-SP-V8: vrintn.f32 +; CHECK-NOSP: bl roundevenf +define float @roundeven_f32(float %x) #0 { + %val = call float @llvm.experimental.constrained.roundeven.f32(float %x, metadata !"fpexcept.strict") #0 + ret float %val +} + +; CHECK-LABEL: uitofp_f32_i32: +; CHECK-NOSP: bl __aeabi_ui2f +; FIXME-CHECK-SP: vcvt.f32.f64 +define float @uitofp_f32_i32(i32 %x) #0 { + %val = call float @llvm.experimental.constrained.uitofp.f32.i32(i32 %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret float %val +} + ; CHECK-LABEL: atan2_f32: ; CHECK: bl atan2f define float @atan2_f32(float %x, float %y) #0 { @@ -617,6 +690,80 @@ define double @tan_f64(double %x) #0 { ret double %val } +; CHECK-LABEL: acos_f64: +; CHECK: bl acos +define double @acos_f64(double %x, double %y) #0 { + %val = call double @llvm.experimental.constrained.acos.f64(double %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret double %val +} + +; CHECK-LABEL: asin_f64: +; CHECK: bl asin +define double @asin_f64(double %x, double %y) #0 { + %val = call double @llvm.experimental.constrained.asin.f64(double %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret double %val +} + +; CHECK-LABEL: atan_f64: +; CHECK: bl atan +define double @atan_f64(double %x, double %y) #0 { + %val = call double @llvm.experimental.constrained.atan.f64(double %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret double %val +} + +; CHECK-LABEL: cosh_f64: +; CHECK: bl cosh +define double @cosh_f64(double %x, double %y) #0 { + %val = call double @llvm.experimental.constrained.cosh.f64(double %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret double %val +} + +; CHECK-LABEL: sinh_f64: +; CHECK: bl sinh +define double @sinh_f64(double %x, double %y) #0 { + %val = call double @llvm.experimental.constrained.sinh.f64(double %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret double %val +} + +; CHECK-LABEL: tanh_f64: +; CHECK: bl tanh +define double @tanh_f64(double %x, double %y) #0 { + %val = call double @llvm.experimental.constrained.tanh.f64(double %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret double %val +} + +; CHECK-LABEL: fmuladd_f64: +; CHECK-DP: vfma.f64 +; CHECK-NODP: bl __aeabi_dmul +; CHECK-NODP: bl __aeabi_dadd +define double @fmuladd_f64(double %x, double %y, double %z) #0 { + %val = call double @llvm.experimental.constrained.fmuladd.f64(double %x, double %y, double %z, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret double %val +} + +; CHECK-LABEL: ldexp_f64: +; CHECK: bl ldexp +define double @ldexp_f64(double %x, i32 %y) #0 { + %val = call double @llvm.experimental.constrained.ldexp.f64.i32(double %x, i32 %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret double %val +} + +; CHECK-LABEL: roundeven_f64: +; CHECK-DP-V8: vrintn.f64 +; CHECK-NODP: bl roundeven +define double @roundeven_f64(double %x) #0 { + %val = call double @llvm.experimental.constrained.roundeven.f64(double %x, metadata !"fpexcept.strict") #0 + ret double %val +} + +; CHECK-LABEL: uitofp_f64_i32: +; CHECK-NOSP: bl __aeabi_ui2d +; FIXME-CHECK-SP: vsub.f64 +define double @uitofp_f64_i32(i32 %x) #0 { + %val = call double @llvm.experimental.constrained.uitofp.f64.i32(i32 %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret double %val +} + ; CHECK-LABEL: atan2_f64: ; CHECK: bl atan2 define double @atan2_f64(double %x, double %y) #0 { @@ -1052,6 +1199,16 @@ declare float @llvm.experimental.constrained.powi.f32(float, i32, metadata, meta declare float @llvm.experimental.constrained.sin.f32(float, metadata, metadata) declare float @llvm.experimental.constrained.cos.f32(float, metadata, metadata) declare float @llvm.experimental.constrained.tan.f32(float, metadata, metadata) +declare float @llvm.experimental.constrained.acos.f32(float, metadata, metadata) +declare float @llvm.experimental.constrained.asin.f32(float, metadata, metadata) +declare float @llvm.experimental.constrained.atan.f32(float, metadata, metadata) +declare float @llvm.experimental.constrained.cosh.f32(float, metadata, metadata) +declare float @llvm.experimental.constrained.sinh.f32(float, metadata, metadata) +declare float @llvm.experimental.constrained.tanh.f32(float, metadata, metadata) +declare float @llvm.experimental.constrained.fmuladd.f32(float, float, float, metadata, metadata) +declare float @llvm.experimental.constrained.ldexp.f32.i32(float, i32, metadata, metadata) +declare float @llvm.experimental.constrained.roundeven.f32(float, metadata) +declare float @llvm.experimental.constrained.uitofp.f32.i32(i32, metadata, metadata) declare float @llvm.experimental.constrained.atan2.f32(float, float, metadata, metadata) declare float @llvm.experimental.constrained.pow.f32(float, float, metadata, metadata) declare float @llvm.experimental.constrained.log.f32(float, metadata, metadata) @@ -1087,6 +1244,16 @@ declare double @llvm.experimental.constrained.powi.f64(double, i32, metadata, me declare double @llvm.experimental.constrained.sin.f64(double, metadata, metadata) declare double @llvm.experimental.constrained.cos.f64(double, metadata, metadata) declare double @llvm.experimental.constrained.tan.f64(double, metadata, metadata) +declare double @llvm.experimental.constrained.acos.f64(double, metadata, metadata) +declare double @llvm.experimental.constrained.asin.f64(double, metadata, metadata) +declare double @llvm.experimental.constrained.atan.f64(double, metadata, metadata) +declare double @llvm.experimental.constrained.cosh.f64(double, metadata, metadata) +declare double @llvm.experimental.constrained.sinh.f64(double, metadata, metadata) +declare double @llvm.experimental.constrained.tanh.f64(double, metadata, metadata) +declare double @llvm.experimental.constrained.fmuladd.f64(double, double, double, metadata, metadata) +declare double @llvm.experimental.constrained.ldexp.f64.i32(double, i32, metadata, metadata) +declare double @llvm.experimental.constrained.roundeven.f64(double, metadata) +declare double @llvm.experimental.constrained.uitofp.f64.i32(i32, metadata, metadata) declare double @llvm.experimental.constrained.atan2.f64(double, double, metadata, metadata) declare double @llvm.experimental.constrained.pow.f64(double, double, metadata, metadata) declare double @llvm.experimental.constrained.log.f64(double, metadata, metadata) diff --git a/llvm/test/CodeGen/ARM/fp16-fullfp16.ll b/llvm/test/CodeGen/ARM/fp16-fullfp16.ll index 200b14b..b4060d5 100644 --- a/llvm/test/CodeGen/ARM/fp16-fullfp16.ll +++ b/llvm/test/CodeGen/ARM/fp16-fullfp16.ll @@ -98,12 +98,18 @@ define i32 @test_fptosi_i32(ptr %p) { ret i32 %r } -; FIXME -;define i64 @test_fptosi_i64(ptr %p) { -; %a = load half, ptr %p, align 2 -; %r = fptosi half %a to i64 -; ret i64 %r -;} +define i64 @test_fptosi_i64(ptr %p) { +; CHECK-LABEL: test_fptosi_i64: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: ldrh r0, [r0] +; CHECK-NEXT: vmov s0, r0 +; CHECK-NEXT: bl __fixhfdi +; CHECK-NEXT: pop {r11, pc} + %a = load half, ptr %p, align 2 + %r = fptosi half %a to i64 + ret i64 %r +} define i32 @test_fptoui_i32(ptr %p) { ; CHECK-LABEL: test_fptoui_i32: @@ -116,12 +122,18 @@ define i32 @test_fptoui_i32(ptr %p) { ret i32 %r } -; FIXME -;define i64 @test_fptoui_i64(ptr %p) { -; %a = load half, ptr %p, align 2 -; %r = fptoui half %a to i64 -; ret i64 %r -;} +define i64 @test_fptoui_i64(ptr %p) { +; CHECK-LABEL: test_fptoui_i64: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: ldrh r0, [r0] +; CHECK-NEXT: vmov s0, r0 +; CHECK-NEXT: bl __fixunshfdi +; CHECK-NEXT: pop {r11, pc} + %a = load half, ptr %p, align 2 + %r = fptoui half %a to i64 + ret i64 %r +} define void @test_sitofp_i32(i32 %a, ptr %p) { ; CHECK-LABEL: test_sitofp_i32: @@ -145,19 +157,31 @@ define void @test_uitofp_i32(i32 %a, ptr %p) { ret void } -; FIXME -;define void @test_sitofp_i64(i64 %a, ptr %p) { -; %r = sitofp i64 %a to half -; store half %r, ptr %p -; ret void -;} +define void @test_sitofp_i64(i64 %a, ptr %p) { +; CHECK-LABEL: test_sitofp_i64: +; CHECK: .save {r4, lr} +; CHECK-NEXT: push {r4, lr} +; CHECK-NEXT: mov r4, r2 +; CHECK-NEXT: bl __floatdihf +; CHECK-NEXT: vstr.16 s0, [r4] +; CHECK-NEXT: pop {r4, pc} + %r = sitofp i64 %a to half + store half %r, ptr %p + ret void +} -; FIXME -;define void @test_uitofp_i64(i64 %a, ptr %p) { -; %r = uitofp i64 %a to half -; store half %r, ptr %p -; ret void -;} +define void @test_uitofp_i64(i64 %a, ptr %p) { +; CHECK-LABEL: test_uitofp_i64: +; CHECK: .save {r4, lr} +; CHECK-NEXT: push {r4, lr} +; CHECK-NEXT: mov r4, r2 +; CHECK-NEXT: bl __floatundihf +; CHECK-NEXT: vstr.16 s0, [r4] +; CHECK-NEXT: pop {r4, pc} + %r = uitofp i64 %a to half + store half %r, ptr %p + ret void +} define void @test_fptrunc_float(float %f, ptr %p) { ; CHECK-LABEL: test_fptrunc_float: @@ -613,6 +637,902 @@ define void @test_fmuladd(ptr %p, ptr %q, ptr %r) { ret void } +; Half-precision intrinsics + +define half @add_f16(half %x, half %y) #0 { +; CHECK-LABEL: add_f16: +; CHECK: vadd.f16 s0, s0, s1 +; CHECK-NEXT: bx lr + %val = call half @llvm.experimental.constrained.fadd.f16(half %x, half %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @sub_f16(half %x, half %y) #0 { +; CHECK-LABEL: sub_f16: +; CHECK: vsub.f16 s0, s0, s1 +; CHECK-NEXT: bx lr + %val = call half @llvm.experimental.constrained.fsub.f16(half %x, half %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @mul_f16(half %x, half %y) #0 { +; CHECK-LABEL: mul_f16: +; CHECK: vmul.f16 s0, s0, s1 +; CHECK-NEXT: bx lr + %val = call half @llvm.experimental.constrained.fmul.f16(half %x, half %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @div_f16(half %x, half %y) #0 { +; CHECK-LABEL: div_f16: +; CHECK: vdiv.f16 s0, s0, s1 +; CHECK-NEXT: bx lr + %val = call half @llvm.experimental.constrained.fdiv.f16(half %x, half %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @frem_f16(half %x, half %y) #0 { +; CHECK-LABEL: frem_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: vcvtb.f32.f16 s1, s1 +; CHECK-NEXT: bl fmodf +; CHECK-NEXT: vcvtb.f16.f32 s0, s0 +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.frem.f16(half %x, half %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @fma_f16(half %x, half %y, half %z) #0 { +; CHECK-LABEL: fma_f16: +; CHECK: vfma.f16 s2, s0, s1 +; CHECK-NEXT: vmov.f32 s0, s2 +; CHECK-NEXT: bx lr + %val = call half @llvm.experimental.constrained.fma.f16(half %x, half %y, half %z, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @fmuladd_f16(half %x, half %y, half %z) #0 { +; CHECK-LABEL: fmuladd_f16: +; CHECK: vfma.f16 s2, s0, s1 +; CHECK-NEXT: vmov.f32 s0, s2 +; CHECK-NEXT: bx lr + %val = call half @llvm.experimental.constrained.fmuladd.f16(half %x, half %y, half %z, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define i32 @fptosi_i32_f16(half %x) #0 { +; CHECK-LABEL: fptosi_i32_f16: +; CHECK: vcvt.s32.f16 s0, s0 +; CHECK-NEXT: vmov r0, s0 +; CHECK-NEXT: bx lr + %val = call i32 @llvm.experimental.constrained.fptosi.i32.f16(half %x, metadata !"fpexcept.strict") #0 + ret i32 %val +} + +define i32 @fptoui_i32_f16(half %x) #0 { +; CHECK-LABEL: fptoui_i32_f16: +; CHECK: vcvt.s32.f16 s0, s0 +; CHECK-NEXT: vmov r0, s0 +; CHECK-NEXT: bx lr + %val = call i32 @llvm.experimental.constrained.fptoui.i32.f16(half %x, metadata !"fpexcept.strict") #0 + ret i32 %val +} + +define i64 @fptosi_i64_f16(half %x) #0 { +; CHECK-LABEL: fptosi_i64_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vmov.f16 r0, s0 +; CHECK-NEXT: vmov s0, r0 +; CHECK-NEXT: bl __fixhfdi +; CHECK-NEXT: pop {r11, pc} + %val = call i64 @llvm.experimental.constrained.fptosi.i64.f16(half %x, metadata !"fpexcept.strict") #0 + ret i64 %val +} + +define i64 @fptoui_i64_f16(half %x) #0 { +; CHECK-LABEL: fptoui_i64_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vmov.f16 r0, s0 +; CHECK-NEXT: vmov s0, r0 +; CHECK-NEXT: bl __fixunshfdi +; CHECK-NEXT: pop {r11, pc} + %val = call i64 @llvm.experimental.constrained.fptoui.i64.f16(half %x, metadata !"fpexcept.strict") #0 + ret i64 %val +} + +define half @sitofp_f16_i32(i32 %x) #0 { +; CHECK-LABEL: sitofp_f16_i32: +; CHECK: .pad #8 +; CHECK-NEXT: sub sp, sp, #8 +; CHECK-NEXT: movw r1, #0 +; CHECK-NEXT: eor r0, r0, #-2147483648 +; CHECK-NEXT: movt r1, #17200 +; CHECK-NEXT: str r0, [sp] +; CHECK-NEXT: str r1, [sp, #4] +; CHECK-NEXT: vldr d16, .LCPI57_0 +; CHECK-NEXT: vldr d17, [sp] +; CHECK-NEXT: vsub.f64 d16, d17, d16 +; CHECK-NEXT: vcvtb.f16.f64 s0, d16 +; CHECK-NEXT: add sp, sp, #8 +; CHECK-NEXT: bx lr +; CHECK-NEXT: .p2align 3 +; CHECK-NEXT: .LCPI57_0: +; CHECK-NEXT: .long 2147483648 +; CHECK-NEXT: .long 1127219200 + %val = call half @llvm.experimental.constrained.sitofp.f16.i32(i32 %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @uitofp_f16_i32(i32 %x) #0 { +; CHECK-LABEL: uitofp_f16_i32: +; CHECK: .pad #8 +; CHECK-NEXT: sub sp, sp, #8 +; CHECK-NEXT: movw r1, #0 +; CHECK-NEXT: str r0, [sp] +; CHECK-NEXT: movt r1, #17200 +; CHECK-NEXT: vldr d16, .LCPI58_0 +; CHECK-NEXT: str r1, [sp, #4] +; CHECK-NEXT: vldr d17, [sp] +; CHECK-NEXT: vsub.f64 d16, d17, d16 +; CHECK-NEXT: vcvtb.f16.f64 s0, d16 +; CHECK-NEXT: add sp, sp, #8 +; CHECK-NEXT: bx lr +; CHECK-NEXT: .p2align 3 +; CHECK-NEXT: .LCPI58_0: +; CHECK-NEXT: .long 0 +; CHECK-NEXT: .long 1127219200 + %val = call half @llvm.experimental.constrained.uitofp.f16.i32(i32 %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @sitofp_f16_i64(i64 %x) #0 { +; CHECK-LABEL: sitofp_f16_i64: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: bl __floatdihf +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.sitofp.f16.i64(i64 %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @uitofp_f16_i64(i64 %x) #0 { +; CHECK-LABEL: uitofp_f16_i64: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: bl __floatundihf +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.uitofp.f16.i64(i64 %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @sitofp_f16_i128(i128 %x) #0 { +; CHECK-LABEL: sitofp_f16_i128: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: bl __floattihf +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.sitofp.f16.i128(i128 %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @uitofp_f16_i128(i128 %x) #0 { +; CHECK-LABEL: uitofp_f16_i128: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: bl __floatuntihf +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.uitofp.f16.i128(i128 %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @sqrt_f16(half %x) #0 { +; CHECK-LABEL: sqrt_f16: +; CHECK: vsqrt.f16 s0, s0 +; CHECK-NEXT: bx lr + %val = call half @llvm.experimental.constrained.sqrt.f16(half %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @powi_f16(half %x, i32 %y) #0 { +; CHECK-LABEL: powi_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bl __powisf2 +; CHECK-NEXT: vcvtb.f16.f32 s0, s0 +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.powi.f16(half %x, i32 %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @sin_f16(half %x) #0 { +; CHECK-LABEL: sin_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bl sinf +; CHECK-NEXT: vcvtb.f16.f32 s0, s0 +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.sin.f16(half %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @cos_f16(half %x) #0 { +; CHECK-LABEL: cos_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bl cosf +; CHECK-NEXT: vcvtb.f16.f32 s0, s0 +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.cos.f16(half %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @tan_f16(half %x) #0 { +; CHECK-LABEL: tan_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bl tanf +; CHECK-NEXT: vcvtb.f16.f32 s0, s0 +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.tan.f16(half %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @asin_f16(half %x) #0 { +; CHECK-LABEL: asin_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bl asinf +; CHECK-NEXT: vcvtb.f16.f32 s0, s0 +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.asin.f16(half %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @acos_f16(half %x) #0 { +; CHECK-LABEL: acos_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bl acosf +; CHECK-NEXT: vcvtb.f16.f32 s0, s0 +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.acos.f16(half %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @atan_f16(half %x) #0 { +; CHECK-LABEL: atan_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bl atanf +; CHECK-NEXT: vcvtb.f16.f32 s0, s0 +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.atan.f16(half %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @atan2_f16(half %x, half %y) #0 { +; CHECK-LABEL: atan2_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: vcvtb.f32.f16 s1, s1 +; CHECK-NEXT: bl atan2f +; CHECK-NEXT: vcvtb.f16.f32 s0, s0 +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.atan2.f16(half %x, half %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @sinh_f16(half %x) #0 { +; CHECK-LABEL: sinh_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bl sinhf +; CHECK-NEXT: vcvtb.f16.f32 s0, s0 +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.sinh.f16(half %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @cosh_f16(half %x) #0 { +; CHECK-LABEL: cosh_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bl coshf +; CHECK-NEXT: vcvtb.f16.f32 s0, s0 +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.cosh.f16(half %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @tanh_f16(half %x) #0 { +; CHECK-LABEL: tanh_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bl tanhf +; CHECK-NEXT: vcvtb.f16.f32 s0, s0 +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.tanh.f16(half %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @pow_f16(half %x, half %y) #0 { +; CHECK-LABEL: pow_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: vcvtb.f32.f16 s1, s1 +; CHECK-NEXT: bl powf +; CHECK-NEXT: vcvtb.f16.f32 s0, s0 +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.pow.f16(half %x, half %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @log_f16(half %x) #0 { +; CHECK-LABEL: log_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bl logf +; CHECK-NEXT: vcvtb.f16.f32 s0, s0 +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.log.f16(half %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @log10_f16(half %x) #0 { +; CHECK-LABEL: log10_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bl log10f +; CHECK-NEXT: vcvtb.f16.f32 s0, s0 +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.log10.f16(half %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @log2_f16(half %x) #0 { +; CHECK-LABEL: log2_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bl log2f +; CHECK-NEXT: vcvtb.f16.f32 s0, s0 +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.log2.f16(half %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @exp_f16(half %x) #0 { +; CHECK-LABEL: exp_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bl expf +; CHECK-NEXT: vcvtb.f16.f32 s0, s0 +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.exp.f16(half %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @exp2_f16(half %x) #0 { +; CHECK-LABEL: exp2_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bl exp2f +; CHECK-NEXT: vcvtb.f16.f32 s0, s0 +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.exp2.f16(half %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @rint_f16(half %x) #0 { +; CHECK-LABEL: rint_f16: +; CHECK: vrintx.f16 s0, s0 +; CHECK-NEXT: bx lr + %val = call half @llvm.experimental.constrained.rint.f16(half %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @nearbyint_f16(half %x) #0 { +; CHECK-LABEL: nearbyint_f16: +; CHECK: vrintr.f16 s0, s0 +; CHECK-NEXT: bx lr + %val = call half @llvm.experimental.constrained.nearbyint.f16(half %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define i32 @lrint_f16(half %x) #0 { +; CHECK-LABEL: lrint_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bl lrintf +; CHECK-NEXT: pop {r11, pc} + %val = call i32 @llvm.experimental.constrained.lrint.i32.f16(half %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret i32 %val +} + +define i64 @llrint_f16(half %x) #0 { +; CHECK-LABEL: llrint_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bl llrintf +; CHECK-NEXT: pop {r11, pc} + %val = call i64 @llvm.experimental.constrained.llrint.i64.f16(half %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret i64 %val +} + +define half @maxnum_f16(half %x, half %y) #0 { +; CHECK-LABEL: maxnum_f16: +; CHECK: vmaxnm.f16 s0, s0, s1 +; CHECK-NEXT: bx lr + %val = call half @llvm.experimental.constrained.maxnum.f16(half %x, half %y, metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @minnum_f16(half %x, half %y) #0 { +; CHECK-LABEL: minnum_f16: +; CHECK: vminnm.f16 s0, s0, s1 +; CHECK-NEXT: bx lr + %val = call half @llvm.experimental.constrained.minnum.f16(half %x, half %y, metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @ceil_f16(half %x) #0 { +; CHECK-LABEL: ceil_f16: +; CHECK: vrintp.f16 s0, s0 +; CHECK-NEXT: bx lr + %val = call half @llvm.experimental.constrained.ceil.f16(half %x, metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @floor_f16(half %x) #0 { +; CHECK-LABEL: floor_f16: +; CHECK: vrintm.f16 s0, s0 +; CHECK-NEXT: bx lr + %val = call half @llvm.experimental.constrained.floor.f16(half %x, metadata !"fpexcept.strict") #0 + ret half %val +} + +define i32 @lround_f16(half %x) #0 { +; CHECK-LABEL: lround_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bl lroundf +; CHECK-NEXT: pop {r11, pc} + %val = call i32 @llvm.experimental.constrained.lround.i32.f16(half %x, metadata !"fpexcept.strict") #0 + ret i32 %val +} + +define i64 @llround_f16(half %x) #0 { +; CHECK-LABEL: llround_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bl llroundf +; CHECK-NEXT: pop {r11, pc} + %val = call i64 @llvm.experimental.constrained.llround.i64.f16(half %x, metadata !"fpexcept.strict") #0 + ret i64 %val +} + +define half @round_f16(half %x) #0 { +; CHECK-LABEL: round_f16: +; CHECK: vrinta.f16 s0, s0 +; CHECK-NEXT: bx lr + %val = call half @llvm.experimental.constrained.round.f16(half %x, metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @roundeven_f16(half %x) #0 { +; CHECK-LABEL: roundeven_f16: +; CHECK: vrintn.f16 s0, s0 +; CHECK-NEXT: bx lr + %val = call half @llvm.experimental.constrained.roundeven.f16(half %x, metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @trunc_f16(half %x) #0 { +; CHECK-LABEL: trunc_f16: +; CHECK: vrintz.f16 s0, s0 +; CHECK-NEXT: bx lr + %val = call half @llvm.experimental.constrained.trunc.f16(half %x, metadata !"fpexcept.strict") #0 + ret half %val +} + +define half @ldexp_f16(half %x, i32 %y) #0 { +; CHECK-LABEL: ldexp_f16: +; CHECK: .save {r11, lr} +; CHECK-NEXT: push {r11, lr} +; CHECK-NEXT: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bl ldexpf +; CHECK-NEXT: vcvtb.f16.f32 s0, s0 +; CHECK-NEXT: pop {r11, pc} + %val = call half @llvm.experimental.constrained.ldexp.f16.i32(half %x, i32 %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define i32 @fcmp_olt_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmp_olt_f16: +; CHECK: vcmp.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movwmi r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmp.f16(half %a, half %b, metadata !"olt", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmp_ole_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmp_ole_f16: +; CHECK: vcmp.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movwls r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmp.f16(half %a, half %b, metadata !"ole", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmp_ogt_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmp_ogt_f16: +; CHECK: vcmp.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movwgt r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmp.f16(half %a, half %b, metadata !"ogt", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmp_oge_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmp_oge_f16: +; CHECK: vcmp.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movwge r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmp.f16(half %a, half %b, metadata !"oge", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmp_oeq_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmp_oeq_f16: +; CHECK: vcmp.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movweq r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmp.f16(half %a, half %b, metadata !"oeq", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmp_one_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmp_one_f16: +; CHECK: vcmp.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movwmi r0, #1 +; CHECK-NEXT: movwgt r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmp.f16(half %a, half %b, metadata !"one", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmp_ult_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmp_ult_f16: +; CHECK: vcmp.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movwlt r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmp.f16(half %a, half %b, metadata !"ult", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmp_ule_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmp_ule_f16: +; CHECK: vcmp.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movwle r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmp.f16(half %a, half %b, metadata !"ule", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmp_ugt_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmp_ugt_f16: +; CHECK: vcmp.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movwhi r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmp.f16(half %a, half %b, metadata !"ugt", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmp_uge_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmp_uge_f16: +; CHECK: vcmp.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movwpl r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmp.f16(half %a, half %b, metadata !"uge", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmp_ueq_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmp_ueq_f16: +; CHECK: vcmp.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movweq r0, #1 +; CHECK-NEXT: movwvs r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmp.f16(half %a, half %b, metadata !"ueq", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmp_une_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmp_une_f16: +; CHECK: vcmp.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movwne r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmp.f16(half %a, half %b, metadata !"une", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmps_olt_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmps_olt_f16: +; CHECK: vcmpe.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movwmi r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmps.f16(half %a, half %b, metadata !"olt", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmps_ole_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmps_ole_f16: +; CHECK: vcmpe.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movwls r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmps.f16(half %a, half %b, metadata !"ole", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmps_ogt_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmps_ogt_f16: +; CHECK: vcmpe.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movwgt r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmps.f16(half %a, half %b, metadata !"ogt", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmps_oge_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmps_oge_f16: +; CHECK: vcmpe.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movwge r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmps.f16(half %a, half %b, metadata !"oge", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmps_oeq_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmps_oeq_f16: +; CHECK: vcmpe.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movweq r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmps.f16(half %a, half %b, metadata !"oeq", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmps_one_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmps_one_f16: +; CHECK: vcmpe.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movwmi r0, #1 +; CHECK-NEXT: movwgt r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmps.f16(half %a, half %b, metadata !"one", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmps_ult_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmps_ult_f16: +; CHECK: vcmpe.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movwlt r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmps.f16(half %a, half %b, metadata !"ult", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmps_ule_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmps_ule_f16: +; CHECK: vcmpe.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movwle r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmps.f16(half %a, half %b, metadata !"ule", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmps_ugt_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmps_ugt_f16: +; CHECK: vcmpe.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movwhi r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmps.f16(half %a, half %b, metadata !"ugt", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmps_uge_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmps_uge_f16: +; CHECK: vcmpe.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movwpl r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmps.f16(half %a, half %b, metadata !"uge", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmps_ueq_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmps_ueq_f16: +; CHECK: vcmpe.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movweq r0, #1 +; CHECK-NEXT: movwvs r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmps.f16(half %a, half %b, metadata !"ueq", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + +define i32 @fcmps_une_f16(half %a, half %b) #0 { +; CHECK-LABEL: fcmps_une_f16: +; CHECK: vcmpe.f16 s0, s1 +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: movwne r0, #1 +; CHECK-NEXT: bx lr + %cmp = call i1 @llvm.experimental.constrained.fcmps.f16(half %a, half %b, metadata !"une", metadata !"fpexcept.strict") #0 + %conv = zext i1 %cmp to i32 + ret i32 %conv +} + + +; Intrinsics to convert between floating-point types + +define half @fptrunc_f16_f32(float %x) #0 { +; CHECK-LABEL: fptrunc_f16_f32: +; CHECK: vcvtb.f16.f32 s0, s0 +; CHECK-NEXT: bx lr + %val = call half @llvm.experimental.constrained.fptrunc.f16.f32(float %x, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret half %val +} + +define float @fpext_f32_f16(half %x) #0 { +; CHECK-LABEL: fpext_f32_f16: +; CHECK: vcvtb.f32.f16 s0, s0 +; CHECK-NEXT: bx lr + %val = call float @llvm.experimental.constrained.fpext.f32.f16(half %x, metadata !"fpexcept.strict") #0 + ret float %val +} + + +attributes #0 = { strictfp } + +declare half @llvm.experimental.constrained.fadd.f16(half, half, metadata, metadata) +declare half @llvm.experimental.constrained.fsub.f16(half, half, metadata, metadata) +declare half @llvm.experimental.constrained.fmul.f16(half, half, metadata, metadata) +declare half @llvm.experimental.constrained.fdiv.f16(half, half, metadata, metadata) +declare half @llvm.experimental.constrained.frem.f16(half, half, metadata, metadata) +declare half @llvm.experimental.constrained.fma.f16(half, half, half, metadata, metadata) +declare half @llvm.experimental.constrained.fmuladd.f16(half, half, half, metadata, metadata) +declare i32 @llvm.experimental.constrained.fptosi.i32.f16(half, metadata) +declare i32 @llvm.experimental.constrained.fptoui.i32.f16(half, metadata) +declare i64 @llvm.experimental.constrained.fptosi.i64.f16(half, metadata) +declare i64 @llvm.experimental.constrained.fptoui.i64.f16(half, metadata) +declare half @llvm.experimental.constrained.sitofp.f16.i32(i32, metadata, metadata) +declare half @llvm.experimental.constrained.uitofp.f16.i32(i32, metadata, metadata) +declare half @llvm.experimental.constrained.sitofp.f16.i64(i64, metadata, metadata) +declare half @llvm.experimental.constrained.uitofp.f16.i64(i64, metadata, metadata) +declare half @llvm.experimental.constrained.sitofp.f16.i128(i128, metadata, metadata) +declare half @llvm.experimental.constrained.uitofp.f16.i128(i128, metadata, metadata) +declare half @llvm.experimental.constrained.sqrt.f16(half, metadata, metadata) +declare half @llvm.experimental.constrained.powi.f16(half, i32, metadata, metadata) +declare half @llvm.experimental.constrained.sin.f16(half, metadata, metadata) +declare half @llvm.experimental.constrained.cos.f16(half, metadata, metadata) +declare half @llvm.experimental.constrained.tan.f16(half, metadata, metadata) +declare half @llvm.experimental.constrained.pow.f16(half, half, metadata, metadata) +declare half @llvm.experimental.constrained.log.f16(half, metadata, metadata) +declare half @llvm.experimental.constrained.log10.f16(half, metadata, metadata) +declare half @llvm.experimental.constrained.log2.f16(half, metadata, metadata) +declare half @llvm.experimental.constrained.exp.f16(half, metadata, metadata) +declare half @llvm.experimental.constrained.exp2.f16(half, metadata, metadata) +declare half @llvm.experimental.constrained.rint.f16(half, metadata, metadata) +declare half @llvm.experimental.constrained.nearbyint.f16(half, metadata, metadata) +declare i32 @llvm.experimental.constrained.lrint.i32.f16(half, metadata, metadata) +declare i64 @llvm.experimental.constrained.llrint.i64.f16(half, metadata, metadata) +declare half @llvm.experimental.constrained.maxnum.f16(half, half, metadata) +declare half @llvm.experimental.constrained.minnum.f16(half, half, metadata) +declare half @llvm.experimental.constrained.ceil.f16(half, metadata) +declare half @llvm.experimental.constrained.floor.f16(half, metadata) +declare i32 @llvm.experimental.constrained.lround.i32.f16(half, metadata) +declare i64 @llvm.experimental.constrained.llround.i64.f16(half, metadata) +declare half @llvm.experimental.constrained.round.f16(half, metadata) +declare half @llvm.experimental.constrained.roundeven.f16(half, metadata) +declare half @llvm.experimental.constrained.trunc.f16(half, metadata) +declare i1 @llvm.experimental.constrained.fcmps.f16(half, half, metadata, metadata) +declare i1 @llvm.experimental.constrained.fcmp.f16(half, half, metadata, metadata) + +declare half @llvm.experimental.constrained.fptrunc.f16.f32(float, metadata, metadata) +declare float @llvm.experimental.constrained.fpext.f32.f16(half, metadata) + + declare half @llvm.sqrt.f16(half %a) declare half @llvm.powi.f16.i32(half %a, i32 %b) declare half @llvm.sin.f16(half %a) diff --git a/llvm/test/CodeGen/ARM/strict-fp-func.ll b/llvm/test/CodeGen/ARM/strict-fp-func.ll new file mode 100644 index 0000000..39bb2b4 --- /dev/null +++ b/llvm/test/CodeGen/ARM/strict-fp-func.ll @@ -0,0 +1,13 @@ +; RUN: llc -mtriple arm-none-eabi -stop-after=finalize-isel %s -o - | FileCheck %s + +define float @func_02(float %x, float %y) strictfp nounwind { + %call = call float @func_01(float %x) strictfp + %res = call float @llvm.experimental.constrained.fadd.f32(float %call, float %y, metadata !"round.dynamic", metadata !"fpexcept.ignore") strictfp + ret float %res +} +; CHECK-LABEL: name: func_02 +; CHECK: BL @func_01, {{.*}}, implicit-def $fpscr_rm + + +declare float @func_01(float) +declare float @llvm.experimental.constrained.fadd.f32(float, float, metadata, metadata) diff --git a/llvm/test/CodeGen/ARM/strict-fp-int-promote.ll b/llvm/test/CodeGen/ARM/strict-fp-int-promote.ll new file mode 100644 index 0000000..6e5b589 --- /dev/null +++ b/llvm/test/CodeGen/ARM/strict-fp-int-promote.ll @@ -0,0 +1,159 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc -mtriple armv7-- -mattr=+vfp4 -O0 -o - %s | FileCheck %s +; RUN: llc -mtriple armv7-- -mattr=+vfp4 -O3 -o - %s | FileCheck %s --check-prefix=CHECK-O3 + +declare float @llvm.experimental.constrained.sitofp.f32.i32(i32, metadata, metadata) +declare float @llvm.experimental.constrained.sitofp.f32.i16(i16, metadata, metadata) +declare i1 @llvm.experimental.constrained.fcmp.f32(float, float, metadata, metadata) +declare float @llvm.experimental.constrained.uitofp.f32.i16(i16, metadata, metadata) + +define i32 @test(i32 %a, i16 %b) #0 { +; CHECK-LABEL: test: +; CHECK: @ %bb.0: @ %entry +; CHECK-NEXT: sub sp, sp, #16 +; CHECK-NEXT: mov r2, r0 +; CHECK-NEXT: sxth r0, r1 +; CHECK-NEXT: movw r1, #0 +; CHECK-NEXT: movt r1, #17200 +; CHECK-NEXT: str r1, [sp, #4] +; CHECK-NEXT: eor r2, r2, #-2147483648 +; CHECK-NEXT: str r2, [sp] +; CHECK-NEXT: vldr d16, [sp] +; CHECK-NEXT: vldr d17, .LCPI0_0 +; CHECK-NEXT: vsub.f64 d16, d16, d17 +; CHECK-NEXT: vcvt.f32.f64 s0, d16 +; CHECK-NEXT: str r1, [sp, #12] +; CHECK-NEXT: eor r0, r0, #-2147483648 +; CHECK-NEXT: str r0, [sp, #8] +; CHECK-NEXT: vldr d16, [sp, #8] +; CHECK-NEXT: vsub.f64 d16, d16, d17 +; CHECK-NEXT: vcvt.f32.f64 s2, d16 +; CHECK-NEXT: vcmp.f32 s0, s2 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: movweq r0, #1 +; CHECK-NEXT: add sp, sp, #16 +; CHECK-NEXT: bx lr +; CHECK-NEXT: .p2align 3 +; CHECK-NEXT: @ %bb.1: +; CHECK-NEXT: .LCPI0_0: +; CHECK-NEXT: .long 2147483648 @ double 4503601774854144 +; CHECK-NEXT: .long 1127219200 +; +; CHECK-O3-LABEL: test: +; CHECK-O3: @ %bb.0: @ %entry +; CHECK-O3-NEXT: sub sp, sp, #16 +; CHECK-O3-NEXT: sxth r1, r1 +; CHECK-O3-NEXT: movw r2, #0 +; CHECK-O3-NEXT: movt r2, #17200 +; CHECK-O3-NEXT: str r2, [sp, #4] +; CHECK-O3-NEXT: eor r0, r0, #-2147483648 +; CHECK-O3-NEXT: str r0, [sp] +; CHECK-O3-NEXT: vldr d16, [sp] +; CHECK-O3-NEXT: vldr d17, .LCPI0_0 +; CHECK-O3-NEXT: vsub.f64 d16, d16, d17 +; CHECK-O3-NEXT: vcvt.f32.f64 s0, d16 +; CHECK-O3-NEXT: str r2, [sp, #12] +; CHECK-O3-NEXT: eor r0, r1, #-2147483648 +; CHECK-O3-NEXT: str r0, [sp, #8] +; CHECK-O3-NEXT: vldr d16, [sp, #8] +; CHECK-O3-NEXT: vsub.f64 d16, d16, d17 +; CHECK-O3-NEXT: vcvt.f32.f64 s2, d16 +; CHECK-O3-NEXT: vcmp.f32 s0, s2 +; CHECK-O3-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-O3-NEXT: mov r0, #0 +; CHECK-O3-NEXT: movweq r0, #1 +; CHECK-O3-NEXT: add sp, sp, #16 +; CHECK-O3-NEXT: bx lr +; CHECK-O3-NEXT: .p2align 3 +; CHECK-O3-NEXT: @ %bb.1: +; CHECK-O3-NEXT: .LCPI0_0: +; CHECK-O3-NEXT: .long 2147483648 @ double 4503601774854144 +; CHECK-O3-NEXT: .long 1127219200 +entry: + %conv = call float @llvm.experimental.constrained.sitofp.f32.i32(i32 %a, metadata !"round.tonearest", metadata !"fpexcept.strict") #1 + %conv1 = call float @llvm.experimental.constrained.sitofp.f32.i16(i16 %b, metadata !"round.tonearest", metadata !"fpexcept.strict") #1 + %cmp = call i1 @llvm.experimental.constrained.fcmp.f32(float %conv, float %conv1, metadata !"oeq", metadata !"fpexcept.strict") #1 + %conv2 = zext i1 %cmp to i32 + ret i32 %conv2 +} + +define i32 @test2(i32 %a, i16 %b) #0 { +; CHECK-LABEL: test2: +; CHECK: @ %bb.0: @ %entry +; CHECK-NEXT: sub sp, sp, #16 +; CHECK-NEXT: mov r2, r0 +; CHECK-NEXT: uxth r0, r1 +; CHECK-NEXT: movw r1, #0 +; CHECK-NEXT: movt r1, #17200 +; CHECK-NEXT: str r1, [sp, #4] +; CHECK-NEXT: eor r2, r2, #-2147483648 +; CHECK-NEXT: str r2, [sp] +; CHECK-NEXT: vldr d16, [sp] +; CHECK-NEXT: vldr d17, .LCPI1_0 +; CHECK-NEXT: vsub.f64 d16, d16, d17 +; CHECK-NEXT: vcvt.f32.f64 s0, d16 +; CHECK-NEXT: str r1, [sp, #12] +; CHECK-NEXT: str r0, [sp, #8] +; CHECK-NEXT: vldr d16, [sp, #8] +; CHECK-NEXT: vldr d17, .LCPI1_1 +; CHECK-NEXT: vsub.f64 d16, d16, d17 +; CHECK-NEXT: vcvt.f32.f64 s2, d16 +; CHECK-NEXT: vcmp.f32 s0, s2 +; CHECK-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-NEXT: mov r0, #0 +; CHECK-NEXT: movweq r0, #1 +; CHECK-NEXT: add sp, sp, #16 +; CHECK-NEXT: bx lr +; CHECK-NEXT: .p2align 3 +; CHECK-NEXT: @ %bb.1: +; CHECK-NEXT: .LCPI1_0: +; CHECK-NEXT: .long 2147483648 @ double 4503601774854144 +; CHECK-NEXT: .long 1127219200 +; CHECK-NEXT: .LCPI1_1: +; CHECK-NEXT: .long 0 @ double 4503599627370496 +; CHECK-NEXT: .long 1127219200 +; +; CHECK-O3-LABEL: test2: +; CHECK-O3: @ %bb.0: @ %entry +; CHECK-O3-NEXT: sub sp, sp, #16 +; CHECK-O3-NEXT: uxth r1, r1 +; CHECK-O3-NEXT: movw r2, #0 +; CHECK-O3-NEXT: movt r2, #17200 +; CHECK-O3-NEXT: str r2, [sp, #4] +; CHECK-O3-NEXT: eor r0, r0, #-2147483648 +; CHECK-O3-NEXT: str r0, [sp] +; CHECK-O3-NEXT: vldr d16, [sp] +; CHECK-O3-NEXT: vldr d17, .LCPI1_0 +; CHECK-O3-NEXT: vsub.f64 d16, d16, d17 +; CHECK-O3-NEXT: vcvt.f32.f64 s0, d16 +; CHECK-O3-NEXT: str r2, [sp, #12] +; CHECK-O3-NEXT: str r1, [sp, #8] +; CHECK-O3-NEXT: vldr d16, [sp, #8] +; CHECK-O3-NEXT: vldr d17, .LCPI1_1 +; CHECK-O3-NEXT: vsub.f64 d16, d16, d17 +; CHECK-O3-NEXT: vcvt.f32.f64 s2, d16 +; CHECK-O3-NEXT: vcmp.f32 s0, s2 +; CHECK-O3-NEXT: vmrs APSR_nzcv, fpscr +; CHECK-O3-NEXT: mov r0, #0 +; CHECK-O3-NEXT: movweq r0, #1 +; CHECK-O3-NEXT: add sp, sp, #16 +; CHECK-O3-NEXT: bx lr +; CHECK-O3-NEXT: .p2align 3 +; CHECK-O3-NEXT: @ %bb.1: +; CHECK-O3-NEXT: .LCPI1_0: +; CHECK-O3-NEXT: .long 2147483648 @ double 4503601774854144 +; CHECK-O3-NEXT: .long 1127219200 +; CHECK-O3-NEXT: .LCPI1_1: +; CHECK-O3-NEXT: .long 0 @ double 4503599627370496 +; CHECK-O3-NEXT: .long 1127219200 +entry: + %conv = call float @llvm.experimental.constrained.sitofp.f32.i32(i32 %a, metadata !"round.tonearest", metadata !"fpexcept.strict") #1 + %conv1 = call float @llvm.experimental.constrained.uitofp.f32.i16(i16 %b, metadata !"round.tonearest", metadata !"fpexcept.strict") #1 + %cmp = call i1 @llvm.experimental.constrained.fcmp.f32(float %conv, float %conv1, metadata !"oeq", metadata !"fpexcept.strict") #1 + %conv2 = zext i1 %cmp to i32 + ret i32 %conv2 +} + +attributes #0 = { strictfp noinline optnone } +attributes #1 = { strictfp } diff --git a/llvm/test/CodeGen/ARM/strict-fp-ops.ll b/llvm/test/CodeGen/ARM/strict-fp-ops.ll new file mode 100644 index 0000000..608ab07 --- /dev/null +++ b/llvm/test/CodeGen/ARM/strict-fp-ops.ll @@ -0,0 +1,202 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 +; RUN: llc -mtriple armv7-- -mattr=+vfp4 %s -o - | FileCheck %s + + +; Div whose result is unused should be removed unless we have strict exceptions + +define void @unused_div(float %x, float %y) { +; CHECK-LABEL: unused_div: +; CHECK: @ %bb.0: @ %entry +; CHECK-NEXT: bx lr +entry: + %add = fdiv float %x, %y + ret void +} + +define void @unused_div_fpexcept_strict(float %x, float %y) #0 { +; CHECK-LABEL: unused_div_fpexcept_strict: +; CHECK: @ %bb.0: @ %entry +; CHECK-NEXT: vmov s0, r1 +; CHECK-NEXT: vmov s2, r0 +; CHECK-NEXT: vdiv.f32 s0, s2, s0 +; CHECK-NEXT: bx lr +entry: + %add = call float @llvm.experimental.constrained.fdiv.f32(float %x, float %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret void +} + +define void @unused_div_round_dynamic(float %x, float %y) #0 { +; CHECK-LABEL: unused_div_round_dynamic: +; CHECK: @ %bb.0: @ %entry +; CHECK-NEXT: bx lr +entry: + %add = call float @llvm.experimental.constrained.fdiv.f32(float %x, float %y, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0 + ret void +} + + +; Machine CSE should eliminate the second add unless we have strict exceptions + +define float @add_twice(float %x, float %y, i32 %n) { +; CHECK-LABEL: add_twice: +; CHECK: @ %bb.0: @ %entry +; CHECK-NEXT: vmov s0, r1 +; CHECK-NEXT: cmp r2, #0 +; CHECK-NEXT: vmov s2, r0 +; CHECK-NEXT: vadd.f32 s0, s2, s0 +; CHECK-NEXT: vmul.f32 s2, s0, s0 +; CHECK-NEXT: vmoveq.f32 s2, s0 +; CHECK-NEXT: vmov r0, s2 +; CHECK-NEXT: bx lr +entry: + %add = fadd float %x, %y + %tobool.not = icmp eq i32 %n, 0 + br i1 %tobool.not, label %if.end, label %if.then + +if.then: + %add1 = fadd float %x, %y + %mul = fmul float %add, %add1 + br label %if.end + +if.end: + %a.0 = phi float [ %mul, %if.then ], [ %add, %entry ] + ret float %a.0 +} + +define float @add_twice_fpexcept_strict(float %x, float %y, i32 %n) #0 { +; CHECK-LABEL: add_twice_fpexcept_strict: +; CHECK: @ %bb.0: @ %entry +; CHECK-NEXT: vmov s2, r1 +; CHECK-NEXT: cmp r2, #0 +; CHECK-NEXT: vmov s4, r0 +; CHECK-NEXT: vadd.f32 s0, s4, s2 +; CHECK-NEXT: vaddne.f32 s2, s4, s2 +; CHECK-NEXT: vmulne.f32 s0, s0, s2 +; CHECK-NEXT: vmov r0, s0 +; CHECK-NEXT: bx lr +entry: + %add = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + %tobool.not = icmp eq i32 %n, 0 + br i1 %tobool.not, label %if.end, label %if.then + +if.then: + %add1 = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + %mul = call float @llvm.experimental.constrained.fmul.f32(float %add, float %add1, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + br label %if.end + +if.end: + %a.0 = phi float [ %mul, %if.then ], [ %add, %entry ] + ret float %a.0 +} + +define float @add_twice_round_dynamic(float %x, float %y, i32 %n) #0 { +; CHECK-LABEL: add_twice_round_dynamic: +; CHECK: @ %bb.0: @ %entry +; CHECK-NEXT: vmov s0, r1 +; CHECK-NEXT: cmp r2, #0 +; CHECK-NEXT: vmov s2, r0 +; CHECK-NEXT: vadd.f32 s0, s2, s0 +; CHECK-NEXT: vmulne.f32 s0, s0, s0 +; CHECK-NEXT: vmov r0, s0 +; CHECK-NEXT: bx lr +entry: + %add = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0 + %tobool.not = icmp eq i32 %n, 0 + br i1 %tobool.not, label %if.end, label %if.then + +if.then: + %add1 = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0 + %mul = call float @llvm.experimental.constrained.fmul.f32(float %add, float %add1, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0 + br label %if.end + +if.end: + %a.0 = phi float [ %mul, %if.then ], [ %add, %entry ] + ret float %a.0 +} + +; Two adds separated by llvm.set.rounding should be preserved when rounding is +; dynamic (as they may give different results) or when we have strict exceptions +; (the llvm.set.rounding is irrelevant, but both could trap). + +define float @set_rounding(float %x, float %y) { +; CHECK-LABEL: set_rounding: +; CHECK: @ %bb.0: @ %entry +; CHECK-NEXT: vmrs r2, fpscr +; CHECK-NEXT: vmov s2, r0 +; CHECK-NEXT: vmov s0, r1 +; CHECK-NEXT: vadd.f32 s0, s2, s0 +; CHECK-NEXT: vsub.f32 s0, s0, s0 +; CHECK-NEXT: orr r0, r2, #12582912 +; CHECK-NEXT: vmsr fpscr, r0 +; CHECK-NEXT: vmov r0, s0 +; CHECK-NEXT: vmrs r1, fpscr +; CHECK-NEXT: bic r1, r1, #12582912 +; CHECK-NEXT: vmsr fpscr, r1 +; CHECK-NEXT: bx lr +entry: + %add1 = fadd float %x, %y + call void @llvm.set.rounding(i32 0) + %add2 = fadd float %x, %y + call void @llvm.set.rounding(i32 1) + %sub = fsub float %add1, %add2 + ret float %sub +} + +define float @set_rounding_fpexcept_strict(float %x, float %y) #0 { +; CHECK-LABEL: set_rounding_fpexcept_strict: +; CHECK: @ %bb.0: @ %entry +; CHECK-NEXT: vmov s0, r1 +; CHECK-NEXT: vmov s2, r0 +; CHECK-NEXT: vadd.f32 s4, s2, s0 +; CHECK-NEXT: vmrs r0, fpscr +; CHECK-NEXT: orr r0, r0, #12582912 +; CHECK-NEXT: vmsr fpscr, r0 +; CHECK-NEXT: vadd.f32 s0, s2, s0 +; CHECK-NEXT: vmrs r0, fpscr +; CHECK-NEXT: bic r0, r0, #12582912 +; CHECK-NEXT: vmsr fpscr, r0 +; CHECK-NEXT: vsub.f32 s0, s4, s0 +; CHECK-NEXT: vmov r0, s0 +; CHECK-NEXT: bx lr +entry: + %add1 = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + call void @llvm.set.rounding(i32 0) #0 + %add2 = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + call void @llvm.set.rounding(i32 1) #0 + %sub = call float @llvm.experimental.constrained.fsub.f32(float %add1, float %add2, metadata !"round.tonearest", metadata !"fpexcept.strict") #0 + ret float %sub +} + +define float @set_rounding_round_dynamic(float %x, float %y) #0 { +; CHECK-LABEL: set_rounding_round_dynamic: +; CHECK: @ %bb.0: @ %entry +; CHECK-NEXT: vmov s2, r0 +; CHECK-NEXT: vmrs r0, fpscr +; CHECK-NEXT: vmov s0, r1 +; CHECK-NEXT: vadd.f32 s4, s2, s0 +; CHECK-NEXT: orr r0, r0, #12582912 +; CHECK-NEXT: vmsr fpscr, r0 +; CHECK-NEXT: vmrs r0, fpscr +; CHECK-NEXT: vadd.f32 s0, s2, s0 +; CHECK-NEXT: bic r0, r0, #12582912 +; CHECK-NEXT: vmsr fpscr, r0 +; CHECK-NEXT: vsub.f32 s0, s4, s0 +; CHECK-NEXT: vmov r0, s0 +; CHECK-NEXT: bx lr +entry: + %add1 = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0 + call void @llvm.set.rounding(i32 0) #0 + %add2 = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0 + call void @llvm.set.rounding(i32 1) #0 + %sub = call float @llvm.experimental.constrained.fsub.f32(float %add1, float %add2, metadata !"round.dynamic", metadata !"fpexcept.ignore") #0 + ret float %sub +} + +declare float @llvm.experimental.constrained.fadd.f32(float, float, metadata, metadata) +declare float @llvm.experimental.constrained.fsub.f32(float, float, metadata, metadata) +declare float @llvm.experimental.constrained.fmul.f32(float, float, metadata, metadata) +declare float @llvm.experimental.constrained.fdiv.f32(float, float, metadata, metadata) +declare i32 @llvm.get.rounding() +declare void @llvm.set.rounding(i32) + +attributes #0 = { strictfp } diff --git a/llvm/test/CodeGen/ARM/strictfp_f16_abi_promote.ll b/llvm/test/CodeGen/ARM/strictfp_f16_abi_promote.ll new file mode 100644 index 0000000..5906c79 --- /dev/null +++ b/llvm/test/CodeGen/ARM/strictfp_f16_abi_promote.ll @@ -0,0 +1,270 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc -mtriple=armv7-- < %s | FileCheck -check-prefix=NOFP16 %s + +declare void @f16_user(half) +declare half @f16_result() + +declare void @v2f16_user(<2 x half>) +declare <2 x half> @v2f16_result() + +declare void @v4f16_user(<4 x half>) +declare <4 x half> @v4f16_result() + +declare void @v8f16_user(<8 x half>) +declare <8 x half> @v8f16_result() + +define void @f16_arg(half %arg, ptr %ptr) #0 { +; NOFP16-LABEL: f16_arg: +; NOFP16: @ %bb.0: +; NOFP16-NEXT: push {r4, lr} +; NOFP16-NEXT: uxth r0, r0 +; NOFP16-NEXT: mov r4, r1 +; NOFP16-NEXT: bl __gnu_h2f_ieee +; NOFP16-NEXT: str r0, [r4] +; NOFP16-NEXT: pop {r4, pc} + %fpext = call float @llvm.experimental.constrained.fpext.f32.f16(half %arg, metadata !"fpexcept.strict") + store float %fpext, ptr %ptr + ret void +} + +define void @v2f16_arg(<2 x half> %arg, ptr %ptr) #0 { +; NOFP16-LABEL: v2f16_arg: +; NOFP16: @ %bb.0: +; NOFP16-NEXT: push {r4, r5, r11, lr} +; NOFP16-NEXT: vpush {d8} +; NOFP16-NEXT: mov r5, r0 +; NOFP16-NEXT: uxth r0, r1 +; NOFP16-NEXT: mov r4, r2 +; NOFP16-NEXT: bl __gnu_h2f_ieee +; NOFP16-NEXT: uxth r1, r5 +; NOFP16-NEXT: vmov s17, r0 +; NOFP16-NEXT: mov r0, r1 +; NOFP16-NEXT: bl __gnu_h2f_ieee +; NOFP16-NEXT: vmov s16, r0 +; NOFP16-NEXT: vstr d8, [r4] +; NOFP16-NEXT: vpop {d8} +; NOFP16-NEXT: pop {r4, r5, r11, pc} + %fpext = call <2 x float> @llvm.experimental.constrained.fpext.v2f32.v2f16(<2 x half> %arg, metadata !"fpexcept.strict") + store <2 x float> %fpext, ptr %ptr + ret void +} + +define void @v3f16_arg(<3 x half> %arg, ptr %ptr) #0 { +; NOFP16-LABEL: v3f16_arg: +; NOFP16: @ %bb.0: +; NOFP16-NEXT: push {r4, r5, r6, lr} +; NOFP16-NEXT: vpush {d8} +; NOFP16-NEXT: mov r6, r0 +; NOFP16-NEXT: uxth r0, r1 +; NOFP16-NEXT: mov r4, r3 +; NOFP16-NEXT: mov r5, r2 +; NOFP16-NEXT: bl __gnu_h2f_ieee +; NOFP16-NEXT: uxth r1, r6 +; NOFP16-NEXT: vmov s17, r0 +; NOFP16-NEXT: mov r0, r1 +; NOFP16-NEXT: bl __gnu_h2f_ieee +; NOFP16-NEXT: vmov s16, r0 +; NOFP16-NEXT: uxth r0, r5 +; NOFP16-NEXT: vst1.32 {d8}, [r4:64]! +; NOFP16-NEXT: bl __gnu_h2f_ieee +; NOFP16-NEXT: str r0, [r4] +; NOFP16-NEXT: vpop {d8} +; NOFP16-NEXT: pop {r4, r5, r6, pc} + %fpext = call <3 x float> @llvm.experimental.constrained.fpext.v3f32.v3f16(<3 x half> %arg, metadata !"fpexcept.strict") + store <3 x float> %fpext, ptr %ptr + ret void +} + +define void @v4f16_arg(<4 x half> %arg, ptr %ptr) #0 { +; NOFP16-LABEL: v4f16_arg: +; NOFP16: @ %bb.0: +; NOFP16-NEXT: push {r4, r5, r6, r7, r11, lr} +; NOFP16-NEXT: vpush {d8, d9} +; NOFP16-NEXT: mov r6, r0 +; NOFP16-NEXT: uxth r0, r1 +; NOFP16-NEXT: mov r4, r3 +; NOFP16-NEXT: mov r5, r2 +; NOFP16-NEXT: bl __gnu_h2f_ieee +; NOFP16-NEXT: mov r7, r0 +; NOFP16-NEXT: uxth r0, r4 +; NOFP16-NEXT: bl __gnu_h2f_ieee +; NOFP16-NEXT: vmov s19, r0 +; NOFP16-NEXT: uxth r0, r5 +; NOFP16-NEXT: ldr r4, [sp, #40] +; NOFP16-NEXT: bl __gnu_h2f_ieee +; NOFP16-NEXT: vmov s18, r0 +; NOFP16-NEXT: uxth r0, r6 +; NOFP16-NEXT: vmov s17, r7 +; NOFP16-NEXT: bl __gnu_h2f_ieee +; NOFP16-NEXT: vmov s16, r0 +; NOFP16-NEXT: vst1.64 {d8, d9}, [r4] +; NOFP16-NEXT: vpop {d8, d9} +; NOFP16-NEXT: pop {r4, r5, r6, r7, r11, pc} + %fpext = call <4 x float> @llvm.experimental.constrained.fpext.v4f32.v4f16(<4 x half> %arg, metadata !"fpexcept.strict") + store <4 x float> %fpext, ptr %ptr + ret void +} + + define half @f16_return(float %arg) #0 { +; NOFP16-LABEL: f16_return: +; NOFP16: @ %bb.0: +; NOFP16-NEXT: push {r11, lr} +; NOFP16-NEXT: bl __gnu_f2h_ieee +; NOFP16-NEXT: pop {r11, pc} + %fptrunc = call half @llvm.experimental.constrained.fptrunc.f16.f32(float %arg, metadata !"round.tonearest", metadata !"fpexcept.strict") + ret half %fptrunc + } + + define <2 x half> @v2f16_return(<2 x float> %arg) #0 { +; NOFP16-LABEL: v2f16_return: +; NOFP16: @ %bb.0: +; NOFP16-NEXT: push {r11, lr} +; NOFP16-NEXT: vpush {d8} +; NOFP16-NEXT: sub sp, sp, #8 +; NOFP16-NEXT: vmov d8, r0, r1 +; NOFP16-NEXT: vmov r0, s17 +; NOFP16-NEXT: bl __gnu_f2h_ieee +; NOFP16-NEXT: vmov r1, s16 +; NOFP16-NEXT: strh r0, [sp, #6] +; NOFP16-NEXT: mov r0, r1 +; NOFP16-NEXT: bl __gnu_f2h_ieee +; NOFP16-NEXT: strh r0, [sp, #4] +; NOFP16-NEXT: add r0, sp, #4 +; NOFP16-NEXT: vld1.32 {d16[0]}, [r0:32] +; NOFP16-NEXT: vmovl.u16 q8, d16 +; NOFP16-NEXT: vmov.32 r0, d16[0] +; NOFP16-NEXT: vmov.32 r1, d16[1] +; NOFP16-NEXT: add sp, sp, #8 +; NOFP16-NEXT: vpop {d8} +; NOFP16-NEXT: pop {r11, pc} + %fptrunc = call <2 x half> @llvm.experimental.constrained.fptrunc.v2f16.v2f32(<2 x float> %arg, metadata !"round.tonearest", metadata !"fpexcept.strict") + ret <2 x half> %fptrunc + } + + define <3 x half> @v3f16_return(<3 x float> %arg) #0 { +; NOFP16-LABEL: v3f16_return: +; NOFP16: @ %bb.0: +; NOFP16-NEXT: push {r4, r5, r6, lr} +; NOFP16-NEXT: vmov d1, r2, r3 +; NOFP16-NEXT: mov r5, r0 +; NOFP16-NEXT: vmov d0, r0, r1 +; NOFP16-NEXT: mov r4, r1 +; NOFP16-NEXT: vmov r0, s2 +; NOFP16-NEXT: bl __gnu_f2h_ieee +; NOFP16-NEXT: uxth r6, r0 +; NOFP16-NEXT: mov r0, r4 +; NOFP16-NEXT: bl __gnu_f2h_ieee +; NOFP16-NEXT: mov r4, r0 +; NOFP16-NEXT: mov r0, r5 +; NOFP16-NEXT: bl __gnu_f2h_ieee +; NOFP16-NEXT: pkhbt r0, r0, r4, lsl #16 +; NOFP16-NEXT: vmov d16, r0, r6 +; NOFP16-NEXT: vmov.u16 r0, d16[0] +; NOFP16-NEXT: vmov.u16 r1, d16[1] +; NOFP16-NEXT: vmov.u16 r2, d16[2] +; NOFP16-NEXT: pop {r4, r5, r6, pc} + %fptrunc = call <3 x half> @llvm.experimental.constrained.fptrunc.v3f16.v3f32(<3 x float> %arg, metadata !"round.tonearest", metadata !"fpexcept.strict") + ret <3 x half> %fptrunc + } + + define <4 x half> @v4f16_return(<4 x float> %arg) #0 { +; NOFP16-LABEL: v4f16_return: +; NOFP16: @ %bb.0: +; NOFP16-NEXT: push {r4, r5, r11, lr} +; NOFP16-NEXT: vpush {d8, d9} +; NOFP16-NEXT: vmov d8, r2, r3 +; NOFP16-NEXT: vmov d9, r0, r1 +; NOFP16-NEXT: vmov r2, s17 +; NOFP16-NEXT: mov r0, r2 +; NOFP16-NEXT: bl __gnu_f2h_ieee +; NOFP16-NEXT: mov r4, r0 +; NOFP16-NEXT: vmov r0, s16 +; NOFP16-NEXT: bl __gnu_f2h_ieee +; NOFP16-NEXT: vmov r1, s19 +; NOFP16-NEXT: pkhbt r5, r0, r4, lsl #16 +; NOFP16-NEXT: mov r0, r1 +; NOFP16-NEXT: bl __gnu_f2h_ieee +; NOFP16-NEXT: mov r4, r0 +; NOFP16-NEXT: vmov r0, s18 +; NOFP16-NEXT: bl __gnu_f2h_ieee +; NOFP16-NEXT: pkhbt r0, r0, r4, lsl #16 +; NOFP16-NEXT: vmov d16, r0, r5 +; NOFP16-NEXT: vmov.u16 r0, d16[0] +; NOFP16-NEXT: vmov.u16 r1, d16[1] +; NOFP16-NEXT: vmov.u16 r2, d16[2] +; NOFP16-NEXT: vmov.u16 r3, d16[3] +; NOFP16-NEXT: vpop {d8, d9} +; NOFP16-NEXT: pop {r4, r5, r11, pc} + %fptrunc = call <4 x half> @llvm.experimental.constrained.fptrunc.v4f16.v4f32(<4 x float> %arg, metadata !"round.tonearest", metadata !"fpexcept.strict") + ret <4 x half> %fptrunc + } + +define void @outgoing_v4f16_return(ptr %ptr) #0 { +; NOFP16-LABEL: outgoing_v4f16_return: +; NOFP16: @ %bb.0: +; NOFP16-NEXT: push {r4, lr} +; NOFP16-NEXT: mov r4, r0 +; NOFP16-NEXT: bl v4f16_result +; NOFP16-NEXT: strh r3, [r4, #6] +; NOFP16-NEXT: strh r2, [r4, #4] +; NOFP16-NEXT: strh r1, [r4, #2] +; NOFP16-NEXT: strh r0, [r4] +; NOFP16-NEXT: pop {r4, pc} + %val = call <4 x half> @v4f16_result() #0 + store <4 x half> %val, ptr %ptr + ret void +} + +define void @outgoing_v8f16_return(ptr %ptr) #0 { +; NOFP16-LABEL: outgoing_v8f16_return: +; NOFP16: @ %bb.0: +; NOFP16-NEXT: push {r4, r10, r11, lr} +; NOFP16-NEXT: add r11, sp, #8 +; NOFP16-NEXT: sub sp, sp, #16 +; NOFP16-NEXT: bfc sp, #0, #4 +; NOFP16-NEXT: mov r4, r0 +; NOFP16-NEXT: mov r0, sp +; NOFP16-NEXT: bl v8f16_result +; NOFP16-NEXT: ldm sp, {r0, r1, r2, r3} +; NOFP16-NEXT: stm r4, {r0, r1, r2, r3} +; NOFP16-NEXT: sub sp, r11, #8 +; NOFP16-NEXT: pop {r4, r10, r11, pc} + %val = call <8 x half> @v8f16_result() #0 + store <8 x half> %val, ptr %ptr + ret void +} + +define half @call_split_type_used_outside_block_v8f16() #0 { +; NOFP16-LABEL: call_split_type_used_outside_block_v8f16: +; NOFP16: @ %bb.0: @ %bb0 +; NOFP16-NEXT: push {r4, r10, r11, lr} +; NOFP16-NEXT: add r11, sp, #8 +; NOFP16-NEXT: sub sp, sp, #16 +; NOFP16-NEXT: bfc sp, #0, #4 +; NOFP16-NEXT: mov r4, sp +; NOFP16-NEXT: mov r0, r4 +; NOFP16-NEXT: bl v8f16_result +; NOFP16-NEXT: vld1.32 {d16[0]}, [r4:32] +; NOFP16-NEXT: vmov.u16 r0, d16[0] +; NOFP16-NEXT: sub sp, r11, #8 +; NOFP16-NEXT: pop {r4, r10, r11, pc} +bb0: + %split.ret.type = call <8 x half> @v8f16_result() #0 + br label %bb1 + +bb1: + %extract = extractelement <8 x half> %split.ret.type, i32 0 + ret half %extract +} + +declare float @llvm.experimental.constrained.fpext.f32.f16(half, metadata) #0 +declare <2 x float> @llvm.experimental.constrained.fpext.v2f32.v2f16(<2 x half>, metadata) #0 +declare <3 x float> @llvm.experimental.constrained.fpext.v3f32.v3f16(<3 x half>, metadata) #0 +declare <4 x float> @llvm.experimental.constrained.fpext.v4f32.v4f16(<4 x half>, metadata) #0 + +declare half @llvm.experimental.constrained.fptrunc.f16.f32(float, metadata, metadata) #0 +declare <2 x half> @llvm.experimental.constrained.fptrunc.v2f16.v2f32(<2 x float>, metadata, metadata) #0 +declare <3 x half> @llvm.experimental.constrained.fptrunc.v3f16.v3f32(<3 x float>, metadata, metadata) #0 +declare <4 x half> @llvm.experimental.constrained.fptrunc.v4f16.v4f32(<4 x float>, metadata, metadata) #0 + +attributes #0 = { strictfp } diff --git a/llvm/test/CodeGen/DirectX/Metadata/loop-md-errs.ll b/llvm/test/CodeGen/DirectX/Metadata/loop-md-errs.ll new file mode 100644 index 0000000..fbe4653 --- /dev/null +++ b/llvm/test/CodeGen/DirectX/Metadata/loop-md-errs.ll @@ -0,0 +1,113 @@ +; RUN: split-file %s %t +; RUN: not opt -S --dxil-translate-metadata %t/args.ll 2>&1 | FileCheck %t/args.ll +; RUN: not opt -S --dxil-translate-metadata %t/bad-count.ll 2>&1 | FileCheck %t/bad-count.ll +; RUN: not opt -S --dxil-translate-metadata %t/invalid-disable.ll 2>&1 | FileCheck %t/invalid-disable.ll +; RUN: not opt -S --dxil-translate-metadata %t/invalid-full.ll 2>&1 | FileCheck %t/invalid-full.ll + +; Test that loop metadata is validated as with the DXIL validator + +;--- args.ll + +; CHECK: Invalid "llvm.loop" metadata: Provided conflicting hints + +target triple = "dxilv1.0-unknown-shadermodel6.0-library" + +define void @example_loop(i32 %n) { +entry: + br label %loop.header + +loop.header: + %i = phi i32 [ 0, %entry ], [ %i.next, %loop.body ] + %cmp = icmp slt i32 %i, %n + br i1 %cmp, label %loop.body, label %exit + +loop.body: + %i.next = add nsw i32 %i, 1 + br label %loop.header, !llvm.loop !1 + +exit: + ret void +} + +!1 = !{!1, !2, !3} ; conflicting args +!2 = !{!"llvm.loop.unroll.full"} +!3 = !{!"llvm.loop.unroll.disable"} + +;--- bad-count.ll + +; CHECK: "llvm.loop.unroll.count" must have 2 operands and the second must be a constant integer + +target triple = "dxilv1.0-unknown-shadermodel6.0-library" + +define void @example_loop(i32 %n) { +entry: + br label %loop.header + +loop.header: + %i = phi i32 [ 0, %entry ], [ %i.next, %loop.body ] + %cmp = icmp slt i32 %i, %n + br i1 %cmp, label %loop.body, label %exit + +loop.body: + %i.next = add nsw i32 %i, 1 + br label %loop.header, !llvm.loop !1 + +exit: + ret void +} + +!1 = !{!1, !2} +!2 = !{!"llvm.loop.unroll.count", !"not an int"} ; invalid count parameters + +;--- invalid-disable.ll + +; CHECK: Invalid "llvm.loop" metadata: "llvm.loop.unroll.disable" and "llvm.loop.unroll.full" must be provided as a single operand + +target triple = "dxilv1.0-unknown-shadermodel6.0-library" + +define void @example_loop(i32 %n) { +entry: + br label %loop.header + +loop.header: + %i = phi i32 [ 0, %entry ], [ %i.next, %loop.body ] + %cmp = icmp slt i32 %i, %n + br i1 %cmp, label %loop.body, label %exit + +loop.body: + %i.next = add nsw i32 %i, 1 + br label %loop.header, !llvm.loop !1 + +exit: + ret void +} + +!1 = !{!1, !2} +!2 = !{!"llvm.loop.unroll.disable", i32 0} ; invalid second operand + + +;--- invalid-full.ll + +; CHECK: Invalid "llvm.loop" metadata: "llvm.loop.unroll.disable" and "llvm.loop.unroll.full" must be provided as a single operand + +target triple = "dxilv1.0-unknown-shadermodel6.0-library" + +define void @example_loop(i32 %n) { +entry: + br label %loop.header + +loop.header: + %i = phi i32 [ 0, %entry ], [ %i.next, %loop.body ] + %cmp = icmp slt i32 %i, %n + br i1 %cmp, label %loop.body, label %exit + +loop.body: + %i.next = add nsw i32 %i, 1 + br label %loop.header, !llvm.loop !1 + +exit: + ret void +} + +!1 = !{!1, !2} +!2 = !{!"llvm.loop.unroll.full", i32 0} ; invalid second operand diff --git a/llvm/test/CodeGen/DirectX/Metadata/loop-md-stripped.ll b/llvm/test/CodeGen/DirectX/Metadata/loop-md-stripped.ll new file mode 100644 index 0000000..09d8aec --- /dev/null +++ b/llvm/test/CodeGen/DirectX/Metadata/loop-md-stripped.ll @@ -0,0 +1,58 @@ +; RUN: split-file %s %t +; RUN: opt -S --dxil-translate-metadata %t/not-distinct.ll 2>&1 | FileCheck %t/not-distinct.ll +; RUN: opt -S --dxil-translate-metadata %t/not-md.ll 2>&1 | FileCheck %t/not-md.ll + +; Test that DXIL incompatible loop metadata is stripped + +;--- not-distinct.ll + +; Ensure it is stripped because it is not provided a distinct loop parent +; CHECK-NOT: {!"llvm.loop.unroll.disable"} + +target triple = "dxilv1.0-unknown-shadermodel6.0-library" + +define void @example_loop(i32 %n) { +entry: + br label %loop.header + +loop.header: + %i = phi i32 [ 0, %entry ], [ %i.next, %loop.body ] + %cmp = icmp slt i32 %i, %n + br i1 %cmp, label %loop.body, label %exit + +loop.body: + %i.next = add nsw i32 %i, 1 + br label %loop.header, !llvm.loop !1 + +exit: + ret void +} + +!1 = !{!"llvm.loop.unroll.disable"} ; first node must be a distinct self-reference + + +;--- not-md.ll + +target triple = "dxilv1.0-unknown-shadermodel6.0-library" + +define void @example_loop(i32 %n) { +entry: + br label %loop.header + +loop.header: + %i = phi i32 [ 0, %entry ], [ %i.next, %loop.body ] + %cmp = icmp slt i32 %i, %n + br i1 %cmp, label %loop.body, label %exit + +loop.body: + %i.next = add nsw i32 %i, 1 + ; CHECK: br label %loop.header, !llvm.loop ![[#LOOP_MD:]] + br label %loop.header, !llvm.loop !1 + +exit: + ret void +} + +; CHECK: ![[#LOOP_MD:]] = distinct !{![[#LOOP_MD]]} + +!1 = !{!1, i32 0} ; second operand is not a metadata node diff --git a/llvm/test/CodeGen/DirectX/Metadata/loop-md-valid.ll b/llvm/test/CodeGen/DirectX/Metadata/loop-md-valid.ll new file mode 100644 index 0000000..a189c0e --- /dev/null +++ b/llvm/test/CodeGen/DirectX/Metadata/loop-md-valid.ll @@ -0,0 +1,95 @@ +; RUN: split-file %s %t +; RUN: opt -S --dxil-translate-metadata %t/count.ll | FileCheck %t/count.ll +; RUN: opt -S --dxil-translate-metadata %t/disable.ll | FileCheck %t/disable.ll +; RUN: opt -S --dxil-translate-metadata %t/full.ll | FileCheck %t/full.ll + +;--- count.ll + +; Test that we collapse a self-referential chain and allow a unroll.count hint + +target triple = "dxilv1.0-unknown-shadermodel6.0-library" + +define void @example_loop(i32 %n) { +entry: + br label %loop.header + +loop.header: + %i = phi i32 [ 0, %entry ], [ %i.next, %loop.body ] + %cmp = icmp slt i32 %i, %n + br i1 %cmp, label %loop.body, label %exit + +loop.body: + %i.next = add nsw i32 %i, 1 + ; CHECK: br label %loop.header, !llvm.loop ![[#LOOP_MD:]] + br label %loop.header, !llvm.loop !0 + +exit: + ret void +} + +; CHECK: ![[#LOOP_MD]] = distinct !{![[#LOOP_MD]], ![[#COUNT:]]} +; CHECK: ![[#COUNT]] = !{!"llvm.loop.unroll.count", i6 4} + +!0 = !{!0, !1} +!1 = !{!1, !2} +!2 = !{!"llvm.loop.unroll.count", i6 4} + +;--- disable.ll + +; Test that we allow a disable hint + +target triple = "dxilv1.0-unknown-shadermodel6.0-library" + +define void @example_loop(i32 %n) { +entry: + br label %loop.header + +loop.header: + %i = phi i32 [ 0, %entry ], [ %i.next, %loop.body ] + %cmp = icmp slt i32 %i, %n + br i1 %cmp, label %loop.body, label %exit + +loop.body: + %i.next = add nsw i32 %i, 1 + ; CHECK: br label %loop.header, !llvm.loop ![[#LOOP_MD:]] + br label %loop.header, !llvm.loop !0 + +exit: + ret void +} + +; CHECK: ![[#LOOP_MD]] = distinct !{![[#LOOP_MD]], ![[#DISABLE:]]} +; CHECK: ![[#DISABLE]] = !{!"llvm.loop.unroll.disable"} + +!0 = !{!0, !1} +!1 = !{!"llvm.loop.unroll.disable"} + +;--- full.ll + +; Test that we allow a full hint + +target triple = "dxilv1.0-unknown-shadermodel6.0-library" + +define void @example_loop(i32 %n) { +entry: + br label %loop.header + +loop.header: + %i = phi i32 [ 0, %entry ], [ %i.next, %loop.body ] + %cmp = icmp slt i32 %i, %n + br i1 %cmp, label %loop.body, label %exit + +loop.body: + %i.next = add nsw i32 %i, 1 + ; CHECK: br label %loop.header, !llvm.loop ![[#LOOP_MD:]] + br label %loop.header, !llvm.loop !0 + +exit: + ret void +} + +; CHECK: ![[#LOOP_MD]] = distinct !{![[#LOOP_MD]], ![[#FULL:]]} +; CHECK: ![[#FULL]] = !{!"llvm.loop.unroll.full"} + +!0 = !{!0, !1} +!1 = !{!"llvm.loop.unroll.full"} diff --git a/llvm/test/CodeGen/DirectX/Metadata/multiple-entries-cs-error.ll b/llvm/test/CodeGen/DirectX/Metadata/multiple-entries-cs-error.ll index 9697d438..5740ee1 100644 --- a/llvm/test/CodeGen/DirectX/Metadata/multiple-entries-cs-error.ll +++ b/llvm/test/CodeGen/DirectX/Metadata/multiple-entries-cs-error.ll @@ -1,4 +1,4 @@ -; RUN: not opt -S -S -dxil-translate-metadata %s 2>&1 | FileCheck %s +; RUN: not opt -S -dxil-translate-metadata %s 2>&1 | FileCheck %s target triple = "dxil-pc-shadermodel6.8-compute" ; CHECK: Non-library shader: One and only one entry expected diff --git a/llvm/test/CodeGen/DirectX/metadata-stripping.ll b/llvm/test/CodeGen/DirectX/metadata-stripping.ll index 531ab6c..53716ff 100644 --- a/llvm/test/CodeGen/DirectX/metadata-stripping.ll +++ b/llvm/test/CodeGen/DirectX/metadata-stripping.ll @@ -14,7 +14,7 @@ entry: %cmp.i = icmp ult i32 1, 2 ; Ensure that the !llvm.loop metadata node gets dropped. - ; CHECK: br i1 %cmp.i, label %_Z4mainDv3_j.exit, label %_Z4mainDv3_j.exit{{$}} + ; CHECK: br i1 %cmp.i, label %_Z4mainDv3_j.exit, label %_Z4mainDv3_j.exit, !llvm.loop [[LOOPMD:![0-9]+]] br i1 %cmp.i, label %_Z4mainDv3_j.exit, label %_Z4mainDv3_j.exit, !llvm.loop !0 _Z4mainDv3_j.exit: ; preds = %for.body.i, %entry @@ -25,7 +25,8 @@ _Z4mainDv3_j.exit: ; preds = %for.body.i, %entry ; No more metadata should be necessary, the rest (the current 0 and 1) ; should be removed. ; CHECK-NOT: !{!"llvm.loop.mustprogress"} -; CHECK: [[RANGEMD]] = !{i32 1, i32 5} +; CHECK-DAG: [[RANGEMD]] = !{i32 1, i32 5} +; CHECK-DAG: [[LOOPMD]] = distinct !{[[LOOPMD]]} ; CHECK-NOT: !{!"llvm.loop.mustprogress"} !0 = distinct !{!0, !1} !1 = !{!"llvm.loop.mustprogress"} diff --git a/llvm/test/CodeGen/DirectX/strip-module-md.ll b/llvm/test/CodeGen/DirectX/strip-module-md.ll new file mode 100644 index 0000000..4d8b9ec --- /dev/null +++ b/llvm/test/CodeGen/DirectX/strip-module-md.ll @@ -0,0 +1,75 @@ +; RUN: opt -S -dxil-translate-metadata < %s | FileCheck %s + +; Ensures that only metadata explictly specified on the allow list, or debug +; related, metadata is emitted + +target triple = "dxil-unknown-shadermodel6.0-compute" + +; CHECK-NOT: !dx.rootsignatures +; CHECK-NOT: !llvm.errno.tbaa + +; CHECK-DAG: !llvm.dbg.cu + +; CHECK-DAG: !llvm.module.flags = !{![[#DWARF_VER:]], ![[#DEBUG_VER:]]} +; CHECK-DAG: !llvm.ident = !{![[#IDENT:]]} + +; CHECK-DAG: !dx.shaderModel +; CHECK-DAG: !dx.version +; CHECK-DAG: !dx.entryPoints +; CHECK-DAG: !dx.valver +; CHECK-DAG: !dx.resources + +; CHECK-NOT: !dx.rootsignatures +; CHECK-NOT: !llvm.errno.tbaa + +; Check allowed llvm metadata structure to ensure it is still DXIL compatible +; If this fails, please ensure that the updated form is DXIL compatible before +; updating the test. + +; CHECK-DAG: ![[#IDENT]] = !{!"clang 22.0.0"} +; CHECK-DAG: ![[#DWARF_VER]] = !{i32 2, !"Dwarf Version", i32 2} +; CHECK-DAG: ![[#DEBUG_VER]] = !{i32 2, !"Debug Info Version", i32 3} + +; CHECK-NOT: !dx.rootsignatures +; CHECK-NOT: !llvm.errno.tbaa + +@BufA.str = private unnamed_addr constant [5 x i8] c"BufA\00", align 1 + +define void @main () #0 { +entry: + %typed0 = call target("dx.TypedBuffer", <4 x float>, 1, 0, 0) + @llvm.dx.resource.handlefrombinding.tdx.TypedBuffer_v4f32_1_0_0( + i32 3, i32 5, i32 1, i32 0, ptr @BufA.str) + ret void +} + +attributes #0 = { noinline nounwind "hlsl.numthreads"="1,1,1" "hlsl.shader"="compute" } + +; Incompatible +!dx.rootsignatures = !{!2} +!llvm.errno.tbaa = !{!5} + +; Compatible +!llvm.dbg.cu = !{!8} +!llvm.module.flags = !{!11, !12} +!llvm.ident = !{!13} +!dx.valver = !{!14} + +!2 = !{ ptr @main, !3, i32 2 } +!3 = !{ !4 } +!4 = !{ !"RootFlags", i32 1 } + +!5 = !{!6, !6, i64 0} +!6 = !{!"omnipotent char", !7} +!7 = !{!"Simple C/C++ TBAA"} + +!8 = distinct !DICompileUnit(language: DW_LANG_C99, file: !9, producer: "Some Compiler", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !10, splitDebugInlining: false, nameTableKind: None) +!9 = !DIFile(filename: "hlsl.hlsl", directory: "/some-path") +!10 = !{} + +!11 = !{i32 2, !"Dwarf Version", i32 2} +!12 = !{i32 2, !"Debug Info Version", i32 3} + +!13 = !{!"clang 22.0.0"} + +!14 = !{i32 1, i32 1} diff --git a/llvm/test/CodeGen/Hexagon/isel-fclass.ll b/llvm/test/CodeGen/Hexagon/isel-fclass.ll new file mode 100644 index 0000000..96b0210 --- /dev/null +++ b/llvm/test/CodeGen/Hexagon/isel-fclass.ll @@ -0,0 +1,86 @@ +; Tests lowering of sfclass/dfclass compares. +; Sub-optimal code +; { +; p0 = sfclass(r0,#16) +; r0 = sfadd(r0,r0) +; } +; { +; r2 = p0 +; } +; { +; if (p0.new) r0 = ##1065353216 +; p0 = cmp.eq(r2,#0) +; jumpr r31 +; } +; With the patterns added, we should be generating +; { +; p0 = sfclass(r0,#16) +; r0 = sfadd(r0,r0) +; } +; { +; if (!p0) r0 = ##1065353216 +; jumpr r31 +; } + +; RUN: llc -march=hexagon -stop-after=hexagon-isel %s -o - | FileCheck %s + +; CHECK: bb.0.entry1 +; CHECK: F2_sfclass +; CHECK-NOT: C2_cmp +; CHECK: C2_not +; CHECK: F2_sfadd +; Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none) +define float @test1(float noundef %x) { +entry1: + %0 = tail call i32 @llvm.hexagon.F2.sfclass(float %x, i32 16) + %tobool.not = icmp eq i32 %0, 0 + %add = fadd float %x, %x + %spec.select = select i1 %tobool.not, float 1.000000e+00, float %add + ret float %spec.select +} + +; CHECK: bb.0.entry2 +; CHECK: F2_sfclass +; CHECK-NOT: C2_cmp +; CHECK: F2_sfadd +define float @test2(float noundef %x) { +entry2: + %0 = tail call i32 @llvm.hexagon.F2.sfclass(float %x, i32 16) + %tobool.not = icmp eq i32 %0, 0 + %add = fadd float %x, %x + %spec.select = select i1 %tobool.not, float %add, float 1.000000e+00 + ret float %spec.select +} + +; CHECK: bb.0.entry3 +; CHECK: F2_dfclass +; CHECK-NOT: C2_cmp +; CHECK: C2_not +; CHECK: F2_dfadd +define double @test3(double noundef %x) { +entry3: + %0 = tail call i32 @llvm.hexagon.F2.dfclass(double %x, i32 16) + %tobool.not = icmp eq i32 %0, 0 + %add = fadd double %x, %x + %spec.select = select i1 %tobool.not, double 1.000000e+00, double %add + ret double %spec.select +} + +; CHECK: bb.0.entry4 +; CHECK: F2_dfclass +; CHECK-NOT: C2_cmp +; CHECK: F2_dfadd +define double @test4(double noundef %x) { +entry4: + %0 = tail call i32 @llvm.hexagon.F2.dfclass(double %x, i32 16) + %tobool.not = icmp eq i32 %0, 0 + %add = fadd double %x, %x + %spec.select = select i1 %tobool.not, double %add, double 1.000000e+00 + ret double %spec.select +} + +; Function Attrs: mustprogress nocallback nofree nosync nounwind willreturn memory(none) +declare i32 @llvm.hexagon.F2.dfclass(double, i32 immarg) + +; Function Attrs: mustprogress nocallback nofree nosync nounwind willreturn memory(none) +declare i32 @llvm.hexagon.F2.sfclass(float, i32 immarg) diff --git a/llvm/test/CodeGen/Hexagon/isel/trunc-vNi1-HVX.ll b/llvm/test/CodeGen/Hexagon/isel/trunc-vNi1-HVX.ll new file mode 100644 index 0000000..1491729 --- /dev/null +++ b/llvm/test/CodeGen/Hexagon/isel/trunc-vNi1-HVX.ll @@ -0,0 +1,18 @@ +; RUN: llc --mtriple=hexagon -mattr=+hvxv79,+hvx-length128b < %s | FileCheck %s + +define void @f5(<64 x i32> %a0, ptr %a1) { +; CHECK-LABEL: f5: +; CHECK: [[REG0:(r[0-9]+)]] = ##16843009 +; CHECK-DAG: q[[Q0:[0-9]+]] = vand(v{{[0-9]+}},[[REG0]]) +; CHECK-DAG: q[[Q1:[0-9]+]] = vand(v{{[0-9]+}},[[REG0]]) +; CHECK: v{{[0-9]+}}.b = vpacke(v{{[0-9]+}}.h,v{{[0-9]+}}.h) +; CHECK: v{{[0-9]+}}.b = vpacke(v{{[0-9]+}}.h,v{{[0-9]+}}.h) +; CHECK: v[[VROR:[0-9]+]] = vror(v{{[0-9]+}},r{{[0-9]+}}) +; CHECK: v[[VOR:[0-9]+]] = vor(v[[VROR]],v{{[0-9]+}}) +; CHECK: q{{[0-9]+}} = vand(v[[VOR]],r{{[0-9]+}}) +b0: + %v0 = trunc <64 x i32> %a0 to <64 x i1> + store <64 x i1> %v0, ptr %a1, align 1 + ret void +} + diff --git a/llvm/test/CodeGen/LoongArch/ir-instruction/flog2.ll b/llvm/test/CodeGen/LoongArch/ir-instruction/flog2.ll index 93fcd42..e02a2e7 100644 --- a/llvm/test/CodeGen/LoongArch/ir-instruction/flog2.ll +++ b/llvm/test/CodeGen/LoongArch/ir-instruction/flog2.ll @@ -12,8 +12,8 @@ define float @flog2_s(float %x) nounwind { ; ; LA64-LABEL: flog2_s: ; LA64: # %bb.0: -; LA64-NEXT: pcaddu18i $t8, %call36(log2f) -; LA64-NEXT: jr $t8 +; LA64-NEXT: flogb.s $fa0, $fa0 +; LA64-NEXT: ret %y = call float @llvm.log2.f32(float %x) ret float %y } @@ -25,8 +25,8 @@ define double @flog2_d(double %x) nounwind { ; ; LA64-LABEL: flog2_d: ; LA64: # %bb.0: -; LA64-NEXT: pcaddu18i $t8, %call36(log2) -; LA64-NEXT: jr $t8 +; LA64-NEXT: flogb.d $fa0, $fa0 +; LA64-NEXT: ret %y = call double @llvm.log2.f64(double %x) ret double %y } diff --git a/llvm/test/CodeGen/LoongArch/lasx/ctpop-ctlz.ll b/llvm/test/CodeGen/LoongArch/lasx/ctpop-ctlz.ll index ba2118f..b3155c9 100644 --- a/llvm/test/CodeGen/LoongArch/lasx/ctpop-ctlz.ll +++ b/llvm/test/CodeGen/LoongArch/lasx/ctpop-ctlz.ll @@ -106,6 +106,69 @@ define void @ctlz_v4i64(ptr %src, ptr %dst) nounwind { ret void } +define void @not_ctlz_v32i8(ptr %src, ptr %dst) nounwind { +; CHECK-LABEL: not_ctlz_v32i8: +; CHECK: # %bb.0: +; CHECK-NEXT: xvld $xr0, $a0, 0 +; CHECK-NEXT: xvxori.b $xr0, $xr0, 255 +; CHECK-NEXT: xvclz.b $xr0, $xr0 +; CHECK-NEXT: xvst $xr0, $a1, 0 +; CHECK-NEXT: ret + %v = load <32 x i8>, ptr %src + %neg = xor <32 x i8> %v, <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1> + %res = call <32 x i8> @llvm.ctlz.v32i8(<32 x i8> %neg, i1 false) + store <32 x i8> %res, ptr %dst + ret void +} + +define void @not_ctlz_v16i16(ptr %src, ptr %dst) nounwind { +; CHECK-LABEL: not_ctlz_v16i16: +; CHECK: # %bb.0: +; CHECK-NEXT: xvld $xr0, $a0, 0 +; CHECK-NEXT: xvrepli.b $xr1, -1 +; CHECK-NEXT: xvxor.v $xr0, $xr0, $xr1 +; CHECK-NEXT: xvclz.h $xr0, $xr0 +; CHECK-NEXT: xvst $xr0, $a1, 0 +; CHECK-NEXT: ret + %v = load <16 x i16>, ptr %src + %neg = xor <16 x i16> %v, <i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1> + %res = call <16 x i16> @llvm.ctlz.v16i16(<16 x i16> %neg, i1 false) + store <16 x i16> %res, ptr %dst + ret void +} + +define void @not_ctlz_v8i32(ptr %src, ptr %dst) nounwind { +; CHECK-LABEL: not_ctlz_v8i32: +; CHECK: # %bb.0: +; CHECK-NEXT: xvld $xr0, $a0, 0 +; CHECK-NEXT: xvrepli.b $xr1, -1 +; CHECK-NEXT: xvxor.v $xr0, $xr0, $xr1 +; CHECK-NEXT: xvclz.w $xr0, $xr0 +; CHECK-NEXT: xvst $xr0, $a1, 0 +; CHECK-NEXT: ret + %v = load <8 x i32>, ptr %src + %neg = xor <8 x i32> %v, <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1> + %res = call <8 x i32> @llvm.ctlz.v8i32(<8 x i32> %neg, i1 false) + store <8 x i32> %res, ptr %dst + ret void +} + +define void @not_ctlz_v4i64(ptr %src, ptr %dst) nounwind { +; CHECK-LABEL: not_ctlz_v4i64: +; CHECK: # %bb.0: +; CHECK-NEXT: xvld $xr0, $a0, 0 +; CHECK-NEXT: xvrepli.b $xr1, -1 +; CHECK-NEXT: xvxor.v $xr0, $xr0, $xr1 +; CHECK-NEXT: xvclz.d $xr0, $xr0 +; CHECK-NEXT: xvst $xr0, $a1, 0 +; CHECK-NEXT: ret + %v = load <4 x i64>, ptr %src + %neg = xor <4 x i64> %v, <i64 -1, i64 -1, i64 -1, i64 -1> + %res = call <4 x i64> @llvm.ctlz.v4i64(<4 x i64> %neg, i1 false) + store <4 x i64> %res, ptr %dst + ret void +} + declare <32 x i8> @llvm.ctpop.v32i8(<32 x i8>) declare <16 x i16> @llvm.ctpop.v16i16(<16 x i16>) declare <8 x i32> @llvm.ctpop.v8i32(<8 x i32>) diff --git a/llvm/test/CodeGen/LoongArch/lasx/fp-rounding.ll b/llvm/test/CodeGen/LoongArch/lasx/fp-rounding.ll new file mode 100644 index 0000000..fa5f27e --- /dev/null +++ b/llvm/test/CodeGen/LoongArch/lasx/fp-rounding.ll @@ -0,0 +1,132 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lasx < %s | FileCheck %s +; RUN: llc --mtriple=loongarch64 --mattr=+lasx < %s | FileCheck %s + +;; ceilf +define void @ceil_v8f32(ptr %res, ptr %a0) nounwind { +; CHECK-LABEL: ceil_v8f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvfrintrp.s $xr0, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %v0 = load <8 x float>, ptr %a0 + %r = call <8 x float> @llvm.ceil.v8f32(<8 x float> %v0) + store <8 x float> %r, ptr %res + ret void +} + +;; ceil +define void @ceil_v4f64(ptr %res, ptr %a0) nounwind { +; CHECK-LABEL: ceil_v4f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvfrintrp.d $xr0, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %v0 = load <4 x double>, ptr %a0 + %r = call <4 x double> @llvm.ceil.v4f64(<4 x double> %v0) + store <4 x double> %r, ptr %res + ret void +} + +;; floorf +define void @floor_v8f32(ptr %res, ptr %a0) nounwind { +; CHECK-LABEL: floor_v8f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvfrintrm.s $xr0, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %v0 = load <8 x float>, ptr %a0 + %r = call <8 x float> @llvm.floor.v8f32(<8 x float> %v0) + store <8 x float> %r, ptr %res + ret void +} + +;; floor +define void @floor_v4f64(ptr %res, ptr %a0) nounwind { +; CHECK-LABEL: floor_v4f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvfrintrm.d $xr0, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %v0 = load <4 x double>, ptr %a0 + %r = call <4 x double> @llvm.floor.v4f64(<4 x double> %v0) + store <4 x double> %r, ptr %res + ret void +} + +;; truncf +define void @trunc_v8f32(ptr %res, ptr %a0) nounwind { +; CHECK-LABEL: trunc_v8f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvfrintrz.s $xr0, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %v0 = load <8 x float>, ptr %a0 + %r = call <8 x float> @llvm.trunc.v8f32(<8 x float> %v0) + store <8 x float> %r, ptr %res + ret void +} + +;; trunc +define void @trunc_v4f64(ptr %res, ptr %a0) nounwind { +; CHECK-LABEL: trunc_v4f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvfrintrz.d $xr0, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %v0 = load <4 x double>, ptr %a0 + %r = call <4 x double> @llvm.trunc.v4f64(<4 x double> %v0) + store <4 x double> %r, ptr %res + ret void +} + +;; roundevenf +define void @roundeven_v8f32(ptr %res, ptr %a0) nounwind { +; CHECK-LABEL: roundeven_v8f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvfrintrne.s $xr0, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %v0 = load <8 x float>, ptr %a0 + %r = call <8 x float> @llvm.roundeven.v8f32(<8 x float> %v0) + store <8 x float> %r, ptr %res + ret void +} + +;; roundeven +define void @roundeven_v4f64(ptr %res, ptr %a0) nounwind { +; CHECK-LABEL: roundeven_v4f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvfrintrne.d $xr0, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %v0 = load <4 x double>, ptr %a0 + %r = call <4 x double> @llvm.roundeven.v4f64(<4 x double> %v0) + store <4 x double> %r, ptr %res + ret void +} + +declare <8 x float> @llvm.ceil.v8f32(<8 x float>) +declare <4 x double> @llvm.ceil.v4f64(<4 x double>) +declare <8 x float> @llvm.floor.v8f32(<8 x float>) +declare <4 x double> @llvm.floor.v4f64(<4 x double>) +declare <8 x float> @llvm.trunc.v8f32(<8 x float>) +declare <4 x double> @llvm.trunc.v4f64(<4 x double>) +declare <8 x float> @llvm.roundeven.v8f32(<8 x float>) +declare <4 x double> @llvm.roundeven.v4f64(<4 x double>) diff --git a/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/avg.ll b/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/avg.ll new file mode 100644 index 0000000..5c5c199 --- /dev/null +++ b/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/avg.ll @@ -0,0 +1,321 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lasx < %s | FileCheck %s --check-prefixes=CHECK,LA32 +; RUN: llc --mtriple=loongarch64 --mattr=+lasx < %s | FileCheck %s --check-prefixes=CHECK,LA64 + +define void @xvavg_b(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavg_b: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvavg.b $xr0, $xr0, $xr1 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <32 x i8>, ptr %a + %vb = load <32 x i8>, ptr %b + %add = add <32 x i8> %va, %vb + %shr = ashr <32 x i8> %add, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1> + store <32 x i8> %shr, ptr %res + ret void +} + +define void @xvavg_h(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavg_h: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvavg.h $xr0, $xr0, $xr1 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <16 x i16>, ptr %a + %vb = load <16 x i16>, ptr %b + %add = add <16 x i16> %va, %vb + %shr = ashr <16 x i16> %add, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + store <16 x i16> %shr, ptr %res + ret void +} + +define void @xvavg_w(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavg_w: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvavg.w $xr0, $xr0, $xr1 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <8 x i32>, ptr %a + %vb = load <8 x i32>, ptr %b + %add = add <8 x i32> %va, %vb + %shr = ashr <8 x i32> %add, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> + store <8 x i32> %shr, ptr %res + ret void +} + +define void @xvavg_d(ptr %res, ptr %a, ptr %b) nounwind { +; LA32-LABEL: xvavg_d: +; LA32: # %bb.0: # %entry +; LA32-NEXT: xvld $xr0, $a1, 0 +; LA32-NEXT: xvld $xr1, $a2, 0 +; LA32-NEXT: xvadd.d $xr0, $xr0, $xr1 +; LA32-NEXT: xvsrai.d $xr0, $xr0, 1 +; LA32-NEXT: xvst $xr0, $a0, 0 +; LA32-NEXT: ret +; +; LA64-LABEL: xvavg_d: +; LA64: # %bb.0: # %entry +; LA64-NEXT: xvld $xr0, $a1, 0 +; LA64-NEXT: xvld $xr1, $a2, 0 +; LA64-NEXT: xvavg.d $xr0, $xr0, $xr1 +; LA64-NEXT: xvst $xr0, $a0, 0 +; LA64-NEXT: ret +entry: + %va = load <4 x i64>, ptr %a + %vb = load <4 x i64>, ptr %b + %add = add <4 x i64> %va, %vb + %shr = ashr <4 x i64> %add, <i64 1, i64 1, i64 1, i64 1> + store <4 x i64> %shr, ptr %res + ret void +} + +define void @xvavg_bu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavg_bu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvavg.bu $xr0, $xr0, $xr1 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <32 x i8>, ptr %a + %vb = load <32 x i8>, ptr %b + %add = add <32 x i8> %va, %vb + %shr = lshr <32 x i8> %add, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1> + store <32 x i8> %shr, ptr %res + ret void +} + +define void @xvavg_hu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavg_hu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvavg.hu $xr0, $xr0, $xr1 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <16 x i16>, ptr %a + %vb = load <16 x i16>, ptr %b + %add = add <16 x i16> %va, %vb + %shr = lshr <16 x i16> %add, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + store <16 x i16> %shr, ptr %res + ret void +} + +define void @xvavg_wu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavg_wu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvavg.wu $xr0, $xr0, $xr1 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <8 x i32>, ptr %a + %vb = load <8 x i32>, ptr %b + %add = add <8 x i32> %va, %vb + %shr = lshr <8 x i32> %add, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> + store <8 x i32> %shr, ptr %res + ret void +} + +define void @xvavg_du(ptr %res, ptr %a, ptr %b) nounwind { +; LA32-LABEL: xvavg_du: +; LA32: # %bb.0: # %entry +; LA32-NEXT: xvld $xr0, $a1, 0 +; LA32-NEXT: xvld $xr1, $a2, 0 +; LA32-NEXT: xvadd.d $xr0, $xr0, $xr1 +; LA32-NEXT: xvsrli.d $xr0, $xr0, 1 +; LA32-NEXT: xvst $xr0, $a0, 0 +; LA32-NEXT: ret +; +; LA64-LABEL: xvavg_du: +; LA64: # %bb.0: # %entry +; LA64-NEXT: xvld $xr0, $a1, 0 +; LA64-NEXT: xvld $xr1, $a2, 0 +; LA64-NEXT: xvavg.du $xr0, $xr0, $xr1 +; LA64-NEXT: xvst $xr0, $a0, 0 +; LA64-NEXT: ret +entry: + %va = load <4 x i64>, ptr %a + %vb = load <4 x i64>, ptr %b + %add = add <4 x i64> %va, %vb + %shr = lshr <4 x i64> %add, <i64 1, i64 1, i64 1, i64 1> + store <4 x i64> %shr, ptr %res + ret void +} + +define void @xvavgr_b(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavgr_b: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvavgr.b $xr0, $xr0, $xr1 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <32 x i8>, ptr %a + %vb = load <32 x i8>, ptr %b + %add = add <32 x i8> %va, %vb + %add1 = add <32 x i8> %add, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1> + %shr = ashr <32 x i8> %add1, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1> + store <32 x i8> %shr, ptr %res + ret void +} + +define void @xvavgr_h(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavgr_h: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvavgr.h $xr0, $xr0, $xr1 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <16 x i16>, ptr %a + %vb = load <16 x i16>, ptr %b + %add = add <16 x i16> %va, %vb + %add1 = add <16 x i16> %add, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + %shr = ashr <16 x i16> %add1, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + store <16 x i16> %shr, ptr %res + ret void +} + +define void @xvavgr_w(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavgr_w: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvavgr.w $xr0, $xr0, $xr1 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <8 x i32>, ptr %a + %vb = load <8 x i32>, ptr %b + %add = add <8 x i32> %va, %vb + %add1 = add <8 x i32> %add, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> + %shr = ashr <8 x i32> %add1, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> + store <8 x i32> %shr, ptr %res + ret void +} + +define void @xvavgr_d(ptr %res, ptr %a, ptr %b) nounwind { +; LA32-LABEL: xvavgr_d: +; LA32: # %bb.0: # %entry +; LA32-NEXT: xvld $xr0, $a1, 0 +; LA32-NEXT: xvld $xr1, $a2, 0 +; LA32-NEXT: xvadd.d $xr0, $xr0, $xr1 +; LA32-NEXT: xvaddi.du $xr0, $xr0, 1 +; LA32-NEXT: xvsrai.d $xr0, $xr0, 1 +; LA32-NEXT: xvst $xr0, $a0, 0 +; LA32-NEXT: ret +; +; LA64-LABEL: xvavgr_d: +; LA64: # %bb.0: # %entry +; LA64-NEXT: xvld $xr0, $a1, 0 +; LA64-NEXT: xvld $xr1, $a2, 0 +; LA64-NEXT: xvavgr.d $xr0, $xr0, $xr1 +; LA64-NEXT: xvst $xr0, $a0, 0 +; LA64-NEXT: ret +entry: + %va = load <4 x i64>, ptr %a + %vb = load <4 x i64>, ptr %b + %add = add <4 x i64> %va, %vb + %add1 = add <4 x i64> %add, <i64 1, i64 1, i64 1, i64 1> + %shr = ashr <4 x i64> %add1, <i64 1, i64 1, i64 1, i64 1> + store <4 x i64> %shr, ptr %res + ret void +} + +define void @xvavgr_bu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavgr_bu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvavgr.bu $xr0, $xr0, $xr1 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <32 x i8>, ptr %a + %vb = load <32 x i8>, ptr %b + %add = add <32 x i8> %va, %vb + %add1 = add <32 x i8> %add, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1> + %shr = lshr <32 x i8> %add1, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1> + store <32 x i8> %shr, ptr %res + ret void +} + +define void @xvavgr_hu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavgr_hu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvavgr.hu $xr0, $xr0, $xr1 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <16 x i16>, ptr %a + %vb = load <16 x i16>, ptr %b + %add = add <16 x i16> %va, %vb + %add1 = add <16 x i16> %add, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + %shr = lshr <16 x i16> %add1, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + store <16 x i16> %shr, ptr %res + ret void +} + +define void @xvavgr_wu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavgr_wu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvavgr.wu $xr0, $xr0, $xr1 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <8 x i32>, ptr %a + %vb = load <8 x i32>, ptr %b + %add = add <8 x i32> %va, %vb + %add1 = add <8 x i32> %add, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> + %shr = lshr <8 x i32> %add1, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> + store <8 x i32> %shr, ptr %res + ret void +} + +define void @xvavgr_du(ptr %res, ptr %a, ptr %b) nounwind { +; LA32-LABEL: xvavgr_du: +; LA32: # %bb.0: # %entry +; LA32-NEXT: xvld $xr0, $a1, 0 +; LA32-NEXT: xvld $xr1, $a2, 0 +; LA32-NEXT: xvadd.d $xr0, $xr0, $xr1 +; LA32-NEXT: xvaddi.du $xr0, $xr0, 1 +; LA32-NEXT: xvsrli.d $xr0, $xr0, 1 +; LA32-NEXT: xvst $xr0, $a0, 0 +; LA32-NEXT: ret +; +; LA64-LABEL: xvavgr_du: +; LA64: # %bb.0: # %entry +; LA64-NEXT: xvld $xr0, $a1, 0 +; LA64-NEXT: xvld $xr1, $a2, 0 +; LA64-NEXT: xvavgr.du $xr0, $xr0, $xr1 +; LA64-NEXT: xvst $xr0, $a0, 0 +; LA64-NEXT: ret +entry: + %va = load <4 x i64>, ptr %a + %vb = load <4 x i64>, ptr %b + %add = add <4 x i64> %va, %vb + %add1 = add <4 x i64> %add, <i64 1, i64 1, i64 1, i64 1> + %shr = lshr <4 x i64> %add1, <i64 1, i64 1, i64 1, i64 1> + store <4 x i64> %shr, ptr %res + ret void +} diff --git a/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/avgfloor-ceil.ll b/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/avgfloor-ceil.ll new file mode 100644 index 0000000..c82adcb --- /dev/null +++ b/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/avgfloor-ceil.ll @@ -0,0 +1,379 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lasx < %s | FileCheck %s +; RUN: llc --mtriple=loongarch64 --mattr=+lasx < %s | FileCheck %s + +define void @xvavg_b(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavg_b: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvand.v $xr2, $xr0, $xr1 +; CHECK-NEXT: xvxor.v $xr0, $xr0, $xr1 +; CHECK-NEXT: xvsrai.b $xr0, $xr0, 1 +; CHECK-NEXT: xvadd.b $xr0, $xr2, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <32 x i8>, ptr %a + %vb = load <32 x i8>, ptr %b + %ea = sext <32 x i8> %va to <32 x i16> + %eb = sext <32 x i8> %vb to <32 x i16> + %add = add <32 x i16> %ea, %eb + %shr = lshr <32 x i16> %add, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + %r = trunc <32 x i16> %shr to <32 x i8> + store <32 x i8> %r, ptr %res + ret void +} + +define void @xvavg_h(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavg_h: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvand.v $xr2, $xr0, $xr1 +; CHECK-NEXT: xvxor.v $xr0, $xr0, $xr1 +; CHECK-NEXT: xvsrai.h $xr0, $xr0, 1 +; CHECK-NEXT: xvadd.h $xr0, $xr2, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <16 x i16>, ptr %a + %vb = load <16 x i16>, ptr %b + %ea = sext <16 x i16> %va to <16 x i32> + %eb = sext <16 x i16> %vb to <16 x i32> + %add = add <16 x i32> %ea, %eb + %shr = lshr <16 x i32> %add, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> + %r = trunc <16 x i32> %shr to <16 x i16> + store <16 x i16> %r, ptr %res + ret void +} + +define void @xvavg_w(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavg_w: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvand.v $xr2, $xr0, $xr1 +; CHECK-NEXT: xvxor.v $xr0, $xr0, $xr1 +; CHECK-NEXT: xvsrai.w $xr0, $xr0, 1 +; CHECK-NEXT: xvadd.w $xr0, $xr2, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <8 x i32>, ptr %a + %vb = load <8 x i32>, ptr %b + %ea = sext <8 x i32> %va to <8 x i64> + %eb = sext <8 x i32> %vb to <8 x i64> + %add = add <8 x i64> %ea, %eb + %shr = lshr <8 x i64> %add, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1> + %r = trunc <8 x i64> %shr to <8 x i32> + store <8 x i32> %r, ptr %res + ret void +} + +define void @xvavg_d(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavg_d: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvand.v $xr2, $xr0, $xr1 +; CHECK-NEXT: xvxor.v $xr0, $xr0, $xr1 +; CHECK-NEXT: xvsrai.d $xr0, $xr0, 1 +; CHECK-NEXT: xvadd.d $xr0, $xr2, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <4 x i64>, ptr %a + %vb = load <4 x i64>, ptr %b + %ea = sext <4 x i64> %va to <4 x i128> + %eb = sext <4 x i64> %vb to <4 x i128> + %add = add <4 x i128> %ea, %eb + %shr = lshr <4 x i128> %add, <i128 1, i128 1, i128 1, i128 1> + %r = trunc <4 x i128> %shr to <4 x i64> + store <4 x i64> %r, ptr %res + ret void +} + +define void @xvavg_bu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavg_bu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvand.v $xr2, $xr0, $xr1 +; CHECK-NEXT: xvxor.v $xr0, $xr0, $xr1 +; CHECK-NEXT: xvsrli.b $xr0, $xr0, 1 +; CHECK-NEXT: xvadd.b $xr0, $xr2, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <32 x i8>, ptr %a + %vb = load <32 x i8>, ptr %b + %ea = zext <32 x i8> %va to <32 x i16> + %eb = zext <32 x i8> %vb to <32 x i16> + %add = add <32 x i16> %ea, %eb + %shr = lshr <32 x i16> %add, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + %r = trunc <32 x i16> %shr to <32 x i8> + store <32 x i8> %r, ptr %res + ret void +} + +define void @xvavg_hu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavg_hu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvand.v $xr2, $xr0, $xr1 +; CHECK-NEXT: xvxor.v $xr0, $xr0, $xr1 +; CHECK-NEXT: xvsrli.h $xr0, $xr0, 1 +; CHECK-NEXT: xvadd.h $xr0, $xr2, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <16 x i16>, ptr %a + %vb = load <16 x i16>, ptr %b + %ea = zext <16 x i16> %va to <16 x i32> + %eb = zext <16 x i16> %vb to <16 x i32> + %add = add <16 x i32> %ea, %eb + %shr = lshr <16 x i32> %add, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> + %r = trunc <16 x i32> %shr to <16 x i16> + store <16 x i16> %r, ptr %res + ret void +} + +define void @xvavg_wu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavg_wu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvand.v $xr2, $xr0, $xr1 +; CHECK-NEXT: xvxor.v $xr0, $xr0, $xr1 +; CHECK-NEXT: xvsrli.w $xr0, $xr0, 1 +; CHECK-NEXT: xvadd.w $xr0, $xr2, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <8 x i32>, ptr %a + %vb = load <8 x i32>, ptr %b + %ea = zext <8 x i32> %va to <8 x i64> + %eb = zext <8 x i32> %vb to <8 x i64> + %add = add <8 x i64> %ea, %eb + %shr = lshr <8 x i64> %add, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1> + %r = trunc <8 x i64> %shr to <8 x i32> + store <8 x i32> %r, ptr %res + ret void +} + +define void @xvavg_du(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavg_du: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvand.v $xr2, $xr0, $xr1 +; CHECK-NEXT: xvxor.v $xr0, $xr0, $xr1 +; CHECK-NEXT: xvsrli.d $xr0, $xr0, 1 +; CHECK-NEXT: xvadd.d $xr0, $xr2, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <4 x i64>, ptr %a + %vb = load <4 x i64>, ptr %b + %ea = zext <4 x i64> %va to <4 x i128> + %eb = zext <4 x i64> %vb to <4 x i128> + %add = add <4 x i128> %ea, %eb + %shr = lshr <4 x i128> %add, <i128 1, i128 1, i128 1, i128 1> + %r = trunc <4 x i128> %shr to <4 x i64> + store <4 x i64> %r, ptr %res + ret void +} + +define void @xvavgr_b(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavgr_b: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvor.v $xr2, $xr0, $xr1 +; CHECK-NEXT: xvxor.v $xr0, $xr0, $xr1 +; CHECK-NEXT: xvsrai.b $xr0, $xr0, 1 +; CHECK-NEXT: xvsub.b $xr0, $xr2, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <32 x i8>, ptr %a + %vb = load <32 x i8>, ptr %b + %ea = sext <32 x i8> %va to <32 x i16> + %eb = sext <32 x i8> %vb to <32 x i16> + %add = add <32 x i16> %ea, %eb + %add1 = add <32 x i16> %add, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + %shr = lshr <32 x i16> %add1, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + %r = trunc <32 x i16> %shr to <32 x i8> + store <32 x i8> %r, ptr %res + ret void +} + +define void @xvavgr_h(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavgr_h: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvor.v $xr2, $xr0, $xr1 +; CHECK-NEXT: xvxor.v $xr0, $xr0, $xr1 +; CHECK-NEXT: xvsrai.h $xr0, $xr0, 1 +; CHECK-NEXT: xvsub.h $xr0, $xr2, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <16 x i16>, ptr %a + %vb = load <16 x i16>, ptr %b + %ea = sext <16 x i16> %va to <16 x i32> + %eb = sext <16 x i16> %vb to <16 x i32> + %add = add <16 x i32> %ea, %eb + %add1 = add <16 x i32> %add, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> + %shr = lshr <16 x i32> %add1, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> + %r = trunc <16 x i32> %shr to <16 x i16> + store <16 x i16> %r, ptr %res + ret void +} + +define void @xvavgr_w(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavgr_w: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvor.v $xr2, $xr0, $xr1 +; CHECK-NEXT: xvxor.v $xr0, $xr0, $xr1 +; CHECK-NEXT: xvsrai.w $xr0, $xr0, 1 +; CHECK-NEXT: xvsub.w $xr0, $xr2, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <8 x i32>, ptr %a + %vb = load <8 x i32>, ptr %b + %ea = sext <8 x i32> %va to <8 x i64> + %eb = sext <8 x i32> %vb to <8 x i64> + %add = add <8 x i64> %ea, %eb + %add1 = add <8 x i64> %add, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1> + %shr = lshr <8 x i64> %add1, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1> + %r = trunc <8 x i64> %shr to <8 x i32> + store <8 x i32> %r, ptr %res + ret void +} + +define void @xvavgr_d(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavgr_d: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvor.v $xr2, $xr0, $xr1 +; CHECK-NEXT: xvxor.v $xr0, $xr0, $xr1 +; CHECK-NEXT: xvsrai.d $xr0, $xr0, 1 +; CHECK-NEXT: xvsub.d $xr0, $xr2, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <4 x i64>, ptr %a + %vb = load <4 x i64>, ptr %b + %ea = sext <4 x i64> %va to <4 x i128> + %eb = sext <4 x i64> %vb to <4 x i128> + %add = add <4 x i128> %ea, %eb + %add1 = add <4 x i128> %add, <i128 1, i128 1, i128 1, i128 1> + %shr = lshr <4 x i128> %add1, <i128 1, i128 1, i128 1, i128 1> + %r = trunc <4 x i128> %shr to <4 x i64> + store <4 x i64> %r, ptr %res + ret void +} + +define void @xvavgr_bu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavgr_bu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvor.v $xr2, $xr0, $xr1 +; CHECK-NEXT: xvxor.v $xr0, $xr0, $xr1 +; CHECK-NEXT: xvsrli.b $xr0, $xr0, 1 +; CHECK-NEXT: xvsub.b $xr0, $xr2, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <32 x i8>, ptr %a + %vb = load <32 x i8>, ptr %b + %ea = zext <32 x i8> %va to <32 x i16> + %eb = zext <32 x i8> %vb to <32 x i16> + %add = add <32 x i16> %ea, %eb + %add1 = add <32 x i16> %add, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + %shr = lshr <32 x i16> %add1, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + %r = trunc <32 x i16> %shr to <32 x i8> + store <32 x i8> %r, ptr %res + ret void +} + +define void @xvavgr_hu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavgr_hu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvor.v $xr2, $xr0, $xr1 +; CHECK-NEXT: xvxor.v $xr0, $xr0, $xr1 +; CHECK-NEXT: xvsrli.h $xr0, $xr0, 1 +; CHECK-NEXT: xvsub.h $xr0, $xr2, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <16 x i16>, ptr %a + %vb = load <16 x i16>, ptr %b + %ea = zext <16 x i16> %va to <16 x i32> + %eb = zext <16 x i16> %vb to <16 x i32> + %add = add <16 x i32> %ea, %eb + %add1 = add <16 x i32> %add, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> + %shr = lshr <16 x i32> %add1, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> + %r = trunc <16 x i32> %shr to <16 x i16> + store <16 x i16> %r, ptr %res + ret void +} + +define void @xvavgr_wu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavgr_wu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvor.v $xr2, $xr0, $xr1 +; CHECK-NEXT: xvxor.v $xr0, $xr0, $xr1 +; CHECK-NEXT: xvsrli.w $xr0, $xr0, 1 +; CHECK-NEXT: xvsub.w $xr0, $xr2, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <8 x i32>, ptr %a + %vb = load <8 x i32>, ptr %b + %ea = zext <8 x i32> %va to <8 x i64> + %eb = zext <8 x i32> %vb to <8 x i64> + %add = add <8 x i64> %ea, %eb + %add1 = add <8 x i64> %add, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1> + %shr = lshr <8 x i64> %add1, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1> + %r = trunc <8 x i64> %shr to <8 x i32> + store <8 x i32> %r, ptr %res + ret void +} + +define void @xvavgr_du(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: xvavgr_du: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvld $xr1, $a2, 0 +; CHECK-NEXT: xvor.v $xr2, $xr0, $xr1 +; CHECK-NEXT: xvxor.v $xr0, $xr0, $xr1 +; CHECK-NEXT: xvsrli.d $xr0, $xr0, 1 +; CHECK-NEXT: xvsub.d $xr0, $xr2, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <4 x i64>, ptr %a + %vb = load <4 x i64>, ptr %b + %ea = zext <4 x i64> %va to <4 x i128> + %eb = zext <4 x i64> %vb to <4 x i128> + %add = add <4 x i128> %ea, %eb + %add1 = add <4 x i128> %add, <i128 1, i128 1, i128 1, i128 1> + %shr = lshr <4 x i128> %add1, <i128 1, i128 1, i128 1, i128 1> + %r = trunc <4 x i128> %shr to <4 x i64> + store <4 x i64> %r, ptr %res + ret void +} diff --git a/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/flog2.ll b/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/flog2.ll index 68f2e3a..6b5f575 100644 --- a/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/flog2.ll +++ b/llvm/test/CodeGen/LoongArch/lasx/ir-instruction/flog2.ll @@ -1,166 +1,17 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 -; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lasx < %s | FileCheck %s --check-prefix=LA32 -; RUN: llc --mtriple=loongarch64 --mattr=+lasx < %s | FileCheck %s --check-prefix=LA64 +; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lasx < %s | FileCheck %s +; RUN: llc --mtriple=loongarch64 --mattr=+lasx < %s | FileCheck %s declare <8 x float> @llvm.log2.v8f32(<8 x float>) declare <4 x double> @llvm.log2.v4f64(<4 x double>) define void @flog2_v8f32(ptr %res, ptr %a) nounwind { -; LA32-LABEL: flog2_v8f32: -; LA32: # %bb.0: # %entry -; LA32-NEXT: addi.w $sp, $sp, -128 -; LA32-NEXT: st.w $ra, $sp, 124 # 4-byte Folded Spill -; LA32-NEXT: st.w $fp, $sp, 120 # 4-byte Folded Spill -; LA32-NEXT: xvld $xr0, $a1, 0 -; LA32-NEXT: xvst $xr0, $sp, 80 # 32-byte Folded Spill -; LA32-NEXT: move $fp, $a0 -; LA32-NEXT: xvpickve.w $xr0, $xr0, 5 -; LA32-NEXT: # kill: def $f0 killed $f0 killed $xr0 -; LA32-NEXT: bl log2f -; LA32-NEXT: # kill: def $f0 killed $f0 def $vr0 -; LA32-NEXT: vst $vr0, $sp, 48 # 16-byte Folded Spill -; LA32-NEXT: xvld $xr0, $sp, 80 # 32-byte Folded Reload -; LA32-NEXT: xvpickve.w $xr0, $xr0, 4 -; LA32-NEXT: # kill: def $f0 killed $f0 killed $xr0 -; LA32-NEXT: bl log2f -; LA32-NEXT: # kill: def $f0 killed $f0 def $xr0 -; LA32-NEXT: vld $vr1, $sp, 48 # 16-byte Folded Reload -; LA32-NEXT: vextrins.w $vr0, $vr1, 16 -; LA32-NEXT: xvst $xr0, $sp, 48 # 32-byte Folded Spill -; LA32-NEXT: xvld $xr0, $sp, 80 # 32-byte Folded Reload -; LA32-NEXT: xvpickve.w $xr0, $xr0, 6 -; LA32-NEXT: # kill: def $f0 killed $f0 killed $xr0 -; LA32-NEXT: bl log2f -; LA32-NEXT: # kill: def $f0 killed $f0 def $vr0 -; LA32-NEXT: xvld $xr1, $sp, 48 # 32-byte Folded Reload -; LA32-NEXT: vextrins.w $vr1, $vr0, 32 -; LA32-NEXT: xvst $xr1, $sp, 48 # 32-byte Folded Spill -; LA32-NEXT: xvld $xr0, $sp, 80 # 32-byte Folded Reload -; LA32-NEXT: xvpickve.w $xr0, $xr0, 7 -; LA32-NEXT: # kill: def $f0 killed $f0 killed $xr0 -; LA32-NEXT: bl log2f -; LA32-NEXT: # kill: def $f0 killed $f0 def $vr0 -; LA32-NEXT: xvld $xr1, $sp, 48 # 32-byte Folded Reload -; LA32-NEXT: vextrins.w $vr1, $vr0, 48 -; LA32-NEXT: xvst $xr1, $sp, 48 # 32-byte Folded Spill -; LA32-NEXT: xvld $xr0, $sp, 80 # 32-byte Folded Reload -; LA32-NEXT: xvpickve.w $xr0, $xr0, 1 -; LA32-NEXT: # kill: def $f0 killed $f0 killed $xr0 -; LA32-NEXT: bl log2f -; LA32-NEXT: # kill: def $f0 killed $f0 def $vr0 -; LA32-NEXT: vst $vr0, $sp, 16 # 16-byte Folded Spill -; LA32-NEXT: xvld $xr0, $sp, 80 # 32-byte Folded Reload -; LA32-NEXT: xvpickve.w $xr0, $xr0, 0 -; LA32-NEXT: # kill: def $f0 killed $f0 killed $xr0 -; LA32-NEXT: bl log2f -; LA32-NEXT: # kill: def $f0 killed $f0 def $xr0 -; LA32-NEXT: vld $vr1, $sp, 16 # 16-byte Folded Reload -; LA32-NEXT: vextrins.w $vr0, $vr1, 16 -; LA32-NEXT: xvst $xr0, $sp, 16 # 32-byte Folded Spill -; LA32-NEXT: xvld $xr0, $sp, 80 # 32-byte Folded Reload -; LA32-NEXT: xvpickve.w $xr0, $xr0, 2 -; LA32-NEXT: # kill: def $f0 killed $f0 killed $xr0 -; LA32-NEXT: bl log2f -; LA32-NEXT: # kill: def $f0 killed $f0 def $vr0 -; LA32-NEXT: xvld $xr1, $sp, 16 # 32-byte Folded Reload -; LA32-NEXT: vextrins.w $vr1, $vr0, 32 -; LA32-NEXT: xvst $xr1, $sp, 16 # 32-byte Folded Spill -; LA32-NEXT: xvld $xr0, $sp, 80 # 32-byte Folded Reload -; LA32-NEXT: xvpickve.w $xr0, $xr0, 3 -; LA32-NEXT: # kill: def $f0 killed $f0 killed $xr0 -; LA32-NEXT: bl log2f -; LA32-NEXT: # kill: def $f0 killed $f0 def $vr0 -; LA32-NEXT: xvld $xr1, $sp, 16 # 32-byte Folded Reload -; LA32-NEXT: vextrins.w $vr1, $vr0, 48 -; LA32-NEXT: xvld $xr0, $sp, 48 # 32-byte Folded Reload -; LA32-NEXT: xvpermi.q $xr1, $xr0, 2 -; LA32-NEXT: xvst $xr1, $fp, 0 -; LA32-NEXT: ld.w $fp, $sp, 120 # 4-byte Folded Reload -; LA32-NEXT: ld.w $ra, $sp, 124 # 4-byte Folded Reload -; LA32-NEXT: addi.w $sp, $sp, 128 -; LA32-NEXT: ret -; -; LA64-LABEL: flog2_v8f32: -; LA64: # %bb.0: # %entry -; LA64-NEXT: addi.d $sp, $sp, -128 -; LA64-NEXT: st.d $ra, $sp, 120 # 8-byte Folded Spill -; LA64-NEXT: st.d $fp, $sp, 112 # 8-byte Folded Spill -; LA64-NEXT: xvld $xr0, $a1, 0 -; LA64-NEXT: xvst $xr0, $sp, 80 # 32-byte Folded Spill -; LA64-NEXT: move $fp, $a0 -; LA64-NEXT: xvpickve.w $xr0, $xr0, 5 -; LA64-NEXT: # kill: def $f0 killed $f0 killed $xr0 -; LA64-NEXT: pcaddu18i $ra, %call36(log2f) -; LA64-NEXT: jirl $ra, $ra, 0 -; LA64-NEXT: # kill: def $f0 killed $f0 def $vr0 -; LA64-NEXT: vst $vr0, $sp, 48 # 16-byte Folded Spill -; LA64-NEXT: xvld $xr0, $sp, 80 # 32-byte Folded Reload -; LA64-NEXT: xvpickve.w $xr0, $xr0, 4 -; LA64-NEXT: # kill: def $f0 killed $f0 killed $xr0 -; LA64-NEXT: pcaddu18i $ra, %call36(log2f) -; LA64-NEXT: jirl $ra, $ra, 0 -; LA64-NEXT: # kill: def $f0 killed $f0 def $xr0 -; LA64-NEXT: vld $vr1, $sp, 48 # 16-byte Folded Reload -; LA64-NEXT: vextrins.w $vr0, $vr1, 16 -; LA64-NEXT: xvst $xr0, $sp, 48 # 32-byte Folded Spill -; LA64-NEXT: xvld $xr0, $sp, 80 # 32-byte Folded Reload -; LA64-NEXT: xvpickve.w $xr0, $xr0, 6 -; LA64-NEXT: # kill: def $f0 killed $f0 killed $xr0 -; LA64-NEXT: pcaddu18i $ra, %call36(log2f) -; LA64-NEXT: jirl $ra, $ra, 0 -; LA64-NEXT: # kill: def $f0 killed $f0 def $vr0 -; LA64-NEXT: xvld $xr1, $sp, 48 # 32-byte Folded Reload -; LA64-NEXT: vextrins.w $vr1, $vr0, 32 -; LA64-NEXT: xvst $xr1, $sp, 48 # 32-byte Folded Spill -; LA64-NEXT: xvld $xr0, $sp, 80 # 32-byte Folded Reload -; LA64-NEXT: xvpickve.w $xr0, $xr0, 7 -; LA64-NEXT: # kill: def $f0 killed $f0 killed $xr0 -; LA64-NEXT: pcaddu18i $ra, %call36(log2f) -; LA64-NEXT: jirl $ra, $ra, 0 -; LA64-NEXT: # kill: def $f0 killed $f0 def $vr0 -; LA64-NEXT: xvld $xr1, $sp, 48 # 32-byte Folded Reload -; LA64-NEXT: vextrins.w $vr1, $vr0, 48 -; LA64-NEXT: xvst $xr1, $sp, 48 # 32-byte Folded Spill -; LA64-NEXT: xvld $xr0, $sp, 80 # 32-byte Folded Reload -; LA64-NEXT: xvpickve.w $xr0, $xr0, 1 -; LA64-NEXT: # kill: def $f0 killed $f0 killed $xr0 -; LA64-NEXT: pcaddu18i $ra, %call36(log2f) -; LA64-NEXT: jirl $ra, $ra, 0 -; LA64-NEXT: # kill: def $f0 killed $f0 def $vr0 -; LA64-NEXT: vst $vr0, $sp, 16 # 16-byte Folded Spill -; LA64-NEXT: xvld $xr0, $sp, 80 # 32-byte Folded Reload -; LA64-NEXT: xvpickve.w $xr0, $xr0, 0 -; LA64-NEXT: # kill: def $f0 killed $f0 killed $xr0 -; LA64-NEXT: pcaddu18i $ra, %call36(log2f) -; LA64-NEXT: jirl $ra, $ra, 0 -; LA64-NEXT: # kill: def $f0 killed $f0 def $xr0 -; LA64-NEXT: vld $vr1, $sp, 16 # 16-byte Folded Reload -; LA64-NEXT: vextrins.w $vr0, $vr1, 16 -; LA64-NEXT: xvst $xr0, $sp, 16 # 32-byte Folded Spill -; LA64-NEXT: xvld $xr0, $sp, 80 # 32-byte Folded Reload -; LA64-NEXT: xvpickve.w $xr0, $xr0, 2 -; LA64-NEXT: # kill: def $f0 killed $f0 killed $xr0 -; LA64-NEXT: pcaddu18i $ra, %call36(log2f) -; LA64-NEXT: jirl $ra, $ra, 0 -; LA64-NEXT: # kill: def $f0 killed $f0 def $vr0 -; LA64-NEXT: xvld $xr1, $sp, 16 # 32-byte Folded Reload -; LA64-NEXT: vextrins.w $vr1, $vr0, 32 -; LA64-NEXT: xvst $xr1, $sp, 16 # 32-byte Folded Spill -; LA64-NEXT: xvld $xr0, $sp, 80 # 32-byte Folded Reload -; LA64-NEXT: xvpickve.w $xr0, $xr0, 3 -; LA64-NEXT: # kill: def $f0 killed $f0 killed $xr0 -; LA64-NEXT: pcaddu18i $ra, %call36(log2f) -; LA64-NEXT: jirl $ra, $ra, 0 -; LA64-NEXT: # kill: def $f0 killed $f0 def $vr0 -; LA64-NEXT: xvld $xr1, $sp, 16 # 32-byte Folded Reload -; LA64-NEXT: vextrins.w $vr1, $vr0, 48 -; LA64-NEXT: xvld $xr0, $sp, 48 # 32-byte Folded Reload -; LA64-NEXT: xvpermi.q $xr1, $xr0, 2 -; LA64-NEXT: xvst $xr1, $fp, 0 -; LA64-NEXT: ld.d $fp, $sp, 112 # 8-byte Folded Reload -; LA64-NEXT: ld.d $ra, $sp, 120 # 8-byte Folded Reload -; LA64-NEXT: addi.d $sp, $sp, 128 -; LA64-NEXT: ret +; CHECK-LABEL: flog2_v8f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvflogb.s $xr0, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret entry: %v = load <8 x float>, ptr %a %r = call <8 x float> @llvm.log2.v8f32(<8 x float> %v) @@ -169,93 +20,12 @@ entry: } define void @flog2_v4f64(ptr %res, ptr %a) nounwind { -; LA32-LABEL: flog2_v4f64: -; LA32: # %bb.0: # %entry -; LA32-NEXT: addi.w $sp, $sp, -112 -; LA32-NEXT: st.w $ra, $sp, 108 # 4-byte Folded Spill -; LA32-NEXT: st.w $fp, $sp, 104 # 4-byte Folded Spill -; LA32-NEXT: xvld $xr0, $a1, 0 -; LA32-NEXT: xvst $xr0, $sp, 64 # 32-byte Folded Spill -; LA32-NEXT: move $fp, $a0 -; LA32-NEXT: xvpickve.d $xr0, $xr0, 3 -; LA32-NEXT: # kill: def $f0_64 killed $f0_64 killed $xr0 -; LA32-NEXT: bl log2 -; LA32-NEXT: # kill: def $f0_64 killed $f0_64 def $vr0 -; LA32-NEXT: vst $vr0, $sp, 32 # 16-byte Folded Spill -; LA32-NEXT: xvld $xr0, $sp, 64 # 32-byte Folded Reload -; LA32-NEXT: xvpickve.d $xr0, $xr0, 2 -; LA32-NEXT: # kill: def $f0_64 killed $f0_64 killed $xr0 -; LA32-NEXT: bl log2 -; LA32-NEXT: # kill: def $f0_64 killed $f0_64 def $xr0 -; LA32-NEXT: vld $vr1, $sp, 32 # 16-byte Folded Reload -; LA32-NEXT: vextrins.d $vr0, $vr1, 16 -; LA32-NEXT: xvst $xr0, $sp, 32 # 32-byte Folded Spill -; LA32-NEXT: xvld $xr0, $sp, 64 # 32-byte Folded Reload -; LA32-NEXT: xvpickve.d $xr0, $xr0, 1 -; LA32-NEXT: # kill: def $f0_64 killed $f0_64 killed $xr0 -; LA32-NEXT: bl log2 -; LA32-NEXT: # kill: def $f0_64 killed $f0_64 def $vr0 -; LA32-NEXT: vst $vr0, $sp, 16 # 16-byte Folded Spill -; LA32-NEXT: xvld $xr0, $sp, 64 # 32-byte Folded Reload -; LA32-NEXT: xvpickve.d $xr0, $xr0, 0 -; LA32-NEXT: # kill: def $f0_64 killed $f0_64 killed $xr0 -; LA32-NEXT: bl log2 -; LA32-NEXT: # kill: def $f0_64 killed $f0_64 def $xr0 -; LA32-NEXT: vld $vr1, $sp, 16 # 16-byte Folded Reload -; LA32-NEXT: vextrins.d $vr0, $vr1, 16 -; LA32-NEXT: xvld $xr1, $sp, 32 # 32-byte Folded Reload -; LA32-NEXT: xvpermi.q $xr0, $xr1, 2 -; LA32-NEXT: xvst $xr0, $fp, 0 -; LA32-NEXT: ld.w $fp, $sp, 104 # 4-byte Folded Reload -; LA32-NEXT: ld.w $ra, $sp, 108 # 4-byte Folded Reload -; LA32-NEXT: addi.w $sp, $sp, 112 -; LA32-NEXT: ret -; -; LA64-LABEL: flog2_v4f64: -; LA64: # %bb.0: # %entry -; LA64-NEXT: addi.d $sp, $sp, -112 -; LA64-NEXT: st.d $ra, $sp, 104 # 8-byte Folded Spill -; LA64-NEXT: st.d $fp, $sp, 96 # 8-byte Folded Spill -; LA64-NEXT: xvld $xr0, $a1, 0 -; LA64-NEXT: xvst $xr0, $sp, 64 # 32-byte Folded Spill -; LA64-NEXT: move $fp, $a0 -; LA64-NEXT: xvpickve.d $xr0, $xr0, 3 -; LA64-NEXT: # kill: def $f0_64 killed $f0_64 killed $xr0 -; LA64-NEXT: pcaddu18i $ra, %call36(log2) -; LA64-NEXT: jirl $ra, $ra, 0 -; LA64-NEXT: # kill: def $f0_64 killed $f0_64 def $vr0 -; LA64-NEXT: vst $vr0, $sp, 32 # 16-byte Folded Spill -; LA64-NEXT: xvld $xr0, $sp, 64 # 32-byte Folded Reload -; LA64-NEXT: xvpickve.d $xr0, $xr0, 2 -; LA64-NEXT: # kill: def $f0_64 killed $f0_64 killed $xr0 -; LA64-NEXT: pcaddu18i $ra, %call36(log2) -; LA64-NEXT: jirl $ra, $ra, 0 -; LA64-NEXT: # kill: def $f0_64 killed $f0_64 def $xr0 -; LA64-NEXT: vld $vr1, $sp, 32 # 16-byte Folded Reload -; LA64-NEXT: vextrins.d $vr0, $vr1, 16 -; LA64-NEXT: xvst $xr0, $sp, 32 # 32-byte Folded Spill -; LA64-NEXT: xvld $xr0, $sp, 64 # 32-byte Folded Reload -; LA64-NEXT: xvpickve.d $xr0, $xr0, 1 -; LA64-NEXT: # kill: def $f0_64 killed $f0_64 killed $xr0 -; LA64-NEXT: pcaddu18i $ra, %call36(log2) -; LA64-NEXT: jirl $ra, $ra, 0 -; LA64-NEXT: # kill: def $f0_64 killed $f0_64 def $vr0 -; LA64-NEXT: vst $vr0, $sp, 16 # 16-byte Folded Spill -; LA64-NEXT: xvld $xr0, $sp, 64 # 32-byte Folded Reload -; LA64-NEXT: xvpickve.d $xr0, $xr0, 0 -; LA64-NEXT: # kill: def $f0_64 killed $f0_64 killed $xr0 -; LA64-NEXT: pcaddu18i $ra, %call36(log2) -; LA64-NEXT: jirl $ra, $ra, 0 -; LA64-NEXT: # kill: def $f0_64 killed $f0_64 def $xr0 -; LA64-NEXT: vld $vr1, $sp, 16 # 16-byte Folded Reload -; LA64-NEXT: vextrins.d $vr0, $vr1, 16 -; LA64-NEXT: xvld $xr1, $sp, 32 # 32-byte Folded Reload -; LA64-NEXT: xvpermi.q $xr0, $xr1, 2 -; LA64-NEXT: xvst $xr0, $fp, 0 -; LA64-NEXT: ld.d $fp, $sp, 96 # 8-byte Folded Reload -; LA64-NEXT: ld.d $ra, $sp, 104 # 8-byte Folded Reload -; LA64-NEXT: addi.d $sp, $sp, 112 -; LA64-NEXT: ret +; CHECK-LABEL: flog2_v4f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: xvld $xr0, $a1, 0 +; CHECK-NEXT: xvflogb.d $xr0, $xr0 +; CHECK-NEXT: xvst $xr0, $a0, 0 +; CHECK-NEXT: ret entry: %v = load <4 x double>, ptr %a %r = call <4 x double> @llvm.log2.v4f64(<4 x double> %v) diff --git a/llvm/test/CodeGen/LoongArch/lsx/ctpop-ctlz.ll b/llvm/test/CodeGen/LoongArch/lsx/ctpop-ctlz.ll index a9a38e8..6ac7d51 100644 --- a/llvm/test/CodeGen/LoongArch/lsx/ctpop-ctlz.ll +++ b/llvm/test/CodeGen/LoongArch/lsx/ctpop-ctlz.ll @@ -106,6 +106,69 @@ define void @ctlz_v2i64(ptr %src, ptr %dst) nounwind { ret void } +define void @not_ctlz_v16i8(ptr %src, ptr %dst) nounwind { +; CHECK-LABEL: not_ctlz_v16i8: +; CHECK: # %bb.0: +; CHECK-NEXT: vld $vr0, $a0, 0 +; CHECK-NEXT: vxori.b $vr0, $vr0, 255 +; CHECK-NEXT: vclz.b $vr0, $vr0 +; CHECK-NEXT: vst $vr0, $a1, 0 +; CHECK-NEXT: ret + %v = load <16 x i8>, ptr %src + %neg = xor <16 x i8> %v, <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1> + %res = call <16 x i8> @llvm.ctlz.v16i8(<16 x i8> %neg, i1 false) + store <16 x i8> %res, ptr %dst + ret void +} + +define void @not_ctlz_v8i16(ptr %src, ptr %dst) nounwind { +; CHECK-LABEL: not_ctlz_v8i16: +; CHECK: # %bb.0: +; CHECK-NEXT: vld $vr0, $a0, 0 +; CHECK-NEXT: vrepli.b $vr1, -1 +; CHECK-NEXT: vxor.v $vr0, $vr0, $vr1 +; CHECK-NEXT: vclz.h $vr0, $vr0 +; CHECK-NEXT: vst $vr0, $a1, 0 +; CHECK-NEXT: ret + %v = load <8 x i16>, ptr %src + %neg = xor <8 x i16> %v, <i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1> + %res = call <8 x i16> @llvm.ctlz.v8i16(<8 x i16> %neg, i1 false) + store <8 x i16> %res, ptr %dst + ret void +} + +define void @not_ctlz_v4i32(ptr %src, ptr %dst) nounwind { +; CHECK-LABEL: not_ctlz_v4i32: +; CHECK: # %bb.0: +; CHECK-NEXT: vld $vr0, $a0, 0 +; CHECK-NEXT: vrepli.b $vr1, -1 +; CHECK-NEXT: vxor.v $vr0, $vr0, $vr1 +; CHECK-NEXT: vclz.w $vr0, $vr0 +; CHECK-NEXT: vst $vr0, $a1, 0 +; CHECK-NEXT: ret + %v = load <4 x i32>, ptr %src + %neg = xor <4 x i32> %v, <i32 -1, i32 -1, i32 -1, i32 -1> + %res = call <4 x i32> @llvm.ctlz.v4i32(<4 x i32> %neg, i1 false) + store <4 x i32> %res, ptr %dst + ret void +} + +define void @not_ctlz_v2i64(ptr %src, ptr %dst) nounwind { +; CHECK-LABEL: not_ctlz_v2i64: +; CHECK: # %bb.0: +; CHECK-NEXT: vld $vr0, $a0, 0 +; CHECK-NEXT: vrepli.b $vr1, -1 +; CHECK-NEXT: vxor.v $vr0, $vr0, $vr1 +; CHECK-NEXT: vclz.d $vr0, $vr0 +; CHECK-NEXT: vst $vr0, $a1, 0 +; CHECK-NEXT: ret + %v = load <2 x i64>, ptr %src + %neg = xor <2 x i64> %v, <i64 -1, i64 -1> + %res = call <2 x i64> @llvm.ctlz.v2i64(<2 x i64> %neg, i1 false) + store <2 x i64> %res, ptr %dst + ret void +} + declare <16 x i8> @llvm.ctpop.v16i8(<16 x i8>) declare <8 x i16> @llvm.ctpop.v8i16(<8 x i16>) declare <4 x i32> @llvm.ctpop.v4i32(<4 x i32>) diff --git a/llvm/test/CodeGen/LoongArch/lsx/fp-rounding.ll b/llvm/test/CodeGen/LoongArch/lsx/fp-rounding.ll new file mode 100644 index 0000000..cb01ac0 --- /dev/null +++ b/llvm/test/CodeGen/LoongArch/lsx/fp-rounding.ll @@ -0,0 +1,132 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lsx < %s | FileCheck %s +; RUN: llc --mtriple=loongarch64 --mattr=+lsx < %s | FileCheck %s + +;; ceilf +define void @ceil_v4f32(ptr %res, ptr %a0) nounwind { +; CHECK-LABEL: ceil_v4f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vfrintrp.s $vr0, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %v0 = load <4 x float>, ptr %a0 + %r = call <4 x float> @llvm.ceil.v4f32(<4 x float> %v0) + store <4 x float> %r, ptr %res + ret void +} + +;; ceil +define void @ceil_v2f64(ptr %res, ptr %a0) nounwind { +; CHECK-LABEL: ceil_v2f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vfrintrp.d $vr0, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %v0 = load <2 x double>, ptr %a0 + %r = call <2 x double> @llvm.ceil.v2f64(<2 x double> %v0) + store <2 x double> %r, ptr %res + ret void +} + +;; floorf +define void @floor_v4f32(ptr %res, ptr %a0) nounwind { +; CHECK-LABEL: floor_v4f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vfrintrm.s $vr0, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %v0 = load <4 x float>, ptr %a0 + %r = call <4 x float> @llvm.floor.v4f32(<4 x float> %v0) + store <4 x float> %r, ptr %res + ret void +} + +;; floor +define void @floor_v2f64(ptr %res, ptr %a0) nounwind { +; CHECK-LABEL: floor_v2f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vfrintrm.d $vr0, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %v0 = load <2 x double>, ptr %a0 + %r = call <2 x double> @llvm.floor.v2f64(<2 x double> %v0) + store <2 x double> %r, ptr %res + ret void +} + +;; truncf +define void @trunc_v4f32(ptr %res, ptr %a0) nounwind { +; CHECK-LABEL: trunc_v4f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vfrintrz.s $vr0, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %v0 = load <4 x float>, ptr %a0 + %r = call <4 x float> @llvm.trunc.v4f32(<4 x float> %v0) + store <4 x float> %r, ptr %res + ret void +} + +;; trunc +define void @trunc_v2f64(ptr %res, ptr %a0) nounwind { +; CHECK-LABEL: trunc_v2f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vfrintrz.d $vr0, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %v0 = load <2 x double>, ptr %a0 + %r = call <2 x double> @llvm.trunc.v2f64(<2 x double> %v0) + store <2 x double> %r, ptr %res + ret void +} + +;; roundevenf +define void @roundeven_v4f32(ptr %res, ptr %a0) nounwind { +; CHECK-LABEL: roundeven_v4f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vfrintrne.s $vr0, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %v0 = load <4 x float>, ptr %a0 + %r = call <4 x float> @llvm.roundeven.v4f32(<4 x float> %v0) + store <4 x float> %r, ptr %res + ret void +} + +;; roundeven +define void @roundeven_v2f64(ptr %res, ptr %a0) nounwind { +; CHECK-LABEL: roundeven_v2f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vfrintrne.d $vr0, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %v0 = load <2 x double>, ptr %a0 + %r = call <2 x double> @llvm.roundeven.v2f64(<2 x double> %v0) + store <2 x double> %r, ptr %res + ret void +} + +declare <4 x float> @llvm.ceil.v4f32(<4 x float>) +declare <2 x double> @llvm.ceil.v2f64(<2 x double>) +declare <4 x float> @llvm.floor.v4f32(<4 x float>) +declare <2 x double> @llvm.floor.v2f64(<2 x double>) +declare <4 x float> @llvm.trunc.v4f32(<4 x float>) +declare <2 x double> @llvm.trunc.v2f64(<2 x double>) +declare <4 x float> @llvm.roundeven.v4f32(<4 x float>) +declare <2 x double> @llvm.roundeven.v2f64(<2 x double>) diff --git a/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/avg.ll b/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/avg.ll new file mode 100644 index 0000000..334af22 --- /dev/null +++ b/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/avg.ll @@ -0,0 +1,321 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lsx < %s | FileCheck %s --check-prefixes=CHECK,LA32 +; RUN: llc --mtriple=loongarch64 --mattr=+lsx < %s | FileCheck %s --check-prefixes=CHECK,LA64 + +define void @vavg_b(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavg_b: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vavg.b $vr0, $vr0, $vr1 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <16 x i8>, ptr %a + %vb = load <16 x i8>, ptr %b + %add = add <16 x i8> %va, %vb + %shr = ashr <16 x i8> %add, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1> + store <16 x i8> %shr, ptr %res + ret void +} + +define void @vavg_h(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavg_h: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vavg.h $vr0, $vr0, $vr1 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <8 x i16>, ptr %a + %vb = load <8 x i16>, ptr %b + %add = add <8 x i16> %va, %vb + %shr = ashr <8 x i16> %add, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + store <8 x i16> %shr, ptr %res + ret void +} + +define void @vavg_w(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavg_w: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vavg.w $vr0, $vr0, $vr1 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <4 x i32>, ptr %a + %vb = load <4 x i32>, ptr %b + %add = add <4 x i32> %va, %vb + %shr = ashr <4 x i32> %add, <i32 1, i32 1, i32 1, i32 1> + store <4 x i32> %shr, ptr %res + ret void +} + +define void @vavg_d(ptr %res, ptr %a, ptr %b) nounwind { +; LA32-LABEL: vavg_d: +; LA32: # %bb.0: # %entry +; LA32-NEXT: vld $vr0, $a1, 0 +; LA32-NEXT: vld $vr1, $a2, 0 +; LA32-NEXT: vadd.d $vr0, $vr0, $vr1 +; LA32-NEXT: vsrai.d $vr0, $vr0, 1 +; LA32-NEXT: vst $vr0, $a0, 0 +; LA32-NEXT: ret +; +; LA64-LABEL: vavg_d: +; LA64: # %bb.0: # %entry +; LA64-NEXT: vld $vr0, $a1, 0 +; LA64-NEXT: vld $vr1, $a2, 0 +; LA64-NEXT: vavg.d $vr0, $vr0, $vr1 +; LA64-NEXT: vst $vr0, $a0, 0 +; LA64-NEXT: ret +entry: + %va = load <2 x i64>, ptr %a + %vb = load <2 x i64>, ptr %b + %add = add <2 x i64> %va, %vb + %shr = ashr <2 x i64> %add, <i64 1, i64 1> + store <2 x i64> %shr, ptr %res + ret void +} + +define void @vavg_bu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavg_bu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vavg.bu $vr0, $vr0, $vr1 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <16 x i8>, ptr %a + %vb = load <16 x i8>, ptr %b + %add = add <16 x i8> %va, %vb + %shr = lshr <16 x i8> %add, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1> + store <16 x i8> %shr, ptr %res + ret void +} + +define void @vavg_hu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavg_hu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vavg.hu $vr0, $vr0, $vr1 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <8 x i16>, ptr %a + %vb = load <8 x i16>, ptr %b + %add = add <8 x i16> %va, %vb + %shr = lshr <8 x i16> %add, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + store <8 x i16> %shr, ptr %res + ret void +} + +define void @vavg_wu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavg_wu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vavg.wu $vr0, $vr0, $vr1 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <4 x i32>, ptr %a + %vb = load <4 x i32>, ptr %b + %add = add <4 x i32> %va, %vb + %shr = lshr <4 x i32> %add, <i32 1, i32 1, i32 1, i32 1> + store <4 x i32> %shr, ptr %res + ret void +} + +define void @vavg_du(ptr %res, ptr %a, ptr %b) nounwind { +; LA32-LABEL: vavg_du: +; LA32: # %bb.0: # %entry +; LA32-NEXT: vld $vr0, $a1, 0 +; LA32-NEXT: vld $vr1, $a2, 0 +; LA32-NEXT: vadd.d $vr0, $vr0, $vr1 +; LA32-NEXT: vsrli.d $vr0, $vr0, 1 +; LA32-NEXT: vst $vr0, $a0, 0 +; LA32-NEXT: ret +; +; LA64-LABEL: vavg_du: +; LA64: # %bb.0: # %entry +; LA64-NEXT: vld $vr0, $a1, 0 +; LA64-NEXT: vld $vr1, $a2, 0 +; LA64-NEXT: vavg.du $vr0, $vr0, $vr1 +; LA64-NEXT: vst $vr0, $a0, 0 +; LA64-NEXT: ret +entry: + %va = load <2 x i64>, ptr %a + %vb = load <2 x i64>, ptr %b + %add = add <2 x i64> %va, %vb + %shr = lshr <2 x i64> %add, <i64 1, i64 1> + store <2 x i64> %shr, ptr %res + ret void +} + +define void @vavgr_b(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavgr_b: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vavgr.b $vr0, $vr0, $vr1 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <16 x i8>, ptr %a + %vb = load <16 x i8>, ptr %b + %add = add <16 x i8> %va, %vb + %add1 = add <16 x i8> %add, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1> + %shr = ashr <16 x i8> %add1, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1> + store <16 x i8> %shr, ptr %res + ret void +} + +define void @vavgr_h(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavgr_h: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vavgr.h $vr0, $vr0, $vr1 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <8 x i16>, ptr %a + %vb = load <8 x i16>, ptr %b + %add = add <8 x i16> %va, %vb + %add1 = add <8 x i16> %add, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + %shr = ashr <8 x i16> %add1, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + store <8 x i16> %shr, ptr %res + ret void +} + +define void @vavgr_w(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavgr_w: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vavgr.w $vr0, $vr0, $vr1 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <4 x i32>, ptr %a + %vb = load <4 x i32>, ptr %b + %add = add <4 x i32> %va, %vb + %add1 = add <4 x i32> %add, <i32 1, i32 1, i32 1, i32 1> + %shr = ashr <4 x i32> %add1, <i32 1, i32 1, i32 1, i32 1> + store <4 x i32> %shr, ptr %res + ret void +} + +define void @vavgr_d(ptr %res, ptr %a, ptr %b) nounwind { +; LA32-LABEL: vavgr_d: +; LA32: # %bb.0: # %entry +; LA32-NEXT: vld $vr0, $a1, 0 +; LA32-NEXT: vld $vr1, $a2, 0 +; LA32-NEXT: vadd.d $vr0, $vr0, $vr1 +; LA32-NEXT: vaddi.du $vr0, $vr0, 1 +; LA32-NEXT: vsrai.d $vr0, $vr0, 1 +; LA32-NEXT: vst $vr0, $a0, 0 +; LA32-NEXT: ret +; +; LA64-LABEL: vavgr_d: +; LA64: # %bb.0: # %entry +; LA64-NEXT: vld $vr0, $a1, 0 +; LA64-NEXT: vld $vr1, $a2, 0 +; LA64-NEXT: vavgr.d $vr0, $vr0, $vr1 +; LA64-NEXT: vst $vr0, $a0, 0 +; LA64-NEXT: ret +entry: + %va = load <2 x i64>, ptr %a + %vb = load <2 x i64>, ptr %b + %add = add <2 x i64> %va, %vb + %add1 = add <2 x i64> %add, <i64 1, i64 1> + %shr = ashr <2 x i64> %add1, <i64 1, i64 1> + store <2 x i64> %shr, ptr %res + ret void +} + +define void @vavgr_bu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavgr_bu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vavgr.bu $vr0, $vr0, $vr1 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <16 x i8>, ptr %a + %vb = load <16 x i8>, ptr %b + %add = add <16 x i8> %va, %vb + %add1 = add <16 x i8> %add, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1> + %shr = lshr <16 x i8> %add1, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1> + store <16 x i8> %shr, ptr %res + ret void +} + +define void @vavgr_hu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavgr_hu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vavgr.hu $vr0, $vr0, $vr1 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <8 x i16>, ptr %a + %vb = load <8 x i16>, ptr %b + %add = add <8 x i16> %va, %vb + %add1 = add <8 x i16> %add, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + %shr = lshr <8 x i16> %add1, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + store <8 x i16> %shr, ptr %res + ret void +} + +define void @vavgr_wu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavgr_wu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vavgr.wu $vr0, $vr0, $vr1 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <4 x i32>, ptr %a + %vb = load <4 x i32>, ptr %b + %add = add <4 x i32> %va, %vb + %add1 = add <4 x i32> %add, <i32 1, i32 1, i32 1, i32 1> + %shr = lshr <4 x i32> %add1, <i32 1, i32 1, i32 1, i32 1> + store <4 x i32> %shr, ptr %res + ret void +} + +define void @vavgr_du(ptr %res, ptr %a, ptr %b) nounwind { +; LA32-LABEL: vavgr_du: +; LA32: # %bb.0: # %entry +; LA32-NEXT: vld $vr0, $a1, 0 +; LA32-NEXT: vld $vr1, $a2, 0 +; LA32-NEXT: vadd.d $vr0, $vr0, $vr1 +; LA32-NEXT: vaddi.du $vr0, $vr0, 1 +; LA32-NEXT: vsrli.d $vr0, $vr0, 1 +; LA32-NEXT: vst $vr0, $a0, 0 +; LA32-NEXT: ret +; +; LA64-LABEL: vavgr_du: +; LA64: # %bb.0: # %entry +; LA64-NEXT: vld $vr0, $a1, 0 +; LA64-NEXT: vld $vr1, $a2, 0 +; LA64-NEXT: vavgr.du $vr0, $vr0, $vr1 +; LA64-NEXT: vst $vr0, $a0, 0 +; LA64-NEXT: ret +entry: + %va = load <2 x i64>, ptr %a + %vb = load <2 x i64>, ptr %b + %add = add <2 x i64> %va, %vb + %add1 = add <2 x i64> %add, <i64 1, i64 1> + %shr = lshr <2 x i64> %add1, <i64 1, i64 1> + store <2 x i64> %shr, ptr %res + ret void +} diff --git a/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/avgfloor-ceil.ll b/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/avgfloor-ceil.ll new file mode 100644 index 0000000..bb4df64 --- /dev/null +++ b/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/avgfloor-ceil.ll @@ -0,0 +1,379 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lsx < %s | FileCheck %s +; RUN: llc --mtriple=loongarch64 --mattr=+lsx < %s | FileCheck %s + +define void @vavg_b(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavg_b: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vand.v $vr2, $vr0, $vr1 +; CHECK-NEXT: vxor.v $vr0, $vr0, $vr1 +; CHECK-NEXT: vsrai.b $vr0, $vr0, 1 +; CHECK-NEXT: vadd.b $vr0, $vr2, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <16 x i8>, ptr %a + %vb = load <16 x i8>, ptr %b + %ea = sext <16 x i8> %va to <16 x i16> + %eb = sext <16 x i8> %vb to <16 x i16> + %add = add <16 x i16> %ea, %eb + %shr = lshr <16 x i16> %add, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + %r = trunc <16 x i16> %shr to <16 x i8> + store <16 x i8> %r, ptr %res + ret void +} + +define void @vavg_h(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavg_h: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vand.v $vr2, $vr0, $vr1 +; CHECK-NEXT: vxor.v $vr0, $vr0, $vr1 +; CHECK-NEXT: vsrai.h $vr0, $vr0, 1 +; CHECK-NEXT: vadd.h $vr0, $vr2, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <8 x i16>, ptr %a + %vb = load <8 x i16>, ptr %b + %ea = sext <8 x i16> %va to <8 x i32> + %eb = sext <8 x i16> %vb to <8 x i32> + %add = add <8 x i32> %ea, %eb + %shr = lshr <8 x i32> %add, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> + %r = trunc <8 x i32> %shr to <8 x i16> + store <8 x i16> %r, ptr %res + ret void +} + +define void @vavg_w(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavg_w: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vand.v $vr2, $vr0, $vr1 +; CHECK-NEXT: vxor.v $vr0, $vr0, $vr1 +; CHECK-NEXT: vsrai.w $vr0, $vr0, 1 +; CHECK-NEXT: vadd.w $vr0, $vr2, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <4 x i32>, ptr %a + %vb = load <4 x i32>, ptr %b + %ea = sext <4 x i32> %va to <4 x i64> + %eb = sext <4 x i32> %vb to <4 x i64> + %add = add <4 x i64> %ea, %eb + %shr = lshr <4 x i64> %add, <i64 1, i64 1, i64 1, i64 1> + %r = trunc <4 x i64> %shr to <4 x i32> + store <4 x i32> %r, ptr %res + ret void +} + +define void @vavg_d(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavg_d: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vand.v $vr2, $vr0, $vr1 +; CHECK-NEXT: vxor.v $vr0, $vr0, $vr1 +; CHECK-NEXT: vsrai.d $vr0, $vr0, 1 +; CHECK-NEXT: vadd.d $vr0, $vr2, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <2 x i64>, ptr %a + %vb = load <2 x i64>, ptr %b + %ea = sext <2 x i64> %va to <2 x i128> + %eb = sext <2 x i64> %vb to <2 x i128> + %add = add <2 x i128> %ea, %eb + %shr = lshr <2 x i128> %add, <i128 1, i128 1> + %r = trunc <2 x i128> %shr to <2 x i64> + store <2 x i64> %r, ptr %res + ret void +} + +define void @vavg_bu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavg_bu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vand.v $vr2, $vr0, $vr1 +; CHECK-NEXT: vxor.v $vr0, $vr0, $vr1 +; CHECK-NEXT: vsrli.b $vr0, $vr0, 1 +; CHECK-NEXT: vadd.b $vr0, $vr2, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <16 x i8>, ptr %a + %vb = load <16 x i8>, ptr %b + %ea = zext <16 x i8> %va to <16 x i16> + %eb = zext <16 x i8> %vb to <16 x i16> + %add = add <16 x i16> %ea, %eb + %shr = lshr <16 x i16> %add, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + %r = trunc <16 x i16> %shr to <16 x i8> + store <16 x i8> %r, ptr %res + ret void +} + +define void @vavg_hu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavg_hu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vand.v $vr2, $vr0, $vr1 +; CHECK-NEXT: vxor.v $vr0, $vr0, $vr1 +; CHECK-NEXT: vsrli.h $vr0, $vr0, 1 +; CHECK-NEXT: vadd.h $vr0, $vr2, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <8 x i16>, ptr %a + %vb = load <8 x i16>, ptr %b + %ea = zext <8 x i16> %va to <8 x i32> + %eb = zext <8 x i16> %vb to <8 x i32> + %add = add <8 x i32> %ea, %eb + %shr = lshr <8 x i32> %add, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> + %r = trunc <8 x i32> %shr to <8 x i16> + store <8 x i16> %r, ptr %res + ret void +} + +define void @vavg_wu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavg_wu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vand.v $vr2, $vr0, $vr1 +; CHECK-NEXT: vxor.v $vr0, $vr0, $vr1 +; CHECK-NEXT: vsrli.w $vr0, $vr0, 1 +; CHECK-NEXT: vadd.w $vr0, $vr2, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <4 x i32>, ptr %a + %vb = load <4 x i32>, ptr %b + %ea = zext <4 x i32> %va to <4 x i64> + %eb = zext <4 x i32> %vb to <4 x i64> + %add = add <4 x i64> %ea, %eb + %shr = lshr <4 x i64> %add, <i64 1, i64 1, i64 1, i64 1> + %r = trunc <4 x i64> %shr to <4 x i32> + store <4 x i32> %r, ptr %res + ret void +} + +define void @vavg_du(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavg_du: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vand.v $vr2, $vr0, $vr1 +; CHECK-NEXT: vxor.v $vr0, $vr0, $vr1 +; CHECK-NEXT: vsrli.d $vr0, $vr0, 1 +; CHECK-NEXT: vadd.d $vr0, $vr2, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <2 x i64>, ptr %a + %vb = load <2 x i64>, ptr %b + %ea = zext <2 x i64> %va to <2 x i128> + %eb = zext <2 x i64> %vb to <2 x i128> + %add = add <2 x i128> %ea, %eb + %shr = lshr <2 x i128> %add, <i128 1, i128 1> + %r = trunc <2 x i128> %shr to <2 x i64> + store <2 x i64> %r, ptr %res + ret void +} + +define void @vavgr_b(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavgr_b: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vor.v $vr2, $vr0, $vr1 +; CHECK-NEXT: vxor.v $vr0, $vr0, $vr1 +; CHECK-NEXT: vsrai.b $vr0, $vr0, 1 +; CHECK-NEXT: vsub.b $vr0, $vr2, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <16 x i8>, ptr %a + %vb = load <16 x i8>, ptr %b + %ea = sext <16 x i8> %va to <16 x i16> + %eb = sext <16 x i8> %vb to <16 x i16> + %add = add <16 x i16> %ea, %eb + %add1 = add <16 x i16> %add, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + %shr = lshr <16 x i16> %add1, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + %r = trunc <16 x i16> %shr to <16 x i8> + store <16 x i8> %r, ptr %res + ret void +} + +define void @vavgr_h(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavgr_h: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vor.v $vr2, $vr0, $vr1 +; CHECK-NEXT: vxor.v $vr0, $vr0, $vr1 +; CHECK-NEXT: vsrai.h $vr0, $vr0, 1 +; CHECK-NEXT: vsub.h $vr0, $vr2, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <8 x i16>, ptr %a + %vb = load <8 x i16>, ptr %b + %ea = sext <8 x i16> %va to <8 x i32> + %eb = sext <8 x i16> %vb to <8 x i32> + %add = add <8 x i32> %ea, %eb + %add1 = add <8 x i32> %add, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> + %shr = lshr <8 x i32> %add1, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> + %r = trunc <8 x i32> %shr to <8 x i16> + store <8 x i16> %r, ptr %res + ret void +} + +define void @vavgr_w(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavgr_w: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vor.v $vr2, $vr0, $vr1 +; CHECK-NEXT: vxor.v $vr0, $vr0, $vr1 +; CHECK-NEXT: vsrai.w $vr0, $vr0, 1 +; CHECK-NEXT: vsub.w $vr0, $vr2, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <4 x i32>, ptr %a + %vb = load <4 x i32>, ptr %b + %ea = sext <4 x i32> %va to <4 x i64> + %eb = sext <4 x i32> %vb to <4 x i64> + %add = add <4 x i64> %ea, %eb + %add1 = add <4 x i64> %add, <i64 1, i64 1, i64 1, i64 1> + %shr = lshr <4 x i64> %add1, <i64 1, i64 1, i64 1, i64 1> + %r = trunc <4 x i64> %shr to <4 x i32> + store <4 x i32> %r, ptr %res + ret void +} + +define void @vavgr_d(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavgr_d: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vor.v $vr2, $vr0, $vr1 +; CHECK-NEXT: vxor.v $vr0, $vr0, $vr1 +; CHECK-NEXT: vsrai.d $vr0, $vr0, 1 +; CHECK-NEXT: vsub.d $vr0, $vr2, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <2 x i64>, ptr %a + %vb = load <2 x i64>, ptr %b + %ea = sext <2 x i64> %va to <2 x i128> + %eb = sext <2 x i64> %vb to <2 x i128> + %add = add <2 x i128> %ea, %eb + %add1 = add <2 x i128> %add, <i128 1, i128 1> + %shr = lshr <2 x i128> %add1, <i128 1, i128 1> + %r = trunc <2 x i128> %shr to <2 x i64> + store <2 x i64> %r, ptr %res + ret void +} + +define void @vavgr_bu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavgr_bu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vor.v $vr2, $vr0, $vr1 +; CHECK-NEXT: vxor.v $vr0, $vr0, $vr1 +; CHECK-NEXT: vsrli.b $vr0, $vr0, 1 +; CHECK-NEXT: vsub.b $vr0, $vr2, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <16 x i8>, ptr %a + %vb = load <16 x i8>, ptr %b + %ea = zext <16 x i8> %va to <16 x i16> + %eb = zext <16 x i8> %vb to <16 x i16> + %add = add <16 x i16> %ea, %eb + %add1 = add <16 x i16> %add, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + %shr = lshr <16 x i16> %add1, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> + %r = trunc <16 x i16> %shr to <16 x i8> + store <16 x i8> %r, ptr %res + ret void +} + +define void @vavgr_hu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavgr_hu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vor.v $vr2, $vr0, $vr1 +; CHECK-NEXT: vxor.v $vr0, $vr0, $vr1 +; CHECK-NEXT: vsrli.h $vr0, $vr0, 1 +; CHECK-NEXT: vsub.h $vr0, $vr2, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <8 x i16>, ptr %a + %vb = load <8 x i16>, ptr %b + %ea = zext <8 x i16> %va to <8 x i32> + %eb = zext <8 x i16> %vb to <8 x i32> + %add = add <8 x i32> %ea, %eb + %add1 = add <8 x i32> %add, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> + %shr = lshr <8 x i32> %add1, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1> + %r = trunc <8 x i32> %shr to <8 x i16> + store <8 x i16> %r, ptr %res + ret void +} + +define void @vavgr_wu(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavgr_wu: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vor.v $vr2, $vr0, $vr1 +; CHECK-NEXT: vxor.v $vr0, $vr0, $vr1 +; CHECK-NEXT: vsrli.w $vr0, $vr0, 1 +; CHECK-NEXT: vsub.w $vr0, $vr2, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <4 x i32>, ptr %a + %vb = load <4 x i32>, ptr %b + %ea = zext <4 x i32> %va to <4 x i64> + %eb = zext <4 x i32> %vb to <4 x i64> + %add = add <4 x i64> %ea, %eb + %add1 = add <4 x i64> %add, <i64 1, i64 1, i64 1, i64 1> + %shr = lshr <4 x i64> %add1, <i64 1, i64 1, i64 1, i64 1> + %r = trunc <4 x i64> %shr to <4 x i32> + store <4 x i32> %r, ptr %res + ret void +} + +define void @vavgr_du(ptr %res, ptr %a, ptr %b) nounwind { +; CHECK-LABEL: vavgr_du: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vld $vr1, $a2, 0 +; CHECK-NEXT: vor.v $vr2, $vr0, $vr1 +; CHECK-NEXT: vxor.v $vr0, $vr0, $vr1 +; CHECK-NEXT: vsrli.d $vr0, $vr0, 1 +; CHECK-NEXT: vsub.d $vr0, $vr2, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret +entry: + %va = load <2 x i64>, ptr %a + %vb = load <2 x i64>, ptr %b + %ea = zext <2 x i64> %va to <2 x i128> + %eb = zext <2 x i64> %vb to <2 x i128> + %add = add <2 x i128> %ea, %eb + %add1 = add <2 x i128> %add, <i128 1, i128 1> + %shr = lshr <2 x i128> %add1, <i128 1, i128 1> + %r = trunc <2 x i128> %shr to <2 x i64> + store <2 x i64> %r, ptr %res + ret void +} diff --git a/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/flog2.ll b/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/flog2.ll index e5e75ec..87cc7c6 100644 --- a/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/flog2.ll +++ b/llvm/test/CodeGen/LoongArch/lsx/ir-instruction/flog2.ll @@ -1,98 +1,17 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 -; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lsx < %s | FileCheck %s --check-prefix=LA32 -; RUN: llc --mtriple=loongarch64 --mattr=+lsx < %s | FileCheck %s --check-prefix=LA64 +; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lsx < %s | FileCheck %s +; RUN: llc --mtriple=loongarch64 --mattr=+lsx < %s | FileCheck %s declare <4 x float> @llvm.log2.v4f32(<4 x float>) declare <2 x double> @llvm.log2.v2f64(<2 x double>) define void @flog2_v4f32(ptr %res, ptr %a) nounwind { -; LA32-LABEL: flog2_v4f32: -; LA32: # %bb.0: # %entry -; LA32-NEXT: addi.w $sp, $sp, -48 -; LA32-NEXT: st.w $ra, $sp, 44 # 4-byte Folded Spill -; LA32-NEXT: st.w $fp, $sp, 40 # 4-byte Folded Spill -; LA32-NEXT: vld $vr0, $a1, 0 -; LA32-NEXT: vst $vr0, $sp, 16 # 16-byte Folded Spill -; LA32-NEXT: move $fp, $a0 -; LA32-NEXT: vreplvei.w $vr0, $vr0, 1 -; LA32-NEXT: # kill: def $f0 killed $f0 killed $vr0 -; LA32-NEXT: bl log2f -; LA32-NEXT: # kill: def $f0 killed $f0 def $vr0 -; LA32-NEXT: vst $vr0, $sp, 0 # 16-byte Folded Spill -; LA32-NEXT: vld $vr0, $sp, 16 # 16-byte Folded Reload -; LA32-NEXT: vreplvei.w $vr0, $vr0, 0 -; LA32-NEXT: # kill: def $f0 killed $f0 killed $vr0 -; LA32-NEXT: bl log2f -; LA32-NEXT: # kill: def $f0 killed $f0 def $vr0 -; LA32-NEXT: vld $vr1, $sp, 0 # 16-byte Folded Reload -; LA32-NEXT: vextrins.w $vr0, $vr1, 16 -; LA32-NEXT: vst $vr0, $sp, 0 # 16-byte Folded Spill -; LA32-NEXT: vld $vr0, $sp, 16 # 16-byte Folded Reload -; LA32-NEXT: vreplvei.w $vr0, $vr0, 2 -; LA32-NEXT: # kill: def $f0 killed $f0 killed $vr0 -; LA32-NEXT: bl log2f -; LA32-NEXT: # kill: def $f0 killed $f0 def $vr0 -; LA32-NEXT: vld $vr1, $sp, 0 # 16-byte Folded Reload -; LA32-NEXT: vextrins.w $vr1, $vr0, 32 -; LA32-NEXT: vst $vr1, $sp, 0 # 16-byte Folded Spill -; LA32-NEXT: vld $vr0, $sp, 16 # 16-byte Folded Reload -; LA32-NEXT: vreplvei.w $vr0, $vr0, 3 -; LA32-NEXT: # kill: def $f0 killed $f0 killed $vr0 -; LA32-NEXT: bl log2f -; LA32-NEXT: # kill: def $f0 killed $f0 def $vr0 -; LA32-NEXT: vld $vr1, $sp, 0 # 16-byte Folded Reload -; LA32-NEXT: vextrins.w $vr1, $vr0, 48 -; LA32-NEXT: vst $vr1, $fp, 0 -; LA32-NEXT: ld.w $fp, $sp, 40 # 4-byte Folded Reload -; LA32-NEXT: ld.w $ra, $sp, 44 # 4-byte Folded Reload -; LA32-NEXT: addi.w $sp, $sp, 48 -; LA32-NEXT: ret -; -; LA64-LABEL: flog2_v4f32: -; LA64: # %bb.0: # %entry -; LA64-NEXT: addi.d $sp, $sp, -48 -; LA64-NEXT: st.d $ra, $sp, 40 # 8-byte Folded Spill -; LA64-NEXT: st.d $fp, $sp, 32 # 8-byte Folded Spill -; LA64-NEXT: vld $vr0, $a1, 0 -; LA64-NEXT: vst $vr0, $sp, 16 # 16-byte Folded Spill -; LA64-NEXT: move $fp, $a0 -; LA64-NEXT: vreplvei.w $vr0, $vr0, 1 -; LA64-NEXT: # kill: def $f0 killed $f0 killed $vr0 -; LA64-NEXT: pcaddu18i $ra, %call36(log2f) -; LA64-NEXT: jirl $ra, $ra, 0 -; LA64-NEXT: # kill: def $f0 killed $f0 def $vr0 -; LA64-NEXT: vst $vr0, $sp, 0 # 16-byte Folded Spill -; LA64-NEXT: vld $vr0, $sp, 16 # 16-byte Folded Reload -; LA64-NEXT: vreplvei.w $vr0, $vr0, 0 -; LA64-NEXT: # kill: def $f0 killed $f0 killed $vr0 -; LA64-NEXT: pcaddu18i $ra, %call36(log2f) -; LA64-NEXT: jirl $ra, $ra, 0 -; LA64-NEXT: # kill: def $f0 killed $f0 def $vr0 -; LA64-NEXT: vld $vr1, $sp, 0 # 16-byte Folded Reload -; LA64-NEXT: vextrins.w $vr0, $vr1, 16 -; LA64-NEXT: vst $vr0, $sp, 0 # 16-byte Folded Spill -; LA64-NEXT: vld $vr0, $sp, 16 # 16-byte Folded Reload -; LA64-NEXT: vreplvei.w $vr0, $vr0, 2 -; LA64-NEXT: # kill: def $f0 killed $f0 killed $vr0 -; LA64-NEXT: pcaddu18i $ra, %call36(log2f) -; LA64-NEXT: jirl $ra, $ra, 0 -; LA64-NEXT: # kill: def $f0 killed $f0 def $vr0 -; LA64-NEXT: vld $vr1, $sp, 0 # 16-byte Folded Reload -; LA64-NEXT: vextrins.w $vr1, $vr0, 32 -; LA64-NEXT: vst $vr1, $sp, 0 # 16-byte Folded Spill -; LA64-NEXT: vld $vr0, $sp, 16 # 16-byte Folded Reload -; LA64-NEXT: vreplvei.w $vr0, $vr0, 3 -; LA64-NEXT: # kill: def $f0 killed $f0 killed $vr0 -; LA64-NEXT: pcaddu18i $ra, %call36(log2f) -; LA64-NEXT: jirl $ra, $ra, 0 -; LA64-NEXT: # kill: def $f0 killed $f0 def $vr0 -; LA64-NEXT: vld $vr1, $sp, 0 # 16-byte Folded Reload -; LA64-NEXT: vextrins.w $vr1, $vr0, 48 -; LA64-NEXT: vst $vr1, $fp, 0 -; LA64-NEXT: ld.d $fp, $sp, 32 # 8-byte Folded Reload -; LA64-NEXT: ld.d $ra, $sp, 40 # 8-byte Folded Reload -; LA64-NEXT: addi.d $sp, $sp, 48 -; LA64-NEXT: ret +; CHECK-LABEL: flog2_v4f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vflogb.s $vr0, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret entry: %v = load <4 x float>, ptr %a %r = call <4 x float> @llvm.log2.v4f32(<4 x float> %v) @@ -101,59 +20,12 @@ entry: } define void @flog2_v2f64(ptr %res, ptr %a) nounwind { -; LA32-LABEL: flog2_v2f64: -; LA32: # %bb.0: # %entry -; LA32-NEXT: addi.w $sp, $sp, -48 -; LA32-NEXT: st.w $ra, $sp, 44 # 4-byte Folded Spill -; LA32-NEXT: st.w $fp, $sp, 40 # 4-byte Folded Spill -; LA32-NEXT: vld $vr0, $a1, 0 -; LA32-NEXT: vst $vr0, $sp, 0 # 16-byte Folded Spill -; LA32-NEXT: move $fp, $a0 -; LA32-NEXT: vreplvei.d $vr0, $vr0, 1 -; LA32-NEXT: # kill: def $f0_64 killed $f0_64 killed $vr0 -; LA32-NEXT: bl log2 -; LA32-NEXT: # kill: def $f0_64 killed $f0_64 def $vr0 -; LA32-NEXT: vst $vr0, $sp, 16 # 16-byte Folded Spill -; LA32-NEXT: vld $vr0, $sp, 0 # 16-byte Folded Reload -; LA32-NEXT: vreplvei.d $vr0, $vr0, 0 -; LA32-NEXT: # kill: def $f0_64 killed $f0_64 killed $vr0 -; LA32-NEXT: bl log2 -; LA32-NEXT: # kill: def $f0_64 killed $f0_64 def $vr0 -; LA32-NEXT: vld $vr1, $sp, 16 # 16-byte Folded Reload -; LA32-NEXT: vextrins.d $vr0, $vr1, 16 -; LA32-NEXT: vst $vr0, $fp, 0 -; LA32-NEXT: ld.w $fp, $sp, 40 # 4-byte Folded Reload -; LA32-NEXT: ld.w $ra, $sp, 44 # 4-byte Folded Reload -; LA32-NEXT: addi.w $sp, $sp, 48 -; LA32-NEXT: ret -; -; LA64-LABEL: flog2_v2f64: -; LA64: # %bb.0: # %entry -; LA64-NEXT: addi.d $sp, $sp, -48 -; LA64-NEXT: st.d $ra, $sp, 40 # 8-byte Folded Spill -; LA64-NEXT: st.d $fp, $sp, 32 # 8-byte Folded Spill -; LA64-NEXT: vld $vr0, $a1, 0 -; LA64-NEXT: vst $vr0, $sp, 0 # 16-byte Folded Spill -; LA64-NEXT: move $fp, $a0 -; LA64-NEXT: vreplvei.d $vr0, $vr0, 1 -; LA64-NEXT: # kill: def $f0_64 killed $f0_64 killed $vr0 -; LA64-NEXT: pcaddu18i $ra, %call36(log2) -; LA64-NEXT: jirl $ra, $ra, 0 -; LA64-NEXT: # kill: def $f0_64 killed $f0_64 def $vr0 -; LA64-NEXT: vst $vr0, $sp, 16 # 16-byte Folded Spill -; LA64-NEXT: vld $vr0, $sp, 0 # 16-byte Folded Reload -; LA64-NEXT: vreplvei.d $vr0, $vr0, 0 -; LA64-NEXT: # kill: def $f0_64 killed $f0_64 killed $vr0 -; LA64-NEXT: pcaddu18i $ra, %call36(log2) -; LA64-NEXT: jirl $ra, $ra, 0 -; LA64-NEXT: # kill: def $f0_64 killed $f0_64 def $vr0 -; LA64-NEXT: vld $vr1, $sp, 16 # 16-byte Folded Reload -; LA64-NEXT: vextrins.d $vr0, $vr1, 16 -; LA64-NEXT: vst $vr0, $fp, 0 -; LA64-NEXT: ld.d $fp, $sp, 32 # 8-byte Folded Reload -; LA64-NEXT: ld.d $ra, $sp, 40 # 8-byte Folded Reload -; LA64-NEXT: addi.d $sp, $sp, 48 -; LA64-NEXT: ret +; CHECK-LABEL: flog2_v2f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vld $vr0, $a1, 0 +; CHECK-NEXT: vflogb.d $vr0, $vr0 +; CHECK-NEXT: vst $vr0, $a0, 0 +; CHECK-NEXT: ret entry: %v = load <2 x double>, ptr %a %r = call <2 x double> @llvm.log2.v2f64(<2 x double> %v) diff --git a/llvm/test/CodeGen/LoongArch/sink-fold-addi.ll b/llvm/test/CodeGen/LoongArch/sink-fold-addi.ll new file mode 100644 index 0000000..9a806a1 --- /dev/null +++ b/llvm/test/CodeGen/LoongArch/sink-fold-addi.ll @@ -0,0 +1,758 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc --mtriple=loongarch32 --mattr=+32s,+lasx --verify-machineinstrs < %s \ +; RUN: | FileCheck --check-prefix=LA32 %s +; RUN: llc --mtriple=loongarch64 --mattr=+lasx --verify-machineinstrs < %s \ +; RUN: | FileCheck --check-prefix=LA64 %s + +%struct.S = type { i64, i64, i8 } +%struct.F = type { float, double, float } +%struct.V = type { <4 x i32>, <4 x i32>, <16 x i16> } + +define void @sink_fold_i64(i64 %k, i64 %n, ptr %a) nounwind { +; LA32-LABEL: sink_fold_i64: +; LA32: # %bb.0: # %entry +; LA32-NEXT: addi.w $sp, $sp, -48 +; LA32-NEXT: st.w $ra, $sp, 44 # 4-byte Folded Spill +; LA32-NEXT: st.w $fp, $sp, 40 # 4-byte Folded Spill +; LA32-NEXT: st.w $s0, $sp, 36 # 4-byte Folded Spill +; LA32-NEXT: st.w $s1, $sp, 32 # 4-byte Folded Spill +; LA32-NEXT: st.w $s2, $sp, 28 # 4-byte Folded Spill +; LA32-NEXT: st.w $s3, $sp, 24 # 4-byte Folded Spill +; LA32-NEXT: st.w $s4, $sp, 20 # 4-byte Folded Spill +; LA32-NEXT: st.w $s5, $sp, 16 # 4-byte Folded Spill +; LA32-NEXT: st.w $s6, $sp, 12 # 4-byte Folded Spill +; LA32-NEXT: move $s0, $a3 +; LA32-NEXT: move $s1, $a2 +; LA32-NEXT: slli.w $a1, $a0, 4 +; LA32-NEXT: alsl.w $a0, $a0, $a1, 3 +; LA32-NEXT: add.w $a0, $a4, $a0 +; LA32-NEXT: sltui $a1, $a3, 1 +; LA32-NEXT: slti $a2, $a3, 0 +; LA32-NEXT: masknez $a2, $a2, $a1 +; LA32-NEXT: sltui $a3, $s1, 1 +; LA32-NEXT: maskeqz $a1, $a3, $a1 +; LA32-NEXT: or $a1, $a1, $a2 +; LA32-NEXT: addi.w $s2, $a0, 8 +; LA32-NEXT: bnez $a1, .LBB0_3 +; LA32-NEXT: # %bb.1: # %for.body.preheader +; LA32-NEXT: move $fp, $a4 +; LA32-NEXT: move $s4, $zero +; LA32-NEXT: move $s5, $zero +; LA32-NEXT: move $s3, $zero +; LA32-NEXT: move $s6, $zero +; LA32-NEXT: .p2align 4, , 16 +; LA32-NEXT: .LBB0_2: # %for.body +; LA32-NEXT: # =>This Inner Loop Header: Depth=1 +; LA32-NEXT: move $a0, $fp +; LA32-NEXT: bl f +; LA32-NEXT: ld.w $a0, $s2, 4 +; LA32-NEXT: ld.w $a1, $s2, 0 +; LA32-NEXT: add.w $a0, $a0, $s6 +; LA32-NEXT: add.w $s3, $a1, $s3 +; LA32-NEXT: sltu $a1, $s3, $a1 +; LA32-NEXT: addi.w $s4, $s4, 1 +; LA32-NEXT: sltui $a2, $s4, 1 +; LA32-NEXT: add.w $s5, $s5, $a2 +; LA32-NEXT: xor $a2, $s4, $s1 +; LA32-NEXT: xor $a3, $s5, $s0 +; LA32-NEXT: or $a2, $a2, $a3 +; LA32-NEXT: add.w $s6, $a0, $a1 +; LA32-NEXT: bnez $a2, .LBB0_2 +; LA32-NEXT: b .LBB0_4 +; LA32-NEXT: .LBB0_3: +; LA32-NEXT: move $s3, $zero +; LA32-NEXT: move $s6, $zero +; LA32-NEXT: .LBB0_4: # %for.cond.cleanup +; LA32-NEXT: st.w $s3, $s2, 0 +; LA32-NEXT: st.w $s6, $s2, 4 +; LA32-NEXT: ld.w $s6, $sp, 12 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s5, $sp, 16 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s4, $sp, 20 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s3, $sp, 24 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s2, $sp, 28 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s1, $sp, 32 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s0, $sp, 36 # 4-byte Folded Reload +; LA32-NEXT: ld.w $fp, $sp, 40 # 4-byte Folded Reload +; LA32-NEXT: ld.w $ra, $sp, 44 # 4-byte Folded Reload +; LA32-NEXT: addi.w $sp, $sp, 48 +; LA32-NEXT: ret +; +; LA64-LABEL: sink_fold_i64: +; LA64: # %bb.0: # %entry +; LA64-NEXT: addi.d $sp, $sp, -48 +; LA64-NEXT: st.d $ra, $sp, 40 # 8-byte Folded Spill +; LA64-NEXT: st.d $fp, $sp, 32 # 8-byte Folded Spill +; LA64-NEXT: st.d $s0, $sp, 24 # 8-byte Folded Spill +; LA64-NEXT: st.d $s1, $sp, 16 # 8-byte Folded Spill +; LA64-NEXT: st.d $s2, $sp, 8 # 8-byte Folded Spill +; LA64-NEXT: move $s0, $a1 +; LA64-NEXT: slli.d $a1, $a0, 4 +; LA64-NEXT: alsl.d $a0, $a0, $a1, 3 +; LA64-NEXT: add.d $a0, $a2, $a0 +; LA64-NEXT: addi.d $s1, $a0, 8 +; LA64-NEXT: blez $s0, .LBB0_3 +; LA64-NEXT: # %bb.1: # %for.body.preheader +; LA64-NEXT: move $fp, $a2 +; LA64-NEXT: move $s2, $zero +; LA64-NEXT: .p2align 4, , 16 +; LA64-NEXT: .LBB0_2: # %for.body +; LA64-NEXT: # =>This Inner Loop Header: Depth=1 +; LA64-NEXT: move $a0, $fp +; LA64-NEXT: pcaddu18i $ra, %call36(f) +; LA64-NEXT: jirl $ra, $ra, 0 +; LA64-NEXT: ld.d $a0, $s1, 0 +; LA64-NEXT: addi.d $s0, $s0, -1 +; LA64-NEXT: add.d $s2, $a0, $s2 +; LA64-NEXT: bnez $s0, .LBB0_2 +; LA64-NEXT: b .LBB0_4 +; LA64-NEXT: .LBB0_3: +; LA64-NEXT: move $s2, $zero +; LA64-NEXT: .LBB0_4: # %for.cond.cleanup +; LA64-NEXT: st.d $s2, $s1, 0 +; LA64-NEXT: ld.d $s2, $sp, 8 # 8-byte Folded Reload +; LA64-NEXT: ld.d $s1, $sp, 16 # 8-byte Folded Reload +; LA64-NEXT: ld.d $s0, $sp, 24 # 8-byte Folded Reload +; LA64-NEXT: ld.d $fp, $sp, 32 # 8-byte Folded Reload +; LA64-NEXT: ld.d $ra, $sp, 40 # 8-byte Folded Reload +; LA64-NEXT: addi.d $sp, $sp, 48 +; LA64-NEXT: ret +entry: + %y = getelementptr inbounds %struct.S, ptr %a, i64 %k, i32 1 + %cmp4 = icmp sgt i64 %n, 0 + br i1 %cmp4, label %for.body, label %for.cond.cleanup + +for.body: ; preds = %entry, %for.body + %i.06 = phi i64 [ 0, %entry ], [ %inc, %for.body ] + %s.05 = phi i64 [ 0, %entry ], [ %add, %for.body ] + call void @f(ptr %a) + %0 = load i64, ptr %y + %add = add nsw i64 %0, %s.05 + %inc = add nuw nsw i64 %i.06, 1 + %exitcond.not = icmp eq i64 %inc, %n + br i1 %exitcond.not, label %for.cond.cleanup, label %for.body + +for.cond.cleanup: ; preds = %for.body, %entry + %s.0.lcssa = phi i64 [ 0, %entry ], [ %add, %for.body ] + store i64 %s.0.lcssa, ptr %y + ret void +} + +define void @sink_fold_f32(i64 %k, i64 %n, ptr %a) nounwind { +; LA32-LABEL: sink_fold_f32: +; LA32: # %bb.0: # %entry +; LA32-NEXT: addi.w $sp, $sp, -48 +; LA32-NEXT: st.w $ra, $sp, 44 # 4-byte Folded Spill +; LA32-NEXT: st.w $fp, $sp, 40 # 4-byte Folded Spill +; LA32-NEXT: st.w $s0, $sp, 36 # 4-byte Folded Spill +; LA32-NEXT: st.w $s1, $sp, 32 # 4-byte Folded Spill +; LA32-NEXT: st.w $s2, $sp, 28 # 4-byte Folded Spill +; LA32-NEXT: st.w $s3, $sp, 24 # 4-byte Folded Spill +; LA32-NEXT: st.w $s4, $sp, 20 # 4-byte Folded Spill +; LA32-NEXT: fst.d $fs0, $sp, 8 # 8-byte Folded Spill +; LA32-NEXT: move $s0, $a3 +; LA32-NEXT: move $s1, $a2 +; LA32-NEXT: slli.w $a1, $a0, 4 +; LA32-NEXT: alsl.w $a0, $a0, $a1, 3 +; LA32-NEXT: add.w $a0, $a4, $a0 +; LA32-NEXT: sltui $a1, $a3, 1 +; LA32-NEXT: slti $a2, $a3, 0 +; LA32-NEXT: masknez $a2, $a2, $a1 +; LA32-NEXT: sltui $a3, $s1, 1 +; LA32-NEXT: maskeqz $a1, $a3, $a1 +; LA32-NEXT: or $a1, $a1, $a2 +; LA32-NEXT: addi.w $s2, $a0, 16 +; LA32-NEXT: bnez $a1, .LBB1_3 +; LA32-NEXT: # %bb.1: # %for.body.preheader +; LA32-NEXT: move $fp, $a4 +; LA32-NEXT: move $s3, $zero +; LA32-NEXT: move $s4, $zero +; LA32-NEXT: movgr2fr.w $fs0, $zero +; LA32-NEXT: .p2align 4, , 16 +; LA32-NEXT: .LBB1_2: # %for.body +; LA32-NEXT: # =>This Inner Loop Header: Depth=1 +; LA32-NEXT: move $a0, $fp +; LA32-NEXT: bl f +; LA32-NEXT: fld.s $fa0, $s2, 0 +; LA32-NEXT: addi.w $s3, $s3, 1 +; LA32-NEXT: sltui $a0, $s3, 1 +; LA32-NEXT: add.w $s4, $s4, $a0 +; LA32-NEXT: xor $a0, $s3, $s1 +; LA32-NEXT: xor $a1, $s4, $s0 +; LA32-NEXT: or $a0, $a0, $a1 +; LA32-NEXT: fadd.s $fs0, $fa0, $fs0 +; LA32-NEXT: bnez $a0, .LBB1_2 +; LA32-NEXT: b .LBB1_4 +; LA32-NEXT: .LBB1_3: +; LA32-NEXT: movgr2fr.w $fs0, $zero +; LA32-NEXT: .LBB1_4: # %for.cond.cleanup +; LA32-NEXT: fst.s $fs0, $s2, 0 +; LA32-NEXT: fld.d $fs0, $sp, 8 # 8-byte Folded Reload +; LA32-NEXT: ld.w $s4, $sp, 20 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s3, $sp, 24 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s2, $sp, 28 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s1, $sp, 32 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s0, $sp, 36 # 4-byte Folded Reload +; LA32-NEXT: ld.w $fp, $sp, 40 # 4-byte Folded Reload +; LA32-NEXT: ld.w $ra, $sp, 44 # 4-byte Folded Reload +; LA32-NEXT: addi.w $sp, $sp, 48 +; LA32-NEXT: ret +; +; LA64-LABEL: sink_fold_f32: +; LA64: # %bb.0: # %entry +; LA64-NEXT: addi.d $sp, $sp, -48 +; LA64-NEXT: st.d $ra, $sp, 40 # 8-byte Folded Spill +; LA64-NEXT: st.d $fp, $sp, 32 # 8-byte Folded Spill +; LA64-NEXT: st.d $s0, $sp, 24 # 8-byte Folded Spill +; LA64-NEXT: st.d $s1, $sp, 16 # 8-byte Folded Spill +; LA64-NEXT: fst.d $fs0, $sp, 8 # 8-byte Folded Spill +; LA64-NEXT: move $s0, $a1 +; LA64-NEXT: slli.d $a1, $a0, 4 +; LA64-NEXT: alsl.d $a0, $a0, $a1, 3 +; LA64-NEXT: add.d $a0, $a2, $a0 +; LA64-NEXT: addi.d $s1, $a0, 16 +; LA64-NEXT: blez $s0, .LBB1_3 +; LA64-NEXT: # %bb.1: # %for.body.preheader +; LA64-NEXT: move $fp, $a2 +; LA64-NEXT: movgr2fr.w $fs0, $zero +; LA64-NEXT: .p2align 4, , 16 +; LA64-NEXT: .LBB1_2: # %for.body +; LA64-NEXT: # =>This Inner Loop Header: Depth=1 +; LA64-NEXT: move $a0, $fp +; LA64-NEXT: pcaddu18i $ra, %call36(f) +; LA64-NEXT: jirl $ra, $ra, 0 +; LA64-NEXT: fld.s $fa0, $s1, 0 +; LA64-NEXT: addi.d $s0, $s0, -1 +; LA64-NEXT: fadd.s $fs0, $fa0, $fs0 +; LA64-NEXT: bnez $s0, .LBB1_2 +; LA64-NEXT: b .LBB1_4 +; LA64-NEXT: .LBB1_3: +; LA64-NEXT: movgr2fr.w $fs0, $zero +; LA64-NEXT: .LBB1_4: # %for.cond.cleanup +; LA64-NEXT: fst.s $fs0, $s1, 0 +; LA64-NEXT: fld.d $fs0, $sp, 8 # 8-byte Folded Reload +; LA64-NEXT: ld.d $s1, $sp, 16 # 8-byte Folded Reload +; LA64-NEXT: ld.d $s0, $sp, 24 # 8-byte Folded Reload +; LA64-NEXT: ld.d $fp, $sp, 32 # 8-byte Folded Reload +; LA64-NEXT: ld.d $ra, $sp, 40 # 8-byte Folded Reload +; LA64-NEXT: addi.d $sp, $sp, 48 +; LA64-NEXT: ret +entry: + %y = getelementptr inbounds %struct.F, ptr %a, i64 %k, i32 2 + %cmp4 = icmp sgt i64 %n, 0 + br i1 %cmp4, label %for.body, label %for.cond.cleanup + +for.body: ; preds = %entry, %for.body + %i.06 = phi i64 [ 0, %entry ], [ %inc, %for.body ] + %s.05 = phi float [ 0.0, %entry ], [ %add, %for.body ] + call void @f(ptr %a) + %0 = load float, ptr %y + %add = fadd float %0, %s.05 + %inc = add nuw nsw i64 %i.06, 1 + %exitcond.not = icmp eq i64 %inc, %n + br i1 %exitcond.not, label %for.cond.cleanup, label %for.body + +for.cond.cleanup: ; preds = %for.body, %entry + %s.0.lcssa = phi float [ 0.0, %entry ], [ %add, %for.body ] + store float %s.0.lcssa, ptr %y + ret void +} + +define void @sink_fold_v4i32(i64 %k, i64 %n, ptr %a) nounwind { +; LA32-LABEL: sink_fold_v4i32: +; LA32: # %bb.0: # %entry +; LA32-NEXT: addi.w $sp, $sp, -48 +; LA32-NEXT: st.w $ra, $sp, 44 # 4-byte Folded Spill +; LA32-NEXT: st.w $fp, $sp, 40 # 4-byte Folded Spill +; LA32-NEXT: st.w $s0, $sp, 36 # 4-byte Folded Spill +; LA32-NEXT: st.w $s1, $sp, 32 # 4-byte Folded Spill +; LA32-NEXT: st.w $s2, $sp, 28 # 4-byte Folded Spill +; LA32-NEXT: st.w $s3, $sp, 24 # 4-byte Folded Spill +; LA32-NEXT: st.w $s4, $sp, 20 # 4-byte Folded Spill +; LA32-NEXT: move $s0, $a3 +; LA32-NEXT: move $s1, $a2 +; LA32-NEXT: slli.w $a0, $a0, 6 +; LA32-NEXT: add.w $a0, $a4, $a0 +; LA32-NEXT: sltui $a1, $a3, 1 +; LA32-NEXT: slti $a2, $a3, 0 +; LA32-NEXT: masknez $a2, $a2, $a1 +; LA32-NEXT: sltui $a3, $s1, 1 +; LA32-NEXT: maskeqz $a1, $a3, $a1 +; LA32-NEXT: or $a1, $a1, $a2 +; LA32-NEXT: addi.w $s2, $a0, 16 +; LA32-NEXT: bnez $a1, .LBB2_3 +; LA32-NEXT: # %bb.1: # %for.body.preheader +; LA32-NEXT: move $fp, $a4 +; LA32-NEXT: move $s3, $zero +; LA32-NEXT: move $s4, $zero +; LA32-NEXT: vrepli.b $vr0, 0 +; LA32-NEXT: .p2align 4, , 16 +; LA32-NEXT: .LBB2_2: # %for.body +; LA32-NEXT: # =>This Inner Loop Header: Depth=1 +; LA32-NEXT: vst $vr0, $sp, 0 # 16-byte Folded Spill +; LA32-NEXT: move $a0, $fp +; LA32-NEXT: bl f +; LA32-NEXT: vld $vr0, $s2, 0 +; LA32-NEXT: addi.w $s3, $s3, 1 +; LA32-NEXT: sltui $a0, $s3, 1 +; LA32-NEXT: add.w $s4, $s4, $a0 +; LA32-NEXT: xor $a0, $s3, $s1 +; LA32-NEXT: xor $a1, $s4, $s0 +; LA32-NEXT: or $a0, $a0, $a1 +; LA32-NEXT: vld $vr1, $sp, 0 # 16-byte Folded Reload +; LA32-NEXT: vadd.w $vr1, $vr0, $vr1 +; LA32-NEXT: vst $vr1, $sp, 0 # 16-byte Folded Spill +; LA32-NEXT: vld $vr0, $sp, 0 # 16-byte Folded Reload +; LA32-NEXT: bnez $a0, .LBB2_2 +; LA32-NEXT: b .LBB2_4 +; LA32-NEXT: .LBB2_3: +; LA32-NEXT: vrepli.b $vr0, 0 +; LA32-NEXT: .LBB2_4: # %for.cond.cleanup +; LA32-NEXT: vst $vr0, $s2, 0 +; LA32-NEXT: ld.w $s4, $sp, 20 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s3, $sp, 24 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s2, $sp, 28 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s1, $sp, 32 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s0, $sp, 36 # 4-byte Folded Reload +; LA32-NEXT: ld.w $fp, $sp, 40 # 4-byte Folded Reload +; LA32-NEXT: ld.w $ra, $sp, 44 # 4-byte Folded Reload +; LA32-NEXT: addi.w $sp, $sp, 48 +; LA32-NEXT: ret +; +; LA64-LABEL: sink_fold_v4i32: +; LA64: # %bb.0: # %entry +; LA64-NEXT: addi.d $sp, $sp, -48 +; LA64-NEXT: st.d $ra, $sp, 40 # 8-byte Folded Spill +; LA64-NEXT: st.d $fp, $sp, 32 # 8-byte Folded Spill +; LA64-NEXT: st.d $s0, $sp, 24 # 8-byte Folded Spill +; LA64-NEXT: st.d $s1, $sp, 16 # 8-byte Folded Spill +; LA64-NEXT: slli.d $a0, $a0, 6 +; LA64-NEXT: add.d $a0, $a2, $a0 +; LA64-NEXT: addi.d $s1, $a0, 16 +; LA64-NEXT: blez $a1, .LBB2_3 +; LA64-NEXT: # %bb.1: # %for.body.preheader +; LA64-NEXT: move $fp, $a2 +; LA64-NEXT: move $s0, $a1 +; LA64-NEXT: vrepli.b $vr0, 0 +; LA64-NEXT: .p2align 4, , 16 +; LA64-NEXT: .LBB2_2: # %for.body +; LA64-NEXT: # =>This Inner Loop Header: Depth=1 +; LA64-NEXT: vst $vr0, $sp, 0 # 16-byte Folded Spill +; LA64-NEXT: move $a0, $fp +; LA64-NEXT: pcaddu18i $ra, %call36(f) +; LA64-NEXT: jirl $ra, $ra, 0 +; LA64-NEXT: vld $vr0, $s1, 0 +; LA64-NEXT: addi.d $s0, $s0, -1 +; LA64-NEXT: vld $vr1, $sp, 0 # 16-byte Folded Reload +; LA64-NEXT: vadd.w $vr1, $vr0, $vr1 +; LA64-NEXT: vst $vr1, $sp, 0 # 16-byte Folded Spill +; LA64-NEXT: vld $vr0, $sp, 0 # 16-byte Folded Reload +; LA64-NEXT: bnez $s0, .LBB2_2 +; LA64-NEXT: b .LBB2_4 +; LA64-NEXT: .LBB2_3: +; LA64-NEXT: vrepli.b $vr0, 0 +; LA64-NEXT: .LBB2_4: # %for.cond.cleanup +; LA64-NEXT: vst $vr0, $s1, 0 +; LA64-NEXT: ld.d $s1, $sp, 16 # 8-byte Folded Reload +; LA64-NEXT: ld.d $s0, $sp, 24 # 8-byte Folded Reload +; LA64-NEXT: ld.d $fp, $sp, 32 # 8-byte Folded Reload +; LA64-NEXT: ld.d $ra, $sp, 40 # 8-byte Folded Reload +; LA64-NEXT: addi.d $sp, $sp, 48 +; LA64-NEXT: ret +entry: + %y = getelementptr inbounds %struct.V, ptr %a, i64 %k, i32 1 + %cmp = icmp sgt i64 %n, 0 + br i1 %cmp, label %for.body, label %for.cond.cleanup + +for.body: ; preds = %entry, %for.body + %i.0 = phi i64 [ 0, %entry ], [ %inc, %for.body ] + %sum.0 = phi <4 x i32> [ zeroinitializer, %entry ], [ %addv, %for.body ] + call void @f(ptr %a) + %v = load <4 x i32>, ptr %y + %addv = add <4 x i32> %v, %sum.0 + %inc = add nuw nsw i64 %i.0, 1 + %exitcond = icmp eq i64 %inc, %n + br i1 %exitcond, label %for.cond.cleanup, label %for.body + +for.cond.cleanup: ; preds = %for.body, %entry + %sum.lcssa = phi <4 x i32> [ zeroinitializer, %entry ], [ %addv, %for.body ] + store <4 x i32> %sum.lcssa, ptr %y + ret void +} + +define void @sink_fold_v16i16(i64 %k, i64 %n, ptr %a) nounwind { +; LA32-LABEL: sink_fold_v16i16: +; LA32: # %bb.0: # %entry +; LA32-NEXT: addi.w $sp, $sp, -80 +; LA32-NEXT: st.w $ra, $sp, 76 # 4-byte Folded Spill +; LA32-NEXT: st.w $fp, $sp, 72 # 4-byte Folded Spill +; LA32-NEXT: st.w $s0, $sp, 68 # 4-byte Folded Spill +; LA32-NEXT: st.w $s1, $sp, 64 # 4-byte Folded Spill +; LA32-NEXT: st.w $s2, $sp, 60 # 4-byte Folded Spill +; LA32-NEXT: st.w $s3, $sp, 56 # 4-byte Folded Spill +; LA32-NEXT: st.w $s4, $sp, 52 # 4-byte Folded Spill +; LA32-NEXT: move $s0, $a3 +; LA32-NEXT: move $s1, $a2 +; LA32-NEXT: slli.w $a0, $a0, 6 +; LA32-NEXT: add.w $a0, $a4, $a0 +; LA32-NEXT: sltui $a1, $a3, 1 +; LA32-NEXT: slti $a2, $a3, 0 +; LA32-NEXT: masknez $a2, $a2, $a1 +; LA32-NEXT: sltui $a3, $s1, 1 +; LA32-NEXT: maskeqz $a1, $a3, $a1 +; LA32-NEXT: or $a1, $a1, $a2 +; LA32-NEXT: addi.w $s2, $a0, 32 +; LA32-NEXT: bnez $a1, .LBB3_3 +; LA32-NEXT: # %bb.1: # %for.body.preheader +; LA32-NEXT: move $fp, $a4 +; LA32-NEXT: move $s3, $zero +; LA32-NEXT: move $s4, $zero +; LA32-NEXT: xvrepli.b $xr0, 0 +; LA32-NEXT: .p2align 4, , 16 +; LA32-NEXT: .LBB3_2: # %for.body +; LA32-NEXT: # =>This Inner Loop Header: Depth=1 +; LA32-NEXT: xvst $xr0, $sp, 16 # 32-byte Folded Spill +; LA32-NEXT: move $a0, $fp +; LA32-NEXT: bl f +; LA32-NEXT: xvld $xr0, $s2, 0 +; LA32-NEXT: addi.w $s3, $s3, 1 +; LA32-NEXT: sltui $a0, $s3, 1 +; LA32-NEXT: add.w $s4, $s4, $a0 +; LA32-NEXT: xor $a0, $s3, $s1 +; LA32-NEXT: xor $a1, $s4, $s0 +; LA32-NEXT: or $a0, $a0, $a1 +; LA32-NEXT: xvld $xr1, $sp, 16 # 32-byte Folded Reload +; LA32-NEXT: xvadd.h $xr1, $xr0, $xr1 +; LA32-NEXT: xvst $xr1, $sp, 16 # 32-byte Folded Spill +; LA32-NEXT: xvld $xr0, $sp, 16 # 32-byte Folded Reload +; LA32-NEXT: bnez $a0, .LBB3_2 +; LA32-NEXT: b .LBB3_4 +; LA32-NEXT: .LBB3_3: +; LA32-NEXT: xvrepli.b $xr0, 0 +; LA32-NEXT: .LBB3_4: # %for.cond.cleanup +; LA32-NEXT: xvst $xr0, $s2, 0 +; LA32-NEXT: ld.w $s4, $sp, 52 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s3, $sp, 56 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s2, $sp, 60 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s1, $sp, 64 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s0, $sp, 68 # 4-byte Folded Reload +; LA32-NEXT: ld.w $fp, $sp, 72 # 4-byte Folded Reload +; LA32-NEXT: ld.w $ra, $sp, 76 # 4-byte Folded Reload +; LA32-NEXT: addi.w $sp, $sp, 80 +; LA32-NEXT: ret +; +; LA64-LABEL: sink_fold_v16i16: +; LA64: # %bb.0: # %entry +; LA64-NEXT: addi.d $sp, $sp, -80 +; LA64-NEXT: st.d $ra, $sp, 72 # 8-byte Folded Spill +; LA64-NEXT: st.d $fp, $sp, 64 # 8-byte Folded Spill +; LA64-NEXT: st.d $s0, $sp, 56 # 8-byte Folded Spill +; LA64-NEXT: st.d $s1, $sp, 48 # 8-byte Folded Spill +; LA64-NEXT: slli.d $a0, $a0, 6 +; LA64-NEXT: add.d $a0, $a2, $a0 +; LA64-NEXT: addi.d $s1, $a0, 32 +; LA64-NEXT: blez $a1, .LBB3_3 +; LA64-NEXT: # %bb.1: # %for.body.preheader +; LA64-NEXT: move $fp, $a2 +; LA64-NEXT: move $s0, $a1 +; LA64-NEXT: xvrepli.b $xr0, 0 +; LA64-NEXT: .p2align 4, , 16 +; LA64-NEXT: .LBB3_2: # %for.body +; LA64-NEXT: # =>This Inner Loop Header: Depth=1 +; LA64-NEXT: xvst $xr0, $sp, 16 # 32-byte Folded Spill +; LA64-NEXT: move $a0, $fp +; LA64-NEXT: pcaddu18i $ra, %call36(f) +; LA64-NEXT: jirl $ra, $ra, 0 +; LA64-NEXT: xvld $xr0, $s1, 0 +; LA64-NEXT: addi.d $s0, $s0, -1 +; LA64-NEXT: xvld $xr1, $sp, 16 # 32-byte Folded Reload +; LA64-NEXT: xvadd.h $xr1, $xr0, $xr1 +; LA64-NEXT: xvst $xr1, $sp, 16 # 32-byte Folded Spill +; LA64-NEXT: xvld $xr0, $sp, 16 # 32-byte Folded Reload +; LA64-NEXT: bnez $s0, .LBB3_2 +; LA64-NEXT: b .LBB3_4 +; LA64-NEXT: .LBB3_3: +; LA64-NEXT: xvrepli.b $xr0, 0 +; LA64-NEXT: .LBB3_4: # %for.cond.cleanup +; LA64-NEXT: xvst $xr0, $s1, 0 +; LA64-NEXT: ld.d $s1, $sp, 48 # 8-byte Folded Reload +; LA64-NEXT: ld.d $s0, $sp, 56 # 8-byte Folded Reload +; LA64-NEXT: ld.d $fp, $sp, 64 # 8-byte Folded Reload +; LA64-NEXT: ld.d $ra, $sp, 72 # 8-byte Folded Reload +; LA64-NEXT: addi.d $sp, $sp, 80 +; LA64-NEXT: ret +entry: + %y = getelementptr inbounds %struct.V, ptr %a, i64 %k, i32 2 + %cmp = icmp sgt i64 %n, 0 + br i1 %cmp, label %for.body, label %for.cond.cleanup + +for.body: ; preds = %entry, %for.body + %i.0 = phi i64 [ 0, %entry ], [ %inc, %for.body ] + %sum.0 = phi <16 x i16> [ zeroinitializer, %entry ], [ %addv, %for.body ] + call void @f(ptr %a) + %v = load <16 x i16>, ptr %y + %addv = add <16 x i16> %v, %sum.0 + %inc = add nuw nsw i64 %i.0, 1 + %exitcond = icmp eq i64 %inc, %n + br i1 %exitcond, label %for.cond.cleanup, label %for.body + +for.cond.cleanup: ; preds = %for.body, %entry + %sum.lcssa = phi <16 x i16> [ zeroinitializer, %entry ], [ %addv, %for.body ] + store <16 x i16> %sum.lcssa, ptr %y + ret void +} + +define void @sink_fold_extracti8(i64 %k, i64 %n, ptr %a) nounwind { +; LA32-LABEL: sink_fold_extracti8: +; LA32: # %bb.0: # %entry +; LA32-NEXT: addi.w $sp, $sp, -48 +; LA32-NEXT: st.w $ra, $sp, 44 # 4-byte Folded Spill +; LA32-NEXT: st.w $fp, $sp, 40 # 4-byte Folded Spill +; LA32-NEXT: st.w $s0, $sp, 36 # 4-byte Folded Spill +; LA32-NEXT: st.w $s1, $sp, 32 # 4-byte Folded Spill +; LA32-NEXT: st.w $s2, $sp, 28 # 4-byte Folded Spill +; LA32-NEXT: st.w $s3, $sp, 24 # 4-byte Folded Spill +; LA32-NEXT: st.w $s4, $sp, 20 # 4-byte Folded Spill +; LA32-NEXT: move $s0, $a3 +; LA32-NEXT: move $s1, $a2 +; LA32-NEXT: slli.w $a1, $a0, 4 +; LA32-NEXT: alsl.w $a0, $a0, $a1, 3 +; LA32-NEXT: add.w $a0, $a4, $a0 +; LA32-NEXT: sltui $a1, $a3, 1 +; LA32-NEXT: slti $a2, $a3, 0 +; LA32-NEXT: masknez $a2, $a2, $a1 +; LA32-NEXT: sltui $a3, $s1, 1 +; LA32-NEXT: maskeqz $a1, $a3, $a1 +; LA32-NEXT: or $a1, $a1, $a2 +; LA32-NEXT: addi.w $s2, $a0, 16 +; LA32-NEXT: bnez $a1, .LBB4_3 +; LA32-NEXT: # %bb.1: # %for.body.preheader +; LA32-NEXT: move $fp, $a4 +; LA32-NEXT: move $s3, $zero +; LA32-NEXT: move $s4, $zero +; LA32-NEXT: vrepli.b $vr0, 0 +; LA32-NEXT: .p2align 4, , 16 +; LA32-NEXT: .LBB4_2: # %for.body +; LA32-NEXT: # =>This Inner Loop Header: Depth=1 +; LA32-NEXT: vst $vr0, $sp, 0 # 16-byte Folded Spill +; LA32-NEXT: move $a0, $fp +; LA32-NEXT: bl f +; LA32-NEXT: vldrepl.b $vr0, $s2, 0 +; LA32-NEXT: addi.w $s3, $s3, 1 +; LA32-NEXT: sltui $a0, $s3, 1 +; LA32-NEXT: add.w $s4, $s4, $a0 +; LA32-NEXT: xor $a0, $s3, $s1 +; LA32-NEXT: xor $a1, $s4, $s0 +; LA32-NEXT: or $a0, $a0, $a1 +; LA32-NEXT: vld $vr1, $sp, 0 # 16-byte Folded Reload +; LA32-NEXT: vadd.b $vr1, $vr0, $vr1 +; LA32-NEXT: vst $vr1, $sp, 0 # 16-byte Folded Spill +; LA32-NEXT: vld $vr0, $sp, 0 # 16-byte Folded Reload +; LA32-NEXT: bnez $a0, .LBB4_2 +; LA32-NEXT: b .LBB4_4 +; LA32-NEXT: .LBB4_3: +; LA32-NEXT: vrepli.b $vr0, 0 +; LA32-NEXT: .LBB4_4: # %for.cond.cleanup +; LA32-NEXT: vstelm.b $vr0, $s2, 0, 1 +; LA32-NEXT: ld.w $s4, $sp, 20 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s3, $sp, 24 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s2, $sp, 28 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s1, $sp, 32 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s0, $sp, 36 # 4-byte Folded Reload +; LA32-NEXT: ld.w $fp, $sp, 40 # 4-byte Folded Reload +; LA32-NEXT: ld.w $ra, $sp, 44 # 4-byte Folded Reload +; LA32-NEXT: addi.w $sp, $sp, 48 +; LA32-NEXT: ret +; +; LA64-LABEL: sink_fold_extracti8: +; LA64: # %bb.0: # %entry +; LA64-NEXT: addi.d $sp, $sp, -48 +; LA64-NEXT: st.d $ra, $sp, 40 # 8-byte Folded Spill +; LA64-NEXT: st.d $fp, $sp, 32 # 8-byte Folded Spill +; LA64-NEXT: st.d $s0, $sp, 24 # 8-byte Folded Spill +; LA64-NEXT: st.d $s1, $sp, 16 # 8-byte Folded Spill +; LA64-NEXT: move $s0, $a1 +; LA64-NEXT: slli.d $a1, $a0, 4 +; LA64-NEXT: alsl.d $a0, $a0, $a1, 3 +; LA64-NEXT: add.d $a0, $a2, $a0 +; LA64-NEXT: addi.d $s1, $a0, 16 +; LA64-NEXT: blez $s0, .LBB4_3 +; LA64-NEXT: # %bb.1: # %for.body.preheader +; LA64-NEXT: move $fp, $a2 +; LA64-NEXT: vrepli.b $vr0, 0 +; LA64-NEXT: .p2align 4, , 16 +; LA64-NEXT: .LBB4_2: # %for.body +; LA64-NEXT: # =>This Inner Loop Header: Depth=1 +; LA64-NEXT: vst $vr0, $sp, 0 # 16-byte Folded Spill +; LA64-NEXT: move $a0, $fp +; LA64-NEXT: pcaddu18i $ra, %call36(f) +; LA64-NEXT: jirl $ra, $ra, 0 +; LA64-NEXT: vldrepl.b $vr0, $s1, 0 +; LA64-NEXT: addi.d $s0, $s0, -1 +; LA64-NEXT: vld $vr1, $sp, 0 # 16-byte Folded Reload +; LA64-NEXT: vadd.b $vr1, $vr0, $vr1 +; LA64-NEXT: vst $vr1, $sp, 0 # 16-byte Folded Spill +; LA64-NEXT: vld $vr0, $sp, 0 # 16-byte Folded Reload +; LA64-NEXT: bnez $s0, .LBB4_2 +; LA64-NEXT: b .LBB4_4 +; LA64-NEXT: .LBB4_3: +; LA64-NEXT: vrepli.b $vr0, 0 +; LA64-NEXT: .LBB4_4: # %for.cond.cleanup +; LA64-NEXT: vstelm.b $vr0, $s1, 0, 1 +; LA64-NEXT: ld.d $s1, $sp, 16 # 8-byte Folded Reload +; LA64-NEXT: ld.d $s0, $sp, 24 # 8-byte Folded Reload +; LA64-NEXT: ld.d $fp, $sp, 32 # 8-byte Folded Reload +; LA64-NEXT: ld.d $ra, $sp, 40 # 8-byte Folded Reload +; LA64-NEXT: addi.d $sp, $sp, 48 +; LA64-NEXT: ret +entry: + %y = getelementptr inbounds %struct.S, ptr %a, i64 %k, i32 2 + %cmp = icmp sgt i64 %n, 0 + br i1 %cmp, label %for.body, label %for.cond.cleanup + +for.body: ; preds = %entry, %for.body + %i.0 = phi i64 [ 0, %entry ], [ %inc, %for.body ] + %sum.0 = phi <16 x i8> [ zeroinitializer, %entry ], [ %addv, %for.body ] + call void @f(ptr %a) + %e = load i8, ptr %y + %ins0 = insertelement <16 x i8> poison, i8 %e, i32 0 + %v = shufflevector <16 x i8> %ins0, <16 x i8> poison, <16 x i32> zeroinitializer + %addv = add <16 x i8> %v, %sum.0 + %inc = add nuw nsw i64 %i.0, 1 + %exitcond = icmp eq i64 %inc, %n + br i1 %exitcond, label %for.cond.cleanup, label %for.body + +for.cond.cleanup: ; preds = %for.body, %entry + %sum.lcssa = phi <16 x i8> [ zeroinitializer, %entry ], [ %addv, %for.body ] + %res = extractelement <16 x i8> %sum.lcssa, i32 1 + store i8 %res, ptr %y + ret void +} + +define void @sink_fold_extractf64(i64 %k, i64 %n, ptr %a) nounwind { +; LA32-LABEL: sink_fold_extractf64: +; LA32: # %bb.0: # %entry +; LA32-NEXT: addi.w $sp, $sp, -80 +; LA32-NEXT: st.w $ra, $sp, 76 # 4-byte Folded Spill +; LA32-NEXT: st.w $fp, $sp, 72 # 4-byte Folded Spill +; LA32-NEXT: st.w $s0, $sp, 68 # 4-byte Folded Spill +; LA32-NEXT: st.w $s1, $sp, 64 # 4-byte Folded Spill +; LA32-NEXT: st.w $s2, $sp, 60 # 4-byte Folded Spill +; LA32-NEXT: st.w $s3, $sp, 56 # 4-byte Folded Spill +; LA32-NEXT: st.w $s4, $sp, 52 # 4-byte Folded Spill +; LA32-NEXT: move $s0, $a3 +; LA32-NEXT: move $s1, $a2 +; LA32-NEXT: slli.w $a1, $a0, 4 +; LA32-NEXT: alsl.w $a0, $a0, $a1, 3 +; LA32-NEXT: add.w $a0, $a4, $a0 +; LA32-NEXT: sltui $a1, $a3, 1 +; LA32-NEXT: slti $a2, $a3, 0 +; LA32-NEXT: masknez $a2, $a2, $a1 +; LA32-NEXT: sltui $a3, $s1, 1 +; LA32-NEXT: maskeqz $a1, $a3, $a1 +; LA32-NEXT: or $a1, $a1, $a2 +; LA32-NEXT: addi.w $s2, $a0, 8 +; LA32-NEXT: bnez $a1, .LBB5_3 +; LA32-NEXT: # %bb.1: # %for.body.preheader +; LA32-NEXT: move $fp, $a4 +; LA32-NEXT: move $s3, $zero +; LA32-NEXT: move $s4, $zero +; LA32-NEXT: xvrepli.b $xr0, 0 +; LA32-NEXT: .p2align 4, , 16 +; LA32-NEXT: .LBB5_2: # %for.body +; LA32-NEXT: # =>This Inner Loop Header: Depth=1 +; LA32-NEXT: xvst $xr0, $sp, 16 # 32-byte Folded Spill +; LA32-NEXT: move $a0, $fp +; LA32-NEXT: bl f +; LA32-NEXT: xvldrepl.d $xr0, $s2, 0 +; LA32-NEXT: addi.w $s3, $s3, 1 +; LA32-NEXT: sltui $a0, $s3, 1 +; LA32-NEXT: add.w $s4, $s4, $a0 +; LA32-NEXT: xor $a0, $s3, $s1 +; LA32-NEXT: xor $a1, $s4, $s0 +; LA32-NEXT: or $a0, $a0, $a1 +; LA32-NEXT: xvld $xr1, $sp, 16 # 32-byte Folded Reload +; LA32-NEXT: xvfadd.d $xr1, $xr0, $xr1 +; LA32-NEXT: xvst $xr1, $sp, 16 # 32-byte Folded Spill +; LA32-NEXT: xvld $xr0, $sp, 16 # 32-byte Folded Reload +; LA32-NEXT: bnez $a0, .LBB5_2 +; LA32-NEXT: b .LBB5_4 +; LA32-NEXT: .LBB5_3: +; LA32-NEXT: xvrepli.b $xr0, 0 +; LA32-NEXT: .LBB5_4: # %for.cond.cleanup +; LA32-NEXT: xvstelm.d $xr0, $s2, 0, 1 +; LA32-NEXT: ld.w $s4, $sp, 52 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s3, $sp, 56 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s2, $sp, 60 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s1, $sp, 64 # 4-byte Folded Reload +; LA32-NEXT: ld.w $s0, $sp, 68 # 4-byte Folded Reload +; LA32-NEXT: ld.w $fp, $sp, 72 # 4-byte Folded Reload +; LA32-NEXT: ld.w $ra, $sp, 76 # 4-byte Folded Reload +; LA32-NEXT: addi.w $sp, $sp, 80 +; LA32-NEXT: ret +; +; LA64-LABEL: sink_fold_extractf64: +; LA64: # %bb.0: # %entry +; LA64-NEXT: addi.d $sp, $sp, -80 +; LA64-NEXT: st.d $ra, $sp, 72 # 8-byte Folded Spill +; LA64-NEXT: st.d $fp, $sp, 64 # 8-byte Folded Spill +; LA64-NEXT: st.d $s0, $sp, 56 # 8-byte Folded Spill +; LA64-NEXT: st.d $s1, $sp, 48 # 8-byte Folded Spill +; LA64-NEXT: move $s0, $a1 +; LA64-NEXT: slli.d $a1, $a0, 4 +; LA64-NEXT: alsl.d $a0, $a0, $a1, 3 +; LA64-NEXT: add.d $a0, $a2, $a0 +; LA64-NEXT: addi.d $s1, $a0, 8 +; LA64-NEXT: blez $s0, .LBB5_3 +; LA64-NEXT: # %bb.1: # %for.body.preheader +; LA64-NEXT: move $fp, $a2 +; LA64-NEXT: xvrepli.b $xr0, 0 +; LA64-NEXT: .p2align 4, , 16 +; LA64-NEXT: .LBB5_2: # %for.body +; LA64-NEXT: # =>This Inner Loop Header: Depth=1 +; LA64-NEXT: xvst $xr0, $sp, 16 # 32-byte Folded Spill +; LA64-NEXT: move $a0, $fp +; LA64-NEXT: pcaddu18i $ra, %call36(f) +; LA64-NEXT: jirl $ra, $ra, 0 +; LA64-NEXT: xvldrepl.d $xr0, $s1, 0 +; LA64-NEXT: addi.d $s0, $s0, -1 +; LA64-NEXT: xvld $xr1, $sp, 16 # 32-byte Folded Reload +; LA64-NEXT: xvfadd.d $xr1, $xr0, $xr1 +; LA64-NEXT: xvst $xr1, $sp, 16 # 32-byte Folded Spill +; LA64-NEXT: xvld $xr0, $sp, 16 # 32-byte Folded Reload +; LA64-NEXT: bnez $s0, .LBB5_2 +; LA64-NEXT: b .LBB5_4 +; LA64-NEXT: .LBB5_3: +; LA64-NEXT: xvrepli.b $xr0, 0 +; LA64-NEXT: .LBB5_4: # %for.cond.cleanup +; LA64-NEXT: xvstelm.d $xr0, $s1, 0, 1 +; LA64-NEXT: ld.d $s1, $sp, 48 # 8-byte Folded Reload +; LA64-NEXT: ld.d $s0, $sp, 56 # 8-byte Folded Reload +; LA64-NEXT: ld.d $fp, $sp, 64 # 8-byte Folded Reload +; LA64-NEXT: ld.d $ra, $sp, 72 # 8-byte Folded Reload +; LA64-NEXT: addi.d $sp, $sp, 80 +; LA64-NEXT: ret +entry: + %y = getelementptr inbounds %struct.F, ptr %a, i64 %k, i32 1 + %cmp = icmp sgt i64 %n, 0 + br i1 %cmp, label %for.body, label %for.cond.cleanup + +for.body: ; preds = %entry, %for.body + %i.0 = phi i64 [ 0, %entry ], [ %inc, %for.body ] + %sum.0 = phi <4 x double> [ zeroinitializer, %entry ], [ %addv, %for.body ] + call void @f(ptr %a) + %e = load double, ptr %y + %ins0 = insertelement <4 x double> poison, double %e, i32 0 + %v = shufflevector <4 x double> %ins0, <4 x double> poison, <4 x i32> zeroinitializer + %addv = fadd <4 x double> %v, %sum.0 + %inc = add nuw nsw i64 %i.0, 1 + %exitcond = icmp eq i64 %inc, %n + br i1 %exitcond, label %for.cond.cleanup, label %for.body + +for.cond.cleanup: ; preds = %for.body, %entry + %sum.lcssa = phi <4 x double> [ zeroinitializer, %entry ], [ %addv, %for.body ] + %res = extractelement <4 x double> %sum.lcssa, i32 1 + store double %res, ptr %y + ret void +} + +declare void @f(ptr) diff --git a/llvm/test/CodeGen/MIR2Vec/Inputs/reference_x86_vocab_print.txt b/llvm/test/CodeGen/MIR2Vec/Inputs/reference_x86_vocab_print.txt index d3c0da9..000c67ef 100644 --- a/llvm/test/CodeGen/MIR2Vec/Inputs/reference_x86_vocab_print.txt +++ b/llvm/test/CodeGen/MIR2Vec/Inputs/reference_x86_vocab_print.txt @@ -1439,11 +1439,8 @@ Key: PSUBWrm: [ 0.00 0.00 ] Key: PSUBWrr: [ 0.00 0.00 ] Key: PSWAPDrm: [ 0.00 0.00 ] Key: PSWAPDrr: [ 0.00 0.00 ] -Key: PT: [ 0.00 0.00 ] Key: PTCMMIMFP: [ 0.00 0.00 ] Key: PTCMMRLFP: [ 0.00 0.00 ] -Key: PTCONJTCMMIMFP: [ 0.00 0.00 ] -Key: PTCONJTFP: [ 0.00 0.00 ] Key: PTCVTROWD: [ 0.00 0.00 ] Key: PTCVTROWPS: [ 0.00 0.00 ] Key: PTDPBF: [ 0.00 0.00 ] @@ -1471,20 +1468,11 @@ Key: PTILEMOVROWrre: [ 0.00 0.00 ] Key: PTILEMOVROWrreV: [ 0.00 0.00 ] Key: PTILEMOVROWrri: [ 0.00 0.00 ] Key: PTILEMOVROWrriV: [ 0.00 0.00 ] -Key: PTILEPAIRLOAD: [ 0.00 0.00 ] -Key: PTILEPAIRSTORE: [ 0.00 0.00 ] Key: PTILESTORED: [ 0.00 0.00 ] Key: PTILESTOREDV: [ 0.00 0.00 ] Key: PTILEZERO: [ 0.00 0.00 ] Key: PTILEZEROV: [ 0.00 0.00 ] Key: PTMMULTF: [ 0.00 0.00 ] -Key: PTTCMMIMFP: [ 0.00 0.00 ] -Key: PTTCMMRLFP: [ 0.00 0.00 ] -Key: PTTDPBF: [ 0.00 0.00 ] -Key: PTTDPFP: [ 0.00 0.00 ] -Key: PTTMMULTF: [ 0.00 0.00 ] -Key: PTTRANSPOSED: [ 0.00 0.00 ] -Key: PTTRANSPOSEDV: [ 0.00 0.00 ] Key: PTWRITE: [ 0.00 0.00 ] Key: PTWRITEm: [ 0.00 0.00 ] Key: PTWRITEr: [ 0.00 0.00 ] @@ -1717,8 +1705,6 @@ Key: TAILJMPm: [ 0.00 0.00 ] Key: TAILJMPr: [ 0.00 0.00 ] Key: TCMMIMFP: [ 0.00 0.00 ] Key: TCMMRLFP: [ 0.00 0.00 ] -Key: TCONJTCMMIMFP: [ 0.00 0.00 ] -Key: TCONJTFP: [ 0.00 0.00 ] Key: TCRETURN_HIPE: [ 0.00 0.00 ] Key: TCRETURN_WIN: [ 0.00 0.00 ] Key: TCRETURN_WINmi: [ 0.00 0.00 ] @@ -1764,12 +1750,6 @@ Key: TPAUSE: [ 0.00 0.00 ] Key: TRAP: [ 0.00 0.00 ] Key: TST_F: [ 0.00 0.00 ] Key: TST_Fp: [ 0.00 0.00 ] -Key: TTCMMIMFP: [ 0.00 0.00 ] -Key: TTCMMRLFP: [ 0.00 0.00 ] -Key: TTDPBF: [ 0.00 0.00 ] -Key: TTDPFP: [ 0.00 0.00 ] -Key: TTMMULTF: [ 0.00 0.00 ] -Key: TTRANSPOSED: [ 0.00 0.00 ] Key: TZCNT: [ 0.00 0.00 ] Key: TZMSK: [ 0.00 0.00 ] Key: UBSAN_UD: [ 0.00 0.00 ] @@ -7034,7 +7014,6 @@ Key: PhyReg_VR256: [ 0.00 0.00 ] Key: PhyReg_VR512: [ 0.00 0.00 ] Key: PhyReg_VR512_0_15: [ 0.00 0.00 ] Key: PhyReg_TILE: [ 0.00 0.00 ] -Key: PhyReg_TILEPAIR: [ 0.00 0.00 ] Key: VirtReg_GR8: [ 0.00 0.00 ] Key: VirtReg_GRH8: [ 0.00 0.00 ] Key: VirtReg_GR8_NOREX2: [ 0.00 0.00 ] @@ -7170,4 +7149,3 @@ Key: VirtReg_VR256: [ 0.00 0.00 ] Key: VirtReg_VR512: [ 0.00 0.00 ] Key: VirtReg_VR512_0_15: [ 0.00 0.00 ] Key: VirtReg_TILE: [ 0.00 0.00 ] -Key: VirtReg_TILEPAIR: [ 0.00 0.00 ] diff --git a/llvm/test/CodeGen/MIR2Vec/Inputs/reference_x86_vocab_wo=0.5_print.txt b/llvm/test/CodeGen/MIR2Vec/Inputs/reference_x86_vocab_wo=0.5_print.txt index c6e5508..bb72886 100644 --- a/llvm/test/CodeGen/MIR2Vec/Inputs/reference_x86_vocab_wo=0.5_print.txt +++ b/llvm/test/CodeGen/MIR2Vec/Inputs/reference_x86_vocab_wo=0.5_print.txt @@ -1439,11 +1439,8 @@ Key: PSUBWrm: [ 0.00 0.00 ] Key: PSUBWrr: [ 0.00 0.00 ] Key: PSWAPDrm: [ 0.00 0.00 ] Key: PSWAPDrr: [ 0.00 0.00 ] -Key: PT: [ 0.00 0.00 ] Key: PTCMMIMFP: [ 0.00 0.00 ] Key: PTCMMRLFP: [ 0.00 0.00 ] -Key: PTCONJTCMMIMFP: [ 0.00 0.00 ] -Key: PTCONJTFP: [ 0.00 0.00 ] Key: PTCVTROWD: [ 0.00 0.00 ] Key: PTCVTROWPS: [ 0.00 0.00 ] Key: PTDPBF: [ 0.00 0.00 ] @@ -1471,20 +1468,11 @@ Key: PTILEMOVROWrre: [ 0.00 0.00 ] Key: PTILEMOVROWrreV: [ 0.00 0.00 ] Key: PTILEMOVROWrri: [ 0.00 0.00 ] Key: PTILEMOVROWrriV: [ 0.00 0.00 ] -Key: PTILEPAIRLOAD: [ 0.00 0.00 ] -Key: PTILEPAIRSTORE: [ 0.00 0.00 ] Key: PTILESTORED: [ 0.00 0.00 ] Key: PTILESTOREDV: [ 0.00 0.00 ] Key: PTILEZERO: [ 0.00 0.00 ] Key: PTILEZEROV: [ 0.00 0.00 ] Key: PTMMULTF: [ 0.00 0.00 ] -Key: PTTCMMIMFP: [ 0.00 0.00 ] -Key: PTTCMMRLFP: [ 0.00 0.00 ] -Key: PTTDPBF: [ 0.00 0.00 ] -Key: PTTDPFP: [ 0.00 0.00 ] -Key: PTTMMULTF: [ 0.00 0.00 ] -Key: PTTRANSPOSED: [ 0.00 0.00 ] -Key: PTTRANSPOSEDV: [ 0.00 0.00 ] Key: PTWRITE: [ 0.00 0.00 ] Key: PTWRITEm: [ 0.00 0.00 ] Key: PTWRITEr: [ 0.00 0.00 ] @@ -1717,8 +1705,6 @@ Key: TAILJMPm: [ 0.00 0.00 ] Key: TAILJMPr: [ 0.00 0.00 ] Key: TCMMIMFP: [ 0.00 0.00 ] Key: TCMMRLFP: [ 0.00 0.00 ] -Key: TCONJTCMMIMFP: [ 0.00 0.00 ] -Key: TCONJTFP: [ 0.00 0.00 ] Key: TCRETURN_HIPE: [ 0.00 0.00 ] Key: TCRETURN_WIN: [ 0.00 0.00 ] Key: TCRETURN_WINmi: [ 0.00 0.00 ] @@ -1764,12 +1750,6 @@ Key: TPAUSE: [ 0.00 0.00 ] Key: TRAP: [ 0.00 0.00 ] Key: TST_F: [ 0.00 0.00 ] Key: TST_Fp: [ 0.00 0.00 ] -Key: TTCMMIMFP: [ 0.00 0.00 ] -Key: TTCMMRLFP: [ 0.00 0.00 ] -Key: TTDPBF: [ 0.00 0.00 ] -Key: TTDPFP: [ 0.00 0.00 ] -Key: TTMMULTF: [ 0.00 0.00 ] -Key: TTRANSPOSED: [ 0.00 0.00 ] Key: TZCNT: [ 0.00 0.00 ] Key: TZMSK: [ 0.00 0.00 ] Key: UBSAN_UD: [ 0.00 0.00 ] @@ -7034,7 +7014,6 @@ Key: PhyReg_VR256: [ 0.00 0.00 ] Key: PhyReg_VR512: [ 0.00 0.00 ] Key: PhyReg_VR512_0_15: [ 0.00 0.00 ] Key: PhyReg_TILE: [ 0.00 0.00 ] -Key: PhyReg_TILEPAIR: [ 0.00 0.00 ] Key: VirtReg_GR8: [ 0.00 0.00 ] Key: VirtReg_GRH8: [ 0.00 0.00 ] Key: VirtReg_GR8_NOREX2: [ 0.00 0.00 ] @@ -7170,4 +7149,3 @@ Key: VirtReg_VR256: [ 0.00 0.00 ] Key: VirtReg_VR512: [ 0.00 0.00 ] Key: VirtReg_VR512_0_15: [ 0.00 0.00 ] Key: VirtReg_TILE: [ 0.00 0.00 ] -Key: VirtReg_TILEPAIR: [ 0.00 0.00 ] diff --git a/llvm/test/CodeGen/MLRegAlloc/dev-mode-extra-features-logging.ll b/llvm/test/CodeGen/MLRegAlloc/dev-mode-extra-features-logging.ll index bd8d882..9dd402d 100644 --- a/llvm/test/CodeGen/MLRegAlloc/dev-mode-extra-features-logging.ll +++ b/llvm/test/CodeGen/MLRegAlloc/dev-mode-extra-features-logging.ll @@ -26,7 +26,7 @@ ; Also, the first eviction problem is significantly less than 300 instructions. Check ; that there is a zero value. ; Note: we're regex-ing some of the opcodes to avoid test flakyness. -; CHECK: instructions: 20,{{([0-9]{4})}},1{{([0-9]{3})}},2{{([0-9]{3})}},{{.*}},0, +; CHECK: instructions: 20,{{([0-9]{4})}},{{([0-9]{4})}},{{([0-9]{4})}},{{.*}},0, ; Only the candidate virtreg and the 10th LR are included in this problem. Make ; sure the other LRs have values of zero. There are 2700 0s followed by some 1s. ; There's a limit to how many repetitions can be matched. diff --git a/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s-1cta.ll b/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s-1cta.ll index b5c43fd2..d653895 100644 --- a/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s-1cta.ll +++ b/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s-1cta.ll @@ -1,8 +1,12 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 ; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_100a -mattr=+ptx86| FileCheck --check-prefixes=CHECK-PTX64 %s ; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_100a -mattr=+ptx86 --nvptx-short-ptr| FileCheck --check-prefixes=CHECK-PTX-SHARED32 %s +; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_100f -mattr=+ptx88 | FileCheck --check-prefixes=CHECK-PTX64 %s +; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_110f -mattr=+ptx90 | FileCheck --check-prefixes=CHECK-PTX64 %s ; RUN: %if ptxas-sm_100a && ptxas-isa-8.6 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_100a -mattr=+ptx86| %ptxas-verify -arch=sm_100a %} ; RUN: %if ptxas-sm_100a && ptxas-isa-8.6 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_100a -mattr=+ptx86 --nvptx-short-ptr| %ptxas-verify -arch=sm_100a %} +; RUN: %if ptxas-sm_100f && ptxas-isa-8.8 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_100f -mattr=+ptx88 | %ptxas-verify -arch=sm_100f %} +; RUN: %if ptxas-sm_110f && ptxas-isa-9.0 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_110f -mattr=+ptx90 | %ptxas-verify -arch=sm_110f %} target triple = "nvptx64-nvidia-cuda" diff --git a/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s-2cta.ll b/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s-2cta.ll index 57342dc..5de1ac8 100644 --- a/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s-2cta.ll +++ b/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s-2cta.ll @@ -1,8 +1,12 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 ; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_100a -mattr=+ptx86| FileCheck --check-prefixes=CHECK-PTX64 %s ; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_100a -mattr=+ptx86 --nvptx-short-ptr| FileCheck --check-prefixes=CHECK-PTX-SHARED32 %s +; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_100f -mattr=+ptx88 | FileCheck --check-prefixes=CHECK-PTX64 %s +; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_110f -mattr=+ptx90 | FileCheck --check-prefixes=CHECK-PTX64 %s ; RUN: %if ptxas-sm_100a && ptxas-isa-8.6 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_100a -mattr=+ptx86| %ptxas-verify -arch=sm_100a %} ; RUN: %if ptxas-sm_100a && ptxas-isa-8.6 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_100a -mattr=+ptx86 --nvptx-short-ptr| %ptxas-verify -arch=sm_100a %} +; RUN: %if ptxas-sm_100f && ptxas-isa-8.8 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_100f -mattr=+ptx88 | %ptxas-verify -arch=sm_100f %} +; RUN: %if ptxas-sm_110f && ptxas-isa-9.0 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_110f -mattr=+ptx90 | %ptxas-verify -arch=sm_110f %} target triple = "nvptx64-nvidia-cuda" diff --git a/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s-gather4.ll b/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s-gather4.ll index 6296d5a..2f5c1ef 100644 --- a/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s-gather4.ll +++ b/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s-gather4.ll @@ -1,8 +1,12 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 ; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_100a -mattr=+ptx86| FileCheck --check-prefixes=CHECK-PTX64 %s ; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_100a -mattr=+ptx86 --nvptx-short-ptr| FileCheck --check-prefixes=CHECK-PTX-SHARED32 %s +; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_100f -mattr=+ptx88 | FileCheck --check-prefixes=CHECK-PTX64 %s +; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_110f -mattr=+ptx90 | FileCheck --check-prefixes=CHECK-PTX64 %s ; RUN: %if ptxas-sm_100a && ptxas-isa-8.6 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_100a -mattr=+ptx86| %ptxas-verify -arch=sm_100a %} ; RUN: %if ptxas-sm_100a && ptxas-isa-8.6 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_100a -mattr=+ptx86 --nvptx-short-ptr| %ptxas-verify -arch=sm_100a %} +; RUN: %if ptxas-sm_100f && ptxas-isa-8.8 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_100f -mattr=+ptx88 | %ptxas-verify -arch=sm_100f %} +; RUN: %if ptxas-sm_110f && ptxas-isa-9.0 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_110f -mattr=+ptx90 | %ptxas-verify -arch=sm_110f %} target triple = "nvptx64-nvidia-cuda" diff --git a/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s-im2colw.ll b/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s-im2colw.ll index e5ae387..a2b2c2f 100644 --- a/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s-im2colw.ll +++ b/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s-im2colw.ll @@ -1,8 +1,12 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 ; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_100a -mattr=+ptx86| FileCheck --check-prefixes=CHECK-PTX64 %s ; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_100a -mattr=+ptx86 --nvptx-short-ptr| FileCheck --check-prefixes=CHECK-PTX-SHARED32 %s +; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_100f -mattr=+ptx88 | FileCheck --check-prefixes=CHECK-PTX64 %s +; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_110f -mattr=+ptx90 | FileCheck --check-prefixes=CHECK-PTX64 %s ; RUN: %if ptxas-sm_100a && ptxas-isa-8.6 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_100a -mattr=+ptx86| %ptxas-verify -arch=sm_100a %} ; RUN: %if ptxas-sm_100a && ptxas-isa-8.6 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_100a -mattr=+ptx86 --nvptx-short-ptr| %ptxas-verify -arch=sm_100a %} +; RUN: %if ptxas-sm_100f && ptxas-isa-8.8 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_100f -mattr=+ptx88 | %ptxas-verify -arch=sm_100f %} +; RUN: %if ptxas-sm_110f && ptxas-isa-9.0 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_110f -mattr=+ptx90 | %ptxas-verify -arch=sm_110f %} target triple = "nvptx64-nvidia-cuda" diff --git a/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s-im2colw128.ll b/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s-im2colw128.ll index 7d04ada..e4c48dd 100644 --- a/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s-im2colw128.ll +++ b/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s-im2colw128.ll @@ -1,8 +1,12 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 ; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_100a -mattr=+ptx86| FileCheck --check-prefixes=CHECK-PTX64 %s ; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_100a -mattr=+ptx86 --nvptx-short-ptr| FileCheck --check-prefixes=CHECK-PTX-SHARED32 %s +; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_100f -mattr=+ptx88 | FileCheck --check-prefixes=CHECK-PTX64 %s +; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_110f -mattr=+ptx90 | FileCheck --check-prefixes=CHECK-PTX64 %s ; RUN: %if ptxas-sm_100a && ptxas-isa-8.6 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_100a -mattr=+ptx86| %ptxas-verify -arch=sm_100a %} ; RUN: %if ptxas-sm_100a && ptxas-isa-8.6 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_100a -mattr=+ptx86 --nvptx-short-ptr| %ptxas-verify -arch=sm_100a %} +; RUN: %if ptxas-sm_100f && ptxas-isa-8.8 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_100f -mattr=+ptx88 | %ptxas-verify -arch=sm_100f %} +; RUN: %if ptxas-sm_110f && ptxas-isa-9.0 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_110f -mattr=+ptx90 | %ptxas-verify -arch=sm_110f %} target triple = "nvptx64-nvidia-cuda" diff --git a/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s.ll b/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s.ll index b0fe77c..727bb3b 100644 --- a/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s.ll +++ b/llvm/test/CodeGen/NVPTX/cp-async-bulk-tensor-g2s.ll @@ -1,8 +1,12 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 ; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_90 -mattr=+ptx80| FileCheck --check-prefixes=CHECK-PTX64 %s ; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_90 -mattr=+ptx80 --nvptx-short-ptr| FileCheck --check-prefixes=CHECK-PTX-SHARED32 %s +; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_100f -mattr=+ptx88 | FileCheck --check-prefixes=CHECK-PTX64 %s +; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_110f -mattr=+ptx90 | FileCheck --check-prefixes=CHECK-PTX64 %s ; RUN: %if ptxas-sm_90 && ptxas-isa-8.0 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_90 -mattr=+ptx80| %ptxas-verify -arch=sm_90 %} ; RUN: %if ptxas-sm_90 && ptxas-isa-8.0 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_90 -mattr=+ptx80 --nvptx-short-ptr| %ptxas-verify -arch=sm_90 %} +; RUN: %if ptxas-sm_100f && ptxas-isa-8.8 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_100f -mattr=+ptx88 | %ptxas-verify -arch=sm_100f %} +; RUN: %if ptxas-sm_110f && ptxas-isa-9.0 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_110f -mattr=+ptx90 | %ptxas-verify -arch=sm_110f %} target triple = "nvptx64-nvidia-cuda" @@ -29,10 +33,10 @@ define void @cp_async_bulk_tensor_g2s_tile_1d(ptr addrspace(7) %d, ptr addrspace ; CHECK-PTX64-NEXT: ld.param.b64 %rd2, [cp_async_bulk_tensor_g2s_tile_1d_param_1]; ; CHECK-PTX64-NEXT: ld.param.b64 %rd3, [cp_async_bulk_tensor_g2s_tile_1d_param_2]; ; CHECK-PTX64-NEXT: ld.param.b32 %r1, [cp_async_bulk_tensor_g2s_tile_1d_param_3]; -; CHECK-PTX64-NEXT: cp.async.bulk.tensor.1d.shared::cluster.global.tile.mbarrier::complete_tx::bytes [%rd1], [%rd3, {%r1}], [%rd2]; +; CHECK-PTX64-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_tile_1d_param_4]; ; CHECK-PTX64-NEXT: ld.param.b64 %rd4, [cp_async_bulk_tensor_g2s_tile_1d_param_5]; +; CHECK-PTX64-NEXT: cp.async.bulk.tensor.1d.shared::cluster.global.tile.mbarrier::complete_tx::bytes [%rd1], [%rd3, {%r1}], [%rd2]; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.1d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.L2::cache_hint [%rd1], [%rd3, {%r1}], [%rd2], %rd4; -; CHECK-PTX64-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_tile_1d_param_4]; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.1d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.multicast::cluster [%rd1], [%rd3, {%r1}], [%rd2], %rs1; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.1d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.multicast::cluster.L2::cache_hint [%rd1], [%rd3, {%r1}], [%rd2], %rs1, %rd4; ; CHECK-PTX64-NEXT: ret; @@ -48,10 +52,10 @@ define void @cp_async_bulk_tensor_g2s_tile_1d(ptr addrspace(7) %d, ptr addrspace ; CHECK-PTX-SHARED32-NEXT: ld.param.b32 %r2, [cp_async_bulk_tensor_g2s_tile_1d_param_1]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b64 %rd1, [cp_async_bulk_tensor_g2s_tile_1d_param_2]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b32 %r3, [cp_async_bulk_tensor_g2s_tile_1d_param_3]; -; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.1d.shared::cluster.global.tile.mbarrier::complete_tx::bytes [%r1], [%rd1, {%r3}], [%r2]; +; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_tile_1d_param_4]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b64 %rd2, [cp_async_bulk_tensor_g2s_tile_1d_param_5]; +; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.1d.shared::cluster.global.tile.mbarrier::complete_tx::bytes [%r1], [%rd1, {%r3}], [%r2]; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.1d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.L2::cache_hint [%r1], [%rd1, {%r3}], [%r2], %rd2; -; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_tile_1d_param_4]; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.1d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.multicast::cluster [%r1], [%rd1, {%r3}], [%r2], %rs1; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.1d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.multicast::cluster.L2::cache_hint [%r1], [%rd1, {%r3}], [%r2], %rs1, %rd2; ; CHECK-PTX-SHARED32-NEXT: ret; @@ -79,10 +83,10 @@ define void @cp_async_bulk_tensor_g2s_tile_2d(ptr addrspace(7) %d, ptr addrspace ; CHECK-PTX64-NEXT: ld.param.b64 %rd3, [cp_async_bulk_tensor_g2s_tile_2d_param_2]; ; CHECK-PTX64-NEXT: ld.param.b32 %r1, [cp_async_bulk_tensor_g2s_tile_2d_param_3]; ; CHECK-PTX64-NEXT: ld.param.b32 %r2, [cp_async_bulk_tensor_g2s_tile_2d_param_4]; -; CHECK-PTX64-NEXT: cp.async.bulk.tensor.2d.shared::cluster.global.tile.mbarrier::complete_tx::bytes [%rd1], [%rd3, {%r1, %r2}], [%rd2]; +; CHECK-PTX64-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_tile_2d_param_5]; ; CHECK-PTX64-NEXT: ld.param.b64 %rd4, [cp_async_bulk_tensor_g2s_tile_2d_param_6]; +; CHECK-PTX64-NEXT: cp.async.bulk.tensor.2d.shared::cluster.global.tile.mbarrier::complete_tx::bytes [%rd1], [%rd3, {%r1, %r2}], [%rd2]; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.2d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.L2::cache_hint [%rd1], [%rd3, {%r1, %r2}], [%rd2], %rd4; -; CHECK-PTX64-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_tile_2d_param_5]; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.2d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.multicast::cluster [%rd1], [%rd3, {%r1, %r2}], [%rd2], %rs1; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.2d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.multicast::cluster.L2::cache_hint [%rd1], [%rd3, {%r1, %r2}], [%rd2], %rs1, %rd4; ; CHECK-PTX64-NEXT: ret; @@ -99,10 +103,10 @@ define void @cp_async_bulk_tensor_g2s_tile_2d(ptr addrspace(7) %d, ptr addrspace ; CHECK-PTX-SHARED32-NEXT: ld.param.b64 %rd1, [cp_async_bulk_tensor_g2s_tile_2d_param_2]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b32 %r3, [cp_async_bulk_tensor_g2s_tile_2d_param_3]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b32 %r4, [cp_async_bulk_tensor_g2s_tile_2d_param_4]; -; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.2d.shared::cluster.global.tile.mbarrier::complete_tx::bytes [%r1], [%rd1, {%r3, %r4}], [%r2]; +; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_tile_2d_param_5]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b64 %rd2, [cp_async_bulk_tensor_g2s_tile_2d_param_6]; +; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.2d.shared::cluster.global.tile.mbarrier::complete_tx::bytes [%r1], [%rd1, {%r3, %r4}], [%r2]; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.2d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.L2::cache_hint [%r1], [%rd1, {%r3, %r4}], [%r2], %rd2; -; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_tile_2d_param_5]; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.2d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.multicast::cluster [%r1], [%rd1, {%r3, %r4}], [%r2], %rs1; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.2d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.multicast::cluster.L2::cache_hint [%r1], [%rd1, {%r3, %r4}], [%r2], %rs1, %rd2; ; CHECK-PTX-SHARED32-NEXT: ret; @@ -131,10 +135,10 @@ define void @cp_async_bulk_tensor_g2s_tile_3d(ptr addrspace(7) %d, ptr addrspace ; CHECK-PTX64-NEXT: ld.param.b32 %r1, [cp_async_bulk_tensor_g2s_tile_3d_param_3]; ; CHECK-PTX64-NEXT: ld.param.b32 %r2, [cp_async_bulk_tensor_g2s_tile_3d_param_4]; ; CHECK-PTX64-NEXT: ld.param.b32 %r3, [cp_async_bulk_tensor_g2s_tile_3d_param_5]; -; CHECK-PTX64-NEXT: cp.async.bulk.tensor.3d.shared::cluster.global.tile.mbarrier::complete_tx::bytes [%rd1], [%rd3, {%r1, %r2, %r3}], [%rd2]; +; CHECK-PTX64-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_tile_3d_param_6]; ; CHECK-PTX64-NEXT: ld.param.b64 %rd4, [cp_async_bulk_tensor_g2s_tile_3d_param_7]; +; CHECK-PTX64-NEXT: cp.async.bulk.tensor.3d.shared::cluster.global.tile.mbarrier::complete_tx::bytes [%rd1], [%rd3, {%r1, %r2, %r3}], [%rd2]; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.3d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.L2::cache_hint [%rd1], [%rd3, {%r1, %r2, %r3}], [%rd2], %rd4; -; CHECK-PTX64-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_tile_3d_param_6]; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.3d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.multicast::cluster [%rd1], [%rd3, {%r1, %r2, %r3}], [%rd2], %rs1; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.3d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.multicast::cluster.L2::cache_hint [%rd1], [%rd3, {%r1, %r2, %r3}], [%rd2], %rs1, %rd4; ; CHECK-PTX64-NEXT: ret; @@ -152,10 +156,10 @@ define void @cp_async_bulk_tensor_g2s_tile_3d(ptr addrspace(7) %d, ptr addrspace ; CHECK-PTX-SHARED32-NEXT: ld.param.b32 %r3, [cp_async_bulk_tensor_g2s_tile_3d_param_3]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b32 %r4, [cp_async_bulk_tensor_g2s_tile_3d_param_4]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b32 %r5, [cp_async_bulk_tensor_g2s_tile_3d_param_5]; -; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.3d.shared::cluster.global.tile.mbarrier::complete_tx::bytes [%r1], [%rd1, {%r3, %r4, %r5}], [%r2]; +; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_tile_3d_param_6]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b64 %rd2, [cp_async_bulk_tensor_g2s_tile_3d_param_7]; +; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.3d.shared::cluster.global.tile.mbarrier::complete_tx::bytes [%r1], [%rd1, {%r3, %r4, %r5}], [%r2]; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.3d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.L2::cache_hint [%r1], [%rd1, {%r3, %r4, %r5}], [%r2], %rd2; -; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_tile_3d_param_6]; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.3d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.multicast::cluster [%r1], [%rd1, {%r3, %r4, %r5}], [%r2], %rs1; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.3d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.multicast::cluster.L2::cache_hint [%r1], [%rd1, {%r3, %r4, %r5}], [%r2], %rs1, %rd2; ; CHECK-PTX-SHARED32-NEXT: ret; @@ -185,10 +189,10 @@ define void @cp_async_bulk_tensor_g2s_tile_4d(ptr addrspace(7) %d, ptr addrspace ; CHECK-PTX64-NEXT: ld.param.b32 %r2, [cp_async_bulk_tensor_g2s_tile_4d_param_4]; ; CHECK-PTX64-NEXT: ld.param.b32 %r3, [cp_async_bulk_tensor_g2s_tile_4d_param_5]; ; CHECK-PTX64-NEXT: ld.param.b32 %r4, [cp_async_bulk_tensor_g2s_tile_4d_param_6]; -; CHECK-PTX64-NEXT: cp.async.bulk.tensor.4d.shared::cluster.global.tile.mbarrier::complete_tx::bytes [%rd1], [%rd3, {%r1, %r2, %r3, %r4}], [%rd2]; +; CHECK-PTX64-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_tile_4d_param_7]; ; CHECK-PTX64-NEXT: ld.param.b64 %rd4, [cp_async_bulk_tensor_g2s_tile_4d_param_8]; +; CHECK-PTX64-NEXT: cp.async.bulk.tensor.4d.shared::cluster.global.tile.mbarrier::complete_tx::bytes [%rd1], [%rd3, {%r1, %r2, %r3, %r4}], [%rd2]; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.4d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.L2::cache_hint [%rd1], [%rd3, {%r1, %r2, %r3, %r4}], [%rd2], %rd4; -; CHECK-PTX64-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_tile_4d_param_7]; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.4d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.multicast::cluster [%rd1], [%rd3, {%r1, %r2, %r3, %r4}], [%rd2], %rs1; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.4d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.multicast::cluster.L2::cache_hint [%rd1], [%rd3, {%r1, %r2, %r3, %r4}], [%rd2], %rs1, %rd4; ; CHECK-PTX64-NEXT: ret; @@ -207,10 +211,10 @@ define void @cp_async_bulk_tensor_g2s_tile_4d(ptr addrspace(7) %d, ptr addrspace ; CHECK-PTX-SHARED32-NEXT: ld.param.b32 %r4, [cp_async_bulk_tensor_g2s_tile_4d_param_4]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b32 %r5, [cp_async_bulk_tensor_g2s_tile_4d_param_5]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b32 %r6, [cp_async_bulk_tensor_g2s_tile_4d_param_6]; -; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.4d.shared::cluster.global.tile.mbarrier::complete_tx::bytes [%r1], [%rd1, {%r3, %r4, %r5, %r6}], [%r2]; +; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_tile_4d_param_7]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b64 %rd2, [cp_async_bulk_tensor_g2s_tile_4d_param_8]; +; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.4d.shared::cluster.global.tile.mbarrier::complete_tx::bytes [%r1], [%rd1, {%r3, %r4, %r5, %r6}], [%r2]; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.4d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.L2::cache_hint [%r1], [%rd1, {%r3, %r4, %r5, %r6}], [%r2], %rd2; -; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_tile_4d_param_7]; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.4d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.multicast::cluster [%r1], [%rd1, {%r3, %r4, %r5, %r6}], [%r2], %rs1; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.4d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.multicast::cluster.L2::cache_hint [%r1], [%rd1, {%r3, %r4, %r5, %r6}], [%r2], %rs1, %rd2; ; CHECK-PTX-SHARED32-NEXT: ret; @@ -241,10 +245,10 @@ define void @cp_async_bulk_tensor_g2s_tile_5d(ptr addrspace(7) %d, ptr addrspace ; CHECK-PTX64-NEXT: ld.param.b32 %r3, [cp_async_bulk_tensor_g2s_tile_5d_param_5]; ; CHECK-PTX64-NEXT: ld.param.b32 %r4, [cp_async_bulk_tensor_g2s_tile_5d_param_6]; ; CHECK-PTX64-NEXT: ld.param.b32 %r5, [cp_async_bulk_tensor_g2s_tile_5d_param_7]; -; CHECK-PTX64-NEXT: cp.async.bulk.tensor.5d.shared::cluster.global.tile.mbarrier::complete_tx::bytes [%rd1], [%rd3, {%r1, %r2, %r3, %r4, %r5}], [%rd2]; +; CHECK-PTX64-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_tile_5d_param_8]; ; CHECK-PTX64-NEXT: ld.param.b64 %rd4, [cp_async_bulk_tensor_g2s_tile_5d_param_9]; +; CHECK-PTX64-NEXT: cp.async.bulk.tensor.5d.shared::cluster.global.tile.mbarrier::complete_tx::bytes [%rd1], [%rd3, {%r1, %r2, %r3, %r4, %r5}], [%rd2]; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.5d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.L2::cache_hint [%rd1], [%rd3, {%r1, %r2, %r3, %r4, %r5}], [%rd2], %rd4; -; CHECK-PTX64-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_tile_5d_param_8]; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.5d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.multicast::cluster [%rd1], [%rd3, {%r1, %r2, %r3, %r4, %r5}], [%rd2], %rs1; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.5d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.multicast::cluster.L2::cache_hint [%rd1], [%rd3, {%r1, %r2, %r3, %r4, %r5}], [%rd2], %rs1, %rd4; ; CHECK-PTX64-NEXT: ret; @@ -264,10 +268,10 @@ define void @cp_async_bulk_tensor_g2s_tile_5d(ptr addrspace(7) %d, ptr addrspace ; CHECK-PTX-SHARED32-NEXT: ld.param.b32 %r5, [cp_async_bulk_tensor_g2s_tile_5d_param_5]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b32 %r6, [cp_async_bulk_tensor_g2s_tile_5d_param_6]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b32 %r7, [cp_async_bulk_tensor_g2s_tile_5d_param_7]; -; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.5d.shared::cluster.global.tile.mbarrier::complete_tx::bytes [%r1], [%rd1, {%r3, %r4, %r5, %r6, %r7}], [%r2]; +; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_tile_5d_param_8]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b64 %rd2, [cp_async_bulk_tensor_g2s_tile_5d_param_9]; +; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.5d.shared::cluster.global.tile.mbarrier::complete_tx::bytes [%r1], [%rd1, {%r3, %r4, %r5, %r6, %r7}], [%r2]; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.5d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.L2::cache_hint [%r1], [%rd1, {%r3, %r4, %r5, %r6, %r7}], [%r2], %rd2; -; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_tile_5d_param_8]; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.5d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.multicast::cluster [%r1], [%rd1, {%r3, %r4, %r5, %r6, %r7}], [%r2], %rs1; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.5d.shared::cluster.global.tile.mbarrier::complete_tx::bytes.multicast::cluster.L2::cache_hint [%r1], [%rd1, {%r3, %r4, %r5, %r6, %r7}], [%r2], %rs1, %rd2; ; CHECK-PTX-SHARED32-NEXT: ret; @@ -297,10 +301,10 @@ define void @cp_async_bulk_tensor_g2s_im2col_3d(ptr addrspace(7) %d, ptr addrspa ; CHECK-PTX64-NEXT: ld.param.b32 %r2, [cp_async_bulk_tensor_g2s_im2col_3d_param_4]; ; CHECK-PTX64-NEXT: ld.param.b32 %r3, [cp_async_bulk_tensor_g2s_im2col_3d_param_5]; ; CHECK-PTX64-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_im2col_3d_param_6]; -; CHECK-PTX64-NEXT: cp.async.bulk.tensor.3d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes [%rd1], [%rd3, {%r1, %r2, %r3}], [%rd2], {%rs1}; +; CHECK-PTX64-NEXT: ld.param.b16 %rs2, [cp_async_bulk_tensor_g2s_im2col_3d_param_7]; ; CHECK-PTX64-NEXT: ld.param.b64 %rd4, [cp_async_bulk_tensor_g2s_im2col_3d_param_8]; +; CHECK-PTX64-NEXT: cp.async.bulk.tensor.3d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes [%rd1], [%rd3, {%r1, %r2, %r3}], [%rd2], {%rs1}; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.3d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes.L2::cache_hint [%rd1], [%rd3, {%r1, %r2, %r3}], [%rd2], {%rs1}, %rd4; -; CHECK-PTX64-NEXT: ld.param.b16 %rs2, [cp_async_bulk_tensor_g2s_im2col_3d_param_7]; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.3d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes.multicast::cluster [%rd1], [%rd3, {%r1, %r2, %r3}], [%rd2], {%rs1}, %rs2; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.3d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes.multicast::cluster.L2::cache_hint [%rd1], [%rd3, {%r1, %r2, %r3}], [%rd2], {%rs1}, %rs2, %rd4; ; CHECK-PTX64-NEXT: ret; @@ -319,10 +323,10 @@ define void @cp_async_bulk_tensor_g2s_im2col_3d(ptr addrspace(7) %d, ptr addrspa ; CHECK-PTX-SHARED32-NEXT: ld.param.b32 %r4, [cp_async_bulk_tensor_g2s_im2col_3d_param_4]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b32 %r5, [cp_async_bulk_tensor_g2s_im2col_3d_param_5]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_im2col_3d_param_6]; -; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.3d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes [%r1], [%rd1, {%r3, %r4, %r5}], [%r2], {%rs1}; +; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs2, [cp_async_bulk_tensor_g2s_im2col_3d_param_7]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b64 %rd2, [cp_async_bulk_tensor_g2s_im2col_3d_param_8]; +; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.3d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes [%r1], [%rd1, {%r3, %r4, %r5}], [%r2], {%rs1}; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.3d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes.L2::cache_hint [%r1], [%rd1, {%r3, %r4, %r5}], [%r2], {%rs1}, %rd2; -; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs2, [cp_async_bulk_tensor_g2s_im2col_3d_param_7]; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.3d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes.multicast::cluster [%r1], [%rd1, {%r3, %r4, %r5}], [%r2], {%rs1}, %rs2; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.3d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes.multicast::cluster.L2::cache_hint [%r1], [%rd1, {%r3, %r4, %r5}], [%r2], {%rs1}, %rs2, %rd2; ; CHECK-PTX-SHARED32-NEXT: ret; @@ -354,10 +358,10 @@ define void @cp_async_bulk_tensor_g2s_im2col_4d(ptr addrspace(7) %d, ptr addrspa ; CHECK-PTX64-NEXT: ld.param.b32 %r4, [cp_async_bulk_tensor_g2s_im2col_4d_param_6]; ; CHECK-PTX64-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_im2col_4d_param_7]; ; CHECK-PTX64-NEXT: ld.param.b16 %rs2, [cp_async_bulk_tensor_g2s_im2col_4d_param_8]; -; CHECK-PTX64-NEXT: cp.async.bulk.tensor.4d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes [%rd1], [%rd3, {%r1, %r2, %r3, %r4}], [%rd2], {%rs1, %rs2}; +; CHECK-PTX64-NEXT: ld.param.b16 %rs3, [cp_async_bulk_tensor_g2s_im2col_4d_param_9]; ; CHECK-PTX64-NEXT: ld.param.b64 %rd4, [cp_async_bulk_tensor_g2s_im2col_4d_param_10]; +; CHECK-PTX64-NEXT: cp.async.bulk.tensor.4d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes [%rd1], [%rd3, {%r1, %r2, %r3, %r4}], [%rd2], {%rs1, %rs2}; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.4d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes.L2::cache_hint [%rd1], [%rd3, {%r1, %r2, %r3, %r4}], [%rd2], {%rs1, %rs2}, %rd4; -; CHECK-PTX64-NEXT: ld.param.b16 %rs3, [cp_async_bulk_tensor_g2s_im2col_4d_param_9]; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.4d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes.multicast::cluster [%rd1], [%rd3, {%r1, %r2, %r3, %r4}], [%rd2], {%rs1, %rs2}, %rs3; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.4d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes.multicast::cluster.L2::cache_hint [%rd1], [%rd3, {%r1, %r2, %r3, %r4}], [%rd2], {%rs1, %rs2}, %rs3, %rd4; ; CHECK-PTX64-NEXT: ret; @@ -378,10 +382,10 @@ define void @cp_async_bulk_tensor_g2s_im2col_4d(ptr addrspace(7) %d, ptr addrspa ; CHECK-PTX-SHARED32-NEXT: ld.param.b32 %r6, [cp_async_bulk_tensor_g2s_im2col_4d_param_6]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_im2col_4d_param_7]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs2, [cp_async_bulk_tensor_g2s_im2col_4d_param_8]; -; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.4d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes [%r1], [%rd1, {%r3, %r4, %r5, %r6}], [%r2], {%rs1, %rs2}; +; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs3, [cp_async_bulk_tensor_g2s_im2col_4d_param_9]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b64 %rd2, [cp_async_bulk_tensor_g2s_im2col_4d_param_10]; +; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.4d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes [%r1], [%rd1, {%r3, %r4, %r5, %r6}], [%r2], {%rs1, %rs2}; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.4d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes.L2::cache_hint [%r1], [%rd1, {%r3, %r4, %r5, %r6}], [%r2], {%rs1, %rs2}, %rd2; -; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs3, [cp_async_bulk_tensor_g2s_im2col_4d_param_9]; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.4d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes.multicast::cluster [%r1], [%rd1, {%r3, %r4, %r5, %r6}], [%r2], {%rs1, %rs2}, %rs3; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.4d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes.multicast::cluster.L2::cache_hint [%r1], [%rd1, {%r3, %r4, %r5, %r6}], [%r2], {%rs1, %rs2}, %rs3, %rd2; ; CHECK-PTX-SHARED32-NEXT: ret; @@ -415,10 +419,10 @@ define void @cp_async_bulk_tensor_g2s_im2col_5d(ptr addrspace(7) %d, ptr addrspa ; CHECK-PTX64-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_im2col_5d_param_8]; ; CHECK-PTX64-NEXT: ld.param.b16 %rs2, [cp_async_bulk_tensor_g2s_im2col_5d_param_9]; ; CHECK-PTX64-NEXT: ld.param.b16 %rs3, [cp_async_bulk_tensor_g2s_im2col_5d_param_10]; -; CHECK-PTX64-NEXT: cp.async.bulk.tensor.5d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes [%rd1], [%rd3, {%r1, %r2, %r3, %r4, %r5}], [%rd2], {%rs1, %rs2, %rs3}; +; CHECK-PTX64-NEXT: ld.param.b16 %rs4, [cp_async_bulk_tensor_g2s_im2col_5d_param_11]; ; CHECK-PTX64-NEXT: ld.param.b64 %rd4, [cp_async_bulk_tensor_g2s_im2col_5d_param_12]; +; CHECK-PTX64-NEXT: cp.async.bulk.tensor.5d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes [%rd1], [%rd3, {%r1, %r2, %r3, %r4, %r5}], [%rd2], {%rs1, %rs2, %rs3}; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.5d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes.L2::cache_hint [%rd1], [%rd3, {%r1, %r2, %r3, %r4, %r5}], [%rd2], {%rs1, %rs2, %rs3}, %rd4; -; CHECK-PTX64-NEXT: ld.param.b16 %rs4, [cp_async_bulk_tensor_g2s_im2col_5d_param_11]; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.5d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes.multicast::cluster [%rd1], [%rd3, {%r1, %r2, %r3, %r4, %r5}], [%rd2], {%rs1, %rs2, %rs3}, %rs4; ; CHECK-PTX64-NEXT: cp.async.bulk.tensor.5d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes.multicast::cluster.L2::cache_hint [%rd1], [%rd3, {%r1, %r2, %r3, %r4, %r5}], [%rd2], {%rs1, %rs2, %rs3}, %rs4, %rd4; ; CHECK-PTX64-NEXT: ret; @@ -441,10 +445,10 @@ define void @cp_async_bulk_tensor_g2s_im2col_5d(ptr addrspace(7) %d, ptr addrspa ; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs1, [cp_async_bulk_tensor_g2s_im2col_5d_param_8]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs2, [cp_async_bulk_tensor_g2s_im2col_5d_param_9]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs3, [cp_async_bulk_tensor_g2s_im2col_5d_param_10]; -; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.5d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes [%r1], [%rd1, {%r3, %r4, %r5, %r6, %r7}], [%r2], {%rs1, %rs2, %rs3}; +; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs4, [cp_async_bulk_tensor_g2s_im2col_5d_param_11]; ; CHECK-PTX-SHARED32-NEXT: ld.param.b64 %rd2, [cp_async_bulk_tensor_g2s_im2col_5d_param_12]; +; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.5d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes [%r1], [%rd1, {%r3, %r4, %r5, %r6, %r7}], [%r2], {%rs1, %rs2, %rs3}; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.5d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes.L2::cache_hint [%r1], [%rd1, {%r3, %r4, %r5, %r6, %r7}], [%r2], {%rs1, %rs2, %rs3}, %rd2; -; CHECK-PTX-SHARED32-NEXT: ld.param.b16 %rs4, [cp_async_bulk_tensor_g2s_im2col_5d_param_11]; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.5d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes.multicast::cluster [%r1], [%rd1, {%r3, %r4, %r5, %r6, %r7}], [%r2], {%rs1, %rs2, %rs3}, %rs4; ; CHECK-PTX-SHARED32-NEXT: cp.async.bulk.tensor.5d.shared::cluster.global.im2col.mbarrier::complete_tx::bytes.multicast::cluster.L2::cache_hint [%r1], [%rd1, {%r3, %r4, %r5, %r6, %r7}], [%r2], {%rs1, %rs2, %rs3}, %rs4, %rd2; ; CHECK-PTX-SHARED32-NEXT: ret; diff --git a/llvm/test/CodeGen/NVPTX/insertelt-dynamic.ll b/llvm/test/CodeGen/NVPTX/insertelt-dynamic.ll new file mode 100644 index 0000000..f2ccf3e --- /dev/null +++ b/llvm/test/CodeGen/NVPTX/insertelt-dynamic.ll @@ -0,0 +1,738 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc < %s -mcpu=sm_20 | FileCheck %s +; RUN: %if ptxas %{ llc < %s -mcpu=sm_20 | %ptxas-verify %} +target triple = "nvptx64-nvidia-cuda" + +; Test dynamic insertelt at the beginning of a chain +define <4 x i32> @dynamic_at_beginning(i32 %idx) { +; CHECK-LABEL: dynamic_at_beginning( +; CHECK: { +; CHECK-NEXT: .local .align 4 .b8 __local_depot0[16]; +; CHECK-NEXT: .reg .b64 %SP; +; CHECK-NEXT: .reg .b64 %SPL; +; CHECK-NEXT: .reg .b32 %r<3>; +; CHECK-NEXT: .reg .b64 %rd<6>; +; CHECK-EMPTY: +; CHECK-NEXT: // %bb.0: +; CHECK-NEXT: mov.b64 %SPL, __local_depot0; +; CHECK-NEXT: cvta.local.u64 %SP, %SPL; +; CHECK-NEXT: ld.param.b32 %rd1, [dynamic_at_beginning_param_0]; +; CHECK-NEXT: and.b64 %rd2, %rd1, 3; +; CHECK-NEXT: shl.b64 %rd3, %rd2, 2; +; CHECK-NEXT: add.u64 %rd4, %SP, 0; +; CHECK-NEXT: add.s64 %rd5, %rd4, %rd3; +; CHECK-NEXT: st.b32 [%rd5], 10; +; CHECK-NEXT: ld.b32 %r1, [%SP+12]; +; CHECK-NEXT: ld.b32 %r2, [%SP]; +; CHECK-NEXT: st.param.v4.b32 [func_retval0], {%r2, 20, 30, %r1}; +; CHECK-NEXT: ret; + %v0 = insertelement <4 x i32> poison, i32 10, i32 %idx + %v1 = insertelement <4 x i32> %v0, i32 20, i32 1 + %v2 = insertelement <4 x i32> %v1, i32 30, i32 2 + ret <4 x i32> %v2 +} + +; Test dynamic insertelt at the end of a chain +define <4 x i32> @dynamic_at_end(i32 %idx) { +; CHECK-LABEL: dynamic_at_end( +; CHECK: { +; CHECK-NEXT: .local .align 4 .b8 __local_depot1[16]; +; CHECK-NEXT: .reg .b64 %SP; +; CHECK-NEXT: .reg .b64 %SPL; +; CHECK-NEXT: .reg .b32 %r<5>; +; CHECK-NEXT: .reg .b64 %rd<6>; +; CHECK-EMPTY: +; CHECK-NEXT: // %bb.0: +; CHECK-NEXT: mov.b64 %SPL, __local_depot1; +; CHECK-NEXT: cvta.local.u64 %SP, %SPL; +; CHECK-NEXT: ld.param.b32 %rd1, [dynamic_at_end_param_0]; +; CHECK-NEXT: and.b64 %rd2, %rd1, 3; +; CHECK-NEXT: shl.b64 %rd3, %rd2, 2; +; CHECK-NEXT: add.u64 %rd4, %SP, 0; +; CHECK-NEXT: add.s64 %rd5, %rd4, %rd3; +; CHECK-NEXT: st.b32 [%SP+4], 20; +; CHECK-NEXT: st.b32 [%SP], 10; +; CHECK-NEXT: st.b32 [%rd5], 30; +; CHECK-NEXT: ld.b32 %r1, [%SP+12]; +; CHECK-NEXT: ld.b32 %r2, [%SP+8]; +; CHECK-NEXT: ld.b32 %r3, [%SP+4]; +; CHECK-NEXT: ld.b32 %r4, [%SP]; +; CHECK-NEXT: st.param.v4.b32 [func_retval0], {%r4, %r3, %r2, %r1}; +; CHECK-NEXT: ret; + %v0 = insertelement <4 x i32> poison, i32 10, i32 0 + %v1 = insertelement <4 x i32> %v0, i32 20, i32 1 + %v2 = insertelement <4 x i32> %v1, i32 30, i32 %idx + ret <4 x i32> %v2 +} + +; Test dynamic insertelt in the middle of a chain +define <4 x i32> @dynamic_in_middle(i32 %idx) { +; CHECK-LABEL: dynamic_in_middle( +; CHECK: { +; CHECK-NEXT: .local .align 4 .b8 __local_depot2[16]; +; CHECK-NEXT: .reg .b64 %SP; +; CHECK-NEXT: .reg .b64 %SPL; +; CHECK-NEXT: .reg .b32 %r<4>; +; CHECK-NEXT: .reg .b64 %rd<6>; +; CHECK-EMPTY: +; CHECK-NEXT: // %bb.0: +; CHECK-NEXT: mov.b64 %SPL, __local_depot2; +; CHECK-NEXT: cvta.local.u64 %SP, %SPL; +; CHECK-NEXT: ld.param.b32 %rd1, [dynamic_in_middle_param_0]; +; CHECK-NEXT: and.b64 %rd2, %rd1, 3; +; CHECK-NEXT: shl.b64 %rd3, %rd2, 2; +; CHECK-NEXT: add.u64 %rd4, %SP, 0; +; CHECK-NEXT: add.s64 %rd5, %rd4, %rd3; +; CHECK-NEXT: st.b32 [%SP], 10; +; CHECK-NEXT: st.b32 [%rd5], 20; +; CHECK-NEXT: ld.b32 %r1, [%SP+12]; +; CHECK-NEXT: ld.b32 %r2, [%SP+4]; +; CHECK-NEXT: ld.b32 %r3, [%SP]; +; CHECK-NEXT: st.param.v4.b32 [func_retval0], {%r3, %r2, 30, %r1}; +; CHECK-NEXT: ret; + %v0 = insertelement <4 x i32> poison, i32 10, i32 0 + %v1 = insertelement <4 x i32> %v0, i32 20, i32 %idx + %v2 = insertelement <4 x i32> %v1, i32 30, i32 2 + ret <4 x i32> %v2 +} + +; Test repeated dynamic insertelt with the same index +define <4 x i32> @repeated_same_index(i32 %idx) { +; CHECK-LABEL: repeated_same_index( +; CHECK: { +; CHECK-NEXT: .local .align 4 .b8 __local_depot3[16]; +; CHECK-NEXT: .reg .b64 %SP; +; CHECK-NEXT: .reg .b64 %SPL; +; CHECK-NEXT: .reg .b32 %r<5>; +; CHECK-NEXT: .reg .b64 %rd<6>; +; CHECK-EMPTY: +; CHECK-NEXT: // %bb.0: +; CHECK-NEXT: mov.b64 %SPL, __local_depot3; +; CHECK-NEXT: cvta.local.u64 %SP, %SPL; +; CHECK-NEXT: ld.param.b32 %rd1, [repeated_same_index_param_0]; +; CHECK-NEXT: and.b64 %rd2, %rd1, 3; +; CHECK-NEXT: shl.b64 %rd3, %rd2, 2; +; CHECK-NEXT: add.u64 %rd4, %SP, 0; +; CHECK-NEXT: add.s64 %rd5, %rd4, %rd3; +; CHECK-NEXT: st.b32 [%rd5], 20; +; CHECK-NEXT: ld.b32 %r1, [%SP+12]; +; CHECK-NEXT: ld.b32 %r2, [%SP+8]; +; CHECK-NEXT: ld.b32 %r3, [%SP+4]; +; CHECK-NEXT: ld.b32 %r4, [%SP]; +; CHECK-NEXT: st.param.v4.b32 [func_retval0], {%r4, %r3, %r2, %r1}; +; CHECK-NEXT: ret; + %v0 = insertelement <4 x i32> poison, i32 10, i32 %idx + %v1 = insertelement <4 x i32> %v0, i32 20, i32 %idx + ret <4 x i32> %v1 +} + +; Test multiple dynamic insertelts +define <4 x i32> @multiple_dynamic(i32 %idx0, i32 %idx1) { +; CHECK-LABEL: multiple_dynamic( +; CHECK: { +; CHECK-NEXT: .local .align 4 .b8 __local_depot4[16]; +; CHECK-NEXT: .reg .b64 %SP; +; CHECK-NEXT: .reg .b64 %SPL; +; CHECK-NEXT: .reg .b32 %r<5>; +; CHECK-NEXT: .reg .b64 %rd<10>; +; CHECK-EMPTY: +; CHECK-NEXT: // %bb.0: +; CHECK-NEXT: mov.b64 %SPL, __local_depot4; +; CHECK-NEXT: cvta.local.u64 %SP, %SPL; +; CHECK-NEXT: ld.param.b32 %rd1, [multiple_dynamic_param_0]; +; CHECK-NEXT: and.b64 %rd2, %rd1, 3; +; CHECK-NEXT: shl.b64 %rd3, %rd2, 2; +; CHECK-NEXT: add.u64 %rd4, %SP, 0; +; CHECK-NEXT: add.s64 %rd5, %rd4, %rd3; +; CHECK-NEXT: st.b32 [%rd5], 10; +; CHECK-NEXT: ld.param.b32 %rd6, [multiple_dynamic_param_1]; +; CHECK-NEXT: and.b64 %rd7, %rd6, 3; +; CHECK-NEXT: shl.b64 %rd8, %rd7, 2; +; CHECK-NEXT: add.s64 %rd9, %rd4, %rd8; +; CHECK-NEXT: st.b32 [%rd9], 20; +; CHECK-NEXT: ld.b32 %r1, [%SP+12]; +; CHECK-NEXT: ld.b32 %r2, [%SP+8]; +; CHECK-NEXT: ld.b32 %r3, [%SP+4]; +; CHECK-NEXT: ld.b32 %r4, [%SP]; +; CHECK-NEXT: st.param.v4.b32 [func_retval0], {%r4, %r3, %r2, %r1}; +; CHECK-NEXT: ret; + %v0 = insertelement <4 x i32> poison, i32 10, i32 %idx0 + %v1 = insertelement <4 x i32> %v0, i32 20, i32 %idx1 + ret <4 x i32> %v1 +} + +; Test chain with all dynamic insertelts +define <4 x i32> @all_dynamic(i32 %idx0, i32 %idx1, i32 %idx2, i32 %idx3) { +; CHECK-LABEL: all_dynamic( +; CHECK: { +; CHECK-NEXT: .local .align 4 .b8 __local_depot5[16]; +; CHECK-NEXT: .reg .b64 %SP; +; CHECK-NEXT: .reg .b64 %SPL; +; CHECK-NEXT: .reg .b32 %r<5>; +; CHECK-NEXT: .reg .b64 %rd<18>; +; CHECK-EMPTY: +; CHECK-NEXT: // %bb.0: +; CHECK-NEXT: mov.b64 %SPL, __local_depot5; +; CHECK-NEXT: cvta.local.u64 %SP, %SPL; +; CHECK-NEXT: ld.param.b32 %rd1, [all_dynamic_param_0]; +; CHECK-NEXT: and.b64 %rd2, %rd1, 3; +; CHECK-NEXT: shl.b64 %rd3, %rd2, 2; +; CHECK-NEXT: add.u64 %rd4, %SP, 0; +; CHECK-NEXT: add.s64 %rd5, %rd4, %rd3; +; CHECK-NEXT: ld.param.b32 %rd6, [all_dynamic_param_1]; +; CHECK-NEXT: and.b64 %rd7, %rd6, 3; +; CHECK-NEXT: shl.b64 %rd8, %rd7, 2; +; CHECK-NEXT: add.s64 %rd9, %rd4, %rd8; +; CHECK-NEXT: ld.param.b32 %rd10, [all_dynamic_param_2]; +; CHECK-NEXT: and.b64 %rd11, %rd10, 3; +; CHECK-NEXT: shl.b64 %rd12, %rd11, 2; +; CHECK-NEXT: add.s64 %rd13, %rd4, %rd12; +; CHECK-NEXT: st.b32 [%rd5], 10; +; CHECK-NEXT: st.b32 [%rd9], 20; +; CHECK-NEXT: st.b32 [%rd13], 30; +; CHECK-NEXT: ld.param.b32 %rd14, [all_dynamic_param_3]; +; CHECK-NEXT: and.b64 %rd15, %rd14, 3; +; CHECK-NEXT: shl.b64 %rd16, %rd15, 2; +; CHECK-NEXT: add.s64 %rd17, %rd4, %rd16; +; CHECK-NEXT: st.b32 [%rd17], 40; +; CHECK-NEXT: ld.b32 %r1, [%SP+12]; +; CHECK-NEXT: ld.b32 %r2, [%SP+8]; +; CHECK-NEXT: ld.b32 %r3, [%SP+4]; +; CHECK-NEXT: ld.b32 %r4, [%SP]; +; CHECK-NEXT: st.param.v4.b32 [func_retval0], {%r4, %r3, %r2, %r1}; +; CHECK-NEXT: ret; + %v0 = insertelement <4 x i32> poison, i32 10, i32 %idx0 + %v1 = insertelement <4 x i32> %v0, i32 20, i32 %idx1 + %v2 = insertelement <4 x i32> %v1, i32 30, i32 %idx2 + %v3 = insertelement <4 x i32> %v2, i32 40, i32 %idx3 + ret <4 x i32> %v3 +} + +; Test mixed constant and dynamic insertelts with high ratio of dynamic ones. +; Should lower all insertelts to stores. +define <4 x i32> @mix_dynamic_constant(i32 %idx0, i32 %idx1) { +; CHECK-LABEL: mix_dynamic_constant( +; CHECK: { +; CHECK-NEXT: .local .align 4 .b8 __local_depot6[16]; +; CHECK-NEXT: .reg .b64 %SP; +; CHECK-NEXT: .reg .b64 %SPL; +; CHECK-NEXT: .reg .b32 %r<5>; +; CHECK-NEXT: .reg .b64 %rd<10>; +; CHECK-EMPTY: +; CHECK-NEXT: // %bb.0: +; CHECK-NEXT: mov.b64 %SPL, __local_depot6; +; CHECK-NEXT: cvta.local.u64 %SP, %SPL; +; CHECK-NEXT: ld.param.b32 %rd1, [mix_dynamic_constant_param_0]; +; CHECK-NEXT: and.b64 %rd2, %rd1, 3; +; CHECK-NEXT: shl.b64 %rd3, %rd2, 2; +; CHECK-NEXT: add.u64 %rd4, %SP, 0; +; CHECK-NEXT: add.s64 %rd5, %rd4, %rd3; +; CHECK-NEXT: st.b32 [%rd5], 10; +; CHECK-NEXT: ld.param.b32 %rd6, [mix_dynamic_constant_param_1]; +; CHECK-NEXT: and.b64 %rd7, %rd6, 3; +; CHECK-NEXT: shl.b64 %rd8, %rd7, 2; +; CHECK-NEXT: add.s64 %rd9, %rd4, %rd8; +; CHECK-NEXT: st.b32 [%SP+4], 20; +; CHECK-NEXT: st.b32 [%rd9], 30; +; CHECK-NEXT: ld.b32 %r1, [%SP+12]; +; CHECK-NEXT: ld.b32 %r2, [%SP+8]; +; CHECK-NEXT: ld.b32 %r3, [%SP+4]; +; CHECK-NEXT: ld.b32 %r4, [%SP]; +; CHECK-NEXT: st.param.v4.b32 [func_retval0], {%r4, %r3, %r2, %r1}; +; CHECK-NEXT: ret; + %v0 = insertelement <4 x i32> poison, i32 10, i32 %idx0 + %v1 = insertelement <4 x i32> %v0, i32 20, i32 1 + %v2 = insertelement <4 x i32> %v1, i32 30, i32 %idx1 + ret <4 x i32> %v2 +} + +; Test two separate chains that don't interfere +define void @two_separate_chains(i32 %idx0, i32 %idx1, ptr %out0, ptr %out1) { +; CHECK-LABEL: two_separate_chains( +; CHECK: { +; CHECK-NEXT: .local .align 4 .b8 __local_depot7[32]; +; CHECK-NEXT: .reg .b64 %SP; +; CHECK-NEXT: .reg .b64 %SPL; +; CHECK-NEXT: .reg .b32 %r<7>; +; CHECK-NEXT: .reg .b64 %rd<13>; +; CHECK-EMPTY: +; CHECK-NEXT: // %bb.0: +; CHECK-NEXT: mov.b64 %SPL, __local_depot7; +; CHECK-NEXT: cvta.local.u64 %SP, %SPL; +; CHECK-NEXT: ld.param.b32 %rd1, [two_separate_chains_param_0]; +; CHECK-NEXT: and.b64 %rd2, %rd1, 3; +; CHECK-NEXT: shl.b64 %rd3, %rd2, 2; +; CHECK-NEXT: add.u64 %rd4, %SP, 16; +; CHECK-NEXT: add.s64 %rd5, %rd4, %rd3; +; CHECK-NEXT: st.b32 [%rd5], 10; +; CHECK-NEXT: ld.param.b32 %rd6, [two_separate_chains_param_1]; +; CHECK-NEXT: and.b64 %rd7, %rd6, 3; +; CHECK-NEXT: shl.b64 %rd8, %rd7, 2; +; CHECK-NEXT: add.u64 %rd9, %SP, 0; +; CHECK-NEXT: add.s64 %rd10, %rd9, %rd8; +; CHECK-NEXT: ld.b32 %r1, [%SP+28]; +; CHECK-NEXT: ld.b32 %r2, [%SP+24]; +; CHECK-NEXT: ld.b32 %r3, [%SP+16]; +; CHECK-NEXT: ld.param.b64 %rd11, [two_separate_chains_param_2]; +; CHECK-NEXT: st.b32 [%rd10], 30; +; CHECK-NEXT: ld.param.b64 %rd12, [two_separate_chains_param_3]; +; CHECK-NEXT: ld.b32 %r4, [%SP+12]; +; CHECK-NEXT: ld.b32 %r5, [%SP+4]; +; CHECK-NEXT: ld.b32 %r6, [%SP]; +; CHECK-NEXT: st.v4.b32 [%rd11], {%r3, 20, %r2, %r1}; +; CHECK-NEXT: st.v4.b32 [%rd12], {%r6, %r5, 40, %r4}; +; CHECK-NEXT: ret; + ; Chain 1 + %v0 = insertelement <4 x i32> poison, i32 10, i32 %idx0 + %v1 = insertelement <4 x i32> %v0, i32 20, i32 1 + + ; Chain 2 + %w0 = insertelement <4 x i32> poison, i32 30, i32 %idx1 + %w1 = insertelement <4 x i32> %w0, i32 40, i32 2 + + store <4 x i32> %v1, ptr %out0 + store <4 x i32> %w1, ptr %out1 + ret void +} + +; Test overlapping chains (chain 2 starts from middle of chain 1) +define void @overlapping_chains(i32 %idx0, i32 %idx1, ptr %out0, ptr %out1) { +; CHECK-LABEL: overlapping_chains( +; CHECK: { +; CHECK-NEXT: .local .align 4 .b8 __local_depot8[32]; +; CHECK-NEXT: .reg .b64 %SP; +; CHECK-NEXT: .reg .b64 %SPL; +; CHECK-NEXT: .reg .b32 %r<7>; +; CHECK-NEXT: .reg .b64 %rd<14>; +; CHECK-EMPTY: +; CHECK-NEXT: // %bb.0: +; CHECK-NEXT: mov.b64 %SPL, __local_depot8; +; CHECK-NEXT: cvta.local.u64 %SP, %SPL; +; CHECK-NEXT: ld.param.b32 %rd1, [overlapping_chains_param_0]; +; CHECK-NEXT: and.b64 %rd2, %rd1, 3; +; CHECK-NEXT: shl.b64 %rd3, %rd2, 2; +; CHECK-NEXT: add.u64 %rd4, %SP, 16; +; CHECK-NEXT: add.s64 %rd5, %rd4, %rd3; +; CHECK-NEXT: st.b32 [%rd5], 10; +; CHECK-NEXT: add.u64 %rd6, %SP, 0; +; CHECK-NEXT: add.s64 %rd7, %rd6, %rd3; +; CHECK-NEXT: ld.b32 %r1, [%SP+28]; +; CHECK-NEXT: ld.b32 %r2, [%SP+16]; +; CHECK-NEXT: ld.param.b64 %rd8, [overlapping_chains_param_2]; +; CHECK-NEXT: st.b32 [%rd7], 10; +; CHECK-NEXT: ld.param.b32 %rd9, [overlapping_chains_param_1]; +; CHECK-NEXT: and.b64 %rd10, %rd9, 3; +; CHECK-NEXT: shl.b64 %rd11, %rd10, 2; +; CHECK-NEXT: add.s64 %rd12, %rd6, %rd11; +; CHECK-NEXT: st.b32 [%SP+4], 20; +; CHECK-NEXT: st.b32 [%rd12], 30; +; CHECK-NEXT: ld.param.b64 %rd13, [overlapping_chains_param_3]; +; CHECK-NEXT: ld.b32 %r3, [%SP+12]; +; CHECK-NEXT: ld.b32 %r4, [%SP+8]; +; CHECK-NEXT: ld.b32 %r5, [%SP+4]; +; CHECK-NEXT: ld.b32 %r6, [%SP]; +; CHECK-NEXT: st.v4.b32 [%rd8], {%r2, 20, 40, %r1}; +; CHECK-NEXT: st.v4.b32 [%rd13], {%r6, %r5, %r4, %r3}; +; CHECK-NEXT: ret; + %v0 = insertelement <4 x i32> poison, i32 10, i32 %idx0 + %v1 = insertelement <4 x i32> %v0, i32 20, i32 1 + + ; Chain 2 starts from v1 + %w0 = insertelement <4 x i32> %v1, i32 30, i32 %idx1 + + ; Continue chain 1 + %v2 = insertelement <4 x i32> %v1, i32 40, i32 2 + + store <4 x i32> %v2, ptr %out0 + store <4 x i32> %w0, ptr %out1 + ret void +} + +; Test with i1 elements (1-bit, non-byte-aligned) +define <8 x i1> @dynamic_i1(i32 %idx) { +; CHECK-LABEL: dynamic_i1( +; CHECK: { +; CHECK-NEXT: .local .align 8 .b8 __local_depot9[8]; +; CHECK-NEXT: .reg .b64 %SP; +; CHECK-NEXT: .reg .b64 %SPL; +; CHECK-NEXT: .reg .b32 %r<9>; +; CHECK-NEXT: .reg .b64 %rd<5>; +; CHECK-EMPTY: +; CHECK-NEXT: // %bb.0: +; CHECK-NEXT: mov.b64 %SPL, __local_depot9; +; CHECK-NEXT: cvta.local.u64 %SP, %SPL; +; CHECK-NEXT: ld.param.b32 %rd1, [dynamic_i1_param_0]; +; CHECK-NEXT: and.b64 %rd2, %rd1, 7; +; CHECK-NEXT: add.u64 %rd3, %SP, 0; +; CHECK-NEXT: or.b64 %rd4, %rd3, %rd2; +; CHECK-NEXT: st.v2.b32 [%SP], {%r1, %r2}; +; CHECK-NEXT: st.b8 [%rd4], 1; +; CHECK-NEXT: ld.b32 %r3, [%SP]; +; CHECK-NEXT: prmt.b32 %r4, %r3, 0, 0x7773U; +; CHECK-NEXT: ld.b32 %r5, [%SP+4]; +; CHECK-NEXT: prmt.b32 %r6, %r5, 0, 0x7771U; +; CHECK-NEXT: prmt.b32 %r7, %r5, 0, 0x7772U; +; CHECK-NEXT: prmt.b32 %r8, %r5, 0, 0x7773U; +; CHECK-NEXT: st.param.b8 [func_retval0+4], %r5; +; CHECK-NEXT: st.param.b8 [func_retval0], %r3; +; CHECK-NEXT: st.param.b8 [func_retval0+7], %r8; +; CHECK-NEXT: st.param.b8 [func_retval0+6], %r7; +; CHECK-NEXT: st.param.b8 [func_retval0+5], %r6; +; CHECK-NEXT: st.param.b8 [func_retval0+3], %r4; +; CHECK-NEXT: st.param.b8 [func_retval0+2], 1; +; CHECK-NEXT: st.param.b8 [func_retval0+1], 0; +; CHECK-NEXT: ret; + %v0 = insertelement <8 x i1> poison, i1 1, i32 %idx + %v1 = insertelement <8 x i1> %v0, i1 0, i32 1 + %v2 = insertelement <8 x i1> %v1, i1 1, i32 2 + ret <8 x i1> %v2 +} + +; Test with i2 elements (2-bit, non-byte-aligned) +define <8 x i2> @dynamic_i2(i32 %idx) { +; CHECK-LABEL: dynamic_i2( +; CHECK: { +; CHECK-NEXT: .local .align 8 .b8 __local_depot10[16]; +; CHECK-NEXT: .reg .b64 %SP; +; CHECK-NEXT: .reg .b64 %SPL; +; CHECK-NEXT: .reg .b16 %rs<24>; +; CHECK-NEXT: .reg .b32 %r<10>; +; CHECK-NEXT: .reg .b64 %rd<5>; +; CHECK-EMPTY: +; CHECK-NEXT: // %bb.0: +; CHECK-NEXT: mov.b64 %SPL, __local_depot10; +; CHECK-NEXT: cvta.local.u64 %SP, %SPL; +; CHECK-NEXT: ld.param.b32 %rd1, [dynamic_i2_param_0]; +; CHECK-NEXT: and.b64 %rd2, %rd1, 7; +; CHECK-NEXT: add.u64 %rd3, %SP, 0; +; CHECK-NEXT: or.b64 %rd4, %rd3, %rd2; +; CHECK-NEXT: st.v2.b32 [%SP], {%r1, %r2}; +; CHECK-NEXT: st.b8 [%rd4], 1; +; CHECK-NEXT: ld.b32 %r3, [%SP+4]; +; CHECK-NEXT: cvt.u16.u32 %rs1, %r3; +; CHECK-NEXT: and.b16 %rs2, %rs1, 3; +; CHECK-NEXT: prmt.b32 %r4, %r3, 0, 0x7771U; +; CHECK-NEXT: cvt.u16.u32 %rs3, %r4; +; CHECK-NEXT: and.b16 %rs4, %rs3, 3; +; CHECK-NEXT: shl.b16 %rs5, %rs4, 2; +; CHECK-NEXT: or.b16 %rs6, %rs2, %rs5; +; CHECK-NEXT: prmt.b32 %r5, %r3, 0, 0x7772U; +; CHECK-NEXT: cvt.u16.u32 %rs7, %r5; +; CHECK-NEXT: and.b16 %rs8, %rs7, 3; +; CHECK-NEXT: shl.b16 %rs9, %rs8, 4; +; CHECK-NEXT: or.b16 %rs10, %rs6, %rs9; +; CHECK-NEXT: prmt.b32 %r6, %r3, 0, 0x7773U; +; CHECK-NEXT: cvt.u16.u32 %rs11, %r6; +; CHECK-NEXT: shl.b16 %rs12, %rs11, 6; +; CHECK-NEXT: or.b16 %rs13, %rs10, %rs12; +; CHECK-NEXT: st.b8 [%SP+8], %rs13; +; CHECK-NEXT: ld.b32 %r7, [%SP]; +; CHECK-NEXT: prmt.b32 %r8, %r7, 0, 0x7773U; +; CHECK-NEXT: cvt.u16.u32 %rs14, %r8; +; CHECK-NEXT: shl.b16 %rs15, %rs14, 6; +; CHECK-NEXT: and.b16 %rs16, %rs15, 192; +; CHECK-NEXT: ld.s8 %rs17, [%SP+8]; +; CHECK-NEXT: shl.b16 %rs18, %rs17, 8; +; CHECK-NEXT: or.b16 %rs19, %rs16, %rs18; +; CHECK-NEXT: prmt.b32 %r9, %r7, 0, 0x7770U; +; CHECK-NEXT: st.param.b16 [func_retval0], %r9; +; CHECK-NEXT: st.param.b16 [func_retval0+8], %rs17; +; CHECK-NEXT: shr.s16 %rs20, %rs18, 14; +; CHECK-NEXT: st.param.b16 [func_retval0+14], %rs20; +; CHECK-NEXT: shr.s16 %rs21, %rs18, 12; +; CHECK-NEXT: st.param.b16 [func_retval0+12], %rs21; +; CHECK-NEXT: shr.s16 %rs22, %rs18, 10; +; CHECK-NEXT: st.param.b16 [func_retval0+10], %rs22; +; CHECK-NEXT: shr.s16 %rs23, %rs19, 6; +; CHECK-NEXT: st.param.b16 [func_retval0+6], %rs23; +; CHECK-NEXT: st.param.b16 [func_retval0+4], 3; +; CHECK-NEXT: st.param.b16 [func_retval0+2], 2; +; CHECK-NEXT: ret; + %v0 = insertelement <8 x i2> poison, i2 1, i32 %idx + %v1 = insertelement <8 x i2> %v0, i2 2, i32 1 + %v2 = insertelement <8 x i2> %v1, i2 3, i32 2 + ret <8 x i2> %v2 +} + +; Test with i3 elements (3-bit, non-byte-aligned) +define <8 x i3> @dynamic_i3(i32 %idx) { +; CHECK-LABEL: dynamic_i3( +; CHECK: { +; CHECK-NEXT: .local .align 8 .b8 __local_depot11[8]; +; CHECK-NEXT: .reg .b64 %SP; +; CHECK-NEXT: .reg .b64 %SPL; +; CHECK-NEXT: .reg .b16 %rs<5>; +; CHECK-NEXT: .reg .b32 %r<15>; +; CHECK-NEXT: .reg .b64 %rd<5>; +; CHECK-EMPTY: +; CHECK-NEXT: // %bb.0: +; CHECK-NEXT: mov.b64 %SPL, __local_depot11; +; CHECK-NEXT: cvta.local.u64 %SP, %SPL; +; CHECK-NEXT: ld.param.b32 %rd1, [dynamic_i3_param_0]; +; CHECK-NEXT: and.b64 %rd2, %rd1, 7; +; CHECK-NEXT: add.u64 %rd3, %SP, 0; +; CHECK-NEXT: or.b64 %rd4, %rd3, %rd2; +; CHECK-NEXT: st.v2.b32 [%SP], {%r1, %r2}; +; CHECK-NEXT: st.b8 [%rd4], 1; +; CHECK-NEXT: ld.b32 %r3, [%SP]; +; CHECK-NEXT: ld.b32 %r4, [%SP+4]; +; CHECK-NEXT: prmt.b32 %r5, %r4, 0, 0x7773U; +; CHECK-NEXT: prmt.b32 %r6, %r4, 0, 0x7772U; +; CHECK-NEXT: prmt.b32 %r7, %r6, %r5, 0x5410U; +; CHECK-NEXT: st.param.b32 [func_retval0+12], %r7; +; CHECK-NEXT: prmt.b32 %r8, %r4, 0, 0x7771U; +; CHECK-NEXT: prmt.b32 %r9, %r4, 0, 0x7770U; +; CHECK-NEXT: prmt.b32 %r10, %r9, %r8, 0x5410U; +; CHECK-NEXT: st.param.b32 [func_retval0+8], %r10; +; CHECK-NEXT: prmt.b32 %r11, %r3, 0, 0x7773U; +; CHECK-NEXT: cvt.u16.u32 %rs1, %r11; +; CHECK-NEXT: mov.b16 %rs2, 3; +; CHECK-NEXT: mov.b32 %r12, {%rs2, %rs1}; +; CHECK-NEXT: st.param.b32 [func_retval0+4], %r12; +; CHECK-NEXT: prmt.b32 %r13, %r3, 0, 0x7770U; +; CHECK-NEXT: cvt.u16.u32 %rs3, %r13; +; CHECK-NEXT: mov.b16 %rs4, 2; +; CHECK-NEXT: mov.b32 %r14, {%rs3, %rs4}; +; CHECK-NEXT: st.param.b32 [func_retval0], %r14; +; CHECK-NEXT: ret; + %v0 = insertelement <8 x i3> poison, i3 1, i32 %idx + %v1 = insertelement <8 x i3> %v0, i3 2, i32 1 + %v2 = insertelement <8 x i3> %v1, i3 3, i32 2 + ret <8 x i3> %v2 +} + +; Test with i4 elements (4-bit, non-byte-aligned) +define <8 x i4> @dynamic_i4(i32 %idx) { +; CHECK-LABEL: dynamic_i4( +; CHECK: { +; CHECK-NEXT: .local .align 8 .b8 __local_depot12[16]; +; CHECK-NEXT: .reg .b64 %SP; +; CHECK-NEXT: .reg .b64 %SPL; +; CHECK-NEXT: .reg .b16 %rs<30>; +; CHECK-NEXT: .reg .b32 %r<22>; +; CHECK-NEXT: .reg .b64 %rd<5>; +; CHECK-EMPTY: +; CHECK-NEXT: // %bb.0: +; CHECK-NEXT: mov.b64 %SPL, __local_depot12; +; CHECK-NEXT: cvta.local.u64 %SP, %SPL; +; CHECK-NEXT: ld.param.b32 %rd1, [dynamic_i4_param_0]; +; CHECK-NEXT: and.b64 %rd2, %rd1, 7; +; CHECK-NEXT: add.u64 %rd3, %SP, 0; +; CHECK-NEXT: or.b64 %rd4, %rd3, %rd2; +; CHECK-NEXT: st.v2.b32 [%SP], {%r1, %r2}; +; CHECK-NEXT: st.b8 [%rd4], 1; +; CHECK-NEXT: ld.b32 %r3, [%SP]; +; CHECK-NEXT: prmt.b32 %r4, %r3, 0, 0x7770U; +; CHECK-NEXT: cvt.u16.u32 %rs1, %r4; +; CHECK-NEXT: and.b16 %rs2, %rs1, 15; +; CHECK-NEXT: prmt.b32 %r5, %r3, 0, 0x7771U; +; CHECK-NEXT: cvt.u16.u32 %rs3, %r5; +; CHECK-NEXT: and.b16 %rs4, %rs3, 15; +; CHECK-NEXT: shl.b16 %rs5, %rs4, 4; +; CHECK-NEXT: or.b16 %rs6, %rs2, %rs5; +; CHECK-NEXT: prmt.b32 %r6, %r3, 0, 0x7772U; +; CHECK-NEXT: cvt.u16.u32 %rs7, %r6; +; CHECK-NEXT: and.b16 %rs8, %rs7, 15; +; CHECK-NEXT: shl.b16 %rs9, %rs8, 8; +; CHECK-NEXT: or.b16 %rs10, %rs6, %rs9; +; CHECK-NEXT: prmt.b32 %r7, %r3, 0, 0x7773U; +; CHECK-NEXT: cvt.u16.u32 %rs11, %r7; +; CHECK-NEXT: shl.b16 %rs12, %rs11, 12; +; CHECK-NEXT: or.b16 %rs13, %rs10, %rs12; +; CHECK-NEXT: cvt.u32.u16 %r8, %rs13; +; CHECK-NEXT: ld.b32 %r9, [%SP+4]; +; CHECK-NEXT: prmt.b32 %r10, %r9, 0, 0x7770U; +; CHECK-NEXT: cvt.u16.u32 %rs14, %r10; +; CHECK-NEXT: and.b16 %rs15, %rs14, 15; +; CHECK-NEXT: prmt.b32 %r11, %r9, 0, 0x7771U; +; CHECK-NEXT: cvt.u16.u32 %rs16, %r11; +; CHECK-NEXT: and.b16 %rs17, %rs16, 15; +; CHECK-NEXT: shl.b16 %rs18, %rs17, 4; +; CHECK-NEXT: or.b16 %rs19, %rs15, %rs18; +; CHECK-NEXT: prmt.b32 %r12, %r9, 0, 0x7772U; +; CHECK-NEXT: cvt.u16.u32 %rs20, %r12; +; CHECK-NEXT: and.b16 %rs21, %rs20, 15; +; CHECK-NEXT: shl.b16 %rs22, %rs21, 8; +; CHECK-NEXT: or.b16 %rs23, %rs19, %rs22; +; CHECK-NEXT: prmt.b32 %r13, %r9, 0, 0x7773U; +; CHECK-NEXT: cvt.u16.u32 %rs24, %r13; +; CHECK-NEXT: shl.b16 %rs25, %rs24, 12; +; CHECK-NEXT: or.b16 %rs26, %rs23, %rs25; +; CHECK-NEXT: cvt.u32.u16 %r14, %rs26; +; CHECK-NEXT: shl.b32 %r15, %r14, 16; +; CHECK-NEXT: or.b32 %r16, %r8, %r15; +; CHECK-NEXT: mov.b32 %r17, {%rs20, %rs24}; +; CHECK-NEXT: st.param.b32 [func_retval0+12], %r17; +; CHECK-NEXT: mov.b32 %r18, {%rs14, %rs16}; +; CHECK-NEXT: st.param.b32 [func_retval0+8], %r18; +; CHECK-NEXT: mov.b16 %rs27, 2; +; CHECK-NEXT: mov.b32 %r19, {%rs1, %rs27}; +; CHECK-NEXT: st.param.b32 [func_retval0], %r19; +; CHECK-NEXT: shr.u32 %r20, %r16, 12; +; CHECK-NEXT: cvt.u16.u32 %rs28, %r20; +; CHECK-NEXT: mov.b16 %rs29, 3; +; CHECK-NEXT: mov.b32 %r21, {%rs29, %rs28}; +; CHECK-NEXT: st.param.b32 [func_retval0+4], %r21; +; CHECK-NEXT: ret; + %v0 = insertelement <8 x i4> poison, i4 1, i32 %idx + %v1 = insertelement <8 x i4> %v0, i4 2, i32 1 + %v2 = insertelement <8 x i4> %v1, i4 3, i32 2 + ret <8 x i4> %v2 +} + +; Test with i5 elements (5-bit, non-byte-aligned) +define <8 x i5> @dynamic_i5(i32 %idx) { +; CHECK-LABEL: dynamic_i5( +; CHECK: { +; CHECK-NEXT: .local .align 8 .b8 __local_depot13[8]; +; CHECK-NEXT: .reg .b64 %SP; +; CHECK-NEXT: .reg .b64 %SPL; +; CHECK-NEXT: .reg .b16 %rs<5>; +; CHECK-NEXT: .reg .b32 %r<15>; +; CHECK-NEXT: .reg .b64 %rd<5>; +; CHECK-EMPTY: +; CHECK-NEXT: // %bb.0: +; CHECK-NEXT: mov.b64 %SPL, __local_depot13; +; CHECK-NEXT: cvta.local.u64 %SP, %SPL; +; CHECK-NEXT: ld.param.b32 %rd1, [dynamic_i5_param_0]; +; CHECK-NEXT: and.b64 %rd2, %rd1, 7; +; CHECK-NEXT: add.u64 %rd3, %SP, 0; +; CHECK-NEXT: or.b64 %rd4, %rd3, %rd2; +; CHECK-NEXT: st.v2.b32 [%SP], {%r1, %r2}; +; CHECK-NEXT: st.b8 [%rd4], 1; +; CHECK-NEXT: ld.b32 %r3, [%SP]; +; CHECK-NEXT: ld.b32 %r4, [%SP+4]; +; CHECK-NEXT: prmt.b32 %r5, %r4, 0, 0x7773U; +; CHECK-NEXT: prmt.b32 %r6, %r4, 0, 0x7772U; +; CHECK-NEXT: prmt.b32 %r7, %r6, %r5, 0x5410U; +; CHECK-NEXT: prmt.b32 %r8, %r4, 0, 0x7771U; +; CHECK-NEXT: prmt.b32 %r9, %r4, 0, 0x7770U; +; CHECK-NEXT: prmt.b32 %r10, %r9, %r8, 0x5410U; +; CHECK-NEXT: st.param.v2.b32 [func_retval0+8], {%r10, %r7}; +; CHECK-NEXT: prmt.b32 %r11, %r3, 0, 0x7773U; +; CHECK-NEXT: cvt.u16.u32 %rs1, %r11; +; CHECK-NEXT: mov.b16 %rs2, 3; +; CHECK-NEXT: mov.b32 %r12, {%rs2, %rs1}; +; CHECK-NEXT: prmt.b32 %r13, %r3, 0, 0x7770U; +; CHECK-NEXT: cvt.u16.u32 %rs3, %r13; +; CHECK-NEXT: mov.b16 %rs4, 2; +; CHECK-NEXT: mov.b32 %r14, {%rs3, %rs4}; +; CHECK-NEXT: st.param.v2.b32 [func_retval0], {%r14, %r12}; +; CHECK-NEXT: ret; + %v0 = insertelement <8 x i5> poison, i5 1, i32 %idx + %v1 = insertelement <8 x i5> %v0, i5 2, i32 1 + %v2 = insertelement <8 x i5> %v1, i5 3, i32 2 + ret <8 x i5> %v2 +} + +; Test with i7 elements (7-bit, non-byte-aligned) +define <8 x i7> @dynamic_i7(i32 %idx) { +; CHECK-LABEL: dynamic_i7( +; CHECK: { +; CHECK-NEXT: .local .align 8 .b8 __local_depot14[8]; +; CHECK-NEXT: .reg .b64 %SP; +; CHECK-NEXT: .reg .b64 %SPL; +; CHECK-NEXT: .reg .b16 %rs<5>; +; CHECK-NEXT: .reg .b32 %r<15>; +; CHECK-NEXT: .reg .b64 %rd<5>; +; CHECK-EMPTY: +; CHECK-NEXT: // %bb.0: +; CHECK-NEXT: mov.b64 %SPL, __local_depot14; +; CHECK-NEXT: cvta.local.u64 %SP, %SPL; +; CHECK-NEXT: ld.param.b32 %rd1, [dynamic_i7_param_0]; +; CHECK-NEXT: and.b64 %rd2, %rd1, 7; +; CHECK-NEXT: add.u64 %rd3, %SP, 0; +; CHECK-NEXT: or.b64 %rd4, %rd3, %rd2; +; CHECK-NEXT: st.v2.b32 [%SP], {%r1, %r2}; +; CHECK-NEXT: st.b8 [%rd4], 1; +; CHECK-NEXT: ld.b32 %r3, [%SP]; +; CHECK-NEXT: ld.b32 %r4, [%SP+4]; +; CHECK-NEXT: prmt.b32 %r5, %r4, 0, 0x7773U; +; CHECK-NEXT: prmt.b32 %r6, %r4, 0, 0x7772U; +; CHECK-NEXT: prmt.b32 %r7, %r6, %r5, 0x5410U; +; CHECK-NEXT: prmt.b32 %r8, %r4, 0, 0x7771U; +; CHECK-NEXT: prmt.b32 %r9, %r4, 0, 0x7770U; +; CHECK-NEXT: prmt.b32 %r10, %r9, %r8, 0x5410U; +; CHECK-NEXT: st.param.v2.b32 [func_retval0+8], {%r10, %r7}; +; CHECK-NEXT: prmt.b32 %r11, %r3, 0, 0x7773U; +; CHECK-NEXT: cvt.u16.u32 %rs1, %r11; +; CHECK-NEXT: mov.b16 %rs2, 3; +; CHECK-NEXT: mov.b32 %r12, {%rs2, %rs1}; +; CHECK-NEXT: prmt.b32 %r13, %r3, 0, 0x7770U; +; CHECK-NEXT: cvt.u16.u32 %rs3, %r13; +; CHECK-NEXT: mov.b16 %rs4, 2; +; CHECK-NEXT: mov.b32 %r14, {%rs3, %rs4}; +; CHECK-NEXT: st.param.v2.b32 [func_retval0], {%r14, %r12}; +; CHECK-NEXT: ret; + %v0 = insertelement <8 x i7> poison, i7 1, i32 %idx + %v1 = insertelement <8 x i7> %v0, i7 2, i32 1 + %v2 = insertelement <8 x i7> %v1, i7 3, i32 2 + ret <8 x i7> %v2 +} + +; Test with i6 elements (6-bit, non-byte-aligned) +define <8 x i6> @dynamic_i6(i32 %idx) { +; CHECK-LABEL: dynamic_i6( +; CHECK: { +; CHECK-NEXT: .local .align 8 .b8 __local_depot15[8]; +; CHECK-NEXT: .reg .b64 %SP; +; CHECK-NEXT: .reg .b64 %SPL; +; CHECK-NEXT: .reg .b16 %rs<5>; +; CHECK-NEXT: .reg .b32 %r<15>; +; CHECK-NEXT: .reg .b64 %rd<5>; +; CHECK-EMPTY: +; CHECK-NEXT: // %bb.0: +; CHECK-NEXT: mov.b64 %SPL, __local_depot15; +; CHECK-NEXT: cvta.local.u64 %SP, %SPL; +; CHECK-NEXT: ld.param.b32 %rd1, [dynamic_i6_param_0]; +; CHECK-NEXT: and.b64 %rd2, %rd1, 7; +; CHECK-NEXT: add.u64 %rd3, %SP, 0; +; CHECK-NEXT: or.b64 %rd4, %rd3, %rd2; +; CHECK-NEXT: st.v2.b32 [%SP], {%r1, %r2}; +; CHECK-NEXT: st.b8 [%rd4], 1; +; CHECK-NEXT: ld.b32 %r3, [%SP]; +; CHECK-NEXT: ld.b32 %r4, [%SP+4]; +; CHECK-NEXT: prmt.b32 %r5, %r4, 0, 0x7773U; +; CHECK-NEXT: prmt.b32 %r6, %r4, 0, 0x7772U; +; CHECK-NEXT: prmt.b32 %r7, %r6, %r5, 0x5410U; +; CHECK-NEXT: prmt.b32 %r8, %r4, 0, 0x7771U; +; CHECK-NEXT: prmt.b32 %r9, %r4, 0, 0x7770U; +; CHECK-NEXT: prmt.b32 %r10, %r9, %r8, 0x5410U; +; CHECK-NEXT: st.param.v2.b32 [func_retval0+8], {%r10, %r7}; +; CHECK-NEXT: prmt.b32 %r11, %r3, 0, 0x7773U; +; CHECK-NEXT: cvt.u16.u32 %rs1, %r11; +; CHECK-NEXT: mov.b16 %rs2, 3; +; CHECK-NEXT: mov.b32 %r12, {%rs2, %rs1}; +; CHECK-NEXT: prmt.b32 %r13, %r3, 0, 0x7770U; +; CHECK-NEXT: cvt.u16.u32 %rs3, %r13; +; CHECK-NEXT: mov.b16 %rs4, 2; +; CHECK-NEXT: mov.b32 %r14, {%rs3, %rs4}; +; CHECK-NEXT: st.param.v2.b32 [func_retval0], {%r14, %r12}; +; CHECK-NEXT: ret; + %v0 = insertelement <8 x i6> poison, i6 1, i32 %idx + %v1 = insertelement <8 x i6> %v0, i6 2, i32 1 + %v2 = insertelement <8 x i6> %v1, i6 3, i32 2 + ret <8 x i6> %v2 +} + +; Test with multiple dynamic insertions on i3 elements +define <4 x i3> @multiple_dynamic_i3(i32 %idx0, i32 %idx1) { +; CHECK-LABEL: multiple_dynamic_i3( +; CHECK: { +; CHECK-NEXT: .reg .b32 %r<9>; +; CHECK-EMPTY: +; CHECK-NEXT: // %bb.0: +; CHECK-NEXT: ld.param.b32 %r1, [multiple_dynamic_i3_param_0]; +; CHECK-NEXT: shl.b32 %r2, %r1, 3; +; CHECK-NEXT: bfi.b32 %r3, 1, %r4, %r2, 8; +; CHECK-NEXT: ld.param.b32 %r5, [multiple_dynamic_i3_param_1]; +; CHECK-NEXT: shl.b32 %r6, %r5, 3; +; CHECK-NEXT: bfi.b32 %r7, 2, %r3, %r6, 8; +; CHECK-NEXT: st.param.b16 [func_retval0], %r7; +; CHECK-NEXT: shr.u32 %r8, %r7, 16; +; CHECK-NEXT: st.param.b16 [func_retval0+2], %r8; +; CHECK-NEXT: ret; + %v0 = insertelement <4 x i3> poison, i3 1, i32 %idx0 + %v1 = insertelement <4 x i3> %v0, i3 2, i32 %idx1 + ret <4 x i3> %v1 +} diff --git a/llvm/test/CodeGen/PowerPC/combine-sext-and-shl-after-isel.ll b/llvm/test/CodeGen/PowerPC/combine-sext-and-shl-after-isel.ll index 00a77f9..530169f 100644 --- a/llvm/test/CodeGen/PowerPC/combine-sext-and-shl-after-isel.ll +++ b/llvm/test/CodeGen/PowerPC/combine-sext-and-shl-after-isel.ll @@ -212,37 +212,33 @@ define hidden void @testCaller(i1 %incond) local_unnamed_addr align 2 nounwind { ; CHECK-NEXT: std r30, 48(r1) # 8-byte Folded Spill ; CHECK-NEXT: andi. r3, r3, 1 ; CHECK-NEXT: li r3, -1 +; CHECK-NEXT: li r4, 0 ; CHECK-NEXT: li r30, 0 ; CHECK-NEXT: crmove 4*cr2+lt, gt ; CHECK-NEXT: std r29, 40(r1) # 8-byte Folded Spill ; CHECK-NEXT: b .LBB3_2 -; CHECK-NEXT: .p2align 4 ; CHECK-NEXT: .LBB3_1: # %if.end116 ; CHECK-NEXT: # ; CHECK-NEXT: bl callee ; CHECK-NEXT: nop ; CHECK-NEXT: mr r3, r29 -; CHECK-NEXT: .LBB3_2: # %cond.end.i.i -; CHECK-NEXT: # =>This Loop Header: Depth=1 -; CHECK-NEXT: # Child Loop BB3_3 Depth 2 -; CHECK-NEXT: lwz r29, 0(r3) -; CHECK-NEXT: li r5, 0 -; CHECK-NEXT: extsw r4, r29 -; CHECK-NEXT: .p2align 5 -; CHECK-NEXT: .LBB3_3: # %while.body5.i -; CHECK-NEXT: # Parent Loop BB3_2 Depth=1 -; CHECK-NEXT: # => This Inner Loop Header: Depth=2 -; CHECK-NEXT: addi r5, r5, -1 -; CHECK-NEXT: cmpwi r5, 0 -; CHECK-NEXT: bgt cr0, .LBB3_3 -; CHECK-NEXT: # %bb.4: # %while.cond12.preheader.i +; CHECK-NEXT: li r4, 0 +; CHECK-NEXT: .p2align 4 +; CHECK-NEXT: .LBB3_2: # %while.body5.i ; CHECK-NEXT: # +; CHECK-NEXT: addi r4, r4, -1 +; CHECK-NEXT: cmpwi r4, 0 +; CHECK-NEXT: bgt cr0, .LBB3_2 +; CHECK-NEXT: # %bb.3: # %while.cond12.preheader.i +; CHECK-NEXT: # +; CHECK-NEXT: lwz r29, 0(r3) ; CHECK-NEXT: bc 12, 4*cr2+lt, .LBB3_1 -; CHECK-NEXT: # %bb.5: # %for.cond99.preheader +; CHECK-NEXT: # %bb.4: # %for.cond99.preheader ; CHECK-NEXT: # +; CHECK-NEXT: extsw r4, r29 ; CHECK-NEXT: ld r5, 0(r3) -; CHECK-NEXT: sldi r4, r4, 2 ; CHECK-NEXT: stw r3, 0(r3) +; CHECK-NEXT: sldi r4, r4, 2 ; CHECK-NEXT: stwx r30, r5, r4 ; CHECK-NEXT: b .LBB3_1 ; @@ -256,37 +252,33 @@ define hidden void @testCaller(i1 %incond) local_unnamed_addr align 2 nounwind { ; CHECK-BE-NEXT: std r30, 64(r1) # 8-byte Folded Spill ; CHECK-BE-NEXT: andi. r3, r3, 1 ; CHECK-BE-NEXT: li r3, -1 +; CHECK-BE-NEXT: li r4, 0 ; CHECK-BE-NEXT: li r30, 0 ; CHECK-BE-NEXT: crmove 4*cr2+lt, gt ; CHECK-BE-NEXT: std r29, 56(r1) # 8-byte Folded Spill ; CHECK-BE-NEXT: b .LBB3_2 -; CHECK-BE-NEXT: .p2align 4 ; CHECK-BE-NEXT: .LBB3_1: # %if.end116 ; CHECK-BE-NEXT: # ; CHECK-BE-NEXT: bl callee ; CHECK-BE-NEXT: nop ; CHECK-BE-NEXT: mr r3, r29 -; CHECK-BE-NEXT: .LBB3_2: # %cond.end.i.i -; CHECK-BE-NEXT: # =>This Loop Header: Depth=1 -; CHECK-BE-NEXT: # Child Loop BB3_3 Depth 2 -; CHECK-BE-NEXT: lwz r29, 0(r3) -; CHECK-BE-NEXT: li r5, 0 -; CHECK-BE-NEXT: extsw r4, r29 -; CHECK-BE-NEXT: .p2align 5 -; CHECK-BE-NEXT: .LBB3_3: # %while.body5.i -; CHECK-BE-NEXT: # Parent Loop BB3_2 Depth=1 -; CHECK-BE-NEXT: # => This Inner Loop Header: Depth=2 -; CHECK-BE-NEXT: addi r5, r5, -1 -; CHECK-BE-NEXT: cmpwi r5, 0 -; CHECK-BE-NEXT: bgt cr0, .LBB3_3 -; CHECK-BE-NEXT: # %bb.4: # %while.cond12.preheader.i +; CHECK-BE-NEXT: li r4, 0 +; CHECK-BE-NEXT: .p2align 4 +; CHECK-BE-NEXT: .LBB3_2: # %while.body5.i +; CHECK-BE-NEXT: # +; CHECK-BE-NEXT: addi r4, r4, -1 +; CHECK-BE-NEXT: cmpwi r4, 0 +; CHECK-BE-NEXT: bgt cr0, .LBB3_2 +; CHECK-BE-NEXT: # %bb.3: # %while.cond12.preheader.i ; CHECK-BE-NEXT: # +; CHECK-BE-NEXT: lwz r29, 0(r3) ; CHECK-BE-NEXT: bc 12, 4*cr2+lt, .LBB3_1 -; CHECK-BE-NEXT: # %bb.5: # %for.cond99.preheader +; CHECK-BE-NEXT: # %bb.4: # %for.cond99.preheader ; CHECK-BE-NEXT: # +; CHECK-BE-NEXT: extsw r4, r29 ; CHECK-BE-NEXT: ld r5, 0(r3) -; CHECK-BE-NEXT: sldi r4, r4, 2 ; CHECK-BE-NEXT: stw r3, 0(r3) +; CHECK-BE-NEXT: sldi r4, r4, 2 ; CHECK-BE-NEXT: stwx r30, r5, r4 ; CHECK-BE-NEXT: b .LBB3_1 ; @@ -300,32 +292,28 @@ define hidden void @testCaller(i1 %incond) local_unnamed_addr align 2 nounwind { ; CHECK-P9-NEXT: std r0, 80(r1) ; CHECK-P9-NEXT: std r30, 48(r1) # 8-byte Folded Spill ; CHECK-P9-NEXT: li r3, -1 +; CHECK-P9-NEXT: li r4, 0 ; CHECK-P9-NEXT: li r30, 0 ; CHECK-P9-NEXT: std r29, 40(r1) # 8-byte Folded Spill ; CHECK-P9-NEXT: crmove 4*cr2+lt, gt ; CHECK-P9-NEXT: b .LBB3_2 -; CHECK-P9-NEXT: .p2align 4 ; CHECK-P9-NEXT: .LBB3_1: # %if.end116 ; CHECK-P9-NEXT: # ; CHECK-P9-NEXT: bl callee ; CHECK-P9-NEXT: nop ; CHECK-P9-NEXT: mr r3, r29 -; CHECK-P9-NEXT: .LBB3_2: # %cond.end.i.i -; CHECK-P9-NEXT: # =>This Loop Header: Depth=1 -; CHECK-P9-NEXT: # Child Loop BB3_3 Depth 2 -; CHECK-P9-NEXT: lwz r29, 0(r3) ; CHECK-P9-NEXT: li r4, 0 -; CHECK-P9-NEXT: .p2align 5 -; CHECK-P9-NEXT: .LBB3_3: # %while.body5.i -; CHECK-P9-NEXT: # Parent Loop BB3_2 Depth=1 -; CHECK-P9-NEXT: # => This Inner Loop Header: Depth=2 +; CHECK-P9-NEXT: .p2align 4 +; CHECK-P9-NEXT: .LBB3_2: # %while.body5.i +; CHECK-P9-NEXT: # ; CHECK-P9-NEXT: addi r4, r4, -1 ; CHECK-P9-NEXT: cmpwi r4, 0 -; CHECK-P9-NEXT: bgt cr0, .LBB3_3 -; CHECK-P9-NEXT: # %bb.4: # %while.cond12.preheader.i +; CHECK-P9-NEXT: bgt cr0, .LBB3_2 +; CHECK-P9-NEXT: # %bb.3: # %while.cond12.preheader.i ; CHECK-P9-NEXT: # +; CHECK-P9-NEXT: lwz r29, 0(r3) ; CHECK-P9-NEXT: bc 12, 4*cr2+lt, .LBB3_1 -; CHECK-P9-NEXT: # %bb.5: # %for.cond99.preheader +; CHECK-P9-NEXT: # %bb.4: # %for.cond99.preheader ; CHECK-P9-NEXT: # ; CHECK-P9-NEXT: ld r4, 0(r3) ; CHECK-P9-NEXT: extswsli r5, r29, 2 @@ -343,32 +331,28 @@ define hidden void @testCaller(i1 %incond) local_unnamed_addr align 2 nounwind { ; CHECK-P9-BE-NEXT: std r0, 96(r1) ; CHECK-P9-BE-NEXT: std r30, 64(r1) # 8-byte Folded Spill ; CHECK-P9-BE-NEXT: li r3, -1 +; CHECK-P9-BE-NEXT: li r4, 0 ; CHECK-P9-BE-NEXT: li r30, 0 ; CHECK-P9-BE-NEXT: std r29, 56(r1) # 8-byte Folded Spill ; CHECK-P9-BE-NEXT: crmove 4*cr2+lt, gt ; CHECK-P9-BE-NEXT: b .LBB3_2 -; CHECK-P9-BE-NEXT: .p2align 4 ; CHECK-P9-BE-NEXT: .LBB3_1: # %if.end116 ; CHECK-P9-BE-NEXT: # ; CHECK-P9-BE-NEXT: bl callee ; CHECK-P9-BE-NEXT: nop ; CHECK-P9-BE-NEXT: mr r3, r29 -; CHECK-P9-BE-NEXT: .LBB3_2: # %cond.end.i.i -; CHECK-P9-BE-NEXT: # =>This Loop Header: Depth=1 -; CHECK-P9-BE-NEXT: # Child Loop BB3_3 Depth 2 -; CHECK-P9-BE-NEXT: lwz r29, 0(r3) ; CHECK-P9-BE-NEXT: li r4, 0 -; CHECK-P9-BE-NEXT: .p2align 5 -; CHECK-P9-BE-NEXT: .LBB3_3: # %while.body5.i -; CHECK-P9-BE-NEXT: # Parent Loop BB3_2 Depth=1 -; CHECK-P9-BE-NEXT: # => This Inner Loop Header: Depth=2 +; CHECK-P9-BE-NEXT: .p2align 4 +; CHECK-P9-BE-NEXT: .LBB3_2: # %while.body5.i +; CHECK-P9-BE-NEXT: # ; CHECK-P9-BE-NEXT: addi r4, r4, -1 ; CHECK-P9-BE-NEXT: cmpwi r4, 0 -; CHECK-P9-BE-NEXT: bgt cr0, .LBB3_3 -; CHECK-P9-BE-NEXT: # %bb.4: # %while.cond12.preheader.i +; CHECK-P9-BE-NEXT: bgt cr0, .LBB3_2 +; CHECK-P9-BE-NEXT: # %bb.3: # %while.cond12.preheader.i ; CHECK-P9-BE-NEXT: # +; CHECK-P9-BE-NEXT: lwz r29, 0(r3) ; CHECK-P9-BE-NEXT: bc 12, 4*cr2+lt, .LBB3_1 -; CHECK-P9-BE-NEXT: # %bb.5: # %for.cond99.preheader +; CHECK-P9-BE-NEXT: # %bb.4: # %for.cond99.preheader ; CHECK-P9-BE-NEXT: # ; CHECK-P9-BE-NEXT: ld r4, 0(r3) ; CHECK-P9-BE-NEXT: extswsli r5, r29, 2 diff --git a/llvm/test/CodeGen/PowerPC/vec_insert_elt.ll b/llvm/test/CodeGen/PowerPC/vec_insert_elt.ll index 291a9c1..b006c78 100644 --- a/llvm/test/CodeGen/PowerPC/vec_insert_elt.ll +++ b/llvm/test/CodeGen/PowerPC/vec_insert_elt.ll @@ -242,17 +242,14 @@ define <2 x i64> @testDoubleword(<2 x i64> %a, i64 %b, i64 %idx) { ; AIX-P8-32-LABEL: testDoubleword: ; AIX-P8-32: # %bb.0: # %entry ; AIX-P8-32-NEXT: add r6, r6, r6 -; AIX-P8-32-NEXT: addi r5, r1, -32 +; AIX-P8-32-NEXT: addi r5, r1, -16 ; AIX-P8-32-NEXT: rlwinm r7, r6, 2, 28, 29 -; AIX-P8-32-NEXT: stxvw4x v2, 0, r5 +; AIX-P8-32-NEXT: stxvd2x v2, 0, r5 ; AIX-P8-32-NEXT: stwx r3, r5, r7 -; AIX-P8-32-NEXT: addi r3, r1, -16 -; AIX-P8-32-NEXT: lxvw4x vs0, 0, r5 -; AIX-P8-32-NEXT: addi r5, r6, 1 -; AIX-P8-32-NEXT: rlwinm r5, r5, 2, 28, 29 -; AIX-P8-32-NEXT: stxvw4x vs0, 0, r3 -; AIX-P8-32-NEXT: stwx r4, r3, r5 -; AIX-P8-32-NEXT: lxvw4x v2, 0, r3 +; AIX-P8-32-NEXT: addi r3, r6, 1 +; AIX-P8-32-NEXT: rlwinm r3, r3, 2, 28, 29 +; AIX-P8-32-NEXT: stwx r4, r5, r3 +; AIX-P8-32-NEXT: lxvd2x v2, 0, r5 ; AIX-P8-32-NEXT: blr entry: %vecins = insertelement <2 x i64> %a, i64 %b, i64 %idx @@ -426,17 +423,14 @@ define <4 x float> @testFloat2(<4 x float> %a, ptr %b, i32 zeroext %idx1, i32 ze ; AIX-P8-LABEL: testFloat2: ; AIX-P8: # %bb.0: # %entry ; AIX-P8-NEXT: lwz r6, 0(r3) -; AIX-P8-NEXT: rlwinm r4, r4, 2, 28, 29 -; AIX-P8-NEXT: addi r7, r1, -32 +; AIX-P8-NEXT: lwz r3, 1(r3) +; AIX-P8-NEXT: addi r7, r1, -16 ; AIX-P8-NEXT: stxvw4x v2, 0, r7 ; AIX-P8-NEXT: rlwinm r5, r5, 2, 28, 29 +; AIX-P8-NEXT: rlwinm r4, r4, 2, 28, 29 ; AIX-P8-NEXT: stwx r6, r7, r4 -; AIX-P8-NEXT: addi r4, r1, -16 -; AIX-P8-NEXT: lxvw4x vs0, 0, r7 -; AIX-P8-NEXT: lwz r3, 1(r3) -; AIX-P8-NEXT: stxvw4x vs0, 0, r4 -; AIX-P8-NEXT: stwx r3, r4, r5 -; AIX-P8-NEXT: lxvw4x v2, 0, r4 +; AIX-P8-NEXT: stwx r3, r7, r5 +; AIX-P8-NEXT: lxvw4x v2, 0, r7 ; AIX-P8-NEXT: blr entry: %add.ptr1 = getelementptr inbounds i8, ptr %b, i64 1 @@ -493,38 +487,32 @@ define <4 x float> @testFloat3(<4 x float> %a, ptr %b, i32 zeroext %idx1, i32 ze ; ; AIX-P8-64-LABEL: testFloat3: ; AIX-P8-64: # %bb.0: # %entry +; AIX-P8-64-NEXT: li r7, 1 ; AIX-P8-64-NEXT: lis r6, 1 -; AIX-P8-64-NEXT: rlwinm r4, r4, 2, 28, 29 -; AIX-P8-64-NEXT: addi r7, r1, -32 ; AIX-P8-64-NEXT: rlwinm r5, r5, 2, 28, 29 +; AIX-P8-64-NEXT: rlwinm r4, r4, 2, 28, 29 +; AIX-P8-64-NEXT: rldic r7, r7, 36, 27 ; AIX-P8-64-NEXT: lwzx r6, r3, r6 +; AIX-P8-64-NEXT: lwzx r3, r3, r7 +; AIX-P8-64-NEXT: addi r7, r1, -16 ; AIX-P8-64-NEXT: stxvw4x v2, 0, r7 ; AIX-P8-64-NEXT: stwx r6, r7, r4 -; AIX-P8-64-NEXT: li r4, 1 -; AIX-P8-64-NEXT: lxvw4x vs0, 0, r7 -; AIX-P8-64-NEXT: rldic r4, r4, 36, 27 -; AIX-P8-64-NEXT: lwzx r3, r3, r4 -; AIX-P8-64-NEXT: addi r4, r1, -16 -; AIX-P8-64-NEXT: stxvw4x vs0, 0, r4 -; AIX-P8-64-NEXT: stwx r3, r4, r5 -; AIX-P8-64-NEXT: lxvw4x v2, 0, r4 +; AIX-P8-64-NEXT: stwx r3, r7, r5 +; AIX-P8-64-NEXT: lxvw4x v2, 0, r7 ; AIX-P8-64-NEXT: blr ; ; AIX-P8-32-LABEL: testFloat3: ; AIX-P8-32: # %bb.0: # %entry ; AIX-P8-32-NEXT: lis r6, 1 -; AIX-P8-32-NEXT: rlwinm r4, r4, 2, 28, 29 -; AIX-P8-32-NEXT: addi r7, r1, -32 ; AIX-P8-32-NEXT: rlwinm r5, r5, 2, 28, 29 +; AIX-P8-32-NEXT: rlwinm r4, r4, 2, 28, 29 +; AIX-P8-32-NEXT: addi r7, r1, -16 ; AIX-P8-32-NEXT: lwzx r6, r3, r6 +; AIX-P8-32-NEXT: lwz r3, 0(r3) ; AIX-P8-32-NEXT: stxvw4x v2, 0, r7 ; AIX-P8-32-NEXT: stwx r6, r7, r4 -; AIX-P8-32-NEXT: addi r4, r1, -16 -; AIX-P8-32-NEXT: lxvw4x vs0, 0, r7 -; AIX-P8-32-NEXT: lwz r3, 0(r3) -; AIX-P8-32-NEXT: stxvw4x vs0, 0, r4 -; AIX-P8-32-NEXT: stwx r3, r4, r5 -; AIX-P8-32-NEXT: lxvw4x v2, 0, r4 +; AIX-P8-32-NEXT: stwx r3, r7, r5 +; AIX-P8-32-NEXT: lxvw4x v2, 0, r7 ; AIX-P8-32-NEXT: blr entry: %add.ptr = getelementptr inbounds i8, ptr %b, i64 65536 diff --git a/llvm/test/CodeGen/RISCV/GlobalISel/rvv/vse.ll b/llvm/test/CodeGen/RISCV/GlobalISel/rvv/vse.ll new file mode 100644 index 0000000..785d9fc --- /dev/null +++ b/llvm/test/CodeGen/RISCV/GlobalISel/rvv/vse.ll @@ -0,0 +1,1575 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v,+zvfhmin,+zvfbfmin \ +; RUN: -global-isel -verify-machineinstrs -target-abi=ilp32d | FileCheck %s +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v,+zvfhmin,+zvfbfmin \ +; RUN: -global-isel -verify-machineinstrs -target-abi=lp64d | FileCheck %s + +declare void @llvm.riscv.vse.nxv1i64( + <vscale x 1 x i64>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv1i64_nxv1i64(<vscale x 1 x i64> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv1i64_nxv1i64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e64, m1, ta, ma +; CHECK-NEXT: vse64.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv1i64( + <vscale x 1 x i64> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv1i64( + <vscale x 1 x i64>, + ptr, + <vscale x 1 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv1i64_nxv1i64(<vscale x 1 x i64> %0, ptr %1, <vscale x 1 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv1i64_nxv1i64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e64, m1, ta, ma +; CHECK-NEXT: vse64.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv1i64( + <vscale x 1 x i64> %0, + ptr %1, + <vscale x 1 x i1> %2, + iXLen %3) + + ret void +} + +define void @intrinsic_vse_allonesmask_v_nxv1i64_nxv1i64(<vscale x 1 x i64> %0, ptr %1, <vscale x 1 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_allonesmask_v_nxv1i64_nxv1i64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e64, m1, ta, ma +; CHECK-NEXT: vse64.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv1i64( + <vscale x 1 x i64> %0, + ptr %1, + <vscale x 1 x i1> splat (i1 true), + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv2i64( + <vscale x 2 x i64>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv2i64_nxv2i64(<vscale x 2 x i64> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv2i64_nxv2i64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e64, m2, ta, ma +; CHECK-NEXT: vse64.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv2i64( + <vscale x 2 x i64> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv2i64( + <vscale x 2 x i64>, + ptr, + <vscale x 2 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv2i64_nxv2i64(<vscale x 2 x i64> %0, ptr %1, <vscale x 2 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv2i64_nxv2i64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e64, m2, ta, ma +; CHECK-NEXT: vse64.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv2i64( + <vscale x 2 x i64> %0, + ptr %1, + <vscale x 2 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv4i64( + <vscale x 4 x i64>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv4i64_nxv4i64(<vscale x 4 x i64> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv4i64_nxv4i64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e64, m4, ta, ma +; CHECK-NEXT: vse64.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv4i64( + <vscale x 4 x i64> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv4i64( + <vscale x 4 x i64>, + ptr, + <vscale x 4 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv4i64_nxv4i64(<vscale x 4 x i64> %0, ptr %1, <vscale x 4 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv4i64_nxv4i64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e64, m4, ta, ma +; CHECK-NEXT: vse64.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv4i64( + <vscale x 4 x i64> %0, + ptr %1, + <vscale x 4 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv8i64( + <vscale x 8 x i64>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv8i64_nxv8i64(<vscale x 8 x i64> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv8i64_nxv8i64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e64, m8, ta, ma +; CHECK-NEXT: vse64.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv8i64( + <vscale x 8 x i64> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv8i64( + <vscale x 8 x i64>, + ptr, + <vscale x 8 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv8i64_nxv8i64(<vscale x 8 x i64> %0, ptr %1, <vscale x 8 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv8i64_nxv8i64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e64, m8, ta, ma +; CHECK-NEXT: vse64.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv8i64( + <vscale x 8 x i64> %0, + ptr %1, + <vscale x 8 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv1f64( + <vscale x 1 x double>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv1f64_nxv1f64(<vscale x 1 x double> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv1f64_nxv1f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e64, m1, ta, ma +; CHECK-NEXT: vse64.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv1f64( + <vscale x 1 x double> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv1f64( + <vscale x 1 x double>, + ptr, + <vscale x 1 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv1f64_nxv1f64(<vscale x 1 x double> %0, ptr %1, <vscale x 1 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv1f64_nxv1f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e64, m1, ta, ma +; CHECK-NEXT: vse64.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv1f64( + <vscale x 1 x double> %0, + ptr %1, + <vscale x 1 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv2f64( + <vscale x 2 x double>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv2f64_nxv2f64(<vscale x 2 x double> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv2f64_nxv2f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e64, m2, ta, ma +; CHECK-NEXT: vse64.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv2f64( + <vscale x 2 x double> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv2f64( + <vscale x 2 x double>, + ptr, + <vscale x 2 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv2f64_nxv2f64(<vscale x 2 x double> %0, ptr %1, <vscale x 2 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv2f64_nxv2f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e64, m2, ta, ma +; CHECK-NEXT: vse64.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv2f64( + <vscale x 2 x double> %0, + ptr %1, + <vscale x 2 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv4f64( + <vscale x 4 x double>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv4f64_nxv4f64(<vscale x 4 x double> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv4f64_nxv4f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e64, m4, ta, ma +; CHECK-NEXT: vse64.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv4f64( + <vscale x 4 x double> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv4f64( + <vscale x 4 x double>, + ptr, + <vscale x 4 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv4f64_nxv4f64(<vscale x 4 x double> %0, ptr %1, <vscale x 4 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv4f64_nxv4f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e64, m4, ta, ma +; CHECK-NEXT: vse64.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv4f64( + <vscale x 4 x double> %0, + ptr %1, + <vscale x 4 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv8f64( + <vscale x 8 x double>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv8f64_nxv8f64(<vscale x 8 x double> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv8f64_nxv8f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e64, m8, ta, ma +; CHECK-NEXT: vse64.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv8f64( + <vscale x 8 x double> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv8f64( + <vscale x 8 x double>, + ptr, + <vscale x 8 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv8f64_nxv8f64(<vscale x 8 x double> %0, ptr %1, <vscale x 8 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv8f64_nxv8f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e64, m8, ta, ma +; CHECK-NEXT: vse64.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv8f64( + <vscale x 8 x double> %0, + ptr %1, + <vscale x 8 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv1i32( + <vscale x 1 x i32>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv1i32_nxv1i32(<vscale x 1 x i32> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv1i32_nxv1i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, mf2, ta, ma +; CHECK-NEXT: vse32.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv1i32( + <vscale x 1 x i32> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv1i32( + <vscale x 1 x i32>, + ptr, + <vscale x 1 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv1i32_nxv1i32(<vscale x 1 x i32> %0, ptr %1, <vscale x 1 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv1i32_nxv1i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, mf2, ta, ma +; CHECK-NEXT: vse32.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv1i32( + <vscale x 1 x i32> %0, + ptr %1, + <vscale x 1 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv2i32( + <vscale x 2 x i32>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv2i32_nxv2i32(<vscale x 2 x i32> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv2i32_nxv2i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, m1, ta, ma +; CHECK-NEXT: vse32.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv2i32( + <vscale x 2 x i32> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv2i32( + <vscale x 2 x i32>, + ptr, + <vscale x 2 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv2i32_nxv2i32(<vscale x 2 x i32> %0, ptr %1, <vscale x 2 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv2i32_nxv2i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, m1, ta, ma +; CHECK-NEXT: vse32.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv2i32( + <vscale x 2 x i32> %0, + ptr %1, + <vscale x 2 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv4i32( + <vscale x 4 x i32>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv4i32_nxv4i32(<vscale x 4 x i32> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv4i32_nxv4i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, m2, ta, ma +; CHECK-NEXT: vse32.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv4i32( + <vscale x 4 x i32> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv4i32( + <vscale x 4 x i32>, + ptr, + <vscale x 4 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv4i32_nxv4i32(<vscale x 4 x i32> %0, ptr %1, <vscale x 4 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv4i32_nxv4i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, m2, ta, ma +; CHECK-NEXT: vse32.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv4i32( + <vscale x 4 x i32> %0, + ptr %1, + <vscale x 4 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv8i32( + <vscale x 8 x i32>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv8i32_nxv8i32(<vscale x 8 x i32> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv8i32_nxv8i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, m4, ta, ma +; CHECK-NEXT: vse32.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv8i32( + <vscale x 8 x i32> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv8i32( + <vscale x 8 x i32>, + ptr, + <vscale x 8 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv8i32_nxv8i32(<vscale x 8 x i32> %0, ptr %1, <vscale x 8 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv8i32_nxv8i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, m4, ta, ma +; CHECK-NEXT: vse32.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv8i32( + <vscale x 8 x i32> %0, + ptr %1, + <vscale x 8 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv16i32( + <vscale x 16 x i32>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv16i32_nxv16i32(<vscale x 16 x i32> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv16i32_nxv16i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, m8, ta, ma +; CHECK-NEXT: vse32.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv16i32( + <vscale x 16 x i32> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv16i32( + <vscale x 16 x i32>, + ptr, + <vscale x 16 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv16i32_nxv16i32(<vscale x 16 x i32> %0, ptr %1, <vscale x 16 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv16i32_nxv16i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, m8, ta, ma +; CHECK-NEXT: vse32.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv16i32( + <vscale x 16 x i32> %0, + ptr %1, + <vscale x 16 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv1f32( + <vscale x 1 x float>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv1f32_nxv1f32(<vscale x 1 x float> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv1f32_nxv1f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, mf2, ta, ma +; CHECK-NEXT: vse32.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv1f32( + <vscale x 1 x float> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv1f32( + <vscale x 1 x float>, + ptr, + <vscale x 1 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv1f32_nxv1f32(<vscale x 1 x float> %0, ptr %1, <vscale x 1 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv1f32_nxv1f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, mf2, ta, ma +; CHECK-NEXT: vse32.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv1f32( + <vscale x 1 x float> %0, + ptr %1, + <vscale x 1 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv2f32( + <vscale x 2 x float>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv2f32_nxv2f32(<vscale x 2 x float> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv2f32_nxv2f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, m1, ta, ma +; CHECK-NEXT: vse32.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv2f32( + <vscale x 2 x float> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv2f32( + <vscale x 2 x float>, + ptr, + <vscale x 2 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv2f32_nxv2f32(<vscale x 2 x float> %0, ptr %1, <vscale x 2 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv2f32_nxv2f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, m1, ta, ma +; CHECK-NEXT: vse32.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv2f32( + <vscale x 2 x float> %0, + ptr %1, + <vscale x 2 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv4f32( + <vscale x 4 x float>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv4f32_nxv4f32(<vscale x 4 x float> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv4f32_nxv4f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, m2, ta, ma +; CHECK-NEXT: vse32.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv4f32( + <vscale x 4 x float> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv4f32( + <vscale x 4 x float>, + ptr, + <vscale x 4 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv4f32_nxv4f32(<vscale x 4 x float> %0, ptr %1, <vscale x 4 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv4f32_nxv4f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, m2, ta, ma +; CHECK-NEXT: vse32.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv4f32( + <vscale x 4 x float> %0, + ptr %1, + <vscale x 4 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv8f32( + <vscale x 8 x float>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv8f32_nxv8f32(<vscale x 8 x float> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv8f32_nxv8f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, m4, ta, ma +; CHECK-NEXT: vse32.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv8f32( + <vscale x 8 x float> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv8f32( + <vscale x 8 x float>, + ptr, + <vscale x 8 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv8f32_nxv8f32(<vscale x 8 x float> %0, ptr %1, <vscale x 8 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv8f32_nxv8f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, m4, ta, ma +; CHECK-NEXT: vse32.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv8f32( + <vscale x 8 x float> %0, + ptr %1, + <vscale x 8 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv16f32( + <vscale x 16 x float>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv16f32_nxv16f32(<vscale x 16 x float> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv16f32_nxv16f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, m8, ta, ma +; CHECK-NEXT: vse32.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv16f32( + <vscale x 16 x float> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv16f32( + <vscale x 16 x float>, + ptr, + <vscale x 16 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv16f32_nxv16f32(<vscale x 16 x float> %0, ptr %1, <vscale x 16 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv16f32_nxv16f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, m8, ta, ma +; CHECK-NEXT: vse32.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv16f32( + <vscale x 16 x float> %0, + ptr %1, + <vscale x 16 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv1i16( + <vscale x 1 x i16>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv1i16_nxv1i16(<vscale x 1 x i16> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv1i16_nxv1i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, mf4, ta, ma +; CHECK-NEXT: vse16.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv1i16( + <vscale x 1 x i16> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv1i16( + <vscale x 1 x i16>, + ptr, + <vscale x 1 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv1i16_nxv1i16(<vscale x 1 x i16> %0, ptr %1, <vscale x 1 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv1i16_nxv1i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, mf4, ta, ma +; CHECK-NEXT: vse16.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv1i16( + <vscale x 1 x i16> %0, + ptr %1, + <vscale x 1 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv2i16( + <vscale x 2 x i16>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv2i16_nxv2i16(<vscale x 2 x i16> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv2i16_nxv2i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, mf2, ta, ma +; CHECK-NEXT: vse16.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv2i16( + <vscale x 2 x i16> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv2i16( + <vscale x 2 x i16>, + ptr, + <vscale x 2 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv2i16_nxv2i16(<vscale x 2 x i16> %0, ptr %1, <vscale x 2 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv2i16_nxv2i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, mf2, ta, ma +; CHECK-NEXT: vse16.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv2i16( + <vscale x 2 x i16> %0, + ptr %1, + <vscale x 2 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv4i16( + <vscale x 4 x i16>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv4i16_nxv4i16(<vscale x 4 x i16> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv4i16_nxv4i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, m1, ta, ma +; CHECK-NEXT: vse16.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv4i16( + <vscale x 4 x i16> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv4i16( + <vscale x 4 x i16>, + ptr, + <vscale x 4 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv4i16_nxv4i16(<vscale x 4 x i16> %0, ptr %1, <vscale x 4 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv4i16_nxv4i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, m1, ta, ma +; CHECK-NEXT: vse16.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv4i16( + <vscale x 4 x i16> %0, + ptr %1, + <vscale x 4 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv8i16( + <vscale x 8 x i16>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv8i16_nxv8i16(<vscale x 8 x i16> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv8i16_nxv8i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, m2, ta, ma +; CHECK-NEXT: vse16.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv8i16( + <vscale x 8 x i16> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv8i16( + <vscale x 8 x i16>, + ptr, + <vscale x 8 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv8i16_nxv8i16(<vscale x 8 x i16> %0, ptr %1, <vscale x 8 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv8i16_nxv8i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, m2, ta, ma +; CHECK-NEXT: vse16.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv8i16( + <vscale x 8 x i16> %0, + ptr %1, + <vscale x 8 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv16i16( + <vscale x 16 x i16>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv16i16_nxv16i16(<vscale x 16 x i16> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv16i16_nxv16i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, m4, ta, ma +; CHECK-NEXT: vse16.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv16i16( + <vscale x 16 x i16> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv16i16( + <vscale x 16 x i16>, + ptr, + <vscale x 16 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv16i16_nxv16i16(<vscale x 16 x i16> %0, ptr %1, <vscale x 16 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv16i16_nxv16i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, m4, ta, ma +; CHECK-NEXT: vse16.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv16i16( + <vscale x 16 x i16> %0, + ptr %1, + <vscale x 16 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv32i16( + <vscale x 32 x i16>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv32i16_nxv32i16(<vscale x 32 x i16> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv32i16_nxv32i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, m8, ta, ma +; CHECK-NEXT: vse16.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv32i16( + <vscale x 32 x i16> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv32i16( + <vscale x 32 x i16>, + ptr, + <vscale x 32 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv32i16_nxv32i16(<vscale x 32 x i16> %0, ptr %1, <vscale x 32 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv32i16_nxv32i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, m8, ta, ma +; CHECK-NEXT: vse16.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv32i16( + <vscale x 32 x i16> %0, + ptr %1, + <vscale x 32 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv1f16( + <vscale x 1 x half>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv1f16_nxv1f16(<vscale x 1 x half> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv1f16_nxv1f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, mf4, ta, ma +; CHECK-NEXT: vse16.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv1f16( + <vscale x 1 x half> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv1f16( + <vscale x 1 x half>, + ptr, + <vscale x 1 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv1f16_nxv1f16(<vscale x 1 x half> %0, ptr %1, <vscale x 1 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv1f16_nxv1f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, mf4, ta, ma +; CHECK-NEXT: vse16.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv1f16( + <vscale x 1 x half> %0, + ptr %1, + <vscale x 1 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv2f16( + <vscale x 2 x half>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv2f16_nxv2f16(<vscale x 2 x half> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv2f16_nxv2f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, mf2, ta, ma +; CHECK-NEXT: vse16.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv2f16( + <vscale x 2 x half> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv2f16( + <vscale x 2 x half>, + ptr, + <vscale x 2 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv2f16_nxv2f16(<vscale x 2 x half> %0, ptr %1, <vscale x 2 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv2f16_nxv2f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, mf2, ta, ma +; CHECK-NEXT: vse16.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv2f16( + <vscale x 2 x half> %0, + ptr %1, + <vscale x 2 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv4f16( + <vscale x 4 x half>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv4f16_nxv4f16(<vscale x 4 x half> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv4f16_nxv4f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, m1, ta, ma +; CHECK-NEXT: vse16.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv4f16( + <vscale x 4 x half> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv4f16( + <vscale x 4 x half>, + ptr, + <vscale x 4 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv4f16_nxv4f16(<vscale x 4 x half> %0, ptr %1, <vscale x 4 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv4f16_nxv4f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, m1, ta, ma +; CHECK-NEXT: vse16.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv4f16( + <vscale x 4 x half> %0, + ptr %1, + <vscale x 4 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv8f16( + <vscale x 8 x half>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv8f16_nxv8f16(<vscale x 8 x half> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv8f16_nxv8f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, m2, ta, ma +; CHECK-NEXT: vse16.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv8f16( + <vscale x 8 x half> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv8f16( + <vscale x 8 x half>, + ptr, + <vscale x 8 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv8f16_nxv8f16(<vscale x 8 x half> %0, ptr %1, <vscale x 8 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv8f16_nxv8f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, m2, ta, ma +; CHECK-NEXT: vse16.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv8f16( + <vscale x 8 x half> %0, + ptr %1, + <vscale x 8 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv16f16( + <vscale x 16 x half>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv16f16_nxv16f16(<vscale x 16 x half> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv16f16_nxv16f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, m4, ta, ma +; CHECK-NEXT: vse16.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv16f16( + <vscale x 16 x half> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv16f16( + <vscale x 16 x half>, + ptr, + <vscale x 16 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv16f16_nxv16f16(<vscale x 16 x half> %0, ptr %1, <vscale x 16 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv16f16_nxv16f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, m4, ta, ma +; CHECK-NEXT: vse16.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv16f16( + <vscale x 16 x half> %0, + ptr %1, + <vscale x 16 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv32f16( + <vscale x 32 x half>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv32f16_nxv32f16(<vscale x 32 x half> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv32f16_nxv32f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, m8, ta, ma +; CHECK-NEXT: vse16.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv32f16( + <vscale x 32 x half> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv32f16( + <vscale x 32 x half>, + ptr, + <vscale x 32 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv32f16_nxv32f16(<vscale x 32 x half> %0, ptr %1, <vscale x 32 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv32f16_nxv32f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, m8, ta, ma +; CHECK-NEXT: vse16.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv32f16( + <vscale x 32 x half> %0, + ptr %1, + <vscale x 32 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv1i8( + <vscale x 1 x i8>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv1i8_nxv1i8(<vscale x 1 x i8> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv1i8_nxv1i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, mf8, ta, ma +; CHECK-NEXT: vse8.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv1i8( + <vscale x 1 x i8> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv1i8( + <vscale x 1 x i8>, + ptr, + <vscale x 1 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv1i8_nxv1i8(<vscale x 1 x i8> %0, ptr %1, <vscale x 1 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv1i8_nxv1i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, mf8, ta, ma +; CHECK-NEXT: vse8.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv1i8( + <vscale x 1 x i8> %0, + ptr %1, + <vscale x 1 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv2i8( + <vscale x 2 x i8>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv2i8_nxv2i8(<vscale x 2 x i8> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv2i8_nxv2i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, mf4, ta, ma +; CHECK-NEXT: vse8.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv2i8( + <vscale x 2 x i8> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv2i8( + <vscale x 2 x i8>, + ptr, + <vscale x 2 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv2i8_nxv2i8(<vscale x 2 x i8> %0, ptr %1, <vscale x 2 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv2i8_nxv2i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, mf4, ta, ma +; CHECK-NEXT: vse8.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv2i8( + <vscale x 2 x i8> %0, + ptr %1, + <vscale x 2 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv4i8( + <vscale x 4 x i8>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv4i8_nxv4i8(<vscale x 4 x i8> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv4i8_nxv4i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, mf2, ta, ma +; CHECK-NEXT: vse8.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv4i8( + <vscale x 4 x i8> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv4i8( + <vscale x 4 x i8>, + ptr, + <vscale x 4 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv4i8_nxv4i8(<vscale x 4 x i8> %0, ptr %1, <vscale x 4 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv4i8_nxv4i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, mf2, ta, ma +; CHECK-NEXT: vse8.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv4i8( + <vscale x 4 x i8> %0, + ptr %1, + <vscale x 4 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv8i8( + <vscale x 8 x i8>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv8i8_nxv8i8(<vscale x 8 x i8> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv8i8_nxv8i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, m1, ta, ma +; CHECK-NEXT: vse8.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv8i8( + <vscale x 8 x i8> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv8i8( + <vscale x 8 x i8>, + ptr, + <vscale x 8 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv8i8_nxv8i8(<vscale x 8 x i8> %0, ptr %1, <vscale x 8 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv8i8_nxv8i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, m1, ta, ma +; CHECK-NEXT: vse8.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv8i8( + <vscale x 8 x i8> %0, + ptr %1, + <vscale x 8 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv16i8( + <vscale x 16 x i8>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv16i8_nxv16i8(<vscale x 16 x i8> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv16i8_nxv16i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, m2, ta, ma +; CHECK-NEXT: vse8.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv16i8( + <vscale x 16 x i8> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv16i8( + <vscale x 16 x i8>, + ptr, + <vscale x 16 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv16i8_nxv16i8(<vscale x 16 x i8> %0, ptr %1, <vscale x 16 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv16i8_nxv16i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, m2, ta, ma +; CHECK-NEXT: vse8.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv16i8( + <vscale x 16 x i8> %0, + ptr %1, + <vscale x 16 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv32i8( + <vscale x 32 x i8>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv32i8_nxv32i8(<vscale x 32 x i8> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv32i8_nxv32i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, m4, ta, ma +; CHECK-NEXT: vse8.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv32i8( + <vscale x 32 x i8> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv32i8( + <vscale x 32 x i8>, + ptr, + <vscale x 32 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv32i8_nxv32i8(<vscale x 32 x i8> %0, ptr %1, <vscale x 32 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv32i8_nxv32i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, m4, ta, ma +; CHECK-NEXT: vse8.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv32i8( + <vscale x 32 x i8> %0, + ptr %1, + <vscale x 32 x i1> %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vse.nxv64i8( + <vscale x 64 x i8>, + ptr, + iXLen); + +define void @intrinsic_vse_v_nxv64i8_nxv64i8(<vscale x 64 x i8> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vse_v_nxv64i8_nxv64i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, m8, ta, ma +; CHECK-NEXT: vse8.v v8, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.nxv64i8( + <vscale x 64 x i8> %0, + ptr %1, + iXLen %2) + + ret void +} + +declare void @llvm.riscv.vse.mask.nxv64i8( + <vscale x 64 x i8>, + ptr, + <vscale x 64 x i1>, + iXLen); + +define void @intrinsic_vse_mask_v_nxv64i8_nxv64i8(<vscale x 64 x i8> %0, ptr %1, <vscale x 64 x i1> %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vse_mask_v_nxv64i8_nxv64i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, m8, ta, ma +; CHECK-NEXT: vse8.v v8, (a0), v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vse.mask.nxv64i8( + <vscale x 64 x i8> %0, + ptr %1, + <vscale x 64 x i1> %2, + iXLen %3) + + ret void +} diff --git a/llvm/test/CodeGen/RISCV/GlobalISel/rvv/vsm.ll b/llvm/test/CodeGen/RISCV/GlobalISel/rvv/vsm.ll new file mode 100644 index 0000000..5237536 --- /dev/null +++ b/llvm/test/CodeGen/RISCV/GlobalISel/rvv/vsm.ll @@ -0,0 +1,139 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v \ +; RUN: -global-isel -verify-machineinstrs | FileCheck %s +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v \ +; RUN: -global-isel -verify-machineinstrs | FileCheck %s + +declare void @llvm.riscv.vsm.nxv1i1(<vscale x 1 x i1>, ptr, iXLen); + +define void @intrinsic_vsm_v_nxv1i1(<vscale x 1 x i1> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vsm_v_nxv1i1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, mf8, ta, ma +; CHECK-NEXT: vsm.v v0, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsm.nxv1i1(<vscale x 1 x i1> %0, ptr %1, iXLen %2) + ret void +} + +declare void @llvm.riscv.vsm.nxv2i1(<vscale x 2 x i1>, ptr, iXLen); + +define void @intrinsic_vsm_v_nxv2i1(<vscale x 2 x i1> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vsm_v_nxv2i1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, mf4, ta, ma +; CHECK-NEXT: vsm.v v0, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsm.nxv2i1(<vscale x 2 x i1> %0, ptr %1, iXLen %2) + ret void +} + +declare void @llvm.riscv.vsm.nxv4i1(<vscale x 4 x i1>, ptr, iXLen); + +define void @intrinsic_vsm_v_nxv4i1(<vscale x 4 x i1> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vsm_v_nxv4i1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, mf2, ta, ma +; CHECK-NEXT: vsm.v v0, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsm.nxv4i1(<vscale x 4 x i1> %0, ptr %1, iXLen %2) + ret void +} + +declare void @llvm.riscv.vsm.nxv8i1(<vscale x 8 x i1>, ptr, iXLen); + +define void @intrinsic_vsm_v_nxv8i1(<vscale x 8 x i1> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vsm_v_nxv8i1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, m1, ta, ma +; CHECK-NEXT: vsm.v v0, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsm.nxv8i1(<vscale x 8 x i1> %0, ptr %1, iXLen %2) + ret void +} + +declare void @llvm.riscv.vsm.nxv16i1(<vscale x 16 x i1>, ptr, iXLen); + +define void @intrinsic_vsm_v_nxv16i1(<vscale x 16 x i1> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vsm_v_nxv16i1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, m2, ta, ma +; CHECK-NEXT: vsm.v v0, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsm.nxv16i1(<vscale x 16 x i1> %0, ptr %1, iXLen %2) + ret void +} + +declare void @llvm.riscv.vsm.nxv32i1(<vscale x 32 x i1>, ptr, iXLen); + +define void @intrinsic_vsm_v_nxv32i1(<vscale x 32 x i1> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vsm_v_nxv32i1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, m4, ta, ma +; CHECK-NEXT: vsm.v v0, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsm.nxv32i1(<vscale x 32 x i1> %0, ptr %1, iXLen %2) + ret void +} + +declare void @llvm.riscv.vsm.nxv64i1(<vscale x 64 x i1>, ptr, iXLen); + +define void @intrinsic_vsm_v_nxv64i1(<vscale x 64 x i1> %0, ptr %1, iXLen %2) nounwind { +; CHECK-LABEL: intrinsic_vsm_v_nxv64i1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e8, m8, ta, ma +; CHECK-NEXT: vsm.v v0, (a0) +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsm.nxv64i1(<vscale x 64 x i1> %0, ptr %1, iXLen %2) + ret void +} + +declare <vscale x 1 x i1> @llvm.riscv.vmseq.nxv1i16( + <vscale x 1 x i16>, + <vscale x 1 x i16>, + iXLen); + +; Make sure we can use the vsetvli from the producing instruction. +define void @test_vsetvli_i16(<vscale x 1 x i16> %0, <vscale x 1 x i16> %1, ptr %2, iXLen %3) nounwind { +; CHECK-LABEL: test_vsetvli_i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e16, mf4, ta, ma +; CHECK-NEXT: vmseq.vv v8, v8, v9 +; CHECK-NEXT: vsm.v v8, (a0) +; CHECK-NEXT: ret +entry: + %a = call <vscale x 1 x i1> @llvm.riscv.vmseq.nxv1i16( + <vscale x 1 x i16> %0, + <vscale x 1 x i16> %1, + iXLen %3) + call void @llvm.riscv.vsm.nxv1i1(<vscale x 1 x i1> %a, ptr %2, iXLen %3) + ret void +} + +declare <vscale x 1 x i1> @llvm.riscv.vmseq.nxv1i32( + <vscale x 1 x i32>, + <vscale x 1 x i32>, + iXLen); + +define void @test_vsetvli_i32(<vscale x 1 x i32> %0, <vscale x 1 x i32> %1, ptr %2, iXLen %3) nounwind { +; CHECK-LABEL: test_vsetvli_i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a1, e32, mf2, ta, ma +; CHECK-NEXT: vmseq.vv v8, v8, v9 +; CHECK-NEXT: vsm.v v8, (a0) +; CHECK-NEXT: ret +entry: + %a = call <vscale x 1 x i1> @llvm.riscv.vmseq.nxv1i32( + <vscale x 1 x i32> %0, + <vscale x 1 x i32> %1, + iXLen %3) + call void @llvm.riscv.vsm.nxv1i1(<vscale x 1 x i1> %a, ptr %2, iXLen %3) + ret void +} diff --git a/llvm/test/CodeGen/RISCV/GlobalISel/rvv/vsse.ll b/llvm/test/CodeGen/RISCV/GlobalISel/rvv/vsse.ll new file mode 100644 index 0000000..b7609ff --- /dev/null +++ b/llvm/test/CodeGen/RISCV/GlobalISel/rvv/vsse.ll @@ -0,0 +1,1724 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py +; RUN: sed 's/iXLen/i32/g' %s | llc -mtriple=riscv32 -mattr=+v,+zvfhmin,+zvfbfmin \ +; RUN: -global-isel -verify-machineinstrs -target-abi=ilp32d | FileCheck %s +; RUN: sed 's/iXLen/i64/g' %s | llc -mtriple=riscv64 -mattr=+v,+zvfhmin,+zvfbfmin \ +; RUN: -global-isel -verify-machineinstrs -target-abi=lp64d | FileCheck %s + +declare void @llvm.riscv.vsse.nxv1i64( + <vscale x 1 x i64>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv1i64_nxv1i64(<vscale x 1 x i64> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv1i64_nxv1i64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e64, m1, ta, ma +; CHECK-NEXT: vsse64.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv1i64( + <vscale x 1 x i64> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv1i64( + <vscale x 1 x i64>, + ptr, + iXLen, + <vscale x 1 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv1i64_nxv1i64(<vscale x 1 x i64> %0, ptr %1, iXLen %2, <vscale x 1 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv1i64_nxv1i64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e64, m1, ta, ma +; CHECK-NEXT: vsse64.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv1i64( + <vscale x 1 x i64> %0, + ptr %1, + iXLen %2, + <vscale x 1 x i1> %3, + iXLen %4) + + ret void +} + +define void @intrinsic_vsse_allonesmask_v_nxv1i64_nxv1i64(<vscale x 1 x i64> %0, ptr %1, iXLen %2, <vscale x 1 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_allonesmask_v_nxv1i64_nxv1i64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e64, m1, ta, ma +; CHECK-NEXT: vsse64.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv1i64( + <vscale x 1 x i64> %0, + ptr %1, + iXLen %2, + <vscale x 1 x i1> splat (i1 true), + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv2i64( + <vscale x 2 x i64>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv2i64_nxv2i64(<vscale x 2 x i64> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv2i64_nxv2i64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e64, m2, ta, ma +; CHECK-NEXT: vsse64.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv2i64( + <vscale x 2 x i64> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv2i64( + <vscale x 2 x i64>, + ptr, + iXLen, + <vscale x 2 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv2i64_nxv2i64(<vscale x 2 x i64> %0, ptr %1, iXLen %2, <vscale x 2 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv2i64_nxv2i64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e64, m2, ta, ma +; CHECK-NEXT: vsse64.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv2i64( + <vscale x 2 x i64> %0, + ptr %1, + iXLen %2, + <vscale x 2 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv4i64( + <vscale x 4 x i64>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv4i64_nxv4i64(<vscale x 4 x i64> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv4i64_nxv4i64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e64, m4, ta, ma +; CHECK-NEXT: vsse64.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv4i64( + <vscale x 4 x i64> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv4i64( + <vscale x 4 x i64>, + ptr, + iXLen, + <vscale x 4 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv4i64_nxv4i64(<vscale x 4 x i64> %0, ptr %1, iXLen %2, <vscale x 4 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv4i64_nxv4i64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e64, m4, ta, ma +; CHECK-NEXT: vsse64.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv4i64( + <vscale x 4 x i64> %0, + ptr %1, + iXLen %2, + <vscale x 4 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv8i64( + <vscale x 8 x i64>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv8i64_nxv8i64(<vscale x 8 x i64> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv8i64_nxv8i64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e64, m8, ta, ma +; CHECK-NEXT: vsse64.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv8i64( + <vscale x 8 x i64> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv8i64( + <vscale x 8 x i64>, + ptr, + iXLen, + <vscale x 8 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv8i64_nxv8i64(<vscale x 8 x i64> %0, ptr %1, iXLen %2, <vscale x 8 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv8i64_nxv8i64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e64, m8, ta, ma +; CHECK-NEXT: vsse64.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv8i64( + <vscale x 8 x i64> %0, + ptr %1, + iXLen %2, + <vscale x 8 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv1f64( + <vscale x 1 x double>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv1f64_nxv1f64(<vscale x 1 x double> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv1f64_nxv1f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e64, m1, ta, ma +; CHECK-NEXT: vsse64.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv1f64( + <vscale x 1 x double> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv1f64( + <vscale x 1 x double>, + ptr, + iXLen, + <vscale x 1 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv1f64_nxv1f64(<vscale x 1 x double> %0, ptr %1, iXLen %2, <vscale x 1 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv1f64_nxv1f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e64, m1, ta, ma +; CHECK-NEXT: vsse64.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv1f64( + <vscale x 1 x double> %0, + ptr %1, + iXLen %2, + <vscale x 1 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv2f64( + <vscale x 2 x double>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv2f64_nxv2f64(<vscale x 2 x double> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv2f64_nxv2f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e64, m2, ta, ma +; CHECK-NEXT: vsse64.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv2f64( + <vscale x 2 x double> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv2f64( + <vscale x 2 x double>, + ptr, + iXLen, + <vscale x 2 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv2f64_nxv2f64(<vscale x 2 x double> %0, ptr %1, iXLen %2, <vscale x 2 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv2f64_nxv2f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e64, m2, ta, ma +; CHECK-NEXT: vsse64.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv2f64( + <vscale x 2 x double> %0, + ptr %1, + iXLen %2, + <vscale x 2 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv4f64( + <vscale x 4 x double>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv4f64_nxv4f64(<vscale x 4 x double> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv4f64_nxv4f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e64, m4, ta, ma +; CHECK-NEXT: vsse64.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv4f64( + <vscale x 4 x double> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv4f64( + <vscale x 4 x double>, + ptr, + iXLen, + <vscale x 4 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv4f64_nxv4f64(<vscale x 4 x double> %0, ptr %1, iXLen %2, <vscale x 4 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv4f64_nxv4f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e64, m4, ta, ma +; CHECK-NEXT: vsse64.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv4f64( + <vscale x 4 x double> %0, + ptr %1, + iXLen %2, + <vscale x 4 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv8f64( + <vscale x 8 x double>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv8f64_nxv8f64(<vscale x 8 x double> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv8f64_nxv8f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e64, m8, ta, ma +; CHECK-NEXT: vsse64.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv8f64( + <vscale x 8 x double> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv8f64( + <vscale x 8 x double>, + ptr, + iXLen, + <vscale x 8 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv8f64_nxv8f64(<vscale x 8 x double> %0, ptr %1, iXLen %2, <vscale x 8 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv8f64_nxv8f64: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e64, m8, ta, ma +; CHECK-NEXT: vsse64.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv8f64( + <vscale x 8 x double> %0, + ptr %1, + iXLen %2, + <vscale x 8 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv1i32( + <vscale x 1 x i32>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv1i32_nxv1i32(<vscale x 1 x i32> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv1i32_nxv1i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e32, mf2, ta, ma +; CHECK-NEXT: vsse32.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv1i32( + <vscale x 1 x i32> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv1i32( + <vscale x 1 x i32>, + ptr, + iXLen, + <vscale x 1 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv1i32_nxv1i32(<vscale x 1 x i32> %0, ptr %1, iXLen %2, <vscale x 1 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv1i32_nxv1i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e32, mf2, ta, ma +; CHECK-NEXT: vsse32.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv1i32( + <vscale x 1 x i32> %0, + ptr %1, + iXLen %2, + <vscale x 1 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv2i32( + <vscale x 2 x i32>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv2i32_nxv2i32(<vscale x 2 x i32> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv2i32_nxv2i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e32, m1, ta, ma +; CHECK-NEXT: vsse32.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv2i32( + <vscale x 2 x i32> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv2i32( + <vscale x 2 x i32>, + ptr, + iXLen, + <vscale x 2 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv2i32_nxv2i32(<vscale x 2 x i32> %0, ptr %1, iXLen %2, <vscale x 2 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv2i32_nxv2i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e32, m1, ta, ma +; CHECK-NEXT: vsse32.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv2i32( + <vscale x 2 x i32> %0, + ptr %1, + iXLen %2, + <vscale x 2 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv4i32( + <vscale x 4 x i32>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv4i32_nxv4i32(<vscale x 4 x i32> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv4i32_nxv4i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e32, m2, ta, ma +; CHECK-NEXT: vsse32.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv4i32( + <vscale x 4 x i32> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv4i32( + <vscale x 4 x i32>, + ptr, + iXLen, + <vscale x 4 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv4i32_nxv4i32(<vscale x 4 x i32> %0, ptr %1, iXLen %2, <vscale x 4 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv4i32_nxv4i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e32, m2, ta, ma +; CHECK-NEXT: vsse32.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv4i32( + <vscale x 4 x i32> %0, + ptr %1, + iXLen %2, + <vscale x 4 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv8i32( + <vscale x 8 x i32>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv8i32_nxv8i32(<vscale x 8 x i32> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv8i32_nxv8i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e32, m4, ta, ma +; CHECK-NEXT: vsse32.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv8i32( + <vscale x 8 x i32> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv8i32( + <vscale x 8 x i32>, + ptr, + iXLen, + <vscale x 8 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv8i32_nxv8i32(<vscale x 8 x i32> %0, ptr %1, iXLen %2, <vscale x 8 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv8i32_nxv8i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e32, m4, ta, ma +; CHECK-NEXT: vsse32.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv8i32( + <vscale x 8 x i32> %0, + ptr %1, + iXLen %2, + <vscale x 8 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv16i32( + <vscale x 16 x i32>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv16i32_nxv16i32(<vscale x 16 x i32> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv16i32_nxv16i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e32, m8, ta, ma +; CHECK-NEXT: vsse32.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv16i32( + <vscale x 16 x i32> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv16i32( + <vscale x 16 x i32>, + ptr, + iXLen, + <vscale x 16 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv16i32_nxv16i32(<vscale x 16 x i32> %0, ptr %1, iXLen %2, <vscale x 16 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv16i32_nxv16i32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e32, m8, ta, ma +; CHECK-NEXT: vsse32.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv16i32( + <vscale x 16 x i32> %0, + ptr %1, + iXLen %2, + <vscale x 16 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv1f32( + <vscale x 1 x float>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv1f32_nxv1f32(<vscale x 1 x float> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv1f32_nxv1f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e32, mf2, ta, ma +; CHECK-NEXT: vsse32.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv1f32( + <vscale x 1 x float> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv1f32( + <vscale x 1 x float>, + ptr, + iXLen, + <vscale x 1 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv1f32_nxv1f32(<vscale x 1 x float> %0, ptr %1, iXLen %2, <vscale x 1 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv1f32_nxv1f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e32, mf2, ta, ma +; CHECK-NEXT: vsse32.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv1f32( + <vscale x 1 x float> %0, + ptr %1, + iXLen %2, + <vscale x 1 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv2f32( + <vscale x 2 x float>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv2f32_nxv2f32(<vscale x 2 x float> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv2f32_nxv2f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e32, m1, ta, ma +; CHECK-NEXT: vsse32.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv2f32( + <vscale x 2 x float> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv2f32( + <vscale x 2 x float>, + ptr, + iXLen, + <vscale x 2 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv2f32_nxv2f32(<vscale x 2 x float> %0, ptr %1, iXLen %2, <vscale x 2 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv2f32_nxv2f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e32, m1, ta, ma +; CHECK-NEXT: vsse32.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv2f32( + <vscale x 2 x float> %0, + ptr %1, + iXLen %2, + <vscale x 2 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv4f32( + <vscale x 4 x float>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv4f32_nxv4f32(<vscale x 4 x float> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv4f32_nxv4f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e32, m2, ta, ma +; CHECK-NEXT: vsse32.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv4f32( + <vscale x 4 x float> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv4f32( + <vscale x 4 x float>, + ptr, + iXLen, + <vscale x 4 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv4f32_nxv4f32(<vscale x 4 x float> %0, ptr %1, iXLen %2, <vscale x 4 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv4f32_nxv4f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e32, m2, ta, ma +; CHECK-NEXT: vsse32.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv4f32( + <vscale x 4 x float> %0, + ptr %1, + iXLen %2, + <vscale x 4 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv8f32( + <vscale x 8 x float>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv8f32_nxv8f32(<vscale x 8 x float> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv8f32_nxv8f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e32, m4, ta, ma +; CHECK-NEXT: vsse32.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv8f32( + <vscale x 8 x float> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv8f32( + <vscale x 8 x float>, + ptr, + iXLen, + <vscale x 8 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv8f32_nxv8f32(<vscale x 8 x float> %0, ptr %1, iXLen %2, <vscale x 8 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv8f32_nxv8f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e32, m4, ta, ma +; CHECK-NEXT: vsse32.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv8f32( + <vscale x 8 x float> %0, + ptr %1, + iXLen %2, + <vscale x 8 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv16f32( + <vscale x 16 x float>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv16f32_nxv16f32(<vscale x 16 x float> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv16f32_nxv16f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e32, m8, ta, ma +; CHECK-NEXT: vsse32.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv16f32( + <vscale x 16 x float> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv16f32( + <vscale x 16 x float>, + ptr, + iXLen, + <vscale x 16 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv16f32_nxv16f32(<vscale x 16 x float> %0, ptr %1, iXLen %2, <vscale x 16 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv16f32_nxv16f32: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e32, m8, ta, ma +; CHECK-NEXT: vsse32.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv16f32( + <vscale x 16 x float> %0, + ptr %1, + iXLen %2, + <vscale x 16 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv1i16( + <vscale x 1 x i16>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv1i16_nxv1i16(<vscale x 1 x i16> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv1i16_nxv1i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, mf4, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv1i16( + <vscale x 1 x i16> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv1i16( + <vscale x 1 x i16>, + ptr, + iXLen, + <vscale x 1 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv1i16_nxv1i16(<vscale x 1 x i16> %0, ptr %1, iXLen %2, <vscale x 1 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv1i16_nxv1i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, mf4, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv1i16( + <vscale x 1 x i16> %0, + ptr %1, + iXLen %2, + <vscale x 1 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv2i16( + <vscale x 2 x i16>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv2i16_nxv2i16(<vscale x 2 x i16> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv2i16_nxv2i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, mf2, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv2i16( + <vscale x 2 x i16> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv2i16( + <vscale x 2 x i16>, + ptr, + iXLen, + <vscale x 2 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv2i16_nxv2i16(<vscale x 2 x i16> %0, ptr %1, iXLen %2, <vscale x 2 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv2i16_nxv2i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, mf2, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv2i16( + <vscale x 2 x i16> %0, + ptr %1, + iXLen %2, + <vscale x 2 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv4i16( + <vscale x 4 x i16>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv4i16_nxv4i16(<vscale x 4 x i16> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv4i16_nxv4i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, m1, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv4i16( + <vscale x 4 x i16> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv4i16( + <vscale x 4 x i16>, + ptr, + iXLen, + <vscale x 4 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv4i16_nxv4i16(<vscale x 4 x i16> %0, ptr %1, iXLen %2, <vscale x 4 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv4i16_nxv4i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, m1, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv4i16( + <vscale x 4 x i16> %0, + ptr %1, + iXLen %2, + <vscale x 4 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv8i16( + <vscale x 8 x i16>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv8i16_nxv8i16(<vscale x 8 x i16> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv8i16_nxv8i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, m2, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv8i16( + <vscale x 8 x i16> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv8i16( + <vscale x 8 x i16>, + ptr, + iXLen, + <vscale x 8 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv8i16_nxv8i16(<vscale x 8 x i16> %0, ptr %1, iXLen %2, <vscale x 8 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv8i16_nxv8i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, m2, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv8i16( + <vscale x 8 x i16> %0, + ptr %1, + iXLen %2, + <vscale x 8 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv16i16( + <vscale x 16 x i16>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv16i16_nxv16i16(<vscale x 16 x i16> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv16i16_nxv16i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, m4, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv16i16( + <vscale x 16 x i16> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv16i16( + <vscale x 16 x i16>, + ptr, + iXLen, + <vscale x 16 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv16i16_nxv16i16(<vscale x 16 x i16> %0, ptr %1, iXLen %2, <vscale x 16 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv16i16_nxv16i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, m4, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv16i16( + <vscale x 16 x i16> %0, + ptr %1, + iXLen %2, + <vscale x 16 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv32i16( + <vscale x 32 x i16>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv32i16_nxv32i16(<vscale x 32 x i16> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv32i16_nxv32i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, m8, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv32i16( + <vscale x 32 x i16> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv32i16( + <vscale x 32 x i16>, + ptr, + iXLen, + <vscale x 32 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv32i16_nxv32i16(<vscale x 32 x i16> %0, ptr %1, iXLen %2, <vscale x 32 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv32i16_nxv32i16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, m8, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv32i16( + <vscale x 32 x i16> %0, + ptr %1, + iXLen %2, + <vscale x 32 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv1f16( + <vscale x 1 x half>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv1f16_nxv1f16(<vscale x 1 x half> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv1f16_nxv1f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, mf4, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv1f16( + <vscale x 1 x half> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv1f16( + <vscale x 1 x half>, + ptr, + iXLen, + <vscale x 1 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv1f16_nxv1f16(<vscale x 1 x half> %0, ptr %1, iXLen %2, <vscale x 1 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv1f16_nxv1f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, mf4, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv1f16( + <vscale x 1 x half> %0, + ptr %1, + iXLen %2, + <vscale x 1 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv2f16( + <vscale x 2 x half>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv2f16_nxv2f16(<vscale x 2 x half> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv2f16_nxv2f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, mf2, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv2f16( + <vscale x 2 x half> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv2f16( + <vscale x 2 x half>, + ptr, + iXLen, + <vscale x 2 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv2f16_nxv2f16(<vscale x 2 x half> %0, ptr %1, iXLen %2, <vscale x 2 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv2f16_nxv2f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, mf2, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv2f16( + <vscale x 2 x half> %0, + ptr %1, + iXLen %2, + <vscale x 2 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv4f16( + <vscale x 4 x half>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv4f16_nxv4f16(<vscale x 4 x half> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv4f16_nxv4f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, m1, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv4f16( + <vscale x 4 x half> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv4f16( + <vscale x 4 x half>, + ptr, + iXLen, + <vscale x 4 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv4f16_nxv4f16(<vscale x 4 x half> %0, ptr %1, iXLen %2, <vscale x 4 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv4f16_nxv4f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, m1, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv4f16( + <vscale x 4 x half> %0, + ptr %1, + iXLen %2, + <vscale x 4 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv8f16( + <vscale x 8 x half>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv8f16_nxv8f16(<vscale x 8 x half> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv8f16_nxv8f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, m2, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv8f16( + <vscale x 8 x half> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv8f16( + <vscale x 8 x half>, + ptr, + iXLen, + <vscale x 8 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv8f16_nxv8f16(<vscale x 8 x half> %0, ptr %1, iXLen %2, <vscale x 8 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv8f16_nxv8f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, m2, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv8f16( + <vscale x 8 x half> %0, + ptr %1, + iXLen %2, + <vscale x 8 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv16f16( + <vscale x 16 x half>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv16f16_nxv16f16(<vscale x 16 x half> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv16f16_nxv16f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, m4, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv16f16( + <vscale x 16 x half> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv16f16( + <vscale x 16 x half>, + ptr, + iXLen, + <vscale x 16 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv16f16_nxv16f16(<vscale x 16 x half> %0, ptr %1, iXLen %2, <vscale x 16 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv16f16_nxv16f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, m4, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv16f16( + <vscale x 16 x half> %0, + ptr %1, + iXLen %2, + <vscale x 16 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv32f16( + <vscale x 32 x half>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv32f16_nxv32f16(<vscale x 32 x half> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv32f16_nxv32f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, m8, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv32f16( + <vscale x 32 x half> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv32f16( + <vscale x 32 x half>, + ptr, + iXLen, + <vscale x 32 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv32f16_nxv32f16(<vscale x 32 x half> %0, ptr %1, iXLen %2, <vscale x 32 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv32f16_nxv32f16: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e16, m8, ta, ma +; CHECK-NEXT: vsse16.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv32f16( + <vscale x 32 x half> %0, + ptr %1, + iXLen %2, + <vscale x 32 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv1i8( + <vscale x 1 x i8>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv1i8_nxv1i8(<vscale x 1 x i8> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv1i8_nxv1i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e8, mf8, ta, ma +; CHECK-NEXT: vsse8.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv1i8( + <vscale x 1 x i8> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv1i8( + <vscale x 1 x i8>, + ptr, + iXLen, + <vscale x 1 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv1i8_nxv1i8(<vscale x 1 x i8> %0, ptr %1, iXLen %2, <vscale x 1 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv1i8_nxv1i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e8, mf8, ta, ma +; CHECK-NEXT: vsse8.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv1i8( + <vscale x 1 x i8> %0, + ptr %1, + iXLen %2, + <vscale x 1 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv2i8( + <vscale x 2 x i8>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv2i8_nxv2i8(<vscale x 2 x i8> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv2i8_nxv2i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e8, mf4, ta, ma +; CHECK-NEXT: vsse8.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv2i8( + <vscale x 2 x i8> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv2i8( + <vscale x 2 x i8>, + ptr, + iXLen, + <vscale x 2 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv2i8_nxv2i8(<vscale x 2 x i8> %0, ptr %1, iXLen %2, <vscale x 2 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv2i8_nxv2i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e8, mf4, ta, ma +; CHECK-NEXT: vsse8.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv2i8( + <vscale x 2 x i8> %0, + ptr %1, + iXLen %2, + <vscale x 2 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv4i8( + <vscale x 4 x i8>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv4i8_nxv4i8(<vscale x 4 x i8> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv4i8_nxv4i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e8, mf2, ta, ma +; CHECK-NEXT: vsse8.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv4i8( + <vscale x 4 x i8> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv4i8( + <vscale x 4 x i8>, + ptr, + iXLen, + <vscale x 4 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv4i8_nxv4i8(<vscale x 4 x i8> %0, ptr %1, iXLen %2, <vscale x 4 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv4i8_nxv4i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e8, mf2, ta, ma +; CHECK-NEXT: vsse8.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv4i8( + <vscale x 4 x i8> %0, + ptr %1, + iXLen %2, + <vscale x 4 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv8i8( + <vscale x 8 x i8>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv8i8_nxv8i8(<vscale x 8 x i8> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv8i8_nxv8i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e8, m1, ta, ma +; CHECK-NEXT: vsse8.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv8i8( + <vscale x 8 x i8> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv8i8( + <vscale x 8 x i8>, + ptr, + iXLen, + <vscale x 8 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv8i8_nxv8i8(<vscale x 8 x i8> %0, ptr %1, iXLen %2, <vscale x 8 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv8i8_nxv8i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e8, m1, ta, ma +; CHECK-NEXT: vsse8.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv8i8( + <vscale x 8 x i8> %0, + ptr %1, + iXLen %2, + <vscale x 8 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv16i8( + <vscale x 16 x i8>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv16i8_nxv16i8(<vscale x 16 x i8> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv16i8_nxv16i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e8, m2, ta, ma +; CHECK-NEXT: vsse8.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv16i8( + <vscale x 16 x i8> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv16i8( + <vscale x 16 x i8>, + ptr, + iXLen, + <vscale x 16 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv16i8_nxv16i8(<vscale x 16 x i8> %0, ptr %1, iXLen %2, <vscale x 16 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv16i8_nxv16i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e8, m2, ta, ma +; CHECK-NEXT: vsse8.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv16i8( + <vscale x 16 x i8> %0, + ptr %1, + iXLen %2, + <vscale x 16 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv32i8( + <vscale x 32 x i8>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv32i8_nxv32i8(<vscale x 32 x i8> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv32i8_nxv32i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e8, m4, ta, ma +; CHECK-NEXT: vsse8.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv32i8( + <vscale x 32 x i8> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv32i8( + <vscale x 32 x i8>, + ptr, + iXLen, + <vscale x 32 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv32i8_nxv32i8(<vscale x 32 x i8> %0, ptr %1, iXLen %2, <vscale x 32 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv32i8_nxv32i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e8, m4, ta, ma +; CHECK-NEXT: vsse8.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv32i8( + <vscale x 32 x i8> %0, + ptr %1, + iXLen %2, + <vscale x 32 x i1> %3, + iXLen %4) + + ret void +} + +declare void @llvm.riscv.vsse.nxv64i8( + <vscale x 64 x i8>, + ptr, + iXLen, + iXLen); + +define void @intrinsic_vsse_v_nxv64i8_nxv64i8(<vscale x 64 x i8> %0, ptr %1, iXLen %2, iXLen %3) nounwind { +; CHECK-LABEL: intrinsic_vsse_v_nxv64i8_nxv64i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e8, m8, ta, ma +; CHECK-NEXT: vsse8.v v8, (a0), a1 +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.nxv64i8( + <vscale x 64 x i8> %0, + ptr %1, + iXLen %2, + iXLen %3) + + ret void +} + +declare void @llvm.riscv.vsse.mask.nxv64i8( + <vscale x 64 x i8>, + ptr, + iXLen, + <vscale x 64 x i1>, + iXLen); + +define void @intrinsic_vsse_mask_v_nxv64i8_nxv64i8(<vscale x 64 x i8> %0, ptr %1, iXLen %2, <vscale x 64 x i1> %3, iXLen %4) nounwind { +; CHECK-LABEL: intrinsic_vsse_mask_v_nxv64i8_nxv64i8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: vsetvli zero, a2, e8, m8, ta, ma +; CHECK-NEXT: vsse8.v v8, (a0), a1, v0.t +; CHECK-NEXT: ret +entry: + call void @llvm.riscv.vsse.mask.nxv64i8( + <vscale x 64 x i8> %0, + ptr %1, + iXLen %2, + <vscale x 64 x i1> %3, + iXLen %4) + + ret void +} diff --git a/llvm/test/CodeGen/RISCV/features-info.ll b/llvm/test/CodeGen/RISCV/features-info.ll index 37e11db..988d049 100644 --- a/llvm/test/CodeGen/RISCV/features-info.ll +++ b/llvm/test/CodeGen/RISCV/features-info.ll @@ -136,6 +136,7 @@ ; CHECK-NEXT: shgatpa - 'Shgatpa' (SvNNx4 mode supported for all modes supported by satp, as well as Bare). ; CHECK-NEXT: shifted-zextw-fusion - Enable SLLI+SRLI to be fused when computing (shifted) word zero extension. ; CHECK-NEXT: shlcofideleg - 'Shlcofideleg' (Delegating LCOFI Interrupts to VS-mode). +; CHECK-NEXT: short-forward-branch-i-minmax - Enable short forward branch optimization for min,max instructions in Zbb. ; CHECK-NEXT: short-forward-branch-opt - Enable short forward branch optimization. ; CHECK-NEXT: shtvala - 'Shtvala' (htval provides all needed values). ; CHECK-NEXT: shvsatpa - 'Shvsatpa' (vsatp supports all modes supported by satp). diff --git a/llvm/test/CodeGen/RISCV/rv64-stackmap.ll b/llvm/test/CodeGen/RISCV/rv64-stackmap.ll index d07f608..c50a0fb3 100644 --- a/llvm/test/CodeGen/RISCV/rv64-stackmap.ll +++ b/llvm/test/CodeGen/RISCV/rv64-stackmap.ll @@ -7,11 +7,11 @@ ; CHECK-NEXT: .byte 0 ; CHECK-NEXT: .half 0 ; Num Functions -; CHECK-NEXT: .word 12 +; CHECK-NEXT: .word 13 ; Num LargeConstants -; CHECK-NEXT: .word 2 +; CHECK-NEXT: .word 3 ; Num Callsites -; CHECK-NEXT: .word 16 +; CHECK-NEXT: .word 17 ; Functions and stack size ; CHECK-NEXT: .quad constantargs @@ -38,8 +38,8 @@ ; CHECK-NEXT: .quad liveConstant ; CHECK-NEXT: .quad 0 ; CHECK-NEXT: .quad 1 -; CHECK-NEXT: .quad spilledValue -; CHECK-NEXT: .quad 144 +; CHECK-NEXT: .quad liveArgs +; CHECK-NEXT: .quad 0 ; CHECK-NEXT: .quad 1 ; CHECK-NEXT: .quad directFrameIdx ; CHECK-NEXT: .quad 48 @@ -50,10 +50,14 @@ ; CHECK-NEXT: .quad needsStackRealignment ; CHECK-NEXT: .quad -1 ; CHECK-NEXT: .quad 1 +; CHECK-NEXT: .quad floats +; CHECK-NEXT: .quad 32 +; CHECK-NEXT: .quad 1 ; Num LargeConstants ; CHECK-NEXT: .quad 4294967295 ; CHECK-NEXT: .quad 4294967296 +; CHECK-NEXT: .quad 4609434218613702656 ; Constant arguments ; @@ -278,7 +282,7 @@ define void @liveConstant() { ; ; Verify 28 stack map entries. ; -; CHECK-LABEL: .word .L{{.*}}-spilledValue +; CHECK-LABEL: .word .L{{.*}}-liveArgs ; CHECK-NEXT: .half 0 ; CHECK-NEXT: .half 28 ; @@ -290,9 +294,9 @@ define void @liveConstant() { ; CHECK-NEXT: .half 2 ; CHECK-NEXT: .half 0 ; CHECK-NEXT: .word -define void @spilledValue(i64 %arg0, i64 %arg1, i64 %arg2, i64 %arg3, i64 %arg4, i64 %l0, i64 %l1, i64 %l2, i64 %l3, i64 %l4, i64 %l5, i64 %l6, i64 %l7, i64 %l8, i64 %l9, i64 %l10, i64 %l11, i64 %l12, i64 %l13, i64 %l14, i64 %l15, i64 %l16, i64 %l17, i64 %l18, i64 %l19, i64 %l20, i64 %l21, i64 %l22, i64 %l23, i64 %l24, i64 %l25, i64 %l26, i64 %l27) { +define void @liveArgs(i64 %arg0, i64 %arg1, i64 %arg2, i64 %arg3, i64 %arg4, i64 %l0, i64 %l1, i64 %l2, i64 %l3, i64 %l4, i64 %l5, i64 %l6, i64 %l7, i64 %l8, i64 %l9, i64 %l10, i64 %l11, i64 %l12, i64 %l13, i64 %l14, i64 %l15, i64 %l16, i64 %l17, i64 %l18, i64 %l19, i64 %l20, i64 %l21, i64 %l22, i64 %l23, i64 %l24, i8 %l25, i16 zeroext %l26, i32 signext %l27) { entry: - call void (i64, i32, ptr, i32, ...) @llvm.experimental.patchpoint.void(i64 11, i32 28, ptr null, i32 5, i64 %arg0, i64 %arg1, i64 %arg2, i64 %arg3, i64 %arg4, i64 %l0, i64 %l1, i64 %l2, i64 %l3, i64 %l4, i64 %l5, i64 %l6, i64 %l7, i64 %l8, i64 %l9, i64 %l10, i64 %l11, i64 %l12, i64 %l13, i64 %l14, i64 %l15, i64 %l16, i64 %l17, i64 %l18, i64 %l19, i64 %l20, i64 %l21, i64 %l22, i64 %l23, i64 %l24, i64 %l25, i64 %l26, i64 %l27) + call void (i64, i32, ptr, i32, ...) @llvm.experimental.patchpoint.void(i64 11, i32 28, ptr null, i32 5, i64 %arg0, i64 %arg1, i64 %arg2, i64 %arg3, i64 %arg4, i64 %l0, i64 %l1, i64 %l2, i64 %l3, i64 %l4, i64 %l5, i64 %l6, i64 %l7, i64 %l8, i64 %l9, i64 %l10, i64 %l11, i64 %l12, i64 %l13, i64 %l14, i64 %l15, i64 %l16, i64 %l17, i64 %l18, i64 %l19, i64 %l20, i64 %l21, i64 %l22, i64 %l23, i64 %l24, i8 %l25, i16 %l26, i32 %l27) ret void } @@ -379,6 +383,104 @@ define void @needsStackRealignment() { } declare void @escape_values(...) +; CHECK-LABEL: .word .L{{.*}}-floats +; CHECK-NEXT: .half 0 +; Num Locations +; CHECK-NEXT: .half 12 +; Loc 0: constant float as constant integer +; CHECK-NEXT: .byte 4 +; CHECK-NEXT: .byte 0 +; CHECK-NEXT: .half 8 +; CHECK-NEXT: .half 0 +; CHECK-NEXT: .half 0 +; CHECK-NEXT: .word +; Loc 1: constant double as large constant integer +; CHECK-NEXT: .byte 5 +; CHECK-NEXT: .byte 0 +; CHECK-NEXT: .half 8 +; CHECK-NEXT: .half 0 +; CHECK-NEXT: .half 0 +; CHECK-NEXT: .word +; Loc 2: constant half as constant integer +; CHECK-NEXT: .byte 4 +; CHECK-NEXT: .byte 0 +; CHECK-NEXT: .half 8 +; CHECK-NEXT: .half 0 +; CHECK-NEXT: .half 0 +; CHECK-NEXT: .word +; Loc 3: constant bfloat as constant integer +; CHECK-NEXT: .byte 4 +; CHECK-NEXT: .byte 0 +; CHECK-NEXT: .half 8 +; CHECK-NEXT: .half 0 +; CHECK-NEXT: .half 0 +; CHECK-NEXT: .word +; Loc 4: float value in X register +; CHECK-NEXT: .byte 1 +; CHECK-NEXT: .byte 0 +; CHECK-NEXT: .half 8 +; CHECK-NEXT: .half 10 +; CHECK-NEXT: .half 0 +; CHECK-NEXT: .word +; Loc 5: double value in X register +; CHECK-NEXT: .byte 1 +; CHECK-NEXT: .byte 0 +; CHECK-NEXT: .half 8 +; CHECK-NEXT: .half 11 +; CHECK-NEXT: .half 0 +; CHECK-NEXT: .word +; Loc 6: half value in X register +; CHECK-NEXT: .byte 1 +; CHECK-NEXT: .byte 0 +; CHECK-NEXT: .half 8 +; CHECK-NEXT: .half 12 +; CHECK-NEXT: .half 0 +; CHECK-NEXT: .word +; Loc 7: bfloat value in X register +; CHECK-NEXT: .byte 1 +; CHECK-NEXT: .byte 0 +; CHECK-NEXT: .half 8 +; CHECK-NEXT: .half 13 +; CHECK-NEXT: .half 0 +; CHECK-NEXT: .word +; Loc 8: float on stack +; CHECK-NEXT: .byte 2 +; CHECK-NEXT: .byte 0 +; CHECK-NEXT: .half 8 +; CHECK-NEXT: .half 2 +; CHECK-NEXT: .half 0 +; CHECK-NEXT: .word +; Loc 9: double on stack +; CHECK-NEXT: .byte 2 +; CHECK-NEXT: .byte 0 +; CHECK-NEXT: .half 8 +; CHECK-NEXT: .half 2 +; CHECK-NEXT: .half 0 +; CHECK-NEXT: .word +; Loc 10: half on stack +; CHECK-NEXT: .byte 2 +; CHECK-NEXT: .byte 0 +; CHECK-NEXT: .half 8 +; CHECK-NEXT: .half 2 +; CHECK-NEXT: .half 0 +; CHECK-NEXT: .word +; Loc 11: bfloat on stack +; CHECK-NEXT: .byte 2 +; CHECK-NEXT: .byte 0 +; CHECK-NEXT: .half 8 +; CHECK-NEXT: .half 2 +; CHECK-NEXT: .half 0 +; CHECK-NEXT: .word +define void @floats(float %f, double %g, half %h, bfloat %i) { + %ff = alloca float + %gg = alloca double + %hh = alloca half + %ii = alloca bfloat + call void (i64, i32, ...) @llvm.experimental.stackmap(i64 888, i32 0, float 1.25, + double 1.5, half 1.5, bfloat 1.5, float %f, double %g, half %h, bfloat %i, ptr %ff, ptr %gg, ptr %hh, ptr %ii) + ret void +} + declare void @llvm.experimental.stackmap(i64, i32, ...) declare void @llvm.experimental.patchpoint.void(i64, i32, ptr, i32, ...) declare i64 @llvm.experimental.patchpoint.i64(i64, i32, ptr, i32, ...) diff --git a/llvm/test/CodeGen/RISCV/rv64p.ll b/llvm/test/CodeGen/RISCV/rv64p.ll index cb07f94..f937f44 100644 --- a/llvm/test/CodeGen/RISCV/rv64p.ll +++ b/llvm/test/CodeGen/RISCV/rv64p.ll @@ -297,8 +297,7 @@ declare i32 @llvm.abs.i32(i32, i1 immarg) define i32 @abs_i32(i32 %x) { ; CHECK-LABEL: abs_i32: ; CHECK: # %bb.0: -; CHECK-NEXT: sext.w a0, a0 -; CHECK-NEXT: abs a0, a0 +; CHECK-NEXT: absw a0, a0 ; CHECK-NEXT: ret %abs = tail call i32 @llvm.abs.i32(i32 %x, i1 true) ret i32 %abs @@ -307,8 +306,7 @@ define i32 @abs_i32(i32 %x) { define signext i32 @abs_i32_sext(i32 signext %x) { ; CHECK-LABEL: abs_i32_sext: ; CHECK: # %bb.0: -; CHECK-NEXT: abs a0, a0 -; CHECK-NEXT: sext.w a0, a0 +; CHECK-NEXT: absw a0, a0 ; CHECK-NEXT: ret %abs = tail call i32 @llvm.abs.i32(i32 %x, i1 true) ret i32 %abs diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-masked-gather.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-masked-gather.ll index 4c35b25..7e6f2c7 100644 --- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-masked-gather.ll +++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-masked-gather.ll @@ -15265,6 +15265,259 @@ define <4 x i32> @masked_gather_widen_sew_negative_stride(ptr %base) { ret <4 x i32> %x } +define <7 x i8> @mgather_baseidx_v7i8(ptr %base, <7 x i8> %idxs, <7 x i1> %m, <7 x i8> %passthru) { +; RV32-LABEL: mgather_baseidx_v7i8: +; RV32: # %bb.0: +; RV32-NEXT: li a1, 127 +; RV32-NEXT: vsetivli zero, 8, e32, m2, ta, ma +; RV32-NEXT: vmv.s.x v10, a1 +; RV32-NEXT: vmand.mm v0, v0, v10 +; RV32-NEXT: vsext.vf4 v10, v8 +; RV32-NEXT: vsetvli zero, zero, e8, mf2, ta, mu +; RV32-NEXT: vluxei32.v v9, (a0), v10, v0.t +; RV32-NEXT: vmv1r.v v8, v9 +; RV32-NEXT: ret +; +; RV64V-LABEL: mgather_baseidx_v7i8: +; RV64V: # %bb.0: +; RV64V-NEXT: li a1, 127 +; RV64V-NEXT: vsetivli zero, 8, e64, m4, ta, ma +; RV64V-NEXT: vmv.s.x v10, a1 +; RV64V-NEXT: vmand.mm v0, v0, v10 +; RV64V-NEXT: vsext.vf8 v12, v8 +; RV64V-NEXT: vsetvli zero, zero, e8, mf2, ta, mu +; RV64V-NEXT: vluxei64.v v9, (a0), v12, v0.t +; RV64V-NEXT: vmv1r.v v8, v9 +; RV64V-NEXT: ret +; +; RV64ZVE32F-LABEL: mgather_baseidx_v7i8: +; RV64ZVE32F: # %bb.0: +; RV64ZVE32F-NEXT: addi sp, sp, -16 +; RV64ZVE32F-NEXT: .cfi_def_cfa_offset 16 +; RV64ZVE32F-NEXT: .cfi_remember_state +; RV64ZVE32F-NEXT: li a1, 64 +; RV64ZVE32F-NEXT: addi a2, sp, 8 +; RV64ZVE32F-NEXT: vsetvli zero, a1, e8, m4, ta, ma +; RV64ZVE32F-NEXT: vsm.v v0, (a2) +; RV64ZVE32F-NEXT: ld a1, 8(sp) +; RV64ZVE32F-NEXT: andi a2, a1, 1 +; RV64ZVE32F-NEXT: beqz a2, .LBB132_2 +; RV64ZVE32F-NEXT: # %bb.1: # %cond.load +; RV64ZVE32F-NEXT: vmv.x.s a2, v8 +; RV64ZVE32F-NEXT: add a2, a0, a2 +; RV64ZVE32F-NEXT: lbu a2, 0(a2) +; RV64ZVE32F-NEXT: vsetivli zero, 8, e8, mf2, ta, ma +; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1 +; RV64ZVE32F-NEXT: vmv.v.x v11, a2 +; RV64ZVE32F-NEXT: vmv.x.s a2, v10 +; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 2 +; RV64ZVE32F-NEXT: vslide1down.vx v11, v11, a2 +; RV64ZVE32F-NEXT: vmv.x.s a2, v10 +; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 3 +; RV64ZVE32F-NEXT: vslide1down.vx v11, v11, a2 +; RV64ZVE32F-NEXT: vmv.x.s a2, v10 +; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 4 +; RV64ZVE32F-NEXT: vslide1down.vx v11, v11, a2 +; RV64ZVE32F-NEXT: vmv.x.s a2, v10 +; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 5 +; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 6 +; RV64ZVE32F-NEXT: vslide1down.vx v11, v11, a2 +; RV64ZVE32F-NEXT: vmv.x.s a2, v10 +; RV64ZVE32F-NEXT: vslide1down.vx v10, v11, a2 +; RV64ZVE32F-NEXT: vmv.x.s a2, v9 +; RV64ZVE32F-NEXT: vslide1down.vx v9, v10, a2 +; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1 +; RV64ZVE32F-NEXT: .LBB132_2: # %else +; RV64ZVE32F-NEXT: andi a2, a1, 2 +; RV64ZVE32F-NEXT: beqz a2, .LBB132_4 +; RV64ZVE32F-NEXT: # %bb.3: # %cond.load1 +; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma +; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 1 +; RV64ZVE32F-NEXT: vmv.x.s a2, v9 +; RV64ZVE32F-NEXT: vsetivli zero, 8, e8, mf2, ta, ma +; RV64ZVE32F-NEXT: vslidedown.vi v11, v9, 2 +; RV64ZVE32F-NEXT: vmv.x.s a3, v10 +; RV64ZVE32F-NEXT: add a3, a0, a3 +; RV64ZVE32F-NEXT: lbu a3, 0(a3) +; RV64ZVE32F-NEXT: vmv.v.x v10, a2 +; RV64ZVE32F-NEXT: vmv.x.s a2, v11 +; RV64ZVE32F-NEXT: vslidedown.vi v11, v9, 3 +; RV64ZVE32F-NEXT: vslide1down.vx v10, v10, a3 +; RV64ZVE32F-NEXT: vmv.x.s a3, v11 +; RV64ZVE32F-NEXT: vslidedown.vi v11, v9, 4 +; RV64ZVE32F-NEXT: vslide1down.vx v10, v10, a2 +; RV64ZVE32F-NEXT: vmv.x.s a2, v11 +; RV64ZVE32F-NEXT: vslidedown.vi v11, v9, 5 +; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 6 +; RV64ZVE32F-NEXT: vslide1down.vx v10, v10, a3 +; RV64ZVE32F-NEXT: vmv.x.s a3, v11 +; RV64ZVE32F-NEXT: vslide1down.vx v10, v10, a2 +; RV64ZVE32F-NEXT: vslide1down.vx v10, v10, a3 +; RV64ZVE32F-NEXT: vmv.x.s a2, v9 +; RV64ZVE32F-NEXT: vslide1down.vx v9, v10, a2 +; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1 +; RV64ZVE32F-NEXT: .LBB132_4: # %else2 +; RV64ZVE32F-NEXT: andi a2, a1, 4 +; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma +; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 2 +; RV64ZVE32F-NEXT: beqz a2, .LBB132_6 +; RV64ZVE32F-NEXT: # %bb.5: # %cond.load4 +; RV64ZVE32F-NEXT: vmv.x.s a2, v10 +; RV64ZVE32F-NEXT: vsetivli zero, 8, e8, mf2, ta, ma +; RV64ZVE32F-NEXT: vslidedown.vi v11, v9, 1 +; RV64ZVE32F-NEXT: vmv.x.s a3, v9 +; RV64ZVE32F-NEXT: vslidedown.vi v12, v9, 3 +; RV64ZVE32F-NEXT: vmv.x.s a4, v11 +; RV64ZVE32F-NEXT: vmv.v.x v11, a3 +; RV64ZVE32F-NEXT: vmv.x.s a3, v12 +; RV64ZVE32F-NEXT: vslidedown.vi v12, v9, 4 +; RV64ZVE32F-NEXT: add a2, a0, a2 +; RV64ZVE32F-NEXT: lbu a2, 0(a2) +; RV64ZVE32F-NEXT: vslide1down.vx v11, v11, a4 +; RV64ZVE32F-NEXT: vmv.x.s a4, v12 +; RV64ZVE32F-NEXT: vslidedown.vi v12, v9, 5 +; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 6 +; RV64ZVE32F-NEXT: vslide1down.vx v11, v11, a2 +; RV64ZVE32F-NEXT: vmv.x.s a2, v12 +; RV64ZVE32F-NEXT: vslide1down.vx v11, v11, a3 +; RV64ZVE32F-NEXT: vslide1down.vx v11, v11, a4 +; RV64ZVE32F-NEXT: vslide1down.vx v11, v11, a2 +; RV64ZVE32F-NEXT: vmv.x.s a2, v9 +; RV64ZVE32F-NEXT: vslide1down.vx v9, v11, a2 +; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1 +; RV64ZVE32F-NEXT: .LBB132_6: # %else5 +; RV64ZVE32F-NEXT: andi a2, a1, 8 +; RV64ZVE32F-NEXT: beqz a2, .LBB132_8 +; RV64ZVE32F-NEXT: # %bb.7: # %cond.load7 +; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma +; RV64ZVE32F-NEXT: vslidedown.vi v10, v10, 1 +; RV64ZVE32F-NEXT: vsetivli zero, 8, e8, mf2, ta, ma +; RV64ZVE32F-NEXT: vslidedown.vi v11, v9, 1 +; RV64ZVE32F-NEXT: vmv.x.s a2, v9 +; RV64ZVE32F-NEXT: vmv.x.s a3, v10 +; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 2 +; RV64ZVE32F-NEXT: vmv.x.s a4, v11 +; RV64ZVE32F-NEXT: vmv.v.x v11, a2 +; RV64ZVE32F-NEXT: vmv.x.s a2, v10 +; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 4 +; RV64ZVE32F-NEXT: vslide1down.vx v11, v11, a4 +; RV64ZVE32F-NEXT: vmv.x.s a4, v10 +; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 5 +; RV64ZVE32F-NEXT: add a3, a0, a3 +; RV64ZVE32F-NEXT: lbu a3, 0(a3) +; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 6 +; RV64ZVE32F-NEXT: vslide1down.vx v11, v11, a2 +; RV64ZVE32F-NEXT: vmv.x.s a2, v10 +; RV64ZVE32F-NEXT: vslide1down.vx v10, v11, a3 +; RV64ZVE32F-NEXT: vslide1down.vx v10, v10, a4 +; RV64ZVE32F-NEXT: vslide1down.vx v10, v10, a2 +; RV64ZVE32F-NEXT: vmv.x.s a2, v9 +; RV64ZVE32F-NEXT: vslide1down.vx v9, v10, a2 +; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1 +; RV64ZVE32F-NEXT: .LBB132_8: # %else8 +; RV64ZVE32F-NEXT: andi a2, a1, 16 +; RV64ZVE32F-NEXT: vsetivli zero, 4, e8, mf2, ta, ma +; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 4 +; RV64ZVE32F-NEXT: bnez a2, .LBB132_13 +; RV64ZVE32F-NEXT: # %bb.9: # %else11 +; RV64ZVE32F-NEXT: andi a2, a1, 32 +; RV64ZVE32F-NEXT: bnez a2, .LBB132_14 +; RV64ZVE32F-NEXT: .LBB132_10: # %else14 +; RV64ZVE32F-NEXT: andi a1, a1, 64 +; RV64ZVE32F-NEXT: beqz a1, .LBB132_12 +; RV64ZVE32F-NEXT: .LBB132_11: # %cond.load16 +; RV64ZVE32F-NEXT: vsetivli zero, 2, e8, mf4, ta, ma +; RV64ZVE32F-NEXT: vslidedown.vi v8, v8, 2 +; RV64ZVE32F-NEXT: vsetivli zero, 8, e8, mf2, ta, ma +; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1 +; RV64ZVE32F-NEXT: vmv.x.s a1, v9 +; RV64ZVE32F-NEXT: vmv.x.s a2, v8 +; RV64ZVE32F-NEXT: vmv.v.x v8, a1 +; RV64ZVE32F-NEXT: vmv.x.s a1, v10 +; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 2 +; RV64ZVE32F-NEXT: add a0, a0, a2 +; RV64ZVE32F-NEXT: vmv.x.s a2, v10 +; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 3 +; RV64ZVE32F-NEXT: vslide1down.vx v8, v8, a1 +; RV64ZVE32F-NEXT: vmv.x.s a1, v10 +; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 4 +; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 5 +; RV64ZVE32F-NEXT: vslide1down.vx v8, v8, a2 +; RV64ZVE32F-NEXT: vmv.x.s a2, v10 +; RV64ZVE32F-NEXT: lbu a0, 0(a0) +; RV64ZVE32F-NEXT: vslide1down.vx v8, v8, a1 +; RV64ZVE32F-NEXT: vslide1down.vx v8, v8, a2 +; RV64ZVE32F-NEXT: vmv.x.s a1, v9 +; RV64ZVE32F-NEXT: vslide1down.vx v8, v8, a1 +; RV64ZVE32F-NEXT: vslide1down.vx v8, v8, a0 +; RV64ZVE32F-NEXT: vslidedown.vi v9, v8, 1 +; RV64ZVE32F-NEXT: .LBB132_12: # %else17 +; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, m1, ta, ma +; RV64ZVE32F-NEXT: vmv1r.v v8, v9 +; RV64ZVE32F-NEXT: addi sp, sp, 16 +; RV64ZVE32F-NEXT: .cfi_def_cfa_offset 0 +; RV64ZVE32F-NEXT: ret +; RV64ZVE32F-NEXT: .LBB132_13: # %cond.load10 +; RV64ZVE32F-NEXT: .cfi_restore_state +; RV64ZVE32F-NEXT: vmv.x.s a2, v8 +; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 1 +; RV64ZVE32F-NEXT: vmv.x.s a3, v9 +; RV64ZVE32F-NEXT: vslidedown.vi v11, v9, 2 +; RV64ZVE32F-NEXT: vmv.x.s a4, v10 +; RV64ZVE32F-NEXT: vsetivli zero, 8, e8, mf2, ta, ma +; RV64ZVE32F-NEXT: vmv.v.x v10, a3 +; RV64ZVE32F-NEXT: vmv.x.s a3, v11 +; RV64ZVE32F-NEXT: vslidedown.vi v11, v9, 3 +; RV64ZVE32F-NEXT: vslide1down.vx v10, v10, a4 +; RV64ZVE32F-NEXT: vmv.x.s a4, v11 +; RV64ZVE32F-NEXT: vslidedown.vi v11, v9, 5 +; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 6 +; RV64ZVE32F-NEXT: add a2, a0, a2 +; RV64ZVE32F-NEXT: lbu a2, 0(a2) +; RV64ZVE32F-NEXT: vslide1down.vx v10, v10, a3 +; RV64ZVE32F-NEXT: vmv.x.s a3, v11 +; RV64ZVE32F-NEXT: vslide1down.vx v10, v10, a4 +; RV64ZVE32F-NEXT: vslide1down.vx v10, v10, a2 +; RV64ZVE32F-NEXT: vslide1down.vx v10, v10, a3 +; RV64ZVE32F-NEXT: vmv.x.s a2, v9 +; RV64ZVE32F-NEXT: vslide1down.vx v9, v10, a2 +; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1 +; RV64ZVE32F-NEXT: andi a2, a1, 32 +; RV64ZVE32F-NEXT: beqz a2, .LBB132_10 +; RV64ZVE32F-NEXT: .LBB132_14: # %cond.load13 +; RV64ZVE32F-NEXT: vsetivli zero, 1, e8, mf4, ta, ma +; RV64ZVE32F-NEXT: vslidedown.vi v10, v8, 1 +; RV64ZVE32F-NEXT: vsetivli zero, 8, e8, mf2, ta, ma +; RV64ZVE32F-NEXT: vslidedown.vi v11, v9, 1 +; RV64ZVE32F-NEXT: vmv.x.s a2, v9 +; RV64ZVE32F-NEXT: vmv.x.s a3, v10 +; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 2 +; RV64ZVE32F-NEXT: vmv.x.s a4, v11 +; RV64ZVE32F-NEXT: vmv.v.x v11, a2 +; RV64ZVE32F-NEXT: vmv.x.s a2, v10 +; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 3 +; RV64ZVE32F-NEXT: vslide1down.vx v11, v11, a4 +; RV64ZVE32F-NEXT: vmv.x.s a4, v10 +; RV64ZVE32F-NEXT: vslidedown.vi v10, v9, 4 +; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 6 +; RV64ZVE32F-NEXT: add a3, a0, a3 +; RV64ZVE32F-NEXT: lbu a3, 0(a3) +; RV64ZVE32F-NEXT: vslide1down.vx v11, v11, a2 +; RV64ZVE32F-NEXT: vmv.x.s a2, v10 +; RV64ZVE32F-NEXT: vslide1down.vx v10, v11, a4 +; RV64ZVE32F-NEXT: vslide1down.vx v10, v10, a2 +; RV64ZVE32F-NEXT: vslide1down.vx v10, v10, a3 +; RV64ZVE32F-NEXT: vmv.x.s a2, v9 +; RV64ZVE32F-NEXT: vslide1down.vx v9, v10, a2 +; RV64ZVE32F-NEXT: vslidedown.vi v9, v9, 1 +; RV64ZVE32F-NEXT: andi a1, a1, 64 +; RV64ZVE32F-NEXT: bnez a1, .LBB132_11 +; RV64ZVE32F-NEXT: j .LBB132_12 + %ptrs = getelementptr inbounds i8, ptr %base, <7 x i8> %idxs + %v = call <7 x i8> @llvm.masked.gather.v7i8.v7p0(<7 x ptr> %ptrs, i32 1, <7 x i1> %m, <7 x i8> %passthru) + ret <7 x i8> %v +} + ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line: ; RV32V-ZVFH: {{.*}} ; RV32V-ZVFHMIN: {{.*}} diff --git a/llvm/test/CodeGen/RISCV/rvv/pr165232.ll b/llvm/test/CodeGen/RISCV/rvv/pr165232.ll new file mode 100644 index 0000000..bef53c6 --- /dev/null +++ b/llvm/test/CodeGen/RISCV/rvv/pr165232.ll @@ -0,0 +1,244 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc < %s -mtriple=riscv64 -mattr=+v | FileCheck %s + +target datalayout = "e-m:e-p:64:64-i64:64-i128:128-n32:64-S128" +target triple = "riscv64-unknown-linux-gnu" + +define i1 @main(ptr %var_117, ptr %arrayinit.element3045, ptr %arrayinit.element3047, ptr %arrayinit.element3049, ptr %arrayinit.element3051, ptr %arrayinit.element3053, ptr %arrayinit.element3055, ptr %arrayinit.element3057, ptr %arrayinit.element3059, ptr %arrayinit.element3061, ptr %arrayinit.element3063, ptr %arrayinit.element3065, ptr %arrayinit.element3067, i64 %var_94_i.07698, target("riscv.vector.tuple", <vscale x 16 x i8>, 2) %0, target("riscv.vector.tuple", <vscale x 16 x i8>, 4) %1) { +; CHECK-LABEL: main: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: addi sp, sp, -16 +; CHECK-NEXT: .cfi_def_cfa_offset 16 +; CHECK-NEXT: csrr t0, vlenb +; CHECK-NEXT: slli t0, t0, 3 +; CHECK-NEXT: mv t1, t0 +; CHECK-NEXT: slli t0, t0, 1 +; CHECK-NEXT: add t0, t0, t1 +; CHECK-NEXT: sub sp, sp, t0 +; CHECK-NEXT: .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x10, 0x22, 0x11, 0x18, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 16 + 24 * vlenb +; CHECK-NEXT: sd a1, 8(sp) # 8-byte Folded Spill +; CHECK-NEXT: sd a2, 0(sp) # 8-byte Folded Spill +; CHECK-NEXT: csrr a1, vlenb +; CHECK-NEXT: slli a1, a1, 4 +; CHECK-NEXT: add a1, sp, a1 +; CHECK-NEXT: addi a1, a1, 16 +; CHECK-NEXT: vs4r.v v12, (a1) # vscale x 32-byte Folded Spill +; CHECK-NEXT: csrr a2, vlenb +; CHECK-NEXT: slli a2, a2, 2 +; CHECK-NEXT: add a1, a1, a2 +; CHECK-NEXT: vs4r.v v16, (a1) # vscale x 32-byte Folded Spill +; CHECK-NEXT: csrr a1, vlenb +; CHECK-NEXT: slli a1, a1, 2 +; CHECK-NEXT: mv a2, a1 +; CHECK-NEXT: slli a1, a1, 1 +; CHECK-NEXT: add a1, a1, a2 +; CHECK-NEXT: add a1, sp, a1 +; CHECK-NEXT: addi a1, a1, 16 +; CHECK-NEXT: vs4r.v v8, (a1) # vscale x 32-byte Folded Spill +; CHECK-NEXT: csrr a1, vlenb +; CHECK-NEXT: slli a1, a1, 3 +; CHECK-NEXT: mv a2, a1 +; CHECK-NEXT: slli a1, a1, 1 +; CHECK-NEXT: add a1, a1, a2 +; CHECK-NEXT: add a1, sp, a1 +; CHECK-NEXT: ld t0, 56(a1) +; CHECK-NEXT: csrr a1, vlenb +; CHECK-NEXT: slli a1, a1, 3 +; CHECK-NEXT: mv a2, a1 +; CHECK-NEXT: slli a1, a1, 1 +; CHECK-NEXT: add a1, a1, a2 +; CHECK-NEXT: add a1, sp, a1 +; CHECK-NEXT: ld t1, 48(a1) +; CHECK-NEXT: vsetvli t2, zero, e8, m1, ta, ma +; CHECK-NEXT: vmv.v.i v9, 0 +; CHECK-NEXT: csrr a1, vlenb +; CHECK-NEXT: slli a1, a1, 3 +; CHECK-NEXT: mv a2, a1 +; CHECK-NEXT: slli a1, a1, 1 +; CHECK-NEXT: add a1, a1, a2 +; CHECK-NEXT: add a1, sp, a1 +; CHECK-NEXT: ld t2, 40(a1) +; CHECK-NEXT: # kill: def $v10 killed $v9 killed $vtype +; CHECK-NEXT: csrr a1, vlenb +; CHECK-NEXT: slli a1, a1, 3 +; CHECK-NEXT: mv a2, a1 +; CHECK-NEXT: slli a1, a1, 1 +; CHECK-NEXT: add a1, a1, a2 +; CHECK-NEXT: add a1, sp, a1 +; CHECK-NEXT: ld t3, 32(a1) +; CHECK-NEXT: vmv.v.i v11, 0 +; CHECK-NEXT: csrr a1, vlenb +; CHECK-NEXT: slli a1, a1, 3 +; CHECK-NEXT: mv a2, a1 +; CHECK-NEXT: slli a1, a1, 1 +; CHECK-NEXT: add a1, a1, a2 +; CHECK-NEXT: add a1, sp, a1 +; CHECK-NEXT: ld t4, 16(a1) +; CHECK-NEXT: vmv.v.i v12, 0 +; CHECK-NEXT: csrr a1, vlenb +; CHECK-NEXT: slli a1, a1, 3 +; CHECK-NEXT: mv a2, a1 +; CHECK-NEXT: slli a1, a1, 1 +; CHECK-NEXT: add a1, a1, a2 +; CHECK-NEXT: add a1, sp, a1 +; CHECK-NEXT: ld t5, 24(a1) +; CHECK-NEXT: vmv.v.i v13, 0 +; CHECK-NEXT: vsetvli t6, zero, e8, m2, ta, ma +; CHECK-NEXT: vmv.v.i v22, 0 +; CHECK-NEXT: vmv1r.v v14, v9 +; CHECK-NEXT: sd zero, 0(a0) +; CHECK-NEXT: vmv.v.i v24, 0 +; CHECK-NEXT: vmv1r.v v15, v9 +; CHECK-NEXT: vmv1r.v v18, v9 +; CHECK-NEXT: li t6, 1023 +; CHECK-NEXT: vmv.v.i v26, 0 +; CHECK-NEXT: vmv1r.v v19, v9 +; CHECK-NEXT: slli t6, t6, 52 +; CHECK-NEXT: vmv.v.i v28, 0 +; CHECK-NEXT: addi a1, sp, 16 +; CHECK-NEXT: vs2r.v v22, (a1) # vscale x 16-byte Folded Spill +; CHECK-NEXT: csrr a2, vlenb +; CHECK-NEXT: slli a2, a2, 1 +; CHECK-NEXT: add a1, a1, a2 +; CHECK-NEXT: vs4r.v v24, (a1) # vscale x 32-byte Folded Spill +; CHECK-NEXT: slli a2, a2, 1 +; CHECK-NEXT: add a1, a1, a2 +; CHECK-NEXT: ld a2, 0(sp) # 8-byte Folded Reload +; CHECK-NEXT: vs2r.v v28, (a1) # vscale x 16-byte Folded Spill +; CHECK-NEXT: ld a1, 8(sp) # 8-byte Folded Reload +; CHECK-NEXT: vmv1r.v v20, v9 +; CHECK-NEXT: sd t6, 0(t5) +; CHECK-NEXT: vmv2r.v v16, v14 +; CHECK-NEXT: vmv2r.v v14, v12 +; CHECK-NEXT: vmv2r.v v12, v10 +; CHECK-NEXT: vmv1r.v v11, v9 +; CHECK-NEXT: vmv1r.v v21, v9 +; CHECK-NEXT: csrr t5, vlenb +; CHECK-NEXT: slli t5, t5, 3 +; CHECK-NEXT: add t5, sp, t5 +; CHECK-NEXT: addi t5, t5, 16 +; CHECK-NEXT: vs2r.v v18, (t5) # vscale x 16-byte Folded Spill +; CHECK-NEXT: csrr t6, vlenb +; CHECK-NEXT: slli t6, t6, 1 +; CHECK-NEXT: add t5, t5, t6 +; CHECK-NEXT: vs2r.v v20, (t5) # vscale x 16-byte Folded Spill +; CHECK-NEXT: vsetivli zero, 0, e32, m1, ta, ma +; CHECK-NEXT: vmv.v.i v19, 0 +; CHECK-NEXT: vmclr.m v10 +; CHECK-NEXT: vsetvli zero, zero, e64, m2, ta, ma +; CHECK-NEXT: vmv.v.i v6, 0 +; CHECK-NEXT: .LBB0_1: # %for.body +; CHECK-NEXT: # =>This Inner Loop Header: Depth=1 +; CHECK-NEXT: vmv1r.v v0, v10 +; CHECK-NEXT: vmv1r.v v20, v19 +; CHECK-NEXT: vmv1r.v v3, v19 +; CHECK-NEXT: vmv1r.v v5, v19 +; CHECK-NEXT: vmv1r.v v2, v19 +; CHECK-NEXT: vmv1r.v v31, v19 +; CHECK-NEXT: vmv1r.v v30, v19 +; CHECK-NEXT: vmv1r.v v4, v19 +; CHECK-NEXT: vmv2r.v v22, v10 +; CHECK-NEXT: vmv4r.v v24, v12 +; CHECK-NEXT: vmv2r.v v28, v16 +; CHECK-NEXT: vmv2r.v v8, v6 +; CHECK-NEXT: vmv1r.v v18, v19 +; CHECK-NEXT: vmv1r.v v21, v10 +; CHECK-NEXT: vsetvli zero, zero, e32, m1, tu, ma +; CHECK-NEXT: vle32.v v20, (t4) +; CHECK-NEXT: vle32.v v3, (t1) +; CHECK-NEXT: vle32.v v30, (a7) +; CHECK-NEXT: vle64.v v8, (a4) +; CHECK-NEXT: vle32.v v5, (t2) +; CHECK-NEXT: vle32.v v2, (t3) +; CHECK-NEXT: vle32.v v31, (a6) +; CHECK-NEXT: vmv1r.v v24, v30 +; CHECK-NEXT: vsetvli zero, zero, e64, m2, ta, mu +; CHECK-NEXT: vmflt.vv v21, v8, v6, v0.t +; CHECK-NEXT: vmv1r.v v8, v19 +; CHECK-NEXT: vsetvli zero, zero, e32, m1, tu, mu +; CHECK-NEXT: vle32.v v18, (a2) +; CHECK-NEXT: vle32.v v8, (a3) +; CHECK-NEXT: vle32.v v4, (a5) +; CHECK-NEXT: vmv1r.v v22, v20 +; CHECK-NEXT: csrr t5, vlenb +; CHECK-NEXT: slli t5, t5, 3 +; CHECK-NEXT: add t5, sp, t5 +; CHECK-NEXT: addi t5, t5, 16 +; CHECK-NEXT: vl1r.v v1, (t5) # vscale x 8-byte Folded Reload +; CHECK-NEXT: csrr t6, vlenb +; CHECK-NEXT: add t5, t5, t6 +; CHECK-NEXT: vl2r.v v2, (t5) # vscale x 16-byte Folded Reload +; CHECK-NEXT: slli t6, t6, 1 +; CHECK-NEXT: add t5, t5, t6 +; CHECK-NEXT: vl1r.v v4, (t5) # vscale x 8-byte Folded Reload +; CHECK-NEXT: vsseg4e32.v v1, (zero) +; CHECK-NEXT: vsseg8e32.v v22, (a1) +; CHECK-NEXT: vmv1r.v v0, v21 +; CHECK-NEXT: vssub.vv v8, v19, v18, v0.t +; CHECK-NEXT: csrr t5, vlenb +; CHECK-NEXT: slli t5, t5, 2 +; CHECK-NEXT: mv t6, t5 +; CHECK-NEXT: slli t5, t5, 1 +; CHECK-NEXT: add t5, t5, t6 +; CHECK-NEXT: add t5, sp, t5 +; CHECK-NEXT: addi t5, t5, 16 +; CHECK-NEXT: vl4r.v v20, (t5) # vscale x 32-byte Folded Reload +; CHECK-NEXT: vsetvli zero, t0, e64, m2, ta, ma +; CHECK-NEXT: vsseg2e64.v v20, (zero) +; CHECK-NEXT: vmv1r.v v0, v10 +; CHECK-NEXT: addi t5, sp, 16 +; CHECK-NEXT: vl4r.v v20, (t5) # vscale x 32-byte Folded Reload +; CHECK-NEXT: csrr t6, vlenb +; CHECK-NEXT: slli t6, t6, 2 +; CHECK-NEXT: add t5, t5, t6 +; CHECK-NEXT: vl4r.v v24, (t5) # vscale x 32-byte Folded Reload +; CHECK-NEXT: vsetivli zero, 0, e64, m2, ta, ma +; CHECK-NEXT: vsseg4e64.v v20, (zero), v0.t +; CHECK-NEXT: vsetvli zero, zero, e32, m1, ta, ma +; CHECK-NEXT: vsseg8e32.v v8, (a0) +; CHECK-NEXT: csrr t5, vlenb +; CHECK-NEXT: slli t5, t5, 4 +; CHECK-NEXT: add t5, sp, t5 +; CHECK-NEXT: addi t5, t5, 16 +; CHECK-NEXT: vl4r.v v20, (t5) # vscale x 32-byte Folded Reload +; CHECK-NEXT: csrr t6, vlenb +; CHECK-NEXT: slli t6, t6, 2 +; CHECK-NEXT: add t5, t5, t6 +; CHECK-NEXT: vl4r.v v24, (t5) # vscale x 32-byte Folded Reload +; CHECK-NEXT: vsetvli zero, zero, e64, m2, ta, ma +; CHECK-NEXT: vsseg4e64.v v20, (zero) +; CHECK-NEXT: j .LBB0_1 +entry: + store double 0.000000e+00, ptr %var_117, align 8 + store double 1.000000e+00, ptr %arrayinit.element3061, align 8 + br label %for.body + +for.body: ; preds = %for.body, %entry + %2 = call <vscale x 2 x float> @llvm.riscv.vle.nxv2f32.p0.i64(<vscale x 2 x float> zeroinitializer, ptr %arrayinit.element3059, i64 0) + %3 = call <vscale x 2 x float> @llvm.riscv.vle.nxv2f32.p0.i64(<vscale x 2 x float> zeroinitializer, ptr %arrayinit.element3067, i64 0) + %4 = call <vscale x 2 x float> @llvm.riscv.vle.nxv2f32.p0.i64(<vscale x 2 x float> zeroinitializer, ptr %arrayinit.element3065, i64 0) + %5 = call <vscale x 2 x float> @llvm.riscv.vle.nxv2f32.p0.i64(<vscale x 2 x float> zeroinitializer, ptr %arrayinit.element3063, i64 0) + %6 = call <vscale x 2 x float> @llvm.riscv.vle.nxv2f32.p0.i64(<vscale x 2 x float> zeroinitializer, ptr %arrayinit.element3055, i64 0) + %7 = call <vscale x 2 x float> @llvm.riscv.vle.nxv2f32.p0.i64(<vscale x 2 x float> zeroinitializer, ptr %arrayinit.element3057, i64 0) + %8 = call <vscale x 2 x float> @llvm.riscv.vle.nxv2f32.p0.i64(<vscale x 2 x float> zeroinitializer, ptr %arrayinit.element3053, i64 0) + %9 = call <vscale x 2 x double> @llvm.riscv.vle.nxv2f64.p0.i64(<vscale x 2 x double> zeroinitializer, ptr %arrayinit.element3051, i64 0) + %10 = tail call <vscale x 2 x i32> @llvm.riscv.vle.nxv2i32.p0.i64(<vscale x 2 x i32> zeroinitializer, ptr %arrayinit.element3047, i64 0) + %11 = tail call <vscale x 2 x i32> @llvm.riscv.vle.nxv2i32.p0.i64(<vscale x 2 x i32> zeroinitializer, ptr %arrayinit.element3049, i64 0) + call void @llvm.riscv.vsseg4.triscv.vector.tuple_nxv8i8_4t.p0.i64(target("riscv.vector.tuple", <vscale x 8 x i8>, 4) zeroinitializer, ptr null, i64 0, i64 5) + %12 = tail call target("riscv.vector.tuple", <vscale x 8 x i8>, 8) @llvm.riscv.tuple.insert.triscv.vector.tuple_nxv8i8_8t.nxv2f32(target("riscv.vector.tuple", <vscale x 8 x i8>, 8) zeroinitializer, <vscale x 2 x float> %8, i32 0) + %13 = tail call target("riscv.vector.tuple", <vscale x 8 x i8>, 8) @llvm.riscv.tuple.insert.triscv.vector.tuple_nxv8i8_8t.nxv2f32(target("riscv.vector.tuple", <vscale x 8 x i8>, 8) %12, <vscale x 2 x float> %7, i32 2) + %14 = tail call target("riscv.vector.tuple", <vscale x 8 x i8>, 8) @llvm.riscv.tuple.insert.triscv.vector.tuple_nxv8i8_8t.nxv2f32(target("riscv.vector.tuple", <vscale x 8 x i8>, 8) %13, <vscale x 2 x float> %6, i32 0) + %15 = tail call target("riscv.vector.tuple", <vscale x 8 x i8>, 8) @llvm.riscv.tuple.insert.triscv.vector.tuple_nxv8i8_8t.nxv2f32(target("riscv.vector.tuple", <vscale x 8 x i8>, 8) %14, <vscale x 2 x float> %5, i32 0) + %16 = tail call target("riscv.vector.tuple", <vscale x 8 x i8>, 8) @llvm.riscv.tuple.insert.triscv.vector.tuple_nxv8i8_8t.nxv2f32(target("riscv.vector.tuple", <vscale x 8 x i8>, 8) %15, <vscale x 2 x float> %4, i32 0) + %17 = tail call target("riscv.vector.tuple", <vscale x 8 x i8>, 8) @llvm.riscv.tuple.insert.triscv.vector.tuple_nxv8i8_8t.nxv2f32(target("riscv.vector.tuple", <vscale x 8 x i8>, 8) %16, <vscale x 2 x float> %3, i32 0) + %18 = tail call target("riscv.vector.tuple", <vscale x 8 x i8>, 8) @llvm.riscv.tuple.insert.triscv.vector.tuple_nxv8i8_8t.nxv2f32(target("riscv.vector.tuple", <vscale x 8 x i8>, 8) %17, <vscale x 2 x float> %2, i32 0) + call void @llvm.riscv.vsseg8.triscv.vector.tuple_nxv8i8_8t.p0.i64(target("riscv.vector.tuple", <vscale x 8 x i8>, 8) %18, ptr %arrayinit.element3045, i64 0, i64 5) + %19 = tail call <vscale x 2 x i1> @llvm.riscv.vmfgt.mask.nxv2f64.nxv2f64.i64(<vscale x 2 x i1> zeroinitializer, <vscale x 2 x double> zeroinitializer, <vscale x 2 x double> %9, <vscale x 2 x i1> zeroinitializer, i64 0) + %20 = tail call <vscale x 2 x i32> @llvm.riscv.vssub.mask.nxv2i32.nxv2i32.i64(<vscale x 2 x i32> %11, <vscale x 2 x i32> zeroinitializer, <vscale x 2 x i32> %10, <vscale x 2 x i1> %19, i64 0, i64 0) + call void @llvm.riscv.vsseg2.triscv.vector.tuple_nxv16i8_2t.p0.i64(target("riscv.vector.tuple", <vscale x 16 x i8>, 2) %0, ptr null, i64 %var_94_i.07698, i64 6) + call void @llvm.riscv.vsseg4.mask.triscv.vector.tuple_nxv16i8_4t.p0.nxv2i1.i64(target("riscv.vector.tuple", <vscale x 16 x i8>, 4) zeroinitializer, ptr null, <vscale x 2 x i1> zeroinitializer, i64 0, i64 6) + %21 = tail call target("riscv.vector.tuple", <vscale x 8 x i8>, 8) @llvm.riscv.tuple.insert.triscv.vector.tuple_nxv8i8_8t.nxv2i32(target("riscv.vector.tuple", <vscale x 8 x i8>, 8) poison, <vscale x 2 x i32> %20, i32 0) + call void @llvm.riscv.vsseg8.triscv.vector.tuple_nxv8i8_8t.p0.i64(target("riscv.vector.tuple", <vscale x 8 x i8>, 8) %21, ptr %var_117, i64 0, i64 5) + call void @llvm.riscv.vsseg4.triscv.vector.tuple_nxv16i8_4t.p0.i64(target("riscv.vector.tuple", <vscale x 16 x i8>, 4) %1, ptr null, i64 0, i64 6) + br label %for.body +} diff --git a/llvm/test/CodeGen/RISCV/rvv/zvlsseg-spill.mir b/llvm/test/CodeGen/RISCV/rvv/zvlsseg-spill.mir index dd9960d..9c2fa9d 100644 --- a/llvm/test/CodeGen/RISCV/rvv/zvlsseg-spill.mir +++ b/llvm/test/CodeGen/RISCV/rvv/zvlsseg-spill.mir @@ -32,10 +32,10 @@ body: | ; CHECK-NEXT: $x11 = ADDI $x2, 16 ; CHECK-NEXT: VS4R_V $v0m4, $x11, implicit $v0_v1_v2_v3_v4_v5_v6 :: (store (<vscale x 1 x s256>) into %stack.0, align 8) ; CHECK-NEXT: $x12 = PseudoReadVLENB - ; CHECK-NEXT: $x13 = SLLI $x12, 2 - ; CHECK-NEXT: $x11 = ADD killed $x11, killed $x13 + ; CHECK-NEXT: $x12 = SLLI killed $x12, 2 + ; CHECK-NEXT: $x11 = ADD killed $x11, $x12 ; CHECK-NEXT: VS2R_V $v4m2, $x11, implicit $v0_v1_v2_v3_v4_v5_v6 :: (store (<vscale x 1 x s128>) into %stack.0, align 8) - ; CHECK-NEXT: $x12 = SLLI killed $x12, 1 + ; CHECK-NEXT: $x12 = SRLI killed $x12, 1 ; CHECK-NEXT: $x11 = ADD killed $x11, killed $x12 ; CHECK-NEXT: VS1R_V $v6, killed $x11, implicit $v0_v1_v2_v3_v4_v5_v6 :: (store (<vscale x 1 x s64>) into %stack.0) ; CHECK-NEXT: $x11 = ADDI $x2, 16 @@ -93,10 +93,10 @@ body: | ; CHECK-NEXT: $x11 = ADDI $x2, 16 ; CHECK-NEXT: $v10m2 = VL2RE8_V $x11 :: (load (<vscale x 1 x s128>) from %stack.0, align 8) ; CHECK-NEXT: $x12 = PseudoReadVLENB - ; CHECK-NEXT: $x13 = SLLI $x12, 1 - ; CHECK-NEXT: $x11 = ADD killed $x11, killed $x13 + ; CHECK-NEXT: $x12 = SLLI killed $x12, 1 + ; CHECK-NEXT: $x11 = ADD killed $x11, $x12 ; CHECK-NEXT: $v12m4 = VL4RE8_V $x11 :: (load (<vscale x 1 x s256>) from %stack.0, align 8) - ; CHECK-NEXT: $x12 = SLLI killed $x12, 2 + ; CHECK-NEXT: $x12 = SLLI killed $x12, 1 ; CHECK-NEXT: $x11 = ADD killed $x11, killed $x12 ; CHECK-NEXT: $v16 = VL1RE8_V killed $x11 :: (load (<vscale x 1 x s64>) from %stack.0) ; CHECK-NEXT: VS1R_V killed $v10, killed renamable $x10 diff --git a/llvm/test/CodeGen/RISCV/short-forward-branch-opt-min-max.ll b/llvm/test/CodeGen/RISCV/short-forward-branch-opt-min-max.ll new file mode 100644 index 0000000..05e06cea --- /dev/null +++ b/llvm/test/CodeGen/RISCV/short-forward-branch-opt-min-max.ll @@ -0,0 +1,703 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc < %s -mtriple=riscv32 -mattr=+zbb | FileCheck %s --check-prefixes=RV32I-ZBB +; RUN: llc < %s -mtriple=riscv64 -mattr=+zbb | FileCheck %s --check-prefixes=RV64I-ZBB +; RUN: llc < %s -mtriple=riscv32 -mattr=+zbb,+short-forward-branch-opt | \ +; RUN: FileCheck %s --check-prefixes=RV32I-SFB-ZBB +; RUN: llc < %s -mtriple=riscv64 -mattr=+zbb,+short-forward-branch-opt | \ +; RUN: FileCheck %s --check-prefixes=RV64I-SFB-ZBB +; RUN: llc < %s -mtriple=riscv32 -mattr=+zbb,+short-forward-branch-i-minmax | \ +; RUN: FileCheck %s --check-prefixes=RV32I-SFBIMinMax-ZBB +; RUN: llc < %s -mtriple=riscv64 -mattr=+zbb,+short-forward-branch-i-minmax | \ +; RUN: FileCheck %s --check-prefixes=RV64I-SFBIMinMax-ZBB + +define i32 @select_example_smax(i32 %a, i32 %b, i1 zeroext %x, i32 %y) { +; RV32I-ZBB-LABEL: select_example_smax: +; RV32I-ZBB: # %bb.0: # %entry +; RV32I-ZBB-NEXT: beqz a2, .LBB0_2 +; RV32I-ZBB-NEXT: # %bb.1: +; RV32I-ZBB-NEXT: max a1, a0, a3 +; RV32I-ZBB-NEXT: .LBB0_2: # %entry +; RV32I-ZBB-NEXT: mv a0, a1 +; RV32I-ZBB-NEXT: ret +; +; RV64I-ZBB-LABEL: select_example_smax: +; RV64I-ZBB: # %bb.0: # %entry +; RV64I-ZBB-NEXT: beqz a2, .LBB0_2 +; RV64I-ZBB-NEXT: # %bb.1: +; RV64I-ZBB-NEXT: sext.w a3, a3 +; RV64I-ZBB-NEXT: sext.w a0, a0 +; RV64I-ZBB-NEXT: max a1, a0, a3 +; RV64I-ZBB-NEXT: .LBB0_2: # %entry +; RV64I-ZBB-NEXT: mv a0, a1 +; RV64I-ZBB-NEXT: ret +; +; RV32I-SFB-ZBB-LABEL: select_example_smax: +; RV32I-SFB-ZBB: # %bb.0: # %entry +; RV32I-SFB-ZBB-NEXT: max a0, a0, a3 +; RV32I-SFB-ZBB-NEXT: bnez a2, .LBB0_2 +; RV32I-SFB-ZBB-NEXT: # %bb.1: # %entry +; RV32I-SFB-ZBB-NEXT: mv a0, a1 +; RV32I-SFB-ZBB-NEXT: .LBB0_2: # %entry +; RV32I-SFB-ZBB-NEXT: ret +; +; RV64I-SFB-ZBB-LABEL: select_example_smax: +; RV64I-SFB-ZBB: # %bb.0: # %entry +; RV64I-SFB-ZBB-NEXT: sext.w a3, a3 +; RV64I-SFB-ZBB-NEXT: sext.w a0, a0 +; RV64I-SFB-ZBB-NEXT: max a0, a0, a3 +; RV64I-SFB-ZBB-NEXT: bnez a2, .LBB0_2 +; RV64I-SFB-ZBB-NEXT: # %bb.1: # %entry +; RV64I-SFB-ZBB-NEXT: mv a0, a1 +; RV64I-SFB-ZBB-NEXT: .LBB0_2: # %entry +; RV64I-SFB-ZBB-NEXT: ret +; +; RV32I-SFBIMinMax-ZBB-LABEL: select_example_smax: +; RV32I-SFBIMinMax-ZBB: # %bb.0: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: beqz a2, .LBB0_2 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.1: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: max a1, a0, a3 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB0_2: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv a0, a1 +; RV32I-SFBIMinMax-ZBB-NEXT: ret +; +; RV64I-SFBIMinMax-ZBB-LABEL: select_example_smax: +; RV64I-SFBIMinMax-ZBB: # %bb.0: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: sext.w a3, a3 +; RV64I-SFBIMinMax-ZBB-NEXT: sext.w a0, a0 +; RV64I-SFBIMinMax-ZBB-NEXT: beqz a2, .LBB0_2 +; RV64I-SFBIMinMax-ZBB-NEXT: # %bb.1: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: max a1, a0, a3 +; RV64I-SFBIMinMax-ZBB-NEXT: .LBB0_2: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: mv a0, a1 +; RV64I-SFBIMinMax-ZBB-NEXT: ret +entry: + %res = call i32 @llvm.smax.i32(i32 %a, i32 %y) + %sel = select i1 %x, i32 %res, i32 %b + ret i32 %sel +} + +define i32 @select_example_smin(i32 %a, i32 %b, i1 zeroext %x, i32 %y) { +; RV32I-ZBB-LABEL: select_example_smin: +; RV32I-ZBB: # %bb.0: # %entry +; RV32I-ZBB-NEXT: beqz a2, .LBB1_2 +; RV32I-ZBB-NEXT: # %bb.1: +; RV32I-ZBB-NEXT: min a1, a0, a3 +; RV32I-ZBB-NEXT: .LBB1_2: # %entry +; RV32I-ZBB-NEXT: mv a0, a1 +; RV32I-ZBB-NEXT: ret +; +; RV64I-ZBB-LABEL: select_example_smin: +; RV64I-ZBB: # %bb.0: # %entry +; RV64I-ZBB-NEXT: beqz a2, .LBB1_2 +; RV64I-ZBB-NEXT: # %bb.1: +; RV64I-ZBB-NEXT: sext.w a3, a3 +; RV64I-ZBB-NEXT: sext.w a0, a0 +; RV64I-ZBB-NEXT: min a1, a0, a3 +; RV64I-ZBB-NEXT: .LBB1_2: # %entry +; RV64I-ZBB-NEXT: mv a0, a1 +; RV64I-ZBB-NEXT: ret +; +; RV32I-SFB-ZBB-LABEL: select_example_smin: +; RV32I-SFB-ZBB: # %bb.0: # %entry +; RV32I-SFB-ZBB-NEXT: min a0, a0, a3 +; RV32I-SFB-ZBB-NEXT: bnez a2, .LBB1_2 +; RV32I-SFB-ZBB-NEXT: # %bb.1: # %entry +; RV32I-SFB-ZBB-NEXT: mv a0, a1 +; RV32I-SFB-ZBB-NEXT: .LBB1_2: # %entry +; RV32I-SFB-ZBB-NEXT: ret +; +; RV64I-SFB-ZBB-LABEL: select_example_smin: +; RV64I-SFB-ZBB: # %bb.0: # %entry +; RV64I-SFB-ZBB-NEXT: sext.w a3, a3 +; RV64I-SFB-ZBB-NEXT: sext.w a0, a0 +; RV64I-SFB-ZBB-NEXT: min a0, a0, a3 +; RV64I-SFB-ZBB-NEXT: bnez a2, .LBB1_2 +; RV64I-SFB-ZBB-NEXT: # %bb.1: # %entry +; RV64I-SFB-ZBB-NEXT: mv a0, a1 +; RV64I-SFB-ZBB-NEXT: .LBB1_2: # %entry +; RV64I-SFB-ZBB-NEXT: ret +; +; RV32I-SFBIMinMax-ZBB-LABEL: select_example_smin: +; RV32I-SFBIMinMax-ZBB: # %bb.0: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: beqz a2, .LBB1_2 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.1: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: min a1, a0, a3 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB1_2: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv a0, a1 +; RV32I-SFBIMinMax-ZBB-NEXT: ret +; +; RV64I-SFBIMinMax-ZBB-LABEL: select_example_smin: +; RV64I-SFBIMinMax-ZBB: # %bb.0: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: sext.w a3, a3 +; RV64I-SFBIMinMax-ZBB-NEXT: sext.w a0, a0 +; RV64I-SFBIMinMax-ZBB-NEXT: beqz a2, .LBB1_2 +; RV64I-SFBIMinMax-ZBB-NEXT: # %bb.1: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: min a1, a0, a3 +; RV64I-SFBIMinMax-ZBB-NEXT: .LBB1_2: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: mv a0, a1 +; RV64I-SFBIMinMax-ZBB-NEXT: ret +entry: + %res = call i32 @llvm.smin.i32(i32 %a, i32 %y) + %sel = select i1 %x, i32 %res, i32 %b + ret i32 %sel +} + +define i32 @select_example_umax(i32 %a, i32 %b, i1 zeroext %x, i32 %y) { +; RV32I-ZBB-LABEL: select_example_umax: +; RV32I-ZBB: # %bb.0: # %entry +; RV32I-ZBB-NEXT: beqz a2, .LBB2_2 +; RV32I-ZBB-NEXT: # %bb.1: +; RV32I-ZBB-NEXT: maxu a1, a0, a3 +; RV32I-ZBB-NEXT: .LBB2_2: # %entry +; RV32I-ZBB-NEXT: mv a0, a1 +; RV32I-ZBB-NEXT: ret +; +; RV64I-ZBB-LABEL: select_example_umax: +; RV64I-ZBB: # %bb.0: # %entry +; RV64I-ZBB-NEXT: beqz a2, .LBB2_2 +; RV64I-ZBB-NEXT: # %bb.1: +; RV64I-ZBB-NEXT: sext.w a3, a3 +; RV64I-ZBB-NEXT: sext.w a0, a0 +; RV64I-ZBB-NEXT: maxu a1, a0, a3 +; RV64I-ZBB-NEXT: .LBB2_2: # %entry +; RV64I-ZBB-NEXT: mv a0, a1 +; RV64I-ZBB-NEXT: ret +; +; RV32I-SFB-ZBB-LABEL: select_example_umax: +; RV32I-SFB-ZBB: # %bb.0: # %entry +; RV32I-SFB-ZBB-NEXT: maxu a0, a0, a3 +; RV32I-SFB-ZBB-NEXT: bnez a2, .LBB2_2 +; RV32I-SFB-ZBB-NEXT: # %bb.1: # %entry +; RV32I-SFB-ZBB-NEXT: mv a0, a1 +; RV32I-SFB-ZBB-NEXT: .LBB2_2: # %entry +; RV32I-SFB-ZBB-NEXT: ret +; +; RV64I-SFB-ZBB-LABEL: select_example_umax: +; RV64I-SFB-ZBB: # %bb.0: # %entry +; RV64I-SFB-ZBB-NEXT: sext.w a3, a3 +; RV64I-SFB-ZBB-NEXT: sext.w a0, a0 +; RV64I-SFB-ZBB-NEXT: maxu a0, a0, a3 +; RV64I-SFB-ZBB-NEXT: bnez a2, .LBB2_2 +; RV64I-SFB-ZBB-NEXT: # %bb.1: # %entry +; RV64I-SFB-ZBB-NEXT: mv a0, a1 +; RV64I-SFB-ZBB-NEXT: .LBB2_2: # %entry +; RV64I-SFB-ZBB-NEXT: ret +; +; RV32I-SFBIMinMax-ZBB-LABEL: select_example_umax: +; RV32I-SFBIMinMax-ZBB: # %bb.0: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: beqz a2, .LBB2_2 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.1: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: maxu a1, a0, a3 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB2_2: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv a0, a1 +; RV32I-SFBIMinMax-ZBB-NEXT: ret +; +; RV64I-SFBIMinMax-ZBB-LABEL: select_example_umax: +; RV64I-SFBIMinMax-ZBB: # %bb.0: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: sext.w a3, a3 +; RV64I-SFBIMinMax-ZBB-NEXT: sext.w a0, a0 +; RV64I-SFBIMinMax-ZBB-NEXT: beqz a2, .LBB2_2 +; RV64I-SFBIMinMax-ZBB-NEXT: # %bb.1: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: maxu a1, a0, a3 +; RV64I-SFBIMinMax-ZBB-NEXT: .LBB2_2: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: mv a0, a1 +; RV64I-SFBIMinMax-ZBB-NEXT: ret +entry: + %res = call i32 @llvm.umax.i32(i32 %a, i32 %y) + %sel = select i1 %x, i32 %res, i32 %b + ret i32 %sel +} + +define i32 @select_example_umin(i32 %a, i32 %b, i1 zeroext %x, i32 %y) { +; RV32I-ZBB-LABEL: select_example_umin: +; RV32I-ZBB: # %bb.0: # %entry +; RV32I-ZBB-NEXT: beqz a2, .LBB3_2 +; RV32I-ZBB-NEXT: # %bb.1: +; RV32I-ZBB-NEXT: minu a1, a0, a3 +; RV32I-ZBB-NEXT: .LBB3_2: # %entry +; RV32I-ZBB-NEXT: mv a0, a1 +; RV32I-ZBB-NEXT: ret +; +; RV64I-ZBB-LABEL: select_example_umin: +; RV64I-ZBB: # %bb.0: # %entry +; RV64I-ZBB-NEXT: beqz a2, .LBB3_2 +; RV64I-ZBB-NEXT: # %bb.1: +; RV64I-ZBB-NEXT: sext.w a3, a3 +; RV64I-ZBB-NEXT: sext.w a0, a0 +; RV64I-ZBB-NEXT: minu a1, a0, a3 +; RV64I-ZBB-NEXT: .LBB3_2: # %entry +; RV64I-ZBB-NEXT: mv a0, a1 +; RV64I-ZBB-NEXT: ret +; +; RV32I-SFB-ZBB-LABEL: select_example_umin: +; RV32I-SFB-ZBB: # %bb.0: # %entry +; RV32I-SFB-ZBB-NEXT: minu a0, a0, a3 +; RV32I-SFB-ZBB-NEXT: bnez a2, .LBB3_2 +; RV32I-SFB-ZBB-NEXT: # %bb.1: # %entry +; RV32I-SFB-ZBB-NEXT: mv a0, a1 +; RV32I-SFB-ZBB-NEXT: .LBB3_2: # %entry +; RV32I-SFB-ZBB-NEXT: ret +; +; RV64I-SFB-ZBB-LABEL: select_example_umin: +; RV64I-SFB-ZBB: # %bb.0: # %entry +; RV64I-SFB-ZBB-NEXT: sext.w a3, a3 +; RV64I-SFB-ZBB-NEXT: sext.w a0, a0 +; RV64I-SFB-ZBB-NEXT: minu a0, a0, a3 +; RV64I-SFB-ZBB-NEXT: bnez a2, .LBB3_2 +; RV64I-SFB-ZBB-NEXT: # %bb.1: # %entry +; RV64I-SFB-ZBB-NEXT: mv a0, a1 +; RV64I-SFB-ZBB-NEXT: .LBB3_2: # %entry +; RV64I-SFB-ZBB-NEXT: ret +; +; RV32I-SFBIMinMax-ZBB-LABEL: select_example_umin: +; RV32I-SFBIMinMax-ZBB: # %bb.0: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: beqz a2, .LBB3_2 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.1: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: minu a1, a0, a3 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB3_2: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv a0, a1 +; RV32I-SFBIMinMax-ZBB-NEXT: ret +; +; RV64I-SFBIMinMax-ZBB-LABEL: select_example_umin: +; RV64I-SFBIMinMax-ZBB: # %bb.0: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: sext.w a3, a3 +; RV64I-SFBIMinMax-ZBB-NEXT: sext.w a0, a0 +; RV64I-SFBIMinMax-ZBB-NEXT: beqz a2, .LBB3_2 +; RV64I-SFBIMinMax-ZBB-NEXT: # %bb.1: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: minu a1, a0, a3 +; RV64I-SFBIMinMax-ZBB-NEXT: .LBB3_2: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: mv a0, a1 +; RV64I-SFBIMinMax-ZBB-NEXT: ret +entry: + %res = call i32 @llvm.umin.i32(i32 %a, i32 %y) + %sel = select i1 %x, i32 %res, i32 %b + ret i32 %sel +} + +define i64 @select_example_smax_1(i64 %a, i64 %b, i1 zeroext %x, i64 %y) { +; RV32I-ZBB-LABEL: select_example_smax_1: +; RV32I-ZBB: # %bb.0: # %entry +; RV32I-ZBB-NEXT: beq a1, a6, .LBB4_2 +; RV32I-ZBB-NEXT: # %bb.1: # %entry +; RV32I-ZBB-NEXT: slt a7, a6, a1 +; RV32I-ZBB-NEXT: beqz a7, .LBB4_3 +; RV32I-ZBB-NEXT: j .LBB4_4 +; RV32I-ZBB-NEXT: .LBB4_2: +; RV32I-ZBB-NEXT: sltu a7, a5, a0 +; RV32I-ZBB-NEXT: bnez a7, .LBB4_4 +; RV32I-ZBB-NEXT: .LBB4_3: # %entry +; RV32I-ZBB-NEXT: mv a1, a6 +; RV32I-ZBB-NEXT: mv a0, a5 +; RV32I-ZBB-NEXT: .LBB4_4: # %entry +; RV32I-ZBB-NEXT: beqz a4, .LBB4_6 +; RV32I-ZBB-NEXT: # %bb.5: # %entry +; RV32I-ZBB-NEXT: ret +; RV32I-ZBB-NEXT: .LBB4_6: # %entry +; RV32I-ZBB-NEXT: mv a0, a2 +; RV32I-ZBB-NEXT: mv a1, a3 +; RV32I-ZBB-NEXT: ret +; +; RV64I-ZBB-LABEL: select_example_smax_1: +; RV64I-ZBB: # %bb.0: # %entry +; RV64I-ZBB-NEXT: beqz a2, .LBB4_2 +; RV64I-ZBB-NEXT: # %bb.1: +; RV64I-ZBB-NEXT: max a1, a0, a3 +; RV64I-ZBB-NEXT: .LBB4_2: # %entry +; RV64I-ZBB-NEXT: mv a0, a1 +; RV64I-ZBB-NEXT: ret +; +; RV32I-SFB-ZBB-LABEL: select_example_smax_1: +; RV32I-SFB-ZBB: # %bb.0: # %entry +; RV32I-SFB-ZBB-NEXT: sltu a7, a5, a0 +; RV32I-SFB-ZBB-NEXT: slt t0, a6, a1 +; RV32I-SFB-ZBB-NEXT: bne a1, a6, .LBB4_2 +; RV32I-SFB-ZBB-NEXT: # %bb.1: # %entry +; RV32I-SFB-ZBB-NEXT: mv t0, a7 +; RV32I-SFB-ZBB-NEXT: .LBB4_2: # %entry +; RV32I-SFB-ZBB-NEXT: bnez t0, .LBB4_4 +; RV32I-SFB-ZBB-NEXT: # %bb.3: # %entry +; RV32I-SFB-ZBB-NEXT: mv a1, a6 +; RV32I-SFB-ZBB-NEXT: .LBB4_4: # %entry +; RV32I-SFB-ZBB-NEXT: bnez t0, .LBB4_6 +; RV32I-SFB-ZBB-NEXT: # %bb.5: # %entry +; RV32I-SFB-ZBB-NEXT: mv a0, a5 +; RV32I-SFB-ZBB-NEXT: .LBB4_6: # %entry +; RV32I-SFB-ZBB-NEXT: bnez a4, .LBB4_8 +; RV32I-SFB-ZBB-NEXT: # %bb.7: # %entry +; RV32I-SFB-ZBB-NEXT: mv a0, a2 +; RV32I-SFB-ZBB-NEXT: .LBB4_8: # %entry +; RV32I-SFB-ZBB-NEXT: bnez a4, .LBB4_10 +; RV32I-SFB-ZBB-NEXT: # %bb.9: # %entry +; RV32I-SFB-ZBB-NEXT: mv a1, a3 +; RV32I-SFB-ZBB-NEXT: .LBB4_10: # %entry +; RV32I-SFB-ZBB-NEXT: ret +; +; RV64I-SFB-ZBB-LABEL: select_example_smax_1: +; RV64I-SFB-ZBB: # %bb.0: # %entry +; RV64I-SFB-ZBB-NEXT: max a0, a0, a3 +; RV64I-SFB-ZBB-NEXT: bnez a2, .LBB4_2 +; RV64I-SFB-ZBB-NEXT: # %bb.1: # %entry +; RV64I-SFB-ZBB-NEXT: mv a0, a1 +; RV64I-SFB-ZBB-NEXT: .LBB4_2: # %entry +; RV64I-SFB-ZBB-NEXT: ret +; +; RV32I-SFBIMinMax-ZBB-LABEL: select_example_smax_1: +; RV32I-SFBIMinMax-ZBB: # %bb.0: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: sltu a7, a5, a0 +; RV32I-SFBIMinMax-ZBB-NEXT: slt t0, a6, a1 +; RV32I-SFBIMinMax-ZBB-NEXT: bne a1, a6, .LBB4_2 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.1: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv t0, a7 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB4_2: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: bnez t0, .LBB4_4 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.3: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv a1, a6 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB4_4: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: bnez t0, .LBB4_6 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.5: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv a0, a5 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB4_6: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: bnez a4, .LBB4_8 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.7: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv a0, a2 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB4_8: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: bnez a4, .LBB4_10 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.9: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv a1, a3 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB4_10: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: ret +; +; RV64I-SFBIMinMax-ZBB-LABEL: select_example_smax_1: +; RV64I-SFBIMinMax-ZBB: # %bb.0: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: beqz a2, .LBB4_2 +; RV64I-SFBIMinMax-ZBB-NEXT: # %bb.1: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: max a1, a0, a3 +; RV64I-SFBIMinMax-ZBB-NEXT: .LBB4_2: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: mv a0, a1 +; RV64I-SFBIMinMax-ZBB-NEXT: ret +entry: + %res = call i64 @llvm.smax.i64(i64 %a, i64 %y) + %sel = select i1 %x, i64 %res, i64 %b + ret i64 %sel +} + +define i64 @select_example_smin_1(i64 %a, i64 %b, i1 zeroext %x, i64 %y) { +; RV32I-ZBB-LABEL: select_example_smin_1: +; RV32I-ZBB: # %bb.0: # %entry +; RV32I-ZBB-NEXT: beq a1, a6, .LBB5_2 +; RV32I-ZBB-NEXT: # %bb.1: # %entry +; RV32I-ZBB-NEXT: slt a7, a1, a6 +; RV32I-ZBB-NEXT: beqz a7, .LBB5_3 +; RV32I-ZBB-NEXT: j .LBB5_4 +; RV32I-ZBB-NEXT: .LBB5_2: +; RV32I-ZBB-NEXT: sltu a7, a0, a5 +; RV32I-ZBB-NEXT: bnez a7, .LBB5_4 +; RV32I-ZBB-NEXT: .LBB5_3: # %entry +; RV32I-ZBB-NEXT: mv a1, a6 +; RV32I-ZBB-NEXT: mv a0, a5 +; RV32I-ZBB-NEXT: .LBB5_4: # %entry +; RV32I-ZBB-NEXT: beqz a4, .LBB5_6 +; RV32I-ZBB-NEXT: # %bb.5: # %entry +; RV32I-ZBB-NEXT: ret +; RV32I-ZBB-NEXT: .LBB5_6: # %entry +; RV32I-ZBB-NEXT: mv a0, a2 +; RV32I-ZBB-NEXT: mv a1, a3 +; RV32I-ZBB-NEXT: ret +; +; RV64I-ZBB-LABEL: select_example_smin_1: +; RV64I-ZBB: # %bb.0: # %entry +; RV64I-ZBB-NEXT: beqz a2, .LBB5_2 +; RV64I-ZBB-NEXT: # %bb.1: +; RV64I-ZBB-NEXT: min a1, a0, a3 +; RV64I-ZBB-NEXT: .LBB5_2: # %entry +; RV64I-ZBB-NEXT: mv a0, a1 +; RV64I-ZBB-NEXT: ret +; +; RV32I-SFB-ZBB-LABEL: select_example_smin_1: +; RV32I-SFB-ZBB: # %bb.0: # %entry +; RV32I-SFB-ZBB-NEXT: sltu a7, a0, a5 +; RV32I-SFB-ZBB-NEXT: slt t0, a1, a6 +; RV32I-SFB-ZBB-NEXT: bne a1, a6, .LBB5_2 +; RV32I-SFB-ZBB-NEXT: # %bb.1: # %entry +; RV32I-SFB-ZBB-NEXT: mv t0, a7 +; RV32I-SFB-ZBB-NEXT: .LBB5_2: # %entry +; RV32I-SFB-ZBB-NEXT: bnez t0, .LBB5_4 +; RV32I-SFB-ZBB-NEXT: # %bb.3: # %entry +; RV32I-SFB-ZBB-NEXT: mv a1, a6 +; RV32I-SFB-ZBB-NEXT: .LBB5_4: # %entry +; RV32I-SFB-ZBB-NEXT: bnez t0, .LBB5_6 +; RV32I-SFB-ZBB-NEXT: # %bb.5: # %entry +; RV32I-SFB-ZBB-NEXT: mv a0, a5 +; RV32I-SFB-ZBB-NEXT: .LBB5_6: # %entry +; RV32I-SFB-ZBB-NEXT: bnez a4, .LBB5_8 +; RV32I-SFB-ZBB-NEXT: # %bb.7: # %entry +; RV32I-SFB-ZBB-NEXT: mv a0, a2 +; RV32I-SFB-ZBB-NEXT: .LBB5_8: # %entry +; RV32I-SFB-ZBB-NEXT: bnez a4, .LBB5_10 +; RV32I-SFB-ZBB-NEXT: # %bb.9: # %entry +; RV32I-SFB-ZBB-NEXT: mv a1, a3 +; RV32I-SFB-ZBB-NEXT: .LBB5_10: # %entry +; RV32I-SFB-ZBB-NEXT: ret +; +; RV64I-SFB-ZBB-LABEL: select_example_smin_1: +; RV64I-SFB-ZBB: # %bb.0: # %entry +; RV64I-SFB-ZBB-NEXT: min a0, a0, a3 +; RV64I-SFB-ZBB-NEXT: bnez a2, .LBB5_2 +; RV64I-SFB-ZBB-NEXT: # %bb.1: # %entry +; RV64I-SFB-ZBB-NEXT: mv a0, a1 +; RV64I-SFB-ZBB-NEXT: .LBB5_2: # %entry +; RV64I-SFB-ZBB-NEXT: ret +; +; RV32I-SFBIMinMax-ZBB-LABEL: select_example_smin_1: +; RV32I-SFBIMinMax-ZBB: # %bb.0: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: sltu a7, a0, a5 +; RV32I-SFBIMinMax-ZBB-NEXT: slt t0, a1, a6 +; RV32I-SFBIMinMax-ZBB-NEXT: bne a1, a6, .LBB5_2 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.1: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv t0, a7 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB5_2: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: bnez t0, .LBB5_4 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.3: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv a1, a6 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB5_4: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: bnez t0, .LBB5_6 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.5: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv a0, a5 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB5_6: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: bnez a4, .LBB5_8 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.7: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv a0, a2 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB5_8: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: bnez a4, .LBB5_10 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.9: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv a1, a3 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB5_10: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: ret +; +; RV64I-SFBIMinMax-ZBB-LABEL: select_example_smin_1: +; RV64I-SFBIMinMax-ZBB: # %bb.0: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: beqz a2, .LBB5_2 +; RV64I-SFBIMinMax-ZBB-NEXT: # %bb.1: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: min a1, a0, a3 +; RV64I-SFBIMinMax-ZBB-NEXT: .LBB5_2: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: mv a0, a1 +; RV64I-SFBIMinMax-ZBB-NEXT: ret +entry: + %res = call i64 @llvm.smin.i64(i64 %a, i64 %y) + %sel = select i1 %x, i64 %res, i64 %b + ret i64 %sel +} + +define i64 @select_example_umax_1(i64 %a, i64 %b, i1 zeroext %x, i64 %y) { +; RV32I-ZBB-LABEL: select_example_umax_1: +; RV32I-ZBB: # %bb.0: # %entry +; RV32I-ZBB-NEXT: beq a1, a6, .LBB6_2 +; RV32I-ZBB-NEXT: # %bb.1: # %entry +; RV32I-ZBB-NEXT: sltu a7, a6, a1 +; RV32I-ZBB-NEXT: beqz a7, .LBB6_3 +; RV32I-ZBB-NEXT: j .LBB6_4 +; RV32I-ZBB-NEXT: .LBB6_2: +; RV32I-ZBB-NEXT: sltu a7, a5, a0 +; RV32I-ZBB-NEXT: bnez a7, .LBB6_4 +; RV32I-ZBB-NEXT: .LBB6_3: # %entry +; RV32I-ZBB-NEXT: mv a1, a6 +; RV32I-ZBB-NEXT: mv a0, a5 +; RV32I-ZBB-NEXT: .LBB6_4: # %entry +; RV32I-ZBB-NEXT: beqz a4, .LBB6_6 +; RV32I-ZBB-NEXT: # %bb.5: # %entry +; RV32I-ZBB-NEXT: ret +; RV32I-ZBB-NEXT: .LBB6_6: # %entry +; RV32I-ZBB-NEXT: mv a0, a2 +; RV32I-ZBB-NEXT: mv a1, a3 +; RV32I-ZBB-NEXT: ret +; +; RV64I-ZBB-LABEL: select_example_umax_1: +; RV64I-ZBB: # %bb.0: # %entry +; RV64I-ZBB-NEXT: beqz a2, .LBB6_2 +; RV64I-ZBB-NEXT: # %bb.1: +; RV64I-ZBB-NEXT: maxu a1, a0, a3 +; RV64I-ZBB-NEXT: .LBB6_2: # %entry +; RV64I-ZBB-NEXT: mv a0, a1 +; RV64I-ZBB-NEXT: ret +; +; RV32I-SFB-ZBB-LABEL: select_example_umax_1: +; RV32I-SFB-ZBB: # %bb.0: # %entry +; RV32I-SFB-ZBB-NEXT: sltu a7, a5, a0 +; RV32I-SFB-ZBB-NEXT: sltu t0, a6, a1 +; RV32I-SFB-ZBB-NEXT: bne a1, a6, .LBB6_2 +; RV32I-SFB-ZBB-NEXT: # %bb.1: # %entry +; RV32I-SFB-ZBB-NEXT: mv t0, a7 +; RV32I-SFB-ZBB-NEXT: .LBB6_2: # %entry +; RV32I-SFB-ZBB-NEXT: bnez t0, .LBB6_4 +; RV32I-SFB-ZBB-NEXT: # %bb.3: # %entry +; RV32I-SFB-ZBB-NEXT: mv a1, a6 +; RV32I-SFB-ZBB-NEXT: .LBB6_4: # %entry +; RV32I-SFB-ZBB-NEXT: bnez t0, .LBB6_6 +; RV32I-SFB-ZBB-NEXT: # %bb.5: # %entry +; RV32I-SFB-ZBB-NEXT: mv a0, a5 +; RV32I-SFB-ZBB-NEXT: .LBB6_6: # %entry +; RV32I-SFB-ZBB-NEXT: bnez a4, .LBB6_8 +; RV32I-SFB-ZBB-NEXT: # %bb.7: # %entry +; RV32I-SFB-ZBB-NEXT: mv a0, a2 +; RV32I-SFB-ZBB-NEXT: .LBB6_8: # %entry +; RV32I-SFB-ZBB-NEXT: bnez a4, .LBB6_10 +; RV32I-SFB-ZBB-NEXT: # %bb.9: # %entry +; RV32I-SFB-ZBB-NEXT: mv a1, a3 +; RV32I-SFB-ZBB-NEXT: .LBB6_10: # %entry +; RV32I-SFB-ZBB-NEXT: ret +; +; RV64I-SFB-ZBB-LABEL: select_example_umax_1: +; RV64I-SFB-ZBB: # %bb.0: # %entry +; RV64I-SFB-ZBB-NEXT: maxu a0, a0, a3 +; RV64I-SFB-ZBB-NEXT: bnez a2, .LBB6_2 +; RV64I-SFB-ZBB-NEXT: # %bb.1: # %entry +; RV64I-SFB-ZBB-NEXT: mv a0, a1 +; RV64I-SFB-ZBB-NEXT: .LBB6_2: # %entry +; RV64I-SFB-ZBB-NEXT: ret +; +; RV32I-SFBIMinMax-ZBB-LABEL: select_example_umax_1: +; RV32I-SFBIMinMax-ZBB: # %bb.0: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: sltu a7, a5, a0 +; RV32I-SFBIMinMax-ZBB-NEXT: sltu t0, a6, a1 +; RV32I-SFBIMinMax-ZBB-NEXT: bne a1, a6, .LBB6_2 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.1: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv t0, a7 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB6_2: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: bnez t0, .LBB6_4 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.3: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv a1, a6 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB6_4: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: bnez t0, .LBB6_6 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.5: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv a0, a5 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB6_6: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: bnez a4, .LBB6_8 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.7: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv a0, a2 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB6_8: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: bnez a4, .LBB6_10 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.9: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv a1, a3 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB6_10: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: ret +; +; RV64I-SFBIMinMax-ZBB-LABEL: select_example_umax_1: +; RV64I-SFBIMinMax-ZBB: # %bb.0: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: beqz a2, .LBB6_2 +; RV64I-SFBIMinMax-ZBB-NEXT: # %bb.1: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: maxu a1, a0, a3 +; RV64I-SFBIMinMax-ZBB-NEXT: .LBB6_2: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: mv a0, a1 +; RV64I-SFBIMinMax-ZBB-NEXT: ret +entry: + %res = call i64 @llvm.umax.i64(i64 %a, i64 %y) + %sel = select i1 %x, i64 %res, i64 %b + ret i64 %sel +} + +define i64 @select_example_umin_1(i64 %a, i64 %b, i1 zeroext %x, i64 %y) { +; RV32I-ZBB-LABEL: select_example_umin_1: +; RV32I-ZBB: # %bb.0: # %entry +; RV32I-ZBB-NEXT: beq a1, a6, .LBB7_2 +; RV32I-ZBB-NEXT: # %bb.1: # %entry +; RV32I-ZBB-NEXT: sltu a7, a1, a6 +; RV32I-ZBB-NEXT: beqz a7, .LBB7_3 +; RV32I-ZBB-NEXT: j .LBB7_4 +; RV32I-ZBB-NEXT: .LBB7_2: +; RV32I-ZBB-NEXT: sltu a7, a0, a5 +; RV32I-ZBB-NEXT: bnez a7, .LBB7_4 +; RV32I-ZBB-NEXT: .LBB7_3: # %entry +; RV32I-ZBB-NEXT: mv a1, a6 +; RV32I-ZBB-NEXT: mv a0, a5 +; RV32I-ZBB-NEXT: .LBB7_4: # %entry +; RV32I-ZBB-NEXT: beqz a4, .LBB7_6 +; RV32I-ZBB-NEXT: # %bb.5: # %entry +; RV32I-ZBB-NEXT: ret +; RV32I-ZBB-NEXT: .LBB7_6: # %entry +; RV32I-ZBB-NEXT: mv a0, a2 +; RV32I-ZBB-NEXT: mv a1, a3 +; RV32I-ZBB-NEXT: ret +; +; RV64I-ZBB-LABEL: select_example_umin_1: +; RV64I-ZBB: # %bb.0: # %entry +; RV64I-ZBB-NEXT: beqz a2, .LBB7_2 +; RV64I-ZBB-NEXT: # %bb.1: +; RV64I-ZBB-NEXT: minu a1, a0, a3 +; RV64I-ZBB-NEXT: .LBB7_2: # %entry +; RV64I-ZBB-NEXT: mv a0, a1 +; RV64I-ZBB-NEXT: ret +; +; RV32I-SFB-ZBB-LABEL: select_example_umin_1: +; RV32I-SFB-ZBB: # %bb.0: # %entry +; RV32I-SFB-ZBB-NEXT: sltu a7, a0, a5 +; RV32I-SFB-ZBB-NEXT: sltu t0, a1, a6 +; RV32I-SFB-ZBB-NEXT: bne a1, a6, .LBB7_2 +; RV32I-SFB-ZBB-NEXT: # %bb.1: # %entry +; RV32I-SFB-ZBB-NEXT: mv t0, a7 +; RV32I-SFB-ZBB-NEXT: .LBB7_2: # %entry +; RV32I-SFB-ZBB-NEXT: bnez t0, .LBB7_4 +; RV32I-SFB-ZBB-NEXT: # %bb.3: # %entry +; RV32I-SFB-ZBB-NEXT: mv a1, a6 +; RV32I-SFB-ZBB-NEXT: .LBB7_4: # %entry +; RV32I-SFB-ZBB-NEXT: bnez t0, .LBB7_6 +; RV32I-SFB-ZBB-NEXT: # %bb.5: # %entry +; RV32I-SFB-ZBB-NEXT: mv a0, a5 +; RV32I-SFB-ZBB-NEXT: .LBB7_6: # %entry +; RV32I-SFB-ZBB-NEXT: bnez a4, .LBB7_8 +; RV32I-SFB-ZBB-NEXT: # %bb.7: # %entry +; RV32I-SFB-ZBB-NEXT: mv a0, a2 +; RV32I-SFB-ZBB-NEXT: .LBB7_8: # %entry +; RV32I-SFB-ZBB-NEXT: bnez a4, .LBB7_10 +; RV32I-SFB-ZBB-NEXT: # %bb.9: # %entry +; RV32I-SFB-ZBB-NEXT: mv a1, a3 +; RV32I-SFB-ZBB-NEXT: .LBB7_10: # %entry +; RV32I-SFB-ZBB-NEXT: ret +; +; RV64I-SFB-ZBB-LABEL: select_example_umin_1: +; RV64I-SFB-ZBB: # %bb.0: # %entry +; RV64I-SFB-ZBB-NEXT: minu a0, a0, a3 +; RV64I-SFB-ZBB-NEXT: bnez a2, .LBB7_2 +; RV64I-SFB-ZBB-NEXT: # %bb.1: # %entry +; RV64I-SFB-ZBB-NEXT: mv a0, a1 +; RV64I-SFB-ZBB-NEXT: .LBB7_2: # %entry +; RV64I-SFB-ZBB-NEXT: ret +; +; RV32I-SFBIMinMax-ZBB-LABEL: select_example_umin_1: +; RV32I-SFBIMinMax-ZBB: # %bb.0: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: sltu a7, a0, a5 +; RV32I-SFBIMinMax-ZBB-NEXT: sltu t0, a1, a6 +; RV32I-SFBIMinMax-ZBB-NEXT: bne a1, a6, .LBB7_2 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.1: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv t0, a7 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB7_2: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: bnez t0, .LBB7_4 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.3: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv a1, a6 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB7_4: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: bnez t0, .LBB7_6 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.5: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv a0, a5 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB7_6: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: bnez a4, .LBB7_8 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.7: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv a0, a2 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB7_8: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: bnez a4, .LBB7_10 +; RV32I-SFBIMinMax-ZBB-NEXT: # %bb.9: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: mv a1, a3 +; RV32I-SFBIMinMax-ZBB-NEXT: .LBB7_10: # %entry +; RV32I-SFBIMinMax-ZBB-NEXT: ret +; +; RV64I-SFBIMinMax-ZBB-LABEL: select_example_umin_1: +; RV64I-SFBIMinMax-ZBB: # %bb.0: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: beqz a2, .LBB7_2 +; RV64I-SFBIMinMax-ZBB-NEXT: # %bb.1: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: minu a1, a0, a3 +; RV64I-SFBIMinMax-ZBB-NEXT: .LBB7_2: # %entry +; RV64I-SFBIMinMax-ZBB-NEXT: mv a0, a1 +; RV64I-SFBIMinMax-ZBB-NEXT: ret +entry: + %res = call i64 @llvm.umin.i64(i64 %a, i64 %y) + %sel = select i1 %x, i64 %res, i64 %b + ret i64 %sel +} diff --git a/llvm/test/CodeGen/SPIRV/hlsl-resources/issue-146942-ptr-cast.ll b/llvm/test/CodeGen/SPIRV/hlsl-resources/issue-146942-ptr-cast.ll index ed67344..4817e74 100644 --- a/llvm/test/CodeGen/SPIRV/hlsl-resources/issue-146942-ptr-cast.ll +++ b/llvm/test/CodeGen/SPIRV/hlsl-resources/issue-146942-ptr-cast.ll @@ -16,7 +16,6 @@ define void @case1() local_unnamed_addr { ; CHECK: %[[#BUFFER_LOAD:]] = OpLoad %[[#FLOAT4]] %{{[0-9]+}} Aligned 16 ; CHECK: %[[#CAST_LOAD:]] = OpBitcast %[[#INT4]] %[[#BUFFER_LOAD]] - ; CHECK: %[[#VEC_SHUFFLE:]] = OpVectorShuffle %[[#INT4]] %[[#CAST_LOAD]] %[[#CAST_LOAD]] 0 1 2 3 %1 = tail call target("spirv.VulkanBuffer", [0 x <4 x float>], 12, 0) @llvm.spv.resource.handlefrombinding.tspirv.VulkanBuffer_a0v4f32_12_0t(i32 0, i32 2, i32 1, i32 0, ptr nonnull @.str) %2 = tail call target("spirv.VulkanBuffer", [0 x <4 x i32>], 12, 1) @llvm.spv.resource.handlefrombinding.tspirv.VulkanBuffer_a0v4i32_12_1t(i32 0, i32 5, i32 1, i32 0, ptr nonnull @.str.2) %3 = tail call noundef align 16 dereferenceable(16) ptr addrspace(11) @llvm.spv.resource.getpointer.p11.tspirv.VulkanBuffer_a0v4f32_12_0t(target("spirv.VulkanBuffer", [0 x <4 x float>], 12, 0) %1, i32 0) @@ -29,8 +28,7 @@ define void @case1() local_unnamed_addr { define void @case2() local_unnamed_addr { ; CHECK: %[[#BUFFER_LOAD:]] = OpLoad %[[#FLOAT4]] %{{[0-9]+}} Aligned 16 ; CHECK: %[[#CAST_LOAD:]] = OpBitcast %[[#INT4]] %[[#BUFFER_LOAD]] - ; CHECK: %[[#VEC_SHUFFLE:]] = OpVectorShuffle %[[#INT4]] %[[#CAST_LOAD]] %[[#CAST_LOAD]] 0 1 2 3 - ; CHECK: %[[#VEC_TRUNCATE:]] = OpVectorShuffle %[[#INT3]] %[[#VEC_SHUFFLE]] %[[#UNDEF_INT4]] 0 1 2 + ; CHECK: %[[#VEC_TRUNCATE:]] = OpVectorShuffle %[[#INT3]] %[[#CAST_LOAD]] %[[#UNDEF_INT4]] 0 1 2 %1 = tail call target("spirv.VulkanBuffer", [0 x <4 x float>], 12, 0) @llvm.spv.resource.handlefrombinding.tspirv.VulkanBuffer_a0v4f32_12_0t(i32 0, i32 2, i32 1, i32 0, ptr nonnull @.str) %2 = tail call target("spirv.VulkanBuffer", [0 x <3 x i32>], 12, 1) @llvm.spv.resource.handlefrombinding.tspirv.VulkanBuffer_a0v3i32_12_1t(i32 0, i32 5, i32 1, i32 0, ptr nonnull @.str.3) %3 = tail call noundef align 16 dereferenceable(16) ptr addrspace(11) @llvm.spv.resource.getpointer.p11.tspirv.VulkanBuffer_a0v4f32_12_0t(target("spirv.VulkanBuffer", [0 x <4 x float>], 12, 0) %1, i32 0) diff --git a/llvm/test/CodeGen/SPIRV/pointers/ptrcast-bitcast.ll b/llvm/test/CodeGen/SPIRV/pointers/ptrcast-bitcast.ll index 8491328..a1ec2cd 100644 --- a/llvm/test/CodeGen/SPIRV/pointers/ptrcast-bitcast.ll +++ b/llvm/test/CodeGen/SPIRV/pointers/ptrcast-bitcast.ll @@ -26,3 +26,25 @@ entry: store <4 x i32> %6, ptr addrspace(11) %7, align 16 ret void } + +; This tests a load from a pointer that has been bitcast between vector types +; which share the same total bit-width but have different numbers of elements. +; Tests that legalize-pointer-casts works correctly by moving the bitcast to +; the element that was loaded. + +define void @main2() local_unnamed_addr #0 { +entry: +; CHECK: %[[LOAD:[0-9]+]] = OpLoad %[[#v2_double]] {{.*}} +; CHECK: %[[BITCAST1:[0-9]+]] = OpBitcast %[[#v4_uint]] %[[LOAD]] +; CHECK: %[[BITCAST2:[0-9]+]] = OpBitcast %[[#v2_double]] %[[BITCAST1]] +; CHECK: OpStore {{%[0-9]+}} %[[BITCAST2]] {{.*}} + + %0 = tail call target("spirv.VulkanBuffer", [0 x <2 x double>], 12, 1) @llvm.spv.resource.handlefrombinding.tspirv.VulkanBuffer_a0v2f64_12_1t(i32 0, i32 2, i32 1, i32 0, ptr nonnull @.str.2) + %2 = tail call noundef align 16 dereferenceable(16) ptr addrspace(11) @llvm.spv.resource.getpointer.p11.tspirv.VulkanBuffer_a0v2f64_12_1t(target("spirv.VulkanBuffer", [0 x <2 x double>], 12, 1) %0, i32 0) + %3 = load <4 x i32>, ptr addrspace(11) %2 + %4 = tail call noundef align 16 dereferenceable(16) ptr addrspace(11) @llvm.spv.resource.getpointer.p11.tspirv.VulkanBuffer_a0v2f64_12_1t(target("spirv.VulkanBuffer", [0 x <2 x double>], 12, 1) %0, i32 1) + store <4 x i32> %3, ptr addrspace(11) %4 + ret void +} + +attributes #0 = { "hlsl.numthreads"="1,1,1" "hlsl.shader"="compute" } diff --git a/llvm/test/CodeGen/SystemZ/stackmap.ll b/llvm/test/CodeGen/SystemZ/stackmap.ll index 05b8de7..f414ea3 100644 --- a/llvm/test/CodeGen/SystemZ/stackmap.ll +++ b/llvm/test/CodeGen/SystemZ/stackmap.ll @@ -84,14 +84,14 @@ ; CHECK-NEXT: .short 8 ; CHECK-NEXT: .short 0 ; CHECK-NEXT: .short 0 -; CHECK-NEXT: .long 65535 +; CHECK-NEXT: .long -1 ; SmallConstant ; CHECK-NEXT: .byte 4 ; CHECK-NEXT: .byte 0 ; CHECK-NEXT: .short 8 ; CHECK-NEXT: .short 0 ; CHECK-NEXT: .short 0 -; CHECK-NEXT: .long 65535 +; CHECK-NEXT: .long -1 ; SmallConstant ; CHECK-NEXT: .byte 4 ; CHECK-NEXT: .byte 0 diff --git a/llvm/test/CodeGen/Thumb2/LowOverheadLoops/minloop.ll b/llvm/test/CodeGen/Thumb2/LowOverheadLoops/minloop.ll index 9c36bae..ec257bc 100644 --- a/llvm/test/CodeGen/Thumb2/LowOverheadLoops/minloop.ll +++ b/llvm/test/CodeGen/Thumb2/LowOverheadLoops/minloop.ll @@ -6,77 +6,81 @@ define void @arm_min_q31(ptr nocapture readonly %pSrc, i32 %blockSize, ptr nocap ; CHECK: @ %bb.0: @ %entry ; CHECK-NEXT: .save {r4, r5, r6, r7, r8, r9, r10, r11, lr} ; CHECK-NEXT: push.w {r4, r5, r6, r7, r8, r9, r10, r11, lr} +; CHECK-NEXT: .pad #4 +; CHECK-NEXT: sub sp, #4 ; CHECK-NEXT: ldr.w r12, [r0] ; CHECK-NEXT: subs.w r9, r1, #1 ; CHECK-NEXT: beq .LBB0_3 ; CHECK-NEXT: @ %bb.1: @ %while.body.preheader -; CHECK-NEXT: and r8, r9, #3 +; CHECK-NEXT: and r6, r9, #3 ; CHECK-NEXT: subs r7, r1, #2 ; CHECK-NEXT: cmp r7, #3 ; CHECK-NEXT: bhs .LBB0_4 ; CHECK-NEXT: @ %bb.2: -; CHECK-NEXT: movs r6, #0 -; CHECK-NEXT: b .LBB0_6 +; CHECK-NEXT: mov.w r10, #0 +; CHECK-NEXT: cbnz r6, .LBB0_7 +; CHECK-NEXT: b .LBB0_10 ; CHECK-NEXT: .LBB0_3: -; CHECK-NEXT: movs r6, #0 +; CHECK-NEXT: mov.w r10, #0 ; CHECK-NEXT: b .LBB0_10 ; CHECK-NEXT: .LBB0_4: @ %while.body.preheader.new ; CHECK-NEXT: bic r7, r9, #3 -; CHECK-NEXT: movs r6, #1 +; CHECK-NEXT: str r6, [sp] @ 4-byte Spill ; CHECK-NEXT: subs r7, #4 +; CHECK-NEXT: movs r6, #1 +; CHECK-NEXT: mov.w r8, #0 +; CHECK-NEXT: mov.w r10, #0 ; CHECK-NEXT: add.w lr, r6, r7, lsr #2 -; CHECK-NEXT: movs r6, #0 -; CHECK-NEXT: movs r7, #4 ; CHECK-NEXT: .LBB0_5: @ %while.body ; CHECK-NEXT: @ =>This Inner Loop Header: Depth=1 -; CHECK-NEXT: ldr r10, [r0, #16]! -; CHECK-NEXT: sub.w r9, r9, #4 -; CHECK-NEXT: ldrd r5, r4, [r0, #-12] -; CHECK-NEXT: ldr r11, [r0, #-4] +; CHECK-NEXT: ldr r11, [r0, #16]! +; CHECK-NEXT: ldrd r5, r7, [r0, #-12] +; CHECK-NEXT: ldr r4, [r0, #-4] ; CHECK-NEXT: cmp r12, r5 -; CHECK-NEXT: it gt -; CHECK-NEXT: subgt r6, r7, #3 ; CHECK-NEXT: csel r5, r5, r12, gt -; CHECK-NEXT: cmp r5, r4 +; CHECK-NEXT: csinc r6, r10, r8, le +; CHECK-NEXT: cmp r5, r7 ; CHECK-NEXT: it gt -; CHECK-NEXT: subgt r6, r7, #2 -; CHECK-NEXT: csel r5, r4, r5, gt -; CHECK-NEXT: cmp r5, r11 +; CHECK-NEXT: addgt.w r6, r8, #2 +; CHECK-NEXT: csel r7, r7, r5, gt +; CHECK-NEXT: cmp r7, r4 ; CHECK-NEXT: it gt -; CHECK-NEXT: subgt r6, r7, #1 -; CHECK-NEXT: csel r5, r11, r5, gt -; CHECK-NEXT: cmp r5, r10 -; CHECK-NEXT: csel r6, r7, r6, gt -; CHECK-NEXT: add.w r7, r7, #4 -; CHECK-NEXT: csel r12, r10, r5, gt +; CHECK-NEXT: addgt.w r6, r8, #3 +; CHECK-NEXT: csel r7, r4, r7, gt +; CHECK-NEXT: add.w r8, r8, #4 +; CHECK-NEXT: cmp r7, r11 +; CHECK-NEXT: csel r10, r8, r6, gt +; CHECK-NEXT: csel r12, r11, r7, gt ; CHECK-NEXT: le lr, .LBB0_5 -; CHECK-NEXT: .LBB0_6: @ %while.end.loopexit.unr-lcssa -; CHECK-NEXT: cmp.w r8, #0 -; CHECK-NEXT: beq .LBB0_10 -; CHECK-NEXT: @ %bb.7: @ %while.body.epil +; CHECK-NEXT: @ %bb.6: @ %while.end.loopexit.unr-lcssa.loopexit +; CHECK-NEXT: ldr r6, [sp] @ 4-byte Reload +; CHECK-NEXT: sub.w r9, r9, r8 +; CHECK-NEXT: cbz r6, .LBB0_10 +; CHECK-NEXT: .LBB0_7: @ %while.body.epil ; CHECK-NEXT: ldr r7, [r0, #4] ; CHECK-NEXT: sub.w r1, r1, r9 ; CHECK-NEXT: cmp r12, r7 -; CHECK-NEXT: csel r6, r1, r6, gt +; CHECK-NEXT: csel r10, r1, r10, gt ; CHECK-NEXT: csel r12, r7, r12, gt -; CHECK-NEXT: cmp.w r8, #1 +; CHECK-NEXT: cmp r6, #1 ; CHECK-NEXT: beq .LBB0_10 ; CHECK-NEXT: @ %bb.8: @ %while.body.epil.1 ; CHECK-NEXT: ldr r7, [r0, #8] ; CHECK-NEXT: cmp r12, r7 -; CHECK-NEXT: csinc r6, r6, r1, le +; CHECK-NEXT: csinc r10, r10, r1, le ; CHECK-NEXT: csel r12, r7, r12, gt -; CHECK-NEXT: cmp.w r8, #2 +; CHECK-NEXT: cmp r6, #2 ; CHECK-NEXT: beq .LBB0_10 ; CHECK-NEXT: @ %bb.9: @ %while.body.epil.2 ; CHECK-NEXT: ldr r0, [r0, #12] ; CHECK-NEXT: cmp r12, r0 ; CHECK-NEXT: it gt -; CHECK-NEXT: addgt r6, r1, #2 +; CHECK-NEXT: addgt.w r10, r1, #2 ; CHECK-NEXT: csel r12, r0, r12, gt ; CHECK-NEXT: .LBB0_10: @ %while.end ; CHECK-NEXT: str.w r12, [r2] -; CHECK-NEXT: str r6, [r3] +; CHECK-NEXT: str.w r10, [r3] +; CHECK-NEXT: add sp, #4 ; CHECK-NEXT: pop.w {r4, r5, r6, r7, r8, r9, r10, r11, pc} entry: %0 = load i32, ptr %pSrc, align 4 diff --git a/llvm/test/CodeGen/X86/amx-tf32-internal.ll b/llvm/test/CodeGen/X86/amx-tf32-internal.ll index 6d0f3c5..caf7a1c 100644 --- a/llvm/test/CodeGen/X86/amx-tf32-internal.ll +++ b/llvm/test/CodeGen/X86/amx-tf32-internal.ll @@ -1,6 +1,6 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+amx-tile,+avx512f, \ -; RUN: -mattr=+amx-tf32,+amx-transpose -verify-machineinstrs | FileCheck %s +; RUN: -mattr=+amx-tf32 -verify-machineinstrs | FileCheck %s define void @test_amx(i8* %pointer, i8* %base, i64 %stride) { ; CHECK-LABEL: test_amx: @@ -20,7 +20,6 @@ define void @test_amx(i8* %pointer, i8* %base, i64 %stride) { ; CHECK-NEXT: tilezero %tmm1 ; CHECK-NEXT: tilezero %tmm2 ; CHECK-NEXT: tmmultf32ps %tmm1, %tmm0, %tmm2 -; CHECK-NEXT: ttmmultf32ps %tmm1, %tmm0, %tmm2 ; CHECK-NEXT: tilestored %tmm2, (%rdi,%rdx) ; CHECK-NEXT: tilerelease ; CHECK-NEXT: vzeroupper @@ -31,9 +30,8 @@ define void @test_amx(i8* %pointer, i8* %base, i64 %stride) { %c = call x86_amx @llvm.x86.tilezero.internal(i16 8, i16 8) %c1 = call x86_amx @llvm.x86.tmmultf32ps.internal(i16 8, i16 8, i16 8, x86_amx %c, x86_amx %a, x86_amx %b) - %c2 = call x86_amx @llvm.x86.ttmmultf32ps.internal(i16 8, i16 8, i16 8, x86_amx %c1, x86_amx %a, x86_amx %b) - call void @llvm.x86.tilestored64.internal(i16 8, i16 8, i8* %pointer, i64 %stride, x86_amx %c2) + call void @llvm.x86.tilestored64.internal(i16 8, i16 8, i8* %pointer, i64 %stride, x86_amx %c1) ret void } @@ -43,4 +41,3 @@ declare void @llvm.x86.tilestored64.internal(i16, i16, i8*, i64, x86_amx) declare x86_amx @llvm.x86.tmmultf32ps.internal(i16, i16, i16, x86_amx, x86_amx, x86_amx) -declare x86_amx @llvm.x86.ttmmultf32ps.internal(i16, i16, i16, x86_amx, x86_amx, x86_amx) diff --git a/llvm/test/CodeGen/X86/amx-tf32-intrinsics.ll b/llvm/test/CodeGen/X86/amx-tf32-intrinsics.ll index af1a7ae..642c1b7 100644 --- a/llvm/test/CodeGen/X86/amx-tf32-intrinsics.ll +++ b/llvm/test/CodeGen/X86/amx-tf32-intrinsics.ll @@ -1,5 +1,5 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc < %s -O0 -mtriple=x86_64-unknown-unknown -mattr=+amx-tile,+amx-tf32,+amx-transpose -verify-machineinstrs | FileCheck %s +; RUN: llc < %s -O0 -mtriple=x86_64-unknown-unknown -mattr=+amx-tile,+amx-tf32 -verify-machineinstrs | FileCheck %s define void @test_tmmultf32ps() { ; CHECK-LABEL: test_tmmultf32ps: @@ -11,13 +11,3 @@ define void @test_tmmultf32ps() { } declare void @llvm.x86.tmmultf32ps(i8 %A, i8 %B, i8 %C) -define void @test_ttmmultf32ps() { -; CHECK-LABEL: test_ttmmultf32ps: -; CHECK: # %bb.0: -; CHECK-NEXT: ttmmultf32ps %tmm3, %tmm2, %tmm1 -; CHECK-NEXT: retq - call void @llvm.x86.ttmmultf32ps(i8 1, i8 2, i8 3) - ret void -} -declare void @llvm.x86.ttmmultf32ps(i8 %A, i8 %B, i8 %C) - diff --git a/llvm/test/CodeGen/X86/amx_movrs_transpose_intrinsics.ll b/llvm/test/CodeGen/X86/amx_movrs_transpose_intrinsics.ll deleted file mode 100755 index 1f5758c..0000000 --- a/llvm/test/CodeGen/X86/amx_movrs_transpose_intrinsics.ll +++ /dev/null @@ -1,122 +0,0 @@ -; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc < %s -O0 -mtriple=x86_64-unknown-unknown -mattr=+amx-transpose,+amx-movrs | FileCheck %s --check-prefixes=CHECK,O0 -; RUN: llc < %s -O2 -mtriple=x86_64-unknown-unknown -mattr=+amx-transpose,+amx-movrs | FileCheck %s --check-prefixes=CHECK,O2 -; RUN: llc < %s -O2 -mtriple=x86_64-unknown-unknown -mattr=+amx-transpose,+amx-movrs,+egpr --show-mc-encoding | FileCheck %s --check-prefix=EGPR - -define void @test_amx(i64 %stride, i8* %addr1) #0 { -; CHECK-LABEL: test_amx: -; CHECK: # %bb.0: -; CHECK-NEXT: t2rpntlvwz0rs (%rsi,%rdi), %tmm0 -; CHECK-NEXT: t2rpntlvwz0rst1 (%rsi,%rdi), %tmm2 -; CHECK-NEXT: t2rpntlvwz1rs (%rsi,%rdi), %tmm0 -; CHECK-NEXT: t2rpntlvwz1rst1 (%rsi,%rdi), %tmm2 -; CHECK-NEXT: retq -; -; EGPR-LABEL: test_amx: -; EGPR: # %bb.0: -; EGPR-NEXT: t2rpntlvwz0rs (%rsi,%rdi), %tmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe5,0x78,0xf8,0x04,0x3e] -; EGPR-NEXT: t2rpntlvwz0rst1 (%rsi,%rdi), %tmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe5,0x78,0xf9,0x14,0x3e] -; EGPR-NEXT: t2rpntlvwz1rs (%rsi,%rdi), %tmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe5,0x79,0xf8,0x04,0x3e] -; EGPR-NEXT: t2rpntlvwz1rst1 (%rsi,%rdi), %tmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe5,0x79,0xf9,0x14,0x3e] -; EGPR-NEXT: retq # encoding: [0xc3] - call void @llvm.x86.t2rpntlvwz0rs(i8 1, i8* %addr1, i64 %stride) - call void @llvm.x86.t2rpntlvwz0rst1(i8 2, i8* %addr1, i64 %stride) - call void @llvm.x86.t2rpntlvwz1rs(i8 1, i8* %addr1, i64 %stride) - call void @llvm.x86.t2rpntlvwz1rst1(i8 2, i8* %addr1, i64 %stride) - ret void -} -declare void @llvm.x86.t2rpntlvwz0rs(i8 , i8* , i64 ) -declare void @llvm.x86.t2rpntlvwz0rst1(i8 , i8* , i64 ) -declare void @llvm.x86.t2rpntlvwz1rs(i8 , i8* , i64 ) -declare void @llvm.x86.t2rpntlvwz1rst1(i8 , i8* , i64 ) - -define void @test_amx2(i8* %base, i64 %stride) #0 { -; O0-LABEL: test_amx2: -; O0: # %bb.0: -; O0-NEXT: xorps %xmm0, %xmm0 -; O0-NEXT: movups %xmm0, -{{[0-9]+}}(%rsp) -; O0-NEXT: movups %xmm0, -{{[0-9]+}}(%rsp) -; O0-NEXT: movups %xmm0, -{{[0-9]+}}(%rsp) -; O0-NEXT: movups %xmm0, -{{[0-9]+}}(%rsp) -; O0-NEXT: movb $1, -{{[0-9]+}}(%rsp) -; O0-NEXT: movw $8, %ax -; O0-NEXT: # implicit-def: $al -; O0-NEXT: movb %al, -{{[0-9]+}}(%rsp) -; O0-NEXT: movw %ax, -{{[0-9]+}}(%rsp) -; O0-NEXT: # implicit-def: $al -; O0-NEXT: movb %al, -{{[0-9]+}}(%rsp) -; O0-NEXT: movw %ax, -{{[0-9]+}}(%rsp) -; O0-NEXT: ldtilecfg -{{[0-9]+}}(%rsp) -; O0-NEXT: t2rpntlvwz0rst1 (%rdi,%rsi), %tmm4 -; O0-NEXT: movw $8, %ax -; O0-NEXT: # implicit-def: $al -; O0-NEXT: movb %al, -{{[0-9]+}}(%rsp) -; O0-NEXT: movw %ax, -{{[0-9]+}}(%rsp) -; O0-NEXT: # implicit-def: $al -; O0-NEXT: movb %al, -{{[0-9]+}}(%rsp) -; O0-NEXT: movw %ax, -{{[0-9]+}}(%rsp) -; O0-NEXT: ldtilecfg -{{[0-9]+}}(%rsp) -; O0-NEXT: t2rpntlvwz1rs (%rdi,%rsi), %tmm4 -; O0-NEXT: movw $8, %ax -; O0-NEXT: # implicit-def: $al -; O0-NEXT: movb %al, -{{[0-9]+}}(%rsp) -; O0-NEXT: movw %ax, -{{[0-9]+}}(%rsp) -; O0-NEXT: # implicit-def: $al -; O0-NEXT: movb %al, -{{[0-9]+}}(%rsp) -; O0-NEXT: movw %ax, -{{[0-9]+}}(%rsp) -; O0-NEXT: ldtilecfg -{{[0-9]+}}(%rsp) -; O0-NEXT: t2rpntlvwz1rst1 (%rdi,%rsi), %tmm4 -; O0-NEXT: tilerelease -; O0-NEXT: retq -; -; O2-LABEL: test_amx2: -; O2: # %bb.0: -; O2-NEXT: xorps %xmm0, %xmm0 -; O2-NEXT: movups %xmm0, -{{[0-9]+}}(%rsp) -; O2-NEXT: movups %xmm0, -{{[0-9]+}}(%rsp) -; O2-NEXT: movups %xmm0, -{{[0-9]+}}(%rsp) -; O2-NEXT: movups %xmm0, -{{[0-9]+}}(%rsp) -; O2-NEXT: movb $1, -{{[0-9]+}}(%rsp) -; O2-NEXT: movb $8, -{{[0-9]+}}(%rsp) -; O2-NEXT: movw $8, -{{[0-9]+}}(%rsp) -; O2-NEXT: movb $8, -{{[0-9]+}}(%rsp) -; O2-NEXT: movw $8, -{{[0-9]+}}(%rsp) -; O2-NEXT: ldtilecfg -{{[0-9]+}}(%rsp) -; O2-NEXT: movw $8, %ax -; O2-NEXT: t2rpntlvwz0rs (%rdi,%rsi), %tmm4 -; O2-NEXT: t2rpntlvwz0rst1 (%rdi,%rsi), %tmm4 -; O2-NEXT: t2rpntlvwz1rs (%rdi,%rsi), %tmm4 -; O2-NEXT: t2rpntlvwz1rst1 (%rdi,%rsi), %tmm4 -; O2-NEXT: tilerelease -; O2-NEXT: retq -; -; EGPR-LABEL: test_amx2: -; EGPR: # %bb.0: -; EGPR-NEXT: xorps %xmm0, %xmm0 # encoding: [0x0f,0x57,0xc0] -; EGPR-NEXT: movups %xmm0, -{{[0-9]+}}(%rsp) # encoding: [0x0f,0x11,0x44,0x24,0xc0] -; EGPR-NEXT: movups %xmm0, -{{[0-9]+}}(%rsp) # encoding: [0x0f,0x11,0x44,0x24,0xd0] -; EGPR-NEXT: movups %xmm0, -{{[0-9]+}}(%rsp) # encoding: [0x0f,0x11,0x44,0x24,0xe0] -; EGPR-NEXT: movups %xmm0, -{{[0-9]+}}(%rsp) # encoding: [0x0f,0x11,0x44,0x24,0xf0] -; EGPR-NEXT: movb $1, -{{[0-9]+}}(%rsp) # encoding: [0xc6,0x44,0x24,0xc0,0x01] -; EGPR-NEXT: movb $8, -{{[0-9]+}}(%rsp) # encoding: [0xc6,0x44,0x24,0xf4,0x08] -; EGPR-NEXT: movw $8, -{{[0-9]+}}(%rsp) # encoding: [0x66,0xc7,0x44,0x24,0xd8,0x08,0x00] -; EGPR-NEXT: movb $8, -{{[0-9]+}}(%rsp) # encoding: [0xc6,0x44,0x24,0xf5,0x08] -; EGPR-NEXT: movw $8, -{{[0-9]+}}(%rsp) # encoding: [0x66,0xc7,0x44,0x24,0xda,0x08,0x00] -; EGPR-NEXT: ldtilecfg -{{[0-9]+}}(%rsp) # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0x49,0x44,0x24,0xc0] -; EGPR-NEXT: movw $8, %ax # encoding: [0x66,0xb8,0x08,0x00] -; EGPR-NEXT: t2rpntlvwz0rs (%rdi,%rsi), %tmm4 # EVEX TO VEX Compression encoding: [0xc4,0xe5,0x78,0xf8,0x24,0x37] -; EGPR-NEXT: t2rpntlvwz0rst1 (%rdi,%rsi), %tmm4 # EVEX TO VEX Compression encoding: [0xc4,0xe5,0x78,0xf9,0x24,0x37] -; EGPR-NEXT: t2rpntlvwz1rs (%rdi,%rsi), %tmm4 # EVEX TO VEX Compression encoding: [0xc4,0xe5,0x79,0xf8,0x24,0x37] -; EGPR-NEXT: t2rpntlvwz1rst1 (%rdi,%rsi), %tmm4 # EVEX TO VEX Compression encoding: [0xc4,0xe5,0x79,0xf9,0x24,0x37] -; EGPR-NEXT: tilerelease # encoding: [0xc4,0xe2,0x78,0x49,0xc0] -; EGPR-NEXT: retq # encoding: [0xc3] - call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0rs.internal(i16 8, i16 8, i16 8, i8* %base, i64 %stride) - call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0rst1.internal(i16 8, i16 8, i16 8, i8* %base, i64 %stride) - call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz1rs.internal(i16 8, i16 8, i16 8, i8* %base, i64 %stride) - call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz1rst1.internal(i16 8, i16 8, i16 8, i8* %base, i64 %stride) - ret void -} -declare { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0rs.internal(i16, i16, i16, i8*, i64) -declare { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0rst1.internal(i16, i16, i16, i8*, i64) -declare { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz1rs.internal(i16, i16, i16, i8*, i64) -declare { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz1rst1.internal(i16, i16, i16, i8*, i64) diff --git a/llvm/test/CodeGen/X86/amx_tile_pair_O2_to_O0.ll b/llvm/test/CodeGen/X86/amx_tile_pair_O2_to_O0.ll deleted file mode 100644 index 4f41410..0000000 --- a/llvm/test/CodeGen/X86/amx_tile_pair_O2_to_O0.ll +++ /dev/null @@ -1,136 +0,0 @@ -; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc < %s -O0 -mtriple=x86_64-unknown-unknown -mattr=+amx-tile,+amx-bf16,+avx512f, \ -; RUN: -mattr=+amx-transpose -verify-machineinstrs | FileCheck %s - -@buf = dso_local global [2048 x i8] zeroinitializer, align 16 -@buf2 = dso_local global [2048 x i8] zeroinitializer, align 16 - -define dso_local void @test_tile_2rpntlvwz0(i16 noundef signext %row, i16 noundef signext %col0, i16 noundef signext %col1) local_unnamed_addr #0 { -; CHECK-LABEL: test_tile_2rpntlvwz0: -; CHECK: # %bb.0: # %entry -; CHECK-NEXT: pushq %rbp -; CHECK-NEXT: .cfi_def_cfa_offset 16 -; CHECK-NEXT: .cfi_offset %rbp, -16 -; CHECK-NEXT: movq %rsp, %rbp -; CHECK-NEXT: .cfi_def_cfa_register %rbp -; CHECK-NEXT: pushq %rbx -; CHECK-NEXT: andq $-1024, %rsp # imm = 0xFC00 -; CHECK-NEXT: subq $8192, %rsp # imm = 0x2000 -; CHECK-NEXT: .cfi_offset %rbx, -24 -; CHECK-NEXT: vxorps %xmm0, %xmm0, %xmm0 -; CHECK-NEXT: vmovups %zmm0, {{[0-9]+}}(%rsp) -; CHECK-NEXT: movb $1, {{[0-9]+}}(%rsp) -; CHECK-NEXT: # kill: def $dx killed $dx killed $edx -; CHECK-NEXT: movw %si, %cx -; CHECK-NEXT: movw %di, %ax -; CHECK-NEXT: # implicit-def: $al -; CHECK-NEXT: movb %al, {{[0-9]+}}(%rsp) -; CHECK-NEXT: movw %dx, {{[0-9]+}}(%rsp) -; CHECK-NEXT: # implicit-def: $al -; CHECK-NEXT: movb %al, {{[0-9]+}}(%rsp) -; CHECK-NEXT: movw %dx, {{[0-9]+}}(%rsp) -; CHECK-NEXT: # implicit-def: $al -; CHECK-NEXT: movb %al, {{[0-9]+}}(%rsp) -; CHECK-NEXT: movw %cx, {{[0-9]+}}(%rsp) -; CHECK-NEXT: # implicit-def: $cl -; CHECK-NEXT: movb %cl, {{[0-9]+}}(%rsp) -; CHECK-NEXT: movw %dx, {{[0-9]+}}(%rsp) -; CHECK-NEXT: # implicit-def: $al -; CHECK-NEXT: movb %al, {{[0-9]+}}(%rsp) -; CHECK-NEXT: movw %cx, {{[0-9]+}}(%rsp) -; CHECK-NEXT: # implicit-def: $al -; CHECK-NEXT: movb %al, {{[0-9]+}}(%rsp) -; CHECK-NEXT: movw %cx, {{[0-9]+}}(%rsp) -; CHECK-NEXT: # implicit-def: $al -; CHECK-NEXT: movb %al, {{[0-9]+}}(%rsp) -; CHECK-NEXT: movw %cx, {{[0-9]+}}(%rsp) -; CHECK-NEXT: # implicit-def: $al -; CHECK-NEXT: movb %al, {{[0-9]+}}(%rsp) -; CHECK-NEXT: movw %dx, {{[0-9]+}}(%rsp) -; CHECK-NEXT: ldtilecfg {{[0-9]+}}(%rsp) -; CHECK-NEXT: movl $buf, %esi -; CHECK-NEXT: movl $32, %edi -; CHECK-NEXT: t2rpntlvwz0 (%rsi,%rdi), %tmm4 -; CHECK-NEXT: movabsq $64, %rbx -; CHECK-NEXT: tilestored %tmm5, (%rsp,%rbx) # 1024-byte Folded Spill -; CHECK-NEXT: tileloadd (%rsp,%rbx), %tmm0 # 1024-byte Folded Reload -; CHECK-NEXT: movabsq $64, %rbx -; CHECK-NEXT: tilestored %tmm4, 1024(%rsp,%rbx) # 1024-byte Folded Spill -; CHECK-NEXT: tileloadd 1024(%rsp,%rbx), %tmm1 # 1024-byte Folded Reload -; CHECK-NEXT: movl $64, %edi -; CHECK-NEXT: leaq {{[0-9]+}}(%rsp), %rsi -; CHECK-NEXT: tilestored %tmm1, (%rsi,%rdi) -; CHECK-NEXT: movl $64, %edi -; CHECK-NEXT: leaq {{[0-9]+}}(%rsp), %rsi -; CHECK-NEXT: tilestored %tmm0, (%rsi,%rdi) -; CHECK-NEXT: tilezero %tmm0 -; CHECK-NEXT: movl $64, %edi -; CHECK-NEXT: leaq {{[0-9]+}}(%rsp), %rsi -; CHECK-NEXT: tilestored %tmm0, (%rsi,%rdi) -; CHECK-NEXT: movl $64, %edi -; CHECK-NEXT: leaq {{[0-9]+}}(%rsp), %rsi -; CHECK-NEXT: tileloadd (%rsi,%rdi), %tmm1 -; CHECK-NEXT: movl $64, %edi -; CHECK-NEXT: leaq {{[0-9]+}}(%rsp), %rsi -; CHECK-NEXT: tileloadd (%rsi,%rdi), %tmm2 -; CHECK-NEXT: movl $64, %edi -; CHECK-NEXT: leaq {{[0-9]+}}(%rsp), %rsi -; CHECK-NEXT: tileloadd (%rsi,%rdi), %tmm0 -; CHECK-NEXT: tdpbssd %tmm2, %tmm1, %tmm0 -; CHECK-NEXT: movl $64, %edi -; CHECK-NEXT: leaq {{[0-9]+}}(%rsp), %rsi -; CHECK-NEXT: tilestored %tmm0, (%rsi,%rdi) -; CHECK-NEXT: movl $64, %edi -; CHECK-NEXT: leaq {{[0-9]+}}(%rsp), %rsi -; CHECK-NEXT: tileloadd (%rsi,%rdi), %tmm0 -; CHECK-NEXT: movl $buf2, %edx -; CHECK-NEXT: movl $32, %esi -; CHECK-NEXT: tilestored %tmm0, (%rdx,%rsi) -; CHECK-NEXT: leaq -8(%rbp), %rsp -; CHECK-NEXT: popq %rbx -; CHECK-NEXT: popq %rbp -; CHECK-NEXT: .cfi_def_cfa %rsp, 8 -; CHECK-NEXT: tilerelease -; CHECK-NEXT: vzeroupper -; CHECK-NEXT: retq -entry: - %0 = tail call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0.internal(i16 %row, i16 %col0, i16 %col1, ptr @buf, i64 32) #3 - %1 = extractvalue { x86_amx, x86_amx } %0, 0 - %2 = tail call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx %1) #3 - %3 = extractvalue { x86_amx, x86_amx } %0, 1 - %4 = tail call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx %3) #3 - %5 = tail call x86_amx @llvm.x86.tilezero.internal(i16 %row, i16 %col0) #3 - %6 = tail call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx %5) #3 - %7 = tail call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> %6) #3 - %8 = tail call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> %2) #3 - %9 = tail call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> %4) #3 - %10 = tail call x86_amx @llvm.x86.tdpbssd.internal(i16 %row, i16 %col1, i16 %col0, x86_amx %7, x86_amx %8, x86_amx %9) #3 - %11 = tail call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx %10) #3 - %12 = tail call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> %11) #3 - tail call void @llvm.x86.tilestored64.internal(i16 %row, i16 %col0, ptr @buf2, i64 32, x86_amx %12) #3 - ret void -} - -declare { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0.internal(i16, i16, i16, ptr, i64) #1 - -declare <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx) #2 - -declare x86_amx @llvm.x86.tilezero.internal(i16, i16) #3 - -declare x86_amx @llvm.x86.tdpbssd.internal(i16, i16, i16, x86_amx, x86_amx, x86_amx) #3 - -declare x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32>) #2 - -declare void @llvm.x86.tilestored64.internal(i16, i16, ptr, i64, x86_amx) #4 - -attributes #0 = { nounwind uwtable "target-cpu"="x86-64" "target-features"="+amx-bf16,+amx-int8,+amx-tile,+amx-transpose" } -attributes #1 = { argmemonly nofree nounwind readonly } -attributes #2 = { nofree nosync nounwind readnone } -attributes #3 = { nounwind } -attributes #4 = { argmemonly nounwind writeonly } - -!llvm.module.flags = !{!0, !1, !2} - -!0 = !{i32 1, !"wchar_size", i32 4} -!1 = !{i32 7, !"uwtable", i32 2} -!2 = !{i32 7, !"frame-pointer", i32 2} diff --git a/llvm/test/CodeGen/X86/amx_tile_pair_configure_O0.mir b/llvm/test/CodeGen/X86/amx_tile_pair_configure_O0.mir deleted file mode 100644 index ab12ab3..0000000 --- a/llvm/test/CodeGen/X86/amx_tile_pair_configure_O0.mir +++ /dev/null @@ -1,165 +0,0 @@ -# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py -# RUN: llc -O0 -mtriple=x86_64-unknown-unknown -mattr=+amx-tile,+amx-bf16,+avx512f, \ -# RUN: -mattr=+amx-transpose -run-pass=fasttileconfig -o - %s | FileCheck %s - ---- -name: test_tile_2rpntlvwz0 -alignment: 16 -exposesReturnsTwice: false -legalized: false -regBankSelected: false -selected: false -failedISel: false -tracksRegLiveness: true -hasWinCFI: false -callsEHReturn: false -callsUnwindInit: false -hasEHContTarget: false -hasEHScopes: false -hasEHFunclets: false -failsVerification: false -tracksDebugUserValues: false -registers: [] -liveins: - - { reg: '$edi', virtual-reg: '' } - - { reg: '$esi', virtual-reg: '' } - - { reg: '$edx', virtual-reg: '' } -frameInfo: - isFrameAddressTaken: false - isReturnAddressTaken: false - hasStackMap: false - hasPatchPoint: false - stackSize: 0 - offsetAdjustment: 0 - maxAlignment: 1024 - adjustsStack: false - hasCalls: true - stackProtector: '' - functionContext: '' - maxCallFrameSize: 4294967295 - cvBytesOfCalleeSavedRegisters: 0 - hasOpaqueSPAdjustment: false - hasVAStart: false - hasMustTailInVarArgFunc: false - hasTailCall: false - localFrameSize: 0 - savePoint: [] - restorePoint: [] -fixedStack: [] -stack: - - { id: 0, name: '', type: default, offset: 0, size: 8, alignment: 8, - stack-id: default, callee-saved-register: '', callee-saved-restored: true, - debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } - - { id: 1, name: '', type: default, offset: 0, size: 8, alignment: 8, - stack-id: default, callee-saved-register: '', callee-saved-restored: true, - debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } - - { id: 2, name: '', type: default, offset: 0, size: 8, alignment: 8, - stack-id: default, callee-saved-register: '', callee-saved-restored: true, - debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } - - { id: 3, name: '', type: default, offset: 0, size: 8, alignment: 8, - stack-id: default, callee-saved-register: '', callee-saved-restored: true, - debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } - - { id: 4, name: '', type: default, offset: 0, size: 64, alignment: 4, - stack-id: default, callee-saved-register: '', callee-saved-restored: true, - debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } - - { id: 5, name: '', type: spill-slot, offset: 0, size: 2, alignment: 2, - stack-id: default, callee-saved-register: '', callee-saved-restored: true, - debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } - - { id: 6, name: '', type: spill-slot, offset: 0, size: 2, alignment: 2, - stack-id: default, callee-saved-register: '', callee-saved-restored: true, - debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } - - { id: 7, name: '', type: spill-slot, offset: 0, size: 8, alignment: 8, - stack-id: default, callee-saved-register: '', callee-saved-restored: true, - debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } -callSites: [] -debugValueSubstitutions: [] -constants: [] -machineFunctionInfo: - amxProgModel: ManagedRA -body: | - bb.0.entry: - liveins: $rdi, $rsi, $rdx, $rax - - ; CHECK-LABEL: name: test_tile_2rpntlvwz0 - ; CHECK: liveins: $rdi, $rsi, $rdx, $rax - ; CHECK-NEXT: {{ $}} - ; CHECK-NEXT: renamable $zmm0 = AVX512_512_SET0 - ; CHECK-NEXT: VMOVUPSZmr %stack.4, 1, $noreg, 0, $noreg, killed renamable $zmm0 :: (store (s512) into %stack.4, align 4) - ; CHECK-NEXT: MOV8mi %stack.4, 1, $noreg, 0, $noreg, 1 :: (store (s512) into %stack.4, align 4) - ; CHECK-NEXT: renamable $rcx = MOV32ri64 64 - ; CHECK-NEXT: MOV64mr %stack.7, 1, $noreg, 0, $noreg, $rcx :: (store (s64) into %stack.7) - ; CHECK-NEXT: renamable $cx = MOV16ri 64 - ; CHECK-NEXT: MOV16mr %stack.5, 1, $noreg, 0, $noreg, $cx :: (store (s16) into %stack.5) - ; CHECK-NEXT: renamable $cx = MOV16ri 16 - ; CHECK-NEXT: renamable $r8w = MOV16ri 16 - ; CHECK-NEXT: MOV16mr %stack.6, 1, $noreg, 0, $noreg, $r8w :: (store (s16) into %stack.6) - ; CHECK-NEXT: $al = IMPLICIT_DEF - ; CHECK-NEXT: MOV8mr %stack.4, 1, $noreg, 48, $noreg, $al :: (store (s512) into %stack.4 + 48, align 4) - ; CHECK-NEXT: MOV16mr %stack.4, 1, $noreg, 16, $noreg, $cx :: (store (s512) into %stack.4 + 16, align 4) - ; CHECK-NEXT: $al = IMPLICIT_DEF - ; CHECK-NEXT: MOV8mr %stack.4, 1, $noreg, 50, $noreg, $al :: (store (s512) into %stack.4 + 50, align 2, basealign 4) - ; CHECK-NEXT: MOV16mr %stack.4, 1, $noreg, 20, $noreg, $cx :: (store (s512) into %stack.4 + 20, align 4) - ; CHECK-NEXT: $al = IMPLICIT_DEF - ; CHECK-NEXT: MOV8mr %stack.4, 1, $noreg, 49, $noreg, $al :: (store (s512) into %stack.4 + 49, align 1, basealign 4) - ; CHECK-NEXT: MOV16mr %stack.4, 1, $noreg, 18, $noreg, $di :: (store (s512) into %stack.4 + 18, align 2, basealign 4) - ; CHECK-NEXT: $al = IMPLICIT_DEF - ; CHECK-NEXT: MOV8mr %stack.4, 1, $noreg, 48, $noreg, $al :: (store (s512) into %stack.4 + 48, align 4) - ; CHECK-NEXT: MOV16mr %stack.4, 1, $noreg, 16, $noreg, $cx :: (store (s512) into %stack.4 + 16, align 4) - ; CHECK-NEXT: $al = IMPLICIT_DEF - ; CHECK-NEXT: MOV8mr %stack.4, 1, $noreg, 48, $noreg, $al :: (store (s512) into %stack.4 + 48, align 4) - ; CHECK-NEXT: MOV16mr %stack.4, 1, $noreg, 16, $noreg, $cx :: (store (s512) into %stack.4 + 16, align 4) - ; CHECK-NEXT: $al = IMPLICIT_DEF - ; CHECK-NEXT: MOV8mr %stack.4, 1, $noreg, 52, $noreg, $al :: (store (s512) into %stack.4 + 52, align 4) - ; CHECK-NEXT: MOV16mr %stack.4, 1, $noreg, 24, $noreg, $cx :: (store (s512) into %stack.4 + 24, align 4) - ; CHECK-NEXT: $al = IMPLICIT_DEF - ; CHECK-NEXT: MOV8mr %stack.4, 1, $noreg, 53, $noreg, $al :: (store (s512) into %stack.4 + 53, align 1, basealign 4) - ; CHECK-NEXT: MOV16mr %stack.4, 1, $noreg, 26, $noreg, $di :: (store (s512) into %stack.4 + 26, align 2, basealign 4) - ; CHECK-NEXT: PLDTILECFGV %stack.4, 1, $noreg, 0, $noreg, implicit-def dead $tmm0, implicit-def dead $tmm1, implicit-def dead $tmm2, implicit-def dead $tmm3, implicit-def dead $tmm4, implicit-def dead $tmm5, implicit-def dead $tmm6, implicit-def dead $tmm7 :: (load (s512) from %stack.4, align 4) - ; CHECK-NEXT: renamable $r9 = COPY $rsi - ; CHECK-NEXT: $rsi = MOV64rm %stack.7, 1, $noreg, 0, $noreg :: (load (s64) from %stack.7) - ; CHECK-NEXT: renamable $r8 = COPY $rdi - ; CHECK-NEXT: $di = MOV16rm %stack.6, 1, $noreg, 0, $noreg :: (load (s16) from %stack.6) - ; CHECK-NEXT: renamable $r10 = COPY $rax - ; CHECK-NEXT: $ax = MOV16rm %stack.5, 1, $noreg, 0, $noreg :: (load (s16) from %stack.5) - ; CHECK-NEXT: renamable $tmm4_tmm5 = PT2RPNTLVWZ0V renamable $ax, renamable $cx, renamable $di, renamable $rdx, 1, killed renamable $r10, 0, $noreg - ; CHECK-NEXT: renamable $tmm0 = COPY renamable $tmm5 - ; CHECK-NEXT: renamable $tmm1 = COPY renamable $tmm4, implicit killed $tmm4_tmm5 - ; CHECK-NEXT: PTILESTOREDV renamable $ax, renamable $cx, renamable $r9, 1, renamable $rsi, 0, $noreg, killed renamable $tmm1 - ; CHECK-NEXT: PTILESTOREDV renamable $ax, renamable $di, renamable $r8, 1, renamable $rsi, 0, $noreg, killed renamable $tmm0 - ; CHECK-NEXT: renamable $tmm0 = PTILEZEROV renamable $ax, renamable $cx - ; CHECK-NEXT: PTILESTOREDV renamable $ax, renamable $cx, renamable $rdx, 1, renamable $rsi, 0, $noreg, killed renamable $tmm0 - ; CHECK-NEXT: renamable $tmm0 = PTILELOADDV renamable $ax, renamable $cx, killed renamable $r9, 1, renamable $rsi, 0, $noreg - ; CHECK-NEXT: renamable $tmm1 = PTILELOADDV renamable $ax, renamable $di, killed renamable $r8, 1, renamable $rsi, 0, $noreg - ; CHECK-NEXT: renamable $tmm2 = PTILELOADDV renamable $ax, renamable $cx, renamable $rdx, 1, renamable $rsi, 0, $noreg - ; CHECK-NEXT: renamable $tmm0 = PTDPBSSDV renamable $ax, renamable $cx, killed renamable $di, renamable $tmm0, killed renamable $tmm1, killed renamable $tmm2 - ; CHECK-NEXT: PTILESTOREDV killed renamable $ax, killed renamable $cx, killed renamable $rdx, 1, killed renamable $rsi, 0, $noreg, killed renamable $tmm0 - renamable $zmm0 = AVX512_512_SET0 - VMOVUPSZmr %stack.4, 1, $noreg, 0, $noreg, killed renamable $zmm0 :: (store (s512) into %stack.4, align 4) - MOV8mi %stack.4, 1, $noreg, 0, $noreg, 1 :: (store (s512) into %stack.4, align 4) - renamable $rcx = MOV32ri64 64 - MOV64mr %stack.7, 1, $noreg, 0, $noreg, $rcx :: (store (s64) into %stack.7) - renamable $cx = MOV16ri 64 - MOV16mr %stack.5, 1, $noreg, 0, $noreg, $cx :: (store (s16) into %stack.5) - renamable $cx = MOV16ri 16 - renamable $r8w = MOV16ri 16 - MOV16mr %stack.6, 1, $noreg, 0, $noreg, $r8w :: (store (s16) into %stack.6) - PLDTILECFGV %stack.4, 1, $noreg, 0, $noreg, implicit-def dead $tmm0, implicit-def dead $tmm1, implicit-def dead $tmm2, implicit-def dead $tmm3, implicit-def dead $tmm4, implicit-def dead $tmm5, implicit-def dead $tmm6, implicit-def dead $tmm7 :: (load (s512) from %stack.4, align 4) - renamable $r9 = COPY $rsi - $rsi = MOV64rm %stack.7, 1, $noreg, 0, $noreg :: (load (s64) from %stack.7) - renamable $r8 = COPY $rdi - $di = MOV16rm %stack.6, 1, $noreg, 0, $noreg :: (load (s16) from %stack.6) - renamable $r10 = COPY $rax - $ax = MOV16rm %stack.5, 1, $noreg, 0, $noreg :: (load (s16) from %stack.5) - renamable $tmm4_tmm5 = PT2RPNTLVWZ0V renamable $ax, renamable $cx, renamable $di, renamable $rdx, 1, killed renamable $r10, 0, $noreg - renamable $tmm0 = COPY renamable $tmm5 - renamable $tmm1 = COPY renamable $tmm4, implicit killed $tmm4_tmm5 - PTILESTOREDV renamable $ax, renamable $cx, renamable $r9, 1, renamable $rsi, 0, $noreg, killed renamable $tmm1 - PTILESTOREDV renamable $ax, renamable $di, renamable $r8, 1, renamable $rsi, 0, $noreg, killed renamable $tmm0 - renamable $tmm0 = PTILEZEROV renamable $ax, renamable $cx - PTILESTOREDV renamable $ax, renamable $cx, renamable $rdx, 1, renamable $rsi, 0, $noreg, killed renamable $tmm0 - renamable $tmm0 = PTILELOADDV renamable $ax, renamable $cx, killed renamable $r9, 1, renamable $rsi, 0, $noreg - renamable $tmm1 = PTILELOADDV renamable $ax, renamable $di, killed renamable $r8, 1, renamable $rsi, 0, $noreg - renamable $tmm2 = PTILELOADDV renamable $ax, renamable $cx, renamable $rdx, 1, renamable $rsi, 0, $noreg - renamable $tmm0 = PTDPBSSDV renamable $ax, renamable $cx, killed renamable $di, renamable $tmm0, killed renamable $tmm1, killed renamable $tmm2 - PTILESTOREDV killed renamable $ax, killed renamable $cx, killed renamable $rdx, 1, killed renamable $rsi, 0, $noreg, killed renamable $tmm0 -... diff --git a/llvm/test/CodeGen/X86/amx_tile_pair_configure_O2.mir b/llvm/test/CodeGen/X86/amx_tile_pair_configure_O2.mir deleted file mode 100644 index c7d241f..0000000 --- a/llvm/test/CodeGen/X86/amx_tile_pair_configure_O2.mir +++ /dev/null @@ -1,153 +0,0 @@ -# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py -# RUN: llc -O2 -mtriple=x86_64-unknown-unknown -mattr=+amx-tile,+amx-bf16,+avx512f, \ -# RUN: -mattr=+amx-transpose -run-pass=greedy,tileconfig -o - %s | FileCheck %s - ---- | - @buf = dso_local global [2048 x i8] zeroinitializer, align 16 - @buf2 = dso_local global [2048 x i8] zeroinitializer, align 16 - - define dso_local void @test_tile_2rpntlvwz0(i16 noundef signext %row, i16 noundef signext %col0, i16 noundef signext %col1) local_unnamed_addr #0 { - entry: - %0 = tail call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0.internal(i16 %row, i16 %col0, i16 %col1, i8* getelementptr inbounds ([2048 x i8], [2048 x i8]* @buf, i64 0, i64 0), i64 32) #5 - %1 = extractvalue { x86_amx, x86_amx } %0, 0 - %2 = extractvalue { x86_amx, x86_amx } %0, 1 - %3 = tail call x86_amx @llvm.x86.tilezero.internal(i16 %row, i16 %col0) #5 - %4 = tail call x86_amx @llvm.x86.tdpbssd.internal(i16 %row, i16 %col1, i16 %col0, x86_amx %3, x86_amx %1, x86_amx %2) #5 - tail call void @llvm.x86.tilestored64.internal(i16 %row, i16 %col0, i8* getelementptr inbounds ([2048 x i8], [2048 x i8]* @buf2, i64 0, i64 0), i64 32, x86_amx %4) #5 - ret void - } - - declare { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0.internal(i16, i16, i16, i8*, i64) #1 - - declare <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx) #2 - - declare x86_amx @llvm.x86.tilezero.internal(i16, i16) #3 - - declare x86_amx @llvm.x86.tdpbssd.internal(i16, i16, i16, x86_amx, x86_amx, x86_amx) #3 - - declare x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32>) #2 - - declare void @llvm.x86.tilestored64.internal(i16, i16, i8*, i64, x86_amx) #4 - - attributes #0 = { nounwind uwtable "frame-pointer"="all" "min-legal-vector-width"="8192" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+amx-bf16,+amx-int8,+amx-tile,+amx-transpose,+avx,+avx2,+avx512f,+crc32,+cx8,+f16c,+fma,+fxsr,+mmx,+popcnt,+sse,+sse2,+sse3,+sse4.1,+sse4.2,+ssse3,+x87,+xsave,+amx-tile,+amx-bf16,+avx512f,+amx-transpose" "tune-cpu"="generic" } - attributes #1 = { argmemonly nounwind readonly "target-features"="+amx-tile,+amx-bf16,+avx512f,+amx-transpose" } - attributes #2 = { nounwind readnone "target-features"="+amx-tile,+amx-bf16,+avx512f,+amx-transpose" } - attributes #3 = { nounwind "target-features"="+amx-tile,+amx-bf16,+avx512f,+amx-transpose" } - attributes #4 = { argmemonly nounwind writeonly "target-features"="+amx-tile,+amx-bf16,+avx512f,+amx-transpose" } - attributes #5 = { nounwind } - -... ---- -name: test_tile_2rpntlvwz0 -alignment: 16 -exposesReturnsTwice: false -legalized: false -regBankSelected: false -selected: false -failedISel: false -tracksRegLiveness: true -hasWinCFI: false -callsEHReturn: false -callsUnwindInit: false -hasEHContTarget: false -hasEHScopes: false -hasEHFunclets: false -failsVerification: false -tracksDebugUserValues: false -registers: - - { id: 0, class: gr32, preferred-register: '' } - - { id: 1, class: gr32, preferred-register: '' } - - { id: 2, class: gr32, preferred-register: '' } - - { id: 3, class: gr16, preferred-register: '' } - - { id: 4, class: gr16, preferred-register: '' } - - { id: 5, class: gr16, preferred-register: '' } - - { id: 6, class: gr64, preferred-register: '' } - - { id: 7, class: gr64_nosp, preferred-register: '' } - - { id: 8, class: tilepair, preferred-register: '' } - - { id: 9, class: tile, preferred-register: '' } - - { id: 10, class: tile, preferred-register: '' } - - { id: 11, class: tile, preferred-register: '' } - - { id: 12, class: tile, preferred-register: '' } - - { id: 13, class: gr64, preferred-register: '' } - - { id: 14, class: vr512, preferred-register: '' } -liveins: - - { reg: '$edi', virtual-reg: '%0' } - - { reg: '$esi', virtual-reg: '%1' } - - { reg: '$edx', virtual-reg: '%2' } -frameInfo: - isFrameAddressTaken: false - isReturnAddressTaken: false - hasStackMap: false - hasPatchPoint: false - stackSize: 0 - offsetAdjustment: 0 - maxAlignment: 4 - adjustsStack: false - hasCalls: false - stackProtector: '' - functionContext: '' - maxCallFrameSize: 4294967295 - cvBytesOfCalleeSavedRegisters: 0 - hasOpaqueSPAdjustment: false - hasVAStart: false - hasMustTailInVarArgFunc: false - hasTailCall: false - localFrameSize: 0 - savePoint: [] - restorePoint: [] -fixedStack: [] -stack: - - { id: 0, name: '', type: default, offset: 0, size: 64, alignment: 4, - stack-id: default, callee-saved-register: '', callee-saved-restored: true, - debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } -callSites: [] -debugValueSubstitutions: [] -constants: [] -machineFunctionInfo: - amxProgModel: ManagedRA -body: | - bb.0.entry: - liveins: $edi, $esi, $edx - - - ; CHECK-LABEL: name: test_tile_2rpntlvwz0 - ; CHECK: liveins: $edi, $esi, $edx - ; CHECK-NEXT: {{ $}} - ; CHECK-NEXT: [[COPY:%[0-9]+]]:gr32 = COPY $edx - ; CHECK-NEXT: [[COPY1:%[0-9]+]]:gr32 = COPY $esi - ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gr32 = COPY $edi - ; CHECK-NEXT: [[AVX512_512_SET0_:%[0-9]+]]:vr512 = AVX512_512_SET0 - ; CHECK-NEXT: VMOVUPSZmr %stack.0, 1, $noreg, 0, $noreg, [[AVX512_512_SET0_]] :: (store (s512) into %stack.0, align 4) - ; CHECK-NEXT: MOV8mi %stack.0, 1, $noreg, 0, $noreg, 1 :: (store (s512) into %stack.0, align 4) - ; CHECK-NEXT: MOV16mr %stack.0, 1, $noreg, 26, $noreg, [[COPY]].sub_16bit :: (store (s512) into %stack.0 + 26, align 2, basealign 4) - ; CHECK-NEXT: MOV8mr %stack.0, 1, $noreg, 53, $noreg, [[COPY2]].sub_8bit :: (store (s512) into %stack.0 + 53, align 1, basealign 4) - ; CHECK-NEXT: MOV16mr %stack.0, 1, $noreg, 24, $noreg, [[COPY1]].sub_16bit :: (store (s512) into %stack.0 + 24, align 4) - ; CHECK-NEXT: MOV8mr %stack.0, 1, $noreg, 52, $noreg, [[COPY2]].sub_8bit :: (store (s512) into %stack.0 + 52, align 4) - ; CHECK-NEXT: MOV16mr %stack.0, 1, $noreg, 16, $noreg, [[COPY]].sub_16bit :: (store (s512) into %stack.0 + 16, align 4) - ; CHECK-NEXT: MOV8mr %stack.0, 1, $noreg, 48, $noreg, [[COPY2]].sub_8bit :: (store (s512) into %stack.0 + 48, align 4) - ; CHECK-NEXT: PLDTILECFGV %stack.0, 1, $noreg, 0, $noreg, implicit-def dead $tmm0, implicit-def dead $tmm1, implicit-def dead $tmm2, implicit-def dead $tmm3, implicit-def dead $tmm4, implicit-def dead $tmm5, implicit-def dead $tmm6, implicit-def dead $tmm7 :: (load (s512) from %stack.0, align 4) - ; CHECK-NEXT: [[MOV32ri64_:%[0-9]+]]:gr64 = MOV32ri64 @buf - ; CHECK-NEXT: [[MOV32ri64_1:%[0-9]+]]:gr64_nosp = MOV32ri64 32 - ; CHECK-NEXT: [[PT2RPNTLVWZ0V:%[0-9]+]]:tilepair = PT2RPNTLVWZ0V [[COPY2]].sub_16bit, [[COPY1]].sub_16bit, [[COPY]].sub_16bit, [[MOV32ri64_]], 1, [[MOV32ri64_1]], 0, $noreg - ; CHECK-NEXT: [[PTILEZEROV:%[0-9]+]]:tile = PTILEZEROV [[COPY2]].sub_16bit, [[COPY1]].sub_16bit - ; CHECK-NEXT: [[PTILEZEROV:%[0-9]+]]:tile = PTDPBSSDV [[COPY2]].sub_16bit, [[COPY]].sub_16bit, [[COPY1]].sub_16bit, [[PTILEZEROV]], [[PT2RPNTLVWZ0V]].sub_t0, [[PT2RPNTLVWZ0V]].sub_t1 - ; CHECK-NEXT: [[MOV32ri64_2:%[0-9]+]]:gr64 = MOV32ri64 @buf2 - ; CHECK-NEXT: PTILESTOREDV [[COPY2]].sub_16bit, [[COPY1]].sub_16bit, [[MOV32ri64_2]], 1, [[MOV32ri64_1]], 0, $noreg, [[PTILEZEROV]] - ; CHECK-NEXT: RET 0 - %2:gr32 = COPY $edx - %1:gr32 = COPY $esi - %0:gr32 = COPY $edi - %14:vr512 = AVX512_512_SET0 - VMOVUPSZmr %stack.0, 1, $noreg, 0, $noreg, %14 :: (store (s512) into %stack.0, align 4) - MOV8mi %stack.0, 1, $noreg, 0, $noreg, 1 :: (store (s512) into %stack.0, align 4) - PLDTILECFGV %stack.0, 1, $noreg, 0, $noreg, implicit-def dead $tmm0, implicit-def dead $tmm1, implicit-def dead $tmm2, implicit-def dead $tmm3, implicit-def dead $tmm4, implicit-def dead $tmm5, implicit-def dead $tmm6, implicit-def dead $tmm7 :: (load (s512) from %stack.0, align 4) - %6:gr64 = MOV32ri64 @buf - %7:gr64_nosp = MOV32ri64 32 - %8:tilepair = PT2RPNTLVWZ0V %0.sub_16bit, %1.sub_16bit, %2.sub_16bit, %6, 1, %7, 0, $noreg - %12:tile = PTILEZEROV %0.sub_16bit, %1.sub_16bit - %12:tile = PTDPBSSDV %0.sub_16bit, %2.sub_16bit, %1.sub_16bit, %12, %8.sub_t0, %8.sub_t1 - %13:gr64 = MOV32ri64 @buf2 - PTILESTOREDV %0.sub_16bit, %1.sub_16bit, %13, 1, %7, 0, $noreg, %12 - RET 0 - -... diff --git a/llvm/test/CodeGen/X86/amx_tile_pair_copy.mir b/llvm/test/CodeGen/X86/amx_tile_pair_copy.mir deleted file mode 100644 index 66b15aa..0000000 --- a/llvm/test/CodeGen/X86/amx_tile_pair_copy.mir +++ /dev/null @@ -1,97 +0,0 @@ -# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py -# RUN: llc -O0 -mtriple=x86_64-unknown-unknown -mattr=+amx-tile,+amx-bf16,+avx512f, \ -# RUN: -mattr=+amx-transpose -run-pass=lowertilecopy -o - %s | FileCheck %s - ---- -name: test_tile_2rpntlvwz0 -alignment: 16 -exposesReturnsTwice: false -legalized: false -regBankSelected: false -selected: false -failedISel: false -tracksRegLiveness: true -hasWinCFI: false -callsEHReturn: false -callsUnwindInit: false -hasEHContTarget: false -hasEHScopes: false -hasEHFunclets: false -failsVerification: false -tracksDebugUserValues: false -registers: [] -liveins: - - { reg: '$edi', virtual-reg: '' } - - { reg: '$esi', virtual-reg: '' } - - { reg: '$edx', virtual-reg: '' } - - { reg: '$cx', virtual-reg: '' } - - { reg: '$r9', virtual-reg: '' } - - { reg: '$r10', virtual-reg: '' } -frameInfo: - isFrameAddressTaken: false - isReturnAddressTaken: false - hasStackMap: false - hasPatchPoint: false - stackSize: 0 - offsetAdjustment: 0 - maxAlignment: 1024 - adjustsStack: false - hasCalls: true - stackProtector: '' - functionContext: '' - maxCallFrameSize: 4294967295 - cvBytesOfCalleeSavedRegisters: 0 - hasOpaqueSPAdjustment: false - hasVAStart: false - hasMustTailInVarArgFunc: false - hasTailCall: false - localFrameSize: 0 - savePoint: [] - restorePoint: [] -fixedStack: [] -stack: - - { id: 43, name: '', type: default, offset: 0, size: 64, alignment: 4, - stack-id: default, callee-saved-register: '', callee-saved-restored: true, - debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } - - { id: 68, name: '', type: spill-slot, offset: 0, size: 8, alignment: 8, - stack-id: default, callee-saved-register: '', callee-saved-restored: true, - debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } -callSites: [] -debugValueSubstitutions: [] -constants: [] -machineFunctionInfo: - amxProgModel: ManagedRA -body: | - bb.0.entry: - liveins: $edi, $esi, $edx, $cx, $di, $r8w, $r11, $r10, $rbx, $r8, $r9 - - - ; CHECK-LABEL: name: test_tile_2rpntlvwz0 - ; CHECK: liveins: $edi, $esi, $edx, $cx, $di, $r8w, $r11, $r10, $rbx, $r8, $r9 - ; CHECK-NEXT: {{ $}} - ; CHECK-NEXT: PLDTILECFGV %stack.0, 1, $noreg, 0, $noreg, implicit-def dead $tmm0, implicit-def dead $tmm1, implicit-def dead $tmm2, implicit-def dead $tmm3, implicit-def dead $tmm4, implicit-def dead $tmm5, implicit-def dead $tmm6, implicit-def dead $tmm7 :: (load (s512) from %stack.0, align 4) - ; CHECK-NEXT: renamable $tmm4_tmm5 = PT2RPNTLVWZ0V killed renamable $cx, killed renamable $di, killed renamable $r8w, killed renamable $r11, 1, killed renamable $rbx, 0, $noreg - ; CHECK-NEXT: $rax = MOV64ri 64 - ; CHECK-NEXT: TILESTORED %stack.3, 1, $rax, 0, $noreg, $tmm5 :: (store (s8192) into %stack.3) - ; CHECK-NEXT: $tmm0 = TILELOADD %stack.3, 1, killed $rax, 0, $noreg :: (load (s8192) from %stack.3) - ; CHECK-NEXT: $rax = MOV64ri 64 - ; CHECK-NEXT: TILESTORED %stack.2, 1, $rax, 0, $noreg, $tmm4 :: (store (s8192) into %stack.2) - ; CHECK-NEXT: $tmm1 = TILELOADD %stack.2, 1, killed $rax, 0, $noreg :: (load (s8192) from %stack.2) - ; CHECK-NEXT: renamable $r8 = MOV32ri64 64 - ; CHECK-NEXT: MOV64mr %stack.1, 1, $noreg, 0, $noreg, $r8 :: (store (s64) into %stack.1) - ; CHECK-NEXT: renamable $di = MOV16ri 64 - ; CHECK-NEXT: renamable $cx = MOV16ri 16 - ; CHECK-NEXT: PTILESTOREDV renamable $cx, renamable $di, killed renamable $r10, 1, renamable $r8, 0, $noreg, killed renamable $tmm1 - ; CHECK-NEXT: PTILESTOREDV killed renamable $cx, killed renamable $di, killed renamable $r9, 1, renamable $r8, 0, $noreg, killed renamable $tmm0 - PLDTILECFGV %stack.43, 1, $noreg, 0, $noreg, implicit-def dead $tmm0, implicit-def dead $tmm1, implicit-def dead $tmm2, implicit-def dead $tmm3, implicit-def dead $tmm4, implicit-def dead $tmm5, implicit-def dead $tmm6, implicit-def dead $tmm7 :: (load (s512) from %stack.43, align 4) - renamable $tmm4_tmm5 = PT2RPNTLVWZ0V killed renamable $cx, killed renamable $di, killed renamable $r8w, killed renamable $r11, 1, killed renamable $rbx, 0, $noreg - renamable $tmm0 = COPY renamable $tmm5 - renamable $tmm1 = COPY renamable $tmm4, implicit killed $tmm4_tmm5 - renamable $r8 = MOV32ri64 64 - MOV64mr %stack.68, 1, $noreg, 0, $noreg, $r8 :: (store (s64) into %stack.68) - renamable $di = MOV16ri 64 - renamable $cx = MOV16ri 16 - PTILESTOREDV renamable $cx, renamable $di, killed renamable $r10, 1, renamable $r8, 0, $noreg, killed renamable $tmm1 - PTILESTOREDV killed renamable $cx, killed renamable $di, killed renamable $r9, 1, renamable $r8, 0, $noreg, killed renamable $tmm0 - -... diff --git a/llvm/test/CodeGen/X86/amx_tile_pair_lower_type_O0.ll b/llvm/test/CodeGen/X86/amx_tile_pair_lower_type_O0.ll deleted file mode 100644 index 3549875..0000000 --- a/llvm/test/CodeGen/X86/amx_tile_pair_lower_type_O0.ll +++ /dev/null @@ -1,87 +0,0 @@ -; NOTE: Assertions have been autogenerated by utils/update_test_checks.py - ; RUN: opt --codegen-opt-level=0 -mtriple=x86_64 -x86-lower-amx-type %s -S | FileCheck %s - ; RUN: opt --codegen-opt-level=0 -mtriple=x86_64 -passes=x86-lower-amx-type %s -S | FileCheck %s - - @buf = dso_local global [2048 x i8] zeroinitializer, align 16 - - ; Function Attrs: noinline nounwind optnone uwtable - define dso_local void @test_tile_2rpntlvwz0(i16 noundef signext %row, i16 noundef signext %col0, i16 noundef signext %col1, ptr %m) #0 { -; CHECK-LABEL: @test_tile_2rpntlvwz0( -; CHECK-NEXT: entry: -; CHECK-NEXT: [[TMP0:%.*]] = udiv i16 [[COL1:%.*]], 4 -; CHECK-NEXT: [[TMP1:%.*]] = call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0.internal(i16 [[ROW:%.*]], i16 [[COL0:%.*]], i16 [[COL1]], ptr @buf, i64 32) #[[ATTR3:[0-9]+]] -; CHECK-NEXT: [[TMP2:%.*]] = extractvalue { x86_amx, x86_amx } [[TMP1]], 0 -; CHECK-NEXT: [[TMP3:%.*]] = sext i16 [[COL0]] to i64 -; CHECK-NEXT: call void @llvm.x86.tilestored64.internal(i16 [[ROW]], i16 [[COL0]], ptr [[M:%.*]], i64 [[TMP3]], x86_amx [[TMP2]]) -; CHECK-NEXT: [[TMP5:%.*]] = extractvalue { x86_amx, x86_amx } [[TMP1]], 1 -; CHECK-NEXT: [[TMP6:%.*]] = sext i16 [[COL1]] to i64 -; CHECK-NEXT: call void @llvm.x86.tilestored64.internal(i16 [[ROW]], i16 [[COL1]], ptr [[M]], i64 [[TMP6]], x86_amx [[TMP5]]) -; CHECK-NEXT: [[TMP8:%.*]] = call x86_amx @llvm.x86.tilezero.internal(i16 [[ROW]], i16 [[COL0]]) #[[ATTR3]] -; CHECK-NEXT: [[TMP9:%.*]] = sext i16 [[COL0]] to i64 -; CHECK-NEXT: call void @llvm.x86.tilestored64.internal(i16 [[ROW]], i16 [[COL0]], ptr [[M]], i64 [[TMP9]], x86_amx [[TMP8]]) -; CHECK-NEXT: [[TMP11:%.*]] = sext i16 [[COL0]] to i64 -; CHECK-NEXT: [[TMP13:%.*]] = call x86_amx @llvm.x86.tileloadd64.internal(i16 [[ROW]], i16 [[COL0]], ptr [[M]], i64 [[TMP11]]) -; CHECK-NEXT: [[TMP14:%.*]] = sext i16 [[COL1]] to i64 -; CHECK-NEXT: [[TMP16:%.*]] = call x86_amx @llvm.x86.tileloadd64.internal(i16 [[ROW]], i16 [[COL1]], ptr [[M]], i64 [[TMP14]]) -; CHECK-NEXT: [[TMP17:%.*]] = sext i16 [[COL0]] to i64 -; CHECK-NEXT: [[TMP19:%.*]] = call x86_amx @llvm.x86.tileloadd64.internal(i16 [[TMP0]], i16 [[COL0]], ptr [[M]], i64 [[TMP17]]) -; CHECK-NEXT: [[TMP20:%.*]] = call x86_amx @llvm.x86.tdpbssd.internal(i16 [[ROW]], i16 [[COL0]], i16 [[COL1]], x86_amx [[TMP13]], x86_amx [[TMP16]], x86_amx [[TMP19]]) #[[ATTR3]] -; CHECK-NEXT: [[TMP21:%.*]] = sext i16 [[COL0]] to i64 -; CHECK-NEXT: call void @llvm.x86.tilestored64.internal(i16 [[ROW]], i16 [[COL0]], ptr [[M]], i64 [[TMP21]], x86_amx [[TMP20]]) -; CHECK-NEXT: ret void -; - entry: - - %0 = call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0.internal(i16 %row, i16 %col0, i16 %col1, ptr getelementptr inbounds ([2048 x i8], ptr @buf, i64 0, i64 0), i64 32) #7 - %1 = extractvalue { x86_amx, x86_amx } %0, 0 - %2 = call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx %1) #7 - store <256 x i32> %2, ptr %m, align 1024 - - %3 = extractvalue { x86_amx, x86_amx } %0, 1 - %4 = call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx %3) #7 - store <256 x i32> %4, ptr %m, align 1024 - - %5 = call x86_amx @llvm.x86.tilezero.internal(i16 %row, i16 %col0) #7 - %6 = call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx %5) #7 - store <256 x i32> %6, ptr %m, align 64 - - %7 = load <256 x i32>, ptr %m, align 64 - %8 = call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> %7) #7 - %9 = load <256 x i32>, ptr %m, align 64 - %10 = call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> %9) #7 - %11 = load <256 x i32>, ptr %m, align 64 - %12 = call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> %11) #7 - - %13 = call x86_amx @llvm.x86.tdpbssd.internal(i16 %row, i16 %col0, i16 %col1, x86_amx %8, x86_amx %10, x86_amx %12) #7 - %14 = call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx %13) #7 - store <256 x i32> %14, ptr %m, align 64 - - ret void - } - - ; Function Attrs: argmemonly nounwind readonly - declare { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0.internal(i16, i16, i16, ptr, i64) #2 - - ; Function Attrs: nounwind readnone - declare <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx) #3 - - ; Function Attrs: nounwind - declare x86_amx @llvm.x86.tilezero.internal(i16, i16) #4 - - ; Function Attrs: nounwind - declare x86_amx @llvm.x86.tdpbssd.internal(i16, i16, i16, x86_amx, x86_amx, x86_amx) #4 - - ; Function Attrs: nounwind readnone - declare x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32>) #3 - - ; Function Attrs: argmemonly nounwind writeonly - declare void @llvm.x86.tilestored64.internal(i16, i16, ptr, i64, x86_amx) #5 - - attributes #0 = { noinline nounwind optnone uwtable "frame-pointer"="all" "min-legal-vector-width"="8192" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+amx-bf16,+amx-int8,+amx-tile,+amx-transpose,+avx,+avx2,+avx512f,+crc32,+cx8,+f16c,+fma,+fxsr,+mmx,+popcnt,+sse,+sse2,+sse3,+sse4.1,+sse4.2,+ssse3,+x87,+xsave,+amx-tile,+amx-bf16,+avx512f,+amx-transpose" "tune-cpu"="generic" } - attributes #1 = { argmemonly nofree nounwind willreturn writeonly "target-features"="+amx-tile,+amx-bf16,+avx512f,+amx-transpose" } - attributes #2 = { argmemonly nounwind readonly "target-features"="+amx-tile,+amx-bf16,+avx512f,+amx-transpose" } - attributes #3 = { nounwind readnone "target-features"="+amx-tile,+amx-bf16,+avx512f,+amx-transpose" } - attributes #4 = { nounwind "target-features"="+amx-tile,+amx-bf16,+avx512f,+amx-transpose" } - attributes #5 = { argmemonly nounwind writeonly "target-features"="+amx-tile,+amx-bf16,+avx512f,+amx-transpose" } - attributes #6 = { argmemonly nofree nounwind willreturn "target-features"="+amx-tile,+amx-bf16,+avx512f,+amx-transpose" } - attributes #7 = { nounwind } diff --git a/llvm/test/CodeGen/X86/amx_tile_pair_lower_type_O2.ll b/llvm/test/CodeGen/X86/amx_tile_pair_lower_type_O2.ll deleted file mode 100644 index 96966264..0000000 --- a/llvm/test/CodeGen/X86/amx_tile_pair_lower_type_O2.ll +++ /dev/null @@ -1,61 +0,0 @@ -; NOTE: Assertions have been autogenerated by utils/update_test_checks.py -; RUN: opt --codegen-opt-level=2 -mtriple=x86_64 -x86-lower-amx-type %s -S | FileCheck %s -; RUN: opt --codegen-opt-level=2 -mtriple=x86_64 -passes=x86-lower-amx-type %s -S | FileCheck %s - - @buf = dso_local global [2048 x i8] zeroinitializer, align 16 - @buf2 = dso_local global [2048 x i8] zeroinitializer, align 16 - - ; Function Attrs: nounwind uwtable - define dso_local void @test_tile_2rpntlvwz0(i16 noundef signext %row, i16 noundef signext %col0, i16 noundef signext %col1) local_unnamed_addr #0 { -; CHECK-LABEL: @test_tile_2rpntlvwz0( -; CHECK-NEXT: entry: -; CHECK-NEXT: [[TMP0:%.*]] = tail call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0.internal(i16 [[ROW:%.*]], i16 [[COL0:%.*]], i16 [[COL1:%.*]], ptr @buf, i64 32) #[[ATTR3:[0-9]+]] -; CHECK-NEXT: [[TMP1:%.*]] = extractvalue { x86_amx, x86_amx } [[TMP0]], 0 -; CHECK-NEXT: [[TMP2:%.*]] = extractvalue { x86_amx, x86_amx } [[TMP0]], 1 -; CHECK-NEXT: [[TMP3:%.*]] = tail call x86_amx @llvm.x86.tilezero.internal(i16 [[ROW]], i16 [[COL0]]) #[[ATTR3]] -; CHECK-NEXT: [[TMP4:%.*]] = tail call x86_amx @llvm.x86.tdpbssd.internal(i16 [[ROW]], i16 [[COL1]], i16 [[COL0]], x86_amx [[TMP3]], x86_amx [[TMP1]], x86_amx [[TMP2]]) #[[ATTR3]] -; CHECK-NEXT: tail call void @llvm.x86.tilestored64.internal(i16 [[ROW]], i16 [[COL0]], ptr @buf2, i64 32, x86_amx [[TMP4]]) #[[ATTR3]] -; CHECK-NEXT: ret void -; - entry: - %0 = tail call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0.internal(i16 %row, i16 %col0, i16 %col1, ptr @buf, i64 32) #5 - %1 = extractvalue { x86_amx, x86_amx } %0, 0 - %2 = tail call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx %1) #5 - %3 = extractvalue { x86_amx, x86_amx } %0, 1 - %4 = tail call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx %3) #5 - %5 = tail call x86_amx @llvm.x86.tilezero.internal(i16 %row, i16 %col0) #5 - %6 = tail call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx %5) #5 - %7 = tail call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> %6) #5 - %8 = tail call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> %2) #5 - %9 = tail call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> %4) #5 - %10 = tail call x86_amx @llvm.x86.tdpbssd.internal(i16 %row, i16 %col1, i16 %col0, x86_amx %7, x86_amx %8, x86_amx %9) #5 - %11 = tail call <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx %10) #5 - %12 = tail call x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32> %11) #5 - tail call void @llvm.x86.tilestored64.internal(i16 %row, i16 %col0, ptr @buf2, i64 32, x86_amx %12) #5 - ret void - } - - ; Function Attrs: argmemonly nounwind readonly - declare { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0.internal(i16, i16, i16, ptr, i64) #1 - - ; Function Attrs: nounwind readnone - declare <256 x i32> @llvm.x86.cast.tile.to.vector.v256i32(x86_amx) #2 - - ; Function Attrs: nounwind - declare x86_amx @llvm.x86.tilezero.internal(i16, i16) #3 - - ; Function Attrs: nounwind - declare x86_amx @llvm.x86.tdpbssd.internal(i16, i16, i16, x86_amx, x86_amx, x86_amx) #3 - - ; Function Attrs: nounwind readnone - declare x86_amx @llvm.x86.cast.vector.to.tile.v256i32(<256 x i32>) #2 - - ; Function Attrs: argmemonly nounwind writeonly - declare void @llvm.x86.tilestored64.internal(i16, i16, ptr, i64, x86_amx) #4 - - attributes #0 = { nounwind uwtable "frame-pointer"="all" "min-legal-vector-width"="8192" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+amx-bf16,+amx-int8,+amx-tile,+amx-transpose,+avx,+avx2,+avx512f,+crc32,+cx8,+f16c,+fma,+fxsr,+mmx,+popcnt,+sse,+sse2,+sse3,+sse4.1,+sse4.2,+ssse3,+x87,+xsave,+amx-tile,+amx-bf16,+avx512f,+amx-transpose" "tune-cpu"="generic" } - attributes #1 = { argmemonly nounwind readonly "target-features"="+amx-tile,+amx-bf16,+avx512f,+amx-transpose" } - attributes #2 = { nounwind readnone "target-features"="+amx-tile,+amx-bf16,+avx512f,+amx-transpose" } - attributes #3 = { nounwind "target-features"="+amx-tile,+amx-bf16,+avx512f,+amx-transpose" } - attributes #4 = { argmemonly nounwind writeonly "target-features"="+amx-tile,+amx-bf16,+avx512f,+amx-transpose" } - attributes #5 = { nounwind } diff --git a/llvm/test/CodeGen/X86/amx_tile_pair_preconfigure_O0.mir b/llvm/test/CodeGen/X86/amx_tile_pair_preconfigure_O0.mir deleted file mode 100644 index 1e3b242..0000000 --- a/llvm/test/CodeGen/X86/amx_tile_pair_preconfigure_O0.mir +++ /dev/null @@ -1,134 +0,0 @@ -# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py -# RUN: llc -O0 -mtriple=x86_64-unknown-unknown -mattr=+amx-tile,+amx-bf16,+avx512f, \ -# RUN: -mattr=+amx-transpose -run-pass=fastpretileconfig -o - %s | FileCheck %s - ---- -name: test_tile_2rpntlvwz0 -alignment: 16 -exposesReturnsTwice: false -legalized: false -regBankSelected: false -selected: false -failedISel: false -tracksRegLiveness: true -hasWinCFI: false -callsEHReturn: false -callsUnwindInit: false -hasEHContTarget: false -hasEHScopes: false -hasEHFunclets: false -failsVerification: false -tracksDebugUserValues: false -registers: - - { id: 0, class: gr64_nosp, preferred-register: '' } - - { id: 1, class: gr16, preferred-register: '' } - - { id: 2, class: gr16, preferred-register: '' } - - { id: 3, class: gr16, preferred-register: '' } - - { id: 4, class: gr64, preferred-register: '' } - - { id: 5, class: gr64, preferred-register: '' } - - { id: 6, class: gr64, preferred-register: '' } - - { id: 7, class: gr64_nosp, preferred-register: '' } - - { id: 8, class: tilepair, preferred-register: '' } - - { id: 9, class: tile, preferred-register: '' } - - { id: 10, class: tile, preferred-register: '' } - - { id: 11, class: tile, preferred-register: '' } - - { id: 181, class: tile, preferred-register: '' } - - { id: 183, class: tile, preferred-register: '' } - - { id: 185, class: tile, preferred-register: '' } - - { id: 186, class: tile, preferred-register: '' } -liveins: - - { reg: '$edi', virtual-reg: '%0' } - - { reg: '$esi', virtual-reg: '%1' } - - { reg: '$edx', virtual-reg: '%2' } -frameInfo: - isFrameAddressTaken: false - isReturnAddressTaken: false - hasStackMap: false - hasPatchPoint: false - stackSize: 0 - offsetAdjustment: 0 - maxAlignment: 1024 - adjustsStack: false - hasCalls: true - stackProtector: '' - functionContext: '' - maxCallFrameSize: 4294967295 - cvBytesOfCalleeSavedRegisters: 0 - hasOpaqueSPAdjustment: false - hasVAStart: false - hasMustTailInVarArgFunc: false - hasTailCall: false - localFrameSize: 0 - savePoint: [] - restorePoint: [] -fixedStack: [] -stack: - - { id: 18, name: '', type: default, offset: 0, size: 8, alignment: 8, - stack-id: default, callee-saved-register: '', callee-saved-restored: true, - debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } - - { id: 19, name: '', type: default, offset: 0, size: 8, alignment: 8, - stack-id: default, callee-saved-register: '', callee-saved-restored: true, - debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } - - { id: 20, name: '', type: default, offset: 0, size: 8, alignment: 8, - stack-id: default, callee-saved-register: '', callee-saved-restored: true, - debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } - - { id: 21, name: '', type: default, offset: 0, size: 8, - alignment: 8, stack-id: default, callee-saved-register: '', callee-saved-restored: true, - debug-info-variable: '', debug-info-expression: '', debug-info-location: '' } -callSites: [] -debugValueSubstitutions: [] -constants: [] -machineFunctionInfo: - amxProgModel: ManagedRA -body: | - bb.0.entry: - liveins: $rdi, $rsi, $rdx, $rax - - ; CHECK-LABEL: name: test_tile_2rpntlvwz0 - ; CHECK: liveins: $rdi, $rsi, $rdx, $rax - ; CHECK-NEXT: {{ $}} - ; CHECK-NEXT: [[AVX512_512_SET0_:%[0-9]+]]:vr512 = AVX512_512_SET0 - ; CHECK-NEXT: VMOVUPSZmr %stack.4, 1, $noreg, 0, $noreg, [[AVX512_512_SET0_]] :: (store (s512) into %stack.4, align 4) - ; CHECK-NEXT: MOV8mi %stack.4, 1, $noreg, 0, $noreg, 1 :: (store (s512) into %stack.4, align 4) - ; CHECK-NEXT: [[MOV32ri64_:%[0-9]+]]:gr64_nosp = MOV32ri64 64 - ; CHECK-NEXT: [[MOV16ri:%[0-9]+]]:gr16 = MOV16ri 64 - ; CHECK-NEXT: [[MOV16ri1:%[0-9]+]]:gr16 = MOV16ri 16 - ; CHECK-NEXT: [[MOV16ri2:%[0-9]+]]:gr16 = MOV16ri 16 - ; CHECK-NEXT: PLDTILECFGV %stack.4, 1, $noreg, 0, $noreg, implicit-def $tmm0, implicit-def $tmm1, implicit-def $tmm2, implicit-def $tmm3, implicit-def $tmm4, implicit-def $tmm5, implicit-def $tmm6, implicit-def $tmm7 :: (load (s512) from %stack.4, align 4) - ; CHECK-NEXT: [[COPY:%[0-9]+]]:gr64 = COPY $rsi - ; CHECK-NEXT: [[COPY1:%[0-9]+]]:gr64 = COPY $rdi - ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gr64 = COPY $rdx - ; CHECK-NEXT: [[COPY3:%[0-9]+]]:gr64_nosp = COPY $rax - ; CHECK-NEXT: [[PT2RPNTLVWZ0V:%[0-9]+]]:tilepair = PT2RPNTLVWZ0V [[MOV16ri]], [[MOV16ri1]], [[MOV16ri2]], [[COPY2]], 1, killed [[COPY3]], 0, $noreg - ; CHECK-NEXT: [[COPY4:%[0-9]+]]:tile = COPY [[PT2RPNTLVWZ0V]].sub_t1 - ; CHECK-NEXT: [[COPY5:%[0-9]+]]:tile = COPY [[PT2RPNTLVWZ0V]].sub_t0 - ; CHECK-NEXT: PTILESTOREDV [[MOV16ri]], [[MOV16ri1]], [[COPY]], 1, [[MOV32ri64_]], 0, $noreg, killed [[COPY5]] - ; CHECK-NEXT: PTILESTOREDV [[MOV16ri]], [[MOV16ri2]], [[COPY1]], 1, [[MOV32ri64_]], 0, $noreg, killed [[COPY4]] - ; CHECK-NEXT: [[PTILEZEROV:%[0-9]+]]:tile = PTILEZEROV [[MOV16ri]], [[MOV16ri1]] - ; CHECK-NEXT: PTILESTOREDV [[MOV16ri]], [[MOV16ri1]], [[COPY2]], 1, [[MOV32ri64_]], 0, $noreg, killed [[PTILEZEROV]] - ; CHECK-NEXT: [[PTILELOADDV:%[0-9]+]]:tile = PTILELOADDV [[MOV16ri]], [[MOV16ri1]], [[COPY]], 1, [[MOV32ri64_]], 0, $noreg - ; CHECK-NEXT: [[PTILELOADDV1:%[0-9]+]]:tile = PTILELOADDV [[MOV16ri]], [[MOV16ri2]], [[COPY1]], 1, [[MOV32ri64_]], 0, $noreg - ; CHECK-NEXT: [[PTILELOADDV2:%[0-9]+]]:tile = PTILELOADDV [[MOV16ri]], [[MOV16ri1]], [[COPY2]], 1, [[MOV32ri64_]], 0, $noreg - ; CHECK-NEXT: [[PTDPBSSDV:%[0-9]+]]:tile = PTDPBSSDV [[MOV16ri]], [[MOV16ri1]], [[MOV16ri2]], [[PTILELOADDV]], killed [[PTILELOADDV1]], killed [[PTILELOADDV2]] - ; CHECK-NEXT: PTILESTOREDV killed [[MOV16ri]], killed [[MOV16ri1]], killed [[COPY2]], 1, killed [[MOV32ri64_]], 0, $noreg, killed [[PTDPBSSDV]] - %0:gr64_nosp = MOV32ri64 64 - %1:gr16 = MOV16ri 64 - %2:gr16 = MOV16ri 16 - %3:gr16 = MOV16ri 16 - %4:gr64 = COPY $rsi - %5:gr64 = COPY $rdi - %6:gr64 = COPY $rdx - %7:gr64_nosp = COPY $rax - %8:tilepair = PT2RPNTLVWZ0V %1, %2, %3, %6, 1, killed %7, 0, $noreg - %9:tile = COPY %8.sub_t1 - %10:tile = COPY %8.sub_t0 - PTILESTOREDV %1, %2, %4, 1, %0, 0, $noreg, killed %10 - PTILESTOREDV %1, %3, %5, 1, %0, 0, $noreg, killed %9 - %11:tile = PTILEZEROV %1, %2 - PTILESTOREDV %1, %2, %6, 1, %0, 0, $noreg, killed %11 - %181:tile = PTILELOADDV %1, %2, %4, 1, %0, 0, $noreg - %183:tile = PTILELOADDV %1, %3, %5, 1, %0, 0, $noreg - %185:tile = PTILELOADDV %1, %2, %6, 1, %0, 0, $noreg - %186:tile = PTDPBSSDV %1, %2, %3, %181, killed %183, killed %185 - PTILESTOREDV killed %1, killed %2, killed %6, 1, killed %0, 0, $noreg, killed %186 -... diff --git a/llvm/test/CodeGen/X86/amx_tile_pair_preconfigure_O2.mir b/llvm/test/CodeGen/X86/amx_tile_pair_preconfigure_O2.mir deleted file mode 100644 index ac2cdb4..0000000 --- a/llvm/test/CodeGen/X86/amx_tile_pair_preconfigure_O2.mir +++ /dev/null @@ -1,113 +0,0 @@ -# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py -# RUN: llc -O2 -mtriple=x86_64-unknown-unknown -mattr=+amx-tile,+amx-bf16,+avx512f, \ -# RUN: -mattr=+amx-transpose -run-pass=tilepreconfig -o - %s | FileCheck %s - ---- -name: test_tile_2rpntlvwz0 -alignment: 16 -exposesReturnsTwice: false -legalized: false -regBankSelected: false -selected: false -failedISel: false -tracksRegLiveness: true -hasWinCFI: false -callsEHReturn: false -callsUnwindInit: false -hasEHContTarget: false -hasEHScopes: false -hasEHFunclets: false -failsVerification: false -tracksDebugUserValues: false -registers: - - { id: 0, class: gr32, preferred-register: '' } - - { id: 1, class: gr32, preferred-register: '' } - - { id: 2, class: gr32, preferred-register: '' } - - { id: 3, class: gr16, preferred-register: '' } - - { id: 4, class: gr16, preferred-register: '' } - - { id: 5, class: gr16, preferred-register: '' } - - { id: 6, class: gr64, preferred-register: '' } - - { id: 7, class: gr64_nosp, preferred-register: '' } - - { id: 8, class: tilepair, preferred-register: '' } - - { id: 9, class: tile, preferred-register: '' } - - { id: 10, class: tile, preferred-register: '' } - - { id: 11, class: tile, preferred-register: '' } - - { id: 12, class: tile, preferred-register: '' } - - { id: 13, class: gr64, preferred-register: '' } -liveins: - - { reg: '$edi', virtual-reg: '%0' } - - { reg: '$esi', virtual-reg: '%1' } - - { reg: '$edx', virtual-reg: '%2' } -frameInfo: - isFrameAddressTaken: false - isReturnAddressTaken: false - hasStackMap: false - hasPatchPoint: false - stackSize: 0 - offsetAdjustment: 0 - maxAlignment: 1 - adjustsStack: false - hasCalls: false - stackProtector: '' - functionContext: '' - maxCallFrameSize: 4294967295 - cvBytesOfCalleeSavedRegisters: 0 - hasOpaqueSPAdjustment: false - hasVAStart: false - hasMustTailInVarArgFunc: false - hasTailCall: false - localFrameSize: 0 - savePoint: [] - restorePoint: [] -fixedStack: [] -stack: [] -callSites: [] -debugValueSubstitutions: [] -constants: [] -machineFunctionInfo: - amxProgModel: ManagedRA -body: | - bb.0.entry: - liveins: $edi, $esi, $edx, $rax, $rbx - - ; CHECK-LABEL: name: test_tile_2rpntlvwz0 - ; CHECK: liveins: $edi, $esi, $edx, $rax, $rbx - ; CHECK-NEXT: {{ $}} - ; CHECK-NEXT: [[AVX512_512_SET0_:%[0-9]+]]:vr512 = AVX512_512_SET0 - ; CHECK-NEXT: VMOVUPSZmr %stack.0, 1, $noreg, 0, $noreg, [[AVX512_512_SET0_]] :: (store (s512) into %stack.0, align 4) - ; CHECK-NEXT: MOV8mi %stack.0, 1, $noreg, 0, $noreg, 1 :: (store (s512) into %stack.0, align 4) - ; CHECK-NEXT: [[COPY:%[0-9]+]]:gr32 = COPY $edx - ; CHECK-NEXT: [[COPY1:%[0-9]+]]:gr32 = COPY $esi - ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gr32 = COPY $edi - ; CHECK-NEXT: [[COPY3:%[0-9]+]]:gr16 = COPY [[COPY]].sub_16bit - ; CHECK-NEXT: [[COPY4:%[0-9]+]]:gr16 = COPY [[COPY1]].sub_16bit - ; CHECK-NEXT: [[COPY5:%[0-9]+]]:gr16 = COPY [[COPY2]].sub_16bit - ; CHECK-NEXT: PLDTILECFGV %stack.0, 1, $noreg, 0, $noreg, implicit-def $tmm0, implicit-def $tmm1, implicit-def $tmm2, implicit-def $tmm3, implicit-def $tmm4, implicit-def $tmm5, implicit-def $tmm6, implicit-def $tmm7 :: (load (s512) from %stack.0, align 4) - ; CHECK-NEXT: [[COPY6:%[0-9]+]]:gr64 = COPY $rax - ; CHECK-NEXT: [[MOV32ri64_:%[0-9]+]]:gr64_nosp = MOV32ri64 32 - ; CHECK-NEXT: [[PT2RPNTLVWZ0V:%[0-9]+]]:tilepair = PT2RPNTLVWZ0V [[COPY5]], [[COPY4]], [[COPY3]], killed [[COPY6]], 1, [[MOV32ri64_]], 0, $noreg - ; CHECK-NEXT: [[COPY7:%[0-9]+]]:tile = COPY [[PT2RPNTLVWZ0V]].sub_t1 - ; CHECK-NEXT: [[COPY8:%[0-9]+]]:tile = COPY [[PT2RPNTLVWZ0V]].sub_t0 - ; CHECK-NEXT: [[PTILEZEROV:%[0-9]+]]:tile = PTILEZEROV [[COPY5]], [[COPY4]] - ; CHECK-NEXT: [[PTDPBSSDV:%[0-9]+]]:tile = PTDPBSSDV [[COPY5]], [[COPY3]], [[COPY4]], [[PTILEZEROV]], killed [[COPY8]], killed [[COPY7]] - ; CHECK-NEXT: [[COPY9:%[0-9]+]]:gr64 = COPY $rbx - ; CHECK-NEXT: PTILESTOREDV [[COPY5]], [[COPY4]], killed [[COPY9]], 1, [[MOV32ri64_]], 0, $noreg, killed [[PTDPBSSDV]] - ; CHECK-NEXT: RET 0 - %2:gr32 = COPY $edx - %1:gr32 = COPY $esi - %0:gr32 = COPY $edi - %3:gr16 = COPY %2.sub_16bit - %4:gr16 = COPY %1.sub_16bit - %5:gr16 = COPY %0.sub_16bit - %6:gr64 = COPY $rax - %7:gr64_nosp = MOV32ri64 32 - %8:tilepair = PT2RPNTLVWZ0V %5, %4, %3, killed %6, 1, %7, 0, $noreg - %9:tile = COPY %8.sub_t1 - %10:tile = COPY %8.sub_t0 - %11:tile = PTILEZEROV %5, %4 - %12:tile = PTDPBSSDV %5, %3, %4, %11, killed %10, killed %9 - %13:gr64 = COPY $rbx - PTILESTOREDV %5, %4, killed %13, 1, %7, 0, $noreg, killed %12 - RET 0 - -... diff --git a/llvm/test/CodeGen/X86/amx_transpose_intrinsics.ll b/llvm/test/CodeGen/X86/amx_transpose_intrinsics.ll deleted file mode 100644 index 4cfd97a..0000000 --- a/llvm/test/CodeGen/X86/amx_transpose_intrinsics.ll +++ /dev/null @@ -1,371 +0,0 @@ -; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+amx-bf16,+amx-fp16,+amx-complex,+amx-transpose | FileCheck %s -; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+amx-bf16,+amx-fp16,+amx-complex,+amx-transpose,+egpr --show-mc-encoding | FileCheck %s --check-prefix=EGPR - -define void @test_amx(i32 %rv32, i64 %stride, i64 %rvalue, i8* %addr1, <4 x float> %xmm) #0 { -; CHECK-LABEL: test_amx: -; CHECK: # %bb.0: -; CHECK-NEXT: t2rpntlvwz0 (%rcx,%rsi), %tmm0 -; CHECK-NEXT: t2rpntlvwz0t1 (%rcx,%rsi), %tmm2 -; CHECK-NEXT: t2rpntlvwz1 (%rcx,%rsi), %tmm0 -; CHECK-NEXT: t2rpntlvwz1t1 (%rcx,%rsi), %tmm2 -; CHECK-NEXT: ttransposed %tmm3, %tmm1 -; CHECK-NEXT: ttdpbf16ps %tmm3, %tmm2, %tmm1 -; CHECK-NEXT: ttdpfp16ps %tmm6, %tmm5, %tmm4 -; CHECK-NEXT: ttcmmimfp16ps %tmm3, %tmm2, %tmm1 -; CHECK-NEXT: ttcmmrlfp16ps %tmm3, %tmm2, %tmm1 -; CHECK-NEXT: tconjtcmmimfp16ps %tmm3, %tmm2, %tmm1 -; CHECK-NEXT: tconjtfp16 %tmm2, %tmm1 -; CHECK-NEXT: retq -; -; EGPR-LABEL: test_amx: -; EGPR: # %bb.0: -; EGPR-NEXT: t2rpntlvwz0 (%rcx,%rsi), %tmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0x6e,0x04,0x31] -; EGPR-NEXT: t2rpntlvwz0t1 (%rcx,%rsi), %tmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0x6f,0x14,0x31] -; EGPR-NEXT: t2rpntlvwz1 (%rcx,%rsi), %tmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x6e,0x04,0x31] -; EGPR-NEXT: t2rpntlvwz1t1 (%rcx,%rsi), %tmm2 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x6f,0x14,0x31] -; EGPR-NEXT: ttransposed %tmm3, %tmm1 # encoding: [0xc4,0xe2,0x7a,0x5f,0xcb] -; EGPR-NEXT: ttdpbf16ps %tmm3, %tmm2, %tmm1 # encoding: [0xc4,0xe2,0x62,0x6c,0xca] -; EGPR-NEXT: ttdpfp16ps %tmm6, %tmm5, %tmm4 # encoding: [0xc4,0xe2,0x4b,0x6c,0xe5] -; EGPR-NEXT: ttcmmimfp16ps %tmm3, %tmm2, %tmm1 # encoding: [0xc4,0xe2,0x63,0x6b,0xca] -; EGPR-NEXT: ttcmmrlfp16ps %tmm3, %tmm2, %tmm1 # encoding: [0xc4,0xe2,0x62,0x6b,0xca] -; EGPR-NEXT: tconjtcmmimfp16ps %tmm3, %tmm2, %tmm1 # encoding: [0xc4,0xe2,0x60,0x6b,0xca] -; EGPR-NEXT: tconjtfp16 %tmm2, %tmm1 # encoding: [0xc4,0xe2,0x79,0x6b,0xca] -; EGPR-NEXT: retq # encoding: [0xc3] - call void @llvm.x86.t2rpntlvwz0(i8 1, i8* %addr1, i64 %stride) - call void @llvm.x86.t2rpntlvwz0t1(i8 2, i8* %addr1, i64 %stride) - call void @llvm.x86.t2rpntlvwz1(i8 1, i8* %addr1, i64 %stride) - call void @llvm.x86.t2rpntlvwz1t1(i8 2, i8* %addr1, i64 %stride) - call void @llvm.x86.ttransposed(i8 1, i8 3) - call void @llvm.x86.ttdpbf16ps(i8 1, i8 2, i8 3) - call void @llvm.x86.ttdpfp16ps(i8 4, i8 5, i8 6) - call void @llvm.x86.ttcmmimfp16ps(i8 1, i8 2, i8 3) - call void @llvm.x86.ttcmmrlfp16ps(i8 1, i8 2, i8 3) - call void @llvm.x86.tconjtcmmimfp16ps(i8 1, i8 2, i8 3) - call void @llvm.x86.tconjtfp16(i8 1, i8 2) - ret void -} - -declare void @llvm.x86.t2rpntlvwz0(i8 %tile1, i8* %addr1, i64 %stride) -declare void @llvm.x86.t2rpntlvwz0t1(i8 %tile1, i8* %addr1, i64 %stride) -declare void @llvm.x86.t2rpntlvwz1(i8 %tile1, i8* %addr1, i64 %stride) -declare void @llvm.x86.t2rpntlvwz1t1(i8 %tile1, i8* %addr1, i64 %stride) -declare void @llvm.x86.ttransposed(i8 %tile0, i8 %tile1) -declare void @llvm.x86.ttdpbf16ps(i8 %tile0, i8 %tile1, i8 %tile2) -declare void @llvm.x86.ttdpfp16ps(i8 %tile0, i8 %tile1, i8 %tile2) -declare void @llvm.x86.ttcmmimfp16ps(i8 %A, i8 %B, i8 %C) -declare void @llvm.x86.ttcmmrlfp16ps(i8 %A, i8 %B, i8 %C) -declare void @llvm.x86.tconjtcmmimfp16ps(i8 %A, i8 %B, i8 %C) -declare void @llvm.x86.tconjtfp16(i8 %A, i8 %B) - -define void @test_amx2(i8* %pointer, i8* %base, i64 %stride) #0 { -; CHECK-LABEL: test_amx2: -; CHECK: # %bb.0: -; CHECK-NEXT: pushq %rbp -; CHECK-NEXT: subq $2928, %rsp # imm = 0xB70 -; CHECK-NEXT: vxorps %xmm0, %xmm0, %xmm0 -; CHECK-NEXT: vmovups %zmm0, {{[0-9]+}}(%rsp) -; CHECK-NEXT: movb $1, {{[0-9]+}}(%rsp) -; CHECK-NEXT: movb $8, {{[0-9]+}}(%rsp) -; CHECK-NEXT: movw $8, {{[0-9]+}}(%rsp) -; CHECK-NEXT: movb $8, {{[0-9]+}}(%rsp) -; CHECK-NEXT: movw $8, {{[0-9]+}}(%rsp) -; CHECK-NEXT: movb $8, {{[0-9]+}}(%rsp) -; CHECK-NEXT: movw $8, {{[0-9]+}}(%rsp) -; CHECK-NEXT: movb $8, {{[0-9]+}}(%rsp) -; CHECK-NEXT: movw $8, {{[0-9]+}}(%rsp) -; CHECK-NEXT: ldtilecfg {{[0-9]+}}(%rsp) -; CHECK-NEXT: movw $8, %ax -; CHECK-NEXT: tileloadd (%rsi,%rdx), %tmm0 -; CHECK-NEXT: tilezero %tmm1 -; CHECK-NEXT: tilezero %tmm2 -; CHECK-NEXT: ttdpbf16ps %tmm1, %tmm0, %tmm2 -; CHECK-NEXT: ttdpfp16ps %tmm1, %tmm0, %tmm2 -; CHECK-NEXT: ttcmmimfp16ps %tmm1, %tmm0, %tmm2 -; CHECK-NEXT: ttcmmrlfp16ps %tmm1, %tmm0, %tmm2 -; CHECK-NEXT: movabsq $64, %rbp -; CHECK-NEXT: tilestored %tmm2, 896(%rsp,%rbp) # 1024-byte Folded Spill -; CHECK-NEXT: tileloadd 896(%rsp,%rbp), %tmm3 # 1024-byte Folded Reload -; CHECK-NEXT: tconjtcmmimfp16ps %tmm1, %tmm0, %tmm3 -; CHECK-NEXT: tconjtfp16 %tmm3, %tmm0 -; CHECK-NEXT: tilestored %tmm2, (%rdi,%rdx) -; CHECK-NEXT: addq $2928, %rsp # imm = 0xB70 -; CHECK-NEXT: popq %rbp -; CHECK-NEXT: tilerelease -; CHECK-NEXT: vzeroupper -; CHECK-NEXT: retq -; -; EGPR-LABEL: test_amx2: -; EGPR: # %bb.0: -; EGPR-NEXT: pushq %rbp # encoding: [0x55] -; EGPR-NEXT: subq $2928, %rsp # encoding: [0x48,0x81,0xec,0x70,0x0b,0x00,0x00] -; EGPR-NEXT: # imm = 0xB70 -; EGPR-NEXT: vxorps %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0x57,0xc0] -; EGPR-NEXT: vmovups %zmm0, {{[0-9]+}}(%rsp) # encoding: [0x62,0xf1,0x7c,0x48,0x11,0x44,0x24,0x0d] -; EGPR-NEXT: movb $1, {{[0-9]+}}(%rsp) # encoding: [0xc6,0x84,0x24,0x40,0x03,0x00,0x00,0x01] -; EGPR-NEXT: movb $8, {{[0-9]+}}(%rsp) # encoding: [0xc6,0x84,0x24,0x70,0x03,0x00,0x00,0x08] -; EGPR-NEXT: movw $8, {{[0-9]+}}(%rsp) # encoding: [0x66,0xc7,0x84,0x24,0x50,0x03,0x00,0x00,0x08,0x00] -; EGPR-NEXT: movb $8, {{[0-9]+}}(%rsp) # encoding: [0xc6,0x84,0x24,0x71,0x03,0x00,0x00,0x08] -; EGPR-NEXT: movw $8, {{[0-9]+}}(%rsp) # encoding: [0x66,0xc7,0x84,0x24,0x52,0x03,0x00,0x00,0x08,0x00] -; EGPR-NEXT: movb $8, {{[0-9]+}}(%rsp) # encoding: [0xc6,0x84,0x24,0x72,0x03,0x00,0x00,0x08] -; EGPR-NEXT: movw $8, {{[0-9]+}}(%rsp) # encoding: [0x66,0xc7,0x84,0x24,0x54,0x03,0x00,0x00,0x08,0x00] -; EGPR-NEXT: movb $8, {{[0-9]+}}(%rsp) # encoding: [0xc6,0x84,0x24,0x73,0x03,0x00,0x00,0x08] -; EGPR-NEXT: movw $8, {{[0-9]+}}(%rsp) # encoding: [0x66,0xc7,0x84,0x24,0x56,0x03,0x00,0x00,0x08,0x00] -; EGPR-NEXT: ldtilecfg {{[0-9]+}}(%rsp) # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0x49,0x84,0x24,0x40,0x03,0x00,0x00] -; EGPR-NEXT: movw $8, %ax # encoding: [0x66,0xb8,0x08,0x00] -; EGPR-NEXT: tileloadd (%rsi,%rdx), %tmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7b,0x4b,0x04,0x16] -; EGPR-NEXT: tilezero %tmm1 # encoding: [0xc4,0xe2,0x7b,0x49,0xc8] -; EGPR-NEXT: tilezero %tmm2 # encoding: [0xc4,0xe2,0x7b,0x49,0xd0] -; EGPR-NEXT: ttdpbf16ps %tmm1, %tmm0, %tmm2 # encoding: [0xc4,0xe2,0x72,0x6c,0xd0] -; EGPR-NEXT: ttdpfp16ps %tmm1, %tmm0, %tmm2 # encoding: [0xc4,0xe2,0x73,0x6c,0xd0] -; EGPR-NEXT: ttcmmimfp16ps %tmm1, %tmm0, %tmm2 # encoding: [0xc4,0xe2,0x73,0x6b,0xd0] -; EGPR-NEXT: ttcmmrlfp16ps %tmm1, %tmm0, %tmm2 # encoding: [0xc4,0xe2,0x72,0x6b,0xd0] -; EGPR-NEXT: movabsq $64, %rbp # encoding: [0x48,0xbd,0x40,0x00,0x00,0x00,0x00,0x00,0x00,0x00] -; EGPR-NEXT: tilestored %tmm2, 896(%rsp,%rbp) # 1024-byte Folded Spill -; EGPR-NEXT: # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7a,0x4b,0x94,0x2c,0x80,0x03,0x00,0x00] -; EGPR-NEXT: tileloadd 896(%rsp,%rbp), %tmm3 # 1024-byte Folded Reload -; EGPR-NEXT: # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7b,0x4b,0x9c,0x2c,0x80,0x03,0x00,0x00] -; EGPR-NEXT: tconjtcmmimfp16ps %tmm1, %tmm0, %tmm3 # encoding: [0xc4,0xe2,0x70,0x6b,0xd8] -; EGPR-NEXT: tconjtfp16 %tmm3, %tmm0 # encoding: [0xc4,0xe2,0x79,0x6b,0xc3] -; EGPR-NEXT: tilestored %tmm2, (%rdi,%rdx) # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7a,0x4b,0x14,0x17] -; EGPR-NEXT: addq $2928, %rsp # encoding: [0x48,0x81,0xc4,0x70,0x0b,0x00,0x00] -; EGPR-NEXT: # imm = 0xB70 -; EGPR-NEXT: popq %rbp # encoding: [0x5d] -; EGPR-NEXT: tilerelease # encoding: [0xc4,0xe2,0x78,0x49,0xc0] -; EGPR-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77] -; EGPR-NEXT: retq # encoding: [0xc3] - - %a = call x86_amx @llvm.x86.tileloadd64.internal(i16 8, i16 8, i8* %base, i64 %stride) - %b = call x86_amx @llvm.x86.tilezero.internal(i16 8, i16 8) - %c = call x86_amx @llvm.x86.tilezero.internal(i16 8, i16 8) - %c1 = call x86_amx @llvm.x86.ttdpbf16ps.internal(i16 8, i16 8, i16 8, x86_amx %c, x86_amx %a, x86_amx %b) - %c2 = call x86_amx @llvm.x86.ttdpfp16ps.internal(i16 8, i16 8, i16 8, x86_amx %c1, x86_amx %a, x86_amx %b) - %c3 = call x86_amx @llvm.x86.ttcmmimfp16ps.internal(i16 8, i16 8, i16 8, x86_amx %c2, x86_amx %a, x86_amx %b) - %c4 = call x86_amx @llvm.x86.ttcmmrlfp16ps.internal(i16 8, i16 8, i16 8, x86_amx %c3, x86_amx %a, x86_amx %b) - %c5 = call x86_amx @llvm.x86.tconjtcmmimfp16ps.internal(i16 8, i16 8, i16 8, x86_amx %c4, x86_amx %a, x86_amx %b) - %c6 = call x86_amx @llvm.x86.tconjtfp16.internal(i16 8, i16 8, x86_amx %c5) - - call void @llvm.x86.tilestored64.internal(i16 8, i16 8, i8* %pointer, i64 %stride, x86_amx %c4) - ret void -} - -define void @test_amx3(i8* %pointer, i8* %base, i64 %stride) #0 { -; CHECK-LABEL: test_amx3: -; CHECK: # %bb.0: -; CHECK-NEXT: vxorps %xmm0, %xmm0, %xmm0 -; CHECK-NEXT: vmovups %zmm0, -{{[0-9]+}}(%rsp) -; CHECK-NEXT: movb $1, -{{[0-9]+}}(%rsp) -; CHECK-NEXT: movb $8, -{{[0-9]+}}(%rsp) -; CHECK-NEXT: movw $8, -{{[0-9]+}}(%rsp) -; CHECK-NEXT: movb $8, -{{[0-9]+}}(%rsp) -; CHECK-NEXT: movw $8, -{{[0-9]+}}(%rsp) -; CHECK-NEXT: movb $0, -{{[0-9]+}}(%rsp) -; CHECK-NEXT: movw $0, -{{[0-9]+}}(%rsp) -; CHECK-NEXT: ldtilecfg -{{[0-9]+}}(%rsp) -; CHECK-NEXT: xorl %eax, %eax -; CHECK-NEXT: movw $8, %cx -; CHECK-NEXT: t2rpntlvwz0 (%rsi,%rdx), %tmm4 -; CHECK-NEXT: t2rpntlvwz0t1 (%rsi,%rdx), %tmm4 -; CHECK-NEXT: t2rpntlvwz1 (%rsi,%rdx), %tmm4 -; CHECK-NEXT: t2rpntlvwz1t1 (%rsi,%rdx), %tmm4 -; CHECK-NEXT: ttransposed %tmm4, %tmm0 -; CHECK-NEXT: tilestored %tmm0, (%rdi,%rdx) -; CHECK-NEXT: tilerelease -; CHECK-NEXT: vzeroupper -; CHECK-NEXT: retq -; -; EGPR-LABEL: test_amx3: -; EGPR: # %bb.0: -; EGPR-NEXT: vxorps %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0x57,0xc0] -; EGPR-NEXT: vmovups %zmm0, -{{[0-9]+}}(%rsp) # encoding: [0x62,0xf1,0x7c,0x48,0x11,0x44,0x24,0xff] -; EGPR-NEXT: movb $1, -{{[0-9]+}}(%rsp) # encoding: [0xc6,0x44,0x24,0xc0,0x01] -; EGPR-NEXT: movb $8, -{{[0-9]+}}(%rsp) # encoding: [0xc6,0x44,0x24,0xf0,0x08] -; EGPR-NEXT: movw $8, -{{[0-9]+}}(%rsp) # encoding: [0x66,0xc7,0x44,0x24,0xd0,0x08,0x00] -; EGPR-NEXT: movb $8, -{{[0-9]+}}(%rsp) # encoding: [0xc6,0x44,0x24,0xf4,0x08] -; EGPR-NEXT: movw $8, -{{[0-9]+}}(%rsp) # encoding: [0x66,0xc7,0x44,0x24,0xd8,0x08,0x00] -; EGPR-NEXT: movb $0, -{{[0-9]+}}(%rsp) # encoding: [0xc6,0x44,0x24,0xf5,0x00] -; EGPR-NEXT: movw $0, -{{[0-9]+}}(%rsp) # encoding: [0x66,0xc7,0x44,0x24,0xda,0x00,0x00] -; EGPR-NEXT: ldtilecfg -{{[0-9]+}}(%rsp) # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0x49,0x44,0x24,0xc0] -; EGPR-NEXT: xorl %eax, %eax # encoding: [0x31,0xc0] -; EGPR-NEXT: movw $8, %cx # encoding: [0x66,0xb9,0x08,0x00] -; EGPR-NEXT: t2rpntlvwz0 (%rsi,%rdx), %tmm4 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0x6e,0x24,0x16] -; EGPR-NEXT: t2rpntlvwz0t1 (%rsi,%rdx), %tmm4 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0x6f,0x24,0x16] -; EGPR-NEXT: t2rpntlvwz1 (%rsi,%rdx), %tmm4 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x6e,0x24,0x16] -; EGPR-NEXT: t2rpntlvwz1t1 (%rsi,%rdx), %tmm4 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x6f,0x24,0x16] -; EGPR-NEXT: ttransposed %tmm4, %tmm0 # encoding: [0xc4,0xe2,0x7a,0x5f,0xc4] -; EGPR-NEXT: tilestored %tmm0, (%rdi,%rdx) # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7a,0x4b,0x04,0x17] -; EGPR-NEXT: tilerelease # encoding: [0xc4,0xe2,0x78,0x49,0xc0] -; EGPR-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77] -; EGPR-NEXT: retq # encoding: [0xc3] - %1 = call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0.internal(i16 8, i16 8, i16 0, i8* %base, i64 %stride) - %2 = call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0t1.internal(i16 8, i16 8, i16 0, i8* %base, i64 %stride) - %3 = call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz1.internal(i16 8, i16 8, i16 0, i8* %base, i64 %stride) - %4 = call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz1t1.internal(i16 8, i16 8, i16 0, i8* %base, i64 %stride) - %5 = extractvalue { x86_amx, x86_amx } %4, 0 - %6 = call x86_amx @llvm.x86.ttransposed.internal(i16 8, i16 8, x86_amx %5) - call void @llvm.x86.tilestored64.internal(i16 8, i16 8, i8* %pointer, i64 %stride, x86_amx %6) - ret void -} - -define void @test_amx_spill(i8* %pointer, i8* %base, i64 %stride) #0 { -; CHECK-LABEL: test_amx_spill: -; CHECK: # %bb.0: -; CHECK-NEXT: subq $6088, %rsp # imm = 0x17C8 -; CHECK-NEXT: vxorps %xmm0, %xmm0, %xmm0 -; CHECK-NEXT: vmovups %zmm0, -{{[0-9]+}}(%rsp) -; CHECK-NEXT: movb $1, -{{[0-9]+}}(%rsp) -; CHECK-NEXT: movb $8, -{{[0-9]+}}(%rsp) -; CHECK-NEXT: movw $8, -{{[0-9]+}}(%rsp) -; CHECK-NEXT: movb $8, -{{[0-9]+}}(%rsp) -; CHECK-NEXT: movw $8, -{{[0-9]+}}(%rsp) -; CHECK-NEXT: movb $8, -{{[0-9]+}}(%rsp) -; CHECK-NEXT: movw $8, -{{[0-9]+}}(%rsp) -; CHECK-NEXT: movb $8, -{{[0-9]+}}(%rsp) -; CHECK-NEXT: movw $8, -{{[0-9]+}}(%rsp) -; CHECK-NEXT: movb $8, -{{[0-9]+}}(%rsp) -; CHECK-NEXT: movw $8, -{{[0-9]+}}(%rsp) -; CHECK-NEXT: ldtilecfg -{{[0-9]+}}(%rsp) -; CHECK-NEXT: movw $8, %ax -; CHECK-NEXT: tileloadd (%rsi,%rdx), %tmm0 -; CHECK-NEXT: t2rpntlvwz0 (%rsi,%rdx), %tmm4 -; CHECK-NEXT: t2rpntlvwz0t1 (%rsi,%rdx), %tmm6 -; CHECK-NEXT: movabsq $64, %rcx -; CHECK-NEXT: tilestored %tmm6, 4032(%rsp,%rcx) # 1024-byte Folded Spill -; CHECK-NEXT: tilestored %tmm7, 5056(%rsp,%rcx) # 1024-byte Folded Spill -; CHECK-NEXT: t2rpntlvwz1 (%rsi,%rdx), %tmm6 -; CHECK-NEXT: tilestored %tmm6, 1984(%rsp,%rcx) # 1024-byte Folded Spill -; CHECK-NEXT: tilestored %tmm7, 3008(%rsp,%rcx) # 1024-byte Folded Spill -; CHECK-NEXT: t2rpntlvwz1t1 (%rsi,%rdx), %tmm6 -; CHECK-NEXT: tilestored %tmm6, -64(%rsp,%rcx) # 1024-byte Folded Spill -; CHECK-NEXT: tilestored %tmm7, 960(%rsp,%rcx) # 1024-byte Folded Spill -; CHECK-NEXT: t2rpntlvwz0 (%rsi,%rdx), %tmm6 -; CHECK-NEXT: tilestored %tmm4, (%rsi,%rdx) -; CHECK-NEXT: tilestored %tmm5, (%rsi,%rdx) -; CHECK-NEXT: tileloadd 4032(%rsp,%rcx), %tmm4 # 1024-byte Folded Reload -; CHECK-NEXT: tileloadd 5056(%rsp,%rcx), %tmm5 # 1024-byte Folded Reload -; CHECK-NEXT: tilestored %tmm4, (%rsi,%rdx) -; CHECK-NEXT: tilestored %tmm5, (%rsi,%rdx) -; CHECK-NEXT: tileloadd 1984(%rsp,%rcx), %tmm4 # 1024-byte Folded Reload -; CHECK-NEXT: tileloadd 3008(%rsp,%rcx), %tmm5 # 1024-byte Folded Reload -; CHECK-NEXT: tilestored %tmm4, (%rsi,%rdx) -; CHECK-NEXT: tilestored %tmm5, (%rsi,%rdx) -; CHECK-NEXT: tileloadd -64(%rsp,%rcx), %tmm4 # 1024-byte Folded Reload -; CHECK-NEXT: tileloadd 960(%rsp,%rcx), %tmm5 # 1024-byte Folded Reload -; CHECK-NEXT: tilestored %tmm4, (%rsi,%rdx) -; CHECK-NEXT: tilestored %tmm5, (%rsi,%rdx) -; CHECK-NEXT: tilestored %tmm6, (%rsi,%rdx) -; CHECK-NEXT: tilestored %tmm7, (%rsi,%rdx) -; CHECK-NEXT: addq $6088, %rsp # imm = 0x17C8 -; CHECK-NEXT: tilerelease -; CHECK-NEXT: vzeroupper -; CHECK-NEXT: retq -; -; EGPR-LABEL: test_amx_spill: -; EGPR: # %bb.0: -; EGPR-NEXT: subq $6088, %rsp # encoding: [0x48,0x81,0xec,0xc8,0x17,0x00,0x00] -; EGPR-NEXT: # imm = 0x17C8 -; EGPR-NEXT: vxorps %xmm0, %xmm0, %xmm0 # encoding: [0xc5,0xf8,0x57,0xc0] -; EGPR-NEXT: vmovups %zmm0, -{{[0-9]+}}(%rsp) # encoding: [0x62,0xf1,0x7c,0x48,0x11,0x44,0x24,0xfe] -; EGPR-NEXT: movb $1, -{{[0-9]+}}(%rsp) # encoding: [0xc6,0x44,0x24,0x80,0x01] -; EGPR-NEXT: movb $8, -{{[0-9]+}}(%rsp) # encoding: [0xc6,0x44,0x24,0xb0,0x08] -; EGPR-NEXT: movw $8, -{{[0-9]+}}(%rsp) # encoding: [0x66,0xc7,0x44,0x24,0x90,0x08,0x00] -; EGPR-NEXT: movb $8, -{{[0-9]+}}(%rsp) # encoding: [0xc6,0x44,0x24,0xb4,0x08] -; EGPR-NEXT: movw $8, -{{[0-9]+}}(%rsp) # encoding: [0x66,0xc7,0x44,0x24,0x98,0x08,0x00] -; EGPR-NEXT: movb $8, -{{[0-9]+}}(%rsp) # encoding: [0xc6,0x44,0x24,0xb5,0x08] -; EGPR-NEXT: movw $8, -{{[0-9]+}}(%rsp) # encoding: [0x66,0xc7,0x44,0x24,0x9a,0x08,0x00] -; EGPR-NEXT: movb $8, -{{[0-9]+}}(%rsp) # encoding: [0xc6,0x44,0x24,0xb6,0x08] -; EGPR-NEXT: movw $8, -{{[0-9]+}}(%rsp) # encoding: [0x66,0xc7,0x44,0x24,0x9c,0x08,0x00] -; EGPR-NEXT: movb $8, -{{[0-9]+}}(%rsp) # encoding: [0xc6,0x44,0x24,0xb7,0x08] -; EGPR-NEXT: movw $8, -{{[0-9]+}}(%rsp) # encoding: [0x66,0xc7,0x44,0x24,0x9e,0x08,0x00] -; EGPR-NEXT: ldtilecfg -{{[0-9]+}}(%rsp) # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0x49,0x44,0x24,0x80] -; EGPR-NEXT: movw $8, %ax # encoding: [0x66,0xb8,0x08,0x00] -; EGPR-NEXT: tileloadd (%rsi,%rdx), %tmm0 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7b,0x4b,0x04,0x16] -; EGPR-NEXT: t2rpntlvwz0 (%rsi,%rdx), %tmm4 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0x6e,0x24,0x16] -; EGPR-NEXT: t2rpntlvwz0t1 (%rsi,%rdx), %tmm6 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0x6f,0x34,0x16] -; EGPR-NEXT: movabsq $64, %rcx # encoding: [0x48,0xb9,0x40,0x00,0x00,0x00,0x00,0x00,0x00,0x00] -; EGPR-NEXT: tilestored %tmm6, 4032(%rsp,%rcx) # 1024-byte Folded Spill -; EGPR-NEXT: # encoding: [0xc4,0xe2,0x7a,0x4b,0xb4,0x0c,0xc0,0x0f,0x00,0x00] -; EGPR-NEXT: tilestored %tmm7, 5056(%rsp,%rcx) # 1024-byte Folded Spill -; EGPR-NEXT: # encoding: [0xc4,0xe2,0x7a,0x4b,0xbc,0x0c,0xc0,0x13,0x00,0x00] -; EGPR-NEXT: t2rpntlvwz1 (%rsi,%rdx), %tmm6 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x6e,0x34,0x16] -; EGPR-NEXT: tilestored %tmm6, 1984(%rsp,%rcx) # 1024-byte Folded Spill -; EGPR-NEXT: # encoding: [0xc4,0xe2,0x7a,0x4b,0xb4,0x0c,0xc0,0x07,0x00,0x00] -; EGPR-NEXT: tilestored %tmm7, 3008(%rsp,%rcx) # 1024-byte Folded Spill -; EGPR-NEXT: # encoding: [0xc4,0xe2,0x7a,0x4b,0xbc,0x0c,0xc0,0x0b,0x00,0x00] -; EGPR-NEXT: t2rpntlvwz1t1 (%rsi,%rdx), %tmm6 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x79,0x6f,0x34,0x16] -; EGPR-NEXT: tilestored %tmm6, -64(%rsp,%rcx) # 1024-byte Folded Spill -; EGPR-NEXT: # encoding: [0xc4,0xe2,0x7a,0x4b,0x74,0x0c,0xc0] -; EGPR-NEXT: tilestored %tmm7, 960(%rsp,%rcx) # 1024-byte Folded Spill -; EGPR-NEXT: # encoding: [0xc4,0xe2,0x7a,0x4b,0xbc,0x0c,0xc0,0x03,0x00,0x00] -; EGPR-NEXT: t2rpntlvwz0 (%rsi,%rdx), %tmm6 # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x78,0x6e,0x34,0x16] -; EGPR-NEXT: tilestored %tmm4, (%rsi,%rdx) # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7a,0x4b,0x24,0x16] -; EGPR-NEXT: tilestored %tmm5, (%rsi,%rdx) # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7a,0x4b,0x2c,0x16] -; EGPR-NEXT: tileloadd 4032(%rsp,%rcx), %tmm4 # 1024-byte Folded Reload -; EGPR-NEXT: # encoding: [0xc4,0xe2,0x7b,0x4b,0xa4,0x0c,0xc0,0x0f,0x00,0x00] -; EGPR-NEXT: tileloadd 5056(%rsp,%rcx), %tmm5 # 1024-byte Folded Reload -; EGPR-NEXT: # encoding: [0xc4,0xe2,0x7b,0x4b,0xac,0x0c,0xc0,0x13,0x00,0x00] -; EGPR-NEXT: tilestored %tmm4, (%rsi,%rdx) # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7a,0x4b,0x24,0x16] -; EGPR-NEXT: tilestored %tmm5, (%rsi,%rdx) # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7a,0x4b,0x2c,0x16] -; EGPR-NEXT: tileloadd 1984(%rsp,%rcx), %tmm4 # 1024-byte Folded Reload -; EGPR-NEXT: # encoding: [0xc4,0xe2,0x7b,0x4b,0xa4,0x0c,0xc0,0x07,0x00,0x00] -; EGPR-NEXT: tileloadd 3008(%rsp,%rcx), %tmm5 # 1024-byte Folded Reload -; EGPR-NEXT: # encoding: [0xc4,0xe2,0x7b,0x4b,0xac,0x0c,0xc0,0x0b,0x00,0x00] -; EGPR-NEXT: tilestored %tmm4, (%rsi,%rdx) # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7a,0x4b,0x24,0x16] -; EGPR-NEXT: tilestored %tmm5, (%rsi,%rdx) # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7a,0x4b,0x2c,0x16] -; EGPR-NEXT: tileloadd -64(%rsp,%rcx), %tmm4 # 1024-byte Folded Reload -; EGPR-NEXT: # encoding: [0xc4,0xe2,0x7b,0x4b,0x64,0x0c,0xc0] -; EGPR-NEXT: tileloadd 960(%rsp,%rcx), %tmm5 # 1024-byte Folded Reload -; EGPR-NEXT: # encoding: [0xc4,0xe2,0x7b,0x4b,0xac,0x0c,0xc0,0x03,0x00,0x00] -; EGPR-NEXT: tilestored %tmm4, (%rsi,%rdx) # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7a,0x4b,0x24,0x16] -; EGPR-NEXT: tilestored %tmm5, (%rsi,%rdx) # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7a,0x4b,0x2c,0x16] -; EGPR-NEXT: tilestored %tmm6, (%rsi,%rdx) # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7a,0x4b,0x34,0x16] -; EGPR-NEXT: tilestored %tmm7, (%rsi,%rdx) # EVEX TO VEX Compression encoding: [0xc4,0xe2,0x7a,0x4b,0x3c,0x16] -; EGPR-NEXT: addq $6088, %rsp # encoding: [0x48,0x81,0xc4,0xc8,0x17,0x00,0x00] -; EGPR-NEXT: # imm = 0x17C8 -; EGPR-NEXT: tilerelease # encoding: [0xc4,0xe2,0x78,0x49,0xc0] -; EGPR-NEXT: vzeroupper # encoding: [0xc5,0xf8,0x77] -; EGPR-NEXT: retq # encoding: [0xc3] - %a = call x86_amx @llvm.x86.tileloadd64.internal(i16 8, i16 8, i8* %base, i64 %stride) - %b1 = call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0.internal(i16 8, i16 8, i16 8, i8* %base, i64 %stride) - %b2 = call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0t1.internal(i16 8, i16 8, i16 8, i8* %base, i64 %stride) - %b3 = call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz1.internal(i16 8, i16 8, i16 8, i8* %base, i64 %stride) - %b4 = call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz1t1.internal(i16 8, i16 8, i16 8, i8* %base, i64 %stride) - %b5 = call { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0.internal(i16 8, i16 8, i16 8, i8* %base, i64 %stride) - %e11 = extractvalue { x86_amx, x86_amx } %b1, 0 - %e12 = extractvalue { x86_amx, x86_amx } %b1, 1 - %e21 = extractvalue { x86_amx, x86_amx } %b2, 0 - %e22 = extractvalue { x86_amx, x86_amx } %b2, 1 - %e31 = extractvalue { x86_amx, x86_amx } %b3, 0 - %e32 = extractvalue { x86_amx, x86_amx } %b3, 1 - %e41 = extractvalue { x86_amx, x86_amx } %b4, 0 - %e42 = extractvalue { x86_amx, x86_amx } %b4, 1 - %e51 = extractvalue { x86_amx, x86_amx } %b5, 0 - %e52 = extractvalue { x86_amx, x86_amx } %b5, 1 - call void @llvm.x86.tilestored64.internal(i16 8, i16 8, i8* %base, i64 %stride, x86_amx %e11) - call void @llvm.x86.tilestored64.internal(i16 8, i16 8, i8* %base, i64 %stride, x86_amx %e12) - call void @llvm.x86.tilestored64.internal(i16 8, i16 8, i8* %base, i64 %stride, x86_amx %e21) - call void @llvm.x86.tilestored64.internal(i16 8, i16 8, i8* %base, i64 %stride, x86_amx %e22) - call void @llvm.x86.tilestored64.internal(i16 8, i16 8, i8* %base, i64 %stride, x86_amx %e31) - call void @llvm.x86.tilestored64.internal(i16 8, i16 8, i8* %base, i64 %stride, x86_amx %e32) - call void @llvm.x86.tilestored64.internal(i16 8, i16 8, i8* %base, i64 %stride, x86_amx %e41) - call void @llvm.x86.tilestored64.internal(i16 8, i16 8, i8* %base, i64 %stride, x86_amx %e42) - call void @llvm.x86.tilestored64.internal(i16 8, i16 8, i8* %base, i64 %stride, x86_amx %e51) - call void @llvm.x86.tilestored64.internal(i16 8, i16 8, i8* %base, i64 %stride, x86_amx %e52) - ret void -} - -declare x86_amx @llvm.x86.tileloadd64.internal(i16, i16, i8*, i64) -declare void @llvm.x86.tilestored64.internal(i16, i16, i8*, i64, x86_amx) -declare { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0.internal(i16, i16, i16, i8*, i64) -declare { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz0t1.internal(i16, i16, i16, i8*, i64) -declare { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz1.internal(i16, i16, i16, i8*, i64) -declare { x86_amx, x86_amx } @llvm.x86.t2rpntlvwz1t1.internal(i16, i16, i16, i8*, i64) -declare x86_amx @llvm.x86.ttransposed.internal(i16, i16, x86_amx) -declare x86_amx @llvm.x86.ttdpbf16ps.internal(i16, i16, i16, x86_amx, x86_amx, x86_amx) -declare x86_amx @llvm.x86.ttdpfp16ps.internal(i16, i16, i16, x86_amx, x86_amx, x86_amx) -declare x86_amx @llvm.x86.ttcmmimfp16ps.internal(i16, i16, i16, x86_amx, x86_amx, x86_amx) -declare x86_amx @llvm.x86.ttcmmrlfp16ps.internal(i16, i16, i16, x86_amx, x86_amx, x86_amx) -declare x86_amx @llvm.x86.tconjtcmmimfp16ps.internal(i16, i16, i16, x86_amx, x86_amx, x86_amx) -declare x86_amx @llvm.x86.tconjtfp16.internal(i16, i16, x86_amx) - -attributes #0 = { nounwind } diff --git a/llvm/test/CodeGen/X86/and-mask-variable.ll b/llvm/test/CodeGen/X86/and-mask-variable.ll index d89f0db..3e5bd69 100644 --- a/llvm/test/CodeGen/X86/and-mask-variable.ll +++ b/llvm/test/CodeGen/X86/and-mask-variable.ll @@ -1,10 +1,10 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc -mtriple=i686-unknown-linux-gnu -mattr=-bmi,-tbm,-bmi2,+fast-bextr < %s | FileCheck %s --check-prefixes=X86-NOBMI -; RUN: llc -mtriple=i686-unknown-linux-gnu -mattr=+bmi,+tbm,+bmi2,+fast-bextr < %s | FileCheck %s --check-prefixes=X86-BMI2 -; RUN: llc -mtriple=i686-unknown-linux-gnu -mattr=+bmi,-tbm,+bmi2,+fast-bextr < %s | FileCheck %s --check-prefixes=X86-BMI2 -; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=-bmi,-tbm,-bmi2,+fast-bextr < %s | FileCheck %s --check-prefixes=X64-NOBMI -; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+bmi,+tbm,+bmi2,+fast-bextr < %s | FileCheck %s --check-prefixes=X64-BMI2 -; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+bmi,-tbm,+bmi2,+fast-bextr < %s | FileCheck %s --check-prefixes=X64-BMI2 +; RUN: llc -mtriple=i686-unknown-linux-gnu -mattr=-bmi,-tbm,-bmi2 < %s | FileCheck %s --check-prefixes=X86-NOBMI +; RUN: llc -mtriple=i686-unknown-linux-gnu -mattr=+bmi,+tbm,+bmi2 < %s | FileCheck %s --check-prefixes=X86-BMI2 +; RUN: llc -mtriple=i686-unknown-linux-gnu -mattr=+bmi,-tbm,+bmi2 < %s | FileCheck %s --check-prefixes=X86-BMI2 +; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=-bmi,-tbm,-bmi2 < %s | FileCheck %s --check-prefixes=X64-NOBMI +; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+bmi,+tbm,+bmi2 < %s | FileCheck %s --check-prefixes=X64-BMI2 +; RUN: llc -mtriple=x86_64-unknown-linux-gnu -mattr=+bmi,-tbm,+bmi2 < %s | FileCheck %s --check-prefixes=X64-BMI2 define i32 @mask_pair(i32 %x, i32 %y) nounwind { ; X86-NOBMI-LABEL: mask_pair: diff --git a/llvm/test/CodeGen/X86/atomic-load-store.ll b/llvm/test/CodeGen/X86/atomic-load-store.ll index 3e7b73a..1173c45 100644 --- a/llvm/test/CodeGen/X86/atomic-load-store.ll +++ b/llvm/test/CodeGen/X86/atomic-load-store.ll @@ -1,12 +1,12 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc < %s -mtriple=x86_64-- -verify-machineinstrs -mcpu=x86-64 | FileCheck %s --check-prefixes=CHECK,CHECK-O3 -; RUN: llc < %s -mtriple=x86_64-- -verify-machineinstrs -mcpu=x86-64-v2 | FileCheck %s --check-prefixes=CHECK,CHECK-SSE-O3 -; RUN: llc < %s -mtriple=x86_64-- -verify-machineinstrs -mcpu=x86-64-v3 | FileCheck %s --check-prefixes=CHECK,CHECK-AVX-O3 -; RUN: llc < %s -mtriple=x86_64-- -verify-machineinstrs -mcpu=x86-64-v4 | FileCheck %s --check-prefixes=CHECK,CHECK-AVX-O3 -; RUN: llc < %s -mtriple=x86_64-- -verify-machineinstrs -O0 -mcpu=x86-64 | FileCheck %s --check-prefixes=CHECK,CHECK-O0 -; RUN: llc < %s -mtriple=x86_64-- -verify-machineinstrs -O0 -mcpu=x86-64-v2 | FileCheck %s --check-prefixes=CHECK,CHECK-SSE-O0 -; RUN: llc < %s -mtriple=x86_64-- -verify-machineinstrs -O0 -mcpu=x86-64-v3 | FileCheck %s --check-prefixes=CHECK,CHECK-AVX-O0 -; RUN: llc < %s -mtriple=x86_64-- -verify-machineinstrs -O0 -mcpu=x86-64-v4 | FileCheck %s --check-prefixes=CHECK,CHECK-AVX-O0 +; RUN: llc < %s -mtriple=x86_64-- -verify-machineinstrs -mcpu=x86-64 | FileCheck %s --check-prefixes=CHECK,CHECK-O3,CHECK-SSE-O3 +; RUN: llc < %s -mtriple=x86_64-- -verify-machineinstrs -mcpu=x86-64-v2 | FileCheck %s --check-prefixes=CHECK,CHECK-O3,CHECK-SSE-O3 +; RUN: llc < %s -mtriple=x86_64-- -verify-machineinstrs -mcpu=x86-64-v3 | FileCheck %s --check-prefixes=CHECK,CHECK-O3,CHECK-AVX-O3,CHECK-AVX2-O3 +; RUN: llc < %s -mtriple=x86_64-- -verify-machineinstrs -mcpu=x86-64-v4 | FileCheck %s --check-prefixes=CHECK,CHECK-O3,CHECK-AVX-O3,CHECK-AVX512-O3 +; RUN: llc < %s -mtriple=x86_64-- -verify-machineinstrs -O0 -mcpu=x86-64 | FileCheck %s --check-prefixes=CHECK,CHECK-O0,CHECK-SSE-O0 +; RUN: llc < %s -mtriple=x86_64-- -verify-machineinstrs -O0 -mcpu=x86-64-v2 | FileCheck %s --check-prefixes=CHECK,CHECK-O0,CHECK-SSE-O0 +; RUN: llc < %s -mtriple=x86_64-- -verify-machineinstrs -O0 -mcpu=x86-64-v3 | FileCheck %s --check-prefixes=CHECK,CHECK-O0,CHECK-AVX-O0,CHECK-AVX2-O0 +; RUN: llc < %s -mtriple=x86_64-- -verify-machineinstrs -O0 -mcpu=x86-64-v4 | FileCheck %s --check-prefixes=CHECK,CHECK-O0,CHECK-AVX-O0,CHECK-AVX512-O0 define void @test1(ptr %ptr, i32 %val1) { ; CHECK-LABEL: test1: @@ -50,30 +50,10 @@ define <1 x i8> @atomic_vec1_i8(ptr %x) { ; CHECK-O3-NEXT: movzbl (%rdi), %eax ; CHECK-O3-NEXT: retq ; -; CHECK-SSE-O3-LABEL: atomic_vec1_i8: -; CHECK-SSE-O3: # %bb.0: -; CHECK-SSE-O3-NEXT: movzbl (%rdi), %eax -; CHECK-SSE-O3-NEXT: retq -; -; CHECK-AVX-O3-LABEL: atomic_vec1_i8: -; CHECK-AVX-O3: # %bb.0: -; CHECK-AVX-O3-NEXT: movzbl (%rdi), %eax -; CHECK-AVX-O3-NEXT: retq -; ; CHECK-O0-LABEL: atomic_vec1_i8: ; CHECK-O0: # %bb.0: ; CHECK-O0-NEXT: movb (%rdi), %al ; CHECK-O0-NEXT: retq -; -; CHECK-SSE-O0-LABEL: atomic_vec1_i8: -; CHECK-SSE-O0: # %bb.0: -; CHECK-SSE-O0-NEXT: movb (%rdi), %al -; CHECK-SSE-O0-NEXT: retq -; -; CHECK-AVX-O0-LABEL: atomic_vec1_i8: -; CHECK-AVX-O0: # %bb.0: -; CHECK-AVX-O0-NEXT: movb (%rdi), %al -; CHECK-AVX-O0-NEXT: retq %ret = load atomic <1 x i8>, ptr %x acquire, align 1 ret <1 x i8> %ret } @@ -84,30 +64,10 @@ define <1 x i16> @atomic_vec1_i16(ptr %x) { ; CHECK-O3-NEXT: movzwl (%rdi), %eax ; CHECK-O3-NEXT: retq ; -; CHECK-SSE-O3-LABEL: atomic_vec1_i16: -; CHECK-SSE-O3: # %bb.0: -; CHECK-SSE-O3-NEXT: movzwl (%rdi), %eax -; CHECK-SSE-O3-NEXT: retq -; -; CHECK-AVX-O3-LABEL: atomic_vec1_i16: -; CHECK-AVX-O3: # %bb.0: -; CHECK-AVX-O3-NEXT: movzwl (%rdi), %eax -; CHECK-AVX-O3-NEXT: retq -; ; CHECK-O0-LABEL: atomic_vec1_i16: ; CHECK-O0: # %bb.0: ; CHECK-O0-NEXT: movw (%rdi), %ax ; CHECK-O0-NEXT: retq -; -; CHECK-SSE-O0-LABEL: atomic_vec1_i16: -; CHECK-SSE-O0: # %bb.0: -; CHECK-SSE-O0-NEXT: movw (%rdi), %ax -; CHECK-SSE-O0-NEXT: retq -; -; CHECK-AVX-O0-LABEL: atomic_vec1_i16: -; CHECK-AVX-O0: # %bb.0: -; CHECK-AVX-O0-NEXT: movw (%rdi), %ax -; CHECK-AVX-O0-NEXT: retq %ret = load atomic <1 x i16>, ptr %x acquire, align 2 ret <1 x i16> %ret } @@ -119,35 +79,11 @@ define <1 x i32> @atomic_vec1_i8_zext(ptr %x) { ; CHECK-O3-NEXT: movzbl %al, %eax ; CHECK-O3-NEXT: retq ; -; CHECK-SSE-O3-LABEL: atomic_vec1_i8_zext: -; CHECK-SSE-O3: # %bb.0: -; CHECK-SSE-O3-NEXT: movzbl (%rdi), %eax -; CHECK-SSE-O3-NEXT: movzbl %al, %eax -; CHECK-SSE-O3-NEXT: retq -; -; CHECK-AVX-O3-LABEL: atomic_vec1_i8_zext: -; CHECK-AVX-O3: # %bb.0: -; CHECK-AVX-O3-NEXT: movzbl (%rdi), %eax -; CHECK-AVX-O3-NEXT: movzbl %al, %eax -; CHECK-AVX-O3-NEXT: retq -; ; CHECK-O0-LABEL: atomic_vec1_i8_zext: ; CHECK-O0: # %bb.0: ; CHECK-O0-NEXT: movb (%rdi), %al ; CHECK-O0-NEXT: movzbl %al, %eax ; CHECK-O0-NEXT: retq -; -; CHECK-SSE-O0-LABEL: atomic_vec1_i8_zext: -; CHECK-SSE-O0: # %bb.0: -; CHECK-SSE-O0-NEXT: movb (%rdi), %al -; CHECK-SSE-O0-NEXT: movzbl %al, %eax -; CHECK-SSE-O0-NEXT: retq -; -; CHECK-AVX-O0-LABEL: atomic_vec1_i8_zext: -; CHECK-AVX-O0: # %bb.0: -; CHECK-AVX-O0-NEXT: movb (%rdi), %al -; CHECK-AVX-O0-NEXT: movzbl %al, %eax -; CHECK-AVX-O0-NEXT: retq %ret = load atomic <1 x i8>, ptr %x acquire, align 1 %zret = zext <1 x i8> %ret to <1 x i32> ret <1 x i32> %zret @@ -160,35 +96,11 @@ define <1 x i64> @atomic_vec1_i16_sext(ptr %x) { ; CHECK-O3-NEXT: movswq %ax, %rax ; CHECK-O3-NEXT: retq ; -; CHECK-SSE-O3-LABEL: atomic_vec1_i16_sext: -; CHECK-SSE-O3: # %bb.0: -; CHECK-SSE-O3-NEXT: movzwl (%rdi), %eax -; CHECK-SSE-O3-NEXT: movswq %ax, %rax -; CHECK-SSE-O3-NEXT: retq -; -; CHECK-AVX-O3-LABEL: atomic_vec1_i16_sext: -; CHECK-AVX-O3: # %bb.0: -; CHECK-AVX-O3-NEXT: movzwl (%rdi), %eax -; CHECK-AVX-O3-NEXT: movswq %ax, %rax -; CHECK-AVX-O3-NEXT: retq -; ; CHECK-O0-LABEL: atomic_vec1_i16_sext: ; CHECK-O0: # %bb.0: ; CHECK-O0-NEXT: movw (%rdi), %ax ; CHECK-O0-NEXT: movswq %ax, %rax ; CHECK-O0-NEXT: retq -; -; CHECK-SSE-O0-LABEL: atomic_vec1_i16_sext: -; CHECK-SSE-O0: # %bb.0: -; CHECK-SSE-O0-NEXT: movw (%rdi), %ax -; CHECK-SSE-O0-NEXT: movswq %ax, %rax -; CHECK-SSE-O0-NEXT: retq -; -; CHECK-AVX-O0-LABEL: atomic_vec1_i16_sext: -; CHECK-AVX-O0: # %bb.0: -; CHECK-AVX-O0-NEXT: movw (%rdi), %ax -; CHECK-AVX-O0-NEXT: movswq %ax, %rax -; CHECK-AVX-O0-NEXT: retq %ret = load atomic <1 x i16>, ptr %x acquire, align 2 %sret = sext <1 x i16> %ret to <1 x i64> ret <1 x i64> %sret @@ -204,12 +116,6 @@ define <1 x ptr addrspace(270)> @atomic_vec1_ptr270(ptr %x) { } define <1 x bfloat> @atomic_vec1_bfloat(ptr %x) { -; CHECK-O3-LABEL: atomic_vec1_bfloat: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: movzwl (%rdi), %eax -; CHECK-O3-NEXT: pinsrw $0, %eax, %xmm0 -; CHECK-O3-NEXT: retq -; ; CHECK-SSE-O3-LABEL: atomic_vec1_bfloat: ; CHECK-SSE-O3: # %bb.0: ; CHECK-SSE-O3-NEXT: movzwl (%rdi), %eax @@ -222,15 +128,6 @@ define <1 x bfloat> @atomic_vec1_bfloat(ptr %x) { ; CHECK-AVX-O3-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 ; CHECK-AVX-O3-NEXT: retq ; -; CHECK-O0-LABEL: atomic_vec1_bfloat: -; CHECK-O0: # %bb.0: -; CHECK-O0-NEXT: movw (%rdi), %cx -; CHECK-O0-NEXT: # implicit-def: $eax -; CHECK-O0-NEXT: movw %cx, %ax -; CHECK-O0-NEXT: # implicit-def: $xmm0 -; CHECK-O0-NEXT: pinsrw $0, %eax, %xmm0 -; CHECK-O0-NEXT: retq -; ; CHECK-SSE-O0-LABEL: atomic_vec1_bfloat: ; CHECK-SSE-O0: # %bb.0: ; CHECK-SSE-O0-NEXT: movw (%rdi), %cx @@ -283,30 +180,6 @@ define <1 x ptr> @atomic_vec1_ptr(ptr %x) nounwind { ; CHECK-O3-NEXT: popq %rcx ; CHECK-O3-NEXT: retq ; -; CHECK-SSE-O3-LABEL: atomic_vec1_ptr: -; CHECK-SSE-O3: # %bb.0: -; CHECK-SSE-O3-NEXT: pushq %rax -; CHECK-SSE-O3-NEXT: movq %rdi, %rsi -; CHECK-SSE-O3-NEXT: movq %rsp, %rdx -; CHECK-SSE-O3-NEXT: movl $8, %edi -; CHECK-SSE-O3-NEXT: movl $2, %ecx -; CHECK-SSE-O3-NEXT: callq __atomic_load@PLT -; CHECK-SSE-O3-NEXT: movq (%rsp), %rax -; CHECK-SSE-O3-NEXT: popq %rcx -; CHECK-SSE-O3-NEXT: retq -; -; CHECK-AVX-O3-LABEL: atomic_vec1_ptr: -; CHECK-AVX-O3: # %bb.0: -; CHECK-AVX-O3-NEXT: pushq %rax -; CHECK-AVX-O3-NEXT: movq %rdi, %rsi -; CHECK-AVX-O3-NEXT: movq %rsp, %rdx -; CHECK-AVX-O3-NEXT: movl $8, %edi -; CHECK-AVX-O3-NEXT: movl $2, %ecx -; CHECK-AVX-O3-NEXT: callq __atomic_load@PLT -; CHECK-AVX-O3-NEXT: movq (%rsp), %rax -; CHECK-AVX-O3-NEXT: popq %rcx -; CHECK-AVX-O3-NEXT: retq -; ; CHECK-O0-LABEL: atomic_vec1_ptr: ; CHECK-O0: # %bb.0: ; CHECK-O0-NEXT: pushq %rax @@ -318,41 +191,11 @@ define <1 x ptr> @atomic_vec1_ptr(ptr %x) nounwind { ; CHECK-O0-NEXT: movq (%rsp), %rax ; CHECK-O0-NEXT: popq %rcx ; CHECK-O0-NEXT: retq -; -; CHECK-SSE-O0-LABEL: atomic_vec1_ptr: -; CHECK-SSE-O0: # %bb.0: -; CHECK-SSE-O0-NEXT: pushq %rax -; CHECK-SSE-O0-NEXT: movq %rdi, %rsi -; CHECK-SSE-O0-NEXT: movl $8, %edi -; CHECK-SSE-O0-NEXT: movq %rsp, %rdx -; CHECK-SSE-O0-NEXT: movl $2, %ecx -; CHECK-SSE-O0-NEXT: callq __atomic_load@PLT -; CHECK-SSE-O0-NEXT: movq (%rsp), %rax -; CHECK-SSE-O0-NEXT: popq %rcx -; CHECK-SSE-O0-NEXT: retq -; -; CHECK-AVX-O0-LABEL: atomic_vec1_ptr: -; CHECK-AVX-O0: # %bb.0: -; CHECK-AVX-O0-NEXT: pushq %rax -; CHECK-AVX-O0-NEXT: movq %rdi, %rsi -; CHECK-AVX-O0-NEXT: movl $8, %edi -; CHECK-AVX-O0-NEXT: movq %rsp, %rdx -; CHECK-AVX-O0-NEXT: movl $2, %ecx -; CHECK-AVX-O0-NEXT: callq __atomic_load@PLT -; CHECK-AVX-O0-NEXT: movq (%rsp), %rax -; CHECK-AVX-O0-NEXT: popq %rcx -; CHECK-AVX-O0-NEXT: retq %ret = load atomic <1 x ptr>, ptr %x acquire, align 4 ret <1 x ptr> %ret } define <1 x half> @atomic_vec1_half(ptr %x) { -; CHECK-O3-LABEL: atomic_vec1_half: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: movzwl (%rdi), %eax -; CHECK-O3-NEXT: pinsrw $0, %eax, %xmm0 -; CHECK-O3-NEXT: retq -; ; CHECK-SSE-O3-LABEL: atomic_vec1_half: ; CHECK-SSE-O3: # %bb.0: ; CHECK-SSE-O3-NEXT: movzwl (%rdi), %eax @@ -365,15 +208,6 @@ define <1 x half> @atomic_vec1_half(ptr %x) { ; CHECK-AVX-O3-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0 ; CHECK-AVX-O3-NEXT: retq ; -; CHECK-O0-LABEL: atomic_vec1_half: -; CHECK-O0: # %bb.0: -; CHECK-O0-NEXT: movw (%rdi), %cx -; CHECK-O0-NEXT: # implicit-def: $eax -; CHECK-O0-NEXT: movw %cx, %ax -; CHECK-O0-NEXT: # implicit-def: $xmm0 -; CHECK-O0-NEXT: pinsrw $0, %eax, %xmm0 -; CHECK-O0-NEXT: retq -; ; CHECK-SSE-O0-LABEL: atomic_vec1_half: ; CHECK-SSE-O0: # %bb.0: ; CHECK-SSE-O0-NEXT: movw (%rdi), %cx @@ -396,11 +230,6 @@ define <1 x half> @atomic_vec1_half(ptr %x) { } define <1 x float> @atomic_vec1_float(ptr %x) { -; CHECK-O3-LABEL: atomic_vec1_float: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero -; CHECK-O3-NEXT: retq -; ; CHECK-SSE-O3-LABEL: atomic_vec1_float: ; CHECK-SSE-O3: # %bb.0: ; CHECK-SSE-O3-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero @@ -411,11 +240,6 @@ define <1 x float> @atomic_vec1_float(ptr %x) { ; CHECK-AVX-O3-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero ; CHECK-AVX-O3-NEXT: retq ; -; CHECK-O0-LABEL: atomic_vec1_float: -; CHECK-O0: # %bb.0: -; CHECK-O0-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero -; CHECK-O0-NEXT: retq -; ; CHECK-SSE-O0-LABEL: atomic_vec1_float: ; CHECK-SSE-O0: # %bb.0: ; CHECK-SSE-O0-NEXT: movss {{.*#+}} xmm0 = mem[0],zero,zero,zero @@ -430,11 +254,6 @@ define <1 x float> @atomic_vec1_float(ptr %x) { } define <1 x double> @atomic_vec1_double_align(ptr %x) nounwind { -; CHECK-O3-LABEL: atomic_vec1_double_align: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero -; CHECK-O3-NEXT: retq -; ; CHECK-SSE-O3-LABEL: atomic_vec1_double_align: ; CHECK-SSE-O3: # %bb.0: ; CHECK-SSE-O3-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero @@ -445,11 +264,6 @@ define <1 x double> @atomic_vec1_double_align(ptr %x) nounwind { ; CHECK-AVX-O3-NEXT: vmovsd {{.*#+}} xmm0 = mem[0],zero ; CHECK-AVX-O3-NEXT: retq ; -; CHECK-O0-LABEL: atomic_vec1_double_align: -; CHECK-O0: # %bb.0: -; CHECK-O0-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero -; CHECK-O0-NEXT: retq -; ; CHECK-SSE-O0-LABEL: atomic_vec1_double_align: ; CHECK-SSE-O0: # %bb.0: ; CHECK-SSE-O0-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero @@ -476,30 +290,6 @@ define <1 x i64> @atomic_vec1_i64(ptr %x) nounwind { ; CHECK-O3-NEXT: popq %rcx ; CHECK-O3-NEXT: retq ; -; CHECK-SSE-O3-LABEL: atomic_vec1_i64: -; CHECK-SSE-O3: # %bb.0: -; CHECK-SSE-O3-NEXT: pushq %rax -; CHECK-SSE-O3-NEXT: movq %rdi, %rsi -; CHECK-SSE-O3-NEXT: movq %rsp, %rdx -; CHECK-SSE-O3-NEXT: movl $8, %edi -; CHECK-SSE-O3-NEXT: movl $2, %ecx -; CHECK-SSE-O3-NEXT: callq __atomic_load@PLT -; CHECK-SSE-O3-NEXT: movq (%rsp), %rax -; CHECK-SSE-O3-NEXT: popq %rcx -; CHECK-SSE-O3-NEXT: retq -; -; CHECK-AVX-O3-LABEL: atomic_vec1_i64: -; CHECK-AVX-O3: # %bb.0: -; CHECK-AVX-O3-NEXT: pushq %rax -; CHECK-AVX-O3-NEXT: movq %rdi, %rsi -; CHECK-AVX-O3-NEXT: movq %rsp, %rdx -; CHECK-AVX-O3-NEXT: movl $8, %edi -; CHECK-AVX-O3-NEXT: movl $2, %ecx -; CHECK-AVX-O3-NEXT: callq __atomic_load@PLT -; CHECK-AVX-O3-NEXT: movq (%rsp), %rax -; CHECK-AVX-O3-NEXT: popq %rcx -; CHECK-AVX-O3-NEXT: retq -; ; CHECK-O0-LABEL: atomic_vec1_i64: ; CHECK-O0: # %bb.0: ; CHECK-O0-NEXT: pushq %rax @@ -511,47 +301,11 @@ define <1 x i64> @atomic_vec1_i64(ptr %x) nounwind { ; CHECK-O0-NEXT: movq (%rsp), %rax ; CHECK-O0-NEXT: popq %rcx ; CHECK-O0-NEXT: retq -; -; CHECK-SSE-O0-LABEL: atomic_vec1_i64: -; CHECK-SSE-O0: # %bb.0: -; CHECK-SSE-O0-NEXT: pushq %rax -; CHECK-SSE-O0-NEXT: movq %rdi, %rsi -; CHECK-SSE-O0-NEXT: movl $8, %edi -; CHECK-SSE-O0-NEXT: movq %rsp, %rdx -; CHECK-SSE-O0-NEXT: movl $2, %ecx -; CHECK-SSE-O0-NEXT: callq __atomic_load@PLT -; CHECK-SSE-O0-NEXT: movq (%rsp), %rax -; CHECK-SSE-O0-NEXT: popq %rcx -; CHECK-SSE-O0-NEXT: retq -; -; CHECK-AVX-O0-LABEL: atomic_vec1_i64: -; CHECK-AVX-O0: # %bb.0: -; CHECK-AVX-O0-NEXT: pushq %rax -; CHECK-AVX-O0-NEXT: movq %rdi, %rsi -; CHECK-AVX-O0-NEXT: movl $8, %edi -; CHECK-AVX-O0-NEXT: movq %rsp, %rdx -; CHECK-AVX-O0-NEXT: movl $2, %ecx -; CHECK-AVX-O0-NEXT: callq __atomic_load@PLT -; CHECK-AVX-O0-NEXT: movq (%rsp), %rax -; CHECK-AVX-O0-NEXT: popq %rcx -; CHECK-AVX-O0-NEXT: retq %ret = load atomic <1 x i64>, ptr %x acquire, align 4 ret <1 x i64> %ret } define <1 x double> @atomic_vec1_double(ptr %x) nounwind { -; CHECK-O3-LABEL: atomic_vec1_double: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: pushq %rax -; CHECK-O3-NEXT: movq %rdi, %rsi -; CHECK-O3-NEXT: movq %rsp, %rdx -; CHECK-O3-NEXT: movl $8, %edi -; CHECK-O3-NEXT: movl $2, %ecx -; CHECK-O3-NEXT: callq __atomic_load@PLT -; CHECK-O3-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero -; CHECK-O3-NEXT: popq %rax -; CHECK-O3-NEXT: retq -; ; CHECK-SSE-O3-LABEL: atomic_vec1_double: ; CHECK-SSE-O3: # %bb.0: ; CHECK-SSE-O3-NEXT: pushq %rax @@ -576,18 +330,6 @@ define <1 x double> @atomic_vec1_double(ptr %x) nounwind { ; CHECK-AVX-O3-NEXT: popq %rax ; CHECK-AVX-O3-NEXT: retq ; -; CHECK-O0-LABEL: atomic_vec1_double: -; CHECK-O0: # %bb.0: -; CHECK-O0-NEXT: pushq %rax -; CHECK-O0-NEXT: movq %rdi, %rsi -; CHECK-O0-NEXT: movl $8, %edi -; CHECK-O0-NEXT: movq %rsp, %rdx -; CHECK-O0-NEXT: movl $2, %ecx -; CHECK-O0-NEXT: callq __atomic_load@PLT -; CHECK-O0-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero -; CHECK-O0-NEXT: popq %rax -; CHECK-O0-NEXT: retq -; ; CHECK-SSE-O0-LABEL: atomic_vec1_double: ; CHECK-SSE-O0: # %bb.0: ; CHECK-SSE-O0-NEXT: pushq %rax @@ -616,18 +358,6 @@ define <1 x double> @atomic_vec1_double(ptr %x) nounwind { } define <2 x i32> @atomic_vec2_i32(ptr %x) nounwind { -; CHECK-O3-LABEL: atomic_vec2_i32: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: pushq %rax -; CHECK-O3-NEXT: movq %rdi, %rsi -; CHECK-O3-NEXT: movq %rsp, %rdx -; CHECK-O3-NEXT: movl $8, %edi -; CHECK-O3-NEXT: movl $2, %ecx -; CHECK-O3-NEXT: callq __atomic_load@PLT -; CHECK-O3-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero -; CHECK-O3-NEXT: popq %rax -; CHECK-O3-NEXT: retq -; ; CHECK-SSE-O3-LABEL: atomic_vec2_i32: ; CHECK-SSE-O3: # %bb.0: ; CHECK-SSE-O3-NEXT: pushq %rax @@ -652,18 +382,6 @@ define <2 x i32> @atomic_vec2_i32(ptr %x) nounwind { ; CHECK-AVX-O3-NEXT: popq %rax ; CHECK-AVX-O3-NEXT: retq ; -; CHECK-O0-LABEL: atomic_vec2_i32: -; CHECK-O0: # %bb.0: -; CHECK-O0-NEXT: pushq %rax -; CHECK-O0-NEXT: movq %rdi, %rsi -; CHECK-O0-NEXT: movl $8, %edi -; CHECK-O0-NEXT: movq %rsp, %rdx -; CHECK-O0-NEXT: movl $2, %ecx -; CHECK-O0-NEXT: callq __atomic_load@PLT -; CHECK-O0-NEXT: movq {{.*#+}} xmm0 = mem[0],zero -; CHECK-O0-NEXT: popq %rax -; CHECK-O0-NEXT: retq -; ; CHECK-SSE-O0-LABEL: atomic_vec2_i32: ; CHECK-SSE-O0: # %bb.0: ; CHECK-SSE-O0-NEXT: pushq %rax @@ -692,18 +410,6 @@ define <2 x i32> @atomic_vec2_i32(ptr %x) nounwind { } define <4 x float> @atomic_vec4_float(ptr %x) nounwind { -; CHECK-O3-LABEL: atomic_vec4_float: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: subq $24, %rsp -; CHECK-O3-NEXT: movq %rdi, %rsi -; CHECK-O3-NEXT: movq %rsp, %rdx -; CHECK-O3-NEXT: movl $16, %edi -; CHECK-O3-NEXT: movl $2, %ecx -; CHECK-O3-NEXT: callq __atomic_load@PLT -; CHECK-O3-NEXT: movaps (%rsp), %xmm0 -; CHECK-O3-NEXT: addq $24, %rsp -; CHECK-O3-NEXT: retq -; ; CHECK-SSE-O3-LABEL: atomic_vec4_float: ; CHECK-SSE-O3: # %bb.0: ; CHECK-SSE-O3-NEXT: subq $24, %rsp @@ -728,18 +434,6 @@ define <4 x float> @atomic_vec4_float(ptr %x) nounwind { ; CHECK-AVX-O3-NEXT: addq $24, %rsp ; CHECK-AVX-O3-NEXT: retq ; -; CHECK-O0-LABEL: atomic_vec4_float: -; CHECK-O0: # %bb.0: -; CHECK-O0-NEXT: subq $24, %rsp -; CHECK-O0-NEXT: movq %rdi, %rsi -; CHECK-O0-NEXT: movl $16, %edi -; CHECK-O0-NEXT: movq %rsp, %rdx -; CHECK-O0-NEXT: movl $2, %ecx -; CHECK-O0-NEXT: callq __atomic_load@PLT -; CHECK-O0-NEXT: movaps (%rsp), %xmm0 -; CHECK-O0-NEXT: addq $24, %rsp -; CHECK-O0-NEXT: retq -; ; CHECK-SSE-O0-LABEL: atomic_vec4_float: ; CHECK-SSE-O0: # %bb.0: ; CHECK-SSE-O0-NEXT: subq $24, %rsp @@ -768,21 +462,6 @@ define <4 x float> @atomic_vec4_float(ptr %x) nounwind { } define <8 x double> @atomic_vec8_double(ptr %x) nounwind { -; CHECK-O3-LABEL: atomic_vec8_double: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: subq $72, %rsp -; CHECK-O3-NEXT: movq %rdi, %rsi -; CHECK-O3-NEXT: movq %rsp, %rdx -; CHECK-O3-NEXT: movl $64, %edi -; CHECK-O3-NEXT: movl $2, %ecx -; CHECK-O3-NEXT: callq __atomic_load@PLT -; CHECK-O3-NEXT: movaps (%rsp), %xmm0 -; CHECK-O3-NEXT: movaps {{[0-9]+}}(%rsp), %xmm1 -; CHECK-O3-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 -; CHECK-O3-NEXT: movaps {{[0-9]+}}(%rsp), %xmm3 -; CHECK-O3-NEXT: addq $72, %rsp -; CHECK-O3-NEXT: retq -; ; CHECK-SSE-O3-LABEL: atomic_vec8_double: ; CHECK-SSE-O3: # %bb.0: ; CHECK-SSE-O3-NEXT: subq $72, %rsp @@ -798,20 +477,30 @@ define <8 x double> @atomic_vec8_double(ptr %x) nounwind { ; CHECK-SSE-O3-NEXT: addq $72, %rsp ; CHECK-SSE-O3-NEXT: retq ; -; CHECK-O0-LABEL: atomic_vec8_double: -; CHECK-O0: # %bb.0: -; CHECK-O0-NEXT: subq $72, %rsp -; CHECK-O0-NEXT: movq %rdi, %rsi -; CHECK-O0-NEXT: movl $64, %edi -; CHECK-O0-NEXT: movq %rsp, %rdx -; CHECK-O0-NEXT: movl $2, %ecx -; CHECK-O0-NEXT: callq __atomic_load@PLT -; CHECK-O0-NEXT: movapd (%rsp), %xmm0 -; CHECK-O0-NEXT: movapd {{[0-9]+}}(%rsp), %xmm1 -; CHECK-O0-NEXT: movapd {{[0-9]+}}(%rsp), %xmm2 -; CHECK-O0-NEXT: movapd {{[0-9]+}}(%rsp), %xmm3 -; CHECK-O0-NEXT: addq $72, %rsp -; CHECK-O0-NEXT: retq +; CHECK-AVX2-O3-LABEL: atomic_vec8_double: +; CHECK-AVX2-O3: # %bb.0: +; CHECK-AVX2-O3-NEXT: subq $72, %rsp +; CHECK-AVX2-O3-NEXT: movq %rdi, %rsi +; CHECK-AVX2-O3-NEXT: movq %rsp, %rdx +; CHECK-AVX2-O3-NEXT: movl $64, %edi +; CHECK-AVX2-O3-NEXT: movl $2, %ecx +; CHECK-AVX2-O3-NEXT: callq __atomic_load@PLT +; CHECK-AVX2-O3-NEXT: vmovups (%rsp), %ymm0 +; CHECK-AVX2-O3-NEXT: vmovups {{[0-9]+}}(%rsp), %ymm1 +; CHECK-AVX2-O3-NEXT: addq $72, %rsp +; CHECK-AVX2-O3-NEXT: retq +; +; CHECK-AVX512-O3-LABEL: atomic_vec8_double: +; CHECK-AVX512-O3: # %bb.0: +; CHECK-AVX512-O3-NEXT: subq $72, %rsp +; CHECK-AVX512-O3-NEXT: movq %rdi, %rsi +; CHECK-AVX512-O3-NEXT: movq %rsp, %rdx +; CHECK-AVX512-O3-NEXT: movl $64, %edi +; CHECK-AVX512-O3-NEXT: movl $2, %ecx +; CHECK-AVX512-O3-NEXT: callq __atomic_load@PLT +; CHECK-AVX512-O3-NEXT: vmovups (%rsp), %zmm0 +; CHECK-AVX512-O3-NEXT: addq $72, %rsp +; CHECK-AVX512-O3-NEXT: retq ; ; CHECK-SSE-O0-LABEL: atomic_vec8_double: ; CHECK-SSE-O0: # %bb.0: @@ -827,24 +516,36 @@ define <8 x double> @atomic_vec8_double(ptr %x) nounwind { ; CHECK-SSE-O0-NEXT: movapd {{[0-9]+}}(%rsp), %xmm3 ; CHECK-SSE-O0-NEXT: addq $72, %rsp ; CHECK-SSE-O0-NEXT: retq +; +; CHECK-AVX2-O0-LABEL: atomic_vec8_double: +; CHECK-AVX2-O0: # %bb.0: +; CHECK-AVX2-O0-NEXT: subq $72, %rsp +; CHECK-AVX2-O0-NEXT: movq %rdi, %rsi +; CHECK-AVX2-O0-NEXT: movl $64, %edi +; CHECK-AVX2-O0-NEXT: movq %rsp, %rdx +; CHECK-AVX2-O0-NEXT: movl $2, %ecx +; CHECK-AVX2-O0-NEXT: callq __atomic_load@PLT +; CHECK-AVX2-O0-NEXT: vmovupd (%rsp), %ymm0 +; CHECK-AVX2-O0-NEXT: vmovupd {{[0-9]+}}(%rsp), %ymm1 +; CHECK-AVX2-O0-NEXT: addq $72, %rsp +; CHECK-AVX2-O0-NEXT: retq +; +; CHECK-AVX512-O0-LABEL: atomic_vec8_double: +; CHECK-AVX512-O0: # %bb.0: +; CHECK-AVX512-O0-NEXT: subq $72, %rsp +; CHECK-AVX512-O0-NEXT: movq %rdi, %rsi +; CHECK-AVX512-O0-NEXT: movl $64, %edi +; CHECK-AVX512-O0-NEXT: movq %rsp, %rdx +; CHECK-AVX512-O0-NEXT: movl $2, %ecx +; CHECK-AVX512-O0-NEXT: callq __atomic_load@PLT +; CHECK-AVX512-O0-NEXT: vmovupd (%rsp), %zmm0 +; CHECK-AVX512-O0-NEXT: addq $72, %rsp +; CHECK-AVX512-O0-NEXT: retq %ret = load atomic <8 x double>, ptr %x acquire, align 4 ret <8 x double> %ret } define <16 x bfloat> @atomic_vec16_bfloat(ptr %x) nounwind { -; CHECK-O3-LABEL: atomic_vec16_bfloat: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: subq $40, %rsp -; CHECK-O3-NEXT: movq %rdi, %rsi -; CHECK-O3-NEXT: movq %rsp, %rdx -; CHECK-O3-NEXT: movl $32, %edi -; CHECK-O3-NEXT: movl $2, %ecx -; CHECK-O3-NEXT: callq __atomic_load@PLT -; CHECK-O3-NEXT: movaps (%rsp), %xmm0 -; CHECK-O3-NEXT: movaps {{[0-9]+}}(%rsp), %xmm1 -; CHECK-O3-NEXT: addq $40, %rsp -; CHECK-O3-NEXT: retq -; ; CHECK-SSE-O3-LABEL: atomic_vec16_bfloat: ; CHECK-SSE-O3: # %bb.0: ; CHECK-SSE-O3-NEXT: subq $40, %rsp @@ -870,19 +571,6 @@ define <16 x bfloat> @atomic_vec16_bfloat(ptr %x) nounwind { ; CHECK-AVX-O3-NEXT: addq $40, %rsp ; CHECK-AVX-O3-NEXT: retq ; -; CHECK-O0-LABEL: atomic_vec16_bfloat: -; CHECK-O0: # %bb.0: -; CHECK-O0-NEXT: subq $40, %rsp -; CHECK-O0-NEXT: movq %rdi, %rsi -; CHECK-O0-NEXT: movl $32, %edi -; CHECK-O0-NEXT: movq %rsp, %rdx -; CHECK-O0-NEXT: movl $2, %ecx -; CHECK-O0-NEXT: callq __atomic_load@PLT -; CHECK-O0-NEXT: movaps (%rsp), %xmm0 -; CHECK-O0-NEXT: movaps {{[0-9]+}}(%rsp), %xmm1 -; CHECK-O0-NEXT: addq $40, %rsp -; CHECK-O0-NEXT: retq -; ; CHECK-SSE-O0-LABEL: atomic_vec16_bfloat: ; CHECK-SSE-O0: # %bb.0: ; CHECK-SSE-O0-NEXT: subq $40, %rsp @@ -912,21 +600,6 @@ define <16 x bfloat> @atomic_vec16_bfloat(ptr %x) nounwind { } define <32 x half> @atomic_vec32_half(ptr %x) nounwind { -; CHECK-O3-LABEL: atomic_vec32_half: -; CHECK-O3: # %bb.0: -; CHECK-O3-NEXT: subq $72, %rsp -; CHECK-O3-NEXT: movq %rdi, %rsi -; CHECK-O3-NEXT: movq %rsp, %rdx -; CHECK-O3-NEXT: movl $64, %edi -; CHECK-O3-NEXT: movl $2, %ecx -; CHECK-O3-NEXT: callq __atomic_load@PLT -; CHECK-O3-NEXT: movaps (%rsp), %xmm0 -; CHECK-O3-NEXT: movaps {{[0-9]+}}(%rsp), %xmm1 -; CHECK-O3-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 -; CHECK-O3-NEXT: movaps {{[0-9]+}}(%rsp), %xmm3 -; CHECK-O3-NEXT: addq $72, %rsp -; CHECK-O3-NEXT: retq -; ; CHECK-SSE-O3-LABEL: atomic_vec32_half: ; CHECK-SSE-O3: # %bb.0: ; CHECK-SSE-O3-NEXT: subq $72, %rsp @@ -942,20 +615,30 @@ define <32 x half> @atomic_vec32_half(ptr %x) nounwind { ; CHECK-SSE-O3-NEXT: addq $72, %rsp ; CHECK-SSE-O3-NEXT: retq ; -; CHECK-O0-LABEL: atomic_vec32_half: -; CHECK-O0: # %bb.0: -; CHECK-O0-NEXT: subq $72, %rsp -; CHECK-O0-NEXT: movq %rdi, %rsi -; CHECK-O0-NEXT: movl $64, %edi -; CHECK-O0-NEXT: movq %rsp, %rdx -; CHECK-O0-NEXT: movl $2, %ecx -; CHECK-O0-NEXT: callq __atomic_load@PLT -; CHECK-O0-NEXT: movaps (%rsp), %xmm0 -; CHECK-O0-NEXT: movaps {{[0-9]+}}(%rsp), %xmm1 -; CHECK-O0-NEXT: movaps {{[0-9]+}}(%rsp), %xmm2 -; CHECK-O0-NEXT: movaps {{[0-9]+}}(%rsp), %xmm3 -; CHECK-O0-NEXT: addq $72, %rsp -; CHECK-O0-NEXT: retq +; CHECK-AVX2-O3-LABEL: atomic_vec32_half: +; CHECK-AVX2-O3: # %bb.0: +; CHECK-AVX2-O3-NEXT: subq $72, %rsp +; CHECK-AVX2-O3-NEXT: movq %rdi, %rsi +; CHECK-AVX2-O3-NEXT: movq %rsp, %rdx +; CHECK-AVX2-O3-NEXT: movl $64, %edi +; CHECK-AVX2-O3-NEXT: movl $2, %ecx +; CHECK-AVX2-O3-NEXT: callq __atomic_load@PLT +; CHECK-AVX2-O3-NEXT: vmovups (%rsp), %ymm0 +; CHECK-AVX2-O3-NEXT: vmovups {{[0-9]+}}(%rsp), %ymm1 +; CHECK-AVX2-O3-NEXT: addq $72, %rsp +; CHECK-AVX2-O3-NEXT: retq +; +; CHECK-AVX512-O3-LABEL: atomic_vec32_half: +; CHECK-AVX512-O3: # %bb.0: +; CHECK-AVX512-O3-NEXT: subq $72, %rsp +; CHECK-AVX512-O3-NEXT: movq %rdi, %rsi +; CHECK-AVX512-O3-NEXT: movq %rsp, %rdx +; CHECK-AVX512-O3-NEXT: movl $64, %edi +; CHECK-AVX512-O3-NEXT: movl $2, %ecx +; CHECK-AVX512-O3-NEXT: callq __atomic_load@PLT +; CHECK-AVX512-O3-NEXT: vmovups (%rsp), %zmm0 +; CHECK-AVX512-O3-NEXT: addq $72, %rsp +; CHECK-AVX512-O3-NEXT: retq ; ; CHECK-SSE-O0-LABEL: atomic_vec32_half: ; CHECK-SSE-O0: # %bb.0: @@ -971,6 +654,31 @@ define <32 x half> @atomic_vec32_half(ptr %x) nounwind { ; CHECK-SSE-O0-NEXT: movaps {{[0-9]+}}(%rsp), %xmm3 ; CHECK-SSE-O0-NEXT: addq $72, %rsp ; CHECK-SSE-O0-NEXT: retq +; +; CHECK-AVX2-O0-LABEL: atomic_vec32_half: +; CHECK-AVX2-O0: # %bb.0: +; CHECK-AVX2-O0-NEXT: subq $72, %rsp +; CHECK-AVX2-O0-NEXT: movq %rdi, %rsi +; CHECK-AVX2-O0-NEXT: movl $64, %edi +; CHECK-AVX2-O0-NEXT: movq %rsp, %rdx +; CHECK-AVX2-O0-NEXT: movl $2, %ecx +; CHECK-AVX2-O0-NEXT: callq __atomic_load@PLT +; CHECK-AVX2-O0-NEXT: vmovups (%rsp), %ymm0 +; CHECK-AVX2-O0-NEXT: vmovups {{[0-9]+}}(%rsp), %ymm1 +; CHECK-AVX2-O0-NEXT: addq $72, %rsp +; CHECK-AVX2-O0-NEXT: retq +; +; CHECK-AVX512-O0-LABEL: atomic_vec32_half: +; CHECK-AVX512-O0: # %bb.0: +; CHECK-AVX512-O0-NEXT: subq $72, %rsp +; CHECK-AVX512-O0-NEXT: movq %rdi, %rsi +; CHECK-AVX512-O0-NEXT: movl $64, %edi +; CHECK-AVX512-O0-NEXT: movq %rsp, %rdx +; CHECK-AVX512-O0-NEXT: movl $2, %ecx +; CHECK-AVX512-O0-NEXT: callq __atomic_load@PLT +; CHECK-AVX512-O0-NEXT: vmovups (%rsp), %zmm0 +; CHECK-AVX512-O0-NEXT: addq $72, %rsp +; CHECK-AVX512-O0-NEXT: retq %ret = load atomic <32 x half>, ptr %x acquire, align 4 ret <32 x half> %ret } diff --git a/llvm/test/CodeGen/X86/basic-block-sections-bb-hash.ll b/llvm/test/CodeGen/X86/basic-block-sections-bb-hash.ll new file mode 100644 index 0000000..293b48d --- /dev/null +++ b/llvm/test/CodeGen/X86/basic-block-sections-bb-hash.ll @@ -0,0 +1,39 @@ +;; BB section test with basic block hashes. + +;; basic block sections Profile with bb hashes +; RUN: echo 'v1' > %t +; RUN: echo 'f foo' >> %t +; RUN: echo 'g 0:10,1:9,2:1 1:8,3:8 2:2,3:2 3:11' >> %t +; RUN: echo 'c 0 2 3' >> %t +; RUN: echo 'h 0:64863A11B5CA0000 1:54F1E80D6B270006 2:54F1F4E66B270008 3:C8BC6041A2CB0009' >> %t +; RUN: llc < %s -O0 -mtriple=x86_64-pc-linux -function-sections -basic-block-sections=%t | FileCheck %s +; +define void @foo(i1 zeroext) nounwind { + %2 = alloca i8, align 1 + %3 = zext i1 %0 to i8 + store i8 %3, ptr %2, align 1 + %4 = load i8, ptr %2, align 1 + %5 = trunc i8 %4 to i1 + br i1 %5, label %6, label %8 + +6: ; preds = %1 + %7 = call i32 @bar() + br label %10 + +8: ; preds = %1 + %9 = call i32 @baz() + br label %10 + +10: ; preds = %8, %6 + ret void +} + +declare i32 @bar() #1 + +declare i32 @baz() #1 + +; CHECK: .section .text.foo,"ax",@progbits +; CHECK: callq baz +; CHECK: retq +; CHECK: .section .text.split.foo,"ax",@progbits +; CHECK: callq bar diff --git a/llvm/test/CodeGen/X86/basic-block-sections-clusters-error.ll b/llvm/test/CodeGen/X86/basic-block-sections-clusters-error.ll index 751ab76..eb0a14b 100644 --- a/llvm/test/CodeGen/X86/basic-block-sections-clusters-error.ll +++ b/llvm/test/CodeGen/X86/basic-block-sections-clusters-error.ll @@ -69,6 +69,20 @@ ; RUN: echo 'g 0:4,1:2:3' >> %t15 ; RUN: not --crash llc < %s -O0 -mtriple=x86_64 -function-sections -basic-block-sections=%t15 2>&1 | FileCheck %s --check-prefix=CHECK-ERROR15 ; CHECK-ERROR15: LLVM ERROR: invalid profile {{.*}} at line 4: unsigned integer expected: '2:3' +; RUN: echo 'v1' > %t16 +; RUN: echo 'f dummy1' >> %t16 +; RUN: echo 'c 0 1' >> %t16 +; RUN: echo 'g 0:4,1:2' >> %t16 +; RUN: echo 'h a:1111111111111111 1:ffffffffffffffff' >> %t16 +; RUN: not --crash llc < %s -O0 -mtriple=x86_64 -function-sections -basic-block-sections=%t16 2>&1 | FileCheck %s --check-prefix=CHECK-ERROR16 +; CHECK-ERROR16: LLVM ERROR: invalid profile {{.*}} at line 5: unsigned integer expected: 'a' +; RUN: echo 'v1' > %t17 +; RUN: echo 'f dummy1' >> %t17 +; RUN: echo 'c 0 1' >> %t17 +; RUN: echo 'g 0:4,1:2' >> %t17 +; RUN: echo 'h 0:111111111111111g 1:ffffffffffffffff' >> %t17 +; RUN: not --crash llc < %s -O0 -mtriple=x86_64 -function-sections -basic-block-sections=%t17 2>&1 | FileCheck %s --check-prefix=CHECK-ERROR17 +; CHECK-ERROR17: LLVM ERROR: invalid profile {{.*}} at line 5: unsigned integer expected in hex format: '111111111111111g' define i32 @dummy1(i32 %x, i32 %y, i32 %z) { diff --git a/llvm/test/CodeGen/X86/bittest-big-integer.ll b/llvm/test/CodeGen/X86/bittest-big-integer.ll index 19d751d1..8007d9d 100644 --- a/llvm/test/CodeGen/X86/bittest-big-integer.ll +++ b/llvm/test/CodeGen/X86/bittest-big-integer.ll @@ -7025,3 +7025,279 @@ define i1 @test_ne_i4096(ptr %word, i32 %position) nounwind { %cmp = icmp ne i4096 %test, 0 ret i1 %cmp } + +; Special Cases + +; Multiple uses of the stored value +define i1 @complement_cmpz_i128(ptr %word, i32 %position) nounwind { +; X86-LABEL: complement_cmpz_i128: +; X86: # %bb.0: +; X86-NEXT: pushl %ebp +; X86-NEXT: movl %esp, %ebp +; X86-NEXT: pushl %ebx +; X86-NEXT: pushl %edi +; X86-NEXT: pushl %esi +; X86-NEXT: andl $-16, %esp +; X86-NEXT: subl $64, %esp +; X86-NEXT: movzbl 12(%ebp), %ecx +; X86-NEXT: movl $0, {{[0-9]+}}(%esp) +; X86-NEXT: movl $0, {{[0-9]+}}(%esp) +; X86-NEXT: movl $0, {{[0-9]+}}(%esp) +; X86-NEXT: movl $1, {{[0-9]+}}(%esp) +; X86-NEXT: movl $0, {{[0-9]+}}(%esp) +; X86-NEXT: movl $0, {{[0-9]+}}(%esp) +; X86-NEXT: movl $0, {{[0-9]+}}(%esp) +; X86-NEXT: movl $0, {{[0-9]+}}(%esp) +; X86-NEXT: movl %ecx, %eax +; X86-NEXT: shrb $3, %al +; X86-NEXT: andb $12, %al +; X86-NEXT: negb %al +; X86-NEXT: movsbl %al, %esi +; X86-NEXT: movl 36(%esp,%esi), %eax +; X86-NEXT: movl 40(%esp,%esi), %edi +; X86-NEXT: movl %edi, %edx +; X86-NEXT: shldl %cl, %eax, %edx +; X86-NEXT: movl 32(%esp,%esi), %ebx +; X86-NEXT: movl %ebx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; X86-NEXT: movl 44(%esp,%esi), %esi +; X86-NEXT: shldl %cl, %edi, %esi +; X86-NEXT: movl %ebx, %edi +; X86-NEXT: shll %cl, %edi +; X86-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ebx # 4-byte Reload +; X86-NEXT: shldl %cl, %ebx, %eax +; X86-NEXT: movl 8(%ebp), %ecx +; X86-NEXT: xorl 12(%ecx), %esi +; X86-NEXT: xorl 8(%ecx), %edx +; X86-NEXT: xorl 4(%ecx), %eax +; X86-NEXT: xorl (%ecx), %edi +; X86-NEXT: movl %edx, 8(%ecx) +; X86-NEXT: movl %esi, 12(%ecx) +; X86-NEXT: movl %edi, (%ecx) +; X86-NEXT: movl %eax, 4(%ecx) +; X86-NEXT: orl %esi, %eax +; X86-NEXT: orl %edx, %edi +; X86-NEXT: orl %eax, %edi +; X86-NEXT: setne %al +; X86-NEXT: leal -12(%ebp), %esp +; X86-NEXT: popl %esi +; X86-NEXT: popl %edi +; X86-NEXT: popl %ebx +; X86-NEXT: popl %ebp +; X86-NEXT: retl +; +; SSE-LABEL: complement_cmpz_i128: +; SSE: # %bb.0: +; SSE-NEXT: movl %esi, %ecx +; SSE-NEXT: movl $1, %eax +; SSE-NEXT: xorl %edx, %edx +; SSE-NEXT: shldq %cl, %rax, %rdx +; SSE-NEXT: shlq %cl, %rax +; SSE-NEXT: xorl %esi, %esi +; SSE-NEXT: testb $64, %cl +; SSE-NEXT: cmovneq %rax, %rdx +; SSE-NEXT: cmovneq %rsi, %rax +; SSE-NEXT: xorq 8(%rdi), %rdx +; SSE-NEXT: xorq (%rdi), %rax +; SSE-NEXT: movq %rax, (%rdi) +; SSE-NEXT: movq %rdx, 8(%rdi) +; SSE-NEXT: orq %rdx, %rax +; SSE-NEXT: setne %al +; SSE-NEXT: retq +; +; AVX2-LABEL: complement_cmpz_i128: +; AVX2: # %bb.0: +; AVX2-NEXT: movl %esi, %ecx +; AVX2-NEXT: movl $1, %eax +; AVX2-NEXT: xorl %edx, %edx +; AVX2-NEXT: shldq %cl, %rax, %rdx +; AVX2-NEXT: xorl %esi, %esi +; AVX2-NEXT: shlxq %rcx, %rax, %rax +; AVX2-NEXT: testb $64, %cl +; AVX2-NEXT: cmovneq %rax, %rdx +; AVX2-NEXT: cmovneq %rsi, %rax +; AVX2-NEXT: xorq 8(%rdi), %rdx +; AVX2-NEXT: xorq (%rdi), %rax +; AVX2-NEXT: movq %rax, (%rdi) +; AVX2-NEXT: movq %rdx, 8(%rdi) +; AVX2-NEXT: orq %rdx, %rax +; AVX2-NEXT: setne %al +; AVX2-NEXT: retq +; +; AVX512-LABEL: complement_cmpz_i128: +; AVX512: # %bb.0: +; AVX512-NEXT: movl %esi, %ecx +; AVX512-NEXT: xorl %eax, %eax +; AVX512-NEXT: movl $1, %edx +; AVX512-NEXT: xorl %esi, %esi +; AVX512-NEXT: shldq %cl, %rdx, %rsi +; AVX512-NEXT: shlxq %rcx, %rdx, %rdx +; AVX512-NEXT: testb $64, %cl +; AVX512-NEXT: cmovneq %rdx, %rsi +; AVX512-NEXT: cmovneq %rax, %rdx +; AVX512-NEXT: xorq 8(%rdi), %rsi +; AVX512-NEXT: xorq (%rdi), %rdx +; AVX512-NEXT: movq %rdx, (%rdi) +; AVX512-NEXT: movq %rsi, 8(%rdi) +; AVX512-NEXT: orq %rsi, %rdx +; AVX512-NEXT: setne %al +; AVX512-NEXT: retq + %rem = and i32 %position, 127 + %ofs = zext nneg i32 %rem to i128 + %bit = shl nuw i128 1, %ofs + %ld = load i128, ptr %word + %res = xor i128 %ld, %bit + store i128 %res, ptr %word + %cmp = icmp ne i128 %res, 0 + ret i1 %cmp +} + +; Multiple loads in store chain +define i32 @reset_multiload_i128(ptr %word, i32 %position, ptr %p) nounwind { +; X86-LABEL: reset_multiload_i128: +; X86: # %bb.0: +; X86-NEXT: pushl %ebp +; X86-NEXT: movl %esp, %ebp +; X86-NEXT: pushl %ebx +; X86-NEXT: pushl %edi +; X86-NEXT: pushl %esi +; X86-NEXT: andl $-16, %esp +; X86-NEXT: subl $80, %esp +; X86-NEXT: movzbl 12(%ebp), %ecx +; X86-NEXT: movl $0, {{[0-9]+}}(%esp) +; X86-NEXT: movl $0, {{[0-9]+}}(%esp) +; X86-NEXT: movl $0, {{[0-9]+}}(%esp) +; X86-NEXT: movl $1, {{[0-9]+}}(%esp) +; X86-NEXT: movl $0, {{[0-9]+}}(%esp) +; X86-NEXT: movl $0, {{[0-9]+}}(%esp) +; X86-NEXT: movl $0, {{[0-9]+}}(%esp) +; X86-NEXT: movl $0, {{[0-9]+}}(%esp) +; X86-NEXT: movl %ecx, %eax +; X86-NEXT: shrb $3, %al +; X86-NEXT: andb $12, %al +; X86-NEXT: negb %al +; X86-NEXT: movsbl %al, %eax +; X86-NEXT: movl 56(%esp,%eax), %esi +; X86-NEXT: movl 60(%esp,%eax), %edx +; X86-NEXT: shldl %cl, %esi, %edx +; X86-NEXT: movl %edx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; X86-NEXT: movl 48(%esp,%eax), %edi +; X86-NEXT: movl 52(%esp,%eax), %eax +; X86-NEXT: shldl %cl, %eax, %esi +; X86-NEXT: movl %esi, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; X86-NEXT: shldl %cl, %edi, %eax +; X86-NEXT: movl 8(%ebp), %ebx +; X86-NEXT: shll %cl, %edi +; X86-NEXT: movl 8(%ebx), %ecx +; X86-NEXT: movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; X86-NEXT: andl %esi, %ecx +; X86-NEXT: movl (%ebx), %esi +; X86-NEXT: movl %esi, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; X86-NEXT: andl %edi, %esi +; X86-NEXT: orl %ecx, %esi +; X86-NEXT: movl 12(%ebx), %ecx +; X86-NEXT: movl %ecx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; X86-NEXT: andl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Folded Reload +; X86-NEXT: movl %eax, %edx +; X86-NEXT: movl 4(%ebx), %ebx +; X86-NEXT: movl %ebx, {{[-0-9]+}}(%e{{[sb]}}p) # 4-byte Spill +; X86-NEXT: andl %ebx, %eax +; X86-NEXT: orl %ecx, %eax +; X86-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Reload +; X86-NEXT: notl %ecx +; X86-NEXT: andl {{[-0-9]+}}(%e{{[sb]}}p), %ecx # 4-byte Folded Reload +; X86-NEXT: movl {{[-0-9]+}}(%e{{[sb]}}p), %ebx # 4-byte Reload +; X86-NEXT: notl %ebx +; X86-NEXT: andl {{[-0-9]+}}(%e{{[sb]}}p), %ebx # 4-byte Folded Reload +; X86-NEXT: notl %edx +; X86-NEXT: andl {{[-0-9]+}}(%e{{[sb]}}p), %edx # 4-byte Folded Reload +; X86-NEXT: notl %edi +; X86-NEXT: andl {{[-0-9]+}}(%e{{[sb]}}p), %edi # 4-byte Folded Reload +; X86-NEXT: orl %esi, %eax +; X86-NEXT: movl 16(%ebp), %eax +; X86-NEXT: movl (%eax), %eax +; X86-NEXT: movl 8(%ebp), %esi +; X86-NEXT: movl %ebx, 8(%esi) +; X86-NEXT: movl %ecx, 12(%esi) +; X86-NEXT: movl %edi, (%esi) +; X86-NEXT: movl %edx, 4(%esi) +; X86-NEXT: je .LBB22_2 +; X86-NEXT: # %bb.1: +; X86-NEXT: xorl %eax, %eax +; X86-NEXT: .LBB22_2: +; X86-NEXT: leal -12(%ebp), %esp +; X86-NEXT: popl %esi +; X86-NEXT: popl %edi +; X86-NEXT: popl %ebx +; X86-NEXT: popl %ebp +; X86-NEXT: retl +; +; SSE-LABEL: reset_multiload_i128: +; SSE: # %bb.0: +; SSE-NEXT: movl %esi, %ecx +; SSE-NEXT: movl $1, %esi +; SSE-NEXT: xorl %r8d, %r8d +; SSE-NEXT: shldq %cl, %rsi, %r8 +; SSE-NEXT: xorl %eax, %eax +; SSE-NEXT: shlq %cl, %rsi +; SSE-NEXT: testb $64, %cl +; SSE-NEXT: cmovneq %rsi, %r8 +; SSE-NEXT: cmovneq %rax, %rsi +; SSE-NEXT: movq (%rdi), %rcx +; SSE-NEXT: movq 8(%rdi), %r9 +; SSE-NEXT: movq %r9, %r10 +; SSE-NEXT: andq %r8, %r10 +; SSE-NEXT: notq %r8 +; SSE-NEXT: movq %rcx, %r11 +; SSE-NEXT: andq %rsi, %r11 +; SSE-NEXT: notq %rsi +; SSE-NEXT: andq %r9, %r8 +; SSE-NEXT: andq %rcx, %rsi +; SSE-NEXT: orq %r10, %r11 +; SSE-NEXT: jne .LBB22_2 +; SSE-NEXT: # %bb.1: +; SSE-NEXT: movl (%rdx), %eax +; SSE-NEXT: .LBB22_2: +; SSE-NEXT: movq %rsi, (%rdi) +; SSE-NEXT: movq %r8, 8(%rdi) +; SSE-NEXT: # kill: def $eax killed $eax killed $rax +; SSE-NEXT: retq +; +; AVX-LABEL: reset_multiload_i128: +; AVX: # %bb.0: +; AVX-NEXT: movl %esi, %ecx +; AVX-NEXT: movl $1, %esi +; AVX-NEXT: xorl %r8d, %r8d +; AVX-NEXT: shldq %cl, %rsi, %r8 +; AVX-NEXT: xorl %eax, %eax +; AVX-NEXT: shlxq %rcx, %rsi, %r9 +; AVX-NEXT: testb $64, %cl +; AVX-NEXT: cmovneq %r9, %r8 +; AVX-NEXT: cmovneq %rax, %r9 +; AVX-NEXT: movq (%rdi), %r10 +; AVX-NEXT: movq 8(%rdi), %r11 +; AVX-NEXT: andnq %r11, %r8, %rcx +; AVX-NEXT: andq %r8, %r11 +; AVX-NEXT: andnq %r10, %r9, %rsi +; AVX-NEXT: andq %r9, %r10 +; AVX-NEXT: orq %r11, %r10 +; AVX-NEXT: jne .LBB22_2 +; AVX-NEXT: # %bb.1: +; AVX-NEXT: movl (%rdx), %eax +; AVX-NEXT: .LBB22_2: +; AVX-NEXT: movq %rsi, (%rdi) +; AVX-NEXT: movq %rcx, 8(%rdi) +; AVX-NEXT: # kill: def $eax killed $eax killed $rax +; AVX-NEXT: retq + %rem = and i32 %position, 127 + %ofs = zext nneg i32 %rem to i128 + %bit = shl nuw i128 1, %ofs + %mask = xor i128 %bit, -1 + %ld = load i128, ptr %word + %sel = load i32, ptr %p + %test = and i128 %ld, %bit + %res = and i128 %ld, %mask + %cmp = icmp eq i128 %test, 0 + store i128 %res, ptr %word + %ret = select i1 %cmp, i32 %sel, i32 0 + ret i32 %ret +} diff --git a/llvm/test/CodeGen/X86/ipra-reg-usage.ll b/llvm/test/CodeGen/X86/ipra-reg-usage.ll index e73ff79..f270f8f 100644 --- a/llvm/test/CodeGen/X86/ipra-reg-usage.ll +++ b/llvm/test/CodeGen/X86/ipra-reg-usage.ll @@ -7,7 +7,7 @@ target triple = "x86_64-unknown-unknown" declare void @bar1() define preserve_allcc void @foo()#0 { -; CHECK: foo Clobbered Registers: $cs $df $ds $eflags $eip $eiz $es $esp $fpcw $fpsw $fs $fs_base $gs $gs_base $hip $hsp $ip $mxcsr $rflags $rip $riz $rsp $sp $sph $spl $ss $ssp $_eflags $cr0 $cr1 $cr2 $cr3 $cr4 $cr5 $cr6 $cr7 $cr8 $cr9 $cr10 $cr11 $cr12 $cr13 $cr14 $cr15 $dr0 $dr1 $dr2 $dr3 $dr4 $dr5 $dr6 $dr7 $dr8 $dr9 $dr10 $dr11 $dr12 $dr13 $dr14 $dr15 $fp0 $fp1 $fp2 $fp3 $fp4 $fp5 $fp6 $fp7 $mm0 $mm1 $mm2 $mm3 $mm4 $mm5 $mm6 $mm7 $r11 $st0 $st1 $st2 $st3 $st4 $st5 $st6 $st7 $r11b $r11bh $r11d $r11w $r11wh $ymm0 $ymm1 $ymm2 $ymm3 $ymm4 $ymm5 $ymm6 $ymm7 $ymm8 $ymm9 $ymm10 $ymm11 $ymm12 $ymm13 $ymm14 $ymm15 $k0 $k1 $k2 $k3 $k4 $k5 $k6 $k7 $xmm16 $xmm17 $xmm18 $xmm19 $xmm20 $xmm21 $xmm22 $xmm23 $xmm24 $xmm25 $xmm26 $xmm27 $xmm28 $xmm29 $xmm30 $xmm31 $ymm16 $ymm17 $ymm18 $ymm19 $ymm20 $ymm21 $ymm22 $ymm23 $ymm24 $ymm25 $ymm26 $ymm27 $ymm28 $ymm29 $ymm30 $ymm31 $zmm0 $zmm1 $zmm2 $zmm3 $zmm4 $zmm5 $zmm6 $zmm7 $zmm8 $zmm9 $zmm10 $zmm11 $zmm12 $zmm13 $zmm14 $zmm15 $zmm16 $zmm17 $zmm18 $zmm19 $zmm20 $zmm21 $zmm22 $zmm23 $zmm24 $zmm25 $zmm26 $zmm27 $zmm28 $zmm29 $zmm30 $zmm31 $k0_k1 $k2_k3 $k4_k5 $k6_k7 $tmmcfg $tmm0 $tmm1 $tmm2 $tmm3 $tmm4 $tmm5 $tmm6 $tmm7 $tmm0_tmm1 $tmm2_tmm3 $tmm4_tmm5 $tmm6_tmm7 $r16 $r17 $r18 $r19 $r20 $r21 $r22 $r23 $r24 $r25 $r26 $r27 $r28 $r29 $r30 $r31 $r16b $r17b $r18b $r19b $r20b $r21b $r22b $r23b $r24b $r25b $r26b $r27b $r28b $r29b $r30b $r31b $r16bh $r17bh $r18bh $r19bh $r20bh $r21bh $r22bh $r23bh $r24bh $r25bh $r26bh $r27bh $r28bh $r29bh $r30bh $r31bh $r16d $r17d $r18d $r19d $r20d $r21d $r22d $r23d $r24d $r25d $r26d $r27d $r28d $r29d $r30d $r31d $r16w $r17w $r18w $r19w $r20w $r21w $r22w $r23w $r24w $r25w $r26w $r27w $r28w $r29w $r30w $r31w $r16wh $r17wh $r18wh $r19wh $r20wh $r21wh $r22wh $r23wh $r24wh $r25wh $r26wh $r27wh $r28wh $r29wh $r30wh $r31wh +; CHECK: foo Clobbered Registers: $cs $df $ds $eflags $eip $eiz $es $esp $fpcw $fpsw $fs $fs_base $gs $gs_base $hip $hsp $ip $mxcsr $rflags $rip $riz $rsp $sp $sph $spl $ss $ssp $_eflags $cr0 $cr1 $cr2 $cr3 $cr4 $cr5 $cr6 $cr7 $cr8 $cr9 $cr10 $cr11 $cr12 $cr13 $cr14 $cr15 $dr0 $dr1 $dr2 $dr3 $dr4 $dr5 $dr6 $dr7 $dr8 $dr9 $dr10 $dr11 $dr12 $dr13 $dr14 $dr15 $fp0 $fp1 $fp2 $fp3 $fp4 $fp5 $fp6 $fp7 $mm0 $mm1 $mm2 $mm3 $mm4 $mm5 $mm6 $mm7 $r11 $st0 $st1 $st2 $st3 $st4 $st5 $st6 $st7 $r11b $r11bh $r11d $r11w $r11wh $ymm0 $ymm1 $ymm2 $ymm3 $ymm4 $ymm5 $ymm6 $ymm7 $ymm8 $ymm9 $ymm10 $ymm11 $ymm12 $ymm13 $ymm14 $ymm15 $k0 $k1 $k2 $k3 $k4 $k5 $k6 $k7 $xmm16 $xmm17 $xmm18 $xmm19 $xmm20 $xmm21 $xmm22 $xmm23 $xmm24 $xmm25 $xmm26 $xmm27 $xmm28 $xmm29 $xmm30 $xmm31 $ymm16 $ymm17 $ymm18 $ymm19 $ymm20 $ymm21 $ymm22 $ymm23 $ymm24 $ymm25 $ymm26 $ymm27 $ymm28 $ymm29 $ymm30 $ymm31 $zmm0 $zmm1 $zmm2 $zmm3 $zmm4 $zmm5 $zmm6 $zmm7 $zmm8 $zmm9 $zmm10 $zmm11 $zmm12 $zmm13 $zmm14 $zmm15 $zmm16 $zmm17 $zmm18 $zmm19 $zmm20 $zmm21 $zmm22 $zmm23 $zmm24 $zmm25 $zmm26 $zmm27 $zmm28 $zmm29 $zmm30 $zmm31 $k0_k1 $k2_k3 $k4_k5 $k6_k7 $tmmcfg $tmm0 $tmm1 $tmm2 $tmm3 $tmm4 $tmm5 $tmm6 $tmm7 $r16 $r17 $r18 $r19 $r20 $r21 $r22 $r23 $r24 $r25 $r26 $r27 $r28 $r29 $r30 $r31 $r16b $r17b $r18b $r19b $r20b $r21b $r22b $r23b $r24b $r25b $r26b $r27b $r28b $r29b $r30b $r31b $r16bh $r17bh $r18bh $r19bh $r20bh $r21bh $r22bh $r23bh $r24bh $r25bh $r26bh $r27bh $r28bh $r29bh $r30bh $r31bh $r16d $r17d $r18d $r19d $r20d $r21d $r22d $r23d $r24d $r25d $r26d $r27d $r28d $r29d $r30d $r31d $r16w $r17w $r18w $r19w $r20w $r21w $r22w $r23w $r24w $r25w $r26w $r27w $r28w $r29w $r30w $r31w $r16wh $r17wh $r18wh $r19wh $r20wh $r21wh $r22wh $r23wh $r24wh $r25wh $r26wh $r27wh $r28wh $r29wh $r30wh $r31wh call void @bar1() call void @bar2() ret void @@ -15,7 +15,7 @@ define preserve_allcc void @foo()#0 { declare void @bar2() define preserve_nonecc void @foo2()#0 { -; CHECK: foo2 Clobbered Registers: $ah $al $ax $ch $cl $cs $cx $df $dh $di $dih $dil $dl $ds $dx $eax $ecx $edi $edx $eflags $eip $eiz $es $esi $esp $fpcw $fpsw $fs $fs_base $gs $gs_base $hax $hcx $hdi $hdx $hip $hsi $hsp $ip $mxcsr $rax $rcx $rdi $rdx $rflags $rip $riz $rsi $rsp $si $sih $sil $sp $sph $spl $ss $ssp $_eflags $cr0 $cr1 $cr2 $cr3 $cr4 $cr5 $cr6 $cr7 $cr8 $cr9 $cr10 $cr11 $cr12 $cr13 $cr14 $cr15 $dr0 $dr1 $dr2 $dr3 $dr4 $dr5 $dr6 $dr7 $dr8 $dr9 $dr10 $dr11 $dr12 $dr13 $dr14 $dr15 $fp0 $fp1 $fp2 $fp3 $fp4 $fp5 $fp6 $fp7 $mm0 $mm1 $mm2 $mm3 $mm4 $mm5 $mm6 $mm7 $r8 $r9 $r10 $r11 $st0 $st1 $st2 $st3 $st4 $st5 $st6 $st7 $xmm0 $xmm1 $xmm2 $xmm3 $xmm4 $xmm5 $xmm6 $xmm7 $xmm8 $xmm9 $xmm10 $xmm11 $xmm12 $xmm13 $xmm14 $xmm15 $r8b $r9b $r10b $r11b $r8bh $r9bh $r10bh $r11bh $r8d $r9d $r10d $r11d $r8w $r9w $r10w $r11w $r8wh $r9wh $r10wh $r11wh $ymm0 $ymm1 $ymm2 $ymm3 $ymm4 $ymm5 $ymm6 $ymm7 $ymm8 $ymm9 $ymm10 $ymm11 $ymm12 $ymm13 $ymm14 $ymm15 $k0 $k1 $k2 $k3 $k4 $k5 $k6 $k7 $xmm16 $xmm17 $xmm18 $xmm19 $xmm20 $xmm21 $xmm22 $xmm23 $xmm24 $xmm25 $xmm26 $xmm27 $xmm28 $xmm29 $xmm30 $xmm31 $ymm16 $ymm17 $ymm18 $ymm19 $ymm20 $ymm21 $ymm22 $ymm23 $ymm24 $ymm25 $ymm26 $ymm27 $ymm28 $ymm29 $ymm30 $ymm31 $zmm0 $zmm1 $zmm2 $zmm3 $zmm4 $zmm5 $zmm6 $zmm7 $zmm8 $zmm9 $zmm10 $zmm11 $zmm12 $zmm13 $zmm14 $zmm15 $zmm16 $zmm17 $zmm18 $zmm19 $zmm20 $zmm21 $zmm22 $zmm23 $zmm24 $zmm25 $zmm26 $zmm27 $zmm28 $zmm29 $zmm30 $zmm31 $k0_k1 $k2_k3 $k4_k5 $k6_k7 $tmmcfg $tmm0 $tmm1 $tmm2 $tmm3 $tmm4 $tmm5 $tmm6 $tmm7 $tmm0_tmm1 $tmm2_tmm3 $tmm4_tmm5 $tmm6_tmm7 $r16 $r17 $r18 $r19 $r20 $r21 $r22 $r23 $r24 $r25 $r26 $r27 $r28 $r29 $r30 $r31 $r16b $r17b $r18b $r19b $r20b $r21b $r22b $r23b $r24b $r25b $r26b $r27b $r28b $r29b $r30b $r31b $r16bh $r17bh $r18bh $r19bh $r20bh $r21bh $r22bh $r23bh $r24bh $r25bh $r26bh $r27bh $r28bh $r29bh $r30bh $r31bh $r16d $r17d $r18d $r19d $r20d $r21d $r22d $r23d $r24d $r25d $r26d $r27d $r28d $r29d $r30d $r31d $r16w $r17w $r18w $r19w $r20w $r21w $r22w $r23w $r24w $r25w $r26w $r27w $r28w $r29w $r30w $r31w $r16wh $r17wh $r18wh $r19wh $r20wh $r21wh $r22wh $r23wh $r24wh $r25wh $r26wh $r27wh $r28wh $r29wh $r30wh $r31wh +; CHECK: foo2 Clobbered Registers: $ah $al $ax $ch $cl $cs $cx $df $dh $di $dih $dil $dl $ds $dx $eax $ecx $edi $edx $eflags $eip $eiz $es $esi $esp $fpcw $fpsw $fs $fs_base $gs $gs_base $hax $hcx $hdi $hdx $hip $hsi $hsp $ip $mxcsr $rax $rcx $rdi $rdx $rflags $rip $riz $rsi $rsp $si $sih $sil $sp $sph $spl $ss $ssp $_eflags $cr0 $cr1 $cr2 $cr3 $cr4 $cr5 $cr6 $cr7 $cr8 $cr9 $cr10 $cr11 $cr12 $cr13 $cr14 $cr15 $dr0 $dr1 $dr2 $dr3 $dr4 $dr5 $dr6 $dr7 $dr8 $dr9 $dr10 $dr11 $dr12 $dr13 $dr14 $dr15 $fp0 $fp1 $fp2 $fp3 $fp4 $fp5 $fp6 $fp7 $mm0 $mm1 $mm2 $mm3 $mm4 $mm5 $mm6 $mm7 $r8 $r9 $r10 $r11 $st0 $st1 $st2 $st3 $st4 $st5 $st6 $st7 $xmm0 $xmm1 $xmm2 $xmm3 $xmm4 $xmm5 $xmm6 $xmm7 $xmm8 $xmm9 $xmm10 $xmm11 $xmm12 $xmm13 $xmm14 $xmm15 $r8b $r9b $r10b $r11b $r8bh $r9bh $r10bh $r11bh $r8d $r9d $r10d $r11d $r8w $r9w $r10w $r11w $r8wh $r9wh $r10wh $r11wh $ymm0 $ymm1 $ymm2 $ymm3 $ymm4 $ymm5 $ymm6 $ymm7 $ymm8 $ymm9 $ymm10 $ymm11 $ymm12 $ymm13 $ymm14 $ymm15 $k0 $k1 $k2 $k3 $k4 $k5 $k6 $k7 $xmm16 $xmm17 $xmm18 $xmm19 $xmm20 $xmm21 $xmm22 $xmm23 $xmm24 $xmm25 $xmm26 $xmm27 $xmm28 $xmm29 $xmm30 $xmm31 $ymm16 $ymm17 $ymm18 $ymm19 $ymm20 $ymm21 $ymm22 $ymm23 $ymm24 $ymm25 $ymm26 $ymm27 $ymm28 $ymm29 $ymm30 $ymm31 $zmm0 $zmm1 $zmm2 $zmm3 $zmm4 $zmm5 $zmm6 $zmm7 $zmm8 $zmm9 $zmm10 $zmm11 $zmm12 $zmm13 $zmm14 $zmm15 $zmm16 $zmm17 $zmm18 $zmm19 $zmm20 $zmm21 $zmm22 $zmm23 $zmm24 $zmm25 $zmm26 $zmm27 $zmm28 $zmm29 $zmm30 $zmm31 $k0_k1 $k2_k3 $k4_k5 $k6_k7 $tmmcfg $tmm0 $tmm1 $tmm2 $tmm3 $tmm4 $tmm5 $tmm6 $tmm7 $r16 $r17 $r18 $r19 $r20 $r21 $r22 $r23 $r24 $r25 $r26 $r27 $r28 $r29 $r30 $r31 $r16b $r17b $r18b $r19b $r20b $r21b $r22b $r23b $r24b $r25b $r26b $r27b $r28b $r29b $r30b $r31b $r16bh $r17bh $r18bh $r19bh $r20bh $r21bh $r22bh $r23bh $r24bh $r25bh $r26bh $r27bh $r28bh $r29bh $r30bh $r31bh $r16d $r17d $r18d $r19d $r20d $r21d $r22d $r23d $r24d $r25d $r26d $r27d $r28d $r29d $r30d $r31d $r16w $r17w $r18w $r19w $r20w $r21w $r22w $r23w $r24w $r25w $r26w $r27w $r28w $r29w $r30w $r31w $r16wh $r17wh $r18wh $r19wh $r20wh $r21wh $r22wh $r23wh $r24wh $r25wh $r26wh $r27wh $r28wh $r29wh $r30wh $r31wh call void @bar1() call void @bar2() ret void diff --git a/llvm/test/CodeGen/X86/ldexp-avx512.ll b/llvm/test/CodeGen/X86/ldexp-avx512.ll new file mode 100644 index 0000000..ea93a91 --- /dev/null +++ b/llvm/test/CodeGen/X86/ldexp-avx512.ll @@ -0,0 +1,467 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3 +; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512f | FileCheck %s --check-prefixes=CHECK,AVX512 +; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX512VL + +define half @test_half(half %x, i32 %exp) nounwind { +; CHECK-LABEL: test_half: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: pushq %rax +; CHECK-NEXT: vcvtph2ps %xmm0, %xmm0 +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vcvtps2ph $4, %xmm0, %xmm0 +; CHECK-NEXT: popq %rax +; CHECK-NEXT: retq +entry: + %r = tail call fast half @llvm.ldexp.f16.i32(half %x, i32 %exp) + ret half %r +} +declare half @llvm.ldexp.f16.i32(half, i32) memory(none) + +define float @test_float(float %x, i32 %exp) nounwind { +; CHECK-LABEL: test_float: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: jmp ldexpf@PLT # TAILCALL +entry: + %r = tail call fast float @ldexpf(float %x, i32 %exp) + ret float %r +} +declare float @ldexpf(float, i32) memory(none) + +define double @test_double(double %x, i32 %exp) nounwind { +; CHECK-LABEL: test_double: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: jmp ldexp@PLT # TAILCALL +entry: + %r = tail call fast double @ldexp(double %x, i32 %exp) + ret double %r +} +declare double @ldexp(double, i32) memory(none) + +define fp128 @testExpl(fp128 %x, i32 %exp) nounwind { +; CHECK-LABEL: testExpl: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: jmp ldexpl@PLT # TAILCALL +entry: + %r = tail call fast fp128 @ldexpl(fp128 %x, i32 %exp) + ret fp128 %r +} +declare fp128 @ldexpl(fp128, i32) memory(none) + +define <4 x float> @test_ldexp_4xfloat(<4 x float> %x, <4 x i32> %exp) nounwind { +; CHECK-LABEL: test_ldexp_4xfloat: +; CHECK: # %bb.0: +; CHECK-NEXT: subq $56, %rsp +; CHECK-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovd %xmm1, %edi +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill +; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload +; CHECK-NEXT: vextractps $1, %xmm0, %edi +; CHECK-NEXT: vmovshdup {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,1,3,3] +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3] +; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill +; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload +; CHECK-NEXT: vextractps $2, %xmm0, %edi +; CHECK-NEXT: vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,0] +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1],xmm0[0],xmm1[3] +; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill +; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload +; CHECK-NEXT: vextractps $3, %xmm0, %edi +; CHECK-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[3,3,3,3] +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[0] +; CHECK-NEXT: addq $56, %rsp +; CHECK-NEXT: retq + %r = call <4 x float> @llvm.ldexp.v4f32.v4i32(<4 x float> %x, <4 x i32> %exp) + ret <4 x float> %r +} +declare <4 x float> @llvm.ldexp.v4f32.v4i32(<4 x float>, <4 x i32>) + +define <2 x double> @test_ldexp_2xdouble(<2 x double> %x, <2 x i32> %exp) nounwind { +; CHECK-LABEL: test_ldexp_2xdouble: +; CHECK: # %bb.0: +; CHECK-NEXT: subq $56, %rsp +; CHECK-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill +; CHECK-NEXT: vmovd %xmm1, %edi +; CHECK-NEXT: callq ldexp@PLT +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload +; CHECK-NEXT: vextractps $1, %xmm0, %edi +; CHECK-NEXT: vpermilpd $1, (%rsp), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,0] +; CHECK-NEXT: callq ldexp@PLT +; CHECK-NEXT: vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload +; CHECK-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm1[0],xmm0[0] +; CHECK-NEXT: addq $56, %rsp +; CHECK-NEXT: retq + %r = call <2 x double> @llvm.ldexp.v2f64.v2i32(<2 x double> %x, <2 x i32> %exp) + ret <2 x double> %r +} +declare <2 x double> @llvm.ldexp.v2f64.v2i32(<2 x double>, <2 x i32>) + +define <8 x float> @test_ldexp_8xfloat(<8 x float> %x, <8 x i32> %exp) nounwind { +; CHECK-LABEL: test_ldexp_8xfloat: +; CHECK: # %bb.0: +; CHECK-NEXT: subq $120, %rsp +; CHECK-NEXT: vmovdqu %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill +; CHECK-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill +; CHECK-NEXT: vextractf128 $1, %ymm0, %xmm0 +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vextracti128 $1, %ymm1, %xmm1 +; CHECK-NEXT: vmovdqa %xmm1, (%rsp) # 16-byte Spill +; CHECK-NEXT: vmovd %xmm1, %edi +; CHECK-NEXT: vzeroupper +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovaps (%rsp), %xmm0 # 16-byte Reload +; CHECK-NEXT: vextractps $1, %xmm0, %edi +; CHECK-NEXT: vmovshdup {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,1,3,3] +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3] +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovaps (%rsp), %xmm0 # 16-byte Reload +; CHECK-NEXT: vextractps $2, %xmm0, %edi +; CHECK-NEXT: vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,0] +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1],xmm0[0],xmm1[3] +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovaps (%rsp), %xmm0 # 16-byte Reload +; CHECK-NEXT: vextractps $3, %xmm0, %edi +; CHECK-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[3,3,3,3] +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[0] +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload +; CHECK-NEXT: vmovd %xmm0, %edi +; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload +; CHECK-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 +; CHECK-NEXT: vzeroupper +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill +; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload +; CHECK-NEXT: vextractps $1, %xmm0, %edi +; CHECK-NEXT: vmovshdup {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,1,3,3] +; CHECK-NEXT: vzeroupper +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3] +; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill +; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload +; CHECK-NEXT: vextractps $2, %xmm0, %edi +; CHECK-NEXT: vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,0] +; CHECK-NEXT: vzeroupper +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1],xmm0[0],xmm1[3] +; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill +; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload +; CHECK-NEXT: vextractps $3, %xmm0, %edi +; CHECK-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[3,3,3,3] +; CHECK-NEXT: vzeroupper +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[0] +; CHECK-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload +; CHECK-NEXT: addq $120, %rsp +; CHECK-NEXT: retq + %r = call <8 x float> @llvm.ldexp.v8f32.v8i32(<8 x float> %x, <8 x i32> %exp) + ret <8 x float> %r +} +declare <8 x float> @llvm.ldexp.v8f32.v8i32(<8 x float>, <8 x i32>) + +define <4 x double> @test_ldexp_4xdouble(<4 x double> %x, <4 x i32> %exp) nounwind { +; CHECK-LABEL: test_ldexp_4xdouble: +; CHECK: # %bb.0: +; CHECK-NEXT: subq $88, %rsp +; CHECK-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill +; CHECK-NEXT: vextractps $2, %xmm1, %edi +; CHECK-NEXT: vextractf128 $1, %ymm0, %xmm0 +; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill +; CHECK-NEXT: vzeroupper +; CHECK-NEXT: callq ldexp@PLT +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload +; CHECK-NEXT: vextractps $3, %xmm0, %edi +; CHECK-NEXT: vpermilpd $1, (%rsp), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,0] +; CHECK-NEXT: callq ldexp@PLT +; CHECK-NEXT: vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload +; CHECK-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm1[0],xmm0[0] +; CHECK-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload +; CHECK-NEXT: vmovd %xmm0, %edi +; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload +; CHECK-NEXT: # kill: def $xmm0 killed $xmm0 killed $ymm0 +; CHECK-NEXT: vzeroupper +; CHECK-NEXT: callq ldexp@PLT +; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill +; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload +; CHECK-NEXT: vextractps $1, %xmm0, %edi +; CHECK-NEXT: vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,0] +; CHECK-NEXT: callq ldexp@PLT +; CHECK-NEXT: vmovapd (%rsp), %xmm1 # 16-byte Reload +; CHECK-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm1[0],xmm0[0] +; CHECK-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload +; CHECK-NEXT: addq $88, %rsp +; CHECK-NEXT: retq + %r = call <4 x double> @llvm.ldexp.v4f64.v4i32(<4 x double> %x, <4 x i32> %exp) + ret <4 x double> %r +} +declare <4 x double> @llvm.ldexp.v4f64.v4i32(<4 x double>, <4 x i32>) + +define <16 x float> @test_ldexp_16xfloat(<16 x float> %x, <16 x i32> %exp) nounwind { +; CHECK-LABEL: test_ldexp_16xfloat: +; CHECK: # %bb.0: +; CHECK-NEXT: subq $216, %rsp +; CHECK-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill +; CHECK-NEXT: vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill +; CHECK-NEXT: vextractf32x4 $3, %zmm0, %xmm0 +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vextracti32x4 $3, %zmm1, %xmm1 +; CHECK-NEXT: vmovdqa %xmm1, (%rsp) # 16-byte Spill +; CHECK-NEXT: vmovd %xmm1, %edi +; CHECK-NEXT: vzeroupper +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovaps (%rsp), %xmm0 # 16-byte Reload +; CHECK-NEXT: vextractps $1, %xmm0, %edi +; CHECK-NEXT: vmovshdup {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,1,3,3] +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3] +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovaps (%rsp), %xmm0 # 16-byte Reload +; CHECK-NEXT: vextractps $2, %xmm0, %edi +; CHECK-NEXT: vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,0] +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1],xmm0[0],xmm1[3] +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovaps (%rsp), %xmm0 # 16-byte Reload +; CHECK-NEXT: vextractps $3, %xmm0, %edi +; CHECK-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[3,3,3,3] +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[0] +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload +; CHECK-NEXT: vextractf32x4 $2, %zmm0, %xmm1 +; CHECK-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload +; CHECK-NEXT: vextracti32x4 $2, %zmm0, %xmm0 +; CHECK-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill +; CHECK-NEXT: vmovd %xmm0, %edi +; CHECK-NEXT: vmovaps %xmm1, %xmm0 +; CHECK-NEXT: vzeroupper +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovaps (%rsp), %xmm0 # 16-byte Reload +; CHECK-NEXT: vextractps $1, %xmm0, %edi +; CHECK-NEXT: vmovshdup {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,1,3,3] +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3] +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovaps (%rsp), %xmm0 # 16-byte Reload +; CHECK-NEXT: vextractps $2, %xmm0, %edi +; CHECK-NEXT: vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,0] +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1],xmm0[0],xmm1[3] +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovaps (%rsp), %xmm0 # 16-byte Reload +; CHECK-NEXT: vextractps $3, %xmm0, %edi +; CHECK-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[3,3,3,3] +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[0] +; CHECK-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload +; CHECK-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill +; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload +; CHECK-NEXT: vextractf128 $1, %ymm0, %xmm1 +; CHECK-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload +; CHECK-NEXT: vextracti128 $1, %ymm0, %xmm0 +; CHECK-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill +; CHECK-NEXT: vmovd %xmm0, %edi +; CHECK-NEXT: vmovaps %xmm1, %xmm0 +; CHECK-NEXT: vzeroupper +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovaps (%rsp), %xmm0 # 16-byte Reload +; CHECK-NEXT: vextractps $1, %xmm0, %edi +; CHECK-NEXT: vmovshdup {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,1,3,3] +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3] +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovaps (%rsp), %xmm0 # 16-byte Reload +; CHECK-NEXT: vextractps $2, %xmm0, %edi +; CHECK-NEXT: vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,0] +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1],xmm0[0],xmm1[3] +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovaps (%rsp), %xmm0 # 16-byte Reload +; CHECK-NEXT: vextractps $3, %xmm0, %edi +; CHECK-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[3,3,3,3] +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[0] +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovdqu64 {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload +; CHECK-NEXT: vmovd %xmm0, %edi +; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload +; CHECK-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0 +; CHECK-NEXT: vzeroupper +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill +; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload +; CHECK-NEXT: vextractps $1, %xmm0, %edi +; CHECK-NEXT: vmovshdup {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,1,3,3] +; CHECK-NEXT: vzeroupper +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[2,3] +; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill +; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload +; CHECK-NEXT: vextractps $2, %xmm0, %edi +; CHECK-NEXT: vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,0] +; CHECK-NEXT: vzeroupper +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1],xmm0[0],xmm1[3] +; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill +; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload +; CHECK-NEXT: vextractps $3, %xmm0, %edi +; CHECK-NEXT: vpermilps $255, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[3,3,3,3] +; CHECK-NEXT: vzeroupper +; CHECK-NEXT: callq ldexpf@PLT +; CHECK-NEXT: vmovaps (%rsp), %xmm1 # 16-byte Reload +; CHECK-NEXT: vinsertps {{.*#+}} xmm0 = xmm1[0,1,2],xmm0[0] +; CHECK-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload +; CHECK-NEXT: vinsertf64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 32-byte Folded Reload +; CHECK-NEXT: addq $216, %rsp +; CHECK-NEXT: retq + %r = call <16 x float> @llvm.ldexp.v16f32.v16i32(<16 x float> %x, <16 x i32> %exp) + ret <16 x float> %r +} +declare <16 x float> @llvm.ldexp.v16f32.v16i32(<16 x float>, <16 x i32>) + +define <8 x double> @test_ldexp_8xdouble(<8 x double> %x, <8 x i32> %exp) nounwind { +; CHECK-LABEL: test_ldexp_8xdouble: +; CHECK: # %bb.0: +; CHECK-NEXT: subq $184, %rsp +; CHECK-NEXT: vmovups %ymm1, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill +; CHECK-NEXT: vmovups %zmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 64-byte Spill +; CHECK-NEXT: vextractf32x4 $3, %zmm0, %xmm0 +; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill +; CHECK-NEXT: vextractf128 $1, %ymm1, %xmm1 +; CHECK-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vextractps $2, %xmm1, %edi +; CHECK-NEXT: vzeroupper +; CHECK-NEXT: callq ldexp@PLT +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload +; CHECK-NEXT: vextractps $3, %xmm0, %edi +; CHECK-NEXT: vpermilpd $1, (%rsp), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,0] +; CHECK-NEXT: callq ldexp@PLT +; CHECK-NEXT: vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload +; CHECK-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm1[0],xmm0[0] +; CHECK-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload +; CHECK-NEXT: vextractf32x4 $2, %zmm0, %xmm1 +; CHECK-NEXT: vmovaps %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload +; CHECK-NEXT: vmovd %xmm0, %edi +; CHECK-NEXT: vmovaps %xmm1, %xmm0 +; CHECK-NEXT: vzeroupper +; CHECK-NEXT: callq ldexp@PLT +; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill +; CHECK-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload +; CHECK-NEXT: vextractps $1, %xmm0, %edi +; CHECK-NEXT: vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,0] +; CHECK-NEXT: callq ldexp@PLT +; CHECK-NEXT: vmovapd (%rsp), %xmm1 # 16-byte Reload +; CHECK-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm1[0],xmm0[0] +; CHECK-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload +; CHECK-NEXT: vmovupd %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill +; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload +; CHECK-NEXT: vextractps $2, %xmm0, %edi +; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload +; CHECK-NEXT: vextractf128 $1, %ymm0, %xmm0 +; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill +; CHECK-NEXT: vzeroupper +; CHECK-NEXT: callq ldexp@PLT +; CHECK-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload +; CHECK-NEXT: vextractps $3, %xmm0, %edi +; CHECK-NEXT: vpermilpd $1, (%rsp), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,0] +; CHECK-NEXT: vzeroupper +; CHECK-NEXT: callq ldexp@PLT +; CHECK-NEXT: vmovapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload +; CHECK-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm1[0],xmm0[0] +; CHECK-NEXT: vmovapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill +; CHECK-NEXT: vmovdqu {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload +; CHECK-NEXT: vmovd %xmm0, %edi +; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %zmm0 # 64-byte Reload +; CHECK-NEXT: # kill: def $xmm0 killed $xmm0 killed $zmm0 +; CHECK-NEXT: vzeroupper +; CHECK-NEXT: callq ldexp@PLT +; CHECK-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill +; CHECK-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload +; CHECK-NEXT: vextractps $1, %xmm0, %edi +; CHECK-NEXT: vpermilpd $1, {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload +; CHECK-NEXT: # xmm0 = mem[1,0] +; CHECK-NEXT: vzeroupper +; CHECK-NEXT: callq ldexp@PLT +; CHECK-NEXT: vmovapd (%rsp), %xmm1 # 16-byte Reload +; CHECK-NEXT: vunpcklpd {{.*#+}} xmm0 = xmm1[0],xmm0[0] +; CHECK-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload +; CHECK-NEXT: vinsertf64x4 $1, {{[-0-9]+}}(%r{{[sb]}}p), %zmm0, %zmm0 # 32-byte Folded Reload +; CHECK-NEXT: addq $184, %rsp +; CHECK-NEXT: retq + %r = call <8 x double> @llvm.ldexp.v8f64.v8i32(<8 x double> %x, <8 x i32> %exp) + ret <8 x double> %r +} +declare <8 x double> @llvm.ldexp.v8f64.v8i32(<8 x double>, <8 x i32>) + +;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line: +; AVX512: {{.*}} +; AVX512VL: {{.*}} diff --git a/llvm/test/CodeGen/X86/pr165755.ll b/llvm/test/CodeGen/X86/pr165755.ll new file mode 100644 index 0000000..3ab484f --- /dev/null +++ b/llvm/test/CodeGen/X86/pr165755.ll @@ -0,0 +1,26 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc < %s -mtriple=i686-- | FileCheck %s --check-prefixes=X86 +; RUN: llc < %s -mtriple=x86_64-- | FileCheck %s --check-prefixes=X64 + +define i32 @PR165755(ptr %p0) { +; X86-LABEL: PR165755: +; X86: # %bb.0: +; X86-NEXT: movl {{[0-9]+}}(%esp), %ecx +; X86-NEXT: movl (%ecx), %eax +; X86-NEXT: movb $0, (%ecx) +; X86-NEXT: retl +; +; X64-LABEL: PR165755: +; X64: # %bb.0: +; X64-NEXT: movl (%rdi), %eax +; X64-NEXT: movb $0, (%rdi) +; X64-NEXT: retq + %ld64 = load i64, ptr %p0, align 8 + store i8 0, ptr %p0, align 1 + %ld32 = load i32, ptr %p0, align 8 + %mask = and i32 %ld32, 32 + %zext = zext i32 %mask to i64 + %srl = lshr i64 %ld64, %zext + %res = trunc i64 %srl to i32 + ret i32 %res +} diff --git a/llvm/test/CodeGen/X86/vector-reduce-or-cmp.ll b/llvm/test/CodeGen/X86/vector-reduce-or-cmp.ll index ab1feba..044327d 100644 --- a/llvm/test/CodeGen/X86/vector-reduce-or-cmp.ll +++ b/llvm/test/CodeGen/X86/vector-reduce-or-cmp.ll @@ -875,28 +875,12 @@ define i1 @mask_v8i32(<8 x i32> %a0) { ; SSE41-NEXT: sete %al ; SSE41-NEXT: retq ; -; AVX1-LABEL: mask_v8i32: -; AVX1: # %bb.0: -; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0 -; AVX1-NEXT: sete %al -; AVX1-NEXT: vzeroupper -; AVX1-NEXT: retq -; -; AVX2-LABEL: mask_v8i32: -; AVX2: # %bb.0: -; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [9223372039002259456,9223372039002259456,9223372039002259456,9223372039002259456] -; AVX2-NEXT: vptest %ymm1, %ymm0 -; AVX2-NEXT: sete %al -; AVX2-NEXT: vzeroupper -; AVX2-NEXT: retq -; -; AVX512-LABEL: mask_v8i32: -; AVX512: # %bb.0: -; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [9223372039002259456,9223372039002259456,9223372039002259456,9223372039002259456] -; AVX512-NEXT: vptest %ymm1, %ymm0 -; AVX512-NEXT: sete %al -; AVX512-NEXT: vzeroupper -; AVX512-NEXT: retq +; AVX-LABEL: mask_v8i32: +; AVX: # %bb.0: +; AVX-NEXT: vtestps %ymm0, %ymm0 +; AVX-NEXT: sete %al +; AVX-NEXT: vzeroupper +; AVX-NEXT: retq %1 = call i32 @llvm.vector.reduce.or.v8i32(<8 x i32> %a0) %2 = and i32 %1, 2147483648 %3 = icmp eq i32 %2, 0 @@ -965,33 +949,46 @@ define i1 @signtest_v8i32(<8 x i32> %a0) { ; SSE41-NEXT: sete %al ; SSE41-NEXT: retq ; -; AVX1-LABEL: signtest_v8i32: -; AVX1: # %bb.0: -; AVX1-NEXT: vptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0 -; AVX1-NEXT: sete %al -; AVX1-NEXT: vzeroupper -; AVX1-NEXT: retq -; -; AVX2-LABEL: signtest_v8i32: -; AVX2: # %bb.0: -; AVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [9223372039002259456,9223372039002259456,9223372039002259456,9223372039002259456] -; AVX2-NEXT: vptest %ymm1, %ymm0 -; AVX2-NEXT: sete %al -; AVX2-NEXT: vzeroupper -; AVX2-NEXT: retq -; -; AVX512-LABEL: signtest_v8i32: -; AVX512: # %bb.0: -; AVX512-NEXT: vpbroadcastq {{.*#+}} ymm1 = [9223372039002259456,9223372039002259456,9223372039002259456,9223372039002259456] -; AVX512-NEXT: vptest %ymm1, %ymm0 -; AVX512-NEXT: sete %al -; AVX512-NEXT: vzeroupper -; AVX512-NEXT: retq +; AVX-LABEL: signtest_v8i32: +; AVX: # %bb.0: +; AVX-NEXT: vtestps %ymm0, %ymm0 +; AVX-NEXT: sete %al +; AVX-NEXT: vzeroupper +; AVX-NEXT: retq %1 = call i32 @llvm.vector.reduce.or.v8i32(<8 x i32> %a0) %2 = icmp sgt i32 %1, -1 ret i1 %2 } +define i1 @signtest_v4i64(<4 x i64> %a0) { +; SSE2-LABEL: signtest_v4i64: +; SSE2: # %bb.0: +; SSE2-NEXT: por %xmm1, %xmm0 +; SSE2-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3] +; SSE2-NEXT: por %xmm0, %xmm1 +; SSE2-NEXT: movq %xmm1, %rax +; SSE2-NEXT: testq %rax, %rax +; SSE2-NEXT: setns %al +; SSE2-NEXT: retq +; +; SSE41-LABEL: signtest_v4i64: +; SSE41: # %bb.0: +; SSE41-NEXT: por %xmm1, %xmm0 +; SSE41-NEXT: ptest {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0 +; SSE41-NEXT: sete %al +; SSE41-NEXT: retq +; +; AVX-LABEL: signtest_v4i64: +; AVX: # %bb.0: +; AVX-NEXT: vtestpd %ymm0, %ymm0 +; AVX-NEXT: sete %al +; AVX-NEXT: vzeroupper +; AVX-NEXT: retq + %1 = call i64 @llvm.vector.reduce.or.v4i64(<4 x i64> %a0) + %2 = icmp sgt i64 %1, -1 + ret i1 %2 +} + define i1 @trunc_v16i16(<16 x i16> %a0) { ; SSE2-LABEL: trunc_v16i16: ; SSE2: # %bb.0: @@ -1162,11 +1159,11 @@ define i32 @mask_v3i1(<3 x i32> %a, <3 x i32> %b) { ; SSE2-NEXT: movd %xmm0, %eax ; SSE2-NEXT: orl %ecx, %eax ; SSE2-NEXT: testb $1, %al -; SSE2-NEXT: je .LBB29_2 +; SSE2-NEXT: je .LBB30_2 ; SSE2-NEXT: # %bb.1: ; SSE2-NEXT: xorl %eax, %eax ; SSE2-NEXT: retq -; SSE2-NEXT: .LBB29_2: +; SSE2-NEXT: .LBB30_2: ; SSE2-NEXT: movl $1, %eax ; SSE2-NEXT: retq ; @@ -1181,11 +1178,11 @@ define i32 @mask_v3i1(<3 x i32> %a, <3 x i32> %b) { ; SSE41-NEXT: pextrd $2, %xmm1, %eax ; SSE41-NEXT: orl %ecx, %eax ; SSE41-NEXT: testb $1, %al -; SSE41-NEXT: je .LBB29_2 +; SSE41-NEXT: je .LBB30_2 ; SSE41-NEXT: # %bb.1: ; SSE41-NEXT: xorl %eax, %eax ; SSE41-NEXT: retq -; SSE41-NEXT: .LBB29_2: +; SSE41-NEXT: .LBB30_2: ; SSE41-NEXT: movl $1, %eax ; SSE41-NEXT: retq ; @@ -1200,11 +1197,11 @@ define i32 @mask_v3i1(<3 x i32> %a, <3 x i32> %b) { ; AVX1OR2-NEXT: vpextrd $2, %xmm0, %eax ; AVX1OR2-NEXT: orl %ecx, %eax ; AVX1OR2-NEXT: testb $1, %al -; AVX1OR2-NEXT: je .LBB29_2 +; AVX1OR2-NEXT: je .LBB30_2 ; AVX1OR2-NEXT: # %bb.1: ; AVX1OR2-NEXT: xorl %eax, %eax ; AVX1OR2-NEXT: retq -; AVX1OR2-NEXT: .LBB29_2: +; AVX1OR2-NEXT: .LBB30_2: ; AVX1OR2-NEXT: movl $1, %eax ; AVX1OR2-NEXT: retq ; @@ -1219,12 +1216,12 @@ define i32 @mask_v3i1(<3 x i32> %a, <3 x i32> %b) { ; AVX512F-NEXT: korw %k0, %k1, %k0 ; AVX512F-NEXT: kmovw %k0, %eax ; AVX512F-NEXT: testb $1, %al -; AVX512F-NEXT: je .LBB29_2 +; AVX512F-NEXT: je .LBB30_2 ; AVX512F-NEXT: # %bb.1: ; AVX512F-NEXT: xorl %eax, %eax ; AVX512F-NEXT: vzeroupper ; AVX512F-NEXT: retq -; AVX512F-NEXT: .LBB29_2: +; AVX512F-NEXT: .LBB30_2: ; AVX512F-NEXT: movl $1, %eax ; AVX512F-NEXT: vzeroupper ; AVX512F-NEXT: retq @@ -1240,12 +1237,12 @@ define i32 @mask_v3i1(<3 x i32> %a, <3 x i32> %b) { ; AVX512BW-NEXT: korw %k0, %k1, %k0 ; AVX512BW-NEXT: kmovd %k0, %eax ; AVX512BW-NEXT: testb $1, %al -; AVX512BW-NEXT: je .LBB29_2 +; AVX512BW-NEXT: je .LBB30_2 ; AVX512BW-NEXT: # %bb.1: ; AVX512BW-NEXT: xorl %eax, %eax ; AVX512BW-NEXT: vzeroupper ; AVX512BW-NEXT: retq -; AVX512BW-NEXT: .LBB29_2: +; AVX512BW-NEXT: .LBB30_2: ; AVX512BW-NEXT: movl $1, %eax ; AVX512BW-NEXT: vzeroupper ; AVX512BW-NEXT: retq @@ -1259,11 +1256,11 @@ define i32 @mask_v3i1(<3 x i32> %a, <3 x i32> %b) { ; AVX512BWVL-NEXT: korw %k0, %k1, %k0 ; AVX512BWVL-NEXT: kmovd %k0, %eax ; AVX512BWVL-NEXT: testb $1, %al -; AVX512BWVL-NEXT: je .LBB29_2 +; AVX512BWVL-NEXT: je .LBB30_2 ; AVX512BWVL-NEXT: # %bb.1: ; AVX512BWVL-NEXT: xorl %eax, %eax ; AVX512BWVL-NEXT: retq -; AVX512BWVL-NEXT: .LBB29_2: +; AVX512BWVL-NEXT: .LBB30_2: ; AVX512BWVL-NEXT: movl $1, %eax ; AVX512BWVL-NEXT: retq %1 = icmp ne <3 x i32> %a, %b |