diff options
Diffstat (limited to 'llvm/test')
87 files changed, 10492 insertions, 5729 deletions
diff --git a/llvm/test/Analysis/CostModel/RISCV/masked_ldst.ll b/llvm/test/Analysis/CostModel/RISCV/masked_ldst.ll index 892277a..68c89c3 100644 --- a/llvm/test/Analysis/CostModel/RISCV/masked_ldst.ll +++ b/llvm/test/Analysis/CostModel/RISCV/masked_ldst.ll @@ -13,14 +13,14 @@ define void @fixed() { ; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i32 = call <2 x i32> @llvm.masked.load.v2i32.p0(ptr undef, i32 8, <2 x i1> undef, <2 x i32> undef) ; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4i32 = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr undef, i32 8, <4 x i1> undef, <4 x i32> undef) ; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2i64 = call <2 x i64> @llvm.masked.load.v2i64.p0(ptr undef, i32 8, <2 x i1> undef, <2 x i64> undef) -; CHECK-NEXT: Cost Model: Found an estimated cost of 9 for instruction: %v2f16 = call <2 x half> @llvm.masked.load.v2f16.p0(ptr undef, i32 8, <2 x i1> undef, <2 x half> undef) -; CHECK-NEXT: Cost Model: Found an estimated cost of 19 for instruction: %v4f16 = call <4 x half> @llvm.masked.load.v4f16.p0(ptr undef, i32 8, <4 x i1> undef, <4 x half> undef) -; CHECK-NEXT: Cost Model: Found an estimated cost of 39 for instruction: %v8f16 = call <8 x half> @llvm.masked.load.v8f16.p0(ptr undef, i32 8, <8 x i1> undef, <8 x half> undef) +; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v2f16 = call <2 x half> @llvm.masked.load.v2f16.p0(ptr undef, i32 8, <2 x i1> undef, <2 x half> undef) +; CHECK-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %v4f16 = call <4 x half> @llvm.masked.load.v4f16.p0(ptr undef, i32 8, <4 x i1> undef, <4 x half> undef) +; CHECK-NEXT: Cost Model: Found an estimated cost of 8 for instruction: %v8f16 = call <8 x half> @llvm.masked.load.v8f16.p0(ptr undef, i32 8, <8 x i1> undef, <8 x half> undef) ; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2f32 = call <2 x float> @llvm.masked.load.v2f32.p0(ptr undef, i32 8, <2 x i1> undef, <2 x float> undef) ; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v4f32 = call <4 x float> @llvm.masked.load.v4f32.p0(ptr undef, i32 8, <4 x i1> undef, <4 x float> undef) ; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %v2f64 = call <2 x double> @llvm.masked.load.v2f64.p0(ptr undef, i32 8, <2 x i1> undef, <2 x double> undef) ; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %v4i64 = call <4 x i64> @llvm.masked.load.v4i64.p0(ptr undef, i32 8, <4 x i1> undef, <4 x i64> undef) -; CHECK-NEXT: Cost Model: Found an estimated cost of 223 for instruction: %v32f16 = call <32 x half> @llvm.masked.load.v32f16.p0(ptr undef, i32 8, <32 x i1> undef, <32 x half> undef) +; CHECK-NEXT: Cost Model: Found an estimated cost of 32 for instruction: %v32f16 = call <32 x half> @llvm.masked.load.v32f16.p0(ptr undef, i32 8, <32 x i1> undef, <32 x half> undef) ; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void ; entry: diff --git a/llvm/test/CodeGen/AArch64/andcompare.ll b/llvm/test/CodeGen/AArch64/andcompare.ll index cbacd17..0e15b94 100644 --- a/llvm/test/CodeGen/AArch64/andcompare.ll +++ b/llvm/test/CodeGen/AArch64/andcompare.ll @@ -1,23 +1,23 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc -mtriple=aarch64-none-elf -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,SDISEL -; RUN: llc -mtriple=aarch64-none-elf -global-isel -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,GISEL +; RUN: llc -mtriple=aarch64-none-elf -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SD +; RUN: llc -mtriple=aarch64-none-elf -global-isel -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-GI define i32 @and_eq_eq(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_eq_eq: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, eq -; SDISEL-NEXT: cset w0, eq -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_eq_eq: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, eq -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, eq -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_eq_eq: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, eq +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_eq_eq: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, eq +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, eq +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp eq i32 %s0, %s1 %c1 = icmp eq i32 %s2, %s3 @@ -27,21 +27,21 @@ entry: } define i32 @and_eq_ne(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_eq_ne: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #4, eq -; SDISEL-NEXT: cset w0, ne -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_eq_ne: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, eq -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ne -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_eq_ne: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #4, eq +; CHECK-SD-NEXT: cset w0, ne +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_eq_ne: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, eq +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ne +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp eq i32 %s0, %s1 %c1 = icmp ne i32 %s2, %s3 @@ -51,21 +51,21 @@ entry: } define i32 @and_eq_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_eq_ult: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, eq -; SDISEL-NEXT: cset w0, lo -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_eq_ult: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, eq -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lo -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_eq_ult: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, eq +; CHECK-SD-NEXT: cset w0, lo +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_eq_ult: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, eq +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lo +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp eq i32 %s0, %s1 %c1 = icmp ult i32 %s2, %s3 @@ -75,21 +75,21 @@ entry: } define i32 @and_eq_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_eq_ule: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, eq -; SDISEL-NEXT: cset w0, ls -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_eq_ule: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, eq -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ls -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_eq_ule: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, eq +; CHECK-SD-NEXT: cset w0, ls +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_eq_ule: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, eq +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ls +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp eq i32 %s0, %s1 %c1 = icmp ule i32 %s2, %s3 @@ -99,21 +99,21 @@ entry: } define i32 @and_eq_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_eq_ugt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, eq -; SDISEL-NEXT: cset w0, hi -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_eq_ugt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, eq -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hi -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_eq_ugt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, eq +; CHECK-SD-NEXT: cset w0, hi +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_eq_ugt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, eq +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hi +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp eq i32 %s0, %s1 %c1 = icmp ugt i32 %s2, %s3 @@ -123,21 +123,21 @@ entry: } define i32 @and_eq_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_eq_uge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, eq -; SDISEL-NEXT: cset w0, hs -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_eq_uge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, eq -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hs -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_eq_uge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, eq +; CHECK-SD-NEXT: cset w0, hs +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_eq_uge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, eq +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hs +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp eq i32 %s0, %s1 %c1 = icmp uge i32 %s2, %s3 @@ -147,21 +147,21 @@ entry: } define i32 @and_eq_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_eq_slt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, eq -; SDISEL-NEXT: cset w0, lt -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_eq_slt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, eq -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lt -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_eq_slt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, eq +; CHECK-SD-NEXT: cset w0, lt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_eq_slt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, eq +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lt +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp eq i32 %s0, %s1 %c1 = icmp slt i32 %s2, %s3 @@ -171,21 +171,21 @@ entry: } define i32 @and_eq_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_eq_sle: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, eq -; SDISEL-NEXT: cset w0, le -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_eq_sle: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, eq -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, le -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_eq_sle: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, eq +; CHECK-SD-NEXT: cset w0, le +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_eq_sle: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, eq +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, le +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp eq i32 %s0, %s1 %c1 = icmp sle i32 %s2, %s3 @@ -195,21 +195,21 @@ entry: } define i32 @and_eq_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_eq_sgt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #4, eq -; SDISEL-NEXT: cset w0, gt -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_eq_sgt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, eq -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, gt -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_eq_sgt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #4, eq +; CHECK-SD-NEXT: cset w0, gt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_eq_sgt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, eq +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, gt +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp eq i32 %s0, %s1 %c1 = icmp sgt i32 %s2, %s3 @@ -219,21 +219,21 @@ entry: } define i32 @and_eq_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_eq_sge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #8, eq -; SDISEL-NEXT: cset w0, ge -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_eq_sge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, eq -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ge -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_eq_sge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #8, eq +; CHECK-SD-NEXT: cset w0, ge +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_eq_sge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, eq +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ge +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp eq i32 %s0, %s1 %c1 = icmp sge i32 %s2, %s3 @@ -243,21 +243,21 @@ entry: } define i32 @and_ne_eq(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ne_eq: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, ne -; SDISEL-NEXT: cset w0, eq -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ne_eq: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ne -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, eq -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ne_eq: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, ne +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ne_eq: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ne +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, eq +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ne i32 %s0, %s1 %c1 = icmp eq i32 %s2, %s3 @@ -267,21 +267,21 @@ entry: } define i32 @and_ne_ne(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ne_ne: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #4, ne -; SDISEL-NEXT: cset w0, ne -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ne_ne: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ne -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ne -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ne_ne: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #4, ne +; CHECK-SD-NEXT: cset w0, ne +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ne_ne: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ne +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ne +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ne i32 %s0, %s1 %c1 = icmp ne i32 %s2, %s3 @@ -291,21 +291,21 @@ entry: } define i32 @and_ne_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ne_ult: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, ne -; SDISEL-NEXT: cset w0, lo -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ne_ult: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ne -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lo -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ne_ult: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, ne +; CHECK-SD-NEXT: cset w0, lo +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ne_ult: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ne +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lo +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ne i32 %s0, %s1 %c1 = icmp ult i32 %s2, %s3 @@ -315,21 +315,21 @@ entry: } define i32 @and_ne_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ne_ule: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, ne -; SDISEL-NEXT: cset w0, ls -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ne_ule: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ne -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ls -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ne_ule: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, ne +; CHECK-SD-NEXT: cset w0, ls +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ne_ule: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ne +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ls +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ne i32 %s0, %s1 %c1 = icmp ule i32 %s2, %s3 @@ -339,21 +339,21 @@ entry: } define i32 @and_ne_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ne_ugt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, ne -; SDISEL-NEXT: cset w0, hi -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ne_ugt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ne -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hi -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ne_ugt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, ne +; CHECK-SD-NEXT: cset w0, hi +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ne_ugt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ne +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hi +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ne i32 %s0, %s1 %c1 = icmp ugt i32 %s2, %s3 @@ -363,21 +363,21 @@ entry: } define i32 @and_ne_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ne_uge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, ne -; SDISEL-NEXT: cset w0, hs -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ne_uge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ne -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hs -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ne_uge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, ne +; CHECK-SD-NEXT: cset w0, hs +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ne_uge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ne +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hs +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ne i32 %s0, %s1 %c1 = icmp uge i32 %s2, %s3 @@ -387,21 +387,21 @@ entry: } define i32 @and_ne_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ne_slt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, ne -; SDISEL-NEXT: cset w0, lt -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ne_slt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ne -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lt -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ne_slt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, ne +; CHECK-SD-NEXT: cset w0, lt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ne_slt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ne +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lt +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ne i32 %s0, %s1 %c1 = icmp slt i32 %s2, %s3 @@ -411,21 +411,21 @@ entry: } define i32 @and_ne_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ne_sle: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, ne -; SDISEL-NEXT: cset w0, le -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ne_sle: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ne -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, le -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ne_sle: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, ne +; CHECK-SD-NEXT: cset w0, le +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ne_sle: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ne +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, le +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ne i32 %s0, %s1 %c1 = icmp sle i32 %s2, %s3 @@ -435,21 +435,21 @@ entry: } define i32 @and_ne_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ne_sgt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #4, ne -; SDISEL-NEXT: cset w0, gt -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ne_sgt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ne -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, gt -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ne_sgt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #4, ne +; CHECK-SD-NEXT: cset w0, gt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ne_sgt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ne +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, gt +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ne i32 %s0, %s1 %c1 = icmp sgt i32 %s2, %s3 @@ -459,21 +459,21 @@ entry: } define i32 @and_ne_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ne_sge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #8, ne -; SDISEL-NEXT: cset w0, ge -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ne_sge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ne -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ge -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ne_sge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #8, ne +; CHECK-SD-NEXT: cset w0, ge +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ne_sge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ne +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ge +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ne i32 %s0, %s1 %c1 = icmp sge i32 %s2, %s3 @@ -483,21 +483,21 @@ entry: } define i32 @and_ult_eq(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ult_eq: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, lo -; SDISEL-NEXT: cset w0, eq -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ult_eq: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lo -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, eq -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ult_eq: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, lo +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ult_eq: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lo +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, eq +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ult i32 %s0, %s1 %c1 = icmp eq i32 %s2, %s3 @@ -507,21 +507,21 @@ entry: } define i32 @and_ult_ne(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ult_ne: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #4, lo -; SDISEL-NEXT: cset w0, ne -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ult_ne: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lo -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ne -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ult_ne: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #4, lo +; CHECK-SD-NEXT: cset w0, ne +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ult_ne: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lo +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ne +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ult i32 %s0, %s1 %c1 = icmp ne i32 %s2, %s3 @@ -531,21 +531,21 @@ entry: } define i32 @and_ult_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ult_ult: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, lo -; SDISEL-NEXT: cset w0, lo -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ult_ult: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lo -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lo -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ult_ult: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, lo +; CHECK-SD-NEXT: cset w0, lo +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ult_ult: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lo +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lo +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ult i32 %s0, %s1 %c1 = icmp ult i32 %s2, %s3 @@ -555,21 +555,21 @@ entry: } define i32 @and_ult_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ult_ule: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, lo -; SDISEL-NEXT: cset w0, ls -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ult_ule: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lo -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ls -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ult_ule: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, lo +; CHECK-SD-NEXT: cset w0, ls +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ult_ule: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lo +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ls +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ult i32 %s0, %s1 %c1 = icmp ule i32 %s2, %s3 @@ -579,21 +579,21 @@ entry: } define i32 @and_ult_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ult_ugt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, lo -; SDISEL-NEXT: cset w0, hi -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ult_ugt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lo -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hi -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ult_ugt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, lo +; CHECK-SD-NEXT: cset w0, hi +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ult_ugt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lo +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hi +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ult i32 %s0, %s1 %c1 = icmp ugt i32 %s2, %s3 @@ -603,21 +603,21 @@ entry: } define i32 @and_ult_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ult_uge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, lo -; SDISEL-NEXT: cset w0, hs -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ult_uge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lo -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hs -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ult_uge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, lo +; CHECK-SD-NEXT: cset w0, hs +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ult_uge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lo +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hs +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ult i32 %s0, %s1 %c1 = icmp uge i32 %s2, %s3 @@ -627,21 +627,21 @@ entry: } define i32 @and_ult_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ult_slt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, lo -; SDISEL-NEXT: cset w0, lt -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ult_slt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lo -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lt -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ult_slt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, lo +; CHECK-SD-NEXT: cset w0, lt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ult_slt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lo +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lt +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ult i32 %s0, %s1 %c1 = icmp slt i32 %s2, %s3 @@ -651,21 +651,21 @@ entry: } define i32 @and_ult_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ult_sle: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, lo -; SDISEL-NEXT: cset w0, le -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ult_sle: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lo -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, le -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ult_sle: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, lo +; CHECK-SD-NEXT: cset w0, le +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ult_sle: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lo +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, le +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ult i32 %s0, %s1 %c1 = icmp sle i32 %s2, %s3 @@ -675,21 +675,21 @@ entry: } define i32 @and_ult_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ult_sgt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #4, lo -; SDISEL-NEXT: cset w0, gt -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ult_sgt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lo -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, gt -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ult_sgt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #4, lo +; CHECK-SD-NEXT: cset w0, gt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ult_sgt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lo +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, gt +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ult i32 %s0, %s1 %c1 = icmp sgt i32 %s2, %s3 @@ -699,21 +699,21 @@ entry: } define i32 @and_ult_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ult_sge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #8, lo -; SDISEL-NEXT: cset w0, ge -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ult_sge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lo -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ge -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ult_sge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #8, lo +; CHECK-SD-NEXT: cset w0, ge +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ult_sge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lo +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ge +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ult i32 %s0, %s1 %c1 = icmp sge i32 %s2, %s3 @@ -723,21 +723,21 @@ entry: } define i32 @and_ule_eq(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ule_eq: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, ls -; SDISEL-NEXT: cset w0, eq -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ule_eq: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ls -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, eq -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ule_eq: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, ls +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ule_eq: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ls +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, eq +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ule i32 %s0, %s1 %c1 = icmp eq i32 %s2, %s3 @@ -747,21 +747,21 @@ entry: } define i32 @and_ule_ne(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ule_ne: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #4, ls -; SDISEL-NEXT: cset w0, ne -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ule_ne: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ls -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ne -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ule_ne: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #4, ls +; CHECK-SD-NEXT: cset w0, ne +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ule_ne: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ls +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ne +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ule i32 %s0, %s1 %c1 = icmp ne i32 %s2, %s3 @@ -771,21 +771,21 @@ entry: } define i32 @and_ule_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ule_ult: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, ls -; SDISEL-NEXT: cset w0, lo -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ule_ult: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ls -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lo -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ule_ult: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, ls +; CHECK-SD-NEXT: cset w0, lo +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ule_ult: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ls +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lo +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ule i32 %s0, %s1 %c1 = icmp ult i32 %s2, %s3 @@ -795,21 +795,21 @@ entry: } define i32 @and_ule_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ule_ule: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, ls -; SDISEL-NEXT: cset w0, ls -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ule_ule: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ls -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ls -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ule_ule: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, ls +; CHECK-SD-NEXT: cset w0, ls +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ule_ule: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ls +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ls +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ule i32 %s0, %s1 %c1 = icmp ule i32 %s2, %s3 @@ -819,21 +819,21 @@ entry: } define i32 @and_ule_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ule_ugt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, ls -; SDISEL-NEXT: cset w0, hi -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ule_ugt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ls -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hi -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ule_ugt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, ls +; CHECK-SD-NEXT: cset w0, hi +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ule_ugt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ls +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hi +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ule i32 %s0, %s1 %c1 = icmp ugt i32 %s2, %s3 @@ -843,21 +843,21 @@ entry: } define i32 @and_ule_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ule_uge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, ls -; SDISEL-NEXT: cset w0, hs -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ule_uge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ls -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hs -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ule_uge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, ls +; CHECK-SD-NEXT: cset w0, hs +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ule_uge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ls +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hs +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ule i32 %s0, %s1 %c1 = icmp uge i32 %s2, %s3 @@ -867,21 +867,21 @@ entry: } define i32 @and_ule_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ule_slt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, ls -; SDISEL-NEXT: cset w0, lt -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ule_slt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ls -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lt -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ule_slt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, ls +; CHECK-SD-NEXT: cset w0, lt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ule_slt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ls +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lt +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ule i32 %s0, %s1 %c1 = icmp slt i32 %s2, %s3 @@ -891,21 +891,21 @@ entry: } define i32 @and_ule_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ule_sle: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, ls -; SDISEL-NEXT: cset w0, le -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ule_sle: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ls -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, le -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ule_sle: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, ls +; CHECK-SD-NEXT: cset w0, le +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ule_sle: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ls +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, le +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ule i32 %s0, %s1 %c1 = icmp sle i32 %s2, %s3 @@ -915,21 +915,21 @@ entry: } define i32 @and_ule_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ule_sgt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #4, ls -; SDISEL-NEXT: cset w0, gt -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ule_sgt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ls -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, gt -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ule_sgt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #4, ls +; CHECK-SD-NEXT: cset w0, gt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ule_sgt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ls +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, gt +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ule i32 %s0, %s1 %c1 = icmp sgt i32 %s2, %s3 @@ -939,21 +939,21 @@ entry: } define i32 @and_ule_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ule_sge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #8, ls -; SDISEL-NEXT: cset w0, ge -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ule_sge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ls -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ge -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ule_sge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #8, ls +; CHECK-SD-NEXT: cset w0, ge +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ule_sge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ls +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ge +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ule i32 %s0, %s1 %c1 = icmp sge i32 %s2, %s3 @@ -963,21 +963,21 @@ entry: } define i32 @and_ugt_eq(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ugt_eq: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, hi -; SDISEL-NEXT: cset w0, eq -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ugt_eq: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hi -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, eq -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ugt_eq: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, hi +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ugt_eq: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hi +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, eq +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ugt i32 %s0, %s1 %c1 = icmp eq i32 %s2, %s3 @@ -987,21 +987,21 @@ entry: } define i32 @and_ugt_ne(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ugt_ne: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #4, hi -; SDISEL-NEXT: cset w0, ne -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ugt_ne: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hi -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ne -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ugt_ne: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #4, hi +; CHECK-SD-NEXT: cset w0, ne +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ugt_ne: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hi +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ne +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ugt i32 %s0, %s1 %c1 = icmp ne i32 %s2, %s3 @@ -1011,21 +1011,21 @@ entry: } define i32 @and_ugt_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ugt_ult: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, hi -; SDISEL-NEXT: cset w0, lo -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ugt_ult: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hi -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lo -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ugt_ult: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, hi +; CHECK-SD-NEXT: cset w0, lo +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ugt_ult: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hi +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lo +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ugt i32 %s0, %s1 %c1 = icmp ult i32 %s2, %s3 @@ -1035,21 +1035,21 @@ entry: } define i32 @and_ugt_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ugt_ule: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, hi -; SDISEL-NEXT: cset w0, ls -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ugt_ule: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hi -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ls -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ugt_ule: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, hi +; CHECK-SD-NEXT: cset w0, ls +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ugt_ule: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hi +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ls +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ugt i32 %s0, %s1 %c1 = icmp ule i32 %s2, %s3 @@ -1059,21 +1059,21 @@ entry: } define i32 @and_ugt_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ugt_ugt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, hi -; SDISEL-NEXT: cset w0, hi -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ugt_ugt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hi -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hi -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ugt_ugt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, hi +; CHECK-SD-NEXT: cset w0, hi +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ugt_ugt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hi +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hi +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ugt i32 %s0, %s1 %c1 = icmp ugt i32 %s2, %s3 @@ -1083,21 +1083,21 @@ entry: } define i32 @and_ugt_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ugt_uge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, hi -; SDISEL-NEXT: cset w0, hs -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ugt_uge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hi -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hs -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ugt_uge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, hi +; CHECK-SD-NEXT: cset w0, hs +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ugt_uge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hi +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hs +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ugt i32 %s0, %s1 %c1 = icmp uge i32 %s2, %s3 @@ -1107,21 +1107,21 @@ entry: } define i32 @and_ugt_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ugt_slt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, hi -; SDISEL-NEXT: cset w0, lt -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ugt_slt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hi -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lt -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ugt_slt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, hi +; CHECK-SD-NEXT: cset w0, lt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ugt_slt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hi +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lt +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ugt i32 %s0, %s1 %c1 = icmp slt i32 %s2, %s3 @@ -1131,21 +1131,21 @@ entry: } define i32 @and_ugt_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ugt_sle: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, hi -; SDISEL-NEXT: cset w0, le -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ugt_sle: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hi -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, le -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ugt_sle: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, hi +; CHECK-SD-NEXT: cset w0, le +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ugt_sle: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hi +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, le +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ugt i32 %s0, %s1 %c1 = icmp sle i32 %s2, %s3 @@ -1155,21 +1155,21 @@ entry: } define i32 @and_ugt_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ugt_sgt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #4, hi -; SDISEL-NEXT: cset w0, gt -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ugt_sgt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hi -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, gt -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ugt_sgt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #4, hi +; CHECK-SD-NEXT: cset w0, gt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ugt_sgt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hi +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, gt +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ugt i32 %s0, %s1 %c1 = icmp sgt i32 %s2, %s3 @@ -1179,21 +1179,21 @@ entry: } define i32 @and_ugt_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_ugt_sge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #8, hi -; SDISEL-NEXT: cset w0, ge -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_ugt_sge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hi -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ge -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_ugt_sge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #8, hi +; CHECK-SD-NEXT: cset w0, ge +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_ugt_sge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hi +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ge +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp ugt i32 %s0, %s1 %c1 = icmp sge i32 %s2, %s3 @@ -1203,21 +1203,21 @@ entry: } define i32 @and_uge_eq(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_uge_eq: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, hs -; SDISEL-NEXT: cset w0, eq -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_uge_eq: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hs -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, eq -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_uge_eq: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, hs +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_uge_eq: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hs +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, eq +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp uge i32 %s0, %s1 %c1 = icmp eq i32 %s2, %s3 @@ -1227,21 +1227,21 @@ entry: } define i32 @and_uge_ne(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_uge_ne: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #4, hs -; SDISEL-NEXT: cset w0, ne -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_uge_ne: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hs -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ne -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_uge_ne: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #4, hs +; CHECK-SD-NEXT: cset w0, ne +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_uge_ne: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hs +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ne +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp uge i32 %s0, %s1 %c1 = icmp ne i32 %s2, %s3 @@ -1251,21 +1251,21 @@ entry: } define i32 @and_uge_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_uge_ult: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, hs -; SDISEL-NEXT: cset w0, lo -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_uge_ult: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hs -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lo -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_uge_ult: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, hs +; CHECK-SD-NEXT: cset w0, lo +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_uge_ult: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hs +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lo +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp uge i32 %s0, %s1 %c1 = icmp ult i32 %s2, %s3 @@ -1275,21 +1275,21 @@ entry: } define i32 @and_uge_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_uge_ule: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, hs -; SDISEL-NEXT: cset w0, ls -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_uge_ule: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hs -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ls -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_uge_ule: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, hs +; CHECK-SD-NEXT: cset w0, ls +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_uge_ule: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hs +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ls +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp uge i32 %s0, %s1 %c1 = icmp ule i32 %s2, %s3 @@ -1299,21 +1299,21 @@ entry: } define i32 @and_uge_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_uge_ugt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, hs -; SDISEL-NEXT: cset w0, hi -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_uge_ugt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hs -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hi -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_uge_ugt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, hs +; CHECK-SD-NEXT: cset w0, hi +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_uge_ugt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hs +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hi +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp uge i32 %s0, %s1 %c1 = icmp ugt i32 %s2, %s3 @@ -1323,21 +1323,21 @@ entry: } define i32 @and_uge_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_uge_uge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, hs -; SDISEL-NEXT: cset w0, hs -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_uge_uge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hs -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hs -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_uge_uge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, hs +; CHECK-SD-NEXT: cset w0, hs +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_uge_uge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hs +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hs +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp uge i32 %s0, %s1 %c1 = icmp uge i32 %s2, %s3 @@ -1347,21 +1347,21 @@ entry: } define i32 @and_uge_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_uge_slt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, hs -; SDISEL-NEXT: cset w0, lt -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_uge_slt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hs -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lt -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_uge_slt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, hs +; CHECK-SD-NEXT: cset w0, lt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_uge_slt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hs +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lt +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp uge i32 %s0, %s1 %c1 = icmp slt i32 %s2, %s3 @@ -1371,21 +1371,21 @@ entry: } define i32 @and_uge_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_uge_sle: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, hs -; SDISEL-NEXT: cset w0, le -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_uge_sle: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hs -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, le -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_uge_sle: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, hs +; CHECK-SD-NEXT: cset w0, le +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_uge_sle: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hs +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, le +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp uge i32 %s0, %s1 %c1 = icmp sle i32 %s2, %s3 @@ -1395,21 +1395,21 @@ entry: } define i32 @and_uge_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_uge_sgt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #4, hs -; SDISEL-NEXT: cset w0, gt -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_uge_sgt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hs -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, gt -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_uge_sgt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #4, hs +; CHECK-SD-NEXT: cset w0, gt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_uge_sgt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hs +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, gt +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp uge i32 %s0, %s1 %c1 = icmp sgt i32 %s2, %s3 @@ -1419,21 +1419,21 @@ entry: } define i32 @and_uge_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_uge_sge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #8, hs -; SDISEL-NEXT: cset w0, ge -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_uge_sge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hs -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ge -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_uge_sge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #8, hs +; CHECK-SD-NEXT: cset w0, ge +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_uge_sge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hs +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ge +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp uge i32 %s0, %s1 %c1 = icmp sge i32 %s2, %s3 @@ -1443,21 +1443,21 @@ entry: } define i32 @and_slt_eq(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_slt_eq: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, lt -; SDISEL-NEXT: cset w0, eq -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_slt_eq: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, eq -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_slt_eq: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, lt +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_slt_eq: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, eq +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp slt i32 %s0, %s1 %c1 = icmp eq i32 %s2, %s3 @@ -1467,21 +1467,21 @@ entry: } define i32 @and_slt_ne(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_slt_ne: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #4, lt -; SDISEL-NEXT: cset w0, ne -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_slt_ne: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ne -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_slt_ne: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #4, lt +; CHECK-SD-NEXT: cset w0, ne +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_slt_ne: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ne +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp slt i32 %s0, %s1 %c1 = icmp ne i32 %s2, %s3 @@ -1491,21 +1491,21 @@ entry: } define i32 @and_slt_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_slt_ult: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, lt -; SDISEL-NEXT: cset w0, lo -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_slt_ult: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lo -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_slt_ult: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, lt +; CHECK-SD-NEXT: cset w0, lo +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_slt_ult: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lo +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp slt i32 %s0, %s1 %c1 = icmp ult i32 %s2, %s3 @@ -1515,21 +1515,21 @@ entry: } define i32 @and_slt_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_slt_ule: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, lt -; SDISEL-NEXT: cset w0, ls -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_slt_ule: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ls -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_slt_ule: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, lt +; CHECK-SD-NEXT: cset w0, ls +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_slt_ule: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ls +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp slt i32 %s0, %s1 %c1 = icmp ule i32 %s2, %s3 @@ -1539,21 +1539,21 @@ entry: } define i32 @and_slt_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_slt_ugt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, lt -; SDISEL-NEXT: cset w0, hi -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_slt_ugt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hi -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_slt_ugt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, lt +; CHECK-SD-NEXT: cset w0, hi +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_slt_ugt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hi +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp slt i32 %s0, %s1 %c1 = icmp ugt i32 %s2, %s3 @@ -1563,21 +1563,21 @@ entry: } define i32 @and_slt_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_slt_uge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, lt -; SDISEL-NEXT: cset w0, hs -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_slt_uge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hs -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_slt_uge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, lt +; CHECK-SD-NEXT: cset w0, hs +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_slt_uge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hs +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp slt i32 %s0, %s1 %c1 = icmp uge i32 %s2, %s3 @@ -1587,21 +1587,21 @@ entry: } define i32 @and_slt_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_slt_slt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, lt -; SDISEL-NEXT: cset w0, lt -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_slt_slt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lt -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_slt_slt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, lt +; CHECK-SD-NEXT: cset w0, lt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_slt_slt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lt +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp slt i32 %s0, %s1 %c1 = icmp slt i32 %s2, %s3 @@ -1611,21 +1611,21 @@ entry: } define i32 @and_slt_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_slt_sle: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, lt -; SDISEL-NEXT: cset w0, le -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_slt_sle: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, le -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_slt_sle: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, lt +; CHECK-SD-NEXT: cset w0, le +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_slt_sle: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, le +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp slt i32 %s0, %s1 %c1 = icmp sle i32 %s2, %s3 @@ -1635,21 +1635,21 @@ entry: } define i32 @and_slt_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_slt_sgt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #4, lt -; SDISEL-NEXT: cset w0, gt -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_slt_sgt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, gt -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_slt_sgt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #4, lt +; CHECK-SD-NEXT: cset w0, gt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_slt_sgt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, gt +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp slt i32 %s0, %s1 %c1 = icmp sgt i32 %s2, %s3 @@ -1659,21 +1659,21 @@ entry: } define i32 @and_slt_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_slt_sge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #8, lt -; SDISEL-NEXT: cset w0, ge -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_slt_sge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ge -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_slt_sge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #8, lt +; CHECK-SD-NEXT: cset w0, ge +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_slt_sge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ge +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp slt i32 %s0, %s1 %c1 = icmp sge i32 %s2, %s3 @@ -1683,21 +1683,21 @@ entry: } define i32 @and_sle_eq(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sle_eq: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, le -; SDISEL-NEXT: cset w0, eq -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sle_eq: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, le -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, eq -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sle_eq: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, le +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sle_eq: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, le +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, eq +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sle i32 %s0, %s1 %c1 = icmp eq i32 %s2, %s3 @@ -1707,21 +1707,21 @@ entry: } define i32 @and_sle_ne(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sle_ne: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #4, le -; SDISEL-NEXT: cset w0, ne -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sle_ne: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, le -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ne -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sle_ne: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #4, le +; CHECK-SD-NEXT: cset w0, ne +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sle_ne: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, le +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ne +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sle i32 %s0, %s1 %c1 = icmp ne i32 %s2, %s3 @@ -1731,21 +1731,21 @@ entry: } define i32 @and_sle_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sle_ult: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, le -; SDISEL-NEXT: cset w0, lo -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sle_ult: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, le -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lo -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sle_ult: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, le +; CHECK-SD-NEXT: cset w0, lo +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sle_ult: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, le +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lo +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sle i32 %s0, %s1 %c1 = icmp ult i32 %s2, %s3 @@ -1755,21 +1755,21 @@ entry: } define i32 @and_sle_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sle_ule: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, le -; SDISEL-NEXT: cset w0, ls -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sle_ule: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, le -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ls -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sle_ule: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, le +; CHECK-SD-NEXT: cset w0, ls +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sle_ule: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, le +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ls +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sle i32 %s0, %s1 %c1 = icmp ule i32 %s2, %s3 @@ -1779,21 +1779,21 @@ entry: } define i32 @and_sle_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sle_ugt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, le -; SDISEL-NEXT: cset w0, hi -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sle_ugt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, le -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hi -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sle_ugt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, le +; CHECK-SD-NEXT: cset w0, hi +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sle_ugt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, le +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hi +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sle i32 %s0, %s1 %c1 = icmp ugt i32 %s2, %s3 @@ -1803,21 +1803,21 @@ entry: } define i32 @and_sle_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sle_uge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, le -; SDISEL-NEXT: cset w0, hs -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sle_uge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, le -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hs -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sle_uge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, le +; CHECK-SD-NEXT: cset w0, hs +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sle_uge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, le +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hs +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sle i32 %s0, %s1 %c1 = icmp uge i32 %s2, %s3 @@ -1827,21 +1827,21 @@ entry: } define i32 @and_sle_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sle_slt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, le -; SDISEL-NEXT: cset w0, lt -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sle_slt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, le -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lt -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sle_slt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, le +; CHECK-SD-NEXT: cset w0, lt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sle_slt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, le +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lt +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sle i32 %s0, %s1 %c1 = icmp slt i32 %s2, %s3 @@ -1851,21 +1851,21 @@ entry: } define i32 @and_sle_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sle_sle: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, le -; SDISEL-NEXT: cset w0, le -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sle_sle: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, le -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, le -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sle_sle: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, le +; CHECK-SD-NEXT: cset w0, le +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sle_sle: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, le +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, le +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sle i32 %s0, %s1 %c1 = icmp sle i32 %s2, %s3 @@ -1875,21 +1875,21 @@ entry: } define i32 @and_sle_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sle_sgt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #4, le -; SDISEL-NEXT: cset w0, gt -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sle_sgt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, le -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, gt -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sle_sgt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #4, le +; CHECK-SD-NEXT: cset w0, gt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sle_sgt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, le +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, gt +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sle i32 %s0, %s1 %c1 = icmp sgt i32 %s2, %s3 @@ -1899,21 +1899,21 @@ entry: } define i32 @and_sle_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sle_sge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #8, le -; SDISEL-NEXT: cset w0, ge -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sle_sge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, le -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ge -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sle_sge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #8, le +; CHECK-SD-NEXT: cset w0, ge +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sle_sge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, le +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ge +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sle i32 %s0, %s1 %c1 = icmp sge i32 %s2, %s3 @@ -1923,21 +1923,21 @@ entry: } define i32 @and_sgt_eq(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sgt_eq: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, gt -; SDISEL-NEXT: cset w0, eq -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sgt_eq: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, gt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, eq -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sgt_eq: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, gt +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sgt_eq: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, gt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, eq +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sgt i32 %s0, %s1 %c1 = icmp eq i32 %s2, %s3 @@ -1947,21 +1947,21 @@ entry: } define i32 @and_sgt_ne(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sgt_ne: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #4, gt -; SDISEL-NEXT: cset w0, ne -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sgt_ne: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, gt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ne -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sgt_ne: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #4, gt +; CHECK-SD-NEXT: cset w0, ne +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sgt_ne: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, gt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ne +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sgt i32 %s0, %s1 %c1 = icmp ne i32 %s2, %s3 @@ -1971,21 +1971,21 @@ entry: } define i32 @and_sgt_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sgt_ult: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, gt -; SDISEL-NEXT: cset w0, lo -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sgt_ult: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, gt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lo -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sgt_ult: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, gt +; CHECK-SD-NEXT: cset w0, lo +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sgt_ult: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, gt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lo +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sgt i32 %s0, %s1 %c1 = icmp ult i32 %s2, %s3 @@ -1995,21 +1995,21 @@ entry: } define i32 @and_sgt_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sgt_ule: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, gt -; SDISEL-NEXT: cset w0, ls -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sgt_ule: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, gt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ls -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sgt_ule: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, gt +; CHECK-SD-NEXT: cset w0, ls +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sgt_ule: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, gt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ls +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sgt i32 %s0, %s1 %c1 = icmp ule i32 %s2, %s3 @@ -2019,21 +2019,21 @@ entry: } define i32 @and_sgt_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sgt_ugt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, gt -; SDISEL-NEXT: cset w0, hi -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sgt_ugt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, gt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hi -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sgt_ugt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, gt +; CHECK-SD-NEXT: cset w0, hi +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sgt_ugt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, gt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hi +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sgt i32 %s0, %s1 %c1 = icmp ugt i32 %s2, %s3 @@ -2043,21 +2043,21 @@ entry: } define i32 @and_sgt_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sgt_uge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, gt -; SDISEL-NEXT: cset w0, hs -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sgt_uge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, gt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hs -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sgt_uge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, gt +; CHECK-SD-NEXT: cset w0, hs +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sgt_uge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, gt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hs +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sgt i32 %s0, %s1 %c1 = icmp uge i32 %s2, %s3 @@ -2067,21 +2067,21 @@ entry: } define i32 @and_sgt_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sgt_slt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, gt -; SDISEL-NEXT: cset w0, lt -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sgt_slt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, gt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lt -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sgt_slt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, gt +; CHECK-SD-NEXT: cset w0, lt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sgt_slt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, gt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lt +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sgt i32 %s0, %s1 %c1 = icmp slt i32 %s2, %s3 @@ -2091,21 +2091,21 @@ entry: } define i32 @and_sgt_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sgt_sle: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, gt -; SDISEL-NEXT: cset w0, le -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sgt_sle: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, gt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, le -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sgt_sle: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, gt +; CHECK-SD-NEXT: cset w0, le +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sgt_sle: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, gt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, le +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sgt i32 %s0, %s1 %c1 = icmp sle i32 %s2, %s3 @@ -2115,21 +2115,21 @@ entry: } define i32 @and_sgt_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sgt_sgt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #4, gt -; SDISEL-NEXT: cset w0, gt -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sgt_sgt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, gt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, gt -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sgt_sgt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #4, gt +; CHECK-SD-NEXT: cset w0, gt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sgt_sgt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, gt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, gt +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sgt i32 %s0, %s1 %c1 = icmp sgt i32 %s2, %s3 @@ -2139,21 +2139,21 @@ entry: } define i32 @and_sgt_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sgt_sge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #8, gt -; SDISEL-NEXT: cset w0, ge -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sgt_sge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, gt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ge -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sgt_sge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #8, gt +; CHECK-SD-NEXT: cset w0, ge +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sgt_sge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, gt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ge +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sgt i32 %s0, %s1 %c1 = icmp sge i32 %s2, %s3 @@ -2163,21 +2163,21 @@ entry: } define i32 @and_sge_eq(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sge_eq: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, ge -; SDISEL-NEXT: cset w0, eq -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sge_eq: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ge -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, eq -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sge_eq: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, ge +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sge_eq: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ge +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, eq +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sge i32 %s0, %s1 %c1 = icmp eq i32 %s2, %s3 @@ -2187,21 +2187,21 @@ entry: } define i32 @and_sge_ne(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sge_ne: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #4, ge -; SDISEL-NEXT: cset w0, ne -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sge_ne: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ge -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ne -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sge_ne: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #4, ge +; CHECK-SD-NEXT: cset w0, ne +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sge_ne: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ge +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ne +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sge i32 %s0, %s1 %c1 = icmp ne i32 %s2, %s3 @@ -2211,21 +2211,21 @@ entry: } define i32 @and_sge_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sge_ult: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, ge -; SDISEL-NEXT: cset w0, lo -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sge_ult: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ge -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lo -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sge_ult: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, ge +; CHECK-SD-NEXT: cset w0, lo +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sge_ult: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ge +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lo +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sge i32 %s0, %s1 %c1 = icmp ult i32 %s2, %s3 @@ -2235,21 +2235,21 @@ entry: } define i32 @and_sge_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sge_ule: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, ge -; SDISEL-NEXT: cset w0, ls -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sge_ule: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ge -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ls -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sge_ule: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, ge +; CHECK-SD-NEXT: cset w0, ls +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sge_ule: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ge +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ls +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sge i32 %s0, %s1 %c1 = icmp ule i32 %s2, %s3 @@ -2259,21 +2259,21 @@ entry: } define i32 @and_sge_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sge_ugt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, ge -; SDISEL-NEXT: cset w0, hi -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sge_ugt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ge -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hi -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sge_ugt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, ge +; CHECK-SD-NEXT: cset w0, hi +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sge_ugt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ge +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hi +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sge i32 %s0, %s1 %c1 = icmp ugt i32 %s2, %s3 @@ -2283,21 +2283,21 @@ entry: } define i32 @and_sge_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sge_uge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, ge -; SDISEL-NEXT: cset w0, hs -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sge_uge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ge -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hs -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sge_uge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, ge +; CHECK-SD-NEXT: cset w0, hs +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sge_uge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ge +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hs +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sge i32 %s0, %s1 %c1 = icmp uge i32 %s2, %s3 @@ -2307,21 +2307,21 @@ entry: } define i32 @and_sge_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sge_slt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, ge -; SDISEL-NEXT: cset w0, lt -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sge_slt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ge -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lt -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sge_slt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, ge +; CHECK-SD-NEXT: cset w0, lt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sge_slt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ge +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lt +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sge i32 %s0, %s1 %c1 = icmp slt i32 %s2, %s3 @@ -2331,21 +2331,21 @@ entry: } define i32 @and_sge_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sge_sle: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, ge -; SDISEL-NEXT: cset w0, le -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sge_sle: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ge -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, le -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sge_sle: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, ge +; CHECK-SD-NEXT: cset w0, le +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sge_sle: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ge +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, le +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sge i32 %s0, %s1 %c1 = icmp sle i32 %s2, %s3 @@ -2355,21 +2355,21 @@ entry: } define i32 @and_sge_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sge_sgt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #4, ge -; SDISEL-NEXT: cset w0, gt -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sge_sgt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ge -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, gt -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sge_sgt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #4, ge +; CHECK-SD-NEXT: cset w0, gt +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sge_sgt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ge +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, gt +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sge i32 %s0, %s1 %c1 = icmp sgt i32 %s2, %s3 @@ -2379,21 +2379,21 @@ entry: } define i32 @and_sge_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) { -; SDISEL-LABEL: and_sge_sge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #8, ge -; SDISEL-NEXT: cset w0, ge -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_sge_sge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ge -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ge -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_sge_sge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #8, ge +; CHECK-SD-NEXT: cset w0, ge +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_sge_sge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ge +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ge +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %c0 = icmp sge i32 %s0, %s1 %c1 = icmp sge i32 %s2, %s3 @@ -2403,19 +2403,19 @@ entry: } define i32 @cmp_to_ands1(i32 %num) { -; SDISEL-LABEL: cmp_to_ands1: -; SDISEL: // %bb.0: -; SDISEL-NEXT: and w8, w0, #0xff -; SDISEL-NEXT: tst w0, #0xfe -; SDISEL-NEXT: csel w0, w8, wzr, ne -; SDISEL-NEXT: ret -; -; GISEL-LABEL: cmp_to_ands1: -; GISEL: // %bb.0: -; GISEL-NEXT: and w8, w0, #0xff -; GISEL-NEXT: cmp w8, #1 -; GISEL-NEXT: csel w0, w8, wzr, hi -; GISEL-NEXT: ret +; CHECK-SD-LABEL: cmp_to_ands1: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: and w8, w0, #0xff +; CHECK-SD-NEXT: tst w0, #0xfe +; CHECK-SD-NEXT: csel w0, w8, wzr, ne +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: cmp_to_ands1: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: and w8, w0, #0xff +; CHECK-GI-NEXT: cmp w8, #1 +; CHECK-GI-NEXT: csel w0, w8, wzr, hi +; CHECK-GI-NEXT: ret %and = and i32 %num, 255 %cmp = icmp ugt i32 %and, 1 %r = select i1 %cmp, i32 %and, i32 0 @@ -2423,19 +2423,19 @@ define i32 @cmp_to_ands1(i32 %num) { } define i32 @cmp_to_ands2(i32 %num) { -; SDISEL-LABEL: cmp_to_ands2: -; SDISEL: // %bb.0: -; SDISEL-NEXT: and w8, w0, #0xfe -; SDISEL-NEXT: tst w0, #0xc0 -; SDISEL-NEXT: csel w0, w8, wzr, ne -; SDISEL-NEXT: ret -; -; GISEL-LABEL: cmp_to_ands2: -; GISEL: // %bb.0: -; GISEL-NEXT: and w8, w0, #0xfe -; GISEL-NEXT: cmp w8, #63 -; GISEL-NEXT: csel w0, w8, wzr, hi -; GISEL-NEXT: ret +; CHECK-SD-LABEL: cmp_to_ands2: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: and w8, w0, #0xfe +; CHECK-SD-NEXT: tst w0, #0xc0 +; CHECK-SD-NEXT: csel w0, w8, wzr, ne +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: cmp_to_ands2: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: and w8, w0, #0xfe +; CHECK-GI-NEXT: cmp w8, #63 +; CHECK-GI-NEXT: csel w0, w8, wzr, hi +; CHECK-GI-NEXT: ret %and = and i32 %num, 254 %cmp = icmp ugt i32 %and, 63 %r = select i1 %cmp, i32 %and, i32 0 @@ -2443,19 +2443,19 @@ define i32 @cmp_to_ands2(i32 %num) { } define i32 @cmp_to_ands3(i32 %num, i32 %a) { -; SDISEL-LABEL: cmp_to_ands3: -; SDISEL: // %bb.0: -; SDISEL-NEXT: tst w0, #0x10 -; SDISEL-NEXT: csel w0, w1, wzr, ne -; SDISEL-NEXT: ret -; -; GISEL-LABEL: cmp_to_ands3: -; GISEL: // %bb.0: -; GISEL-NEXT: mov w8, #23 // =0x17 -; GISEL-NEXT: and w8, w0, w8 -; GISEL-NEXT: cmp w8, #7 -; GISEL-NEXT: csel w0, w1, wzr, hi -; GISEL-NEXT: ret +; CHECK-SD-LABEL: cmp_to_ands3: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: tst w0, #0x10 +; CHECK-SD-NEXT: csel w0, w1, wzr, ne +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: cmp_to_ands3: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: mov w8, #23 // =0x17 +; CHECK-GI-NEXT: and w8, w0, w8 +; CHECK-GI-NEXT: cmp w8, #7 +; CHECK-GI-NEXT: csel w0, w1, wzr, hi +; CHECK-GI-NEXT: ret %and = and i32 %num, 23 %cmp = icmp ugt i32 %and, 7 %r = select i1 %cmp, i32 %a, i32 0 @@ -2463,19 +2463,19 @@ define i32 @cmp_to_ands3(i32 %num, i32 %a) { } define i32 @cmp_to_ands4(i32 %num, i32 %a) { -; SDISEL-LABEL: cmp_to_ands4: -; SDISEL: // %bb.0: -; SDISEL-NEXT: and w8, w0, #0x30 -; SDISEL-NEXT: tst w0, #0x20 -; SDISEL-NEXT: csel w0, w8, w1, eq -; SDISEL-NEXT: ret -; -; GISEL-LABEL: cmp_to_ands4: -; GISEL: // %bb.0: -; GISEL-NEXT: and w8, w0, #0x30 -; GISEL-NEXT: cmp w8, #31 -; GISEL-NEXT: csel w0, w8, w1, ls -; GISEL-NEXT: ret +; CHECK-SD-LABEL: cmp_to_ands4: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: and w8, w0, #0x30 +; CHECK-SD-NEXT: tst w0, #0x20 +; CHECK-SD-NEXT: csel w0, w8, w1, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: cmp_to_ands4: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: and w8, w0, #0x30 +; CHECK-GI-NEXT: cmp w8, #31 +; CHECK-GI-NEXT: csel w0, w8, w1, ls +; CHECK-GI-NEXT: ret %and = and i32 %num, 48 %cmp = icmp ule i32 %and, 31 %r = select i1 %cmp, i32 %and, i32 %a @@ -2483,19 +2483,19 @@ define i32 @cmp_to_ands4(i32 %num, i32 %a) { } define i32 @cmp_to_ands5(i32 %num, i32 %a) { -; SDISEL-LABEL: cmp_to_ands5: -; SDISEL: // %bb.0: -; SDISEL-NEXT: and w8, w0, #0xf8 -; SDISEL-NEXT: tst w0, #0xc0 -; SDISEL-NEXT: csel w0, w8, w1, eq -; SDISEL-NEXT: ret -; -; GISEL-LABEL: cmp_to_ands5: -; GISEL: // %bb.0: -; GISEL-NEXT: and w8, w0, #0xf8 -; GISEL-NEXT: cmp w8, #64 -; GISEL-NEXT: csel w0, w8, w1, lo -; GISEL-NEXT: ret +; CHECK-SD-LABEL: cmp_to_ands5: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: and w8, w0, #0xf8 +; CHECK-SD-NEXT: tst w0, #0xc0 +; CHECK-SD-NEXT: csel w0, w8, w1, eq +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: cmp_to_ands5: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: and w8, w0, #0xf8 +; CHECK-GI-NEXT: cmp w8, #64 +; CHECK-GI-NEXT: csel w0, w8, w1, lo +; CHECK-GI-NEXT: ret %and = and i32 %num, 248 %cmp = icmp ult i32 %and, 64 %r = select i1 %cmp, i32 %and, i32 %a @@ -2503,19 +2503,19 @@ define i32 @cmp_to_ands5(i32 %num, i32 %a) { } define i32 @cmp_to_ands6(i32 %num) { -; SDISEL-LABEL: cmp_to_ands6: -; SDISEL: // %bb.0: -; SDISEL-NEXT: and w8, w0, #0xfe -; SDISEL-NEXT: tst w0, #0xf0 -; SDISEL-NEXT: csel w0, w8, wzr, ne -; SDISEL-NEXT: ret -; -; GISEL-LABEL: cmp_to_ands6: -; GISEL: // %bb.0: -; GISEL-NEXT: and w8, w0, #0xfe -; GISEL-NEXT: cmp w8, #16 -; GISEL-NEXT: csel w0, w8, wzr, hs -; GISEL-NEXT: ret +; CHECK-SD-LABEL: cmp_to_ands6: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: and w8, w0, #0xfe +; CHECK-SD-NEXT: tst w0, #0xf0 +; CHECK-SD-NEXT: csel w0, w8, wzr, ne +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: cmp_to_ands6: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: and w8, w0, #0xfe +; CHECK-GI-NEXT: cmp w8, #16 +; CHECK-GI-NEXT: csel w0, w8, wzr, hs +; CHECK-GI-NEXT: ret %and = and i32 %num, 254 %cmp = icmp uge i32 %and, 16 %r = select i1 %cmp, i32 %and, i32 0 @@ -2523,21 +2523,21 @@ define i32 @cmp_to_ands6(i32 %num) { } define i1 @and_fcmp(float %0, float %1) { -; SDISEL-LABEL: and_fcmp: -; SDISEL: // %bb.0: -; SDISEL-NEXT: fcmp s1, s1 -; SDISEL-NEXT: fccmp s0, s0, #0, vs -; SDISEL-NEXT: cset w0, vs -; SDISEL-NEXT: ret -; -; GISEL-LABEL: and_fcmp: -; GISEL: // %bb.0: -; GISEL-NEXT: fcmp s0, #0.0 -; GISEL-NEXT: cset w8, vs -; GISEL-NEXT: fcmp s1, #0.0 -; GISEL-NEXT: cset w9, vs -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-SD-LABEL: and_fcmp: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: fcmp s1, s1 +; CHECK-SD-NEXT: fccmp s0, s0, #0, vs +; CHECK-SD-NEXT: cset w0, vs +; CHECK-SD-NEXT: ret +; +; CHECK-GI-LABEL: and_fcmp: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: fcmp s0, #0.0 +; CHECK-GI-NEXT: cset w8, vs +; CHECK-GI-NEXT: fcmp s1, #0.0 +; CHECK-GI-NEXT: cset w9, vs +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret %3 = fcmp uno float %0, 0.000000e+00 %4 = fcmp uno float %1, 0.000000e+00 diff --git a/llvm/test/CodeGen/AArch64/andorbrcompare.ll b/llvm/test/CodeGen/AArch64/andorbrcompare.ll index 951a5cd..5bc06ec 100644 --- a/llvm/test/CodeGen/AArch64/andorbrcompare.ll +++ b/llvm/test/CodeGen/AArch64/andorbrcompare.ll @@ -1,44 +1,44 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc -mtriple=aarch64-none-elf -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,SDISEL -; RUN: llc -mtriple=aarch64-none-elf -global-isel -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,GISEL +; RUN: llc -mtriple=aarch64-none-elf -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SD +; RUN: llc -mtriple=aarch64-none-elf -global-isel -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-GI declare void @dummy() define i32 @and_eq_ne_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %s4, i32 %s5, ptr %p) { -; SDISEL-LABEL: and_eq_ne_ult: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w2, w3 -; SDISEL-NEXT: ccmp w0, w1, #0, ne -; SDISEL-NEXT: b.eq .LBB0_3 -; SDISEL-NEXT: // %bb.1: // %entry -; SDISEL-NEXT: cmp w4, w5 -; SDISEL-NEXT: b.lo .LBB0_3 -; SDISEL-NEXT: // %bb.2: -; SDISEL-NEXT: mov w0, wzr -; SDISEL-NEXT: ret -; SDISEL-NEXT: .LBB0_3: // %if -; SDISEL-NEXT: mov w0, #1 // =0x1 -; SDISEL-NEXT: str w0, [x6] -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: and_eq_ne_ult: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w2, w3 +; CHECK-SD-NEXT: ccmp w0, w1, #0, ne +; CHECK-SD-NEXT: b.eq .LBB0_3 +; CHECK-SD-NEXT: // %bb.1: // %entry +; CHECK-SD-NEXT: cmp w4, w5 +; CHECK-SD-NEXT: b.lo .LBB0_3 +; CHECK-SD-NEXT: // %bb.2: +; CHECK-SD-NEXT: mov w0, wzr +; CHECK-SD-NEXT: ret +; CHECK-SD-NEXT: .LBB0_3: // %if +; CHECK-SD-NEXT: mov w0, #1 // =0x1 +; CHECK-SD-NEXT: str w0, [x6] +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: and_eq_ne_ult: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, eq -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ne -; GISEL-NEXT: and w8, w8, w9 -; GISEL-NEXT: tbnz w8, #0, .LBB0_3 -; GISEL-NEXT: // %bb.1: // %entry -; GISEL-NEXT: cmp w4, w5 -; GISEL-NEXT: mov w0, wzr -; GISEL-NEXT: b.lo .LBB0_3 -; GISEL-NEXT: // %bb.2: // %common.ret -; GISEL-NEXT: ret -; GISEL-NEXT: .LBB0_3: // %if -; GISEL-NEXT: mov w0, #1 // =0x1 -; GISEL-NEXT: str w0, [x6] -; GISEL-NEXT: ret +; CHECK-GI-LABEL: and_eq_ne_ult: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, eq +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ne +; CHECK-GI-NEXT: and w8, w8, w9 +; CHECK-GI-NEXT: tbnz w8, #0, .LBB0_3 +; CHECK-GI-NEXT: // %bb.1: // %entry +; CHECK-GI-NEXT: cmp w4, w5 +; CHECK-GI-NEXT: mov w0, wzr +; CHECK-GI-NEXT: b.lo .LBB0_3 +; CHECK-GI-NEXT: // %bb.2: // %common.ret +; CHECK-GI-NEXT: ret +; CHECK-GI-NEXT: .LBB0_3: // %if +; CHECK-GI-NEXT: mov w0, #1 // =0x1 +; CHECK-GI-NEXT: str w0, [x6] +; CHECK-GI-NEXT: ret entry: %c0 = icmp eq i32 %s0, %s1 %c1 = icmp ne i32 %s2, %s3 @@ -56,40 +56,40 @@ else: } define i32 @and_ne_ult_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %s4, i32 %s5, ptr %p) { -; SDISEL-LABEL: and_ne_ult_ule: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w2, w3 -; SDISEL-NEXT: ccmp w0, w1, #4, lo -; SDISEL-NEXT: b.ne .LBB1_3 -; SDISEL-NEXT: // %bb.1: // %entry -; SDISEL-NEXT: cmp w4, w5 -; SDISEL-NEXT: b.ls .LBB1_3 -; SDISEL-NEXT: // %bb.2: -; SDISEL-NEXT: mov w0, wzr -; SDISEL-NEXT: ret -; SDISEL-NEXT: .LBB1_3: // %if -; SDISEL-NEXT: mov w0, #1 // =0x1 -; SDISEL-NEXT: str w0, [x6] -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: and_ne_ult_ule: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w2, w3 +; CHECK-SD-NEXT: ccmp w0, w1, #4, lo +; CHECK-SD-NEXT: b.ne .LBB1_3 +; CHECK-SD-NEXT: // %bb.1: // %entry +; CHECK-SD-NEXT: cmp w4, w5 +; CHECK-SD-NEXT: b.ls .LBB1_3 +; CHECK-SD-NEXT: // %bb.2: +; CHECK-SD-NEXT: mov w0, wzr +; CHECK-SD-NEXT: ret +; CHECK-SD-NEXT: .LBB1_3: // %if +; CHECK-SD-NEXT: mov w0, #1 // =0x1 +; CHECK-SD-NEXT: str w0, [x6] +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: and_ne_ult_ule: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ne -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lo -; GISEL-NEXT: and w8, w8, w9 -; GISEL-NEXT: tbnz w8, #0, .LBB1_3 -; GISEL-NEXT: // %bb.1: // %entry -; GISEL-NEXT: cmp w4, w5 -; GISEL-NEXT: mov w0, wzr -; GISEL-NEXT: b.ls .LBB1_3 -; GISEL-NEXT: // %bb.2: // %common.ret -; GISEL-NEXT: ret -; GISEL-NEXT: .LBB1_3: // %if -; GISEL-NEXT: mov w0, #1 // =0x1 -; GISEL-NEXT: str w0, [x6] -; GISEL-NEXT: ret +; CHECK-GI-LABEL: and_ne_ult_ule: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ne +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lo +; CHECK-GI-NEXT: and w8, w8, w9 +; CHECK-GI-NEXT: tbnz w8, #0, .LBB1_3 +; CHECK-GI-NEXT: // %bb.1: // %entry +; CHECK-GI-NEXT: cmp w4, w5 +; CHECK-GI-NEXT: mov w0, wzr +; CHECK-GI-NEXT: b.ls .LBB1_3 +; CHECK-GI-NEXT: // %bb.2: // %common.ret +; CHECK-GI-NEXT: ret +; CHECK-GI-NEXT: .LBB1_3: // %if +; CHECK-GI-NEXT: mov w0, #1 // =0x1 +; CHECK-GI-NEXT: str w0, [x6] +; CHECK-GI-NEXT: ret entry: %c0 = icmp ne i32 %s0, %s1 %c1 = icmp ult i32 %s2, %s3 @@ -107,40 +107,40 @@ else: } define i32 @and_ult_ule_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %s4, i32 %s5, ptr %p) { -; SDISEL-LABEL: and_ult_ule_ugt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w2, w3 -; SDISEL-NEXT: ccmp w0, w1, #2, ls -; SDISEL-NEXT: b.lo .LBB2_3 -; SDISEL-NEXT: // %bb.1: // %entry -; SDISEL-NEXT: cmp w4, w5 -; SDISEL-NEXT: b.hi .LBB2_3 -; SDISEL-NEXT: // %bb.2: -; SDISEL-NEXT: mov w0, wzr -; SDISEL-NEXT: ret -; SDISEL-NEXT: .LBB2_3: // %if -; SDISEL-NEXT: mov w0, #1 // =0x1 -; SDISEL-NEXT: str w0, [x6] -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: and_ult_ule_ugt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w2, w3 +; CHECK-SD-NEXT: ccmp w0, w1, #2, ls +; CHECK-SD-NEXT: b.lo .LBB2_3 +; CHECK-SD-NEXT: // %bb.1: // %entry +; CHECK-SD-NEXT: cmp w4, w5 +; CHECK-SD-NEXT: b.hi .LBB2_3 +; CHECK-SD-NEXT: // %bb.2: +; CHECK-SD-NEXT: mov w0, wzr +; CHECK-SD-NEXT: ret +; CHECK-SD-NEXT: .LBB2_3: // %if +; CHECK-SD-NEXT: mov w0, #1 // =0x1 +; CHECK-SD-NEXT: str w0, [x6] +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: and_ult_ule_ugt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lo -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ls -; GISEL-NEXT: and w8, w8, w9 -; GISEL-NEXT: tbnz w8, #0, .LBB2_3 -; GISEL-NEXT: // %bb.1: // %entry -; GISEL-NEXT: cmp w4, w5 -; GISEL-NEXT: mov w0, wzr -; GISEL-NEXT: b.hi .LBB2_3 -; GISEL-NEXT: // %bb.2: // %common.ret -; GISEL-NEXT: ret -; GISEL-NEXT: .LBB2_3: // %if -; GISEL-NEXT: mov w0, #1 // =0x1 -; GISEL-NEXT: str w0, [x6] -; GISEL-NEXT: ret +; CHECK-GI-LABEL: and_ult_ule_ugt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lo +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ls +; CHECK-GI-NEXT: and w8, w8, w9 +; CHECK-GI-NEXT: tbnz w8, #0, .LBB2_3 +; CHECK-GI-NEXT: // %bb.1: // %entry +; CHECK-GI-NEXT: cmp w4, w5 +; CHECK-GI-NEXT: mov w0, wzr +; CHECK-GI-NEXT: b.hi .LBB2_3 +; CHECK-GI-NEXT: // %bb.2: // %common.ret +; CHECK-GI-NEXT: ret +; CHECK-GI-NEXT: .LBB2_3: // %if +; CHECK-GI-NEXT: mov w0, #1 // =0x1 +; CHECK-GI-NEXT: str w0, [x6] +; CHECK-GI-NEXT: ret entry: %c0 = icmp ult i32 %s0, %s1 %c1 = icmp ule i32 %s2, %s3 @@ -158,40 +158,40 @@ else: } define i32 @and_ule_ugt_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %s4, i32 %s5, ptr %p) { -; SDISEL-LABEL: and_ule_ugt_uge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w2, w3 -; SDISEL-NEXT: ccmp w0, w1, #2, hi -; SDISEL-NEXT: b.ls .LBB3_3 -; SDISEL-NEXT: // %bb.1: // %entry -; SDISEL-NEXT: cmp w4, w5 -; SDISEL-NEXT: b.hs .LBB3_3 -; SDISEL-NEXT: // %bb.2: -; SDISEL-NEXT: mov w0, wzr -; SDISEL-NEXT: ret -; SDISEL-NEXT: .LBB3_3: // %if -; SDISEL-NEXT: mov w0, #1 // =0x1 -; SDISEL-NEXT: str w0, [x6] -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: and_ule_ugt_uge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w2, w3 +; CHECK-SD-NEXT: ccmp w0, w1, #2, hi +; CHECK-SD-NEXT: b.ls .LBB3_3 +; CHECK-SD-NEXT: // %bb.1: // %entry +; CHECK-SD-NEXT: cmp w4, w5 +; CHECK-SD-NEXT: b.hs .LBB3_3 +; CHECK-SD-NEXT: // %bb.2: +; CHECK-SD-NEXT: mov w0, wzr +; CHECK-SD-NEXT: ret +; CHECK-SD-NEXT: .LBB3_3: // %if +; CHECK-SD-NEXT: mov w0, #1 // =0x1 +; CHECK-SD-NEXT: str w0, [x6] +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: and_ule_ugt_uge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, ls -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hi -; GISEL-NEXT: and w8, w8, w9 -; GISEL-NEXT: tbnz w8, #0, .LBB3_3 -; GISEL-NEXT: // %bb.1: // %entry -; GISEL-NEXT: cmp w4, w5 -; GISEL-NEXT: mov w0, wzr -; GISEL-NEXT: b.hs .LBB3_3 -; GISEL-NEXT: // %bb.2: // %common.ret -; GISEL-NEXT: ret -; GISEL-NEXT: .LBB3_3: // %if -; GISEL-NEXT: mov w0, #1 // =0x1 -; GISEL-NEXT: str w0, [x6] -; GISEL-NEXT: ret +; CHECK-GI-LABEL: and_ule_ugt_uge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, ls +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hi +; CHECK-GI-NEXT: and w8, w8, w9 +; CHECK-GI-NEXT: tbnz w8, #0, .LBB3_3 +; CHECK-GI-NEXT: // %bb.1: // %entry +; CHECK-GI-NEXT: cmp w4, w5 +; CHECK-GI-NEXT: mov w0, wzr +; CHECK-GI-NEXT: b.hs .LBB3_3 +; CHECK-GI-NEXT: // %bb.2: // %common.ret +; CHECK-GI-NEXT: ret +; CHECK-GI-NEXT: .LBB3_3: // %if +; CHECK-GI-NEXT: mov w0, #1 // =0x1 +; CHECK-GI-NEXT: str w0, [x6] +; CHECK-GI-NEXT: ret entry: %c0 = icmp ule i32 %s0, %s1 %c1 = icmp ugt i32 %s2, %s3 @@ -209,40 +209,40 @@ else: } define i32 @and_ugt_uge_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %s4, i32 %s5, ptr %p) { -; SDISEL-LABEL: and_ugt_uge_slt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w2, w3 -; SDISEL-NEXT: ccmp w0, w1, #0, hs -; SDISEL-NEXT: b.hi .LBB4_3 -; SDISEL-NEXT: // %bb.1: // %entry -; SDISEL-NEXT: cmp w4, w5 -; SDISEL-NEXT: b.lt .LBB4_3 -; SDISEL-NEXT: // %bb.2: -; SDISEL-NEXT: mov w0, wzr -; SDISEL-NEXT: ret -; SDISEL-NEXT: .LBB4_3: // %if -; SDISEL-NEXT: mov w0, #1 // =0x1 -; SDISEL-NEXT: str w0, [x6] -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: and_ugt_uge_slt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w2, w3 +; CHECK-SD-NEXT: ccmp w0, w1, #0, hs +; CHECK-SD-NEXT: b.hi .LBB4_3 +; CHECK-SD-NEXT: // %bb.1: // %entry +; CHECK-SD-NEXT: cmp w4, w5 +; CHECK-SD-NEXT: b.lt .LBB4_3 +; CHECK-SD-NEXT: // %bb.2: +; CHECK-SD-NEXT: mov w0, wzr +; CHECK-SD-NEXT: ret +; CHECK-SD-NEXT: .LBB4_3: // %if +; CHECK-SD-NEXT: mov w0, #1 // =0x1 +; CHECK-SD-NEXT: str w0, [x6] +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: and_ugt_uge_slt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hi -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hs -; GISEL-NEXT: and w8, w8, w9 -; GISEL-NEXT: tbnz w8, #0, .LBB4_3 -; GISEL-NEXT: // %bb.1: // %entry -; GISEL-NEXT: cmp w4, w5 -; GISEL-NEXT: mov w0, wzr -; GISEL-NEXT: b.lt .LBB4_3 -; GISEL-NEXT: // %bb.2: // %common.ret -; GISEL-NEXT: ret -; GISEL-NEXT: .LBB4_3: // %if -; GISEL-NEXT: mov w0, #1 // =0x1 -; GISEL-NEXT: str w0, [x6] -; GISEL-NEXT: ret +; CHECK-GI-LABEL: and_ugt_uge_slt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hi +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hs +; CHECK-GI-NEXT: and w8, w8, w9 +; CHECK-GI-NEXT: tbnz w8, #0, .LBB4_3 +; CHECK-GI-NEXT: // %bb.1: // %entry +; CHECK-GI-NEXT: cmp w4, w5 +; CHECK-GI-NEXT: mov w0, wzr +; CHECK-GI-NEXT: b.lt .LBB4_3 +; CHECK-GI-NEXT: // %bb.2: // %common.ret +; CHECK-GI-NEXT: ret +; CHECK-GI-NEXT: .LBB4_3: // %if +; CHECK-GI-NEXT: mov w0, #1 // =0x1 +; CHECK-GI-NEXT: str w0, [x6] +; CHECK-GI-NEXT: ret entry: %c0 = icmp ugt i32 %s0, %s1 %c1 = icmp uge i32 %s2, %s3 @@ -260,40 +260,40 @@ else: } define i32 @and_uge_slt_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %s4, i32 %s5, ptr %p) { -; SDISEL-LABEL: and_uge_slt_sle: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w2, w3 -; SDISEL-NEXT: ccmp w0, w1, #0, lt -; SDISEL-NEXT: b.hs .LBB5_3 -; SDISEL-NEXT: // %bb.1: // %entry -; SDISEL-NEXT: cmp w4, w5 -; SDISEL-NEXT: b.le .LBB5_3 -; SDISEL-NEXT: // %bb.2: -; SDISEL-NEXT: mov w0, wzr -; SDISEL-NEXT: ret -; SDISEL-NEXT: .LBB5_3: // %if -; SDISEL-NEXT: mov w0, #1 // =0x1 -; SDISEL-NEXT: str w0, [x6] -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: and_uge_slt_sle: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w2, w3 +; CHECK-SD-NEXT: ccmp w0, w1, #0, lt +; CHECK-SD-NEXT: b.hs .LBB5_3 +; CHECK-SD-NEXT: // %bb.1: // %entry +; CHECK-SD-NEXT: cmp w4, w5 +; CHECK-SD-NEXT: b.le .LBB5_3 +; CHECK-SD-NEXT: // %bb.2: +; CHECK-SD-NEXT: mov w0, wzr +; CHECK-SD-NEXT: ret +; CHECK-SD-NEXT: .LBB5_3: // %if +; CHECK-SD-NEXT: mov w0, #1 // =0x1 +; CHECK-SD-NEXT: str w0, [x6] +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: and_uge_slt_sle: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, hs -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, lt -; GISEL-NEXT: and w8, w8, w9 -; GISEL-NEXT: tbnz w8, #0, .LBB5_3 -; GISEL-NEXT: // %bb.1: // %entry -; GISEL-NEXT: cmp w4, w5 -; GISEL-NEXT: mov w0, wzr -; GISEL-NEXT: b.le .LBB5_3 -; GISEL-NEXT: // %bb.2: // %common.ret -; GISEL-NEXT: ret -; GISEL-NEXT: .LBB5_3: // %if -; GISEL-NEXT: mov w0, #1 // =0x1 -; GISEL-NEXT: str w0, [x6] -; GISEL-NEXT: ret +; CHECK-GI-LABEL: and_uge_slt_sle: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, hs +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, lt +; CHECK-GI-NEXT: and w8, w8, w9 +; CHECK-GI-NEXT: tbnz w8, #0, .LBB5_3 +; CHECK-GI-NEXT: // %bb.1: // %entry +; CHECK-GI-NEXT: cmp w4, w5 +; CHECK-GI-NEXT: mov w0, wzr +; CHECK-GI-NEXT: b.le .LBB5_3 +; CHECK-GI-NEXT: // %bb.2: // %common.ret +; CHECK-GI-NEXT: ret +; CHECK-GI-NEXT: .LBB5_3: // %if +; CHECK-GI-NEXT: mov w0, #1 // =0x1 +; CHECK-GI-NEXT: str w0, [x6] +; CHECK-GI-NEXT: ret entry: %c0 = icmp uge i32 %s0, %s1 %c1 = icmp slt i32 %s2, %s3 @@ -311,40 +311,40 @@ else: } define i32 @and_slt_sle_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %s4, i32 %s5, ptr %p) { -; SDISEL-LABEL: and_slt_sle_sgt: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w2, w3 -; SDISEL-NEXT: ccmp w0, w1, #0, le -; SDISEL-NEXT: b.lt .LBB6_3 -; SDISEL-NEXT: // %bb.1: // %entry -; SDISEL-NEXT: cmp w4, w5 -; SDISEL-NEXT: b.gt .LBB6_3 -; SDISEL-NEXT: // %bb.2: -; SDISEL-NEXT: mov w0, wzr -; SDISEL-NEXT: ret -; SDISEL-NEXT: .LBB6_3: // %if -; SDISEL-NEXT: mov w0, #1 // =0x1 -; SDISEL-NEXT: str w0, [x6] -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: and_slt_sle_sgt: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w2, w3 +; CHECK-SD-NEXT: ccmp w0, w1, #0, le +; CHECK-SD-NEXT: b.lt .LBB6_3 +; CHECK-SD-NEXT: // %bb.1: // %entry +; CHECK-SD-NEXT: cmp w4, w5 +; CHECK-SD-NEXT: b.gt .LBB6_3 +; CHECK-SD-NEXT: // %bb.2: +; CHECK-SD-NEXT: mov w0, wzr +; CHECK-SD-NEXT: ret +; CHECK-SD-NEXT: .LBB6_3: // %if +; CHECK-SD-NEXT: mov w0, #1 // =0x1 +; CHECK-SD-NEXT: str w0, [x6] +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: and_slt_sle_sgt: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, le -; GISEL-NEXT: and w8, w8, w9 -; GISEL-NEXT: tbnz w8, #0, .LBB6_3 -; GISEL-NEXT: // %bb.1: // %entry -; GISEL-NEXT: cmp w4, w5 -; GISEL-NEXT: mov w0, wzr -; GISEL-NEXT: b.gt .LBB6_3 -; GISEL-NEXT: // %bb.2: // %common.ret -; GISEL-NEXT: ret -; GISEL-NEXT: .LBB6_3: // %if -; GISEL-NEXT: mov w0, #1 // =0x1 -; GISEL-NEXT: str w0, [x6] -; GISEL-NEXT: ret +; CHECK-GI-LABEL: and_slt_sle_sgt: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, le +; CHECK-GI-NEXT: and w8, w8, w9 +; CHECK-GI-NEXT: tbnz w8, #0, .LBB6_3 +; CHECK-GI-NEXT: // %bb.1: // %entry +; CHECK-GI-NEXT: cmp w4, w5 +; CHECK-GI-NEXT: mov w0, wzr +; CHECK-GI-NEXT: b.gt .LBB6_3 +; CHECK-GI-NEXT: // %bb.2: // %common.ret +; CHECK-GI-NEXT: ret +; CHECK-GI-NEXT: .LBB6_3: // %if +; CHECK-GI-NEXT: mov w0, #1 // =0x1 +; CHECK-GI-NEXT: str w0, [x6] +; CHECK-GI-NEXT: ret entry: %c0 = icmp slt i32 %s0, %s1 %c1 = icmp sle i32 %s2, %s3 @@ -362,40 +362,40 @@ else: } define i32 @and_sle_sgt_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %s4, i32 %s5, ptr %p) { -; SDISEL-LABEL: and_sle_sgt_sge: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: cmp w2, w3 -; SDISEL-NEXT: ccmp w0, w1, #0, gt -; SDISEL-NEXT: b.le .LBB7_3 -; SDISEL-NEXT: // %bb.1: // %entry -; SDISEL-NEXT: cmp w4, w5 -; SDISEL-NEXT: b.ge .LBB7_3 -; SDISEL-NEXT: // %bb.2: -; SDISEL-NEXT: mov w0, wzr -; SDISEL-NEXT: ret -; SDISEL-NEXT: .LBB7_3: // %if -; SDISEL-NEXT: mov w0, #1 // =0x1 -; SDISEL-NEXT: str w0, [x6] -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: and_sle_sgt_sge: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: cmp w2, w3 +; CHECK-SD-NEXT: ccmp w0, w1, #0, gt +; CHECK-SD-NEXT: b.le .LBB7_3 +; CHECK-SD-NEXT: // %bb.1: // %entry +; CHECK-SD-NEXT: cmp w4, w5 +; CHECK-SD-NEXT: b.ge .LBB7_3 +; CHECK-SD-NEXT: // %bb.2: +; CHECK-SD-NEXT: mov w0, wzr +; CHECK-SD-NEXT: ret +; CHECK-SD-NEXT: .LBB7_3: // %if +; CHECK-SD-NEXT: mov w0, #1 // =0x1 +; CHECK-SD-NEXT: str w0, [x6] +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: and_sle_sgt_sge: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, le -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, gt -; GISEL-NEXT: and w8, w8, w9 -; GISEL-NEXT: tbnz w8, #0, .LBB7_3 -; GISEL-NEXT: // %bb.1: // %entry -; GISEL-NEXT: cmp w4, w5 -; GISEL-NEXT: mov w0, wzr -; GISEL-NEXT: b.ge .LBB7_3 -; GISEL-NEXT: // %bb.2: // %common.ret -; GISEL-NEXT: ret -; GISEL-NEXT: .LBB7_3: // %if -; GISEL-NEXT: mov w0, #1 // =0x1 -; GISEL-NEXT: str w0, [x6] -; GISEL-NEXT: ret +; CHECK-GI-LABEL: and_sle_sgt_sge: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, le +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, gt +; CHECK-GI-NEXT: and w8, w8, w9 +; CHECK-GI-NEXT: tbnz w8, #0, .LBB7_3 +; CHECK-GI-NEXT: // %bb.1: // %entry +; CHECK-GI-NEXT: cmp w4, w5 +; CHECK-GI-NEXT: mov w0, wzr +; CHECK-GI-NEXT: b.ge .LBB7_3 +; CHECK-GI-NEXT: // %bb.2: // %common.ret +; CHECK-GI-NEXT: ret +; CHECK-GI-NEXT: .LBB7_3: // %if +; CHECK-GI-NEXT: mov w0, #1 // =0x1 +; CHECK-GI-NEXT: str w0, [x6] +; CHECK-GI-NEXT: ret entry: %c0 = icmp sle i32 %s0, %s1 %c1 = icmp sgt i32 %s2, %s3 diff --git a/llvm/test/CodeGen/AArch64/arm64-ccmp.ll b/llvm/test/CodeGen/AArch64/arm64-ccmp.ll index 06e957f..a546ffd 100644 --- a/llvm/test/CodeGen/AArch64/arm64-ccmp.ll +++ b/llvm/test/CodeGen/AArch64/arm64-ccmp.ll @@ -1,6 +1,6 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc < %s -debugify-and-strip-all-safe -mcpu=cyclone -verify-machineinstrs -aarch64-enable-ccmp -aarch64-stress-ccmp | FileCheck %s --check-prefixes=CHECK,SDISEL -; RUN: llc < %s -debugify-and-strip-all-safe -mcpu=cyclone -verify-machineinstrs -aarch64-enable-ccmp -aarch64-stress-ccmp -global-isel | FileCheck %s --check-prefixes=CHECK,GISEL +; RUN: llc < %s -debugify-and-strip-all-safe -mcpu=cyclone -verify-machineinstrs -aarch64-enable-ccmp -aarch64-stress-ccmp | FileCheck %s --check-prefixes=CHECK,CHECK-SD +; RUN: llc < %s -debugify-and-strip-all-safe -mcpu=cyclone -verify-machineinstrs -aarch64-enable-ccmp -aarch64-stress-ccmp -global-isel | FileCheck %s --check-prefixes=CHECK,CHECK-GI target triple = "arm64-apple-ios" define i32 @single_same(i32 %a, i32 %b) nounwind ssp { @@ -32,31 +32,31 @@ if.end: ; Different condition codes for the two compares. define i32 @single_different(i32 %a, i32 %b) nounwind ssp { -; SDISEL-LABEL: single_different: -; SDISEL: ; %bb.0: ; %entry -; SDISEL-NEXT: cmp w0, #6 -; SDISEL-NEXT: ccmp w1, #17, #0, ge -; SDISEL-NEXT: b.eq LBB1_2 -; SDISEL-NEXT: ; %bb.1: ; %if.then -; SDISEL-NEXT: stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill -; SDISEL-NEXT: bl _foo -; SDISEL-NEXT: ldp x29, x30, [sp], #16 ; 16-byte Folded Reload -; SDISEL-NEXT: LBB1_2: ; %if.end -; SDISEL-NEXT: mov w0, #7 ; =0x7 -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: single_different: +; CHECK-SD: ; %bb.0: ; %entry +; CHECK-SD-NEXT: cmp w0, #6 +; CHECK-SD-NEXT: ccmp w1, #17, #0, ge +; CHECK-SD-NEXT: b.eq LBB1_2 +; CHECK-SD-NEXT: ; %bb.1: ; %if.then +; CHECK-SD-NEXT: stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill +; CHECK-SD-NEXT: bl _foo +; CHECK-SD-NEXT: ldp x29, x30, [sp], #16 ; 16-byte Folded Reload +; CHECK-SD-NEXT: LBB1_2: ; %if.end +; CHECK-SD-NEXT: mov w0, #7 ; =0x7 +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: single_different: -; GISEL: ; %bb.0: ; %entry -; GISEL-NEXT: cmp w0, #5 -; GISEL-NEXT: ccmp w1, #17, #0, gt -; GISEL-NEXT: b.eq LBB1_2 -; GISEL-NEXT: ; %bb.1: ; %if.then -; GISEL-NEXT: stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill -; GISEL-NEXT: bl _foo -; GISEL-NEXT: ldp x29, x30, [sp], #16 ; 16-byte Folded Reload -; GISEL-NEXT: LBB1_2: ; %if.end -; GISEL-NEXT: mov w0, #7 ; =0x7 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: single_different: +; CHECK-GI: ; %bb.0: ; %entry +; CHECK-GI-NEXT: cmp w0, #5 +; CHECK-GI-NEXT: ccmp w1, #17, #0, gt +; CHECK-GI-NEXT: b.eq LBB1_2 +; CHECK-GI-NEXT: ; %bb.1: ; %if.then +; CHECK-GI-NEXT: stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill +; CHECK-GI-NEXT: bl _foo +; CHECK-GI-NEXT: ldp x29, x30, [sp], #16 ; 16-byte Folded Reload +; CHECK-GI-NEXT: LBB1_2: ; %if.end +; CHECK-GI-NEXT: mov w0, #7 ; =0x7 +; CHECK-GI-NEXT: ret entry: %cmp = icmp sle i32 %a, 5 %cmp1 = icmp ne i32 %b, 17 @@ -73,41 +73,41 @@ if.end: ; Second block clobbers the flags, can't convert (easily). define i32 @single_flagclobber(i32 %a, i32 %b) nounwind ssp { -; SDISEL-LABEL: single_flagclobber: -; SDISEL: ; %bb.0: ; %entry -; SDISEL-NEXT: cmp w0, #5 -; SDISEL-NEXT: b.eq LBB2_2 -; SDISEL-NEXT: ; %bb.1: ; %lor.lhs.false -; SDISEL-NEXT: lsl w8, w1, #1 -; SDISEL-NEXT: cmp w1, #7 -; SDISEL-NEXT: csinc w8, w8, w1, lt -; SDISEL-NEXT: cmp w8, #16 -; SDISEL-NEXT: b.gt LBB2_3 -; SDISEL-NEXT: LBB2_2: ; %if.then -; SDISEL-NEXT: stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill -; SDISEL-NEXT: bl _foo -; SDISEL-NEXT: ldp x29, x30, [sp], #16 ; 16-byte Folded Reload -; SDISEL-NEXT: LBB2_3: ; %if.end -; SDISEL-NEXT: mov w0, #7 ; =0x7 -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: single_flagclobber: +; CHECK-SD: ; %bb.0: ; %entry +; CHECK-SD-NEXT: cmp w0, #5 +; CHECK-SD-NEXT: b.eq LBB2_2 +; CHECK-SD-NEXT: ; %bb.1: ; %lor.lhs.false +; CHECK-SD-NEXT: lsl w8, w1, #1 +; CHECK-SD-NEXT: cmp w1, #7 +; CHECK-SD-NEXT: csinc w8, w8, w1, lt +; CHECK-SD-NEXT: cmp w8, #16 +; CHECK-SD-NEXT: b.gt LBB2_3 +; CHECK-SD-NEXT: LBB2_2: ; %if.then +; CHECK-SD-NEXT: stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill +; CHECK-SD-NEXT: bl _foo +; CHECK-SD-NEXT: ldp x29, x30, [sp], #16 ; 16-byte Folded Reload +; CHECK-SD-NEXT: LBB2_3: ; %if.end +; CHECK-SD-NEXT: mov w0, #7 ; =0x7 +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: single_flagclobber: -; GISEL: ; %bb.0: ; %entry -; GISEL-NEXT: cmp w0, #5 -; GISEL-NEXT: b.eq LBB2_2 -; GISEL-NEXT: ; %bb.1: ; %lor.lhs.false -; GISEL-NEXT: lsl w8, w1, #1 -; GISEL-NEXT: cmp w1, #7 -; GISEL-NEXT: csinc w8, w8, w1, lt -; GISEL-NEXT: cmp w8, #17 -; GISEL-NEXT: b.ge LBB2_3 -; GISEL-NEXT: LBB2_2: ; %if.then -; GISEL-NEXT: stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill -; GISEL-NEXT: bl _foo -; GISEL-NEXT: ldp x29, x30, [sp], #16 ; 16-byte Folded Reload -; GISEL-NEXT: LBB2_3: ; %if.end -; GISEL-NEXT: mov w0, #7 ; =0x7 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: single_flagclobber: +; CHECK-GI: ; %bb.0: ; %entry +; CHECK-GI-NEXT: cmp w0, #5 +; CHECK-GI-NEXT: b.eq LBB2_2 +; CHECK-GI-NEXT: ; %bb.1: ; %lor.lhs.false +; CHECK-GI-NEXT: lsl w8, w1, #1 +; CHECK-GI-NEXT: cmp w1, #7 +; CHECK-GI-NEXT: csinc w8, w8, w1, lt +; CHECK-GI-NEXT: cmp w8, #17 +; CHECK-GI-NEXT: b.ge LBB2_3 +; CHECK-GI-NEXT: LBB2_2: ; %if.then +; CHECK-GI-NEXT: stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill +; CHECK-GI-NEXT: bl _foo +; CHECK-GI-NEXT: ldp x29, x30, [sp], #16 ; 16-byte Folded Reload +; CHECK-GI-NEXT: LBB2_3: ; %if.end +; CHECK-GI-NEXT: mov w0, #7 ; =0x7 +; CHECK-GI-NEXT: ret entry: %cmp = icmp eq i32 %a, 5 br i1 %cmp, label %if.then, label %lor.lhs.false @@ -171,37 +171,37 @@ if.end: ; preds = %if.then, %lor.lhs.f ; The sdiv/udiv instructions do not trap when the divisor is zero, so they are ; safe to speculate. define i32 @speculate_division(i32 %a, i32 %b) nounwind ssp { -; SDISEL-LABEL: speculate_division: -; SDISEL: ; %bb.0: ; %entry -; SDISEL-NEXT: cmp w0, #1 -; SDISEL-NEXT: sdiv w8, w1, w0 -; SDISEL-NEXT: ccmp w8, #16, #0, ge -; SDISEL-NEXT: b.le LBB4_2 -; SDISEL-NEXT: ; %bb.1: ; %if.end -; SDISEL-NEXT: mov w0, #7 ; =0x7 -; SDISEL-NEXT: ret -; SDISEL-NEXT: LBB4_2: ; %if.then -; SDISEL-NEXT: stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill -; SDISEL-NEXT: bl _foo -; SDISEL-NEXT: ldp x29, x30, [sp], #16 ; 16-byte Folded Reload -; SDISEL-NEXT: mov w0, #7 ; =0x7 -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: speculate_division: +; CHECK-SD: ; %bb.0: ; %entry +; CHECK-SD-NEXT: cmp w0, #1 +; CHECK-SD-NEXT: sdiv w8, w1, w0 +; CHECK-SD-NEXT: ccmp w8, #16, #0, ge +; CHECK-SD-NEXT: b.le LBB4_2 +; CHECK-SD-NEXT: ; %bb.1: ; %if.end +; CHECK-SD-NEXT: mov w0, #7 ; =0x7 +; CHECK-SD-NEXT: ret +; CHECK-SD-NEXT: LBB4_2: ; %if.then +; CHECK-SD-NEXT: stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill +; CHECK-SD-NEXT: bl _foo +; CHECK-SD-NEXT: ldp x29, x30, [sp], #16 ; 16-byte Folded Reload +; CHECK-SD-NEXT: mov w0, #7 ; =0x7 +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: speculate_division: -; GISEL: ; %bb.0: ; %entry -; GISEL-NEXT: cmp w0, #0 -; GISEL-NEXT: sdiv w8, w1, w0 -; GISEL-NEXT: ccmp w8, #17, #0, gt -; GISEL-NEXT: b.lt LBB4_2 -; GISEL-NEXT: ; %bb.1: ; %if.end -; GISEL-NEXT: mov w0, #7 ; =0x7 -; GISEL-NEXT: ret -; GISEL-NEXT: LBB4_2: ; %if.then -; GISEL-NEXT: stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill -; GISEL-NEXT: bl _foo -; GISEL-NEXT: ldp x29, x30, [sp], #16 ; 16-byte Folded Reload -; GISEL-NEXT: mov w0, #7 ; =0x7 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: speculate_division: +; CHECK-GI: ; %bb.0: ; %entry +; CHECK-GI-NEXT: cmp w0, #0 +; CHECK-GI-NEXT: sdiv w8, w1, w0 +; CHECK-GI-NEXT: ccmp w8, #17, #0, gt +; CHECK-GI-NEXT: b.lt LBB4_2 +; CHECK-GI-NEXT: ; %bb.1: ; %if.end +; CHECK-GI-NEXT: mov w0, #7 ; =0x7 +; CHECK-GI-NEXT: ret +; CHECK-GI-NEXT: LBB4_2: ; %if.then +; CHECK-GI-NEXT: stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill +; CHECK-GI-NEXT: bl _foo +; CHECK-GI-NEXT: ldp x29, x30, [sp], #16 ; 16-byte Folded Reload +; CHECK-GI-NEXT: mov w0, #7 ; =0x7 +; CHECK-GI-NEXT: ret entry: %cmp = icmp sgt i32 %a, 0 br i1 %cmp, label %land.lhs.true, label %if.end @@ -221,41 +221,41 @@ if.end: ; Floating point compare. define i32 @single_fcmp(i32 %a, float %b) nounwind ssp { -; SDISEL-LABEL: single_fcmp: -; SDISEL: ; %bb.0: ; %entry -; SDISEL-NEXT: cmp w0, #1 -; SDISEL-NEXT: scvtf s1, w0 -; SDISEL-NEXT: fdiv s0, s0, s1 -; SDISEL-NEXT: fmov s1, #17.00000000 -; SDISEL-NEXT: fccmp s0, s1, #8, ge -; SDISEL-NEXT: b.ge LBB5_2 -; SDISEL-NEXT: ; %bb.1: ; %if.end -; SDISEL-NEXT: mov w0, #7 ; =0x7 -; SDISEL-NEXT: ret -; SDISEL-NEXT: LBB5_2: ; %if.then -; SDISEL-NEXT: stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill -; SDISEL-NEXT: bl _foo -; SDISEL-NEXT: ldp x29, x30, [sp], #16 ; 16-byte Folded Reload -; SDISEL-NEXT: mov w0, #7 ; =0x7 -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: single_fcmp: +; CHECK-SD: ; %bb.0: ; %entry +; CHECK-SD-NEXT: cmp w0, #1 +; CHECK-SD-NEXT: scvtf s1, w0 +; CHECK-SD-NEXT: fdiv s0, s0, s1 +; CHECK-SD-NEXT: fmov s1, #17.00000000 +; CHECK-SD-NEXT: fccmp s0, s1, #8, ge +; CHECK-SD-NEXT: b.ge LBB5_2 +; CHECK-SD-NEXT: ; %bb.1: ; %if.end +; CHECK-SD-NEXT: mov w0, #7 ; =0x7 +; CHECK-SD-NEXT: ret +; CHECK-SD-NEXT: LBB5_2: ; %if.then +; CHECK-SD-NEXT: stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill +; CHECK-SD-NEXT: bl _foo +; CHECK-SD-NEXT: ldp x29, x30, [sp], #16 ; 16-byte Folded Reload +; CHECK-SD-NEXT: mov w0, #7 ; =0x7 +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: single_fcmp: -; GISEL: ; %bb.0: ; %entry -; GISEL-NEXT: cmp w0, #0 -; GISEL-NEXT: scvtf s1, w0 -; GISEL-NEXT: fdiv s0, s0, s1 -; GISEL-NEXT: fmov s1, #17.00000000 -; GISEL-NEXT: fccmp s0, s1, #8, gt -; GISEL-NEXT: b.ge LBB5_2 -; GISEL-NEXT: ; %bb.1: ; %if.end -; GISEL-NEXT: mov w0, #7 ; =0x7 -; GISEL-NEXT: ret -; GISEL-NEXT: LBB5_2: ; %if.then -; GISEL-NEXT: stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill -; GISEL-NEXT: bl _foo -; GISEL-NEXT: ldp x29, x30, [sp], #16 ; 16-byte Folded Reload -; GISEL-NEXT: mov w0, #7 ; =0x7 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: single_fcmp: +; CHECK-GI: ; %bb.0: ; %entry +; CHECK-GI-NEXT: cmp w0, #0 +; CHECK-GI-NEXT: scvtf s1, w0 +; CHECK-GI-NEXT: fdiv s0, s0, s1 +; CHECK-GI-NEXT: fmov s1, #17.00000000 +; CHECK-GI-NEXT: fccmp s0, s1, #8, gt +; CHECK-GI-NEXT: b.ge LBB5_2 +; CHECK-GI-NEXT: ; %bb.1: ; %if.end +; CHECK-GI-NEXT: mov w0, #7 ; =0x7 +; CHECK-GI-NEXT: ret +; CHECK-GI-NEXT: LBB5_2: ; %if.then +; CHECK-GI-NEXT: stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill +; CHECK-GI-NEXT: bl _foo +; CHECK-GI-NEXT: ldp x29, x30, [sp], #16 ; 16-byte Folded Reload +; CHECK-GI-NEXT: mov w0, #7 ; =0x7 +; CHECK-GI-NEXT: ret entry: %cmp = icmp sgt i32 %a, 0 br i1 %cmp, label %land.lhs.true, label %if.end @@ -499,28 +499,28 @@ define float @select_or_float(i32 %w0, i32 %w1, float %x2, float %x3) { } define i64 @gccbug(i64 %x0, i64 %x1) { -; SDISEL-LABEL: gccbug: -; SDISEL: ; %bb.0: -; SDISEL-NEXT: cmp x0, #2 -; SDISEL-NEXT: ccmp x0, #4, #4, ne -; SDISEL-NEXT: ccmp x1, #0, #0, eq -; SDISEL-NEXT: mov w8, #1 ; =0x1 -; SDISEL-NEXT: cinc x0, x8, eq -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: gccbug: +; CHECK-SD: ; %bb.0: +; CHECK-SD-NEXT: cmp x0, #2 +; CHECK-SD-NEXT: ccmp x0, #4, #4, ne +; CHECK-SD-NEXT: ccmp x1, #0, #0, eq +; CHECK-SD-NEXT: mov w8, #1 ; =0x1 +; CHECK-SD-NEXT: cinc x0, x8, eq +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: gccbug: -; GISEL: ; %bb.0: -; GISEL-NEXT: cmp x1, #0 -; GISEL-NEXT: cset w8, eq -; GISEL-NEXT: cmp x0, #2 -; GISEL-NEXT: cset w9, eq -; GISEL-NEXT: cmp x0, #4 -; GISEL-NEXT: cset w10, eq -; GISEL-NEXT: orr w9, w10, w9 -; GISEL-NEXT: and w8, w9, w8 -; GISEL-NEXT: and x8, x8, #0x1 -; GISEL-NEXT: add x0, x8, #1 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: gccbug: +; CHECK-GI: ; %bb.0: +; CHECK-GI-NEXT: cmp x1, #0 +; CHECK-GI-NEXT: cset w8, eq +; CHECK-GI-NEXT: cmp x0, #2 +; CHECK-GI-NEXT: cset w9, eq +; CHECK-GI-NEXT: cmp x0, #4 +; CHECK-GI-NEXT: cset w10, eq +; CHECK-GI-NEXT: orr w9, w10, w9 +; CHECK-GI-NEXT: and w8, w9, w8 +; CHECK-GI-NEXT: and x8, x8, #0x1 +; CHECK-GI-NEXT: add x0, x8, #1 +; CHECK-GI-NEXT: ret %cmp0 = icmp eq i64 %x1, 0 %cmp1 = icmp eq i64 %x0, 2 %cmp2 = icmp eq i64 %x0, 4 @@ -570,23 +570,23 @@ define i32 @select_andor(i32 %v1, i32 %v2, i32 %v3) { } define i32 @select_andor32(i32 %v1, i32 %v2, i32 %v3) { -; SDISEL-LABEL: select_andor32: -; SDISEL: ; %bb.0: -; SDISEL-NEXT: cmp w1, w2 -; SDISEL-NEXT: mov w8, #32 ; =0x20 -; SDISEL-NEXT: ccmp w0, w8, #4, lt -; SDISEL-NEXT: ccmp w0, w1, #0, eq -; SDISEL-NEXT: csel w0, w0, w1, eq -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: select_andor32: +; CHECK-SD: ; %bb.0: +; CHECK-SD-NEXT: cmp w1, w2 +; CHECK-SD-NEXT: mov w8, #32 ; =0x20 +; CHECK-SD-NEXT: ccmp w0, w8, #4, lt +; CHECK-SD-NEXT: ccmp w0, w1, #0, eq +; CHECK-SD-NEXT: csel w0, w0, w1, eq +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: select_andor32: -; GISEL: ; %bb.0: -; GISEL-NEXT: mov w8, #32 ; =0x20 -; GISEL-NEXT: cmp w1, w2 -; GISEL-NEXT: ccmp w0, w8, #4, lt -; GISEL-NEXT: ccmp w0, w1, #0, eq -; GISEL-NEXT: csel w0, w0, w1, eq -; GISEL-NEXT: ret +; CHECK-GI-LABEL: select_andor32: +; CHECK-GI: ; %bb.0: +; CHECK-GI-NEXT: mov w8, #32 ; =0x20 +; CHECK-GI-NEXT: cmp w1, w2 +; CHECK-GI-NEXT: ccmp w0, w8, #4, lt +; CHECK-GI-NEXT: ccmp w0, w1, #0, eq +; CHECK-GI-NEXT: csel w0, w0, w1, eq +; CHECK-GI-NEXT: ret %c0 = icmp eq i32 %v1, %v2 %c1 = icmp sge i32 %v2, %v3 %c2 = icmp eq i32 %v1, 32 @@ -597,22 +597,22 @@ define i32 @select_andor32(i32 %v1, i32 %v2, i32 %v3) { } define i64 @select_noccmp1(i64 %v1, i64 %v2, i64 %v3, i64 %r) { -; SDISEL-LABEL: select_noccmp1: -; SDISEL: ; %bb.0: -; SDISEL-NEXT: cmp x0, #0 -; SDISEL-NEXT: ccmp x0, #13, #4, lt -; SDISEL-NEXT: cset w8, gt -; SDISEL-NEXT: cmp x2, #2 -; SDISEL-NEXT: ccmp x2, #4, #4, lt -; SDISEL-NEXT: csinc w8, w8, wzr, le -; SDISEL-NEXT: cmp w8, #0 -; SDISEL-NEXT: csel x0, xzr, x3, ne -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: select_noccmp1: +; CHECK-SD: ; %bb.0: +; CHECK-SD-NEXT: cmp x0, #0 +; CHECK-SD-NEXT: ccmp x0, #13, #4, lt +; CHECK-SD-NEXT: cset w8, gt +; CHECK-SD-NEXT: cmp x2, #2 +; CHECK-SD-NEXT: ccmp x2, #4, #4, lt +; CHECK-SD-NEXT: csinc w8, w8, wzr, le +; CHECK-SD-NEXT: cmp w8, #0 +; CHECK-SD-NEXT: csel x0, xzr, x3, ne +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: select_noccmp1: -; GISEL: ; %bb.0: -; GISEL-NEXT: mov x0, x3 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: select_noccmp1: +; CHECK-GI: ; %bb.0: +; CHECK-GI-NEXT: mov x0, x3 +; CHECK-GI-NEXT: ret %c0 = icmp slt i64 %v1, 0 %c1 = icmp sgt i64 %v1, 13 %c2 = icmp slt i64 %v3, 2 @@ -627,28 +627,28 @@ define i64 @select_noccmp1(i64 %v1, i64 %v2, i64 %v3, i64 %r) { @g = global i32 0 define i64 @select_noccmp2(i64 %v1, i64 %v2, i64 %v3, i64 %r) { -; SDISEL-LABEL: select_noccmp2: -; SDISEL: ; %bb.0: -; SDISEL-NEXT: cmp x0, #0 -; SDISEL-NEXT: ccmp x0, #13, #0, ge -; SDISEL-NEXT: cset w8, gt -; SDISEL-NEXT: cmp w8, #0 -; SDISEL-NEXT: csel x0, xzr, x3, ne -; SDISEL-NEXT: sbfx w8, w8, #0, #1 -; SDISEL-NEXT: adrp x9, _g@PAGE -; SDISEL-NEXT: str w8, [x9, _g@PAGEOFF] -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: select_noccmp2: +; CHECK-SD: ; %bb.0: +; CHECK-SD-NEXT: cmp x0, #0 +; CHECK-SD-NEXT: ccmp x0, #13, #0, ge +; CHECK-SD-NEXT: cset w8, gt +; CHECK-SD-NEXT: cmp w8, #0 +; CHECK-SD-NEXT: csel x0, xzr, x3, ne +; CHECK-SD-NEXT: sbfx w8, w8, #0, #1 +; CHECK-SD-NEXT: adrp x9, _g@PAGE +; CHECK-SD-NEXT: str w8, [x9, _g@PAGEOFF] +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: select_noccmp2: -; GISEL: ; %bb.0: -; GISEL-NEXT: cmp x0, #14 -; GISEL-NEXT: cset w8, hs -; GISEL-NEXT: tst w8, #0x1 -; GISEL-NEXT: csel x0, xzr, x3, ne -; GISEL-NEXT: sbfx w8, w8, #0, #1 -; GISEL-NEXT: adrp x9, _g@PAGE -; GISEL-NEXT: str w8, [x9, _g@PAGEOFF] -; GISEL-NEXT: ret +; CHECK-GI-LABEL: select_noccmp2: +; CHECK-GI: ; %bb.0: +; CHECK-GI-NEXT: cmp x0, #14 +; CHECK-GI-NEXT: cset w8, hs +; CHECK-GI-NEXT: tst w8, #0x1 +; CHECK-GI-NEXT: csel x0, xzr, x3, ne +; CHECK-GI-NEXT: sbfx w8, w8, #0, #1 +; CHECK-GI-NEXT: adrp x9, _g@PAGE +; CHECK-GI-NEXT: str w8, [x9, _g@PAGEOFF] +; CHECK-GI-NEXT: ret %c0 = icmp slt i64 %v1, 0 %c1 = icmp sgt i64 %v1, 13 %or = or i1 %c0, %c1 @@ -661,33 +661,33 @@ define i64 @select_noccmp2(i64 %v1, i64 %v2, i64 %v3, i64 %r) { ; The following is not possible to implement with a single cmp;ccmp;csel ; sequence. define i32 @select_noccmp3(i32 %v0, i32 %v1, i32 %v2) { -; SDISEL-LABEL: select_noccmp3: -; SDISEL: ; %bb.0: -; SDISEL-NEXT: cmp w0, #0 -; SDISEL-NEXT: ccmp w0, #13, #0, ge -; SDISEL-NEXT: cset w8, gt -; SDISEL-NEXT: cmp w0, #22 -; SDISEL-NEXT: mov w9, #44 ; =0x2c -; SDISEL-NEXT: ccmp w0, w9, #0, ge -; SDISEL-NEXT: csel w8, wzr, w8, le -; SDISEL-NEXT: cmp w0, #99 -; SDISEL-NEXT: mov w9, #77 ; =0x4d -; SDISEL-NEXT: ccmp w0, w9, #4, ne -; SDISEL-NEXT: cset w9, eq -; SDISEL-NEXT: tst w8, w9 -; SDISEL-NEXT: csel w0, w1, w2, ne -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: select_noccmp3: +; CHECK-SD: ; %bb.0: +; CHECK-SD-NEXT: cmp w0, #0 +; CHECK-SD-NEXT: ccmp w0, #13, #0, ge +; CHECK-SD-NEXT: cset w8, gt +; CHECK-SD-NEXT: cmp w0, #22 +; CHECK-SD-NEXT: mov w9, #44 ; =0x2c +; CHECK-SD-NEXT: ccmp w0, w9, #0, ge +; CHECK-SD-NEXT: csel w8, wzr, w8, le +; CHECK-SD-NEXT: cmp w0, #99 +; CHECK-SD-NEXT: mov w9, #77 ; =0x4d +; CHECK-SD-NEXT: ccmp w0, w9, #4, ne +; CHECK-SD-NEXT: cset w9, eq +; CHECK-SD-NEXT: tst w8, w9 +; CHECK-SD-NEXT: csel w0, w1, w2, ne +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: select_noccmp3: -; GISEL: ; %bb.0: -; GISEL-NEXT: mov w8, #99 ; =0x63 -; GISEL-NEXT: sub w9, w0, #45 -; GISEL-NEXT: cmp w0, #77 -; GISEL-NEXT: ccmp w0, w8, #4, ne -; GISEL-NEXT: ccmn w9, #23, #2, eq -; GISEL-NEXT: ccmp w0, #14, #0, lo -; GISEL-NEXT: csel w0, w1, w2, hs -; GISEL-NEXT: ret +; CHECK-GI-LABEL: select_noccmp3: +; CHECK-GI: ; %bb.0: +; CHECK-GI-NEXT: mov w8, #99 ; =0x63 +; CHECK-GI-NEXT: sub w9, w0, #45 +; CHECK-GI-NEXT: cmp w0, #77 +; CHECK-GI-NEXT: ccmp w0, w8, #4, ne +; CHECK-GI-NEXT: ccmn w9, #23, #2, eq +; CHECK-GI-NEXT: ccmp w0, #14, #0, lo +; CHECK-GI-NEXT: csel w0, w1, w2, hs +; CHECK-GI-NEXT: ret %c0 = icmp slt i32 %v0, 0 %c1 = icmp sgt i32 %v0, 13 %c2 = icmp slt i32 %v0, 22 @@ -864,27 +864,27 @@ define i32 @select_or_olt_ueq_ogt(double %v0, double %v1, double %v2, double %v3 ; Verify that we correctly promote f16. define i32 @half_select_and_olt_oge(half %v0, half %v1, half %v2, half %v3, i32 %a, i32 %b) #0 { -; SDISEL-LABEL: half_select_and_olt_oge: -; SDISEL: ; %bb.0: -; SDISEL-NEXT: fcvt s1, h1 -; SDISEL-NEXT: fcvt s0, h0 -; SDISEL-NEXT: fcmp s0, s1 -; SDISEL-NEXT: fcvt s0, h3 -; SDISEL-NEXT: fcvt s1, h2 -; SDISEL-NEXT: fccmp s1, s0, #8, mi -; SDISEL-NEXT: csel w0, w0, w1, ge -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: half_select_and_olt_oge: +; CHECK-SD: ; %bb.0: +; CHECK-SD-NEXT: fcvt s1, h1 +; CHECK-SD-NEXT: fcvt s0, h0 +; CHECK-SD-NEXT: fcmp s0, s1 +; CHECK-SD-NEXT: fcvt s0, h3 +; CHECK-SD-NEXT: fcvt s1, h2 +; CHECK-SD-NEXT: fccmp s1, s0, #8, mi +; CHECK-SD-NEXT: csel w0, w0, w1, ge +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: half_select_and_olt_oge: -; GISEL: ; %bb.0: -; GISEL-NEXT: fcvt s0, h0 -; GISEL-NEXT: fcvt s1, h1 -; GISEL-NEXT: fcvt s2, h2 -; GISEL-NEXT: fcvt s3, h3 -; GISEL-NEXT: fcmp s0, s1 -; GISEL-NEXT: fccmp s2, s3, #8, mi -; GISEL-NEXT: csel w0, w0, w1, ge -; GISEL-NEXT: ret +; CHECK-GI-LABEL: half_select_and_olt_oge: +; CHECK-GI: ; %bb.0: +; CHECK-GI-NEXT: fcvt s0, h0 +; CHECK-GI-NEXT: fcvt s1, h1 +; CHECK-GI-NEXT: fcvt s2, h2 +; CHECK-GI-NEXT: fcvt s3, h3 +; CHECK-GI-NEXT: fcmp s0, s1 +; CHECK-GI-NEXT: fccmp s2, s3, #8, mi +; CHECK-GI-NEXT: csel w0, w0, w1, ge +; CHECK-GI-NEXT: ret %c0 = fcmp olt half %v0, %v1 %c1 = fcmp oge half %v2, %v3 %cr = and i1 %c1, %c0 @@ -893,29 +893,29 @@ define i32 @half_select_and_olt_oge(half %v0, half %v1, half %v2, half %v3, i32 } define i32 @half_select_and_olt_one(half %v0, half %v1, half %v2, half %v3, i32 %a, i32 %b) #0 { -; SDISEL-LABEL: half_select_and_olt_one: -; SDISEL: ; %bb.0: -; SDISEL-NEXT: fcvt s1, h1 -; SDISEL-NEXT: fcvt s0, h0 -; SDISEL-NEXT: fcmp s0, s1 -; SDISEL-NEXT: fcvt s0, h3 -; SDISEL-NEXT: fcvt s1, h2 -; SDISEL-NEXT: fccmp s1, s0, #4, mi -; SDISEL-NEXT: fccmp s1, s0, #1, ne -; SDISEL-NEXT: csel w0, w0, w1, vc -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: half_select_and_olt_one: +; CHECK-SD: ; %bb.0: +; CHECK-SD-NEXT: fcvt s1, h1 +; CHECK-SD-NEXT: fcvt s0, h0 +; CHECK-SD-NEXT: fcmp s0, s1 +; CHECK-SD-NEXT: fcvt s0, h3 +; CHECK-SD-NEXT: fcvt s1, h2 +; CHECK-SD-NEXT: fccmp s1, s0, #4, mi +; CHECK-SD-NEXT: fccmp s1, s0, #1, ne +; CHECK-SD-NEXT: csel w0, w0, w1, vc +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: half_select_and_olt_one: -; GISEL: ; %bb.0: -; GISEL-NEXT: fcvt s0, h0 -; GISEL-NEXT: fcvt s1, h1 -; GISEL-NEXT: fcvt s2, h2 -; GISEL-NEXT: fcvt s3, h3 -; GISEL-NEXT: fcmp s0, s1 -; GISEL-NEXT: fccmp s2, s3, #4, mi -; GISEL-NEXT: fccmp s2, s3, #1, ne -; GISEL-NEXT: csel w0, w0, w1, vc -; GISEL-NEXT: ret +; CHECK-GI-LABEL: half_select_and_olt_one: +; CHECK-GI: ; %bb.0: +; CHECK-GI-NEXT: fcvt s0, h0 +; CHECK-GI-NEXT: fcvt s1, h1 +; CHECK-GI-NEXT: fcvt s2, h2 +; CHECK-GI-NEXT: fcvt s3, h3 +; CHECK-GI-NEXT: fcmp s0, s1 +; CHECK-GI-NEXT: fccmp s2, s3, #4, mi +; CHECK-GI-NEXT: fccmp s2, s3, #1, ne +; CHECK-GI-NEXT: csel w0, w0, w1, vc +; CHECK-GI-NEXT: ret %c0 = fcmp olt half %v0, %v1 %c1 = fcmp one half %v2, %v3 %cr = and i1 %c1, %c0 @@ -926,51 +926,51 @@ define i32 @half_select_and_olt_one(half %v0, half %v1, half %v2, half %v3, i32 ; Also verify that we don't try to generate f128 FCCMPs, using RT calls instead. define i32 @f128_select_and_olt_oge(fp128 %v0, fp128 %v1, fp128 %v2, fp128 %v3, i32 %a, i32 %b) #0 { -; SDISEL-LABEL: f128_select_and_olt_oge: -; SDISEL: ; %bb.0: -; SDISEL-NEXT: sub sp, sp, #80 -; SDISEL-NEXT: stp x22, x21, [sp, #32] ; 16-byte Folded Spill -; SDISEL-NEXT: stp x20, x19, [sp, #48] ; 16-byte Folded Spill -; SDISEL-NEXT: stp x29, x30, [sp, #64] ; 16-byte Folded Spill -; SDISEL-NEXT: mov x19, x1 -; SDISEL-NEXT: mov x20, x0 -; SDISEL-NEXT: stp q2, q3, [sp] ; 32-byte Folded Spill -; SDISEL-NEXT: bl ___lttf2 -; SDISEL-NEXT: cmp w0, #0 -; SDISEL-NEXT: cset w21, lt -; SDISEL-NEXT: ldp q0, q1, [sp] ; 32-byte Folded Reload -; SDISEL-NEXT: bl ___getf2 -; SDISEL-NEXT: cmp w0, #0 -; SDISEL-NEXT: cset w8, ge -; SDISEL-NEXT: tst w8, w21 -; SDISEL-NEXT: csel w0, w20, w19, ne -; SDISEL-NEXT: ldp x29, x30, [sp, #64] ; 16-byte Folded Reload -; SDISEL-NEXT: ldp x20, x19, [sp, #48] ; 16-byte Folded Reload -; SDISEL-NEXT: ldp x22, x21, [sp, #32] ; 16-byte Folded Reload -; SDISEL-NEXT: add sp, sp, #80 -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: f128_select_and_olt_oge: +; CHECK-SD: ; %bb.0: +; CHECK-SD-NEXT: sub sp, sp, #80 +; CHECK-SD-NEXT: stp x22, x21, [sp, #32] ; 16-byte Folded Spill +; CHECK-SD-NEXT: stp x20, x19, [sp, #48] ; 16-byte Folded Spill +; CHECK-SD-NEXT: stp x29, x30, [sp, #64] ; 16-byte Folded Spill +; CHECK-SD-NEXT: mov x19, x1 +; CHECK-SD-NEXT: mov x20, x0 +; CHECK-SD-NEXT: stp q2, q3, [sp] ; 32-byte Folded Spill +; CHECK-SD-NEXT: bl ___lttf2 +; CHECK-SD-NEXT: cmp w0, #0 +; CHECK-SD-NEXT: cset w21, lt +; CHECK-SD-NEXT: ldp q0, q1, [sp] ; 32-byte Folded Reload +; CHECK-SD-NEXT: bl ___getf2 +; CHECK-SD-NEXT: cmp w0, #0 +; CHECK-SD-NEXT: cset w8, ge +; CHECK-SD-NEXT: tst w8, w21 +; CHECK-SD-NEXT: csel w0, w20, w19, ne +; CHECK-SD-NEXT: ldp x29, x30, [sp, #64] ; 16-byte Folded Reload +; CHECK-SD-NEXT: ldp x20, x19, [sp, #48] ; 16-byte Folded Reload +; CHECK-SD-NEXT: ldp x22, x21, [sp, #32] ; 16-byte Folded Reload +; CHECK-SD-NEXT: add sp, sp, #80 +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: f128_select_and_olt_oge: -; GISEL: ; %bb.0: -; GISEL-NEXT: sub sp, sp, #80 -; GISEL-NEXT: stp x22, x21, [sp, #32] ; 16-byte Folded Spill -; GISEL-NEXT: stp x20, x19, [sp, #48] ; 16-byte Folded Spill -; GISEL-NEXT: stp x29, x30, [sp, #64] ; 16-byte Folded Spill -; GISEL-NEXT: stp q3, q2, [sp] ; 32-byte Folded Spill -; GISEL-NEXT: mov x19, x0 -; GISEL-NEXT: mov x20, x1 -; GISEL-NEXT: bl ___lttf2 -; GISEL-NEXT: mov x21, x0 -; GISEL-NEXT: ldp q1, q0, [sp] ; 32-byte Folded Reload -; GISEL-NEXT: bl ___getf2 -; GISEL-NEXT: cmp w21, #0 -; GISEL-NEXT: ccmp w0, #0, #8, lt -; GISEL-NEXT: csel w0, w19, w20, ge -; GISEL-NEXT: ldp x29, x30, [sp, #64] ; 16-byte Folded Reload -; GISEL-NEXT: ldp x20, x19, [sp, #48] ; 16-byte Folded Reload -; GISEL-NEXT: ldp x22, x21, [sp, #32] ; 16-byte Folded Reload -; GISEL-NEXT: add sp, sp, #80 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: f128_select_and_olt_oge: +; CHECK-GI: ; %bb.0: +; CHECK-GI-NEXT: sub sp, sp, #80 +; CHECK-GI-NEXT: stp x22, x21, [sp, #32] ; 16-byte Folded Spill +; CHECK-GI-NEXT: stp x20, x19, [sp, #48] ; 16-byte Folded Spill +; CHECK-GI-NEXT: stp x29, x30, [sp, #64] ; 16-byte Folded Spill +; CHECK-GI-NEXT: stp q3, q2, [sp] ; 32-byte Folded Spill +; CHECK-GI-NEXT: mov x19, x0 +; CHECK-GI-NEXT: mov x20, x1 +; CHECK-GI-NEXT: bl ___lttf2 +; CHECK-GI-NEXT: mov x21, x0 +; CHECK-GI-NEXT: ldp q1, q0, [sp] ; 32-byte Folded Reload +; CHECK-GI-NEXT: bl ___getf2 +; CHECK-GI-NEXT: cmp w21, #0 +; CHECK-GI-NEXT: ccmp w0, #0, #8, lt +; CHECK-GI-NEXT: csel w0, w19, w20, ge +; CHECK-GI-NEXT: ldp x29, x30, [sp, #64] ; 16-byte Folded Reload +; CHECK-GI-NEXT: ldp x20, x19, [sp, #48] ; 16-byte Folded Reload +; CHECK-GI-NEXT: ldp x22, x21, [sp, #32] ; 16-byte Folded Reload +; CHECK-GI-NEXT: add sp, sp, #80 +; CHECK-GI-NEXT: ret %c0 = fcmp olt fp128 %v0, %v1 %c1 = fcmp oge fp128 %v2, %v3 %cr = and i1 %c1, %c0 @@ -1048,46 +1048,46 @@ define i32 @deep_or2(i32 %a0, i32 %a1, i32 %a2, i32 %a3, i32 %x, i32 %y) { ; This test is trying to test that multiple ccmp's don't get created in a way ; that they would have multiple uses. It doesn't seem to. define i32 @multiccmp(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %x, i32 %y) #0 { -; SDISEL-LABEL: multiccmp: -; SDISEL: ; %bb.0: ; %entry -; SDISEL-NEXT: stp x22, x21, [sp, #-48]! ; 16-byte Folded Spill -; SDISEL-NEXT: stp x20, x19, [sp, #16] ; 16-byte Folded Spill -; SDISEL-NEXT: stp x29, x30, [sp, #32] ; 16-byte Folded Spill -; SDISEL-NEXT: mov x19, x5 -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: cset w20, gt -; SDISEL-NEXT: cmp w2, w3 -; SDISEL-NEXT: cset w21, ne -; SDISEL-NEXT: tst w20, w21 -; SDISEL-NEXT: csel w0, w5, w4, ne -; SDISEL-NEXT: bl _callee -; SDISEL-NEXT: tst w20, w21 -; SDISEL-NEXT: csel w0, w0, w19, ne -; SDISEL-NEXT: bl _callee -; SDISEL-NEXT: ldp x29, x30, [sp, #32] ; 16-byte Folded Reload -; SDISEL-NEXT: ldp x20, x19, [sp, #16] ; 16-byte Folded Reload -; SDISEL-NEXT: ldp x22, x21, [sp], #48 ; 16-byte Folded Reload -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: multiccmp: +; CHECK-SD: ; %bb.0: ; %entry +; CHECK-SD-NEXT: stp x22, x21, [sp, #-48]! ; 16-byte Folded Spill +; CHECK-SD-NEXT: stp x20, x19, [sp, #16] ; 16-byte Folded Spill +; CHECK-SD-NEXT: stp x29, x30, [sp, #32] ; 16-byte Folded Spill +; CHECK-SD-NEXT: mov x19, x5 +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: cset w20, gt +; CHECK-SD-NEXT: cmp w2, w3 +; CHECK-SD-NEXT: cset w21, ne +; CHECK-SD-NEXT: tst w20, w21 +; CHECK-SD-NEXT: csel w0, w5, w4, ne +; CHECK-SD-NEXT: bl _callee +; CHECK-SD-NEXT: tst w20, w21 +; CHECK-SD-NEXT: csel w0, w0, w19, ne +; CHECK-SD-NEXT: bl _callee +; CHECK-SD-NEXT: ldp x29, x30, [sp, #32] ; 16-byte Folded Reload +; CHECK-SD-NEXT: ldp x20, x19, [sp, #16] ; 16-byte Folded Reload +; CHECK-SD-NEXT: ldp x22, x21, [sp], #48 ; 16-byte Folded Reload +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: multiccmp: -; GISEL: ; %bb.0: ; %entry -; GISEL-NEXT: stp x20, x19, [sp, #-32]! ; 16-byte Folded Spill -; GISEL-NEXT: stp x29, x30, [sp, #16] ; 16-byte Folded Spill -; GISEL-NEXT: mov x19, x5 -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, gt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ne -; GISEL-NEXT: and w20, w8, w9 -; GISEL-NEXT: tst w20, #0x1 -; GISEL-NEXT: csel w0, w5, w4, ne -; GISEL-NEXT: bl _callee -; GISEL-NEXT: tst w20, #0x1 -; GISEL-NEXT: csel w0, w0, w19, ne -; GISEL-NEXT: bl _callee -; GISEL-NEXT: ldp x29, x30, [sp, #16] ; 16-byte Folded Reload -; GISEL-NEXT: ldp x20, x19, [sp], #32 ; 16-byte Folded Reload -; GISEL-NEXT: ret +; CHECK-GI-LABEL: multiccmp: +; CHECK-GI: ; %bb.0: ; %entry +; CHECK-GI-NEXT: stp x20, x19, [sp, #-32]! ; 16-byte Folded Spill +; CHECK-GI-NEXT: stp x29, x30, [sp, #16] ; 16-byte Folded Spill +; CHECK-GI-NEXT: mov x19, x5 +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, gt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ne +; CHECK-GI-NEXT: and w20, w8, w9 +; CHECK-GI-NEXT: tst w20, #0x1 +; CHECK-GI-NEXT: csel w0, w5, w4, ne +; CHECK-GI-NEXT: bl _callee +; CHECK-GI-NEXT: tst w20, #0x1 +; CHECK-GI-NEXT: csel w0, w0, w19, ne +; CHECK-GI-NEXT: bl _callee +; CHECK-GI-NEXT: ldp x29, x30, [sp, #16] ; 16-byte Folded Reload +; CHECK-GI-NEXT: ldp x20, x19, [sp], #32 ; 16-byte Folded Reload +; CHECK-GI-NEXT: ret entry: %c0 = icmp sgt i32 %s0, %s1 %c1 = icmp ne i32 %s2, %s3 @@ -1100,57 +1100,57 @@ entry: } define i32 @multiccmp2(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %x, i32 %y) #0 { -; SDISEL-LABEL: multiccmp2: -; SDISEL: ; %bb.0: ; %entry -; SDISEL-NEXT: stp x22, x21, [sp, #-48]! ; 16-byte Folded Spill -; SDISEL-NEXT: stp x20, x19, [sp, #16] ; 16-byte Folded Spill -; SDISEL-NEXT: stp x29, x30, [sp, #32] ; 16-byte Folded Spill -; SDISEL-NEXT: mov x19, x5 -; SDISEL-NEXT: mov x20, x3 -; SDISEL-NEXT: mov x21, x0 -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: cset w8, gt -; SDISEL-NEXT: cmp w2, w3 -; SDISEL-NEXT: cset w22, ne -; SDISEL-NEXT: tst w8, w22 -; SDISEL-NEXT: csel w0, w5, w4, ne -; SDISEL-NEXT: bl _callee -; SDISEL-NEXT: cmp w21, w20 -; SDISEL-NEXT: cset w8, eq -; SDISEL-NEXT: tst w22, w8 -; SDISEL-NEXT: csel w0, w0, w19, ne -; SDISEL-NEXT: bl _callee -; SDISEL-NEXT: ldp x29, x30, [sp, #32] ; 16-byte Folded Reload -; SDISEL-NEXT: ldp x20, x19, [sp, #16] ; 16-byte Folded Reload -; SDISEL-NEXT: ldp x22, x21, [sp], #48 ; 16-byte Folded Reload -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: multiccmp2: +; CHECK-SD: ; %bb.0: ; %entry +; CHECK-SD-NEXT: stp x22, x21, [sp, #-48]! ; 16-byte Folded Spill +; CHECK-SD-NEXT: stp x20, x19, [sp, #16] ; 16-byte Folded Spill +; CHECK-SD-NEXT: stp x29, x30, [sp, #32] ; 16-byte Folded Spill +; CHECK-SD-NEXT: mov x19, x5 +; CHECK-SD-NEXT: mov x20, x3 +; CHECK-SD-NEXT: mov x21, x0 +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: cset w8, gt +; CHECK-SD-NEXT: cmp w2, w3 +; CHECK-SD-NEXT: cset w22, ne +; CHECK-SD-NEXT: tst w8, w22 +; CHECK-SD-NEXT: csel w0, w5, w4, ne +; CHECK-SD-NEXT: bl _callee +; CHECK-SD-NEXT: cmp w21, w20 +; CHECK-SD-NEXT: cset w8, eq +; CHECK-SD-NEXT: tst w22, w8 +; CHECK-SD-NEXT: csel w0, w0, w19, ne +; CHECK-SD-NEXT: bl _callee +; CHECK-SD-NEXT: ldp x29, x30, [sp, #32] ; 16-byte Folded Reload +; CHECK-SD-NEXT: ldp x20, x19, [sp, #16] ; 16-byte Folded Reload +; CHECK-SD-NEXT: ldp x22, x21, [sp], #48 ; 16-byte Folded Reload +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: multiccmp2: -; GISEL: ; %bb.0: ; %entry -; GISEL-NEXT: stp x22, x21, [sp, #-48]! ; 16-byte Folded Spill -; GISEL-NEXT: stp x20, x19, [sp, #16] ; 16-byte Folded Spill -; GISEL-NEXT: stp x29, x30, [sp, #32] ; 16-byte Folded Spill -; GISEL-NEXT: mov x19, x0 -; GISEL-NEXT: mov x20, x3 -; GISEL-NEXT: mov x21, x5 -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, gt -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w22, ne -; GISEL-NEXT: and w8, w8, w22 -; GISEL-NEXT: tst w8, #0x1 -; GISEL-NEXT: csel w0, w5, w4, ne -; GISEL-NEXT: bl _callee -; GISEL-NEXT: cmp w19, w20 -; GISEL-NEXT: cset w8, eq -; GISEL-NEXT: and w8, w22, w8 -; GISEL-NEXT: tst w8, #0x1 -; GISEL-NEXT: csel w0, w0, w21, ne -; GISEL-NEXT: bl _callee -; GISEL-NEXT: ldp x29, x30, [sp, #32] ; 16-byte Folded Reload -; GISEL-NEXT: ldp x20, x19, [sp, #16] ; 16-byte Folded Reload -; GISEL-NEXT: ldp x22, x21, [sp], #48 ; 16-byte Folded Reload -; GISEL-NEXT: ret +; CHECK-GI-LABEL: multiccmp2: +; CHECK-GI: ; %bb.0: ; %entry +; CHECK-GI-NEXT: stp x22, x21, [sp, #-48]! ; 16-byte Folded Spill +; CHECK-GI-NEXT: stp x20, x19, [sp, #16] ; 16-byte Folded Spill +; CHECK-GI-NEXT: stp x29, x30, [sp, #32] ; 16-byte Folded Spill +; CHECK-GI-NEXT: mov x19, x0 +; CHECK-GI-NEXT: mov x20, x3 +; CHECK-GI-NEXT: mov x21, x5 +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, gt +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w22, ne +; CHECK-GI-NEXT: and w8, w8, w22 +; CHECK-GI-NEXT: tst w8, #0x1 +; CHECK-GI-NEXT: csel w0, w5, w4, ne +; CHECK-GI-NEXT: bl _callee +; CHECK-GI-NEXT: cmp w19, w20 +; CHECK-GI-NEXT: cset w8, eq +; CHECK-GI-NEXT: and w8, w22, w8 +; CHECK-GI-NEXT: tst w8, #0x1 +; CHECK-GI-NEXT: csel w0, w0, w21, ne +; CHECK-GI-NEXT: bl _callee +; CHECK-GI-NEXT: ldp x29, x30, [sp, #32] ; 16-byte Folded Reload +; CHECK-GI-NEXT: ldp x20, x19, [sp, #16] ; 16-byte Folded Reload +; CHECK-GI-NEXT: ldp x22, x21, [sp], #48 ; 16-byte Folded Reload +; CHECK-GI-NEXT: ret entry: %c0 = icmp sgt i32 %s0, %s1 %c1 = icmp ne i32 %s2, %s3 @@ -1168,21 +1168,21 @@ entry: declare i32 @callee(i32) define i1 @cmp_and_negative_const(i32 %0, i32 %1) { -; SDISEL-LABEL: cmp_and_negative_const: -; SDISEL: ; %bb.0: -; SDISEL-NEXT: cmn w0, #1 -; SDISEL-NEXT: ccmn w1, #2, #0, eq -; SDISEL-NEXT: cset w0, eq -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: cmp_and_negative_const: +; CHECK-SD: ; %bb.0: +; CHECK-SD-NEXT: cmn w0, #1 +; CHECK-SD-NEXT: ccmn w1, #2, #0, eq +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: cmp_and_negative_const: -; GISEL: ; %bb.0: -; GISEL-NEXT: cmn w0, #1 -; GISEL-NEXT: cset w8, eq -; GISEL-NEXT: cmn w1, #2 -; GISEL-NEXT: cset w9, eq -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: cmp_and_negative_const: +; CHECK-GI: ; %bb.0: +; CHECK-GI-NEXT: cmn w0, #1 +; CHECK-GI-NEXT: cset w8, eq +; CHECK-GI-NEXT: cmn w1, #2 +; CHECK-GI-NEXT: cset w9, eq +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret %3 = icmp eq i32 %0, -1 %4 = icmp eq i32 %1, -2 %5 = and i1 %3, %4 @@ -1190,21 +1190,21 @@ define i1 @cmp_and_negative_const(i32 %0, i32 %1) { } define i1 @cmp_or_negative_const(i32 %a, i32 %b) { -; SDISEL-LABEL: cmp_or_negative_const: -; SDISEL: ; %bb.0: -; SDISEL-NEXT: cmn w0, #1 -; SDISEL-NEXT: ccmn w1, #2, #4, ne -; SDISEL-NEXT: cset w0, eq -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: cmp_or_negative_const: +; CHECK-SD: ; %bb.0: +; CHECK-SD-NEXT: cmn w0, #1 +; CHECK-SD-NEXT: ccmn w1, #2, #4, ne +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: cmp_or_negative_const: -; GISEL: ; %bb.0: -; GISEL-NEXT: cmn w0, #1 -; GISEL-NEXT: cset w8, eq -; GISEL-NEXT: cmn w1, #2 -; GISEL-NEXT: cset w9, eq -; GISEL-NEXT: orr w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: cmp_or_negative_const: +; CHECK-GI: ; %bb.0: +; CHECK-GI-NEXT: cmn w0, #1 +; CHECK-GI-NEXT: cset w8, eq +; CHECK-GI-NEXT: cmn w1, #2 +; CHECK-GI-NEXT: cset w9, eq +; CHECK-GI-NEXT: orr w0, w8, w9 +; CHECK-GI-NEXT: ret %cmp = icmp eq i32 %a, -1 %cmp1 = icmp eq i32 %b, -2 %or.cond = or i1 %cmp, %cmp1 diff --git a/llvm/test/CodeGen/AArch64/cmp-chains.ll b/llvm/test/CodeGen/AArch64/cmp-chains.ll index 4b816df..3620444 100644 --- a/llvm/test/CodeGen/AArch64/cmp-chains.ll +++ b/llvm/test/CodeGen/AArch64/cmp-chains.ll @@ -1,26 +1,26 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc < %s -mtriple=aarch64-- | FileCheck %s --check-prefixes=CHECK,SDISEL -; RUN: llc < %s -mtriple=aarch64-- -global-isel | FileCheck %s --check-prefixes=CHECK,GISEL +; RUN: llc < %s -mtriple=aarch64-- | FileCheck %s --check-prefixes=CHECK,CHECK-SD +; RUN: llc < %s -mtriple=aarch64-- -global-isel | FileCheck %s --check-prefixes=CHECK,CHECK-GI ; Ensure chains of comparisons produce chains of `ccmp` ; (x0 < x1) && (x2 > x3) define i32 @cmp_and2(i32 %0, i32 %1, i32 %2, i32 %3) { -; SDISEL-LABEL: cmp_and2: -; SDISEL: // %bb.0: -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, lo -; SDISEL-NEXT: cset w0, hi -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: cmp_and2: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, lo +; CHECK-SD-NEXT: cset w0, hi +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: cmp_and2: -; GISEL: // %bb.0: -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lo -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hi -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: cmp_and2: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lo +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hi +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret %5 = icmp ult i32 %0, %1 %6 = icmp ugt i32 %2, %3 %7 = select i1 %5, i1 %6, i1 false @@ -30,25 +30,25 @@ define i32 @cmp_and2(i32 %0, i32 %1, i32 %2, i32 %3) { ; (x0 < x1) && (x2 > x3) && (x4 != x5) define i32 @cmp_and3(i32 %0, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5) { -; SDISEL-LABEL: cmp_and3: -; SDISEL: // %bb.0: -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, lo -; SDISEL-NEXT: ccmp w4, w5, #4, hi -; SDISEL-NEXT: cset w0, ne -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: cmp_and3: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, lo +; CHECK-SD-NEXT: ccmp w4, w5, #4, hi +; CHECK-SD-NEXT: cset w0, ne +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: cmp_and3: -; GISEL: // %bb.0: -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lo -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hi -; GISEL-NEXT: cmp w4, w5 -; GISEL-NEXT: and w8, w8, w9 -; GISEL-NEXT: cset w9, ne -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: cmp_and3: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lo +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hi +; CHECK-GI-NEXT: cmp w4, w5 +; CHECK-GI-NEXT: and w8, w8, w9 +; CHECK-GI-NEXT: cset w9, ne +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret %7 = icmp ult i32 %0, %1 %8 = icmp ugt i32 %2, %3 %9 = select i1 %7, i1 %8, i1 false @@ -60,29 +60,29 @@ define i32 @cmp_and3(i32 %0, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5) { ; (x0 < x1) && (x2 > x3) && (x4 != x5) && (x6 == x7) define i32 @cmp_and4(i32 %0, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5, i32 %6, i32 %7) { -; SDISEL-LABEL: cmp_and4: -; SDISEL: // %bb.0: -; SDISEL-NEXT: cmp w2, w3 -; SDISEL-NEXT: ccmp w0, w1, #2, hi -; SDISEL-NEXT: ccmp w4, w5, #4, lo -; SDISEL-NEXT: ccmp w6, w7, #0, ne -; SDISEL-NEXT: cset w0, eq -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: cmp_and4: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: cmp w2, w3 +; CHECK-SD-NEXT: ccmp w0, w1, #2, hi +; CHECK-SD-NEXT: ccmp w4, w5, #4, lo +; CHECK-SD-NEXT: ccmp w6, w7, #0, ne +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: cmp_and4: -; GISEL: // %bb.0: -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w8, hi -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w9, lo -; GISEL-NEXT: cmp w4, w5 -; GISEL-NEXT: cset w10, ne -; GISEL-NEXT: cmp w6, w7 -; GISEL-NEXT: and w8, w8, w9 -; GISEL-NEXT: cset w11, eq -; GISEL-NEXT: and w9, w10, w11 -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: cmp_and4: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w8, hi +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w9, lo +; CHECK-GI-NEXT: cmp w4, w5 +; CHECK-GI-NEXT: cset w10, ne +; CHECK-GI-NEXT: cmp w6, w7 +; CHECK-GI-NEXT: and w8, w8, w9 +; CHECK-GI-NEXT: cset w11, eq +; CHECK-GI-NEXT: and w9, w10, w11 +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret %9 = icmp ugt i32 %2, %3 %10 = icmp ult i32 %0, %1 %11 = select i1 %9, i1 %10, i1 false @@ -96,22 +96,22 @@ define i32 @cmp_and4(i32 %0, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5, i32 %6, i32 ; (x0 < x1) || (x2 > x3) define i32 @cmp_or2(i32 %0, i32 %1, i32 %2, i32 %3) { -; SDISEL-LABEL: cmp_or2: -; SDISEL: // %bb.0: -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #0, hs -; SDISEL-NEXT: cset w0, ne -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: cmp_or2: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #0, hs +; CHECK-SD-NEXT: cset w0, ne +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: cmp_or2: -; GISEL: // %bb.0: -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lo -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, ne -; GISEL-NEXT: orr w8, w8, w9 -; GISEL-NEXT: and w0, w8, #0x1 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: cmp_or2: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lo +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, ne +; CHECK-GI-NEXT: orr w8, w8, w9 +; CHECK-GI-NEXT: and w0, w8, #0x1 +; CHECK-GI-NEXT: ret %5 = icmp ult i32 %0, %1 %6 = icmp ne i32 %2, %3 %7 = select i1 %5, i1 true, i1 %6 @@ -121,26 +121,26 @@ define i32 @cmp_or2(i32 %0, i32 %1, i32 %2, i32 %3) { ; (x0 < x1) || (x2 > x3) || (x4 != x5) define i32 @cmp_or3(i32 %0, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5) { -; SDISEL-LABEL: cmp_or3: -; SDISEL: // %bb.0: -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, hs -; SDISEL-NEXT: ccmp w4, w5, #0, ls -; SDISEL-NEXT: cset w0, ne -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: cmp_or3: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, hs +; CHECK-SD-NEXT: ccmp w4, w5, #0, ls +; CHECK-SD-NEXT: cset w0, ne +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: cmp_or3: -; GISEL: // %bb.0: -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lo -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hi -; GISEL-NEXT: cmp w4, w5 -; GISEL-NEXT: orr w8, w8, w9 -; GISEL-NEXT: cset w9, ne -; GISEL-NEXT: orr w8, w8, w9 -; GISEL-NEXT: and w0, w8, #0x1 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: cmp_or3: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lo +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hi +; CHECK-GI-NEXT: cmp w4, w5 +; CHECK-GI-NEXT: orr w8, w8, w9 +; CHECK-GI-NEXT: cset w9, ne +; CHECK-GI-NEXT: orr w8, w8, w9 +; CHECK-GI-NEXT: and w0, w8, #0x1 +; CHECK-GI-NEXT: ret %7 = icmp ult i32 %0, %1 %8 = icmp ugt i32 %2, %3 %9 = select i1 %7, i1 true, i1 %8 @@ -152,30 +152,30 @@ define i32 @cmp_or3(i32 %0, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5) { ; (x0 < x1) || (x2 > x3) || (x4 != x5) || (x6 == x7) define i32 @cmp_or4(i32 %0, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5, i32 %6, i32 %7) { -; SDISEL-LABEL: cmp_or4: -; SDISEL: // %bb.0: -; SDISEL-NEXT: cmp w0, w1 -; SDISEL-NEXT: ccmp w2, w3, #2, hs -; SDISEL-NEXT: ccmp w4, w5, #0, ls -; SDISEL-NEXT: ccmp w6, w7, #4, eq -; SDISEL-NEXT: cset w0, eq -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: cmp_or4: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: cmp w0, w1 +; CHECK-SD-NEXT: ccmp w2, w3, #2, hs +; CHECK-SD-NEXT: ccmp w4, w5, #0, ls +; CHECK-SD-NEXT: ccmp w6, w7, #4, eq +; CHECK-SD-NEXT: cset w0, eq +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: cmp_or4: -; GISEL: // %bb.0: -; GISEL-NEXT: cmp w0, w1 -; GISEL-NEXT: cset w8, lo -; GISEL-NEXT: cmp w2, w3 -; GISEL-NEXT: cset w9, hi -; GISEL-NEXT: cmp w4, w5 -; GISEL-NEXT: cset w10, ne -; GISEL-NEXT: cmp w6, w7 -; GISEL-NEXT: orr w8, w8, w9 -; GISEL-NEXT: cset w11, eq -; GISEL-NEXT: orr w9, w10, w11 -; GISEL-NEXT: orr w8, w8, w9 -; GISEL-NEXT: and w0, w8, #0x1 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: cmp_or4: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: cmp w0, w1 +; CHECK-GI-NEXT: cset w8, lo +; CHECK-GI-NEXT: cmp w2, w3 +; CHECK-GI-NEXT: cset w9, hi +; CHECK-GI-NEXT: cmp w4, w5 +; CHECK-GI-NEXT: cset w10, ne +; CHECK-GI-NEXT: cmp w6, w7 +; CHECK-GI-NEXT: orr w8, w8, w9 +; CHECK-GI-NEXT: cset w11, eq +; CHECK-GI-NEXT: orr w9, w10, w11 +; CHECK-GI-NEXT: orr w8, w8, w9 +; CHECK-GI-NEXT: and w0, w8, #0x1 +; CHECK-GI-NEXT: ret %9 = icmp ult i32 %0, %1 %10 = icmp ugt i32 %2, %3 %11 = select i1 %9, i1 true, i1 %10 @@ -189,22 +189,22 @@ define i32 @cmp_or4(i32 %0, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5, i32 %6, i32 ; (x0 != 0) || (x1 != 0) define i32 @true_or2(i32 %0, i32 %1) { -; SDISEL-LABEL: true_or2: -; SDISEL: // %bb.0: -; SDISEL-NEXT: orr w8, w0, w1 -; SDISEL-NEXT: cmp w8, #0 -; SDISEL-NEXT: cset w0, ne -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: true_or2: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: orr w8, w0, w1 +; CHECK-SD-NEXT: cmp w8, #0 +; CHECK-SD-NEXT: cset w0, ne +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: true_or2: -; GISEL: // %bb.0: -; GISEL-NEXT: cmp w0, #0 -; GISEL-NEXT: cset w8, ne -; GISEL-NEXT: cmp w1, #0 -; GISEL-NEXT: cset w9, ne -; GISEL-NEXT: orr w8, w8, w9 -; GISEL-NEXT: and w0, w8, #0x1 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: true_or2: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: cmp w0, #0 +; CHECK-GI-NEXT: cset w8, ne +; CHECK-GI-NEXT: cmp w1, #0 +; CHECK-GI-NEXT: cset w9, ne +; CHECK-GI-NEXT: orr w8, w8, w9 +; CHECK-GI-NEXT: and w0, w8, #0x1 +; CHECK-GI-NEXT: ret %3 = icmp ne i32 %0, 0 %4 = icmp ne i32 %1, 0 %5 = select i1 %3, i1 true, i1 %4 @@ -214,26 +214,26 @@ define i32 @true_or2(i32 %0, i32 %1) { ; (x0 != 0) || (x1 != 0) || (x2 != 0) define i32 @true_or3(i32 %0, i32 %1, i32 %2) { -; SDISEL-LABEL: true_or3: -; SDISEL: // %bb.0: -; SDISEL-NEXT: orr w8, w0, w1 -; SDISEL-NEXT: orr w8, w8, w2 -; SDISEL-NEXT: cmp w8, #0 -; SDISEL-NEXT: cset w0, ne -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: true_or3: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: orr w8, w0, w1 +; CHECK-SD-NEXT: orr w8, w8, w2 +; CHECK-SD-NEXT: cmp w8, #0 +; CHECK-SD-NEXT: cset w0, ne +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: true_or3: -; GISEL: // %bb.0: -; GISEL-NEXT: cmp w0, #0 -; GISEL-NEXT: cset w8, ne -; GISEL-NEXT: cmp w1, #0 -; GISEL-NEXT: cset w9, ne -; GISEL-NEXT: cmp w2, #0 -; GISEL-NEXT: orr w8, w8, w9 -; GISEL-NEXT: cset w9, ne -; GISEL-NEXT: orr w8, w8, w9 -; GISEL-NEXT: and w0, w8, #0x1 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: true_or3: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: cmp w0, #0 +; CHECK-GI-NEXT: cset w8, ne +; CHECK-GI-NEXT: cmp w1, #0 +; CHECK-GI-NEXT: cset w9, ne +; CHECK-GI-NEXT: cmp w2, #0 +; CHECK-GI-NEXT: orr w8, w8, w9 +; CHECK-GI-NEXT: cset w9, ne +; CHECK-GI-NEXT: orr w8, w8, w9 +; CHECK-GI-NEXT: and w0, w8, #0x1 +; CHECK-GI-NEXT: ret %4 = icmp ne i32 %0, 0 %5 = icmp ne i32 %1, 0 %6 = select i1 %4, i1 true, i1 %5 @@ -260,22 +260,22 @@ define i32 @neg_range_int(i32 %a, i32 %b, i32 %c) { ; (b > -(d | 1) && a < c) define i32 @neg_range_int_comp(i32 %a, i32 %b, i32 %c, i32 %d) { -; SDISEL-LABEL: neg_range_int_comp: -; SDISEL: // %bb.0: -; SDISEL-NEXT: orr w8, w3, #0x1 -; SDISEL-NEXT: cmp w0, w2 -; SDISEL-NEXT: ccmn w1, w8, #4, lt -; SDISEL-NEXT: csel w0, w1, w0, gt -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: neg_range_int_comp: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: orr w8, w3, #0x1 +; CHECK-SD-NEXT: cmp w0, w2 +; CHECK-SD-NEXT: ccmn w1, w8, #4, lt +; CHECK-SD-NEXT: csel w0, w1, w0, gt +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: neg_range_int_comp: -; GISEL: // %bb.0: -; GISEL-NEXT: orr w8, w3, #0x1 -; GISEL-NEXT: cmp w0, w2 -; GISEL-NEXT: neg w8, w8 -; GISEL-NEXT: ccmp w1, w8, #4, lt -; GISEL-NEXT: csel w0, w1, w0, gt -; GISEL-NEXT: ret +; CHECK-GI-LABEL: neg_range_int_comp: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: orr w8, w3, #0x1 +; CHECK-GI-NEXT: cmp w0, w2 +; CHECK-GI-NEXT: neg w8, w8 +; CHECK-GI-NEXT: ccmp w1, w8, #4, lt +; CHECK-GI-NEXT: csel w0, w1, w0, gt +; CHECK-GI-NEXT: ret %dor = or i32 %d, 1 %negd = sub i32 0, %dor %cmp = icmp sgt i32 %b, %negd @@ -287,22 +287,22 @@ define i32 @neg_range_int_comp(i32 %a, i32 %b, i32 %c, i32 %d) { ; (b >u -(d | 1) && a < c) define i32 @neg_range_int_comp_u(i32 %a, i32 %b, i32 %c, i32 %d) { -; SDISEL-LABEL: neg_range_int_comp_u: -; SDISEL: // %bb.0: -; SDISEL-NEXT: orr w8, w3, #0x1 -; SDISEL-NEXT: cmp w0, w2 -; SDISEL-NEXT: ccmn w1, w8, #0, lt -; SDISEL-NEXT: csel w0, w1, w0, hi -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: neg_range_int_comp_u: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: orr w8, w3, #0x1 +; CHECK-SD-NEXT: cmp w0, w2 +; CHECK-SD-NEXT: ccmn w1, w8, #0, lt +; CHECK-SD-NEXT: csel w0, w1, w0, hi +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: neg_range_int_comp_u: -; GISEL: // %bb.0: -; GISEL-NEXT: orr w8, w3, #0x1 -; GISEL-NEXT: cmp w0, w2 -; GISEL-NEXT: neg w8, w8 -; GISEL-NEXT: ccmp w1, w8, #0, lt -; GISEL-NEXT: csel w0, w1, w0, hi -; GISEL-NEXT: ret +; CHECK-GI-LABEL: neg_range_int_comp_u: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: orr w8, w3, #0x1 +; CHECK-GI-NEXT: cmp w0, w2 +; CHECK-GI-NEXT: neg w8, w8 +; CHECK-GI-NEXT: ccmp w1, w8, #0, lt +; CHECK-GI-NEXT: csel w0, w1, w0, hi +; CHECK-GI-NEXT: ret %dor = or i32 %d, 1 %negd = sub i32 0, %dor %cmp = icmp ugt i32 %b, %negd @@ -314,22 +314,22 @@ define i32 @neg_range_int_comp_u(i32 %a, i32 %b, i32 %c, i32 %d) { ; (b > -(d | 1) && a u < c) define i32 @neg_range_int_comp_ua(i32 %a, i32 %b, i32 %c, i32 %d) { -; SDISEL-LABEL: neg_range_int_comp_ua: -; SDISEL: // %bb.0: -; SDISEL-NEXT: orr w8, w3, #0x1 -; SDISEL-NEXT: cmp w0, w2 -; SDISEL-NEXT: ccmn w1, w8, #4, lo -; SDISEL-NEXT: csel w0, w1, w0, gt -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: neg_range_int_comp_ua: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: orr w8, w3, #0x1 +; CHECK-SD-NEXT: cmp w0, w2 +; CHECK-SD-NEXT: ccmn w1, w8, #4, lo +; CHECK-SD-NEXT: csel w0, w1, w0, gt +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: neg_range_int_comp_ua: -; GISEL: // %bb.0: -; GISEL-NEXT: orr w8, w3, #0x1 -; GISEL-NEXT: cmp w0, w2 -; GISEL-NEXT: neg w8, w8 -; GISEL-NEXT: ccmp w1, w8, #4, lo -; GISEL-NEXT: csel w0, w1, w0, gt -; GISEL-NEXT: ret +; CHECK-GI-LABEL: neg_range_int_comp_ua: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: orr w8, w3, #0x1 +; CHECK-GI-NEXT: cmp w0, w2 +; CHECK-GI-NEXT: neg w8, w8 +; CHECK-GI-NEXT: ccmp w1, w8, #4, lo +; CHECK-GI-NEXT: csel w0, w1, w0, gt +; CHECK-GI-NEXT: ret %dor = or i32 %d, 1 %negd = sub i32 0, %dor %cmp = icmp sgt i32 %b, %negd @@ -341,19 +341,19 @@ define i32 @neg_range_int_comp_ua(i32 %a, i32 %b, i32 %c, i32 %d) { ; (b <= -3 && a > c) define i32 @neg_range_int_2(i32 %a, i32 %b, i32 %c) { -; SDISEL-LABEL: neg_range_int_2: -; SDISEL: // %bb.0: -; SDISEL-NEXT: cmp w0, w2 -; SDISEL-NEXT: ccmn w1, #4, #4, gt -; SDISEL-NEXT: csel w0, w1, w0, gt -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: neg_range_int_2: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: cmp w0, w2 +; CHECK-SD-NEXT: ccmn w1, #4, #4, gt +; CHECK-SD-NEXT: csel w0, w1, w0, gt +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: neg_range_int_2: -; GISEL: // %bb.0: -; GISEL-NEXT: cmp w0, w2 -; GISEL-NEXT: ccmn w1, #3, #8, gt -; GISEL-NEXT: csel w0, w1, w0, ge -; GISEL-NEXT: ret +; CHECK-GI-LABEL: neg_range_int_2: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: cmp w0, w2 +; CHECK-GI-NEXT: ccmn w1, #3, #8, gt +; CHECK-GI-NEXT: csel w0, w1, w0, ge +; CHECK-GI-NEXT: ret %cmp = icmp sge i32 %b, -3 %cmp1 = icmp sgt i32 %a, %c %or.cond = and i1 %cmp, %cmp1 @@ -363,22 +363,22 @@ define i32 @neg_range_int_2(i32 %a, i32 %b, i32 %c) { ; (b < -(d | 1) && a >= c) define i32 @neg_range_int_comp2(i32 %a, i32 %b, i32 %c, i32 %d) { -; SDISEL-LABEL: neg_range_int_comp2: -; SDISEL: // %bb.0: -; SDISEL-NEXT: orr w8, w3, #0x1 -; SDISEL-NEXT: cmp w0, w2 -; SDISEL-NEXT: ccmn w1, w8, #0, ge -; SDISEL-NEXT: csel w0, w1, w0, lt -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: neg_range_int_comp2: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: orr w8, w3, #0x1 +; CHECK-SD-NEXT: cmp w0, w2 +; CHECK-SD-NEXT: ccmn w1, w8, #0, ge +; CHECK-SD-NEXT: csel w0, w1, w0, lt +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: neg_range_int_comp2: -; GISEL: // %bb.0: -; GISEL-NEXT: orr w8, w3, #0x1 -; GISEL-NEXT: cmp w0, w2 -; GISEL-NEXT: neg w8, w8 -; GISEL-NEXT: ccmp w1, w8, #0, ge -; GISEL-NEXT: csel w0, w1, w0, lt -; GISEL-NEXT: ret +; CHECK-GI-LABEL: neg_range_int_comp2: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: orr w8, w3, #0x1 +; CHECK-GI-NEXT: cmp w0, w2 +; CHECK-GI-NEXT: neg w8, w8 +; CHECK-GI-NEXT: ccmp w1, w8, #0, ge +; CHECK-GI-NEXT: csel w0, w1, w0, lt +; CHECK-GI-NEXT: ret %dor = or i32 %d, 1 %negd = sub i32 0, %dor %cmp = icmp slt i32 %b, %negd @@ -390,22 +390,22 @@ define i32 @neg_range_int_comp2(i32 %a, i32 %b, i32 %c, i32 %d) { ; (b <u -(d | 1) && a > c) define i32 @neg_range_int_comp_u2(i32 %a, i32 %b, i32 %c, i32 %d) { -; SDISEL-LABEL: neg_range_int_comp_u2: -; SDISEL: // %bb.0: -; SDISEL-NEXT: orr w8, w3, #0x1 -; SDISEL-NEXT: cmp w0, w2 -; SDISEL-NEXT: ccmn w1, w8, #2, gt -; SDISEL-NEXT: csel w0, w1, w0, lo -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: neg_range_int_comp_u2: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: orr w8, w3, #0x1 +; CHECK-SD-NEXT: cmp w0, w2 +; CHECK-SD-NEXT: ccmn w1, w8, #2, gt +; CHECK-SD-NEXT: csel w0, w1, w0, lo +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: neg_range_int_comp_u2: -; GISEL: // %bb.0: -; GISEL-NEXT: orr w8, w3, #0x1 -; GISEL-NEXT: cmp w0, w2 -; GISEL-NEXT: neg w8, w8 -; GISEL-NEXT: ccmp w1, w8, #2, gt -; GISEL-NEXT: csel w0, w1, w0, lo -; GISEL-NEXT: ret +; CHECK-GI-LABEL: neg_range_int_comp_u2: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: orr w8, w3, #0x1 +; CHECK-GI-NEXT: cmp w0, w2 +; CHECK-GI-NEXT: neg w8, w8 +; CHECK-GI-NEXT: ccmp w1, w8, #2, gt +; CHECK-GI-NEXT: csel w0, w1, w0, lo +; CHECK-GI-NEXT: ret %dor = or i32 %d, 1 %negd = sub i32 0, %dor %cmp = icmp ult i32 %b, %negd @@ -417,22 +417,22 @@ define i32 @neg_range_int_comp_u2(i32 %a, i32 %b, i32 %c, i32 %d) { ; (b > -(d | 1) && a u > c) define i32 @neg_range_int_comp_ua2(i32 %a, i32 %b, i32 %c, i32 %d) { -; SDISEL-LABEL: neg_range_int_comp_ua2: -; SDISEL: // %bb.0: -; SDISEL-NEXT: orr w8, w3, #0x1 -; SDISEL-NEXT: cmp w0, w2 -; SDISEL-NEXT: ccmn w1, w8, #4, hi -; SDISEL-NEXT: csel w0, w1, w0, gt -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: neg_range_int_comp_ua2: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: orr w8, w3, #0x1 +; CHECK-SD-NEXT: cmp w0, w2 +; CHECK-SD-NEXT: ccmn w1, w8, #4, hi +; CHECK-SD-NEXT: csel w0, w1, w0, gt +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: neg_range_int_comp_ua2: -; GISEL: // %bb.0: -; GISEL-NEXT: orr w8, w3, #0x1 -; GISEL-NEXT: cmp w0, w2 -; GISEL-NEXT: neg w8, w8 -; GISEL-NEXT: ccmp w1, w8, #4, hi -; GISEL-NEXT: csel w0, w1, w0, gt -; GISEL-NEXT: ret +; CHECK-GI-LABEL: neg_range_int_comp_ua2: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: orr w8, w3, #0x1 +; CHECK-GI-NEXT: cmp w0, w2 +; CHECK-GI-NEXT: neg w8, w8 +; CHECK-GI-NEXT: ccmp w1, w8, #4, hi +; CHECK-GI-NEXT: csel w0, w1, w0, gt +; CHECK-GI-NEXT: ret %dor = or i32 %d, 1 %negd = sub i32 0, %dor %cmp = icmp sgt i32 %b, %negd @@ -444,22 +444,22 @@ define i32 @neg_range_int_comp_ua2(i32 %a, i32 %b, i32 %c, i32 %d) { ; (b > -(d | 1) && a u == c) define i32 @neg_range_int_comp_ua3(i32 %a, i32 %b, i32 %c, i32 %d) { -; SDISEL-LABEL: neg_range_int_comp_ua3: -; SDISEL: // %bb.0: -; SDISEL-NEXT: orr w8, w3, #0x1 -; SDISEL-NEXT: cmp w0, w2 -; SDISEL-NEXT: ccmn w1, w8, #4, eq -; SDISEL-NEXT: csel w0, w1, w0, gt -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: neg_range_int_comp_ua3: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: orr w8, w3, #0x1 +; CHECK-SD-NEXT: cmp w0, w2 +; CHECK-SD-NEXT: ccmn w1, w8, #4, eq +; CHECK-SD-NEXT: csel w0, w1, w0, gt +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: neg_range_int_comp_ua3: -; GISEL: // %bb.0: -; GISEL-NEXT: orr w8, w3, #0x1 -; GISEL-NEXT: cmp w0, w2 -; GISEL-NEXT: neg w8, w8 -; GISEL-NEXT: ccmp w1, w8, #4, eq -; GISEL-NEXT: csel w0, w1, w0, gt -; GISEL-NEXT: ret +; CHECK-GI-LABEL: neg_range_int_comp_ua3: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: orr w8, w3, #0x1 +; CHECK-GI-NEXT: cmp w0, w2 +; CHECK-GI-NEXT: neg w8, w8 +; CHECK-GI-NEXT: ccmp w1, w8, #4, eq +; CHECK-GI-NEXT: csel w0, w1, w0, gt +; CHECK-GI-NEXT: ret %dor = or i32 %d, 1 %negd = sub i32 0, %dor %cmp = icmp sgt i32 %b, %negd @@ -471,26 +471,26 @@ define i32 @neg_range_int_comp_ua3(i32 %a, i32 %b, i32 %c, i32 %d) { ; -(a | 1) > (b | 3) && a < c define i32 @neg_range_int_c(i32 %a, i32 %b, i32 %c) { -; SDISEL-LABEL: neg_range_int_c: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: orr w8, w0, #0x1 -; SDISEL-NEXT: orr w9, w1, #0x3 -; SDISEL-NEXT: cmn w9, w8 -; SDISEL-NEXT: ccmp w2, w0, #2, lo -; SDISEL-NEXT: cset w0, lo -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: neg_range_int_c: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: orr w8, w0, #0x1 +; CHECK-SD-NEXT: orr w9, w1, #0x3 +; CHECK-SD-NEXT: cmn w9, w8 +; CHECK-SD-NEXT: ccmp w2, w0, #2, lo +; CHECK-SD-NEXT: cset w0, lo +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: neg_range_int_c: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: orr w8, w0, #0x1 -; GISEL-NEXT: orr w9, w1, #0x3 -; GISEL-NEXT: neg w8, w8 -; GISEL-NEXT: cmp w9, w8 -; GISEL-NEXT: cset w8, lo -; GISEL-NEXT: cmp w2, w0 -; GISEL-NEXT: cset w9, lo -; GISEL-NEXT: and w0, w8, w9 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: neg_range_int_c: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: orr w8, w0, #0x1 +; CHECK-GI-NEXT: orr w9, w1, #0x3 +; CHECK-GI-NEXT: neg w8, w8 +; CHECK-GI-NEXT: cmp w9, w8 +; CHECK-GI-NEXT: cset w8, lo +; CHECK-GI-NEXT: cmp w2, w0 +; CHECK-GI-NEXT: cset w9, lo +; CHECK-GI-NEXT: and w0, w8, w9 +; CHECK-GI-NEXT: ret entry: %or = or i32 %a, 1 %sub = sub i32 0, %or diff --git a/llvm/test/CodeGen/AArch64/dag-combine-select.ll b/llvm/test/CodeGen/AArch64/dag-combine-select.ll index 56208f1..02b0077 100644 --- a/llvm/test/CodeGen/AArch64/dag-combine-select.ll +++ b/llvm/test/CodeGen/AArch64/dag-combine-select.ll @@ -1,26 +1,26 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc -mtriple arm64-none-eabi -o - %s | FileCheck %s --check-prefixes=CHECK,SDISEL -; RUN: llc -mtriple arm64-none-eabi -global-isel -o - %s | FileCheck %s --check-prefixes=CHECK,GISEL +; RUN: llc -mtriple arm64-none-eabi -o - %s | FileCheck %s --check-prefixes=CHECK,CHECK-SD +; RUN: llc -mtriple arm64-none-eabi -global-isel -o - %s | FileCheck %s --check-prefixes=CHECK,CHECK-GI @out = internal global i32 0, align 4 ; Ensure that we transform select(C0, x, select(C1, x, y)) towards ; select(C0 | C1, x, y) so we can use CMP;CCMP for the implementation. define i32 @test0(i32 %v0, i32 %v1, i32 %v2) { -; SDISEL-LABEL: test0: -; SDISEL: // %bb.0: -; SDISEL-NEXT: cmp w0, #7 -; SDISEL-NEXT: ccmp w1, #0, #0, ne -; SDISEL-NEXT: csel w0, w1, w2, gt -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: test0: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: cmp w0, #7 +; CHECK-SD-NEXT: ccmp w1, #0, #0, ne +; CHECK-SD-NEXT: csel w0, w1, w2, gt +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: test0: -; GISEL: // %bb.0: -; GISEL-NEXT: cmp w0, #7 -; GISEL-NEXT: csel w8, w1, w2, eq -; GISEL-NEXT: cmp w1, #0 -; GISEL-NEXT: csel w0, w1, w8, gt -; GISEL-NEXT: ret +; CHECK-GI-LABEL: test0: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: cmp w0, #7 +; CHECK-GI-NEXT: csel w8, w1, w2, eq +; CHECK-GI-NEXT: cmp w1, #0 +; CHECK-GI-NEXT: csel w0, w1, w8, gt +; CHECK-GI-NEXT: ret %cmp1 = icmp eq i32 %v0, 7 %cmp2 = icmp sgt i32 %v1, 0 %sel0 = select i1 %cmp1, i32 %v1, i32 %v2 @@ -32,36 +32,36 @@ define i32 @test0(i32 %v0, i32 %v1, i32 %v2) { ; sequences. This case should be transformed to select(C0, select(C1, x, y), y) ; anyway to get CSE effects. define void @test1(i32 %bitset, i32 %val0, i32 %val1) { -; SDISEL-LABEL: test1: -; SDISEL: // %bb.0: -; SDISEL-NEXT: cmp w0, #7 -; SDISEL-NEXT: adrp x9, out -; SDISEL-NEXT: csel w8, w1, w2, eq -; SDISEL-NEXT: cmp w8, #13 -; SDISEL-NEXT: csel w8, w1, w2, lo -; SDISEL-NEXT: cmp w0, #42 -; SDISEL-NEXT: csel w10, w1, w8, eq -; SDISEL-NEXT: str w8, [x9, :lo12:out] -; SDISEL-NEXT: str w10, [x9, :lo12:out] -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: test1: +; CHECK-SD: // %bb.0: +; CHECK-SD-NEXT: cmp w0, #7 +; CHECK-SD-NEXT: adrp x9, out +; CHECK-SD-NEXT: csel w8, w1, w2, eq +; CHECK-SD-NEXT: cmp w8, #13 +; CHECK-SD-NEXT: csel w8, w1, w2, lo +; CHECK-SD-NEXT: cmp w0, #42 +; CHECK-SD-NEXT: csel w10, w1, w8, eq +; CHECK-SD-NEXT: str w8, [x9, :lo12:out] +; CHECK-SD-NEXT: str w10, [x9, :lo12:out] +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: test1: -; GISEL: // %bb.0: -; GISEL-NEXT: cmp w0, #7 -; GISEL-NEXT: csel w8, w1, w2, eq -; GISEL-NEXT: cmp w8, #13 -; GISEL-NEXT: cset w8, lo -; GISEL-NEXT: tst w8, #0x1 -; GISEL-NEXT: csel w9, w1, w2, ne -; GISEL-NEXT: cmp w0, #42 -; GISEL-NEXT: cset w10, eq -; GISEL-NEXT: orr w8, w10, w8 -; GISEL-NEXT: tst w8, #0x1 -; GISEL-NEXT: adrp x8, out -; GISEL-NEXT: csel w10, w1, w2, ne -; GISEL-NEXT: str w9, [x8, :lo12:out] -; GISEL-NEXT: str w10, [x8, :lo12:out] -; GISEL-NEXT: ret +; CHECK-GI-LABEL: test1: +; CHECK-GI: // %bb.0: +; CHECK-GI-NEXT: cmp w0, #7 +; CHECK-GI-NEXT: csel w8, w1, w2, eq +; CHECK-GI-NEXT: cmp w8, #13 +; CHECK-GI-NEXT: cset w8, lo +; CHECK-GI-NEXT: tst w8, #0x1 +; CHECK-GI-NEXT: csel w9, w1, w2, ne +; CHECK-GI-NEXT: cmp w0, #42 +; CHECK-GI-NEXT: cset w10, eq +; CHECK-GI-NEXT: orr w8, w10, w8 +; CHECK-GI-NEXT: tst w8, #0x1 +; CHECK-GI-NEXT: adrp x8, out +; CHECK-GI-NEXT: csel w10, w1, w2, ne +; CHECK-GI-NEXT: str w9, [x8, :lo12:out] +; CHECK-GI-NEXT: str w10, [x8, :lo12:out] +; CHECK-GI-NEXT: ret %cmp1 = icmp eq i32 %bitset, 7 %cond = select i1 %cmp1, i32 %val0, i32 %val1 %cmp5 = icmp ult i32 %cond, 13 diff --git a/llvm/test/CodeGen/AArch64/fp16_intrinsic_scalar_1op.ll b/llvm/test/CodeGen/AArch64/fp16_intrinsic_scalar_1op.ll index 1b98954..b056460 100644 --- a/llvm/test/CodeGen/AArch64/fp16_intrinsic_scalar_1op.ll +++ b/llvm/test/CodeGen/AArch64/fp16_intrinsic_scalar_1op.ll @@ -1,6 +1,6 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc < %s -mtriple=aarch64 -global-isel=0 -mattr=+v8.2a,+fullfp16 | FileCheck %s --check-prefixes=CHECK,SDISEL -; RUN: llc < %s -mtriple=aarch64 -global-isel=1 -mattr=+v8.2a,+fullfp16 | FileCheck %s --check-prefixes=CHECK,GISEL +; RUN: llc < %s -mtriple=aarch64 -global-isel=0 -mattr=+v8.2a,+fullfp16 | FileCheck %s --check-prefixes=CHECK,CHECK-SD +; RUN: llc < %s -mtriple=aarch64 -global-isel=1 -mattr=+v8.2a,+fullfp16 | FileCheck %s --check-prefixes=CHECK,CHECK-GI declare i64 @llvm.aarch64.neon.fcvtpu.i64.f16(half) declare i32 @llvm.aarch64.neon.fcvtpu.i32.f16(half) @@ -27,18 +27,18 @@ declare half @llvm.aarch64.neon.frecpx.f16(half) declare half @llvm.aarch64.neon.frecpe.f16(half) define dso_local i16 @t2(half %a) { -; SDISEL-LABEL: t2: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: fcmp h0, #0.0 -; SDISEL-NEXT: csetm w0, eq -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: t2: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: fcmp h0, #0.0 +; CHECK-SD-NEXT: csetm w0, eq +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: t2: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: fcmp h0, #0.0 -; GISEL-NEXT: cset w8, eq -; GISEL-NEXT: sbfx w0, w8, #0, #1 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: t2: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: fcmp h0, #0.0 +; CHECK-GI-NEXT: cset w8, eq +; CHECK-GI-NEXT: sbfx w0, w8, #0, #1 +; CHECK-GI-NEXT: ret entry: %0 = fcmp oeq half %a, 0xH0000 %vceqz = sext i1 %0 to i16 @@ -46,18 +46,18 @@ entry: } define dso_local i16 @t3(half %a) { -; SDISEL-LABEL: t3: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: fcmp h0, #0.0 -; SDISEL-NEXT: csetm w0, ge -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: t3: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: fcmp h0, #0.0 +; CHECK-SD-NEXT: csetm w0, ge +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: t3: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: fcmp h0, #0.0 -; GISEL-NEXT: cset w8, ge -; GISEL-NEXT: sbfx w0, w8, #0, #1 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: t3: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: fcmp h0, #0.0 +; CHECK-GI-NEXT: cset w8, ge +; CHECK-GI-NEXT: sbfx w0, w8, #0, #1 +; CHECK-GI-NEXT: ret entry: %0 = fcmp oge half %a, 0xH0000 %vcgez = sext i1 %0 to i16 @@ -65,18 +65,18 @@ entry: } define dso_local i16 @t4(half %a) { -; SDISEL-LABEL: t4: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: fcmp h0, #0.0 -; SDISEL-NEXT: csetm w0, gt -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: t4: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: fcmp h0, #0.0 +; CHECK-SD-NEXT: csetm w0, gt +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: t4: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: fcmp h0, #0.0 -; GISEL-NEXT: cset w8, gt -; GISEL-NEXT: sbfx w0, w8, #0, #1 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: t4: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: fcmp h0, #0.0 +; CHECK-GI-NEXT: cset w8, gt +; CHECK-GI-NEXT: sbfx w0, w8, #0, #1 +; CHECK-GI-NEXT: ret entry: %0 = fcmp ogt half %a, 0xH0000 %vcgtz = sext i1 %0 to i16 @@ -84,18 +84,18 @@ entry: } define dso_local i16 @t5(half %a) { -; SDISEL-LABEL: t5: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: fcmp h0, #0.0 -; SDISEL-NEXT: csetm w0, ls -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: t5: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: fcmp h0, #0.0 +; CHECK-SD-NEXT: csetm w0, ls +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: t5: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: fcmp h0, #0.0 -; GISEL-NEXT: cset w8, ls -; GISEL-NEXT: sbfx w0, w8, #0, #1 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: t5: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: fcmp h0, #0.0 +; CHECK-GI-NEXT: cset w8, ls +; CHECK-GI-NEXT: sbfx w0, w8, #0, #1 +; CHECK-GI-NEXT: ret entry: %0 = fcmp ole half %a, 0xH0000 %vclez = sext i1 %0 to i16 @@ -103,18 +103,18 @@ entry: } define dso_local i16 @t6(half %a) { -; SDISEL-LABEL: t6: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: fcmp h0, #0.0 -; SDISEL-NEXT: csetm w0, mi -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: t6: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: fcmp h0, #0.0 +; CHECK-SD-NEXT: csetm w0, mi +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: t6: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: fcmp h0, #0.0 -; GISEL-NEXT: cset w8, mi -; GISEL-NEXT: sbfx w0, w8, #0, #1 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: t6: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: fcmp h0, #0.0 +; CHECK-GI-NEXT: cset w8, mi +; CHECK-GI-NEXT: sbfx w0, w8, #0, #1 +; CHECK-GI-NEXT: ret entry: %0 = fcmp olt half %a, 0xH0000 %vcltz = sext i1 %0 to i16 @@ -172,15 +172,15 @@ entry: } define dso_local i16 @t16(half %a) { -; SDISEL-LABEL: t16: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: fcvtzs w0, h0 -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: t16: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: fcvtzs w0, h0 +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: t16: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: fcvtzu w0, h0 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: t16: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: fcvtzu w0, h0 +; CHECK-GI-NEXT: ret entry: %0 = fptoui half %a to i16 ret i16 %0 diff --git a/llvm/test/CodeGen/AArch64/fp16_intrinsic_scalar_2op.ll b/llvm/test/CodeGen/AArch64/fp16_intrinsic_scalar_2op.ll index 5b08ef2..da70599 100644 --- a/llvm/test/CodeGen/AArch64/fp16_intrinsic_scalar_2op.ll +++ b/llvm/test/CodeGen/AArch64/fp16_intrinsic_scalar_2op.ll @@ -1,6 +1,6 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc < %s -mtriple=aarch64 -global-isel=0 -mattr=+v8.2a,+fullfp16 | FileCheck %s --check-prefixes=CHECK,SDISEL -; RUN: llc < %s -mtriple=aarch64 -global-isel=1 -mattr=+v8.2a,+fullfp16 | FileCheck %s --check-prefixes=CHECK,GISEL +; RUN: llc < %s -mtriple=aarch64 -global-isel=0 -mattr=+v8.2a,+fullfp16 | FileCheck %s --check-prefixes=CHECK,CHECK-SD +; RUN: llc < %s -mtriple=aarch64 -global-isel=1 -mattr=+v8.2a,+fullfp16 | FileCheck %s --check-prefixes=CHECK,CHECK-GI declare half @llvm.aarch64.sisd.fabd.f16(half, half) @@ -35,18 +35,18 @@ entry: } define dso_local i16 @t_vceqh_f16(half %a, half %b) { -; SDISEL-LABEL: t_vceqh_f16: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: fcmp h0, h1 -; SDISEL-NEXT: csetm w0, eq -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: t_vceqh_f16: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: fcmp h0, h1 +; CHECK-SD-NEXT: csetm w0, eq +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: t_vceqh_f16: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: fcmp h0, h1 -; GISEL-NEXT: cset w8, eq -; GISEL-NEXT: sbfx w0, w8, #0, #1 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: t_vceqh_f16: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: fcmp h0, h1 +; CHECK-GI-NEXT: cset w8, eq +; CHECK-GI-NEXT: sbfx w0, w8, #0, #1 +; CHECK-GI-NEXT: ret entry: %0 = fcmp oeq half %a, %b %vcmpd = sext i1 %0 to i16 @@ -54,18 +54,18 @@ entry: } define dso_local i16 @t_vcgeh_f16(half %a, half %b) { -; SDISEL-LABEL: t_vcgeh_f16: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: fcmp h0, h1 -; SDISEL-NEXT: csetm w0, ge -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: t_vcgeh_f16: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: fcmp h0, h1 +; CHECK-SD-NEXT: csetm w0, ge +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: t_vcgeh_f16: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: fcmp h0, h1 -; GISEL-NEXT: cset w8, ge -; GISEL-NEXT: sbfx w0, w8, #0, #1 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: t_vcgeh_f16: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: fcmp h0, h1 +; CHECK-GI-NEXT: cset w8, ge +; CHECK-GI-NEXT: sbfx w0, w8, #0, #1 +; CHECK-GI-NEXT: ret entry: %0 = fcmp oge half %a, %b %vcmpd = sext i1 %0 to i16 @@ -73,18 +73,18 @@ entry: } define dso_local i16 @t_vcgth_f16(half %a, half %b) { -; SDISEL-LABEL: t_vcgth_f16: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: fcmp h0, h1 -; SDISEL-NEXT: csetm w0, gt -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: t_vcgth_f16: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: fcmp h0, h1 +; CHECK-SD-NEXT: csetm w0, gt +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: t_vcgth_f16: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: fcmp h0, h1 -; GISEL-NEXT: cset w8, gt -; GISEL-NEXT: sbfx w0, w8, #0, #1 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: t_vcgth_f16: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: fcmp h0, h1 +; CHECK-GI-NEXT: cset w8, gt +; CHECK-GI-NEXT: sbfx w0, w8, #0, #1 +; CHECK-GI-NEXT: ret entry: %0 = fcmp ogt half %a, %b %vcmpd = sext i1 %0 to i16 @@ -92,18 +92,18 @@ entry: } define dso_local i16 @t_vcleh_f16(half %a, half %b) { -; SDISEL-LABEL: t_vcleh_f16: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: fcmp h0, h1 -; SDISEL-NEXT: csetm w0, ls -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: t_vcleh_f16: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: fcmp h0, h1 +; CHECK-SD-NEXT: csetm w0, ls +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: t_vcleh_f16: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: fcmp h0, h1 -; GISEL-NEXT: cset w8, ls -; GISEL-NEXT: sbfx w0, w8, #0, #1 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: t_vcleh_f16: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: fcmp h0, h1 +; CHECK-GI-NEXT: cset w8, ls +; CHECK-GI-NEXT: sbfx w0, w8, #0, #1 +; CHECK-GI-NEXT: ret entry: %0 = fcmp ole half %a, %b %vcmpd = sext i1 %0 to i16 @@ -111,18 +111,18 @@ entry: } define dso_local i16 @t_vclth_f16(half %a, half %b) { -; SDISEL-LABEL: t_vclth_f16: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: fcmp h0, h1 -; SDISEL-NEXT: csetm w0, mi -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: t_vclth_f16: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: fcmp h0, h1 +; CHECK-SD-NEXT: csetm w0, mi +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: t_vclth_f16: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: fcmp h0, h1 -; GISEL-NEXT: cset w8, mi -; GISEL-NEXT: sbfx w0, w8, #0, #1 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: t_vclth_f16: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: fcmp h0, h1 +; CHECK-GI-NEXT: cset w8, mi +; CHECK-GI-NEXT: sbfx w0, w8, #0, #1 +; CHECK-GI-NEXT: ret entry: %0 = fcmp olt half %a, %b %vcmpd = sext i1 %0 to i16 @@ -187,18 +187,18 @@ declare half @llvm.aarch64.neon.vcvtfxu2fp.f16.i32(i32, i32) #1 declare i32 @llvm.aarch64.neon.vcvtfp2fxu.i32.f16(half, i32) #1 define dso_local half @test_vcvth_n_f16_s16_1(i16 %a) { -; SDISEL-LABEL: test_vcvth_n_f16_s16_1: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: fmov s0, w0 -; SDISEL-NEXT: scvtf h0, h0, #1 -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: test_vcvth_n_f16_s16_1: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: fmov s0, w0 +; CHECK-SD-NEXT: scvtf h0, h0, #1 +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: test_vcvth_n_f16_s16_1: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: sxth w8, w0 -; GISEL-NEXT: fmov s0, w8 -; GISEL-NEXT: scvtf h0, h0, #1 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: test_vcvth_n_f16_s16_1: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: sxth w8, w0 +; CHECK-GI-NEXT: fmov s0, w8 +; CHECK-GI-NEXT: scvtf h0, h0, #1 +; CHECK-GI-NEXT: ret entry: %sext = sext i16 %a to i32 %fcvth_n = tail call half @llvm.aarch64.neon.vcvtfxs2fp.f16.i32(i32 %sext, i32 1) @@ -206,18 +206,18 @@ entry: } define dso_local half @test_vcvth_n_f16_s16_16(i16 %a) { -; SDISEL-LABEL: test_vcvth_n_f16_s16_16: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: fmov s0, w0 -; SDISEL-NEXT: scvtf h0, h0, #16 -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: test_vcvth_n_f16_s16_16: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: fmov s0, w0 +; CHECK-SD-NEXT: scvtf h0, h0, #16 +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: test_vcvth_n_f16_s16_16: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: sxth w8, w0 -; GISEL-NEXT: fmov s0, w8 -; GISEL-NEXT: scvtf h0, h0, #16 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: test_vcvth_n_f16_s16_16: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: sxth w8, w0 +; CHECK-GI-NEXT: fmov s0, w8 +; CHECK-GI-NEXT: scvtf h0, h0, #16 +; CHECK-GI-NEXT: ret entry: %sext = sext i16 %a to i32 %fcvth_n = tail call half @llvm.aarch64.neon.vcvtfxs2fp.f16.i32(i32 %sext, i32 16) @@ -315,18 +315,18 @@ entry: } define dso_local half @test_vcvth_n_f16_u16_1(i16 %a) { -; SDISEL-LABEL: test_vcvth_n_f16_u16_1: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: fmov s0, w0 -; SDISEL-NEXT: ucvtf h0, h0, #1 -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: test_vcvth_n_f16_u16_1: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: fmov s0, w0 +; CHECK-SD-NEXT: ucvtf h0, h0, #1 +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: test_vcvth_n_f16_u16_1: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: and w8, w0, #0xffff -; GISEL-NEXT: fmov s0, w8 -; GISEL-NEXT: ucvtf h0, h0, #1 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: test_vcvth_n_f16_u16_1: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: and w8, w0, #0xffff +; CHECK-GI-NEXT: fmov s0, w8 +; CHECK-GI-NEXT: ucvtf h0, h0, #1 +; CHECK-GI-NEXT: ret entry: %0 = zext i16 %a to i32 %fcvth_n = tail call half @llvm.aarch64.neon.vcvtfxu2fp.f16.i32(i32 %0, i32 1) @@ -334,18 +334,18 @@ entry: } define dso_local half @test_vcvth_n_f16_u16_16(i16 %a) { -; SDISEL-LABEL: test_vcvth_n_f16_u16_16: -; SDISEL: // %bb.0: // %entry -; SDISEL-NEXT: fmov s0, w0 -; SDISEL-NEXT: ucvtf h0, h0, #16 -; SDISEL-NEXT: ret +; CHECK-SD-LABEL: test_vcvth_n_f16_u16_16: +; CHECK-SD: // %bb.0: // %entry +; CHECK-SD-NEXT: fmov s0, w0 +; CHECK-SD-NEXT: ucvtf h0, h0, #16 +; CHECK-SD-NEXT: ret ; -; GISEL-LABEL: test_vcvth_n_f16_u16_16: -; GISEL: // %bb.0: // %entry -; GISEL-NEXT: and w8, w0, #0xffff -; GISEL-NEXT: fmov s0, w8 -; GISEL-NEXT: ucvtf h0, h0, #16 -; GISEL-NEXT: ret +; CHECK-GI-LABEL: test_vcvth_n_f16_u16_16: +; CHECK-GI: // %bb.0: // %entry +; CHECK-GI-NEXT: and w8, w0, #0xffff +; CHECK-GI-NEXT: fmov s0, w8 +; CHECK-GI-NEXT: ucvtf h0, h0, #16 +; CHECK-GI-NEXT: ret entry: %0 = zext i16 %a to i32 %fcvth_n = tail call half @llvm.aarch64.neon.vcvtfxu2fp.f16.i32(i32 %0, i32 16) diff --git a/llvm/test/CodeGen/AArch64/late-taildup-computed-goto.ll b/llvm/test/CodeGen/AArch64/late-taildup-computed-goto.ll new file mode 100644 index 0000000..c4a027c --- /dev/null +++ b/llvm/test/CodeGen/AArch64/late-taildup-computed-goto.ll @@ -0,0 +1,162 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 +; RUN: llc -tail-dup-pred-size=2 -tail-dup-succ-size=2 -o - %s | FileCheck %s + +target triple = "arm64-apple-macosx13.0.0" + +@opcode.targets = local_unnamed_addr constant [6 x ptr] [ptr blockaddress(@test_interp, %op1.bb), ptr blockaddress(@test_interp, %op6.bb), ptr blockaddress(@test_interp, %loop.header), ptr blockaddress(@test_interp, %op2.bb), ptr blockaddress(@test_interp, %op4.bb), ptr blockaddress(@test_interp, %op5.bb)] + +define void @test_interp(ptr %frame, ptr %dst) { +; CHECK-LABEL: test_interp: +; CHECK: ; %bb.0: ; %entry +; CHECK-NEXT: stp x24, x23, [sp, #-64]! ; 16-byte Folded Spill +; CHECK-NEXT: stp x22, x21, [sp, #16] ; 16-byte Folded Spill +; CHECK-NEXT: stp x20, x19, [sp, #32] ; 16-byte Folded Spill +; CHECK-NEXT: stp x29, x30, [sp, #48] ; 16-byte Folded Spill +; CHECK-NEXT: .cfi_def_cfa_offset 64 +; CHECK-NEXT: .cfi_offset w30, -8 +; CHECK-NEXT: .cfi_offset w29, -16 +; CHECK-NEXT: .cfi_offset w19, -24 +; CHECK-NEXT: .cfi_offset w20, -32 +; CHECK-NEXT: .cfi_offset w21, -40 +; CHECK-NEXT: .cfi_offset w22, -48 +; CHECK-NEXT: .cfi_offset w23, -56 +; CHECK-NEXT: .cfi_offset w24, -64 +; CHECK-NEXT: Lloh0: +; CHECK-NEXT: adrp x21, _opcode.targets@PAGE +; CHECK-NEXT: Lloh1: +; CHECK-NEXT: add x21, x21, _opcode.targets@PAGEOFF +; CHECK-NEXT: mov x22, xzr +; CHECK-NEXT: add x8, x21, xzr, lsl #3 +; CHECK-NEXT: mov x19, x1 +; CHECK-NEXT: mov x20, x0 +; CHECK-NEXT: add x23, x22, #1 +; CHECK-NEXT: br x8 +; CHECK-NEXT: Ltmp0: ; Block address taken +; CHECK-NEXT: LBB0_1: ; %loop.header +; CHECK-NEXT: ; =>This Inner Loop Header: Depth=1 +; CHECK-NEXT: add x8, x21, x23, lsl #3 +; CHECK-NEXT: mov x20, xzr +; CHECK-NEXT: mov x22, xzr +; CHECK-NEXT: add x23, x23, #1 +; CHECK-NEXT: br x8 +; CHECK-NEXT: Ltmp1: ; Block address taken +; CHECK-NEXT: LBB0_2: ; %op1.bb +; CHECK-NEXT: ; =>This Inner Loop Header: Depth=1 +; CHECK-NEXT: str xzr, [x19] +; CHECK-NEXT: mov w8, #1 ; =0x1 +; CHECK-NEXT: ldr x0, [x20, #-8]! +; CHECK-NEXT: ldr x9, [x0, #8] +; CHECK-NEXT: str x8, [x0] +; CHECK-NEXT: ldr x8, [x9, #48] +; CHECK-NEXT: blr x8 +; CHECK-NEXT: add x8, x21, x23, lsl #3 +; CHECK-NEXT: add x23, x23, #1 +; CHECK-NEXT: br x8 +; CHECK-NEXT: Ltmp2: ; Block address taken +; CHECK-NEXT: LBB0_3: ; %op2.bb +; CHECK-NEXT: ; =>This Inner Loop Header: Depth=1 +; CHECK-NEXT: add x8, x21, x23, lsl #3 +; CHECK-NEXT: mov x20, xzr +; CHECK-NEXT: add x23, x23, #1 +; CHECK-NEXT: str x22, [x19] +; CHECK-NEXT: mov x22, xzr +; CHECK-NEXT: br x8 +; CHECK-NEXT: Ltmp3: ; Block address taken +; CHECK-NEXT: LBB0_4: ; %op4.bb +; CHECK-NEXT: ; =>This Inner Loop Header: Depth=1 +; CHECK-NEXT: str x22, [x19] +; CHECK-NEXT: add x10, x21, x23, lsl #3 +; CHECK-NEXT: add x23, x23, #1 +; CHECK-NEXT: ldur x8, [x22, #12] +; CHECK-NEXT: ldur x9, [x20, #-8] +; CHECK-NEXT: add x22, x22, #20 +; CHECK-NEXT: stp x8, x9, [x20, #-8] +; CHECK-NEXT: add x20, x20, #8 +; CHECK-NEXT: br x10 +; CHECK-NEXT: Ltmp4: ; Block address taken +; CHECK-NEXT: LBB0_5: ; %op5.bb +; CHECK-NEXT: ; =>This Inner Loop Header: Depth=1 +; CHECK-NEXT: str x22, [x19] +; CHECK-NEXT: add x10, x21, x23, lsl #3 +; CHECK-NEXT: add x23, x23, #1 +; CHECK-NEXT: ldur x8, [x22, #12] +; CHECK-NEXT: ldur x9, [x20, #-8] +; CHECK-NEXT: add x22, x22, #20 +; CHECK-NEXT: stp x8, x9, [x20, #-8] +; CHECK-NEXT: add x20, x20, #8 +; CHECK-NEXT: br x10 +; CHECK-NEXT: Ltmp5: ; Block address taken +; CHECK-NEXT: LBB0_6: ; %op6.bb +; CHECK-NEXT: ; =>This Inner Loop Header: Depth=1 +; CHECK-NEXT: ldr x0, [x20, #-8]! +; CHECK-NEXT: mov w8, #1 ; =0x1 +; CHECK-NEXT: ldr x9, [x0, #8] +; CHECK-NEXT: str x8, [x0] +; CHECK-NEXT: ldr x8, [x9, #48] +; CHECK-NEXT: blr x8 +; CHECK-NEXT: add x8, x21, x23, lsl #3 +; CHECK-NEXT: add x23, x23, #1 +; CHECK-NEXT: br x8 +; CHECK-NEXT: .loh AdrpAdd Lloh0, Lloh1 +entry: + br label %loop.header + +loop.header: + %iv = phi i64 [ 0, %entry ], [ %iv.next, %op1.bb ], [ %iv.next, %op2.bb ], [ %iv.next, %op4.bb ], [ %iv.next, %op5.bb ], [ %iv.next, %op6.bb ], [ %iv.next, %loop.header ] + %stack.pointer = phi ptr [ %frame, %entry ], [ %stack.8, %op1.bb ], [ null, %op2.bb ], [ %stack.next, %op4.bb ], [ %stack.next.2, %op5.bb ], [ %stack.4, %op6.bb ], [ null, %loop.header ] + %next.instr = phi ptr [ null, %entry ], [ %next.instr, %op1.bb ], [ null, %op2.bb ], [ %next.instr.20, %op4.bb ], [ %next.instr.21, %op5.bb ], [ %next.instr, %op6.bb ], [ null, %loop.header ] + %iv.next = add i64 %iv, 1 + %next_op = getelementptr [6 x ptr], ptr @opcode.targets, i64 0, i64 %iv + indirectbr ptr %next_op, [label %op1.bb, label %op6.bb, label %loop.header, label %op2.bb, label %op4.bb, label %op5.bb] + +op1.bb: + store ptr null, ptr %dst, align 8 + %stack.8 = getelementptr i8, ptr %stack.pointer, i64 -8 + %l.0 = load ptr, ptr %stack.8, align 8 + store i64 1, ptr %l.0, align 8 + %gep.0 = getelementptr i8, ptr %l.0, i64 8 + %l.1 = load ptr, ptr %gep.0, align 8 + %gep.1 = getelementptr i8, ptr %l.1, i64 48 + %l.2 = load ptr, ptr %gep.1, align 8 + tail call void %l.2(ptr nonnull %l.0) + br label %loop.header + +op2.bb: + store ptr %next.instr, ptr %dst, align 8 + br label %loop.header + +op4.bb: + store ptr %next.instr, ptr %dst, align 8 + %next.instr.20 = getelementptr i8, ptr %next.instr, i64 20 + %stack.2 = getelementptr i8, ptr %stack.pointer, i64 -8 + %l.3 = load ptr, ptr %stack.2, align 8 + %next.instr.12 = getelementptr i8, ptr %next.instr, i64 12 + %next.instr.12.val = load ptr, ptr %next.instr.12, align 2 + store ptr %next.instr.12.val, ptr %stack.2, align 8 + store ptr %l.3, ptr %stack.pointer, align 8 + %stack.next = getelementptr i8, ptr %stack.pointer, i64 8 + br label %loop.header + +op5.bb: + store ptr %next.instr, ptr %dst, align 8 + %next.instr.21 = getelementptr i8, ptr %next.instr, i64 20 + %stack.3 = getelementptr i8, ptr %stack.pointer, i64 -8 + %l.4 = load ptr, ptr %stack.3, align 8 + %next.instr.2 = getelementptr i8, ptr %next.instr, i64 12 + %next.instr.2.val = load ptr, ptr %next.instr.2, align 2 + store ptr %next.instr.2.val, ptr %stack.3, align 8 + store ptr %l.4, ptr %stack.pointer, align 8 + %stack.next.2 = getelementptr i8, ptr %stack.pointer, i64 8 + br label %loop.header + +op6.bb: + %stack.4 = getelementptr i8, ptr %stack.pointer, i64 -8 + %l.5 = load ptr, ptr %stack.4, align 8 + store i64 1, ptr %l.5, align 8 + %gep.5 = getelementptr i8, ptr %l.5, i64 8 + %l.6 = load ptr, ptr %gep.5, align 8 + %gep.6 = getelementptr i8, ptr %l.6, i64 48 + %l.7 = load ptr, ptr %gep.6, align 8 + tail call void %l.7(ptr nonnull %l.5) + br label %loop.header +} diff --git a/llvm/test/CodeGen/AArch64/preferred-function-alignment.ll b/llvm/test/CodeGen/AArch64/preferred-function-alignment.ll index 05f4fb1..a6cb712 100644 --- a/llvm/test/CodeGen/AArch64/preferred-function-alignment.ll +++ b/llvm/test/CodeGen/AArch64/preferred-function-alignment.ll @@ -40,3 +40,10 @@ define void @test_optsize() optsize { ; CHECK-LABEL: test_optsize ; CHECK-NEXT: .p2align 2 + +define void @test_minsize() minsize { + ret void +} + +; CHECK-LABEL: test_minsize +; CHECK-NEXT: .p2align 2 diff --git a/llvm/test/CodeGen/AMDGPU/gfx1250-scratch-scope-se.ll b/llvm/test/CodeGen/AMDGPU/gfx1250-scratch-scope-se.ll new file mode 100644 index 0000000..d1e82a0 --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/gfx1250-scratch-scope-se.ll @@ -0,0 +1,95 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 +; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1250 < %s | FileCheck -check-prefixes=GCN,GCN-SDAG %s +; RUN: llc -global-isel -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1250 < %s | FileCheck -check-prefixes=GCN,GCN-GISEL %s + +; Test that stores that may hit scratch are correctly promoted to SCOPE_SE. + +define void @test_scratch_store(ptr addrspace(5) %ptr, i32 %val) { +; GCN-LABEL: test_scratch_store: +; GCN: ; %bb.0: +; GCN-NEXT: s_wait_loadcnt_dscnt 0x0 +; GCN-NEXT: s_wait_kmcnt 0x0 +; GCN-NEXT: scratch_store_b32 v0, v1, off scope:SCOPE_SE +; GCN-NEXT: s_set_pc_i64 s[30:31] + store i32 %val, ptr addrspace(5) %ptr + ret void +} + +define void @test_unknown_flat_store(ptr %ptr, i32 %val) { +; GCN-LABEL: test_unknown_flat_store: +; GCN: ; %bb.0: +; GCN-NEXT: s_wait_loadcnt_dscnt 0x0 +; GCN-NEXT: s_wait_kmcnt 0x0 +; GCN-NEXT: flat_store_b32 v[0:1], v2 scope:SCOPE_SE +; GCN-NEXT: s_wait_dscnt 0x0 +; GCN-NEXT: s_set_pc_i64 s[30:31] + store i32 %val, ptr %ptr + ret void +} + +define void @test_flat_store_no_scratch_alloc(ptr %ptr, i32 %val) #0 { +; GCN-LABEL: test_flat_store_no_scratch_alloc: +; GCN: ; %bb.0: +; GCN-NEXT: s_wait_loadcnt_dscnt 0x0 +; GCN-NEXT: s_wait_kmcnt 0x0 +; GCN-NEXT: flat_store_b32 v[0:1], v2 +; GCN-NEXT: s_wait_dscnt 0x0 +; GCN-NEXT: s_set_pc_i64 s[30:31] + store i32 %val, ptr %ptr + ret void +} + +; TODO: handle +define void @test_flat_store_noalias_addrspace(ptr %ptr, i32 %val) { +; GCN-LABEL: test_flat_store_noalias_addrspace: +; GCN: ; %bb.0: +; GCN-NEXT: s_wait_loadcnt_dscnt 0x0 +; GCN-NEXT: s_wait_kmcnt 0x0 +; GCN-NEXT: flat_store_b32 v[0:1], v2 scope:SCOPE_SE +; GCN-NEXT: s_wait_dscnt 0x0 +; GCN-NEXT: s_set_pc_i64 s[30:31] + store i32 %val, ptr %ptr, !noalias.addrspace !{i32 5, i32 6} + ret void +} + +; TODO: would be nice to handle too +define void @test_flat_store_select(ptr addrspace(1) %a, ptr addrspace(3) %b, i1 %cond, i32 %val) { +; GCN-SDAG-LABEL: test_flat_store_select: +; GCN-SDAG: ; %bb.0: +; GCN-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0 +; GCN-SDAG-NEXT: s_wait_kmcnt 0x0 +; GCN-SDAG-NEXT: v_cmp_ne_u32_e32 vcc_lo, -1, v2 +; GCN-SDAG-NEXT: v_and_b32_e32 v3, 1, v3 +; GCN-SDAG-NEXT: s_mov_b64 s[0:1], src_shared_base +; GCN-SDAG-NEXT: v_cndmask_b32_e32 v2, 0, v2, vcc_lo +; GCN-SDAG-NEXT: v_cndmask_b32_e64 v5, 0, s1, vcc_lo +; GCN-SDAG-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2) +; GCN-SDAG-NEXT: v_cmp_eq_u32_e32 vcc_lo, 1, v3 +; GCN-SDAG-NEXT: v_dual_cndmask_b32 v1, v5, v1 :: v_dual_cndmask_b32 v0, v2, v0 +; GCN-SDAG-NEXT: flat_store_b32 v[0:1], v4 scope:SCOPE_SE +; GCN-SDAG-NEXT: s_wait_dscnt 0x0 +; GCN-SDAG-NEXT: s_set_pc_i64 s[30:31] +; +; GCN-GISEL-LABEL: test_flat_store_select: +; GCN-GISEL: ; %bb.0: +; GCN-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0 +; GCN-GISEL-NEXT: s_wait_kmcnt 0x0 +; GCN-GISEL-NEXT: v_cmp_ne_u32_e32 vcc_lo, -1, v2 +; GCN-GISEL-NEXT: v_and_b32_e32 v3, 1, v3 +; GCN-GISEL-NEXT: s_mov_b64 s[0:1], src_shared_base +; GCN-GISEL-NEXT: v_cndmask_b32_e32 v2, 0, v2, vcc_lo +; GCN-GISEL-NEXT: v_cndmask_b32_e64 v5, 0, s1, vcc_lo +; GCN-GISEL-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2) +; GCN-GISEL-NEXT: v_cmp_ne_u32_e32 vcc_lo, 0, v3 +; GCN-GISEL-NEXT: v_dual_cndmask_b32 v0, v2, v0 :: v_dual_cndmask_b32 v1, v5, v1 +; GCN-GISEL-NEXT: flat_store_b32 v[0:1], v4 scope:SCOPE_SE +; GCN-GISEL-NEXT: s_wait_dscnt 0x0 +; GCN-GISEL-NEXT: s_set_pc_i64 s[30:31] + %a.ascast = addrspacecast ptr addrspace(1) %a to ptr + %b.ascast = addrspacecast ptr addrspace(3) %b to ptr + %ptr = select i1 %cond, ptr %a.ascast, ptr %b.ascast + store i32 %val, ptr %ptr + ret void +} + +attributes #0 = { "amdgpu-no-flat-scratch-init" } diff --git a/llvm/test/CodeGen/AMDGPU/global-load-xcnt.ll b/llvm/test/CodeGen/AMDGPU/global-load-xcnt.ll index fd644a3..3a898a9 100644 --- a/llvm/test/CodeGen/AMDGPU/global-load-xcnt.ll +++ b/llvm/test/CodeGen/AMDGPU/global-load-xcnt.ll @@ -124,27 +124,27 @@ define i32 @test_v64i32_load_store(ptr addrspace(1) %ptr, i32 %idx, ptr addrspac ; GCN-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0 ; GCN-SDAG-NEXT: s_wait_kmcnt 0x0 ; GCN-SDAG-NEXT: s_clause 0xd -; GCN-SDAG-NEXT: scratch_store_b32 off, v40, s32 offset:52 -; GCN-SDAG-NEXT: scratch_store_b32 off, v41, s32 offset:48 -; GCN-SDAG-NEXT: scratch_store_b32 off, v42, s32 offset:44 -; GCN-SDAG-NEXT: scratch_store_b32 off, v43, s32 offset:40 -; GCN-SDAG-NEXT: scratch_store_b32 off, v44, s32 offset:36 -; GCN-SDAG-NEXT: scratch_store_b32 off, v45, s32 offset:32 -; GCN-SDAG-NEXT: scratch_store_b32 off, v56, s32 offset:28 -; GCN-SDAG-NEXT: scratch_store_b32 off, v57, s32 offset:24 -; GCN-SDAG-NEXT: scratch_store_b32 off, v58, s32 offset:20 -; GCN-SDAG-NEXT: scratch_store_b32 off, v59, s32 offset:16 -; GCN-SDAG-NEXT: scratch_store_b32 off, v60, s32 offset:12 -; GCN-SDAG-NEXT: scratch_store_b32 off, v61, s32 offset:8 -; GCN-SDAG-NEXT: scratch_store_b32 off, v62, s32 offset:4 -; GCN-SDAG-NEXT: scratch_store_b32 off, v63, s32 +; GCN-SDAG-NEXT: scratch_store_b32 off, v40, s32 offset:52 scope:SCOPE_SE +; GCN-SDAG-NEXT: scratch_store_b32 off, v41, s32 offset:48 scope:SCOPE_SE +; GCN-SDAG-NEXT: scratch_store_b32 off, v42, s32 offset:44 scope:SCOPE_SE +; GCN-SDAG-NEXT: scratch_store_b32 off, v43, s32 offset:40 scope:SCOPE_SE +; GCN-SDAG-NEXT: scratch_store_b32 off, v44, s32 offset:36 scope:SCOPE_SE +; GCN-SDAG-NEXT: scratch_store_b32 off, v45, s32 offset:32 scope:SCOPE_SE +; GCN-SDAG-NEXT: scratch_store_b32 off, v56, s32 offset:28 scope:SCOPE_SE +; GCN-SDAG-NEXT: scratch_store_b32 off, v57, s32 offset:24 scope:SCOPE_SE +; GCN-SDAG-NEXT: scratch_store_b32 off, v58, s32 offset:20 scope:SCOPE_SE +; GCN-SDAG-NEXT: scratch_store_b32 off, v59, s32 offset:16 scope:SCOPE_SE +; GCN-SDAG-NEXT: scratch_store_b32 off, v60, s32 offset:12 scope:SCOPE_SE +; GCN-SDAG-NEXT: scratch_store_b32 off, v61, s32 offset:8 scope:SCOPE_SE +; GCN-SDAG-NEXT: scratch_store_b32 off, v62, s32 offset:4 scope:SCOPE_SE +; GCN-SDAG-NEXT: scratch_store_b32 off, v63, s32 scope:SCOPE_SE ; GCN-SDAG-NEXT: global_load_b128 v[6:9], v[0:1], off offset:224 ; GCN-SDAG-NEXT: v_dual_mov_b32 v5, v4 :: v_dual_mov_b32 v4, v3 ; GCN-SDAG-NEXT: s_wait_loadcnt 0x0 -; GCN-SDAG-NEXT: scratch_store_b128 off, v[6:9], s32 offset:56 ; 16-byte Folded Spill +; GCN-SDAG-NEXT: scratch_store_b128 off, v[6:9], s32 offset:56 scope:SCOPE_SE ; 16-byte Folded Spill ; GCN-SDAG-NEXT: global_load_b128 v[6:9], v[0:1], off offset:240 ; GCN-SDAG-NEXT: s_wait_loadcnt 0x0 -; GCN-SDAG-NEXT: scratch_store_b128 off, v[6:9], s32 offset:72 ; 16-byte Folded Spill +; GCN-SDAG-NEXT: scratch_store_b128 off, v[6:9], s32 offset:72 scope:SCOPE_SE ; 16-byte Folded Spill ; GCN-SDAG-NEXT: s_clause 0xd ; GCN-SDAG-NEXT: global_load_b128 v[10:13], v[0:1], off offset:192 ; GCN-SDAG-NEXT: global_load_b128 v[14:17], v[0:1], off offset:208 @@ -206,27 +206,27 @@ define i32 @test_v64i32_load_store(ptr addrspace(1) %ptr, i32 %idx, ptr addrspac ; GCN-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0 ; GCN-GISEL-NEXT: s_wait_kmcnt 0x0 ; GCN-GISEL-NEXT: s_clause 0xf -; GCN-GISEL-NEXT: scratch_store_b32 off, v40, s32 offset:60 -; GCN-GISEL-NEXT: scratch_store_b32 off, v41, s32 offset:56 -; GCN-GISEL-NEXT: scratch_store_b32 off, v42, s32 offset:52 -; GCN-GISEL-NEXT: scratch_store_b32 off, v43, s32 offset:48 -; GCN-GISEL-NEXT: scratch_store_b32 off, v44, s32 offset:44 -; GCN-GISEL-NEXT: scratch_store_b32 off, v45, s32 offset:40 -; GCN-GISEL-NEXT: scratch_store_b32 off, v46, s32 offset:36 -; GCN-GISEL-NEXT: scratch_store_b32 off, v47, s32 offset:32 -; GCN-GISEL-NEXT: scratch_store_b32 off, v56, s32 offset:28 -; GCN-GISEL-NEXT: scratch_store_b32 off, v57, s32 offset:24 -; GCN-GISEL-NEXT: scratch_store_b32 off, v58, s32 offset:20 -; GCN-GISEL-NEXT: scratch_store_b32 off, v59, s32 offset:16 -; GCN-GISEL-NEXT: scratch_store_b32 off, v60, s32 offset:12 -; GCN-GISEL-NEXT: scratch_store_b32 off, v61, s32 offset:8 -; GCN-GISEL-NEXT: scratch_store_b32 off, v62, s32 offset:4 -; GCN-GISEL-NEXT: scratch_store_b32 off, v63, s32 +; GCN-GISEL-NEXT: scratch_store_b32 off, v40, s32 offset:60 scope:SCOPE_SE +; GCN-GISEL-NEXT: scratch_store_b32 off, v41, s32 offset:56 scope:SCOPE_SE +; GCN-GISEL-NEXT: scratch_store_b32 off, v42, s32 offset:52 scope:SCOPE_SE +; GCN-GISEL-NEXT: scratch_store_b32 off, v43, s32 offset:48 scope:SCOPE_SE +; GCN-GISEL-NEXT: scratch_store_b32 off, v44, s32 offset:44 scope:SCOPE_SE +; GCN-GISEL-NEXT: scratch_store_b32 off, v45, s32 offset:40 scope:SCOPE_SE +; GCN-GISEL-NEXT: scratch_store_b32 off, v46, s32 offset:36 scope:SCOPE_SE +; GCN-GISEL-NEXT: scratch_store_b32 off, v47, s32 offset:32 scope:SCOPE_SE +; GCN-GISEL-NEXT: scratch_store_b32 off, v56, s32 offset:28 scope:SCOPE_SE +; GCN-GISEL-NEXT: scratch_store_b32 off, v57, s32 offset:24 scope:SCOPE_SE +; GCN-GISEL-NEXT: scratch_store_b32 off, v58, s32 offset:20 scope:SCOPE_SE +; GCN-GISEL-NEXT: scratch_store_b32 off, v59, s32 offset:16 scope:SCOPE_SE +; GCN-GISEL-NEXT: scratch_store_b32 off, v60, s32 offset:12 scope:SCOPE_SE +; GCN-GISEL-NEXT: scratch_store_b32 off, v61, s32 offset:8 scope:SCOPE_SE +; GCN-GISEL-NEXT: scratch_store_b32 off, v62, s32 offset:4 scope:SCOPE_SE +; GCN-GISEL-NEXT: scratch_store_b32 off, v63, s32 scope:SCOPE_SE ; GCN-GISEL-NEXT: s_wait_xcnt 0x8 ; GCN-GISEL-NEXT: v_dual_mov_b32 v46, v3 :: v_dual_mov_b32 v47, v4 ; GCN-GISEL-NEXT: global_load_b128 v[2:5], v[0:1], off offset:32 ; GCN-GISEL-NEXT: s_wait_loadcnt 0x0 -; GCN-GISEL-NEXT: scratch_store_b128 off, v[2:5], s32 offset:80 ; 16-byte Folded Spill +; GCN-GISEL-NEXT: scratch_store_b128 off, v[2:5], s32 offset:80 scope:SCOPE_SE ; 16-byte Folded Spill ; GCN-GISEL-NEXT: s_clause 0xe ; GCN-GISEL-NEXT: global_load_b128 v[6:9], v[0:1], off offset:48 ; GCN-GISEL-NEXT: global_load_b128 v[10:13], v[0:1], off offset:64 @@ -244,7 +244,7 @@ define i32 @test_v64i32_load_store(ptr addrspace(1) %ptr, i32 %idx, ptr addrspac ; GCN-GISEL-NEXT: global_load_b128 v[60:63], v[0:1], off offset:16 ; GCN-GISEL-NEXT: global_load_b128 v[0:3], v[0:1], off offset:240 ; GCN-GISEL-NEXT: s_wait_loadcnt 0x0 -; GCN-GISEL-NEXT: scratch_store_b128 off, v[0:3], s32 offset:64 ; 16-byte Folded Spill +; GCN-GISEL-NEXT: scratch_store_b128 off, v[0:3], s32 offset:64 scope:SCOPE_SE ; 16-byte Folded Spill ; GCN-GISEL-NEXT: scratch_load_b128 v[0:3], off, s32 offset:80 th:TH_LOAD_LU ; 16-byte Folded Reload ; GCN-GISEL-NEXT: s_wait_loadcnt 0x0 ; GCN-GISEL-NEXT: s_clause 0xe @@ -299,10 +299,10 @@ define i64 @test_v16i64_load_store(ptr addrspace(1) %ptr_a, ptr addrspace(1) %pt ; GCN-SDAG-NEXT: s_wait_loadcnt_dscnt 0x0 ; GCN-SDAG-NEXT: s_wait_kmcnt 0x0 ; GCN-SDAG-NEXT: s_clause 0x3 -; GCN-SDAG-NEXT: scratch_store_b32 off, v40, s32 offset:12 -; GCN-SDAG-NEXT: scratch_store_b32 off, v41, s32 offset:8 -; GCN-SDAG-NEXT: scratch_store_b32 off, v42, s32 offset:4 -; GCN-SDAG-NEXT: scratch_store_b32 off, v43, s32 +; GCN-SDAG-NEXT: scratch_store_b32 off, v40, s32 offset:12 scope:SCOPE_SE +; GCN-SDAG-NEXT: scratch_store_b32 off, v41, s32 offset:8 scope:SCOPE_SE +; GCN-SDAG-NEXT: scratch_store_b32 off, v42, s32 offset:4 scope:SCOPE_SE +; GCN-SDAG-NEXT: scratch_store_b32 off, v43, s32 scope:SCOPE_SE ; GCN-SDAG-NEXT: s_clause 0x7 ; GCN-SDAG-NEXT: global_load_b128 v[10:13], v[0:1], off offset:112 ; GCN-SDAG-NEXT: global_load_b128 v[18:21], v[0:1], off offset:96 @@ -385,12 +385,12 @@ define i64 @test_v16i64_load_store(ptr addrspace(1) %ptr_a, ptr addrspace(1) %pt ; GCN-GISEL-NEXT: s_wait_loadcnt_dscnt 0x0 ; GCN-GISEL-NEXT: s_wait_kmcnt 0x0 ; GCN-GISEL-NEXT: s_clause 0x5 -; GCN-GISEL-NEXT: scratch_store_b32 off, v40, s32 offset:20 -; GCN-GISEL-NEXT: scratch_store_b32 off, v41, s32 offset:16 -; GCN-GISEL-NEXT: scratch_store_b32 off, v42, s32 offset:12 -; GCN-GISEL-NEXT: scratch_store_b32 off, v43, s32 offset:8 -; GCN-GISEL-NEXT: scratch_store_b32 off, v44, s32 offset:4 -; GCN-GISEL-NEXT: scratch_store_b32 off, v45, s32 +; GCN-GISEL-NEXT: scratch_store_b32 off, v40, s32 offset:20 scope:SCOPE_SE +; GCN-GISEL-NEXT: scratch_store_b32 off, v41, s32 offset:16 scope:SCOPE_SE +; GCN-GISEL-NEXT: scratch_store_b32 off, v42, s32 offset:12 scope:SCOPE_SE +; GCN-GISEL-NEXT: scratch_store_b32 off, v43, s32 offset:8 scope:SCOPE_SE +; GCN-GISEL-NEXT: scratch_store_b32 off, v44, s32 offset:4 scope:SCOPE_SE +; GCN-GISEL-NEXT: scratch_store_b32 off, v45, s32 scope:SCOPE_SE ; GCN-GISEL-NEXT: s_clause 0x7 ; GCN-GISEL-NEXT: global_load_b128 v[6:9], v[0:1], off offset:80 ; GCN-GISEL-NEXT: global_load_b128 v[10:13], v[0:1], off diff --git a/llvm/test/CodeGen/AMDGPU/llc-pipeline-npm.ll b/llvm/test/CodeGen/AMDGPU/llc-pipeline-npm.ll index ceed41f..4f81d35 100644 --- a/llvm/test/CodeGen/AMDGPU/llc-pipeline-npm.ll +++ b/llvm/test/CodeGen/AMDGPU/llc-pipeline-npm.ll @@ -10,9 +10,9 @@ ; GCN-O0: require<MachineModuleAnalysis>,require<profile-summary>,require<collector-metadata>,pre-isel-intrinsic-lowering,function(expand-large-div-rem,expand-fp),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,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,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),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),amdgpu-lower-buffer-fat-pointers,cgscc(function(codegenprepare,load-store-vectorizer,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-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),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,verify,loop-mssa(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),amdgpu-lower-buffer-fat-pointers,cgscc(function(codegenprepare,load-store-vectorizer,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>,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-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),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),amdgpu-lower-buffer-fat-pointers,cgscc(function(codegenprepare,load-store-vectorizer,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-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),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,verify,loop-mssa(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),amdgpu-lower-buffer-fat-pointers,cgscc(function(codegenprepare,load-store-vectorizer,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>,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-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/llvm.amdgcn.mfma.gfx942.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx942.ll index 7d85d34..beda16c 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx942.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx942.ll @@ -1,13 +1,9 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 ; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck --check-prefixes=GFX942,GFX942-VGPRCD,GFX942-SDAG,GFX942-VGPRCD-SDAG %s ; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck --check-prefixes=GFX942,GFX942-VGPRCD,GFX942-GISEL,GFX942-VGPRCD-GISEL %s -; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx942 -stress-regalloc=10 < %s | FileCheck --check-prefixes=GFX942,GFX942-AGPRCD,GFX942-SDAG,GFX942-AGPRCD-SDAG %s -; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx942 -stress-regalloc=10 < %s | FileCheck --check-prefixes=GFX942,GFX942-AGPRCD,GFX942-GISEL,GFX942-AGPRCD-GISEL %s ; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx950 < %s | FileCheck --check-prefixes=GFX950,GFX950-VGPRCD,GFX950-SDAG,GFX950-VGPRCD-SDAG %s ; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx950 < %s | FileCheck --check-prefixes=GFX950,GFX950-VGPRCD,GFX950-GISEL,GFX950-VGPRCD-GISEL %s -; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx950 -stress-regalloc=10 < %s | FileCheck --check-prefixes=GFX950,GFX950-AGPRCD,GFX950-SDAG,GFX950-AGPRCD-SDAG %s -; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx950 -stress-regalloc=10 < %s | FileCheck --check-prefixes=GFX950,GFX950-AGPRCD,GFX950-GISEL,GFX950-AGPRCD-GISEL %s declare <4 x i32> @llvm.amdgcn.mfma.i32.16x16x32.i8(i64, i64, <4 x i32>, i32, i32, i32) declare <16 x i32> @llvm.amdgcn.mfma.i32.32x32x16.i8(i64, i64, <16 x i32>, i32, i32, i32) diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.bf16.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.bf16.ll index d358837..8081a15 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.bf16.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.bf16.ll @@ -252,62 +252,55 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_bf16__vgprcd(<8 x bfloat> %arg ; GCN-NEXT: s_load_dwordx8 s[24:31], s[4:5], 0x24 ; GCN-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 -; GCN-NEXT: v_mov_b32_e32 v8, 0 +; GCN-NEXT: v_mov_b32_e32 v44, 0 ; GCN-NEXT: s_waitcnt lgkmcnt(0) -; GCN-NEXT: v_mov_b64_e32 v[0:1], s[24:25] -; GCN-NEXT: v_mov_b64_e32 v[2:3], s[26:27] -; GCN-NEXT: v_mov_b64_e32 v[4:5], s[28:29] -; GCN-NEXT: v_accvgpr_write_b32 a31, s23 -; GCN-NEXT: v_mov_b64_e32 v[6:7], s[30:31] -; GCN-NEXT: v_accvgpr_write_b32 a30, s22 -; GCN-NEXT: v_accvgpr_write_b32 a29, s21 -; GCN-NEXT: v_accvgpr_write_b32 a28, s20 -; GCN-NEXT: v_accvgpr_write_b32 a27, s19 -; GCN-NEXT: v_accvgpr_write_b32 a26, s18 -; GCN-NEXT: v_accvgpr_write_b32 a25, s17 -; GCN-NEXT: v_accvgpr_write_b32 a24, s16 -; GCN-NEXT: v_accvgpr_write_b32 a23, s15 -; GCN-NEXT: v_accvgpr_write_b32 a22, s14 -; GCN-NEXT: v_accvgpr_write_b32 a21, s13 -; GCN-NEXT: v_accvgpr_write_b32 a20, s12 -; GCN-NEXT: v_accvgpr_write_b32 a19, s11 -; GCN-NEXT: v_accvgpr_write_b32 a18, s10 -; GCN-NEXT: v_accvgpr_write_b32 a17, s9 -; GCN-NEXT: v_accvgpr_write_b32 a16, s8 -; GCN-NEXT: v_mov_b32_e32 v10, s20 -; GCN-NEXT: v_mov_b32_e32 v11, s21 -; GCN-NEXT: v_mfma_f32_32x32x16_bf16 a[0:15], v[0:3], v[4:7], a[16:31] -; GCN-NEXT: v_mov_b32_e32 v12, s22 -; GCN-NEXT: v_mov_b32_e32 v13, s23 -; GCN-NEXT: v_mov_b32_e32 v0, s16 -; GCN-NEXT: v_mov_b32_e32 v1, s17 -; GCN-NEXT: v_mov_b32_e32 v2, s18 -; GCN-NEXT: v_mov_b32_e32 v3, s19 -; GCN-NEXT: global_store_dwordx4 v8, v[10:13], s[0:1] offset:48 sc0 sc1 -; GCN-NEXT: s_waitcnt vmcnt(0) -; GCN-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1] offset:32 sc0 sc1 +; GCN-NEXT: v_mov_b64_e32 v[34:35], s[26:27] +; GCN-NEXT: v_mov_b64_e32 v[32:33], s[24:25] +; GCN-NEXT: v_mov_b64_e32 v[38:39], s[30:31] +; GCN-NEXT: v_mov_b64_e32 v[30:31], s[22:23] +; GCN-NEXT: v_mov_b64_e32 v[36:37], s[28:29] +; GCN-NEXT: v_mov_b64_e32 v[28:29], s[20:21] +; GCN-NEXT: v_mov_b64_e32 v[26:27], s[18:19] +; GCN-NEXT: v_mov_b64_e32 v[24:25], s[16:17] +; GCN-NEXT: v_mov_b64_e32 v[22:23], s[14:15] +; GCN-NEXT: v_mov_b64_e32 v[20:21], s[12:13] +; GCN-NEXT: v_mov_b64_e32 v[18:19], s[10:11] +; GCN-NEXT: v_mov_b64_e32 v[16:17], s[8:9] +; GCN-NEXT: v_mov_b32_e32 v40, s20 +; GCN-NEXT: v_mov_b32_e32 v41, s21 +; GCN-NEXT: v_mfma_f32_32x32x16_bf16 v[0:15], v[32:35], v[36:39], v[16:31] +; GCN-NEXT: v_mov_b32_e32 v42, s22 +; GCN-NEXT: v_mov_b32_e32 v43, s23 +; GCN-NEXT: global_store_dwordx4 v44, v[40:43], s[0:1] offset:48 sc0 sc1 +; GCN-NEXT: s_waitcnt vmcnt(0) +; GCN-NEXT: s_nop 2 +; GCN-NEXT: v_mov_b32_e32 v16, s16 +; GCN-NEXT: v_mov_b32_e32 v17, s17 +; GCN-NEXT: v_mov_b32_e32 v18, s18 +; GCN-NEXT: v_mov_b32_e32 v19, s19 +; GCN-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] offset:32 sc0 sc1 ; GCN-NEXT: s_waitcnt vmcnt(0) ; GCN-NEXT: s_nop 0 -; GCN-NEXT: v_mov_b32_e32 v0, s12 -; GCN-NEXT: v_mov_b32_e32 v1, s13 -; GCN-NEXT: v_mov_b32_e32 v2, s14 -; GCN-NEXT: v_mov_b32_e32 v3, s15 -; GCN-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1] offset:16 sc0 sc1 +; GCN-NEXT: v_mov_b32_e32 v16, s12 +; GCN-NEXT: v_mov_b32_e32 v17, s13 +; GCN-NEXT: v_mov_b32_e32 v18, s14 +; GCN-NEXT: v_mov_b32_e32 v19, s15 +; GCN-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] offset:16 sc0 sc1 ; GCN-NEXT: s_waitcnt vmcnt(0) ; GCN-NEXT: s_nop 0 -; GCN-NEXT: v_mov_b32_e32 v0, s8 -; GCN-NEXT: v_mov_b32_e32 v1, s9 -; GCN-NEXT: v_mov_b32_e32 v2, s10 -; GCN-NEXT: v_mov_b32_e32 v3, s11 -; GCN-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1] sc0 sc1 +; GCN-NEXT: v_mov_b32_e32 v16, s8 +; GCN-NEXT: v_mov_b32_e32 v17, s9 +; GCN-NEXT: v_mov_b32_e32 v18, s10 +; GCN-NEXT: v_mov_b32_e32 v19, s11 +; GCN-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] sc0 sc1 ; GCN-NEXT: s_waitcnt vmcnt(0) -; GCN-NEXT: global_store_dwordx4 v8, a[8:11], s[0:1] offset:32 sc0 sc1 +; GCN-NEXT: global_store_dwordx4 v44, v[8:11], s[0:1] offset:32 sc0 sc1 ; GCN-NEXT: s_waitcnt vmcnt(0) -; GCN-NEXT: global_store_dwordx4 v8, a[12:15], s[0:1] offset:48 sc0 sc1 +; GCN-NEXT: global_store_dwordx4 v44, v[12:15], s[0:1] offset:48 sc0 sc1 ; GCN-NEXT: s_waitcnt vmcnt(0) -; GCN-NEXT: global_store_dwordx4 v8, a[0:3], s[0:1] sc0 sc1 +; GCN-NEXT: global_store_dwordx4 v44, v[0:3], s[0:1] sc0 sc1 ; GCN-NEXT: s_waitcnt vmcnt(0) -; GCN-NEXT: global_store_dwordx4 v8, a[4:7], s[0:1] offset:16 sc0 sc1 +; GCN-NEXT: global_store_dwordx4 v44, v[4:7], s[0:1] offset:16 sc0 sc1 ; GCN-NEXT: s_waitcnt vmcnt(0) ; GCN-NEXT: s_endpgm %result = call <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.bf16(<8 x bfloat> %arg0, <8 x bfloat> %arg1, <16 x float> %arg2, i32 0, i32 0, i32 0) @@ -322,62 +315,55 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_bf16__vgprcd__flags(<8 x bfloa ; GCN-NEXT: s_load_dwordx8 s[24:31], s[4:5], 0x24 ; GCN-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 -; GCN-NEXT: v_mov_b32_e32 v8, 0 +; GCN-NEXT: v_mov_b32_e32 v44, 0 ; GCN-NEXT: s_waitcnt lgkmcnt(0) -; GCN-NEXT: v_mov_b64_e32 v[0:1], s[24:25] -; GCN-NEXT: v_mov_b64_e32 v[2:3], s[26:27] -; GCN-NEXT: v_mov_b64_e32 v[4:5], s[28:29] -; GCN-NEXT: v_accvgpr_write_b32 a31, s23 -; GCN-NEXT: v_mov_b64_e32 v[6:7], s[30:31] -; GCN-NEXT: v_accvgpr_write_b32 a30, s22 -; GCN-NEXT: v_accvgpr_write_b32 a29, s21 -; GCN-NEXT: v_accvgpr_write_b32 a28, s20 -; GCN-NEXT: v_accvgpr_write_b32 a27, s19 -; GCN-NEXT: v_accvgpr_write_b32 a26, s18 -; GCN-NEXT: v_accvgpr_write_b32 a25, s17 -; GCN-NEXT: v_accvgpr_write_b32 a24, s16 -; GCN-NEXT: v_accvgpr_write_b32 a23, s15 -; GCN-NEXT: v_accvgpr_write_b32 a22, s14 -; GCN-NEXT: v_accvgpr_write_b32 a21, s13 -; GCN-NEXT: v_accvgpr_write_b32 a20, s12 -; GCN-NEXT: v_accvgpr_write_b32 a19, s11 -; GCN-NEXT: v_accvgpr_write_b32 a18, s10 -; GCN-NEXT: v_accvgpr_write_b32 a17, s9 -; GCN-NEXT: v_accvgpr_write_b32 a16, s8 -; GCN-NEXT: v_mov_b32_e32 v10, s20 -; GCN-NEXT: v_mov_b32_e32 v11, s21 -; GCN-NEXT: v_mfma_f32_32x32x16_bf16 a[0:15], v[0:3], v[4:7], a[16:31] cbsz:1 abid:2 blgp:3 -; GCN-NEXT: v_mov_b32_e32 v12, s22 -; GCN-NEXT: v_mov_b32_e32 v13, s23 -; GCN-NEXT: v_mov_b32_e32 v0, s16 -; GCN-NEXT: v_mov_b32_e32 v1, s17 -; GCN-NEXT: v_mov_b32_e32 v2, s18 -; GCN-NEXT: v_mov_b32_e32 v3, s19 -; GCN-NEXT: global_store_dwordx4 v8, v[10:13], s[0:1] offset:48 sc0 sc1 -; GCN-NEXT: s_waitcnt vmcnt(0) -; GCN-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1] offset:32 sc0 sc1 +; GCN-NEXT: v_mov_b64_e32 v[34:35], s[26:27] +; GCN-NEXT: v_mov_b64_e32 v[32:33], s[24:25] +; GCN-NEXT: v_mov_b64_e32 v[38:39], s[30:31] +; GCN-NEXT: v_mov_b64_e32 v[30:31], s[22:23] +; GCN-NEXT: v_mov_b64_e32 v[36:37], s[28:29] +; GCN-NEXT: v_mov_b64_e32 v[28:29], s[20:21] +; GCN-NEXT: v_mov_b64_e32 v[26:27], s[18:19] +; GCN-NEXT: v_mov_b64_e32 v[24:25], s[16:17] +; GCN-NEXT: v_mov_b64_e32 v[22:23], s[14:15] +; GCN-NEXT: v_mov_b64_e32 v[20:21], s[12:13] +; GCN-NEXT: v_mov_b64_e32 v[18:19], s[10:11] +; GCN-NEXT: v_mov_b64_e32 v[16:17], s[8:9] +; GCN-NEXT: v_mov_b32_e32 v40, s20 +; GCN-NEXT: v_mov_b32_e32 v41, s21 +; GCN-NEXT: v_mfma_f32_32x32x16_bf16 v[0:15], v[32:35], v[36:39], v[16:31] cbsz:1 abid:2 blgp:3 +; GCN-NEXT: v_mov_b32_e32 v42, s22 +; GCN-NEXT: v_mov_b32_e32 v43, s23 +; GCN-NEXT: global_store_dwordx4 v44, v[40:43], s[0:1] offset:48 sc0 sc1 +; GCN-NEXT: s_waitcnt vmcnt(0) +; GCN-NEXT: s_nop 2 +; GCN-NEXT: v_mov_b32_e32 v16, s16 +; GCN-NEXT: v_mov_b32_e32 v17, s17 +; GCN-NEXT: v_mov_b32_e32 v18, s18 +; GCN-NEXT: v_mov_b32_e32 v19, s19 +; GCN-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] offset:32 sc0 sc1 ; GCN-NEXT: s_waitcnt vmcnt(0) ; GCN-NEXT: s_nop 0 -; GCN-NEXT: v_mov_b32_e32 v0, s12 -; GCN-NEXT: v_mov_b32_e32 v1, s13 -; GCN-NEXT: v_mov_b32_e32 v2, s14 -; GCN-NEXT: v_mov_b32_e32 v3, s15 -; GCN-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1] offset:16 sc0 sc1 +; GCN-NEXT: v_mov_b32_e32 v16, s12 +; GCN-NEXT: v_mov_b32_e32 v17, s13 +; GCN-NEXT: v_mov_b32_e32 v18, s14 +; GCN-NEXT: v_mov_b32_e32 v19, s15 +; GCN-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] offset:16 sc0 sc1 ; GCN-NEXT: s_waitcnt vmcnt(0) ; GCN-NEXT: s_nop 0 -; GCN-NEXT: v_mov_b32_e32 v0, s8 -; GCN-NEXT: v_mov_b32_e32 v1, s9 -; GCN-NEXT: v_mov_b32_e32 v2, s10 -; GCN-NEXT: v_mov_b32_e32 v3, s11 -; GCN-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1] sc0 sc1 +; GCN-NEXT: v_mov_b32_e32 v16, s8 +; GCN-NEXT: v_mov_b32_e32 v17, s9 +; GCN-NEXT: v_mov_b32_e32 v18, s10 +; GCN-NEXT: v_mov_b32_e32 v19, s11 +; GCN-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] sc0 sc1 ; GCN-NEXT: s_waitcnt vmcnt(0) -; GCN-NEXT: global_store_dwordx4 v8, a[8:11], s[0:1] offset:32 sc0 sc1 +; GCN-NEXT: global_store_dwordx4 v44, v[8:11], s[0:1] offset:32 sc0 sc1 ; GCN-NEXT: s_waitcnt vmcnt(0) -; GCN-NEXT: global_store_dwordx4 v8, a[12:15], s[0:1] offset:48 sc0 sc1 +; GCN-NEXT: global_store_dwordx4 v44, v[12:15], s[0:1] offset:48 sc0 sc1 ; GCN-NEXT: s_waitcnt vmcnt(0) -; GCN-NEXT: global_store_dwordx4 v8, a[0:3], s[0:1] sc0 sc1 +; GCN-NEXT: global_store_dwordx4 v44, v[0:3], s[0:1] sc0 sc1 ; GCN-NEXT: s_waitcnt vmcnt(0) -; GCN-NEXT: global_store_dwordx4 v8, a[4:7], s[0:1] offset:16 sc0 sc1 +; GCN-NEXT: global_store_dwordx4 v44, v[4:7], s[0:1] offset:16 sc0 sc1 ; GCN-NEXT: s_waitcnt vmcnt(0) ; GCN-NEXT: s_endpgm %result = call <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.bf16(<8 x bfloat> %arg0, <8 x bfloat> %arg1, <16 x float> %arg2, i32 1, i32 2, i32 3) @@ -393,35 +379,27 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_bf16__vgprcd_mac(<8 x bfloat> ; GCN-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 ; GCN-NEXT: s_waitcnt lgkmcnt(0) -; GCN-NEXT: v_mov_b64_e32 v[0:1], s[24:25] -; GCN-NEXT: v_mov_b64_e32 v[2:3], s[26:27] -; GCN-NEXT: v_mov_b64_e32 v[4:5], s[28:29] -; GCN-NEXT: v_accvgpr_write_b32 a0, s8 -; GCN-NEXT: v_mov_b64_e32 v[6:7], s[30:31] -; GCN-NEXT: v_accvgpr_write_b32 a1, s9 -; GCN-NEXT: v_accvgpr_write_b32 a2, s10 -; GCN-NEXT: v_accvgpr_write_b32 a3, s11 -; GCN-NEXT: v_accvgpr_write_b32 a4, s12 -; GCN-NEXT: v_accvgpr_write_b32 a5, s13 -; GCN-NEXT: v_accvgpr_write_b32 a6, s14 -; GCN-NEXT: v_accvgpr_write_b32 a7, s15 -; GCN-NEXT: v_accvgpr_write_b32 a8, s16 -; GCN-NEXT: v_accvgpr_write_b32 a9, s17 -; GCN-NEXT: v_accvgpr_write_b32 a10, s18 -; GCN-NEXT: v_accvgpr_write_b32 a11, s19 -; GCN-NEXT: v_accvgpr_write_b32 a12, s20 -; GCN-NEXT: v_accvgpr_write_b32 a13, s21 -; GCN-NEXT: v_accvgpr_write_b32 a14, s22 -; GCN-NEXT: v_accvgpr_write_b32 a15, s23 +; GCN-NEXT: v_mov_b64_e32 v[16:17], s[24:25] +; GCN-NEXT: v_mov_b64_e32 v[18:19], s[26:27] +; GCN-NEXT: v_mov_b64_e32 v[20:21], s[28:29] +; GCN-NEXT: v_mov_b64_e32 v[0:1], s[8:9] +; GCN-NEXT: v_mov_b64_e32 v[22:23], s[30:31] +; GCN-NEXT: v_mov_b64_e32 v[2:3], s[10:11] +; GCN-NEXT: v_mov_b64_e32 v[4:5], s[12:13] +; GCN-NEXT: v_mov_b64_e32 v[6:7], s[14:15] +; GCN-NEXT: v_mov_b64_e32 v[8:9], s[16:17] +; GCN-NEXT: v_mov_b64_e32 v[10:11], s[18:19] +; GCN-NEXT: v_mov_b64_e32 v[12:13], s[20:21] +; GCN-NEXT: v_mov_b64_e32 v[14:15], s[22:23] ; GCN-NEXT: s_nop 1 -; GCN-NEXT: v_mfma_f32_32x32x16_bf16 a[0:15], v[0:3], v[4:7], a[0:15] -; GCN-NEXT: v_mov_b32_e32 v0, 0 +; GCN-NEXT: v_mfma_f32_32x32x16_bf16 v[0:15], v[16:19], v[20:23], v[0:15] +; GCN-NEXT: v_mov_b32_e32 v16, 0 ; GCN-NEXT: s_nop 7 ; GCN-NEXT: s_nop 2 -; GCN-NEXT: global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 -; GCN-NEXT: global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 -; GCN-NEXT: global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 -; GCN-NEXT: global_store_dwordx4 v0, a[0:3], s[0:1] +; GCN-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1] offset:48 +; GCN-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1] offset:32 +; GCN-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1] offset:16 +; GCN-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1] ; GCN-NEXT: s_endpgm %result = call <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.bf16(<8 x bfloat> %arg0, <8 x bfloat> %arg1, <16 x float> %arg2, i32 0, i32 0, i32 0) store <16 x float> %result, ptr addrspace(1) %out @@ -435,40 +413,32 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_bf16__vgprcd_mac_flags(<8 x bf ; GCN-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 ; GCN-NEXT: s_waitcnt lgkmcnt(0) -; GCN-NEXT: v_mov_b64_e32 v[0:1], s[24:25] -; GCN-NEXT: v_mov_b64_e32 v[2:3], s[26:27] -; GCN-NEXT: v_mov_b64_e32 v[4:5], s[28:29] -; GCN-NEXT: v_accvgpr_write_b32 a0, s8 -; GCN-NEXT: v_mov_b64_e32 v[6:7], s[30:31] -; GCN-NEXT: v_accvgpr_write_b32 a1, s9 -; GCN-NEXT: v_accvgpr_write_b32 a2, s10 -; GCN-NEXT: v_accvgpr_write_b32 a3, s11 -; GCN-NEXT: v_accvgpr_write_b32 a4, s12 -; GCN-NEXT: v_accvgpr_write_b32 a5, s13 -; GCN-NEXT: v_accvgpr_write_b32 a6, s14 -; GCN-NEXT: v_accvgpr_write_b32 a7, s15 -; GCN-NEXT: v_accvgpr_write_b32 a8, s16 -; GCN-NEXT: v_accvgpr_write_b32 a9, s17 -; GCN-NEXT: v_accvgpr_write_b32 a10, s18 -; GCN-NEXT: v_accvgpr_write_b32 a11, s19 -; GCN-NEXT: v_accvgpr_write_b32 a12, s20 -; GCN-NEXT: v_accvgpr_write_b32 a13, s21 -; GCN-NEXT: v_accvgpr_write_b32 a14, s22 -; GCN-NEXT: v_accvgpr_write_b32 a15, s23 +; GCN-NEXT: v_mov_b64_e32 v[16:17], s[24:25] +; GCN-NEXT: v_mov_b64_e32 v[18:19], s[26:27] +; GCN-NEXT: v_mov_b64_e32 v[20:21], s[28:29] +; GCN-NEXT: v_mov_b64_e32 v[0:1], s[8:9] +; GCN-NEXT: v_mov_b64_e32 v[22:23], s[30:31] +; GCN-NEXT: v_mov_b64_e32 v[2:3], s[10:11] +; GCN-NEXT: v_mov_b64_e32 v[4:5], s[12:13] +; GCN-NEXT: v_mov_b64_e32 v[6:7], s[14:15] +; GCN-NEXT: v_mov_b64_e32 v[8:9], s[16:17] +; GCN-NEXT: v_mov_b64_e32 v[10:11], s[18:19] +; GCN-NEXT: v_mov_b64_e32 v[12:13], s[20:21] +; GCN-NEXT: v_mov_b64_e32 v[14:15], s[22:23] ; GCN-NEXT: s_nop 1 -; GCN-NEXT: v_mfma_f32_32x32x16_bf16 a[0:15], v[0:3], v[4:7], a[0:15] cbsz:3 abid:2 blgp:1 -; GCN-NEXT: v_mov_b32_e32 v0, 0 +; GCN-NEXT: v_mfma_f32_32x32x16_bf16 v[0:15], v[16:19], v[20:23], v[0:15] cbsz:3 abid:2 blgp:1 +; GCN-NEXT: v_mov_b32_e32 v16, 0 ; GCN-NEXT: s_nop 7 ; GCN-NEXT: s_nop 2 -; GCN-NEXT: global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 -; GCN-NEXT: global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 -; GCN-NEXT: global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 -; GCN-NEXT: global_store_dwordx4 v0, a[0:3], s[0:1] +; GCN-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1] offset:48 +; GCN-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1] offset:32 +; GCN-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1] offset:16 +; GCN-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1] ; GCN-NEXT: s_endpgm %result = call <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.bf16(<8 x bfloat> %arg0, <8 x bfloat> %arg1, <16 x float> %arg2, i32 3, i32 2, i32 1) store <16 x float> %result, ptr addrspace(1) %out ret void } -attributes #0 = { "amdgpu-flat-work-group-size"="512,512" } +attributes #0 = { "amdgpu-flat-work-group-size"="512,512" "amdgpu-agpr-alloc"="0,0" } attributes #1 = { "amdgpu-flat-work-group-size"="1,64" } diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.ll index 21465be..d81ec1c 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.ll @@ -141,20 +141,18 @@ define amdgpu_kernel void @test_mfma_f32_16x16x32_f16_no_agpr__vgprcd(ptr addrsp ; SDAG-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x34 ; SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 ; SDAG-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x24 -; SDAG-NEXT: v_mov_b32_e32 v8, 0 +; SDAG-NEXT: v_mov_b32_e32 v12, 0 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) ; SDAG-NEXT: v_mov_b64_e32 v[0:1], s[8:9] ; SDAG-NEXT: v_mov_b64_e32 v[2:3], s[10:11] ; SDAG-NEXT: v_mov_b64_e32 v[4:5], s[12:13] -; SDAG-NEXT: v_accvgpr_write_b32 a0, s0 +; SDAG-NEXT: v_mov_b64_e32 v[10:11], s[2:3] ; SDAG-NEXT: v_mov_b64_e32 v[6:7], s[14:15] -; SDAG-NEXT: v_accvgpr_write_b32 a1, s1 -; SDAG-NEXT: v_accvgpr_write_b32 a2, s2 -; SDAG-NEXT: v_accvgpr_write_b32 a3, s3 +; SDAG-NEXT: v_mov_b64_e32 v[8:9], s[0:1] ; SDAG-NEXT: s_nop 1 -; SDAG-NEXT: v_mfma_f32_16x16x32_f16 a[0:3], v[0:3], v[4:7], a[0:3] +; SDAG-NEXT: v_mfma_f32_16x16x32_f16 v[0:3], v[0:3], v[4:7], v[8:11] ; SDAG-NEXT: s_nop 7 -; SDAG-NEXT: global_store_dwordx4 v8, a[0:3], s[6:7] +; SDAG-NEXT: global_store_dwordx4 v12, v[0:3], s[6:7] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_mfma_f32_16x16x32_f16_no_agpr__vgprcd: @@ -166,16 +164,14 @@ define amdgpu_kernel void @test_mfma_f32_16x16x32_f16_no_agpr__vgprcd(ptr addrsp ; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[8:9] ; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[10:11] ; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] -; GISEL-NEXT: v_accvgpr_write_b32 a0, s0 +; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[2:3] ; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] -; GISEL-NEXT: v_accvgpr_write_b32 a1, s1 -; GISEL-NEXT: v_accvgpr_write_b32 a2, s2 -; GISEL-NEXT: v_accvgpr_write_b32 a3, s3 +; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[0:1] ; GISEL-NEXT: s_nop 1 -; GISEL-NEXT: v_mfma_f32_16x16x32_f16 a[0:3], v[0:3], v[4:7], a[0:3] -; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: v_mfma_f32_16x16x32_f16 v[0:3], v[0:3], v[4:7], v[8:11] +; GISEL-NEXT: v_mov_b32_e32 v4, 0 ; GISEL-NEXT: s_nop 6 -; GISEL-NEXT: global_store_dwordx4 v0, a[0:3], s[6:7] +; GISEL-NEXT: global_store_dwordx4 v4, v[0:3], s[6:7] ; GISEL-NEXT: s_endpgm ; ; HEURRC-LABEL: test_mfma_f32_16x16x32_f16_no_agpr__vgprcd: @@ -183,20 +179,18 @@ define amdgpu_kernel void @test_mfma_f32_16x16x32_f16_no_agpr__vgprcd(ptr addrsp ; HEURRC-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x34 ; HEURRC-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 ; HEURRC-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x24 -; HEURRC-NEXT: v_mov_b32_e32 v8, 0 +; HEURRC-NEXT: v_mov_b32_e32 v12, 0 ; HEURRC-NEXT: s_waitcnt lgkmcnt(0) ; HEURRC-NEXT: v_mov_b64_e32 v[0:1], s[8:9] ; HEURRC-NEXT: v_mov_b64_e32 v[2:3], s[10:11] ; HEURRC-NEXT: v_mov_b64_e32 v[4:5], s[12:13] -; HEURRC-NEXT: v_accvgpr_write_b32 a0, s0 +; HEURRC-NEXT: v_mov_b64_e32 v[10:11], s[2:3] ; HEURRC-NEXT: v_mov_b64_e32 v[6:7], s[14:15] -; HEURRC-NEXT: v_accvgpr_write_b32 a1, s1 -; HEURRC-NEXT: v_accvgpr_write_b32 a2, s2 -; HEURRC-NEXT: v_accvgpr_write_b32 a3, s3 +; HEURRC-NEXT: v_mov_b64_e32 v[8:9], s[0:1] ; HEURRC-NEXT: s_nop 1 -; HEURRC-NEXT: v_mfma_f32_16x16x32_f16 a[0:3], v[0:3], v[4:7], a[0:3] +; HEURRC-NEXT: v_mfma_f32_16x16x32_f16 v[0:3], v[0:3], v[4:7], v[8:11] ; HEURRC-NEXT: s_nop 7 -; HEURRC-NEXT: global_store_dwordx4 v8, a[0:3], s[6:7] +; HEURRC-NEXT: global_store_dwordx4 v12, v[0:3], s[6:7] ; HEURRC-NEXT: s_endpgm ; ; VGPRRC-LABEL: test_mfma_f32_16x16x32_f16_no_agpr__vgprcd: @@ -266,20 +260,18 @@ define amdgpu_kernel void @test_mfma_f32_16x16x32_f16_no_agpr__vgprcd__flags(ptr ; SDAG-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x34 ; SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 ; SDAG-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x24 -; SDAG-NEXT: v_mov_b32_e32 v8, 0 +; SDAG-NEXT: v_mov_b32_e32 v12, 0 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) ; SDAG-NEXT: v_mov_b64_e32 v[0:1], s[8:9] ; SDAG-NEXT: v_mov_b64_e32 v[2:3], s[10:11] ; SDAG-NEXT: v_mov_b64_e32 v[4:5], s[12:13] -; SDAG-NEXT: v_accvgpr_write_b32 a0, s0 +; SDAG-NEXT: v_mov_b64_e32 v[10:11], s[2:3] ; SDAG-NEXT: v_mov_b64_e32 v[6:7], s[14:15] -; SDAG-NEXT: v_accvgpr_write_b32 a1, s1 -; SDAG-NEXT: v_accvgpr_write_b32 a2, s2 -; SDAG-NEXT: v_accvgpr_write_b32 a3, s3 +; SDAG-NEXT: v_mov_b64_e32 v[8:9], s[0:1] ; SDAG-NEXT: s_nop 1 -; SDAG-NEXT: v_mfma_f32_16x16x32_f16 a[0:3], v[0:3], v[4:7], a[0:3] cbsz:3 abid:2 blgp:1 +; SDAG-NEXT: v_mfma_f32_16x16x32_f16 v[0:3], v[0:3], v[4:7], v[8:11] cbsz:3 abid:2 blgp:1 ; SDAG-NEXT: s_nop 7 -; SDAG-NEXT: global_store_dwordx4 v8, a[0:3], s[6:7] +; SDAG-NEXT: global_store_dwordx4 v12, v[0:3], s[6:7] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_mfma_f32_16x16x32_f16_no_agpr__vgprcd__flags: @@ -291,16 +283,14 @@ define amdgpu_kernel void @test_mfma_f32_16x16x32_f16_no_agpr__vgprcd__flags(ptr ; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[8:9] ; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[10:11] ; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] -; GISEL-NEXT: v_accvgpr_write_b32 a0, s0 +; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[2:3] ; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] -; GISEL-NEXT: v_accvgpr_write_b32 a1, s1 -; GISEL-NEXT: v_accvgpr_write_b32 a2, s2 -; GISEL-NEXT: v_accvgpr_write_b32 a3, s3 +; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[0:1] ; GISEL-NEXT: s_nop 1 -; GISEL-NEXT: v_mfma_f32_16x16x32_f16 a[0:3], v[0:3], v[4:7], a[0:3] cbsz:3 abid:2 blgp:1 -; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: v_mfma_f32_16x16x32_f16 v[0:3], v[0:3], v[4:7], v[8:11] cbsz:3 abid:2 blgp:1 +; GISEL-NEXT: v_mov_b32_e32 v4, 0 ; GISEL-NEXT: s_nop 6 -; GISEL-NEXT: global_store_dwordx4 v0, a[0:3], s[6:7] +; GISEL-NEXT: global_store_dwordx4 v4, v[0:3], s[6:7] ; GISEL-NEXT: s_endpgm ; ; HEURRC-LABEL: test_mfma_f32_16x16x32_f16_no_agpr__vgprcd__flags: @@ -308,20 +298,18 @@ define amdgpu_kernel void @test_mfma_f32_16x16x32_f16_no_agpr__vgprcd__flags(ptr ; HEURRC-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x34 ; HEURRC-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 ; HEURRC-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x24 -; HEURRC-NEXT: v_mov_b32_e32 v8, 0 +; HEURRC-NEXT: v_mov_b32_e32 v12, 0 ; HEURRC-NEXT: s_waitcnt lgkmcnt(0) ; HEURRC-NEXT: v_mov_b64_e32 v[0:1], s[8:9] ; HEURRC-NEXT: v_mov_b64_e32 v[2:3], s[10:11] ; HEURRC-NEXT: v_mov_b64_e32 v[4:5], s[12:13] -; HEURRC-NEXT: v_accvgpr_write_b32 a0, s0 +; HEURRC-NEXT: v_mov_b64_e32 v[10:11], s[2:3] ; HEURRC-NEXT: v_mov_b64_e32 v[6:7], s[14:15] -; HEURRC-NEXT: v_accvgpr_write_b32 a1, s1 -; HEURRC-NEXT: v_accvgpr_write_b32 a2, s2 -; HEURRC-NEXT: v_accvgpr_write_b32 a3, s3 +; HEURRC-NEXT: v_mov_b64_e32 v[8:9], s[0:1] ; HEURRC-NEXT: s_nop 1 -; HEURRC-NEXT: v_mfma_f32_16x16x32_f16 a[0:3], v[0:3], v[4:7], a[0:3] cbsz:3 abid:2 blgp:1 +; HEURRC-NEXT: v_mfma_f32_16x16x32_f16 v[0:3], v[0:3], v[4:7], v[8:11] cbsz:3 abid:2 blgp:1 ; HEURRC-NEXT: s_nop 7 -; HEURRC-NEXT: global_store_dwordx4 v8, a[0:3], s[6:7] +; HEURRC-NEXT: global_store_dwordx4 v12, v[0:3], s[6:7] ; HEURRC-NEXT: s_endpgm ; ; VGPRRC-LABEL: test_mfma_f32_16x16x32_f16_no_agpr__vgprcd__flags: @@ -1505,62 +1493,55 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd(<8 x half> %arg0, ; SDAG-NEXT: s_load_dwordx8 s[24:31], s[4:5], 0x24 ; SDAG-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; SDAG-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 -; SDAG-NEXT: v_mov_b32_e32 v8, 0 +; SDAG-NEXT: v_mov_b32_e32 v44, 0 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b64_e32 v[0:1], s[24:25] -; SDAG-NEXT: v_mov_b64_e32 v[2:3], s[26:27] -; SDAG-NEXT: v_mov_b64_e32 v[4:5], s[28:29] -; SDAG-NEXT: v_accvgpr_write_b32 a31, s23 -; SDAG-NEXT: v_mov_b64_e32 v[6:7], s[30:31] -; SDAG-NEXT: v_accvgpr_write_b32 a30, s22 -; SDAG-NEXT: v_accvgpr_write_b32 a29, s21 -; SDAG-NEXT: v_accvgpr_write_b32 a28, s20 -; SDAG-NEXT: v_accvgpr_write_b32 a27, s19 -; SDAG-NEXT: v_accvgpr_write_b32 a26, s18 -; SDAG-NEXT: v_accvgpr_write_b32 a25, s17 -; SDAG-NEXT: v_accvgpr_write_b32 a24, s16 -; SDAG-NEXT: v_accvgpr_write_b32 a23, s15 -; SDAG-NEXT: v_accvgpr_write_b32 a22, s14 -; SDAG-NEXT: v_accvgpr_write_b32 a21, s13 -; SDAG-NEXT: v_accvgpr_write_b32 a20, s12 -; SDAG-NEXT: v_accvgpr_write_b32 a19, s11 -; SDAG-NEXT: v_accvgpr_write_b32 a18, s10 -; SDAG-NEXT: v_accvgpr_write_b32 a17, s9 -; SDAG-NEXT: v_accvgpr_write_b32 a16, s8 -; SDAG-NEXT: v_mov_b32_e32 v10, s20 -; SDAG-NEXT: v_mov_b32_e32 v11, s21 -; SDAG-NEXT: v_mfma_f32_32x32x16_f16 a[0:15], v[0:3], v[4:7], a[16:31] -; SDAG-NEXT: v_mov_b32_e32 v12, s22 -; SDAG-NEXT: v_mov_b32_e32 v13, s23 -; SDAG-NEXT: v_mov_b32_e32 v0, s16 -; SDAG-NEXT: v_mov_b32_e32 v1, s17 -; SDAG-NEXT: v_mov_b32_e32 v2, s18 -; SDAG-NEXT: v_mov_b32_e32 v3, s19 -; SDAG-NEXT: global_store_dwordx4 v8, v[10:13], s[0:1] offset:48 sc0 sc1 +; SDAG-NEXT: v_mov_b64_e32 v[34:35], s[26:27] +; SDAG-NEXT: v_mov_b64_e32 v[32:33], s[24:25] +; SDAG-NEXT: v_mov_b64_e32 v[38:39], s[30:31] +; SDAG-NEXT: v_mov_b64_e32 v[30:31], s[22:23] +; SDAG-NEXT: v_mov_b64_e32 v[36:37], s[28:29] +; SDAG-NEXT: v_mov_b64_e32 v[28:29], s[20:21] +; SDAG-NEXT: v_mov_b64_e32 v[26:27], s[18:19] +; SDAG-NEXT: v_mov_b64_e32 v[24:25], s[16:17] +; SDAG-NEXT: v_mov_b64_e32 v[22:23], s[14:15] +; SDAG-NEXT: v_mov_b64_e32 v[20:21], s[12:13] +; SDAG-NEXT: v_mov_b64_e32 v[18:19], s[10:11] +; SDAG-NEXT: v_mov_b64_e32 v[16:17], s[8:9] +; SDAG-NEXT: v_mov_b32_e32 v40, s20 +; SDAG-NEXT: v_mov_b32_e32 v41, s21 +; SDAG-NEXT: v_mfma_f32_32x32x16_f16 v[0:15], v[32:35], v[36:39], v[16:31] +; SDAG-NEXT: v_mov_b32_e32 v42, s22 +; SDAG-NEXT: v_mov_b32_e32 v43, s23 +; SDAG-NEXT: global_store_dwordx4 v44, v[40:43], s[0:1] offset:48 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1] offset:32 sc0 sc1 +; SDAG-NEXT: s_nop 2 +; SDAG-NEXT: v_mov_b32_e32 v16, s16 +; SDAG-NEXT: v_mov_b32_e32 v17, s17 +; SDAG-NEXT: v_mov_b32_e32 v18, s18 +; SDAG-NEXT: v_mov_b32_e32 v19, s19 +; SDAG-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] offset:32 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_mov_b32_e32 v0, s12 -; SDAG-NEXT: v_mov_b32_e32 v1, s13 -; SDAG-NEXT: v_mov_b32_e32 v2, s14 -; SDAG-NEXT: v_mov_b32_e32 v3, s15 -; SDAG-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1] offset:16 sc0 sc1 +; SDAG-NEXT: v_mov_b32_e32 v16, s12 +; SDAG-NEXT: v_mov_b32_e32 v17, s13 +; SDAG-NEXT: v_mov_b32_e32 v18, s14 +; SDAG-NEXT: v_mov_b32_e32 v19, s15 +; SDAG-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] offset:16 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_mov_b32_e32 v0, s8 -; SDAG-NEXT: v_mov_b32_e32 v1, s9 -; SDAG-NEXT: v_mov_b32_e32 v2, s10 -; SDAG-NEXT: v_mov_b32_e32 v3, s11 -; SDAG-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1] sc0 sc1 +; SDAG-NEXT: v_mov_b32_e32 v16, s8 +; SDAG-NEXT: v_mov_b32_e32 v17, s9 +; SDAG-NEXT: v_mov_b32_e32 v18, s10 +; SDAG-NEXT: v_mov_b32_e32 v19, s11 +; SDAG-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v8, a[8:11], s[0:1] offset:32 sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v44, v[8:11], s[0:1] offset:32 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v8, a[12:15], s[0:1] offset:48 sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v44, v[12:15], s[0:1] offset:48 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v8, a[0:3], s[0:1] sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v44, v[0:3], s[0:1] sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v8, a[4:7], s[0:1] offset:16 sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v44, v[4:7], s[0:1] offset:16 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_endpgm ; @@ -1569,52 +1550,44 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd(<8 x half> %arg0, ; GISEL-NEXT: s_load_dwordx8 s[24:31], s[4:5], 0x24 ; GISEL-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; GISEL-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 -; GISEL-NEXT: v_mov_b32_e32 v24, 0 +; GISEL-NEXT: v_mov_b32_e32 v56, 0 ; GISEL-NEXT: s_waitcnt lgkmcnt(0) -; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[24:25] -; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[26:27] -; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[28:29] -; GISEL-NEXT: v_accvgpr_write_b32 a0, s8 -; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[30:31] -; GISEL-NEXT: v_accvgpr_write_b32 a1, s9 -; GISEL-NEXT: v_accvgpr_write_b32 a2, s10 -; GISEL-NEXT: v_accvgpr_write_b32 a3, s11 -; GISEL-NEXT: v_accvgpr_write_b32 a4, s12 -; GISEL-NEXT: v_accvgpr_write_b32 a5, s13 -; GISEL-NEXT: v_accvgpr_write_b32 a6, s14 -; GISEL-NEXT: v_accvgpr_write_b32 a7, s15 -; GISEL-NEXT: v_accvgpr_write_b32 a8, s16 -; GISEL-NEXT: v_accvgpr_write_b32 a9, s17 -; GISEL-NEXT: v_accvgpr_write_b32 a10, s18 -; GISEL-NEXT: v_accvgpr_write_b32 a11, s19 -; GISEL-NEXT: v_accvgpr_write_b32 a12, s20 -; GISEL-NEXT: v_accvgpr_write_b32 a13, s21 -; GISEL-NEXT: v_accvgpr_write_b32 a14, s22 -; GISEL-NEXT: v_accvgpr_write_b32 a15, s23 -; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[8:9] -; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[10:11] -; GISEL-NEXT: v_mfma_f32_32x32x16_f16 a[16:31], v[0:3], v[4:7], a[0:15] -; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[12:13] -; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[16:17] -; GISEL-NEXT: v_mov_b64_e32 v[20:21], s[20:21] -; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[14:15] -; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[18:19] -; GISEL-NEXT: v_mov_b64_e32 v[22:23], s[22:23] -; GISEL-NEXT: global_store_dwordx4 v24, v[8:11], s[0:1] sc0 sc1 +; GISEL-NEXT: v_mov_b64_e32 v[34:35], s[26:27] +; GISEL-NEXT: v_mov_b64_e32 v[32:33], s[24:25] +; GISEL-NEXT: v_mov_b64_e32 v[38:39], s[30:31] +; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[8:9] +; GISEL-NEXT: v_mov_b64_e32 v[36:37], s[28:29] +; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[10:11] +; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] +; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] +; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[16:17] +; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[18:19] +; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[20:21] +; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[22:23] +; GISEL-NEXT: v_mov_b64_e32 v[42:43], s[10:11] +; GISEL-NEXT: v_mov_b64_e32 v[40:41], s[8:9] +; GISEL-NEXT: v_mfma_f32_32x32x16_f16 v[16:31], v[32:35], v[36:39], v[0:15] +; GISEL-NEXT: v_mov_b64_e32 v[46:47], s[14:15] +; GISEL-NEXT: v_mov_b64_e32 v[50:51], s[18:19] +; GISEL-NEXT: v_mov_b64_e32 v[54:55], s[22:23] +; GISEL-NEXT: v_mov_b64_e32 v[44:45], s[12:13] +; GISEL-NEXT: v_mov_b64_e32 v[48:49], s[16:17] +; GISEL-NEXT: v_mov_b64_e32 v[52:53], s[20:21] +; GISEL-NEXT: global_store_dwordx4 v56, v[40:43], s[0:1] sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, v[12:15], s[0:1] offset:16 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[44:47], s[0:1] offset:16 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, v[16:19], s[0:1] offset:32 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[48:51], s[0:1] offset:32 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, v[20:23], s[0:1] offset:48 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[52:55], s[0:1] offset:48 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, a[16:19], s[0:1] sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[16:19], s[0:1] sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, a[20:23], s[0:1] offset:16 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[20:23], s[0:1] offset:16 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, a[24:27], s[0:1] offset:32 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[24:27], s[0:1] offset:32 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, a[28:31], s[0:1] offset:48 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[28:31], s[0:1] offset:48 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) ; GISEL-NEXT: s_endpgm ; @@ -1623,62 +1596,55 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd(<8 x half> %arg0, ; HEURRC-NEXT: s_load_dwordx8 s[24:31], s[4:5], 0x24 ; HEURRC-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; HEURRC-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 -; HEURRC-NEXT: v_mov_b32_e32 v8, 0 +; HEURRC-NEXT: v_mov_b32_e32 v44, 0 ; HEURRC-NEXT: s_waitcnt lgkmcnt(0) -; HEURRC-NEXT: v_mov_b64_e32 v[0:1], s[24:25] -; HEURRC-NEXT: v_mov_b64_e32 v[2:3], s[26:27] -; HEURRC-NEXT: v_mov_b64_e32 v[4:5], s[28:29] -; HEURRC-NEXT: v_accvgpr_write_b32 a31, s23 -; HEURRC-NEXT: v_mov_b64_e32 v[6:7], s[30:31] -; HEURRC-NEXT: v_accvgpr_write_b32 a30, s22 -; HEURRC-NEXT: v_accvgpr_write_b32 a29, s21 -; HEURRC-NEXT: v_accvgpr_write_b32 a28, s20 -; HEURRC-NEXT: v_accvgpr_write_b32 a27, s19 -; HEURRC-NEXT: v_accvgpr_write_b32 a26, s18 -; HEURRC-NEXT: v_accvgpr_write_b32 a25, s17 -; HEURRC-NEXT: v_accvgpr_write_b32 a24, s16 -; HEURRC-NEXT: v_accvgpr_write_b32 a23, s15 -; HEURRC-NEXT: v_accvgpr_write_b32 a22, s14 -; HEURRC-NEXT: v_accvgpr_write_b32 a21, s13 -; HEURRC-NEXT: v_accvgpr_write_b32 a20, s12 -; HEURRC-NEXT: v_accvgpr_write_b32 a19, s11 -; HEURRC-NEXT: v_accvgpr_write_b32 a18, s10 -; HEURRC-NEXT: v_accvgpr_write_b32 a17, s9 -; HEURRC-NEXT: v_accvgpr_write_b32 a16, s8 -; HEURRC-NEXT: v_mov_b32_e32 v10, s20 -; HEURRC-NEXT: v_mov_b32_e32 v11, s21 -; HEURRC-NEXT: v_mfma_f32_32x32x16_f16 a[0:15], v[0:3], v[4:7], a[16:31] -; HEURRC-NEXT: v_mov_b32_e32 v12, s22 -; HEURRC-NEXT: v_mov_b32_e32 v13, s23 -; HEURRC-NEXT: v_mov_b32_e32 v0, s16 -; HEURRC-NEXT: v_mov_b32_e32 v1, s17 -; HEURRC-NEXT: v_mov_b32_e32 v2, s18 -; HEURRC-NEXT: v_mov_b32_e32 v3, s19 -; HEURRC-NEXT: global_store_dwordx4 v8, v[10:13], s[0:1] offset:48 sc0 sc1 +; HEURRC-NEXT: v_mov_b64_e32 v[34:35], s[26:27] +; HEURRC-NEXT: v_mov_b64_e32 v[32:33], s[24:25] +; HEURRC-NEXT: v_mov_b64_e32 v[38:39], s[30:31] +; HEURRC-NEXT: v_mov_b64_e32 v[30:31], s[22:23] +; HEURRC-NEXT: v_mov_b64_e32 v[36:37], s[28:29] +; HEURRC-NEXT: v_mov_b64_e32 v[28:29], s[20:21] +; HEURRC-NEXT: v_mov_b64_e32 v[26:27], s[18:19] +; HEURRC-NEXT: v_mov_b64_e32 v[24:25], s[16:17] +; HEURRC-NEXT: v_mov_b64_e32 v[22:23], s[14:15] +; HEURRC-NEXT: v_mov_b64_e32 v[20:21], s[12:13] +; HEURRC-NEXT: v_mov_b64_e32 v[18:19], s[10:11] +; HEURRC-NEXT: v_mov_b64_e32 v[16:17], s[8:9] +; HEURRC-NEXT: v_mov_b32_e32 v40, s20 +; HEURRC-NEXT: v_mov_b32_e32 v41, s21 +; HEURRC-NEXT: v_mfma_f32_32x32x16_f16 v[0:15], v[32:35], v[36:39], v[16:31] +; HEURRC-NEXT: v_mov_b32_e32 v42, s22 +; HEURRC-NEXT: v_mov_b32_e32 v43, s23 +; HEURRC-NEXT: global_store_dwordx4 v44, v[40:43], s[0:1] offset:48 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) -; HEURRC-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1] offset:32 sc0 sc1 +; HEURRC-NEXT: s_nop 2 +; HEURRC-NEXT: v_mov_b32_e32 v16, s16 +; HEURRC-NEXT: v_mov_b32_e32 v17, s17 +; HEURRC-NEXT: v_mov_b32_e32 v18, s18 +; HEURRC-NEXT: v_mov_b32_e32 v19, s19 +; HEURRC-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] offset:32 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) ; HEURRC-NEXT: s_nop 0 -; HEURRC-NEXT: v_mov_b32_e32 v0, s12 -; HEURRC-NEXT: v_mov_b32_e32 v1, s13 -; HEURRC-NEXT: v_mov_b32_e32 v2, s14 -; HEURRC-NEXT: v_mov_b32_e32 v3, s15 -; HEURRC-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1] offset:16 sc0 sc1 +; HEURRC-NEXT: v_mov_b32_e32 v16, s12 +; HEURRC-NEXT: v_mov_b32_e32 v17, s13 +; HEURRC-NEXT: v_mov_b32_e32 v18, s14 +; HEURRC-NEXT: v_mov_b32_e32 v19, s15 +; HEURRC-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] offset:16 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) ; HEURRC-NEXT: s_nop 0 -; HEURRC-NEXT: v_mov_b32_e32 v0, s8 -; HEURRC-NEXT: v_mov_b32_e32 v1, s9 -; HEURRC-NEXT: v_mov_b32_e32 v2, s10 -; HEURRC-NEXT: v_mov_b32_e32 v3, s11 -; HEURRC-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1] sc0 sc1 +; HEURRC-NEXT: v_mov_b32_e32 v16, s8 +; HEURRC-NEXT: v_mov_b32_e32 v17, s9 +; HEURRC-NEXT: v_mov_b32_e32 v18, s10 +; HEURRC-NEXT: v_mov_b32_e32 v19, s11 +; HEURRC-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) -; HEURRC-NEXT: global_store_dwordx4 v8, a[8:11], s[0:1] offset:32 sc0 sc1 +; HEURRC-NEXT: global_store_dwordx4 v44, v[8:11], s[0:1] offset:32 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) -; HEURRC-NEXT: global_store_dwordx4 v8, a[12:15], s[0:1] offset:48 sc0 sc1 +; HEURRC-NEXT: global_store_dwordx4 v44, v[12:15], s[0:1] offset:48 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) -; HEURRC-NEXT: global_store_dwordx4 v8, a[0:3], s[0:1] sc0 sc1 +; HEURRC-NEXT: global_store_dwordx4 v44, v[0:3], s[0:1] sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) -; HEURRC-NEXT: global_store_dwordx4 v8, a[4:7], s[0:1] offset:16 sc0 sc1 +; HEURRC-NEXT: global_store_dwordx4 v44, v[4:7], s[0:1] offset:16 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) ; HEURRC-NEXT: s_endpgm ; @@ -1687,7 +1653,7 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd(<8 x half> %arg0, ; VGPRRC-NEXT: s_load_dwordx8 s[24:31], s[4:5], 0x24 ; VGPRRC-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; VGPRRC-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 -; VGPRRC-NEXT: v_mov_b32_e32 v40, 0 +; VGPRRC-NEXT: v_mov_b32_e32 v44, 0 ; VGPRRC-NEXT: s_waitcnt lgkmcnt(0) ; VGPRRC-NEXT: v_mov_b64_e32 v[34:35], s[26:27] ; VGPRRC-NEXT: v_mov_b64_e32 v[32:33], s[24:25] @@ -1701,41 +1667,41 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd(<8 x half> %arg0, ; VGPRRC-NEXT: v_mov_b64_e32 v[20:21], s[12:13] ; VGPRRC-NEXT: v_mov_b64_e32 v[18:19], s[10:11] ; VGPRRC-NEXT: v_mov_b64_e32 v[16:17], s[8:9] -; VGPRRC-NEXT: v_mov_b32_e32 v42, s20 -; VGPRRC-NEXT: v_mov_b32_e32 v43, s21 +; VGPRRC-NEXT: v_mov_b32_e32 v40, s20 +; VGPRRC-NEXT: v_mov_b32_e32 v41, s21 ; VGPRRC-NEXT: v_mfma_f32_32x32x16_f16 v[0:15], v[32:35], v[36:39], v[16:31] -; VGPRRC-NEXT: v_mov_b32_e32 v44, s22 -; VGPRRC-NEXT: v_mov_b32_e32 v45, s23 -; VGPRRC-NEXT: global_store_dwordx4 v40, v[42:45], s[0:1] offset:48 sc0 sc1 +; VGPRRC-NEXT: v_mov_b32_e32 v42, s22 +; VGPRRC-NEXT: v_mov_b32_e32 v43, s23 +; VGPRRC-NEXT: global_store_dwordx4 v44, v[40:43], s[0:1] offset:48 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) ; VGPRRC-NEXT: s_nop 2 ; VGPRRC-NEXT: v_mov_b32_e32 v16, s16 ; VGPRRC-NEXT: v_mov_b32_e32 v17, s17 ; VGPRRC-NEXT: v_mov_b32_e32 v18, s18 ; VGPRRC-NEXT: v_mov_b32_e32 v19, s19 -; VGPRRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:32 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] offset:32 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) ; VGPRRC-NEXT: s_nop 0 ; VGPRRC-NEXT: v_mov_b32_e32 v16, s12 ; VGPRRC-NEXT: v_mov_b32_e32 v17, s13 ; VGPRRC-NEXT: v_mov_b32_e32 v18, s14 ; VGPRRC-NEXT: v_mov_b32_e32 v19, s15 -; VGPRRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:16 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] offset:16 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) ; VGPRRC-NEXT: s_nop 0 ; VGPRRC-NEXT: v_mov_b32_e32 v16, s8 ; VGPRRC-NEXT: v_mov_b32_e32 v17, s9 ; VGPRRC-NEXT: v_mov_b32_e32 v18, s10 ; VGPRRC-NEXT: v_mov_b32_e32 v19, s11 -; VGPRRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) -; VGPRRC-NEXT: global_store_dwordx4 v40, v[8:11], s[0:1] offset:32 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v44, v[8:11], s[0:1] offset:32 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) -; VGPRRC-NEXT: global_store_dwordx4 v40, v[12:15], s[0:1] offset:48 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v44, v[12:15], s[0:1] offset:48 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) -; VGPRRC-NEXT: global_store_dwordx4 v40, v[0:3], s[0:1] sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v44, v[0:3], s[0:1] sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) -; VGPRRC-NEXT: global_store_dwordx4 v40, v[4:7], s[0:1] offset:16 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v44, v[4:7], s[0:1] offset:16 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) ; VGPRRC-NEXT: s_endpgm ; AGPR-LABEL: test_mfma_f32_32x32x16_f16__vgprcd: @@ -1869,62 +1835,55 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd__flags(<8 x half> ; SDAG-NEXT: s_load_dwordx8 s[24:31], s[4:5], 0x24 ; SDAG-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; SDAG-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 -; SDAG-NEXT: v_mov_b32_e32 v8, 0 +; SDAG-NEXT: v_mov_b32_e32 v44, 0 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b64_e32 v[0:1], s[24:25] -; SDAG-NEXT: v_mov_b64_e32 v[2:3], s[26:27] -; SDAG-NEXT: v_mov_b64_e32 v[4:5], s[28:29] -; SDAG-NEXT: v_accvgpr_write_b32 a31, s23 -; SDAG-NEXT: v_mov_b64_e32 v[6:7], s[30:31] -; SDAG-NEXT: v_accvgpr_write_b32 a30, s22 -; SDAG-NEXT: v_accvgpr_write_b32 a29, s21 -; SDAG-NEXT: v_accvgpr_write_b32 a28, s20 -; SDAG-NEXT: v_accvgpr_write_b32 a27, s19 -; SDAG-NEXT: v_accvgpr_write_b32 a26, s18 -; SDAG-NEXT: v_accvgpr_write_b32 a25, s17 -; SDAG-NEXT: v_accvgpr_write_b32 a24, s16 -; SDAG-NEXT: v_accvgpr_write_b32 a23, s15 -; SDAG-NEXT: v_accvgpr_write_b32 a22, s14 -; SDAG-NEXT: v_accvgpr_write_b32 a21, s13 -; SDAG-NEXT: v_accvgpr_write_b32 a20, s12 -; SDAG-NEXT: v_accvgpr_write_b32 a19, s11 -; SDAG-NEXT: v_accvgpr_write_b32 a18, s10 -; SDAG-NEXT: v_accvgpr_write_b32 a17, s9 -; SDAG-NEXT: v_accvgpr_write_b32 a16, s8 -; SDAG-NEXT: v_mov_b32_e32 v10, s20 -; SDAG-NEXT: v_mov_b32_e32 v11, s21 -; SDAG-NEXT: v_mfma_f32_32x32x16_f16 a[0:15], v[0:3], v[4:7], a[16:31] cbsz:1 abid:2 blgp:3 -; SDAG-NEXT: v_mov_b32_e32 v12, s22 -; SDAG-NEXT: v_mov_b32_e32 v13, s23 -; SDAG-NEXT: v_mov_b32_e32 v0, s16 -; SDAG-NEXT: v_mov_b32_e32 v1, s17 -; SDAG-NEXT: v_mov_b32_e32 v2, s18 -; SDAG-NEXT: v_mov_b32_e32 v3, s19 -; SDAG-NEXT: global_store_dwordx4 v8, v[10:13], s[0:1] offset:48 sc0 sc1 +; SDAG-NEXT: v_mov_b64_e32 v[34:35], s[26:27] +; SDAG-NEXT: v_mov_b64_e32 v[32:33], s[24:25] +; SDAG-NEXT: v_mov_b64_e32 v[38:39], s[30:31] +; SDAG-NEXT: v_mov_b64_e32 v[30:31], s[22:23] +; SDAG-NEXT: v_mov_b64_e32 v[36:37], s[28:29] +; SDAG-NEXT: v_mov_b64_e32 v[28:29], s[20:21] +; SDAG-NEXT: v_mov_b64_e32 v[26:27], s[18:19] +; SDAG-NEXT: v_mov_b64_e32 v[24:25], s[16:17] +; SDAG-NEXT: v_mov_b64_e32 v[22:23], s[14:15] +; SDAG-NEXT: v_mov_b64_e32 v[20:21], s[12:13] +; SDAG-NEXT: v_mov_b64_e32 v[18:19], s[10:11] +; SDAG-NEXT: v_mov_b64_e32 v[16:17], s[8:9] +; SDAG-NEXT: v_mov_b32_e32 v40, s20 +; SDAG-NEXT: v_mov_b32_e32 v41, s21 +; SDAG-NEXT: v_mfma_f32_32x32x16_f16 v[0:15], v[32:35], v[36:39], v[16:31] cbsz:1 abid:2 blgp:3 +; SDAG-NEXT: v_mov_b32_e32 v42, s22 +; SDAG-NEXT: v_mov_b32_e32 v43, s23 +; SDAG-NEXT: global_store_dwordx4 v44, v[40:43], s[0:1] offset:48 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1] offset:32 sc0 sc1 +; SDAG-NEXT: s_nop 2 +; SDAG-NEXT: v_mov_b32_e32 v16, s16 +; SDAG-NEXT: v_mov_b32_e32 v17, s17 +; SDAG-NEXT: v_mov_b32_e32 v18, s18 +; SDAG-NEXT: v_mov_b32_e32 v19, s19 +; SDAG-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] offset:32 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_mov_b32_e32 v0, s12 -; SDAG-NEXT: v_mov_b32_e32 v1, s13 -; SDAG-NEXT: v_mov_b32_e32 v2, s14 -; SDAG-NEXT: v_mov_b32_e32 v3, s15 -; SDAG-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1] offset:16 sc0 sc1 +; SDAG-NEXT: v_mov_b32_e32 v16, s12 +; SDAG-NEXT: v_mov_b32_e32 v17, s13 +; SDAG-NEXT: v_mov_b32_e32 v18, s14 +; SDAG-NEXT: v_mov_b32_e32 v19, s15 +; SDAG-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] offset:16 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_mov_b32_e32 v0, s8 -; SDAG-NEXT: v_mov_b32_e32 v1, s9 -; SDAG-NEXT: v_mov_b32_e32 v2, s10 -; SDAG-NEXT: v_mov_b32_e32 v3, s11 -; SDAG-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1] sc0 sc1 +; SDAG-NEXT: v_mov_b32_e32 v16, s8 +; SDAG-NEXT: v_mov_b32_e32 v17, s9 +; SDAG-NEXT: v_mov_b32_e32 v18, s10 +; SDAG-NEXT: v_mov_b32_e32 v19, s11 +; SDAG-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v8, a[8:11], s[0:1] offset:32 sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v44, v[8:11], s[0:1] offset:32 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v8, a[12:15], s[0:1] offset:48 sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v44, v[12:15], s[0:1] offset:48 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v8, a[0:3], s[0:1] sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v44, v[0:3], s[0:1] sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v8, a[4:7], s[0:1] offset:16 sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v44, v[4:7], s[0:1] offset:16 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_endpgm ; @@ -1933,52 +1892,44 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd__flags(<8 x half> ; GISEL-NEXT: s_load_dwordx8 s[24:31], s[4:5], 0x24 ; GISEL-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; GISEL-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 -; GISEL-NEXT: v_mov_b32_e32 v24, 0 +; GISEL-NEXT: v_mov_b32_e32 v56, 0 ; GISEL-NEXT: s_waitcnt lgkmcnt(0) -; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[24:25] -; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[26:27] -; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[28:29] -; GISEL-NEXT: v_accvgpr_write_b32 a0, s8 -; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[30:31] -; GISEL-NEXT: v_accvgpr_write_b32 a1, s9 -; GISEL-NEXT: v_accvgpr_write_b32 a2, s10 -; GISEL-NEXT: v_accvgpr_write_b32 a3, s11 -; GISEL-NEXT: v_accvgpr_write_b32 a4, s12 -; GISEL-NEXT: v_accvgpr_write_b32 a5, s13 -; GISEL-NEXT: v_accvgpr_write_b32 a6, s14 -; GISEL-NEXT: v_accvgpr_write_b32 a7, s15 -; GISEL-NEXT: v_accvgpr_write_b32 a8, s16 -; GISEL-NEXT: v_accvgpr_write_b32 a9, s17 -; GISEL-NEXT: v_accvgpr_write_b32 a10, s18 -; GISEL-NEXT: v_accvgpr_write_b32 a11, s19 -; GISEL-NEXT: v_accvgpr_write_b32 a12, s20 -; GISEL-NEXT: v_accvgpr_write_b32 a13, s21 -; GISEL-NEXT: v_accvgpr_write_b32 a14, s22 -; GISEL-NEXT: v_accvgpr_write_b32 a15, s23 -; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[8:9] -; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[10:11] -; GISEL-NEXT: v_mfma_f32_32x32x16_f16 a[16:31], v[0:3], v[4:7], a[0:15] cbsz:1 abid:2 blgp:3 -; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[12:13] -; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[16:17] -; GISEL-NEXT: v_mov_b64_e32 v[20:21], s[20:21] -; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[14:15] -; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[18:19] -; GISEL-NEXT: v_mov_b64_e32 v[22:23], s[22:23] -; GISEL-NEXT: global_store_dwordx4 v24, v[8:11], s[0:1] sc0 sc1 +; GISEL-NEXT: v_mov_b64_e32 v[34:35], s[26:27] +; GISEL-NEXT: v_mov_b64_e32 v[32:33], s[24:25] +; GISEL-NEXT: v_mov_b64_e32 v[38:39], s[30:31] +; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[8:9] +; GISEL-NEXT: v_mov_b64_e32 v[36:37], s[28:29] +; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[10:11] +; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] +; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] +; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[16:17] +; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[18:19] +; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[20:21] +; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[22:23] +; GISEL-NEXT: v_mov_b64_e32 v[42:43], s[10:11] +; GISEL-NEXT: v_mov_b64_e32 v[40:41], s[8:9] +; GISEL-NEXT: v_mfma_f32_32x32x16_f16 v[16:31], v[32:35], v[36:39], v[0:15] cbsz:1 abid:2 blgp:3 +; GISEL-NEXT: v_mov_b64_e32 v[46:47], s[14:15] +; GISEL-NEXT: v_mov_b64_e32 v[50:51], s[18:19] +; GISEL-NEXT: v_mov_b64_e32 v[54:55], s[22:23] +; GISEL-NEXT: v_mov_b64_e32 v[44:45], s[12:13] +; GISEL-NEXT: v_mov_b64_e32 v[48:49], s[16:17] +; GISEL-NEXT: v_mov_b64_e32 v[52:53], s[20:21] +; GISEL-NEXT: global_store_dwordx4 v56, v[40:43], s[0:1] sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, v[12:15], s[0:1] offset:16 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[44:47], s[0:1] offset:16 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, v[16:19], s[0:1] offset:32 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[48:51], s[0:1] offset:32 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, v[20:23], s[0:1] offset:48 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[52:55], s[0:1] offset:48 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, a[16:19], s[0:1] sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[16:19], s[0:1] sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, a[20:23], s[0:1] offset:16 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[20:23], s[0:1] offset:16 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, a[24:27], s[0:1] offset:32 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[24:27], s[0:1] offset:32 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, a[28:31], s[0:1] offset:48 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[28:31], s[0:1] offset:48 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) ; GISEL-NEXT: s_endpgm ; @@ -1987,62 +1938,55 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd__flags(<8 x half> ; HEURRC-NEXT: s_load_dwordx8 s[24:31], s[4:5], 0x24 ; HEURRC-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; HEURRC-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 -; HEURRC-NEXT: v_mov_b32_e32 v8, 0 +; HEURRC-NEXT: v_mov_b32_e32 v44, 0 ; HEURRC-NEXT: s_waitcnt lgkmcnt(0) -; HEURRC-NEXT: v_mov_b64_e32 v[0:1], s[24:25] -; HEURRC-NEXT: v_mov_b64_e32 v[2:3], s[26:27] -; HEURRC-NEXT: v_mov_b64_e32 v[4:5], s[28:29] -; HEURRC-NEXT: v_accvgpr_write_b32 a31, s23 -; HEURRC-NEXT: v_mov_b64_e32 v[6:7], s[30:31] -; HEURRC-NEXT: v_accvgpr_write_b32 a30, s22 -; HEURRC-NEXT: v_accvgpr_write_b32 a29, s21 -; HEURRC-NEXT: v_accvgpr_write_b32 a28, s20 -; HEURRC-NEXT: v_accvgpr_write_b32 a27, s19 -; HEURRC-NEXT: v_accvgpr_write_b32 a26, s18 -; HEURRC-NEXT: v_accvgpr_write_b32 a25, s17 -; HEURRC-NEXT: v_accvgpr_write_b32 a24, s16 -; HEURRC-NEXT: v_accvgpr_write_b32 a23, s15 -; HEURRC-NEXT: v_accvgpr_write_b32 a22, s14 -; HEURRC-NEXT: v_accvgpr_write_b32 a21, s13 -; HEURRC-NEXT: v_accvgpr_write_b32 a20, s12 -; HEURRC-NEXT: v_accvgpr_write_b32 a19, s11 -; HEURRC-NEXT: v_accvgpr_write_b32 a18, s10 -; HEURRC-NEXT: v_accvgpr_write_b32 a17, s9 -; HEURRC-NEXT: v_accvgpr_write_b32 a16, s8 -; HEURRC-NEXT: v_mov_b32_e32 v10, s20 -; HEURRC-NEXT: v_mov_b32_e32 v11, s21 -; HEURRC-NEXT: v_mfma_f32_32x32x16_f16 a[0:15], v[0:3], v[4:7], a[16:31] cbsz:1 abid:2 blgp:3 -; HEURRC-NEXT: v_mov_b32_e32 v12, s22 -; HEURRC-NEXT: v_mov_b32_e32 v13, s23 -; HEURRC-NEXT: v_mov_b32_e32 v0, s16 -; HEURRC-NEXT: v_mov_b32_e32 v1, s17 -; HEURRC-NEXT: v_mov_b32_e32 v2, s18 -; HEURRC-NEXT: v_mov_b32_e32 v3, s19 -; HEURRC-NEXT: global_store_dwordx4 v8, v[10:13], s[0:1] offset:48 sc0 sc1 +; HEURRC-NEXT: v_mov_b64_e32 v[34:35], s[26:27] +; HEURRC-NEXT: v_mov_b64_e32 v[32:33], s[24:25] +; HEURRC-NEXT: v_mov_b64_e32 v[38:39], s[30:31] +; HEURRC-NEXT: v_mov_b64_e32 v[30:31], s[22:23] +; HEURRC-NEXT: v_mov_b64_e32 v[36:37], s[28:29] +; HEURRC-NEXT: v_mov_b64_e32 v[28:29], s[20:21] +; HEURRC-NEXT: v_mov_b64_e32 v[26:27], s[18:19] +; HEURRC-NEXT: v_mov_b64_e32 v[24:25], s[16:17] +; HEURRC-NEXT: v_mov_b64_e32 v[22:23], s[14:15] +; HEURRC-NEXT: v_mov_b64_e32 v[20:21], s[12:13] +; HEURRC-NEXT: v_mov_b64_e32 v[18:19], s[10:11] +; HEURRC-NEXT: v_mov_b64_e32 v[16:17], s[8:9] +; HEURRC-NEXT: v_mov_b32_e32 v40, s20 +; HEURRC-NEXT: v_mov_b32_e32 v41, s21 +; HEURRC-NEXT: v_mfma_f32_32x32x16_f16 v[0:15], v[32:35], v[36:39], v[16:31] cbsz:1 abid:2 blgp:3 +; HEURRC-NEXT: v_mov_b32_e32 v42, s22 +; HEURRC-NEXT: v_mov_b32_e32 v43, s23 +; HEURRC-NEXT: global_store_dwordx4 v44, v[40:43], s[0:1] offset:48 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) -; HEURRC-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1] offset:32 sc0 sc1 +; HEURRC-NEXT: s_nop 2 +; HEURRC-NEXT: v_mov_b32_e32 v16, s16 +; HEURRC-NEXT: v_mov_b32_e32 v17, s17 +; HEURRC-NEXT: v_mov_b32_e32 v18, s18 +; HEURRC-NEXT: v_mov_b32_e32 v19, s19 +; HEURRC-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] offset:32 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) ; HEURRC-NEXT: s_nop 0 -; HEURRC-NEXT: v_mov_b32_e32 v0, s12 -; HEURRC-NEXT: v_mov_b32_e32 v1, s13 -; HEURRC-NEXT: v_mov_b32_e32 v2, s14 -; HEURRC-NEXT: v_mov_b32_e32 v3, s15 -; HEURRC-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1] offset:16 sc0 sc1 +; HEURRC-NEXT: v_mov_b32_e32 v16, s12 +; HEURRC-NEXT: v_mov_b32_e32 v17, s13 +; HEURRC-NEXT: v_mov_b32_e32 v18, s14 +; HEURRC-NEXT: v_mov_b32_e32 v19, s15 +; HEURRC-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] offset:16 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) ; HEURRC-NEXT: s_nop 0 -; HEURRC-NEXT: v_mov_b32_e32 v0, s8 -; HEURRC-NEXT: v_mov_b32_e32 v1, s9 -; HEURRC-NEXT: v_mov_b32_e32 v2, s10 -; HEURRC-NEXT: v_mov_b32_e32 v3, s11 -; HEURRC-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1] sc0 sc1 +; HEURRC-NEXT: v_mov_b32_e32 v16, s8 +; HEURRC-NEXT: v_mov_b32_e32 v17, s9 +; HEURRC-NEXT: v_mov_b32_e32 v18, s10 +; HEURRC-NEXT: v_mov_b32_e32 v19, s11 +; HEURRC-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) -; HEURRC-NEXT: global_store_dwordx4 v8, a[8:11], s[0:1] offset:32 sc0 sc1 +; HEURRC-NEXT: global_store_dwordx4 v44, v[8:11], s[0:1] offset:32 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) -; HEURRC-NEXT: global_store_dwordx4 v8, a[12:15], s[0:1] offset:48 sc0 sc1 +; HEURRC-NEXT: global_store_dwordx4 v44, v[12:15], s[0:1] offset:48 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) -; HEURRC-NEXT: global_store_dwordx4 v8, a[0:3], s[0:1] sc0 sc1 +; HEURRC-NEXT: global_store_dwordx4 v44, v[0:3], s[0:1] sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) -; HEURRC-NEXT: global_store_dwordx4 v8, a[4:7], s[0:1] offset:16 sc0 sc1 +; HEURRC-NEXT: global_store_dwordx4 v44, v[4:7], s[0:1] offset:16 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) ; HEURRC-NEXT: s_endpgm ; @@ -2051,7 +1995,7 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd__flags(<8 x half> ; VGPRRC-NEXT: s_load_dwordx8 s[24:31], s[4:5], 0x24 ; VGPRRC-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; VGPRRC-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 -; VGPRRC-NEXT: v_mov_b32_e32 v40, 0 +; VGPRRC-NEXT: v_mov_b32_e32 v44, 0 ; VGPRRC-NEXT: s_waitcnt lgkmcnt(0) ; VGPRRC-NEXT: v_mov_b64_e32 v[34:35], s[26:27] ; VGPRRC-NEXT: v_mov_b64_e32 v[32:33], s[24:25] @@ -2065,41 +2009,41 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd__flags(<8 x half> ; VGPRRC-NEXT: v_mov_b64_e32 v[20:21], s[12:13] ; VGPRRC-NEXT: v_mov_b64_e32 v[18:19], s[10:11] ; VGPRRC-NEXT: v_mov_b64_e32 v[16:17], s[8:9] -; VGPRRC-NEXT: v_mov_b32_e32 v42, s20 -; VGPRRC-NEXT: v_mov_b32_e32 v43, s21 +; VGPRRC-NEXT: v_mov_b32_e32 v40, s20 +; VGPRRC-NEXT: v_mov_b32_e32 v41, s21 ; VGPRRC-NEXT: v_mfma_f32_32x32x16_f16 v[0:15], v[32:35], v[36:39], v[16:31] cbsz:1 abid:2 blgp:3 -; VGPRRC-NEXT: v_mov_b32_e32 v44, s22 -; VGPRRC-NEXT: v_mov_b32_e32 v45, s23 -; VGPRRC-NEXT: global_store_dwordx4 v40, v[42:45], s[0:1] offset:48 sc0 sc1 +; VGPRRC-NEXT: v_mov_b32_e32 v42, s22 +; VGPRRC-NEXT: v_mov_b32_e32 v43, s23 +; VGPRRC-NEXT: global_store_dwordx4 v44, v[40:43], s[0:1] offset:48 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) ; VGPRRC-NEXT: s_nop 2 ; VGPRRC-NEXT: v_mov_b32_e32 v16, s16 ; VGPRRC-NEXT: v_mov_b32_e32 v17, s17 ; VGPRRC-NEXT: v_mov_b32_e32 v18, s18 ; VGPRRC-NEXT: v_mov_b32_e32 v19, s19 -; VGPRRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:32 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] offset:32 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) ; VGPRRC-NEXT: s_nop 0 ; VGPRRC-NEXT: v_mov_b32_e32 v16, s12 ; VGPRRC-NEXT: v_mov_b32_e32 v17, s13 ; VGPRRC-NEXT: v_mov_b32_e32 v18, s14 ; VGPRRC-NEXT: v_mov_b32_e32 v19, s15 -; VGPRRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:16 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] offset:16 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) ; VGPRRC-NEXT: s_nop 0 ; VGPRRC-NEXT: v_mov_b32_e32 v16, s8 ; VGPRRC-NEXT: v_mov_b32_e32 v17, s9 ; VGPRRC-NEXT: v_mov_b32_e32 v18, s10 ; VGPRRC-NEXT: v_mov_b32_e32 v19, s11 -; VGPRRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v44, v[16:19], s[0:1] sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) -; VGPRRC-NEXT: global_store_dwordx4 v40, v[8:11], s[0:1] offset:32 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v44, v[8:11], s[0:1] offset:32 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) -; VGPRRC-NEXT: global_store_dwordx4 v40, v[12:15], s[0:1] offset:48 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v44, v[12:15], s[0:1] offset:48 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) -; VGPRRC-NEXT: global_store_dwordx4 v40, v[0:3], s[0:1] sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v44, v[0:3], s[0:1] sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) -; VGPRRC-NEXT: global_store_dwordx4 v40, v[4:7], s[0:1] offset:16 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v44, v[4:7], s[0:1] offset:16 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) ; VGPRRC-NEXT: s_endpgm ; AGPR-LABEL: test_mfma_f32_32x32x16_f16__vgprcd__flags: @@ -2234,35 +2178,27 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd_mac(<8 x half> %ar ; SDAG-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; SDAG-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b64_e32 v[0:1], s[24:25] -; SDAG-NEXT: v_mov_b64_e32 v[2:3], s[26:27] -; SDAG-NEXT: v_mov_b64_e32 v[4:5], s[28:29] -; SDAG-NEXT: v_accvgpr_write_b32 a0, s8 -; SDAG-NEXT: v_mov_b64_e32 v[6:7], s[30:31] -; SDAG-NEXT: v_accvgpr_write_b32 a1, s9 -; SDAG-NEXT: v_accvgpr_write_b32 a2, s10 -; SDAG-NEXT: v_accvgpr_write_b32 a3, s11 -; SDAG-NEXT: v_accvgpr_write_b32 a4, s12 -; SDAG-NEXT: v_accvgpr_write_b32 a5, s13 -; SDAG-NEXT: v_accvgpr_write_b32 a6, s14 -; SDAG-NEXT: v_accvgpr_write_b32 a7, s15 -; SDAG-NEXT: v_accvgpr_write_b32 a8, s16 -; SDAG-NEXT: v_accvgpr_write_b32 a9, s17 -; SDAG-NEXT: v_accvgpr_write_b32 a10, s18 -; SDAG-NEXT: v_accvgpr_write_b32 a11, s19 -; SDAG-NEXT: v_accvgpr_write_b32 a12, s20 -; SDAG-NEXT: v_accvgpr_write_b32 a13, s21 -; SDAG-NEXT: v_accvgpr_write_b32 a14, s22 -; SDAG-NEXT: v_accvgpr_write_b32 a15, s23 +; SDAG-NEXT: v_mov_b64_e32 v[16:17], s[24:25] +; SDAG-NEXT: v_mov_b64_e32 v[18:19], s[26:27] +; SDAG-NEXT: v_mov_b64_e32 v[20:21], s[28:29] +; SDAG-NEXT: v_mov_b64_e32 v[0:1], s[8:9] +; SDAG-NEXT: v_mov_b64_e32 v[22:23], s[30:31] +; SDAG-NEXT: v_mov_b64_e32 v[2:3], s[10:11] +; SDAG-NEXT: v_mov_b64_e32 v[4:5], s[12:13] +; SDAG-NEXT: v_mov_b64_e32 v[6:7], s[14:15] +; SDAG-NEXT: v_mov_b64_e32 v[8:9], s[16:17] +; SDAG-NEXT: v_mov_b64_e32 v[10:11], s[18:19] +; SDAG-NEXT: v_mov_b64_e32 v[12:13], s[20:21] +; SDAG-NEXT: v_mov_b64_e32 v[14:15], s[22:23] ; SDAG-NEXT: s_nop 1 -; SDAG-NEXT: v_mfma_f32_32x32x16_f16 a[0:15], v[0:3], v[4:7], a[0:15] -; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: v_mfma_f32_32x32x16_f16 v[0:15], v[16:19], v[20:23], v[0:15] +; SDAG-NEXT: v_mov_b32_e32 v16, 0 ; SDAG-NEXT: s_nop 7 ; SDAG-NEXT: s_nop 2 -; SDAG-NEXT: global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 -; SDAG-NEXT: global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 -; SDAG-NEXT: global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 -; SDAG-NEXT: global_store_dwordx4 v0, a[0:3], s[0:1] +; SDAG-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1] offset:48 +; SDAG-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1] offset:32 +; SDAG-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1] offset:16 +; SDAG-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_mfma_f32_32x32x16_f16__vgprcd_mac: @@ -2271,35 +2207,27 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd_mac(<8 x half> %ar ; GISEL-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; GISEL-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 ; GISEL-NEXT: s_waitcnt lgkmcnt(0) -; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[24:25] -; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[26:27] -; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[28:29] -; GISEL-NEXT: v_accvgpr_write_b32 a0, s8 -; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[30:31] -; GISEL-NEXT: v_accvgpr_write_b32 a1, s9 -; GISEL-NEXT: v_accvgpr_write_b32 a2, s10 -; GISEL-NEXT: v_accvgpr_write_b32 a3, s11 -; GISEL-NEXT: v_accvgpr_write_b32 a4, s12 -; GISEL-NEXT: v_accvgpr_write_b32 a5, s13 -; GISEL-NEXT: v_accvgpr_write_b32 a6, s14 -; GISEL-NEXT: v_accvgpr_write_b32 a7, s15 -; GISEL-NEXT: v_accvgpr_write_b32 a8, s16 -; GISEL-NEXT: v_accvgpr_write_b32 a9, s17 -; GISEL-NEXT: v_accvgpr_write_b32 a10, s18 -; GISEL-NEXT: v_accvgpr_write_b32 a11, s19 -; GISEL-NEXT: v_accvgpr_write_b32 a12, s20 -; GISEL-NEXT: v_accvgpr_write_b32 a13, s21 -; GISEL-NEXT: v_accvgpr_write_b32 a14, s22 -; GISEL-NEXT: v_accvgpr_write_b32 a15, s23 +; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[24:25] +; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[26:27] +; GISEL-NEXT: v_mov_b64_e32 v[20:21], s[28:29] +; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[8:9] +; GISEL-NEXT: v_mov_b64_e32 v[22:23], s[30:31] +; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[10:11] +; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] +; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] +; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[16:17] +; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[18:19] +; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[20:21] +; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[22:23] ; GISEL-NEXT: s_nop 1 -; GISEL-NEXT: v_mfma_f32_32x32x16_f16 a[0:15], v[0:3], v[4:7], a[0:15] -; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: v_mfma_f32_32x32x16_f16 v[0:15], v[16:19], v[20:23], v[0:15] +; GISEL-NEXT: v_mov_b32_e32 v16, 0 ; GISEL-NEXT: s_nop 7 ; GISEL-NEXT: s_nop 2 -; GISEL-NEXT: global_store_dwordx4 v0, a[0:3], s[0:1] -; GISEL-NEXT: global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 -; GISEL-NEXT: global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 -; GISEL-NEXT: global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 +; GISEL-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1] +; GISEL-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1] offset:16 +; GISEL-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1] offset:32 +; GISEL-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1] offset:48 ; GISEL-NEXT: s_endpgm ; ; HEURRC-LABEL: test_mfma_f32_32x32x16_f16__vgprcd_mac: @@ -2308,35 +2236,27 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd_mac(<8 x half> %ar ; HEURRC-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; HEURRC-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 ; HEURRC-NEXT: s_waitcnt lgkmcnt(0) -; HEURRC-NEXT: v_mov_b64_e32 v[0:1], s[24:25] -; HEURRC-NEXT: v_mov_b64_e32 v[2:3], s[26:27] -; HEURRC-NEXT: v_mov_b64_e32 v[4:5], s[28:29] -; HEURRC-NEXT: v_accvgpr_write_b32 a0, s8 -; HEURRC-NEXT: v_mov_b64_e32 v[6:7], s[30:31] -; HEURRC-NEXT: v_accvgpr_write_b32 a1, s9 -; HEURRC-NEXT: v_accvgpr_write_b32 a2, s10 -; HEURRC-NEXT: v_accvgpr_write_b32 a3, s11 -; HEURRC-NEXT: v_accvgpr_write_b32 a4, s12 -; HEURRC-NEXT: v_accvgpr_write_b32 a5, s13 -; HEURRC-NEXT: v_accvgpr_write_b32 a6, s14 -; HEURRC-NEXT: v_accvgpr_write_b32 a7, s15 -; HEURRC-NEXT: v_accvgpr_write_b32 a8, s16 -; HEURRC-NEXT: v_accvgpr_write_b32 a9, s17 -; HEURRC-NEXT: v_accvgpr_write_b32 a10, s18 -; HEURRC-NEXT: v_accvgpr_write_b32 a11, s19 -; HEURRC-NEXT: v_accvgpr_write_b32 a12, s20 -; HEURRC-NEXT: v_accvgpr_write_b32 a13, s21 -; HEURRC-NEXT: v_accvgpr_write_b32 a14, s22 -; HEURRC-NEXT: v_accvgpr_write_b32 a15, s23 +; HEURRC-NEXT: v_mov_b64_e32 v[16:17], s[24:25] +; HEURRC-NEXT: v_mov_b64_e32 v[18:19], s[26:27] +; HEURRC-NEXT: v_mov_b64_e32 v[20:21], s[28:29] +; HEURRC-NEXT: v_mov_b64_e32 v[0:1], s[8:9] +; HEURRC-NEXT: v_mov_b64_e32 v[22:23], s[30:31] +; HEURRC-NEXT: v_mov_b64_e32 v[2:3], s[10:11] +; HEURRC-NEXT: v_mov_b64_e32 v[4:5], s[12:13] +; HEURRC-NEXT: v_mov_b64_e32 v[6:7], s[14:15] +; HEURRC-NEXT: v_mov_b64_e32 v[8:9], s[16:17] +; HEURRC-NEXT: v_mov_b64_e32 v[10:11], s[18:19] +; HEURRC-NEXT: v_mov_b64_e32 v[12:13], s[20:21] +; HEURRC-NEXT: v_mov_b64_e32 v[14:15], s[22:23] ; HEURRC-NEXT: s_nop 1 -; HEURRC-NEXT: v_mfma_f32_32x32x16_f16 a[0:15], v[0:3], v[4:7], a[0:15] -; HEURRC-NEXT: v_mov_b32_e32 v0, 0 +; HEURRC-NEXT: v_mfma_f32_32x32x16_f16 v[0:15], v[16:19], v[20:23], v[0:15] +; HEURRC-NEXT: v_mov_b32_e32 v16, 0 ; HEURRC-NEXT: s_nop 7 ; HEURRC-NEXT: s_nop 2 -; HEURRC-NEXT: global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 -; HEURRC-NEXT: global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 -; HEURRC-NEXT: global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 -; HEURRC-NEXT: global_store_dwordx4 v0, a[0:3], s[0:1] +; HEURRC-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1] offset:48 +; HEURRC-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1] offset:32 +; HEURRC-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1] offset:16 +; HEURRC-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1] ; HEURRC-NEXT: s_endpgm ; ; VGPRRC-LABEL: test_mfma_f32_32x32x16_f16__vgprcd_mac: @@ -2443,35 +2363,27 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd_mac_flags(<8 x hal ; SDAG-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; SDAG-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b64_e32 v[0:1], s[24:25] -; SDAG-NEXT: v_mov_b64_e32 v[2:3], s[26:27] -; SDAG-NEXT: v_mov_b64_e32 v[4:5], s[28:29] -; SDAG-NEXT: v_accvgpr_write_b32 a0, s8 -; SDAG-NEXT: v_mov_b64_e32 v[6:7], s[30:31] -; SDAG-NEXT: v_accvgpr_write_b32 a1, s9 -; SDAG-NEXT: v_accvgpr_write_b32 a2, s10 -; SDAG-NEXT: v_accvgpr_write_b32 a3, s11 -; SDAG-NEXT: v_accvgpr_write_b32 a4, s12 -; SDAG-NEXT: v_accvgpr_write_b32 a5, s13 -; SDAG-NEXT: v_accvgpr_write_b32 a6, s14 -; SDAG-NEXT: v_accvgpr_write_b32 a7, s15 -; SDAG-NEXT: v_accvgpr_write_b32 a8, s16 -; SDAG-NEXT: v_accvgpr_write_b32 a9, s17 -; SDAG-NEXT: v_accvgpr_write_b32 a10, s18 -; SDAG-NEXT: v_accvgpr_write_b32 a11, s19 -; SDAG-NEXT: v_accvgpr_write_b32 a12, s20 -; SDAG-NEXT: v_accvgpr_write_b32 a13, s21 -; SDAG-NEXT: v_accvgpr_write_b32 a14, s22 -; SDAG-NEXT: v_accvgpr_write_b32 a15, s23 +; SDAG-NEXT: v_mov_b64_e32 v[16:17], s[24:25] +; SDAG-NEXT: v_mov_b64_e32 v[18:19], s[26:27] +; SDAG-NEXT: v_mov_b64_e32 v[20:21], s[28:29] +; SDAG-NEXT: v_mov_b64_e32 v[0:1], s[8:9] +; SDAG-NEXT: v_mov_b64_e32 v[22:23], s[30:31] +; SDAG-NEXT: v_mov_b64_e32 v[2:3], s[10:11] +; SDAG-NEXT: v_mov_b64_e32 v[4:5], s[12:13] +; SDAG-NEXT: v_mov_b64_e32 v[6:7], s[14:15] +; SDAG-NEXT: v_mov_b64_e32 v[8:9], s[16:17] +; SDAG-NEXT: v_mov_b64_e32 v[10:11], s[18:19] +; SDAG-NEXT: v_mov_b64_e32 v[12:13], s[20:21] +; SDAG-NEXT: v_mov_b64_e32 v[14:15], s[22:23] ; SDAG-NEXT: s_nop 1 -; SDAG-NEXT: v_mfma_f32_32x32x16_f16 a[0:15], v[0:3], v[4:7], a[0:15] cbsz:3 abid:2 blgp:1 -; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: v_mfma_f32_32x32x16_f16 v[0:15], v[16:19], v[20:23], v[0:15] cbsz:3 abid:2 blgp:1 +; SDAG-NEXT: v_mov_b32_e32 v16, 0 ; SDAG-NEXT: s_nop 7 ; SDAG-NEXT: s_nop 2 -; SDAG-NEXT: global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 -; SDAG-NEXT: global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 -; SDAG-NEXT: global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 -; SDAG-NEXT: global_store_dwordx4 v0, a[0:3], s[0:1] +; SDAG-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1] offset:48 +; SDAG-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1] offset:32 +; SDAG-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1] offset:16 +; SDAG-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_mfma_f32_32x32x16_f16__vgprcd_mac_flags: @@ -2480,35 +2392,27 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd_mac_flags(<8 x hal ; GISEL-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; GISEL-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 ; GISEL-NEXT: s_waitcnt lgkmcnt(0) -; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[24:25] -; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[26:27] -; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[28:29] -; GISEL-NEXT: v_accvgpr_write_b32 a0, s8 -; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[30:31] -; GISEL-NEXT: v_accvgpr_write_b32 a1, s9 -; GISEL-NEXT: v_accvgpr_write_b32 a2, s10 -; GISEL-NEXT: v_accvgpr_write_b32 a3, s11 -; GISEL-NEXT: v_accvgpr_write_b32 a4, s12 -; GISEL-NEXT: v_accvgpr_write_b32 a5, s13 -; GISEL-NEXT: v_accvgpr_write_b32 a6, s14 -; GISEL-NEXT: v_accvgpr_write_b32 a7, s15 -; GISEL-NEXT: v_accvgpr_write_b32 a8, s16 -; GISEL-NEXT: v_accvgpr_write_b32 a9, s17 -; GISEL-NEXT: v_accvgpr_write_b32 a10, s18 -; GISEL-NEXT: v_accvgpr_write_b32 a11, s19 -; GISEL-NEXT: v_accvgpr_write_b32 a12, s20 -; GISEL-NEXT: v_accvgpr_write_b32 a13, s21 -; GISEL-NEXT: v_accvgpr_write_b32 a14, s22 -; GISEL-NEXT: v_accvgpr_write_b32 a15, s23 +; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[24:25] +; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[26:27] +; GISEL-NEXT: v_mov_b64_e32 v[20:21], s[28:29] +; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[8:9] +; GISEL-NEXT: v_mov_b64_e32 v[22:23], s[30:31] +; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[10:11] +; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] +; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] +; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[16:17] +; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[18:19] +; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[20:21] +; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[22:23] ; GISEL-NEXT: s_nop 1 -; GISEL-NEXT: v_mfma_f32_32x32x16_f16 a[0:15], v[0:3], v[4:7], a[0:15] cbsz:3 abid:2 blgp:1 -; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: v_mfma_f32_32x32x16_f16 v[0:15], v[16:19], v[20:23], v[0:15] cbsz:3 abid:2 blgp:1 +; GISEL-NEXT: v_mov_b32_e32 v16, 0 ; GISEL-NEXT: s_nop 7 ; GISEL-NEXT: s_nop 2 -; GISEL-NEXT: global_store_dwordx4 v0, a[0:3], s[0:1] -; GISEL-NEXT: global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 -; GISEL-NEXT: global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 -; GISEL-NEXT: global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 +; GISEL-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1] +; GISEL-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1] offset:16 +; GISEL-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1] offset:32 +; GISEL-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1] offset:48 ; GISEL-NEXT: s_endpgm ; ; HEURRC-LABEL: test_mfma_f32_32x32x16_f16__vgprcd_mac_flags: @@ -2517,35 +2421,27 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd_mac_flags(<8 x hal ; HEURRC-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; HEURRC-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 ; HEURRC-NEXT: s_waitcnt lgkmcnt(0) -; HEURRC-NEXT: v_mov_b64_e32 v[0:1], s[24:25] -; HEURRC-NEXT: v_mov_b64_e32 v[2:3], s[26:27] -; HEURRC-NEXT: v_mov_b64_e32 v[4:5], s[28:29] -; HEURRC-NEXT: v_accvgpr_write_b32 a0, s8 -; HEURRC-NEXT: v_mov_b64_e32 v[6:7], s[30:31] -; HEURRC-NEXT: v_accvgpr_write_b32 a1, s9 -; HEURRC-NEXT: v_accvgpr_write_b32 a2, s10 -; HEURRC-NEXT: v_accvgpr_write_b32 a3, s11 -; HEURRC-NEXT: v_accvgpr_write_b32 a4, s12 -; HEURRC-NEXT: v_accvgpr_write_b32 a5, s13 -; HEURRC-NEXT: v_accvgpr_write_b32 a6, s14 -; HEURRC-NEXT: v_accvgpr_write_b32 a7, s15 -; HEURRC-NEXT: v_accvgpr_write_b32 a8, s16 -; HEURRC-NEXT: v_accvgpr_write_b32 a9, s17 -; HEURRC-NEXT: v_accvgpr_write_b32 a10, s18 -; HEURRC-NEXT: v_accvgpr_write_b32 a11, s19 -; HEURRC-NEXT: v_accvgpr_write_b32 a12, s20 -; HEURRC-NEXT: v_accvgpr_write_b32 a13, s21 -; HEURRC-NEXT: v_accvgpr_write_b32 a14, s22 -; HEURRC-NEXT: v_accvgpr_write_b32 a15, s23 +; HEURRC-NEXT: v_mov_b64_e32 v[16:17], s[24:25] +; HEURRC-NEXT: v_mov_b64_e32 v[18:19], s[26:27] +; HEURRC-NEXT: v_mov_b64_e32 v[20:21], s[28:29] +; HEURRC-NEXT: v_mov_b64_e32 v[0:1], s[8:9] +; HEURRC-NEXT: v_mov_b64_e32 v[22:23], s[30:31] +; HEURRC-NEXT: v_mov_b64_e32 v[2:3], s[10:11] +; HEURRC-NEXT: v_mov_b64_e32 v[4:5], s[12:13] +; HEURRC-NEXT: v_mov_b64_e32 v[6:7], s[14:15] +; HEURRC-NEXT: v_mov_b64_e32 v[8:9], s[16:17] +; HEURRC-NEXT: v_mov_b64_e32 v[10:11], s[18:19] +; HEURRC-NEXT: v_mov_b64_e32 v[12:13], s[20:21] +; HEURRC-NEXT: v_mov_b64_e32 v[14:15], s[22:23] ; HEURRC-NEXT: s_nop 1 -; HEURRC-NEXT: v_mfma_f32_32x32x16_f16 a[0:15], v[0:3], v[4:7], a[0:15] cbsz:3 abid:2 blgp:1 -; HEURRC-NEXT: v_mov_b32_e32 v0, 0 +; HEURRC-NEXT: v_mfma_f32_32x32x16_f16 v[0:15], v[16:19], v[20:23], v[0:15] cbsz:3 abid:2 blgp:1 +; HEURRC-NEXT: v_mov_b32_e32 v16, 0 ; HEURRC-NEXT: s_nop 7 ; HEURRC-NEXT: s_nop 2 -; HEURRC-NEXT: global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 -; HEURRC-NEXT: global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 -; HEURRC-NEXT: global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 -; HEURRC-NEXT: global_store_dwordx4 v0, a[0:3], s[0:1] +; HEURRC-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1] offset:48 +; HEURRC-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1] offset:32 +; HEURRC-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1] offset:16 +; HEURRC-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1] ; HEURRC-NEXT: s_endpgm ; ; VGPRRC-LABEL: test_mfma_f32_32x32x16_f16__vgprcd_mac_flags: @@ -2781,24 +2677,24 @@ define amdgpu_kernel void @test_mfma_i32_16x16x64_i8_no_agpr__vgprcd(ptr addrspa ; SDAG-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x34 ; SDAG-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x24 ; SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 -; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: v_mov_b32_e32 v12, 0 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v2, s8 -; SDAG-NEXT: v_mov_b32_e32 v3, s9 -; SDAG-NEXT: v_mov_b32_e32 v4, s10 -; SDAG-NEXT: v_mov_b32_e32 v5, s11 -; SDAG-NEXT: v_mov_b32_e32 v6, s12 -; SDAG-NEXT: v_mov_b32_e32 v7, s13 -; SDAG-NEXT: v_mov_b32_e32 v8, s14 -; SDAG-NEXT: v_mov_b32_e32 v9, s15 -; SDAG-NEXT: v_accvgpr_write_b32 a0, s0 -; SDAG-NEXT: v_accvgpr_write_b32 a1, s1 -; SDAG-NEXT: v_accvgpr_write_b32 a2, s2 -; SDAG-NEXT: v_accvgpr_write_b32 a3, s3 +; SDAG-NEXT: v_mov_b32_e32 v0, s8 +; SDAG-NEXT: v_mov_b32_e32 v1, s9 +; SDAG-NEXT: v_mov_b32_e32 v2, s10 +; SDAG-NEXT: v_mov_b32_e32 v3, s11 +; SDAG-NEXT: v_mov_b32_e32 v4, s12 +; SDAG-NEXT: v_mov_b32_e32 v5, s13 +; SDAG-NEXT: v_mov_b32_e32 v6, s14 +; SDAG-NEXT: v_mov_b32_e32 v7, s15 +; SDAG-NEXT: v_mov_b32_e32 v8, s0 +; SDAG-NEXT: v_mov_b32_e32 v9, s1 +; SDAG-NEXT: v_mov_b32_e32 v10, s2 +; SDAG-NEXT: v_mov_b32_e32 v11, s3 ; SDAG-NEXT: s_nop 1 -; SDAG-NEXT: v_mfma_i32_16x16x64_i8 a[0:3], v[2:5], v[6:9], a[0:3] +; SDAG-NEXT: v_mfma_i32_16x16x64_i8 v[0:3], v[0:3], v[4:7], v[8:11] ; SDAG-NEXT: s_nop 7 -; SDAG-NEXT: global_store_dwordx4 v0, a[0:3], s[6:7] +; SDAG-NEXT: global_store_dwordx4 v12, v[0:3], s[6:7] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_mfma_i32_16x16x64_i8_no_agpr__vgprcd: @@ -2810,16 +2706,14 @@ define amdgpu_kernel void @test_mfma_i32_16x16x64_i8_no_agpr__vgprcd(ptr addrspa ; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[8:9] ; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[10:11] ; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] -; GISEL-NEXT: v_accvgpr_write_b32 a0, s0 +; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[2:3] ; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] -; GISEL-NEXT: v_accvgpr_write_b32 a1, s1 -; GISEL-NEXT: v_accvgpr_write_b32 a2, s2 -; GISEL-NEXT: v_accvgpr_write_b32 a3, s3 +; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[0:1] ; GISEL-NEXT: s_nop 1 -; GISEL-NEXT: v_mfma_i32_16x16x64_i8 a[0:3], v[0:3], v[4:7], a[0:3] -; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: v_mfma_i32_16x16x64_i8 v[0:3], v[0:3], v[4:7], v[8:11] +; GISEL-NEXT: v_mov_b32_e32 v4, 0 ; GISEL-NEXT: s_nop 6 -; GISEL-NEXT: global_store_dwordx4 v0, a[0:3], s[6:7] +; GISEL-NEXT: global_store_dwordx4 v4, v[0:3], s[6:7] ; GISEL-NEXT: s_endpgm ; ; HEURRC-LABEL: test_mfma_i32_16x16x64_i8_no_agpr__vgprcd: @@ -2827,24 +2721,24 @@ define amdgpu_kernel void @test_mfma_i32_16x16x64_i8_no_agpr__vgprcd(ptr addrspa ; HEURRC-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x34 ; HEURRC-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x24 ; HEURRC-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 -; HEURRC-NEXT: v_mov_b32_e32 v0, 0 +; HEURRC-NEXT: v_mov_b32_e32 v12, 0 ; HEURRC-NEXT: s_waitcnt lgkmcnt(0) -; HEURRC-NEXT: v_mov_b32_e32 v2, s8 -; HEURRC-NEXT: v_mov_b32_e32 v3, s9 -; HEURRC-NEXT: v_mov_b32_e32 v4, s10 -; HEURRC-NEXT: v_mov_b32_e32 v5, s11 -; HEURRC-NEXT: v_mov_b32_e32 v6, s12 -; HEURRC-NEXT: v_mov_b32_e32 v7, s13 -; HEURRC-NEXT: v_mov_b32_e32 v8, s14 -; HEURRC-NEXT: v_mov_b32_e32 v9, s15 -; HEURRC-NEXT: v_accvgpr_write_b32 a0, s0 -; HEURRC-NEXT: v_accvgpr_write_b32 a1, s1 -; HEURRC-NEXT: v_accvgpr_write_b32 a2, s2 -; HEURRC-NEXT: v_accvgpr_write_b32 a3, s3 +; HEURRC-NEXT: v_mov_b32_e32 v0, s8 +; HEURRC-NEXT: v_mov_b32_e32 v1, s9 +; HEURRC-NEXT: v_mov_b32_e32 v2, s10 +; HEURRC-NEXT: v_mov_b32_e32 v3, s11 +; HEURRC-NEXT: v_mov_b32_e32 v4, s12 +; HEURRC-NEXT: v_mov_b32_e32 v5, s13 +; HEURRC-NEXT: v_mov_b32_e32 v6, s14 +; HEURRC-NEXT: v_mov_b32_e32 v7, s15 +; HEURRC-NEXT: v_mov_b32_e32 v8, s0 +; HEURRC-NEXT: v_mov_b32_e32 v9, s1 +; HEURRC-NEXT: v_mov_b32_e32 v10, s2 +; HEURRC-NEXT: v_mov_b32_e32 v11, s3 ; HEURRC-NEXT: s_nop 1 -; HEURRC-NEXT: v_mfma_i32_16x16x64_i8 a[0:3], v[2:5], v[6:9], a[0:3] +; HEURRC-NEXT: v_mfma_i32_16x16x64_i8 v[0:3], v[0:3], v[4:7], v[8:11] ; HEURRC-NEXT: s_nop 7 -; HEURRC-NEXT: global_store_dwordx4 v0, a[0:3], s[6:7] +; HEURRC-NEXT: global_store_dwordx4 v12, v[0:3], s[6:7] ; HEURRC-NEXT: s_endpgm ; ; VGPRRC-LABEL: test_mfma_i32_16x16x64_i8_no_agpr__vgprcd: @@ -2852,24 +2746,24 @@ define amdgpu_kernel void @test_mfma_i32_16x16x64_i8_no_agpr__vgprcd(ptr addrspa ; VGPRRC-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x34 ; VGPRRC-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x24 ; VGPRRC-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 -; VGPRRC-NEXT: v_mov_b32_e32 v4, 0 +; VGPRRC-NEXT: v_mov_b32_e32 v12, 0 ; VGPRRC-NEXT: s_waitcnt lgkmcnt(0) -; VGPRRC-NEXT: v_mov_b32_e32 v6, s8 -; VGPRRC-NEXT: v_mov_b32_e32 v7, s9 -; VGPRRC-NEXT: v_mov_b32_e32 v8, s10 -; VGPRRC-NEXT: v_mov_b32_e32 v9, s11 -; VGPRRC-NEXT: v_mov_b32_e32 v10, s12 -; VGPRRC-NEXT: v_mov_b32_e32 v11, s13 -; VGPRRC-NEXT: v_mov_b32_e32 v12, s14 -; VGPRRC-NEXT: v_mov_b32_e32 v13, s15 -; VGPRRC-NEXT: v_mov_b32_e32 v0, s0 -; VGPRRC-NEXT: v_mov_b32_e32 v1, s1 -; VGPRRC-NEXT: v_mov_b32_e32 v2, s2 -; VGPRRC-NEXT: v_mov_b32_e32 v3, s3 +; VGPRRC-NEXT: v_mov_b32_e32 v0, s8 +; VGPRRC-NEXT: v_mov_b32_e32 v1, s9 +; VGPRRC-NEXT: v_mov_b32_e32 v2, s10 +; VGPRRC-NEXT: v_mov_b32_e32 v3, s11 +; VGPRRC-NEXT: v_mov_b32_e32 v4, s12 +; VGPRRC-NEXT: v_mov_b32_e32 v5, s13 +; VGPRRC-NEXT: v_mov_b32_e32 v6, s14 +; VGPRRC-NEXT: v_mov_b32_e32 v7, s15 +; VGPRRC-NEXT: v_mov_b32_e32 v8, s0 +; VGPRRC-NEXT: v_mov_b32_e32 v9, s1 +; VGPRRC-NEXT: v_mov_b32_e32 v10, s2 +; VGPRRC-NEXT: v_mov_b32_e32 v11, s3 ; VGPRRC-NEXT: s_nop 1 -; VGPRRC-NEXT: v_mfma_i32_16x16x64_i8 v[0:3], v[6:9], v[10:13], v[0:3] +; VGPRRC-NEXT: v_mfma_i32_16x16x64_i8 v[0:3], v[0:3], v[4:7], v[8:11] ; VGPRRC-NEXT: s_nop 7 -; VGPRRC-NEXT: global_store_dwordx4 v4, v[0:3], s[6:7] +; VGPRRC-NEXT: global_store_dwordx4 v12, v[0:3], s[6:7] ; VGPRRC-NEXT: s_endpgm ; AGPR-LABEL: test_mfma_i32_16x16x64_i8_no_agpr__vgprcd: ; AGPR: ; %bb.0: @@ -2930,24 +2824,24 @@ define amdgpu_kernel void @test_mfma_i32_16x16x64_i8_no_agpr__vgprcd__flags(ptr ; SDAG-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x34 ; SDAG-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x24 ; SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 -; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: v_mov_b32_e32 v12, 0 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v2, s8 -; SDAG-NEXT: v_mov_b32_e32 v3, s9 -; SDAG-NEXT: v_mov_b32_e32 v4, s10 -; SDAG-NEXT: v_mov_b32_e32 v5, s11 -; SDAG-NEXT: v_mov_b32_e32 v6, s12 -; SDAG-NEXT: v_mov_b32_e32 v7, s13 -; SDAG-NEXT: v_mov_b32_e32 v8, s14 -; SDAG-NEXT: v_mov_b32_e32 v9, s15 -; SDAG-NEXT: v_accvgpr_write_b32 a0, s0 -; SDAG-NEXT: v_accvgpr_write_b32 a1, s1 -; SDAG-NEXT: v_accvgpr_write_b32 a2, s2 -; SDAG-NEXT: v_accvgpr_write_b32 a3, s3 +; SDAG-NEXT: v_mov_b32_e32 v0, s8 +; SDAG-NEXT: v_mov_b32_e32 v1, s9 +; SDAG-NEXT: v_mov_b32_e32 v2, s10 +; SDAG-NEXT: v_mov_b32_e32 v3, s11 +; SDAG-NEXT: v_mov_b32_e32 v4, s12 +; SDAG-NEXT: v_mov_b32_e32 v5, s13 +; SDAG-NEXT: v_mov_b32_e32 v6, s14 +; SDAG-NEXT: v_mov_b32_e32 v7, s15 +; SDAG-NEXT: v_mov_b32_e32 v8, s0 +; SDAG-NEXT: v_mov_b32_e32 v9, s1 +; SDAG-NEXT: v_mov_b32_e32 v10, s2 +; SDAG-NEXT: v_mov_b32_e32 v11, s3 ; SDAG-NEXT: s_nop 1 -; SDAG-NEXT: v_mfma_i32_16x16x64_i8 a[0:3], v[2:5], v[6:9], a[0:3] cbsz:3 abid:2 blgp:1 +; SDAG-NEXT: v_mfma_i32_16x16x64_i8 v[0:3], v[0:3], v[4:7], v[8:11] cbsz:3 abid:2 blgp:1 ; SDAG-NEXT: s_nop 7 -; SDAG-NEXT: global_store_dwordx4 v0, a[0:3], s[6:7] +; SDAG-NEXT: global_store_dwordx4 v12, v[0:3], s[6:7] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_mfma_i32_16x16x64_i8_no_agpr__vgprcd__flags: @@ -2959,16 +2853,14 @@ define amdgpu_kernel void @test_mfma_i32_16x16x64_i8_no_agpr__vgprcd__flags(ptr ; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[8:9] ; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[10:11] ; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] -; GISEL-NEXT: v_accvgpr_write_b32 a0, s0 +; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[2:3] ; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] -; GISEL-NEXT: v_accvgpr_write_b32 a1, s1 -; GISEL-NEXT: v_accvgpr_write_b32 a2, s2 -; GISEL-NEXT: v_accvgpr_write_b32 a3, s3 +; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[0:1] ; GISEL-NEXT: s_nop 1 -; GISEL-NEXT: v_mfma_i32_16x16x64_i8 a[0:3], v[0:3], v[4:7], a[0:3] cbsz:3 abid:2 blgp:1 -; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: v_mfma_i32_16x16x64_i8 v[0:3], v[0:3], v[4:7], v[8:11] cbsz:3 abid:2 blgp:1 +; GISEL-NEXT: v_mov_b32_e32 v4, 0 ; GISEL-NEXT: s_nop 6 -; GISEL-NEXT: global_store_dwordx4 v0, a[0:3], s[6:7] +; GISEL-NEXT: global_store_dwordx4 v4, v[0:3], s[6:7] ; GISEL-NEXT: s_endpgm ; ; HEURRC-LABEL: test_mfma_i32_16x16x64_i8_no_agpr__vgprcd__flags: @@ -2976,24 +2868,24 @@ define amdgpu_kernel void @test_mfma_i32_16x16x64_i8_no_agpr__vgprcd__flags(ptr ; HEURRC-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x34 ; HEURRC-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x24 ; HEURRC-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 -; HEURRC-NEXT: v_mov_b32_e32 v0, 0 +; HEURRC-NEXT: v_mov_b32_e32 v12, 0 ; HEURRC-NEXT: s_waitcnt lgkmcnt(0) -; HEURRC-NEXT: v_mov_b32_e32 v2, s8 -; HEURRC-NEXT: v_mov_b32_e32 v3, s9 -; HEURRC-NEXT: v_mov_b32_e32 v4, s10 -; HEURRC-NEXT: v_mov_b32_e32 v5, s11 -; HEURRC-NEXT: v_mov_b32_e32 v6, s12 -; HEURRC-NEXT: v_mov_b32_e32 v7, s13 -; HEURRC-NEXT: v_mov_b32_e32 v8, s14 -; HEURRC-NEXT: v_mov_b32_e32 v9, s15 -; HEURRC-NEXT: v_accvgpr_write_b32 a0, s0 -; HEURRC-NEXT: v_accvgpr_write_b32 a1, s1 -; HEURRC-NEXT: v_accvgpr_write_b32 a2, s2 -; HEURRC-NEXT: v_accvgpr_write_b32 a3, s3 +; HEURRC-NEXT: v_mov_b32_e32 v0, s8 +; HEURRC-NEXT: v_mov_b32_e32 v1, s9 +; HEURRC-NEXT: v_mov_b32_e32 v2, s10 +; HEURRC-NEXT: v_mov_b32_e32 v3, s11 +; HEURRC-NEXT: v_mov_b32_e32 v4, s12 +; HEURRC-NEXT: v_mov_b32_e32 v5, s13 +; HEURRC-NEXT: v_mov_b32_e32 v6, s14 +; HEURRC-NEXT: v_mov_b32_e32 v7, s15 +; HEURRC-NEXT: v_mov_b32_e32 v8, s0 +; HEURRC-NEXT: v_mov_b32_e32 v9, s1 +; HEURRC-NEXT: v_mov_b32_e32 v10, s2 +; HEURRC-NEXT: v_mov_b32_e32 v11, s3 ; HEURRC-NEXT: s_nop 1 -; HEURRC-NEXT: v_mfma_i32_16x16x64_i8 a[0:3], v[2:5], v[6:9], a[0:3] cbsz:3 abid:2 blgp:1 +; HEURRC-NEXT: v_mfma_i32_16x16x64_i8 v[0:3], v[0:3], v[4:7], v[8:11] cbsz:3 abid:2 blgp:1 ; HEURRC-NEXT: s_nop 7 -; HEURRC-NEXT: global_store_dwordx4 v0, a[0:3], s[6:7] +; HEURRC-NEXT: global_store_dwordx4 v12, v[0:3], s[6:7] ; HEURRC-NEXT: s_endpgm ; ; VGPRRC-LABEL: test_mfma_i32_16x16x64_i8_no_agpr__vgprcd__flags: @@ -3001,24 +2893,24 @@ define amdgpu_kernel void @test_mfma_i32_16x16x64_i8_no_agpr__vgprcd__flags(ptr ; VGPRRC-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x34 ; VGPRRC-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x24 ; VGPRRC-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 -; VGPRRC-NEXT: v_mov_b32_e32 v4, 0 +; VGPRRC-NEXT: v_mov_b32_e32 v12, 0 ; VGPRRC-NEXT: s_waitcnt lgkmcnt(0) -; VGPRRC-NEXT: v_mov_b32_e32 v6, s8 -; VGPRRC-NEXT: v_mov_b32_e32 v7, s9 -; VGPRRC-NEXT: v_mov_b32_e32 v8, s10 -; VGPRRC-NEXT: v_mov_b32_e32 v9, s11 -; VGPRRC-NEXT: v_mov_b32_e32 v10, s12 -; VGPRRC-NEXT: v_mov_b32_e32 v11, s13 -; VGPRRC-NEXT: v_mov_b32_e32 v12, s14 -; VGPRRC-NEXT: v_mov_b32_e32 v13, s15 -; VGPRRC-NEXT: v_mov_b32_e32 v0, s0 -; VGPRRC-NEXT: v_mov_b32_e32 v1, s1 -; VGPRRC-NEXT: v_mov_b32_e32 v2, s2 -; VGPRRC-NEXT: v_mov_b32_e32 v3, s3 +; VGPRRC-NEXT: v_mov_b32_e32 v0, s8 +; VGPRRC-NEXT: v_mov_b32_e32 v1, s9 +; VGPRRC-NEXT: v_mov_b32_e32 v2, s10 +; VGPRRC-NEXT: v_mov_b32_e32 v3, s11 +; VGPRRC-NEXT: v_mov_b32_e32 v4, s12 +; VGPRRC-NEXT: v_mov_b32_e32 v5, s13 +; VGPRRC-NEXT: v_mov_b32_e32 v6, s14 +; VGPRRC-NEXT: v_mov_b32_e32 v7, s15 +; VGPRRC-NEXT: v_mov_b32_e32 v8, s0 +; VGPRRC-NEXT: v_mov_b32_e32 v9, s1 +; VGPRRC-NEXT: v_mov_b32_e32 v10, s2 +; VGPRRC-NEXT: v_mov_b32_e32 v11, s3 ; VGPRRC-NEXT: s_nop 1 -; VGPRRC-NEXT: v_mfma_i32_16x16x64_i8 v[0:3], v[6:9], v[10:13], v[0:3] cbsz:3 abid:2 blgp:1 +; VGPRRC-NEXT: v_mfma_i32_16x16x64_i8 v[0:3], v[0:3], v[4:7], v[8:11] cbsz:3 abid:2 blgp:1 ; VGPRRC-NEXT: s_nop 7 -; VGPRRC-NEXT: global_store_dwordx4 v4, v[0:3], s[6:7] +; VGPRRC-NEXT: global_store_dwordx4 v12, v[0:3], s[6:7] ; VGPRRC-NEXT: s_endpgm ; AGPR-LABEL: test_mfma_i32_16x16x64_i8_no_agpr__vgprcd__flags: ; AGPR: ; %bb.0: @@ -4246,70 +4138,63 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd(<4 x i32> %arg0, <4 ; SDAG: ; %bb.0: ; SDAG-NEXT: s_load_dwordx8 s[20:27], s[4:5], 0x24 ; SDAG-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 -; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: v_mov_b32_e32 v40, 0 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v2, s20 -; SDAG-NEXT: v_mov_b32_e32 v3, s21 -; SDAG-NEXT: v_mov_b32_e32 v4, s22 -; SDAG-NEXT: v_mov_b32_e32 v5, s23 +; SDAG-NEXT: v_mov_b32_e32 v32, s20 +; SDAG-NEXT: v_mov_b32_e32 v33, s21 +; SDAG-NEXT: v_mov_b32_e32 v34, s22 +; SDAG-NEXT: v_mov_b32_e32 v35, s23 ; SDAG-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 -; SDAG-NEXT: v_mov_b32_e32 v6, s24 -; SDAG-NEXT: v_mov_b32_e32 v7, s25 -; SDAG-NEXT: v_mov_b32_e32 v8, s26 -; SDAG-NEXT: v_mov_b32_e32 v9, s27 +; SDAG-NEXT: v_mov_b32_e32 v36, s24 +; SDAG-NEXT: v_mov_b32_e32 v37, s25 +; SDAG-NEXT: v_mov_b32_e32 v38, s26 +; SDAG-NEXT: v_mov_b32_e32 v39, s27 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_accvgpr_write_b32 a31, s23 -; SDAG-NEXT: v_accvgpr_write_b32 a30, s22 -; SDAG-NEXT: v_accvgpr_write_b32 a29, s21 -; SDAG-NEXT: v_accvgpr_write_b32 a28, s20 -; SDAG-NEXT: v_accvgpr_write_b32 a27, s19 -; SDAG-NEXT: v_accvgpr_write_b32 a26, s18 -; SDAG-NEXT: v_accvgpr_write_b32 a25, s17 -; SDAG-NEXT: v_accvgpr_write_b32 a24, s16 -; SDAG-NEXT: v_accvgpr_write_b32 a23, s15 -; SDAG-NEXT: v_accvgpr_write_b32 a22, s14 -; SDAG-NEXT: v_accvgpr_write_b32 a21, s13 -; SDAG-NEXT: v_accvgpr_write_b32 a20, s12 -; SDAG-NEXT: v_accvgpr_write_b32 a19, s11 -; SDAG-NEXT: v_accvgpr_write_b32 a18, s10 -; SDAG-NEXT: v_accvgpr_write_b32 a17, s9 -; SDAG-NEXT: v_accvgpr_write_b32 a16, s8 +; SDAG-NEXT: v_mov_b64_e32 v[30:31], s[22:23] +; SDAG-NEXT: v_mov_b64_e32 v[28:29], s[20:21] +; SDAG-NEXT: v_mov_b64_e32 v[26:27], s[18:19] +; SDAG-NEXT: v_mov_b64_e32 v[24:25], s[16:17] +; SDAG-NEXT: v_mov_b64_e32 v[22:23], s[14:15] +; SDAG-NEXT: v_mov_b64_e32 v[20:21], s[12:13] +; SDAG-NEXT: v_mov_b64_e32 v[18:19], s[10:11] +; SDAG-NEXT: v_mov_b64_e32 v[16:17], s[8:9] ; SDAG-NEXT: s_nop 1 -; SDAG-NEXT: v_mfma_i32_32x32x32_i8 a[0:15], v[2:5], v[6:9], a[16:31] -; SDAG-NEXT: v_mov_b32_e32 v2, s20 -; SDAG-NEXT: v_mov_b32_e32 v3, s21 -; SDAG-NEXT: v_mov_b32_e32 v4, s22 -; SDAG-NEXT: v_mov_b32_e32 v5, s23 -; SDAG-NEXT: global_store_dwordx4 v0, v[2:5], s[0:1] offset:48 sc0 sc1 +; SDAG-NEXT: v_mfma_i32_32x32x32_i8 v[0:15], v[32:35], v[36:39], v[16:31] +; SDAG-NEXT: s_nop 6 +; SDAG-NEXT: v_mov_b32_e32 v16, s20 +; SDAG-NEXT: v_mov_b32_e32 v17, s21 +; SDAG-NEXT: v_mov_b32_e32 v18, s22 +; SDAG-NEXT: v_mov_b32_e32 v19, s23 +; SDAG-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:48 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_mov_b32_e32 v2, s16 -; SDAG-NEXT: v_mov_b32_e32 v3, s17 -; SDAG-NEXT: v_mov_b32_e32 v4, s18 -; SDAG-NEXT: v_mov_b32_e32 v5, s19 -; SDAG-NEXT: global_store_dwordx4 v0, v[2:5], s[0:1] offset:32 sc0 sc1 +; SDAG-NEXT: v_mov_b32_e32 v16, s16 +; SDAG-NEXT: v_mov_b32_e32 v17, s17 +; SDAG-NEXT: v_mov_b32_e32 v18, s18 +; SDAG-NEXT: v_mov_b32_e32 v19, s19 +; SDAG-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:32 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_mov_b32_e32 v2, s12 -; SDAG-NEXT: v_mov_b32_e32 v3, s13 -; SDAG-NEXT: v_mov_b32_e32 v4, s14 -; SDAG-NEXT: v_mov_b32_e32 v5, s15 -; SDAG-NEXT: global_store_dwordx4 v0, v[2:5], s[0:1] offset:16 sc0 sc1 +; SDAG-NEXT: v_mov_b32_e32 v16, s12 +; SDAG-NEXT: v_mov_b32_e32 v17, s13 +; SDAG-NEXT: v_mov_b32_e32 v18, s14 +; SDAG-NEXT: v_mov_b32_e32 v19, s15 +; SDAG-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:16 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_mov_b32_e32 v2, s8 -; SDAG-NEXT: v_mov_b32_e32 v3, s9 -; SDAG-NEXT: v_mov_b32_e32 v4, s10 -; SDAG-NEXT: v_mov_b32_e32 v5, s11 -; SDAG-NEXT: global_store_dwordx4 v0, v[2:5], s[0:1] sc0 sc1 +; SDAG-NEXT: v_mov_b32_e32 v16, s8 +; SDAG-NEXT: v_mov_b32_e32 v17, s9 +; SDAG-NEXT: v_mov_b32_e32 v18, s10 +; SDAG-NEXT: v_mov_b32_e32 v19, s11 +; SDAG-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v40, v[8:11], s[0:1] offset:32 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v40, v[12:15], s[0:1] offset:48 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v0, a[0:3], s[0:1] sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v40, v[0:3], s[0:1] sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v40, v[4:7], s[0:1] offset:16 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_endpgm ; @@ -4318,52 +4203,44 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd(<4 x i32> %arg0, <4 ; GISEL-NEXT: s_load_dwordx8 s[24:31], s[4:5], 0x24 ; GISEL-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; GISEL-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 -; GISEL-NEXT: v_mov_b32_e32 v24, 0 +; GISEL-NEXT: v_mov_b32_e32 v56, 0 ; GISEL-NEXT: s_waitcnt lgkmcnt(0) -; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[24:25] -; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[26:27] -; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[28:29] -; GISEL-NEXT: v_accvgpr_write_b32 a0, s8 -; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[30:31] -; GISEL-NEXT: v_accvgpr_write_b32 a1, s9 -; GISEL-NEXT: v_accvgpr_write_b32 a2, s10 -; GISEL-NEXT: v_accvgpr_write_b32 a3, s11 -; GISEL-NEXT: v_accvgpr_write_b32 a4, s12 -; GISEL-NEXT: v_accvgpr_write_b32 a5, s13 -; GISEL-NEXT: v_accvgpr_write_b32 a6, s14 -; GISEL-NEXT: v_accvgpr_write_b32 a7, s15 -; GISEL-NEXT: v_accvgpr_write_b32 a8, s16 -; GISEL-NEXT: v_accvgpr_write_b32 a9, s17 -; GISEL-NEXT: v_accvgpr_write_b32 a10, s18 -; GISEL-NEXT: v_accvgpr_write_b32 a11, s19 -; GISEL-NEXT: v_accvgpr_write_b32 a12, s20 -; GISEL-NEXT: v_accvgpr_write_b32 a13, s21 -; GISEL-NEXT: v_accvgpr_write_b32 a14, s22 -; GISEL-NEXT: v_accvgpr_write_b32 a15, s23 -; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[8:9] -; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[10:11] -; GISEL-NEXT: v_mfma_i32_32x32x32_i8 a[16:31], v[0:3], v[4:7], a[0:15] -; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[12:13] -; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[16:17] -; GISEL-NEXT: v_mov_b64_e32 v[20:21], s[20:21] -; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[14:15] -; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[18:19] -; GISEL-NEXT: v_mov_b64_e32 v[22:23], s[22:23] -; GISEL-NEXT: global_store_dwordx4 v24, v[8:11], s[0:1] sc0 sc1 +; GISEL-NEXT: v_mov_b64_e32 v[34:35], s[26:27] +; GISEL-NEXT: v_mov_b64_e32 v[32:33], s[24:25] +; GISEL-NEXT: v_mov_b64_e32 v[38:39], s[30:31] +; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[8:9] +; GISEL-NEXT: v_mov_b64_e32 v[36:37], s[28:29] +; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[10:11] +; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] +; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] +; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[16:17] +; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[18:19] +; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[20:21] +; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[22:23] +; GISEL-NEXT: v_mov_b64_e32 v[42:43], s[10:11] +; GISEL-NEXT: v_mov_b64_e32 v[40:41], s[8:9] +; GISEL-NEXT: v_mfma_i32_32x32x32_i8 v[16:31], v[32:35], v[36:39], v[0:15] +; GISEL-NEXT: v_mov_b64_e32 v[46:47], s[14:15] +; GISEL-NEXT: v_mov_b64_e32 v[50:51], s[18:19] +; GISEL-NEXT: v_mov_b64_e32 v[54:55], s[22:23] +; GISEL-NEXT: v_mov_b64_e32 v[44:45], s[12:13] +; GISEL-NEXT: v_mov_b64_e32 v[48:49], s[16:17] +; GISEL-NEXT: v_mov_b64_e32 v[52:53], s[20:21] +; GISEL-NEXT: global_store_dwordx4 v56, v[40:43], s[0:1] sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, v[12:15], s[0:1] offset:16 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[44:47], s[0:1] offset:16 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, v[16:19], s[0:1] offset:32 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[48:51], s[0:1] offset:32 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, v[20:23], s[0:1] offset:48 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[52:55], s[0:1] offset:48 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, a[16:19], s[0:1] sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[16:19], s[0:1] sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, a[20:23], s[0:1] offset:16 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[20:23], s[0:1] offset:16 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, a[24:27], s[0:1] offset:32 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[24:27], s[0:1] offset:32 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, a[28:31], s[0:1] offset:48 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[28:31], s[0:1] offset:48 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) ; GISEL-NEXT: s_endpgm ; @@ -4371,70 +4248,63 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd(<4 x i32> %arg0, <4 ; HEURRC: ; %bb.0: ; HEURRC-NEXT: s_load_dwordx8 s[20:27], s[4:5], 0x24 ; HEURRC-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 -; HEURRC-NEXT: v_mov_b32_e32 v0, 0 +; HEURRC-NEXT: v_mov_b32_e32 v40, 0 ; HEURRC-NEXT: s_waitcnt lgkmcnt(0) -; HEURRC-NEXT: v_mov_b32_e32 v2, s20 -; HEURRC-NEXT: v_mov_b32_e32 v3, s21 -; HEURRC-NEXT: v_mov_b32_e32 v4, s22 -; HEURRC-NEXT: v_mov_b32_e32 v5, s23 +; HEURRC-NEXT: v_mov_b32_e32 v32, s20 +; HEURRC-NEXT: v_mov_b32_e32 v33, s21 +; HEURRC-NEXT: v_mov_b32_e32 v34, s22 +; HEURRC-NEXT: v_mov_b32_e32 v35, s23 ; HEURRC-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 -; HEURRC-NEXT: v_mov_b32_e32 v6, s24 -; HEURRC-NEXT: v_mov_b32_e32 v7, s25 -; HEURRC-NEXT: v_mov_b32_e32 v8, s26 -; HEURRC-NEXT: v_mov_b32_e32 v9, s27 +; HEURRC-NEXT: v_mov_b32_e32 v36, s24 +; HEURRC-NEXT: v_mov_b32_e32 v37, s25 +; HEURRC-NEXT: v_mov_b32_e32 v38, s26 +; HEURRC-NEXT: v_mov_b32_e32 v39, s27 ; HEURRC-NEXT: s_waitcnt lgkmcnt(0) -; HEURRC-NEXT: v_accvgpr_write_b32 a31, s23 -; HEURRC-NEXT: v_accvgpr_write_b32 a30, s22 -; HEURRC-NEXT: v_accvgpr_write_b32 a29, s21 -; HEURRC-NEXT: v_accvgpr_write_b32 a28, s20 -; HEURRC-NEXT: v_accvgpr_write_b32 a27, s19 -; HEURRC-NEXT: v_accvgpr_write_b32 a26, s18 -; HEURRC-NEXT: v_accvgpr_write_b32 a25, s17 -; HEURRC-NEXT: v_accvgpr_write_b32 a24, s16 -; HEURRC-NEXT: v_accvgpr_write_b32 a23, s15 -; HEURRC-NEXT: v_accvgpr_write_b32 a22, s14 -; HEURRC-NEXT: v_accvgpr_write_b32 a21, s13 -; HEURRC-NEXT: v_accvgpr_write_b32 a20, s12 -; HEURRC-NEXT: v_accvgpr_write_b32 a19, s11 -; HEURRC-NEXT: v_accvgpr_write_b32 a18, s10 -; HEURRC-NEXT: v_accvgpr_write_b32 a17, s9 -; HEURRC-NEXT: v_accvgpr_write_b32 a16, s8 +; HEURRC-NEXT: v_mov_b64_e32 v[30:31], s[22:23] +; HEURRC-NEXT: v_mov_b64_e32 v[28:29], s[20:21] +; HEURRC-NEXT: v_mov_b64_e32 v[26:27], s[18:19] +; HEURRC-NEXT: v_mov_b64_e32 v[24:25], s[16:17] +; HEURRC-NEXT: v_mov_b64_e32 v[22:23], s[14:15] +; HEURRC-NEXT: v_mov_b64_e32 v[20:21], s[12:13] +; HEURRC-NEXT: v_mov_b64_e32 v[18:19], s[10:11] +; HEURRC-NEXT: v_mov_b64_e32 v[16:17], s[8:9] ; HEURRC-NEXT: s_nop 1 -; HEURRC-NEXT: v_mfma_i32_32x32x32_i8 a[0:15], v[2:5], v[6:9], a[16:31] -; HEURRC-NEXT: v_mov_b32_e32 v2, s20 -; HEURRC-NEXT: v_mov_b32_e32 v3, s21 -; HEURRC-NEXT: v_mov_b32_e32 v4, s22 -; HEURRC-NEXT: v_mov_b32_e32 v5, s23 -; HEURRC-NEXT: global_store_dwordx4 v0, v[2:5], s[0:1] offset:48 sc0 sc1 +; HEURRC-NEXT: v_mfma_i32_32x32x32_i8 v[0:15], v[32:35], v[36:39], v[16:31] +; HEURRC-NEXT: s_nop 6 +; HEURRC-NEXT: v_mov_b32_e32 v16, s20 +; HEURRC-NEXT: v_mov_b32_e32 v17, s21 +; HEURRC-NEXT: v_mov_b32_e32 v18, s22 +; HEURRC-NEXT: v_mov_b32_e32 v19, s23 +; HEURRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:48 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) ; HEURRC-NEXT: s_nop 0 -; HEURRC-NEXT: v_mov_b32_e32 v2, s16 -; HEURRC-NEXT: v_mov_b32_e32 v3, s17 -; HEURRC-NEXT: v_mov_b32_e32 v4, s18 -; HEURRC-NEXT: v_mov_b32_e32 v5, s19 -; HEURRC-NEXT: global_store_dwordx4 v0, v[2:5], s[0:1] offset:32 sc0 sc1 +; HEURRC-NEXT: v_mov_b32_e32 v16, s16 +; HEURRC-NEXT: v_mov_b32_e32 v17, s17 +; HEURRC-NEXT: v_mov_b32_e32 v18, s18 +; HEURRC-NEXT: v_mov_b32_e32 v19, s19 +; HEURRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:32 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) ; HEURRC-NEXT: s_nop 0 -; HEURRC-NEXT: v_mov_b32_e32 v2, s12 -; HEURRC-NEXT: v_mov_b32_e32 v3, s13 -; HEURRC-NEXT: v_mov_b32_e32 v4, s14 -; HEURRC-NEXT: v_mov_b32_e32 v5, s15 -; HEURRC-NEXT: global_store_dwordx4 v0, v[2:5], s[0:1] offset:16 sc0 sc1 +; HEURRC-NEXT: v_mov_b32_e32 v16, s12 +; HEURRC-NEXT: v_mov_b32_e32 v17, s13 +; HEURRC-NEXT: v_mov_b32_e32 v18, s14 +; HEURRC-NEXT: v_mov_b32_e32 v19, s15 +; HEURRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:16 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) ; HEURRC-NEXT: s_nop 0 -; HEURRC-NEXT: v_mov_b32_e32 v2, s8 -; HEURRC-NEXT: v_mov_b32_e32 v3, s9 -; HEURRC-NEXT: v_mov_b32_e32 v4, s10 -; HEURRC-NEXT: v_mov_b32_e32 v5, s11 -; HEURRC-NEXT: global_store_dwordx4 v0, v[2:5], s[0:1] sc0 sc1 +; HEURRC-NEXT: v_mov_b32_e32 v16, s8 +; HEURRC-NEXT: v_mov_b32_e32 v17, s9 +; HEURRC-NEXT: v_mov_b32_e32 v18, s10 +; HEURRC-NEXT: v_mov_b32_e32 v19, s11 +; HEURRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) -; HEURRC-NEXT: global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 sc0 sc1 +; HEURRC-NEXT: global_store_dwordx4 v40, v[8:11], s[0:1] offset:32 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) -; HEURRC-NEXT: global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 sc0 sc1 +; HEURRC-NEXT: global_store_dwordx4 v40, v[12:15], s[0:1] offset:48 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) -; HEURRC-NEXT: global_store_dwordx4 v0, a[0:3], s[0:1] sc0 sc1 +; HEURRC-NEXT: global_store_dwordx4 v40, v[0:3], s[0:1] sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) -; HEURRC-NEXT: global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 sc0 sc1 +; HEURRC-NEXT: global_store_dwordx4 v40, v[4:7], s[0:1] offset:16 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) ; HEURRC-NEXT: s_endpgm ; @@ -4442,17 +4312,17 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd(<4 x i32> %arg0, <4 ; VGPRRC: ; %bb.0: ; VGPRRC-NEXT: s_load_dwordx8 s[20:27], s[4:5], 0x24 ; VGPRRC-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 -; VGPRRC-NEXT: v_mov_b32_e32 v32, 0 +; VGPRRC-NEXT: v_mov_b32_e32 v40, 0 ; VGPRRC-NEXT: s_waitcnt lgkmcnt(0) -; VGPRRC-NEXT: v_mov_b32_e32 v34, s20 -; VGPRRC-NEXT: v_mov_b32_e32 v35, s21 -; VGPRRC-NEXT: v_mov_b32_e32 v36, s22 -; VGPRRC-NEXT: v_mov_b32_e32 v37, s23 +; VGPRRC-NEXT: v_mov_b32_e32 v32, s20 +; VGPRRC-NEXT: v_mov_b32_e32 v33, s21 +; VGPRRC-NEXT: v_mov_b32_e32 v34, s22 +; VGPRRC-NEXT: v_mov_b32_e32 v35, s23 ; VGPRRC-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 -; VGPRRC-NEXT: v_mov_b32_e32 v38, s24 -; VGPRRC-NEXT: v_mov_b32_e32 v39, s25 -; VGPRRC-NEXT: v_mov_b32_e32 v40, s26 -; VGPRRC-NEXT: v_mov_b32_e32 v41, s27 +; VGPRRC-NEXT: v_mov_b32_e32 v36, s24 +; VGPRRC-NEXT: v_mov_b32_e32 v37, s25 +; VGPRRC-NEXT: v_mov_b32_e32 v38, s26 +; VGPRRC-NEXT: v_mov_b32_e32 v39, s27 ; VGPRRC-NEXT: s_waitcnt lgkmcnt(0) ; VGPRRC-NEXT: v_mov_b64_e32 v[30:31], s[22:23] ; VGPRRC-NEXT: v_mov_b64_e32 v[28:29], s[20:21] @@ -4463,42 +4333,42 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd(<4 x i32> %arg0, <4 ; VGPRRC-NEXT: v_mov_b64_e32 v[18:19], s[10:11] ; VGPRRC-NEXT: v_mov_b64_e32 v[16:17], s[8:9] ; VGPRRC-NEXT: s_nop 1 -; VGPRRC-NEXT: v_mfma_i32_32x32x32_i8 v[0:15], v[34:37], v[38:41], v[16:31] +; VGPRRC-NEXT: v_mfma_i32_32x32x32_i8 v[0:15], v[32:35], v[36:39], v[16:31] ; VGPRRC-NEXT: s_nop 6 ; VGPRRC-NEXT: v_mov_b32_e32 v16, s20 ; VGPRRC-NEXT: v_mov_b32_e32 v17, s21 ; VGPRRC-NEXT: v_mov_b32_e32 v18, s22 ; VGPRRC-NEXT: v_mov_b32_e32 v19, s23 -; VGPRRC-NEXT: global_store_dwordx4 v32, v[16:19], s[0:1] offset:48 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:48 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) ; VGPRRC-NEXT: s_nop 0 ; VGPRRC-NEXT: v_mov_b32_e32 v16, s16 ; VGPRRC-NEXT: v_mov_b32_e32 v17, s17 ; VGPRRC-NEXT: v_mov_b32_e32 v18, s18 ; VGPRRC-NEXT: v_mov_b32_e32 v19, s19 -; VGPRRC-NEXT: global_store_dwordx4 v32, v[16:19], s[0:1] offset:32 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:32 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) ; VGPRRC-NEXT: s_nop 0 ; VGPRRC-NEXT: v_mov_b32_e32 v16, s12 ; VGPRRC-NEXT: v_mov_b32_e32 v17, s13 ; VGPRRC-NEXT: v_mov_b32_e32 v18, s14 ; VGPRRC-NEXT: v_mov_b32_e32 v19, s15 -; VGPRRC-NEXT: global_store_dwordx4 v32, v[16:19], s[0:1] offset:16 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:16 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) ; VGPRRC-NEXT: s_nop 0 ; VGPRRC-NEXT: v_mov_b32_e32 v16, s8 ; VGPRRC-NEXT: v_mov_b32_e32 v17, s9 ; VGPRRC-NEXT: v_mov_b32_e32 v18, s10 ; VGPRRC-NEXT: v_mov_b32_e32 v19, s11 -; VGPRRC-NEXT: global_store_dwordx4 v32, v[16:19], s[0:1] sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) -; VGPRRC-NEXT: global_store_dwordx4 v32, v[8:11], s[0:1] offset:32 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v40, v[8:11], s[0:1] offset:32 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) -; VGPRRC-NEXT: global_store_dwordx4 v32, v[12:15], s[0:1] offset:48 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v40, v[12:15], s[0:1] offset:48 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) -; VGPRRC-NEXT: global_store_dwordx4 v32, v[0:3], s[0:1] sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v40, v[0:3], s[0:1] sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) -; VGPRRC-NEXT: global_store_dwordx4 v32, v[4:7], s[0:1] offset:16 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v40, v[4:7], s[0:1] offset:16 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) ; VGPRRC-NEXT: s_endpgm ; AGPR-LABEL: test_mfma_i32_32x32x32_i8__vgprcd: @@ -4645,70 +4515,63 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd__flags(<4 x i32> %a ; SDAG: ; %bb.0: ; SDAG-NEXT: s_load_dwordx8 s[20:27], s[4:5], 0x24 ; SDAG-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 -; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: v_mov_b32_e32 v40, 0 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v2, s20 -; SDAG-NEXT: v_mov_b32_e32 v3, s21 -; SDAG-NEXT: v_mov_b32_e32 v4, s22 -; SDAG-NEXT: v_mov_b32_e32 v5, s23 +; SDAG-NEXT: v_mov_b32_e32 v32, s20 +; SDAG-NEXT: v_mov_b32_e32 v33, s21 +; SDAG-NEXT: v_mov_b32_e32 v34, s22 +; SDAG-NEXT: v_mov_b32_e32 v35, s23 ; SDAG-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 -; SDAG-NEXT: v_mov_b32_e32 v6, s24 -; SDAG-NEXT: v_mov_b32_e32 v7, s25 -; SDAG-NEXT: v_mov_b32_e32 v8, s26 -; SDAG-NEXT: v_mov_b32_e32 v9, s27 +; SDAG-NEXT: v_mov_b32_e32 v36, s24 +; SDAG-NEXT: v_mov_b32_e32 v37, s25 +; SDAG-NEXT: v_mov_b32_e32 v38, s26 +; SDAG-NEXT: v_mov_b32_e32 v39, s27 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_accvgpr_write_b32 a31, s23 -; SDAG-NEXT: v_accvgpr_write_b32 a30, s22 -; SDAG-NEXT: v_accvgpr_write_b32 a29, s21 -; SDAG-NEXT: v_accvgpr_write_b32 a28, s20 -; SDAG-NEXT: v_accvgpr_write_b32 a27, s19 -; SDAG-NEXT: v_accvgpr_write_b32 a26, s18 -; SDAG-NEXT: v_accvgpr_write_b32 a25, s17 -; SDAG-NEXT: v_accvgpr_write_b32 a24, s16 -; SDAG-NEXT: v_accvgpr_write_b32 a23, s15 -; SDAG-NEXT: v_accvgpr_write_b32 a22, s14 -; SDAG-NEXT: v_accvgpr_write_b32 a21, s13 -; SDAG-NEXT: v_accvgpr_write_b32 a20, s12 -; SDAG-NEXT: v_accvgpr_write_b32 a19, s11 -; SDAG-NEXT: v_accvgpr_write_b32 a18, s10 -; SDAG-NEXT: v_accvgpr_write_b32 a17, s9 -; SDAG-NEXT: v_accvgpr_write_b32 a16, s8 +; SDAG-NEXT: v_mov_b64_e32 v[30:31], s[22:23] +; SDAG-NEXT: v_mov_b64_e32 v[28:29], s[20:21] +; SDAG-NEXT: v_mov_b64_e32 v[26:27], s[18:19] +; SDAG-NEXT: v_mov_b64_e32 v[24:25], s[16:17] +; SDAG-NEXT: v_mov_b64_e32 v[22:23], s[14:15] +; SDAG-NEXT: v_mov_b64_e32 v[20:21], s[12:13] +; SDAG-NEXT: v_mov_b64_e32 v[18:19], s[10:11] +; SDAG-NEXT: v_mov_b64_e32 v[16:17], s[8:9] ; SDAG-NEXT: s_nop 1 -; SDAG-NEXT: v_mfma_i32_32x32x32_i8 a[0:15], v[2:5], v[6:9], a[16:31] cbsz:1 abid:2 blgp:3 -; SDAG-NEXT: v_mov_b32_e32 v2, s20 -; SDAG-NEXT: v_mov_b32_e32 v3, s21 -; SDAG-NEXT: v_mov_b32_e32 v4, s22 -; SDAG-NEXT: v_mov_b32_e32 v5, s23 -; SDAG-NEXT: global_store_dwordx4 v0, v[2:5], s[0:1] offset:48 sc0 sc1 +; SDAG-NEXT: v_mfma_i32_32x32x32_i8 v[0:15], v[32:35], v[36:39], v[16:31] cbsz:1 abid:2 blgp:3 +; SDAG-NEXT: s_nop 6 +; SDAG-NEXT: v_mov_b32_e32 v16, s20 +; SDAG-NEXT: v_mov_b32_e32 v17, s21 +; SDAG-NEXT: v_mov_b32_e32 v18, s22 +; SDAG-NEXT: v_mov_b32_e32 v19, s23 +; SDAG-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:48 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_mov_b32_e32 v2, s16 -; SDAG-NEXT: v_mov_b32_e32 v3, s17 -; SDAG-NEXT: v_mov_b32_e32 v4, s18 -; SDAG-NEXT: v_mov_b32_e32 v5, s19 -; SDAG-NEXT: global_store_dwordx4 v0, v[2:5], s[0:1] offset:32 sc0 sc1 +; SDAG-NEXT: v_mov_b32_e32 v16, s16 +; SDAG-NEXT: v_mov_b32_e32 v17, s17 +; SDAG-NEXT: v_mov_b32_e32 v18, s18 +; SDAG-NEXT: v_mov_b32_e32 v19, s19 +; SDAG-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:32 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_mov_b32_e32 v2, s12 -; SDAG-NEXT: v_mov_b32_e32 v3, s13 -; SDAG-NEXT: v_mov_b32_e32 v4, s14 -; SDAG-NEXT: v_mov_b32_e32 v5, s15 -; SDAG-NEXT: global_store_dwordx4 v0, v[2:5], s[0:1] offset:16 sc0 sc1 +; SDAG-NEXT: v_mov_b32_e32 v16, s12 +; SDAG-NEXT: v_mov_b32_e32 v17, s13 +; SDAG-NEXT: v_mov_b32_e32 v18, s14 +; SDAG-NEXT: v_mov_b32_e32 v19, s15 +; SDAG-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:16 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_mov_b32_e32 v2, s8 -; SDAG-NEXT: v_mov_b32_e32 v3, s9 -; SDAG-NEXT: v_mov_b32_e32 v4, s10 -; SDAG-NEXT: v_mov_b32_e32 v5, s11 -; SDAG-NEXT: global_store_dwordx4 v0, v[2:5], s[0:1] sc0 sc1 +; SDAG-NEXT: v_mov_b32_e32 v16, s8 +; SDAG-NEXT: v_mov_b32_e32 v17, s9 +; SDAG-NEXT: v_mov_b32_e32 v18, s10 +; SDAG-NEXT: v_mov_b32_e32 v19, s11 +; SDAG-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v40, v[8:11], s[0:1] offset:32 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v40, v[12:15], s[0:1] offset:48 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v0, a[0:3], s[0:1] sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v40, v[0:3], s[0:1] sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v40, v[4:7], s[0:1] offset:16 sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_endpgm ; @@ -4717,52 +4580,44 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd__flags(<4 x i32> %a ; GISEL-NEXT: s_load_dwordx8 s[24:31], s[4:5], 0x24 ; GISEL-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; GISEL-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 -; GISEL-NEXT: v_mov_b32_e32 v24, 0 +; GISEL-NEXT: v_mov_b32_e32 v56, 0 ; GISEL-NEXT: s_waitcnt lgkmcnt(0) -; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[24:25] -; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[26:27] -; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[28:29] -; GISEL-NEXT: v_accvgpr_write_b32 a0, s8 -; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[30:31] -; GISEL-NEXT: v_accvgpr_write_b32 a1, s9 -; GISEL-NEXT: v_accvgpr_write_b32 a2, s10 -; GISEL-NEXT: v_accvgpr_write_b32 a3, s11 -; GISEL-NEXT: v_accvgpr_write_b32 a4, s12 -; GISEL-NEXT: v_accvgpr_write_b32 a5, s13 -; GISEL-NEXT: v_accvgpr_write_b32 a6, s14 -; GISEL-NEXT: v_accvgpr_write_b32 a7, s15 -; GISEL-NEXT: v_accvgpr_write_b32 a8, s16 -; GISEL-NEXT: v_accvgpr_write_b32 a9, s17 -; GISEL-NEXT: v_accvgpr_write_b32 a10, s18 -; GISEL-NEXT: v_accvgpr_write_b32 a11, s19 -; GISEL-NEXT: v_accvgpr_write_b32 a12, s20 -; GISEL-NEXT: v_accvgpr_write_b32 a13, s21 -; GISEL-NEXT: v_accvgpr_write_b32 a14, s22 -; GISEL-NEXT: v_accvgpr_write_b32 a15, s23 -; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[8:9] -; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[10:11] -; GISEL-NEXT: v_mfma_i32_32x32x32_i8 a[16:31], v[0:3], v[4:7], a[0:15] cbsz:1 abid:2 blgp:3 -; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[12:13] -; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[16:17] -; GISEL-NEXT: v_mov_b64_e32 v[20:21], s[20:21] -; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[14:15] -; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[18:19] -; GISEL-NEXT: v_mov_b64_e32 v[22:23], s[22:23] -; GISEL-NEXT: global_store_dwordx4 v24, v[8:11], s[0:1] sc0 sc1 +; GISEL-NEXT: v_mov_b64_e32 v[34:35], s[26:27] +; GISEL-NEXT: v_mov_b64_e32 v[32:33], s[24:25] +; GISEL-NEXT: v_mov_b64_e32 v[38:39], s[30:31] +; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[8:9] +; GISEL-NEXT: v_mov_b64_e32 v[36:37], s[28:29] +; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[10:11] +; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] +; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] +; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[16:17] +; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[18:19] +; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[20:21] +; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[22:23] +; GISEL-NEXT: v_mov_b64_e32 v[42:43], s[10:11] +; GISEL-NEXT: v_mov_b64_e32 v[40:41], s[8:9] +; GISEL-NEXT: v_mfma_i32_32x32x32_i8 v[16:31], v[32:35], v[36:39], v[0:15] cbsz:1 abid:2 blgp:3 +; GISEL-NEXT: v_mov_b64_e32 v[46:47], s[14:15] +; GISEL-NEXT: v_mov_b64_e32 v[50:51], s[18:19] +; GISEL-NEXT: v_mov_b64_e32 v[54:55], s[22:23] +; GISEL-NEXT: v_mov_b64_e32 v[44:45], s[12:13] +; GISEL-NEXT: v_mov_b64_e32 v[48:49], s[16:17] +; GISEL-NEXT: v_mov_b64_e32 v[52:53], s[20:21] +; GISEL-NEXT: global_store_dwordx4 v56, v[40:43], s[0:1] sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, v[12:15], s[0:1] offset:16 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[44:47], s[0:1] offset:16 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, v[16:19], s[0:1] offset:32 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[48:51], s[0:1] offset:32 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, v[20:23], s[0:1] offset:48 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[52:55], s[0:1] offset:48 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, a[16:19], s[0:1] sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[16:19], s[0:1] sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, a[20:23], s[0:1] offset:16 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[20:23], s[0:1] offset:16 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, a[24:27], s[0:1] offset:32 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[24:27], s[0:1] offset:32 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v24, a[28:31], s[0:1] offset:48 sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v56, v[28:31], s[0:1] offset:48 sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) ; GISEL-NEXT: s_endpgm ; @@ -4770,70 +4625,63 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd__flags(<4 x i32> %a ; HEURRC: ; %bb.0: ; HEURRC-NEXT: s_load_dwordx8 s[20:27], s[4:5], 0x24 ; HEURRC-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 -; HEURRC-NEXT: v_mov_b32_e32 v0, 0 +; HEURRC-NEXT: v_mov_b32_e32 v40, 0 ; HEURRC-NEXT: s_waitcnt lgkmcnt(0) -; HEURRC-NEXT: v_mov_b32_e32 v2, s20 -; HEURRC-NEXT: v_mov_b32_e32 v3, s21 -; HEURRC-NEXT: v_mov_b32_e32 v4, s22 -; HEURRC-NEXT: v_mov_b32_e32 v5, s23 +; HEURRC-NEXT: v_mov_b32_e32 v32, s20 +; HEURRC-NEXT: v_mov_b32_e32 v33, s21 +; HEURRC-NEXT: v_mov_b32_e32 v34, s22 +; HEURRC-NEXT: v_mov_b32_e32 v35, s23 ; HEURRC-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 -; HEURRC-NEXT: v_mov_b32_e32 v6, s24 -; HEURRC-NEXT: v_mov_b32_e32 v7, s25 -; HEURRC-NEXT: v_mov_b32_e32 v8, s26 -; HEURRC-NEXT: v_mov_b32_e32 v9, s27 +; HEURRC-NEXT: v_mov_b32_e32 v36, s24 +; HEURRC-NEXT: v_mov_b32_e32 v37, s25 +; HEURRC-NEXT: v_mov_b32_e32 v38, s26 +; HEURRC-NEXT: v_mov_b32_e32 v39, s27 ; HEURRC-NEXT: s_waitcnt lgkmcnt(0) -; HEURRC-NEXT: v_accvgpr_write_b32 a31, s23 -; HEURRC-NEXT: v_accvgpr_write_b32 a30, s22 -; HEURRC-NEXT: v_accvgpr_write_b32 a29, s21 -; HEURRC-NEXT: v_accvgpr_write_b32 a28, s20 -; HEURRC-NEXT: v_accvgpr_write_b32 a27, s19 -; HEURRC-NEXT: v_accvgpr_write_b32 a26, s18 -; HEURRC-NEXT: v_accvgpr_write_b32 a25, s17 -; HEURRC-NEXT: v_accvgpr_write_b32 a24, s16 -; HEURRC-NEXT: v_accvgpr_write_b32 a23, s15 -; HEURRC-NEXT: v_accvgpr_write_b32 a22, s14 -; HEURRC-NEXT: v_accvgpr_write_b32 a21, s13 -; HEURRC-NEXT: v_accvgpr_write_b32 a20, s12 -; HEURRC-NEXT: v_accvgpr_write_b32 a19, s11 -; HEURRC-NEXT: v_accvgpr_write_b32 a18, s10 -; HEURRC-NEXT: v_accvgpr_write_b32 a17, s9 -; HEURRC-NEXT: v_accvgpr_write_b32 a16, s8 +; HEURRC-NEXT: v_mov_b64_e32 v[30:31], s[22:23] +; HEURRC-NEXT: v_mov_b64_e32 v[28:29], s[20:21] +; HEURRC-NEXT: v_mov_b64_e32 v[26:27], s[18:19] +; HEURRC-NEXT: v_mov_b64_e32 v[24:25], s[16:17] +; HEURRC-NEXT: v_mov_b64_e32 v[22:23], s[14:15] +; HEURRC-NEXT: v_mov_b64_e32 v[20:21], s[12:13] +; HEURRC-NEXT: v_mov_b64_e32 v[18:19], s[10:11] +; HEURRC-NEXT: v_mov_b64_e32 v[16:17], s[8:9] ; HEURRC-NEXT: s_nop 1 -; HEURRC-NEXT: v_mfma_i32_32x32x32_i8 a[0:15], v[2:5], v[6:9], a[16:31] cbsz:1 abid:2 blgp:3 -; HEURRC-NEXT: v_mov_b32_e32 v2, s20 -; HEURRC-NEXT: v_mov_b32_e32 v3, s21 -; HEURRC-NEXT: v_mov_b32_e32 v4, s22 -; HEURRC-NEXT: v_mov_b32_e32 v5, s23 -; HEURRC-NEXT: global_store_dwordx4 v0, v[2:5], s[0:1] offset:48 sc0 sc1 +; HEURRC-NEXT: v_mfma_i32_32x32x32_i8 v[0:15], v[32:35], v[36:39], v[16:31] cbsz:1 abid:2 blgp:3 +; HEURRC-NEXT: s_nop 6 +; HEURRC-NEXT: v_mov_b32_e32 v16, s20 +; HEURRC-NEXT: v_mov_b32_e32 v17, s21 +; HEURRC-NEXT: v_mov_b32_e32 v18, s22 +; HEURRC-NEXT: v_mov_b32_e32 v19, s23 +; HEURRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:48 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) ; HEURRC-NEXT: s_nop 0 -; HEURRC-NEXT: v_mov_b32_e32 v2, s16 -; HEURRC-NEXT: v_mov_b32_e32 v3, s17 -; HEURRC-NEXT: v_mov_b32_e32 v4, s18 -; HEURRC-NEXT: v_mov_b32_e32 v5, s19 -; HEURRC-NEXT: global_store_dwordx4 v0, v[2:5], s[0:1] offset:32 sc0 sc1 +; HEURRC-NEXT: v_mov_b32_e32 v16, s16 +; HEURRC-NEXT: v_mov_b32_e32 v17, s17 +; HEURRC-NEXT: v_mov_b32_e32 v18, s18 +; HEURRC-NEXT: v_mov_b32_e32 v19, s19 +; HEURRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:32 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) ; HEURRC-NEXT: s_nop 0 -; HEURRC-NEXT: v_mov_b32_e32 v2, s12 -; HEURRC-NEXT: v_mov_b32_e32 v3, s13 -; HEURRC-NEXT: v_mov_b32_e32 v4, s14 -; HEURRC-NEXT: v_mov_b32_e32 v5, s15 -; HEURRC-NEXT: global_store_dwordx4 v0, v[2:5], s[0:1] offset:16 sc0 sc1 +; HEURRC-NEXT: v_mov_b32_e32 v16, s12 +; HEURRC-NEXT: v_mov_b32_e32 v17, s13 +; HEURRC-NEXT: v_mov_b32_e32 v18, s14 +; HEURRC-NEXT: v_mov_b32_e32 v19, s15 +; HEURRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:16 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) ; HEURRC-NEXT: s_nop 0 -; HEURRC-NEXT: v_mov_b32_e32 v2, s8 -; HEURRC-NEXT: v_mov_b32_e32 v3, s9 -; HEURRC-NEXT: v_mov_b32_e32 v4, s10 -; HEURRC-NEXT: v_mov_b32_e32 v5, s11 -; HEURRC-NEXT: global_store_dwordx4 v0, v[2:5], s[0:1] sc0 sc1 +; HEURRC-NEXT: v_mov_b32_e32 v16, s8 +; HEURRC-NEXT: v_mov_b32_e32 v17, s9 +; HEURRC-NEXT: v_mov_b32_e32 v18, s10 +; HEURRC-NEXT: v_mov_b32_e32 v19, s11 +; HEURRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) -; HEURRC-NEXT: global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 sc0 sc1 +; HEURRC-NEXT: global_store_dwordx4 v40, v[8:11], s[0:1] offset:32 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) -; HEURRC-NEXT: global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 sc0 sc1 +; HEURRC-NEXT: global_store_dwordx4 v40, v[12:15], s[0:1] offset:48 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) -; HEURRC-NEXT: global_store_dwordx4 v0, a[0:3], s[0:1] sc0 sc1 +; HEURRC-NEXT: global_store_dwordx4 v40, v[0:3], s[0:1] sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) -; HEURRC-NEXT: global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 sc0 sc1 +; HEURRC-NEXT: global_store_dwordx4 v40, v[4:7], s[0:1] offset:16 sc0 sc1 ; HEURRC-NEXT: s_waitcnt vmcnt(0) ; HEURRC-NEXT: s_endpgm ; @@ -4841,17 +4689,17 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd__flags(<4 x i32> %a ; VGPRRC: ; %bb.0: ; VGPRRC-NEXT: s_load_dwordx8 s[20:27], s[4:5], 0x24 ; VGPRRC-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 -; VGPRRC-NEXT: v_mov_b32_e32 v32, 0 +; VGPRRC-NEXT: v_mov_b32_e32 v40, 0 ; VGPRRC-NEXT: s_waitcnt lgkmcnt(0) -; VGPRRC-NEXT: v_mov_b32_e32 v34, s20 -; VGPRRC-NEXT: v_mov_b32_e32 v35, s21 -; VGPRRC-NEXT: v_mov_b32_e32 v36, s22 -; VGPRRC-NEXT: v_mov_b32_e32 v37, s23 +; VGPRRC-NEXT: v_mov_b32_e32 v32, s20 +; VGPRRC-NEXT: v_mov_b32_e32 v33, s21 +; VGPRRC-NEXT: v_mov_b32_e32 v34, s22 +; VGPRRC-NEXT: v_mov_b32_e32 v35, s23 ; VGPRRC-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 -; VGPRRC-NEXT: v_mov_b32_e32 v38, s24 -; VGPRRC-NEXT: v_mov_b32_e32 v39, s25 -; VGPRRC-NEXT: v_mov_b32_e32 v40, s26 -; VGPRRC-NEXT: v_mov_b32_e32 v41, s27 +; VGPRRC-NEXT: v_mov_b32_e32 v36, s24 +; VGPRRC-NEXT: v_mov_b32_e32 v37, s25 +; VGPRRC-NEXT: v_mov_b32_e32 v38, s26 +; VGPRRC-NEXT: v_mov_b32_e32 v39, s27 ; VGPRRC-NEXT: s_waitcnt lgkmcnt(0) ; VGPRRC-NEXT: v_mov_b64_e32 v[30:31], s[22:23] ; VGPRRC-NEXT: v_mov_b64_e32 v[28:29], s[20:21] @@ -4862,42 +4710,42 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd__flags(<4 x i32> %a ; VGPRRC-NEXT: v_mov_b64_e32 v[18:19], s[10:11] ; VGPRRC-NEXT: v_mov_b64_e32 v[16:17], s[8:9] ; VGPRRC-NEXT: s_nop 1 -; VGPRRC-NEXT: v_mfma_i32_32x32x32_i8 v[0:15], v[34:37], v[38:41], v[16:31] cbsz:1 abid:2 blgp:3 +; VGPRRC-NEXT: v_mfma_i32_32x32x32_i8 v[0:15], v[32:35], v[36:39], v[16:31] cbsz:1 abid:2 blgp:3 ; VGPRRC-NEXT: s_nop 6 ; VGPRRC-NEXT: v_mov_b32_e32 v16, s20 ; VGPRRC-NEXT: v_mov_b32_e32 v17, s21 ; VGPRRC-NEXT: v_mov_b32_e32 v18, s22 ; VGPRRC-NEXT: v_mov_b32_e32 v19, s23 -; VGPRRC-NEXT: global_store_dwordx4 v32, v[16:19], s[0:1] offset:48 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:48 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) ; VGPRRC-NEXT: s_nop 0 ; VGPRRC-NEXT: v_mov_b32_e32 v16, s16 ; VGPRRC-NEXT: v_mov_b32_e32 v17, s17 ; VGPRRC-NEXT: v_mov_b32_e32 v18, s18 ; VGPRRC-NEXT: v_mov_b32_e32 v19, s19 -; VGPRRC-NEXT: global_store_dwordx4 v32, v[16:19], s[0:1] offset:32 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:32 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) ; VGPRRC-NEXT: s_nop 0 ; VGPRRC-NEXT: v_mov_b32_e32 v16, s12 ; VGPRRC-NEXT: v_mov_b32_e32 v17, s13 ; VGPRRC-NEXT: v_mov_b32_e32 v18, s14 ; VGPRRC-NEXT: v_mov_b32_e32 v19, s15 -; VGPRRC-NEXT: global_store_dwordx4 v32, v[16:19], s[0:1] offset:16 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] offset:16 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) ; VGPRRC-NEXT: s_nop 0 ; VGPRRC-NEXT: v_mov_b32_e32 v16, s8 ; VGPRRC-NEXT: v_mov_b32_e32 v17, s9 ; VGPRRC-NEXT: v_mov_b32_e32 v18, s10 ; VGPRRC-NEXT: v_mov_b32_e32 v19, s11 -; VGPRRC-NEXT: global_store_dwordx4 v32, v[16:19], s[0:1] sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v40, v[16:19], s[0:1] sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) -; VGPRRC-NEXT: global_store_dwordx4 v32, v[8:11], s[0:1] offset:32 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v40, v[8:11], s[0:1] offset:32 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) -; VGPRRC-NEXT: global_store_dwordx4 v32, v[12:15], s[0:1] offset:48 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v40, v[12:15], s[0:1] offset:48 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) -; VGPRRC-NEXT: global_store_dwordx4 v32, v[0:3], s[0:1] sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v40, v[0:3], s[0:1] sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) -; VGPRRC-NEXT: global_store_dwordx4 v32, v[4:7], s[0:1] offset:16 sc0 sc1 +; VGPRRC-NEXT: global_store_dwordx4 v40, v[4:7], s[0:1] offset:16 sc0 sc1 ; VGPRRC-NEXT: s_waitcnt vmcnt(0) ; VGPRRC-NEXT: s_endpgm ; AGPR-LABEL: test_mfma_i32_32x32x32_i8__vgprcd__flags: @@ -5045,41 +4893,33 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd_mac(<4 x i32> %arg0 ; SDAG-NEXT: s_load_dwordx8 s[20:27], s[4:5], 0x24 ; SDAG-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v0, s20 -; SDAG-NEXT: v_mov_b32_e32 v1, s21 -; SDAG-NEXT: v_mov_b32_e32 v2, s22 -; SDAG-NEXT: v_mov_b32_e32 v3, s23 +; SDAG-NEXT: v_mov_b32_e32 v16, s20 +; SDAG-NEXT: v_mov_b32_e32 v17, s21 +; SDAG-NEXT: v_mov_b32_e32 v18, s22 +; SDAG-NEXT: v_mov_b32_e32 v19, s23 ; SDAG-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 -; SDAG-NEXT: v_mov_b32_e32 v4, s24 -; SDAG-NEXT: v_mov_b32_e32 v5, s25 -; SDAG-NEXT: v_mov_b32_e32 v6, s26 -; SDAG-NEXT: v_mov_b32_e32 v7, s27 +; SDAG-NEXT: v_mov_b32_e32 v20, s24 +; SDAG-NEXT: v_mov_b32_e32 v21, s25 +; SDAG-NEXT: v_mov_b32_e32 v22, s26 +; SDAG-NEXT: v_mov_b32_e32 v23, s27 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_accvgpr_write_b32 a0, s8 -; SDAG-NEXT: v_accvgpr_write_b32 a1, s9 -; SDAG-NEXT: v_accvgpr_write_b32 a2, s10 -; SDAG-NEXT: v_accvgpr_write_b32 a3, s11 -; SDAG-NEXT: v_accvgpr_write_b32 a4, s12 -; SDAG-NEXT: v_accvgpr_write_b32 a5, s13 -; SDAG-NEXT: v_accvgpr_write_b32 a6, s14 -; SDAG-NEXT: v_accvgpr_write_b32 a7, s15 -; SDAG-NEXT: v_accvgpr_write_b32 a8, s16 -; SDAG-NEXT: v_accvgpr_write_b32 a9, s17 -; SDAG-NEXT: v_accvgpr_write_b32 a10, s18 -; SDAG-NEXT: v_accvgpr_write_b32 a11, s19 -; SDAG-NEXT: v_accvgpr_write_b32 a12, s20 -; SDAG-NEXT: v_accvgpr_write_b32 a13, s21 -; SDAG-NEXT: v_accvgpr_write_b32 a14, s22 -; SDAG-NEXT: v_accvgpr_write_b32 a15, s23 +; SDAG-NEXT: v_mov_b64_e32 v[0:1], s[8:9] +; SDAG-NEXT: v_mov_b64_e32 v[2:3], s[10:11] +; SDAG-NEXT: v_mov_b64_e32 v[4:5], s[12:13] +; SDAG-NEXT: v_mov_b64_e32 v[6:7], s[14:15] +; SDAG-NEXT: v_mov_b64_e32 v[8:9], s[16:17] +; SDAG-NEXT: v_mov_b64_e32 v[10:11], s[18:19] +; SDAG-NEXT: v_mov_b64_e32 v[12:13], s[20:21] +; SDAG-NEXT: v_mov_b64_e32 v[14:15], s[22:23] ; SDAG-NEXT: s_nop 1 -; SDAG-NEXT: v_mfma_i32_32x32x32_i8 a[0:15], v[0:3], v[4:7], a[0:15] -; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: v_mfma_i32_32x32x32_i8 v[0:15], v[16:19], v[20:23], v[0:15] +; SDAG-NEXT: v_mov_b32_e32 v16, 0 ; SDAG-NEXT: s_nop 7 ; SDAG-NEXT: s_nop 2 -; SDAG-NEXT: global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 -; SDAG-NEXT: global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 -; SDAG-NEXT: global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 -; SDAG-NEXT: global_store_dwordx4 v0, a[0:3], s[0:1] +; SDAG-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1] offset:48 +; SDAG-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1] offset:32 +; SDAG-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1] offset:16 +; SDAG-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_mfma_i32_32x32x32_i8__vgprcd_mac: @@ -5088,35 +4928,27 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd_mac(<4 x i32> %arg0 ; GISEL-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; GISEL-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 ; GISEL-NEXT: s_waitcnt lgkmcnt(0) -; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[24:25] -; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[26:27] -; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[28:29] -; GISEL-NEXT: v_accvgpr_write_b32 a0, s8 -; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[30:31] -; GISEL-NEXT: v_accvgpr_write_b32 a1, s9 -; GISEL-NEXT: v_accvgpr_write_b32 a2, s10 -; GISEL-NEXT: v_accvgpr_write_b32 a3, s11 -; GISEL-NEXT: v_accvgpr_write_b32 a4, s12 -; GISEL-NEXT: v_accvgpr_write_b32 a5, s13 -; GISEL-NEXT: v_accvgpr_write_b32 a6, s14 -; GISEL-NEXT: v_accvgpr_write_b32 a7, s15 -; GISEL-NEXT: v_accvgpr_write_b32 a8, s16 -; GISEL-NEXT: v_accvgpr_write_b32 a9, s17 -; GISEL-NEXT: v_accvgpr_write_b32 a10, s18 -; GISEL-NEXT: v_accvgpr_write_b32 a11, s19 -; GISEL-NEXT: v_accvgpr_write_b32 a12, s20 -; GISEL-NEXT: v_accvgpr_write_b32 a13, s21 -; GISEL-NEXT: v_accvgpr_write_b32 a14, s22 -; GISEL-NEXT: v_accvgpr_write_b32 a15, s23 +; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[24:25] +; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[26:27] +; GISEL-NEXT: v_mov_b64_e32 v[20:21], s[28:29] +; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[8:9] +; GISEL-NEXT: v_mov_b64_e32 v[22:23], s[30:31] +; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[10:11] +; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] +; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] +; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[16:17] +; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[18:19] +; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[20:21] +; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[22:23] ; GISEL-NEXT: s_nop 1 -; GISEL-NEXT: v_mfma_i32_32x32x32_i8 a[0:15], v[0:3], v[4:7], a[0:15] -; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: v_mfma_i32_32x32x32_i8 v[0:15], v[16:19], v[20:23], v[0:15] +; GISEL-NEXT: v_mov_b32_e32 v16, 0 ; GISEL-NEXT: s_nop 7 ; GISEL-NEXT: s_nop 2 -; GISEL-NEXT: global_store_dwordx4 v0, a[0:3], s[0:1] -; GISEL-NEXT: global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 -; GISEL-NEXT: global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 -; GISEL-NEXT: global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 +; GISEL-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1] +; GISEL-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1] offset:16 +; GISEL-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1] offset:32 +; GISEL-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1] offset:48 ; GISEL-NEXT: s_endpgm ; ; HEURRC-LABEL: test_mfma_i32_32x32x32_i8__vgprcd_mac: @@ -5124,41 +4956,33 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd_mac(<4 x i32> %arg0 ; HEURRC-NEXT: s_load_dwordx8 s[20:27], s[4:5], 0x24 ; HEURRC-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 ; HEURRC-NEXT: s_waitcnt lgkmcnt(0) -; HEURRC-NEXT: v_mov_b32_e32 v0, s20 -; HEURRC-NEXT: v_mov_b32_e32 v1, s21 -; HEURRC-NEXT: v_mov_b32_e32 v2, s22 -; HEURRC-NEXT: v_mov_b32_e32 v3, s23 +; HEURRC-NEXT: v_mov_b32_e32 v16, s20 +; HEURRC-NEXT: v_mov_b32_e32 v17, s21 +; HEURRC-NEXT: v_mov_b32_e32 v18, s22 +; HEURRC-NEXT: v_mov_b32_e32 v19, s23 ; HEURRC-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 -; HEURRC-NEXT: v_mov_b32_e32 v4, s24 -; HEURRC-NEXT: v_mov_b32_e32 v5, s25 -; HEURRC-NEXT: v_mov_b32_e32 v6, s26 -; HEURRC-NEXT: v_mov_b32_e32 v7, s27 +; HEURRC-NEXT: v_mov_b32_e32 v20, s24 +; HEURRC-NEXT: v_mov_b32_e32 v21, s25 +; HEURRC-NEXT: v_mov_b32_e32 v22, s26 +; HEURRC-NEXT: v_mov_b32_e32 v23, s27 ; HEURRC-NEXT: s_waitcnt lgkmcnt(0) -; HEURRC-NEXT: v_accvgpr_write_b32 a0, s8 -; HEURRC-NEXT: v_accvgpr_write_b32 a1, s9 -; HEURRC-NEXT: v_accvgpr_write_b32 a2, s10 -; HEURRC-NEXT: v_accvgpr_write_b32 a3, s11 -; HEURRC-NEXT: v_accvgpr_write_b32 a4, s12 -; HEURRC-NEXT: v_accvgpr_write_b32 a5, s13 -; HEURRC-NEXT: v_accvgpr_write_b32 a6, s14 -; HEURRC-NEXT: v_accvgpr_write_b32 a7, s15 -; HEURRC-NEXT: v_accvgpr_write_b32 a8, s16 -; HEURRC-NEXT: v_accvgpr_write_b32 a9, s17 -; HEURRC-NEXT: v_accvgpr_write_b32 a10, s18 -; HEURRC-NEXT: v_accvgpr_write_b32 a11, s19 -; HEURRC-NEXT: v_accvgpr_write_b32 a12, s20 -; HEURRC-NEXT: v_accvgpr_write_b32 a13, s21 -; HEURRC-NEXT: v_accvgpr_write_b32 a14, s22 -; HEURRC-NEXT: v_accvgpr_write_b32 a15, s23 +; HEURRC-NEXT: v_mov_b64_e32 v[0:1], s[8:9] +; HEURRC-NEXT: v_mov_b64_e32 v[2:3], s[10:11] +; HEURRC-NEXT: v_mov_b64_e32 v[4:5], s[12:13] +; HEURRC-NEXT: v_mov_b64_e32 v[6:7], s[14:15] +; HEURRC-NEXT: v_mov_b64_e32 v[8:9], s[16:17] +; HEURRC-NEXT: v_mov_b64_e32 v[10:11], s[18:19] +; HEURRC-NEXT: v_mov_b64_e32 v[12:13], s[20:21] +; HEURRC-NEXT: v_mov_b64_e32 v[14:15], s[22:23] ; HEURRC-NEXT: s_nop 1 -; HEURRC-NEXT: v_mfma_i32_32x32x32_i8 a[0:15], v[0:3], v[4:7], a[0:15] -; HEURRC-NEXT: v_mov_b32_e32 v0, 0 +; HEURRC-NEXT: v_mfma_i32_32x32x32_i8 v[0:15], v[16:19], v[20:23], v[0:15] +; HEURRC-NEXT: v_mov_b32_e32 v16, 0 ; HEURRC-NEXT: s_nop 7 ; HEURRC-NEXT: s_nop 2 -; HEURRC-NEXT: global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 -; HEURRC-NEXT: global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 -; HEURRC-NEXT: global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 -; HEURRC-NEXT: global_store_dwordx4 v0, a[0:3], s[0:1] +; HEURRC-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1] offset:48 +; HEURRC-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1] offset:32 +; HEURRC-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1] offset:16 +; HEURRC-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1] ; HEURRC-NEXT: s_endpgm ; ; VGPRRC-LABEL: test_mfma_i32_32x32x32_i8__vgprcd_mac: @@ -5279,41 +5103,33 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd_mac_flags(<4 x i32> ; SDAG-NEXT: s_load_dwordx8 s[20:27], s[4:5], 0x24 ; SDAG-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v0, s20 -; SDAG-NEXT: v_mov_b32_e32 v1, s21 -; SDAG-NEXT: v_mov_b32_e32 v2, s22 -; SDAG-NEXT: v_mov_b32_e32 v3, s23 +; SDAG-NEXT: v_mov_b32_e32 v16, s20 +; SDAG-NEXT: v_mov_b32_e32 v17, s21 +; SDAG-NEXT: v_mov_b32_e32 v18, s22 +; SDAG-NEXT: v_mov_b32_e32 v19, s23 ; SDAG-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 -; SDAG-NEXT: v_mov_b32_e32 v4, s24 -; SDAG-NEXT: v_mov_b32_e32 v5, s25 -; SDAG-NEXT: v_mov_b32_e32 v6, s26 -; SDAG-NEXT: v_mov_b32_e32 v7, s27 +; SDAG-NEXT: v_mov_b32_e32 v20, s24 +; SDAG-NEXT: v_mov_b32_e32 v21, s25 +; SDAG-NEXT: v_mov_b32_e32 v22, s26 +; SDAG-NEXT: v_mov_b32_e32 v23, s27 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_accvgpr_write_b32 a0, s8 -; SDAG-NEXT: v_accvgpr_write_b32 a1, s9 -; SDAG-NEXT: v_accvgpr_write_b32 a2, s10 -; SDAG-NEXT: v_accvgpr_write_b32 a3, s11 -; SDAG-NEXT: v_accvgpr_write_b32 a4, s12 -; SDAG-NEXT: v_accvgpr_write_b32 a5, s13 -; SDAG-NEXT: v_accvgpr_write_b32 a6, s14 -; SDAG-NEXT: v_accvgpr_write_b32 a7, s15 -; SDAG-NEXT: v_accvgpr_write_b32 a8, s16 -; SDAG-NEXT: v_accvgpr_write_b32 a9, s17 -; SDAG-NEXT: v_accvgpr_write_b32 a10, s18 -; SDAG-NEXT: v_accvgpr_write_b32 a11, s19 -; SDAG-NEXT: v_accvgpr_write_b32 a12, s20 -; SDAG-NEXT: v_accvgpr_write_b32 a13, s21 -; SDAG-NEXT: v_accvgpr_write_b32 a14, s22 -; SDAG-NEXT: v_accvgpr_write_b32 a15, s23 +; SDAG-NEXT: v_mov_b64_e32 v[0:1], s[8:9] +; SDAG-NEXT: v_mov_b64_e32 v[2:3], s[10:11] +; SDAG-NEXT: v_mov_b64_e32 v[4:5], s[12:13] +; SDAG-NEXT: v_mov_b64_e32 v[6:7], s[14:15] +; SDAG-NEXT: v_mov_b64_e32 v[8:9], s[16:17] +; SDAG-NEXT: v_mov_b64_e32 v[10:11], s[18:19] +; SDAG-NEXT: v_mov_b64_e32 v[12:13], s[20:21] +; SDAG-NEXT: v_mov_b64_e32 v[14:15], s[22:23] ; SDAG-NEXT: s_nop 1 -; SDAG-NEXT: v_mfma_i32_32x32x32_i8 a[0:15], v[0:3], v[4:7], a[0:15] cbsz:3 abid:2 blgp:1 -; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: v_mfma_i32_32x32x32_i8 v[0:15], v[16:19], v[20:23], v[0:15] cbsz:3 abid:2 blgp:1 +; SDAG-NEXT: v_mov_b32_e32 v16, 0 ; SDAG-NEXT: s_nop 7 ; SDAG-NEXT: s_nop 2 -; SDAG-NEXT: global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 -; SDAG-NEXT: global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 -; SDAG-NEXT: global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 -; SDAG-NEXT: global_store_dwordx4 v0, a[0:3], s[0:1] +; SDAG-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1] offset:48 +; SDAG-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1] offset:32 +; SDAG-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1] offset:16 +; SDAG-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_mfma_i32_32x32x32_i8__vgprcd_mac_flags: @@ -5322,35 +5138,27 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd_mac_flags(<4 x i32> ; GISEL-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 ; GISEL-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 ; GISEL-NEXT: s_waitcnt lgkmcnt(0) -; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[24:25] -; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[26:27] -; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[28:29] -; GISEL-NEXT: v_accvgpr_write_b32 a0, s8 -; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[30:31] -; GISEL-NEXT: v_accvgpr_write_b32 a1, s9 -; GISEL-NEXT: v_accvgpr_write_b32 a2, s10 -; GISEL-NEXT: v_accvgpr_write_b32 a3, s11 -; GISEL-NEXT: v_accvgpr_write_b32 a4, s12 -; GISEL-NEXT: v_accvgpr_write_b32 a5, s13 -; GISEL-NEXT: v_accvgpr_write_b32 a6, s14 -; GISEL-NEXT: v_accvgpr_write_b32 a7, s15 -; GISEL-NEXT: v_accvgpr_write_b32 a8, s16 -; GISEL-NEXT: v_accvgpr_write_b32 a9, s17 -; GISEL-NEXT: v_accvgpr_write_b32 a10, s18 -; GISEL-NEXT: v_accvgpr_write_b32 a11, s19 -; GISEL-NEXT: v_accvgpr_write_b32 a12, s20 -; GISEL-NEXT: v_accvgpr_write_b32 a13, s21 -; GISEL-NEXT: v_accvgpr_write_b32 a14, s22 -; GISEL-NEXT: v_accvgpr_write_b32 a15, s23 +; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[24:25] +; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[26:27] +; GISEL-NEXT: v_mov_b64_e32 v[20:21], s[28:29] +; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[8:9] +; GISEL-NEXT: v_mov_b64_e32 v[22:23], s[30:31] +; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[10:11] +; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] +; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] +; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[16:17] +; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[18:19] +; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[20:21] +; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[22:23] ; GISEL-NEXT: s_nop 1 -; GISEL-NEXT: v_mfma_i32_32x32x32_i8 a[0:15], v[0:3], v[4:7], a[0:15] cbsz:3 abid:2 blgp:1 -; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: v_mfma_i32_32x32x32_i8 v[0:15], v[16:19], v[20:23], v[0:15] cbsz:3 abid:2 blgp:1 +; GISEL-NEXT: v_mov_b32_e32 v16, 0 ; GISEL-NEXT: s_nop 7 ; GISEL-NEXT: s_nop 2 -; GISEL-NEXT: global_store_dwordx4 v0, a[0:3], s[0:1] -; GISEL-NEXT: global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 -; GISEL-NEXT: global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 -; GISEL-NEXT: global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 +; GISEL-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1] +; GISEL-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1] offset:16 +; GISEL-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1] offset:32 +; GISEL-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1] offset:48 ; GISEL-NEXT: s_endpgm ; ; HEURRC-LABEL: test_mfma_i32_32x32x32_i8__vgprcd_mac_flags: @@ -5358,41 +5166,33 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd_mac_flags(<4 x i32> ; HEURRC-NEXT: s_load_dwordx8 s[20:27], s[4:5], 0x24 ; HEURRC-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xa4 ; HEURRC-NEXT: s_waitcnt lgkmcnt(0) -; HEURRC-NEXT: v_mov_b32_e32 v0, s20 -; HEURRC-NEXT: v_mov_b32_e32 v1, s21 -; HEURRC-NEXT: v_mov_b32_e32 v2, s22 -; HEURRC-NEXT: v_mov_b32_e32 v3, s23 +; HEURRC-NEXT: v_mov_b32_e32 v16, s20 +; HEURRC-NEXT: v_mov_b32_e32 v17, s21 +; HEURRC-NEXT: v_mov_b32_e32 v18, s22 +; HEURRC-NEXT: v_mov_b32_e32 v19, s23 ; HEURRC-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x64 -; HEURRC-NEXT: v_mov_b32_e32 v4, s24 -; HEURRC-NEXT: v_mov_b32_e32 v5, s25 -; HEURRC-NEXT: v_mov_b32_e32 v6, s26 -; HEURRC-NEXT: v_mov_b32_e32 v7, s27 +; HEURRC-NEXT: v_mov_b32_e32 v20, s24 +; HEURRC-NEXT: v_mov_b32_e32 v21, s25 +; HEURRC-NEXT: v_mov_b32_e32 v22, s26 +; HEURRC-NEXT: v_mov_b32_e32 v23, s27 ; HEURRC-NEXT: s_waitcnt lgkmcnt(0) -; HEURRC-NEXT: v_accvgpr_write_b32 a0, s8 -; HEURRC-NEXT: v_accvgpr_write_b32 a1, s9 -; HEURRC-NEXT: v_accvgpr_write_b32 a2, s10 -; HEURRC-NEXT: v_accvgpr_write_b32 a3, s11 -; HEURRC-NEXT: v_accvgpr_write_b32 a4, s12 -; HEURRC-NEXT: v_accvgpr_write_b32 a5, s13 -; HEURRC-NEXT: v_accvgpr_write_b32 a6, s14 -; HEURRC-NEXT: v_accvgpr_write_b32 a7, s15 -; HEURRC-NEXT: v_accvgpr_write_b32 a8, s16 -; HEURRC-NEXT: v_accvgpr_write_b32 a9, s17 -; HEURRC-NEXT: v_accvgpr_write_b32 a10, s18 -; HEURRC-NEXT: v_accvgpr_write_b32 a11, s19 -; HEURRC-NEXT: v_accvgpr_write_b32 a12, s20 -; HEURRC-NEXT: v_accvgpr_write_b32 a13, s21 -; HEURRC-NEXT: v_accvgpr_write_b32 a14, s22 -; HEURRC-NEXT: v_accvgpr_write_b32 a15, s23 +; HEURRC-NEXT: v_mov_b64_e32 v[0:1], s[8:9] +; HEURRC-NEXT: v_mov_b64_e32 v[2:3], s[10:11] +; HEURRC-NEXT: v_mov_b64_e32 v[4:5], s[12:13] +; HEURRC-NEXT: v_mov_b64_e32 v[6:7], s[14:15] +; HEURRC-NEXT: v_mov_b64_e32 v[8:9], s[16:17] +; HEURRC-NEXT: v_mov_b64_e32 v[10:11], s[18:19] +; HEURRC-NEXT: v_mov_b64_e32 v[12:13], s[20:21] +; HEURRC-NEXT: v_mov_b64_e32 v[14:15], s[22:23] ; HEURRC-NEXT: s_nop 1 -; HEURRC-NEXT: v_mfma_i32_32x32x32_i8 a[0:15], v[0:3], v[4:7], a[0:15] cbsz:3 abid:2 blgp:1 -; HEURRC-NEXT: v_mov_b32_e32 v0, 0 +; HEURRC-NEXT: v_mfma_i32_32x32x32_i8 v[0:15], v[16:19], v[20:23], v[0:15] cbsz:3 abid:2 blgp:1 +; HEURRC-NEXT: v_mov_b32_e32 v16, 0 ; HEURRC-NEXT: s_nop 7 ; HEURRC-NEXT: s_nop 2 -; HEURRC-NEXT: global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 -; HEURRC-NEXT: global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 -; HEURRC-NEXT: global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 -; HEURRC-NEXT: global_store_dwordx4 v0, a[0:3], s[0:1] +; HEURRC-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1] offset:48 +; HEURRC-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1] offset:32 +; HEURRC-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1] offset:16 +; HEURRC-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1] ; HEURRC-NEXT: s_endpgm ; ; VGPRRC-LABEL: test_mfma_i32_32x32x32_i8__vgprcd_mac_flags: @@ -5643,20 +5443,18 @@ define amdgpu_kernel void @test_mfma_f32_16x16x32_bf16_no_agpr__vgprcd(ptr addrs ; GCN-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x34 ; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 ; GCN-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x24 -; GCN-NEXT: v_mov_b32_e32 v8, 0 +; GCN-NEXT: v_mov_b32_e32 v12, 0 ; GCN-NEXT: s_waitcnt lgkmcnt(0) ; GCN-NEXT: v_mov_b64_e32 v[0:1], s[8:9] ; GCN-NEXT: v_mov_b64_e32 v[2:3], s[10:11] ; GCN-NEXT: v_mov_b64_e32 v[4:5], s[12:13] -; GCN-NEXT: v_accvgpr_write_b32 a0, s0 +; GCN-NEXT: v_mov_b64_e32 v[10:11], s[2:3] ; GCN-NEXT: v_mov_b64_e32 v[6:7], s[14:15] -; GCN-NEXT: v_accvgpr_write_b32 a1, s1 -; GCN-NEXT: v_accvgpr_write_b32 a2, s2 -; GCN-NEXT: v_accvgpr_write_b32 a3, s3 +; GCN-NEXT: v_mov_b64_e32 v[8:9], s[0:1] ; GCN-NEXT: s_nop 1 -; GCN-NEXT: v_mfma_f32_16x16x32_bf16 a[0:3], v[0:3], v[4:7], a[0:3] +; GCN-NEXT: v_mfma_f32_16x16x32_bf16 v[0:3], v[0:3], v[4:7], v[8:11] ; GCN-NEXT: s_nop 7 -; GCN-NEXT: global_store_dwordx4 v8, a[0:3], s[6:7] +; GCN-NEXT: global_store_dwordx4 v12, v[0:3], s[6:7] ; GCN-NEXT: s_endpgm ; ; HEURRC-LABEL: test_mfma_f32_16x16x32_bf16_no_agpr__vgprcd: @@ -5664,20 +5462,18 @@ define amdgpu_kernel void @test_mfma_f32_16x16x32_bf16_no_agpr__vgprcd(ptr addrs ; HEURRC-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x34 ; HEURRC-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 ; HEURRC-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x24 -; HEURRC-NEXT: v_mov_b32_e32 v8, 0 +; HEURRC-NEXT: v_mov_b32_e32 v12, 0 ; HEURRC-NEXT: s_waitcnt lgkmcnt(0) ; HEURRC-NEXT: v_mov_b64_e32 v[0:1], s[8:9] ; HEURRC-NEXT: v_mov_b64_e32 v[2:3], s[10:11] ; HEURRC-NEXT: v_mov_b64_e32 v[4:5], s[12:13] -; HEURRC-NEXT: v_accvgpr_write_b32 a0, s0 +; HEURRC-NEXT: v_mov_b64_e32 v[10:11], s[2:3] ; HEURRC-NEXT: v_mov_b64_e32 v[6:7], s[14:15] -; HEURRC-NEXT: v_accvgpr_write_b32 a1, s1 -; HEURRC-NEXT: v_accvgpr_write_b32 a2, s2 -; HEURRC-NEXT: v_accvgpr_write_b32 a3, s3 +; HEURRC-NEXT: v_mov_b64_e32 v[8:9], s[0:1] ; HEURRC-NEXT: s_nop 1 -; HEURRC-NEXT: v_mfma_f32_16x16x32_bf16 a[0:3], v[0:3], v[4:7], a[0:3] +; HEURRC-NEXT: v_mfma_f32_16x16x32_bf16 v[0:3], v[0:3], v[4:7], v[8:11] ; HEURRC-NEXT: s_nop 7 -; HEURRC-NEXT: global_store_dwordx4 v8, a[0:3], s[6:7] +; HEURRC-NEXT: global_store_dwordx4 v12, v[0:3], s[6:7] ; HEURRC-NEXT: s_endpgm ; ; VGPRRC-LABEL: test_mfma_f32_16x16x32_bf16_no_agpr__vgprcd: @@ -5747,20 +5543,18 @@ define amdgpu_kernel void @test_mfma_f32_16x16x32_bf16_no_agpr__vgprcd__flags(pt ; GCN-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x34 ; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 ; GCN-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x24 -; GCN-NEXT: v_mov_b32_e32 v8, 0 +; GCN-NEXT: v_mov_b32_e32 v12, 0 ; GCN-NEXT: s_waitcnt lgkmcnt(0) ; GCN-NEXT: v_mov_b64_e32 v[0:1], s[8:9] ; GCN-NEXT: v_mov_b64_e32 v[2:3], s[10:11] ; GCN-NEXT: v_mov_b64_e32 v[4:5], s[12:13] -; GCN-NEXT: v_accvgpr_write_b32 a0, s0 +; GCN-NEXT: v_mov_b64_e32 v[10:11], s[2:3] ; GCN-NEXT: v_mov_b64_e32 v[6:7], s[14:15] -; GCN-NEXT: v_accvgpr_write_b32 a1, s1 -; GCN-NEXT: v_accvgpr_write_b32 a2, s2 -; GCN-NEXT: v_accvgpr_write_b32 a3, s3 +; GCN-NEXT: v_mov_b64_e32 v[8:9], s[0:1] ; GCN-NEXT: s_nop 1 -; GCN-NEXT: v_mfma_f32_16x16x32_bf16 a[0:3], v[0:3], v[4:7], a[0:3] cbsz:3 abid:2 blgp:1 +; GCN-NEXT: v_mfma_f32_16x16x32_bf16 v[0:3], v[0:3], v[4:7], v[8:11] cbsz:3 abid:2 blgp:1 ; GCN-NEXT: s_nop 7 -; GCN-NEXT: global_store_dwordx4 v8, a[0:3], s[6:7] +; GCN-NEXT: global_store_dwordx4 v12, v[0:3], s[6:7] ; GCN-NEXT: s_endpgm ; ; HEURRC-LABEL: test_mfma_f32_16x16x32_bf16_no_agpr__vgprcd__flags: @@ -5768,20 +5562,18 @@ define amdgpu_kernel void @test_mfma_f32_16x16x32_bf16_no_agpr__vgprcd__flags(pt ; HEURRC-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x34 ; HEURRC-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 ; HEURRC-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x24 -; HEURRC-NEXT: v_mov_b32_e32 v8, 0 +; HEURRC-NEXT: v_mov_b32_e32 v12, 0 ; HEURRC-NEXT: s_waitcnt lgkmcnt(0) ; HEURRC-NEXT: v_mov_b64_e32 v[0:1], s[8:9] ; HEURRC-NEXT: v_mov_b64_e32 v[2:3], s[10:11] ; HEURRC-NEXT: v_mov_b64_e32 v[4:5], s[12:13] -; HEURRC-NEXT: v_accvgpr_write_b32 a0, s0 +; HEURRC-NEXT: v_mov_b64_e32 v[10:11], s[2:3] ; HEURRC-NEXT: v_mov_b64_e32 v[6:7], s[14:15] -; HEURRC-NEXT: v_accvgpr_write_b32 a1, s1 -; HEURRC-NEXT: v_accvgpr_write_b32 a2, s2 -; HEURRC-NEXT: v_accvgpr_write_b32 a3, s3 +; HEURRC-NEXT: v_mov_b64_e32 v[8:9], s[0:1] ; HEURRC-NEXT: s_nop 1 -; HEURRC-NEXT: v_mfma_f32_16x16x32_bf16 a[0:3], v[0:3], v[4:7], a[0:3] cbsz:3 abid:2 blgp:1 +; HEURRC-NEXT: v_mfma_f32_16x16x32_bf16 v[0:3], v[0:3], v[4:7], v[8:11] cbsz:3 abid:2 blgp:1 ; HEURRC-NEXT: s_nop 7 -; HEURRC-NEXT: global_store_dwordx4 v8, a[0:3], s[6:7] +; HEURRC-NEXT: global_store_dwordx4 v12, v[0:3], s[6:7] ; HEURRC-NEXT: s_endpgm ; ; VGPRRC-LABEL: test_mfma_f32_16x16x32_bf16_no_agpr__vgprcd__flags: @@ -5845,5 +5637,5 @@ define amdgpu_kernel void @test_mfma_f32_16x16x32_bf16_no_agpr__vgprcd__flags(pt ret void } -attributes #0 = { "amdgpu-flat-work-group-size"="512,512" } +attributes #0 = { "amdgpu-flat-work-group-size"="512,512" "amdgpu-agpr-alloc"="0,0" } attributes #1 = { "amdgpu-flat-work-group-size"="1,64" } diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.scale.f32.16x16x128.f8f6f4.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.scale.f32.16x16x128.f8f6f4.ll index 37809da..f78ea92 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.scale.f32.16x16x128.f8f6f4.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.scale.f32.16x16x128.f8f6f4.ll @@ -1895,36 +1895,36 @@ define amdgpu_kernel void @test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd(<8 x i32 ; SDAG-LABEL: test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd: ; SDAG: ; %bb.0: ; SDAG-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x0 -; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: v_mov_b32_e32 v20, 0 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v2, s8 -; SDAG-NEXT: v_mov_b32_e32 v3, s9 -; SDAG-NEXT: v_mov_b32_e32 v4, s10 -; SDAG-NEXT: v_mov_b32_e32 v5, s11 -; SDAG-NEXT: v_mov_b32_e32 v6, s12 -; SDAG-NEXT: v_mov_b32_e32 v7, s13 -; SDAG-NEXT: v_mov_b32_e32 v8, s14 -; SDAG-NEXT: v_mov_b32_e32 v9, s15 +; SDAG-NEXT: v_mov_b32_e32 v0, s8 +; SDAG-NEXT: v_mov_b32_e32 v1, s9 +; SDAG-NEXT: v_mov_b32_e32 v2, s10 +; SDAG-NEXT: v_mov_b32_e32 v3, s11 +; SDAG-NEXT: v_mov_b32_e32 v4, s12 +; SDAG-NEXT: v_mov_b32_e32 v5, s13 +; SDAG-NEXT: v_mov_b32_e32 v6, s14 +; SDAG-NEXT: v_mov_b32_e32 v7, s15 ; SDAG-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x40 -; SDAG-NEXT: v_mov_b32_e32 v10, s16 -; SDAG-NEXT: v_mov_b32_e32 v11, s17 -; SDAG-NEXT: v_mov_b32_e32 v12, s18 -; SDAG-NEXT: v_mov_b32_e32 v13, s19 -; SDAG-NEXT: v_mov_b32_e32 v14, s20 -; SDAG-NEXT: v_mov_b32_e32 v15, s21 -; SDAG-NEXT: v_mov_b32_e32 v16, s22 -; SDAG-NEXT: v_mov_b32_e32 v17, s23 +; SDAG-NEXT: v_mov_b32_e32 v8, s16 +; SDAG-NEXT: v_mov_b32_e32 v9, s17 +; SDAG-NEXT: v_mov_b32_e32 v10, s18 +; SDAG-NEXT: v_mov_b32_e32 v11, s19 +; SDAG-NEXT: v_mov_b32_e32 v12, s20 +; SDAG-NEXT: v_mov_b32_e32 v13, s21 +; SDAG-NEXT: v_mov_b32_e32 v14, s22 +; SDAG-NEXT: v_mov_b32_e32 v15, s23 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_accvgpr_write_b32 a0, s8 -; SDAG-NEXT: v_accvgpr_write_b32 a1, s9 -; SDAG-NEXT: v_accvgpr_write_b32 a2, s10 -; SDAG-NEXT: v_accvgpr_write_b32 a3, s11 -; SDAG-NEXT: v_mov_b32_e32 v1, s13 +; SDAG-NEXT: v_mov_b32_e32 v16, s8 +; SDAG-NEXT: v_mov_b32_e32 v17, s9 +; SDAG-NEXT: v_mov_b32_e32 v18, s10 +; SDAG-NEXT: v_mov_b32_e32 v19, s11 +; SDAG-NEXT: v_mov_b32_e32 v21, s13 ; SDAG-NEXT: s_nop 1 -; SDAG-NEXT: v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[2:9], v[10:17], a[0:3], s12, v1 op_sel:[1,1,0] op_sel_hi:[1,0,0] blgp:2 +; SDAG-NEXT: v_mfma_scale_f32_16x16x128_f8f6f4 v[0:3], v[0:7], v[8:15], v[16:19], s12, v21 op_sel:[1,1,0] op_sel_hi:[1,0,0] blgp:2 ; SDAG-NEXT: s_nop 7 ; SDAG-NEXT: s_nop 3 -; SDAG-NEXT: global_store_dwordx4 v0, a[0:3], s[14:15] +; SDAG-NEXT: global_store_dwordx4 v20, v[0:3], s[14:15] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd: @@ -1937,20 +1937,18 @@ define amdgpu_kernel void @test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd(<8 x i32 ; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] ; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] ; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[16:17] -; GISEL-NEXT: v_accvgpr_write_b32 a0, s24 +; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[24:25] ; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[18:19] ; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[20:21] ; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[22:23] -; GISEL-NEXT: v_accvgpr_write_b32 a1, s25 -; GISEL-NEXT: v_accvgpr_write_b32 a2, s26 -; GISEL-NEXT: v_accvgpr_write_b32 a3, s27 -; GISEL-NEXT: v_mov_b32_e32 v16, s29 +; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[26:27] +; GISEL-NEXT: v_mov_b32_e32 v20, s29 ; GISEL-NEXT: s_nop 1 -; GISEL-NEXT: v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[0:7], v[8:15], a[0:3], s28, v16 op_sel:[1,1,0] op_sel_hi:[1,0,0] blgp:2 -; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: v_mfma_scale_f32_16x16x128_f8f6f4 v[0:3], v[0:7], v[8:15], v[16:19], s28, v20 op_sel:[1,1,0] op_sel_hi:[1,0,0] blgp:2 +; GISEL-NEXT: v_mov_b32_e32 v4, 0 ; GISEL-NEXT: s_nop 7 ; GISEL-NEXT: s_nop 2 -; GISEL-NEXT: global_store_dwordx4 v0, a[0:3], s[30:31] +; GISEL-NEXT: global_store_dwordx4 v4, v[0:3], s[30:31] ; GISEL-NEXT: s_endpgm %result = call <4 x float> @llvm.amdgcn.mfma.scale.f32.16x16x128.f8f6f4.v8i32.v8i32(<8 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 0, i32 2, i32 3, i32 %scale0, i32 1, i32 %scale1) store <4 x float> %result, ptr addrspace(1) %ptr, align 16 @@ -1964,40 +1962,38 @@ define amdgpu_kernel void @test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd___scaleA ; SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x40 ; SDAG-NEXT: s_movk_i32 s6, 0x41 ; SDAG-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x50 -; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: v_mov_b32_e32 v20, 0 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v2, s8 -; SDAG-NEXT: v_mov_b32_e32 v3, s9 -; SDAG-NEXT: v_mov_b32_e32 v4, s10 -; SDAG-NEXT: v_mov_b32_e32 v5, s11 -; SDAG-NEXT: v_mov_b32_e32 v6, s12 -; SDAG-NEXT: v_mov_b32_e32 v7, s13 -; SDAG-NEXT: v_mov_b32_e32 v8, s14 -; SDAG-NEXT: v_mov_b32_e32 v9, s15 -; SDAG-NEXT: v_accvgpr_write_b32 a0, s0 -; SDAG-NEXT: v_mov_b32_e32 v10, s16 -; SDAG-NEXT: v_mov_b32_e32 v11, s17 -; SDAG-NEXT: v_mov_b32_e32 v12, s18 -; SDAG-NEXT: v_mov_b32_e32 v13, s19 -; SDAG-NEXT: v_mov_b32_e32 v14, s20 -; SDAG-NEXT: v_mov_b32_e32 v15, s21 -; SDAG-NEXT: v_mov_b32_e32 v16, s22 -; SDAG-NEXT: v_mov_b32_e32 v17, s23 -; SDAG-NEXT: v_accvgpr_write_b32 a1, s1 -; SDAG-NEXT: v_accvgpr_write_b32 a2, s2 -; SDAG-NEXT: v_accvgpr_write_b32 a3, s3 +; SDAG-NEXT: v_mov_b32_e32 v0, s8 +; SDAG-NEXT: v_mov_b32_e32 v1, s9 +; SDAG-NEXT: v_mov_b32_e32 v2, s10 +; SDAG-NEXT: v_mov_b32_e32 v3, s11 +; SDAG-NEXT: v_mov_b32_e32 v4, s12 +; SDAG-NEXT: v_mov_b32_e32 v5, s13 +; SDAG-NEXT: v_mov_b32_e32 v6, s14 +; SDAG-NEXT: v_mov_b32_e32 v7, s15 +; SDAG-NEXT: v_mov_b64_e32 v[18:19], s[2:3] +; SDAG-NEXT: v_mov_b32_e32 v8, s16 +; SDAG-NEXT: v_mov_b32_e32 v9, s17 +; SDAG-NEXT: v_mov_b32_e32 v10, s18 +; SDAG-NEXT: v_mov_b32_e32 v11, s19 +; SDAG-NEXT: v_mov_b32_e32 v12, s20 +; SDAG-NEXT: v_mov_b32_e32 v13, s21 +; SDAG-NEXT: v_mov_b32_e32 v14, s22 +; SDAG-NEXT: v_mov_b32_e32 v15, s23 +; SDAG-NEXT: v_mov_b64_e32 v[16:17], s[0:1] ; SDAG-NEXT: s_nop 1 -; SDAG-NEXT: v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[2:9], v[10:17], a[0:3], s6, -2 op_sel:[1,1,0] op_sel_hi:[1,0,0] +; SDAG-NEXT: v_mfma_scale_f32_16x16x128_f8f6f4 v[0:3], v[0:7], v[8:15], v[16:19], s6, -2 op_sel:[1,1,0] op_sel_hi:[1,0,0] ; SDAG-NEXT: s_nop 7 ; SDAG-NEXT: s_nop 3 -; SDAG-NEXT: global_store_dwordx4 v0, a[0:3], s[4:5] +; SDAG-NEXT: global_store_dwordx4 v20, v[0:3], s[4:5] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd___scaleA_kimm__scaleB__inlineimm: ; GISEL: ; %bb.0: ; GISEL-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x0 ; GISEL-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x40 -; GISEL-NEXT: v_mov_b32_e32 v16, 0x41 +; GISEL-NEXT: v_mov_b32_e32 v20, 0x41 ; GISEL-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x50 ; GISEL-NEXT: s_waitcnt lgkmcnt(0) ; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[8:9] @@ -2005,19 +2001,17 @@ define amdgpu_kernel void @test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd___scaleA ; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] ; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] ; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[16:17] -; GISEL-NEXT: v_accvgpr_write_b32 a0, s0 +; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[2:3] ; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[18:19] ; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[20:21] ; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[22:23] -; GISEL-NEXT: v_accvgpr_write_b32 a1, s1 -; GISEL-NEXT: v_accvgpr_write_b32 a2, s2 -; GISEL-NEXT: v_accvgpr_write_b32 a3, s3 +; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[0:1] ; GISEL-NEXT: s_nop 1 -; GISEL-NEXT: v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[0:7], v[8:15], a[0:3], v16, -2 op_sel:[1,1,0] op_sel_hi:[1,0,0] -; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: v_mfma_scale_f32_16x16x128_f8f6f4 v[0:3], v[0:7], v[8:15], v[16:19], v20, -2 op_sel:[1,1,0] op_sel_hi:[1,0,0] +; GISEL-NEXT: v_mov_b32_e32 v4, 0 ; GISEL-NEXT: s_nop 7 ; GISEL-NEXT: s_nop 2 -; GISEL-NEXT: global_store_dwordx4 v0, a[0:3], s[4:5] +; GISEL-NEXT: global_store_dwordx4 v4, v[0:3], s[4:5] ; GISEL-NEXT: s_endpgm %result = call <4 x float> @llvm.amdgcn.mfma.scale.f32.16x16x128.f8f6f4.v8i32.v8i32(<8 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 0, i32 0, i32 3, i32 65, i32 1, i32 -2) store <4 x float> %result, ptr addrspace(1) %ptr, align 16 @@ -2031,40 +2025,38 @@ define amdgpu_kernel void @test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd___scaleA ; SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x40 ; SDAG-NEXT: s_movk_i32 s6, 0x41 ; SDAG-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x50 -; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: v_mov_b32_e32 v20, 0 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v2, s8 -; SDAG-NEXT: v_mov_b32_e32 v3, s9 -; SDAG-NEXT: v_mov_b32_e32 v4, s10 -; SDAG-NEXT: v_mov_b32_e32 v5, s11 -; SDAG-NEXT: v_mov_b32_e32 v6, s12 -; SDAG-NEXT: v_mov_b32_e32 v7, s13 -; SDAG-NEXT: v_mov_b32_e32 v8, s14 -; SDAG-NEXT: v_mov_b32_e32 v9, s15 -; SDAG-NEXT: v_accvgpr_write_b32 a0, s0 -; SDAG-NEXT: v_mov_b32_e32 v10, s16 -; SDAG-NEXT: v_mov_b32_e32 v11, s17 -; SDAG-NEXT: v_mov_b32_e32 v12, s18 -; SDAG-NEXT: v_mov_b32_e32 v13, s19 -; SDAG-NEXT: v_mov_b32_e32 v14, s20 -; SDAG-NEXT: v_mov_b32_e32 v15, s21 -; SDAG-NEXT: v_mov_b32_e32 v16, s22 -; SDAG-NEXT: v_mov_b32_e32 v17, s23 -; SDAG-NEXT: v_accvgpr_write_b32 a1, s1 -; SDAG-NEXT: v_accvgpr_write_b32 a2, s2 -; SDAG-NEXT: v_accvgpr_write_b32 a3, s3 +; SDAG-NEXT: v_mov_b32_e32 v0, s8 +; SDAG-NEXT: v_mov_b32_e32 v1, s9 +; SDAG-NEXT: v_mov_b32_e32 v2, s10 +; SDAG-NEXT: v_mov_b32_e32 v3, s11 +; SDAG-NEXT: v_mov_b32_e32 v4, s12 +; SDAG-NEXT: v_mov_b32_e32 v5, s13 +; SDAG-NEXT: v_mov_b32_e32 v6, s14 +; SDAG-NEXT: v_mov_b32_e32 v7, s15 +; SDAG-NEXT: v_mov_b64_e32 v[18:19], s[2:3] +; SDAG-NEXT: v_mov_b32_e32 v8, s16 +; SDAG-NEXT: v_mov_b32_e32 v9, s17 +; SDAG-NEXT: v_mov_b32_e32 v10, s18 +; SDAG-NEXT: v_mov_b32_e32 v11, s19 +; SDAG-NEXT: v_mov_b32_e32 v12, s20 +; SDAG-NEXT: v_mov_b32_e32 v13, s21 +; SDAG-NEXT: v_mov_b32_e32 v14, s22 +; SDAG-NEXT: v_mov_b32_e32 v15, s23 +; SDAG-NEXT: v_mov_b64_e32 v[16:17], s[0:1] ; SDAG-NEXT: s_nop 1 -; SDAG-NEXT: v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[2:9], v[10:17], a[0:3], s6, 1.0 op_sel:[1,1,0] op_sel_hi:[1,0,0] +; SDAG-NEXT: v_mfma_scale_f32_16x16x128_f8f6f4 v[0:3], v[0:7], v[8:15], v[16:19], s6, 1.0 op_sel:[1,1,0] op_sel_hi:[1,0,0] ; SDAG-NEXT: s_nop 7 ; SDAG-NEXT: s_nop 3 -; SDAG-NEXT: global_store_dwordx4 v0, a[0:3], s[4:5] +; SDAG-NEXT: global_store_dwordx4 v20, v[0:3], s[4:5] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd___scaleA_kimm__scaleB__FP_literal: ; GISEL: ; %bb.0: ; GISEL-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x0 ; GISEL-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x40 -; GISEL-NEXT: v_mov_b32_e32 v16, 0x41 +; GISEL-NEXT: v_mov_b32_e32 v20, 0x41 ; GISEL-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x50 ; GISEL-NEXT: s_waitcnt lgkmcnt(0) ; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[8:9] @@ -2072,19 +2064,17 @@ define amdgpu_kernel void @test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd___scaleA ; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] ; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] ; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[16:17] -; GISEL-NEXT: v_accvgpr_write_b32 a0, s0 +; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[2:3] ; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[18:19] ; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[20:21] ; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[22:23] -; GISEL-NEXT: v_accvgpr_write_b32 a1, s1 -; GISEL-NEXT: v_accvgpr_write_b32 a2, s2 -; GISEL-NEXT: v_accvgpr_write_b32 a3, s3 +; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[0:1] ; GISEL-NEXT: s_nop 1 -; GISEL-NEXT: v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[0:7], v[8:15], a[0:3], v16, 1.0 op_sel:[1,1,0] op_sel_hi:[1,0,0] -; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: v_mfma_scale_f32_16x16x128_f8f6f4 v[0:3], v[0:7], v[8:15], v[16:19], v20, 1.0 op_sel:[1,1,0] op_sel_hi:[1,0,0] +; GISEL-NEXT: v_mov_b32_e32 v4, 0 ; GISEL-NEXT: s_nop 7 ; GISEL-NEXT: s_nop 2 -; GISEL-NEXT: global_store_dwordx4 v0, a[0:3], s[4:5] +; GISEL-NEXT: global_store_dwordx4 v4, v[0:3], s[4:5] ; GISEL-NEXT: s_endpgm %result = call <4 x float> @llvm.amdgcn.mfma.scale.f32.16x16x128.f8f6f4.v8i32.v8i32(<8 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 0, i32 0, i32 3, i32 65, i32 1, i32 1065353216) store <4 x float> %result, ptr addrspace(1) %ptr, align 16 @@ -2096,34 +2086,32 @@ define amdgpu_kernel void @test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd___scaleA ; SDAG: ; %bb.0: ; SDAG-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x0 ; SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x40 -; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: v_mov_b32_e32 v20, 0 ; SDAG-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x50 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v2, s8 -; SDAG-NEXT: v_mov_b32_e32 v3, s9 -; SDAG-NEXT: v_mov_b32_e32 v4, s10 -; SDAG-NEXT: v_mov_b32_e32 v5, s11 -; SDAG-NEXT: v_mov_b32_e32 v6, s12 -; SDAG-NEXT: v_mov_b32_e32 v7, s13 -; SDAG-NEXT: v_mov_b32_e32 v8, s14 -; SDAG-NEXT: v_mov_b32_e32 v9, s15 -; SDAG-NEXT: v_accvgpr_write_b32 a0, s0 -; SDAG-NEXT: v_mov_b32_e32 v10, s16 -; SDAG-NEXT: v_mov_b32_e32 v11, s17 -; SDAG-NEXT: v_mov_b32_e32 v12, s18 -; SDAG-NEXT: v_mov_b32_e32 v13, s19 -; SDAG-NEXT: v_mov_b32_e32 v14, s20 -; SDAG-NEXT: v_mov_b32_e32 v15, s21 -; SDAG-NEXT: v_mov_b32_e32 v16, s22 -; SDAG-NEXT: v_mov_b32_e32 v17, s23 -; SDAG-NEXT: v_accvgpr_write_b32 a1, s1 -; SDAG-NEXT: v_accvgpr_write_b32 a2, s2 -; SDAG-NEXT: v_accvgpr_write_b32 a3, s3 +; SDAG-NEXT: v_mov_b32_e32 v0, s8 +; SDAG-NEXT: v_mov_b32_e32 v1, s9 +; SDAG-NEXT: v_mov_b32_e32 v2, s10 +; SDAG-NEXT: v_mov_b32_e32 v3, s11 +; SDAG-NEXT: v_mov_b32_e32 v4, s12 +; SDAG-NEXT: v_mov_b32_e32 v5, s13 +; SDAG-NEXT: v_mov_b32_e32 v6, s14 +; SDAG-NEXT: v_mov_b32_e32 v7, s15 +; SDAG-NEXT: v_mov_b64_e32 v[18:19], s[2:3] +; SDAG-NEXT: v_mov_b32_e32 v8, s16 +; SDAG-NEXT: v_mov_b32_e32 v9, s17 +; SDAG-NEXT: v_mov_b32_e32 v10, s18 +; SDAG-NEXT: v_mov_b32_e32 v11, s19 +; SDAG-NEXT: v_mov_b32_e32 v12, s20 +; SDAG-NEXT: v_mov_b32_e32 v13, s21 +; SDAG-NEXT: v_mov_b32_e32 v14, s22 +; SDAG-NEXT: v_mov_b32_e32 v15, s23 +; SDAG-NEXT: v_mov_b64_e32 v[16:17], s[0:1] ; SDAG-NEXT: s_nop 1 -; SDAG-NEXT: v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[2:9], v[10:17], a[0:3], 1.0, -2 op_sel:[1,1,0] op_sel_hi:[1,0,0] +; SDAG-NEXT: v_mfma_scale_f32_16x16x128_f8f6f4 v[0:3], v[0:7], v[8:15], v[16:19], 1.0, -2 op_sel:[1,1,0] op_sel_hi:[1,0,0] ; SDAG-NEXT: s_nop 7 ; SDAG-NEXT: s_nop 3 -; SDAG-NEXT: global_store_dwordx4 v0, a[0:3], s[4:5] +; SDAG-NEXT: global_store_dwordx4 v20, v[0:3], s[4:5] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd___scaleA_FP_literal__scaleB__inline_imm: @@ -2136,21 +2124,19 @@ define amdgpu_kernel void @test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd___scaleA ; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] ; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] ; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[16:17] -; GISEL-NEXT: v_accvgpr_write_b32 a0, s0 +; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[2:3] ; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[18:19] ; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[20:21] ; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[22:23] -; GISEL-NEXT: v_accvgpr_write_b32 a1, s1 -; GISEL-NEXT: v_accvgpr_write_b32 a2, s2 -; GISEL-NEXT: v_accvgpr_write_b32 a3, s3 +; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[0:1] ; GISEL-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x50 ; GISEL-NEXT: s_nop 0 -; GISEL-NEXT: v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[0:7], v[8:15], a[0:3], 1.0, -2 op_sel:[1,1,0] op_sel_hi:[1,0,0] -; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: v_mfma_scale_f32_16x16x128_f8f6f4 v[0:3], v[0:7], v[8:15], v[16:19], 1.0, -2 op_sel:[1,1,0] op_sel_hi:[1,0,0] +; GISEL-NEXT: v_mov_b32_e32 v4, 0 ; GISEL-NEXT: s_waitcnt lgkmcnt(0) ; GISEL-NEXT: s_nop 7 ; GISEL-NEXT: s_nop 1 -; GISEL-NEXT: global_store_dwordx4 v0, a[0:3], s[4:5] +; GISEL-NEXT: global_store_dwordx4 v4, v[0:3], s[4:5] ; GISEL-NEXT: s_endpgm %result = call <4 x float> @llvm.amdgcn.mfma.scale.f32.16x16x128.f8f6f4.v8i32.v8i32(<8 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 0, i32 0, i32 3, i32 1065353216, i32 1, i32 -2) store <4 x float> %result, ptr addrspace(1) %ptr, align 16 @@ -2162,34 +2148,32 @@ define amdgpu_kernel void @test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd___scaleA ; SDAG: ; %bb.0: ; SDAG-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x0 ; SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x40 -; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: v_mov_b32_e32 v20, 0 ; SDAG-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x50 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v2, s8 -; SDAG-NEXT: v_mov_b32_e32 v3, s9 -; SDAG-NEXT: v_mov_b32_e32 v4, s10 -; SDAG-NEXT: v_mov_b32_e32 v5, s11 -; SDAG-NEXT: v_mov_b32_e32 v6, s12 -; SDAG-NEXT: v_mov_b32_e32 v7, s13 -; SDAG-NEXT: v_mov_b32_e32 v8, s14 -; SDAG-NEXT: v_mov_b32_e32 v9, s15 -; SDAG-NEXT: v_accvgpr_write_b32 a0, s0 -; SDAG-NEXT: v_mov_b32_e32 v10, s16 -; SDAG-NEXT: v_mov_b32_e32 v11, s17 -; SDAG-NEXT: v_mov_b32_e32 v12, s18 -; SDAG-NEXT: v_mov_b32_e32 v13, s19 -; SDAG-NEXT: v_mov_b32_e32 v14, s20 -; SDAG-NEXT: v_mov_b32_e32 v15, s21 -; SDAG-NEXT: v_mov_b32_e32 v16, s22 -; SDAG-NEXT: v_mov_b32_e32 v17, s23 -; SDAG-NEXT: v_accvgpr_write_b32 a1, s1 -; SDAG-NEXT: v_accvgpr_write_b32 a2, s2 -; SDAG-NEXT: v_accvgpr_write_b32 a3, s3 +; SDAG-NEXT: v_mov_b32_e32 v0, s8 +; SDAG-NEXT: v_mov_b32_e32 v1, s9 +; SDAG-NEXT: v_mov_b32_e32 v2, s10 +; SDAG-NEXT: v_mov_b32_e32 v3, s11 +; SDAG-NEXT: v_mov_b32_e32 v4, s12 +; SDAG-NEXT: v_mov_b32_e32 v5, s13 +; SDAG-NEXT: v_mov_b32_e32 v6, s14 +; SDAG-NEXT: v_mov_b32_e32 v7, s15 +; SDAG-NEXT: v_mov_b64_e32 v[18:19], s[2:3] +; SDAG-NEXT: v_mov_b32_e32 v8, s16 +; SDAG-NEXT: v_mov_b32_e32 v9, s17 +; SDAG-NEXT: v_mov_b32_e32 v10, s18 +; SDAG-NEXT: v_mov_b32_e32 v11, s19 +; SDAG-NEXT: v_mov_b32_e32 v12, s20 +; SDAG-NEXT: v_mov_b32_e32 v13, s21 +; SDAG-NEXT: v_mov_b32_e32 v14, s22 +; SDAG-NEXT: v_mov_b32_e32 v15, s23 +; SDAG-NEXT: v_mov_b64_e32 v[16:17], s[0:1] ; SDAG-NEXT: s_nop 1 -; SDAG-NEXT: v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[2:9], v[10:17], a[0:3], 1.0, 0.15915494 op_sel:[1,1,0] op_sel_hi:[1,0,0] +; SDAG-NEXT: v_mfma_scale_f32_16x16x128_f8f6f4 v[0:3], v[0:7], v[8:15], v[16:19], 1.0, 0.15915494 op_sel:[1,1,0] op_sel_hi:[1,0,0] ; SDAG-NEXT: s_nop 7 ; SDAG-NEXT: s_nop 3 -; SDAG-NEXT: global_store_dwordx4 v0, a[0:3], s[4:5] +; SDAG-NEXT: global_store_dwordx4 v20, v[0:3], s[4:5] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd___scaleA_FP_literal__scaleB__FP_literal: @@ -2202,21 +2186,19 @@ define amdgpu_kernel void @test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd___scaleA ; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] ; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] ; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[16:17] -; GISEL-NEXT: v_accvgpr_write_b32 a0, s0 +; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[2:3] ; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[18:19] ; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[20:21] ; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[22:23] -; GISEL-NEXT: v_accvgpr_write_b32 a1, s1 -; GISEL-NEXT: v_accvgpr_write_b32 a2, s2 -; GISEL-NEXT: v_accvgpr_write_b32 a3, s3 +; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[0:1] ; GISEL-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x50 ; GISEL-NEXT: s_nop 0 -; GISEL-NEXT: v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[0:7], v[8:15], a[0:3], 1.0, 0.15915494 op_sel:[1,1,0] op_sel_hi:[1,0,0] -; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: v_mfma_scale_f32_16x16x128_f8f6f4 v[0:3], v[0:7], v[8:15], v[16:19], 1.0, 0.15915494 op_sel:[1,1,0] op_sel_hi:[1,0,0] +; GISEL-NEXT: v_mov_b32_e32 v4, 0 ; GISEL-NEXT: s_waitcnt lgkmcnt(0) ; GISEL-NEXT: s_nop 7 ; GISEL-NEXT: s_nop 1 -; GISEL-NEXT: global_store_dwordx4 v0, a[0:3], s[4:5] +; GISEL-NEXT: global_store_dwordx4 v4, v[0:3], s[4:5] ; GISEL-NEXT: s_endpgm %result = call <4 x float> @llvm.amdgcn.mfma.scale.f32.16x16x128.f8f6f4.v8i32.v8i32(<8 x i32> %arg0, <8 x i32> %arg1, <4 x float> %arg2, i32 0, i32 0, i32 3, i32 1065353216, i32 1, i32 1042479491) store <4 x float> %result, ptr addrspace(1) %ptr, align 16 @@ -2559,5 +2541,5 @@ declare <4 x float> @llvm.amdgcn.mfma.scale.f32.16x16x128.f8f6f4.v6i32.v8i32(<6 declare <4 x float> @llvm.amdgcn.mfma.scale.f32.16x16x128.f8f6f4.v8i32.v4i32(<8 x i32>, <4 x i32>, <4 x float>, i32 immarg, i32 immarg, i32 immarg, i32, i32 immarg, i32) #1 declare <4 x float> @llvm.amdgcn.mfma.scale.f32.16x16x128.f8f6f4.v8i32.v6i32(<8 x i32>, <6 x i32>, <4 x float>, i32 immarg, i32 immarg, i32 immarg, i32, i32 immarg, i32) #1 -attributes #0 = { "amdgpu-flat-work-group-size"="512,512" } +attributes #0 = { "amdgpu-flat-work-group-size"="512,512" "amdgpu-agpr-alloc"="0,0" } attributes #1 = { convergent nocallback nofree nosync nounwind willreturn memory(none) } diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.ll index bc50058..0b2818f 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.ll @@ -4539,49 +4539,41 @@ define amdgpu_kernel void @test_mfma_scale_f32_32x32x64_f8f6f4__vgprcd(<8 x i32> ; SDAG-NEXT: s_load_dwordx16 s[36:51], s[4:5], 0x40 ; SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x80 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_accvgpr_write_b32 a0, s36 -; SDAG-NEXT: v_mov_b32_e32 v2, s8 -; SDAG-NEXT: v_mov_b32_e32 v3, s9 -; SDAG-NEXT: v_mov_b32_e32 v4, s10 -; SDAG-NEXT: v_mov_b32_e32 v5, s11 -; SDAG-NEXT: v_mov_b32_e32 v6, s12 -; SDAG-NEXT: v_mov_b32_e32 v7, s13 -; SDAG-NEXT: v_mov_b32_e32 v8, s14 -; SDAG-NEXT: v_mov_b32_e32 v9, s15 -; SDAG-NEXT: v_mov_b32_e32 v10, s16 -; SDAG-NEXT: v_mov_b32_e32 v11, s17 -; SDAG-NEXT: v_mov_b32_e32 v12, s18 -; SDAG-NEXT: v_mov_b32_e32 v13, s19 -; SDAG-NEXT: v_mov_b32_e32 v14, s20 -; SDAG-NEXT: v_mov_b32_e32 v15, s21 -; SDAG-NEXT: v_mov_b32_e32 v16, s22 -; SDAG-NEXT: v_mov_b32_e32 v17, s23 -; SDAG-NEXT: v_accvgpr_write_b32 a1, s37 -; SDAG-NEXT: v_accvgpr_write_b32 a2, s38 -; SDAG-NEXT: v_accvgpr_write_b32 a3, s39 -; SDAG-NEXT: v_accvgpr_write_b32 a4, s40 -; SDAG-NEXT: v_accvgpr_write_b32 a5, s41 -; SDAG-NEXT: v_accvgpr_write_b32 a6, s42 -; SDAG-NEXT: v_accvgpr_write_b32 a7, s43 -; SDAG-NEXT: v_accvgpr_write_b32 a8, s44 -; SDAG-NEXT: v_accvgpr_write_b32 a9, s45 -; SDAG-NEXT: v_accvgpr_write_b32 a10, s46 -; SDAG-NEXT: v_accvgpr_write_b32 a11, s47 -; SDAG-NEXT: v_accvgpr_write_b32 a12, s48 -; SDAG-NEXT: v_accvgpr_write_b32 a13, s49 -; SDAG-NEXT: v_accvgpr_write_b32 a14, s50 -; SDAG-NEXT: v_accvgpr_write_b32 a15, s51 -; SDAG-NEXT: v_mov_b32_e32 v0, s1 +; SDAG-NEXT: v_mov_b64_e32 v[0:1], s[36:37] +; SDAG-NEXT: v_mov_b32_e32 v16, s8 +; SDAG-NEXT: v_mov_b32_e32 v17, s9 +; SDAG-NEXT: v_mov_b32_e32 v18, s10 +; SDAG-NEXT: v_mov_b32_e32 v19, s11 +; SDAG-NEXT: v_mov_b32_e32 v20, s12 +; SDAG-NEXT: v_mov_b32_e32 v21, s13 +; SDAG-NEXT: v_mov_b32_e32 v22, s14 +; SDAG-NEXT: v_mov_b32_e32 v23, s15 +; SDAG-NEXT: v_mov_b32_e32 v24, s16 +; SDAG-NEXT: v_mov_b32_e32 v25, s17 +; SDAG-NEXT: v_mov_b32_e32 v26, s18 +; SDAG-NEXT: v_mov_b32_e32 v27, s19 +; SDAG-NEXT: v_mov_b32_e32 v28, s20 +; SDAG-NEXT: v_mov_b32_e32 v29, s21 +; SDAG-NEXT: v_mov_b32_e32 v30, s22 +; SDAG-NEXT: v_mov_b32_e32 v31, s23 +; SDAG-NEXT: v_mov_b64_e32 v[2:3], s[38:39] +; SDAG-NEXT: v_mov_b64_e32 v[4:5], s[40:41] +; SDAG-NEXT: v_mov_b64_e32 v[6:7], s[42:43] +; SDAG-NEXT: v_mov_b64_e32 v[8:9], s[44:45] +; SDAG-NEXT: v_mov_b64_e32 v[10:11], s[46:47] +; SDAG-NEXT: v_mov_b64_e32 v[12:13], s[48:49] +; SDAG-NEXT: v_mov_b64_e32 v[14:15], s[50:51] +; SDAG-NEXT: v_mov_b32_e32 v32, s1 ; SDAG-NEXT: s_nop 1 -; SDAG-NEXT: v_mfma_scale_f32_32x32x64_f8f6f4 a[0:15], v[2:9], v[10:17], a[0:15], s0, v0 op_sel:[1,1,0] op_sel_hi:[1,0,0] blgp:2 -; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: v_mfma_scale_f32_32x32x64_f8f6f4 v[0:15], v[16:23], v[24:31], v[0:15], s0, v32 op_sel:[1,1,0] op_sel_hi:[1,0,0] blgp:2 +; SDAG-NEXT: v_mov_b32_e32 v16, 0 ; SDAG-NEXT: s_nop 7 ; SDAG-NEXT: s_nop 7 ; SDAG-NEXT: s_nop 2 -; SDAG-NEXT: global_store_dwordx4 v0, a[12:15], s[2:3] offset:48 -; SDAG-NEXT: global_store_dwordx4 v0, a[8:11], s[2:3] offset:32 -; SDAG-NEXT: global_store_dwordx4 v0, a[4:7], s[2:3] offset:16 -; SDAG-NEXT: global_store_dwordx4 v0, a[0:3], s[2:3] +; SDAG-NEXT: global_store_dwordx4 v16, v[12:15], s[2:3] offset:48 +; SDAG-NEXT: global_store_dwordx4 v16, v[8:11], s[2:3] offset:32 +; SDAG-NEXT: global_store_dwordx4 v16, v[4:7], s[2:3] offset:16 +; SDAG-NEXT: global_store_dwordx4 v16, v[0:3], s[2:3] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_mfma_scale_f32_32x32x64_f8f6f4__vgprcd: @@ -4590,41 +4582,33 @@ define amdgpu_kernel void @test_mfma_scale_f32_32x32x64_f8f6f4__vgprcd(<8 x i32> ; GISEL-NEXT: s_load_dwordx16 s[36:51], s[4:5], 0x40 ; GISEL-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x80 ; GISEL-NEXT: s_waitcnt lgkmcnt(0) -; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[8:9] -; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[10:11] -; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] -; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] -; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[16:17] -; GISEL-NEXT: v_accvgpr_write_b32 a0, s36 -; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[18:19] -; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[20:21] -; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[22:23] -; GISEL-NEXT: v_accvgpr_write_b32 a1, s37 -; GISEL-NEXT: v_accvgpr_write_b32 a2, s38 -; GISEL-NEXT: v_accvgpr_write_b32 a3, s39 -; GISEL-NEXT: v_accvgpr_write_b32 a4, s40 -; GISEL-NEXT: v_accvgpr_write_b32 a5, s41 -; GISEL-NEXT: v_accvgpr_write_b32 a6, s42 -; GISEL-NEXT: v_accvgpr_write_b32 a7, s43 -; GISEL-NEXT: v_accvgpr_write_b32 a8, s44 -; GISEL-NEXT: v_accvgpr_write_b32 a9, s45 -; GISEL-NEXT: v_accvgpr_write_b32 a10, s46 -; GISEL-NEXT: v_accvgpr_write_b32 a11, s47 -; GISEL-NEXT: v_accvgpr_write_b32 a12, s48 -; GISEL-NEXT: v_accvgpr_write_b32 a13, s49 -; GISEL-NEXT: v_accvgpr_write_b32 a14, s50 -; GISEL-NEXT: v_accvgpr_write_b32 a15, s51 -; GISEL-NEXT: v_mov_b32_e32 v16, s1 +; GISEL-NEXT: v_mov_b64_e32 v[22:23], s[14:15] +; GISEL-NEXT: v_mov_b64_e32 v[20:21], s[12:13] +; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[10:11] +; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[8:9] +; GISEL-NEXT: v_mov_b64_e32 v[30:31], s[22:23] +; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[36:37] +; GISEL-NEXT: v_mov_b64_e32 v[28:29], s[20:21] +; GISEL-NEXT: v_mov_b64_e32 v[26:27], s[18:19] +; GISEL-NEXT: v_mov_b64_e32 v[24:25], s[16:17] +; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[38:39] +; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[40:41] +; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[42:43] +; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[44:45] +; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[46:47] +; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[48:49] +; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[50:51] +; GISEL-NEXT: v_mov_b32_e32 v32, s1 ; GISEL-NEXT: s_nop 1 -; GISEL-NEXT: v_mfma_scale_f32_32x32x64_f8f6f4 a[0:15], v[0:7], v[8:15], a[0:15], s0, v16 op_sel:[1,1,0] op_sel_hi:[1,0,0] blgp:2 -; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: v_mfma_scale_f32_32x32x64_f8f6f4 v[0:15], v[16:23], v[24:31], v[0:15], s0, v32 op_sel:[1,1,0] op_sel_hi:[1,0,0] blgp:2 +; GISEL-NEXT: v_mov_b32_e32 v16, 0 ; GISEL-NEXT: s_nop 7 ; GISEL-NEXT: s_nop 7 ; GISEL-NEXT: s_nop 2 -; GISEL-NEXT: global_store_dwordx4 v0, a[0:3], s[2:3] -; GISEL-NEXT: global_store_dwordx4 v0, a[4:7], s[2:3] offset:16 -; GISEL-NEXT: global_store_dwordx4 v0, a[8:11], s[2:3] offset:32 -; GISEL-NEXT: global_store_dwordx4 v0, a[12:15], s[2:3] offset:48 +; GISEL-NEXT: global_store_dwordx4 v16, v[0:3], s[2:3] +; GISEL-NEXT: global_store_dwordx4 v16, v[4:7], s[2:3] offset:16 +; GISEL-NEXT: global_store_dwordx4 v16, v[8:11], s[2:3] offset:32 +; GISEL-NEXT: global_store_dwordx4 v16, v[12:15], s[2:3] offset:48 ; GISEL-NEXT: s_endpgm %result = call <16 x float> @llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.v8i32.v8i32(<8 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 0, i32 2, i32 3, i32 %scale0, i32 1, i32 %scale1) store <16 x float> %result, ptr addrspace(1) %ptr, align 64 @@ -4639,91 +4623,75 @@ define amdgpu_kernel void @test_mfma_scale_f32_32x32x64_f8f6f4__vgprcd___scaleA_ ; SDAG-NEXT: s_movk_i32 s2, 0x41 ; SDAG-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x80 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v0, s8 -; SDAG-NEXT: v_mov_b32_e32 v1, s9 -; SDAG-NEXT: v_mov_b32_e32 v2, s10 -; SDAG-NEXT: v_mov_b32_e32 v3, s11 -; SDAG-NEXT: v_mov_b32_e32 v4, s12 -; SDAG-NEXT: v_mov_b32_e32 v5, s13 -; SDAG-NEXT: v_mov_b32_e32 v6, s14 -; SDAG-NEXT: v_mov_b32_e32 v7, s15 -; SDAG-NEXT: v_accvgpr_write_b32 a0, s36 -; SDAG-NEXT: v_mov_b32_e32 v8, s16 -; SDAG-NEXT: v_mov_b32_e32 v9, s17 -; SDAG-NEXT: v_mov_b32_e32 v10, s18 -; SDAG-NEXT: v_mov_b32_e32 v11, s19 -; SDAG-NEXT: v_mov_b32_e32 v12, s20 -; SDAG-NEXT: v_mov_b32_e32 v13, s21 -; SDAG-NEXT: v_mov_b32_e32 v14, s22 -; SDAG-NEXT: v_mov_b32_e32 v15, s23 -; SDAG-NEXT: v_accvgpr_write_b32 a1, s37 -; SDAG-NEXT: v_accvgpr_write_b32 a2, s38 -; SDAG-NEXT: v_accvgpr_write_b32 a3, s39 -; SDAG-NEXT: v_accvgpr_write_b32 a4, s40 -; SDAG-NEXT: v_accvgpr_write_b32 a5, s41 -; SDAG-NEXT: v_accvgpr_write_b32 a6, s42 -; SDAG-NEXT: v_accvgpr_write_b32 a7, s43 -; SDAG-NEXT: v_accvgpr_write_b32 a8, s44 -; SDAG-NEXT: v_accvgpr_write_b32 a9, s45 -; SDAG-NEXT: v_accvgpr_write_b32 a10, s46 -; SDAG-NEXT: v_accvgpr_write_b32 a11, s47 -; SDAG-NEXT: v_accvgpr_write_b32 a12, s48 -; SDAG-NEXT: v_accvgpr_write_b32 a13, s49 -; SDAG-NEXT: v_accvgpr_write_b32 a14, s50 -; SDAG-NEXT: v_accvgpr_write_b32 a15, s51 +; SDAG-NEXT: v_mov_b32_e32 v16, s8 +; SDAG-NEXT: v_mov_b32_e32 v17, s9 +; SDAG-NEXT: v_mov_b32_e32 v18, s10 +; SDAG-NEXT: v_mov_b32_e32 v19, s11 +; SDAG-NEXT: v_mov_b32_e32 v20, s12 +; SDAG-NEXT: v_mov_b32_e32 v21, s13 +; SDAG-NEXT: v_mov_b32_e32 v22, s14 +; SDAG-NEXT: v_mov_b32_e32 v23, s15 +; SDAG-NEXT: v_mov_b64_e32 v[0:1], s[36:37] +; SDAG-NEXT: v_mov_b32_e32 v24, s16 +; SDAG-NEXT: v_mov_b32_e32 v25, s17 +; SDAG-NEXT: v_mov_b32_e32 v26, s18 +; SDAG-NEXT: v_mov_b32_e32 v27, s19 +; SDAG-NEXT: v_mov_b32_e32 v28, s20 +; SDAG-NEXT: v_mov_b32_e32 v29, s21 +; SDAG-NEXT: v_mov_b32_e32 v30, s22 +; SDAG-NEXT: v_mov_b32_e32 v31, s23 +; SDAG-NEXT: v_mov_b64_e32 v[2:3], s[38:39] +; SDAG-NEXT: v_mov_b64_e32 v[4:5], s[40:41] +; SDAG-NEXT: v_mov_b64_e32 v[6:7], s[42:43] +; SDAG-NEXT: v_mov_b64_e32 v[8:9], s[44:45] +; SDAG-NEXT: v_mov_b64_e32 v[10:11], s[46:47] +; SDAG-NEXT: v_mov_b64_e32 v[12:13], s[48:49] +; SDAG-NEXT: v_mov_b64_e32 v[14:15], s[50:51] ; SDAG-NEXT: s_nop 1 -; SDAG-NEXT: v_mfma_scale_f32_32x32x64_f8f6f4 a[0:15], v[0:7], v[8:15], a[0:15], s2, -2 op_sel:[1,1,0] op_sel_hi:[1,0,0] blgp:2 -; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: v_mfma_scale_f32_32x32x64_f8f6f4 v[0:15], v[16:23], v[24:31], v[0:15], s2, -2 op_sel:[1,1,0] op_sel_hi:[1,0,0] blgp:2 +; SDAG-NEXT: v_mov_b32_e32 v16, 0 ; SDAG-NEXT: s_nop 7 ; SDAG-NEXT: s_nop 7 ; SDAG-NEXT: s_nop 2 -; SDAG-NEXT: global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 -; SDAG-NEXT: global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 -; SDAG-NEXT: global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 -; SDAG-NEXT: global_store_dwordx4 v0, a[0:3], s[0:1] +; SDAG-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1] offset:48 +; SDAG-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1] offset:32 +; SDAG-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1] offset:16 +; SDAG-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_mfma_scale_f32_32x32x64_f8f6f4__vgprcd___scaleA_kimm__scaleB__inlineimm: ; GISEL: ; %bb.0: ; GISEL-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x0 ; GISEL-NEXT: s_load_dwordx16 s[36:51], s[4:5], 0x40 -; GISEL-NEXT: v_mov_b32_e32 v16, 0x41 +; GISEL-NEXT: v_mov_b32_e32 v32, 0x41 ; GISEL-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x80 ; GISEL-NEXT: s_waitcnt lgkmcnt(0) -; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[8:9] -; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[10:11] -; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] -; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] -; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[16:17] -; GISEL-NEXT: v_accvgpr_write_b32 a0, s36 -; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[18:19] -; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[20:21] -; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[22:23] -; GISEL-NEXT: v_accvgpr_write_b32 a1, s37 -; GISEL-NEXT: v_accvgpr_write_b32 a2, s38 -; GISEL-NEXT: v_accvgpr_write_b32 a3, s39 -; GISEL-NEXT: v_accvgpr_write_b32 a4, s40 -; GISEL-NEXT: v_accvgpr_write_b32 a5, s41 -; GISEL-NEXT: v_accvgpr_write_b32 a6, s42 -; GISEL-NEXT: v_accvgpr_write_b32 a7, s43 -; GISEL-NEXT: v_accvgpr_write_b32 a8, s44 -; GISEL-NEXT: v_accvgpr_write_b32 a9, s45 -; GISEL-NEXT: v_accvgpr_write_b32 a10, s46 -; GISEL-NEXT: v_accvgpr_write_b32 a11, s47 -; GISEL-NEXT: v_accvgpr_write_b32 a12, s48 -; GISEL-NEXT: v_accvgpr_write_b32 a13, s49 -; GISEL-NEXT: v_accvgpr_write_b32 a14, s50 -; GISEL-NEXT: v_accvgpr_write_b32 a15, s51 +; GISEL-NEXT: v_mov_b64_e32 v[22:23], s[14:15] +; GISEL-NEXT: v_mov_b64_e32 v[20:21], s[12:13] +; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[10:11] +; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[8:9] +; GISEL-NEXT: v_mov_b64_e32 v[30:31], s[22:23] +; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[36:37] +; GISEL-NEXT: v_mov_b64_e32 v[28:29], s[20:21] +; GISEL-NEXT: v_mov_b64_e32 v[26:27], s[18:19] +; GISEL-NEXT: v_mov_b64_e32 v[24:25], s[16:17] +; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[38:39] +; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[40:41] +; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[42:43] +; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[44:45] +; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[46:47] +; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[48:49] +; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[50:51] ; GISEL-NEXT: s_nop 1 -; GISEL-NEXT: v_mfma_scale_f32_32x32x64_f8f6f4 a[0:15], v[0:7], v[8:15], a[0:15], v16, -2 op_sel:[1,1,0] op_sel_hi:[1,0,0] blgp:2 -; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: v_mfma_scale_f32_32x32x64_f8f6f4 v[0:15], v[16:23], v[24:31], v[0:15], v32, -2 op_sel:[1,1,0] op_sel_hi:[1,0,0] blgp:2 +; GISEL-NEXT: v_mov_b32_e32 v16, 0 ; GISEL-NEXT: s_nop 7 ; GISEL-NEXT: s_nop 7 ; GISEL-NEXT: s_nop 2 -; GISEL-NEXT: global_store_dwordx4 v0, a[0:3], s[0:1] -; GISEL-NEXT: global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 -; GISEL-NEXT: global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 -; GISEL-NEXT: global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 +; GISEL-NEXT: global_store_dwordx4 v16, v[0:3], s[0:1] +; GISEL-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1] offset:16 +; GISEL-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1] offset:32 +; GISEL-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1] offset:48 ; GISEL-NEXT: s_endpgm %result = call <16 x float> @llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.v8i32.v8i32(<8 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 0, i32 2, i32 3, i32 65, i32 1, i32 -2) store <16 x float> %result, ptr addrspace(1) %ptr, align 64 @@ -5031,77 +4999,72 @@ define amdgpu_kernel void @test_mfma_scale_f32_32x32x64_f8f6f4_0_0__vgprcd_nonma ; SDAG: ; %bb.0: ; SDAG-NEXT: s_load_dwordx16 s[12:27], s[4:5], 0x0 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v0, s12 -; SDAG-NEXT: v_mov_b32_e32 v1, s13 -; SDAG-NEXT: v_mov_b32_e32 v2, s14 -; SDAG-NEXT: v_mov_b32_e32 v3, s15 -; SDAG-NEXT: v_mov_b32_e32 v4, s16 -; SDAG-NEXT: v_mov_b32_e32 v5, s17 -; SDAG-NEXT: v_mov_b32_e32 v6, s18 -; SDAG-NEXT: v_mov_b32_e32 v7, s19 -; SDAG-NEXT: v_mov_b32_e32 v8, s20 -; SDAG-NEXT: v_mov_b32_e32 v9, s21 -; SDAG-NEXT: v_mov_b32_e32 v10, s22 -; SDAG-NEXT: v_mov_b32_e32 v11, s23 +; SDAG-NEXT: v_mov_b32_e32 v32, s12 +; SDAG-NEXT: v_mov_b32_e32 v33, s13 +; SDAG-NEXT: v_mov_b32_e32 v34, s14 +; SDAG-NEXT: v_mov_b32_e32 v35, s15 +; SDAG-NEXT: v_mov_b32_e32 v36, s16 +; SDAG-NEXT: v_mov_b32_e32 v37, s17 +; SDAG-NEXT: v_mov_b32_e32 v38, s18 +; SDAG-NEXT: v_mov_b32_e32 v39, s19 +; SDAG-NEXT: v_mov_b32_e32 v40, s20 +; SDAG-NEXT: v_mov_b32_e32 v41, s21 +; SDAG-NEXT: v_mov_b32_e32 v42, s22 +; SDAG-NEXT: v_mov_b32_e32 v43, s23 ; SDAG-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x40 -; SDAG-NEXT: v_mov_b32_e32 v12, s24 -; SDAG-NEXT: v_mov_b32_e32 v13, s25 -; SDAG-NEXT: v_mov_b32_e32 v14, s26 -; SDAG-NEXT: v_mov_b32_e32 v15, s27 +; SDAG-NEXT: v_mov_b32_e32 v44, s24 +; SDAG-NEXT: v_mov_b32_e32 v45, s25 +; SDAG-NEXT: v_mov_b32_e32 v46, s26 +; SDAG-NEXT: v_mov_b32_e32 v47, s27 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_accvgpr_write_b32 a31, s23 -; SDAG-NEXT: v_accvgpr_write_b32 a30, s22 -; SDAG-NEXT: v_accvgpr_write_b32 a29, s21 -; SDAG-NEXT: v_accvgpr_write_b32 a28, s20 -; SDAG-NEXT: v_accvgpr_write_b32 a27, s19 -; SDAG-NEXT: v_accvgpr_write_b32 a26, s18 -; SDAG-NEXT: v_accvgpr_write_b32 a25, s17 -; SDAG-NEXT: v_accvgpr_write_b32 a24, s16 -; SDAG-NEXT: v_accvgpr_write_b32 a23, s15 -; SDAG-NEXT: v_accvgpr_write_b32 a22, s14 -; SDAG-NEXT: v_accvgpr_write_b32 a21, s13 -; SDAG-NEXT: v_accvgpr_write_b32 a20, s12 -; SDAG-NEXT: v_accvgpr_write_b32 a19, s11 -; SDAG-NEXT: v_accvgpr_write_b32 a18, s10 -; SDAG-NEXT: v_accvgpr_write_b32 a17, s9 -; SDAG-NEXT: v_accvgpr_write_b32 a16, s8 +; SDAG-NEXT: v_mov_b64_e32 v[30:31], s[22:23] +; SDAG-NEXT: v_mov_b64_e32 v[28:29], s[20:21] +; SDAG-NEXT: v_mov_b64_e32 v[26:27], s[18:19] +; SDAG-NEXT: v_mov_b64_e32 v[24:25], s[16:17] +; SDAG-NEXT: v_mov_b64_e32 v[22:23], s[14:15] +; SDAG-NEXT: v_mov_b64_e32 v[20:21], s[12:13] +; SDAG-NEXT: v_mov_b64_e32 v[18:19], s[10:11] +; SDAG-NEXT: v_mov_b64_e32 v[16:17], s[8:9] ; SDAG-NEXT: s_nop 1 -; SDAG-NEXT: v_mfma_f32_32x32x64_f8f6f4 a[0:15], v[0:7], v[8:15], a[16:31] blgp:2 -; SDAG-NEXT: v_mov_b32_e32 v2, s20 -; SDAG-NEXT: v_mov_b32_e32 v3, s21 -; SDAG-NEXT: v_mov_b32_e32 v4, s22 -; SDAG-NEXT: v_mov_b32_e32 v5, s23 -; SDAG-NEXT: v_mov_b64_e32 v[0:1], 48 -; SDAG-NEXT: global_store_dwordx4 v[0:1], v[2:5], off sc0 sc1 +; SDAG-NEXT: v_mfma_f32_32x32x64_f8f6f4 v[0:15], v[32:39], v[40:47], v[16:31] blgp:2 +; SDAG-NEXT: s_nop 7 +; SDAG-NEXT: s_nop 6 +; SDAG-NEXT: v_mov_b32_e32 v16, s20 +; SDAG-NEXT: v_mov_b32_e32 v17, s21 +; SDAG-NEXT: v_mov_b32_e32 v18, s22 +; SDAG-NEXT: v_mov_b32_e32 v19, s23 +; SDAG-NEXT: v_mov_b64_e32 v[20:21], 48 +; SDAG-NEXT: global_store_dwordx4 v[20:21], v[16:19], off sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v6, s18 -; SDAG-NEXT: v_mov_b32_e32 v7, s19 -; SDAG-NEXT: v_mov_b32_e32 v4, s16 -; SDAG-NEXT: v_mov_b32_e32 v5, s17 -; SDAG-NEXT: v_mov_b64_e32 v[2:3], 32 -; SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off sc0 sc1 +; SDAG-NEXT: v_mov_b64_e32 v[22:23], 32 +; SDAG-NEXT: v_mov_b64_e32 v[24:25], 16 +; SDAG-NEXT: v_mov_b32_e32 v16, s16 +; SDAG-NEXT: v_mov_b32_e32 v17, s17 +; SDAG-NEXT: v_mov_b32_e32 v18, s18 +; SDAG-NEXT: v_mov_b32_e32 v19, s19 +; SDAG-NEXT: global_store_dwordx4 v[22:23], v[16:19], off sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v8, s14 -; SDAG-NEXT: v_mov_b32_e32 v9, s15 -; SDAG-NEXT: v_mov_b32_e32 v6, s12 -; SDAG-NEXT: v_mov_b32_e32 v7, s13 -; SDAG-NEXT: v_mov_b64_e32 v[4:5], 16 -; SDAG-NEXT: global_store_dwordx4 v[4:5], v[6:9], off sc0 sc1 +; SDAG-NEXT: v_mov_b64_e32 v[26:27], 0 +; SDAG-NEXT: v_mov_b32_e32 v16, s12 +; SDAG-NEXT: v_mov_b32_e32 v17, s13 +; SDAG-NEXT: v_mov_b32_e32 v18, s14 +; SDAG-NEXT: v_mov_b32_e32 v19, s15 +; SDAG-NEXT: global_store_dwordx4 v[24:25], v[16:19], off sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v10, s10 -; SDAG-NEXT: v_mov_b32_e32 v11, s11 -; SDAG-NEXT: v_mov_b32_e32 v8, s8 -; SDAG-NEXT: v_mov_b32_e32 v9, s9 -; SDAG-NEXT: v_mov_b64_e32 v[6:7], 0 -; SDAG-NEXT: global_store_dwordx4 v[6:7], v[8:11], off sc0 sc1 +; SDAG-NEXT: s_nop 0 +; SDAG-NEXT: v_mov_b32_e32 v16, s8 +; SDAG-NEXT: v_mov_b32_e32 v17, s9 +; SDAG-NEXT: v_mov_b32_e32 v18, s10 +; SDAG-NEXT: v_mov_b32_e32 v19, s11 +; SDAG-NEXT: global_store_dwordx4 v[26:27], v[16:19], off sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v[2:3], a[8:11], off sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v[22:23], v[8:11], off sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v[0:1], a[12:15], off sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v[20:21], v[12:15], off sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v[6:7], a[0:3], off sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v[26:27], v[0:3], off sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v[4:5], a[4:7], off sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v[24:25], v[4:7], off sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_endpgm ; @@ -5109,61 +5072,45 @@ define amdgpu_kernel void @test_mfma_scale_f32_32x32x64_f8f6f4_0_0__vgprcd_nonma ; GISEL: ; %bb.0: ; GISEL-NEXT: s_load_dwordx16 s[36:51], s[4:5], 0x0 ; GISEL-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x40 -; GISEL-NEXT: v_mov_b64_e32 v[16:17], 0 -; GISEL-NEXT: v_mov_b64_e32 v[18:19], 16 -; GISEL-NEXT: v_mov_b64_e32 v[20:21], 32 ; GISEL-NEXT: s_waitcnt lgkmcnt(0) -; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[36:37] -; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[38:39] -; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[40:41] -; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[42:43] -; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[44:45] -; GISEL-NEXT: v_accvgpr_write_b32 a31, s23 -; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[46:47] -; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[48:49] -; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[50:51] -; GISEL-NEXT: v_accvgpr_write_b32 a30, s22 -; GISEL-NEXT: v_accvgpr_write_b32 a29, s21 -; GISEL-NEXT: v_accvgpr_write_b32 a28, s20 -; GISEL-NEXT: v_accvgpr_write_b32 a27, s19 -; GISEL-NEXT: v_accvgpr_write_b32 a26, s18 -; GISEL-NEXT: v_accvgpr_write_b32 a25, s17 -; GISEL-NEXT: v_accvgpr_write_b32 a24, s16 -; GISEL-NEXT: v_accvgpr_write_b32 a23, s15 -; GISEL-NEXT: v_accvgpr_write_b32 a22, s14 -; GISEL-NEXT: v_accvgpr_write_b32 a21, s13 -; GISEL-NEXT: v_accvgpr_write_b32 a20, s12 -; GISEL-NEXT: v_accvgpr_write_b32 a19, s11 -; GISEL-NEXT: v_accvgpr_write_b32 a18, s10 -; GISEL-NEXT: v_accvgpr_write_b32 a17, s9 -; GISEL-NEXT: v_accvgpr_write_b32 a16, s8 -; GISEL-NEXT: v_mov_b64_e32 v[22:23], 48 -; GISEL-NEXT: s_nop 0 -; GISEL-NEXT: v_mfma_f32_32x32x64_f8f6f4 a[0:15], v[0:7], v[8:15], a[16:31] blgp:2 -; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[8:9] -; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[10:11] -; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] -; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[16:17] -; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[20:21] -; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] -; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[18:19] -; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[22:23] -; GISEL-NEXT: global_store_dwordx4 v[16:17], v[0:3], off sc0 sc1 +; GISEL-NEXT: v_mov_b64_e32 v[32:33], s[36:37] +; GISEL-NEXT: v_mov_b64_e32 v[34:35], s[38:39] +; GISEL-NEXT: v_mov_b64_e32 v[36:37], s[40:41] +; GISEL-NEXT: v_mov_b64_e32 v[38:39], s[42:43] +; GISEL-NEXT: v_mov_b64_e32 v[40:41], s[44:45] +; GISEL-NEXT: v_mov_b64_e32 v[30:31], s[22:23] +; GISEL-NEXT: v_mov_b64_e32 v[42:43], s[46:47] +; GISEL-NEXT: v_mov_b64_e32 v[44:45], s[48:49] +; GISEL-NEXT: v_mov_b64_e32 v[46:47], s[50:51] +; GISEL-NEXT: v_mov_b64_e32 v[28:29], s[20:21] +; GISEL-NEXT: v_mov_b64_e32 v[26:27], s[18:19] +; GISEL-NEXT: v_mov_b64_e32 v[24:25], s[16:17] +; GISEL-NEXT: v_mov_b64_e32 v[22:23], s[14:15] +; GISEL-NEXT: v_mov_b64_e32 v[20:21], s[12:13] +; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[10:11] +; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[8:9] +; GISEL-NEXT: s_nop 1 +; GISEL-NEXT: v_mfma_f32_32x32x64_f8f6f4 v[0:15], v[32:39], v[40:47], v[16:31] blgp:2 +; GISEL-NEXT: v_mov_b64_e32 v[32:33], 0 +; GISEL-NEXT: v_mov_b64_e32 v[34:35], 16 +; GISEL-NEXT: v_mov_b64_e32 v[36:37], 32 +; GISEL-NEXT: v_mov_b64_e32 v[38:39], 48 +; GISEL-NEXT: global_store_dwordx4 v[32:33], v[16:19], off sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v[18:19], v[4:7], off sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v[34:35], v[20:23], off sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v[20:21], v[8:11], off sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v[36:37], v[24:27], off sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v[22:23], v[12:15], off sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v[38:39], v[28:31], off sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: s_nop 3 -; GISEL-NEXT: global_store_dwordx4 v[16:17], a[0:3], off sc0 sc1 +; GISEL-NEXT: s_nop 7 +; GISEL-NEXT: global_store_dwordx4 v[32:33], v[0:3], off sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v[18:19], a[4:7], off sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v[34:35], v[4:7], off sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v[20:21], a[8:11], off sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v[36:37], v[8:11], off sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v[22:23], a[12:15], off sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v[38:39], v[12:15], off sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) ; GISEL-NEXT: s_endpgm %result = call <16 x float> @llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.v8i32.v8i32(<8 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 0, i32 2, i32 0, i32 0, i32 0, i32 0) @@ -5177,77 +5124,70 @@ define amdgpu_kernel void @test_mfma_scale_f32_32x32x64_f8f6f4_25_42__vgprcd_non ; SDAG: ; %bb.0: ; SDAG-NEXT: s_load_dwordx16 s[12:27], s[4:5], 0x0 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v0, s12 -; SDAG-NEXT: v_mov_b32_e32 v1, s13 -; SDAG-NEXT: v_mov_b32_e32 v2, s14 -; SDAG-NEXT: v_mov_b32_e32 v3, s15 -; SDAG-NEXT: v_mov_b32_e32 v4, s16 -; SDAG-NEXT: v_mov_b32_e32 v5, s17 -; SDAG-NEXT: v_mov_b32_e32 v6, s18 -; SDAG-NEXT: v_mov_b32_e32 v7, s19 -; SDAG-NEXT: v_mov_b32_e32 v8, s20 -; SDAG-NEXT: v_mov_b32_e32 v9, s21 -; SDAG-NEXT: v_mov_b32_e32 v10, s22 -; SDAG-NEXT: v_mov_b32_e32 v11, s23 +; SDAG-NEXT: v_mov_b32_e32 v16, s12 +; SDAG-NEXT: v_mov_b32_e32 v17, s13 +; SDAG-NEXT: v_mov_b32_e32 v18, s14 +; SDAG-NEXT: v_mov_b32_e32 v19, s15 +; SDAG-NEXT: v_mov_b32_e32 v20, s16 +; SDAG-NEXT: v_mov_b32_e32 v21, s17 +; SDAG-NEXT: v_mov_b32_e32 v22, s18 +; SDAG-NEXT: v_mov_b32_e32 v23, s19 +; SDAG-NEXT: v_mov_b32_e32 v24, s20 +; SDAG-NEXT: v_mov_b32_e32 v25, s21 +; SDAG-NEXT: v_mov_b32_e32 v26, s22 +; SDAG-NEXT: v_mov_b32_e32 v27, s23 ; SDAG-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x40 -; SDAG-NEXT: v_mov_b32_e32 v12, s24 -; SDAG-NEXT: v_mov_b32_e32 v13, s25 -; SDAG-NEXT: v_mov_b32_e32 v14, s26 -; SDAG-NEXT: v_mov_b32_e32 v15, s27 +; SDAG-NEXT: v_mov_b32_e32 v28, s24 +; SDAG-NEXT: v_mov_b32_e32 v29, s25 +; SDAG-NEXT: v_mov_b32_e32 v30, s26 +; SDAG-NEXT: v_mov_b32_e32 v31, s27 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_accvgpr_write_b32 a0, s8 -; SDAG-NEXT: v_accvgpr_write_b32 a1, s9 -; SDAG-NEXT: v_accvgpr_write_b32 a2, s10 -; SDAG-NEXT: v_accvgpr_write_b32 a3, s11 -; SDAG-NEXT: v_accvgpr_write_b32 a4, s12 -; SDAG-NEXT: v_accvgpr_write_b32 a5, s13 -; SDAG-NEXT: v_accvgpr_write_b32 a6, s14 -; SDAG-NEXT: v_accvgpr_write_b32 a7, s15 -; SDAG-NEXT: v_accvgpr_write_b32 a8, s16 -; SDAG-NEXT: v_accvgpr_write_b32 a9, s17 -; SDAG-NEXT: v_accvgpr_write_b32 a10, s18 -; SDAG-NEXT: v_accvgpr_write_b32 a11, s19 -; SDAG-NEXT: v_accvgpr_write_b32 a12, s20 -; SDAG-NEXT: v_accvgpr_write_b32 a13, s21 -; SDAG-NEXT: v_accvgpr_write_b32 a14, s22 -; SDAG-NEXT: v_accvgpr_write_b32 a15, s23 +; SDAG-NEXT: v_mov_b64_e32 v[0:1], s[8:9] +; SDAG-NEXT: v_mov_b64_e32 v[2:3], s[10:11] +; SDAG-NEXT: v_mov_b64_e32 v[4:5], s[12:13] +; SDAG-NEXT: v_mov_b64_e32 v[6:7], s[14:15] +; SDAG-NEXT: v_mov_b64_e32 v[8:9], s[16:17] +; SDAG-NEXT: v_mov_b64_e32 v[10:11], s[18:19] +; SDAG-NEXT: v_mov_b64_e32 v[12:13], s[20:21] +; SDAG-NEXT: v_mov_b64_e32 v[14:15], s[22:23] ; SDAG-NEXT: s_nop 1 -; SDAG-NEXT: v_mfma_scale_f32_32x32x64_f8f6f4 a[0:15], v[0:7], v[8:15], a[0:15], 25, 42 op_sel_hi:[0,0,0] blgp:2 -; SDAG-NEXT: v_mov_b32_e32 v2, s20 -; SDAG-NEXT: v_mov_b32_e32 v3, s21 -; SDAG-NEXT: v_mov_b32_e32 v4, s22 -; SDAG-NEXT: v_mov_b32_e32 v5, s23 -; SDAG-NEXT: v_mov_b64_e32 v[0:1], 48 -; SDAG-NEXT: global_store_dwordx4 v[0:1], v[2:5], off sc0 sc1 +; SDAG-NEXT: v_mfma_scale_f32_32x32x64_f8f6f4 v[0:15], v[16:23], v[24:31], v[0:15], 25, 42 op_sel_hi:[0,0,0] blgp:2 +; SDAG-NEXT: v_mov_b32_e32 v16, s20 +; SDAG-NEXT: v_mov_b32_e32 v17, s21 +; SDAG-NEXT: v_mov_b32_e32 v18, s22 +; SDAG-NEXT: v_mov_b32_e32 v19, s23 +; SDAG-NEXT: v_mov_b64_e32 v[20:21], 48 +; SDAG-NEXT: global_store_dwordx4 v[20:21], v[16:19], off sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v6, s18 -; SDAG-NEXT: v_mov_b32_e32 v7, s19 -; SDAG-NEXT: v_mov_b32_e32 v4, s16 -; SDAG-NEXT: v_mov_b32_e32 v5, s17 -; SDAG-NEXT: v_mov_b64_e32 v[2:3], 32 -; SDAG-NEXT: global_store_dwordx4 v[2:3], v[4:7], off sc0 sc1 +; SDAG-NEXT: v_mov_b64_e32 v[22:23], 32 +; SDAG-NEXT: v_mov_b64_e32 v[24:25], 16 +; SDAG-NEXT: v_mov_b32_e32 v16, s16 +; SDAG-NEXT: v_mov_b32_e32 v17, s17 +; SDAG-NEXT: v_mov_b32_e32 v18, s18 +; SDAG-NEXT: v_mov_b32_e32 v19, s19 +; SDAG-NEXT: global_store_dwordx4 v[22:23], v[16:19], off sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v8, s14 -; SDAG-NEXT: v_mov_b32_e32 v9, s15 -; SDAG-NEXT: v_mov_b32_e32 v6, s12 -; SDAG-NEXT: v_mov_b32_e32 v7, s13 -; SDAG-NEXT: v_mov_b64_e32 v[4:5], 16 -; SDAG-NEXT: global_store_dwordx4 v[4:5], v[6:9], off sc0 sc1 +; SDAG-NEXT: v_mov_b64_e32 v[26:27], 0 +; SDAG-NEXT: v_mov_b32_e32 v16, s12 +; SDAG-NEXT: v_mov_b32_e32 v17, s13 +; SDAG-NEXT: v_mov_b32_e32 v18, s14 +; SDAG-NEXT: v_mov_b32_e32 v19, s15 +; SDAG-NEXT: global_store_dwordx4 v[24:25], v[16:19], off sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v10, s10 -; SDAG-NEXT: v_mov_b32_e32 v11, s11 -; SDAG-NEXT: v_mov_b32_e32 v8, s8 -; SDAG-NEXT: v_mov_b32_e32 v9, s9 -; SDAG-NEXT: v_mov_b64_e32 v[6:7], 0 -; SDAG-NEXT: global_store_dwordx4 v[6:7], v[8:11], off sc0 sc1 +; SDAG-NEXT: s_nop 0 +; SDAG-NEXT: v_mov_b32_e32 v16, s8 +; SDAG-NEXT: v_mov_b32_e32 v17, s9 +; SDAG-NEXT: v_mov_b32_e32 v18, s10 +; SDAG-NEXT: v_mov_b32_e32 v19, s11 +; SDAG-NEXT: global_store_dwordx4 v[26:27], v[16:19], off sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v[2:3], a[8:11], off sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v[22:23], v[8:11], off sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v[0:1], a[12:15], off sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v[20:21], v[12:15], off sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v[6:7], a[0:3], off sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v[26:27], v[0:3], off sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) -; SDAG-NEXT: global_store_dwordx4 v[4:5], a[4:7], off sc0 sc1 +; SDAG-NEXT: global_store_dwordx4 v[24:25], v[4:7], off sc0 sc1 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_endpgm ; @@ -5255,61 +5195,53 @@ define amdgpu_kernel void @test_mfma_scale_f32_32x32x64_f8f6f4_25_42__vgprcd_non ; GISEL: ; %bb.0: ; GISEL-NEXT: s_load_dwordx16 s[36:51], s[4:5], 0x0 ; GISEL-NEXT: s_load_dwordx16 s[8:23], s[4:5], 0x40 -; GISEL-NEXT: v_mov_b64_e32 v[16:17], 0 -; GISEL-NEXT: v_mov_b64_e32 v[18:19], 16 -; GISEL-NEXT: v_mov_b64_e32 v[20:21], 32 +; GISEL-NEXT: v_mov_b64_e32 v[32:33], 0 +; GISEL-NEXT: v_mov_b64_e32 v[34:35], 16 +; GISEL-NEXT: v_mov_b64_e32 v[36:37], 32 ; GISEL-NEXT: s_waitcnt lgkmcnt(0) -; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[36:37] -; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[38:39] -; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[40:41] -; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[42:43] -; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[44:45] -; GISEL-NEXT: v_accvgpr_write_b32 a0, s8 -; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[46:47] -; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[48:49] -; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[50:51] -; GISEL-NEXT: v_accvgpr_write_b32 a1, s9 -; GISEL-NEXT: v_accvgpr_write_b32 a2, s10 -; GISEL-NEXT: v_accvgpr_write_b32 a3, s11 -; GISEL-NEXT: v_accvgpr_write_b32 a4, s12 -; GISEL-NEXT: v_accvgpr_write_b32 a5, s13 -; GISEL-NEXT: v_accvgpr_write_b32 a6, s14 -; GISEL-NEXT: v_accvgpr_write_b32 a7, s15 -; GISEL-NEXT: v_accvgpr_write_b32 a8, s16 -; GISEL-NEXT: v_accvgpr_write_b32 a9, s17 -; GISEL-NEXT: v_accvgpr_write_b32 a10, s18 -; GISEL-NEXT: v_accvgpr_write_b32 a11, s19 -; GISEL-NEXT: v_accvgpr_write_b32 a12, s20 -; GISEL-NEXT: v_accvgpr_write_b32 a13, s21 -; GISEL-NEXT: v_accvgpr_write_b32 a14, s22 -; GISEL-NEXT: v_accvgpr_write_b32 a15, s23 -; GISEL-NEXT: v_mov_b64_e32 v[22:23], 48 -; GISEL-NEXT: s_nop 0 -; GISEL-NEXT: v_mfma_scale_f32_32x32x64_f8f6f4 a[0:15], v[0:7], v[8:15], a[0:15], 25, 42 op_sel_hi:[0,0,0] blgp:2 +; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[36:37] +; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[38:39] +; GISEL-NEXT: v_mov_b64_e32 v[20:21], s[40:41] +; GISEL-NEXT: v_mov_b64_e32 v[22:23], s[42:43] +; GISEL-NEXT: v_mov_b64_e32 v[24:25], s[44:45] ; GISEL-NEXT: v_mov_b64_e32 v[0:1], s[8:9] +; GISEL-NEXT: v_mov_b64_e32 v[26:27], s[46:47] +; GISEL-NEXT: v_mov_b64_e32 v[28:29], s[48:49] +; GISEL-NEXT: v_mov_b64_e32 v[30:31], s[50:51] ; GISEL-NEXT: v_mov_b64_e32 v[2:3], s[10:11] ; GISEL-NEXT: v_mov_b64_e32 v[4:5], s[12:13] -; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[16:17] -; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[20:21] ; GISEL-NEXT: v_mov_b64_e32 v[6:7], s[14:15] +; GISEL-NEXT: v_mov_b64_e32 v[8:9], s[16:17] ; GISEL-NEXT: v_mov_b64_e32 v[10:11], s[18:19] +; GISEL-NEXT: v_mov_b64_e32 v[12:13], s[20:21] ; GISEL-NEXT: v_mov_b64_e32 v[14:15], s[22:23] -; GISEL-NEXT: global_store_dwordx4 v[16:17], v[0:3], off sc0 sc1 +; GISEL-NEXT: v_mov_b64_e32 v[38:39], 48 +; GISEL-NEXT: s_nop 0 +; GISEL-NEXT: v_mfma_scale_f32_32x32x64_f8f6f4 v[0:15], v[16:23], v[24:31], v[0:15], 25, 42 op_sel_hi:[0,0,0] blgp:2 +; GISEL-NEXT: v_mov_b64_e32 v[18:19], s[10:11] +; GISEL-NEXT: v_mov_b64_e32 v[16:17], s[8:9] +; GISEL-NEXT: v_mov_b64_e32 v[22:23], s[14:15] +; GISEL-NEXT: v_mov_b64_e32 v[26:27], s[18:19] +; GISEL-NEXT: v_mov_b64_e32 v[30:31], s[22:23] +; GISEL-NEXT: v_mov_b64_e32 v[20:21], s[12:13] +; GISEL-NEXT: v_mov_b64_e32 v[24:25], s[16:17] +; GISEL-NEXT: v_mov_b64_e32 v[28:29], s[20:21] +; GISEL-NEXT: global_store_dwordx4 v[32:33], v[16:19], off sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v[18:19], v[4:7], off sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v[34:35], v[20:23], off sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v[20:21], v[8:11], off sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v[36:37], v[24:27], off sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v[22:23], v[12:15], off sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v[38:39], v[28:31], off sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) ; GISEL-NEXT: s_nop 3 -; GISEL-NEXT: global_store_dwordx4 v[16:17], a[0:3], off sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v[32:33], v[0:3], off sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v[18:19], a[4:7], off sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v[34:35], v[4:7], off sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v[20:21], a[8:11], off sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v[36:37], v[8:11], off sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) -; GISEL-NEXT: global_store_dwordx4 v[22:23], a[12:15], off sc0 sc1 +; GISEL-NEXT: global_store_dwordx4 v[38:39], v[12:15], off sc0 sc1 ; GISEL-NEXT: s_waitcnt vmcnt(0) ; GISEL-NEXT: s_endpgm %result = call <16 x float> @llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.v8i32.v8i32(<8 x i32> %arg0, <8 x i32> %arg1, <16 x float> %arg2, i32 0, i32 2, i32 0, i32 25, i32 0, i32 42) @@ -6298,6 +6230,6 @@ declare <16 x float> @llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.v6i32.v8i32(<6 declare <16 x float> @llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.v8i32.v4i32(<8 x i32>, <4 x i32>, <16 x float>, i32 immarg, i32 immarg, i32 immarg, i32, i32 immarg, i32) #2 declare <16 x float> @llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.v8i32.v6i32(<8 x i32>, <6 x i32>, <16 x float>, i32 immarg, i32 immarg, i32 immarg, i32, i32 immarg, i32) #2 -attributes #0 = { "amdgpu-flat-work-group-size"="512,512" } +attributes #0 = { "amdgpu-flat-work-group-size"="512,512" "amdgpu-agpr-alloc"="0,0" } attributes #1 = { "amdgpu-flat-work-group-size"="128,128" } attributes #2 = { convergent nocallback nofree nosync nounwind willreturn memory(none) } diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.xf32.gfx942.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.xf32.gfx942.ll index ea9334a..31a48de 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.xf32.gfx942.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.xf32.gfx942.ll @@ -1,8 +1,6 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 ; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck --check-prefixes=GFX942,GFX942-SDAG %s ; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck --check-prefixes=GFX942,GFX942-GISEL %s -; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx942 -stress-regalloc=10 < %s | FileCheck --check-prefixes=GFX942-STRESS,GFX942-SDAG-STRESS %s -; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx942 -stress-regalloc=10 < %s | FileCheck --check-prefixes=GFX942-STRESS,GFX942-GISEL-STRESS %s declare <4 x float> @llvm.amdgcn.mfma.f32.16x16x8.xf32(<2 x float>, <2 x float>, <4 x float>, i32, i32, i32) declare <16 x float> @llvm.amdgcn.mfma.f32.32x32x4.xf32(<2 x float>, <2 x float>, <16 x float>, i32, i32, i32) @@ -51,50 +49,6 @@ define amdgpu_kernel void @test_mfma_f32_16x16x8xf32(ptr addrspace(1) %arg) #0 { ; GFX942-GISEL-NEXT: s_nop 5 ; GFX942-GISEL-NEXT: global_store_dwordx4 v0, a[0:3], s[6:7] ; GFX942-GISEL-NEXT: s_endpgm -; -; GFX942-SDAG-STRESS-LABEL: test_mfma_f32_16x16x8xf32: -; GFX942-SDAG-STRESS: ; %bb.0: ; %bb -; GFX942-SDAG-STRESS-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x24 -; GFX942-SDAG-STRESS-NEXT: v_mov_b32_e32 v0, 1.0 -; GFX942-SDAG-STRESS-NEXT: v_mov_b32_e32 v1, 2.0 -; GFX942-SDAG-STRESS-NEXT: v_mov_b32_e32 v2, 0x40400000 -; GFX942-SDAG-STRESS-NEXT: v_mov_b32_e32 v3, 4.0 -; GFX942-SDAG-STRESS-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-SDAG-STRESS-NEXT: s_load_dwordx4 s[0:3], s[6:7], 0x0 -; GFX942-SDAG-STRESS-NEXT: v_mov_b32_e32 v4, 0 -; GFX942-SDAG-STRESS-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-SDAG-STRESS-NEXT: v_accvgpr_write_b32 a0, s0 -; GFX942-SDAG-STRESS-NEXT: v_accvgpr_write_b32 a1, s1 -; GFX942-SDAG-STRESS-NEXT: v_accvgpr_write_b32 a2, s2 -; GFX942-SDAG-STRESS-NEXT: v_accvgpr_write_b32 a3, s3 -; GFX942-SDAG-STRESS-NEXT: s_nop 1 -; GFX942-SDAG-STRESS-NEXT: v_mfma_f32_16x16x8_xf32 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3 -; GFX942-SDAG-STRESS-NEXT: s_nop 6 -; GFX942-SDAG-STRESS-NEXT: global_store_dwordx4 v4, a[0:3], s[6:7] -; GFX942-SDAG-STRESS-NEXT: s_endpgm -; -; GFX942-GISEL-STRESS-LABEL: test_mfma_f32_16x16x8xf32: -; GFX942-GISEL-STRESS: ; %bb.0: ; %bb -; GFX942-GISEL-STRESS-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x24 -; GFX942-GISEL-STRESS-NEXT: s_mov_b32 s0, 1.0 -; GFX942-GISEL-STRESS-NEXT: s_mov_b32 s2, 0x40400000 -; GFX942-GISEL-STRESS-NEXT: s_mov_b32 s1, 2.0 -; GFX942-GISEL-STRESS-NEXT: s_mov_b32 s3, 4.0 -; GFX942-GISEL-STRESS-NEXT: v_mov_b64_e32 v[0:1], s[0:1] -; GFX942-GISEL-STRESS-NEXT: v_mov_b64_e32 v[2:3], s[2:3] -; GFX942-GISEL-STRESS-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-GISEL-STRESS-NEXT: s_load_dwordx4 s[0:3], s[6:7], 0x0 -; GFX942-GISEL-STRESS-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-GISEL-STRESS-NEXT: v_accvgpr_write_b32 a0, s0 -; GFX942-GISEL-STRESS-NEXT: v_accvgpr_write_b32 a1, s1 -; GFX942-GISEL-STRESS-NEXT: v_accvgpr_write_b32 a2, s2 -; GFX942-GISEL-STRESS-NEXT: v_accvgpr_write_b32 a3, s3 -; GFX942-GISEL-STRESS-NEXT: s_nop 1 -; GFX942-GISEL-STRESS-NEXT: v_mfma_f32_16x16x8_xf32 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3 -; GFX942-GISEL-STRESS-NEXT: v_mov_b32_e32 v0, 0 -; GFX942-GISEL-STRESS-NEXT: s_nop 5 -; GFX942-GISEL-STRESS-NEXT: global_store_dwordx4 v0, a[0:3], s[6:7] -; GFX942-GISEL-STRESS-NEXT: s_endpgm bb: %in.1 = load <4 x float>, ptr addrspace(1) %arg %mai.1 = tail call <4 x float> @llvm.amdgcn.mfma.f32.16x16x8.xf32(<2 x float> <float 1.0, float 2.0>, <2 x float> <float 3.0, float 4.0>, <4 x float> %in.1, i32 1, i32 2, i32 3) @@ -178,82 +132,6 @@ define amdgpu_kernel void @test_mfma_f32_32x32x4xf32(ptr addrspace(1) %arg) #0 { ; GFX942-GISEL-NEXT: global_store_dwordx4 v0, a[8:11], s[16:17] offset:32 ; GFX942-GISEL-NEXT: global_store_dwordx4 v0, a[12:15], s[16:17] offset:48 ; GFX942-GISEL-NEXT: s_endpgm -; -; GFX942-SDAG-STRESS-LABEL: test_mfma_f32_32x32x4xf32: -; GFX942-SDAG-STRESS: ; %bb.0: ; %bb -; GFX942-SDAG-STRESS-NEXT: s_load_dwordx2 s[16:17], s[4:5], 0x24 -; GFX942-SDAG-STRESS-NEXT: v_mov_b32_e32 v0, 1.0 -; GFX942-SDAG-STRESS-NEXT: v_mov_b32_e32 v1, 2.0 -; GFX942-SDAG-STRESS-NEXT: v_mov_b32_e32 v2, 0x40400000 -; GFX942-SDAG-STRESS-NEXT: v_mov_b32_e32 v3, 4.0 -; GFX942-SDAG-STRESS-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-SDAG-STRESS-NEXT: s_load_dwordx16 s[0:15], s[16:17], 0x0 -; GFX942-SDAG-STRESS-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-SDAG-STRESS-NEXT: v_accvgpr_write_b32 a0, s0 -; GFX942-SDAG-STRESS-NEXT: v_accvgpr_write_b32 a1, s1 -; GFX942-SDAG-STRESS-NEXT: v_accvgpr_write_b32 a2, s2 -; GFX942-SDAG-STRESS-NEXT: v_accvgpr_write_b32 a3, s3 -; GFX942-SDAG-STRESS-NEXT: v_accvgpr_write_b32 a4, s4 -; GFX942-SDAG-STRESS-NEXT: v_accvgpr_write_b32 a5, s5 -; GFX942-SDAG-STRESS-NEXT: v_accvgpr_write_b32 a6, s6 -; GFX942-SDAG-STRESS-NEXT: v_accvgpr_write_b32 a7, s7 -; GFX942-SDAG-STRESS-NEXT: v_accvgpr_write_b32 a8, s8 -; GFX942-SDAG-STRESS-NEXT: v_accvgpr_write_b32 a9, s9 -; GFX942-SDAG-STRESS-NEXT: v_accvgpr_write_b32 a10, s10 -; GFX942-SDAG-STRESS-NEXT: v_accvgpr_write_b32 a11, s11 -; GFX942-SDAG-STRESS-NEXT: v_accvgpr_write_b32 a12, s12 -; GFX942-SDAG-STRESS-NEXT: v_accvgpr_write_b32 a13, s13 -; GFX942-SDAG-STRESS-NEXT: v_accvgpr_write_b32 a14, s14 -; GFX942-SDAG-STRESS-NEXT: v_accvgpr_write_b32 a15, s15 -; GFX942-SDAG-STRESS-NEXT: s_nop 1 -; GFX942-SDAG-STRESS-NEXT: v_mfma_f32_32x32x4_xf32 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3 -; GFX942-SDAG-STRESS-NEXT: v_mov_b32_e32 v0, 0 -; GFX942-SDAG-STRESS-NEXT: s_nop 7 -; GFX942-SDAG-STRESS-NEXT: s_nop 1 -; GFX942-SDAG-STRESS-NEXT: global_store_dwordx4 v0, a[12:15], s[16:17] offset:48 -; GFX942-SDAG-STRESS-NEXT: global_store_dwordx4 v0, a[8:11], s[16:17] offset:32 -; GFX942-SDAG-STRESS-NEXT: global_store_dwordx4 v0, a[4:7], s[16:17] offset:16 -; GFX942-SDAG-STRESS-NEXT: global_store_dwordx4 v0, a[0:3], s[16:17] -; GFX942-SDAG-STRESS-NEXT: s_endpgm -; -; GFX942-GISEL-STRESS-LABEL: test_mfma_f32_32x32x4xf32: -; GFX942-GISEL-STRESS: ; %bb.0: ; %bb -; GFX942-GISEL-STRESS-NEXT: s_load_dwordx2 s[16:17], s[4:5], 0x24 -; GFX942-GISEL-STRESS-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-GISEL-STRESS-NEXT: s_load_dwordx16 s[0:15], s[16:17], 0x0 -; GFX942-GISEL-STRESS-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-GISEL-STRESS-NEXT: v_accvgpr_write_b32 a0, s0 -; GFX942-GISEL-STRESS-NEXT: v_accvgpr_write_b32 a1, s1 -; GFX942-GISEL-STRESS-NEXT: v_accvgpr_write_b32 a2, s2 -; GFX942-GISEL-STRESS-NEXT: v_accvgpr_write_b32 a3, s3 -; GFX942-GISEL-STRESS-NEXT: v_accvgpr_write_b32 a4, s4 -; GFX942-GISEL-STRESS-NEXT: v_accvgpr_write_b32 a5, s5 -; GFX942-GISEL-STRESS-NEXT: v_accvgpr_write_b32 a6, s6 -; GFX942-GISEL-STRESS-NEXT: v_accvgpr_write_b32 a7, s7 -; GFX942-GISEL-STRESS-NEXT: v_accvgpr_write_b32 a8, s8 -; GFX942-GISEL-STRESS-NEXT: v_accvgpr_write_b32 a9, s9 -; GFX942-GISEL-STRESS-NEXT: v_accvgpr_write_b32 a10, s10 -; GFX942-GISEL-STRESS-NEXT: v_accvgpr_write_b32 a11, s11 -; GFX942-GISEL-STRESS-NEXT: v_accvgpr_write_b32 a12, s12 -; GFX942-GISEL-STRESS-NEXT: v_accvgpr_write_b32 a13, s13 -; GFX942-GISEL-STRESS-NEXT: v_accvgpr_write_b32 a14, s14 -; GFX942-GISEL-STRESS-NEXT: v_accvgpr_write_b32 a15, s15 -; GFX942-GISEL-STRESS-NEXT: s_mov_b32 s0, 1.0 -; GFX942-GISEL-STRESS-NEXT: s_mov_b32 s1, 2.0 -; GFX942-GISEL-STRESS-NEXT: v_mov_b64_e32 v[0:1], s[0:1] -; GFX942-GISEL-STRESS-NEXT: s_mov_b32 s0, 0x40400000 -; GFX942-GISEL-STRESS-NEXT: s_mov_b32 s1, 4.0 -; GFX942-GISEL-STRESS-NEXT: v_mov_b64_e32 v[2:3], s[0:1] -; GFX942-GISEL-STRESS-NEXT: s_nop 1 -; GFX942-GISEL-STRESS-NEXT: v_mfma_f32_32x32x4_xf32 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3 -; GFX942-GISEL-STRESS-NEXT: v_mov_b32_e32 v0, 0 -; GFX942-GISEL-STRESS-NEXT: s_nop 7 -; GFX942-GISEL-STRESS-NEXT: s_nop 1 -; GFX942-GISEL-STRESS-NEXT: global_store_dwordx4 v0, a[0:3], s[16:17] -; GFX942-GISEL-STRESS-NEXT: global_store_dwordx4 v0, a[4:7], s[16:17] offset:16 -; GFX942-GISEL-STRESS-NEXT: global_store_dwordx4 v0, a[8:11], s[16:17] offset:32 -; GFX942-GISEL-STRESS-NEXT: global_store_dwordx4 v0, a[12:15], s[16:17] offset:48 -; GFX942-GISEL-STRESS-NEXT: s_endpgm bb: %in.1 = load <16 x float>, ptr addrspace(1) %arg %mai.1 = tail call <16 x float> @llvm.amdgcn.mfma.f32.32x32x4.xf32(<2 x float> <float 1.0, float 2.0>, <2 x float> <float 3.0, float 4.0>, <16 x float> %in.1, i32 1, i32 2, i32 3) @@ -264,4 +142,3 @@ bb: attributes #0 = { "amdgpu-flat-work-group-size"="1,256" } ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line: ; GFX942: {{.*}} -; GFX942-STRESS: {{.*}} diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.smfmac.gfx950.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.smfmac.gfx950.ll index 8056881..b25fe83 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.smfmac.gfx950.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.smfmac.gfx950.ll @@ -17,24 +17,24 @@ define amdgpu_kernel void @test_smfmac_f32_16x16x64_f16__vgpr(ptr addrspace(1) % ; SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x34 ; SDAG-NEXT: v_and_b32_e32 v0, 0x3ff, v0 ; SDAG-NEXT: v_lshlrev_b32_e32 v0, 4, v0 -; SDAG-NEXT: v_mov_b32_e32 v12, 0 +; SDAG-NEXT: v_mov_b32_e32 v16, 0 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: global_load_dwordx4 v[14:17], v0, s[6:7] +; SDAG-NEXT: global_load_dwordx4 v[8:11], v0, s[6:7] ; SDAG-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x44 ; SDAG-NEXT: s_load_dword s16, s[4:5], 0x64 -; SDAG-NEXT: v_mov_b64_e32 v[10:11], s[2:3] -; SDAG-NEXT: v_mov_b64_e32 v[8:9], s[0:1] +; SDAG-NEXT: v_mov_b64_e32 v[14:15], s[2:3] +; SDAG-NEXT: v_mov_b64_e32 v[12:13], s[0:1] ; SDAG-NEXT: s_waitcnt lgkmcnt(0) ; SDAG-NEXT: v_mov_b64_e32 v[0:1], s[8:9] ; SDAG-NEXT: v_mov_b64_e32 v[2:3], s[10:11] ; SDAG-NEXT: v_mov_b64_e32 v[4:5], s[12:13] ; SDAG-NEXT: v_mov_b64_e32 v[6:7], s[14:15] -; SDAG-NEXT: v_mov_b32_e32 v13, s16 +; SDAG-NEXT: v_mov_b32_e32 v17, s16 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_smfmac_f32_16x16x64_f16 v[14:17], v[8:11], v[0:7], v13 cbsz:1 abid:2 +; SDAG-NEXT: v_smfmac_f32_16x16x64_f16 v[8:11], v[12:15], v[0:7], v17 cbsz:1 abid:2 ; SDAG-NEXT: s_nop 7 -; SDAG-NEXT: global_store_dwordx4 v12, v[14:17], s[6:7] +; SDAG-NEXT: global_store_dwordx4 v16, v[8:11], s[6:7] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_smfmac_f32_16x16x64_f16__vgpr: @@ -547,24 +547,24 @@ define amdgpu_kernel void @test_smfmac_f32_16x16x64_bf16__vgpr(ptr addrspace(1) ; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x34 ; GCN-NEXT: v_and_b32_e32 v0, 0x3ff, v0 ; GCN-NEXT: v_lshlrev_b32_e32 v0, 4, v0 -; GCN-NEXT: v_mov_b32_e32 v12, 0 +; GCN-NEXT: v_mov_b32_e32 v16, 0 ; GCN-NEXT: s_waitcnt lgkmcnt(0) -; GCN-NEXT: global_load_dwordx4 v[14:17], v0, s[6:7] +; GCN-NEXT: global_load_dwordx4 v[8:11], v0, s[6:7] ; GCN-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x44 ; GCN-NEXT: s_load_dword s16, s[4:5], 0x64 -; GCN-NEXT: v_mov_b64_e32 v[10:11], s[2:3] -; GCN-NEXT: v_mov_b64_e32 v[8:9], s[0:1] +; GCN-NEXT: v_mov_b64_e32 v[14:15], s[2:3] +; GCN-NEXT: v_mov_b64_e32 v[12:13], s[0:1] ; GCN-NEXT: s_waitcnt lgkmcnt(0) ; GCN-NEXT: v_mov_b64_e32 v[0:1], s[8:9] ; GCN-NEXT: v_mov_b64_e32 v[2:3], s[10:11] ; GCN-NEXT: v_mov_b64_e32 v[4:5], s[12:13] ; GCN-NEXT: v_mov_b64_e32 v[6:7], s[14:15] -; GCN-NEXT: v_mov_b32_e32 v13, s16 +; GCN-NEXT: v_mov_b32_e32 v17, s16 ; GCN-NEXT: s_waitcnt vmcnt(0) ; GCN-NEXT: s_nop 0 -; GCN-NEXT: v_smfmac_f32_16x16x64_bf16 v[14:17], v[8:11], v[0:7], v13 cbsz:1 abid:2 +; GCN-NEXT: v_smfmac_f32_16x16x64_bf16 v[8:11], v[12:15], v[0:7], v17 cbsz:1 abid:2 ; GCN-NEXT: s_nop 7 -; GCN-NEXT: global_store_dwordx4 v12, v[14:17], s[6:7] +; GCN-NEXT: global_store_dwordx4 v16, v[8:11], s[6:7] ; GCN-NEXT: s_endpgm bb: %id = call i32 @llvm.amdgcn.workitem.id.x() @@ -855,30 +855,30 @@ define amdgpu_kernel void @test_smfmac_i32_16x16x128_i8__vgpr(ptr addrspace(1) % ; SDAG-NEXT: v_and_b32_e32 v0, 0x3ff, v0 ; SDAG-NEXT: v_lshlrev_b32_e32 v0, 4, v0 ; SDAG-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x34 +; SDAG-NEXT: v_mov_b32_e32 v16, 0 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: global_load_dwordx4 v[10:13], v0, s[6:7] +; SDAG-NEXT: global_load_dwordx4 v[8:11], v0, s[6:7] ; SDAG-NEXT: s_load_dword s16, s[4:5], 0x64 ; SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 -; SDAG-NEXT: v_mov_b32_e32 v0, 0 -; SDAG-NEXT: v_mov_b32_e32 v14, s8 -; SDAG-NEXT: v_mov_b32_e32 v15, s9 -; SDAG-NEXT: v_mov_b32_e32 v16, s10 -; SDAG-NEXT: v_mov_b32_e32 v17, s11 -; SDAG-NEXT: v_mov_b32_e32 v2, s12 -; SDAG-NEXT: v_mov_b32_e32 v3, s13 -; SDAG-NEXT: v_mov_b32_e32 v4, s14 -; SDAG-NEXT: v_mov_b32_e32 v5, s15 +; SDAG-NEXT: v_mov_b32_e32 v12, s8 +; SDAG-NEXT: v_mov_b32_e32 v13, s9 +; SDAG-NEXT: v_mov_b32_e32 v14, s10 +; SDAG-NEXT: v_mov_b32_e32 v15, s11 +; SDAG-NEXT: v_mov_b32_e32 v0, s12 +; SDAG-NEXT: v_mov_b32_e32 v1, s13 +; SDAG-NEXT: v_mov_b32_e32 v2, s14 +; SDAG-NEXT: v_mov_b32_e32 v3, s15 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v6, s0 -; SDAG-NEXT: v_mov_b32_e32 v7, s1 -; SDAG-NEXT: v_mov_b32_e32 v8, s2 -; SDAG-NEXT: v_mov_b32_e32 v9, s3 -; SDAG-NEXT: v_mov_b32_e32 v1, s16 +; SDAG-NEXT: v_mov_b32_e32 v4, s0 +; SDAG-NEXT: v_mov_b32_e32 v5, s1 +; SDAG-NEXT: v_mov_b32_e32 v6, s2 +; SDAG-NEXT: v_mov_b32_e32 v7, s3 +; SDAG-NEXT: v_mov_b32_e32 v17, s16 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_smfmac_i32_16x16x128_i8 v[10:13], v[14:17], v[2:9], v1 cbsz:1 abid:2 +; SDAG-NEXT: v_smfmac_i32_16x16x128_i8 v[8:11], v[12:15], v[0:7], v17 cbsz:1 abid:2 ; SDAG-NEXT: s_nop 7 -; SDAG-NEXT: global_store_dwordx4 v0, v[10:13], s[6:7] +; SDAG-NEXT: global_store_dwordx4 v16, v[8:11], s[6:7] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_smfmac_i32_16x16x128_i8__vgpr: @@ -1032,22 +1032,22 @@ define amdgpu_kernel void @test_smfmac_i32_32x32x64_i8__vgpr(ptr addrspace(1) %a ; SDAG-NEXT: s_load_dword s16, s[4:5], 0x64 ; SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v26, s8 -; SDAG-NEXT: v_mov_b32_e32 v27, s9 -; SDAG-NEXT: v_mov_b32_e32 v28, s10 -; SDAG-NEXT: v_mov_b32_e32 v29, s11 -; SDAG-NEXT: v_mov_b32_e32 v18, s12 -; SDAG-NEXT: v_mov_b32_e32 v19, s13 -; SDAG-NEXT: v_mov_b32_e32 v20, s14 -; SDAG-NEXT: v_mov_b32_e32 v21, s15 -; SDAG-NEXT: v_mov_b32_e32 v22, s0 -; SDAG-NEXT: v_mov_b32_e32 v23, s1 -; SDAG-NEXT: v_mov_b32_e32 v24, s2 -; SDAG-NEXT: v_mov_b32_e32 v25, s3 -; SDAG-NEXT: v_mov_b32_e32 v16, s16 +; SDAG-NEXT: v_mov_b32_e32 v24, s8 +; SDAG-NEXT: v_mov_b32_e32 v25, s9 +; SDAG-NEXT: v_mov_b32_e32 v26, s10 +; SDAG-NEXT: v_mov_b32_e32 v27, s11 +; SDAG-NEXT: v_mov_b32_e32 v16, s12 +; SDAG-NEXT: v_mov_b32_e32 v17, s13 +; SDAG-NEXT: v_mov_b32_e32 v18, s14 +; SDAG-NEXT: v_mov_b32_e32 v19, s15 +; SDAG-NEXT: v_mov_b32_e32 v20, s0 +; SDAG-NEXT: v_mov_b32_e32 v21, s1 +; SDAG-NEXT: v_mov_b32_e32 v22, s2 +; SDAG-NEXT: v_mov_b32_e32 v23, s3 +; SDAG-NEXT: v_mov_b32_e32 v28, s16 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_smfmac_i32_32x32x64_i8 v[0:15], v[26:29], v[18:25], v16 cbsz:1 abid:2 +; SDAG-NEXT: v_smfmac_i32_32x32x64_i8 v[0:15], v[24:27], v[16:23], v28 cbsz:1 abid:2 ; SDAG-NEXT: v_mov_b32_e32 v16, 0 ; SDAG-NEXT: s_nop 7 ; SDAG-NEXT: s_nop 2 @@ -1397,30 +1397,30 @@ define amdgpu_kernel void @test_smfmac_f32_16x16x128_bf8_bf8__vgpr(ptr addrspace ; SDAG-NEXT: v_and_b32_e32 v0, 0x3ff, v0 ; SDAG-NEXT: v_lshlrev_b32_e32 v0, 4, v0 ; SDAG-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x34 +; SDAG-NEXT: v_mov_b32_e32 v16, 0 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: global_load_dwordx4 v[10:13], v0, s[6:7] +; SDAG-NEXT: global_load_dwordx4 v[8:11], v0, s[6:7] ; SDAG-NEXT: s_load_dword s16, s[4:5], 0x64 ; SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 -; SDAG-NEXT: v_mov_b32_e32 v0, 0 -; SDAG-NEXT: v_mov_b32_e32 v14, s8 -; SDAG-NEXT: v_mov_b32_e32 v15, s9 -; SDAG-NEXT: v_mov_b32_e32 v16, s10 -; SDAG-NEXT: v_mov_b32_e32 v17, s11 -; SDAG-NEXT: v_mov_b32_e32 v2, s12 -; SDAG-NEXT: v_mov_b32_e32 v3, s13 -; SDAG-NEXT: v_mov_b32_e32 v4, s14 -; SDAG-NEXT: v_mov_b32_e32 v5, s15 +; SDAG-NEXT: v_mov_b32_e32 v12, s8 +; SDAG-NEXT: v_mov_b32_e32 v13, s9 +; SDAG-NEXT: v_mov_b32_e32 v14, s10 +; SDAG-NEXT: v_mov_b32_e32 v15, s11 +; SDAG-NEXT: v_mov_b32_e32 v0, s12 +; SDAG-NEXT: v_mov_b32_e32 v1, s13 +; SDAG-NEXT: v_mov_b32_e32 v2, s14 +; SDAG-NEXT: v_mov_b32_e32 v3, s15 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v6, s0 -; SDAG-NEXT: v_mov_b32_e32 v7, s1 -; SDAG-NEXT: v_mov_b32_e32 v8, s2 -; SDAG-NEXT: v_mov_b32_e32 v9, s3 -; SDAG-NEXT: v_mov_b32_e32 v1, s16 +; SDAG-NEXT: v_mov_b32_e32 v4, s0 +; SDAG-NEXT: v_mov_b32_e32 v5, s1 +; SDAG-NEXT: v_mov_b32_e32 v6, s2 +; SDAG-NEXT: v_mov_b32_e32 v7, s3 +; SDAG-NEXT: v_mov_b32_e32 v17, s16 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_smfmac_f32_16x16x128_bf8_bf8 v[10:13], v[14:17], v[2:9], v1 cbsz:1 abid:2 +; SDAG-NEXT: v_smfmac_f32_16x16x128_bf8_bf8 v[8:11], v[12:15], v[0:7], v17 cbsz:1 abid:2 ; SDAG-NEXT: s_nop 7 -; SDAG-NEXT: global_store_dwordx4 v0, v[10:13], s[6:7] +; SDAG-NEXT: global_store_dwordx4 v16, v[8:11], s[6:7] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_smfmac_f32_16x16x128_bf8_bf8__vgpr: @@ -1566,30 +1566,30 @@ define amdgpu_kernel void @test_smfmac_f32_16x16x128_bf8_fp8__vgpr(ptr addrspace ; SDAG-NEXT: v_and_b32_e32 v0, 0x3ff, v0 ; SDAG-NEXT: v_lshlrev_b32_e32 v0, 4, v0 ; SDAG-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x34 +; SDAG-NEXT: v_mov_b32_e32 v16, 0 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: global_load_dwordx4 v[10:13], v0, s[6:7] +; SDAG-NEXT: global_load_dwordx4 v[8:11], v0, s[6:7] ; SDAG-NEXT: s_load_dword s16, s[4:5], 0x64 ; SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 -; SDAG-NEXT: v_mov_b32_e32 v0, 0 -; SDAG-NEXT: v_mov_b32_e32 v14, s8 -; SDAG-NEXT: v_mov_b32_e32 v15, s9 -; SDAG-NEXT: v_mov_b32_e32 v16, s10 -; SDAG-NEXT: v_mov_b32_e32 v17, s11 -; SDAG-NEXT: v_mov_b32_e32 v2, s12 -; SDAG-NEXT: v_mov_b32_e32 v3, s13 -; SDAG-NEXT: v_mov_b32_e32 v4, s14 -; SDAG-NEXT: v_mov_b32_e32 v5, s15 +; SDAG-NEXT: v_mov_b32_e32 v12, s8 +; SDAG-NEXT: v_mov_b32_e32 v13, s9 +; SDAG-NEXT: v_mov_b32_e32 v14, s10 +; SDAG-NEXT: v_mov_b32_e32 v15, s11 +; SDAG-NEXT: v_mov_b32_e32 v0, s12 +; SDAG-NEXT: v_mov_b32_e32 v1, s13 +; SDAG-NEXT: v_mov_b32_e32 v2, s14 +; SDAG-NEXT: v_mov_b32_e32 v3, s15 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v6, s0 -; SDAG-NEXT: v_mov_b32_e32 v7, s1 -; SDAG-NEXT: v_mov_b32_e32 v8, s2 -; SDAG-NEXT: v_mov_b32_e32 v9, s3 -; SDAG-NEXT: v_mov_b32_e32 v1, s16 +; SDAG-NEXT: v_mov_b32_e32 v4, s0 +; SDAG-NEXT: v_mov_b32_e32 v5, s1 +; SDAG-NEXT: v_mov_b32_e32 v6, s2 +; SDAG-NEXT: v_mov_b32_e32 v7, s3 +; SDAG-NEXT: v_mov_b32_e32 v17, s16 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_smfmac_f32_16x16x128_bf8_fp8 v[10:13], v[14:17], v[2:9], v1 cbsz:1 abid:2 +; SDAG-NEXT: v_smfmac_f32_16x16x128_bf8_fp8 v[8:11], v[12:15], v[0:7], v17 cbsz:1 abid:2 ; SDAG-NEXT: s_nop 7 -; SDAG-NEXT: global_store_dwordx4 v0, v[10:13], s[6:7] +; SDAG-NEXT: global_store_dwordx4 v16, v[8:11], s[6:7] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_smfmac_f32_16x16x128_bf8_fp8__vgpr: @@ -1735,30 +1735,30 @@ define amdgpu_kernel void @test_smfmac_f32_16x16x128_fp8_bf8__vgpr(ptr addrspace ; SDAG-NEXT: v_and_b32_e32 v0, 0x3ff, v0 ; SDAG-NEXT: v_lshlrev_b32_e32 v0, 4, v0 ; SDAG-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x34 +; SDAG-NEXT: v_mov_b32_e32 v16, 0 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: global_load_dwordx4 v[10:13], v0, s[6:7] +; SDAG-NEXT: global_load_dwordx4 v[8:11], v0, s[6:7] ; SDAG-NEXT: s_load_dword s16, s[4:5], 0x64 ; SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 -; SDAG-NEXT: v_mov_b32_e32 v0, 0 -; SDAG-NEXT: v_mov_b32_e32 v14, s8 -; SDAG-NEXT: v_mov_b32_e32 v15, s9 -; SDAG-NEXT: v_mov_b32_e32 v16, s10 -; SDAG-NEXT: v_mov_b32_e32 v17, s11 -; SDAG-NEXT: v_mov_b32_e32 v2, s12 -; SDAG-NEXT: v_mov_b32_e32 v3, s13 -; SDAG-NEXT: v_mov_b32_e32 v4, s14 -; SDAG-NEXT: v_mov_b32_e32 v5, s15 +; SDAG-NEXT: v_mov_b32_e32 v12, s8 +; SDAG-NEXT: v_mov_b32_e32 v13, s9 +; SDAG-NEXT: v_mov_b32_e32 v14, s10 +; SDAG-NEXT: v_mov_b32_e32 v15, s11 +; SDAG-NEXT: v_mov_b32_e32 v0, s12 +; SDAG-NEXT: v_mov_b32_e32 v1, s13 +; SDAG-NEXT: v_mov_b32_e32 v2, s14 +; SDAG-NEXT: v_mov_b32_e32 v3, s15 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v6, s0 -; SDAG-NEXT: v_mov_b32_e32 v7, s1 -; SDAG-NEXT: v_mov_b32_e32 v8, s2 -; SDAG-NEXT: v_mov_b32_e32 v9, s3 -; SDAG-NEXT: v_mov_b32_e32 v1, s16 +; SDAG-NEXT: v_mov_b32_e32 v4, s0 +; SDAG-NEXT: v_mov_b32_e32 v5, s1 +; SDAG-NEXT: v_mov_b32_e32 v6, s2 +; SDAG-NEXT: v_mov_b32_e32 v7, s3 +; SDAG-NEXT: v_mov_b32_e32 v17, s16 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_smfmac_f32_16x16x128_fp8_bf8 v[10:13], v[14:17], v[2:9], v1 cbsz:1 abid:2 +; SDAG-NEXT: v_smfmac_f32_16x16x128_fp8_bf8 v[8:11], v[12:15], v[0:7], v17 cbsz:1 abid:2 ; SDAG-NEXT: s_nop 7 -; SDAG-NEXT: global_store_dwordx4 v0, v[10:13], s[6:7] +; SDAG-NEXT: global_store_dwordx4 v16, v[8:11], s[6:7] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_smfmac_f32_16x16x128_fp8_bf8__vgpr: @@ -1904,30 +1904,30 @@ define amdgpu_kernel void @test_smfmac_f32_16x16x128_fp8_fp8__vgpr(ptr addrspace ; SDAG-NEXT: v_and_b32_e32 v0, 0x3ff, v0 ; SDAG-NEXT: v_lshlrev_b32_e32 v0, 4, v0 ; SDAG-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x34 +; SDAG-NEXT: v_mov_b32_e32 v16, 0 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: global_load_dwordx4 v[10:13], v0, s[6:7] +; SDAG-NEXT: global_load_dwordx4 v[8:11], v0, s[6:7] ; SDAG-NEXT: s_load_dword s16, s[4:5], 0x64 ; SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 -; SDAG-NEXT: v_mov_b32_e32 v0, 0 -; SDAG-NEXT: v_mov_b32_e32 v14, s8 -; SDAG-NEXT: v_mov_b32_e32 v15, s9 -; SDAG-NEXT: v_mov_b32_e32 v16, s10 -; SDAG-NEXT: v_mov_b32_e32 v17, s11 -; SDAG-NEXT: v_mov_b32_e32 v2, s12 -; SDAG-NEXT: v_mov_b32_e32 v3, s13 -; SDAG-NEXT: v_mov_b32_e32 v4, s14 -; SDAG-NEXT: v_mov_b32_e32 v5, s15 +; SDAG-NEXT: v_mov_b32_e32 v12, s8 +; SDAG-NEXT: v_mov_b32_e32 v13, s9 +; SDAG-NEXT: v_mov_b32_e32 v14, s10 +; SDAG-NEXT: v_mov_b32_e32 v15, s11 +; SDAG-NEXT: v_mov_b32_e32 v0, s12 +; SDAG-NEXT: v_mov_b32_e32 v1, s13 +; SDAG-NEXT: v_mov_b32_e32 v2, s14 +; SDAG-NEXT: v_mov_b32_e32 v3, s15 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v6, s0 -; SDAG-NEXT: v_mov_b32_e32 v7, s1 -; SDAG-NEXT: v_mov_b32_e32 v8, s2 -; SDAG-NEXT: v_mov_b32_e32 v9, s3 -; SDAG-NEXT: v_mov_b32_e32 v1, s16 +; SDAG-NEXT: v_mov_b32_e32 v4, s0 +; SDAG-NEXT: v_mov_b32_e32 v5, s1 +; SDAG-NEXT: v_mov_b32_e32 v6, s2 +; SDAG-NEXT: v_mov_b32_e32 v7, s3 +; SDAG-NEXT: v_mov_b32_e32 v17, s16 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_smfmac_f32_16x16x128_fp8_fp8 v[10:13], v[14:17], v[2:9], v1 cbsz:1 abid:2 +; SDAG-NEXT: v_smfmac_f32_16x16x128_fp8_fp8 v[8:11], v[12:15], v[0:7], v17 cbsz:1 abid:2 ; SDAG-NEXT: s_nop 7 -; SDAG-NEXT: global_store_dwordx4 v0, v[10:13], s[6:7] +; SDAG-NEXT: global_store_dwordx4 v16, v[8:11], s[6:7] ; SDAG-NEXT: s_endpgm ; ; GISEL-LABEL: test_smfmac_f32_16x16x128_fp8_fp8__vgpr: @@ -2081,22 +2081,22 @@ define amdgpu_kernel void @test_smfmac_f32_32x32x64_bf8_bf8__vgpr(ptr addrspace( ; SDAG-NEXT: s_load_dword s16, s[4:5], 0x64 ; SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v26, s8 -; SDAG-NEXT: v_mov_b32_e32 v27, s9 -; SDAG-NEXT: v_mov_b32_e32 v28, s10 -; SDAG-NEXT: v_mov_b32_e32 v29, s11 -; SDAG-NEXT: v_mov_b32_e32 v18, s12 -; SDAG-NEXT: v_mov_b32_e32 v19, s13 -; SDAG-NEXT: v_mov_b32_e32 v20, s14 -; SDAG-NEXT: v_mov_b32_e32 v21, s15 -; SDAG-NEXT: v_mov_b32_e32 v22, s0 -; SDAG-NEXT: v_mov_b32_e32 v23, s1 -; SDAG-NEXT: v_mov_b32_e32 v24, s2 -; SDAG-NEXT: v_mov_b32_e32 v25, s3 -; SDAG-NEXT: v_mov_b32_e32 v16, s16 +; SDAG-NEXT: v_mov_b32_e32 v24, s8 +; SDAG-NEXT: v_mov_b32_e32 v25, s9 +; SDAG-NEXT: v_mov_b32_e32 v26, s10 +; SDAG-NEXT: v_mov_b32_e32 v27, s11 +; SDAG-NEXT: v_mov_b32_e32 v16, s12 +; SDAG-NEXT: v_mov_b32_e32 v17, s13 +; SDAG-NEXT: v_mov_b32_e32 v18, s14 +; SDAG-NEXT: v_mov_b32_e32 v19, s15 +; SDAG-NEXT: v_mov_b32_e32 v20, s0 +; SDAG-NEXT: v_mov_b32_e32 v21, s1 +; SDAG-NEXT: v_mov_b32_e32 v22, s2 +; SDAG-NEXT: v_mov_b32_e32 v23, s3 +; SDAG-NEXT: v_mov_b32_e32 v28, s16 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_smfmac_f32_32x32x64_bf8_bf8 v[0:15], v[26:29], v[18:25], v16 cbsz:1 abid:2 +; SDAG-NEXT: v_smfmac_f32_32x32x64_bf8_bf8 v[0:15], v[24:27], v[16:23], v28 cbsz:1 abid:2 ; SDAG-NEXT: v_mov_b32_e32 v16, 0 ; SDAG-NEXT: s_nop 7 ; SDAG-NEXT: s_nop 2 @@ -2454,22 +2454,22 @@ define amdgpu_kernel void @test_smfmac_f32_32x32x64_bf8_fp8__vgpr(ptr addrspace( ; SDAG-NEXT: s_load_dword s16, s[4:5], 0x64 ; SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v26, s8 -; SDAG-NEXT: v_mov_b32_e32 v27, s9 -; SDAG-NEXT: v_mov_b32_e32 v28, s10 -; SDAG-NEXT: v_mov_b32_e32 v29, s11 -; SDAG-NEXT: v_mov_b32_e32 v18, s12 -; SDAG-NEXT: v_mov_b32_e32 v19, s13 -; SDAG-NEXT: v_mov_b32_e32 v20, s14 -; SDAG-NEXT: v_mov_b32_e32 v21, s15 -; SDAG-NEXT: v_mov_b32_e32 v22, s0 -; SDAG-NEXT: v_mov_b32_e32 v23, s1 -; SDAG-NEXT: v_mov_b32_e32 v24, s2 -; SDAG-NEXT: v_mov_b32_e32 v25, s3 -; SDAG-NEXT: v_mov_b32_e32 v16, s16 +; SDAG-NEXT: v_mov_b32_e32 v24, s8 +; SDAG-NEXT: v_mov_b32_e32 v25, s9 +; SDAG-NEXT: v_mov_b32_e32 v26, s10 +; SDAG-NEXT: v_mov_b32_e32 v27, s11 +; SDAG-NEXT: v_mov_b32_e32 v16, s12 +; SDAG-NEXT: v_mov_b32_e32 v17, s13 +; SDAG-NEXT: v_mov_b32_e32 v18, s14 +; SDAG-NEXT: v_mov_b32_e32 v19, s15 +; SDAG-NEXT: v_mov_b32_e32 v20, s0 +; SDAG-NEXT: v_mov_b32_e32 v21, s1 +; SDAG-NEXT: v_mov_b32_e32 v22, s2 +; SDAG-NEXT: v_mov_b32_e32 v23, s3 +; SDAG-NEXT: v_mov_b32_e32 v28, s16 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_smfmac_f32_32x32x64_bf8_fp8 v[0:15], v[26:29], v[18:25], v16 cbsz:1 abid:2 +; SDAG-NEXT: v_smfmac_f32_32x32x64_bf8_fp8 v[0:15], v[24:27], v[16:23], v28 cbsz:1 abid:2 ; SDAG-NEXT: v_mov_b32_e32 v16, 0 ; SDAG-NEXT: s_nop 7 ; SDAG-NEXT: s_nop 2 @@ -2827,22 +2827,22 @@ define amdgpu_kernel void @test_smfmac_f32_32x32x64_fp8_bf8__vgpr(ptr addrspace( ; SDAG-NEXT: s_load_dword s16, s[4:5], 0x64 ; SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v26, s8 -; SDAG-NEXT: v_mov_b32_e32 v27, s9 -; SDAG-NEXT: v_mov_b32_e32 v28, s10 -; SDAG-NEXT: v_mov_b32_e32 v29, s11 -; SDAG-NEXT: v_mov_b32_e32 v18, s12 -; SDAG-NEXT: v_mov_b32_e32 v19, s13 -; SDAG-NEXT: v_mov_b32_e32 v20, s14 -; SDAG-NEXT: v_mov_b32_e32 v21, s15 -; SDAG-NEXT: v_mov_b32_e32 v22, s0 -; SDAG-NEXT: v_mov_b32_e32 v23, s1 -; SDAG-NEXT: v_mov_b32_e32 v24, s2 -; SDAG-NEXT: v_mov_b32_e32 v25, s3 -; SDAG-NEXT: v_mov_b32_e32 v16, s16 +; SDAG-NEXT: v_mov_b32_e32 v24, s8 +; SDAG-NEXT: v_mov_b32_e32 v25, s9 +; SDAG-NEXT: v_mov_b32_e32 v26, s10 +; SDAG-NEXT: v_mov_b32_e32 v27, s11 +; SDAG-NEXT: v_mov_b32_e32 v16, s12 +; SDAG-NEXT: v_mov_b32_e32 v17, s13 +; SDAG-NEXT: v_mov_b32_e32 v18, s14 +; SDAG-NEXT: v_mov_b32_e32 v19, s15 +; SDAG-NEXT: v_mov_b32_e32 v20, s0 +; SDAG-NEXT: v_mov_b32_e32 v21, s1 +; SDAG-NEXT: v_mov_b32_e32 v22, s2 +; SDAG-NEXT: v_mov_b32_e32 v23, s3 +; SDAG-NEXT: v_mov_b32_e32 v28, s16 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_smfmac_f32_32x32x64_fp8_bf8 v[0:15], v[26:29], v[18:25], v16 cbsz:1 abid:2 +; SDAG-NEXT: v_smfmac_f32_32x32x64_fp8_bf8 v[0:15], v[24:27], v[16:23], v28 cbsz:1 abid:2 ; SDAG-NEXT: v_mov_b32_e32 v16, 0 ; SDAG-NEXT: s_nop 7 ; SDAG-NEXT: s_nop 2 @@ -3200,22 +3200,22 @@ define amdgpu_kernel void @test_smfmac_f32_32x32x64_fp8_fp8__vgpr(ptr addrspace( ; SDAG-NEXT: s_load_dword s16, s[4:5], 0x64 ; SDAG-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x54 ; SDAG-NEXT: s_waitcnt lgkmcnt(0) -; SDAG-NEXT: v_mov_b32_e32 v26, s8 -; SDAG-NEXT: v_mov_b32_e32 v27, s9 -; SDAG-NEXT: v_mov_b32_e32 v28, s10 -; SDAG-NEXT: v_mov_b32_e32 v29, s11 -; SDAG-NEXT: v_mov_b32_e32 v18, s12 -; SDAG-NEXT: v_mov_b32_e32 v19, s13 -; SDAG-NEXT: v_mov_b32_e32 v20, s14 -; SDAG-NEXT: v_mov_b32_e32 v21, s15 -; SDAG-NEXT: v_mov_b32_e32 v22, s0 -; SDAG-NEXT: v_mov_b32_e32 v23, s1 -; SDAG-NEXT: v_mov_b32_e32 v24, s2 -; SDAG-NEXT: v_mov_b32_e32 v25, s3 -; SDAG-NEXT: v_mov_b32_e32 v16, s16 +; SDAG-NEXT: v_mov_b32_e32 v24, s8 +; SDAG-NEXT: v_mov_b32_e32 v25, s9 +; SDAG-NEXT: v_mov_b32_e32 v26, s10 +; SDAG-NEXT: v_mov_b32_e32 v27, s11 +; SDAG-NEXT: v_mov_b32_e32 v16, s12 +; SDAG-NEXT: v_mov_b32_e32 v17, s13 +; SDAG-NEXT: v_mov_b32_e32 v18, s14 +; SDAG-NEXT: v_mov_b32_e32 v19, s15 +; SDAG-NEXT: v_mov_b32_e32 v20, s0 +; SDAG-NEXT: v_mov_b32_e32 v21, s1 +; SDAG-NEXT: v_mov_b32_e32 v22, s2 +; SDAG-NEXT: v_mov_b32_e32 v23, s3 +; SDAG-NEXT: v_mov_b32_e32 v28, s16 ; SDAG-NEXT: s_waitcnt vmcnt(0) ; SDAG-NEXT: s_nop 0 -; SDAG-NEXT: v_smfmac_f32_32x32x64_fp8_fp8 v[0:15], v[26:29], v[18:25], v16 cbsz:1 abid:2 +; SDAG-NEXT: v_smfmac_f32_32x32x64_fp8_fp8 v[0:15], v[24:27], v[16:23], v28 cbsz:1 abid:2 ; SDAG-NEXT: v_mov_b32_e32 v16, 0 ; SDAG-NEXT: s_nop 7 ; SDAG-NEXT: s_nop 2 @@ -3552,4 +3552,4 @@ define <16 x float> @test_smfmac_f32_32x32x64_fp8_fp8__sgpr(<4 x i32> inreg %arg ret <16 x float> %result } -attributes #0 = { "amdgpu-flat-work-group-size"="1,256" } +attributes #0 = { "amdgpu-flat-work-group-size"="1,256" "amdgpu-agpr-alloc"="0,0" } diff --git a/llvm/test/CodeGen/AMDGPU/mad-mix-hi-bf16.ll b/llvm/test/CodeGen/AMDGPU/mad-mix-hi-bf16.ll index 84123e6..393581f 100644 --- a/llvm/test/CodeGen/AMDGPU/mad-mix-hi-bf16.ll +++ b/llvm/test/CodeGen/AMDGPU/mad-mix-hi-bf16.ll @@ -141,7 +141,6 @@ define <2 x bfloat> @v_mad_mixhi_bf16_bf16lo_bf16lo_bf16lo_undeflo_clamp_postcvt ; GFX1250-NEXT: s_wait_kmcnt 0x0 ; GFX1250-NEXT: v_fma_mixlo_bf16 v3, v0, v1, v2 op_sel_hi:[1,1,1] ; GFX1250-NEXT: v_fma_mixhi_bf16 v0, v0, v1, v2 op_sel_hi:[1,1,1] clamp -; GFX1250-NEXT: s_wait_storecnt 0x0 ; GFX1250-NEXT: global_store_b16 v[0:1], v3, off scope:SCOPE_SYS ; GFX1250-NEXT: s_wait_storecnt 0x0 ; GFX1250-NEXT: s_set_pc_i64 s[30:31] diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-volatile.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-volatile.ll index bc25084..5e5e3bf 100644 --- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-volatile.ll +++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-volatile.ll @@ -415,11 +415,6 @@ define amdgpu_kernel void @local_volatile_store_0( ; GFX12-WGP-NEXT: v_mov_b32_e32 v0, s1 ; GFX12-WGP-NEXT: s_wait_kmcnt 0x0 ; GFX12-WGP-NEXT: v_mov_b32_e32 v1, s0 -; GFX12-WGP-NEXT: s_wait_loadcnt 0x0 -; GFX12-WGP-NEXT: s_wait_samplecnt 0x0 -; GFX12-WGP-NEXT: s_wait_bvhcnt 0x0 -; GFX12-WGP-NEXT: s_wait_kmcnt 0x0 -; GFX12-WGP-NEXT: s_wait_storecnt 0x0 ; GFX12-WGP-NEXT: ds_store_b32 v0, v1 ; GFX12-WGP-NEXT: s_endpgm ; @@ -432,11 +427,6 @@ define amdgpu_kernel void @local_volatile_store_0( ; GFX12-CU-NEXT: v_mov_b32_e32 v0, s1 ; GFX12-CU-NEXT: s_wait_kmcnt 0x0 ; GFX12-CU-NEXT: v_mov_b32_e32 v1, s0 -; GFX12-CU-NEXT: s_wait_loadcnt 0x0 -; GFX12-CU-NEXT: s_wait_samplecnt 0x0 -; GFX12-CU-NEXT: s_wait_bvhcnt 0x0 -; GFX12-CU-NEXT: s_wait_kmcnt 0x0 -; GFX12-CU-NEXT: s_wait_storecnt 0x0 ; GFX12-CU-NEXT: ds_store_b32 v0, v1 ; GFX12-CU-NEXT: s_endpgm ptr addrspace(1) %in, ptr addrspace(3) %out) { @@ -562,11 +552,6 @@ define amdgpu_kernel void @local_volatile_store_1( ; GFX12-WGP-NEXT: v_lshl_add_u32 v0, v0, s1, s2 ; GFX12-WGP-NEXT: s_wait_kmcnt 0x0 ; GFX12-WGP-NEXT: v_mov_b32_e32 v1, s0 -; GFX12-WGP-NEXT: s_wait_loadcnt 0x0 -; GFX12-WGP-NEXT: s_wait_samplecnt 0x0 -; GFX12-WGP-NEXT: s_wait_bvhcnt 0x0 -; GFX12-WGP-NEXT: s_wait_kmcnt 0x0 -; GFX12-WGP-NEXT: s_wait_storecnt 0x0 ; GFX12-WGP-NEXT: ds_store_b32 v0, v1 ; GFX12-WGP-NEXT: s_endpgm ; @@ -583,11 +568,6 @@ define amdgpu_kernel void @local_volatile_store_1( ; GFX12-CU-NEXT: v_lshl_add_u32 v0, v0, s1, s2 ; GFX12-CU-NEXT: s_wait_kmcnt 0x0 ; GFX12-CU-NEXT: v_mov_b32_e32 v1, s0 -; GFX12-CU-NEXT: s_wait_loadcnt 0x0 -; GFX12-CU-NEXT: s_wait_samplecnt 0x0 -; GFX12-CU-NEXT: s_wait_bvhcnt 0x0 -; GFX12-CU-NEXT: s_wait_kmcnt 0x0 -; GFX12-CU-NEXT: s_wait_storecnt 0x0 ; GFX12-CU-NEXT: ds_store_b32 v0, v1 ; GFX12-CU-NEXT: s_endpgm ptr addrspace(1) %in, ptr addrspace(3) %out) { diff --git a/llvm/test/CodeGen/AMDGPU/partial-regcopy-and-spill-missed-at-regalloc.ll b/llvm/test/CodeGen/AMDGPU/partial-regcopy-and-spill-missed-at-regalloc.ll index c3164b8..f54a383 100644 --- a/llvm/test/CodeGen/AMDGPU/partial-regcopy-and-spill-missed-at-regalloc.ll +++ b/llvm/test/CodeGen/AMDGPU/partial-regcopy-and-spill-missed-at-regalloc.ll @@ -1,7 +1,7 @@ ; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py -;RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx908 --stop-after=greedy,1 < %s | FileCheck -check-prefix=REGALLOC-GFX908 %s +;RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx908 --stop-after=greedy,2 < %s | FileCheck -check-prefix=REGALLOC-GFX908 %s ;RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx908 --stop-after=prologepilog < %s | FileCheck -check-prefix=PEI-GFX908 %s -;RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx90a --stop-after=greedy,1 < %s | FileCheck -check-prefix=REGALLOC-GFX90A %s +;RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx90a --stop-after=greedy,2 < %s | FileCheck -check-prefix=REGALLOC-GFX90A %s ;RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx90a --stop-after=prologepilog < %s | FileCheck -check-prefix=PEI-GFX90A %s ; Partial reg copy and spill missed during regalloc handled later at frame lowering. @@ -12,17 +12,21 @@ define amdgpu_kernel void @partial_copy(<4 x i32> %arg) #0 { ; REGALLOC-GFX908-NEXT: liveins: $sgpr4_sgpr5 ; REGALLOC-GFX908-NEXT: {{ $}} ; REGALLOC-GFX908-NEXT: INLINEASM &"; use $0", 1 /* sideeffect attdialect */, 2162697 /* reguse:AGPR_32 */, undef %6:agpr_32 - ; REGALLOC-GFX908-NEXT: INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 6094858 /* regdef:VReg_128 */, def %7 - ; REGALLOC-GFX908-NEXT: INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 3538954 /* regdef:VReg_64 */, def %8 - ; REGALLOC-GFX908-NEXT: GLOBAL_STORE_DWORDX4 undef %15:vreg_64, %7, 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1) + ; REGALLOC-GFX908-NEXT: INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 6094858 /* regdef:VReg_128 */, def %25 + ; REGALLOC-GFX908-NEXT: [[COPY:%[0-9]+]]:av_128 = COPY %25 + ; REGALLOC-GFX908-NEXT: INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 3538954 /* regdef:VReg_64 */, def %27 + ; REGALLOC-GFX908-NEXT: SI_SPILL_AV64_SAVE %27, %stack.0, $sgpr32, 0, implicit $exec :: (store (s64) into %stack.0, align 4, addrspace 5) + ; REGALLOC-GFX908-NEXT: [[COPY1:%[0-9]+]]:vreg_128 = COPY [[COPY]] + ; REGALLOC-GFX908-NEXT: GLOBAL_STORE_DWORDX4 undef %15:vreg_64, [[COPY1]], 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1) ; REGALLOC-GFX908-NEXT: renamable $sgpr0_sgpr1_sgpr2_sgpr3 = S_LOAD_DWORDX4_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (s128) from %ir.arg.kernarg.offset1, addrspace 4) - ; REGALLOC-GFX908-NEXT: [[COPY:%[0-9]+]]:areg_128 = COPY killed renamable $sgpr0_sgpr1_sgpr2_sgpr3 + ; REGALLOC-GFX908-NEXT: [[COPY2:%[0-9]+]]:areg_128 = COPY killed renamable $sgpr0_sgpr1_sgpr2_sgpr3 ; REGALLOC-GFX908-NEXT: [[AV_MOV_:%[0-9]+]]:vgpr_32 = AV_MOV_B32_IMM_PSEUDO 1, implicit $exec ; REGALLOC-GFX908-NEXT: [[AV_MOV_1:%[0-9]+]]:vgpr_32 = AV_MOV_B32_IMM_PSEUDO 2, implicit $exec - ; REGALLOC-GFX908-NEXT: [[V_MFMA_I32_4X4X4I8_e64_:%[0-9]+]]:areg_128 = V_MFMA_I32_4X4X4I8_e64 [[AV_MOV_]], [[AV_MOV_1]], [[COPY]], 0, 0, 0, implicit $mode, implicit $exec - ; REGALLOC-GFX908-NEXT: GLOBAL_STORE_DWORDX2 undef %17:vreg_64, %8, 0, 0, implicit $exec :: (volatile store (s64) into `ptr addrspace(1) poison`, addrspace 1) - ; REGALLOC-GFX908-NEXT: [[COPY1:%[0-9]+]]:vreg_128 = COPY [[V_MFMA_I32_4X4X4I8_e64_]] - ; REGALLOC-GFX908-NEXT: GLOBAL_STORE_DWORDX4 undef %19:vreg_64, [[COPY1]], 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1) + ; REGALLOC-GFX908-NEXT: [[V_MFMA_I32_4X4X4I8_e64_:%[0-9]+]]:areg_128 = V_MFMA_I32_4X4X4I8_e64 [[AV_MOV_]], [[AV_MOV_1]], [[COPY2]], 0, 0, 0, implicit $mode, implicit $exec + ; REGALLOC-GFX908-NEXT: [[SI_SPILL_AV64_RESTORE:%[0-9]+]]:vreg_64 = SI_SPILL_AV64_RESTORE %stack.0, $sgpr32, 0, implicit $exec :: (load (s64) from %stack.0, align 4, addrspace 5) + ; REGALLOC-GFX908-NEXT: GLOBAL_STORE_DWORDX2 undef %17:vreg_64, [[SI_SPILL_AV64_RESTORE]], 0, 0, implicit $exec :: (volatile store (s64) into `ptr addrspace(1) poison`, addrspace 1) + ; REGALLOC-GFX908-NEXT: [[COPY3:%[0-9]+]]:vreg_128 = COPY [[V_MFMA_I32_4X4X4I8_e64_]] + ; REGALLOC-GFX908-NEXT: GLOBAL_STORE_DWORDX4 undef %19:vreg_64, [[COPY3]], 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1) ; REGALLOC-GFX908-NEXT: S_ENDPGM 0 ; ; PEI-GFX908-LABEL: name: partial_copy @@ -57,15 +61,17 @@ define amdgpu_kernel void @partial_copy(<4 x i32> %arg) #0 { ; REGALLOC-GFX90A-NEXT: liveins: $sgpr4_sgpr5 ; REGALLOC-GFX90A-NEXT: {{ $}} ; REGALLOC-GFX90A-NEXT: INLINEASM &"; use $0", 1 /* sideeffect attdialect */, 2162697 /* reguse:AGPR_32 */, undef %6:agpr_32 - ; REGALLOC-GFX90A-NEXT: INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 6422538 /* regdef:VReg_128_Align2 */, def %7 - ; REGALLOC-GFX90A-NEXT: INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 3866634 /* regdef:VReg_64_Align2 */, def %8 - ; REGALLOC-GFX90A-NEXT: GLOBAL_STORE_DWORDX4 undef %15:vreg_64_align2, %7, 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1) + ; REGALLOC-GFX90A-NEXT: INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 6422538 /* regdef:VReg_128_Align2 */, def %24 + ; REGALLOC-GFX90A-NEXT: [[COPY:%[0-9]+]]:av_128_align2 = COPY %24 + ; REGALLOC-GFX90A-NEXT: INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 3866634 /* regdef:VReg_64_Align2 */, def %22 + ; REGALLOC-GFX90A-NEXT: [[COPY1:%[0-9]+]]:av_64_align2 = COPY %22 + ; REGALLOC-GFX90A-NEXT: GLOBAL_STORE_DWORDX4 undef %15:vreg_64_align2, [[COPY]], 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1) ; REGALLOC-GFX90A-NEXT: renamable $sgpr0_sgpr1_sgpr2_sgpr3 = S_LOAD_DWORDX4_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (s128) from %ir.arg.kernarg.offset1, addrspace 4) - ; REGALLOC-GFX90A-NEXT: [[COPY:%[0-9]+]]:areg_128_align2 = COPY killed renamable $sgpr0_sgpr1_sgpr2_sgpr3 + ; REGALLOC-GFX90A-NEXT: [[COPY2:%[0-9]+]]:areg_128_align2 = COPY killed renamable $sgpr0_sgpr1_sgpr2_sgpr3 ; REGALLOC-GFX90A-NEXT: [[AV_MOV_:%[0-9]+]]:vgpr_32 = AV_MOV_B32_IMM_PSEUDO 1, implicit $exec ; REGALLOC-GFX90A-NEXT: [[AV_MOV_1:%[0-9]+]]:vgpr_32 = AV_MOV_B32_IMM_PSEUDO 2, implicit $exec - ; REGALLOC-GFX90A-NEXT: [[V_MFMA_I32_4X4X4I8_e64_:%[0-9]+]]:areg_128_align2 = V_MFMA_I32_4X4X4I8_e64 [[AV_MOV_]], [[AV_MOV_1]], [[COPY]], 0, 0, 0, implicit $mode, implicit $exec - ; REGALLOC-GFX90A-NEXT: GLOBAL_STORE_DWORDX2 undef %17:vreg_64_align2, %8, 0, 0, implicit $exec :: (volatile store (s64) into `ptr addrspace(1) poison`, addrspace 1) + ; REGALLOC-GFX90A-NEXT: [[V_MFMA_I32_4X4X4I8_e64_:%[0-9]+]]:areg_128_align2 = V_MFMA_I32_4X4X4I8_e64 [[AV_MOV_]], [[AV_MOV_1]], [[COPY2]], 0, 0, 0, implicit $mode, implicit $exec + ; REGALLOC-GFX90A-NEXT: GLOBAL_STORE_DWORDX2 undef %17:vreg_64_align2, [[COPY1]], 0, 0, implicit $exec :: (volatile store (s64) into `ptr addrspace(1) poison`, addrspace 1) ; REGALLOC-GFX90A-NEXT: GLOBAL_STORE_DWORDX4 undef %19:vreg_64_align2, [[V_MFMA_I32_4X4X4I8_e64_]], 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1) ; REGALLOC-GFX90A-NEXT: S_ENDPGM 0 ; diff --git a/llvm/test/CodeGen/AMDGPU/ssubo.ll b/llvm/test/CodeGen/AMDGPU/ssubo.ll index 053038d..382d892 100644 --- a/llvm/test/CodeGen/AMDGPU/ssubo.ll +++ b/llvm/test/CodeGen/AMDGPU/ssubo.ll @@ -1,14 +1,116 @@ -; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tahiti < %s | FileCheck -check-prefixes=GCN,SI,FUNC %s -; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tonga < %s | FileCheck -check-prefixes=GCN,VI,FUNC %s -; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx900 < %s | FileCheck -check-prefixes=GCN,GFX9,FUNC %s - +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 +; RUN: llc < %s -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tahiti | FileCheck %s --check-prefix=SI +; RUN: llc < %s -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tonga | FileCheck %s --check-prefix=VI +; RUN: llc < %s -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx900 | FileCheck %s --check-prefix=GFX9 +; RUN: llc < %s -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1010 | FileCheck %s --check-prefix=GFX10 +; RUN: llc < %s -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 | FileCheck %s --check-prefix=GFX11 declare { i32, i1 } @llvm.ssub.with.overflow.i32(i32, i32) nounwind readnone declare { i64, i1 } @llvm.ssub.with.overflow.i64(i64, i64) nounwind readnone declare { <2 x i32>, <2 x i1> } @llvm.ssub.with.overflow.v2i32(<2 x i32>, <2 x i32>) nounwind readnone -; FUNC-LABEL: {{^}}ssubo_i64_zext: define amdgpu_kernel void @ssubo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 %b) nounwind { +; SI-LABEL: ssubo_i64_zext: +; SI: ; %bb.0: +; SI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 +; SI-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd +; SI-NEXT: s_mov_b32 s7, 0xf000 +; SI-NEXT: s_mov_b32 s6, -1 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: v_mov_b32_e32 v0, s2 +; SI-NEXT: s_sub_u32 s10, s2, s8 +; SI-NEXT: s_subb_u32 s11, s3, s9 +; SI-NEXT: v_mov_b32_e32 v1, s3 +; SI-NEXT: v_cmp_lt_i64_e32 vcc, s[10:11], v[0:1] +; SI-NEXT: v_cmp_gt_i64_e64 s[2:3], s[8:9], 0 +; SI-NEXT: s_mov_b32 s4, s0 +; SI-NEXT: s_mov_b32 s5, s1 +; SI-NEXT: s_xor_b64 s[0:1], s[2:3], vcc +; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; SI-NEXT: v_mov_b32_e32 v1, s11 +; SI-NEXT: v_add_i32_e32 v0, vcc, s10, v0 +; SI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc +; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0 +; SI-NEXT: s_endpgm +; +; VI-LABEL: ssubo_i64_zext: +; VI: ; %bb.0: +; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 +; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x34 +; VI-NEXT: s_waitcnt lgkmcnt(0) +; VI-NEXT: v_mov_b32_e32 v1, s2 +; VI-NEXT: s_sub_u32 s6, s2, s4 +; VI-NEXT: v_mov_b32_e32 v2, s3 +; VI-NEXT: s_subb_u32 s7, s3, s5 +; VI-NEXT: v_cmp_gt_i64_e64 s[8:9], s[4:5], 0 +; VI-NEXT: v_cmp_lt_i64_e32 vcc, s[6:7], v[1:2] +; VI-NEXT: v_mov_b32_e32 v0, s0 +; VI-NEXT: v_mov_b32_e32 v1, s1 +; VI-NEXT: s_xor_b64 s[0:1], s[8:9], vcc +; VI-NEXT: v_cndmask_b32_e64 v2, 0, 1, s[0:1] +; VI-NEXT: v_mov_b32_e32 v3, s7 +; VI-NEXT: v_add_u32_e32 v2, vcc, s6, v2 +; VI-NEXT: v_addc_u32_e32 v3, vcc, 0, v3, vcc +; VI-NEXT: flat_store_dwordx2 v[0:1], v[2:3] +; VI-NEXT: s_endpgm +; +; GFX9-LABEL: ssubo_i64_zext: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 +; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34 +; GFX9-NEXT: v_mov_b32_e32 v2, 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: v_mov_b32_e32 v0, s2 +; GFX9-NEXT: s_sub_u32 s4, s2, s6 +; GFX9-NEXT: v_mov_b32_e32 v1, s3 +; GFX9-NEXT: s_subb_u32 s5, s3, s7 +; GFX9-NEXT: v_cmp_gt_i64_e64 s[8:9], s[6:7], 0 +; GFX9-NEXT: v_cmp_lt_i64_e32 vcc, s[4:5], v[0:1] +; GFX9-NEXT: v_mov_b32_e32 v1, s5 +; GFX9-NEXT: s_xor_b64 s[2:3], s[8:9], vcc +; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[2:3] +; GFX9-NEXT: v_add_co_u32_e32 v0, vcc, s4, v0 +; GFX9-NEXT: v_addc_co_u32_e32 v1, vcc, 0, v1, vcc +; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] +; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: ssubo_i64_zext: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 +; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34 +; GFX10-NEXT: v_mov_b32_e32 v2, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_sub_u32 s4, s2, s6 +; GFX10-NEXT: s_subb_u32 s5, s3, s7 +; GFX10-NEXT: v_cmp_gt_i64_e64 s6, s[6:7], 0 +; GFX10-NEXT: v_cmp_lt_i64_e64 s2, s[4:5], s[2:3] +; GFX10-NEXT: s_xor_b32 s2, s6, s2 +; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, s2 +; GFX10-NEXT: v_add_co_u32 v0, s2, s4, v0 +; GFX10-NEXT: v_add_co_ci_u32_e64 v1, s2, s5, 0, s2 +; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: ssubo_i64_zext: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 +; GFX11-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 +; GFX11-NEXT: v_mov_b32_e32 v2, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_sub_u32 s6, s2, s4 +; GFX11-NEXT: s_subb_u32 s7, s3, s5 +; GFX11-NEXT: v_cmp_gt_i64_e64 s4, s[4:5], 0 +; GFX11-NEXT: v_cmp_lt_i64_e64 s2, s[6:7], s[2:3] +; GFX11-NEXT: s_xor_b32 s2, s4, s2 +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_cndmask_b32_e64 v0, 0, 1, s2 +; GFX11-NEXT: v_add_co_u32 v0, s2, s6, v0 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_add_co_ci_u32_e64 v1, null, s7, 0, s2 +; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1] +; GFX11-NEXT: s_endpgm %ssub = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 %a, i64 %b) nounwind %val = extractvalue { i64, i1 } %ssub, 0 %carry = extractvalue { i64, i1 } %ssub, 1 @@ -18,8 +120,102 @@ define amdgpu_kernel void @ssubo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 %b) ret void } -; FUNC-LABEL: {{^}}s_ssubo_i32: define amdgpu_kernel void @s_ssubo_i32(ptr addrspace(1) %out, ptr addrspace(1) %carryout, i32 %a, i32 %b) nounwind { +; SI-LABEL: s_ssubo_i32: +; SI: ; %bb.0: +; SI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 +; SI-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd +; SI-NEXT: s_mov_b32 s7, 0xf000 +; SI-NEXT: s_mov_b32 s6, -1 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s4, s0 +; SI-NEXT: s_sub_i32 s12, s8, s9 +; SI-NEXT: s_cmp_gt_i32 s9, 0 +; SI-NEXT: s_cselect_b64 s[10:11], -1, 0 +; SI-NEXT: s_cmp_lt_i32 s12, s8 +; SI-NEXT: s_mov_b32 s5, s1 +; SI-NEXT: s_cselect_b64 s[8:9], -1, 0 +; SI-NEXT: v_mov_b32_e32 v0, s12 +; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0 +; SI-NEXT: s_xor_b64 s[4:5], s[10:11], s[8:9] +; SI-NEXT: s_mov_b32 s0, s2 +; SI-NEXT: s_mov_b32 s1, s3 +; SI-NEXT: s_mov_b32 s2, s6 +; SI-NEXT: s_mov_b32 s3, s7 +; SI-NEXT: s_waitcnt expcnt(0) +; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[4:5] +; SI-NEXT: buffer_store_byte v0, off, s[0:3], 0 +; SI-NEXT: s_endpgm +; +; VI-LABEL: s_ssubo_i32: +; VI: ; %bb.0: +; VI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 +; VI-NEXT: s_load_dwordx2 s[4:5], s[4:5], 0x34 +; VI-NEXT: s_waitcnt lgkmcnt(0) +; VI-NEXT: v_mov_b32_e32 v0, s0 +; VI-NEXT: s_sub_i32 s6, s4, s5 +; VI-NEXT: s_cmp_gt_i32 s5, 0 +; VI-NEXT: v_mov_b32_e32 v1, s1 +; VI-NEXT: s_cselect_b64 s[0:1], -1, 0 +; VI-NEXT: s_cmp_lt_i32 s6, s4 +; VI-NEXT: v_mov_b32_e32 v2, s2 +; VI-NEXT: v_mov_b32_e32 v3, s3 +; VI-NEXT: s_cselect_b64 s[2:3], -1, 0 +; VI-NEXT: v_mov_b32_e32 v4, s6 +; VI-NEXT: s_xor_b64 s[0:1], s[0:1], s[2:3] +; VI-NEXT: flat_store_dword v[0:1], v4 +; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; VI-NEXT: flat_store_byte v[2:3], v0 +; VI-NEXT: s_endpgm +; +; GFX9-LABEL: s_ssubo_i32: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34 +; GFX9-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 +; GFX9-NEXT: v_mov_b32_e32 v0, 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: v_mov_b32_e32 v1, s7 +; GFX9-NEXT: s_sub_i32 s4, s6, s7 +; GFX9-NEXT: v_sub_i32 v1, s6, v1 clamp +; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, s4, v1 +; GFX9-NEXT: v_mov_b32_e32 v2, s4 +; GFX9-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc +; GFX9-NEXT: global_store_dword v0, v2, s[0:1] +; GFX9-NEXT: global_store_byte v0, v1, s[2:3] +; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: s_ssubo_i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34 +; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v1, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: v_sub_nc_i32 v0, s6, s7 clamp +; GFX10-NEXT: s_sub_i32 s4, s6, s7 +; GFX10-NEXT: v_mov_b32_e32 v2, s4 +; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, s4, v0 +; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo +; GFX10-NEXT: global_store_dword v1, v2, s[0:1] +; GFX10-NEXT: global_store_byte v1, v0, s[2:3] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: s_ssubo_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: s_load_b64 s[6:7], s[4:5], 0x34 +; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: v_sub_nc_i32 v0, s6, s7 clamp +; GFX11-NEXT: s_sub_i32 s4, s6, s7 +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_2) +; GFX11-NEXT: v_dual_mov_b32 v1, 0 :: v_dual_mov_b32 v2, s4 +; GFX11-NEXT: v_cmp_ne_u32_e32 vcc_lo, s4, v0 +; GFX11-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b32 v1, v2, s[0:1] +; GFX11-NEXT: global_store_b8 v1, v0, s[2:3] +; GFX11-NEXT: s_endpgm %ssub = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 %a, i32 %b) nounwind %val = extractvalue { i32, i1 } %ssub, 0 %carry = extractvalue { i32, i1 } %ssub, 1 @@ -28,8 +224,112 @@ define amdgpu_kernel void @s_ssubo_i32(ptr addrspace(1) %out, ptr addrspace(1) % ret void } -; FUNC-LABEL: {{^}}v_ssubo_i32: define amdgpu_kernel void @v_ssubo_i32(ptr addrspace(1) %out, ptr addrspace(1) %carryout, ptr addrspace(1) %aptr, ptr addrspace(1) %bptr) nounwind { +; SI-LABEL: v_ssubo_i32: +; SI: ; %bb.0: +; SI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x9 +; SI-NEXT: s_mov_b32 s11, 0xf000 +; SI-NEXT: s_mov_b32 s10, -1 +; SI-NEXT: s_mov_b32 s14, s10 +; SI-NEXT: s_mov_b32 s15, s11 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s12, s4 +; SI-NEXT: s_mov_b32 s13, s5 +; SI-NEXT: s_mov_b32 s4, s6 +; SI-NEXT: s_mov_b32 s5, s7 +; SI-NEXT: s_mov_b32 s6, s10 +; SI-NEXT: s_mov_b32 s7, s11 +; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 +; SI-NEXT: buffer_load_dword v1, off, s[4:7], 0 +; SI-NEXT: s_mov_b32 s8, s0 +; SI-NEXT: s_mov_b32 s9, s1 +; SI-NEXT: s_mov_b32 s4, s2 +; SI-NEXT: s_mov_b32 s5, s3 +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: v_sub_i32_e32 v2, vcc, v0, v1 +; SI-NEXT: v_cmp_lt_i32_e32 vcc, 0, v1 +; SI-NEXT: v_cmp_lt_i32_e64 s[0:1], v2, v0 +; SI-NEXT: s_xor_b64 s[0:1], vcc, s[0:1] +; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; SI-NEXT: buffer_store_dword v2, off, s[8:11], 0 +; SI-NEXT: buffer_store_byte v0, off, s[4:7], 0 +; SI-NEXT: s_endpgm +; +; VI-LABEL: v_ssubo_i32: +; VI: ; %bb.0: +; VI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x24 +; VI-NEXT: s_waitcnt lgkmcnt(0) +; VI-NEXT: v_mov_b32_e32 v0, s4 +; VI-NEXT: v_mov_b32_e32 v1, s5 +; VI-NEXT: v_mov_b32_e32 v2, s6 +; VI-NEXT: v_mov_b32_e32 v3, s7 +; VI-NEXT: flat_load_dword v4, v[0:1] +; VI-NEXT: flat_load_dword v5, v[2:3] +; VI-NEXT: v_mov_b32_e32 v0, s0 +; VI-NEXT: v_mov_b32_e32 v1, s1 +; VI-NEXT: v_mov_b32_e32 v2, s2 +; VI-NEXT: v_mov_b32_e32 v3, s3 +; VI-NEXT: s_waitcnt vmcnt(0) +; VI-NEXT: v_sub_u32_e32 v6, vcc, v4, v5 +; VI-NEXT: v_cmp_lt_i32_e32 vcc, 0, v5 +; VI-NEXT: v_cmp_lt_i32_e64 s[0:1], v6, v4 +; VI-NEXT: s_xor_b64 s[0:1], vcc, s[0:1] +; VI-NEXT: flat_store_dword v[0:1], v6 +; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; VI-NEXT: flat_store_byte v[2:3], v0 +; VI-NEXT: s_endpgm +; +; GFX9-LABEL: v_ssubo_i32: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX9-NEXT: v_mov_b32_e32 v0, 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: global_load_dword v1, v0, s[12:13] +; GFX9-NEXT: global_load_dword v2, v0, s[14:15] +; GFX9-NEXT: s_waitcnt vmcnt(0) +; GFX9-NEXT: v_sub_i32 v3, v1, v2 clamp +; GFX9-NEXT: v_sub_u32_e32 v1, v1, v2 +; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, v1, v3 +; GFX9-NEXT: global_store_dword v0, v1, s[8:9] +; GFX9-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc +; GFX9-NEXT: global_store_byte v0, v1, s[10:11] +; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: v_ssubo_i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v0, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: global_load_dword v1, v0, s[12:13] +; GFX10-NEXT: global_load_dword v2, v0, s[14:15] +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: v_sub_nc_i32 v3, v1, v2 clamp +; GFX10-NEXT: v_sub_nc_u32_e32 v1, v1, v2 +; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, v1, v3 +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo +; GFX10-NEXT: global_store_dword v0, v1, s[8:9] +; GFX10-NEXT: global_store_byte v0, v2, s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: v_ssubo_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[4:5] +; GFX11-NEXT: global_load_b32 v2, v0, s[6:7] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_sub_nc_i32 v3, v1, v2 clamp +; GFX11-NEXT: v_sub_nc_u32_e32 v1, v1, v2 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_cmp_ne_u32_e32 vcc_lo, v1, v3 +; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: global_store_b8 v0, v2, s[2:3] +; GFX11-NEXT: s_endpgm %a = load i32, ptr addrspace(1) %aptr, align 4 %b = load i32, ptr addrspace(1) %bptr, align 4 %ssub = call { i32, i1 } @llvm.ssub.with.overflow.i32(i32 %a, i32 %b) nounwind @@ -40,10 +340,109 @@ define amdgpu_kernel void @v_ssubo_i32(ptr addrspace(1) %out, ptr addrspace(1) % ret void } -; FUNC-LABEL: {{^}}s_ssubo_i64: -; GCN: s_sub_u32 -; GCN: s_subb_u32 define amdgpu_kernel void @s_ssubo_i64(ptr addrspace(1) %out, ptr addrspace(1) %carryout, i64 %a, i64 %b) nounwind { +; SI-LABEL: s_ssubo_i64: +; SI: ; %bb.0: +; SI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x9 +; SI-NEXT: s_mov_b32 s11, 0xf000 +; SI-NEXT: s_mov_b32 s10, -1 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_sub_u32 s12, s4, s6 +; SI-NEXT: v_mov_b32_e32 v0, s4 +; SI-NEXT: s_subb_u32 s13, s5, s7 +; SI-NEXT: v_mov_b32_e32 v1, s5 +; SI-NEXT: v_cmp_lt_i64_e32 vcc, s[12:13], v[0:1] +; SI-NEXT: v_cmp_gt_i64_e64 s[4:5], s[6:7], 0 +; SI-NEXT: v_mov_b32_e32 v0, s12 +; SI-NEXT: s_mov_b32 s8, s0 +; SI-NEXT: s_mov_b32 s9, s1 +; SI-NEXT: v_mov_b32_e32 v1, s13 +; SI-NEXT: s_xor_b64 s[4:5], s[4:5], vcc +; SI-NEXT: s_mov_b32 s0, s2 +; SI-NEXT: s_mov_b32 s1, s3 +; SI-NEXT: s_mov_b32 s2, s10 +; SI-NEXT: s_mov_b32 s3, s11 +; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[8:11], 0 +; SI-NEXT: s_waitcnt expcnt(0) +; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[4:5] +; SI-NEXT: buffer_store_byte v0, off, s[0:3], 0 +; SI-NEXT: s_endpgm +; +; VI-LABEL: s_ssubo_i64: +; VI: ; %bb.0: +; VI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x24 +; VI-NEXT: s_waitcnt lgkmcnt(0) +; VI-NEXT: v_mov_b32_e32 v0, s0 +; VI-NEXT: s_sub_u32 s0, s4, s6 +; VI-NEXT: v_mov_b32_e32 v4, s4 +; VI-NEXT: v_mov_b32_e32 v1, s1 +; VI-NEXT: s_subb_u32 s1, s5, s7 +; VI-NEXT: v_mov_b32_e32 v5, s5 +; VI-NEXT: v_mov_b32_e32 v2, s2 +; VI-NEXT: v_mov_b32_e32 v3, s3 +; VI-NEXT: v_cmp_lt_i64_e32 vcc, s[0:1], v[4:5] +; VI-NEXT: v_cmp_gt_i64_e64 s[2:3], s[6:7], 0 +; VI-NEXT: v_mov_b32_e32 v5, s1 +; VI-NEXT: v_mov_b32_e32 v4, s0 +; VI-NEXT: s_xor_b64 s[0:1], s[2:3], vcc +; VI-NEXT: flat_store_dwordx2 v[0:1], v[4:5] +; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; VI-NEXT: flat_store_byte v[2:3], v0 +; VI-NEXT: s_endpgm +; +; GFX9-LABEL: s_ssubo_i64: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX9-NEXT: v_mov_b32_e32 v2, 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: s_sub_u32 s0, s12, s14 +; GFX9-NEXT: v_mov_b32_e32 v0, s12 +; GFX9-NEXT: v_mov_b32_e32 v1, s13 +; GFX9-NEXT: s_subb_u32 s1, s13, s15 +; GFX9-NEXT: v_cmp_gt_i64_e64 s[2:3], s[14:15], 0 +; GFX9-NEXT: v_cmp_lt_i64_e32 vcc, s[0:1], v[0:1] +; GFX9-NEXT: v_mov_b32_e32 v0, s0 +; GFX9-NEXT: v_mov_b32_e32 v1, s1 +; GFX9-NEXT: s_xor_b64 s[0:1], s[2:3], vcc +; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[8:9] +; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; GFX9-NEXT: global_store_byte v2, v0, s[10:11] +; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: s_ssubo_i64: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v2, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_sub_u32 s0, s12, s14 +; GFX10-NEXT: s_subb_u32 s1, s13, s15 +; GFX10-NEXT: v_cmp_gt_i64_e64 s2, s[14:15], 0 +; GFX10-NEXT: v_cmp_lt_i64_e64 s3, s[0:1], s[12:13] +; GFX10-NEXT: v_mov_b32_e32 v0, s0 +; GFX10-NEXT: v_mov_b32_e32 v1, s1 +; GFX10-NEXT: s_xor_b32 s0, s2, s3 +; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, 1, s0 +; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[8:9] +; GFX10-NEXT: global_store_byte v2, v3, s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: s_ssubo_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_sub_u32 s8, s4, s6 +; GFX11-NEXT: s_subb_u32 s9, s5, s7 +; GFX11-NEXT: v_cmp_gt_i64_e64 s6, s[6:7], 0 +; GFX11-NEXT: v_cmp_lt_i64_e64 s4, s[8:9], s[4:5] +; GFX11-NEXT: v_mov_b32_e32 v0, s8 +; GFX11-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s9 +; GFX11-NEXT: s_xor_b32 s4, s6, s4 +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) +; GFX11-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1] +; GFX11-NEXT: global_store_b8 v2, v3, s[2:3] +; GFX11-NEXT: s_endpgm %ssub = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 %a, i64 %b) nounwind %val = extractvalue { i64, i1 } %ssub, 0 %carry = extractvalue { i64, i1 } %ssub, 1 @@ -52,16 +451,121 @@ define amdgpu_kernel void @s_ssubo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ret void } -; FUNC-LABEL: {{^}}v_ssubo_i64: -; SI: v_sub_i32_e32 v{{[0-9]+}}, vcc, -; SI: v_subb_u32_e32 v{{[0-9]+}}, vcc, - -; VI: v_sub_u32_e32 v{{[0-9]+}}, vcc, -; VI: v_subb_u32_e32 v{{[0-9]+}}, vcc, - -; GFX9: v_sub_co_u32_e32 v{{[0-9]+}}, vcc, -; GFX9: v_subb_co_u32_e32 v{{[0-9]+}}, vcc, define amdgpu_kernel void @v_ssubo_i64(ptr addrspace(1) %out, ptr addrspace(1) %carryout, ptr addrspace(1) %aptr, ptr addrspace(1) %bptr) nounwind { +; SI-LABEL: v_ssubo_i64: +; SI: ; %bb.0: +; SI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x9 +; SI-NEXT: s_mov_b32 s11, 0xf000 +; SI-NEXT: s_mov_b32 s10, -1 +; SI-NEXT: s_mov_b32 s14, s10 +; SI-NEXT: s_mov_b32 s15, s11 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s12, s4 +; SI-NEXT: s_mov_b32 s13, s5 +; SI-NEXT: s_mov_b32 s4, s6 +; SI-NEXT: s_mov_b32 s5, s7 +; SI-NEXT: s_mov_b32 s6, s10 +; SI-NEXT: s_mov_b32 s7, s11 +; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[12:15], 0 +; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[4:7], 0 +; SI-NEXT: s_mov_b32 s8, s0 +; SI-NEXT: s_mov_b32 s9, s1 +; SI-NEXT: s_mov_b32 s4, s2 +; SI-NEXT: s_mov_b32 s5, s3 +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: v_sub_i32_e32 v4, vcc, v0, v2 +; SI-NEXT: v_subb_u32_e32 v5, vcc, v1, v3, vcc +; SI-NEXT: v_cmp_lt_i64_e32 vcc, 0, v[2:3] +; SI-NEXT: v_cmp_lt_i64_e64 s[0:1], v[4:5], v[0:1] +; SI-NEXT: buffer_store_dwordx2 v[4:5], off, s[8:11], 0 +; SI-NEXT: s_xor_b64 s[0:1], vcc, s[0:1] +; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; SI-NEXT: buffer_store_byte v0, off, s[4:7], 0 +; SI-NEXT: s_endpgm +; +; VI-LABEL: v_ssubo_i64: +; VI: ; %bb.0: +; VI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x24 +; VI-NEXT: s_waitcnt lgkmcnt(0) +; VI-NEXT: v_mov_b32_e32 v0, s4 +; VI-NEXT: v_mov_b32_e32 v1, s5 +; VI-NEXT: v_mov_b32_e32 v2, s6 +; VI-NEXT: v_mov_b32_e32 v3, s7 +; VI-NEXT: flat_load_dwordx2 v[0:1], v[0:1] +; VI-NEXT: flat_load_dwordx2 v[2:3], v[2:3] +; VI-NEXT: v_mov_b32_e32 v4, s0 +; VI-NEXT: v_mov_b32_e32 v5, s1 +; VI-NEXT: v_mov_b32_e32 v6, s2 +; VI-NEXT: v_mov_b32_e32 v7, s3 +; VI-NEXT: s_waitcnt vmcnt(0) +; VI-NEXT: v_sub_u32_e32 v8, vcc, v0, v2 +; VI-NEXT: v_subb_u32_e32 v9, vcc, v1, v3, vcc +; VI-NEXT: v_cmp_lt_i64_e32 vcc, 0, v[2:3] +; VI-NEXT: v_cmp_lt_i64_e64 s[0:1], v[8:9], v[0:1] +; VI-NEXT: flat_store_dwordx2 v[4:5], v[8:9] +; VI-NEXT: s_xor_b64 s[0:1], vcc, s[0:1] +; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; VI-NEXT: flat_store_byte v[6:7], v0 +; VI-NEXT: s_endpgm +; +; GFX9-LABEL: v_ssubo_i64: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX9-NEXT: v_mov_b32_e32 v6, 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: global_load_dwordx2 v[0:1], v6, s[12:13] +; GFX9-NEXT: global_load_dwordx2 v[2:3], v6, s[14:15] +; GFX9-NEXT: s_waitcnt vmcnt(0) +; GFX9-NEXT: v_sub_co_u32_e32 v4, vcc, v0, v2 +; GFX9-NEXT: v_subb_co_u32_e32 v5, vcc, v1, v3, vcc +; GFX9-NEXT: v_cmp_lt_i64_e32 vcc, 0, v[2:3] +; GFX9-NEXT: v_cmp_lt_i64_e64 s[0:1], v[4:5], v[0:1] +; GFX9-NEXT: global_store_dwordx2 v6, v[4:5], s[8:9] +; GFX9-NEXT: s_xor_b64 s[0:1], vcc, s[0:1] +; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; GFX9-NEXT: global_store_byte v6, v0, s[10:11] +; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: v_ssubo_i64: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v6, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: global_load_dwordx2 v[0:1], v6, s[12:13] +; GFX10-NEXT: global_load_dwordx2 v[2:3], v6, s[14:15] +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: v_sub_co_u32 v4, vcc_lo, v0, v2 +; GFX10-NEXT: v_sub_co_ci_u32_e32 v5, vcc_lo, v1, v3, vcc_lo +; GFX10-NEXT: v_cmp_lt_i64_e32 vcc_lo, 0, v[2:3] +; GFX10-NEXT: v_cmp_lt_i64_e64 s0, v[4:5], v[0:1] +; GFX10-NEXT: s_xor_b32 s0, vcc_lo, s0 +; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, s0 +; GFX10-NEXT: global_store_dwordx2 v6, v[4:5], s[8:9] +; GFX10-NEXT: global_store_byte v6, v0, s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: v_ssubo_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[4:11], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v6, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b64 v[0:1], v6, s[8:9] +; GFX11-NEXT: global_load_b64 v[2:3], v6, s[10:11] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_sub_co_u32 v4, vcc_lo, v0, v2 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2) +; GFX11-NEXT: v_sub_co_ci_u32_e64 v5, null, v1, v3, vcc_lo +; GFX11-NEXT: v_cmp_lt_i64_e32 vcc_lo, 0, v[2:3] +; GFX11-NEXT: v_cmp_lt_i64_e64 s0, v[4:5], v[0:1] +; GFX11-NEXT: s_xor_b32 s0, vcc_lo, s0 +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) +; GFX11-NEXT: v_cndmask_b32_e64 v0, 0, 1, s0 +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b64 v6, v[4:5], s[4:5] +; GFX11-NEXT: global_store_b8 v6, v0, s[6:7] +; GFX11-NEXT: s_endpgm %a = load i64, ptr addrspace(1) %aptr, align 4 %b = load i64, ptr addrspace(1) %bptr, align 4 %ssub = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 %a, i64 %b) nounwind @@ -72,14 +576,134 @@ define amdgpu_kernel void @v_ssubo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ret void } -; FUNC-LABEL: {{^}}v_ssubo_v2i32: -; SICIVI: v_cmp_lt_i32 -; SICIVI: v_cmp_lt_i32 -; SICIVI: v_sub_{{[iu]}}32 -; SICIVI: v_cmp_lt_i32 -; SICIVI: v_cmp_lt_i32 -; SICIVI: v_sub_{{[iu]}}32 define amdgpu_kernel void @v_ssubo_v2i32(ptr addrspace(1) %out, ptr addrspace(1) %carryout, ptr addrspace(1) %aptr, ptr addrspace(1) %bptr) nounwind { +; SI-LABEL: v_ssubo_v2i32: +; SI: ; %bb.0: +; SI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x9 +; SI-NEXT: s_mov_b32 s11, 0xf000 +; SI-NEXT: s_mov_b32 s10, -1 +; SI-NEXT: s_mov_b32 s14, s10 +; SI-NEXT: s_mov_b32 s15, s11 +; SI-NEXT: s_waitcnt lgkmcnt(0) +; SI-NEXT: s_mov_b32 s12, s4 +; SI-NEXT: s_mov_b32 s13, s5 +; SI-NEXT: s_mov_b32 s4, s6 +; SI-NEXT: s_mov_b32 s5, s7 +; SI-NEXT: s_mov_b32 s6, s10 +; SI-NEXT: s_mov_b32 s7, s11 +; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[12:15], 0 +; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[4:7], 0 +; SI-NEXT: s_mov_b32 s8, s0 +; SI-NEXT: s_mov_b32 s9, s1 +; SI-NEXT: s_mov_b32 s12, s2 +; SI-NEXT: s_mov_b32 s13, s3 +; SI-NEXT: s_waitcnt vmcnt(0) +; SI-NEXT: v_sub_i32_e32 v5, vcc, v1, v3 +; SI-NEXT: v_sub_i32_e32 v4, vcc, v0, v2 +; SI-NEXT: v_cmp_lt_i32_e64 s[0:1], 0, v3 +; SI-NEXT: v_cmp_lt_i32_e64 s[4:5], v5, v1 +; SI-NEXT: v_cmp_lt_i32_e32 vcc, 0, v2 +; SI-NEXT: v_cmp_lt_i32_e64 s[2:3], v4, v0 +; SI-NEXT: s_xor_b64 s[0:1], s[0:1], s[4:5] +; SI-NEXT: v_cndmask_b32_e64 v1, 0, 1, s[0:1] +; SI-NEXT: s_xor_b64 s[0:1], vcc, s[2:3] +; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; SI-NEXT: buffer_store_dwordx2 v[4:5], off, s[8:11], 0 +; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[12:15], 0 +; SI-NEXT: s_endpgm +; +; VI-LABEL: v_ssubo_v2i32: +; VI: ; %bb.0: +; VI-NEXT: s_load_dwordx8 s[0:7], s[4:5], 0x24 +; VI-NEXT: s_waitcnt lgkmcnt(0) +; VI-NEXT: v_mov_b32_e32 v0, s4 +; VI-NEXT: v_mov_b32_e32 v1, s5 +; VI-NEXT: v_mov_b32_e32 v2, s6 +; VI-NEXT: v_mov_b32_e32 v3, s7 +; VI-NEXT: flat_load_dwordx2 v[0:1], v[0:1] +; VI-NEXT: flat_load_dwordx2 v[2:3], v[2:3] +; VI-NEXT: v_mov_b32_e32 v4, s0 +; VI-NEXT: v_mov_b32_e32 v5, s1 +; VI-NEXT: v_mov_b32_e32 v6, s2 +; VI-NEXT: v_mov_b32_e32 v7, s3 +; VI-NEXT: s_waitcnt vmcnt(0) +; VI-NEXT: v_sub_u32_e32 v9, vcc, v1, v3 +; VI-NEXT: v_sub_u32_e32 v8, vcc, v0, v2 +; VI-NEXT: v_cmp_lt_i32_e64 s[0:1], 0, v3 +; VI-NEXT: v_cmp_lt_i32_e64 s[4:5], v9, v1 +; VI-NEXT: v_cmp_lt_i32_e32 vcc, 0, v2 +; VI-NEXT: v_cmp_lt_i32_e64 s[2:3], v8, v0 +; VI-NEXT: s_xor_b64 s[0:1], s[0:1], s[4:5] +; VI-NEXT: v_cndmask_b32_e64 v1, 0, 1, s[0:1] +; VI-NEXT: s_xor_b64 s[0:1], vcc, s[2:3] +; VI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; VI-NEXT: flat_store_dwordx2 v[4:5], v[8:9] +; VI-NEXT: flat_store_dwordx2 v[6:7], v[0:1] +; VI-NEXT: s_endpgm +; +; GFX9-LABEL: v_ssubo_v2i32: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX9-NEXT: v_mov_b32_e32 v6, 0 +; GFX9-NEXT: s_waitcnt lgkmcnt(0) +; GFX9-NEXT: global_load_dwordx2 v[0:1], v6, s[12:13] +; GFX9-NEXT: global_load_dwordx2 v[2:3], v6, s[14:15] +; GFX9-NEXT: s_waitcnt vmcnt(0) +; GFX9-NEXT: v_sub_u32_e32 v5, v1, v3 +; GFX9-NEXT: v_sub_i32 v1, v1, v3 clamp +; GFX9-NEXT: v_sub_u32_e32 v4, v0, v2 +; GFX9-NEXT: v_sub_i32 v0, v0, v2 clamp +; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, v5, v1 +; GFX9-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc +; GFX9-NEXT: v_cmp_ne_u32_e32 vcc, v4, v0 +; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc +; GFX9-NEXT: global_store_dwordx2 v6, v[4:5], s[8:9] +; GFX9-NEXT: global_store_dwordx2 v6, v[0:1], s[10:11] +; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: v_ssubo_v2i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v5, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: global_load_dwordx2 v[0:1], v5, s[12:13] +; GFX10-NEXT: global_load_dwordx2 v[2:3], v5, s[14:15] +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: v_sub_nc_u32_e32 v4, v1, v3 +; GFX10-NEXT: v_sub_nc_i32 v1, v1, v3 clamp +; GFX10-NEXT: v_sub_nc_u32_e32 v3, v0, v2 +; GFX10-NEXT: v_sub_nc_i32 v0, v0, v2 clamp +; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, v4, v1 +; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo +; GFX10-NEXT: v_cmp_ne_u32_e32 vcc_lo, v3, v0 +; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo +; GFX10-NEXT: global_store_dwordx2 v5, v[3:4], s[8:9] +; GFX10-NEXT: global_store_dwordx2 v5, v[0:1], s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: v_ssubo_v2i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v5, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b64 v[0:1], v5, s[4:5] +; GFX11-NEXT: global_load_b64 v[2:3], v5, s[6:7] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_sub_nc_u32_e32 v4, v1, v3 +; GFX11-NEXT: v_sub_nc_i32 v1, v1, v3 clamp +; GFX11-NEXT: v_sub_nc_u32_e32 v3, v0, v2 +; GFX11-NEXT: v_sub_nc_i32 v0, v0, v2 clamp +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_3) +; GFX11-NEXT: v_cmp_ne_u32_e32 vcc_lo, v4, v1 +; GFX11-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo +; GFX11-NEXT: v_cmp_ne_u32_e32 vcc_lo, v3, v0 +; GFX11-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b64 v5, v[3:4], s[0:1] +; GFX11-NEXT: global_store_b64 v5, v[0:1], s[2:3] +; GFX11-NEXT: s_endpgm %a = load <2 x i32>, ptr addrspace(1) %aptr, align 4 %b = load <2 x i32>, ptr addrspace(1) %bptr, align 4 %sadd = call { <2 x i32>, <2 x i1> } @llvm.ssub.with.overflow.v2i32(<2 x i32> %a, <2 x i32> %b) nounwind diff --git a/llvm/test/CodeGen/AMDGPU/uaddo.ll b/llvm/test/CodeGen/AMDGPU/uaddo.ll index d230ff5..e1574dc 100644 --- a/llvm/test/CodeGen/AMDGPU/uaddo.ll +++ b/llvm/test/CodeGen/AMDGPU/uaddo.ll @@ -1,7 +1,9 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn < %s | FileCheck -check-prefixes=SI %s -; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tonga < %s | FileCheck -check-prefixes=VI %s -; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx900 < %s | FileCheck -check-prefixes=GFX9 %s +; RUN: llc < %s -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tahiti | FileCheck %s --check-prefix=SI +; RUN: llc < %s -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tonga | FileCheck %s --check-prefix=VI +; RUN: llc < %s -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx900 | FileCheck %s --check-prefix=GFX9 +; RUN: llc < %s -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1010 | FileCheck %s --check-prefix=GFX10 +; RUN: llc < %s -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 | FileCheck %s --check-prefix=GFX11 define amdgpu_kernel void @s_uaddo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 %b) #0 { ; SI-LABEL: s_uaddo_i64_zext: @@ -12,14 +14,14 @@ define amdgpu_kernel void @s_uaddo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 % ; SI-NEXT: s_mov_b32 s6, -1 ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_mov_b32 s4, s0 -; SI-NEXT: s_mov_b32 s5, s1 ; SI-NEXT: s_add_u32 s0, s2, s8 ; SI-NEXT: v_mov_b32_e32 v0, s2 -; SI-NEXT: v_mov_b32_e32 v1, s3 +; SI-NEXT: s_mov_b32 s5, s1 ; SI-NEXT: s_addc_u32 s1, s3, s9 +; SI-NEXT: v_mov_b32_e32 v1, s3 ; SI-NEXT: v_cmp_lt_u64_e32 vcc, s[0:1], v[0:1] -; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc ; SI-NEXT: v_mov_b32_e32 v1, s1 +; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc ; SI-NEXT: v_add_i32_e32 v0, vcc, s0, v0 ; SI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc ; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0 @@ -61,6 +63,40 @@ define amdgpu_kernel void @s_uaddo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 % ; GFX9-NEXT: v_addc_co_u32_e32 v1, vcc, 0, v1, vcc ; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: s_uaddo_i64_zext: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 +; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34 +; GFX10-NEXT: v_mov_b32_e32 v2, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_add_u32 s4, s2, s6 +; GFX10-NEXT: s_addc_u32 s5, s3, s7 +; GFX10-NEXT: v_cmp_lt_u64_e64 s2, s[4:5], s[2:3] +; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, s2 +; GFX10-NEXT: v_add_co_u32 v0, s2, s4, v0 +; GFX10-NEXT: v_add_co_ci_u32_e64 v1, s2, s5, 0, s2 +; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: s_uaddo_i64_zext: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 +; GFX11-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 +; GFX11-NEXT: v_mov_b32_e32 v2, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_add_u32 s4, s2, s4 +; GFX11-NEXT: s_addc_u32 s5, s3, s5 +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_cmp_lt_u64_e64 s2, s[4:5], s[2:3] +; GFX11-NEXT: v_cndmask_b32_e64 v0, 0, 1, s2 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_add_co_u32 v0, s2, s4, v0 +; GFX11-NEXT: v_add_co_ci_u32_e64 v1, null, s5, 0, s2 +; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1] +; GFX11-NEXT: s_endpgm %uadd = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %a, i64 %b) %val = extractvalue { i64, i1 } %uadd, 0 %carry = extractvalue { i64, i1 } %uadd, 1 @@ -76,21 +112,21 @@ define amdgpu_kernel void @s_uaddo_i32(ptr addrspace(1) %out, ptr addrspace(1) % ; SI-LABEL: s_uaddo_i32: ; SI: ; %bb.0: ; SI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 -; SI-NEXT: s_load_dwordx2 s[12:13], s[4:5], 0xd +; SI-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd ; SI-NEXT: s_mov_b32 s7, 0xf000 ; SI-NEXT: s_mov_b32 s6, -1 -; SI-NEXT: s_mov_b32 s10, s6 -; SI-NEXT: s_mov_b32 s11, s7 ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_mov_b32 s4, s0 +; SI-NEXT: v_mov_b32_e32 v0, s9 ; SI-NEXT: s_mov_b32 s5, s1 -; SI-NEXT: s_mov_b32 s8, s2 -; SI-NEXT: s_mov_b32 s9, s3 -; SI-NEXT: v_mov_b32_e32 v0, s13 -; SI-NEXT: v_add_i32_e32 v0, vcc, s12, v0 +; SI-NEXT: v_add_i32_e32 v0, vcc, s8, v0 +; SI-NEXT: s_mov_b32 s0, s2 +; SI-NEXT: s_mov_b32 s1, s3 +; SI-NEXT: s_mov_b32 s2, s6 +; SI-NEXT: s_mov_b32 s3, s7 ; SI-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc ; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0 -; SI-NEXT: buffer_store_byte v1, off, s[8:11], 0 +; SI-NEXT: buffer_store_byte v1, off, s[0:3], 0 ; SI-NEXT: s_endpgm ; ; VI-LABEL: s_uaddo_i32: @@ -121,6 +157,34 @@ define amdgpu_kernel void @s_uaddo_i32(ptr addrspace(1) %out, ptr addrspace(1) % ; GFX9-NEXT: global_store_dword v0, v1, s[0:1] ; GFX9-NEXT: global_store_byte v0, v2, s[2:3] ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: s_uaddo_i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34 +; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v0, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: v_add_co_u32 v1, s4, s6, s7 +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, s4 +; GFX10-NEXT: global_store_dword v0, v1, s[0:1] +; GFX10-NEXT: global_store_byte v0, v2, s[2:3] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: s_uaddo_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: s_load_b64 s[6:7], s[4:5], 0x34 +; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: v_add_co_u32 v1, s4, s6, s7 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, s4 +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: global_store_b8 v0, v2, s[2:3] +; GFX11-NEXT: s_endpgm %uadd = call { i32, i1 } @llvm.uadd.with.overflow.i32(i32 %a, i32 %b) %val = extractvalue { i32, i1 } %uadd, 0 %carry = extractvalue { i32, i1 } %uadd, 1 @@ -137,17 +201,15 @@ define amdgpu_kernel void @v_uaddo_i32(ptr addrspace(1) %out, ptr addrspace(1) % ; SI-NEXT: s_mov_b32 s10, -1 ; SI-NEXT: s_mov_b32 s14, s10 ; SI-NEXT: s_mov_b32 s15, s11 -; SI-NEXT: s_mov_b32 s18, s10 -; SI-NEXT: s_mov_b32 s19, s11 ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_mov_b32 s12, s4 ; SI-NEXT: s_mov_b32 s13, s5 -; SI-NEXT: s_mov_b32 s16, s6 -; SI-NEXT: s_mov_b32 s17, s7 -; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 -; SI-NEXT: buffer_load_dword v1, off, s[16:19], 0 +; SI-NEXT: s_mov_b32 s4, s6 +; SI-NEXT: s_mov_b32 s5, s7 ; SI-NEXT: s_mov_b32 s6, s10 ; SI-NEXT: s_mov_b32 s7, s11 +; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 +; SI-NEXT: buffer_load_dword v1, off, s[4:7], 0 ; SI-NEXT: s_mov_b32 s8, s0 ; SI-NEXT: s_mov_b32 s9, s1 ; SI-NEXT: s_mov_b32 s4, s2 @@ -193,6 +255,38 @@ define amdgpu_kernel void @v_uaddo_i32(ptr addrspace(1) %out, ptr addrspace(1) % ; GFX9-NEXT: global_store_dword v0, v1, s[8:9] ; GFX9-NEXT: global_store_byte v0, v2, s[10:11] ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: v_uaddo_i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v0, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: global_load_dword v1, v0, s[12:13] +; GFX10-NEXT: global_load_dword v2, v0, s[14:15] +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: v_add_co_u32 v1, s0, v1, v2 +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, s0 +; GFX10-NEXT: global_store_dword v0, v1, s[8:9] +; GFX10-NEXT: global_store_byte v0, v2, s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: v_uaddo_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[4:5] +; GFX11-NEXT: global_load_b32 v2, v0, s[6:7] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_add_co_u32 v1, s4, v1, v2 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, s4 +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: global_store_b8 v0, v2, s[2:3] +; GFX11-NEXT: s_endpgm %tid = call i32 @llvm.amdgcn.workitem.id.x() %tid.ext = sext i32 %tid to i64 %a.gep = getelementptr inbounds i32, ptr addrspace(1) %a.ptr @@ -215,17 +309,15 @@ define amdgpu_kernel void @v_uaddo_i32_novcc(ptr addrspace(1) %out, ptr addrspac ; SI-NEXT: s_mov_b32 s10, -1 ; SI-NEXT: s_mov_b32 s14, s10 ; SI-NEXT: s_mov_b32 s15, s11 -; SI-NEXT: s_mov_b32 s18, s10 -; SI-NEXT: s_mov_b32 s19, s11 ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_mov_b32 s12, s4 ; SI-NEXT: s_mov_b32 s13, s5 -; SI-NEXT: s_mov_b32 s16, s6 -; SI-NEXT: s_mov_b32 s17, s7 -; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 -; SI-NEXT: buffer_load_dword v1, off, s[16:19], 0 +; SI-NEXT: s_mov_b32 s4, s6 +; SI-NEXT: s_mov_b32 s5, s7 ; SI-NEXT: s_mov_b32 s6, s10 ; SI-NEXT: s_mov_b32 s7, s11 +; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 +; SI-NEXT: buffer_load_dword v1, off, s[4:7], 0 ; SI-NEXT: s_mov_b32 s8, s0 ; SI-NEXT: s_mov_b32 s9, s1 ; SI-NEXT: s_mov_b32 s4, s2 @@ -283,6 +375,45 @@ define amdgpu_kernel void @v_uaddo_i32_novcc(ptr addrspace(1) %out, ptr addrspac ; GFX9-NEXT: global_store_byte v0, v2, s[10:11] ; GFX9-NEXT: s_waitcnt vmcnt(0) ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: v_uaddo_i32_novcc: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v0, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: global_load_dword v1, v0, s[12:13] +; GFX10-NEXT: global_load_dword v2, v0, s[14:15] +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: v_add_co_u32 v1, s0, v1, v2 +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, s0 +; GFX10-NEXT: global_store_dword v0, v1, s[8:9] +; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX10-NEXT: ;;#ASMSTART +; GFX10-NEXT: ;;#ASMEND +; GFX10-NEXT: global_store_byte v0, v2, s[10:11] +; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: v_uaddo_i32_novcc: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[4:5] +; GFX11-NEXT: global_load_b32 v2, v0, s[6:7] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_add_co_u32 v1, s4, v1, v2 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, s4 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] dlc +; GFX11-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX11-NEXT: ;;#ASMSTART +; GFX11-NEXT: ;;#ASMEND +; GFX11-NEXT: global_store_b8 v0, v2, s[2:3] dlc +; GFX11-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX11-NEXT: s_endpgm %tid = call i32 @llvm.amdgcn.workitem.id.x() %tid.ext = sext i32 %tid to i64 %a.gep = getelementptr inbounds i32, ptr addrspace(1) %a.ptr @@ -306,21 +437,21 @@ define amdgpu_kernel void @s_uaddo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; SI-NEXT: s_mov_b32 s10, -1 ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_add_u32 s6, s4, s6 -; SI-NEXT: s_addc_u32 s7, s5, s7 -; SI-NEXT: s_mov_b32 s14, s10 -; SI-NEXT: s_mov_b32 s15, s11 -; SI-NEXT: s_mov_b32 s8, s0 -; SI-NEXT: s_mov_b32 s9, s1 -; SI-NEXT: s_mov_b32 s12, s2 -; SI-NEXT: s_mov_b32 s13, s3 ; SI-NEXT: v_mov_b32_e32 v0, s4 +; SI-NEXT: s_addc_u32 s7, s5, s7 ; SI-NEXT: v_mov_b32_e32 v1, s5 +; SI-NEXT: v_cmp_lt_u64_e32 vcc, s[6:7], v[0:1] ; SI-NEXT: v_mov_b32_e32 v2, s6 +; SI-NEXT: s_mov_b32 s8, s0 +; SI-NEXT: s_mov_b32 s9, s1 +; SI-NEXT: s_mov_b32 s0, s2 +; SI-NEXT: s_mov_b32 s1, s3 +; SI-NEXT: s_mov_b32 s2, s10 +; SI-NEXT: s_mov_b32 s3, s11 ; SI-NEXT: v_mov_b32_e32 v3, s7 -; SI-NEXT: buffer_store_dwordx2 v[2:3], off, s[8:11], 0 -; SI-NEXT: v_cmp_lt_u64_e32 vcc, s[6:7], v[0:1] ; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc -; SI-NEXT: buffer_store_byte v0, off, s[12:15], 0 +; SI-NEXT: buffer_store_dwordx2 v[2:3], off, s[8:11], 0 +; SI-NEXT: buffer_store_byte v0, off, s[0:3], 0 ; SI-NEXT: s_endpgm ; ; VI-LABEL: s_uaddo_i64: @@ -359,6 +490,37 @@ define amdgpu_kernel void @s_uaddo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; GFX9-NEXT: global_store_dwordx2 v4, v[2:3], s[8:9] ; GFX9-NEXT: global_store_byte v4, v0, s[10:11] ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: s_uaddo_i64: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v2, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_add_u32 s0, s12, s14 +; GFX10-NEXT: s_addc_u32 s1, s13, s15 +; GFX10-NEXT: v_mov_b32_e32 v0, s0 +; GFX10-NEXT: v_mov_b32_e32 v1, s1 +; GFX10-NEXT: v_cmp_lt_u64_e64 s0, s[0:1], s[12:13] +; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, 1, s0 +; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[8:9] +; GFX10-NEXT: global_store_byte v2, v3, s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: s_uaddo_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_add_u32 s6, s4, s6 +; GFX11-NEXT: s_addc_u32 s7, s5, s7 +; GFX11-NEXT: v_mov_b32_e32 v0, s6 +; GFX11-NEXT: v_cmp_lt_u64_e64 s4, s[6:7], s[4:5] +; GFX11-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) +; GFX11-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1] +; GFX11-NEXT: global_store_b8 v2, v3, s[2:3] +; GFX11-NEXT: s_endpgm %uadd = call { i64, i1 } @llvm.uadd.with.overflow.i64(i64 %a, i64 %b) %val = extractvalue { i64, i1 } %uadd, 0 %carry = extractvalue { i64, i1 } %uadd, 1 @@ -375,17 +537,15 @@ define amdgpu_kernel void @v_uaddo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; SI-NEXT: s_mov_b32 s10, -1 ; SI-NEXT: s_mov_b32 s14, s10 ; SI-NEXT: s_mov_b32 s15, s11 -; SI-NEXT: s_mov_b32 s18, s10 -; SI-NEXT: s_mov_b32 s19, s11 ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_mov_b32 s12, s4 ; SI-NEXT: s_mov_b32 s13, s5 -; SI-NEXT: s_mov_b32 s16, s6 -; SI-NEXT: s_mov_b32 s17, s7 -; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[12:15], 0 -; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[16:19], 0 +; SI-NEXT: s_mov_b32 s4, s6 +; SI-NEXT: s_mov_b32 s5, s7 ; SI-NEXT: s_mov_b32 s6, s10 ; SI-NEXT: s_mov_b32 s7, s11 +; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[12:15], 0 +; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[4:7], 0 ; SI-NEXT: s_mov_b32 s8, s0 ; SI-NEXT: s_mov_b32 s9, s1 ; SI-NEXT: s_mov_b32 s4, s2 @@ -393,8 +553,8 @@ define amdgpu_kernel void @v_uaddo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; SI-NEXT: s_waitcnt vmcnt(0) ; SI-NEXT: v_add_i32_e32 v2, vcc, v0, v2 ; SI-NEXT: v_addc_u32_e32 v3, vcc, v1, v3, vcc -; SI-NEXT: buffer_store_dwordx2 v[2:3], off, s[8:11], 0 ; SI-NEXT: v_cmp_lt_u64_e32 vcc, v[2:3], v[0:1] +; SI-NEXT: buffer_store_dwordx2 v[2:3], off, s[8:11], 0 ; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc ; SI-NEXT: buffer_store_byte v0, off, s[4:7], 0 ; SI-NEXT: s_endpgm @@ -437,6 +597,42 @@ define amdgpu_kernel void @v_uaddo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc ; GFX9-NEXT: global_store_byte v4, v0, s[10:11] ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: v_uaddo_i64: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v4, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: global_load_dwordx2 v[0:1], v4, s[12:13] +; GFX10-NEXT: global_load_dwordx2 v[2:3], v4, s[14:15] +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: v_add_co_u32 v2, vcc_lo, v0, v2 +; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, v1, v3, vcc_lo +; GFX10-NEXT: v_cmp_lt_u64_e32 vcc_lo, v[2:3], v[0:1] +; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo +; GFX10-NEXT: global_store_dwordx2 v4, v[2:3], s[8:9] +; GFX10-NEXT: global_store_byte v4, v0, s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: v_uaddo_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v4, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b64 v[0:1], v4, s[4:5] +; GFX11-NEXT: global_load_b64 v[2:3], v4, s[6:7] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_add_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_add_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_cndmask_b32_e64 v0, 0, 1, vcc_lo +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b64 v4, v[2:3], s[0:1] +; GFX11-NEXT: global_store_b8 v4, v0, s[2:3] +; GFX11-NEXT: s_endpgm %tid = call i32 @llvm.amdgcn.workitem.id.x() %tid.ext = sext i32 %tid to i64 %a.gep = getelementptr inbounds i64, ptr addrspace(1) %a.ptr @@ -459,17 +655,15 @@ define amdgpu_kernel void @v_uaddo_i16(ptr addrspace(1) %out, ptr addrspace(1) % ; SI-NEXT: s_mov_b32 s10, -1 ; SI-NEXT: s_mov_b32 s14, s10 ; SI-NEXT: s_mov_b32 s15, s11 -; SI-NEXT: s_mov_b32 s18, s10 -; SI-NEXT: s_mov_b32 s19, s11 ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_mov_b32 s12, s4 ; SI-NEXT: s_mov_b32 s13, s5 -; SI-NEXT: s_mov_b32 s16, s6 -; SI-NEXT: s_mov_b32 s17, s7 -; SI-NEXT: buffer_load_ushort v0, off, s[12:15], 0 -; SI-NEXT: buffer_load_ushort v1, off, s[16:19], 0 +; SI-NEXT: s_mov_b32 s4, s6 +; SI-NEXT: s_mov_b32 s5, s7 ; SI-NEXT: s_mov_b32 s6, s10 ; SI-NEXT: s_mov_b32 s7, s11 +; SI-NEXT: buffer_load_ushort v0, off, s[12:15], 0 +; SI-NEXT: buffer_load_ushort v1, off, s[4:7], 0 ; SI-NEXT: s_mov_b32 s8, s0 ; SI-NEXT: s_mov_b32 s9, s1 ; SI-NEXT: s_mov_b32 s4, s2 @@ -477,8 +671,8 @@ define amdgpu_kernel void @v_uaddo_i16(ptr addrspace(1) %out, ptr addrspace(1) % ; SI-NEXT: s_waitcnt vmcnt(0) ; SI-NEXT: v_add_i32_e32 v0, vcc, v0, v1 ; SI-NEXT: v_and_b32_e32 v1, 0xffff, v0 -; SI-NEXT: buffer_store_short v0, off, s[8:11], 0 ; SI-NEXT: v_cmp_ne_u32_e32 vcc, v1, v0 +; SI-NEXT: buffer_store_short v0, off, s[8:11], 0 ; SI-NEXT: s_waitcnt expcnt(0) ; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc ; SI-NEXT: buffer_store_byte v0, off, s[4:7], 0 @@ -522,6 +716,42 @@ define amdgpu_kernel void @v_uaddo_i16(ptr addrspace(1) %out, ptr addrspace(1) % ; GFX9-NEXT: global_store_short v0, v2, s[8:9] ; GFX9-NEXT: global_store_byte v0, v1, s[10:11] ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: v_uaddo_i16: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v0, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: global_load_ushort v1, v0, s[12:13] +; GFX10-NEXT: global_load_ushort v2, v0, s[14:15] +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: v_add_nc_u32_e32 v2, v1, v2 +; GFX10-NEXT: v_cmp_lt_u32_sdwa s0, v2, v1 src0_sel:WORD_0 src1_sel:WORD_0 +; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, s0 +; GFX10-NEXT: global_store_short v0, v2, s[8:9] +; GFX10-NEXT: global_store_byte v0, v1, s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: v_uaddo_i16: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_d16_b16 v1, v0, s[4:5] +; GFX11-NEXT: global_load_u16 v2, v0, s[6:7] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_add_nc_u32_e32 v2, v1, v2 +; GFX11-NEXT: v_and_b32_e32 v1, 0xffff, v1 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_and_b32_e32 v3, 0xffff, v2 +; GFX11-NEXT: v_cmp_lt_u32_e32 vcc_lo, v3, v1 +; GFX11-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b16 v0, v2, s[0:1] +; GFX11-NEXT: global_store_b8 v0, v1, s[2:3] +; GFX11-NEXT: s_endpgm %tid = call i32 @llvm.amdgcn.workitem.id.x() %tid.ext = sext i32 %tid to i64 %a.gep = getelementptr inbounds i16, ptr addrspace(1) %a.ptr @@ -544,17 +774,15 @@ define amdgpu_kernel void @v_uaddo_v2i32(ptr addrspace(1) %out, ptr addrspace(1) ; SI-NEXT: s_mov_b32 s10, -1 ; SI-NEXT: s_mov_b32 s14, s10 ; SI-NEXT: s_mov_b32 s15, s11 -; SI-NEXT: s_mov_b32 s18, s10 -; SI-NEXT: s_mov_b32 s19, s11 ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_mov_b32 s12, s4 ; SI-NEXT: s_mov_b32 s13, s5 -; SI-NEXT: s_mov_b32 s16, s6 -; SI-NEXT: s_mov_b32 s17, s7 -; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[12:15], 0 -; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[16:19], 0 +; SI-NEXT: s_mov_b32 s4, s6 +; SI-NEXT: s_mov_b32 s5, s7 ; SI-NEXT: s_mov_b32 s6, s10 ; SI-NEXT: s_mov_b32 s7, s11 +; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[12:15], 0 +; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[4:7], 0 ; SI-NEXT: s_mov_b32 s8, s0 ; SI-NEXT: s_mov_b32 s9, s1 ; SI-NEXT: s_mov_b32 s4, s2 @@ -606,6 +834,42 @@ define amdgpu_kernel void @v_uaddo_v2i32(ptr addrspace(1) %out, ptr addrspace(1) ; GFX9-NEXT: global_store_dwordx2 v4, v[0:1], s[8:9] ; GFX9-NEXT: global_store_dwordx2 v4, v[2:3], s[10:11] ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: v_uaddo_v2i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v4, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: global_load_dwordx2 v[0:1], v4, s[12:13] +; GFX10-NEXT: global_load_dwordx2 v[2:3], v4, s[14:15] +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: v_add_co_u32 v1, s0, v1, v3 +; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, 1, s0 +; GFX10-NEXT: v_add_co_u32 v0, s0, v0, v2 +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, s0 +; GFX10-NEXT: global_store_dwordx2 v4, v[0:1], s[8:9] +; GFX10-NEXT: global_store_dwordx2 v4, v[2:3], s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: v_uaddo_v2i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v4, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b64 v[0:1], v4, s[4:5] +; GFX11-NEXT: global_load_b64 v[2:3], v4, s[6:7] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_add_co_u32 v1, s4, v1, v3 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 +; GFX11-NEXT: v_add_co_u32 v0, s4, v0, v2 +; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, s4 +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b64 v4, v[0:1], s[0:1] +; GFX11-NEXT: global_store_b64 v4, v[2:3], s[2:3] +; GFX11-NEXT: s_endpgm %a = load <2 x i32>, ptr addrspace(1) %aptr, align 4 %b = load <2 x i32>, ptr addrspace(1) %bptr, align 4 %sadd = call { <2 x i32>, <2 x i1> } @llvm.uadd.with.overflow.v2i32(<2 x i32> %a, <2 x i32> %b) nounwind @@ -623,26 +887,27 @@ define amdgpu_kernel void @s_uaddo_clamp_bit(ptr addrspace(1) %out, ptr addrspac ; SI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xd ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: v_mov_b32_e32 v0, s1 -; SI-NEXT: v_add_i32_e32 v0, vcc, s0, v0 ; SI-NEXT: s_cmp_eq_u32 s0, s1 +; SI-NEXT: v_add_i32_e32 v0, vcc, s0, v0 ; SI-NEXT: s_mov_b64 s[0:1], 0 ; SI-NEXT: s_cbranch_scc1 .LBB8_2 ; SI-NEXT: ; %bb.1: ; %if ; SI-NEXT: s_xor_b64 s[0:1], vcc, -1 ; SI-NEXT: .LBB8_2: ; %exit ; SI-NEXT: s_load_dwordx4 s[4:7], s[4:5], 0x9 -; SI-NEXT: s_mov_b32 s3, 0xf000 -; SI-NEXT: s_mov_b32 s2, -1 -; SI-NEXT: v_cndmask_b32_e64 v1, 0, 1, s[0:1] -; SI-NEXT: s_mov_b32 s10, s2 -; SI-NEXT: s_mov_b32 s11, s3 +; SI-NEXT: s_mov_b32 s11, 0xf000 +; SI-NEXT: s_mov_b32 s10, -1 ; SI-NEXT: s_waitcnt lgkmcnt(0) -; SI-NEXT: s_mov_b32 s0, s4 -; SI-NEXT: s_mov_b32 s1, s5 -; SI-NEXT: s_mov_b32 s8, s6 -; SI-NEXT: s_mov_b32 s9, s7 -; SI-NEXT: buffer_store_dword v0, off, s[0:3], 0 -; SI-NEXT: buffer_store_byte v1, off, s[8:11], 0 +; SI-NEXT: s_mov_b32 s8, s4 +; SI-NEXT: s_mov_b32 s9, s5 +; SI-NEXT: s_mov_b32 s4, s6 +; SI-NEXT: s_mov_b32 s5, s7 +; SI-NEXT: s_mov_b32 s6, s10 +; SI-NEXT: s_mov_b32 s7, s11 +; SI-NEXT: buffer_store_dword v0, off, s[8:11], 0 +; SI-NEXT: s_waitcnt expcnt(0) +; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; SI-NEXT: buffer_store_byte v0, off, s[4:7], 0 ; SI-NEXT: s_endpgm ; ; VI-LABEL: s_uaddo_clamp_bit: @@ -687,6 +952,45 @@ define amdgpu_kernel void @s_uaddo_clamp_bit(ptr addrspace(1) %out, ptr addrspac ; GFX9-NEXT: global_store_dword v1, v0, s[8:9] ; GFX9-NEXT: global_store_byte v1, v2, s[10:11] ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: s_uaddo_clamp_bit: +; GFX10: ; %bb.0: ; %entry +; GFX10-NEXT: s_load_dwordx2 s[2:3], s[4:5], 0x34 +; GFX10-NEXT: s_mov_b32 s0, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: v_add_co_u32 v0, s1, s2, s3 +; GFX10-NEXT: s_cmp_eq_u32 s2, s3 +; GFX10-NEXT: s_cbranch_scc1 .LBB8_2 +; GFX10-NEXT: ; %bb.1: ; %if +; GFX10-NEXT: s_xor_b32 s0, s1, -1 +; GFX10-NEXT: .LBB8_2: ; %exit +; GFX10-NEXT: s_load_dwordx4 s[8:11], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v1, 0 +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, s0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: global_store_dword v1, v0, s[8:9] +; GFX10-NEXT: global_store_byte v1, v2, s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: s_uaddo_clamp_bit: +; GFX11: ; %bb.0: ; %entry +; GFX11-NEXT: s_load_b64 s[2:3], s[4:5], 0x34 +; GFX11-NEXT: s_mov_b32 s0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: v_add_co_u32 v0, s1, s2, s3 +; GFX11-NEXT: s_cmp_eq_u32 s2, s3 +; GFX11-NEXT: s_cbranch_scc1 .LBB8_2 +; GFX11-NEXT: ; %bb.1: ; %if +; GFX11-NEXT: s_xor_b32 s0, s1, -1 +; GFX11-NEXT: .LBB8_2: ; %exit +; GFX11-NEXT: s_load_b128 s[4:7], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v1, 0 +; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, s0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b32 v1, v0, s[4:5] +; GFX11-NEXT: global_store_b8 v1, v2, s[6:7] +; GFX11-NEXT: s_endpgm entry: %uadd = call { i32, i1 } @llvm.uadd.with.overflow.i32(i32 %a, i32 %b) %val = extractvalue { i32, i1 } %uadd, 0 @@ -711,19 +1015,19 @@ define amdgpu_kernel void @v_uaddo_clamp_bit(ptr addrspace(1) %out, ptr addrspac ; SI-NEXT: s_load_dwordx8 s[4:11], s[4:5], 0x9 ; SI-NEXT: s_mov_b32 s3, 0xf000 ; SI-NEXT: s_mov_b32 s2, -1 -; SI-NEXT: s_mov_b32 s14, s2 -; SI-NEXT: s_mov_b32 s15, s3 ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_mov_b32 s0, s8 ; SI-NEXT: s_mov_b32 s1, s9 -; SI-NEXT: s_mov_b32 s12, s10 -; SI-NEXT: s_mov_b32 s13, s11 +; SI-NEXT: s_mov_b32 s8, s10 +; SI-NEXT: s_mov_b32 s9, s11 +; SI-NEXT: s_mov_b32 s10, s2 +; SI-NEXT: s_mov_b32 s11, s3 ; SI-NEXT: buffer_load_dword v1, off, s[0:3], 0 -; SI-NEXT: buffer_load_dword v2, off, s[12:15], 0 +; SI-NEXT: buffer_load_dword v2, off, s[8:11], 0 +; SI-NEXT: s_mov_b64 s[8:9], 0 ; SI-NEXT: s_waitcnt vmcnt(0) -; SI-NEXT: v_add_i32_e64 v0, s[0:1], v1, v2 ; SI-NEXT: v_cmp_eq_u32_e32 vcc, v1, v2 -; SI-NEXT: s_mov_b64 s[8:9], 0 +; SI-NEXT: v_add_i32_e64 v0, s[0:1], v1, v2 ; SI-NEXT: s_cbranch_vccnz .LBB9_2 ; SI-NEXT: ; %bb.1: ; %if ; SI-NEXT: s_xor_b64 s[8:9], s[0:1], -1 @@ -786,6 +1090,50 @@ define amdgpu_kernel void @v_uaddo_clamp_bit(ptr addrspace(1) %out, ptr addrspac ; GFX9-NEXT: v_cndmask_b32_e64 v1, 0, 1, s[2:3] ; GFX9-NEXT: global_store_byte v0, v1, s[10:11] ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: v_uaddo_clamp_bit: +; GFX10: ; %bb.0: ; %entry +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v0, 0 +; GFX10-NEXT: s_mov_b32 s0, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: global_load_dword v1, v0, s[12:13] +; GFX10-NEXT: global_load_dword v2, v0, s[14:15] +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v2 +; GFX10-NEXT: v_add_co_u32 v1, s1, v1, v2 +; GFX10-NEXT: s_cbranch_vccnz .LBB9_2 +; GFX10-NEXT: ; %bb.1: ; %if +; GFX10-NEXT: s_xor_b32 s0, s1, -1 +; GFX10-NEXT: .LBB9_2: ; %exit +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, s0 +; GFX10-NEXT: global_store_dword v0, v1, s[8:9] +; GFX10-NEXT: global_store_byte v0, v2, s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: v_uaddo_clamp_bit: +; GFX11: ; %bb.0: ; %entry +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[4:5] +; GFX11-NEXT: global_load_b32 v2, v0, s[6:7] +; GFX11-NEXT: s_mov_b32 s4, 0 +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v2 +; GFX11-NEXT: v_add_co_u32 v1, s5, v1, v2 +; GFX11-NEXT: s_cbranch_vccnz .LBB9_2 +; GFX11-NEXT: ; %bb.1: ; %if +; GFX11-NEXT: s_xor_b32 s4, s5, -1 +; GFX11-NEXT: .LBB9_2: ; %exit +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) +; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, s4 +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: global_store_b8 v0, v2, s[2:3] +; GFX11-NEXT: s_endpgm entry: %tid = call i32 @llvm.amdgcn.workitem.id.x() %tid.ext = sext i32 %tid to i64 @@ -813,23 +1161,23 @@ exit: define amdgpu_cs void @sv_uaddo_i128(ptr addrspace(1) %out, i128 inreg %a, i128 %b) { ; SI-LABEL: sv_uaddo_i128: ; SI: ; %bb.0: -; SI-NEXT: s_mov_b32 s7, 0xf000 -; SI-NEXT: s_mov_b32 s6, 0 ; SI-NEXT: v_add_i32_e32 v2, vcc, s0, v2 ; SI-NEXT: v_mov_b32_e32 v6, s1 -; SI-NEXT: v_mov_b32_e32 v7, s2 -; SI-NEXT: v_mov_b32_e32 v8, s3 -; SI-NEXT: s_mov_b32 s4, s6 -; SI-NEXT: s_mov_b32 s5, s6 ; SI-NEXT: v_addc_u32_e32 v3, vcc, v6, v3, vcc -; SI-NEXT: v_addc_u32_e32 v4, vcc, v7, v4, vcc -; SI-NEXT: v_cmp_gt_u64_e64 s[0:1], s[0:1], v[2:3] -; SI-NEXT: v_cndmask_b32_e64 v2, 0, 1, s[0:1] -; SI-NEXT: v_addc_u32_e32 v5, vcc, v8, v5, vcc +; SI-NEXT: v_mov_b32_e32 v6, s2 +; SI-NEXT: v_addc_u32_e32 v4, vcc, v6, v4, vcc +; SI-NEXT: v_mov_b32_e32 v6, s3 +; SI-NEXT: v_addc_u32_e32 v5, vcc, v6, v5, vcc +; SI-NEXT: v_cmp_gt_u64_e32 vcc, s[0:1], v[2:3] +; SI-NEXT: s_mov_b32 s6, 0 +; SI-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc ; SI-NEXT: v_cmp_gt_u64_e32 vcc, s[2:3], v[4:5] +; SI-NEXT: s_mov_b32 s7, 0xf000 ; SI-NEXT: v_cndmask_b32_e64 v3, 0, 1, vcc ; SI-NEXT: v_cmp_eq_u64_e32 vcc, s[2:3], v[4:5] +; SI-NEXT: s_mov_b32 s4, s6 ; SI-NEXT: v_cndmask_b32_e32 v2, v3, v2, vcc +; SI-NEXT: s_mov_b32 s5, s6 ; SI-NEXT: v_and_b32_e32 v2, 1, v2 ; SI-NEXT: buffer_store_dword v2, v[0:1], s[4:7], 0 addr64 ; SI-NEXT: s_endpgm @@ -871,6 +1219,41 @@ define amdgpu_cs void @sv_uaddo_i128(ptr addrspace(1) %out, i128 inreg %a, i128 ; GFX9-NEXT: v_and_b32_e32 v2, 1, v2 ; GFX9-NEXT: global_store_dword v[0:1], v2, off ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: sv_uaddo_i128: +; GFX10: ; %bb.0: +; GFX10-NEXT: v_add_co_u32 v2, vcc_lo, s0, v2 +; GFX10-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, s1, v3, vcc_lo +; GFX10-NEXT: v_add_co_ci_u32_e32 v4, vcc_lo, s2, v4, vcc_lo +; GFX10-NEXT: v_add_co_ci_u32_e32 v5, vcc_lo, s3, v5, vcc_lo +; GFX10-NEXT: v_cmp_gt_u64_e32 vcc_lo, s[0:1], v[2:3] +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, vcc_lo +; GFX10-NEXT: v_cmp_gt_u64_e32 vcc_lo, s[2:3], v[4:5] +; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, 1, vcc_lo +; GFX10-NEXT: v_cmp_eq_u64_e32 vcc_lo, s[2:3], v[4:5] +; GFX10-NEXT: v_cndmask_b32_e32 v2, v3, v2, vcc_lo +; GFX10-NEXT: v_and_b32_e32 v2, 1, v2 +; GFX10-NEXT: global_store_dword v[0:1], v2, off +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: sv_uaddo_i128: +; GFX11: ; %bb.0: +; GFX11-NEXT: v_add_co_u32 v2, vcc_lo, s0, v2 +; GFX11-NEXT: v_add_co_ci_u32_e32 v3, vcc_lo, s1, v3, vcc_lo +; GFX11-NEXT: v_add_co_ci_u32_e32 v4, vcc_lo, s2, v4, vcc_lo +; GFX11-NEXT: v_add_co_ci_u32_e32 v5, vcc_lo, s3, v5, vcc_lo +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_2) +; GFX11-NEXT: v_cmp_gt_u64_e32 vcc_lo, s[2:3], v[4:5] +; GFX11-NEXT: v_cndmask_b32_e64 v6, 0, 1, vcc_lo +; GFX11-NEXT: v_cmp_gt_u64_e32 vcc_lo, s[0:1], v[2:3] +; GFX11-NEXT: v_mov_b16_e32 v2.l, v6.l +; GFX11-NEXT: v_cndmask_b32_e64 v3, 0, 1, vcc_lo +; GFX11-NEXT: v_cmp_eq_u64_e32 vcc_lo, s[2:3], v[4:5] +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_cndmask_b16 v2.l, v2.l, v3.l, vcc_lo +; GFX11-NEXT: v_and_b32_e32 v2, 1, v2 +; GFX11-NEXT: global_store_b32 v[0:1], v2, off +; GFX11-NEXT: s_endpgm %uadd = call { i128, i1 } @llvm.uadd.with.overflow.i128(i128 %a, i128 %b) %carry = extractvalue { i128, i1 } %uadd, 1 %carry.ext = zext i1 %carry to i32 diff --git a/llvm/test/CodeGen/AMDGPU/usubo.ll b/llvm/test/CodeGen/AMDGPU/usubo.ll index 7d7f1b4..0289dab 100644 --- a/llvm/test/CodeGen/AMDGPU/usubo.ll +++ b/llvm/test/CodeGen/AMDGPU/usubo.ll @@ -1,8 +1,9 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn < %s | FileCheck -check-prefixes=SI %s -; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tonga < %s | FileCheck -check-prefixes=VI %s -; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx900 < %s | FileCheck -check-prefixes=GFX9 %s - +; RUN: llc < %s -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tahiti | FileCheck %s --check-prefix=SI +; RUN: llc < %s -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tonga | FileCheck %s --check-prefix=VI +; RUN: llc < %s -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx900 | FileCheck %s --check-prefix=GFX9 +; RUN: llc < %s -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1010 | FileCheck %s --check-prefix=GFX10 +; RUN: llc < %s -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=gfx1100 | FileCheck %s --check-prefix=GFX11 define amdgpu_kernel void @s_usubo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 %b) #0 { ; SI-LABEL: s_usubo_i64_zext: @@ -13,14 +14,14 @@ define amdgpu_kernel void @s_usubo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 % ; SI-NEXT: s_mov_b32 s6, -1 ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_mov_b32 s4, s0 -; SI-NEXT: s_mov_b32 s5, s1 ; SI-NEXT: s_sub_u32 s0, s2, s8 ; SI-NEXT: v_mov_b32_e32 v0, s2 -; SI-NEXT: v_mov_b32_e32 v1, s3 +; SI-NEXT: s_mov_b32 s5, s1 ; SI-NEXT: s_subb_u32 s1, s3, s9 +; SI-NEXT: v_mov_b32_e32 v1, s3 ; SI-NEXT: v_cmp_gt_u64_e32 vcc, s[0:1], v[0:1] -; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc ; SI-NEXT: v_mov_b32_e32 v1, s1 +; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc ; SI-NEXT: v_add_i32_e32 v0, vcc, s0, v0 ; SI-NEXT: v_addc_u32_e32 v1, vcc, 0, v1, vcc ; SI-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0 @@ -62,6 +63,40 @@ define amdgpu_kernel void @s_usubo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 % ; GFX9-NEXT: v_addc_co_u32_e32 v1, vcc, 0, v1, vcc ; GFX9-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: s_usubo_i64_zext: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 +; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34 +; GFX10-NEXT: v_mov_b32_e32 v2, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_sub_u32 s4, s2, s6 +; GFX10-NEXT: s_subb_u32 s5, s3, s7 +; GFX10-NEXT: v_cmp_gt_u64_e64 s2, s[4:5], s[2:3] +; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, s2 +; GFX10-NEXT: v_add_co_u32 v0, s2, s4, v0 +; GFX10-NEXT: v_add_co_ci_u32_e64 v1, s2, s5, 0, s2 +; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[0:1] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: s_usubo_i64_zext: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 +; GFX11-NEXT: s_load_b64 s[4:5], s[4:5], 0x34 +; GFX11-NEXT: v_mov_b32_e32 v2, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_sub_u32 s4, s2, s4 +; GFX11-NEXT: s_subb_u32 s5, s3, s5 +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_cmp_gt_u64_e64 s2, s[4:5], s[2:3] +; GFX11-NEXT: v_cndmask_b32_e64 v0, 0, 1, s2 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_add_co_u32 v0, s2, s4, v0 +; GFX11-NEXT: v_add_co_ci_u32_e64 v1, null, s5, 0, s2 +; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1] +; GFX11-NEXT: s_endpgm %usub = call { i64, i1 } @llvm.usub.with.overflow.i64(i64 %a, i64 %b) #0 %val = extractvalue { i64, i1 } %usub, 0 %carry = extractvalue { i64, i1 } %usub, 1 @@ -76,21 +111,21 @@ define amdgpu_kernel void @s_usubo_i32(ptr addrspace(1) %out, ptr addrspace(1) % ; SI-LABEL: s_usubo_i32: ; SI: ; %bb.0: ; SI-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 -; SI-NEXT: s_load_dwordx2 s[12:13], s[4:5], 0xd +; SI-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd ; SI-NEXT: s_mov_b32 s7, 0xf000 ; SI-NEXT: s_mov_b32 s6, -1 -; SI-NEXT: s_mov_b32 s10, s6 -; SI-NEXT: s_mov_b32 s11, s7 ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_mov_b32 s4, s0 +; SI-NEXT: v_mov_b32_e32 v0, s9 ; SI-NEXT: s_mov_b32 s5, s1 -; SI-NEXT: s_mov_b32 s8, s2 -; SI-NEXT: s_mov_b32 s9, s3 -; SI-NEXT: v_mov_b32_e32 v0, s13 -; SI-NEXT: v_sub_i32_e32 v0, vcc, s12, v0 +; SI-NEXT: v_sub_i32_e32 v0, vcc, s8, v0 +; SI-NEXT: s_mov_b32 s0, s2 +; SI-NEXT: s_mov_b32 s1, s3 +; SI-NEXT: s_mov_b32 s2, s6 +; SI-NEXT: s_mov_b32 s3, s7 ; SI-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc ; SI-NEXT: buffer_store_dword v0, off, s[4:7], 0 -; SI-NEXT: buffer_store_byte v1, off, s[8:11], 0 +; SI-NEXT: buffer_store_byte v1, off, s[0:3], 0 ; SI-NEXT: s_endpgm ; ; VI-LABEL: s_usubo_i32: @@ -121,6 +156,34 @@ define amdgpu_kernel void @s_usubo_i32(ptr addrspace(1) %out, ptr addrspace(1) % ; GFX9-NEXT: global_store_dword v0, v1, s[0:1] ; GFX9-NEXT: global_store_byte v0, v2, s[2:3] ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: s_usubo_i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: s_load_dwordx2 s[6:7], s[4:5], 0x34 +; GFX10-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v0, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: v_sub_co_u32 v1, s4, s6, s7 +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, s4 +; GFX10-NEXT: global_store_dword v0, v1, s[0:1] +; GFX10-NEXT: global_store_byte v0, v2, s[2:3] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: s_usubo_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: s_load_b64 s[6:7], s[4:5], 0x34 +; GFX11-NEXT: s_load_b128 s[0:3], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: v_sub_co_u32 v1, s4, s6, s7 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, s4 +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: global_store_b8 v0, v2, s[2:3] +; GFX11-NEXT: s_endpgm %usub = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 %a, i32 %b) %val = extractvalue { i32, i1 } %usub, 0 %carry = extractvalue { i32, i1 } %usub, 1 @@ -137,17 +200,15 @@ define amdgpu_kernel void @v_usubo_i32(ptr addrspace(1) %out, ptr addrspace(1) % ; SI-NEXT: s_mov_b32 s10, -1 ; SI-NEXT: s_mov_b32 s14, s10 ; SI-NEXT: s_mov_b32 s15, s11 -; SI-NEXT: s_mov_b32 s18, s10 -; SI-NEXT: s_mov_b32 s19, s11 ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_mov_b32 s12, s4 ; SI-NEXT: s_mov_b32 s13, s5 -; SI-NEXT: s_mov_b32 s16, s6 -; SI-NEXT: s_mov_b32 s17, s7 -; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 -; SI-NEXT: buffer_load_dword v1, off, s[16:19], 0 +; SI-NEXT: s_mov_b32 s4, s6 +; SI-NEXT: s_mov_b32 s5, s7 ; SI-NEXT: s_mov_b32 s6, s10 ; SI-NEXT: s_mov_b32 s7, s11 +; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 +; SI-NEXT: buffer_load_dword v1, off, s[4:7], 0 ; SI-NEXT: s_mov_b32 s8, s0 ; SI-NEXT: s_mov_b32 s9, s1 ; SI-NEXT: s_mov_b32 s4, s2 @@ -193,6 +254,38 @@ define amdgpu_kernel void @v_usubo_i32(ptr addrspace(1) %out, ptr addrspace(1) % ; GFX9-NEXT: global_store_dword v0, v1, s[8:9] ; GFX9-NEXT: global_store_byte v0, v2, s[10:11] ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: v_usubo_i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v0, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: global_load_dword v1, v0, s[12:13] +; GFX10-NEXT: global_load_dword v2, v0, s[14:15] +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: v_sub_co_u32 v1, s0, v1, v2 +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, s0 +; GFX10-NEXT: global_store_dword v0, v1, s[8:9] +; GFX10-NEXT: global_store_byte v0, v2, s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: v_usubo_i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[4:5] +; GFX11-NEXT: global_load_b32 v2, v0, s[6:7] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_sub_co_u32 v1, s4, v1, v2 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, s4 +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: global_store_b8 v0, v2, s[2:3] +; GFX11-NEXT: s_endpgm %tid = call i32 @llvm.amdgcn.workitem.id.x() %tid.ext = sext i32 %tid to i64 %a.gep = getelementptr inbounds i32, ptr addrspace(1) %a.ptr @@ -215,17 +308,15 @@ define amdgpu_kernel void @v_usubo_i32_novcc(ptr addrspace(1) %out, ptr addrspac ; SI-NEXT: s_mov_b32 s10, -1 ; SI-NEXT: s_mov_b32 s14, s10 ; SI-NEXT: s_mov_b32 s15, s11 -; SI-NEXT: s_mov_b32 s18, s10 -; SI-NEXT: s_mov_b32 s19, s11 ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_mov_b32 s12, s4 ; SI-NEXT: s_mov_b32 s13, s5 -; SI-NEXT: s_mov_b32 s16, s6 -; SI-NEXT: s_mov_b32 s17, s7 -; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 -; SI-NEXT: buffer_load_dword v1, off, s[16:19], 0 +; SI-NEXT: s_mov_b32 s4, s6 +; SI-NEXT: s_mov_b32 s5, s7 ; SI-NEXT: s_mov_b32 s6, s10 ; SI-NEXT: s_mov_b32 s7, s11 +; SI-NEXT: buffer_load_dword v0, off, s[12:15], 0 +; SI-NEXT: buffer_load_dword v1, off, s[4:7], 0 ; SI-NEXT: s_mov_b32 s8, s0 ; SI-NEXT: s_mov_b32 s9, s1 ; SI-NEXT: s_mov_b32 s4, s2 @@ -283,6 +374,45 @@ define amdgpu_kernel void @v_usubo_i32_novcc(ptr addrspace(1) %out, ptr addrspac ; GFX9-NEXT: global_store_byte v0, v2, s[10:11] ; GFX9-NEXT: s_waitcnt vmcnt(0) ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: v_usubo_i32_novcc: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v0, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: global_load_dword v1, v0, s[12:13] +; GFX10-NEXT: global_load_dword v2, v0, s[14:15] +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: v_sub_co_u32 v1, s0, v1, v2 +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, s0 +; GFX10-NEXT: global_store_dword v0, v1, s[8:9] +; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX10-NEXT: ;;#ASMSTART +; GFX10-NEXT: ;;#ASMEND +; GFX10-NEXT: global_store_byte v0, v2, s[10:11] +; GFX10-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: v_usubo_i32_novcc: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[4:5] +; GFX11-NEXT: global_load_b32 v2, v0, s[6:7] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_sub_co_u32 v1, s4, v1, v2 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, s4 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] dlc +; GFX11-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX11-NEXT: ;;#ASMSTART +; GFX11-NEXT: ;;#ASMEND +; GFX11-NEXT: global_store_b8 v0, v2, s[2:3] dlc +; GFX11-NEXT: s_waitcnt_vscnt null, 0x0 +; GFX11-NEXT: s_endpgm %tid = call i32 @llvm.amdgcn.workitem.id.x() %tid.ext = sext i32 %tid to i64 %a.gep = getelementptr inbounds i32, ptr addrspace(1) %a.ptr @@ -306,21 +436,21 @@ define amdgpu_kernel void @s_usubo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; SI-NEXT: s_mov_b32 s10, -1 ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_sub_u32 s6, s4, s6 -; SI-NEXT: s_subb_u32 s7, s5, s7 -; SI-NEXT: s_mov_b32 s14, s10 -; SI-NEXT: s_mov_b32 s15, s11 -; SI-NEXT: s_mov_b32 s8, s0 -; SI-NEXT: s_mov_b32 s9, s1 -; SI-NEXT: s_mov_b32 s12, s2 -; SI-NEXT: s_mov_b32 s13, s3 ; SI-NEXT: v_mov_b32_e32 v0, s4 +; SI-NEXT: s_subb_u32 s7, s5, s7 ; SI-NEXT: v_mov_b32_e32 v1, s5 +; SI-NEXT: v_cmp_gt_u64_e32 vcc, s[6:7], v[0:1] ; SI-NEXT: v_mov_b32_e32 v2, s6 +; SI-NEXT: s_mov_b32 s8, s0 +; SI-NEXT: s_mov_b32 s9, s1 +; SI-NEXT: s_mov_b32 s0, s2 +; SI-NEXT: s_mov_b32 s1, s3 +; SI-NEXT: s_mov_b32 s2, s10 +; SI-NEXT: s_mov_b32 s3, s11 ; SI-NEXT: v_mov_b32_e32 v3, s7 -; SI-NEXT: buffer_store_dwordx2 v[2:3], off, s[8:11], 0 -; SI-NEXT: v_cmp_gt_u64_e32 vcc, s[6:7], v[0:1] ; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc -; SI-NEXT: buffer_store_byte v0, off, s[12:15], 0 +; SI-NEXT: buffer_store_dwordx2 v[2:3], off, s[8:11], 0 +; SI-NEXT: buffer_store_byte v0, off, s[0:3], 0 ; SI-NEXT: s_endpgm ; ; VI-LABEL: s_usubo_i64: @@ -359,6 +489,37 @@ define amdgpu_kernel void @s_usubo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; GFX9-NEXT: global_store_dwordx2 v4, v[2:3], s[8:9] ; GFX9-NEXT: global_store_byte v4, v0, s[10:11] ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: s_usubo_i64: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v2, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_sub_u32 s0, s12, s14 +; GFX10-NEXT: s_subb_u32 s1, s13, s15 +; GFX10-NEXT: v_mov_b32_e32 v0, s0 +; GFX10-NEXT: v_mov_b32_e32 v1, s1 +; GFX10-NEXT: v_cmp_gt_u64_e64 s0, s[0:1], s[12:13] +; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, 1, s0 +; GFX10-NEXT: global_store_dwordx2 v2, v[0:1], s[8:9] +; GFX10-NEXT: global_store_byte v2, v3, s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: s_usubo_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_sub_u32 s6, s4, s6 +; GFX11-NEXT: s_subb_u32 s7, s5, s7 +; GFX11-NEXT: v_mov_b32_e32 v0, s6 +; GFX11-NEXT: v_cmp_gt_u64_e64 s4, s[6:7], s[4:5] +; GFX11-NEXT: v_dual_mov_b32 v2, 0 :: v_dual_mov_b32 v1, s7 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) +; GFX11-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b64 v2, v[0:1], s[0:1] +; GFX11-NEXT: global_store_b8 v2, v3, s[2:3] +; GFX11-NEXT: s_endpgm %usub = call { i64, i1 } @llvm.usub.with.overflow.i64(i64 %a, i64 %b) %val = extractvalue { i64, i1 } %usub, 0 %carry = extractvalue { i64, i1 } %usub, 1 @@ -375,17 +536,15 @@ define amdgpu_kernel void @v_usubo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; SI-NEXT: s_mov_b32 s10, -1 ; SI-NEXT: s_mov_b32 s14, s10 ; SI-NEXT: s_mov_b32 s15, s11 -; SI-NEXT: s_mov_b32 s18, s10 -; SI-NEXT: s_mov_b32 s19, s11 ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_mov_b32 s12, s4 ; SI-NEXT: s_mov_b32 s13, s5 -; SI-NEXT: s_mov_b32 s16, s6 -; SI-NEXT: s_mov_b32 s17, s7 -; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[12:15], 0 -; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[16:19], 0 +; SI-NEXT: s_mov_b32 s4, s6 +; SI-NEXT: s_mov_b32 s5, s7 ; SI-NEXT: s_mov_b32 s6, s10 ; SI-NEXT: s_mov_b32 s7, s11 +; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[12:15], 0 +; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[4:7], 0 ; SI-NEXT: s_mov_b32 s8, s0 ; SI-NEXT: s_mov_b32 s9, s1 ; SI-NEXT: s_mov_b32 s4, s2 @@ -393,8 +552,8 @@ define amdgpu_kernel void @v_usubo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; SI-NEXT: s_waitcnt vmcnt(0) ; SI-NEXT: v_sub_i32_e32 v2, vcc, v0, v2 ; SI-NEXT: v_subb_u32_e32 v3, vcc, v1, v3, vcc -; SI-NEXT: buffer_store_dwordx2 v[2:3], off, s[8:11], 0 ; SI-NEXT: v_cmp_gt_u64_e32 vcc, v[2:3], v[0:1] +; SI-NEXT: buffer_store_dwordx2 v[2:3], off, s[8:11], 0 ; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc ; SI-NEXT: buffer_store_byte v0, off, s[4:7], 0 ; SI-NEXT: s_endpgm @@ -437,6 +596,42 @@ define amdgpu_kernel void @v_usubo_i64(ptr addrspace(1) %out, ptr addrspace(1) % ; GFX9-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc ; GFX9-NEXT: global_store_byte v4, v0, s[10:11] ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: v_usubo_i64: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v4, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: global_load_dwordx2 v[0:1], v4, s[12:13] +; GFX10-NEXT: global_load_dwordx2 v[2:3], v4, s[14:15] +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: v_sub_co_u32 v2, vcc_lo, v0, v2 +; GFX10-NEXT: v_sub_co_ci_u32_e32 v3, vcc_lo, v1, v3, vcc_lo +; GFX10-NEXT: v_cmp_gt_u64_e32 vcc_lo, v[2:3], v[0:1] +; GFX10-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo +; GFX10-NEXT: global_store_dwordx2 v4, v[2:3], s[8:9] +; GFX10-NEXT: global_store_byte v4, v0, s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: v_usubo_i64: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v4, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b64 v[0:1], v4, s[4:5] +; GFX11-NEXT: global_load_b64 v[2:3], v4, s[6:7] +; GFX11-NEXT: s_waitcnt vmcnt(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_gt_u64_e32 vcc_lo, v[2:3], v[0:1] +; GFX11-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc_lo +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b64 v4, v[2:3], s[0:1] +; GFX11-NEXT: global_store_b8 v4, v0, s[2:3] +; GFX11-NEXT: s_endpgm %tid = call i32 @llvm.amdgcn.workitem.id.x() %tid.ext = sext i32 %tid to i64 %a.gep = getelementptr inbounds i64, ptr addrspace(1) %a.ptr @@ -459,17 +654,15 @@ define amdgpu_kernel void @v_usubo_i16(ptr addrspace(1) %out, ptr addrspace(1) % ; SI-NEXT: s_mov_b32 s10, -1 ; SI-NEXT: s_mov_b32 s14, s10 ; SI-NEXT: s_mov_b32 s15, s11 -; SI-NEXT: s_mov_b32 s18, s10 -; SI-NEXT: s_mov_b32 s19, s11 ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_mov_b32 s12, s4 ; SI-NEXT: s_mov_b32 s13, s5 -; SI-NEXT: s_mov_b32 s16, s6 -; SI-NEXT: s_mov_b32 s17, s7 -; SI-NEXT: buffer_load_ushort v0, off, s[12:15], 0 -; SI-NEXT: buffer_load_ushort v1, off, s[16:19], 0 +; SI-NEXT: s_mov_b32 s4, s6 +; SI-NEXT: s_mov_b32 s5, s7 ; SI-NEXT: s_mov_b32 s6, s10 ; SI-NEXT: s_mov_b32 s7, s11 +; SI-NEXT: buffer_load_ushort v0, off, s[12:15], 0 +; SI-NEXT: buffer_load_ushort v1, off, s[4:7], 0 ; SI-NEXT: s_mov_b32 s8, s0 ; SI-NEXT: s_mov_b32 s9, s1 ; SI-NEXT: s_mov_b32 s4, s2 @@ -477,8 +670,8 @@ define amdgpu_kernel void @v_usubo_i16(ptr addrspace(1) %out, ptr addrspace(1) % ; SI-NEXT: s_waitcnt vmcnt(0) ; SI-NEXT: v_sub_i32_e32 v0, vcc, v0, v1 ; SI-NEXT: v_and_b32_e32 v1, 0xffff, v0 -; SI-NEXT: buffer_store_short v0, off, s[8:11], 0 ; SI-NEXT: v_cmp_ne_u32_e32 vcc, v1, v0 +; SI-NEXT: buffer_store_short v0, off, s[8:11], 0 ; SI-NEXT: s_waitcnt expcnt(0) ; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, vcc ; SI-NEXT: buffer_store_byte v0, off, s[4:7], 0 @@ -522,6 +715,42 @@ define amdgpu_kernel void @v_usubo_i16(ptr addrspace(1) %out, ptr addrspace(1) % ; GFX9-NEXT: global_store_short v0, v2, s[8:9] ; GFX9-NEXT: global_store_byte v0, v1, s[10:11] ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: v_usubo_i16: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v0, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: global_load_ushort v1, v0, s[12:13] +; GFX10-NEXT: global_load_ushort v2, v0, s[14:15] +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: v_sub_nc_u32_e32 v2, v1, v2 +; GFX10-NEXT: v_cmp_gt_u32_sdwa s0, v2, v1 src0_sel:WORD_0 src1_sel:WORD_0 +; GFX10-NEXT: v_cndmask_b32_e64 v1, 0, 1, s0 +; GFX10-NEXT: global_store_short v0, v2, s[8:9] +; GFX10-NEXT: global_store_byte v0, v1, s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: v_usubo_i16: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_d16_b16 v1, v0, s[4:5] +; GFX11-NEXT: global_load_u16 v2, v0, s[6:7] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_sub_nc_u32_e32 v2, v1, v2 +; GFX11-NEXT: v_and_b32_e32 v1, 0xffff, v1 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_and_b32_e32 v3, 0xffff, v2 +; GFX11-NEXT: v_cmp_gt_u32_e32 vcc_lo, v3, v1 +; GFX11-NEXT: v_cndmask_b32_e64 v1, 0, 1, vcc_lo +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b16 v0, v2, s[0:1] +; GFX11-NEXT: global_store_b8 v0, v1, s[2:3] +; GFX11-NEXT: s_endpgm %tid = call i32 @llvm.amdgcn.workitem.id.x() %tid.ext = sext i32 %tid to i64 %a.gep = getelementptr inbounds i16, ptr addrspace(1) %a.ptr @@ -544,17 +773,15 @@ define amdgpu_kernel void @v_usubo_v2i32(ptr addrspace(1) %out, ptr addrspace(1) ; SI-NEXT: s_mov_b32 s10, -1 ; SI-NEXT: s_mov_b32 s14, s10 ; SI-NEXT: s_mov_b32 s15, s11 -; SI-NEXT: s_mov_b32 s18, s10 -; SI-NEXT: s_mov_b32 s19, s11 ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_mov_b32 s12, s4 ; SI-NEXT: s_mov_b32 s13, s5 -; SI-NEXT: s_mov_b32 s16, s6 -; SI-NEXT: s_mov_b32 s17, s7 -; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[12:15], 0 -; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[16:19], 0 +; SI-NEXT: s_mov_b32 s4, s6 +; SI-NEXT: s_mov_b32 s5, s7 ; SI-NEXT: s_mov_b32 s6, s10 ; SI-NEXT: s_mov_b32 s7, s11 +; SI-NEXT: buffer_load_dwordx2 v[0:1], off, s[12:15], 0 +; SI-NEXT: buffer_load_dwordx2 v[2:3], off, s[4:7], 0 ; SI-NEXT: s_mov_b32 s8, s0 ; SI-NEXT: s_mov_b32 s9, s1 ; SI-NEXT: s_mov_b32 s4, s2 @@ -606,6 +833,42 @@ define amdgpu_kernel void @v_usubo_v2i32(ptr addrspace(1) %out, ptr addrspace(1) ; GFX9-NEXT: global_store_dwordx2 v4, v[0:1], s[8:9] ; GFX9-NEXT: global_store_dwordx2 v4, v[2:3], s[10:11] ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: v_usubo_v2i32: +; GFX10: ; %bb.0: +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v4, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: global_load_dwordx2 v[0:1], v4, s[12:13] +; GFX10-NEXT: global_load_dwordx2 v[2:3], v4, s[14:15] +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: v_sub_co_u32 v1, s0, v1, v3 +; GFX10-NEXT: v_cndmask_b32_e64 v3, 0, 1, s0 +; GFX10-NEXT: v_sub_co_u32 v0, s0, v0, v2 +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, s0 +; GFX10-NEXT: global_store_dwordx2 v4, v[0:1], s[8:9] +; GFX10-NEXT: global_store_dwordx2 v4, v[2:3], s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: v_usubo_v2i32: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v4, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b64 v[0:1], v4, s[4:5] +; GFX11-NEXT: global_load_b64 v[2:3], v4, s[6:7] +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_sub_co_u32 v1, s4, v1, v3 +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_1) +; GFX11-NEXT: v_cndmask_b32_e64 v3, 0, 1, s4 +; GFX11-NEXT: v_sub_co_u32 v0, s4, v0, v2 +; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, s4 +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b64 v4, v[0:1], s[0:1] +; GFX11-NEXT: global_store_b64 v4, v[2:3], s[2:3] +; GFX11-NEXT: s_endpgm %a = load <2 x i32>, ptr addrspace(1) %aptr, align 4 %b = load <2 x i32>, ptr addrspace(1) %bptr, align 4 %sadd = call { <2 x i32>, <2 x i1> } @llvm.usub.with.overflow.v2i32(<2 x i32> %a, <2 x i32> %b) nounwind @@ -623,26 +886,27 @@ define amdgpu_kernel void @s_usubo_clamp_bit(ptr addrspace(1) %out, ptr addrspac ; SI-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xd ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: v_mov_b32_e32 v0, s1 -; SI-NEXT: v_sub_i32_e32 v0, vcc, s0, v0 ; SI-NEXT: s_cmp_eq_u32 s0, s1 +; SI-NEXT: v_sub_i32_e32 v0, vcc, s0, v0 ; SI-NEXT: s_mov_b64 s[0:1], 0 ; SI-NEXT: s_cbranch_scc1 .LBB8_2 ; SI-NEXT: ; %bb.1: ; %if ; SI-NEXT: s_xor_b64 s[0:1], vcc, -1 ; SI-NEXT: .LBB8_2: ; %exit ; SI-NEXT: s_load_dwordx4 s[4:7], s[4:5], 0x9 -; SI-NEXT: s_mov_b32 s3, 0xf000 -; SI-NEXT: s_mov_b32 s2, -1 -; SI-NEXT: v_cndmask_b32_e64 v1, 0, 1, s[0:1] -; SI-NEXT: s_mov_b32 s10, s2 -; SI-NEXT: s_mov_b32 s11, s3 +; SI-NEXT: s_mov_b32 s11, 0xf000 +; SI-NEXT: s_mov_b32 s10, -1 ; SI-NEXT: s_waitcnt lgkmcnt(0) -; SI-NEXT: s_mov_b32 s0, s4 -; SI-NEXT: s_mov_b32 s1, s5 -; SI-NEXT: s_mov_b32 s8, s6 -; SI-NEXT: s_mov_b32 s9, s7 -; SI-NEXT: buffer_store_dword v0, off, s[0:3], 0 -; SI-NEXT: buffer_store_byte v1, off, s[8:11], 0 +; SI-NEXT: s_mov_b32 s8, s4 +; SI-NEXT: s_mov_b32 s9, s5 +; SI-NEXT: s_mov_b32 s4, s6 +; SI-NEXT: s_mov_b32 s5, s7 +; SI-NEXT: s_mov_b32 s6, s10 +; SI-NEXT: s_mov_b32 s7, s11 +; SI-NEXT: buffer_store_dword v0, off, s[8:11], 0 +; SI-NEXT: s_waitcnt expcnt(0) +; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] +; SI-NEXT: buffer_store_byte v0, off, s[4:7], 0 ; SI-NEXT: s_endpgm ; ; VI-LABEL: s_usubo_clamp_bit: @@ -687,6 +951,45 @@ define amdgpu_kernel void @s_usubo_clamp_bit(ptr addrspace(1) %out, ptr addrspac ; GFX9-NEXT: global_store_dword v1, v0, s[8:9] ; GFX9-NEXT: global_store_byte v1, v2, s[10:11] ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: s_usubo_clamp_bit: +; GFX10: ; %bb.0: ; %entry +; GFX10-NEXT: s_load_dwordx2 s[2:3], s[4:5], 0x34 +; GFX10-NEXT: s_mov_b32 s0, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: v_sub_co_u32 v0, s1, s2, s3 +; GFX10-NEXT: s_cmp_eq_u32 s2, s3 +; GFX10-NEXT: s_cbranch_scc1 .LBB8_2 +; GFX10-NEXT: ; %bb.1: ; %if +; GFX10-NEXT: s_xor_b32 s0, s1, -1 +; GFX10-NEXT: .LBB8_2: ; %exit +; GFX10-NEXT: s_load_dwordx4 s[8:11], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v1, 0 +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, s0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: global_store_dword v1, v0, s[8:9] +; GFX10-NEXT: global_store_byte v1, v2, s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: s_usubo_clamp_bit: +; GFX11: ; %bb.0: ; %entry +; GFX11-NEXT: s_load_b64 s[2:3], s[4:5], 0x34 +; GFX11-NEXT: s_mov_b32 s0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: v_sub_co_u32 v0, s1, s2, s3 +; GFX11-NEXT: s_cmp_eq_u32 s2, s3 +; GFX11-NEXT: s_cbranch_scc1 .LBB8_2 +; GFX11-NEXT: ; %bb.1: ; %if +; GFX11-NEXT: s_xor_b32 s0, s1, -1 +; GFX11-NEXT: .LBB8_2: ; %exit +; GFX11-NEXT: s_load_b128 s[4:7], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v1, 0 +; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, s0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b32 v1, v0, s[4:5] +; GFX11-NEXT: global_store_b8 v1, v2, s[6:7] +; GFX11-NEXT: s_endpgm entry: %usub = call { i32, i1 } @llvm.usub.with.overflow.i32(i32 %a, i32 %b) %val = extractvalue { i32, i1 } %usub, 0 @@ -712,19 +1015,19 @@ define amdgpu_kernel void @v_usubo_clamp_bit(ptr addrspace(1) %out, ptr addrspac ; SI-NEXT: s_load_dwordx8 s[4:11], s[4:5], 0x9 ; SI-NEXT: s_mov_b32 s3, 0xf000 ; SI-NEXT: s_mov_b32 s2, -1 -; SI-NEXT: s_mov_b32 s14, s2 -; SI-NEXT: s_mov_b32 s15, s3 ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_mov_b32 s0, s8 ; SI-NEXT: s_mov_b32 s1, s9 -; SI-NEXT: s_mov_b32 s12, s10 -; SI-NEXT: s_mov_b32 s13, s11 +; SI-NEXT: s_mov_b32 s8, s10 +; SI-NEXT: s_mov_b32 s9, s11 +; SI-NEXT: s_mov_b32 s10, s2 +; SI-NEXT: s_mov_b32 s11, s3 ; SI-NEXT: buffer_load_dword v1, off, s[0:3], 0 -; SI-NEXT: buffer_load_dword v2, off, s[12:15], 0 +; SI-NEXT: buffer_load_dword v2, off, s[8:11], 0 +; SI-NEXT: s_mov_b64 s[8:9], 0 ; SI-NEXT: s_waitcnt vmcnt(0) -; SI-NEXT: v_sub_i32_e64 v0, s[0:1], v1, v2 ; SI-NEXT: v_cmp_eq_u32_e32 vcc, v1, v2 -; SI-NEXT: s_mov_b64 s[8:9], 0 +; SI-NEXT: v_sub_i32_e64 v0, s[0:1], v1, v2 ; SI-NEXT: s_cbranch_vccnz .LBB9_2 ; SI-NEXT: ; %bb.1: ; %if ; SI-NEXT: s_xor_b64 s[8:9], s[0:1], -1 @@ -787,6 +1090,50 @@ define amdgpu_kernel void @v_usubo_clamp_bit(ptr addrspace(1) %out, ptr addrspac ; GFX9-NEXT: v_cndmask_b32_e64 v1, 0, 1, s[2:3] ; GFX9-NEXT: global_store_byte v0, v1, s[10:11] ; GFX9-NEXT: s_endpgm +; +; GFX10-LABEL: v_usubo_clamp_bit: +; GFX10: ; %bb.0: ; %entry +; GFX10-NEXT: s_load_dwordx8 s[8:15], s[4:5], 0x24 +; GFX10-NEXT: v_mov_b32_e32 v0, 0 +; GFX10-NEXT: s_mov_b32 s0, 0 +; GFX10-NEXT: s_waitcnt lgkmcnt(0) +; GFX10-NEXT: s_clause 0x1 +; GFX10-NEXT: global_load_dword v1, v0, s[12:13] +; GFX10-NEXT: global_load_dword v2, v0, s[14:15] +; GFX10-NEXT: s_waitcnt vmcnt(0) +; GFX10-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v2 +; GFX10-NEXT: v_sub_co_u32 v1, s1, v1, v2 +; GFX10-NEXT: s_cbranch_vccnz .LBB9_2 +; GFX10-NEXT: ; %bb.1: ; %if +; GFX10-NEXT: s_xor_b32 s0, s1, -1 +; GFX10-NEXT: .LBB9_2: ; %exit +; GFX10-NEXT: v_cndmask_b32_e64 v2, 0, 1, s0 +; GFX10-NEXT: global_store_dword v0, v1, s[8:9] +; GFX10-NEXT: global_store_byte v0, v2, s[10:11] +; GFX10-NEXT: s_endpgm +; +; GFX11-LABEL: v_usubo_clamp_bit: +; GFX11: ; %bb.0: ; %entry +; GFX11-NEXT: s_load_b256 s[0:7], s[4:5], 0x24 +; GFX11-NEXT: v_mov_b32_e32 v0, 0 +; GFX11-NEXT: s_waitcnt lgkmcnt(0) +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_load_b32 v1, v0, s[4:5] +; GFX11-NEXT: global_load_b32 v2, v0, s[6:7] +; GFX11-NEXT: s_mov_b32 s4, 0 +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: v_cmp_eq_u32_e32 vcc_lo, v1, v2 +; GFX11-NEXT: v_sub_co_u32 v1, s5, v1, v2 +; GFX11-NEXT: s_cbranch_vccnz .LBB9_2 +; GFX11-NEXT: ; %bb.1: ; %if +; GFX11-NEXT: s_xor_b32 s4, s5, -1 +; GFX11-NEXT: .LBB9_2: ; %exit +; GFX11-NEXT: s_delay_alu instid0(SALU_CYCLE_1) +; GFX11-NEXT: v_cndmask_b32_e64 v2, 0, 1, s4 +; GFX11-NEXT: s_clause 0x1 +; GFX11-NEXT: global_store_b32 v0, v1, s[0:1] +; GFX11-NEXT: global_store_b8 v0, v2, s[2:3] +; GFX11-NEXT: s_endpgm entry: %tid = call i32 @llvm.amdgcn.workitem.id.x() %tid.ext = sext i32 %tid to i64 diff --git a/llvm/test/CodeGen/AMDGPU/wwm-regalloc-error.ll b/llvm/test/CodeGen/AMDGPU/wwm-regalloc-error.ll index 145f1e4..ff18b32 100644 --- a/llvm/test/CodeGen/AMDGPU/wwm-regalloc-error.ll +++ b/llvm/test/CodeGen/AMDGPU/wwm-regalloc-error.ll @@ -2,7 +2,7 @@ ; A negative test to capture the expected error when the VGPRs are insufficient for wwm-regalloc. -; CHECK: error: can't find enough VGPRs for wwm-regalloc +; CHECK: error: cannot find enough VGPRs for wwm-regalloc define amdgpu_kernel void @test(i32 %in) { entry: diff --git a/llvm/test/CodeGen/ARM/preferred-function-alignment.ll b/llvm/test/CodeGen/ARM/preferred-function-alignment.ll index f3a227c..2fc6790 100644 --- a/llvm/test/CodeGen/ARM/preferred-function-alignment.ll +++ b/llvm/test/CodeGen/ARM/preferred-function-alignment.ll @@ -22,3 +22,11 @@ define void @test() { define void @test_optsize() optsize { ret void } + +; CHECK-LABEL: test_minsize +; ALIGN-CS-16: .p2align 1 +; ALIGN-CS-32: .p2align 2 + +define void @test_minsize() minsize { + ret void +} diff --git a/llvm/test/CodeGen/Hexagon/hvx-reuse-fi-base.ll b/llvm/test/CodeGen/Hexagon/hvx-reuse-fi-base.ll index 16cc1f3..e5a6aa4 100644 --- a/llvm/test/CodeGen/Hexagon/hvx-reuse-fi-base.ll +++ b/llvm/test/CodeGen/Hexagon/hvx-reuse-fi-base.ll @@ -183,7 +183,7 @@ b0: %v11 = call <64 x i32> @llvm.hexagon.V6.vaddubh.128B(<32 x i32> %v10, <32 x i32> undef) %v12 = call <64 x i32> @llvm.hexagon.V6.vrmpyubi.128B(<64 x i32> %v11, i32 2147483647, i32 1) store <64 x i32> %v12, ptr @g0, align 128 - call void (ptr, ...) @f1(ptr @g3) #2 + call void (ptr, ...) @f1(ptr @g3) #3 %v13 = call <32 x i32> @llvm.hexagon.V6.lvsplatw.128B(i32 2) %v14 = call <64 x i32> @llvm.hexagon.V6.vaddubh.128B(<32 x i32> undef, <32 x i32> %v13) %v15 = call <64 x i32> @llvm.hexagon.V6.vrmpyubi.128B(<64 x i32> %v14, i32 -2147483648, i32 1) @@ -193,7 +193,7 @@ b0: %v17 = call <64 x i32> @llvm.hexagon.V6.vaddubh.128B(<32 x i32> undef, <32 x i32> %v16) %v18 = call <64 x i32> @llvm.hexagon.V6.vrmpyubi.128B(<64 x i32> %v17, i32 0, i32 1) store <64 x i32> %v18, ptr @g0, align 128 - call void @f0() #2 + call void @f0() #3 %v19 = call <32 x i32> @llvm.hexagon.V6.lvsplatw.128B(i32 1) %v20 = call <32 x i32> @llvm.hexagon.V6.lvsplatw.128B(i32 2) %v21 = call <64 x i32> @llvm.hexagon.V6.vaddubh.128B(<32 x i32> %v19, <32 x i32> %v20) @@ -205,3 +205,4 @@ b0: attributes #0 = { nounwind "use-soft-float"="false" "target-cpu"="hexagonv66" "target-features"="+hvxv66,+hvx-length128b" } attributes #1 = { nounwind readnone } attributes #2 = { nounwind optsize } +attributes #3 = { nounwind minsize } diff --git a/llvm/test/CodeGen/Mips/nan_lowering.ll b/llvm/test/CodeGen/Mips/nan_lowering.ll new file mode 100644 index 0000000..2a11278 --- /dev/null +++ b/llvm/test/CodeGen/Mips/nan_lowering.ll @@ -0,0 +1,25 @@ +; RUN: llc -mtriple=mips-linux-gnu -mattr=-nan2008 < %s | FileCheck %s +; RUN: llc -mtriple=mips-linux-gnu -mattr=+nan2008 < %s | FileCheck %s + +; Make sure that lowering does not corrupt the value of NaN values, +; regardless of what the NaN mode is. + +define float @test1() { +; CHECK: .4byte 0x7fc00000 + ret float bitcast (i32 u0x7fc00000 to float) +} + +define float @test2() { +; CHECK: .4byte 0x7fc00001 + ret float bitcast (i32 u0x7fc00001 to float) +} + +define float @test3() { +; CHECK: .4byte 0x7f800000 + ret float bitcast (i32 u0x7f800000 to float) +} + +define float @test4() { +; CHECK: .4byte 0x7f800001 + ret float bitcast (i32 u0x7f800001 to float) +} diff --git a/llvm/test/CodeGen/Mips/qnan.ll b/llvm/test/CodeGen/Mips/qnan.ll deleted file mode 100644 index e5b4aa1..0000000 --- a/llvm/test/CodeGen/Mips/qnan.ll +++ /dev/null @@ -1,14 +0,0 @@ -; RUN: llc -O3 -mcpu=mips32r2 -mtriple=mips-linux-gnu < %s -o - | FileCheck %s -check-prefixes=MIPS_Legacy -; RUN: llc -O3 -mcpu=mips32r2 -mtriple=mips-linux-gnu -mattr=+nan2008 < %s -o - | FileCheck %s -check-prefixes=MIPS_NaN2008 - -define dso_local float @nan(float noundef %a, float noundef %b) local_unnamed_addr #0 { -; MIPS_Legacy: $CPI0_0: -; MIPS_Legacy-NEXT: .4byte 0x7fa00000 # float NaN - -; MIPS_NaN2008: $CPI0_0: -; MIPS_NaN2008-NEXT: .4byte 0x7fc00000 # float NaN - -entry: - %0 = tail call float @llvm.minimum.f32(float %a, float %b) - ret float %0 -} diff --git a/llvm/test/CodeGen/PowerPC/froundeven-legalization.ll b/llvm/test/CodeGen/PowerPC/froundeven-legalization.ll new file mode 100644 index 0000000..238e200 --- /dev/null +++ b/llvm/test/CodeGen/PowerPC/froundeven-legalization.ll @@ -0,0 +1,111 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 +; RUN: llc -mtriple=powerpc64le < %s | FileCheck %s + +define void @test(ptr %p1, ptr %p2) nounwind { +; CHECK-LABEL: test: +; CHECK: # %bb.0: +; CHECK-NEXT: mflr 0 +; CHECK-NEXT: stdu 1, -224(1) +; CHECK-NEXT: li 5, 48 +; CHECK-NEXT: std 0, 240(1) +; CHECK-NEXT: std 27, 184(1) # 8-byte Folded Spill +; CHECK-NEXT: li 27, 16 +; CHECK-NEXT: std 28, 192(1) # 8-byte Folded Spill +; CHECK-NEXT: std 29, 200(1) # 8-byte Folded Spill +; CHECK-NEXT: li 29, 32 +; CHECK-NEXT: li 28, 48 +; CHECK-NEXT: stxvd2x 56, 1, 5 # 16-byte Folded Spill +; CHECK-NEXT: li 5, 64 +; CHECK-NEXT: std 30, 208(1) # 8-byte Folded Spill +; CHECK-NEXT: mr 30, 4 +; CHECK-NEXT: stxvd2x 57, 1, 5 # 16-byte Folded Spill +; CHECK-NEXT: li 5, 80 +; CHECK-NEXT: stxvd2x 58, 1, 5 # 16-byte Folded Spill +; CHECK-NEXT: li 5, 96 +; CHECK-NEXT: lxvd2x 58, 0, 3 +; CHECK-NEXT: stxvd2x 59, 1, 5 # 16-byte Folded Spill +; CHECK-NEXT: li 5, 112 +; CHECK-NEXT: lxvd2x 59, 3, 27 +; CHECK-NEXT: stxvd2x 60, 1, 5 # 16-byte Folded Spill +; CHECK-NEXT: li 5, 128 +; CHECK-NEXT: stxvd2x 61, 1, 5 # 16-byte Folded Spill +; CHECK-NEXT: li 5, 144 +; CHECK-NEXT: stxvd2x 62, 1, 5 # 16-byte Folded Spill +; CHECK-NEXT: li 5, 160 +; CHECK-NEXT: lxvd2x 62, 3, 28 +; CHECK-NEXT: stxvd2x 63, 1, 5 # 16-byte Folded Spill +; CHECK-NEXT: lxvd2x 63, 3, 29 +; CHECK-NEXT: xxswapd 57, 58 +; CHECK-NEXT: xxswapd 1, 59 +; CHECK-NEXT: xxswapd 60, 62 +; CHECK-NEXT: xxswapd 61, 63 +; CHECK-NEXT: bl roundeven +; CHECK-NEXT: nop +; CHECK-NEXT: xxswapd 56, 1 +; CHECK-NEXT: xxlor 1, 59, 59 +; CHECK-NEXT: bl roundeven +; CHECK-NEXT: nop +; CHECK-NEXT: xxswapd 0, 1 +; CHECK-NEXT: xxlor 1, 60, 60 +; CHECK-NEXT: xxmrgld 59, 0, 56 +; CHECK-NEXT: bl roundeven +; CHECK-NEXT: nop +; CHECK-NEXT: xxswapd 60, 1 +; CHECK-NEXT: xxlor 1, 62, 62 +; CHECK-NEXT: bl roundeven +; CHECK-NEXT: nop +; CHECK-NEXT: xxswapd 0, 1 +; CHECK-NEXT: xxlor 1, 61, 61 +; CHECK-NEXT: xxmrgld 62, 0, 60 +; CHECK-NEXT: bl roundeven +; CHECK-NEXT: nop +; CHECK-NEXT: xxswapd 61, 1 +; CHECK-NEXT: xxlor 1, 63, 63 +; CHECK-NEXT: bl roundeven +; CHECK-NEXT: nop +; CHECK-NEXT: xxswapd 0, 1 +; CHECK-NEXT: xxlor 1, 57, 57 +; CHECK-NEXT: xxmrgld 63, 0, 61 +; CHECK-NEXT: bl roundeven +; CHECK-NEXT: nop +; CHECK-NEXT: xxswapd 61, 1 +; CHECK-NEXT: xxlor 1, 58, 58 +; CHECK-NEXT: bl roundeven +; CHECK-NEXT: nop +; CHECK-NEXT: li 3, 160 +; CHECK-NEXT: stxvd2x 63, 30, 29 +; CHECK-NEXT: xxswapd 0, 1 +; CHECK-NEXT: stxvd2x 62, 30, 28 +; CHECK-NEXT: stxvd2x 59, 30, 27 +; CHECK-NEXT: ld 29, 200(1) # 8-byte Folded Reload +; CHECK-NEXT: ld 28, 192(1) # 8-byte Folded Reload +; CHECK-NEXT: ld 27, 184(1) # 8-byte Folded Reload +; CHECK-NEXT: lxvd2x 63, 1, 3 # 16-byte Folded Reload +; CHECK-NEXT: li 3, 144 +; CHECK-NEXT: xxmrgld 0, 0, 61 +; CHECK-NEXT: lxvd2x 62, 1, 3 # 16-byte Folded Reload +; CHECK-NEXT: li 3, 128 +; CHECK-NEXT: stxvd2x 0, 0, 30 +; CHECK-NEXT: ld 30, 208(1) # 8-byte Folded Reload +; CHECK-NEXT: lxvd2x 61, 1, 3 # 16-byte Folded Reload +; CHECK-NEXT: li 3, 112 +; CHECK-NEXT: lxvd2x 60, 1, 3 # 16-byte Folded Reload +; CHECK-NEXT: li 3, 96 +; CHECK-NEXT: lxvd2x 59, 1, 3 # 16-byte Folded Reload +; CHECK-NEXT: li 3, 80 +; CHECK-NEXT: lxvd2x 58, 1, 3 # 16-byte Folded Reload +; CHECK-NEXT: li 3, 64 +; CHECK-NEXT: lxvd2x 57, 1, 3 # 16-byte Folded Reload +; CHECK-NEXT: li 3, 48 +; CHECK-NEXT: lxvd2x 56, 1, 3 # 16-byte Folded Reload +; CHECK-NEXT: addi 1, 1, 224 +; CHECK-NEXT: ld 0, 16(1) +; CHECK-NEXT: mtlr 0 +; CHECK-NEXT: blr + %v = load <8 x double>, ptr %p1, align 64 + %res = call <8 x double> @llvm.roundeven.v8f64(<8 x double> %v) + store <8 x double> %res, ptr %p2, align 64 + ret void +} + +declare <8 x double> @llvm.roundeven.v8f64(<8 x double>) diff --git a/llvm/test/CodeGen/X86/embed-bitcode.ll b/llvm/test/CodeGen/X86/embed-bitcode.ll index 0d66ba8..d4af954 100644 --- a/llvm/test/CodeGen/X86/embed-bitcode.ll +++ b/llvm/test/CodeGen/X86/embed-bitcode.ll @@ -1,10 +1,23 @@ ; RUN: llc -filetype=obj -mtriple=x86_64 %s -o %t ; RUN: llvm-readelf -S %t | FileCheck %s +; RUN: llc -filetype=obj -mtriple=x86_64-pc-windows-msvc %s -o %t +; RUN: llvm-readobj -S %t | FileCheck %s --check-prefix=COFF ; CHECK: .text PROGBITS 0000000000000000 [[#%x,OFF:]] 000000 00 AX 0 ; CHECK-NEXT: .llvmbc PROGBITS 0000000000000000 [[#%x,OFF:]] 000004 00 0 ; CHECK-NEXT: .llvmcmd PROGBITS 0000000000000000 [[#%x,OFF:]] 000005 00 0 +; COFF: Name: .llvmbc (2E 6C 6C 76 6D 62 63 00) +; COFF: Characteristics [ +; COFF-NEXT: IMAGE_SCN_ALIGN_1BYTES +; COFF-NEXT: IMAGE_SCN_MEM_DISCARDABLE +; COFF-NEXT: ] +; COFF: Name: .llvmcmd (2E 6C 6C 76 6D 63 6D 64) +; COFF: Characteristics [ +; COFF-NEXT: IMAGE_SCN_ALIGN_1BYTES +; COFF-NEXT: IMAGE_SCN_MEM_DISCARDABLE +; COFF-NEXT: ] + @llvm.embedded.module = private constant [4 x i8] c"BC\C0\DE", section ".llvmbc", align 1 @llvm.cmdline = private constant [5 x i8] c"-cc1\00", section ".llvmcmd", align 1 @llvm.compiler.used = appending global [2 x ptr] [ptr @llvm.embedded.module, ptr @llvm.cmdline], section "llvm.metadata" diff --git a/llvm/test/CodeGen/X86/late-tail-dup-computed-goto.mir b/llvm/test/CodeGen/X86/late-tail-dup-computed-goto.mir new file mode 100644 index 0000000..e272e7e --- /dev/null +++ b/llvm/test/CodeGen/X86/late-tail-dup-computed-goto.mir @@ -0,0 +1,128 @@ +# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 5 +# RUN: llc -mtriple=x86_64-unknown-linux-gnu -run-pass=tailduplication -tail-dup-pred-size=1 -tail-dup-succ-size=1 %s -o - | FileCheck %s +# +# Check that only the computed gotos are duplicated aggressively. +# +--- | + @computed_goto.dispatch = constant [5 x ptr] [ptr null, ptr blockaddress(@computed_goto, %bb1), ptr blockaddress(@computed_goto, %bb2), ptr blockaddress(@computed_goto, %bb3), ptr blockaddress(@computed_goto, %bb4)] + declare i64 @f0() + declare i64 @f1() + declare i64 @f2() + declare i64 @f3() + declare i64 @f4() + declare i64 @f5() + define void @computed_goto() { + start: + ret void + bb1: + ret void + bb2: + ret void + bb3: + ret void + bb4: + ret void + } + define void @jump_table() { ret void } + define void @jump_table_pic() { ret void } +... +--- +name: computed_goto +alignment: 1 +exposesReturnsTwice: false +legalized: false +regBankSelected: false +selected: false +failedISel: false +tracksRegLiveness: true +body: | + ; CHECK-LABEL: name: computed_goto + ; CHECK: bb.0: + ; CHECK-NEXT: successors: %bb.1(0x20000000), %bb.2(0x20000000), %bb.3(0x20000000), %bb.4(0x20000000) + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f0, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax + ; CHECK-NEXT: [[COPY:%[0-9]+]]:gr64 = COPY $rax + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:gr64_nosp = COPY [[COPY]] + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gr64_nosp = COPY [[COPY1]] + ; CHECK-NEXT: JMP64m $noreg, 8, [[COPY2]], @computed_goto.dispatch, $noreg + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: bb.1.bb1 (ir-block-address-taken %ir-block.bb1): + ; CHECK-NEXT: successors: %bb.1(0x20000000), %bb.2(0x20000000), %bb.3(0x20000000), %bb.4(0x20000000) + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f1, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax + ; CHECK-NEXT: [[COPY3:%[0-9]+]]:gr64 = COPY $rax + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:gr64_nosp = COPY [[COPY3]] + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gr64_nosp = COPY [[COPY1]] + ; CHECK-NEXT: JMP64m $noreg, 8, [[COPY2]], @computed_goto.dispatch, $noreg + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: bb.2.bb2 (ir-block-address-taken %ir-block.bb2): + ; CHECK-NEXT: successors: %bb.1(0x20000000), %bb.2(0x20000000), %bb.3(0x20000000), %bb.4(0x20000000) + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f2, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax + ; CHECK-NEXT: [[COPY4:%[0-9]+]]:gr64 = COPY $rax + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:gr64_nosp = COPY [[COPY4]] + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gr64_nosp = COPY [[COPY1]] + ; CHECK-NEXT: JMP64m $noreg, 8, [[COPY2]], @computed_goto.dispatch, $noreg + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: bb.3.bb3 (ir-block-address-taken %ir-block.bb3): + ; CHECK-NEXT: successors: %bb.1(0x20000000), %bb.2(0x20000000), %bb.3(0x20000000), %bb.4(0x20000000) + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f3, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax + ; CHECK-NEXT: [[COPY5:%[0-9]+]]:gr64 = COPY $rax + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:gr64_nosp = COPY [[COPY5]] + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gr64_nosp = COPY [[COPY1]] + ; CHECK-NEXT: JMP64m $noreg, 8, [[COPY2]], @computed_goto.dispatch, $noreg + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: bb.4.bb4 (ir-block-address-taken %ir-block.bb4): + ; CHECK-NEXT: successors: %bb.1(0x20000000), %bb.2(0x20000000), %bb.3(0x20000000), %bb.4(0x20000000) + ; CHECK-NEXT: {{ $}} + ; CHECK-NEXT: CALL64pcrel32 target-flags(x86-plt) @f4, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax + ; CHECK-NEXT: [[COPY6:%[0-9]+]]:gr64 = COPY $rax + ; CHECK-NEXT: [[COPY1:%[0-9]+]]:gr64_nosp = COPY [[COPY6]] + ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gr64_nosp = COPY [[COPY1]] + ; CHECK-NEXT: JMP64m $noreg, 8, [[COPY2]], @computed_goto.dispatch, $noreg + bb.0: + successors: %bb.5(0x80000000) + + CALL64pcrel32 target-flags(x86-plt) @f0, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax + %0:gr64 = COPY $rax + %6:gr64_nosp = COPY %0 + JMP_1 %bb.5 + + bb.1.bb1 (ir-block-address-taken %ir-block.bb1): + successors: %bb.5(0x80000000) + + CALL64pcrel32 target-flags(x86-plt) @f1, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax + %1:gr64 = COPY $rax + %6:gr64_nosp = COPY %1 + JMP_1 %bb.5 + + bb.2.bb2 (ir-block-address-taken %ir-block.bb2): + successors: %bb.5(0x80000000) + + CALL64pcrel32 target-flags(x86-plt) @f2, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax + %2:gr64 = COPY $rax + %6:gr64_nosp = COPY %2 + JMP_1 %bb.5 + + bb.3.bb3 (ir-block-address-taken %ir-block.bb3): + successors: %bb.5(0x80000000) + + CALL64pcrel32 target-flags(x86-plt) @f3, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax + %3:gr64 = COPY $rax + %6:gr64_nosp = COPY %3 + JMP_1 %bb.5 + + bb.4.bb4 (ir-block-address-taken %ir-block.bb4): + successors: %bb.5(0x80000000) + + CALL64pcrel32 target-flags(x86-plt) @f4, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax + %4:gr64 = COPY $rax + %6:gr64_nosp = COPY %4 + + bb.5: + successors: %bb.1(0x20000000), %bb.2(0x20000000), %bb.3(0x20000000), %bb.4(0x20000000) + + %5:gr64_nosp = COPY %6 + JMP64m $noreg, 8, %5, @computed_goto.dispatch, $noreg +... diff --git a/llvm/test/DebugInfo/X86/branch-folder-dbg-after-end.mir b/llvm/test/DebugInfo/X86/branch-folder-dbg-after-end.mir new file mode 100644 index 0000000..743851c --- /dev/null +++ b/llvm/test/DebugInfo/X86/branch-folder-dbg-after-end.mir @@ -0,0 +1,108 @@ +# RUN: llc %s --start-before=branch-folder --stop-after=branch-folder -o - \ +# RUN: | FileCheck %s --implicit-check-not=DBG_PHI + +## Common instructions are hoisted. Check that trailing debug instructions in +## the range are also hoisted, and don't cause a crash. +## +## Note the MIR doesn't match the IR as it's modified from: +## /home/och/dev/llvm-project/llvm/test/DebugInfo/X86/branch-folder-dbg.mir + +# CHECK: bb.0 +# CHECK: CALL64pcrel32 @f, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax +## --- Start splice from bb.2.if.else (and debug instructions from bb.1.if.then) --- +# CHECK-NEXT: $edi = MOV32r0 implicit-def dead $eflags, debug-location !DILocation(line: 0, scope: ![[#]]) +# CHECK-NEXT: DBG_VALUE $noreg, $noreg, ![[#]], !DIExpression(), debug-location +# CHECK-NEXT: DBG_VALUE $noreg, $noreg, ![[#]], !DIExpression(), debug-location +## --- End splice ------------------------------------------------------------------ +# CHECK-NEXT: TEST64rr killed renamable $rax, renamable $rax, implicit-def $eflags +# CHECK-NEXT: JCC_1 %bb.2, 8, implicit $eflags +# CHECK: bb.1 + +--- | + target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128" + target triple = "x86_64-unknown-linux-gnu" + + declare dso_local noundef i64 @f() local_unnamed_addr + + define dso_local noundef i32 @g() local_unnamed_addr !dbg !7 { + %call = tail call noundef i64 @f() + %cmp1 = icmp sgt i64 0, %call + %conv2 = trunc i64 0 to i32 + br i1 %cmp1, label %if.then, label %if.else + + if.then: ; preds = %0 + tail call void @_Z3fooii(i32 noundef %conv2, i32 noundef 0), !dbg !14 + br label %if.end, !dbg !15 + + if.else: ; preds = %0 + tail call void @_Z3barii(i32 noundef %conv2, i32 noundef 1), !dbg !16 + br label %if.end, !dbg !17 + + if.end: ; preds = %if.else, %if.then + ret i32 2 + } + + declare void @_Z3fooii(i32 noundef, i32 noundef) local_unnamed_addr + + declare void @_Z3barii(i32 noundef, i32 noundef) local_unnamed_addr + + !llvm.module.flags = !{!0, !1} + !llvm.ident = !{!2} + !llvm.dbg.cu = !{!3} + !llvm.debugify = !{!5, !6} + + !0 = !{i32 7, !"Dwarf Version", i32 5} + !1 = !{i32 2, !"Debug Info Version", i32 3} + !2 = !{!"clang version 21.0.0"} + !3 = distinct !DICompileUnit(language: DW_LANG_C, file: !4, producer: "debugify", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug) + !4 = !DIFile(filename: "test.nodbg.ll", directory: "/") + !5 = !{i32 15} + !6 = !{i32 7} + !7 = distinct !DISubprogram(name: "g", linkageName: "g", scope: null, file: !4, line: 1, type: !8, scopeLine: 1, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !3, retainedNodes: !10) + !8 = !DISubroutineType(types: !9) + !9 = !{} + !10 = !{!11} + !11 = !DILocalVariable(name: "1", scope: !7, file: !4, line: 3, type: !12) + !12 = !DIBasicType(name: "ty64", size: 64, encoding: DW_ATE_unsigned) + !13 = !DILocation(line: 3, column: 1, scope: !7) + !14 = !DILocation(line: 9, column: 1, scope: !7) + !15 = !DILocation(line: 10, column: 1, scope: !7) + !16 = !DILocation(line: 11, column: 1, scope: !7) + !17 = !DILocation(line: 12, column: 1, scope: !7) +... +--- +name: g +tracksRegLiveness: true +isSSA: false +body: | + bb.0 (%ir-block.0): + successors: %bb.1(0x40000000), %bb.2(0x40000000) + + frame-setup PUSH64r undef $rax, implicit-def $rsp, implicit $rsp + frame-setup CFI_INSTRUCTION def_cfa_offset 16 + CALL64pcrel32 @f, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax + TEST64rr killed renamable $rax, renamable $rax, implicit-def $eflags + JCC_1 %bb.2, 9, implicit killed $eflags + JMP_1 %bb.1 + + bb.1.if.then: + successors: %bb.3(0x80000000) + + $edi = MOV32r0 implicit-def dead $eflags, debug-location !14 + DBG_VALUE $edi, $noreg, !11, !DIExpression(), debug-location !13 + + bb.3.if.end: + $eax = MOV32ri 2 + $rcx = frame-destroy POP64r implicit-def $rsp, implicit $rsp + frame-destroy CFI_INSTRUCTION def_cfa_offset 8 + RET 0, $eax + + bb.2.if.else: + successors: %bb.3(0x80000000) + + $edi = MOV32r0 implicit-def dead $eflags, debug-location !16 + DBG_VALUE $edi, $noreg, !11, !DIExpression(), debug-location !13 + CALL64pcrel32 target-flags(x86-plt) @_Z3barii, csr_64, implicit $rsp, implicit $ssp, implicit killed $edi, implicit killed $edi, implicit-def $rsp, implicit-def $ssp, debug-location !16 + JMP_1 %bb.3, debug-location !15 + +... diff --git a/llvm/test/DebugInfo/X86/branch-folder-dbg.mir b/llvm/test/DebugInfo/X86/branch-folder-dbg.mir index 5c38fd2..7832598 100644 --- a/llvm/test/DebugInfo/X86/branch-folder-dbg.mir +++ b/llvm/test/DebugInfo/X86/branch-folder-dbg.mir @@ -1,16 +1,24 @@ -# RUN: llc %s --start-before=branch-folder --stop-after=branch-folder -o - | FileCheck %s +# RUN: llc %s --start-before=branch-folder --stop-after=branch-folder -o - \ +# RUN: | FileCheck %s --implicit-check-not=DBG_PHI ## Check that common instructions hoisted from `if.then` and `if.else` into -## common pred `entry` get merged debug locations. - -## FIXME: The debug instructions handling here is wrong. +## common pred `entry` get merged debug locations. The debug instructions from +## both branches should get hoisted and killed. +## +## The MIR debug instructions have been modified by hand in order to check they +## can be killed. +## +## Check DBG_PHIs are deleted rather than hoisted (implicit-check-not). # CHECK: bb.0 # CHECK: CALL64pcrel32 @f, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax -## --- Start splice from bb.2.if.else --- -# CHECK-NEXT: DBG_VALUE 2, $noreg, ![[#]], !DIExpression(), debug-location ![[#]] -# CHECK-NEXT: $edi = MOV32r0 implicit-def dead $eflags, debug-location !DILocation(line: 0, scope: ![[#]]) -## --- End splice -------------- +## --- Start splice from bb.2.if.else (and debug instructions from bb.1.if.then) --- +# CHECK-NEXT: DBG_VALUE $noreg, $noreg, ![[#]], !DIExpression(), debug-location ![[#]] +# CHECK-NEXT: DBG_VALUE $noreg, $noreg, ![[#]], !DIExpression(), debug-location ![[#]] +# CHECK-NEXT: $edi = MOV32r0 implicit-def dead $eflags, debug-instr-number 2, debug-location !DILocation(line: 0, scope: ![[#]]) +# CHECK-NEXT: DBG_VALUE $noreg, $noreg, ![[#]], !DIExpression(DW_OP_LLVM_arg, 0), debug-location ![[#]] +# CHECK-NEXT: DBG_VALUE $noreg, $noreg, ![[#]], !DIExpression(DW_OP_LLVM_arg, 0), debug-location ![[#]] +## --- End splice ------------------------------------------------------------------ # CHECK-NEXT: TEST64rr killed renamable $rax, renamable $rax, implicit-def $eflags # CHECK-NEXT: JCC_1 %bb.2, 9, implicit killed $eflags # CHECK: bb.1 @@ -73,6 +81,8 @@ ... --- name: g +tracksRegLiveness: true +isSSA: false body: | bb.0 (%ir-block.0): successors: %bb.1(0x40000000), %bb.2(0x40000000) @@ -87,21 +97,23 @@ body: | bb.1.if.then: successors: %bb.3(0x80000000) - DBG_VALUE 0, $noreg, !11, !DIExpression(), debug-location !13 - $edi = MOV32r0 implicit-def dead $eflags, debug-location !14 + DBG_PHI $esp, 3 + DBG_VALUE $esi, $noreg, !11, !DIExpression(), debug-location !13 + $edi = MOV32r0 implicit-def dead $eflags, debug-instr-number 1, debug-location !14 + DBG_INSTR_REF !11, !DIExpression(DW_OP_LLVM_arg, 0), dbg-instr-ref(1, 0), debug-location !13 $esi = MOV32r0 implicit-def dead $eflags, debug-location !14 CALL64pcrel32 target-flags(x86-plt) @_Z3fooii, csr_64, implicit $rsp, implicit $ssp, implicit killed $edi, implicit killed $esi, implicit-def $rsp, implicit-def $ssp, debug-location !14 - DBG_VALUE 1, $noreg, !11, !DIExpression(), debug-location !13 JMP_1 %bb.3, debug-location !15 bb.2.if.else: successors: %bb.3(0x80000000) - DBG_VALUE 2, $noreg, !11, !DIExpression(), debug-location !13 - $edi = MOV32r0 implicit-def dead $eflags, debug-location !16 + DBG_PHI $esp, 4 + DBG_VALUE $esp, $noreg, !11, !DIExpression(), debug-location !13 + $edi = MOV32r0 implicit-def dead $eflags, debug-instr-number 2, debug-location !16 + DBG_INSTR_REF !11, !DIExpression(DW_OP_LLVM_arg, 0), dbg-instr-ref(2, 0), debug-location !13 $esi = MOV32ri 1, debug-location !16 CALL64pcrel32 target-flags(x86-plt) @_Z3barii, csr_64, implicit $rsp, implicit $ssp, implicit killed $edi, implicit killed $esi, implicit-def $rsp, implicit-def $ssp, debug-location !16 - DBG_VALUE 3, $noreg, !11, !DIExpression(), debug-location !13 bb.3.if.end: $eax = MOV32ri 2 diff --git a/llvm/test/Transforms/FunctionAttrs/noalias.ll b/llvm/test/Transforms/FunctionAttrs/noalias.ll new file mode 100644 index 0000000..8beb6fe --- /dev/null +++ b/llvm/test/Transforms/FunctionAttrs/noalias.ll @@ -0,0 +1,245 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5 +; RUN: opt -S -passes=function-attrs < %s | FileCheck %s + +declare noalias ptr @malloc(i64 %size) +declare ptr @not_malloc(i64 %size) +declare void @capture(ptr) + +@g = external global i8 + +define ptr @return_malloc(i64 %size) { +; CHECK-LABEL: define noalias ptr @return_malloc( +; CHECK-SAME: i64 [[SIZE:%.*]]) { +; CHECK-NEXT: [[A:%.*]] = call ptr @malloc(i64 [[SIZE]]) +; CHECK-NEXT: ret ptr [[A]] +; + %a = call ptr @malloc(i64 %size) + ret ptr %a +} + +define ptr @return_not_malloc(i64 %size) { +; CHECK-LABEL: define ptr @return_not_malloc( +; CHECK-SAME: i64 [[SIZE:%.*]]) { +; CHECK-NEXT: [[A:%.*]] = call ptr @not_malloc(i64 [[SIZE]]) +; CHECK-NEXT: ret ptr [[A]] +; + %a = call ptr @not_malloc(i64 %size) + ret ptr %a +} + +define ptr @return_null() { +; CHECK-LABEL: define noalias noundef ptr @return_null( +; CHECK-SAME: ) #[[ATTR0:[0-9]+]] { +; CHECK-NEXT: ret ptr null +; + ret ptr null +} + +define ptr @return_poison() { +; CHECK-LABEL: define noalias ptr @return_poison( +; CHECK-SAME: ) #[[ATTR0]] { +; CHECK-NEXT: ret ptr poison +; + ret ptr poison +} + +define ptr @return_alloca() { +; CHECK-LABEL: define noalias noundef nonnull ptr @return_alloca( +; CHECK-SAME: ) #[[ATTR0]] { +; CHECK-NEXT: [[A:%.*]] = alloca i8, align 1 +; CHECK-NEXT: ret ptr [[A]] +; + %a = alloca i8 + ret ptr %a +} + +; noalias arg does not imply noalias return +define ptr @return_noalias_arg(ptr noalias %arg) { +; CHECK-LABEL: define ptr @return_noalias_arg( +; CHECK-SAME: ptr noalias readnone returned captures(ret: address, provenance) [[ARG:%.*]]) #[[ATTR0]] { +; CHECK-NEXT: ret ptr [[ARG]] +; + ret ptr %arg +} + +define ptr @return_global() { +; CHECK-LABEL: define noundef nonnull ptr @return_global( +; CHECK-SAME: ) #[[ATTR0]] { +; CHECK-NEXT: ret ptr @g +; + ret ptr @g +} + +define ptr @no_return() { +; CHECK-LABEL: define noalias noundef nonnull ptr @no_return( +; CHECK-SAME: ) #[[ATTR1:[0-9]+]] { +; CHECK-NEXT: unreachable +; + unreachable +} + +define ptr @return_multiple(i1 %c, i64 %size) { +; CHECK-LABEL: define noalias ptr @return_multiple( +; CHECK-SAME: i1 [[C:%.*]], i64 [[SIZE:%.*]]) { +; CHECK-NEXT: br i1 [[C]], label %[[IF:.*]], label %[[ELSE:.*]] +; CHECK: [[IF]]: +; CHECK-NEXT: [[A:%.*]] = call ptr @malloc(i64 [[SIZE]]) +; CHECK-NEXT: ret ptr [[A]] +; CHECK: [[ELSE]]: +; CHECK-NEXT: [[B:%.*]] = call ptr @malloc(i64 [[SIZE]]) +; CHECK-NEXT: ret ptr [[B]] +; +br i1 %c, label %if, label %else + +if: + %a = call ptr @malloc(i64 %size) + ret ptr %a + +else: + %b = call ptr @malloc(i64 %size) + ret ptr %b +} + +define ptr @return_select(i1 %c, i64 %size) { +; CHECK-LABEL: define noalias ptr @return_select( +; CHECK-SAME: i1 [[C:%.*]], i64 [[SIZE:%.*]]) { +; CHECK-NEXT: [[A:%.*]] = call ptr @malloc(i64 [[SIZE]]) +; CHECK-NEXT: [[SEL:%.*]] = select i1 [[C]], ptr [[A]], ptr null +; CHECK-NEXT: ret ptr [[SEL]] +; + %a = call ptr @malloc(i64 %size) + %sel = select i1 %c, ptr %a, ptr null + ret ptr %sel +} + +define ptr @return_phi(i1 %c, i64 %size) { +; CHECK-LABEL: define noalias ptr @return_phi( +; CHECK-SAME: i1 [[C:%.*]], i64 [[SIZE:%.*]]) { +; CHECK-NEXT: br i1 [[C]], label %[[IF:.*]], label %[[ELSE:.*]] +; CHECK: [[IF]]: +; CHECK-NEXT: [[A:%.*]] = call ptr @malloc(i64 [[SIZE]]) +; CHECK-NEXT: br label %[[JOIN:.*]] +; CHECK: [[ELSE]]: +; CHECK-NEXT: br label %[[JOIN]] +; CHECK: [[JOIN]]: +; CHECK-NEXT: [[PHI:%.*]] = phi ptr [ [[A]], %[[IF]] ], [ null, %[[ELSE]] ] +; CHECK-NEXT: ret ptr [[PHI]] +; + br i1 %c, label %if, label %else + +if: + %a = call ptr @malloc(i64 %size) + br label %join + +else: + br label %join + +join: + %phi = phi ptr [ %a, %if ], [ null, %else ] + ret ptr %phi +} + +define ptr @return_phi_wrong(i1 %c, i64 %size) { +; CHECK-LABEL: define ptr @return_phi_wrong( +; CHECK-SAME: i1 [[C:%.*]], i64 [[SIZE:%.*]]) { +; CHECK-NEXT: br i1 [[C]], label %[[IF:.*]], label %[[ELSE:.*]] +; CHECK: [[IF]]: +; CHECK-NEXT: [[A:%.*]] = call ptr @malloc(i64 [[SIZE]]) +; CHECK-NEXT: br label %[[JOIN:.*]] +; CHECK: [[ELSE]]: +; CHECK-NEXT: [[B:%.*]] = call ptr @not_malloc(i64 [[SIZE]]) +; CHECK-NEXT: br label %[[JOIN]] +; CHECK: [[JOIN]]: +; CHECK-NEXT: [[PHI:%.*]] = phi ptr [ [[A]], %[[IF]] ], [ [[B]], %[[ELSE]] ] +; CHECK-NEXT: ret ptr [[PHI]] +; + br i1 %c, label %if, label %else + +if: + %a = call ptr @malloc(i64 %size) + br label %join + +else: + %b = call ptr @not_malloc(i64 %size) + br label %join + +join: + %phi = phi ptr [ %a, %if ], [ %b, %else ] + ret ptr %phi +} + +define ptr @return_malloc_with_store(i64 %size) { +; CHECK-LABEL: define noalias noundef ptr @return_malloc_with_store( +; CHECK-SAME: i64 [[SIZE:%.*]]) { +; CHECK-NEXT: [[A:%.*]] = call ptr @malloc(i64 [[SIZE]]) +; CHECK-NEXT: store i8 0, ptr [[A]], align 1 +; CHECK-NEXT: ret ptr [[A]] +; + %a = call ptr @malloc(i64 %size) + store i8 0, ptr %a + ret ptr %a +} + +define ptr @return_malloc_captured(i64 %size) { +; CHECK-LABEL: define ptr @return_malloc_captured( +; CHECK-SAME: i64 [[SIZE:%.*]]) { +; CHECK-NEXT: [[A:%.*]] = call ptr @malloc(i64 [[SIZE]]) +; CHECK-NEXT: call void @capture(ptr [[A]]) +; CHECK-NEXT: ret ptr [[A]] +; + %a = call ptr @malloc(i64 %size) + call void @capture(ptr %a) + ret ptr %a +} + +define ptr @scc1(i1 %c) { +; CHECK-LABEL: define noalias ptr @scc1( +; CHECK-SAME: i1 [[C:%.*]]) { +; CHECK-NEXT: br i1 [[C]], label %[[IF:.*]], label %[[ELSE:.*]] +; CHECK: [[IF]]: +; CHECK-NEXT: [[A:%.*]] = call ptr @malloc(i64 4) +; CHECK-NEXT: ret ptr [[A]] +; CHECK: [[ELSE]]: +; CHECK-NEXT: [[B:%.*]] = call ptr @scc2(i1 [[C]]) +; CHECK-NEXT: ret ptr [[B]] +; + br i1 %c, label %if, label %else + +if: + %a = call ptr @malloc(i64 4) + ret ptr %a + +else: + %b = call ptr @scc2(i1 %c) + ret ptr %b +} + +define ptr @scc2(i1 %c) { +; CHECK-LABEL: define noalias ptr @scc2( +; CHECK-SAME: i1 [[C:%.*]]) { +; CHECK-NEXT: [[A:%.*]] = call ptr @scc1(i1 [[C]]) +; CHECK-NEXT: ret ptr [[A]] +; + %a = call ptr @scc1(i1 %c) + ret ptr %a +} + +define ptr @return_unknown_call(ptr %fn) { +; CHECK-LABEL: define ptr @return_unknown_call( +; CHECK-SAME: ptr readonly captures(none) [[FN:%.*]]) { +; CHECK-NEXT: [[A:%.*]] = call ptr [[FN]]() +; CHECK-NEXT: ret ptr [[A]] +; + %a = call ptr %fn() + ret ptr %a +} + +define ptr @return_unknown_noalias_call(ptr %fn) { +; CHECK-LABEL: define ptr @return_unknown_noalias_call( +; CHECK-SAME: ptr readonly captures(none) [[FN:%.*]]) { +; CHECK-NEXT: [[A:%.*]] = call noalias ptr [[FN]]() +; CHECK-NEXT: ret ptr [[A]] +; + %a = call noalias ptr %fn() + ret ptr %a +} diff --git a/llvm/test/Transforms/FunctionAttrs/nofree.ll b/llvm/test/Transforms/FunctionAttrs/nofree.ll index 1671189..89f030d 100644 --- a/llvm/test/Transforms/FunctionAttrs/nofree.ll +++ b/llvm/test/Transforms/FunctionAttrs/nofree.ll @@ -156,6 +156,24 @@ entry: ret void } +define void @unknown_call(ptr %fn) { +; CHECK-LABEL: @unknown_call( +; CHECK-NEXT: call void [[FN:%.*]]() +; CHECK-NEXT: ret void +; + call void %fn() + ret void +} + +define void @unknown_nofree_call(ptr %fn) { +; CHECK-LABEL: @unknown_nofree_call( +; CHECK-NEXT: call void [[FN:%.*]]() #[[ATTR5]] +; CHECK-NEXT: ret void +; + call void %fn() nofree + ret void +} + declare void @_ZdaPv(ptr) local_unnamed_addr #4 attributes #0 = { uwtable } diff --git a/llvm/test/Transforms/FunctionAttrs/nonnull.ll b/llvm/test/Transforms/FunctionAttrs/nonnull.ll index 483b560..9b17ded 100644 --- a/llvm/test/Transforms/FunctionAttrs/nonnull.ll +++ b/llvm/test/Transforms/FunctionAttrs/nonnull.ll @@ -1396,5 +1396,35 @@ define ptr @pr91177_non_inbounds_gep(ptr nonnull %arg) { ret ptr %res } +define ptr @unknown_func(ptr %fn) { +; FNATTRS-LABEL: define ptr @unknown_func( +; FNATTRS-SAME: ptr readonly captures(none) [[FN:%.*]]) { +; FNATTRS-NEXT: [[RES:%.*]] = call ptr [[FN]]() +; FNATTRS-NEXT: ret ptr [[RES]] +; +; ATTRIBUTOR-LABEL: define ptr @unknown_func( +; ATTRIBUTOR-SAME: ptr nofree nonnull captures(none) [[FN:%.*]]) { +; ATTRIBUTOR-NEXT: [[RES:%.*]] = call ptr [[FN]]() +; ATTRIBUTOR-NEXT: ret ptr [[RES]] +; + %res = call ptr %fn() + ret ptr %res +} + +define ptr @unknown_nonnull_func(ptr %fn) { +; FNATTRS-LABEL: define ptr @unknown_nonnull_func( +; FNATTRS-SAME: ptr readonly captures(none) [[FN:%.*]]) { +; FNATTRS-NEXT: [[RES:%.*]] = call nonnull ptr [[FN]]() +; FNATTRS-NEXT: ret ptr [[RES]] +; +; ATTRIBUTOR-LABEL: define nonnull ptr @unknown_nonnull_func( +; ATTRIBUTOR-SAME: ptr nofree nonnull captures(none) [[FN:%.*]]) { +; ATTRIBUTOR-NEXT: [[RES:%.*]] = call nonnull ptr [[FN]]() +; ATTRIBUTOR-NEXT: ret ptr [[RES]] +; + %res = call nonnull ptr %fn() + ret ptr %res +} + attributes #0 = { null_pointer_is_valid } attributes #1 = { nounwind willreturn} diff --git a/llvm/test/Transforms/FunctionAttrs/norecurse.ll b/llvm/test/Transforms/FunctionAttrs/norecurse.ll index 7a089f6..5cb8ac0 100644 --- a/llvm/test/Transforms/FunctionAttrs/norecurse.ll +++ b/llvm/test/Transforms/FunctionAttrs/norecurse.ll @@ -241,6 +241,37 @@ define void @r() norecurse { call void @q() ret void } + +define void @unknown_call(ptr %fn) { +; FNATTRS-LABEL: define {{[^@]+}}@unknown_call +; FNATTRS-SAME: (ptr readonly captures(none) [[FN:%.*]]) { +; FNATTRS-NEXT: call void [[FN]]() +; FNATTRS-NEXT: ret void +; +; ATTRIBUTOR-LABEL: define {{[^@]+}}@unknown_call +; ATTRIBUTOR-SAME: (ptr nofree nonnull captures(none) [[FN:%.*]]) { +; ATTRIBUTOR-NEXT: call void [[FN]]() +; ATTRIBUTOR-NEXT: ret void +; + call void %fn() + ret void +} + +define void @unknown_norecurse_call(ptr %fn) { +; FNATTRS-LABEL: define {{[^@]+}}@unknown_norecurse_call +; FNATTRS-SAME: (ptr readonly captures(none) [[FN:%.*]]) { +; FNATTRS-NEXT: call void [[FN]]() #[[ATTR7:[0-9]+]] +; FNATTRS-NEXT: ret void +; +; ATTRIBUTOR-LABEL: define {{[^@]+}}@unknown_norecurse_call +; ATTRIBUTOR-SAME: (ptr nofree nonnull captures(none) [[FN:%.*]]) { +; ATTRIBUTOR-NEXT: call void [[FN]]() #[[ATTR9:[0-9]+]] +; ATTRIBUTOR-NEXT: ret void +; + call void %fn() norecurse + ret void +} + ;. ; FNATTRS: attributes #[[ATTR0]] = { mustprogress nofree norecurse nosync nounwind willreturn memory(none) } ; FNATTRS: attributes #[[ATTR1]] = { nofree nosync nounwind memory(none) } @@ -249,6 +280,7 @@ define void @r() norecurse { ; FNATTRS: attributes #[[ATTR4]] = { mustprogress nofree norecurse nosync nounwind willreturn memory(argmem: readwrite) } ; FNATTRS: attributes #[[ATTR5:[0-9]+]] = { nocallback nofree nounwind willreturn memory(argmem: readwrite) } ; FNATTRS: attributes #[[ATTR6]] = { nofree norecurse nosync memory(none) } +; FNATTRS: attributes #[[ATTR7]] = { norecurse } ;. ; ATTRIBUTOR: attributes #[[ATTR0]] = { mustprogress nofree norecurse nosync nounwind willreturn memory(none) } ; ATTRIBUTOR: attributes #[[ATTR1]] = { nofree nosync nounwind memory(none) } @@ -259,6 +291,7 @@ define void @r() norecurse { ; ATTRIBUTOR: attributes #[[ATTR6]] = { norecurse nosync memory(none) } ; ATTRIBUTOR: attributes #[[ATTR7]] = { nosync } ; ATTRIBUTOR: attributes #[[ATTR8]] = { nofree willreturn } +; ATTRIBUTOR: attributes #[[ATTR9]] = { norecurse } ;. ;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line: ; COMMON: {{.*}} diff --git a/llvm/test/Transforms/FunctionAttrs/nounwind.ll b/llvm/test/Transforms/FunctionAttrs/nounwind.ll index afa9ae3..a64d9a6 100644 --- a/llvm/test/Transforms/FunctionAttrs/nounwind.ll +++ b/llvm/test/Transforms/FunctionAttrs/nounwind.ll @@ -4,10 +4,15 @@ ; TEST 1 define i32 @foo1() { -; COMMON: Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none) -; COMMON-LABEL: define {{[^@]+}}@foo1 -; COMMON-SAME: () #[[ATTR0:[0-9]+]] { -; COMMON-NEXT: ret i32 1 +; FNATTRS: Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none) +; FNATTRS-LABEL: define {{[^@]+}}@foo1 +; FNATTRS-SAME: () #[[ATTR0:[0-9]+]] { +; FNATTRS-NEXT: ret i32 1 +; +; ATTRIBUTOR: Function Attrs: mustprogress nofree norecurse nosync nounwind willreturn memory(none) +; ATTRIBUTOR-LABEL: define {{[^@]+}}@foo1 +; ATTRIBUTOR-SAME: () #[[ATTR0:[0-9]+]] { +; ATTRIBUTOR-NEXT: ret i32 1 ; ret i32 1 } @@ -70,14 +75,23 @@ define void @call_non_nounwind(){ ; } define i32 @maybe_throw(i1 zeroext %0) { -; COMMON-LABEL: define {{[^@]+}}@maybe_throw -; COMMON-SAME: (i1 zeroext [[TMP0:%.*]]) { -; COMMON-NEXT: br i1 [[TMP0]], label [[TMP2:%.*]], label [[TMP3:%.*]] -; COMMON: 2: -; COMMON-NEXT: tail call void @__cxa_rethrow() -; COMMON-NEXT: unreachable -; COMMON: 3: -; COMMON-NEXT: ret i32 -1 +; FNATTRS-LABEL: define {{[^@]+}}@maybe_throw +; FNATTRS-SAME: (i1 zeroext [[TMP0:%.*]]) { +; FNATTRS-NEXT: br i1 [[TMP0]], label [[TMP2:%.*]], label [[TMP3:%.*]] +; FNATTRS: 2: +; FNATTRS-NEXT: tail call void @__cxa_rethrow() +; FNATTRS-NEXT: unreachable +; FNATTRS: 3: +; FNATTRS-NEXT: ret i32 -1 +; +; ATTRIBUTOR-LABEL: define {{[^@]+}}@maybe_throw +; ATTRIBUTOR-SAME: (i1 zeroext [[TMP0:%.*]]) { +; ATTRIBUTOR-NEXT: br i1 [[TMP0]], label [[TMP2:%.*]], label [[TMP3:%.*]] +; ATTRIBUTOR: 2: +; ATTRIBUTOR-NEXT: tail call void @__cxa_rethrow() +; ATTRIBUTOR-NEXT: unreachable +; ATTRIBUTOR: 3: +; ATTRIBUTOR-NEXT: ret i32 -1 ; br i1 %0, label %2, label %3 @@ -101,18 +115,31 @@ declare void @__cxa_rethrow() ; } define i32 @catch_thing() personality ptr @__gxx_personality_v0 { -; COMMON-LABEL: define {{[^@]+}}@catch_thing() personality ptr @__gxx_personality_v0 { -; COMMON-NEXT: invoke void @__cxa_rethrow() -; COMMON-NEXT: to label [[TMP1:%.*]] unwind label [[TMP2:%.*]] -; COMMON: 1: -; COMMON-NEXT: unreachable -; COMMON: 2: -; COMMON-NEXT: [[TMP3:%.*]] = landingpad { ptr, i32 } -; COMMON-NEXT: catch ptr null -; COMMON-NEXT: [[TMP4:%.*]] = extractvalue { ptr, i32 } [[TMP3]], 0 -; COMMON-NEXT: [[TMP5:%.*]] = tail call ptr @__cxa_begin_catch(ptr [[TMP4]]) -; COMMON-NEXT: tail call void @__cxa_end_catch() -; COMMON-NEXT: ret i32 -1 +; FNATTRS-LABEL: define {{[^@]+}}@catch_thing() personality ptr @__gxx_personality_v0 { +; FNATTRS-NEXT: invoke void @__cxa_rethrow() +; FNATTRS-NEXT: to label [[TMP1:%.*]] unwind label [[TMP2:%.*]] +; FNATTRS: 1: +; FNATTRS-NEXT: unreachable +; FNATTRS: 2: +; FNATTRS-NEXT: [[TMP3:%.*]] = landingpad { ptr, i32 } +; FNATTRS-NEXT: catch ptr null +; FNATTRS-NEXT: [[TMP4:%.*]] = extractvalue { ptr, i32 } [[TMP3]], 0 +; FNATTRS-NEXT: [[TMP5:%.*]] = tail call ptr @__cxa_begin_catch(ptr [[TMP4]]) +; FNATTRS-NEXT: tail call void @__cxa_end_catch() +; FNATTRS-NEXT: ret i32 -1 +; +; ATTRIBUTOR-LABEL: define {{[^@]+}}@catch_thing() personality ptr @__gxx_personality_v0 { +; ATTRIBUTOR-NEXT: invoke void @__cxa_rethrow() +; ATTRIBUTOR-NEXT: to label [[TMP1:%.*]] unwind label [[TMP2:%.*]] +; ATTRIBUTOR: 1: +; ATTRIBUTOR-NEXT: unreachable +; ATTRIBUTOR: 2: +; ATTRIBUTOR-NEXT: [[TMP3:%.*]] = landingpad { ptr, i32 } +; ATTRIBUTOR-NEXT: catch ptr null +; ATTRIBUTOR-NEXT: [[TMP4:%.*]] = extractvalue { ptr, i32 } [[TMP3]], 0 +; ATTRIBUTOR-NEXT: [[TMP5:%.*]] = tail call ptr @__cxa_begin_catch(ptr [[TMP4]]) +; ATTRIBUTOR-NEXT: tail call void @__cxa_end_catch() +; ATTRIBUTOR-NEXT: ret i32 -1 ; invoke void @__cxa_rethrow() #1 to label %1 unwind label %2 @@ -130,9 +157,13 @@ define i32 @catch_thing() personality ptr @__gxx_personality_v0 { } define i32 @catch_thing_user() { -; COMMON-LABEL: define {{[^@]+}}@catch_thing_user() { -; COMMON-NEXT: [[CATCH_THING_CALL:%.*]] = call i32 @catch_thing() -; COMMON-NEXT: ret i32 [[CATCH_THING_CALL]] +; FNATTRS-LABEL: define {{[^@]+}}@catch_thing_user() { +; FNATTRS-NEXT: [[CATCH_THING_CALL:%.*]] = call i32 @catch_thing() +; FNATTRS-NEXT: ret i32 [[CATCH_THING_CALL]] +; +; ATTRIBUTOR-LABEL: define {{[^@]+}}@catch_thing_user() { +; ATTRIBUTOR-NEXT: [[CATCH_THING_CALL:%.*]] = call i32 @catch_thing() +; ATTRIBUTOR-NEXT: ret i32 [[CATCH_THING_CALL]] ; %catch_thing_call = call i32 @catch_thing() ret i32 %catch_thing_call @@ -147,10 +178,10 @@ define void @catch_specific_landingpad() personality ptr @__gxx_personality_v0 { ; COMMON-LABEL: define {{[^@]+}}@catch_specific_landingpad ; COMMON-SAME: () #[[ATTR3:[0-9]+]] personality ptr @__gxx_personality_v0 { ; COMMON-NEXT: invoke void @do_throw() -; COMMON-NEXT: to label [[UNREACHABLE:%.*]] unwind label [[LPAD:%.*]] +; COMMON-NEXT: to label [[UNREACHABLE:%.*]] unwind label [[LPAD:%.*]] ; COMMON: lpad: ; COMMON-NEXT: [[LP:%.*]] = landingpad { ptr, i32 } -; COMMON-NEXT: catch ptr @catch_ty +; COMMON-NEXT: catch ptr @catch_ty ; COMMON-NEXT: call void @abort() ; COMMON-NEXT: unreachable ; COMMON: unreachable: @@ -174,10 +205,10 @@ define void @catch_all_landingpad() personality ptr @__gxx_personality_v0 { ; COMMON-LABEL: define {{[^@]+}}@catch_all_landingpad ; COMMON-SAME: () #[[ATTR4:[0-9]+]] personality ptr @__gxx_personality_v0 { ; COMMON-NEXT: invoke void @do_throw() -; COMMON-NEXT: to label [[UNREACHABLE:%.*]] unwind label [[LPAD:%.*]] +; COMMON-NEXT: to label [[UNREACHABLE:%.*]] unwind label [[LPAD:%.*]] ; COMMON: lpad: ; COMMON-NEXT: [[LP:%.*]] = landingpad { ptr, i32 } -; COMMON-NEXT: catch ptr null +; COMMON-NEXT: catch ptr null ; COMMON-NEXT: call void @abort() ; COMMON-NEXT: unreachable ; COMMON: unreachable: @@ -201,10 +232,10 @@ define void @filter_specific_landingpad() personality ptr @__gxx_personality_v0 ; COMMON-LABEL: define {{[^@]+}}@filter_specific_landingpad ; COMMON-SAME: () #[[ATTR3]] personality ptr @__gxx_personality_v0 { ; COMMON-NEXT: invoke void @do_throw() -; COMMON-NEXT: to label [[UNREACHABLE:%.*]] unwind label [[LPAD:%.*]] +; COMMON-NEXT: to label [[UNREACHABLE:%.*]] unwind label [[LPAD:%.*]] ; COMMON: lpad: ; COMMON-NEXT: [[LP:%.*]] = landingpad { ptr, i32 } -; COMMON-NEXT: filter [1 x ptr] [ptr @catch_ty] +; COMMON-NEXT: filter [1 x ptr] [ptr @catch_ty] ; COMMON-NEXT: call void @abort() ; COMMON-NEXT: unreachable ; COMMON: unreachable: @@ -228,10 +259,10 @@ define void @filter_none_landingpad() personality ptr @__gxx_personality_v0 { ; COMMON-LABEL: define {{[^@]+}}@filter_none_landingpad ; COMMON-SAME: () #[[ATTR4]] personality ptr @__gxx_personality_v0 { ; COMMON-NEXT: invoke void @do_throw() -; COMMON-NEXT: to label [[UNREACHABLE:%.*]] unwind label [[LPAD:%.*]] +; COMMON-NEXT: to label [[UNREACHABLE:%.*]] unwind label [[LPAD:%.*]] ; COMMON: lpad: ; COMMON-NEXT: [[LP:%.*]] = landingpad { ptr, i32 } -; COMMON-NEXT: filter [0 x ptr] zeroinitializer +; COMMON-NEXT: filter [0 x ptr] zeroinitializer ; COMMON-NEXT: call void @abort() ; COMMON-NEXT: unreachable ; COMMON: unreachable: @@ -255,10 +286,10 @@ define void @cleanup_landingpad() personality ptr @__gxx_personality_v0 { ; COMMON-LABEL: define {{[^@]+}}@cleanup_landingpad ; COMMON-SAME: () #[[ATTR3]] personality ptr @__gxx_personality_v0 { ; COMMON-NEXT: invoke void @do_throw() -; COMMON-NEXT: to label [[UNREACHABLE:%.*]] unwind label [[LPAD:%.*]] +; COMMON-NEXT: to label [[UNREACHABLE:%.*]] unwind label [[LPAD:%.*]] ; COMMON: lpad: ; COMMON-NEXT: [[LP:%.*]] = landingpad { ptr, i32 } -; COMMON-NEXT: cleanup +; COMMON-NEXT: cleanup ; COMMON-NEXT: call void @abort() ; COMMON-NEXT: unreachable ; COMMON: unreachable: @@ -282,7 +313,7 @@ define void @cleanuppad() personality ptr @__gxx_personality_v0 { ; FNATTRS-LABEL: define {{[^@]+}}@cleanuppad ; FNATTRS-SAME: () #[[ATTR3]] personality ptr @__gxx_personality_v0 { ; FNATTRS-NEXT: invoke void @do_throw() -; FNATTRS-NEXT: to label [[UNREACHABLE:%.*]] unwind label [[CPAD:%.*]] +; FNATTRS-NEXT: to label [[UNREACHABLE:%.*]] unwind label [[CPAD:%.*]] ; FNATTRS: cpad: ; FNATTRS-NEXT: [[CP:%.*]] = cleanuppad within none [] ; FNATTRS-NEXT: call void @abort() @@ -294,7 +325,7 @@ define void @cleanuppad() personality ptr @__gxx_personality_v0 { ; ATTRIBUTOR-LABEL: define {{[^@]+}}@cleanuppad ; ATTRIBUTOR-SAME: () #[[ATTR4]] personality ptr @__gxx_personality_v0 { ; ATTRIBUTOR-NEXT: invoke void @do_throw() -; ATTRIBUTOR-NEXT: to label [[UNREACHABLE:%.*]] unwind label [[CPAD:%.*]] +; ATTRIBUTOR-NEXT: to label [[UNREACHABLE:%.*]] unwind label [[CPAD:%.*]] ; ATTRIBUTOR: cpad: ; ATTRIBUTOR-NEXT: [[CP:%.*]] = cleanuppad within none [] ; ATTRIBUTOR-NEXT: call void @abort() @@ -319,7 +350,7 @@ define void @catchswitch_cleanuppad() personality ptr @__gxx_personality_v0 { ; FNATTRS-LABEL: define {{[^@]+}}@catchswitch_cleanuppad ; FNATTRS-SAME: () #[[ATTR3]] personality ptr @__gxx_personality_v0 { ; FNATTRS-NEXT: invoke void @do_throw() -; FNATTRS-NEXT: to label [[UNREACHABLE:%.*]] unwind label [[CS:%.*]] +; FNATTRS-NEXT: to label [[UNREACHABLE:%.*]] unwind label [[CS:%.*]] ; FNATTRS: cs: ; FNATTRS-NEXT: [[TOK:%.*]] = catchswitch within none [label %catch] unwind label [[CPAD:%.*]] ; FNATTRS: catch: @@ -337,7 +368,7 @@ define void @catchswitch_cleanuppad() personality ptr @__gxx_personality_v0 { ; ATTRIBUTOR-LABEL: define {{[^@]+}}@catchswitch_cleanuppad ; ATTRIBUTOR-SAME: () #[[ATTR4]] personality ptr @__gxx_personality_v0 { ; ATTRIBUTOR-NEXT: invoke void @do_throw() -; ATTRIBUTOR-NEXT: to label [[UNREACHABLE:%.*]] unwind label [[CS:%.*]] +; ATTRIBUTOR-NEXT: to label [[UNREACHABLE:%.*]] unwind label [[CS:%.*]] ; ATTRIBUTOR: cs: ; ATTRIBUTOR-NEXT: [[TOK:%.*]] = catchswitch within none [label %catch] unwind label [[CPAD:%.*]] ; ATTRIBUTOR: catch: @@ -371,6 +402,37 @@ unreachable: unreachable } +define void @unknown_call(ptr %fn) { +; FNATTRS-LABEL: define {{[^@]+}}@unknown_call +; FNATTRS-SAME: (ptr readonly captures(none) [[FN:%.*]]) { +; FNATTRS-NEXT: call void [[FN]]() +; FNATTRS-NEXT: ret void +; +; ATTRIBUTOR-LABEL: define {{[^@]+}}@unknown_call +; ATTRIBUTOR-SAME: (ptr nofree nonnull captures(none) [[FN:%.*]]) { +; ATTRIBUTOR-NEXT: call void [[FN]]() +; ATTRIBUTOR-NEXT: ret void +; + call void %fn() + ret void +} + +define void @unknown_nounwind_call(ptr %fn) { +; FNATTRS-LABEL: define {{[^@]+}}@unknown_nounwind_call +; FNATTRS-SAME: (ptr readonly captures(none) [[FN:%.*]]) { +; FNATTRS-NEXT: call void [[FN]]() #[[ATTR2:[0-9]+]] +; FNATTRS-NEXT: ret void +; +; ATTRIBUTOR: Function Attrs: nounwind +; ATTRIBUTOR-LABEL: define {{[^@]+}}@unknown_nounwind_call +; ATTRIBUTOR-SAME: (ptr nofree nonnull captures(none) [[FN:%.*]]) #[[ATTR2:[0-9]+]] { +; ATTRIBUTOR-NEXT: call void [[FN]]() #[[ATTR2]] +; ATTRIBUTOR-NEXT: ret void +; + call void %fn() nounwind + ret void +} + declare i32 @__gxx_personality_v0(...) declare ptr @__cxa_begin_catch(ptr) diff --git a/llvm/test/Transforms/FunctionAttrs/sendmsg-nocallback.ll b/llvm/test/Transforms/FunctionAttrs/sendmsg-nocallback.ll index 4d5db32..04575e4 100644 --- a/llvm/test/Transforms/FunctionAttrs/sendmsg-nocallback.ll +++ b/llvm/test/Transforms/FunctionAttrs/sendmsg-nocallback.ll @@ -50,10 +50,12 @@ define internal i32 @sendmsg_rtn_is_norecurse() { } define void @user() { -; FNATTRS-LABEL: define void @user() { +; FNATTRS: Function Attrs: norecurse nounwind +; FNATTRS-LABEL: define void @user( +; FNATTRS-SAME: ) #[[ATTR1]] { ; FNATTRS-NEXT: call void @sendmsg_is_norecurse() ; FNATTRS-NEXT: call void @sendmsghalt_is_norecurse() -; FNATTRS-NEXT: call void @sendmsg_rtn_is_norecurse() +; FNATTRS-NEXT: [[TMP1:%.*]] = call i32 @sendmsg_rtn_is_norecurse() ; FNATTRS-NEXT: ret void ; ; ATTRIBUTOR: Function Attrs: norecurse nounwind @@ -61,12 +63,12 @@ define void @user() { ; ATTRIBUTOR-SAME: ) #[[ATTR1]] { ; ATTRIBUTOR-NEXT: call void @sendmsg_is_norecurse() #[[ATTR5:[0-9]+]] ; ATTRIBUTOR-NEXT: call void @sendmsghalt_is_norecurse() #[[ATTR6:[0-9]+]] -; ATTRIBUTOR-NEXT: call void @sendmsg_rtn_is_norecurse() #[[ATTR6]] +; ATTRIBUTOR-NEXT: [[TMP1:%.*]] = call i32 @sendmsg_rtn_is_norecurse() #[[ATTR6]] ; ATTRIBUTOR-NEXT: ret void ; call void @sendmsg_is_norecurse() call void @sendmsghalt_is_norecurse() - call void @sendmsg_rtn_is_norecurse() + call i32 @sendmsg_rtn_is_norecurse() ret void } ;. diff --git a/llvm/test/Transforms/InstCombine/trunc-inseltpoison.ll b/llvm/test/Transforms/InstCombine/trunc-inseltpoison.ll index 33fa2c3..f83352c 100644 --- a/llvm/test/Transforms/InstCombine/trunc-inseltpoison.ll +++ b/llvm/test/Transforms/InstCombine/trunc-inseltpoison.ll @@ -959,8 +959,8 @@ define <3 x i31> @wide_splat3(<3 x i33> %x) { define <8 x i8> @wide_lengthening_splat(<4 x i16> %v) { ; CHECK-LABEL: @wide_lengthening_splat( -; CHECK-NEXT: [[SHUF:%.*]] = shufflevector <4 x i16> [[V:%.*]], <4 x i16> poison, <8 x i32> zeroinitializer -; CHECK-NEXT: [[TR:%.*]] = trunc <8 x i16> [[SHUF]] to <8 x i8> +; CHECK-NEXT: [[TMP1:%.*]] = trunc <4 x i16> [[V:%.*]] to <4 x i8> +; CHECK-NEXT: [[TR:%.*]] = shufflevector <4 x i8> [[TMP1]], <4 x i8> poison, <8 x i32> zeroinitializer ; CHECK-NEXT: ret <8 x i8> [[TR]] ; %shuf = shufflevector <4 x i16> %v, <4 x i16> %v, <8 x i32> zeroinitializer diff --git a/llvm/test/Transforms/InstCombine/trunc.ll b/llvm/test/Transforms/InstCombine/trunc.ll index a85ce71..dfe9d94 100644 --- a/llvm/test/Transforms/InstCombine/trunc.ll +++ b/llvm/test/Transforms/InstCombine/trunc.ll @@ -960,8 +960,8 @@ define <3 x i31> @wide_splat3(<3 x i33> %x) { define <8 x i8> @wide_lengthening_splat(<4 x i16> %v) { ; CHECK-LABEL: @wide_lengthening_splat( -; CHECK-NEXT: [[SHUF:%.*]] = shufflevector <4 x i16> [[V:%.*]], <4 x i16> poison, <8 x i32> zeroinitializer -; CHECK-NEXT: [[TR:%.*]] = trunc <8 x i16> [[SHUF]] to <8 x i8> +; CHECK-NEXT: [[TMP1:%.*]] = trunc <4 x i16> [[V:%.*]] to <4 x i8> +; CHECK-NEXT: [[TR:%.*]] = shufflevector <4 x i8> [[TMP1]], <4 x i8> poison, <8 x i32> zeroinitializer ; CHECK-NEXT: ret <8 x i8> [[TR]] ; %shuf = shufflevector <4 x i16> %v, <4 x i16> %v, <8 x i32> zeroinitializer @@ -969,6 +969,19 @@ define <8 x i8> @wide_lengthening_splat(<4 x i16> %v) { ret <8 x i8> %tr } +; This is a negative test, we expect the trunc to remain after the shuffle as it +; might not be beneficial to preform trunc on a wider type +define <4 x i8> @wide_shortening_splat(<8 x i16> %v) { +; CHECK-LABEL: @wide_shortening_splat( +; CHECK-NEXT: [[SHUF:%.*]] = shufflevector <8 x i16> [[V:%.*]], <8 x i16> poison, <4 x i32> zeroinitializer +; CHECK-NEXT: [[TR:%.*]] = trunc <4 x i16> [[SHUF]] to <4 x i8> +; CHECK-NEXT: ret <4 x i8> [[TR]] +; + %shuf = shufflevector <8 x i16> %v, <8 x i16> %v, <4 x i32> zeroinitializer + %tr = trunc <4 x i16> %shuf to <4 x i8> + ret <4 x i8> %tr +} + define <2 x i8> @narrow_add_vec_constant(<2 x i32> %x) { ; CHECK-LABEL: @narrow_add_vec_constant( ; CHECK-NEXT: [[TMP1:%.*]] = trunc <2 x i32> [[X:%.*]] to <2 x i8> diff --git a/llvm/test/Transforms/InterleavedAccess/RISCV/interleaved-accesses.ll b/llvm/test/Transforms/InterleavedAccess/RISCV/interleaved-accesses.ll index 672e949..b505917 100644 --- a/llvm/test/Transforms/InterleavedAccess/RISCV/interleaved-accesses.ll +++ b/llvm/test/Transforms/InterleavedAccess/RISCV/interleaved-accesses.ll @@ -874,3 +874,79 @@ define void @load_factor2_fp128(ptr %ptr) { %v1 = shufflevector <4 x fp128> %interleaved.vec, <4 x fp128> poison, <2 x i32> <i32 1, i32 3> ret void } + +define void @load_factor2_f32(ptr %ptr) { +; RV32-LABEL: @load_factor2_f32( +; RV32-NEXT: [[TMP1:%.*]] = call { <8 x float>, <8 x float> } @llvm.riscv.seg2.load.mask.v8f32.p0.i32(ptr [[PTR:%.*]], <8 x i1> splat (i1 true), i32 8) +; RV32-NEXT: [[TMP2:%.*]] = extractvalue { <8 x float>, <8 x float> } [[TMP1]], 1 +; RV32-NEXT: [[TMP3:%.*]] = extractvalue { <8 x float>, <8 x float> } [[TMP1]], 0 +; RV32-NEXT: ret void +; +; RV64-LABEL: @load_factor2_f32( +; RV64-NEXT: [[TMP1:%.*]] = call { <8 x float>, <8 x float> } @llvm.riscv.seg2.load.mask.v8f32.p0.i64(ptr [[PTR:%.*]], <8 x i1> splat (i1 true), i64 8) +; RV64-NEXT: [[TMP2:%.*]] = extractvalue { <8 x float>, <8 x float> } [[TMP1]], 1 +; RV64-NEXT: [[TMP3:%.*]] = extractvalue { <8 x float>, <8 x float> } [[TMP1]], 0 +; RV64-NEXT: ret void +; + %interleaved.vec = load <16 x float>, ptr %ptr + %v0 = shufflevector <16 x float> %interleaved.vec, <16 x float> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14> + %v1 = shufflevector <16 x float> %interleaved.vec, <16 x float> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15> + ret void +} + +define void @load_factor2_f64(ptr %ptr) { +; RV32-LABEL: @load_factor2_f64( +; RV32-NEXT: [[TMP1:%.*]] = call { <8 x double>, <8 x double> } @llvm.riscv.seg2.load.mask.v8f64.p0.i32(ptr [[PTR:%.*]], <8 x i1> splat (i1 true), i32 8) +; RV32-NEXT: [[TMP2:%.*]] = extractvalue { <8 x double>, <8 x double> } [[TMP1]], 1 +; RV32-NEXT: [[TMP3:%.*]] = extractvalue { <8 x double>, <8 x double> } [[TMP1]], 0 +; RV32-NEXT: ret void +; +; RV64-LABEL: @load_factor2_f64( +; RV64-NEXT: [[TMP1:%.*]] = call { <8 x double>, <8 x double> } @llvm.riscv.seg2.load.mask.v8f64.p0.i64(ptr [[PTR:%.*]], <8 x i1> splat (i1 true), i64 8) +; RV64-NEXT: [[TMP2:%.*]] = extractvalue { <8 x double>, <8 x double> } [[TMP1]], 1 +; RV64-NEXT: [[TMP3:%.*]] = extractvalue { <8 x double>, <8 x double> } [[TMP1]], 0 +; RV64-NEXT: ret void +; + %interleaved.vec = load <16 x double>, ptr %ptr + %v0 = shufflevector <16 x double> %interleaved.vec, <16 x double> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14> + %v1 = shufflevector <16 x double> %interleaved.vec, <16 x double> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15> + ret void +} + +define void @load_factor2_bf16(ptr %ptr) { +; RV32-LABEL: @load_factor2_bf16( +; RV32-NEXT: [[INTERLEAVED_VEC:%.*]] = load <16 x bfloat>, ptr [[PTR:%.*]], align 32 +; RV32-NEXT: [[V0:%.*]] = shufflevector <16 x bfloat> [[INTERLEAVED_VEC]], <16 x bfloat> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14> +; RV32-NEXT: [[V1:%.*]] = shufflevector <16 x bfloat> [[INTERLEAVED_VEC]], <16 x bfloat> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15> +; RV32-NEXT: ret void +; +; RV64-LABEL: @load_factor2_bf16( +; RV64-NEXT: [[INTERLEAVED_VEC:%.*]] = load <16 x bfloat>, ptr [[PTR:%.*]], align 32 +; RV64-NEXT: [[V0:%.*]] = shufflevector <16 x bfloat> [[INTERLEAVED_VEC]], <16 x bfloat> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14> +; RV64-NEXT: [[V1:%.*]] = shufflevector <16 x bfloat> [[INTERLEAVED_VEC]], <16 x bfloat> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15> +; RV64-NEXT: ret void +; + %interleaved.vec = load <16 x bfloat>, ptr %ptr + %v0 = shufflevector <16 x bfloat> %interleaved.vec, <16 x bfloat> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14> + %v1 = shufflevector <16 x bfloat> %interleaved.vec, <16 x bfloat> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15> + ret void +} + +define void @load_factor2_f16(ptr %ptr) { +; RV32-LABEL: @load_factor2_f16( +; RV32-NEXT: [[INTERLEAVED_VEC:%.*]] = load <16 x half>, ptr [[PTR:%.*]], align 32 +; RV32-NEXT: [[V0:%.*]] = shufflevector <16 x half> [[INTERLEAVED_VEC]], <16 x half> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14> +; RV32-NEXT: [[V1:%.*]] = shufflevector <16 x half> [[INTERLEAVED_VEC]], <16 x half> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15> +; RV32-NEXT: ret void +; +; RV64-LABEL: @load_factor2_f16( +; RV64-NEXT: [[INTERLEAVED_VEC:%.*]] = load <16 x half>, ptr [[PTR:%.*]], align 32 +; RV64-NEXT: [[V0:%.*]] = shufflevector <16 x half> [[INTERLEAVED_VEC]], <16 x half> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14> +; RV64-NEXT: [[V1:%.*]] = shufflevector <16 x half> [[INTERLEAVED_VEC]], <16 x half> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15> +; RV64-NEXT: ret void +; + %interleaved.vec = load <16 x half>, ptr %ptr + %v0 = shufflevector <16 x half> %interleaved.vec, <16 x half> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14> + %v1 = shufflevector <16 x half> %interleaved.vec, <16 x half> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15> + ret void +} diff --git a/llvm/test/Transforms/LoopFusion/sunk-phi-nodes.ll b/llvm/test/Transforms/LoopFusion/sunk-phi-nodes.ll new file mode 100644 index 0000000..36c6bdd --- /dev/null +++ b/llvm/test/Transforms/LoopFusion/sunk-phi-nodes.ll @@ -0,0 +1,65 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5 +; RUN: opt -passes=loop-fusion -S < %s 2>&1 | FileCheck %s +define dso_local i32 @check_sunk_phi_nodes() { +; CHECK-LABEL: define dso_local i32 @check_sunk_phi_nodes() { +; CHECK-NEXT: [[ENTRY:.*]]: +; CHECK-NEXT: br label %[[FOR_BODY:.*]] +; CHECK: [[FOR_BODY]]: +; CHECK-NEXT: [[SUM1_02:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[ADD:%.*]], %[[FOR_INC6:.*]] ] +; CHECK-NEXT: [[I_01:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[INC:%.*]], %[[FOR_INC6]] ] +; CHECK-NEXT: [[I1_04:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[INC7:%.*]], %[[FOR_INC6]] ] +; CHECK-NEXT: [[SUM2_03:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[ADD5:%.*]], %[[FOR_INC6]] ] +; CHECK-NEXT: [[ADD]] = add nsw i32 [[SUM1_02]], [[I_01]] +; CHECK-NEXT: br label %[[FOR_INC:.*]] +; CHECK: [[FOR_INC]]: +; CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 [[I1_04]], [[I1_04]] +; CHECK-NEXT: [[ADD5]] = add nsw i32 [[SUM2_03]], [[MUL]] +; CHECK-NEXT: br label %[[FOR_INC6]] +; CHECK: [[FOR_INC6]]: +; CHECK-NEXT: [[INC]] = add nsw i32 [[I_01]], 1 +; CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 [[INC]], 10 +; CHECK-NEXT: [[INC7]] = add nsw i32 [[I1_04]], 1 +; CHECK-NEXT: [[CMP3:%.*]] = icmp slt i32 [[INC7]], 10 +; CHECK-NEXT: br i1 [[CMP3]], label %[[FOR_BODY]], label %[[FOR_END8:.*]] +; CHECK: [[FOR_END8]]: +; CHECK-NEXT: [[SUM2_0_LCSSA:%.*]] = phi i32 [ [[ADD5]], %[[FOR_INC6]] ] +; CHECK-NEXT: [[SUM1_0_LCSSA:%.*]] = phi i32 [ [[ADD]], %[[FOR_INC6]] ] +; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[SUM1_0_LCSSA]], [[SUM2_0_LCSSA]] +; CHECK-NEXT: ret i32 [[TMP0]] +; +entry: + br label %for.body + +for.body: ; preds = %entry, %for.inc + %sum1.02 = phi i32 [ 0, %entry ], [ %add, %for.inc ] + %i.01 = phi i32 [ 0, %entry ], [ %inc, %for.inc ] + %add = add nsw i32 %sum1.02, %i.01 + br label %for.inc + +for.inc: ; preds = %for.body + %inc = add nsw i32 %i.01, 1 + %cmp = icmp slt i32 %inc, 10 + br i1 %cmp, label %for.body, label %for.end + +for.end: ; preds = %for.inc + %sum1.0.lcssa = phi i32 [ %add, %for.inc ] + br label %for.body4 + +for.body4: ; preds = %for.end, %for.inc6 + %i1.04 = phi i32 [ 0, %for.end ], [ %inc7, %for.inc6 ] + %sum2.03 = phi i32 [ 0, %for.end ], [ %add5, %for.inc6 ] + %mul = mul nsw i32 %i1.04, %i1.04 + %add5 = add nsw i32 %sum2.03, %mul + br label %for.inc6 + +for.inc6: ; preds = %for.body4 + %inc7 = add nsw i32 %i1.04, 1 + %cmp3 = icmp slt i32 %inc7, 10 + br i1 %cmp3, label %for.body4, label %for.end8 + +for.end8: ; preds = %for.inc6 + %sum2.0.lcssa = phi i32 [ %add5, %for.inc6 ] + %0 = add i32 %sum1.0.lcssa, %sum2.0.lcssa + ret i32 %0 +} + diff --git a/llvm/test/Transforms/LoopIdiom/reuse-lcssa-phi-scev-expansion.ll b/llvm/test/Transforms/LoopIdiom/reuse-lcssa-phi-scev-expansion.ll index 357a7b6..65aaf72 100644 --- a/llvm/test/Transforms/LoopIdiom/reuse-lcssa-phi-scev-expansion.ll +++ b/llvm/test/Transforms/LoopIdiom/reuse-lcssa-phi-scev-expansion.ll @@ -96,3 +96,68 @@ loop.3: exit: ret void } + +declare i1 @cond() + +define ptr @test_lcssa_reuse_preserve_lcssa() { +; CHECK-LABEL: define ptr @test_lcssa_reuse_preserve_lcssa() { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: br label %[[LOOP_0_HEADER:.*]] +; CHECK: [[LOOP_0_HEADER]]: +; CHECK-NEXT: br label %[[LOOP_1:.*]] +; CHECK: [[LOOP_1]]: +; CHECK-NEXT: [[IV_1:%.*]] = phi ptr [ null, %[[LOOP_0_HEADER]] ], [ [[IV_1_NEXT:%.*]], %[[LOOP_1]] ] +; CHECK-NEXT: [[IV_1_NEXT]] = getelementptr i8, ptr [[IV_1]], i64 1 +; CHECK-NEXT: [[EC_1:%.*]] = call i1 @cond() +; CHECK-NEXT: br i1 [[EC_1]], label %[[THEN:.*]], label %[[LOOP_1]] +; CHECK: [[THEN]]: +; CHECK-NEXT: [[IV_1_LCSSA1:%.*]] = phi ptr [ [[IV_1]], %[[LOOP_1]] ] +; CHECK-NEXT: [[C_2:%.*]] = call i1 @cond() +; CHECK-NEXT: br i1 [[C_2]], label %[[LOOP_2_PREHEADER:.*]], label %[[LOOP_0_LATCH:.*]] +; CHECK: [[LOOP_2_PREHEADER]]: +; CHECK-NEXT: [[IV_1_LCSSA:%.*]] = phi ptr [ [[IV_1_LCSSA1]], %[[THEN]] ] +; CHECK-NEXT: [[IV_1_LCSSA_LCSSA:%.*]] = phi ptr [ [[IV_1_LCSSA1]], %[[THEN]] ] +; CHECK-NEXT: [[STRLEN:%.*]] = call i64 @strlen(ptr null) +; CHECK-NEXT: [[TMP0:%.*]] = getelementptr i8, ptr [[IV_1_LCSSA]], i64 1 +; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[TMP0]], i64 [[STRLEN]] +; CHECK-NEXT: br label %[[LOOP_2:.*]] +; CHECK: [[LOOP_2]]: +; CHECK-NEXT: [[IV_2:%.*]] = phi ptr [ [[RES:%.*]], %[[LOOP_2]] ], [ [[IV_1_LCSSA_LCSSA]], %[[LOOP_2_PREHEADER]] ] +; CHECK-NEXT: [[RES]] = getelementptr i8, ptr [[IV_2]], i64 1 +; CHECK-NEXT: [[L:%.*]] = load i8, ptr [[IV_1_LCSSA_LCSSA]], align 1 +; CHECK-NEXT: [[EC_2:%.*]] = icmp eq i8 [[L]], 0 +; CHECK-NEXT: br i1 true, label %[[EXIT:.*]], label %[[LOOP_2]] +; CHECK: [[LOOP_0_LATCH]]: +; CHECK-NEXT: br label %[[LOOP_0_HEADER]] +; CHECK: [[EXIT]]: +; CHECK-NEXT: ret ptr [[SCEVGEP]] +; +entry: + br label %loop.0.header + +loop.0.header: + br label %loop.1 + +loop.1: + %iv.1 = phi ptr [ null, %loop.0.header ], [ %iv.1.next, %loop.1 ] + %iv.1.next = getelementptr i8, ptr %iv.1, i64 1 + %ec.1 = call i1 @cond() + br i1 %ec.1, label %then, label %loop.1 + +then: + %c.2 = call i1 @cond() + br i1 %c.2, label %loop.2, label %loop.0.latch + +loop.2: + %iv.2 = phi ptr [ %res, %loop.2 ], [ %iv.1, %then ] + %res = getelementptr i8, ptr %iv.2, i64 1 + %l = load i8, ptr %iv.1, align 1 + %ec.2 = icmp eq i8 %l, 0 + br i1 %ec.2, label %exit, label %loop.2 + +loop.0.latch: + br label %loop.0.header + +exit: + ret ptr %res +} diff --git a/llvm/test/Transforms/LoopUnroll/Hexagon/reuse-lcssa-phi-scev-expansion.ll b/llvm/test/Transforms/LoopUnroll/Hexagon/reuse-lcssa-phi-scev-expansion.ll new file mode 100644 index 0000000..f74fb14 --- /dev/null +++ b/llvm/test/Transforms/LoopUnroll/Hexagon/reuse-lcssa-phi-scev-expansion.ll @@ -0,0 +1,108 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5 +; RUN: opt -p loop-unroll -S %s | FileCheck %s + +target triple = "hexagon-unknown-linux" + +declare void @foo() + +define void @preserve_lcssa_when_reusing_existing_phi() { +; CHECK-LABEL: define void @preserve_lcssa_when_reusing_existing_phi() { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: br label %[[LOOP_1_HEADER:.*]] +; CHECK: [[LOOP_1_HEADER]]: +; CHECK-NEXT: br label %[[LOOP_2_HEADER:.*]] +; CHECK: [[LOOP_2_HEADER]]: +; CHECK-NEXT: br label %[[LOOP_3:.*]] +; CHECK: [[LOOP_3]]: +; CHECK-NEXT: [[IV_3:%.*]] = phi i32 [ [[IV_3_NEXT:%.*]], %[[LOOP_3]] ], [ 0, %[[LOOP_2_HEADER]] ] +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: [[IV_3_NEXT]] = add i32 [[IV_3]], 1 +; CHECK-NEXT: br i1 false, label %[[PH:.*]], label %[[LOOP_3]] +; CHECK: [[PH]]: +; CHECK-NEXT: [[IV_3_LCSSA:%.*]] = phi i32 [ [[IV_3]], %[[LOOP_3]] ] +; CHECK-NEXT: br i1 true, label %[[LOOP_2_LATCH:.*]], label %[[LOOP_4_PREHEADER:.*]] +; CHECK: [[LOOP_4_PREHEADER]]: +; CHECK-NEXT: [[IV_3_LCSSA_LCSSA1:%.*]] = phi i32 [ [[IV_3_LCSSA]], %[[PH]] ] +; CHECK-NEXT: [[IV_3_LCSSA_LCSSA:%.*]] = phi i32 [ [[IV_3_LCSSA]], %[[PH]] ] +; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[IV_3_LCSSA_LCSSA1]], 1 +; CHECK-NEXT: [[TMP1:%.*]] = freeze i32 [[TMP0]] +; CHECK-NEXT: [[TMP2:%.*]] = add i32 [[TMP1]], -1 +; CHECK-NEXT: [[XTRAITER:%.*]] = and i32 [[TMP1]], 7 +; CHECK-NEXT: [[TMP3:%.*]] = icmp ult i32 [[TMP2]], 7 +; CHECK-NEXT: br i1 [[TMP3]], label %[[LOOP_1_LATCH_UNR_LCSSA:.*]], label %[[LOOP_4_PREHEADER_NEW:.*]] +; CHECK: [[LOOP_4_PREHEADER_NEW]]: +; CHECK-NEXT: br label %[[LOOP_4:.*]] +; CHECK: [[LOOP_2_LATCH]]: +; CHECK-NEXT: br label %[[LOOP_2_HEADER]] +; CHECK: [[LOOP_4]]: +; CHECK-NEXT: [[IV_4:%.*]] = phi i32 [ 0, %[[LOOP_4_PREHEADER_NEW]] ], [ [[INC_I_7:%.*]], %[[LOOP_4]] ] +; CHECK-NEXT: [[NITER:%.*]] = phi i32 [ 0, %[[LOOP_4_PREHEADER_NEW]] ], [ [[NITER_NEXT_7:%.*]], %[[LOOP_4]] ] +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: [[INC_I_7]] = add nuw nsw i32 [[IV_4]], 8 +; CHECK-NEXT: [[NITER_NEXT_7]] = add nuw nsw i32 [[NITER]], 8 +; CHECK-NEXT: br i1 true, label %[[LOOP_1_LATCH_UNR_LCSSA_LOOPEXIT:.*]], label %[[LOOP_4]] +; CHECK: [[LOOP_1_LATCH_UNR_LCSSA_LOOPEXIT]]: +; CHECK-NEXT: [[IV_4_UNR_PH:%.*]] = phi i32 [ [[INC_I_7]], %[[LOOP_4]] ] +; CHECK-NEXT: br label %[[LOOP_1_LATCH_UNR_LCSSA]] +; CHECK: [[LOOP_1_LATCH_UNR_LCSSA]]: +; CHECK-NEXT: [[IV_4_UNR:%.*]] = phi i32 [ 0, %[[LOOP_4_PREHEADER]] ], [ [[IV_4_UNR_PH]], %[[LOOP_1_LATCH_UNR_LCSSA_LOOPEXIT]] ] +; CHECK-NEXT: [[LCMP_MOD:%.*]] = icmp ne i32 [[XTRAITER]], 0 +; CHECK-NEXT: br i1 [[LCMP_MOD]], label %[[LOOP_4_EPIL_PREHEADER:.*]], label %[[LOOP_1_LATCH:.*]] +; CHECK: [[LOOP_4_EPIL_PREHEADER]]: +; CHECK-NEXT: br label %[[LOOP_4_EPIL:.*]] +; CHECK: [[LOOP_4_EPIL]]: +; CHECK-NEXT: [[IV_4_EPIL:%.*]] = phi i32 [ [[INC_I_EPIL:%.*]], %[[LOOP_4_EPIL]] ], [ [[IV_4_UNR]], %[[LOOP_4_EPIL_PREHEADER]] ] +; CHECK-NEXT: [[EPIL_ITER:%.*]] = phi i32 [ 0, %[[LOOP_4_EPIL_PREHEADER]] ], [ [[EPIL_ITER_NEXT:%.*]], %[[LOOP_4_EPIL]] ] +; CHECK-NEXT: call void @foo() +; CHECK-NEXT: [[INC_I_EPIL]] = add i32 [[IV_4_EPIL]], 1 +; CHECK-NEXT: [[EC_EPIL:%.*]] = icmp eq i32 [[IV_4_EPIL]], [[IV_3_LCSSA_LCSSA]] +; CHECK-NEXT: [[EPIL_ITER_NEXT]] = add i32 [[EPIL_ITER]], 1 +; CHECK-NEXT: [[EPIL_ITER_CMP:%.*]] = icmp ne i32 [[EPIL_ITER_NEXT]], [[XTRAITER]] +; CHECK-NEXT: br i1 [[EPIL_ITER_CMP]], label %[[LOOP_4_EPIL]], label %[[LOOP_1_LATCH_EPILOG_LCSSA:.*]], !llvm.loop [[LOOP0:![0-9]+]] +; CHECK: [[LOOP_1_LATCH_EPILOG_LCSSA]]: +; CHECK-NEXT: br label %[[LOOP_1_LATCH]] +; CHECK: [[LOOP_1_LATCH]]: +; CHECK-NEXT: br label %[[LOOP_1_HEADER]] +; +entry: + br label %loop.1.header + +loop.1.header: + br label %loop.2.header + +loop.2.header: + br label %loop.3 + +loop.3: + %iv.3 = phi i32 [ %iv.3.next, %loop.3 ], [ 0, %loop.2.header ] + call void @foo() + %iv.3.next = add i32 %iv.3, 1 + br i1 false, label %ph, label %loop.3 + +ph: + br i1 true, label %loop.2.latch, label %loop.4 + +loop.2.latch: + br label %loop.2.header + +loop.4: + %iv.4 = phi i32 [ 0, %ph ], [ %inc.i, %loop.4 ] + call void @foo() + %inc.i = add i32 %iv.4, 1 + %ec = icmp eq i32 %iv.4, %iv.3 + br i1 %ec, label %loop.1.latch, label %loop.4 + +loop.1.latch: + br label %loop.1.header +} +;. +; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]]} +; CHECK: [[META1]] = !{!"llvm.loop.unroll.disable"} +;. diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-epilog-vect.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-epilog-vect.ll index 18cc3a8..0f407cd 100644 --- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-epilog-vect.ll +++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-epilog-vect.ll @@ -148,6 +148,104 @@ exit: ret void } +define void @main_vf_vscale_x_2_no_epi_iteration(ptr %A) #0 vscale_range(8, 8) { +; CHECK-LABEL: @main_vf_vscale_x_2_no_epi_iteration( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 4 +; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] +; CHECK: vector.ph: +; CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 4 +; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]] +; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]] +; CHECK-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4 +; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] +; CHECK: vector.body: +; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] +; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]] +; CHECK-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 2 +; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i64 [[TMP8]] +; CHECK-NEXT: store <vscale x 2 x i64> splat (i64 1), ptr [[TMP6]], align 1 +; CHECK-NEXT: store <vscale x 2 x i64> splat (i64 1), ptr [[TMP9]], align 1 +; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] +; CHECK-NEXT: [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; CHECK-NEXT: br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]] +; CHECK: middle.block: +; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]] +; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]] +; CHECK: scalar.ph: +; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] +; CHECK-NEXT: br label [[FOR_BODY:%.*]] +; CHECK: for.body: +; +; CHECK-VF8-LABEL: @main_vf_vscale_x_2_no_epi_iteration( +; CHECK-VF8-NEXT: iter.check: +; CHECK-VF8-NEXT: br i1 false, label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]] +; CHECK-VF8: vector.main.loop.iter.check: +; CHECK-VF8-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-VF8-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 4 +; CHECK-VF8-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP1]] +; CHECK-VF8-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]] +; CHECK-VF8: vector.ph: +; CHECK-VF8-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-VF8-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 4 +; CHECK-VF8-NEXT: [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]] +; CHECK-VF8-NEXT: [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]] +; CHECK-VF8-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-VF8-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4 +; CHECK-VF8-NEXT: br label [[VECTOR_BODY:%.*]] +; CHECK-VF8: vector.body: +; CHECK-VF8-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] +; CHECK-VF8-NEXT: [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]] +; CHECK-VF8-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-VF8-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 2 +; CHECK-VF8-NEXT: [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i64 [[TMP8]] +; CHECK-VF8-NEXT: store <vscale x 2 x i64> splat (i64 1), ptr [[TMP6]], align 1 +; CHECK-VF8-NEXT: store <vscale x 2 x i64> splat (i64 1), ptr [[TMP9]], align 1 +; CHECK-VF8-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] +; CHECK-VF8-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; CHECK-VF8-NEXT: br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]] +; CHECK-VF8: middle.block: +; CHECK-VF8-NEXT: [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]] +; CHECK-VF8-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]] +; CHECK-VF8: vec.epilog.iter.check: +; CHECK-VF8-NEXT: [[N_VEC_REMAINING:%.*]] = sub i64 1024, [[N_VEC]] +; CHECK-VF8-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 8 +; CHECK-VF8-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]] +; CHECK-VF8: vec.epilog.ph: +; CHECK-VF8-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ] +; CHECK-VF8-NEXT: br label [[VEC_EPILOG_VECTOR_BODY:%.*]] +; CHECK-VF8: vec.epilog.vector.body: +; CHECK-VF8-NEXT: [[INDEX1:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ] +; CHECK-VF8-NEXT: [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX1]] +; CHECK-VF8-NEXT: store <8 x i64> splat (i64 1), ptr [[TMP11]], align 1 +; CHECK-VF8-NEXT: [[INDEX_NEXT2]] = add nuw i64 [[INDEX1]], 8 +; CHECK-VF8-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT2]], 1024 +; CHECK-VF8-NEXT: br i1 [[TMP14]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]] +; CHECK-VF8: vec.epilog.middle.block: +; CHECK-VF8-NEXT: br i1 true, label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]] +; CHECK-VF8: vec.epilog.scalar.ph: +; CHECK-VF8-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ] +; CHECK-VF8-NEXT: br label [[FOR_BODY:%.*]] +; CHECK-VF8: for.body: +; +entry: + br label %for.body + +for.body: + %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ] + %arrayidx = getelementptr inbounds i64, ptr %A, i64 %iv + store i64 1, ptr %arrayidx, align 1 + %iv.next = add nuw nsw i64 %iv, 1 + %exitcond = icmp ne i64 %iv.next, 1024 + br i1 %exitcond, label %for.body, label %exit + +exit: + ret void +} ; DEBUG: LV: Checking a loop in 'main_vf_vscale_x_2' ; DEBUG: Create Skeleton for epilogue vectorized loop (first pass) @@ -163,20 +261,21 @@ exit: ; fixed-width VF=8 for the epilogue if the vectors are known to be ; sufficiently wide. This information can be deduced from vscale_range or ; VScaleForTuning (set by mcpu/mtune). -define void @main_vf_vscale_x_2(ptr %A) #0 vscale_range(8, 8) { +define void @main_vf_vscale_x_2(ptr %A, i64 %n) #0 vscale_range(8, 8) { ; CHECK-LABEL: @main_vf_vscale_x_2( ; CHECK-NEXT: iter.check: -; CHECK-NEXT: br i1 false, label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]] +; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 8 +; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]] ; CHECK: vector.main.loop.iter.check: ; CHECK-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() ; CHECK-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 4 -; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP1]] -; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]] +; CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[N]], [[TMP1]] +; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]] ; CHECK: vector.ph: ; CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() ; CHECK-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 4 -; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]] -; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]] +; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]] +; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] ; CHECK-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() ; CHECK-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4 ; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] @@ -190,44 +289,48 @@ define void @main_vf_vscale_x_2(ptr %A) #0 vscale_range(8, 8) { ; CHECK-NEXT: store <vscale x 2 x i64> splat (i64 1), ptr [[TMP17]], align 1 ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] ; CHECK-NEXT: [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] -; CHECK-NEXT: br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]] +; CHECK-NEXT: br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]] ; CHECK: middle.block: -; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]] -; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]] +; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH:%.*]] ; CHECK: vec.epilog.iter.check: -; CHECK-NEXT: [[N_VEC_REMAINING:%.*]] = sub i64 1024, [[N_VEC]] +; CHECK-NEXT: [[N_VEC_REMAINING:%.*]] = sub i64 [[N]], [[N_VEC]] ; CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 8 ; CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]] ; CHECK: vec.epilog.ph: -; CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ] -; CHECK-NEXT: br label [[VEC_EPILOG_VECTOR_BODY:%.*]] +; CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[SCALAR_PH]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ] +; CHECK-NEXT: [[N_MOD_VF2:%.*]] = urem i64 [[N]], 8 +; CHECK-NEXT: [[N_VEC3:%.*]] = sub i64 [[N]], [[N_MOD_VF2]] +; CHECK-NEXT: br label [[FOR_BODY:%.*]] ; CHECK: vec.epilog.vector.body: -; CHECK-NEXT: [[INDEX1:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ] -; CHECK-NEXT: [[TMP20:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX1]] -; CHECK-NEXT: store <8 x i64> splat (i64 1), ptr [[TMP20]], align 1 -; CHECK-NEXT: [[INDEX_NEXT2]] = add nuw i64 [[INDEX1]], 8 -; CHECK-NEXT: [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT2]], 1024 -; CHECK-NEXT: br i1 [[TMP22]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]] +; CHECK-NEXT: [[INDEX4:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT5:%.*]], [[FOR_BODY]] ] +; CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX4]] +; CHECK-NEXT: store <8 x i64> splat (i64 1), ptr [[TMP19]], align 1 +; CHECK-NEXT: [[INDEX_NEXT5]] = add nuw i64 [[INDEX4]], 8 +; CHECK-NEXT: [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT5]], [[N_VEC3]] +; CHECK-NEXT: br i1 [[TMP20]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]] ; CHECK: vec.epilog.middle.block: -; CHECK-NEXT: br i1 true, label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]] +; CHECK-NEXT: [[CMP_N6:%.*]] = icmp eq i64 [[N]], [[N_VEC3]] +; CHECK-NEXT: br i1 [[CMP_N6]], label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]] ; CHECK: vec.epilog.scalar.ph: -; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ] -; CHECK-NEXT: br label [[FOR_BODY:%.*]] +; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC3]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[SCALAR_PH]] ], [ 0, [[ITER_CHECK:%.*]] ] +; CHECK-NEXT: br label [[FOR_BODY1:%.*]] ; CHECK: for.body: ; ; CHECK-VF8-LABEL: @main_vf_vscale_x_2( ; CHECK-VF8-NEXT: iter.check: -; CHECK-VF8-NEXT: br i1 false, label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]] +; CHECK-VF8-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[N:%.*]], 8 +; CHECK-VF8-NEXT: br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]] ; CHECK-VF8: vector.main.loop.iter.check: ; CHECK-VF8-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() ; CHECK-VF8-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 4 -; CHECK-VF8-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP1]] +; CHECK-VF8-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP1]] ; CHECK-VF8-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]] ; CHECK-VF8: vector.ph: ; CHECK-VF8-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() ; CHECK-VF8-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 4 -; CHECK-VF8-NEXT: [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]] -; CHECK-VF8-NEXT: [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]] +; CHECK-VF8-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]] +; CHECK-VF8-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] ; CHECK-VF8-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() ; CHECK-VF8-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4 ; CHECK-VF8-NEXT: br label [[VECTOR_BODY:%.*]] @@ -241,28 +344,31 @@ define void @main_vf_vscale_x_2(ptr %A) #0 vscale_range(8, 8) { ; CHECK-VF8-NEXT: store <vscale x 2 x i64> splat (i64 1), ptr [[TMP17]], align 1 ; CHECK-VF8-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] ; CHECK-VF8-NEXT: [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] -; CHECK-VF8-NEXT: br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]] +; CHECK-VF8-NEXT: br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]] ; CHECK-VF8: middle.block: -; CHECK-VF8-NEXT: [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]] +; CHECK-VF8-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] ; CHECK-VF8-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]] ; CHECK-VF8: vec.epilog.iter.check: -; CHECK-VF8-NEXT: [[N_VEC_REMAINING:%.*]] = sub i64 1024, [[N_VEC]] +; CHECK-VF8-NEXT: [[N_VEC_REMAINING:%.*]] = sub i64 [[N]], [[N_VEC]] ; CHECK-VF8-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 8 ; CHECK-VF8-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]] ; CHECK-VF8: vec.epilog.ph: ; CHECK-VF8-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ] +; CHECK-VF8-NEXT: [[N_MOD_VF2:%.*]] = urem i64 [[N]], 8 +; CHECK-VF8-NEXT: [[N_VEC3:%.*]] = sub i64 [[N]], [[N_MOD_VF2]] ; CHECK-VF8-NEXT: br label [[VEC_EPILOG_VECTOR_BODY:%.*]] ; CHECK-VF8: vec.epilog.vector.body: ; CHECK-VF8-NEXT: [[INDEX1:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ] ; CHECK-VF8-NEXT: [[TMP20:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX1]] ; CHECK-VF8-NEXT: store <8 x i64> splat (i64 1), ptr [[TMP20]], align 1 ; CHECK-VF8-NEXT: [[INDEX_NEXT2]] = add nuw i64 [[INDEX1]], 8 -; CHECK-VF8-NEXT: [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT2]], 1024 -; CHECK-VF8-NEXT: br i1 [[TMP22]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]] +; CHECK-VF8-NEXT: [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT2]], [[N_VEC3]] +; CHECK-VF8-NEXT: br i1 [[TMP19]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]] ; CHECK-VF8: vec.epilog.middle.block: -; CHECK-VF8-NEXT: br i1 true, label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]] +; CHECK-VF8-NEXT: [[CMP_N6:%.*]] = icmp eq i64 [[N]], [[N_VEC3]] +; CHECK-VF8-NEXT: br i1 [[CMP_N6]], label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]] ; CHECK-VF8: vec.epilog.scalar.ph: -; CHECK-VF8-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ] +; CHECK-VF8-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC3]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ] ; CHECK-VF8-NEXT: br label [[FOR_BODY:%.*]] ; CHECK-VF8: for.body: ; @@ -274,7 +380,7 @@ for.body: %arrayidx = getelementptr inbounds i64, ptr %A, i64 %iv store i64 1, ptr %arrayidx, align 1 %iv.next = add nuw nsw i64 %iv, 1 - %exitcond = icmp ne i64 %iv.next, 1024 + %exitcond = icmp ne i64 %iv.next, %n br i1 %exitcond, label %for.body, label %exit exit: @@ -313,7 +419,7 @@ define void @test_pr57912_pointer_induction(ptr %start) #0 { ; CHECK-NEXT: store <vscale x 16 x i8> zeroinitializer, ptr [[TMP19]], align 1 ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP7]] ; CHECK-NEXT: [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] -; CHECK-NEXT: br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]] +; CHECK-NEXT: br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]] ; CHECK: middle.block: ; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 10000, [[N_VEC]] ; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]] @@ -340,7 +446,7 @@ define void @test_pr57912_pointer_induction(ptr %start) #0 { ; CHECK-NEXT: store <vscale x 8 x i8> zeroinitializer, ptr [[TMP28]], align 1 ; CHECK-NEXT: [[INDEX_NEXT8]] = add nuw i64 [[INDEX7]], [[TMP26]] ; CHECK-NEXT: [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT8]], [[N_VEC3]] -; CHECK-NEXT: br i1 [[TMP30]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]] +; CHECK-NEXT: br i1 [[TMP30]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]] ; CHECK: vec.epilog.middle.block: ; CHECK-NEXT: [[CMP_N6:%.*]] = icmp eq i64 10000, [[N_VEC3]] ; CHECK-NEXT: br i1 [[CMP_N6]], label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]] @@ -376,7 +482,7 @@ define void @test_pr57912_pointer_induction(ptr %start) #0 { ; CHECK-VF8-NEXT: store <vscale x 16 x i8> zeroinitializer, ptr [[TMP17]], align 1 ; CHECK-VF8-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] ; CHECK-VF8-NEXT: [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] -; CHECK-VF8-NEXT: br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]] +; CHECK-VF8-NEXT: br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]] ; CHECK-VF8: middle.block: ; CHECK-VF8-NEXT: [[CMP_N:%.*]] = icmp eq i64 10000, [[N_VEC]] ; CHECK-VF8-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]] @@ -395,7 +501,7 @@ define void @test_pr57912_pointer_induction(ptr %start) #0 { ; CHECK-VF8-NEXT: store <8 x i8> zeroinitializer, ptr [[TMP20]], align 1 ; CHECK-VF8-NEXT: [[INDEX_NEXT4]] = add nuw i64 [[INDEX3]], 8 ; CHECK-VF8-NEXT: [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT4]], 10000 -; CHECK-VF8-NEXT: br i1 [[TMP22]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]] +; CHECK-VF8-NEXT: br i1 [[TMP22]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]] ; CHECK-VF8: vec.epilog.middle.block: ; CHECK-VF8-NEXT: br i1 true, label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]] ; CHECK-VF8: vec.epilog.scalar.ph: @@ -420,4 +526,307 @@ exit: ret void } +; Loop with vscale-based trip count vscale x 1033. +define void @trip_count_vscale(ptr noalias %a, ptr noalias %b) vscale_range(1, 16) #0 { +; CHECK-LABEL: @trip_count_vscale( +; CHECK-NEXT: iter.check: +; CHECK-NEXT: [[V:%.*]] = tail call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[N:%.*]] = mul nuw nsw i64 [[V]], 1033 +; CHECK-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 2 +; CHECK-NEXT: br i1 false, label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]] +; CHECK: vector.main.loop.iter.check: +; CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 8 +; CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[N]], [[TMP3]] +; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]] +; CHECK: vector.ph: +; CHECK-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 8 +; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP5]] +; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; CHECK-NEXT: [[TMP6:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP7:%.*]] = mul nuw i64 [[TMP6]], 8 +; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] +; CHECK: vector.body: +; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] +; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds nuw float, ptr [[A:%.*]], i64 [[INDEX]] +; CHECK-NEXT: [[TMP10:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP11:%.*]] = mul nuw i64 [[TMP10]], 4 +; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds nuw float, ptr [[TMP8]], i64 [[TMP11]] +; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4 +; CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP12]], align 4 +; CHECK-NEXT: [[TMP13:%.*]] = getelementptr inbounds nuw float, ptr [[B:%.*]], i64 [[INDEX]] +; CHECK-NEXT: [[TMP15:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP16:%.*]] = mul nuw i64 [[TMP15]], 4 +; CHECK-NEXT: [[TMP17:%.*]] = getelementptr inbounds nuw float, ptr [[TMP13]], i64 [[TMP16]] +; CHECK-NEXT: [[WIDE_LOAD3:%.*]] = load <vscale x 4 x float>, ptr [[TMP13]], align 4 +; CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <vscale x 4 x float>, ptr [[TMP17]], align 4 +; CHECK-NEXT: [[TMP18:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD3]] +; CHECK-NEXT: [[TMP19:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD2]], [[WIDE_LOAD4]] +; CHECK-NEXT: [[TMP20:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP21:%.*]] = mul nuw i64 [[TMP20]], 4 +; CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds nuw float, ptr [[TMP13]], i64 [[TMP21]] +; CHECK-NEXT: store <vscale x 4 x float> [[TMP18]], ptr [[TMP13]], align 4 +; CHECK-NEXT: store <vscale x 4 x float> [[TMP19]], ptr [[TMP22]], align 4 +; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP7]] +; CHECK-NEXT: [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; CHECK-NEXT: br i1 [[TMP23]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]] +; CHECK: middle.block: +; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]] +; CHECK: vec.epilog.iter.check: +; CHECK-NEXT: [[N_VEC_REMAINING:%.*]] = sub i64 [[N]], [[N_VEC]] +; CHECK-NEXT: [[TMP24:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP25:%.*]] = mul nuw i64 [[TMP24]], 2 +; CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], [[TMP25]] +; CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]] +; CHECK: vec.epilog.ph: +; CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ] +; CHECK-NEXT: [[TMP26:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP27:%.*]] = mul nuw i64 [[TMP26]], 2 +; CHECK-NEXT: [[N_MOD_VF5:%.*]] = urem i64 [[N]], [[TMP27]] +; CHECK-NEXT: [[N_VEC6:%.*]] = sub i64 [[N]], [[N_MOD_VF5]] +; CHECK-NEXT: [[TMP28:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP29:%.*]] = mul nuw i64 [[TMP28]], 2 +; CHECK-NEXT: br label [[VEC_EPILOG_VECTOR_BODY:%.*]] +; CHECK: vec.epilog.vector.body: +; CHECK-NEXT: [[INDEX7:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT10:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ] +; CHECK-NEXT: [[TMP30:%.*]] = getelementptr inbounds nuw float, ptr [[A]], i64 [[INDEX7]] +; CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <vscale x 2 x float>, ptr [[TMP30]], align 4 +; CHECK-NEXT: [[TMP32:%.*]] = getelementptr inbounds nuw float, ptr [[B]], i64 [[INDEX7]] +; CHECK-NEXT: [[WIDE_LOAD9:%.*]] = load <vscale x 2 x float>, ptr [[TMP32]], align 4 +; CHECK-NEXT: [[TMP34:%.*]] = fmul <vscale x 2 x float> [[WIDE_LOAD8]], [[WIDE_LOAD9]] +; CHECK-NEXT: store <vscale x 2 x float> [[TMP34]], ptr [[TMP32]], align 4 +; CHECK-NEXT: [[INDEX_NEXT10]] = add nuw i64 [[INDEX7]], [[TMP29]] +; CHECK-NEXT: [[TMP35:%.*]] = icmp eq i64 [[INDEX_NEXT10]], [[N_VEC6]] +; CHECK-NEXT: br i1 [[TMP35]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]] +; CHECK: vec.epilog.middle.block: +; CHECK-NEXT: [[CMP_N11:%.*]] = icmp eq i64 [[N]], [[N_VEC6]] +; CHECK-NEXT: br i1 [[CMP_N11]], label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]] +; CHECK: vec.epilog.scalar.ph: +; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC6]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ] +; CHECK-NEXT: br label [[FOR_BODY:%.*]] +; CHECK: for.body: +; +; CHECK-VF8-LABEL: @trip_count_vscale( +; CHECK-VF8-NEXT: entry: +; CHECK-VF8-NEXT: [[V:%.*]] = tail call i64 @llvm.vscale.i64() +; CHECK-VF8-NEXT: [[N:%.*]] = mul nuw nsw i64 [[V]], 1033 +; CHECK-VF8-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-VF8-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 8 +; CHECK-VF8-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] +; CHECK-VF8: vector.ph: +; CHECK-VF8-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-VF8-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 8 +; CHECK-VF8-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]] +; CHECK-VF8-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; CHECK-VF8-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-VF8-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 8 +; CHECK-VF8-NEXT: br label [[VECTOR_BODY:%.*]] +; CHECK-VF8: vector.body: +; CHECK-VF8-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] +; CHECK-VF8-NEXT: [[TMP6:%.*]] = getelementptr inbounds nuw float, ptr [[A:%.*]], i64 [[INDEX]] +; CHECK-VF8-NEXT: [[TMP8:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-VF8-NEXT: [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 4 +; CHECK-VF8-NEXT: [[TMP10:%.*]] = getelementptr inbounds nuw float, ptr [[TMP6]], i64 [[TMP9]] +; CHECK-VF8-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP6]], align 4 +; CHECK-VF8-NEXT: [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP10]], align 4 +; CHECK-VF8-NEXT: [[TMP11:%.*]] = getelementptr inbounds nuw float, ptr [[B:%.*]], i64 [[INDEX]] +; CHECK-VF8-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-VF8-NEXT: [[TMP14:%.*]] = mul nuw i64 [[TMP13]], 4 +; CHECK-VF8-NEXT: [[TMP15:%.*]] = getelementptr inbounds nuw float, ptr [[TMP11]], i64 [[TMP14]] +; CHECK-VF8-NEXT: [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4 +; CHECK-VF8-NEXT: [[WIDE_LOAD3:%.*]] = load <vscale x 4 x float>, ptr [[TMP15]], align 4 +; CHECK-VF8-NEXT: [[TMP16:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD2]] +; CHECK-VF8-NEXT: [[TMP17:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD1]], [[WIDE_LOAD3]] +; CHECK-VF8-NEXT: [[TMP18:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-VF8-NEXT: [[TMP19:%.*]] = mul nuw i64 [[TMP18]], 4 +; CHECK-VF8-NEXT: [[TMP20:%.*]] = getelementptr inbounds nuw float, ptr [[TMP11]], i64 [[TMP19]] +; CHECK-VF8-NEXT: store <vscale x 4 x float> [[TMP16]], ptr [[TMP11]], align 4 +; CHECK-VF8-NEXT: store <vscale x 4 x float> [[TMP17]], ptr [[TMP20]], align 4 +; CHECK-VF8-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] +; CHECK-VF8-NEXT: [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; CHECK-VF8-NEXT: br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]] +; CHECK-VF8: middle.block: +; CHECK-VF8-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; CHECK-VF8-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]] +; CHECK-VF8: scalar.ph: +; CHECK-VF8-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] +; CHECK-VF8-NEXT: br label [[FOR_BODY:%.*]] +; CHECK-VF8: for.body: +; +entry: + %v = tail call i64 @llvm.vscale.i64() + %n = mul nuw nsw i64 %v, 1033 + br label %for.body + +for.body: + %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ] + %arrayidx = getelementptr inbounds nuw float, ptr %a, i64 %iv + %l.a = load float, ptr %arrayidx, align 4 + %arrayidx3 = getelementptr inbounds nuw float, ptr %b, i64 %iv + %l.b = load float, ptr %arrayidx3, align 4 + %mul4 = fmul float %l.a, %l.b + store float %mul4, ptr %arrayidx3, align 4 + %iv.next = add nuw nsw i64 %iv, 1 + %ec = icmp eq i64 %iv.next, %n + br i1 %ec, label %exit, label %for.body + +exit: + ret void +} + +; Loop with vscale-based trip count vscale x 1024. +; TODO: No epilogue vectorizations should remain when choosing VF = vscale x 4. +define void @trip_count_vscale_no_epilogue_iterations(ptr noalias %a, ptr noalias %b) vscale_range(1, 16) #0 { +; CHECK-LABEL: @trip_count_vscale_no_epilogue_iterations( +; CHECK-NEXT: iter.check: +; CHECK-NEXT: [[V:%.*]] = tail call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[N:%.*]] = mul nuw nsw i64 [[V]], 1024 +; CHECK-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 2 +; CHECK-NEXT: br i1 false, label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]] +; CHECK: vector.main.loop.iter.check: +; CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 8 +; CHECK-NEXT: [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[N]], [[TMP3]] +; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]] +; CHECK: vector.ph: +; CHECK-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 8 +; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP5]] +; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; CHECK-NEXT: [[TMP6:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP7:%.*]] = mul nuw i64 [[TMP6]], 8 +; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] +; CHECK: vector.body: +; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] +; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds nuw float, ptr [[A:%.*]], i64 [[INDEX]] +; CHECK-NEXT: [[TMP10:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP11:%.*]] = mul nuw i64 [[TMP10]], 4 +; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds nuw float, ptr [[TMP8]], i64 [[TMP11]] +; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4 +; CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP12]], align 4 +; CHECK-NEXT: [[TMP13:%.*]] = getelementptr inbounds nuw float, ptr [[B:%.*]], i64 [[INDEX]] +; CHECK-NEXT: [[TMP15:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP16:%.*]] = mul nuw i64 [[TMP15]], 4 +; CHECK-NEXT: [[TMP17:%.*]] = getelementptr inbounds nuw float, ptr [[TMP13]], i64 [[TMP16]] +; CHECK-NEXT: [[WIDE_LOAD3:%.*]] = load <vscale x 4 x float>, ptr [[TMP13]], align 4 +; CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <vscale x 4 x float>, ptr [[TMP17]], align 4 +; CHECK-NEXT: [[TMP18:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD3]] +; CHECK-NEXT: [[TMP19:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD2]], [[WIDE_LOAD4]] +; CHECK-NEXT: [[TMP20:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP21:%.*]] = mul nuw i64 [[TMP20]], 4 +; CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds nuw float, ptr [[TMP13]], i64 [[TMP21]] +; CHECK-NEXT: store <vscale x 4 x float> [[TMP18]], ptr [[TMP13]], align 4 +; CHECK-NEXT: store <vscale x 4 x float> [[TMP19]], ptr [[TMP22]], align 4 +; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP7]] +; CHECK-NEXT: [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; CHECK-NEXT: br i1 [[TMP23]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]] +; CHECK: middle.block: +; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]] +; CHECK: vec.epilog.iter.check: +; CHECK-NEXT: [[N_VEC_REMAINING:%.*]] = sub i64 [[N]], [[N_VEC]] +; CHECK-NEXT: [[TMP24:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP25:%.*]] = mul nuw i64 [[TMP24]], 2 +; CHECK-NEXT: [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], [[TMP25]] +; CHECK-NEXT: br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]] +; CHECK: vec.epilog.ph: +; CHECK-NEXT: [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ] +; CHECK-NEXT: [[TMP26:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP27:%.*]] = mul nuw i64 [[TMP26]], 2 +; CHECK-NEXT: [[N_MOD_VF5:%.*]] = urem i64 [[N]], [[TMP27]] +; CHECK-NEXT: [[N_VEC6:%.*]] = sub i64 [[N]], [[N_MOD_VF5]] +; CHECK-NEXT: [[TMP28:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP29:%.*]] = mul nuw i64 [[TMP28]], 2 +; CHECK-NEXT: br label [[VEC_EPILOG_VECTOR_BODY:%.*]] +; CHECK: vec.epilog.vector.body: +; CHECK-NEXT: [[INDEX7:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT10:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ] +; CHECK-NEXT: [[TMP30:%.*]] = getelementptr inbounds nuw float, ptr [[A]], i64 [[INDEX7]] +; CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <vscale x 2 x float>, ptr [[TMP30]], align 4 +; CHECK-NEXT: [[TMP32:%.*]] = getelementptr inbounds nuw float, ptr [[B]], i64 [[INDEX7]] +; CHECK-NEXT: [[WIDE_LOAD9:%.*]] = load <vscale x 2 x float>, ptr [[TMP32]], align 4 +; CHECK-NEXT: [[TMP34:%.*]] = fmul <vscale x 2 x float> [[WIDE_LOAD8]], [[WIDE_LOAD9]] +; CHECK-NEXT: store <vscale x 2 x float> [[TMP34]], ptr [[TMP32]], align 4 +; CHECK-NEXT: [[INDEX_NEXT10]] = add nuw i64 [[INDEX7]], [[TMP29]] +; CHECK-NEXT: [[TMP35:%.*]] = icmp eq i64 [[INDEX_NEXT10]], [[N_VEC6]] +; CHECK-NEXT: br i1 [[TMP35]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]] +; CHECK: vec.epilog.middle.block: +; CHECK-NEXT: [[CMP_N11:%.*]] = icmp eq i64 [[N]], [[N_VEC6]] +; CHECK-NEXT: br i1 [[CMP_N11]], label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]] +; CHECK: vec.epilog.scalar.ph: +; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC6]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ] +; CHECK-NEXT: br label [[FOR_BODY:%.*]] +; CHECK: for.body: +; +; CHECK-VF8-LABEL: @trip_count_vscale_no_epilogue_iterations( +; CHECK-VF8-NEXT: entry: +; CHECK-VF8-NEXT: [[V:%.*]] = tail call i64 @llvm.vscale.i64() +; CHECK-VF8-NEXT: [[N:%.*]] = mul nuw nsw i64 [[V]], 1024 +; CHECK-VF8-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-VF8-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 8 +; CHECK-VF8-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] +; CHECK-VF8: vector.ph: +; CHECK-VF8-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-VF8-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 8 +; CHECK-VF8-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]] +; CHECK-VF8-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; CHECK-VF8-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-VF8-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 8 +; CHECK-VF8-NEXT: br label [[VECTOR_BODY:%.*]] +; CHECK-VF8: vector.body: +; CHECK-VF8-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] +; CHECK-VF8-NEXT: [[TMP6:%.*]] = getelementptr inbounds nuw float, ptr [[A:%.*]], i64 [[INDEX]] +; CHECK-VF8-NEXT: [[TMP8:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-VF8-NEXT: [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 4 +; CHECK-VF8-NEXT: [[TMP10:%.*]] = getelementptr inbounds nuw float, ptr [[TMP6]], i64 [[TMP9]] +; CHECK-VF8-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP6]], align 4 +; CHECK-VF8-NEXT: [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP10]], align 4 +; CHECK-VF8-NEXT: [[TMP11:%.*]] = getelementptr inbounds nuw float, ptr [[B:%.*]], i64 [[INDEX]] +; CHECK-VF8-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-VF8-NEXT: [[TMP14:%.*]] = mul nuw i64 [[TMP13]], 4 +; CHECK-VF8-NEXT: [[TMP15:%.*]] = getelementptr inbounds nuw float, ptr [[TMP11]], i64 [[TMP14]] +; CHECK-VF8-NEXT: [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4 +; CHECK-VF8-NEXT: [[WIDE_LOAD3:%.*]] = load <vscale x 4 x float>, ptr [[TMP15]], align 4 +; CHECK-VF8-NEXT: [[TMP16:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD2]] +; CHECK-VF8-NEXT: [[TMP17:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD1]], [[WIDE_LOAD3]] +; CHECK-VF8-NEXT: [[TMP18:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-VF8-NEXT: [[TMP19:%.*]] = mul nuw i64 [[TMP18]], 4 +; CHECK-VF8-NEXT: [[TMP20:%.*]] = getelementptr inbounds nuw float, ptr [[TMP11]], i64 [[TMP19]] +; CHECK-VF8-NEXT: store <vscale x 4 x float> [[TMP16]], ptr [[TMP11]], align 4 +; CHECK-VF8-NEXT: store <vscale x 4 x float> [[TMP17]], ptr [[TMP20]], align 4 +; CHECK-VF8-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] +; CHECK-VF8-NEXT: [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; CHECK-VF8-NEXT: br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]] +; CHECK-VF8: middle.block: +; CHECK-VF8-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; CHECK-VF8-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]] +; CHECK-VF8: scalar.ph: +; CHECK-VF8-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] +; CHECK-VF8-NEXT: br label [[FOR_BODY:%.*]] +; CHECK-VF8: for.body: +; +entry: + %v = tail call i64 @llvm.vscale.i64() + %n = mul nuw nsw i64 %v, 1024 + br label %for.body + +for.body: + %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ] + %arrayidx = getelementptr inbounds nuw float, ptr %a, i64 %iv + %l.a = load float, ptr %arrayidx, align 4 + %arrayidx3 = getelementptr inbounds nuw float, ptr %b, i64 %iv + %l.b = load float, ptr %arrayidx3, align 4 + %mul4 = fmul float %l.a, %l.b + store float %mul4, ptr %arrayidx3, align 4 + %iv.next = add nuw nsw i64 %iv, 1 + %ec = icmp eq i64 %iv.next, %n + br i1 %ec, label %exit, label %for.body + +exit: + ret void +} + attributes #0 = { "target-features"="+sve" } diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/bf16.ll b/llvm/test/Transforms/LoopVectorize/RISCV/bf16.ll index d5b25bf..21266e5 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/bf16.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/bf16.ll @@ -1,5 +1,6 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5 ; RUN: opt < %s -passes=loop-vectorize -mtriple riscv64 -mattr=+v -S | FileCheck %s -check-prefix=NO-ZVFBFMIN +; RUN: opt < %s -passes=loop-vectorize -mtriple riscv64 -mattr=+v -S -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue | FileCheck %s -check-prefix=NO-ZVFBFMIN-PREDICATED ; RUN: opt < %s -passes=loop-vectorize -mtriple riscv64 -mattr=+v,+zvfbfmin -S | FileCheck %s -check-prefix=ZVFBFMIN define void @fadd(ptr noalias %a, ptr noalias %b, i64 %n) { @@ -21,6 +22,24 @@ define void @fadd(ptr noalias %a, ptr noalias %b, i64 %n) { ; NO-ZVFBFMIN: [[EXIT]]: ; NO-ZVFBFMIN-NEXT: ret void ; +; NO-ZVFBFMIN-PREDICATED-LABEL: define void @fadd( +; NO-ZVFBFMIN-PREDICATED-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] { +; NO-ZVFBFMIN-PREDICATED-NEXT: [[ENTRY:.*]]: +; NO-ZVFBFMIN-PREDICATED-NEXT: br label %[[LOOP:.*]] +; NO-ZVFBFMIN-PREDICATED: [[LOOP]]: +; NO-ZVFBFMIN-PREDICATED-NEXT: [[I:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[I_NEXT:%.*]], %[[LOOP]] ] +; NO-ZVFBFMIN-PREDICATED-NEXT: [[A_GEP:%.*]] = getelementptr bfloat, ptr [[A]], i64 [[I]] +; NO-ZVFBFMIN-PREDICATED-NEXT: [[B_GEP:%.*]] = getelementptr bfloat, ptr [[B]], i64 [[I]] +; NO-ZVFBFMIN-PREDICATED-NEXT: [[X:%.*]] = load bfloat, ptr [[A_GEP]], align 2 +; NO-ZVFBFMIN-PREDICATED-NEXT: [[Y:%.*]] = load bfloat, ptr [[B_GEP]], align 2 +; NO-ZVFBFMIN-PREDICATED-NEXT: [[Z:%.*]] = fadd bfloat [[X]], [[Y]] +; NO-ZVFBFMIN-PREDICATED-NEXT: store bfloat [[Z]], ptr [[A_GEP]], align 2 +; NO-ZVFBFMIN-PREDICATED-NEXT: [[I_NEXT]] = add i64 [[I]], 1 +; NO-ZVFBFMIN-PREDICATED-NEXT: [[DONE:%.*]] = icmp eq i64 [[I_NEXT]], [[N]] +; NO-ZVFBFMIN-PREDICATED-NEXT: br i1 [[DONE]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-ZVFBFMIN-PREDICATED: [[EXIT]]: +; NO-ZVFBFMIN-PREDICATED-NEXT: ret void +; ; ZVFBFMIN-LABEL: define void @fadd( ; ZVFBFMIN-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] { ; ZVFBFMIN-NEXT: [[ENTRY:.*]]: @@ -133,6 +152,54 @@ define void @vfwmaccbf16.vv(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 ; NO-ZVFBFMIN: [[EXIT]]: ; NO-ZVFBFMIN-NEXT: ret void ; +; NO-ZVFBFMIN-PREDICATED-LABEL: define void @vfwmaccbf16.vv( +; NO-ZVFBFMIN-PREDICATED-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { +; NO-ZVFBFMIN-PREDICATED-NEXT: [[ENTRY:.*]]: +; NO-ZVFBFMIN-PREDICATED-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4 +; NO-ZVFBFMIN-PREDICATED-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]] +; NO-ZVFBFMIN-PREDICATED: [[VECTOR_PH]]: +; NO-ZVFBFMIN-PREDICATED-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 4 +; NO-ZVFBFMIN-PREDICATED-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-ZVFBFMIN-PREDICATED-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-ZVFBFMIN-PREDICATED: [[VECTOR_BODY]]: +; NO-ZVFBFMIN-PREDICATED-NEXT: [[I:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-ZVFBFMIN-PREDICATED-NEXT: [[A_GEP:%.*]] = getelementptr bfloat, ptr [[A]], i64 [[I]] +; NO-ZVFBFMIN-PREDICATED-NEXT: [[B_GEP:%.*]] = getelementptr bfloat, ptr [[B]], i64 [[I]] +; NO-ZVFBFMIN-PREDICATED-NEXT: [[C_GEP:%.*]] = getelementptr float, ptr [[C]], i64 [[I]] +; NO-ZVFBFMIN-PREDICATED-NEXT: [[WIDE_MASKED_LOAD:%.*]] = load <4 x bfloat>, ptr [[A_GEP]], align 2 +; NO-ZVFBFMIN-PREDICATED-NEXT: [[WIDE_MASKED_LOAD3:%.*]] = load <4 x bfloat>, ptr [[B_GEP]], align 2 +; NO-ZVFBFMIN-PREDICATED-NEXT: [[WIDE_MASKED_LOAD4:%.*]] = load <4 x float>, ptr [[C_GEP]], align 4 +; NO-ZVFBFMIN-PREDICATED-NEXT: [[TMP4:%.*]] = fpext <4 x bfloat> [[WIDE_MASKED_LOAD]] to <4 x float> +; NO-ZVFBFMIN-PREDICATED-NEXT: [[TMP5:%.*]] = fpext <4 x bfloat> [[WIDE_MASKED_LOAD3]] to <4 x float> +; NO-ZVFBFMIN-PREDICATED-NEXT: [[TMP6:%.*]] = call <4 x float> @llvm.fmuladd.v4f32(<4 x float> [[TMP4]], <4 x float> [[TMP5]], <4 x float> [[WIDE_MASKED_LOAD4]]) +; NO-ZVFBFMIN-PREDICATED-NEXT: store <4 x float> [[TMP6]], ptr [[C_GEP]], align 4 +; NO-ZVFBFMIN-PREDICATED-NEXT: [[INDEX_NEXT]] = add nuw i64 [[I]], 4 +; NO-ZVFBFMIN-PREDICATED-NEXT: [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-ZVFBFMIN-PREDICATED-NEXT: br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] +; NO-ZVFBFMIN-PREDICATED: [[MIDDLE_BLOCK]]: +; NO-ZVFBFMIN-PREDICATED-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-ZVFBFMIN-PREDICATED-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-ZVFBFMIN-PREDICATED: [[SCALAR_PH]]: +; NO-ZVFBFMIN-PREDICATED-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ] +; NO-ZVFBFMIN-PREDICATED-NEXT: br label %[[LOOP:.*]] +; NO-ZVFBFMIN-PREDICATED: [[LOOP]]: +; NO-ZVFBFMIN-PREDICATED-NEXT: [[I1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[I_NEXT:%.*]], %[[LOOP]] ] +; NO-ZVFBFMIN-PREDICATED-NEXT: [[A_GEP1:%.*]] = getelementptr bfloat, ptr [[A]], i64 [[I1]] +; NO-ZVFBFMIN-PREDICATED-NEXT: [[B_GEP1:%.*]] = getelementptr bfloat, ptr [[B]], i64 [[I1]] +; NO-ZVFBFMIN-PREDICATED-NEXT: [[C_GEP1:%.*]] = getelementptr float, ptr [[C]], i64 [[I1]] +; NO-ZVFBFMIN-PREDICATED-NEXT: [[X:%.*]] = load bfloat, ptr [[A_GEP1]], align 2 +; NO-ZVFBFMIN-PREDICATED-NEXT: [[Y:%.*]] = load bfloat, ptr [[B_GEP1]], align 2 +; NO-ZVFBFMIN-PREDICATED-NEXT: [[Z:%.*]] = load float, ptr [[C_GEP1]], align 4 +; NO-ZVFBFMIN-PREDICATED-NEXT: [[X_EXT:%.*]] = fpext bfloat [[X]] to float +; NO-ZVFBFMIN-PREDICATED-NEXT: [[Y_EXT:%.*]] = fpext bfloat [[Y]] to float +; NO-ZVFBFMIN-PREDICATED-NEXT: [[FMULADD:%.*]] = call float @llvm.fmuladd.f32(float [[X_EXT]], float [[Y_EXT]], float [[Z]]) +; NO-ZVFBFMIN-PREDICATED-NEXT: store float [[FMULADD]], ptr [[C_GEP1]], align 4 +; NO-ZVFBFMIN-PREDICATED-NEXT: [[I_NEXT]] = add i64 [[I1]], 1 +; NO-ZVFBFMIN-PREDICATED-NEXT: [[DONE:%.*]] = icmp eq i64 [[I_NEXT]], [[N]] +; NO-ZVFBFMIN-PREDICATED-NEXT: br i1 [[DONE]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]] +; NO-ZVFBFMIN-PREDICATED: [[EXIT]]: +; NO-ZVFBFMIN-PREDICATED-NEXT: ret void +; ; ZVFBFMIN-LABEL: define void @vfwmaccbf16.vv( ; ZVFBFMIN-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; ZVFBFMIN-NEXT: [[ENTRY:.*]]: @@ -213,6 +280,11 @@ exit: ; NO-ZVFBFMIN: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"} ; NO-ZVFBFMIN: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]} ;. +; NO-ZVFBFMIN-PREDICATED: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]} +; NO-ZVFBFMIN-PREDICATED: [[META1]] = !{!"llvm.loop.isvectorized", i32 1} +; NO-ZVFBFMIN-PREDICATED: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"} +; NO-ZVFBFMIN-PREDICATED: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]} +;. ; ZVFBFMIN: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]} ; ZVFBFMIN: [[META1]] = !{!"llvm.loop.isvectorized", i32 1} ; ZVFBFMIN: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"} diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/evl-compatible-loops.ll b/llvm/test/Transforms/LoopVectorize/RISCV/evl-compatible-loops.ll index 32d17b9..5d032e6 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/evl-compatible-loops.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/evl-compatible-loops.ll @@ -1,6 +1,6 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4 -; RUN: opt -passes=loop-vectorize -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: opt -passes=loop-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s ; Make sure we do not vectorize a loop with a widened int induction. diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/f16.ll b/llvm/test/Transforms/LoopVectorize/RISCV/f16.ll index 5b56552..53e43e1 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/f16.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/f16.ll @@ -1,5 +1,6 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5 ; RUN: opt < %s -passes=loop-vectorize -mtriple riscv64 -mattr=+v -S | FileCheck %s -check-prefix=NO-ZVFHMIN +; RUN: opt < %s -passes=loop-vectorize -mtriple riscv64 -mattr=+v -S -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue | FileCheck %s -check-prefix=NO-ZVFHMIN-PREDICATED ; RUN: opt < %s -passes=loop-vectorize -mtriple riscv64 -mattr=+v,+zvfhmin -S | FileCheck %s -check-prefix=ZVFHMIN define void @fadd(ptr noalias %a, ptr noalias %b, i64 %n) { @@ -21,6 +22,24 @@ define void @fadd(ptr noalias %a, ptr noalias %b, i64 %n) { ; NO-ZVFHMIN: [[EXIT]]: ; NO-ZVFHMIN-NEXT: ret void ; +; NO-ZVFHMIN-PREDICATED-LABEL: define void @fadd( +; NO-ZVFHMIN-PREDICATED-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] { +; NO-ZVFHMIN-PREDICATED-NEXT: [[ENTRY:.*]]: +; NO-ZVFHMIN-PREDICATED-NEXT: br label %[[LOOP:.*]] +; NO-ZVFHMIN-PREDICATED: [[LOOP]]: +; NO-ZVFHMIN-PREDICATED-NEXT: [[I:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[I_NEXT:%.*]], %[[LOOP]] ] +; NO-ZVFHMIN-PREDICATED-NEXT: [[A_GEP:%.*]] = getelementptr half, ptr [[A]], i64 [[I]] +; NO-ZVFHMIN-PREDICATED-NEXT: [[B_GEP:%.*]] = getelementptr half, ptr [[B]], i64 [[I]] +; NO-ZVFHMIN-PREDICATED-NEXT: [[X:%.*]] = load half, ptr [[A_GEP]], align 2 +; NO-ZVFHMIN-PREDICATED-NEXT: [[Y:%.*]] = load half, ptr [[B_GEP]], align 2 +; NO-ZVFHMIN-PREDICATED-NEXT: [[Z:%.*]] = fadd half [[X]], [[Y]] +; NO-ZVFHMIN-PREDICATED-NEXT: store half [[Z]], ptr [[A_GEP]], align 2 +; NO-ZVFHMIN-PREDICATED-NEXT: [[I_NEXT]] = add i64 [[I]], 1 +; NO-ZVFHMIN-PREDICATED-NEXT: [[DONE:%.*]] = icmp eq i64 [[I_NEXT]], [[N]] +; NO-ZVFHMIN-PREDICATED-NEXT: br i1 [[DONE]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-ZVFHMIN-PREDICATED: [[EXIT]]: +; NO-ZVFHMIN-PREDICATED-NEXT: ret void +; ; ZVFHMIN-LABEL: define void @fadd( ; ZVFHMIN-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] { ; ZVFHMIN-NEXT: [[ENTRY:.*]]: diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll index f44cbf2..5b2d308 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll @@ -1,8 +1,8 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py ; RUN: opt -mtriple riscv64-linux-gnu -mattr=+v,+d -passes=loop-vectorize < %s -S -o - | FileCheck %s -check-prefix=OUTLOOP ; RUN: opt -mtriple riscv64-linux-gnu -mattr=+v,+d -passes=loop-vectorize -prefer-inloop-reductions < %s -S -o - | FileCheck %s -check-prefix=INLOOP -; RUN: opt -passes=loop-vectorize -force-tail-folding-style=data-with-evl -prefer-predicate-over-epilogue=predicate-dont-vectorize -mtriple=riscv64 -mattr=+v -S < %s 2>&1 | FileCheck --check-prefix=IF-EVL-OUTLOOP %s -; RUN: opt -passes=loop-vectorize -prefer-inloop-reductions -force-tail-folding-style=data-with-evl -prefer-predicate-over-epilogue=predicate-dont-vectorize -mtriple=riscv64 -mattr=+v -S < %s 2>&1 | FileCheck --check-prefix=IF-EVL-INLOOP %s +; RUN: opt -passes=loop-vectorize -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue -mtriple=riscv64 -mattr=+v -S < %s 2>&1 | FileCheck --check-prefix=IF-EVL-OUTLOOP %s +; RUN: opt -passes=loop-vectorize -prefer-inloop-reductions -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue -mtriple=riscv64 -mattr=+v -S < %s 2>&1 | FileCheck --check-prefix=IF-EVL-INLOOP %s target datalayout = "e-m:e-p:64:64-i64:64-i128:128-n64-S128" diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/interleaved-masked-access.ll b/llvm/test/Transforms/LoopVectorize/RISCV/interleaved-masked-access.ll index dbe6f27..c35a6ab 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/interleaved-masked-access.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/interleaved-masked-access.ll @@ -1,7 +1,7 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --filter-out-after "^scalar.ph:" --version 2 ; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize,instcombine -mattr=+v -prefer-predicate-over-epilogue=scalar-epilogue %s 2>&1 | FileCheck %s -check-prefix=SCALAR_EPILOGUE -; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize,instcombine -mattr=+v -prefer-predicate-over-epilogue=predicate-dont-vectorize -force-tail-folding-style=data %s 2>&1 | FileCheck %s -check-prefix=PREDICATED_DATA -; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize,instcombine -mattr=+v -prefer-predicate-over-epilogue=predicate-dont-vectorize -force-tail-folding-style=data-with-evl %s 2>&1 | FileCheck %s -check-prefix=PREDICATED_DATA-WITH-EVL +; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize,instcombine -mattr=+v -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue -force-tail-folding-style=data %s 2>&1 | FileCheck %s -check-prefix=PREDICATED_DATA +; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize,instcombine -mattr=+v -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue %s 2>&1 | FileCheck %s -check-prefix=PREDICATED_DATA-WITH-EVL target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128" diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/only-compute-cost-for-vplan-vfs.ll b/llvm/test/Transforms/LoopVectorize/RISCV/only-compute-cost-for-vplan-vfs.ll index 4f3b8d8..0afe04e 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/only-compute-cost-for-vplan-vfs.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/only-compute-cost-for-vplan-vfs.ll @@ -1,5 +1,5 @@ -; RUN: opt -passes=loop-vectorize -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: opt -passes=loop-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S -debug %s 2>&1 | FileCheck %s ; REQUIRES: asserts diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/preserve-dbg-loc.ll b/llvm/test/Transforms/LoopVectorize/RISCV/preserve-dbg-loc.ll index e4892db..f4817bb 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/preserve-dbg-loc.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/preserve-dbg-loc.ll @@ -1,6 +1,5 @@ ; RUN: opt -passes=debugify,loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -S < %s 2>&1 | FileCheck --check-prefix=DEBUGLOC %s ; Testing the debug locations of the generated vector intrinsic is same as diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/scalable-tailfold.ll b/llvm/test/Transforms/LoopVectorize/RISCV/scalable-tailfold.ll index 8d8ea5a..f6f8f6a 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/scalable-tailfold.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/scalable-tailfold.ll @@ -1,5 +1,5 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py -; RUN: opt < %s -passes=loop-vectorize -scalable-vectorization=on -prefer-predicate-over-epilogue=predicate-dont-vectorize -mtriple riscv64-linux-gnu -mattr=+v,+f -S 2>%t | FileCheck %s -check-prefix=CHECK +; RUN: opt < %s -passes=loop-vectorize -scalable-vectorization=on -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue -mtriple riscv64-linux-gnu -mattr=+v,+f -S 2>%t | FileCheck %s -check-prefix=CHECK ; Exercise tail folding on RISCV w/scalable vectors. @@ -330,17 +330,44 @@ for.end: define i64 @uniform_load(ptr noalias nocapture %a, ptr noalias nocapture %b, i64 %n) { ; CHECK-LABEL: @uniform_load( ; CHECK-NEXT: entry: +; CHECK-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 2 +; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1025, [[TMP1]] +; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] +; CHECK: vector.ph: +; CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 2 +; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 1025, [[TMP3]] +; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 1025, [[N_MOD_VF]] +; CHECK-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2 ; CHECK-NEXT: br label [[FOR_BODY:%.*]] -; CHECK: for.body: -; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ] +; CHECK: vector.body: +; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ] ; CHECK-NEXT: [[V:%.*]] = load i64, ptr [[B:%.*]], align 8 +; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[V]], i64 0 +; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer ; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[IV]] -; CHECK-NEXT: store i64 [[V]], ptr [[ARRAYIDX]], align 8 -; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 -; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025 -; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END:%.*]], label [[FOR_BODY]] +; CHECK-NEXT: store <vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr [[ARRAYIDX]], align 8 +; CHECK-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], [[TMP5]] +; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]] +; CHECK-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]] +; CHECK: middle.block: +; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 1025, [[N_VEC]] +; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]] +; CHECK: scalar.ph: +; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ] +; CHECK-NEXT: br label [[FOR_BODY1:%.*]] +; CHECK: for.body: +; CHECK-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ] +; CHECK-NEXT: [[V1:%.*]] = load i64, ptr [[B]], align 8 +; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV1]] +; CHECK-NEXT: store i64 [[V1]], ptr [[ARRAYIDX1]], align 8 +; CHECK-NEXT: [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1 +; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], 1025 +; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY1]], !llvm.loop [[LOOP14:![0-9]+]] ; CHECK: for.end: -; CHECK-NEXT: [[V_LCSSA:%.*]] = phi i64 [ [[V]], [[FOR_BODY]] ] +; CHECK-NEXT: [[V_LCSSA:%.*]] = phi i64 [ [[V1]], [[FOR_BODY1]] ], [ [[V]], [[MIDDLE_BLOCK]] ] ; CHECK-NEXT: ret i64 [[V_LCSSA]] ; entry: @@ -389,7 +416,7 @@ define void @vector_add_trip1024(ptr noalias nocapture %a, i64 %v, i64 %n) { ; CHECK-NEXT: [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP8]], [[EVL_BASED_IV]] ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]] ; CHECK-NEXT: [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] -; CHECK-NEXT: br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]] +; CHECK-NEXT: br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]] ; CHECK: middle.block: ; CHECK-NEXT: br label [[FOR_END:%.*]] ; CHECK: scalar.ph: @@ -403,7 +430,7 @@ define void @vector_add_trip1024(ptr noalias nocapture %a, i64 %v, i64 %n) { ; CHECK-NEXT: store i64 [[ADD]], ptr [[ARRAYIDX]], align 8 ; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 ; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1024 -; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]] +; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]] ; CHECK: for.end: ; CHECK-NEXT: ret void ; diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-bin-unary-ops-args.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-bin-unary-ops-args.ll index f90c450..f9395aa 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-bin-unary-ops-args.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-bin-unary-ops-args.ll @@ -1,12 +1,10 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5 ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s --check-prefix=IF-EVL ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s --check-prefix=NO-VP @@ -69,9 +67,45 @@ define void @test_and(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-LABEL: define void @test_and( ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0:[0-9]+]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[A2:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[B1:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16 +; NO-VP-NEXT: [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16 +; NO-VP-NEXT: [[TMP4:%.*]] = sub i64 [[B1]], [[A2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP5:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1 +; NO-VP-NEXT: [[TMP10:%.*]] = and <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 1) +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]] +; NO-VP-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[DEC]] = add nsw i64 [[LEN]], 1 ; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1 @@ -79,7 +113,7 @@ define void @test_and(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]] ; NO-VP-NEXT: store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1 ; NO-VP-NEXT: [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100 -; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]] ; NO-VP: [[FINISH_LOOPEXIT]]: ; NO-VP-NEXT: ret void ; @@ -160,9 +194,45 @@ define void @test_or(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-LABEL: define void @test_or( ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[A2:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[B1:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16 +; NO-VP-NEXT: [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16 +; NO-VP-NEXT: [[TMP4:%.*]] = sub i64 [[B1]], [[A2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP5:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1 +; NO-VP-NEXT: [[TMP10:%.*]] = or <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 1) +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]] +; NO-VP-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[DEC]] = add nsw i64 [[LEN]], 1 ; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1 @@ -170,7 +240,7 @@ define void @test_or(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]] ; NO-VP-NEXT: store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1 ; NO-VP-NEXT: [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100 -; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP5:![0-9]+]] ; NO-VP: [[FINISH_LOOPEXIT]]: ; NO-VP-NEXT: ret void ; @@ -251,9 +321,45 @@ define void @test_xor(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-LABEL: define void @test_xor( ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[A2:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[B1:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16 +; NO-VP-NEXT: [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16 +; NO-VP-NEXT: [[TMP4:%.*]] = sub i64 [[B1]], [[A2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP5:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1 +; NO-VP-NEXT: [[TMP10:%.*]] = xor <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 1) +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]] +; NO-VP-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[DEC]] = add nsw i64 [[LEN]], 1 ; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1 @@ -261,7 +367,7 @@ define void @test_xor(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]] ; NO-VP-NEXT: store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1 ; NO-VP-NEXT: [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100 -; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP7:![0-9]+]] ; NO-VP: [[FINISH_LOOPEXIT]]: ; NO-VP-NEXT: ret void ; @@ -342,9 +448,45 @@ define void @test_shl(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-LABEL: define void @test_shl( ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[A2:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[B1:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16 +; NO-VP-NEXT: [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16 +; NO-VP-NEXT: [[TMP4:%.*]] = sub i64 [[B1]], [[A2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP5:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1 +; NO-VP-NEXT: [[TMP10:%.*]] = shl <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 1) +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]] +; NO-VP-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[DEC]] = add nsw i64 [[LEN]], 1 ; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1 @@ -352,7 +494,7 @@ define void @test_shl(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]] ; NO-VP-NEXT: store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1 ; NO-VP-NEXT: [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100 -; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP9:![0-9]+]] ; NO-VP: [[FINISH_LOOPEXIT]]: ; NO-VP-NEXT: ret void ; @@ -433,9 +575,45 @@ define void @test_lshr(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-LABEL: define void @test_lshr( ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[A2:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[B1:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16 +; NO-VP-NEXT: [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16 +; NO-VP-NEXT: [[TMP4:%.*]] = sub i64 [[B1]], [[A2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP5:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1 +; NO-VP-NEXT: [[TMP10:%.*]] = lshr <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 1) +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]] +; NO-VP-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[DEC]] = add nsw i64 [[LEN]], 1 ; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1 @@ -443,7 +621,7 @@ define void @test_lshr(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]] ; NO-VP-NEXT: store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1 ; NO-VP-NEXT: [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100 -; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP11:![0-9]+]] ; NO-VP: [[FINISH_LOOPEXIT]]: ; NO-VP-NEXT: ret void ; @@ -524,9 +702,45 @@ define void @test_ashr(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-LABEL: define void @test_ashr( ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[A2:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[B1:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16 +; NO-VP-NEXT: [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16 +; NO-VP-NEXT: [[TMP4:%.*]] = sub i64 [[B1]], [[A2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP5:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1 +; NO-VP-NEXT: [[TMP10:%.*]] = ashr <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 1) +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]] +; NO-VP-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[DEC]] = add nsw i64 [[LEN]], 1 ; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1 @@ -534,7 +748,7 @@ define void @test_ashr(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]] ; NO-VP-NEXT: store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1 ; NO-VP-NEXT: [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100 -; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP13:![0-9]+]] ; NO-VP: [[FINISH_LOOPEXIT]]: ; NO-VP-NEXT: ret void ; @@ -615,9 +829,45 @@ define void @test_add(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-LABEL: define void @test_add( ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[A2:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[B1:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16 +; NO-VP-NEXT: [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16 +; NO-VP-NEXT: [[TMP4:%.*]] = sub i64 [[B1]], [[A2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP5:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1 +; NO-VP-NEXT: [[TMP10:%.*]] = add <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 1) +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]] +; NO-VP-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[DEC]] = add nsw i64 [[LEN]], 1 ; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1 @@ -625,7 +875,7 @@ define void @test_add(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]] ; NO-VP-NEXT: store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1 ; NO-VP-NEXT: [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100 -; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP15:![0-9]+]] ; NO-VP: [[FINISH_LOOPEXIT]]: ; NO-VP-NEXT: ret void ; @@ -706,9 +956,45 @@ define void @test_sub(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-LABEL: define void @test_sub( ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[A2:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[B1:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16 +; NO-VP-NEXT: [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16 +; NO-VP-NEXT: [[TMP4:%.*]] = sub i64 [[B1]], [[A2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP5:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1 +; NO-VP-NEXT: [[TMP10:%.*]] = sub <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 1) +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]] +; NO-VP-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[DEC]] = add nsw i64 [[LEN]], 1 ; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1 @@ -716,7 +1002,7 @@ define void @test_sub(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]] ; NO-VP-NEXT: store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1 ; NO-VP-NEXT: [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100 -; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP17:![0-9]+]] ; NO-VP: [[FINISH_LOOPEXIT]]: ; NO-VP-NEXT: ret void ; @@ -797,9 +1083,45 @@ define void @test_mul(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-LABEL: define void @test_mul( ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[A2:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[B1:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16 +; NO-VP-NEXT: [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16 +; NO-VP-NEXT: [[TMP4:%.*]] = sub i64 [[B1]], [[A2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP5:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1 +; NO-VP-NEXT: [[TMP10:%.*]] = mul <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 3) +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]] +; NO-VP-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[DEC]] = add nsw i64 [[LEN]], 1 ; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1 @@ -807,7 +1129,7 @@ define void @test_mul(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]] ; NO-VP-NEXT: store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1 ; NO-VP-NEXT: [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100 -; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP19:![0-9]+]] ; NO-VP: [[FINISH_LOOPEXIT]]: ; NO-VP-NEXT: ret void ; @@ -888,9 +1210,45 @@ define void @test_sdiv(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-LABEL: define void @test_sdiv( ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[A2:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[B1:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16 +; NO-VP-NEXT: [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16 +; NO-VP-NEXT: [[TMP4:%.*]] = sub i64 [[B1]], [[A2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP5:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1 +; NO-VP-NEXT: [[TMP10:%.*]] = sdiv <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 3) +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]] +; NO-VP-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[DEC]] = add nsw i64 [[LEN]], 1 ; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1 @@ -898,7 +1256,7 @@ define void @test_sdiv(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]] ; NO-VP-NEXT: store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1 ; NO-VP-NEXT: [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100 -; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP21:![0-9]+]] ; NO-VP: [[FINISH_LOOPEXIT]]: ; NO-VP-NEXT: ret void ; @@ -979,9 +1337,45 @@ define void @test_udiv(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-LABEL: define void @test_udiv( ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[A2:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[B1:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16 +; NO-VP-NEXT: [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16 +; NO-VP-NEXT: [[TMP4:%.*]] = sub i64 [[B1]], [[A2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP5:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1 +; NO-VP-NEXT: [[TMP10:%.*]] = udiv <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 3) +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]] +; NO-VP-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[DEC]] = add nsw i64 [[LEN]], 1 ; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1 @@ -989,7 +1383,7 @@ define void @test_udiv(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]] ; NO-VP-NEXT: store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1 ; NO-VP-NEXT: [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100 -; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP23:![0-9]+]] ; NO-VP: [[FINISH_LOOPEXIT]]: ; NO-VP-NEXT: ret void ; @@ -1070,9 +1464,45 @@ define void @test_srem(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-LABEL: define void @test_srem( ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[A2:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[B1:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16 +; NO-VP-NEXT: [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16 +; NO-VP-NEXT: [[TMP4:%.*]] = sub i64 [[B1]], [[A2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP5:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1 +; NO-VP-NEXT: [[TMP10:%.*]] = srem <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 3) +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]] +; NO-VP-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[DEC]] = add nsw i64 [[LEN]], 1 ; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1 @@ -1080,7 +1510,7 @@ define void @test_srem(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]] ; NO-VP-NEXT: store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1 ; NO-VP-NEXT: [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100 -; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP25:![0-9]+]] ; NO-VP: [[FINISH_LOOPEXIT]]: ; NO-VP-NEXT: ret void ; @@ -1161,9 +1591,45 @@ define void @test_urem(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-LABEL: define void @test_urem( ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[A2:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[B1:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16 +; NO-VP-NEXT: [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16 +; NO-VP-NEXT: [[TMP4:%.*]] = sub i64 [[B1]], [[A2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP5:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1 +; NO-VP-NEXT: [[TMP10:%.*]] = urem <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 3) +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]] +; NO-VP-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[DEC]] = add nsw i64 [[LEN]], 1 ; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1 @@ -1171,7 +1637,7 @@ define void @test_urem(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]] ; NO-VP-NEXT: store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1 ; NO-VP-NEXT: [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100 -; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP27:![0-9]+]] ; NO-VP: [[FINISH_LOOPEXIT]]: ; NO-VP-NEXT: ret void ; @@ -1255,9 +1721,46 @@ define void @test_fadd(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-LABEL: define void @test_fadd( ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[A2:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[B1:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[TMP15:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP15]], 4 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4 +; NO-VP-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[B1]], [[A2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 100, [[TMP8]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4 +; NO-VP-NEXT: [[TMP12:%.*]] = fadd fast <vscale x 4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00) +; NO-VP-NEXT: [[TMP13:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 4 x float> [[TMP12]], ptr [[TMP13]], align 4 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP10]] +; NO-VP-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[DEC]] = add nsw i64 [[LEN]], 1 ; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[LEN]] ; NO-VP-NEXT: [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4 @@ -1265,7 +1768,7 @@ define void @test_fadd(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[LEN]] ; NO-VP-NEXT: store float [[TMP]], ptr [[ARRAYIDX1]], align 4 ; NO-VP-NEXT: [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100 -; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP29:![0-9]+]] ; NO-VP: [[FINISH_LOOPEXIT]]: ; NO-VP-NEXT: ret void ; @@ -1347,9 +1850,46 @@ define void @test_fsub(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-LABEL: define void @test_fsub( ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[A2:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[B1:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[TMP15:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP15]], 4 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4 +; NO-VP-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[B1]], [[A2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 100, [[TMP8]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4 +; NO-VP-NEXT: [[TMP12:%.*]] = fsub fast <vscale x 4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00) +; NO-VP-NEXT: [[TMP13:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 4 x float> [[TMP12]], ptr [[TMP13]], align 4 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP10]] +; NO-VP-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[DEC]] = add nsw i64 [[LEN]], 1 ; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[LEN]] ; NO-VP-NEXT: [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4 @@ -1357,7 +1897,7 @@ define void @test_fsub(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[LEN]] ; NO-VP-NEXT: store float [[TMP]], ptr [[ARRAYIDX1]], align 4 ; NO-VP-NEXT: [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100 -; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP31:![0-9]+]] ; NO-VP: [[FINISH_LOOPEXIT]]: ; NO-VP-NEXT: ret void ; @@ -1439,9 +1979,46 @@ define void @test_fmul(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-LABEL: define void @test_fmul( ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[A2:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[B1:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[TMP15:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP15]], 4 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4 +; NO-VP-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[B1]], [[A2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 100, [[TMP8]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4 +; NO-VP-NEXT: [[TMP12:%.*]] = fmul fast <vscale x 4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00) +; NO-VP-NEXT: [[TMP13:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 4 x float> [[TMP12]], ptr [[TMP13]], align 4 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP10]] +; NO-VP-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[DEC]] = add nsw i64 [[LEN]], 1 ; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[LEN]] ; NO-VP-NEXT: [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4 @@ -1449,7 +2026,7 @@ define void @test_fmul(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[LEN]] ; NO-VP-NEXT: store float [[TMP]], ptr [[ARRAYIDX1]], align 4 ; NO-VP-NEXT: [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100 -; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP33:![0-9]+]] ; NO-VP: [[FINISH_LOOPEXIT]]: ; NO-VP-NEXT: ret void ; @@ -1531,9 +2108,46 @@ define void @test_fdiv(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-LABEL: define void @test_fdiv( ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[A2:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[B1:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[TMP15:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP15]], 4 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4 +; NO-VP-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[B1]], [[A2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 100, [[TMP8]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4 +; NO-VP-NEXT: [[TMP12:%.*]] = fdiv fast <vscale x 4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00) +; NO-VP-NEXT: [[TMP13:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 4 x float> [[TMP12]], ptr [[TMP13]], align 4 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP10]] +; NO-VP-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[DEC]] = add nsw i64 [[LEN]], 1 ; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[LEN]] ; NO-VP-NEXT: [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4 @@ -1541,7 +2155,7 @@ define void @test_fdiv(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[LEN]] ; NO-VP-NEXT: store float [[TMP]], ptr [[ARRAYIDX1]], align 4 ; NO-VP-NEXT: [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100 -; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP35:![0-9]+]] ; NO-VP: [[FINISH_LOOPEXIT]]: ; NO-VP-NEXT: ret void ; @@ -1676,9 +2290,46 @@ define void @test_fneg(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-LABEL: define void @test_fneg( ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[A2:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[B1:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[TMP15:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP15]], 4 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4 +; NO-VP-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[B1]], [[A2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 100, [[TMP8]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4 +; NO-VP-NEXT: [[TMP12:%.*]] = fneg fast <vscale x 4 x float> [[WIDE_LOAD]] +; NO-VP-NEXT: [[TMP13:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 4 x float> [[TMP12]], ptr [[TMP13]], align 4 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP10]] +; NO-VP-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[DEC]] = add nsw i64 [[LEN]], 1 ; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[LEN]] ; NO-VP-NEXT: [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4 @@ -1686,7 +2337,7 @@ define void @test_fneg(ptr nocapture %a, ptr nocapture readonly %b) { ; NO-VP-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[LEN]] ; NO-VP-NEXT: store float [[TMP]], ptr [[ARRAYIDX1]], align 4 ; NO-VP-NEXT: [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100 -; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP37:![0-9]+]] ; NO-VP: [[FINISH_LOOPEXIT]]: ; NO-VP-NEXT: ret void ; @@ -1748,3 +2399,42 @@ finish.loopexit: ; IF-EVL: [[LOOP37]] = distinct !{[[LOOP37]], [[META1]], [[META2]], [[META3]]} ; IF-EVL: [[LOOP38]] = distinct !{[[LOOP38]], [[META1]]} ;. +; NO-VP: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]} +; NO-VP: [[META1]] = !{!"llvm.loop.isvectorized", i32 1} +; NO-VP: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"} +; NO-VP: [[LOOP3]] = distinct !{[[LOOP3]], [[META1]]} +; NO-VP: [[LOOP4]] = distinct !{[[LOOP4]], [[META1]], [[META2]]} +; NO-VP: [[LOOP5]] = distinct !{[[LOOP5]], [[META1]]} +; NO-VP: [[LOOP6]] = distinct !{[[LOOP6]], [[META1]], [[META2]]} +; NO-VP: [[LOOP7]] = distinct !{[[LOOP7]], [[META1]]} +; NO-VP: [[LOOP8]] = distinct !{[[LOOP8]], [[META1]], [[META2]]} +; NO-VP: [[LOOP9]] = distinct !{[[LOOP9]], [[META1]]} +; NO-VP: [[LOOP10]] = distinct !{[[LOOP10]], [[META1]], [[META2]]} +; NO-VP: [[LOOP11]] = distinct !{[[LOOP11]], [[META1]]} +; NO-VP: [[LOOP12]] = distinct !{[[LOOP12]], [[META1]], [[META2]]} +; NO-VP: [[LOOP13]] = distinct !{[[LOOP13]], [[META1]]} +; NO-VP: [[LOOP14]] = distinct !{[[LOOP14]], [[META1]], [[META2]]} +; NO-VP: [[LOOP15]] = distinct !{[[LOOP15]], [[META1]]} +; NO-VP: [[LOOP16]] = distinct !{[[LOOP16]], [[META1]], [[META2]]} +; NO-VP: [[LOOP17]] = distinct !{[[LOOP17]], [[META1]]} +; NO-VP: [[LOOP18]] = distinct !{[[LOOP18]], [[META1]], [[META2]]} +; NO-VP: [[LOOP19]] = distinct !{[[LOOP19]], [[META1]]} +; NO-VP: [[LOOP20]] = distinct !{[[LOOP20]], [[META1]], [[META2]]} +; NO-VP: [[LOOP21]] = distinct !{[[LOOP21]], [[META1]]} +; NO-VP: [[LOOP22]] = distinct !{[[LOOP22]], [[META1]], [[META2]]} +; NO-VP: [[LOOP23]] = distinct !{[[LOOP23]], [[META1]]} +; NO-VP: [[LOOP24]] = distinct !{[[LOOP24]], [[META1]], [[META2]]} +; NO-VP: [[LOOP25]] = distinct !{[[LOOP25]], [[META1]]} +; NO-VP: [[LOOP26]] = distinct !{[[LOOP26]], [[META1]], [[META2]]} +; NO-VP: [[LOOP27]] = distinct !{[[LOOP27]], [[META1]]} +; NO-VP: [[LOOP28]] = distinct !{[[LOOP28]], [[META1]], [[META2]]} +; NO-VP: [[LOOP29]] = distinct !{[[LOOP29]], [[META1]]} +; NO-VP: [[LOOP30]] = distinct !{[[LOOP30]], [[META1]], [[META2]]} +; NO-VP: [[LOOP31]] = distinct !{[[LOOP31]], [[META1]]} +; NO-VP: [[LOOP32]] = distinct !{[[LOOP32]], [[META1]], [[META2]]} +; NO-VP: [[LOOP33]] = distinct !{[[LOOP33]], [[META1]]} +; NO-VP: [[LOOP34]] = distinct !{[[LOOP34]], [[META1]], [[META2]]} +; NO-VP: [[LOOP35]] = distinct !{[[LOOP35]], [[META1]]} +; NO-VP: [[LOOP36]] = distinct !{[[LOOP36]], [[META1]], [[META2]]} +; NO-VP: [[LOOP37]] = distinct !{[[LOOP37]], [[META1]]} +;. diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-call-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-call-intrinsics.ll index dfa01a5..7d21c4d 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-call-intrinsics.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-call-intrinsics.ll @@ -1,12 +1,10 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5 ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s --check-prefix=IF-EVL ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s --check-prefix=NO-VP define void @vp_smax(ptr %a, ptr %b, ptr %c, i64 %N) { @@ -78,9 +76,53 @@ define void @vp_smax(ptr %a, ptr %b, ptr %c, i64 %N) { ; NO-VP-LABEL: define void @vp_smax( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], ptr [[C:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[C3:%.*]] = ptrtoint ptr [[C]] to i64 +; NO-VP-NEXT: [[B2:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[A1:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[TMP18:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP19:%.*]] = mul nuw i64 [[TMP18]], 4 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP19]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4 +; NO-VP-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[A1]], [[B2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]] +; NO-VP-NEXT: [[TMP7:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP8:%.*]] = sub i64 [[A1]], [[C3]] +; NO-VP-NEXT: [[DIFF_CHECK4:%.*]] = icmp ult i64 [[TMP8]], [[TMP7]] +; NO-VP-NEXT: [[CONFLICT_RDX:%.*]] = or i1 [[DIFF_CHECK]], [[DIFF_CHECK4]] +; NO-VP-NEXT: br i1 [[CONFLICT_RDX]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP10]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 4 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP13]], align 4 +; NO-VP-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD5:%.*]] = load <vscale x 4 x i32>, ptr [[TMP14]], align 4 +; NO-VP-NEXT: [[TMP15:%.*]] = call <vscale x 4 x i32> @llvm.smax.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[WIDE_LOAD5]]) +; NO-VP-NEXT: [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 4 x i32> [[TMP15]], ptr [[TMP16]], align 4 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]] +; NO-VP-NEXT: [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP17]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ] +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4 ; NO-VP-NEXT: [[GEP3:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[IV]] @@ -90,7 +132,7 @@ define void @vp_smax(ptr %a, ptr %b, ptr %c, i64 %N) { ; NO-VP-NEXT: store i32 [[DOT]], ptr [[GEP11]], align 4 ; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -184,9 +226,53 @@ define void @vp_smin(ptr %a, ptr %b, ptr %c, i64 %N) { ; NO-VP-LABEL: define void @vp_smin( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], ptr [[C:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[C3:%.*]] = ptrtoint ptr [[C]] to i64 +; NO-VP-NEXT: [[B2:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[A1:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[TMP18:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP19:%.*]] = mul nuw i64 [[TMP18]], 4 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP19]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4 +; NO-VP-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[A1]], [[B2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]] +; NO-VP-NEXT: [[TMP7:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP8:%.*]] = sub i64 [[A1]], [[C3]] +; NO-VP-NEXT: [[DIFF_CHECK4:%.*]] = icmp ult i64 [[TMP8]], [[TMP7]] +; NO-VP-NEXT: [[CONFLICT_RDX:%.*]] = or i1 [[DIFF_CHECK]], [[DIFF_CHECK4]] +; NO-VP-NEXT: br i1 [[CONFLICT_RDX]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP10]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 4 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP13]], align 4 +; NO-VP-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD5:%.*]] = load <vscale x 4 x i32>, ptr [[TMP14]], align 4 +; NO-VP-NEXT: [[TMP15:%.*]] = call <vscale x 4 x i32> @llvm.smin.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[WIDE_LOAD5]]) +; NO-VP-NEXT: [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 4 x i32> [[TMP15]], ptr [[TMP16]], align 4 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]] +; NO-VP-NEXT: [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP17]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ] +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4 ; NO-VP-NEXT: [[GEP3:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[IV]] @@ -196,7 +282,7 @@ define void @vp_smin(ptr %a, ptr %b, ptr %c, i64 %N) { ; NO-VP-NEXT: store i32 [[DOT]], ptr [[GEP11]], align 4 ; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP5:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -290,9 +376,53 @@ define void @vp_umax(ptr %a, ptr %b, ptr %c, i64 %N) { ; NO-VP-LABEL: define void @vp_umax( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], ptr [[C:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[C3:%.*]] = ptrtoint ptr [[C]] to i64 +; NO-VP-NEXT: [[B2:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[A1:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[TMP18:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP19:%.*]] = mul nuw i64 [[TMP18]], 4 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP19]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4 +; NO-VP-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[A1]], [[B2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]] +; NO-VP-NEXT: [[TMP7:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP8:%.*]] = sub i64 [[A1]], [[C3]] +; NO-VP-NEXT: [[DIFF_CHECK4:%.*]] = icmp ult i64 [[TMP8]], [[TMP7]] +; NO-VP-NEXT: [[CONFLICT_RDX:%.*]] = or i1 [[DIFF_CHECK]], [[DIFF_CHECK4]] +; NO-VP-NEXT: br i1 [[CONFLICT_RDX]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP10]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 4 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP13]], align 4 +; NO-VP-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD5:%.*]] = load <vscale x 4 x i32>, ptr [[TMP14]], align 4 +; NO-VP-NEXT: [[TMP15:%.*]] = call <vscale x 4 x i32> @llvm.umax.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[WIDE_LOAD5]]) +; NO-VP-NEXT: [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 4 x i32> [[TMP15]], ptr [[TMP16]], align 4 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]] +; NO-VP-NEXT: [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP17]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ] +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4 ; NO-VP-NEXT: [[GEP3:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[IV]] @@ -302,7 +432,7 @@ define void @vp_umax(ptr %a, ptr %b, ptr %c, i64 %N) { ; NO-VP-NEXT: store i32 [[DOT]], ptr [[GEP11]], align 4 ; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP7:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -396,9 +526,53 @@ define void @vp_umin(ptr %a, ptr %b, ptr %c, i64 %N) { ; NO-VP-LABEL: define void @vp_umin( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], ptr [[C:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[C3:%.*]] = ptrtoint ptr [[C]] to i64 +; NO-VP-NEXT: [[B2:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[A1:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[TMP18:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP19:%.*]] = mul nuw i64 [[TMP18]], 4 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP19]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4 +; NO-VP-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[A1]], [[B2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]] +; NO-VP-NEXT: [[TMP7:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP8:%.*]] = sub i64 [[A1]], [[C3]] +; NO-VP-NEXT: [[DIFF_CHECK4:%.*]] = icmp ult i64 [[TMP8]], [[TMP7]] +; NO-VP-NEXT: [[CONFLICT_RDX:%.*]] = or i1 [[DIFF_CHECK]], [[DIFF_CHECK4]] +; NO-VP-NEXT: br i1 [[CONFLICT_RDX]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP10]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 4 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP13]], align 4 +; NO-VP-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD5:%.*]] = load <vscale x 4 x i32>, ptr [[TMP14]], align 4 +; NO-VP-NEXT: [[TMP15:%.*]] = call <vscale x 4 x i32> @llvm.umin.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[WIDE_LOAD5]]) +; NO-VP-NEXT: [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 4 x i32> [[TMP15]], ptr [[TMP16]], align 4 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]] +; NO-VP-NEXT: [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP17]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ] +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4 ; NO-VP-NEXT: [[GEP3:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[IV]] @@ -408,7 +582,7 @@ define void @vp_umin(ptr %a, ptr %b, ptr %c, i64 %N) { ; NO-VP-NEXT: store i32 [[DOT]], ptr [[GEP11]], align 4 ; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP9:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -494,17 +668,54 @@ define void @vp_ctlz(ptr %a, ptr %b, i64 %N) { ; NO-VP-LABEL: define void @vp_ctlz( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[B2:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[A1:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP15:%.*]] = mul nuw i64 [[TMP11]], 4 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 8, i64 [[TMP15]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4 +; NO-VP-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[A1]], [[B2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP8]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[TMP12:%.*]] = call <vscale x 4 x i32> @llvm.ctlz.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], i1 true) +; NO-VP-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: store <vscale x 4 x i32> [[TMP12]], ptr [[TMP13]], align 4 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP10]] +; NO-VP-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ] -; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]] -; NO-VP-NEXT: [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[IV1:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] +; NO-VP-NEXT: [[GEP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV1]] +; NO-VP-NEXT: [[TMP0:%.*]] = load i32, ptr [[GEP1]], align 4 ; NO-VP-NEXT: [[TMP1:%.*]] = tail call range(i32 0, 33) i32 @llvm.ctlz.i32(i32 [[TMP0]], i1 true) -; NO-VP-NEXT: [[GEP3:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: [[GEP3:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]] ; NO-VP-NEXT: store i32 [[TMP1]], ptr [[GEP3]], align 4 -; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 +; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP11:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -587,17 +798,54 @@ define void @vp_cttz(ptr %a, ptr %b, i64 %N) { ; NO-VP-LABEL: define void @vp_cttz( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[B2:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[A1:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP15:%.*]] = mul nuw i64 [[TMP11]], 4 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 8, i64 [[TMP15]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4 +; NO-VP-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[A1]], [[B2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP8]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[TMP12:%.*]] = call <vscale x 4 x i32> @llvm.cttz.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], i1 true) +; NO-VP-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: store <vscale x 4 x i32> [[TMP12]], ptr [[TMP13]], align 4 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP10]] +; NO-VP-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ] -; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]] -; NO-VP-NEXT: [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[IV1:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] +; NO-VP-NEXT: [[GEP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV1]] +; NO-VP-NEXT: [[TMP0:%.*]] = load i32, ptr [[GEP1]], align 4 ; NO-VP-NEXT: [[TMP1:%.*]] = tail call range(i32 0, 33) i32 @llvm.cttz.i32(i32 [[TMP0]], i1 true) -; NO-VP-NEXT: [[GEP3:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: [[GEP3:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]] ; NO-VP-NEXT: store i32 [[TMP1]], ptr [[GEP3]], align 4 -; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 +; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP13:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -684,19 +932,58 @@ define void @vp_lrint(ptr %a, ptr %b, i64 %N) { ; NO-VP-LABEL: define void @vp_lrint( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[B2:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[A1:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP17:%.*]] = mul nuw i64 [[TMP11]], 4 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP17]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4 +; NO-VP-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[A1]], [[B2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP8]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[TMP12:%.*]] = fpext <vscale x 4 x float> [[WIDE_LOAD]] to <vscale x 4 x double> +; NO-VP-NEXT: [[TMP13:%.*]] = call <vscale x 4 x i64> @llvm.lrint.nxv4i64.nxv4f64(<vscale x 4 x double> [[TMP12]]) +; NO-VP-NEXT: [[TMP14:%.*]] = trunc <vscale x 4 x i64> [[TMP13]] to <vscale x 4 x i32> +; NO-VP-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: store <vscale x 4 x i32> [[TMP14]], ptr [[TMP15]], align 4 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP10]] +; NO-VP-NEXT: [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP16]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ] -; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]] -; NO-VP-NEXT: [[TMP0:%.*]] = load float, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[IV1:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] +; NO-VP-NEXT: [[GEP1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV1]] +; NO-VP-NEXT: [[TMP0:%.*]] = load float, ptr [[GEP1]], align 4 ; NO-VP-NEXT: [[CONV2:%.*]] = fpext float [[TMP0]] to double ; NO-VP-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.lrint.i64.f64(double [[CONV2]]) ; NO-VP-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP1]] to i32 -; NO-VP-NEXT: [[GEP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: [[GEP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]] ; NO-VP-NEXT: store i32 [[CONV3]], ptr [[GEP5]], align 4 -; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 +; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP15:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -785,19 +1072,58 @@ define void @vp_llrint(ptr %a, ptr %b, i64 %N) { ; NO-VP-LABEL: define void @vp_llrint( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[B2:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[A1:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP17:%.*]] = mul nuw i64 [[TMP11]], 4 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP17]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4 +; NO-VP-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[A1]], [[B2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP8]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[TMP12:%.*]] = fpext <vscale x 4 x float> [[WIDE_LOAD]] to <vscale x 4 x double> +; NO-VP-NEXT: [[TMP13:%.*]] = call <vscale x 4 x i64> @llvm.llrint.nxv4i64.nxv4f64(<vscale x 4 x double> [[TMP12]]) +; NO-VP-NEXT: [[TMP14:%.*]] = trunc <vscale x 4 x i64> [[TMP13]] to <vscale x 4 x i32> +; NO-VP-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: store <vscale x 4 x i32> [[TMP14]], ptr [[TMP15]], align 4 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP10]] +; NO-VP-NEXT: [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP16]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ] -; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]] -; NO-VP-NEXT: [[TMP0:%.*]] = load float, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[IV1:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] +; NO-VP-NEXT: [[GEP1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV1]] +; NO-VP-NEXT: [[TMP0:%.*]] = load float, ptr [[GEP1]], align 4 ; NO-VP-NEXT: [[CONV2:%.*]] = fpext float [[TMP0]] to double ; NO-VP-NEXT: [[TMP1:%.*]] = tail call i64 @llvm.llrint.i64.f64(double [[CONV2]]) ; NO-VP-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP1]] to i32 -; NO-VP-NEXT: [[GEP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: [[GEP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]] ; NO-VP-NEXT: store i32 [[CONV3]], ptr [[GEP5]], align 4 -; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 +; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP17:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -882,17 +1208,54 @@ define void @vp_abs(ptr %a, ptr %b, i64 %N) { ; NO-VP-LABEL: define void @vp_abs( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[B2:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[A1:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP11]], 4 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 8, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4 +; NO-VP-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[A1]], [[B2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP8]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[TMP12:%.*]] = call <vscale x 4 x i32> @llvm.abs.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], i1 true) +; NO-VP-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: store <vscale x 4 x i32> [[TMP12]], ptr [[TMP13]], align 4 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP10]] +; NO-VP-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ] -; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]] -; NO-VP-NEXT: [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[IV1:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] +; NO-VP-NEXT: [[GEP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV1]] +; NO-VP-NEXT: [[TMP0:%.*]] = load i32, ptr [[GEP1]], align 4 ; NO-VP-NEXT: [[COND:%.*]] = tail call i32 @llvm.abs.i32(i32 [[TMP0]], i1 true) -; NO-VP-NEXT: [[GEP9:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: [[GEP9:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]] ; NO-VP-NEXT: store i32 [[COND]], ptr [[GEP9]], align 4 -; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 +; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP19:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -937,17 +1300,44 @@ define void @log10(ptr %a, ptr %b, i64 %N) { ; NO-VP-LABEL: define void @log10( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[B2:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[A1:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 8 +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP1:%.*]] = sub i64 [[A1]], [[B2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP1]], 16 +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 4 +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[TMP2:%.*]] = call <4 x float> @llvm.log10.v4f32(<4 x float> [[WIDE_LOAD]]) +; NO-VP-NEXT: [[TMP3:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: store <4 x float> [[TMP2]], ptr [[TMP3]], align 4 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV]], 4 +; NO-VP-NEXT: [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ] -; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]] -; NO-VP-NEXT: [[TMP0:%.*]] = load float, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[IV1:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] +; NO-VP-NEXT: [[GEP1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV1]] +; NO-VP-NEXT: [[TMP0:%.*]] = load float, ptr [[GEP1]], align 4 ; NO-VP-NEXT: [[COND:%.*]] = tail call float @llvm.log10.f32(float [[TMP0]]) -; NO-VP-NEXT: [[GEP9:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: [[GEP9:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV1]] ; NO-VP-NEXT: store float [[COND]], ptr [[GEP9]], align 4 -; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 +; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP21:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -1003,3 +1393,26 @@ declare i32 @llvm.abs.i32(i32, i1 immarg) ; IF-EVL: [[LOOP19]] = distinct !{[[LOOP19]], [[META1]], [[META2]], [[META3]]} ; IF-EVL: [[LOOP20]] = distinct !{[[LOOP20]], [[META1]]} ;. +; NO-VP: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]} +; NO-VP: [[META1]] = !{!"llvm.loop.isvectorized", i32 1} +; NO-VP: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"} +; NO-VP: [[LOOP3]] = distinct !{[[LOOP3]], [[META1]]} +; NO-VP: [[LOOP4]] = distinct !{[[LOOP4]], [[META1]], [[META2]]} +; NO-VP: [[LOOP5]] = distinct !{[[LOOP5]], [[META1]]} +; NO-VP: [[LOOP6]] = distinct !{[[LOOP6]], [[META1]], [[META2]]} +; NO-VP: [[LOOP7]] = distinct !{[[LOOP7]], [[META1]]} +; NO-VP: [[LOOP8]] = distinct !{[[LOOP8]], [[META1]], [[META2]]} +; NO-VP: [[LOOP9]] = distinct !{[[LOOP9]], [[META1]]} +; NO-VP: [[LOOP10]] = distinct !{[[LOOP10]], [[META1]], [[META2]]} +; NO-VP: [[LOOP11]] = distinct !{[[LOOP11]], [[META1]]} +; NO-VP: [[LOOP12]] = distinct !{[[LOOP12]], [[META1]], [[META2]]} +; NO-VP: [[LOOP13]] = distinct !{[[LOOP13]], [[META1]]} +; NO-VP: [[LOOP14]] = distinct !{[[LOOP14]], [[META1]], [[META2]]} +; NO-VP: [[LOOP15]] = distinct !{[[LOOP15]], [[META1]]} +; NO-VP: [[LOOP16]] = distinct !{[[LOOP16]], [[META1]], [[META2]]} +; NO-VP: [[LOOP17]] = distinct !{[[LOOP17]], [[META1]]} +; NO-VP: [[LOOP18]] = distinct !{[[LOOP18]], [[META1]], [[META2]]} +; NO-VP: [[LOOP19]] = distinct !{[[LOOP19]], [[META1]]} +; NO-VP: [[LOOP20]] = distinct !{[[LOOP20]], [[META1]], [[META2]]} +; NO-VP: [[LOOP21]] = distinct !{[[LOOP21]], [[META1]]} +;. diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-cast-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-cast-intrinsics.ll index 7139170..4b718dd 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-cast-intrinsics.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-cast-intrinsics.ll @@ -1,12 +1,10 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5 ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s --check-prefix=IF-EVL ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s --check-prefix=NO-VP define void @vp_sext(ptr %a, ptr %b, i64 %N) { @@ -69,17 +67,54 @@ define void @vp_sext(ptr %a, ptr %b, i64 %N) { ; NO-VP-LABEL: define void @vp_sext( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP9]], 2 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 20, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = shl i64 [[N]], 3 +; NO-VP-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]] +; NO-VP-NEXT: [[TMP4:%.*]] = shl i64 [[N]], 2 +; NO-VP-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP4]] +; NO-VP-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[A]], [[SCEVGEP1]] +; NO-VP-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[B]], [[SCEVGEP]] +; NO-VP-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] +; NO-VP-NEXT: br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP5:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 2 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP6]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 2 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 2 x i32>, ptr [[GEP]], align 4, !alias.scope [[META0:![0-9]+]] +; NO-VP-NEXT: [[TMP10:%.*]] = sext <vscale x 2 x i32> [[WIDE_LOAD]] to <vscale x 2 x i64> +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: store <vscale x 2 x i64> [[TMP10]], ptr [[TMP11]], align 8, !alias.scope [[META3:![0-9]+]], !noalias [[META0]] +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP8]] +; NO-VP-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ] -; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]] -; NO-VP-NEXT: [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[IV1:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] +; NO-VP-NEXT: [[GEP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV1]] +; NO-VP-NEXT: [[TMP0:%.*]] = load i32, ptr [[GEP1]], align 4 ; NO-VP-NEXT: [[CONV2:%.*]] = sext i32 [[TMP0]] to i64 -; NO-VP-NEXT: [[GEP4:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: [[GEP4:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV1]] ; NO-VP-NEXT: store i64 [[CONV2]], ptr [[GEP4]], align 8 -; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 +; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP8:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -161,17 +196,54 @@ define void @vp_zext(ptr %a, ptr %b, i64 %N) { ; NO-VP-LABEL: define void @vp_zext( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP9]], 2 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 20, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = shl i64 [[N]], 3 +; NO-VP-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]] +; NO-VP-NEXT: [[TMP4:%.*]] = shl i64 [[N]], 2 +; NO-VP-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP4]] +; NO-VP-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[A]], [[SCEVGEP1]] +; NO-VP-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[B]], [[SCEVGEP]] +; NO-VP-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] +; NO-VP-NEXT: br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP5:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 2 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP6]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 2 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 2 x i32>, ptr [[GEP]], align 4, !alias.scope [[META9:![0-9]+]] +; NO-VP-NEXT: [[TMP10:%.*]] = zext <vscale x 2 x i32> [[WIDE_LOAD]] to <vscale x 2 x i64> +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: store <vscale x 2 x i64> [[TMP10]], ptr [[TMP11]], align 8, !alias.scope [[META12:![0-9]+]], !noalias [[META9]] +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP8]] +; NO-VP-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] -; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]] -; NO-VP-NEXT: [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] +; NO-VP-NEXT: [[GEP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV1]] +; NO-VP-NEXT: [[TMP0:%.*]] = load i32, ptr [[GEP1]], align 4 ; NO-VP-NEXT: [[CONV:%.*]] = zext i32 [[TMP0]] to i64 -; NO-VP-NEXT: [[GEP2:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: [[GEP2:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV1]] ; NO-VP-NEXT: store i64 [[CONV]], ptr [[GEP2]], align 8 -; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 +; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP15:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -253,17 +325,54 @@ define void @vp_trunc(ptr %a, ptr %b, i64 %N) { ; NO-VP-LABEL: define void @vp_trunc( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP9]], 2 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 20, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = shl i64 [[N]], 2 +; NO-VP-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]] +; NO-VP-NEXT: [[TMP4:%.*]] = shl i64 [[N]], 3 +; NO-VP-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP4]] +; NO-VP-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[A]], [[SCEVGEP1]] +; NO-VP-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[B]], [[SCEVGEP]] +; NO-VP-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] +; NO-VP-NEXT: br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP5:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 2 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP6]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 2 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[IV]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[GEP]], align 8, !alias.scope [[META16:![0-9]+]] +; NO-VP-NEXT: [[TMP10:%.*]] = trunc <vscale x 2 x i64> [[WIDE_LOAD]] to <vscale x 2 x i32> +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: store <vscale x 2 x i32> [[TMP10]], ptr [[TMP11]], align 4, !alias.scope [[META19:![0-9]+]], !noalias [[META16]] +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP8]] +; NO-VP-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] -; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[IV]] -; NO-VP-NEXT: [[TMP0:%.*]] = load i64, ptr [[GEP]], align 8 +; NO-VP-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] +; NO-VP-NEXT: [[GEP1:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[IV1]] +; NO-VP-NEXT: [[TMP0:%.*]] = load i64, ptr [[GEP1]], align 8 ; NO-VP-NEXT: [[CONV:%.*]] = trunc i64 [[TMP0]] to i32 -; NO-VP-NEXT: [[GEP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: [[GEP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]] ; NO-VP-NEXT: store i32 [[CONV]], ptr [[GEP2]], align 4 -; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 +; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP22:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -345,17 +454,54 @@ define void @vp_fpext(ptr %a, ptr %b, i64 %N) { ; NO-VP-LABEL: define void @vp_fpext( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP9]], 2 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = shl i64 [[N]], 3 +; NO-VP-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]] +; NO-VP-NEXT: [[TMP4:%.*]] = shl i64 [[N]], 2 +; NO-VP-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP4]] +; NO-VP-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[A]], [[SCEVGEP1]] +; NO-VP-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[B]], [[SCEVGEP]] +; NO-VP-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] +; NO-VP-NEXT: br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP5:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 2 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP6]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 2 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 2 x float>, ptr [[GEP]], align 4, !alias.scope [[META23:![0-9]+]] +; NO-VP-NEXT: [[TMP10:%.*]] = fpext <vscale x 2 x float> [[WIDE_LOAD]] to <vscale x 2 x double> +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: store <vscale x 2 x double> [[TMP10]], ptr [[TMP11]], align 8, !alias.scope [[META26:![0-9]+]], !noalias [[META23]] +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP8]] +; NO-VP-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] -; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]] -; NO-VP-NEXT: [[TMP0:%.*]] = load float, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] +; NO-VP-NEXT: [[GEP1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV1]] +; NO-VP-NEXT: [[TMP0:%.*]] = load float, ptr [[GEP1]], align 4 ; NO-VP-NEXT: [[CONV:%.*]] = fpext float [[TMP0]] to double -; NO-VP-NEXT: [[GEP2:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: [[GEP2:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[IV1]] ; NO-VP-NEXT: store double [[CONV]], ptr [[GEP2]], align 8 -; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 +; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP29:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -437,17 +583,54 @@ define void @vp_fptrunc(ptr %a, ptr %b, i64 %N) { ; NO-VP-LABEL: define void @vp_fptrunc( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP9]], 2 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = shl i64 [[N]], 2 +; NO-VP-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]] +; NO-VP-NEXT: [[TMP4:%.*]] = shl i64 [[N]], 3 +; NO-VP-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP4]] +; NO-VP-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[A]], [[SCEVGEP1]] +; NO-VP-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[B]], [[SCEVGEP]] +; NO-VP-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] +; NO-VP-NEXT: br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP5:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 2 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP6]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 2 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[IV]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 2 x double>, ptr [[GEP]], align 8, !alias.scope [[META30:![0-9]+]] +; NO-VP-NEXT: [[TMP10:%.*]] = fptrunc <vscale x 2 x double> [[WIDE_LOAD]] to <vscale x 2 x float> +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: store <vscale x 2 x float> [[TMP10]], ptr [[TMP11]], align 4, !alias.scope [[META33:![0-9]+]], !noalias [[META30]] +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP8]] +; NO-VP-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP35:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] -; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[IV]] -; NO-VP-NEXT: [[TMP0:%.*]] = load double, ptr [[GEP]], align 8 +; NO-VP-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] +; NO-VP-NEXT: [[GEP1:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[IV1]] +; NO-VP-NEXT: [[TMP0:%.*]] = load double, ptr [[GEP1]], align 8 ; NO-VP-NEXT: [[CONV:%.*]] = fptrunc double [[TMP0]] to float -; NO-VP-NEXT: [[GEP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: [[GEP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV1]] ; NO-VP-NEXT: store float [[CONV]], ptr [[GEP2]], align 4 -; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 +; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP36:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -529,17 +712,54 @@ define void @vp_sitofp(ptr %a, ptr %b, i64 %N) { ; NO-VP-LABEL: define void @vp_sitofp( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[B2:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[A1:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP11]], 4 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4 +; NO-VP-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[A1]], [[B2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP8]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[TMP12:%.*]] = sitofp <vscale x 4 x i32> [[WIDE_LOAD]] to <vscale x 4 x float> +; NO-VP-NEXT: [[TMP13:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: store <vscale x 4 x float> [[TMP12]], ptr [[TMP13]], align 4 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP10]] +; NO-VP-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP37:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] -; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]] -; NO-VP-NEXT: [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] +; NO-VP-NEXT: [[GEP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV1]] +; NO-VP-NEXT: [[TMP0:%.*]] = load i32, ptr [[GEP1]], align 4 ; NO-VP-NEXT: [[CONV:%.*]] = sitofp i32 [[TMP0]] to float -; NO-VP-NEXT: [[GEP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: [[GEP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV1]] ; NO-VP-NEXT: store float [[CONV]], ptr [[GEP2]], align 4 -; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 +; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP38:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -621,17 +841,54 @@ define void @vp_uitofp(ptr %a, ptr %b, i64 %N) { ; NO-VP-LABEL: define void @vp_uitofp( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[B2:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[A1:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP11]], 4 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4 +; NO-VP-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[A1]], [[B2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP8]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[TMP12:%.*]] = uitofp <vscale x 4 x i32> [[WIDE_LOAD]] to <vscale x 4 x float> +; NO-VP-NEXT: [[TMP13:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: store <vscale x 4 x float> [[TMP12]], ptr [[TMP13]], align 4 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP10]] +; NO-VP-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP39:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] -; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]] -; NO-VP-NEXT: [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] +; NO-VP-NEXT: [[GEP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV1]] +; NO-VP-NEXT: [[TMP0:%.*]] = load i32, ptr [[GEP1]], align 4 ; NO-VP-NEXT: [[CONV:%.*]] = uitofp i32 [[TMP0]] to float -; NO-VP-NEXT: [[GEP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: [[GEP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV1]] ; NO-VP-NEXT: store float [[CONV]], ptr [[GEP2]], align 4 -; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 +; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP40:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -713,17 +970,54 @@ define void @vp_fptosi(ptr %a, ptr %b, i64 %N) { ; NO-VP-LABEL: define void @vp_fptosi( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[B2:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[A1:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP11]], 4 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4 +; NO-VP-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[A1]], [[B2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP8]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[TMP12:%.*]] = fptosi <vscale x 4 x float> [[WIDE_LOAD]] to <vscale x 4 x i32> +; NO-VP-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: store <vscale x 4 x i32> [[TMP12]], ptr [[TMP13]], align 4 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP10]] +; NO-VP-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] -; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]] -; NO-VP-NEXT: [[TMP0:%.*]] = load float, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] +; NO-VP-NEXT: [[GEP1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV1]] +; NO-VP-NEXT: [[TMP0:%.*]] = load float, ptr [[GEP1]], align 4 ; NO-VP-NEXT: [[CONV:%.*]] = fptosi float [[TMP0]] to i32 -; NO-VP-NEXT: [[GEP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: [[GEP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]] ; NO-VP-NEXT: store i32 [[CONV]], ptr [[GEP2]], align 4 -; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 +; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP42:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -805,17 +1099,54 @@ define void @vp_fptoui(ptr %a, ptr %b, i64 %N) { ; NO-VP-LABEL: define void @vp_fptoui( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[B2:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[A1:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP11]], 4 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4 +; NO-VP-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[A1]], [[B2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP8]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[TMP12:%.*]] = fptoui <vscale x 4 x float> [[WIDE_LOAD]] to <vscale x 4 x i32> +; NO-VP-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: store <vscale x 4 x i32> [[TMP12]], ptr [[TMP13]], align 4 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP10]] +; NO-VP-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP43:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] -; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]] -; NO-VP-NEXT: [[TMP0:%.*]] = load float, ptr [[GEP]], align 4 +; NO-VP-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] +; NO-VP-NEXT: [[GEP1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV1]] +; NO-VP-NEXT: [[TMP0:%.*]] = load float, ptr [[GEP1]], align 4 ; NO-VP-NEXT: [[CONV:%.*]] = fptoui float [[TMP0]] to i32 -; NO-VP-NEXT: [[GEP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: [[GEP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]] ; NO-VP-NEXT: store i32 [[CONV]], ptr [[GEP2]], align 4 -; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 +; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP44:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -897,17 +1228,54 @@ define void @vp_inttoptr(ptr %a, ptr %b, i64 %N) { ; NO-VP-LABEL: define void @vp_inttoptr( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[B2:%.*]] = ptrtoint ptr [[B]] to i64 +; NO-VP-NEXT: [[A1:%.*]] = ptrtoint ptr [[A]] to i64 +; NO-VP-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP15:%.*]] = mul nuw i64 [[TMP11]], 2 +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP15]]) +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]] +; NO-VP: [[VECTOR_MEMCHECK]]: +; NO-VP-NEXT: [[TMP3:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 2 +; NO-VP-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 8 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[A1]], [[B2]] +; NO-VP-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]] +; NO-VP-NEXT: br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 2 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP8]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 2 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[IV]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[GEP]], align 8 +; NO-VP-NEXT: [[TMP12:%.*]] = inttoptr <vscale x 2 x i64> [[WIDE_LOAD]] to <vscale x 2 x ptr> +; NO-VP-NEXT: [[TMP13:%.*]] = getelementptr inbounds ptr, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: store <vscale x 2 x ptr> [[TMP12]], ptr [[TMP13]], align 8 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP10]] +; NO-VP-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP45:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] -; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[IV]] -; NO-VP-NEXT: [[TMP0:%.*]] = load i64, ptr [[GEP]], align 8 +; NO-VP-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] +; NO-VP-NEXT: [[GEP1:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[IV1]] +; NO-VP-NEXT: [[TMP0:%.*]] = load i64, ptr [[GEP1]], align 8 ; NO-VP-NEXT: [[TMP1:%.*]] = inttoptr i64 [[TMP0]] to ptr -; NO-VP-NEXT: [[GEP2:%.*]] = getelementptr inbounds ptr, ptr [[A]], i64 [[IV]] +; NO-VP-NEXT: [[GEP2:%.*]] = getelementptr inbounds ptr, ptr [[A]], i64 [[IV1]] ; NO-VP-NEXT: store ptr [[TMP1]], ptr [[GEP2]], align 8 -; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 +; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP46:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -987,16 +1355,50 @@ define void @vp_ptrtoint(ptr %a, ptr %b, i64 %N) { ; NO-VP-LABEL: define void @vp_ptrtoint( ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[ENTRY:.*]]: +; NO-VP-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 2 +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP1]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 2 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2 +; NO-VP-NEXT: [[TMP6:%.*]] = call <vscale x 2 x i64> @llvm.stepvector.nxv2i64() +; NO-VP-NEXT: [[TMP7:%.*]] = mul <vscale x 2 x i64> [[TMP6]], splat (i64 1) +; NO-VP-NEXT: [[INDUCTION:%.*]] = add <vscale x 2 x i64> zeroinitializer, [[TMP7]] +; NO-VP-NEXT: [[TMP8:%.*]] = mul i64 1, [[TMP5]] +; NO-VP-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP8]], i64 0 +; NO-VP-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[VEC_IND:%.*]] = phi <vscale x 2 x i64> [ [[INDUCTION]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[B]], <vscale x 2 x i64> [[VEC_IND]] +; NO-VP-NEXT: [[TMP10:%.*]] = ptrtoint <vscale x 2 x ptr> [[TMP9]] to <vscale x 2 x i64> +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 2 x i64> [[TMP10]], ptr [[TMP11]], align 8 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] +; NO-VP-NEXT: [[VEC_IND_NEXT]] = add <vscale x 2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT]] +; NO-VP-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP47:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ] ; NO-VP-NEXT: [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]] ; NO-VP-NEXT: [[TMP0:%.*]] = ptrtoint ptr [[GEP]] to i64 ; NO-VP-NEXT: [[GEP2:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]] ; NO-VP-NEXT: store i64 [[TMP0]], ptr [[GEP2]], align 8 ; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP48:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -1068,3 +1470,53 @@ exit: ; IF-EVL: [[LOOP48]] = distinct !{[[LOOP48]], [[META6]], [[META7]], [[META8]]} ; IF-EVL: [[LOOP49]] = distinct !{[[LOOP49]], [[META8]], [[META6]]} ;. +; NO-VP: [[META0]] = !{[[META1:![0-9]+]]} +; NO-VP: [[META1]] = distinct !{[[META1]], [[META2:![0-9]+]]} +; NO-VP: [[META2]] = distinct !{[[META2]], !"LVerDomain"} +; NO-VP: [[META3]] = !{[[META4:![0-9]+]]} +; NO-VP: [[META4]] = distinct !{[[META4]], [[META2]]} +; NO-VP: [[LOOP5]] = distinct !{[[LOOP5]], [[META6:![0-9]+]], [[META7:![0-9]+]]} +; NO-VP: [[META6]] = !{!"llvm.loop.isvectorized", i32 1} +; NO-VP: [[META7]] = !{!"llvm.loop.unroll.runtime.disable"} +; NO-VP: [[LOOP8]] = distinct !{[[LOOP8]], [[META6]]} +; NO-VP: [[META9]] = !{[[META10:![0-9]+]]} +; NO-VP: [[META10]] = distinct !{[[META10]], [[META11:![0-9]+]]} +; NO-VP: [[META11]] = distinct !{[[META11]], !"LVerDomain"} +; NO-VP: [[META12]] = !{[[META13:![0-9]+]]} +; NO-VP: [[META13]] = distinct !{[[META13]], [[META11]]} +; NO-VP: [[LOOP14]] = distinct !{[[LOOP14]], [[META6]], [[META7]]} +; NO-VP: [[LOOP15]] = distinct !{[[LOOP15]], [[META6]]} +; NO-VP: [[META16]] = !{[[META17:![0-9]+]]} +; NO-VP: [[META17]] = distinct !{[[META17]], [[META18:![0-9]+]]} +; NO-VP: [[META18]] = distinct !{[[META18]], !"LVerDomain"} +; NO-VP: [[META19]] = !{[[META20:![0-9]+]]} +; NO-VP: [[META20]] = distinct !{[[META20]], [[META18]]} +; NO-VP: [[LOOP21]] = distinct !{[[LOOP21]], [[META6]], [[META7]]} +; NO-VP: [[LOOP22]] = distinct !{[[LOOP22]], [[META6]]} +; NO-VP: [[META23]] = !{[[META24:![0-9]+]]} +; NO-VP: [[META24]] = distinct !{[[META24]], [[META25:![0-9]+]]} +; NO-VP: [[META25]] = distinct !{[[META25]], !"LVerDomain"} +; NO-VP: [[META26]] = !{[[META27:![0-9]+]]} +; NO-VP: [[META27]] = distinct !{[[META27]], [[META25]]} +; NO-VP: [[LOOP28]] = distinct !{[[LOOP28]], [[META6]], [[META7]]} +; NO-VP: [[LOOP29]] = distinct !{[[LOOP29]], [[META6]]} +; NO-VP: [[META30]] = !{[[META31:![0-9]+]]} +; NO-VP: [[META31]] = distinct !{[[META31]], [[META32:![0-9]+]]} +; NO-VP: [[META32]] = distinct !{[[META32]], !"LVerDomain"} +; NO-VP: [[META33]] = !{[[META34:![0-9]+]]} +; NO-VP: [[META34]] = distinct !{[[META34]], [[META32]]} +; NO-VP: [[LOOP35]] = distinct !{[[LOOP35]], [[META6]], [[META7]]} +; NO-VP: [[LOOP36]] = distinct !{[[LOOP36]], [[META6]]} +; NO-VP: [[LOOP37]] = distinct !{[[LOOP37]], [[META6]], [[META7]]} +; NO-VP: [[LOOP38]] = distinct !{[[LOOP38]], [[META6]]} +; NO-VP: [[LOOP39]] = distinct !{[[LOOP39]], [[META6]], [[META7]]} +; NO-VP: [[LOOP40]] = distinct !{[[LOOP40]], [[META6]]} +; NO-VP: [[LOOP41]] = distinct !{[[LOOP41]], [[META6]], [[META7]]} +; NO-VP: [[LOOP42]] = distinct !{[[LOOP42]], [[META6]]} +; NO-VP: [[LOOP43]] = distinct !{[[LOOP43]], [[META6]], [[META7]]} +; NO-VP: [[LOOP44]] = distinct !{[[LOOP44]], [[META6]]} +; NO-VP: [[LOOP45]] = distinct !{[[LOOP45]], [[META6]], [[META7]]} +; NO-VP: [[LOOP46]] = distinct !{[[LOOP46]], [[META6]]} +; NO-VP: [[LOOP47]] = distinct !{[[LOOP47]], [[META6]], [[META7]]} +; NO-VP: [[LOOP48]] = distinct !{[[LOOP48]], [[META7]], [[META6]]} +;. diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-cond-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-cond-reduction.ll index f8e8435..5f53cb6 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-cond-reduction.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-cond-reduction.ll @@ -1,24 +1,20 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4 ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefixes=IF-EVL-OUTLOOP ; RUN: opt -passes=loop-vectorize \ ; RUN: -prefer-inloop-reductions \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefixes=IF-EVL-INLOOP ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefixes=NO-VP-OUTLOOP ; RUN: opt -passes=loop-vectorize \ ; RUN: -prefer-inloop-reductions \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefixes=NO-VP-INLOOP define i32 @cond_add(ptr %a, i64 %n, i32 %start) { diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-cost.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-cost.ll index cd53ea0..238aed1 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-cost.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-cost.ll @@ -3,7 +3,7 @@ ; RUN: -mtriple riscv64-linux-gnu -mattr=+v,+f -S -disable-output -debug-only=loop-vectorize 2>&1 | FileCheck %s --check-prefix=DATA ; RUN: opt < %s -passes=loop-vectorize -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ -; RUN: -mtriple riscv64-linux-gnu -force-tail-folding-style=data-with-evl -mattr=+v,+f -S \ +; RUN: -mtriple riscv64-linux-gnu -mattr=+v,+f -S \ ; RUN: -disable-output -debug-only=loop-vectorize 2>&1 | FileCheck %s --check-prefix=EVL ; DATA: Cost of 2 for VF 2: EMIT{{.*}} = active lane mask diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-div.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-div.ll index 0a66ce8..f27f49e 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-div.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-div.ll @@ -1,12 +1,10 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5 ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s --check-prefix=IF-EVL ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s --check-prefix=NO-VP define void @test_sdiv(ptr noalias %a, ptr noalias %b, ptr noalias %c) { @@ -65,9 +63,38 @@ define void @test_sdiv(ptr noalias %a, ptr noalias %b, ptr noalias %c) { ; NO-VP-LABEL: define void @test_sdiv( ; NO-VP-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR0:[0-9]+]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 2 +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP12]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP13]], 2 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP6:%.*]] = getelementptr i64, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP6]], align 8 +; NO-VP-NEXT: [[TMP7:%.*]] = getelementptr i64, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD1:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 8 +; NO-VP-NEXT: [[TMP8:%.*]] = sdiv <vscale x 2 x i64> [[WIDE_LOAD]], [[WIDE_LOAD1]] +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr i64, ptr [[C]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 2 x i64> [[TMP8]], ptr [[TMP9]], align 8 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] +; NO-VP-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[A_GEP:%.*]] = getelementptr i64, ptr [[A]], i64 [[IV]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i64, ptr [[A_GEP]], align 8 ; NO-VP-NEXT: [[B_GEP:%.*]] = getelementptr i64, ptr [[B]], i64 [[IV]] @@ -77,7 +104,7 @@ define void @test_sdiv(ptr noalias %a, ptr noalias %b, ptr noalias %c) { ; NO-VP-NEXT: store i64 [[TMP2]], ptr [[C_GEP]], align 8 ; NO-VP-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1 ; NO-VP-NEXT: [[DONE:%.*]] = icmp eq i64 [[IV_NEXT]], 1024 -; NO-VP-NEXT: br i1 [[DONE]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DONE]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -158,9 +185,38 @@ define void @test_udiv(ptr noalias %a, ptr noalias %b, ptr noalias %c) { ; NO-VP-LABEL: define void @test_udiv( ; NO-VP-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 2 +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP12]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP13]], 2 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP6:%.*]] = getelementptr i64, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP6]], align 8 +; NO-VP-NEXT: [[TMP7:%.*]] = getelementptr i64, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD1:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 8 +; NO-VP-NEXT: [[TMP8:%.*]] = udiv <vscale x 2 x i64> [[WIDE_LOAD]], [[WIDE_LOAD1]] +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr i64, ptr [[C]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 2 x i64> [[TMP8]], ptr [[TMP9]], align 8 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] +; NO-VP-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[A_GEP:%.*]] = getelementptr i64, ptr [[A]], i64 [[IV]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i64, ptr [[A_GEP]], align 8 ; NO-VP-NEXT: [[B_GEP:%.*]] = getelementptr i64, ptr [[B]], i64 [[IV]] @@ -170,7 +226,7 @@ define void @test_udiv(ptr noalias %a, ptr noalias %b, ptr noalias %c) { ; NO-VP-NEXT: store i64 [[TMP2]], ptr [[C_GEP]], align 8 ; NO-VP-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1 ; NO-VP-NEXT: [[DONE:%.*]] = icmp eq i64 [[IV_NEXT]], 1024 -; NO-VP-NEXT: br i1 [[DONE]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DONE]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP5:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -250,9 +306,38 @@ define void @test_srem(ptr noalias %a, ptr noalias %b, ptr noalias %c) { ; NO-VP-LABEL: define void @test_srem( ; NO-VP-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 2 +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP12]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP13]], 2 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP6:%.*]] = getelementptr i64, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP6]], align 8 +; NO-VP-NEXT: [[TMP7:%.*]] = getelementptr i64, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD1:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 8 +; NO-VP-NEXT: [[TMP8:%.*]] = srem <vscale x 2 x i64> [[WIDE_LOAD]], [[WIDE_LOAD1]] +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr i64, ptr [[C]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 2 x i64> [[TMP8]], ptr [[TMP9]], align 8 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] +; NO-VP-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[A_GEP:%.*]] = getelementptr i64, ptr [[A]], i64 [[IV]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i64, ptr [[A_GEP]], align 8 ; NO-VP-NEXT: [[B_GEP:%.*]] = getelementptr i64, ptr [[B]], i64 [[IV]] @@ -262,7 +347,7 @@ define void @test_srem(ptr noalias %a, ptr noalias %b, ptr noalias %c) { ; NO-VP-NEXT: store i64 [[TMP2]], ptr [[C_GEP]], align 8 ; NO-VP-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1 ; NO-VP-NEXT: [[DONE:%.*]] = icmp eq i64 [[IV_NEXT]], 1024 -; NO-VP-NEXT: br i1 [[DONE]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DONE]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP7:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -342,9 +427,38 @@ define void @test_urem(ptr noalias %a, ptr noalias %b, ptr noalias %c) { ; NO-VP-LABEL: define void @test_urem( ; NO-VP-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR0]] { ; NO-VP-NEXT: [[LOOP_PREHEADER:.*]]: +; NO-VP-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 2 +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP12]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]] +; NO-VP: [[VECTOR_PH]]: +; NO-VP-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP13]], 2 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2 +; NO-VP-NEXT: br label %[[VECTOR_BODY:.*]] +; NO-VP: [[VECTOR_BODY]]: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; NO-VP-NEXT: [[TMP6:%.*]] = getelementptr i64, ptr [[A]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP6]], align 8 +; NO-VP-NEXT: [[TMP7:%.*]] = getelementptr i64, ptr [[B]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD1:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 8 +; NO-VP-NEXT: [[TMP8:%.*]] = urem <vscale x 2 x i64> [[WIDE_LOAD]], [[WIDE_LOAD1]] +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr i64, ptr [[C]], i64 [[INDEX]] +; NO-VP-NEXT: store <vscale x 2 x i64> [[TMP8]], ptr [[TMP9]], align 8 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] +; NO-VP-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]] +; NO-VP: [[MIDDLE_BLOCK]]: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]] +; NO-VP: [[SCALAR_PH]]: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ] ; NO-VP-NEXT: br label %[[LOOP:.*]] ; NO-VP: [[LOOP]]: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ] +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] ; NO-VP-NEXT: [[A_GEP:%.*]] = getelementptr i64, ptr [[A]], i64 [[IV]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i64, ptr [[A_GEP]], align 8 ; NO-VP-NEXT: [[B_GEP:%.*]] = getelementptr i64, ptr [[B]], i64 [[IV]] @@ -354,7 +468,7 @@ define void @test_urem(ptr noalias %a, ptr noalias %b, ptr noalias %c) { ; NO-VP-NEXT: store i64 [[TMP2]], ptr [[C_GEP]], align 8 ; NO-VP-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1 ; NO-VP-NEXT: [[DONE:%.*]] = icmp eq i64 [[IV_NEXT]], 1024 -; NO-VP-NEXT: br i1 [[DONE]], label %[[EXIT:.*]], label %[[LOOP]] +; NO-VP-NEXT: br i1 [[DONE]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP9:![0-9]+]] ; NO-VP: [[EXIT]]: ; NO-VP-NEXT: ret void ; @@ -390,3 +504,14 @@ exit: ; IF-EVL: [[LOOP9]] = distinct !{[[LOOP9]], [[META1]], [[META2]], [[META3]]} ; IF-EVL: [[LOOP10]] = distinct !{[[LOOP10]], [[META3]], [[META1]]} ;. +; NO-VP: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]} +; NO-VP: [[META1]] = !{!"llvm.loop.isvectorized", i32 1} +; NO-VP: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"} +; NO-VP: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]} +; NO-VP: [[LOOP4]] = distinct !{[[LOOP4]], [[META1]], [[META2]]} +; NO-VP: [[LOOP5]] = distinct !{[[LOOP5]], [[META2]], [[META1]]} +; NO-VP: [[LOOP6]] = distinct !{[[LOOP6]], [[META1]], [[META2]]} +; NO-VP: [[LOOP7]] = distinct !{[[LOOP7]], [[META2]], [[META1]]} +; NO-VP: [[LOOP8]] = distinct !{[[LOOP8]], [[META1]], [[META2]]} +; NO-VP: [[LOOP9]] = distinct !{[[LOOP9]], [[META2]], [[META1]]} +;. diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-fixed-order-recurrence.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-fixed-order-recurrence.ll index 7c8f111a..6a6bca2d 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-fixed-order-recurrence.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-fixed-order-recurrence.ll @@ -1,14 +1,12 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5 ; RUN: opt -passes=loop-vectorize \ ; RUN: -prefer-inloop-reductions \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v,+f -S < %s| FileCheck %s --check-prefix=IF-EVL ; RUN: opt -passes=loop-vectorize \ ; RUN: -prefer-inloop-reductions \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v,+f -S < %s| FileCheck %s --check-prefix=NO-VP define void @first_order_recurrence(ptr noalias %A, ptr noalias %B, i64 %TC) { @@ -482,20 +480,62 @@ define i32 @FOR_reduction(ptr noalias %A, ptr noalias %B, i64 %TC) { ; IF-EVL-LABEL: define i32 @FOR_reduction( ; IF-EVL-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], i64 [[TC:%.*]]) #[[ATTR0]] { ; IF-EVL-NEXT: [[ENTRY:.*]]: +; IF-EVL-NEXT: [[TMP9:%.*]] = call i64 @llvm.vscale.i64() +; IF-EVL-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP9]], 4 +; IF-EVL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TC]], [[TMP1]] +; IF-EVL-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]] +; IF-EVL: [[VECTOR_PH]]: +; IF-EVL-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; IF-EVL-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 4 +; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TC]], [[TMP3]] +; IF-EVL-NEXT: [[N_VEC:%.*]] = sub i64 [[TC]], [[N_MOD_VF]] +; IF-EVL-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; IF-EVL-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4 +; IF-EVL-NEXT: [[TMP6:%.*]] = call i32 @llvm.vscale.i32() +; IF-EVL-NEXT: [[TMP7:%.*]] = mul nuw i32 [[TMP6]], 4 +; IF-EVL-NEXT: [[TMP8:%.*]] = sub i32 [[TMP7]], 1 +; IF-EVL-NEXT: [[VECTOR_RECUR_INIT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 33, i32 [[TMP8]] +; IF-EVL-NEXT: br label %[[VECTOR_BODY:.*]] +; IF-EVL: [[VECTOR_BODY]]: +; IF-EVL-NEXT: [[INDVARS:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] +; IF-EVL-NEXT: [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i32> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ] +; IF-EVL-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[INDVARS]] +; IF-EVL-NEXT: [[WIDE_LOAD]] = load <vscale x 4 x i32>, ptr [[ARRAYIDX]], align 4 +; IF-EVL-NEXT: [[TMP10:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR]], <vscale x 4 x i32> [[WIDE_LOAD]], i32 -1) +; IF-EVL-NEXT: [[TMP11:%.*]] = add nsw <vscale x 4 x i32> [[TMP10]], [[WIDE_LOAD]] +; IF-EVL-NEXT: [[TMP12:%.*]] = getelementptr inbounds nuw i32, ptr [[B]], i64 [[INDVARS]] +; IF-EVL-NEXT: store <vscale x 4 x i32> [[TMP11]], ptr [[TMP12]], align 4 +; IF-EVL-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDVARS]], [[TMP5]] +; IF-EVL-NEXT: [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; IF-EVL-NEXT: br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]] +; IF-EVL: [[MIDDLE_BLOCK]]: +; IF-EVL-NEXT: [[TMP14:%.*]] = call i32 @llvm.vscale.i32() +; IF-EVL-NEXT: [[TMP15:%.*]] = mul nuw i32 [[TMP14]], 4 +; IF-EVL-NEXT: [[TMP16:%.*]] = sub i32 [[TMP15]], 1 +; IF-EVL-NEXT: [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <vscale x 4 x i32> [[WIDE_LOAD]], i32 [[TMP16]] +; IF-EVL-NEXT: [[TMP17:%.*]] = call i32 @llvm.vscale.i32() +; IF-EVL-NEXT: [[TMP18:%.*]] = mul nuw i32 [[TMP17]], 4 +; IF-EVL-NEXT: [[TMP19:%.*]] = sub i32 [[TMP18]], 2 +; IF-EVL-NEXT: [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <vscale x 4 x i32> [[WIDE_LOAD]], i32 [[TMP19]] +; IF-EVL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TC]], [[N_VEC]] +; IF-EVL-NEXT: br i1 [[CMP_N]], label %[[FOR_END:.*]], label %[[SCALAR_PH]] +; IF-EVL: [[SCALAR_PH]]: +; IF-EVL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ] +; IF-EVL-NEXT: [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ] ; IF-EVL-NEXT: br label %[[FOR_BODY:.*]] ; IF-EVL: [[FOR_BODY]]: -; IF-EVL-NEXT: [[INDVARS:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[INDVARS_NEXT:%.*]], %[[FOR_BODY]] ] -; IF-EVL-NEXT: [[FOR1:%.*]] = phi i32 [ 33, %[[ENTRY]] ], [ [[TMP0:%.*]], %[[FOR_BODY]] ] -; IF-EVL-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[INDVARS]] -; IF-EVL-NEXT: [[TMP0]] = load i32, ptr [[ARRAYIDX]], align 4 +; IF-EVL-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_NEXT:%.*]], %[[FOR_BODY]] ] +; IF-EVL-NEXT: [[FOR1:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], %[[SCALAR_PH]] ], [ [[TMP0:%.*]], %[[FOR_BODY]] ] +; IF-EVL-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[IV]] +; IF-EVL-NEXT: [[TMP0]] = load i32, ptr [[ARRAYIDX1]], align 4 ; IF-EVL-NEXT: [[ADD:%.*]] = add nsw i32 [[FOR1]], [[TMP0]] -; IF-EVL-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds nuw i32, ptr [[B]], i64 [[INDVARS]] +; IF-EVL-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds nuw i32, ptr [[B]], i64 [[IV]] ; IF-EVL-NEXT: store i32 [[ADD]], ptr [[ARRAYIDX2]], align 4 -; IF-EVL-NEXT: [[INDVARS_NEXT]] = add nuw nsw i64 [[INDVARS]], 1 +; IF-EVL-NEXT: [[INDVARS_NEXT]] = add nuw nsw i64 [[IV]], 1 ; IF-EVL-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_NEXT]], [[TC]] -; IF-EVL-NEXT: br i1 [[EXITCOND_NOT]], label %[[FOR_END:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]] +; IF-EVL-NEXT: br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]] ; IF-EVL: [[FOR_END]]: -; IF-EVL-NEXT: [[FOR1_LCSSA:%.*]] = phi i32 [ [[FOR1]], %[[FOR_BODY]] ] +; IF-EVL-NEXT: [[FOR1_LCSSA:%.*]] = phi i32 [ [[FOR1]], %[[FOR_BODY]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ] ; IF-EVL-NEXT: ret i32 [[FOR1_LCSSA]] ; ; NO-VP-LABEL: define i32 @FOR_reduction( @@ -733,8 +773,8 @@ for.end: ; IF-EVL: [[LOOP6]] = distinct !{[[LOOP6]], [[META3]], [[META1]]} ; IF-EVL: [[LOOP7]] = distinct !{[[LOOP7]], [[META1]], [[META2]], [[META3]]} ; IF-EVL: [[LOOP8]] = distinct !{[[LOOP8]], [[META3]], [[META1]]} -; IF-EVL: [[LOOP9]] = distinct !{[[LOOP9]], [[META10:![0-9]+]]} -; IF-EVL: [[META10]] = !{!"llvm.loop.vectorize.enable", i1 true} +; IF-EVL: [[LOOP9]] = distinct !{[[LOOP9]], [[META1]], [[META3]]} +; IF-EVL: [[LOOP10]] = distinct !{[[LOOP10]], [[META3]], [[META1]]} ; IF-EVL: [[LOOP11]] = distinct !{[[LOOP11]], [[META1]], [[META2]], [[META3]]} ; IF-EVL: [[LOOP12]] = distinct !{[[LOOP12]], [[META3]], [[META1]]} ;. diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-gather-scatter.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-gather-scatter.ll index 2b7a9fb..892293b 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-gather-scatter.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-gather-scatter.ll @@ -1,12 +1,10 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefix=IF-EVL ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefix=NO-VP define void @gather_scatter(ptr noalias %in, ptr noalias %out, ptr noalias %index, i64 %n) { @@ -69,18 +67,54 @@ define void @gather_scatter(ptr noalias %in, ptr noalias %out, ptr noalias %inde ; ; NO-VP-LABEL: @gather_scatter( ; NO-VP-NEXT: entry: +; NO-VP-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP14:%.*]] = mul nuw i64 [[TMP13]], 2 +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP14]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] +; NO-VP: vector.ph: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 2 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2 +; NO-VP-NEXT: [[TMP6:%.*]] = call <vscale x 2 x i64> @llvm.stepvector.nxv2i64() +; NO-VP-NEXT: [[TMP7:%.*]] = mul <vscale x 2 x i64> [[TMP6]], splat (i64 1) +; NO-VP-NEXT: [[INDUCTION:%.*]] = add <vscale x 2 x i64> zeroinitializer, [[TMP7]] +; NO-VP-NEXT: [[TMP8:%.*]] = mul i64 1, [[TMP5]] +; NO-VP-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP8]], i64 0 +; NO-VP-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer ; NO-VP-NEXT: br label [[FOR_BODY:%.*]] +; NO-VP: vector.body: +; NO-VP-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] +; NO-VP-NEXT: [[VEC_IND:%.*]] = phi <vscale x 2 x i64> [ [[INDUCTION]], [[ENTRY]] ], [ [[VEC_IND_NEXT:%.*]], [[FOR_BODY]] ] +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[INDEX:%.*]], <vscale x 2 x i64> [[VEC_IND]] +; NO-VP-NEXT: [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 2 x i64> @llvm.masked.gather.nxv2i64.nxv2p0(<vscale x 2 x ptr> [[TMP9]], i32 8, <vscale x 2 x i1> splat (i1 true), <vscale x 2 x i64> poison) +; NO-VP-NEXT: [[TMP10:%.*]] = getelementptr inbounds float, ptr [[IN:%.*]], <vscale x 2 x i64> [[WIDE_MASKED_GATHER]] +; NO-VP-NEXT: [[WIDE_MASKED_GATHER2:%.*]] = call <vscale x 2 x float> @llvm.masked.gather.nxv2f32.nxv2p0(<vscale x 2 x ptr> [[TMP10]], i32 4, <vscale x 2 x i1> splat (i1 true), <vscale x 2 x float> poison) +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr inbounds float, ptr [[OUT:%.*]], <vscale x 2 x i64> [[WIDE_MASKED_GATHER]] +; NO-VP-NEXT: call void @llvm.masked.scatter.nxv2f32.nxv2p0(<vscale x 2 x float> [[WIDE_MASKED_GATHER2]], <vscale x 2 x ptr> [[TMP11]], i32 4, <vscale x 2 x i1> splat (i1 true)) +; NO-VP-NEXT: [[INDVARS_IV_NEXT]] = add nuw i64 [[INDVARS_IV]], [[TMP5]] +; NO-VP-NEXT: [[VEC_IND_NEXT]] = add <vscale x 2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT]] +; NO-VP-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] +; NO-VP: middle.block: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]] +; NO-VP: scalar.ph: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ] +; NO-VP-NEXT: br label [[FOR_BODY1:%.*]] ; NO-VP: for.body: -; NO-VP-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] -; NO-VP-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, ptr [[INDEX:%.*]], i64 [[INDVARS_IV]] +; NO-VP-NEXT: [[INDVARS_IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT1:%.*]], [[FOR_BODY1]] ] +; NO-VP-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, ptr [[INDEX]], i64 [[INDVARS_IV1]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i64, ptr [[ARRAYIDX3]], align 8 -; NO-VP-NEXT: [[ARRAYIDX5:%.*]] = getelementptr inbounds float, ptr [[IN:%.*]], i64 [[TMP0]] +; NO-VP-NEXT: [[ARRAYIDX5:%.*]] = getelementptr inbounds float, ptr [[IN]], i64 [[TMP0]] ; NO-VP-NEXT: [[TMP1:%.*]] = load float, ptr [[ARRAYIDX5]], align 4 -; NO-VP-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds float, ptr [[OUT:%.*]], i64 [[TMP0]] +; NO-VP-NEXT: [[ARRAYIDX7:%.*]] = getelementptr inbounds float, ptr [[OUT]], i64 [[TMP0]] ; NO-VP-NEXT: store float [[TMP1]], ptr [[ARRAYIDX7]], align 4 -; NO-VP-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 -; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N:%.*]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END:%.*]], label [[FOR_BODY]] +; NO-VP-NEXT: [[INDVARS_IV_NEXT1]] = add nuw nsw i64 [[INDVARS_IV1]], 1 +; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT1]], [[N]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY1]], !llvm.loop [[LOOP3:![0-9]+]] ; NO-VP: for.end: ; NO-VP-NEXT: ret void ; diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-inloop-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-inloop-reduction.ll index a7955a5..e0594ad 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-inloop-reduction.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-inloop-reduction.ll @@ -1,14 +1,12 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py ; RUN: opt -passes=loop-vectorize \ ; RUN: -prefer-inloop-reductions \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v,+f -S < %s| FileCheck %s --check-prefix=IF-EVL ; RUN: opt -passes=loop-vectorize \ ; RUN: -prefer-inloop-reductions \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v,+f -S < %s| FileCheck %s --check-prefix=NO-VP define i32 @add(ptr %a, i64 %n, i32 %start) { @@ -124,35 +122,34 @@ for.end: define i32 @mul(ptr %a, i64 %n, i32 %start) { ; IF-EVL-LABEL: @mul( ; IF-EVL-NEXT: entry: -; IF-EVL-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] +; IF-EVL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N_RND_UP:%.*]], 8 +; IF-EVL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] ; IF-EVL: vector.ph: -; IF-EVL-NEXT: [[N_RND_UP:%.*]] = add i64 [[N:%.*]], 3 -; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 4 +; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 8 ; IF-EVL-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]] -; IF-EVL-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1 -; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0 -; IF-EVL-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT1]], <4 x i64> poison, <4 x i32> zeroinitializer ; IF-EVL-NEXT: br label [[FOR_BODY:%.*]] ; IF-EVL: vector.body: ; IF-EVL-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ] ; IF-EVL-NEXT: [[RDX:%.*]] = phi i32 [ [[START:%.*]], [[ENTRY]] ], [ [[MUL:%.*]], [[FOR_BODY]] ] -; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[IV]], i64 0 -; IF-EVL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer -; IF-EVL-NEXT: [[VEC_IV:%.*]] = add <4 x i64> [[BROADCAST_SPLAT]], <i64 0, i64 1, i64 2, i64 3> -; IF-EVL-NEXT: [[TMP1:%.*]] = icmp ule <4 x i64> [[VEC_IV]], [[BROADCAST_SPLAT2]] +; IF-EVL-NEXT: [[VEC_PHI1:%.*]] = phi i32 [ 1, [[ENTRY]] ], [ [[TMP6:%.*]], [[FOR_BODY]] ] ; IF-EVL-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IV]] -; IF-EVL-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP2]], i32 4, <4 x i1> [[TMP1]], <4 x i32> poison) -; IF-EVL-NEXT: [[TMP4:%.*]] = select <4 x i1> [[TMP1]], <4 x i32> [[WIDE_MASKED_LOAD]], <4 x i32> splat (i32 1) +; IF-EVL-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 4 +; IF-EVL-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4 +; IF-EVL-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4 ; IF-EVL-NEXT: [[TMP5:%.*]] = call i32 @llvm.vector.reduce.mul.v4i32(<4 x i32> [[TMP4]]) ; IF-EVL-NEXT: [[MUL]] = mul i32 [[TMP5]], [[RDX]] -; IF-EVL-NEXT: [[IV_NEXT]] = add i64 [[IV]], 4 +; IF-EVL-NEXT: [[TMP8:%.*]] = call i32 @llvm.vector.reduce.mul.v4i32(<4 x i32> [[WIDE_LOAD2]]) +; IF-EVL-NEXT: [[TMP6]] = mul i32 [[TMP8]], [[VEC_PHI1]] +; IF-EVL-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], 8 ; IF-EVL-NEXT: [[TMP7:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]] ; IF-EVL-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]] ; IF-EVL: middle.block: -; IF-EVL-NEXT: br label [[FOR_END:%.*]] +; IF-EVL-NEXT: [[BIN_RDX:%.*]] = mul i32 [[TMP6]], [[MUL]] +; IF-EVL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_RND_UP]], [[N_VEC]] +; IF-EVL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]] ; IF-EVL: scalar.ph: -; IF-EVL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY1:%.*]] ] -; IF-EVL-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[START]], [[ENTRY1]] ] +; IF-EVL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ] +; IF-EVL-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY1]] ] ; IF-EVL-NEXT: br label [[FOR_BODY1:%.*]] ; IF-EVL: for.body: ; IF-EVL-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ] @@ -161,10 +158,10 @@ define i32 @mul(ptr %a, i64 %n, i32 %start) { ; IF-EVL-NEXT: [[TMP0:%.*]] = load i32, ptr [[ARRAYIDX]], align 4 ; IF-EVL-NEXT: [[MUL1]] = mul nsw i32 [[TMP0]], [[RDX1]] ; IF-EVL-NEXT: [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1 -; IF-EVL-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N]] +; IF-EVL-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N_RND_UP]] ; IF-EVL-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY1]], !llvm.loop [[LOOP6:![0-9]+]] ; IF-EVL: for.end: -; IF-EVL-NEXT: [[MUL_LCSSA:%.*]] = phi i32 [ [[MUL1]], [[FOR_BODY1]] ], [ [[MUL]], [[MIDDLE_BLOCK]] ] +; IF-EVL-NEXT: [[MUL_LCSSA:%.*]] = phi i32 [ [[MUL1]], [[FOR_BODY1]] ], [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ] ; IF-EVL-NEXT: ret i32 [[MUL_LCSSA]] ; ; NO-VP-LABEL: @mul( @@ -1116,35 +1113,34 @@ for.end: define float @fmul(ptr %a, i64 %n, float %start) { ; IF-EVL-LABEL: @fmul( ; IF-EVL-NEXT: entry: -; IF-EVL-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] +; IF-EVL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N_RND_UP:%.*]], 8 +; IF-EVL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] ; IF-EVL: vector.ph: -; IF-EVL-NEXT: [[N_RND_UP:%.*]] = add i64 [[N:%.*]], 3 -; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 4 +; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 8 ; IF-EVL-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]] -; IF-EVL-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1 -; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0 -; IF-EVL-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT1]], <4 x i64> poison, <4 x i32> zeroinitializer ; IF-EVL-NEXT: br label [[FOR_BODY:%.*]] ; IF-EVL: vector.body: ; IF-EVL-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ] ; IF-EVL-NEXT: [[RDX:%.*]] = phi float [ [[START:%.*]], [[ENTRY]] ], [ [[MUL:%.*]], [[FOR_BODY]] ] -; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[IV]], i64 0 -; IF-EVL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer -; IF-EVL-NEXT: [[VEC_IV:%.*]] = add <4 x i64> [[BROADCAST_SPLAT]], <i64 0, i64 1, i64 2, i64 3> -; IF-EVL-NEXT: [[TMP1:%.*]] = icmp ule <4 x i64> [[VEC_IV]], [[BROADCAST_SPLAT2]] +; IF-EVL-NEXT: [[VEC_PHI1:%.*]] = phi float [ 1.000000e+00, [[ENTRY]] ], [ [[TMP6:%.*]], [[FOR_BODY]] ] ; IF-EVL-NEXT: [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[IV]] -; IF-EVL-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP2]], i32 4, <4 x i1> [[TMP1]], <4 x float> poison) -; IF-EVL-NEXT: [[TMP4:%.*]] = select reassoc <4 x i1> [[TMP1]], <4 x float> [[WIDE_MASKED_LOAD]], <4 x float> splat (float 1.000000e+00) +; IF-EVL-NEXT: [[TMP1:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 4 +; IF-EVL-NEXT: [[TMP4:%.*]] = load <4 x float>, ptr [[TMP2]], align 4 +; IF-EVL-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP1]], align 4 ; IF-EVL-NEXT: [[TMP5:%.*]] = call reassoc float @llvm.vector.reduce.fmul.v4f32(float 1.000000e+00, <4 x float> [[TMP4]]) ; IF-EVL-NEXT: [[MUL]] = fmul reassoc float [[TMP5]], [[RDX]] -; IF-EVL-NEXT: [[IV_NEXT]] = add i64 [[IV]], 4 +; IF-EVL-NEXT: [[TMP8:%.*]] = call reassoc float @llvm.vector.reduce.fmul.v4f32(float 1.000000e+00, <4 x float> [[WIDE_LOAD2]]) +; IF-EVL-NEXT: [[TMP6]] = fmul reassoc float [[TMP8]], [[VEC_PHI1]] +; IF-EVL-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], 8 ; IF-EVL-NEXT: [[TMP7:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]] ; IF-EVL-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]] ; IF-EVL: middle.block: -; IF-EVL-NEXT: br label [[FOR_END:%.*]] +; IF-EVL-NEXT: [[BIN_RDX:%.*]] = fmul reassoc float [[TMP6]], [[MUL]] +; IF-EVL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_RND_UP]], [[N_VEC]] +; IF-EVL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]] ; IF-EVL: scalar.ph: -; IF-EVL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY1:%.*]] ] -; IF-EVL-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[START]], [[ENTRY1]] ] +; IF-EVL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ] +; IF-EVL-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY1]] ] ; IF-EVL-NEXT: br label [[FOR_BODY1:%.*]] ; IF-EVL: for.body: ; IF-EVL-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ] @@ -1153,10 +1149,10 @@ define float @fmul(ptr %a, i64 %n, float %start) { ; IF-EVL-NEXT: [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4 ; IF-EVL-NEXT: [[MUL1]] = fmul reassoc float [[TMP0]], [[RDX1]] ; IF-EVL-NEXT: [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1 -; IF-EVL-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N]] +; IF-EVL-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N_RND_UP]] ; IF-EVL-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY1]], !llvm.loop [[LOOP24:![0-9]+]] ; IF-EVL: for.end: -; IF-EVL-NEXT: [[MUL_LCSSA:%.*]] = phi float [ [[MUL1]], [[FOR_BODY1]] ], [ [[MUL]], [[MIDDLE_BLOCK]] ] +; IF-EVL-NEXT: [[MUL_LCSSA:%.*]] = phi float [ [[MUL1]], [[FOR_BODY1]] ], [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ] ; IF-EVL-NEXT: ret float [[MUL_LCSSA]] ; ; NO-VP-LABEL: @fmul( @@ -1451,37 +1447,35 @@ for.end: define float @fminimum(ptr %a, i64 %n, float %start) { ; IF-EVL-LABEL: @fminimum( ; IF-EVL-NEXT: entry: -; IF-EVL-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] +; IF-EVL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N_RND_UP:%.*]], 16 +; IF-EVL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] ; IF-EVL: vector.ph: -; IF-EVL-NEXT: [[N_RND_UP:%.*]] = add i64 [[N:%.*]], 7 -; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 8 +; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 16 ; IF-EVL-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]] -; IF-EVL-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1 -; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <8 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0 -; IF-EVL-NEXT: [[BROADCAST_SPLAT4:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT3]], <8 x i64> poison, <8 x i32> zeroinitializer ; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <8 x float> poison, float [[START:%.*]], i64 0 ; IF-EVL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x float> [[BROADCAST_SPLATINSERT2]], <8 x float> poison, <8 x i32> zeroinitializer ; IF-EVL-NEXT: br label [[FOR_BODY:%.*]] ; IF-EVL: vector.body: ; IF-EVL-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ] ; IF-EVL-NEXT: [[VEC_PHI:%.*]] = phi <8 x float> [ [[BROADCAST_SPLAT]], [[ENTRY]] ], [ [[TMP4:%.*]], [[FOR_BODY]] ] -; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <8 x i64> poison, i64 [[IV]], i64 0 -; IF-EVL-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT1]], <8 x i64> poison, <8 x i32> zeroinitializer -; IF-EVL-NEXT: [[VEC_IV:%.*]] = add <8 x i64> [[BROADCAST_SPLAT2]], <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7> -; IF-EVL-NEXT: [[TMP1:%.*]] = icmp ule <8 x i64> [[VEC_IV]], [[BROADCAST_SPLAT4]] +; IF-EVL-NEXT: [[VEC_PHI1:%.*]] = phi <8 x float> [ [[BROADCAST_SPLAT]], [[ENTRY]] ], [ [[TMP3:%.*]], [[FOR_BODY]] ] ; IF-EVL-NEXT: [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[IV]] -; IF-EVL-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP2]], i32 4, <8 x i1> [[TMP1]], <8 x float> poison) +; IF-EVL-NEXT: [[TMP1:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 8 +; IF-EVL-NEXT: [[WIDE_MASKED_LOAD:%.*]] = load <8 x float>, ptr [[TMP2]], align 4 +; IF-EVL-NEXT: [[WIDE_LOAD2:%.*]] = load <8 x float>, ptr [[TMP1]], align 4 ; IF-EVL-NEXT: [[TMP4]] = call <8 x float> @llvm.minimum.v8f32(<8 x float> [[VEC_PHI]], <8 x float> [[WIDE_MASKED_LOAD]]) -; IF-EVL-NEXT: [[TMP5:%.*]] = select <8 x i1> [[TMP1]], <8 x float> [[TMP4]], <8 x float> [[VEC_PHI]] -; IF-EVL-NEXT: [[IV_NEXT]] = add i64 [[IV]], 8 +; IF-EVL-NEXT: [[TMP3]] = call <8 x float> @llvm.minimum.v8f32(<8 x float> [[VEC_PHI1]], <8 x float> [[WIDE_LOAD2]]) +; IF-EVL-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], 16 ; IF-EVL-NEXT: [[TMP6:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]] ; IF-EVL-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP29:![0-9]+]] ; IF-EVL: middle.block: +; IF-EVL-NEXT: [[TMP5:%.*]] = call <8 x float> @llvm.minimum.v8f32(<8 x float> [[TMP4]], <8 x float> [[TMP3]]) ; IF-EVL-NEXT: [[TMP7:%.*]] = call float @llvm.vector.reduce.fminimum.v8f32(<8 x float> [[TMP5]]) -; IF-EVL-NEXT: br label [[FOR_END:%.*]] +; IF-EVL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_RND_UP]], [[N_VEC]] +; IF-EVL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]] ; IF-EVL: scalar.ph: -; IF-EVL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY1:%.*]] ] -; IF-EVL-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[START]], [[ENTRY1]] ] +; IF-EVL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ] +; IF-EVL-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY1]] ] ; IF-EVL-NEXT: br label [[FOR_BODY1:%.*]] ; IF-EVL: for.body: ; IF-EVL-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ] @@ -1490,7 +1484,7 @@ define float @fminimum(ptr %a, i64 %n, float %start) { ; IF-EVL-NEXT: [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4 ; IF-EVL-NEXT: [[MIN]] = tail call float @llvm.minimum.f32(float [[RDX]], float [[TMP0]]) ; IF-EVL-NEXT: [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1 -; IF-EVL-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N]] +; IF-EVL-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N_RND_UP]] ; IF-EVL-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY1]], !llvm.loop [[LOOP30:![0-9]+]] ; IF-EVL: for.end: ; IF-EVL-NEXT: [[MIN_LCSSA:%.*]] = phi float [ [[MIN]], [[FOR_BODY1]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ] @@ -1561,37 +1555,35 @@ for.end: define float @fmaximum(ptr %a, i64 %n, float %start) { ; IF-EVL-LABEL: @fmaximum( ; IF-EVL-NEXT: entry: -; IF-EVL-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] +; IF-EVL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N_RND_UP:%.*]], 16 +; IF-EVL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] ; IF-EVL: vector.ph: -; IF-EVL-NEXT: [[N_RND_UP:%.*]] = add i64 [[N:%.*]], 7 -; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 8 +; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 16 ; IF-EVL-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]] -; IF-EVL-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1 -; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <8 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0 -; IF-EVL-NEXT: [[BROADCAST_SPLAT4:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT3]], <8 x i64> poison, <8 x i32> zeroinitializer ; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <8 x float> poison, float [[START:%.*]], i64 0 ; IF-EVL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x float> [[BROADCAST_SPLATINSERT2]], <8 x float> poison, <8 x i32> zeroinitializer ; IF-EVL-NEXT: br label [[FOR_BODY:%.*]] ; IF-EVL: vector.body: ; IF-EVL-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ] ; IF-EVL-NEXT: [[VEC_PHI:%.*]] = phi <8 x float> [ [[BROADCAST_SPLAT]], [[ENTRY]] ], [ [[TMP4:%.*]], [[FOR_BODY]] ] -; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <8 x i64> poison, i64 [[IV]], i64 0 -; IF-EVL-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT1]], <8 x i64> poison, <8 x i32> zeroinitializer -; IF-EVL-NEXT: [[VEC_IV:%.*]] = add <8 x i64> [[BROADCAST_SPLAT2]], <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7> -; IF-EVL-NEXT: [[TMP1:%.*]] = icmp ule <8 x i64> [[VEC_IV]], [[BROADCAST_SPLAT4]] +; IF-EVL-NEXT: [[VEC_PHI1:%.*]] = phi <8 x float> [ [[BROADCAST_SPLAT]], [[ENTRY]] ], [ [[TMP3:%.*]], [[FOR_BODY]] ] ; IF-EVL-NEXT: [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[IV]] -; IF-EVL-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP2]], i32 4, <8 x i1> [[TMP1]], <8 x float> poison) +; IF-EVL-NEXT: [[TMP1:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 8 +; IF-EVL-NEXT: [[WIDE_MASKED_LOAD:%.*]] = load <8 x float>, ptr [[TMP2]], align 4 +; IF-EVL-NEXT: [[WIDE_LOAD2:%.*]] = load <8 x float>, ptr [[TMP1]], align 4 ; IF-EVL-NEXT: [[TMP4]] = call <8 x float> @llvm.maximum.v8f32(<8 x float> [[VEC_PHI]], <8 x float> [[WIDE_MASKED_LOAD]]) -; IF-EVL-NEXT: [[TMP5:%.*]] = select <8 x i1> [[TMP1]], <8 x float> [[TMP4]], <8 x float> [[VEC_PHI]] -; IF-EVL-NEXT: [[IV_NEXT]] = add i64 [[IV]], 8 +; IF-EVL-NEXT: [[TMP3]] = call <8 x float> @llvm.maximum.v8f32(<8 x float> [[VEC_PHI1]], <8 x float> [[WIDE_LOAD2]]) +; IF-EVL-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], 16 ; IF-EVL-NEXT: [[TMP6:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]] ; IF-EVL-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP31:![0-9]+]] ; IF-EVL: middle.block: +; IF-EVL-NEXT: [[TMP5:%.*]] = call <8 x float> @llvm.maximum.v8f32(<8 x float> [[TMP4]], <8 x float> [[TMP3]]) ; IF-EVL-NEXT: [[TMP7:%.*]] = call float @llvm.vector.reduce.fmaximum.v8f32(<8 x float> [[TMP5]]) -; IF-EVL-NEXT: br label [[FOR_END:%.*]] +; IF-EVL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_RND_UP]], [[N_VEC]] +; IF-EVL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]] ; IF-EVL: scalar.ph: -; IF-EVL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY1:%.*]] ] -; IF-EVL-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[START]], [[ENTRY1]] ] +; IF-EVL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ] +; IF-EVL-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY1]] ] ; IF-EVL-NEXT: br label [[FOR_BODY1:%.*]] ; IF-EVL: for.body: ; IF-EVL-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ] @@ -1600,7 +1592,7 @@ define float @fmaximum(ptr %a, i64 %n, float %start) { ; IF-EVL-NEXT: [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4 ; IF-EVL-NEXT: [[MAX]] = tail call float @llvm.maximum.f32(float [[RDX]], float [[TMP0]]) ; IF-EVL-NEXT: [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1 -; IF-EVL-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N]] +; IF-EVL-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N_RND_UP]] ; IF-EVL-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY1]], !llvm.loop [[LOOP32:![0-9]+]] ; IF-EVL: for.end: ; IF-EVL-NEXT: [[MAX_LCSSA:%.*]] = phi float [ [[MAX]], [[FOR_BODY1]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ] diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-interleave.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-interleave.ll index b2ebe6f..80b0714 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-interleave.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-interleave.ll @@ -1,12 +1,10 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck --check-prefix=IF-EVL %s ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck --check-prefix=NO-VP %s ; FIXME: interleaved accesses are not supported yet with predicated vectorization. diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-intermediate-store.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-intermediate-store.ll index 314c201..c7808b1 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-intermediate-store.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-intermediate-store.ll @@ -1,24 +1,20 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4 ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefixes=IF-EVL-OUTLOOP ; RUN: opt -passes=loop-vectorize \ ; RUN: -prefer-inloop-reductions \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefixes=IF-EVL-INLOOP ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefixes=NO-VP-OUTLOOP ; RUN: opt -passes=loop-vectorize \ ; RUN: -prefer-inloop-reductions \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefixes=NO-VP-INLOOP define void @reduction_intermediate_store(ptr %a, i64 %n, i32 %start, ptr %addr) { diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-iv32.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-iv32.ll index 2c265c9..bda9145 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-iv32.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-iv32.ll @@ -1,12 +1,10 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck --check-prefix=IF-EVL %s ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck --check-prefix=NO-VP %s define void @iv32(ptr noalias %a, ptr noalias %b, i32 %N) { diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-known-no-overflow.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-known-no-overflow.ll index 2c77d78..2601bfd 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-known-no-overflow.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-known-no-overflow.ll @@ -1,7 +1,6 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5 ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s ; TODO: We know the IV will never overflow here so we can skip the overflow diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-masked-loadstore.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-masked-loadstore.ll index 1efd539..226372f 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-masked-loadstore.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-masked-loadstore.ll @@ -1,12 +1,10 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefix=IF-EVL ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefix=NO-VP define void @masked_loadstore(ptr noalias %a, ptr noalias %b, i64 %n) { @@ -66,23 +64,52 @@ define void @masked_loadstore(ptr noalias %a, ptr noalias %b, i64 %n) { ; ; NO-VP-LABEL: @masked_loadstore( ; NO-VP-NEXT: entry: +; NO-VP-NEXT: [[TMP11:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 4 +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP12]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] +; NO-VP: vector.ph: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]] +; NO-VP-NEXT: [[INC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4 ; NO-VP-NEXT: br label [[FOR_BODY:%.*]] +; NO-VP: vector.body: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[FOR_BODY]] ] +; NO-VP-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP6]], align 4 +; NO-VP-NEXT: [[TMP7:%.*]] = icmp ne <vscale x 4 x i32> [[WIDE_LOAD]], zeroinitializer +; NO-VP-NEXT: [[TMP8:%.*]] = getelementptr i32, ptr [[A:%.*]], i64 [[INDEX]] +; NO-VP-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP8]], i32 4, <vscale x 4 x i1> [[TMP7]], <vscale x 4 x i32> poison) +; NO-VP-NEXT: [[TMP9:%.*]] = add <vscale x 4 x i32> [[WIDE_LOAD]], [[WIDE_MASKED_LOAD]] +; NO-VP-NEXT: call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP9]], ptr [[TMP8]], i32 4, <vscale x 4 x i1> [[TMP7]]) +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] +; NO-VP-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[INC]] +; NO-VP-NEXT: br i1 [[TMP10]], label [[FOR_INC:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] +; NO-VP: middle.block: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[INC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]] +; NO-VP: scalar.ph: +; NO-VP-NEXT: [[I_011:%.*]] = phi i64 [ [[INC]], [[FOR_INC]] ], [ 0, [[ENTRY:%.*]] ] +; NO-VP-NEXT: br label [[FOR_BODY1:%.*]] ; NO-VP: for.body: -; NO-VP-NEXT: [[I_011:%.*]] = phi i64 [ [[INC:%.*]], [[FOR_INC:%.*]] ], [ 0, [[ENTRY:%.*]] ] -; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[I_011]] +; NO-VP-NEXT: [[I_11:%.*]] = phi i64 [ [[INC1:%.*]], [[FOR_INC1:%.*]] ], [ [[I_011]], [[SCALAR_PH]] ] +; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[I_11]] ; NO-VP-NEXT: [[TMP0:%.*]] = load i32, ptr [[ARRAYIDX]], align 4 ; NO-VP-NEXT: [[CMP1:%.*]] = icmp ne i32 [[TMP0]], 0 -; NO-VP-NEXT: br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[FOR_INC]] +; NO-VP-NEXT: br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[FOR_INC1]] ; NO-VP: if.then: -; NO-VP-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[I_011]] +; NO-VP-NEXT: [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[I_11]] ; NO-VP-NEXT: [[TMP1:%.*]] = load i32, ptr [[ARRAYIDX3]], align 4 ; NO-VP-NEXT: [[ADD:%.*]] = add i32 [[TMP0]], [[TMP1]] ; NO-VP-NEXT: store i32 [[ADD]], ptr [[ARRAYIDX3]], align 4 -; NO-VP-NEXT: br label [[FOR_INC]] +; NO-VP-NEXT: br label [[FOR_INC1]] ; NO-VP: for.inc: -; NO-VP-NEXT: [[INC]] = add nuw nsw i64 [[I_011]], 1 -; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INC]], [[N:%.*]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label [[EXIT:%.*]], label [[FOR_BODY]] +; NO-VP-NEXT: [[INC1]] = add nuw nsw i64 [[I_11]], 1 +; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INC1]], [[N]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label [[EXIT]], label [[FOR_BODY1]], !llvm.loop [[LOOP3:![0-9]+]] ; NO-VP: exit: ; NO-VP-NEXT: ret void ; diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-no-masking.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-no-masking.ll index 1c49fba..6285a6f 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-no-masking.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-no-masking.ll @@ -1,12 +1,10 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s ; No need to emit predicated vector code if the vector instructions with masking are not required. diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-ordered-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-ordered-reduction.ll index 81e0f75..41bddaa 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-ordered-reduction.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-ordered-reduction.ll @@ -1,14 +1,12 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py ; RUN: opt -passes=loop-vectorize \ ; RUN: -force-ordered-reductions=true -hints-allow-reordering=false \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v,+f -S < %s| FileCheck %s --check-prefix=IF-EVL ; RUN: opt -passes=loop-vectorize \ ; RUN: -force-ordered-reductions=true -hints-allow-reordering=false \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v,+f -S < %s| FileCheck %s --check-prefix=NO-VP define float @fadd(ptr noalias nocapture readonly %a, i64 %n) { @@ -60,18 +58,45 @@ define float @fadd(ptr noalias nocapture readonly %a, i64 %n) { ; ; NO-VP-LABEL: @fadd( ; NO-VP-NEXT: entry: +; NO-VP-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 4 +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N1:%.*]], [[TMP1]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] +; NO-VP: vector.ph: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N1]], [[TMP3]] +; NO-VP-NEXT: [[N:%.*]] = sub i64 [[N1]], [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4 ; NO-VP-NEXT: br label [[FOR_BODY:%.*]] -; NO-VP: for.body: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ] +; NO-VP: vector.body: +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ] ; NO-VP-NEXT: [[SUM_07:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[ADD:%.*]], [[FOR_BODY]] ] ; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[IV]] -; NO-VP-NEXT: [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4 -; NO-VP-NEXT: [[ADD]] = fadd float [[TMP0]], [[SUM_07]] -; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 -; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N:%.*]] -; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[ARRAYIDX]], align 4 +; NO-VP-NEXT: [[ADD]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[SUM_07]], <vscale x 4 x float> [[WIDE_LOAD]]) +; NO-VP-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], [[TMP5]] +; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] +; NO-VP: middle.block: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N1]], [[N]] +; NO-VP-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]] +; NO-VP: scalar.ph: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ] +; NO-VP-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[ADD]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY1]] ] +; NO-VP-NEXT: br label [[FOR_BODY1:%.*]] +; NO-VP: for.body: +; NO-VP-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ] +; NO-VP-NEXT: [[SUM_7:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[ADD1:%.*]], [[FOR_BODY1]] ] +; NO-VP-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV1]] +; NO-VP-NEXT: [[TMP9:%.*]] = load float, ptr [[ARRAYIDX1]], align 4 +; NO-VP-NEXT: [[ADD1]] = fadd float [[TMP9]], [[SUM_7]] +; NO-VP-NEXT: [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1 +; NO-VP-NEXT: [[EXITCOND_NOT1:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N1]] +; NO-VP-NEXT: br i1 [[EXITCOND_NOT1]], label [[FOR_END]], label [[FOR_BODY1]], !llvm.loop [[LOOP3:![0-9]+]] ; NO-VP: for.end: -; NO-VP-NEXT: [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ] +; NO-VP-NEXT: [[ADD_LCSSA:%.*]] = phi float [ [[ADD1]], [[FOR_BODY1]] ], [ [[ADD]], [[MIDDLE_BLOCK]] ] ; NO-VP-NEXT: ret float [[ADD_LCSSA]] ; entry: diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-reduction-cost.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-reduction-cost.ll index 15f9993..10d83f4 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-reduction-cost.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-reduction-cost.ll @@ -1,7 +1,6 @@ ; REQUIRES: asserts ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize --disable-output \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s 2>&1 | FileCheck %s ; CHECK: Cost of 2 for VF vscale x 4: WIDEN-INTRINSIC vp<%{{.+}}> = call llvm.vp.merge(ir<true>, ir<%add>, ir<%rdx>, vp<%{{.+}}>) diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-reduction.ll index 73fd389..3a963b0 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-reduction.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-reduction.ll @@ -1,12 +1,10 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v,+f -S < %s| FileCheck %s --check-prefix=IF-EVL ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v,+f -S < %s| FileCheck %s --check-prefix=NO-VP define i32 @add(ptr %a, i64 %n, i32 %start) { @@ -125,36 +123,34 @@ for.end: define i32 @mul(ptr %a, i64 %n, i32 %start) { ; IF-EVL-LABEL: @mul( ; IF-EVL-NEXT: entry: -; IF-EVL-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] +; IF-EVL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N_RND_UP:%.*]], 16 +; IF-EVL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] ; IF-EVL: vector.ph: -; IF-EVL-NEXT: [[N_RND_UP:%.*]] = add i64 [[N:%.*]], 7 -; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 8 +; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 16 ; IF-EVL-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]] -; IF-EVL-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1 ; IF-EVL-NEXT: [[TMP9:%.*]] = insertelement <8 x i32> splat (i32 1), i32 [[START:%.*]], i32 0 -; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <8 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0 -; IF-EVL-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT1]], <8 x i64> poison, <8 x i32> zeroinitializer ; IF-EVL-NEXT: br label [[FOR_BODY:%.*]] ; IF-EVL: vector.body: ; IF-EVL-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ] ; IF-EVL-NEXT: [[VEC_PHI:%.*]] = phi <8 x i32> [ [[TMP9]], [[ENTRY]] ], [ [[TMP5:%.*]], [[FOR_BODY]] ] -; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <8 x i64> poison, i64 [[IV]], i64 0 -; IF-EVL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT]], <8 x i64> poison, <8 x i32> zeroinitializer -; IF-EVL-NEXT: [[VEC_IV:%.*]] = add <8 x i64> [[BROADCAST_SPLAT]], <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7> -; IF-EVL-NEXT: [[TMP2:%.*]] = icmp ule <8 x i64> [[VEC_IV]], [[BROADCAST_SPLAT2]] +; IF-EVL-NEXT: [[VEC_PHI1:%.*]] = phi <8 x i32> [ splat (i32 1), [[ENTRY]] ], [ [[TMP4:%.*]], [[FOR_BODY]] ] ; IF-EVL-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IV]] -; IF-EVL-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP3]], i32 4, <8 x i1> [[TMP2]], <8 x i32> poison) +; IF-EVL-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 8 +; IF-EVL-NEXT: [[WIDE_MASKED_LOAD:%.*]] = load <8 x i32>, ptr [[TMP3]], align 4 +; IF-EVL-NEXT: [[WIDE_LOAD2:%.*]] = load <8 x i32>, ptr [[TMP2]], align 4 ; IF-EVL-NEXT: [[TMP5]] = mul <8 x i32> [[WIDE_MASKED_LOAD]], [[VEC_PHI]] -; IF-EVL-NEXT: [[TMP6:%.*]] = select <8 x i1> [[TMP2]], <8 x i32> [[TMP5]], <8 x i32> [[VEC_PHI]] -; IF-EVL-NEXT: [[IV_NEXT]] = add i64 [[IV]], 8 +; IF-EVL-NEXT: [[TMP4]] = mul <8 x i32> [[WIDE_LOAD2]], [[VEC_PHI1]] +; IF-EVL-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], 16 ; IF-EVL-NEXT: [[TMP7:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]] ; IF-EVL-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]] ; IF-EVL: middle.block: +; IF-EVL-NEXT: [[TMP6:%.*]] = mul <8 x i32> [[TMP4]], [[TMP5]] ; IF-EVL-NEXT: [[TMP8:%.*]] = call i32 @llvm.vector.reduce.mul.v8i32(<8 x i32> [[TMP6]]) -; IF-EVL-NEXT: br label [[FOR_END:%.*]] +; IF-EVL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_RND_UP]], [[N_VEC]] +; IF-EVL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]] ; IF-EVL: scalar.ph: -; IF-EVL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY1:%.*]] ] -; IF-EVL-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[START]], [[ENTRY1]] ] +; IF-EVL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ] +; IF-EVL-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP8]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY1]] ] ; IF-EVL-NEXT: br label [[FOR_BODY1:%.*]] ; IF-EVL: for.body: ; IF-EVL-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ] @@ -163,7 +159,7 @@ define i32 @mul(ptr %a, i64 %n, i32 %start) { ; IF-EVL-NEXT: [[TMP0:%.*]] = load i32, ptr [[ARRAYIDX]], align 4 ; IF-EVL-NEXT: [[MUL]] = mul nsw i32 [[TMP0]], [[RDX]] ; IF-EVL-NEXT: [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1 -; IF-EVL-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N]] +; IF-EVL-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N_RND_UP]] ; IF-EVL-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY1]], !llvm.loop [[LOOP6:![0-9]+]] ; IF-EVL: for.end: ; IF-EVL-NEXT: [[MUL_LCSSA:%.*]] = phi i32 [ [[MUL]], [[FOR_BODY1]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ] @@ -1158,36 +1154,34 @@ for.end: define float @fmul(ptr %a, i64 %n, float %start) { ; IF-EVL-LABEL: @fmul( ; IF-EVL-NEXT: entry: -; IF-EVL-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] +; IF-EVL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N_RND_UP:%.*]], 16 +; IF-EVL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] ; IF-EVL: vector.ph: -; IF-EVL-NEXT: [[N_RND_UP:%.*]] = add i64 [[N:%.*]], 7 -; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 8 +; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 16 ; IF-EVL-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]] -; IF-EVL-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1 ; IF-EVL-NEXT: [[TMP9:%.*]] = insertelement <8 x float> splat (float 1.000000e+00), float [[START:%.*]], i32 0 -; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <8 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0 -; IF-EVL-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT1]], <8 x i64> poison, <8 x i32> zeroinitializer ; IF-EVL-NEXT: br label [[FOR_BODY:%.*]] ; IF-EVL: vector.body: ; IF-EVL-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ] ; IF-EVL-NEXT: [[VEC_PHI:%.*]] = phi <8 x float> [ [[TMP9]], [[ENTRY]] ], [ [[TMP5:%.*]], [[FOR_BODY]] ] -; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <8 x i64> poison, i64 [[IV]], i64 0 -; IF-EVL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT]], <8 x i64> poison, <8 x i32> zeroinitializer -; IF-EVL-NEXT: [[VEC_IV:%.*]] = add <8 x i64> [[BROADCAST_SPLAT]], <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7> -; IF-EVL-NEXT: [[TMP2:%.*]] = icmp ule <8 x i64> [[VEC_IV]], [[BROADCAST_SPLAT2]] +; IF-EVL-NEXT: [[VEC_PHI1:%.*]] = phi <8 x float> [ splat (float 1.000000e+00), [[ENTRY]] ], [ [[TMP4:%.*]], [[FOR_BODY]] ] ; IF-EVL-NEXT: [[TMP3:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[IV]] -; IF-EVL-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP3]], i32 4, <8 x i1> [[TMP2]], <8 x float> poison) +; IF-EVL-NEXT: [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP3]], i32 8 +; IF-EVL-NEXT: [[WIDE_MASKED_LOAD:%.*]] = load <8 x float>, ptr [[TMP3]], align 4 +; IF-EVL-NEXT: [[WIDE_LOAD2:%.*]] = load <8 x float>, ptr [[TMP2]], align 4 ; IF-EVL-NEXT: [[TMP5]] = fmul reassoc <8 x float> [[WIDE_MASKED_LOAD]], [[VEC_PHI]] -; IF-EVL-NEXT: [[TMP6:%.*]] = select reassoc <8 x i1> [[TMP2]], <8 x float> [[TMP5]], <8 x float> [[VEC_PHI]] -; IF-EVL-NEXT: [[IV_NEXT]] = add i64 [[IV]], 8 +; IF-EVL-NEXT: [[TMP4]] = fmul reassoc <8 x float> [[WIDE_LOAD2]], [[VEC_PHI1]] +; IF-EVL-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], 16 ; IF-EVL-NEXT: [[TMP7:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]] ; IF-EVL-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]] ; IF-EVL: middle.block: +; IF-EVL-NEXT: [[TMP6:%.*]] = fmul reassoc <8 x float> [[TMP4]], [[TMP5]] ; IF-EVL-NEXT: [[TMP8:%.*]] = call reassoc float @llvm.vector.reduce.fmul.v8f32(float 1.000000e+00, <8 x float> [[TMP6]]) -; IF-EVL-NEXT: br label [[FOR_END:%.*]] +; IF-EVL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_RND_UP]], [[N_VEC]] +; IF-EVL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]] ; IF-EVL: scalar.ph: -; IF-EVL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY1:%.*]] ] -; IF-EVL-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[START]], [[ENTRY1]] ] +; IF-EVL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ] +; IF-EVL-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP8]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY1]] ] ; IF-EVL-NEXT: br label [[FOR_BODY1:%.*]] ; IF-EVL: for.body: ; IF-EVL-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ] @@ -1196,7 +1190,7 @@ define float @fmul(ptr %a, i64 %n, float %start) { ; IF-EVL-NEXT: [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4 ; IF-EVL-NEXT: [[MUL]] = fmul reassoc float [[TMP0]], [[RDX]] ; IF-EVL-NEXT: [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1 -; IF-EVL-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N]] +; IF-EVL-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N_RND_UP]] ; IF-EVL-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY1]], !llvm.loop [[LOOP24:![0-9]+]] ; IF-EVL: for.end: ; IF-EVL-NEXT: [[MUL_LCSSA:%.*]] = phi float [ [[MUL]], [[FOR_BODY1]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ] @@ -1504,37 +1498,35 @@ for.end: define float @fminimum(ptr %a, i64 %n, float %start) { ; IF-EVL-LABEL: @fminimum( ; IF-EVL-NEXT: entry: -; IF-EVL-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] +; IF-EVL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N_RND_UP:%.*]], 16 +; IF-EVL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] ; IF-EVL: vector.ph: -; IF-EVL-NEXT: [[N_RND_UP:%.*]] = add i64 [[N:%.*]], 7 -; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 8 +; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 16 ; IF-EVL-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]] -; IF-EVL-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1 -; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <8 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0 -; IF-EVL-NEXT: [[BROADCAST_SPLAT4:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT3]], <8 x i64> poison, <8 x i32> zeroinitializer ; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <8 x float> poison, float [[START:%.*]], i64 0 ; IF-EVL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x float> [[BROADCAST_SPLATINSERT2]], <8 x float> poison, <8 x i32> zeroinitializer ; IF-EVL-NEXT: br label [[FOR_BODY:%.*]] ; IF-EVL: vector.body: ; IF-EVL-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ] ; IF-EVL-NEXT: [[VEC_PHI:%.*]] = phi <8 x float> [ [[BROADCAST_SPLAT]], [[ENTRY]] ], [ [[TMP4:%.*]], [[FOR_BODY]] ] -; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <8 x i64> poison, i64 [[IV]], i64 0 -; IF-EVL-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT1]], <8 x i64> poison, <8 x i32> zeroinitializer -; IF-EVL-NEXT: [[VEC_IV:%.*]] = add <8 x i64> [[BROADCAST_SPLAT2]], <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7> -; IF-EVL-NEXT: [[TMP1:%.*]] = icmp ule <8 x i64> [[VEC_IV]], [[BROADCAST_SPLAT4]] +; IF-EVL-NEXT: [[VEC_PHI1:%.*]] = phi <8 x float> [ [[BROADCAST_SPLAT]], [[ENTRY]] ], [ [[TMP3:%.*]], [[FOR_BODY]] ] ; IF-EVL-NEXT: [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[IV]] -; IF-EVL-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP2]], i32 4, <8 x i1> [[TMP1]], <8 x float> poison) +; IF-EVL-NEXT: [[TMP1:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 8 +; IF-EVL-NEXT: [[WIDE_MASKED_LOAD:%.*]] = load <8 x float>, ptr [[TMP2]], align 4 +; IF-EVL-NEXT: [[WIDE_LOAD2:%.*]] = load <8 x float>, ptr [[TMP1]], align 4 ; IF-EVL-NEXT: [[TMP4]] = call <8 x float> @llvm.minimum.v8f32(<8 x float> [[VEC_PHI]], <8 x float> [[WIDE_MASKED_LOAD]]) -; IF-EVL-NEXT: [[TMP5:%.*]] = select <8 x i1> [[TMP1]], <8 x float> [[TMP4]], <8 x float> [[VEC_PHI]] -; IF-EVL-NEXT: [[IV_NEXT]] = add i64 [[IV]], 8 +; IF-EVL-NEXT: [[TMP3]] = call <8 x float> @llvm.minimum.v8f32(<8 x float> [[VEC_PHI1]], <8 x float> [[WIDE_LOAD2]]) +; IF-EVL-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], 16 ; IF-EVL-NEXT: [[TMP6:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]] ; IF-EVL-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP29:![0-9]+]] ; IF-EVL: middle.block: +; IF-EVL-NEXT: [[TMP5:%.*]] = call <8 x float> @llvm.minimum.v8f32(<8 x float> [[TMP4]], <8 x float> [[TMP3]]) ; IF-EVL-NEXT: [[TMP7:%.*]] = call float @llvm.vector.reduce.fminimum.v8f32(<8 x float> [[TMP5]]) -; IF-EVL-NEXT: br label [[FOR_END:%.*]] +; IF-EVL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_RND_UP]], [[N_VEC]] +; IF-EVL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]] ; IF-EVL: scalar.ph: -; IF-EVL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY1:%.*]] ] -; IF-EVL-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[START]], [[ENTRY1]] ] +; IF-EVL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ] +; IF-EVL-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY1]] ] ; IF-EVL-NEXT: br label [[FOR_BODY1:%.*]] ; IF-EVL: for.body: ; IF-EVL-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ] @@ -1543,7 +1535,7 @@ define float @fminimum(ptr %a, i64 %n, float %start) { ; IF-EVL-NEXT: [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4 ; IF-EVL-NEXT: [[MIN]] = tail call float @llvm.minimum.f32(float [[RDX]], float [[TMP0]]) ; IF-EVL-NEXT: [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1 -; IF-EVL-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N]] +; IF-EVL-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N_RND_UP]] ; IF-EVL-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY1]], !llvm.loop [[LOOP30:![0-9]+]] ; IF-EVL: for.end: ; IF-EVL-NEXT: [[MIN_LCSSA:%.*]] = phi float [ [[MIN]], [[FOR_BODY1]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ] @@ -1614,37 +1606,35 @@ for.end: define float @fmaximum(ptr %a, i64 %n, float %start) { ; IF-EVL-LABEL: @fmaximum( ; IF-EVL-NEXT: entry: -; IF-EVL-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] +; IF-EVL-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N_RND_UP:%.*]], 16 +; IF-EVL-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] ; IF-EVL: vector.ph: -; IF-EVL-NEXT: [[N_RND_UP:%.*]] = add i64 [[N:%.*]], 7 -; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 8 +; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], 16 ; IF-EVL-NEXT: [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]] -; IF-EVL-NEXT: [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1 -; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <8 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0 -; IF-EVL-NEXT: [[BROADCAST_SPLAT4:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT3]], <8 x i64> poison, <8 x i32> zeroinitializer ; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <8 x float> poison, float [[START:%.*]], i64 0 ; IF-EVL-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x float> [[BROADCAST_SPLATINSERT2]], <8 x float> poison, <8 x i32> zeroinitializer ; IF-EVL-NEXT: br label [[FOR_BODY:%.*]] ; IF-EVL: vector.body: ; IF-EVL-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ] ; IF-EVL-NEXT: [[VEC_PHI:%.*]] = phi <8 x float> [ [[BROADCAST_SPLAT]], [[ENTRY]] ], [ [[TMP4:%.*]], [[FOR_BODY]] ] -; IF-EVL-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <8 x i64> poison, i64 [[IV]], i64 0 -; IF-EVL-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT1]], <8 x i64> poison, <8 x i32> zeroinitializer -; IF-EVL-NEXT: [[VEC_IV:%.*]] = add <8 x i64> [[BROADCAST_SPLAT2]], <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7> -; IF-EVL-NEXT: [[TMP1:%.*]] = icmp ule <8 x i64> [[VEC_IV]], [[BROADCAST_SPLAT4]] +; IF-EVL-NEXT: [[VEC_PHI1:%.*]] = phi <8 x float> [ [[BROADCAST_SPLAT]], [[ENTRY]] ], [ [[TMP3:%.*]], [[FOR_BODY]] ] ; IF-EVL-NEXT: [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[IV]] -; IF-EVL-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP2]], i32 4, <8 x i1> [[TMP1]], <8 x float> poison) +; IF-EVL-NEXT: [[TMP1:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 8 +; IF-EVL-NEXT: [[WIDE_MASKED_LOAD:%.*]] = load <8 x float>, ptr [[TMP2]], align 4 +; IF-EVL-NEXT: [[WIDE_LOAD2:%.*]] = load <8 x float>, ptr [[TMP1]], align 4 ; IF-EVL-NEXT: [[TMP4]] = call <8 x float> @llvm.maximum.v8f32(<8 x float> [[VEC_PHI]], <8 x float> [[WIDE_MASKED_LOAD]]) -; IF-EVL-NEXT: [[TMP5:%.*]] = select <8 x i1> [[TMP1]], <8 x float> [[TMP4]], <8 x float> [[VEC_PHI]] -; IF-EVL-NEXT: [[IV_NEXT]] = add i64 [[IV]], 8 +; IF-EVL-NEXT: [[TMP3]] = call <8 x float> @llvm.maximum.v8f32(<8 x float> [[VEC_PHI1]], <8 x float> [[WIDE_LOAD2]]) +; IF-EVL-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], 16 ; IF-EVL-NEXT: [[TMP6:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]] ; IF-EVL-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP31:![0-9]+]] ; IF-EVL: middle.block: +; IF-EVL-NEXT: [[TMP5:%.*]] = call <8 x float> @llvm.maximum.v8f32(<8 x float> [[TMP4]], <8 x float> [[TMP3]]) ; IF-EVL-NEXT: [[TMP7:%.*]] = call float @llvm.vector.reduce.fmaximum.v8f32(<8 x float> [[TMP5]]) -; IF-EVL-NEXT: br label [[FOR_END:%.*]] +; IF-EVL-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_RND_UP]], [[N_VEC]] +; IF-EVL-NEXT: br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]] ; IF-EVL: scalar.ph: -; IF-EVL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY1:%.*]] ] -; IF-EVL-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[START]], [[ENTRY1]] ] +; IF-EVL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ] +; IF-EVL-NEXT: [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY1]] ] ; IF-EVL-NEXT: br label [[FOR_BODY1:%.*]] ; IF-EVL: for.body: ; IF-EVL-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ] @@ -1653,7 +1643,7 @@ define float @fmaximum(ptr %a, i64 %n, float %start) { ; IF-EVL-NEXT: [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4 ; IF-EVL-NEXT: [[MAX]] = tail call float @llvm.maximum.f32(float [[RDX]], float [[TMP0]]) ; IF-EVL-NEXT: [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1 -; IF-EVL-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N]] +; IF-EVL-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N_RND_UP]] ; IF-EVL-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY1]], !llvm.loop [[LOOP32:![0-9]+]] ; IF-EVL: for.end: ; IF-EVL-NEXT: [[MAX_LCSSA:%.*]] = phi float [ [[MAX]], [[FOR_BODY1]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ] diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-reverse-load-store.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-reverse-load-store.ll index 01033c8..62eef6c 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-reverse-load-store.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-reverse-load-store.ll @@ -1,12 +1,10 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefix=IF-EVL ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefix=NO-VP define void @reverse_load_store(i64 %startval, ptr noalias %ptr, ptr noalias %ptr2) { @@ -75,18 +73,61 @@ define void @reverse_load_store(i64 %startval, ptr noalias %ptr, ptr noalias %pt ; ; NO-VP-LABEL: @reverse_load_store( ; NO-VP-NEXT: entry: +; NO-VP-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 4 +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP1]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] +; NO-VP: vector.ph: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[STARTVAL:%.*]], [[N_VEC]] +; NO-VP-NEXT: [[TMP7:%.*]] = trunc i64 [[N_VEC]] to i32 ; NO-VP-NEXT: br label [[FOR_BODY:%.*]] +; NO-VP: vector.body: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[FOR_BODY]] ] +; NO-VP-NEXT: [[OFFSET_IDX:%.*]] = sub i64 [[STARTVAL]], [[INDEX]] +; NO-VP-NEXT: [[TMP8:%.*]] = add i64 [[OFFSET_IDX]], -1 +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[PTR:%.*]], i64 [[TMP8]] +; NO-VP-NEXT: [[TMP10:%.*]] = mul i64 0, [[TMP5]] +; NO-VP-NEXT: [[TMP11:%.*]] = sub i64 [[TMP5]], 1 +; NO-VP-NEXT: [[TMP12:%.*]] = mul i64 -1, [[TMP11]] +; NO-VP-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i64 [[TMP10]] +; NO-VP-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i64 [[TMP12]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP14]], align 4 +; NO-VP-NEXT: [[REVERSE:%.*]] = call <vscale x 4 x i32> @llvm.vector.reverse.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]]) +; NO-VP-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[PTR2:%.*]], i64 [[TMP8]] +; NO-VP-NEXT: [[TMP16:%.*]] = mul i64 0, [[TMP5]] +; NO-VP-NEXT: [[TMP17:%.*]] = sub i64 [[TMP5]], 1 +; NO-VP-NEXT: [[TMP18:%.*]] = mul i64 -1, [[TMP17]] +; NO-VP-NEXT: [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[TMP15]], i64 [[TMP16]] +; NO-VP-NEXT: [[TMP20:%.*]] = getelementptr inbounds i32, ptr [[TMP19]], i64 [[TMP18]] +; NO-VP-NEXT: [[REVERSE1:%.*]] = call <vscale x 4 x i32> @llvm.vector.reverse.nxv4i32(<vscale x 4 x i32> [[REVERSE]]) +; NO-VP-NEXT: store <vscale x 4 x i32> [[REVERSE1]], ptr [[TMP20]], align 4 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] +; NO-VP-NEXT: [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] +; NO-VP: middle.block: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label [[LOOPEND:%.*]], label [[SCALAR_PH]] +; NO-VP: scalar.ph: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[TMP6]], [[MIDDLE_BLOCK]] ], [ [[STARTVAL]], [[ENTRY:%.*]] ] +; NO-VP-NEXT: [[BC_RESUME_VAL2:%.*]] = phi i32 [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ] +; NO-VP-NEXT: br label [[FOR_BODY1:%.*]] ; NO-VP: for.body: -; NO-VP-NEXT: [[ADD_PHI:%.*]] = phi i64 [ [[STARTVAL:%.*]], [[ENTRY:%.*]] ], [ [[ADD:%.*]], [[FOR_BODY]] ] -; NO-VP-NEXT: [[I:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[INC:%.*]], [[FOR_BODY]] ] +; NO-VP-NEXT: [[ADD_PHI:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[FOR_BODY1]] ] +; NO-VP-NEXT: [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY1]] ] ; NO-VP-NEXT: [[ADD]] = add i64 [[ADD_PHI]], -1 -; NO-VP-NEXT: [[GEPL:%.*]] = getelementptr inbounds i32, ptr [[PTR:%.*]], i64 [[ADD]] +; NO-VP-NEXT: [[GEPL:%.*]] = getelementptr inbounds i32, ptr [[PTR]], i64 [[ADD]] ; NO-VP-NEXT: [[TMP:%.*]] = load i32, ptr [[GEPL]], align 4 -; NO-VP-NEXT: [[GEPS:%.*]] = getelementptr inbounds i32, ptr [[PTR2:%.*]], i64 [[ADD]] +; NO-VP-NEXT: [[GEPS:%.*]] = getelementptr inbounds i32, ptr [[PTR2]], i64 [[ADD]] ; NO-VP-NEXT: store i32 [[TMP]], ptr [[GEPS]], align 4 ; NO-VP-NEXT: [[INC]] = add i32 [[I]], 1 ; NO-VP-NEXT: [[EXITCOND:%.*]] = icmp ne i32 [[INC]], 1024 -; NO-VP-NEXT: br i1 [[EXITCOND]], label [[FOR_BODY]], label [[LOOPEND:%.*]] +; NO-VP-NEXT: br i1 [[EXITCOND]], label [[FOR_BODY1]], label [[LOOPEND]], !llvm.loop [[LOOP3:![0-9]+]] ; NO-VP: loopend: ; NO-VP-NEXT: ret void ; @@ -188,25 +229,74 @@ define void @reverse_load_store_masked(i64 %startval, ptr noalias %ptr, ptr noal ; ; NO-VP-LABEL: @reverse_load_store_masked( ; NO-VP-NEXT: entry: +; NO-VP-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 4 +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP1]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[ENTRY:%.*]], label [[VECTOR_PH:%.*]] +; NO-VP: vector.ph: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 4 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4 +; NO-VP-NEXT: [[TMP6:%.*]] = sub i64 [[STARTVAL1:%.*]], [[N_VEC]] +; NO-VP-NEXT: [[TMP7:%.*]] = trunc i64 [[N_VEC]] to i32 ; NO-VP-NEXT: br label [[FOR_BODY:%.*]] +; NO-VP: vector.body: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[FOR_BODY]] ] +; NO-VP-NEXT: [[OFFSET_IDX:%.*]] = sub i64 [[STARTVAL1]], [[INDEX]] +; NO-VP-NEXT: [[OFFSET_IDX1:%.*]] = trunc i64 [[INDEX]] to i32 +; NO-VP-NEXT: [[TMP8:%.*]] = add i64 [[OFFSET_IDX]], -1 +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[PTR:%.*]], i32 [[OFFSET_IDX1]] +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP9]], align 4 +; NO-VP-NEXT: [[TMP10:%.*]] = icmp slt <vscale x 4 x i32> [[WIDE_LOAD]], splat (i32 100) +; NO-VP-NEXT: [[TMP11:%.*]] = getelementptr i32, ptr [[PTR1:%.*]], i64 [[TMP8]] +; NO-VP-NEXT: [[TMP12:%.*]] = mul i64 0, [[TMP5]] +; NO-VP-NEXT: [[TMP13:%.*]] = sub i64 [[TMP5]], 1 +; NO-VP-NEXT: [[TMP14:%.*]] = mul i64 -1, [[TMP13]] +; NO-VP-NEXT: [[TMP15:%.*]] = getelementptr i32, ptr [[TMP11]], i64 [[TMP12]] +; NO-VP-NEXT: [[TMP16:%.*]] = getelementptr i32, ptr [[TMP15]], i64 [[TMP14]] +; NO-VP-NEXT: [[REVERSE:%.*]] = call <vscale x 4 x i1> @llvm.vector.reverse.nxv4i1(<vscale x 4 x i1> [[TMP10]]) +; NO-VP-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP16]], i32 4, <vscale x 4 x i1> [[REVERSE]], <vscale x 4 x i32> poison) +; NO-VP-NEXT: [[REVERSE2:%.*]] = call <vscale x 4 x i32> @llvm.vector.reverse.nxv4i32(<vscale x 4 x i32> [[WIDE_MASKED_LOAD]]) +; NO-VP-NEXT: [[TMP17:%.*]] = getelementptr i32, ptr [[PTR2:%.*]], i64 [[TMP8]] +; NO-VP-NEXT: [[TMP18:%.*]] = mul i64 0, [[TMP5]] +; NO-VP-NEXT: [[TMP19:%.*]] = sub i64 [[TMP5]], 1 +; NO-VP-NEXT: [[TMP20:%.*]] = mul i64 -1, [[TMP19]] +; NO-VP-NEXT: [[TMP21:%.*]] = getelementptr i32, ptr [[TMP17]], i64 [[TMP18]] +; NO-VP-NEXT: [[TMP22:%.*]] = getelementptr i32, ptr [[TMP21]], i64 [[TMP20]] +; NO-VP-NEXT: [[REVERSE3:%.*]] = call <vscale x 4 x i1> @llvm.vector.reverse.nxv4i1(<vscale x 4 x i1> [[TMP10]]) +; NO-VP-NEXT: [[REVERSE4:%.*]] = call <vscale x 4 x i32> @llvm.vector.reverse.nxv4i32(<vscale x 4 x i32> [[REVERSE2]]) +; NO-VP-NEXT: call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[REVERSE4]], ptr [[TMP22]], i32 4, <vscale x 4 x i1> [[REVERSE3]]) +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] +; NO-VP-NEXT: [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP23]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]] +; NO-VP: middle.block: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label [[LOOPEND:%.*]], label [[ENTRY]] +; NO-VP: scalar.ph: +; NO-VP-NEXT: [[STARTVAL:%.*]] = phi i64 [ [[TMP6]], [[MIDDLE_BLOCK]] ], [ [[STARTVAL1]], [[ENTRY1:%.*]] ] +; NO-VP-NEXT: [[BC_RESUME_VAL5:%.*]] = phi i32 [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1]] ] +; NO-VP-NEXT: br label [[FOR_BODY1:%.*]] ; NO-VP: for.body: -; NO-VP-NEXT: [[ADD_PHI:%.*]] = phi i64 [ [[STARTVAL:%.*]], [[ENTRY:%.*]] ], [ [[ADD:%.*]], [[FOR_INC:%.*]] ] -; NO-VP-NEXT: [[I:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[INC:%.*]], [[FOR_INC]] ] +; NO-VP-NEXT: [[ADD_PHI:%.*]] = phi i64 [ [[STARTVAL]], [[ENTRY]] ], [ [[ADD:%.*]], [[FOR_INC:%.*]] ] +; NO-VP-NEXT: [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL5]], [[ENTRY]] ], [ [[INC:%.*]], [[FOR_INC]] ] ; NO-VP-NEXT: [[ADD]] = add i64 [[ADD_PHI]], -1 -; NO-VP-NEXT: [[GEPL:%.*]] = getelementptr inbounds i32, ptr [[PTR:%.*]], i32 [[I]] +; NO-VP-NEXT: [[GEPL:%.*]] = getelementptr inbounds i32, ptr [[PTR]], i32 [[I]] ; NO-VP-NEXT: [[TMP:%.*]] = load i32, ptr [[GEPL]], align 4 ; NO-VP-NEXT: [[CMP1:%.*]] = icmp slt i32 [[TMP]], 100 ; NO-VP-NEXT: br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[FOR_INC]] ; NO-VP: if.then: -; NO-VP-NEXT: [[GEPL1:%.*]] = getelementptr inbounds i32, ptr [[PTR1:%.*]], i64 [[ADD]] +; NO-VP-NEXT: [[GEPL1:%.*]] = getelementptr inbounds i32, ptr [[PTR1]], i64 [[ADD]] ; NO-VP-NEXT: [[V:%.*]] = load i32, ptr [[GEPL1]], align 4 -; NO-VP-NEXT: [[GEPS:%.*]] = getelementptr inbounds i32, ptr [[PTR2:%.*]], i64 [[ADD]] +; NO-VP-NEXT: [[GEPS:%.*]] = getelementptr inbounds i32, ptr [[PTR2]], i64 [[ADD]] ; NO-VP-NEXT: store i32 [[V]], ptr [[GEPS]], align 4 ; NO-VP-NEXT: br label [[FOR_INC]] ; NO-VP: for.inc: ; NO-VP-NEXT: [[INC]] = add i32 [[I]], 1 ; NO-VP-NEXT: [[EXITCOND:%.*]] = icmp ne i32 [[INC]], 1024 -; NO-VP-NEXT: br i1 [[EXITCOND]], label [[FOR_BODY]], label [[LOOPEND:%.*]] +; NO-VP-NEXT: br i1 [[EXITCOND]], label [[FOR_BODY1]], label [[LOOPEND]], !llvm.loop [[LOOP5:![0-9]+]] ; NO-VP: loopend: ; NO-VP-NEXT: ret void ; @@ -318,20 +408,49 @@ define void @multiple_reverse_vector_pointer(ptr noalias %a, ptr noalias %b, ptr ; ; NO-VP-LABEL: @multiple_reverse_vector_pointer( ; NO-VP-NEXT: entry: +; NO-VP-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] +; NO-VP: vector.ph: ; NO-VP-NEXT: br label [[LOOP:%.*]] +; NO-VP: vector.body: +; NO-VP-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[LOOP]] ] +; NO-VP-NEXT: [[OFFSET_IDX:%.*]] = sub i64 1024, [[INDEX]] +; NO-VP-NEXT: [[TMP0:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[OFFSET_IDX]] +; NO-VP-NEXT: [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0 +; NO-VP-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 -15 +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1 +; NO-VP-NEXT: [[REVERSE:%.*]] = shufflevector <16 x i8> [[WIDE_LOAD]], <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0> +; NO-VP-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[B:%.*]], <16 x i8> [[REVERSE]] +; NO-VP-NEXT: [[WIDE_MASKED_GATHER:%.*]] = call <16 x i8> @llvm.masked.gather.v16i8.v16p0(<16 x ptr> [[TMP3]], i32 1, <16 x i1> splat (i1 true), <16 x i8> poison) +; NO-VP-NEXT: [[TMP4:%.*]] = getelementptr i8, ptr [[C:%.*]], i64 [[OFFSET_IDX]] +; NO-VP-NEXT: [[TMP5:%.*]] = getelementptr i8, ptr [[TMP4]], i32 0 +; NO-VP-NEXT: [[TMP6:%.*]] = getelementptr i8, ptr [[TMP5]], i32 -15 +; NO-VP-NEXT: [[REVERSE1:%.*]] = shufflevector <16 x i8> [[WIDE_MASKED_GATHER]], <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0> +; NO-VP-NEXT: store <16 x i8> [[REVERSE1]], ptr [[TMP6]], align 1 +; NO-VP-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[D:%.*]], i64 [[OFFSET_IDX]] +; NO-VP-NEXT: [[TMP8:%.*]] = getelementptr i8, ptr [[TMP7]], i32 0 +; NO-VP-NEXT: [[TMP9:%.*]] = getelementptr i8, ptr [[TMP8]], i32 -15 +; NO-VP-NEXT: store <16 x i8> [[REVERSE1]], ptr [[TMP9]], align 1 +; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16 +; NO-VP-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024 +; NO-VP-NEXT: br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP]], !llvm.loop [[LOOP6:![0-9]+]] +; NO-VP: middle.block: +; NO-VP-NEXT: br label [[SCALAR_PH]] +; NO-VP: scalar.ph: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[MIDDLE_BLOCK]] ], [ 1024, [[ENTRY:%.*]] ] +; NO-VP-NEXT: br label [[LOOP1:%.*]] ; NO-VP: loop: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 1024, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ] -; NO-VP-NEXT: [[GEP_A:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[IV]] +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP1]] ] +; NO-VP-NEXT: [[GEP_A:%.*]] = getelementptr i8, ptr [[A]], i64 [[IV]] ; NO-VP-NEXT: [[X:%.*]] = load i8, ptr [[GEP_A]], align 1 -; NO-VP-NEXT: [[GEP_B:%.*]] = getelementptr i8, ptr [[B:%.*]], i8 [[X]] +; NO-VP-NEXT: [[GEP_B:%.*]] = getelementptr i8, ptr [[B]], i8 [[X]] ; NO-VP-NEXT: [[Y:%.*]] = load i8, ptr [[GEP_B]], align 1 -; NO-VP-NEXT: [[GEP_C:%.*]] = getelementptr i8, ptr [[C:%.*]], i64 [[IV]] +; NO-VP-NEXT: [[GEP_C:%.*]] = getelementptr i8, ptr [[C]], i64 [[IV]] ; NO-VP-NEXT: store i8 [[Y]], ptr [[GEP_C]], align 1 -; NO-VP-NEXT: [[GEP_D:%.*]] = getelementptr i8, ptr [[D:%.*]], i64 [[IV]] +; NO-VP-NEXT: [[GEP_D:%.*]] = getelementptr i8, ptr [[D]], i64 [[IV]] ; NO-VP-NEXT: store i8 [[Y]], ptr [[GEP_D]], align 1 ; NO-VP-NEXT: [[IV_NEXT]] = add i64 [[IV]], -1 ; NO-VP-NEXT: [[CMP_NOT:%.*]] = icmp eq i64 [[IV]], 0 -; NO-VP-NEXT: br i1 [[CMP_NOT]], label [[EXIT:%.*]], label [[LOOP]] +; NO-VP-NEXT: br i1 [[CMP_NOT]], label [[EXIT:%.*]], label [[LOOP1]], !llvm.loop [[LOOP7:![0-9]+]] ; NO-VP: exit: ; NO-VP-NEXT: ret void ; diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-safe-dep-distance.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-safe-dep-distance.ll index 2dd017c..a891eef 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-safe-dep-distance.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-safe-dep-distance.ll @@ -1,12 +1,10 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefix=IF-EVL ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefix=NO-VP ; Dependence distance between read and write is greater than the trip @@ -61,17 +59,44 @@ define void @test(ptr %p) { ; ; NO-VP-LABEL: @test( ; NO-VP-NEXT: entry: +; NO-VP-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 2 +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 200, [[TMP1]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] +; NO-VP: vector.ph: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 2 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 200, [[TMP3]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 200, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2 ; NO-VP-NEXT: br label [[LOOP:%.*]] -; NO-VP: loop: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ] +; NO-VP: vector.body: +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ] ; NO-VP-NEXT: [[A1:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[IV]] -; NO-VP-NEXT: [[V:%.*]] = load i64, ptr [[A1]], align 8 -; NO-VP-NEXT: [[OFFSET:%.*]] = add i64 [[IV]], 200 +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[A1]], align 8 +; NO-VP-NEXT: [[TMP7:%.*]] = add i64 [[IV]], 200 +; NO-VP-NEXT: [[TMP8:%.*]] = getelementptr i64, ptr [[P]], i64 [[TMP7]] +; NO-VP-NEXT: store <vscale x 2 x i64> [[WIDE_LOAD]], ptr [[TMP8]], align 8 +; NO-VP-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], [[TMP5]] +; NO-VP-NEXT: [[TMP9:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP]], !llvm.loop [[LOOP0:![0-9]+]] +; NO-VP: middle.block: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 200, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]] +; NO-VP: scalar.ph: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ] +; NO-VP-NEXT: br label [[LOOP1:%.*]] +; NO-VP: loop: +; NO-VP-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[LOOP1]] ] +; NO-VP-NEXT: [[A3:%.*]] = getelementptr i64, ptr [[P]], i64 [[IV1]] +; NO-VP-NEXT: [[V:%.*]] = load i64, ptr [[A3]], align 8 +; NO-VP-NEXT: [[OFFSET:%.*]] = add i64 [[IV1]], 200 ; NO-VP-NEXT: [[A2:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET]] ; NO-VP-NEXT: store i64 [[V]], ptr [[A2]], align 8 -; NO-VP-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1 -; NO-VP-NEXT: [[CMP:%.*]] = icmp ne i64 [[IV]], 199 -; NO-VP-NEXT: br i1 [[CMP]], label [[LOOP]], label [[EXIT:%.*]] +; NO-VP-NEXT: [[IV_NEXT1]] = add i64 [[IV1]], 1 +; NO-VP-NEXT: [[CMP:%.*]] = icmp ne i64 [[IV1]], 199 +; NO-VP-NEXT: br i1 [[CMP]], label [[LOOP1]], label [[EXIT]], !llvm.loop [[LOOP3:![0-9]+]] ; NO-VP: exit: ; NO-VP-NEXT: ret void ; @@ -143,7 +168,7 @@ define void @test_may_clobber1(ptr %p) { ; NO-VP-NEXT: store <4 x i64> [[WIDE_LOAD]], ptr [[TMP4]], align 32 ; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4 ; NO-VP-NEXT: [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 200 -; NO-VP-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] +; NO-VP-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]] ; NO-VP: middle.block: ; NO-VP-NEXT: br label [[EXIT:%.*]] ; NO-VP: scalar.ph: @@ -158,7 +183,7 @@ define void @test_may_clobber1(ptr %p) { ; NO-VP-NEXT: store i64 [[V]], ptr [[A2]], align 32 ; NO-VP-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1 ; NO-VP-NEXT: [[CMP:%.*]] = icmp ne i64 [[IV]], 199 -; NO-VP-NEXT: br i1 [[CMP]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP3:![0-9]+]] +; NO-VP-NEXT: br i1 [[CMP]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP5:![0-9]+]] ; NO-VP: exit: ; NO-VP-NEXT: ret void ; @@ -279,7 +304,7 @@ define void @test_may_clobber3(ptr %p) { ; NO-VP-NEXT: store <2 x i64> [[WIDE_LOAD]], ptr [[TMP4]], align 32 ; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2 ; NO-VP-NEXT: [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 200 -; NO-VP-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]] +; NO-VP-NEXT: br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]] ; NO-VP: middle.block: ; NO-VP-NEXT: br label [[EXIT:%.*]] ; NO-VP: scalar.ph: @@ -294,7 +319,7 @@ define void @test_may_clobber3(ptr %p) { ; NO-VP-NEXT: store i64 [[V]], ptr [[A2]], align 32 ; NO-VP-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1 ; NO-VP-NEXT: [[CMP:%.*]] = icmp ne i64 [[IV]], 199 -; NO-VP-NEXT: br i1 [[CMP]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP5:![0-9]+]] +; NO-VP-NEXT: br i1 [[CMP]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP7:![0-9]+]] ; NO-VP: exit: ; NO-VP-NEXT: ret void ; @@ -366,17 +391,44 @@ define void @trivial_due_max_vscale(ptr %p) { ; ; NO-VP-LABEL: @trivial_due_max_vscale( ; NO-VP-NEXT: entry: +; NO-VP-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 2 +; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 200, [[TMP1]] +; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] +; NO-VP: vector.ph: +; NO-VP-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 2 +; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i64 200, [[TMP3]] +; NO-VP-NEXT: [[N_VEC:%.*]] = sub i64 200, [[N_MOD_VF]] +; NO-VP-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; NO-VP-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2 ; NO-VP-NEXT: br label [[LOOP:%.*]] -; NO-VP: loop: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ] +; NO-VP: vector.body: +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ] ; NO-VP-NEXT: [[A1:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[IV]] -; NO-VP-NEXT: [[V:%.*]] = load i64, ptr [[A1]], align 32 -; NO-VP-NEXT: [[OFFSET:%.*]] = add i64 [[IV]], 8192 +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[A1]], align 32 +; NO-VP-NEXT: [[TMP7:%.*]] = add i64 [[IV]], 8192 +; NO-VP-NEXT: [[TMP8:%.*]] = getelementptr i64, ptr [[P]], i64 [[TMP7]] +; NO-VP-NEXT: store <vscale x 2 x i64> [[WIDE_LOAD]], ptr [[TMP8]], align 32 +; NO-VP-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], [[TMP5]] +; NO-VP-NEXT: [[TMP9:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]] +; NO-VP-NEXT: br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP]], !llvm.loop [[LOOP8:![0-9]+]] +; NO-VP: middle.block: +; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i64 200, [[N_VEC]] +; NO-VP-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]] +; NO-VP: scalar.ph: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ] +; NO-VP-NEXT: br label [[LOOP1:%.*]] +; NO-VP: loop: +; NO-VP-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[LOOP1]] ] +; NO-VP-NEXT: [[A3:%.*]] = getelementptr i64, ptr [[P]], i64 [[IV1]] +; NO-VP-NEXT: [[V:%.*]] = load i64, ptr [[A3]], align 32 +; NO-VP-NEXT: [[OFFSET:%.*]] = add i64 [[IV1]], 8192 ; NO-VP-NEXT: [[A2:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET]] ; NO-VP-NEXT: store i64 [[V]], ptr [[A2]], align 32 -; NO-VP-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1 -; NO-VP-NEXT: [[CMP:%.*]] = icmp ne i64 [[IV]], 199 -; NO-VP-NEXT: br i1 [[CMP]], label [[LOOP]], label [[EXIT:%.*]] +; NO-VP-NEXT: [[IV_NEXT1]] = add i64 [[IV1]], 1 +; NO-VP-NEXT: [[CMP:%.*]] = icmp ne i64 [[IV1]], 199 +; NO-VP-NEXT: br i1 [[CMP]], label [[LOOP1]], label [[EXIT]], !llvm.loop [[LOOP9:![0-9]+]] ; NO-VP: exit: ; NO-VP-NEXT: ret void ; @@ -448,17 +500,34 @@ define void @no_high_lmul_or_interleave(ptr %p) { ; ; NO-VP-LABEL: @no_high_lmul_or_interleave( ; NO-VP-NEXT: entry: +; NO-VP-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]] +; NO-VP: vector.ph: ; NO-VP-NEXT: br label [[LOOP:%.*]] -; NO-VP: loop: -; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ] +; NO-VP: vector.body: +; NO-VP-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ] ; NO-VP-NEXT: [[A1:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[IV]] -; NO-VP-NEXT: [[V:%.*]] = load i64, ptr [[A1]], align 32 -; NO-VP-NEXT: [[OFFSET:%.*]] = add i64 [[IV]], 1024 +; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[A1]], align 32 +; NO-VP-NEXT: [[TMP1:%.*]] = add i64 [[IV]], 1024 +; NO-VP-NEXT: [[TMP2:%.*]] = getelementptr i64, ptr [[P]], i64 [[TMP1]] +; NO-VP-NEXT: store <4 x i64> [[WIDE_LOAD]], ptr [[TMP2]], align 32 +; NO-VP-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], 4 +; NO-VP-NEXT: [[TMP3:%.*]] = icmp eq i64 [[IV_NEXT]], 3000 +; NO-VP-NEXT: br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP]], !llvm.loop [[LOOP10:![0-9]+]] +; NO-VP: middle.block: +; NO-VP-NEXT: br label [[SCALAR_PH]] +; NO-VP: scalar.ph: +; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 3000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ] +; NO-VP-NEXT: br label [[LOOP1:%.*]] +; NO-VP: loop: +; NO-VP-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[LOOP1]] ] +; NO-VP-NEXT: [[A3:%.*]] = getelementptr i64, ptr [[P]], i64 [[IV1]] +; NO-VP-NEXT: [[V:%.*]] = load i64, ptr [[A3]], align 32 +; NO-VP-NEXT: [[OFFSET:%.*]] = add i64 [[IV1]], 1024 ; NO-VP-NEXT: [[A2:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET]] ; NO-VP-NEXT: store i64 [[V]], ptr [[A2]], align 32 -; NO-VP-NEXT: [[IV_NEXT]] = add i64 [[IV]], 1 -; NO-VP-NEXT: [[CMP:%.*]] = icmp ne i64 [[IV]], 3001 -; NO-VP-NEXT: br i1 [[CMP]], label [[LOOP]], label [[EXIT:%.*]] +; NO-VP-NEXT: [[IV_NEXT1]] = add i64 [[IV1]], 1 +; NO-VP-NEXT: [[CMP:%.*]] = icmp ne i64 [[IV1]], 3001 +; NO-VP-NEXT: br i1 [[CMP]], label [[LOOP1]], label [[EXIT:%.*]], !llvm.loop [[LOOP11:![0-9]+]] ; NO-VP: exit: ; NO-VP-NEXT: ret void ; diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-uniform-store.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-uniform-store.ll index 984b64c..8ab3817 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-uniform-store.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-uniform-store.ll @@ -1,5 +1,5 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 5 -; RUN: opt < %s --prefer-predicate-over-epilogue=predicate-dont-vectorize --passes=loop-vectorize -mcpu=sifive-p470 -mattr=+v,+f -force-tail-folding-style=data-with-evl -S | FileCheck %s +; RUN: opt < %s --prefer-predicate-over-epilogue=predicate-else-scalar-epilogue --passes=loop-vectorize -mcpu=sifive-p470 -mattr=+v,+f -S | FileCheck %s ; Generated from issue #109468. ; In this test case, the vector store with tail mask will transfer to the vp intrinsic with EVL. diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/truncate-to-minimal-bitwidth-evl-crash.ll b/llvm/test/Transforms/LoopVectorize/RISCV/truncate-to-minimal-bitwidth-evl-crash.ll index 1e26d18..27d7bd0 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/truncate-to-minimal-bitwidth-evl-crash.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/truncate-to-minimal-bitwidth-evl-crash.ll @@ -1,5 +1,5 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5 -; RUN: opt -passes=loop-vectorize -force-tail-folding-style=data-with-evl -prefer-predicate-over-epilogue=predicate-dont-vectorize -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s +; RUN: opt -passes=loop-vectorize -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s ; Make sure we don't crash when transforming a VPWidenCastRecipe created without ; an underlying value to an EVL recipe. This occurs in this test via diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/type-info-cache-evl-crash.ll b/llvm/test/Transforms/LoopVectorize/RISCV/type-info-cache-evl-crash.ll index 0e1ec57..27abddf 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/type-info-cache-evl-crash.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/type-info-cache-evl-crash.ll @@ -1,5 +1,5 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5 -; RUN: opt -passes=loop-vectorize -force-tail-folding-style=data-with-evl -prefer-predicate-over-epilogue=predicate-dont-vectorize -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s +; RUN: opt -passes=loop-vectorize -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s ; This test tries to recreate the conditions for a crash that occurred when the ; VPTypeAnalysis cache wasn't cleared after a recipe was erased and clobbered diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/uniform-load-store.ll b/llvm/test/Transforms/LoopVectorize/RISCV/uniform-load-store.ll index 9062542..d3c2ccf 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/uniform-load-store.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/uniform-load-store.ll @@ -1,7 +1,7 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 5 ; RUN: opt < %s -passes=loop-vectorize -scalable-vectorization=on -riscv-v-vector-bits-min=0 -mtriple riscv64-linux-gnu -mattr=+v,+f -S 2>%t | FileCheck %s -check-prefix=SCALABLE ; RUN: opt < %s -passes=loop-vectorize -scalable-vectorization=off -mtriple riscv64-linux-gnu -mattr=+v,+f -S 2>%t | FileCheck %s -check-prefix=FIXEDLEN -; RUN: opt < %s -passes=loop-vectorize -scalable-vectorization=on -riscv-v-vector-bits-min=0 -prefer-predicate-over-epilogue=predicate-dont-vectorize -mtriple riscv64-linux-gnu -mattr=+v,+f -S 2>%t | FileCheck %s -check-prefix=TF-SCALABLE +; RUN: opt < %s -passes=loop-vectorize -scalable-vectorization=on -riscv-v-vector-bits-min=0 -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue -mtriple riscv64-linux-gnu -mattr=+v,+f -S 2>%t | FileCheck %s -check-prefix=TF-SCALABLE target datalayout = "e-m:e-p:64:64-i64:64-i128:128-n64-S128" target triple = "riscv64" @@ -226,17 +226,44 @@ define i64 @uniform_load_outside_use(ptr noalias nocapture %a, ptr noalias nocap ; TF-SCALABLE-LABEL: define i64 @uniform_load_outside_use( ; TF-SCALABLE-SAME: ptr noalias captures(none) [[A:%.*]], ptr noalias captures(none) [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { ; TF-SCALABLE-NEXT: [[ENTRY:.*]]: -; TF-SCALABLE-NEXT: br label %[[FOR_BODY:.*]] -; TF-SCALABLE: [[FOR_BODY]]: -; TF-SCALABLE-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[FOR_BODY]] ] +; TF-SCALABLE-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() +; TF-SCALABLE-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 2 +; TF-SCALABLE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1025, [[TMP1]] +; TF-SCALABLE-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]] +; TF-SCALABLE: [[VECTOR_PH]]: +; TF-SCALABLE-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; TF-SCALABLE-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 2 +; TF-SCALABLE-NEXT: [[N_MOD_VF:%.*]] = urem i64 1025, [[TMP3]] +; TF-SCALABLE-NEXT: [[N_VEC:%.*]] = sub i64 1025, [[N_MOD_VF]] +; TF-SCALABLE-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() +; TF-SCALABLE-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2 +; TF-SCALABLE-NEXT: br label %[[VECTOR_BODY:.*]] +; TF-SCALABLE: [[VECTOR_BODY]]: +; TF-SCALABLE-NEXT: [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] ; TF-SCALABLE-NEXT: [[V:%.*]] = load i64, ptr [[B]], align 8 +; TF-SCALABLE-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[V]], i64 0 +; TF-SCALABLE-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer ; TF-SCALABLE-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]] -; TF-SCALABLE-NEXT: store i64 [[V]], ptr [[ARRAYIDX]], align 8 -; TF-SCALABLE-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 +; TF-SCALABLE-NEXT: store <vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr [[ARRAYIDX]], align 8 +; TF-SCALABLE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP5]] +; TF-SCALABLE-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; TF-SCALABLE-NEXT: br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]] +; TF-SCALABLE: [[MIDDLE_BLOCK]]: +; TF-SCALABLE-NEXT: [[CMP_N:%.*]] = icmp eq i64 1025, [[N_VEC]] +; TF-SCALABLE-NEXT: br i1 [[CMP_N]], label %[[FOR_END:.*]], label %[[SCALAR_PH]] +; TF-SCALABLE: [[SCALAR_PH]]: +; TF-SCALABLE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ] +; TF-SCALABLE-NEXT: br label %[[FOR_BODY:.*]] +; TF-SCALABLE: [[FOR_BODY]]: +; TF-SCALABLE-NEXT: [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[FOR_BODY]] ] +; TF-SCALABLE-NEXT: [[V1:%.*]] = load i64, ptr [[B]], align 8 +; TF-SCALABLE-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV1]] +; TF-SCALABLE-NEXT: store i64 [[V1]], ptr [[ARRAYIDX1]], align 8 +; TF-SCALABLE-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1 ; TF-SCALABLE-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025 -; TF-SCALABLE-NEXT: br i1 [[EXITCOND_NOT]], label %[[FOR_END:.*]], label %[[FOR_BODY]] +; TF-SCALABLE-NEXT: br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]] ; TF-SCALABLE: [[FOR_END]]: -; TF-SCALABLE-NEXT: [[V_LCSSA:%.*]] = phi i64 [ [[V]], %[[FOR_BODY]] ] +; TF-SCALABLE-NEXT: [[V_LCSSA:%.*]] = phi i64 [ [[V1]], %[[FOR_BODY]] ], [ [[V]], %[[MIDDLE_BLOCK]] ] ; TF-SCALABLE-NEXT: ret i64 [[V_LCSSA]] ; entry: @@ -404,7 +431,7 @@ define void @conditional_uniform_load(ptr noalias nocapture %a, ptr noalias noca ; TF-SCALABLE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX1]], [[TMP4]] ; TF-SCALABLE-NEXT: [[VEC_IND_NEXT]] = add <vscale x 4 x i64> [[VEC_IND]], [[DOTSPLAT]] ; TF-SCALABLE-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] -; TF-SCALABLE-NEXT: br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]] +; TF-SCALABLE-NEXT: br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]] ; TF-SCALABLE: [[MIDDLE_BLOCK]]: ; TF-SCALABLE-NEXT: br label %[[FOR_END:.*]] ; TF-SCALABLE: [[SCALAR_PH]]: @@ -423,7 +450,7 @@ define void @conditional_uniform_load(ptr noalias nocapture %a, ptr noalias noca ; TF-SCALABLE-NEXT: store i64 [[PHI]], ptr [[ARRAYIDX]], align 8 ; TF-SCALABLE-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 ; TF-SCALABLE-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025 -; TF-SCALABLE-NEXT: br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]] +; TF-SCALABLE-NEXT: br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]] ; TF-SCALABLE: [[FOR_END]]: ; TF-SCALABLE-NEXT: ret void ; @@ -555,7 +582,7 @@ define void @uniform_load_unaligned(ptr noalias nocapture %a, ptr noalias nocapt ; TF-SCALABLE-NEXT: [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP8]], [[INDEX]] ; TF-SCALABLE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX1]], [[TMP4]] ; TF-SCALABLE-NEXT: [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] -; TF-SCALABLE-NEXT: br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]] +; TF-SCALABLE-NEXT: br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]] ; TF-SCALABLE: [[MIDDLE_BLOCK]]: ; TF-SCALABLE-NEXT: br label %[[FOR_END:.*]] ; TF-SCALABLE: [[SCALAR_PH]]: @@ -568,7 +595,7 @@ define void @uniform_load_unaligned(ptr noalias nocapture %a, ptr noalias nocapt ; TF-SCALABLE-NEXT: store i64 [[V]], ptr [[ARRAYIDX]], align 8 ; TF-SCALABLE-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 ; TF-SCALABLE-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025 -; TF-SCALABLE-NEXT: br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]] +; TF-SCALABLE-NEXT: br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]] ; TF-SCALABLE: [[FOR_END]]: ; TF-SCALABLE-NEXT: ret void ; @@ -693,7 +720,7 @@ define void @uniform_store(ptr noalias nocapture %a, ptr noalias nocapture %b, i ; TF-SCALABLE-NEXT: [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP7]], [[INDEX]] ; TF-SCALABLE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX1]], [[TMP4]] ; TF-SCALABLE-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] -; TF-SCALABLE-NEXT: br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]] +; TF-SCALABLE-NEXT: br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]] ; TF-SCALABLE: [[MIDDLE_BLOCK]]: ; TF-SCALABLE-NEXT: br label %[[FOR_END:.*]] ; TF-SCALABLE: [[SCALAR_PH]]: @@ -706,7 +733,7 @@ define void @uniform_store(ptr noalias nocapture %a, ptr noalias nocapture %b, i ; TF-SCALABLE-NEXT: store i64 [[V]], ptr [[ARRAYIDX]], align 8 ; TF-SCALABLE-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 ; TF-SCALABLE-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025 -; TF-SCALABLE-NEXT: br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]] +; TF-SCALABLE-NEXT: br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]] ; TF-SCALABLE: [[FOR_END]]: ; TF-SCALABLE-NEXT: ret void ; @@ -858,7 +885,7 @@ define void @uniform_store_of_loop_varying(ptr noalias nocapture %a, ptr noalias ; TF-SCALABLE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX1]], [[TMP4]] ; TF-SCALABLE-NEXT: [[VEC_IND_NEXT]] = add <vscale x 2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT2]] ; TF-SCALABLE-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] -; TF-SCALABLE-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]] +; TF-SCALABLE-NEXT: br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]] ; TF-SCALABLE: [[MIDDLE_BLOCK]]: ; TF-SCALABLE-NEXT: br label %[[FOR_END:.*]] ; TF-SCALABLE: [[SCALAR_PH]]: @@ -871,7 +898,7 @@ define void @uniform_store_of_loop_varying(ptr noalias nocapture %a, ptr noalias ; TF-SCALABLE-NEXT: store i64 [[V]], ptr [[ARRAYIDX]], align 8 ; TF-SCALABLE-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 ; TF-SCALABLE-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025 -; TF-SCALABLE-NEXT: br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]] +; TF-SCALABLE-NEXT: br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]] ; TF-SCALABLE: [[FOR_END]]: ; TF-SCALABLE-NEXT: ret void ; @@ -1037,7 +1064,7 @@ define void @conditional_uniform_store(ptr noalias nocapture %a, ptr noalias noc ; TF-SCALABLE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX1]], [[TMP4]] ; TF-SCALABLE-NEXT: [[VEC_IND_NEXT]] = add <vscale x 2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT]] ; TF-SCALABLE-NEXT: [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] -; TF-SCALABLE-NEXT: br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]] +; TF-SCALABLE-NEXT: br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]] ; TF-SCALABLE: [[MIDDLE_BLOCK]]: ; TF-SCALABLE-NEXT: br label %[[FOR_END:.*]] ; TF-SCALABLE: [[SCALAR_PH]]: @@ -1055,7 +1082,7 @@ define void @conditional_uniform_store(ptr noalias nocapture %a, ptr noalias noc ; TF-SCALABLE-NEXT: store i64 [[V]], ptr [[ARRAYIDX]], align 8 ; TF-SCALABLE-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 ; TF-SCALABLE-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025 -; TF-SCALABLE-NEXT: br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]] +; TF-SCALABLE-NEXT: br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]] ; TF-SCALABLE: [[FOR_END]]: ; TF-SCALABLE-NEXT: ret void ; @@ -1186,7 +1213,7 @@ define void @uniform_store_unaligned(ptr noalias nocapture %a, ptr noalias nocap ; TF-SCALABLE-NEXT: [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP7]], [[INDEX]] ; TF-SCALABLE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX1]], [[TMP4]] ; TF-SCALABLE-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] -; TF-SCALABLE-NEXT: br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]] +; TF-SCALABLE-NEXT: br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]] ; TF-SCALABLE: [[MIDDLE_BLOCK]]: ; TF-SCALABLE-NEXT: br label %[[FOR_END:.*]] ; TF-SCALABLE: [[SCALAR_PH]]: @@ -1199,7 +1226,7 @@ define void @uniform_store_unaligned(ptr noalias nocapture %a, ptr noalias nocap ; TF-SCALABLE-NEXT: store i64 [[V]], ptr [[ARRAYIDX]], align 8 ; TF-SCALABLE-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1 ; TF-SCALABLE-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025 -; TF-SCALABLE-NEXT: br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]] +; TF-SCALABLE-NEXT: br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]] ; TF-SCALABLE: [[FOR_END]]: ; TF-SCALABLE-NEXT: ret void ; diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-vp-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-vp-intrinsics.ll index 74eb094..7537a43 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-vp-intrinsics.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-vp-intrinsics.ll @@ -1,12 +1,10 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck --check-prefix=IF-EVL %s ; RUN: opt -passes=loop-vectorize \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck --check-prefix=NO-VP %s define void @foo(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %N) { diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-call-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-call-intrinsics.ll index 9652351..7f1066c 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-call-intrinsics.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-call-intrinsics.ll @@ -1,8 +1,7 @@ ; REQUIRES: asserts ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefix=IF-EVL %s define void @vp_smax(ptr %a, ptr %b, ptr %c, i64 %N) { diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll index 3ec48ef..c1b656a 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll @@ -1,7 +1,6 @@ ; REQUIRES: asserts ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefix=IF-EVL %s define void @vp_sext(ptr %a, ptr %b, i64 %N) { diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics-fixed-order-recurrence.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics-fixed-order-recurrence.ll index 7f29213..9900602 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics-fixed-order-recurrence.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics-fixed-order-recurrence.ll @@ -1,8 +1,7 @@ ; REQUIRES: asserts ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefix=IF-EVL %s define void @first_order_recurrence(ptr noalias %A, ptr noalias %B, i64 %TC) { diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics-reduction.ll index baf546b..1c9554d 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics-reduction.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics-reduction.ll @@ -1,25 +1,21 @@ ; REQUIRES: asserts ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefixes=IF-EVL-OUTLOOP,IF-EVL %s ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \ ; RUN: -prefer-inloop-reductions \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefixes=IF-EVL-INLOOP,IF-EVL %s ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefixes=NO-VP-OUTLOOP %s ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \ ; RUN: -prefer-inloop-reductions \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefixes=NO-VP-INLOOP %s diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics.ll index 97a6130..42a846a 100644 --- a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics.ll +++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics.ll @@ -1,13 +1,11 @@ ; REQUIRES: asserts ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \ -; RUN: -force-tail-folding-style=data-with-evl \ -; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \ +; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefixes=IF-EVL,CHECK %s ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \ -; RUN: -force-tail-folding-style=none \ -; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \ +; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \ ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefixes=NO-VP,CHECK %s define void @foo(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %N) { |