diff options
Diffstat (limited to 'llvm/test/CodeGen')
39 files changed, 5560 insertions, 3359 deletions
diff --git a/llvm/test/CodeGen/AArch64/aarch64-matmul-fp16.ll b/llvm/test/CodeGen/AArch64/aarch64-matmul-fp16.ll new file mode 100644 index 0000000..8d1abdd --- /dev/null +++ b/llvm/test/CodeGen/AArch64/aarch64-matmul-fp16.ll @@ -0,0 +1,14 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc -mtriple aarch64-none-linux-gnu -mattr=+neon,+f8f16mm < %s | FileCheck %s +; RUN: llc -mtriple aarch64-none-linux-gnu -mattr=+neon,+f8f16mm -global-isel < %s | FileCheck %s + +define <8 x half> @fmmla.v8f16.v16i8(<8 x half> %r, <16 x i8> %a, <16 x i8> %b) { +; CHECK-LABEL: fmmla.v8f16.v16i8: +; CHECK: // %bb.0: // %entry +; CHECK-NEXT: fmmla v0.8h, v1.16b, v2.16b +; CHECK-NEXT: ret +entry: + %vfmmla1.i = tail call <8 x half> @llvm.aarch64.neon.fmmla.v8f16.v16i8(<8 x half> %r, <16 x i8> %a, <16 x i8> %b) #3 + ret <8 x half> %vfmmla1.i +} + diff --git a/llvm/test/CodeGen/AArch64/aarch64-matmul-fp32.ll b/llvm/test/CodeGen/AArch64/aarch64-matmul-fp32.ll new file mode 100644 index 0000000..4c33567 --- /dev/null +++ b/llvm/test/CodeGen/AArch64/aarch64-matmul-fp32.ll @@ -0,0 +1,13 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc -mtriple aarch64-none-linux-gnu -mattr=+neon,+f8f32mm < %s | FileCheck %s +; RUN: llc -mtriple aarch64-none-linux-gnu -mattr=+neon,+f8f32mm -global-isel < %s | FileCheck %s + +define <4 x float> @fmmla.v4f32.v16i8(<4 x float> %r, <16 x i8> %a, <16 x i8> %b) { +; CHECK-LABEL: fmmla.v4f32.v16i8: +; CHECK: // %bb.0: // %entry +; CHECK-NEXT: fmmla v0.4s, v1.16b, v2.16b +; CHECK-NEXT: ret +entry: + %vfmmla1.i = tail call <4 x float> @llvm.aarch64.neon.fmmla.v4f32.v16i8(<4 x float> %r, <16 x i8> %a, <16 x i8> %b) #3 + ret <4 x float> %vfmmla1.i +} diff --git a/llvm/test/CodeGen/AArch64/sve2p1-fdot.ll b/llvm/test/CodeGen/AArch64/sve2p1-fdot.ll new file mode 100644 index 0000000..9dbe096 --- /dev/null +++ b/llvm/test/CodeGen/AArch64/sve2p1-fdot.ll @@ -0,0 +1,93 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc -mattr=+sve2 < %s | FileCheck %s --check-prefixes=CHECK,SVE2 +; RUN: llc -global-isel -global-isel-abort=2 -mattr=+sve2 < %s | FileCheck %s --check-prefixes=CHECK,SVE2 +; RUN: llc -mattr=+sve2p1 < %s | FileCheck %s --check-prefixes=CHECK,SVE2P1 +; RUN: llc -global-isel -global-isel-abort=2 -mattr=+sve2p1 < %s | FileCheck %s --check-prefixes=CHECK,SVE2P1 + +target triple = "aarch64-linux-gnu" + +define <vscale x 4 x float> @fdot_wide_nxv4f32(<vscale x 4 x float> %acc, <vscale x 8 x half> %a, <vscale x 8 x half> %b) { +; SVE2-LABEL: fdot_wide_nxv4f32: +; SVE2: // %bb.0: // %entry +; SVE2-NEXT: uunpklo z3.s, z1.h +; SVE2-NEXT: uunpklo z4.s, z2.h +; SVE2-NEXT: ptrue p0.s +; SVE2-NEXT: uunpkhi z1.s, z1.h +; SVE2-NEXT: uunpkhi z2.s, z2.h +; SVE2-NEXT: fcvt z3.s, p0/m, z3.h +; SVE2-NEXT: fcvt z4.s, p0/m, z4.h +; SVE2-NEXT: fcvt z1.s, p0/m, z1.h +; SVE2-NEXT: fcvt z2.s, p0/m, z2.h +; SVE2-NEXT: fmul z3.s, z3.s, z4.s +; SVE2-NEXT: fmul z1.s, z1.s, z2.s +; SVE2-NEXT: fadd z0.s, z0.s, z3.s +; SVE2-NEXT: fadd z0.s, z0.s, z1.s +; SVE2-NEXT: ret +; +; SVE2P1-LABEL: fdot_wide_nxv4f32: +; SVE2P1: // %bb.0: // %entry +; SVE2P1-NEXT: fdot z0.s, z1.h, z2.h +; SVE2P1-NEXT: ret +entry: + %a.wide = fpext <vscale x 8 x half> %a to <vscale x 8 x float> + %b.wide = fpext <vscale x 8 x half> %b to <vscale x 8 x float> + %mult = fmul <vscale x 8 x float> %a.wide, %b.wide + %partial.reduce = call <vscale x 4 x float> @llvm.vector.partial.reduce.fadd(<vscale x 4 x float> %acc, <vscale x 8 x float> %mult) + ret <vscale x 4 x float> %partial.reduce +} + +define <vscale x 4 x float> @fdot_splat_nxv4f32(<vscale x 4 x float> %acc, <vscale x 8 x half> %a) { +; SVE2-LABEL: fdot_splat_nxv4f32: +; SVE2: // %bb.0: // %entry +; SVE2-NEXT: uunpklo z2.s, z1.h +; SVE2-NEXT: ptrue p0.s +; SVE2-NEXT: uunpkhi z1.s, z1.h +; SVE2-NEXT: fcvt z2.s, p0/m, z2.h +; SVE2-NEXT: fcvt z1.s, p0/m, z1.h +; SVE2-NEXT: fadd z0.s, z0.s, z2.s +; SVE2-NEXT: fadd z0.s, z0.s, z1.s +; SVE2-NEXT: ret +; +; SVE2P1-LABEL: fdot_splat_nxv4f32: +; SVE2P1: // %bb.0: // %entry +; SVE2P1-NEXT: fmov z2.h, #1.00000000 +; SVE2P1-NEXT: fdot z0.s, z1.h, z2.h +; SVE2P1-NEXT: ret +entry: + %a.wide = fpext <vscale x 8 x half> %a to <vscale x 8 x float> + %partial.reduce = call <vscale x 4 x float> @llvm.vector.partial.reduce.fadd(<vscale x 4 x float> %acc, <vscale x 8 x float> %a.wide) + ret <vscale x 4 x float> %partial.reduce +} + +define <vscale x 8 x half> @partial_reduce_nxv8f16(<vscale x 8 x half> %acc, <vscale x 16 x half> %a) { +; CHECK-LABEL: partial_reduce_nxv8f16: +; CHECK: // %bb.0: // %entry +; CHECK-NEXT: fadd z0.h, z0.h, z1.h +; CHECK-NEXT: fadd z0.h, z0.h, z2.h +; CHECK-NEXT: ret +entry: + %partial.reduce = call <vscale x 8 x half> @llvm.vector.partial.reduce.fadd(<vscale x 8 x half> %acc, <vscale x 16 x half> %a) + ret <vscale x 8 x half> %partial.reduce +} + +define <vscale x 4 x float> @partial_reduce_nxv4f32(<vscale x 4 x float> %acc, <vscale x 8 x float> %a) { +; CHECK-LABEL: partial_reduce_nxv4f32: +; CHECK: // %bb.0: // %entry +; CHECK-NEXT: fadd z0.s, z0.s, z1.s +; CHECK-NEXT: fadd z0.s, z0.s, z2.s +; CHECK-NEXT: ret +entry: + %partial.reduce = call <vscale x 4 x float> @llvm.vector.partial.reduce.fadd(<vscale x 4 x float> %acc, <vscale x 8 x float> %a) + ret <vscale x 4 x float> %partial.reduce +} + +define <vscale x 2 x double> @partial_reduce_nxv2f64(<vscale x 2 x double> %acc, <vscale x 4 x double> %a) { +; CHECK-LABEL: partial_reduce_nxv2f64: +; CHECK: // %bb.0: // %entry +; CHECK-NEXT: fadd z0.d, z0.d, z1.d +; CHECK-NEXT: fadd z0.d, z0.d, z2.d +; CHECK-NEXT: ret +entry: + %partial.reduce = call <vscale x 2 x double> @llvm.vector.partial.reduce.fadd(<vscale x 2 x double> %acc, <vscale x 4 x double> %a) + ret <vscale x 2 x double> %partial.reduce +} diff --git a/llvm/test/CodeGen/AArch64/sve2p1-fixed-length-fdot.ll b/llvm/test/CodeGen/AArch64/sve2p1-fixed-length-fdot.ll new file mode 100644 index 0000000..89216ce --- /dev/null +++ b/llvm/test/CodeGen/AArch64/sve2p1-fixed-length-fdot.ll @@ -0,0 +1,230 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc -mattr=+sve2 < %s | FileCheck %s --check-prefixes=CHECK,SVE2 +; RUN: llc -mattr=+sve2p1 < %s | FileCheck %s --check-prefixes=CHECK,SVE2P1 + +target triple = "aarch64-linux-gnu" + +define void @fdot_wide_v8f32(ptr %accptr, ptr %aptr, ptr %bptr) vscale_range(2,0) { +; SVE2-LABEL: fdot_wide_v8f32: +; SVE2: // %bb.0: // %entry +; SVE2-NEXT: ptrue p0.s, vl8 +; SVE2-NEXT: mov x8, #8 // =0x8 +; SVE2-NEXT: ld1h { z0.s }, p0/z, [x1] +; SVE2-NEXT: ld1h { z1.s }, p0/z, [x2] +; SVE2-NEXT: ld1h { z2.s }, p0/z, [x1, x8, lsl #1] +; SVE2-NEXT: ld1h { z3.s }, p0/z, [x2, x8, lsl #1] +; SVE2-NEXT: fcvt z0.s, p0/m, z0.h +; SVE2-NEXT: fcvt z1.s, p0/m, z1.h +; SVE2-NEXT: fcvt z2.s, p0/m, z2.h +; SVE2-NEXT: fcvt z3.s, p0/m, z3.h +; SVE2-NEXT: fmul z0.s, p0/m, z0.s, z1.s +; SVE2-NEXT: ld1w { z1.s }, p0/z, [x0] +; SVE2-NEXT: fmul z2.s, p0/m, z2.s, z3.s +; SVE2-NEXT: fadd z0.s, p0/m, z0.s, z1.s +; SVE2-NEXT: fadd z0.s, p0/m, z0.s, z2.s +; SVE2-NEXT: st1w { z0.s }, p0, [x0] +; SVE2-NEXT: ret +; +; SVE2P1-LABEL: fdot_wide_v8f32: +; SVE2P1: // %bb.0: // %entry +; SVE2P1-NEXT: ptrue p0.s, vl8 +; SVE2P1-NEXT: ptrue p1.h, vl16 +; SVE2P1-NEXT: ld1w { z0.s }, p0/z, [x0] +; SVE2P1-NEXT: ld1h { z1.h }, p1/z, [x1] +; SVE2P1-NEXT: ld1h { z2.h }, p1/z, [x2] +; SVE2P1-NEXT: fdot z0.s, z1.h, z2.h +; SVE2P1-NEXT: st1w { z0.s }, p0, [x0] +; SVE2P1-NEXT: ret +entry: + %acc = load <8 x float>, ptr %accptr + %a = load <16 x half>, ptr %aptr + %b = load <16 x half>, ptr %bptr + %a.wide = fpext <16 x half> %a to <16 x float> + %b.wide = fpext <16 x half> %b to <16 x float> + %mult = fmul <16 x float> %a.wide, %b.wide + %partial.reduce = call <8 x float> @llvm.vector.partial.reduce.fadd(<8 x float> %acc, <16 x float> %mult) + store <8 x float> %partial.reduce, ptr %accptr + ret void +} + +define void @fdot_wide_v16f32(ptr %accptr, ptr %aptr, ptr %bptr) vscale_range(4,0) { +; SVE2-LABEL: fdot_wide_v16f32: +; SVE2: // %bb.0: // %entry +; SVE2-NEXT: ptrue p0.s, vl16 +; SVE2-NEXT: mov x8, #16 // =0x10 +; SVE2-NEXT: ld1h { z0.s }, p0/z, [x1] +; SVE2-NEXT: ld1h { z1.s }, p0/z, [x2] +; SVE2-NEXT: ld1h { z2.s }, p0/z, [x1, x8, lsl #1] +; SVE2-NEXT: ld1h { z3.s }, p0/z, [x2, x8, lsl #1] +; SVE2-NEXT: fcvt z0.s, p0/m, z0.h +; SVE2-NEXT: fcvt z1.s, p0/m, z1.h +; SVE2-NEXT: fcvt z2.s, p0/m, z2.h +; SVE2-NEXT: fcvt z3.s, p0/m, z3.h +; SVE2-NEXT: fmul z0.s, p0/m, z0.s, z1.s +; SVE2-NEXT: ld1w { z1.s }, p0/z, [x0] +; SVE2-NEXT: fmul z2.s, p0/m, z2.s, z3.s +; SVE2-NEXT: fadd z0.s, p0/m, z0.s, z1.s +; SVE2-NEXT: fadd z0.s, p0/m, z0.s, z2.s +; SVE2-NEXT: st1w { z0.s }, p0, [x0] +; SVE2-NEXT: ret +; +; SVE2P1-LABEL: fdot_wide_v16f32: +; SVE2P1: // %bb.0: // %entry +; SVE2P1-NEXT: ptrue p0.s, vl16 +; SVE2P1-NEXT: ptrue p1.h, vl32 +; SVE2P1-NEXT: ld1w { z0.s }, p0/z, [x0] +; SVE2P1-NEXT: ld1h { z1.h }, p1/z, [x1] +; SVE2P1-NEXT: ld1h { z2.h }, p1/z, [x2] +; SVE2P1-NEXT: fdot z0.s, z1.h, z2.h +; SVE2P1-NEXT: st1w { z0.s }, p0, [x0] +; SVE2P1-NEXT: ret +entry: + %acc = load <16 x float>, ptr %accptr + %a = load <32 x half>, ptr %aptr + %b = load <32 x half>, ptr %bptr + %a.wide = fpext <32 x half> %a to <32 x float> + %b.wide = fpext <32 x half> %b to <32 x float> + %mult = fmul <32 x float> %a.wide, %b.wide + %partial.reduce = call <16 x float> @llvm.vector.partial.reduce.fadd(<16 x float> %acc, <32 x float> %mult) + store <16 x float> %partial.reduce, ptr %accptr + ret void +} + +define void @fdot_wide_v32f32(ptr %accptr, ptr %aptr, ptr %bptr) vscale_range(8,0) { +; SVE2-LABEL: fdot_wide_v32f32: +; SVE2: // %bb.0: // %entry +; SVE2-NEXT: ptrue p0.s, vl32 +; SVE2-NEXT: mov x8, #32 // =0x20 +; SVE2-NEXT: ld1h { z0.s }, p0/z, [x1] +; SVE2-NEXT: ld1h { z1.s }, p0/z, [x2] +; SVE2-NEXT: ld1h { z2.s }, p0/z, [x1, x8, lsl #1] +; SVE2-NEXT: ld1h { z3.s }, p0/z, [x2, x8, lsl #1] +; SVE2-NEXT: fcvt z0.s, p0/m, z0.h +; SVE2-NEXT: fcvt z1.s, p0/m, z1.h +; SVE2-NEXT: fcvt z2.s, p0/m, z2.h +; SVE2-NEXT: fcvt z3.s, p0/m, z3.h +; SVE2-NEXT: fmul z0.s, p0/m, z0.s, z1.s +; SVE2-NEXT: ld1w { z1.s }, p0/z, [x0] +; SVE2-NEXT: fmul z2.s, p0/m, z2.s, z3.s +; SVE2-NEXT: fadd z0.s, p0/m, z0.s, z1.s +; SVE2-NEXT: fadd z0.s, p0/m, z0.s, z2.s +; SVE2-NEXT: st1w { z0.s }, p0, [x0] +; SVE2-NEXT: ret +; +; SVE2P1-LABEL: fdot_wide_v32f32: +; SVE2P1: // %bb.0: // %entry +; SVE2P1-NEXT: ptrue p0.s, vl32 +; SVE2P1-NEXT: ptrue p1.h, vl64 +; SVE2P1-NEXT: ld1w { z0.s }, p0/z, [x0] +; SVE2P1-NEXT: ld1h { z1.h }, p1/z, [x1] +; SVE2P1-NEXT: ld1h { z2.h }, p1/z, [x2] +; SVE2P1-NEXT: fdot z0.s, z1.h, z2.h +; SVE2P1-NEXT: st1w { z0.s }, p0, [x0] +; SVE2P1-NEXT: ret +entry: + %acc = load <32 x float>, ptr %accptr + %a = load <64 x half>, ptr %aptr + %b = load <64 x half>, ptr %bptr + %a.wide = fpext <64 x half> %a to <64 x float> + %b.wide = fpext <64 x half> %b to <64 x float> + %mult = fmul <64 x float> %a.wide, %b.wide + %partial.reduce = call <32 x float> @llvm.vector.partial.reduce.fadd(<32 x float> %acc, <64 x float> %mult) + store <32 x float> %partial.reduce, ptr %accptr + ret void +} + +define void @fdot_wide_v64f32(ptr %accptr, ptr %aptr, ptr %bptr) vscale_range(16,0) { +; SVE2-LABEL: fdot_wide_v64f32: +; SVE2: // %bb.0: // %entry +; SVE2-NEXT: ptrue p0.s, vl64 +; SVE2-NEXT: mov x8, #64 // =0x40 +; SVE2-NEXT: ld1h { z0.s }, p0/z, [x1] +; SVE2-NEXT: ld1h { z1.s }, p0/z, [x2] +; SVE2-NEXT: ld1h { z2.s }, p0/z, [x1, x8, lsl #1] +; SVE2-NEXT: ld1h { z3.s }, p0/z, [x2, x8, lsl #1] +; SVE2-NEXT: fcvt z0.s, p0/m, z0.h +; SVE2-NEXT: fcvt z1.s, p0/m, z1.h +; SVE2-NEXT: fcvt z2.s, p0/m, z2.h +; SVE2-NEXT: fcvt z3.s, p0/m, z3.h +; SVE2-NEXT: fmul z0.s, p0/m, z0.s, z1.s +; SVE2-NEXT: ld1w { z1.s }, p0/z, [x0] +; SVE2-NEXT: fmul z2.s, p0/m, z2.s, z3.s +; SVE2-NEXT: fadd z0.s, p0/m, z0.s, z1.s +; SVE2-NEXT: fadd z0.s, p0/m, z0.s, z2.s +; SVE2-NEXT: st1w { z0.s }, p0, [x0] +; SVE2-NEXT: ret +; +; SVE2P1-LABEL: fdot_wide_v64f32: +; SVE2P1: // %bb.0: // %entry +; SVE2P1-NEXT: ptrue p0.s, vl64 +; SVE2P1-NEXT: ptrue p1.h, vl128 +; SVE2P1-NEXT: ld1w { z0.s }, p0/z, [x0] +; SVE2P1-NEXT: ld1h { z1.h }, p1/z, [x1] +; SVE2P1-NEXT: ld1h { z2.h }, p1/z, [x2] +; SVE2P1-NEXT: fdot z0.s, z1.h, z2.h +; SVE2P1-NEXT: st1w { z0.s }, p0, [x0] +; SVE2P1-NEXT: ret +entry: + %acc = load <64 x float>, ptr %accptr + %a = load <128 x half>, ptr %aptr + %b = load <128 x half>, ptr %bptr + %a.wide = fpext <128 x half> %a to <128 x float> + %b.wide = fpext <128 x half> %b to <128 x float> + %mult = fmul <128 x float> %a.wide, %b.wide + %partial.reduce = call <64 x float> @llvm.vector.partial.reduce.fadd(<64 x float> %acc, <128 x float> %mult) + store <64 x float> %partial.reduce, ptr %accptr + ret void +} + +define <4 x float> @fixed_fdot_wide(<4 x float> %acc, <8 x half> %a, <8 x half> %b) { +; CHECK-LABEL: fixed_fdot_wide: +; CHECK: // %bb.0: // %entry +; CHECK-NEXT: fcvtl v3.4s, v1.4h +; CHECK-NEXT: fcvtl v4.4s, v2.4h +; CHECK-NEXT: fcvtl2 v1.4s, v1.8h +; CHECK-NEXT: fcvtl2 v2.4s, v2.8h +; CHECK-NEXT: fmul v3.4s, v3.4s, v4.4s +; CHECK-NEXT: fmul v1.4s, v1.4s, v2.4s +; CHECK-NEXT: fadd v0.4s, v0.4s, v3.4s +; CHECK-NEXT: fadd v0.4s, v0.4s, v1.4s +; CHECK-NEXT: ret +entry: + %a.wide = fpext <8 x half> %a to <8 x float> + %b.wide = fpext <8 x half> %b to <8 x float> + %mult = fmul <8 x float> %a.wide, %b.wide + %partial.reduce = call <4 x float> @llvm.vector.partial.reduce.fadd(<4 x float> %acc, <8 x float> %mult) + ret <4 x float> %partial.reduce +} + +define <8 x half> @partial_reduce_half(<8 x half> %acc, <16 x half> %a) { +; CHECK-LABEL: partial_reduce_half: +; CHECK: // %bb.0: // %entry +; CHECK-NEXT: fadd v0.8h, v0.8h, v1.8h +; CHECK-NEXT: fadd v0.8h, v0.8h, v2.8h +; CHECK-NEXT: ret +entry: + %partial.reduce = call <8 x half> @llvm.vector.partial.reduce.fadd(<8 x half> %acc, <16 x half> %a) + ret <8 x half> %partial.reduce +} + +define <4 x float> @partial_reduce_float(<4 x float> %acc, <8 x float> %a) { +; CHECK-LABEL: partial_reduce_float: +; CHECK: // %bb.0: // %entry +; CHECK-NEXT: fadd v0.4s, v0.4s, v1.4s +; CHECK-NEXT: fadd v0.4s, v0.4s, v2.4s +; CHECK-NEXT: ret +entry: + %partial.reduce = call <4 x float> @llvm.vector.partial.reduce.fadd(<4 x float> %acc, <8 x float> %a) + ret <4 x float> %partial.reduce +} + +define <2 x double> @partial_reduce_double(<2 x double> %acc, <4 x double> %a) { +; CHECK-LABEL: partial_reduce_double: +; CHECK: // %bb.0: // %entry +; CHECK-NEXT: fadd v0.2d, v0.2d, v1.2d +; CHECK-NEXT: fadd v0.2d, v0.2d, v2.2d +; CHECK-NEXT: ret +entry: + %partial.reduce = call <2 x double> @llvm.vector.partial.reduce.fadd(<2 x double> %acc, <4 x double> %a) + ret <2 x double> %partial.reduce +} diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.mfma.gfx90a.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.mfma.gfx90a.ll index 5720b88..cc21305 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.mfma.gfx90a.ll +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.mfma.gfx90a.ll @@ -1,5 +1,5 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck --check-prefixes=GCN %s +; RUN: llc -verify-machineinstrs -global-isel -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck --check-prefixes=GCN %s declare <32 x float> @llvm.amdgcn.mfma.f32.32x32x4bf16.1k(<4 x i16>, <4 x i16>, <32 x float>, i32, i32, i32) declare <16 x float> @llvm.amdgcn.mfma.f32.16x16x4bf16.1k(<4 x i16>, <4 x i16>, <16 x float>, i32, i32, i32) diff --git a/llvm/test/CodeGen/AMDGPU/amdgpu-codegenprepare-idiv.ll b/llvm/test/CodeGen/AMDGPU/amdgpu-codegenprepare-idiv.ll index 51df8c3..54b1554 100644 --- a/llvm/test/CodeGen/AMDGPU/amdgpu-codegenprepare-idiv.ll +++ b/llvm/test/CodeGen/AMDGPU/amdgpu-codegenprepare-idiv.ll @@ -7772,7 +7772,6 @@ define amdgpu_kernel void @sdiv_i64_pow2_shl_denom(ptr addrspace(1) %out, i64 %x ; GFX6-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xd ; GFX6-NEXT: s_load_dwordx4 s[4:7], s[4:5], 0x9 ; GFX6-NEXT: s_mov_b32 s3, 0xf000 -; GFX6-NEXT: s_mov_b32 s2, -1 ; GFX6-NEXT: s_waitcnt lgkmcnt(0) ; GFX6-NEXT: s_lshl_b64 s[0:1], 0x1000, s0 ; GFX6-NEXT: s_ashr_i32 s8, s1, 31 @@ -7782,8 +7781,8 @@ define amdgpu_kernel void @sdiv_i64_pow2_shl_denom(ptr addrspace(1) %out, i64 %x ; GFX6-NEXT: s_xor_b64 s[10:11], s[0:1], s[8:9] ; GFX6-NEXT: v_cvt_f32_u32_e32 v0, s10 ; GFX6-NEXT: v_cvt_f32_u32_e32 v1, s11 -; GFX6-NEXT: s_sub_u32 s12, 0, s10 -; GFX6-NEXT: s_subb_u32 s13, 0, s11 +; GFX6-NEXT: s_sub_u32 s0, 0, s10 +; GFX6-NEXT: s_subb_u32 s1, 0, s11 ; GFX6-NEXT: v_madmk_f32 v0, v1, 0x4f800000, v0 ; GFX6-NEXT: v_rcp_f32_e32 v0, v0 ; GFX6-NEXT: v_mul_f32_e32 v0, 0x5f7ffffc, v0 @@ -7792,128 +7791,121 @@ define amdgpu_kernel void @sdiv_i64_pow2_shl_denom(ptr addrspace(1) %out, i64 %x ; GFX6-NEXT: v_madmk_f32 v0, v1, 0xcf800000, v0 ; GFX6-NEXT: v_cvt_u32_f32_e32 v0, v0 ; GFX6-NEXT: v_cvt_u32_f32_e32 v1, v1 -; GFX6-NEXT: v_mul_hi_u32 v2, s12, v0 -; GFX6-NEXT: v_readfirstlane_b32 s14, v1 -; GFX6-NEXT: v_readfirstlane_b32 s0, v0 -; GFX6-NEXT: s_mul_i32 s1, s12, s14 -; GFX6-NEXT: v_readfirstlane_b32 s17, v2 -; GFX6-NEXT: s_mul_i32 s15, s13, s0 -; GFX6-NEXT: s_mul_i32 s16, s12, s0 -; GFX6-NEXT: s_add_i32 s1, s17, s1 -; GFX6-NEXT: v_mul_hi_u32 v3, v0, s16 -; GFX6-NEXT: s_add_i32 s1, s1, s15 -; GFX6-NEXT: v_mul_hi_u32 v0, v0, s1 -; GFX6-NEXT: v_mul_hi_u32 v4, v1, s16 -; GFX6-NEXT: v_readfirstlane_b32 s15, v3 -; GFX6-NEXT: s_mul_i32 s17, s0, s1 -; GFX6-NEXT: v_mul_hi_u32 v1, v1, s1 -; GFX6-NEXT: s_add_u32 s15, s15, s17 -; GFX6-NEXT: v_readfirstlane_b32 s17, v0 -; GFX6-NEXT: s_addc_u32 s17, 0, s17 -; GFX6-NEXT: s_mul_i32 s16, s14, s16 -; GFX6-NEXT: v_readfirstlane_b32 s18, v4 -; GFX6-NEXT: s_add_u32 s15, s15, s16 -; GFX6-NEXT: s_addc_u32 s15, s17, s18 -; GFX6-NEXT: v_readfirstlane_b32 s16, v1 -; GFX6-NEXT: s_addc_u32 s16, s16, 0 -; GFX6-NEXT: s_mul_i32 s1, s14, s1 -; GFX6-NEXT: s_add_u32 s1, s15, s1 -; GFX6-NEXT: s_addc_u32 s15, 0, s16 -; GFX6-NEXT: s_add_u32 s16, s0, s1 -; GFX6-NEXT: v_mov_b32_e32 v0, s16 -; GFX6-NEXT: v_mul_hi_u32 v0, s12, v0 -; GFX6-NEXT: s_cselect_b64 s[0:1], -1, 0 -; GFX6-NEXT: s_or_b32 s0, s0, s1 -; GFX6-NEXT: s_addc_u32 s14, s14, s15 -; GFX6-NEXT: s_mul_i32 s0, s12, s14 -; GFX6-NEXT: v_readfirstlane_b32 s1, v0 -; GFX6-NEXT: s_add_i32 s0, s1, s0 -; GFX6-NEXT: s_mul_i32 s13, s13, s16 -; GFX6-NEXT: s_mul_i32 s1, s12, s16 -; GFX6-NEXT: s_add_i32 s0, s0, s13 -; GFX6-NEXT: v_mov_b32_e32 v2, s1 -; GFX6-NEXT: v_mov_b32_e32 v0, s0 -; GFX6-NEXT: v_mul_hi_u32 v3, s14, v2 -; GFX6-NEXT: v_mul_hi_u32 v2, s16, v2 -; GFX6-NEXT: v_mul_hi_u32 v1, s14, v0 -; GFX6-NEXT: v_mul_hi_u32 v0, s16, v0 -; GFX6-NEXT: s_mul_i32 s13, s16, s0 -; GFX6-NEXT: v_readfirstlane_b32 s17, v2 -; GFX6-NEXT: s_add_u32 s13, s17, s13 -; GFX6-NEXT: v_readfirstlane_b32 s15, v0 -; GFX6-NEXT: s_mul_i32 s1, s14, s1 -; GFX6-NEXT: s_addc_u32 s15, 0, s15 -; GFX6-NEXT: v_readfirstlane_b32 s12, v3 -; GFX6-NEXT: s_add_u32 s1, s13, s1 -; GFX6-NEXT: s_addc_u32 s1, s15, s12 +; GFX6-NEXT: v_mul_hi_u32 v2, s0, v0 ; GFX6-NEXT: v_readfirstlane_b32 s12, v1 -; GFX6-NEXT: s_addc_u32 s12, s12, 0 -; GFX6-NEXT: s_mul_i32 s0, s14, s0 -; GFX6-NEXT: s_add_u32 s0, s1, s0 -; GFX6-NEXT: s_addc_u32 s12, 0, s12 -; GFX6-NEXT: s_add_u32 s15, s16, s0 -; GFX6-NEXT: s_cselect_b64 s[0:1], -1, 0 -; GFX6-NEXT: s_or_b32 s0, s0, s1 -; GFX6-NEXT: s_addc_u32 s14, s14, s12 +; GFX6-NEXT: v_readfirstlane_b32 s2, v0 +; GFX6-NEXT: s_mul_i32 s13, s0, s12 +; GFX6-NEXT: v_readfirstlane_b32 s16, v2 +; GFX6-NEXT: s_mul_i32 s14, s1, s2 +; GFX6-NEXT: s_mul_i32 s15, s0, s2 +; GFX6-NEXT: s_add_i32 s13, s16, s13 +; GFX6-NEXT: v_mul_hi_u32 v3, v0, s15 +; GFX6-NEXT: s_add_i32 s13, s13, s14 +; GFX6-NEXT: v_mul_hi_u32 v0, v0, s13 +; GFX6-NEXT: v_mul_hi_u32 v4, v1, s15 +; GFX6-NEXT: v_readfirstlane_b32 s14, v3 +; GFX6-NEXT: s_mul_i32 s16, s2, s13 +; GFX6-NEXT: v_mul_hi_u32 v1, v1, s13 +; GFX6-NEXT: s_add_u32 s14, s14, s16 +; GFX6-NEXT: v_readfirstlane_b32 s16, v0 +; GFX6-NEXT: s_mul_i32 s15, s12, s15 +; GFX6-NEXT: s_addc_u32 s16, 0, s16 +; GFX6-NEXT: v_readfirstlane_b32 s17, v4 +; GFX6-NEXT: s_add_u32 s14, s14, s15 +; GFX6-NEXT: s_addc_u32 s14, s16, s17 +; GFX6-NEXT: v_readfirstlane_b32 s15, v1 +; GFX6-NEXT: s_addc_u32 s15, s15, 0 +; GFX6-NEXT: s_mul_i32 s13, s12, s13 +; GFX6-NEXT: s_add_u32 s13, s14, s13 +; GFX6-NEXT: s_addc_u32 s14, 0, s15 +; GFX6-NEXT: s_add_u32 s13, s2, s13 +; GFX6-NEXT: v_mov_b32_e32 v0, s13 +; GFX6-NEXT: v_mul_hi_u32 v0, s0, v0 +; GFX6-NEXT: s_addc_u32 s12, s12, s14 +; GFX6-NEXT: s_mul_i32 s14, s0, s12 +; GFX6-NEXT: s_mul_i32 s1, s1, s13 +; GFX6-NEXT: v_readfirstlane_b32 s15, v0 +; GFX6-NEXT: s_add_i32 s14, s15, s14 +; GFX6-NEXT: s_mul_i32 s0, s0, s13 +; GFX6-NEXT: s_add_i32 s1, s14, s1 +; GFX6-NEXT: v_mov_b32_e32 v2, s0 +; GFX6-NEXT: v_mov_b32_e32 v0, s1 +; GFX6-NEXT: v_mul_hi_u32 v3, s12, v2 +; GFX6-NEXT: v_mul_hi_u32 v2, s13, v2 +; GFX6-NEXT: v_mul_hi_u32 v1, s12, v0 +; GFX6-NEXT: v_mul_hi_u32 v0, s13, v0 +; GFX6-NEXT: s_mul_i32 s15, s13, s1 +; GFX6-NEXT: v_readfirstlane_b32 s17, v2 +; GFX6-NEXT: s_add_u32 s15, s17, s15 +; GFX6-NEXT: v_readfirstlane_b32 s16, v0 +; GFX6-NEXT: s_mul_i32 s0, s12, s0 +; GFX6-NEXT: s_addc_u32 s16, 0, s16 +; GFX6-NEXT: v_readfirstlane_b32 s14, v3 +; GFX6-NEXT: s_add_u32 s0, s15, s0 +; GFX6-NEXT: s_addc_u32 s0, s16, s14 +; GFX6-NEXT: v_readfirstlane_b32 s14, v1 +; GFX6-NEXT: s_addc_u32 s14, s14, 0 +; GFX6-NEXT: s_mul_i32 s1, s12, s1 +; GFX6-NEXT: s_add_u32 s0, s0, s1 +; GFX6-NEXT: s_addc_u32 s1, 0, s14 +; GFX6-NEXT: s_add_u32 s14, s13, s0 +; GFX6-NEXT: s_addc_u32 s15, s12, s1 ; GFX6-NEXT: s_ashr_i32 s12, s7, 31 ; GFX6-NEXT: s_add_u32 s0, s6, s12 ; GFX6-NEXT: s_mov_b32 s13, s12 ; GFX6-NEXT: s_addc_u32 s1, s7, s12 ; GFX6-NEXT: s_xor_b64 s[6:7], s[0:1], s[12:13] -; GFX6-NEXT: v_mov_b32_e32 v0, s14 +; GFX6-NEXT: v_mov_b32_e32 v0, s15 ; GFX6-NEXT: v_mul_hi_u32 v1, s6, v0 -; GFX6-NEXT: v_mov_b32_e32 v2, s15 +; GFX6-NEXT: v_mov_b32_e32 v2, s14 ; GFX6-NEXT: v_mul_hi_u32 v3, s6, v2 ; GFX6-NEXT: s_mov_b32 s0, s4 ; GFX6-NEXT: v_readfirstlane_b32 s4, v1 ; GFX6-NEXT: v_mul_hi_u32 v1, s7, v2 -; GFX6-NEXT: s_mul_i32 s1, s6, s14 +; GFX6-NEXT: s_mul_i32 s1, s6, s15 ; GFX6-NEXT: v_readfirstlane_b32 s16, v3 ; GFX6-NEXT: v_mul_hi_u32 v0, s7, v0 ; GFX6-NEXT: s_add_u32 s1, s16, s1 ; GFX6-NEXT: s_addc_u32 s4, 0, s4 -; GFX6-NEXT: s_mul_i32 s15, s7, s15 +; GFX6-NEXT: s_mul_i32 s14, s7, s14 ; GFX6-NEXT: v_readfirstlane_b32 s16, v1 -; GFX6-NEXT: s_add_u32 s1, s1, s15 +; GFX6-NEXT: s_add_u32 s1, s1, s14 ; GFX6-NEXT: s_addc_u32 s1, s4, s16 ; GFX6-NEXT: v_readfirstlane_b32 s4, v0 ; GFX6-NEXT: s_addc_u32 s4, s4, 0 -; GFX6-NEXT: s_mul_i32 s14, s7, s14 -; GFX6-NEXT: s_add_u32 s16, s1, s14 -; GFX6-NEXT: v_mov_b32_e32 v0, s16 +; GFX6-NEXT: s_mul_i32 s14, s7, s15 +; GFX6-NEXT: s_add_u32 s14, s1, s14 +; GFX6-NEXT: v_mov_b32_e32 v0, s14 ; GFX6-NEXT: v_mul_hi_u32 v0, s10, v0 -; GFX6-NEXT: s_addc_u32 s17, 0, s4 +; GFX6-NEXT: s_addc_u32 s15, 0, s4 ; GFX6-NEXT: s_mov_b32 s1, s5 -; GFX6-NEXT: s_mul_i32 s4, s10, s17 +; GFX6-NEXT: s_mul_i32 s4, s10, s15 ; GFX6-NEXT: v_readfirstlane_b32 s5, v0 ; GFX6-NEXT: s_add_i32 s4, s5, s4 -; GFX6-NEXT: s_mul_i32 s5, s11, s16 -; GFX6-NEXT: s_add_i32 s18, s4, s5 -; GFX6-NEXT: s_sub_i32 s14, s7, s18 -; GFX6-NEXT: s_mul_i32 s4, s10, s16 +; GFX6-NEXT: s_mul_i32 s5, s11, s14 +; GFX6-NEXT: s_add_i32 s16, s4, s5 +; GFX6-NEXT: s_sub_i32 s17, s7, s16 +; GFX6-NEXT: s_mul_i32 s4, s10, s14 ; GFX6-NEXT: s_sub_u32 s6, s6, s4 ; GFX6-NEXT: s_cselect_b64 s[4:5], -1, 0 -; GFX6-NEXT: s_or_b32 s15, s4, s5 -; GFX6-NEXT: s_subb_u32 s19, s14, s11 -; GFX6-NEXT: s_sub_u32 s20, s6, s10 -; GFX6-NEXT: s_cselect_b64 s[14:15], -1, 0 -; GFX6-NEXT: s_or_b32 s14, s14, s15 -; GFX6-NEXT: s_subb_u32 s14, s19, 0 -; GFX6-NEXT: s_cmp_ge_u32 s14, s11 -; GFX6-NEXT: s_cselect_b32 s15, -1, 0 -; GFX6-NEXT: s_cmp_ge_u32 s20, s10 +; GFX6-NEXT: s_subb_u32 s17, s17, s11 +; GFX6-NEXT: s_sub_u32 s18, s6, s10 +; GFX6-NEXT: s_subb_u32 s17, s17, 0 +; GFX6-NEXT: s_cmp_ge_u32 s17, s11 ; GFX6-NEXT: s_cselect_b32 s19, -1, 0 -; GFX6-NEXT: s_cmp_eq_u32 s14, s11 -; GFX6-NEXT: s_cselect_b32 s14, s19, s15 -; GFX6-NEXT: s_add_u32 s15, s16, 1 -; GFX6-NEXT: s_addc_u32 s19, s17, 0 -; GFX6-NEXT: s_add_u32 s20, s16, 2 -; GFX6-NEXT: s_addc_u32 s21, s17, 0 -; GFX6-NEXT: s_cmp_lg_u32 s14, 0 -; GFX6-NEXT: s_cselect_b32 s14, s20, s15 -; GFX6-NEXT: s_cselect_b32 s15, s21, s19 +; GFX6-NEXT: s_cmp_ge_u32 s18, s10 +; GFX6-NEXT: s_cselect_b32 s18, -1, 0 +; GFX6-NEXT: s_cmp_eq_u32 s17, s11 +; GFX6-NEXT: s_cselect_b32 s17, s18, s19 +; GFX6-NEXT: s_add_u32 s18, s14, 1 +; GFX6-NEXT: s_addc_u32 s19, s15, 0 +; GFX6-NEXT: s_add_u32 s20, s14, 2 +; GFX6-NEXT: s_addc_u32 s21, s15, 0 +; GFX6-NEXT: s_cmp_lg_u32 s17, 0 +; GFX6-NEXT: s_cselect_b32 s17, s20, s18 +; GFX6-NEXT: s_cselect_b32 s18, s21, s19 ; GFX6-NEXT: s_or_b32 s4, s4, s5 -; GFX6-NEXT: s_subb_u32 s4, s7, s18 +; GFX6-NEXT: s_subb_u32 s4, s7, s16 ; GFX6-NEXT: s_cmp_ge_u32 s4, s11 ; GFX6-NEXT: s_cselect_b32 s5, -1, 0 ; GFX6-NEXT: s_cmp_ge_u32 s6, s10 @@ -7921,13 +7913,14 @@ define amdgpu_kernel void @sdiv_i64_pow2_shl_denom(ptr addrspace(1) %out, i64 %x ; GFX6-NEXT: s_cmp_eq_u32 s4, s11 ; GFX6-NEXT: s_cselect_b32 s4, s6, s5 ; GFX6-NEXT: s_cmp_lg_u32 s4, 0 -; GFX6-NEXT: s_cselect_b32 s5, s15, s17 -; GFX6-NEXT: s_cselect_b32 s4, s14, s16 +; GFX6-NEXT: s_cselect_b32 s5, s18, s15 +; GFX6-NEXT: s_cselect_b32 s4, s17, s14 ; GFX6-NEXT: s_xor_b64 s[6:7], s[12:13], s[8:9] ; GFX6-NEXT: s_xor_b64 s[4:5], s[4:5], s[6:7] ; GFX6-NEXT: s_sub_u32 s4, s4, s6 ; GFX6-NEXT: s_subb_u32 s5, s5, s7 ; GFX6-NEXT: v_mov_b32_e32 v0, s4 +; GFX6-NEXT: s_mov_b32 s2, -1 ; GFX6-NEXT: v_mov_b32_e32 v1, s5 ; GFX6-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 ; GFX6-NEXT: s_endpgm @@ -8278,8 +8271,8 @@ define amdgpu_kernel void @sdiv_v2i64_pow2_shl_denom(ptr addrspace(1) %out, <2 x ; GFX6-NEXT: s_xor_b64 s[6:7], s[6:7], s[2:3] ; GFX6-NEXT: v_cvt_f32_u32_e32 v0, s6 ; GFX6-NEXT: v_cvt_f32_u32_e32 v1, s7 -; GFX6-NEXT: s_sub_u32 s14, 0, s6 -; GFX6-NEXT: s_subb_u32 s15, 0, s7 +; GFX6-NEXT: s_sub_u32 s12, 0, s6 +; GFX6-NEXT: s_subb_u32 s13, 0, s7 ; GFX6-NEXT: v_mac_f32_e32 v0, 0x4f800000, v1 ; GFX6-NEXT: v_rcp_f32_e32 v0, v0 ; GFX6-NEXT: v_mul_f32_e32 v0, 0x5f7ffffc, v0 @@ -8288,69 +8281,65 @@ define amdgpu_kernel void @sdiv_v2i64_pow2_shl_denom(ptr addrspace(1) %out, <2 x ; GFX6-NEXT: v_mac_f32_e32 v0, 0xcf800000, v1 ; GFX6-NEXT: v_cvt_u32_f32_e32 v0, v0 ; GFX6-NEXT: v_cvt_u32_f32_e32 v1, v1 -; GFX6-NEXT: v_mul_hi_u32 v2, s14, v0 -; GFX6-NEXT: v_readfirstlane_b32 s16, v1 -; GFX6-NEXT: v_readfirstlane_b32 s12, v0 -; GFX6-NEXT: s_mul_i32 s13, s14, s16 +; GFX6-NEXT: v_mul_hi_u32 v2, s12, v0 +; GFX6-NEXT: v_readfirstlane_b32 s14, v1 +; GFX6-NEXT: v_readfirstlane_b32 s15, v0 +; GFX6-NEXT: s_mul_i32 s16, s12, s14 ; GFX6-NEXT: v_readfirstlane_b32 s19, v2 -; GFX6-NEXT: s_mul_i32 s17, s15, s12 -; GFX6-NEXT: s_mul_i32 s18, s14, s12 -; GFX6-NEXT: s_add_i32 s13, s19, s13 +; GFX6-NEXT: s_mul_i32 s17, s13, s15 +; GFX6-NEXT: s_mul_i32 s18, s12, s15 +; GFX6-NEXT: s_add_i32 s16, s19, s16 ; GFX6-NEXT: v_mul_hi_u32 v3, v0, s18 -; GFX6-NEXT: s_add_i32 s13, s13, s17 -; GFX6-NEXT: v_mul_hi_u32 v0, v0, s13 +; GFX6-NEXT: s_add_i32 s16, s16, s17 +; GFX6-NEXT: v_mul_hi_u32 v0, v0, s16 ; GFX6-NEXT: v_mul_hi_u32 v4, v1, s18 ; GFX6-NEXT: v_readfirstlane_b32 s17, v3 -; GFX6-NEXT: s_mul_i32 s20, s12, s13 -; GFX6-NEXT: v_mul_hi_u32 v1, v1, s13 +; GFX6-NEXT: s_mul_i32 s20, s15, s16 +; GFX6-NEXT: v_mul_hi_u32 v1, v1, s16 ; GFX6-NEXT: s_add_u32 s17, s17, s20 ; GFX6-NEXT: v_readfirstlane_b32 s20, v0 -; GFX6-NEXT: s_mul_i32 s18, s16, s18 +; GFX6-NEXT: s_mul_i32 s18, s14, s18 ; GFX6-NEXT: s_addc_u32 s20, 0, s20 ; GFX6-NEXT: v_readfirstlane_b32 s19, v4 ; GFX6-NEXT: s_add_u32 s17, s17, s18 ; GFX6-NEXT: s_addc_u32 s17, s20, s19 ; GFX6-NEXT: v_readfirstlane_b32 s18, v1 ; GFX6-NEXT: s_addc_u32 s18, s18, 0 -; GFX6-NEXT: s_mul_i32 s13, s16, s13 -; GFX6-NEXT: s_add_u32 s13, s17, s13 +; GFX6-NEXT: s_mul_i32 s16, s14, s16 +; GFX6-NEXT: s_add_u32 s16, s17, s16 ; GFX6-NEXT: s_addc_u32 s17, 0, s18 -; GFX6-NEXT: s_add_u32 s18, s12, s13 -; GFX6-NEXT: v_mov_b32_e32 v0, s18 -; GFX6-NEXT: v_mul_hi_u32 v0, s14, v0 -; GFX6-NEXT: s_cselect_b64 s[12:13], -1, 0 -; GFX6-NEXT: s_or_b32 s12, s12, s13 -; GFX6-NEXT: s_addc_u32 s16, s16, s17 -; GFX6-NEXT: s_mul_i32 s12, s14, s16 -; GFX6-NEXT: v_readfirstlane_b32 s13, v0 -; GFX6-NEXT: s_add_i32 s12, s13, s12 -; GFX6-NEXT: s_mul_i32 s15, s15, s18 -; GFX6-NEXT: s_mul_i32 s13, s14, s18 -; GFX6-NEXT: s_add_i32 s12, s12, s15 -; GFX6-NEXT: v_mov_b32_e32 v2, s13 -; GFX6-NEXT: v_mov_b32_e32 v0, s12 -; GFX6-NEXT: v_mul_hi_u32 v3, s16, v2 -; GFX6-NEXT: v_mul_hi_u32 v2, s18, v2 -; GFX6-NEXT: v_mul_hi_u32 v1, s16, v0 -; GFX6-NEXT: v_mul_hi_u32 v0, s18, v0 -; GFX6-NEXT: s_mul_i32 s15, s18, s12 -; GFX6-NEXT: v_readfirstlane_b32 s19, v2 -; GFX6-NEXT: s_add_u32 s15, s19, s15 +; GFX6-NEXT: s_add_u32 s15, s15, s16 +; GFX6-NEXT: v_mov_b32_e32 v0, s15 +; GFX6-NEXT: v_mul_hi_u32 v0, s12, v0 +; GFX6-NEXT: s_addc_u32 s14, s14, s17 +; GFX6-NEXT: s_mul_i32 s16, s12, s14 +; GFX6-NEXT: s_mul_i32 s13, s13, s15 ; GFX6-NEXT: v_readfirstlane_b32 s17, v0 -; GFX6-NEXT: s_mul_i32 s13, s16, s13 -; GFX6-NEXT: s_addc_u32 s17, 0, s17 -; GFX6-NEXT: v_readfirstlane_b32 s14, v3 -; GFX6-NEXT: s_add_u32 s13, s15, s13 -; GFX6-NEXT: s_addc_u32 s13, s17, s14 -; GFX6-NEXT: v_readfirstlane_b32 s14, v1 -; GFX6-NEXT: s_addc_u32 s14, s14, 0 -; GFX6-NEXT: s_mul_i32 s12, s16, s12 -; GFX6-NEXT: s_add_u32 s12, s13, s12 -; GFX6-NEXT: s_addc_u32 s14, 0, s14 -; GFX6-NEXT: s_add_u32 s15, s18, s12 -; GFX6-NEXT: s_cselect_b64 s[12:13], -1, 0 -; GFX6-NEXT: s_or_b32 s12, s12, s13 -; GFX6-NEXT: s_addc_u32 s14, s16, s14 +; GFX6-NEXT: s_add_i32 s16, s17, s16 +; GFX6-NEXT: s_mul_i32 s12, s12, s15 +; GFX6-NEXT: s_add_i32 s13, s16, s13 +; GFX6-NEXT: v_mov_b32_e32 v2, s12 +; GFX6-NEXT: v_mov_b32_e32 v0, s13 +; GFX6-NEXT: v_mul_hi_u32 v3, s14, v2 +; GFX6-NEXT: v_mul_hi_u32 v2, s15, v2 +; GFX6-NEXT: v_mul_hi_u32 v1, s14, v0 +; GFX6-NEXT: v_mul_hi_u32 v0, s15, v0 +; GFX6-NEXT: s_mul_i32 s17, s15, s13 +; GFX6-NEXT: v_readfirstlane_b32 s19, v2 +; GFX6-NEXT: s_add_u32 s17, s19, s17 +; GFX6-NEXT: v_readfirstlane_b32 s18, v0 +; GFX6-NEXT: s_mul_i32 s12, s14, s12 +; GFX6-NEXT: s_addc_u32 s18, 0, s18 +; GFX6-NEXT: v_readfirstlane_b32 s16, v3 +; GFX6-NEXT: s_add_u32 s12, s17, s12 +; GFX6-NEXT: s_addc_u32 s12, s18, s16 +; GFX6-NEXT: v_readfirstlane_b32 s16, v1 +; GFX6-NEXT: s_addc_u32 s16, s16, 0 +; GFX6-NEXT: s_mul_i32 s13, s14, s13 +; GFX6-NEXT: s_add_u32 s12, s12, s13 +; GFX6-NEXT: s_addc_u32 s13, 0, s16 +; GFX6-NEXT: s_add_u32 s15, s15, s12 +; GFX6-NEXT: s_addc_u32 s14, s14, s13 ; GFX6-NEXT: s_ashr_i32 s12, s9, 31 ; GFX6-NEXT: s_add_u32 s8, s8, s12 ; GFX6-NEXT: s_mov_b32 s13, s12 @@ -8374,40 +8363,37 @@ define amdgpu_kernel void @sdiv_v2i64_pow2_shl_denom(ptr addrspace(1) %out, <2 x ; GFX6-NEXT: v_readfirstlane_b32 s16, v0 ; GFX6-NEXT: s_addc_u32 s16, s16, 0 ; GFX6-NEXT: s_mul_i32 s14, s9, s14 -; GFX6-NEXT: s_add_u32 s18, s15, s14 -; GFX6-NEXT: v_mov_b32_e32 v0, s18 +; GFX6-NEXT: s_add_u32 s17, s15, s14 +; GFX6-NEXT: v_mov_b32_e32 v0, s17 ; GFX6-NEXT: v_mul_hi_u32 v0, s6, v0 -; GFX6-NEXT: s_addc_u32 s19, 0, s16 -; GFX6-NEXT: s_mul_i32 s14, s6, s19 +; GFX6-NEXT: s_addc_u32 s16, 0, s16 +; GFX6-NEXT: s_mul_i32 s14, s6, s16 ; GFX6-NEXT: v_readfirstlane_b32 s15, v0 ; GFX6-NEXT: s_add_i32 s14, s15, s14 -; GFX6-NEXT: s_mul_i32 s15, s7, s18 -; GFX6-NEXT: s_add_i32 s20, s14, s15 -; GFX6-NEXT: s_sub_i32 s16, s9, s20 -; GFX6-NEXT: s_mul_i32 s14, s6, s18 +; GFX6-NEXT: s_mul_i32 s15, s7, s17 +; GFX6-NEXT: s_add_i32 s18, s14, s15 +; GFX6-NEXT: s_sub_i32 s19, s9, s18 +; GFX6-NEXT: s_mul_i32 s14, s6, s17 ; GFX6-NEXT: s_sub_u32 s8, s8, s14 ; GFX6-NEXT: s_cselect_b64 s[14:15], -1, 0 -; GFX6-NEXT: s_or_b32 s17, s14, s15 -; GFX6-NEXT: s_subb_u32 s21, s16, s7 -; GFX6-NEXT: s_sub_u32 s22, s8, s6 -; GFX6-NEXT: s_cselect_b64 s[16:17], -1, 0 -; GFX6-NEXT: s_or_b32 s16, s16, s17 -; GFX6-NEXT: s_subb_u32 s16, s21, 0 -; GFX6-NEXT: s_cmp_ge_u32 s16, s7 -; GFX6-NEXT: s_cselect_b32 s17, -1, 0 -; GFX6-NEXT: s_cmp_ge_u32 s22, s6 +; GFX6-NEXT: s_subb_u32 s19, s19, s7 +; GFX6-NEXT: s_sub_u32 s20, s8, s6 +; GFX6-NEXT: s_subb_u32 s19, s19, 0 +; GFX6-NEXT: s_cmp_ge_u32 s19, s7 ; GFX6-NEXT: s_cselect_b32 s21, -1, 0 -; GFX6-NEXT: s_cmp_eq_u32 s16, s7 -; GFX6-NEXT: s_cselect_b32 s16, s21, s17 -; GFX6-NEXT: s_add_u32 s17, s18, 1 -; GFX6-NEXT: s_addc_u32 s21, s19, 0 -; GFX6-NEXT: s_add_u32 s22, s18, 2 -; GFX6-NEXT: s_addc_u32 s23, s19, 0 -; GFX6-NEXT: s_cmp_lg_u32 s16, 0 -; GFX6-NEXT: s_cselect_b32 s16, s22, s17 -; GFX6-NEXT: s_cselect_b32 s17, s23, s21 +; GFX6-NEXT: s_cmp_ge_u32 s20, s6 +; GFX6-NEXT: s_cselect_b32 s20, -1, 0 +; GFX6-NEXT: s_cmp_eq_u32 s19, s7 +; GFX6-NEXT: s_cselect_b32 s19, s20, s21 +; GFX6-NEXT: s_add_u32 s20, s17, 1 +; GFX6-NEXT: s_addc_u32 s21, s16, 0 +; GFX6-NEXT: s_add_u32 s22, s17, 2 +; GFX6-NEXT: s_addc_u32 s23, s16, 0 +; GFX6-NEXT: s_cmp_lg_u32 s19, 0 +; GFX6-NEXT: s_cselect_b32 s19, s22, s20 +; GFX6-NEXT: s_cselect_b32 s20, s23, s21 ; GFX6-NEXT: s_or_b32 s14, s14, s15 -; GFX6-NEXT: s_subb_u32 s9, s9, s20 +; GFX6-NEXT: s_subb_u32 s9, s9, s18 ; GFX6-NEXT: s_cmp_ge_u32 s9, s7 ; GFX6-NEXT: s_cselect_b32 s14, -1, 0 ; GFX6-NEXT: s_cmp_ge_u32 s8, s6 @@ -8415,12 +8401,12 @@ define amdgpu_kernel void @sdiv_v2i64_pow2_shl_denom(ptr addrspace(1) %out, <2 x ; GFX6-NEXT: s_cmp_eq_u32 s9, s7 ; GFX6-NEXT: s_cselect_b32 s6, s6, s14 ; GFX6-NEXT: s_cmp_lg_u32 s6, 0 -; GFX6-NEXT: s_cselect_b32 s7, s17, s19 -; GFX6-NEXT: s_cselect_b32 s6, s16, s18 +; GFX6-NEXT: s_cselect_b32 s7, s20, s16 +; GFX6-NEXT: s_cselect_b32 s6, s19, s17 ; GFX6-NEXT: s_xor_b64 s[2:3], s[12:13], s[2:3] ; GFX6-NEXT: s_xor_b64 s[6:7], s[6:7], s[2:3] -; GFX6-NEXT: s_sub_u32 s16, s6, s2 -; GFX6-NEXT: s_subb_u32 s17, s7, s3 +; GFX6-NEXT: s_sub_u32 s14, s6, s2 +; GFX6-NEXT: s_subb_u32 s15, s7, s3 ; GFX6-NEXT: s_ashr_i32 s6, s1, 31 ; GFX6-NEXT: s_add_u32 s0, s0, s6 ; GFX6-NEXT: s_mov_b32 s7, s6 @@ -8428,8 +8414,8 @@ define amdgpu_kernel void @sdiv_v2i64_pow2_shl_denom(ptr addrspace(1) %out, <2 x ; GFX6-NEXT: s_xor_b64 s[8:9], s[0:1], s[6:7] ; GFX6-NEXT: v_cvt_f32_u32_e32 v0, s8 ; GFX6-NEXT: v_cvt_f32_u32_e32 v1, s9 -; GFX6-NEXT: s_sub_u32 s12, 0, s8 -; GFX6-NEXT: s_subb_u32 s13, 0, s9 +; GFX6-NEXT: s_sub_u32 s2, 0, s8 +; GFX6-NEXT: s_subb_u32 s3, 0, s9 ; GFX6-NEXT: v_mac_f32_e32 v0, 0x4f800000, v1 ; GFX6-NEXT: v_rcp_f32_e32 v0, v0 ; GFX6-NEXT: v_mul_f32_e32 v0, 0x5f7ffffc, v0 @@ -8438,128 +8424,121 @@ define amdgpu_kernel void @sdiv_v2i64_pow2_shl_denom(ptr addrspace(1) %out, <2 x ; GFX6-NEXT: v_mac_f32_e32 v0, 0xcf800000, v1 ; GFX6-NEXT: v_cvt_u32_f32_e32 v0, v0 ; GFX6-NEXT: v_cvt_u32_f32_e32 v1, v1 -; GFX6-NEXT: v_mul_hi_u32 v2, s12, v0 -; GFX6-NEXT: v_readfirstlane_b32 s14, v1 -; GFX6-NEXT: v_readfirstlane_b32 s2, v0 -; GFX6-NEXT: s_mul_i32 s1, s12, s14 -; GFX6-NEXT: v_readfirstlane_b32 s3, v2 -; GFX6-NEXT: s_mul_i32 s0, s13, s2 -; GFX6-NEXT: s_add_i32 s1, s3, s1 -; GFX6-NEXT: s_add_i32 s3, s1, s0 -; GFX6-NEXT: s_mul_i32 s15, s12, s2 -; GFX6-NEXT: v_mul_hi_u32 v2, v0, s3 -; GFX6-NEXT: v_mul_hi_u32 v0, v0, s15 -; GFX6-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9 -; GFX6-NEXT: s_mul_i32 s4, s2, s3 -; GFX6-NEXT: v_readfirstlane_b32 s5, v2 +; GFX6-NEXT: v_mul_hi_u32 v2, s2, v0 +; GFX6-NEXT: v_readfirstlane_b32 s12, v1 +; GFX6-NEXT: v_readfirstlane_b32 s0, v0 +; GFX6-NEXT: s_mul_i32 s13, s2, s12 +; GFX6-NEXT: v_readfirstlane_b32 s16, v2 +; GFX6-NEXT: s_mul_i32 s1, s3, s0 +; GFX6-NEXT: s_add_i32 s13, s16, s13 +; GFX6-NEXT: s_add_i32 s13, s13, s1 +; GFX6-NEXT: s_mul_i32 s1, s2, s0 +; GFX6-NEXT: v_mul_hi_u32 v2, v0, s13 +; GFX6-NEXT: v_mul_hi_u32 v0, v0, s1 +; GFX6-NEXT: s_mul_i32 s16, s0, s13 +; GFX6-NEXT: v_readfirstlane_b32 s17, v2 ; GFX6-NEXT: v_readfirstlane_b32 s18, v0 -; GFX6-NEXT: v_mul_hi_u32 v0, v1, s15 -; GFX6-NEXT: v_mul_hi_u32 v1, v1, s3 -; GFX6-NEXT: s_add_u32 s4, s18, s4 -; GFX6-NEXT: s_addc_u32 s5, 0, s5 -; GFX6-NEXT: s_mul_i32 s15, s14, s15 +; GFX6-NEXT: v_mul_hi_u32 v0, v1, s1 +; GFX6-NEXT: v_mul_hi_u32 v1, v1, s13 +; GFX6-NEXT: s_add_u32 s16, s18, s16 +; GFX6-NEXT: s_addc_u32 s17, 0, s17 +; GFX6-NEXT: s_mul_i32 s1, s12, s1 ; GFX6-NEXT: v_readfirstlane_b32 s18, v0 -; GFX6-NEXT: s_add_u32 s4, s4, s15 -; GFX6-NEXT: s_addc_u32 s4, s5, s18 -; GFX6-NEXT: v_readfirstlane_b32 s5, v1 -; GFX6-NEXT: s_addc_u32 s5, s5, 0 -; GFX6-NEXT: s_mul_i32 s3, s14, s3 -; GFX6-NEXT: s_add_u32 s3, s4, s3 -; GFX6-NEXT: s_addc_u32 s4, 0, s5 -; GFX6-NEXT: s_add_u32 s5, s2, s3 -; GFX6-NEXT: v_mov_b32_e32 v0, s5 -; GFX6-NEXT: v_mul_hi_u32 v0, s12, v0 -; GFX6-NEXT: s_cselect_b64 s[2:3], -1, 0 -; GFX6-NEXT: s_or_b32 s2, s2, s3 -; GFX6-NEXT: s_addc_u32 s4, s14, s4 -; GFX6-NEXT: s_mul_i32 s2, s12, s4 -; GFX6-NEXT: v_readfirstlane_b32 s3, v0 -; GFX6-NEXT: s_add_i32 s2, s3, s2 -; GFX6-NEXT: s_mul_i32 s13, s13, s5 -; GFX6-NEXT: s_mul_i32 s3, s12, s5 -; GFX6-NEXT: s_add_i32 s2, s2, s13 -; GFX6-NEXT: v_mov_b32_e32 v2, s3 -; GFX6-NEXT: v_mov_b32_e32 v0, s2 +; GFX6-NEXT: s_add_u32 s1, s16, s1 +; GFX6-NEXT: s_addc_u32 s1, s17, s18 +; GFX6-NEXT: v_readfirstlane_b32 s16, v1 +; GFX6-NEXT: s_addc_u32 s16, s16, 0 +; GFX6-NEXT: s_mul_i32 s13, s12, s13 +; GFX6-NEXT: s_add_u32 s1, s1, s13 +; GFX6-NEXT: s_addc_u32 s13, 0, s16 +; GFX6-NEXT: s_add_u32 s16, s0, s1 +; GFX6-NEXT: v_mov_b32_e32 v0, s16 +; GFX6-NEXT: v_mul_hi_u32 v0, s2, v0 +; GFX6-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9 +; GFX6-NEXT: s_addc_u32 s4, s12, s13 +; GFX6-NEXT: s_mul_i32 s5, s2, s4 +; GFX6-NEXT: v_readfirstlane_b32 s12, v0 +; GFX6-NEXT: s_add_i32 s5, s12, s5 +; GFX6-NEXT: s_mul_i32 s3, s3, s16 +; GFX6-NEXT: s_mul_i32 s2, s2, s16 +; GFX6-NEXT: s_add_i32 s3, s5, s3 +; GFX6-NEXT: v_mov_b32_e32 v2, s2 +; GFX6-NEXT: v_mov_b32_e32 v0, s3 ; GFX6-NEXT: v_mul_hi_u32 v3, s4, v2 -; GFX6-NEXT: v_mul_hi_u32 v2, s5, v2 +; GFX6-NEXT: v_mul_hi_u32 v2, s16, v2 ; GFX6-NEXT: v_mul_hi_u32 v1, s4, v0 -; GFX6-NEXT: v_mul_hi_u32 v0, s5, v0 -; GFX6-NEXT: s_mul_i32 s13, s5, s2 -; GFX6-NEXT: v_readfirstlane_b32 s15, v2 -; GFX6-NEXT: s_add_u32 s13, s15, s13 -; GFX6-NEXT: v_readfirstlane_b32 s14, v0 -; GFX6-NEXT: s_mul_i32 s3, s4, s3 -; GFX6-NEXT: s_addc_u32 s14, 0, s14 -; GFX6-NEXT: v_readfirstlane_b32 s12, v3 -; GFX6-NEXT: s_add_u32 s3, s13, s3 -; GFX6-NEXT: s_addc_u32 s3, s14, s12 -; GFX6-NEXT: v_readfirstlane_b32 s12, v1 -; GFX6-NEXT: s_addc_u32 s12, s12, 0 +; GFX6-NEXT: v_mul_hi_u32 v0, s16, v0 +; GFX6-NEXT: s_mul_i32 s12, s16, s3 +; GFX6-NEXT: v_readfirstlane_b32 s17, v2 +; GFX6-NEXT: s_add_u32 s12, s17, s12 +; GFX6-NEXT: v_readfirstlane_b32 s13, v0 ; GFX6-NEXT: s_mul_i32 s2, s4, s2 -; GFX6-NEXT: s_add_u32 s2, s3, s2 -; GFX6-NEXT: s_addc_u32 s12, 0, s12 -; GFX6-NEXT: s_add_u32 s13, s5, s2 -; GFX6-NEXT: s_cselect_b64 s[2:3], -1, 0 -; GFX6-NEXT: s_or_b32 s2, s2, s3 -; GFX6-NEXT: s_addc_u32 s12, s4, s12 +; GFX6-NEXT: s_addc_u32 s13, 0, s13 +; GFX6-NEXT: v_readfirstlane_b32 s5, v3 +; GFX6-NEXT: s_add_u32 s2, s12, s2 +; GFX6-NEXT: s_addc_u32 s2, s13, s5 +; GFX6-NEXT: v_readfirstlane_b32 s5, v1 +; GFX6-NEXT: s_addc_u32 s5, s5, 0 +; GFX6-NEXT: s_mul_i32 s3, s4, s3 +; GFX6-NEXT: s_add_u32 s2, s2, s3 +; GFX6-NEXT: s_addc_u32 s3, 0, s5 +; GFX6-NEXT: s_add_u32 s12, s16, s2 +; GFX6-NEXT: s_addc_u32 s13, s4, s3 ; GFX6-NEXT: s_ashr_i32 s4, s11, 31 ; GFX6-NEXT: s_add_u32 s2, s10, s4 ; GFX6-NEXT: s_mov_b32 s5, s4 ; GFX6-NEXT: s_addc_u32 s3, s11, s4 ; GFX6-NEXT: s_xor_b64 s[10:11], s[2:3], s[4:5] -; GFX6-NEXT: v_mov_b32_e32 v0, s12 +; GFX6-NEXT: v_mov_b32_e32 v0, s13 ; GFX6-NEXT: v_mul_hi_u32 v1, s10, v0 -; GFX6-NEXT: v_mov_b32_e32 v2, s13 +; GFX6-NEXT: v_mov_b32_e32 v2, s12 ; GFX6-NEXT: v_mul_hi_u32 v3, s10, v2 -; GFX6-NEXT: s_mul_i32 s2, s10, s12 -; GFX6-NEXT: v_readfirstlane_b32 s14, v1 +; GFX6-NEXT: s_mul_i32 s2, s10, s13 +; GFX6-NEXT: v_readfirstlane_b32 s16, v1 ; GFX6-NEXT: v_mul_hi_u32 v1, s11, v2 -; GFX6-NEXT: v_readfirstlane_b32 s15, v3 +; GFX6-NEXT: v_readfirstlane_b32 s17, v3 ; GFX6-NEXT: v_mul_hi_u32 v0, s11, v0 -; GFX6-NEXT: s_add_u32 s2, s15, s2 -; GFX6-NEXT: s_addc_u32 s14, 0, s14 -; GFX6-NEXT: s_mul_i32 s13, s11, s13 -; GFX6-NEXT: v_readfirstlane_b32 s15, v1 -; GFX6-NEXT: s_add_u32 s2, s2, s13 -; GFX6-NEXT: s_addc_u32 s2, s14, s15 -; GFX6-NEXT: v_readfirstlane_b32 s13, v0 -; GFX6-NEXT: s_addc_u32 s13, s13, 0 +; GFX6-NEXT: s_add_u32 s2, s17, s2 +; GFX6-NEXT: s_addc_u32 s16, 0, s16 ; GFX6-NEXT: s_mul_i32 s12, s11, s12 -; GFX6-NEXT: s_add_u32 s18, s2, s12 -; GFX6-NEXT: v_mov_b32_e32 v0, s18 +; GFX6-NEXT: v_readfirstlane_b32 s17, v1 +; GFX6-NEXT: s_add_u32 s2, s2, s12 +; GFX6-NEXT: s_addc_u32 s2, s16, s17 +; GFX6-NEXT: v_readfirstlane_b32 s12, v0 +; GFX6-NEXT: s_addc_u32 s12, s12, 0 +; GFX6-NEXT: s_mul_i32 s13, s11, s13 +; GFX6-NEXT: s_add_u32 s16, s2, s13 +; GFX6-NEXT: v_mov_b32_e32 v0, s16 ; GFX6-NEXT: v_mul_hi_u32 v0, s8, v0 -; GFX6-NEXT: s_addc_u32 s19, 0, s13 -; GFX6-NEXT: s_mul_i32 s12, s8, s19 +; GFX6-NEXT: s_addc_u32 s17, 0, s12 +; GFX6-NEXT: s_mul_i32 s12, s8, s17 ; GFX6-NEXT: s_mov_b32 s3, 0xf000 ; GFX6-NEXT: v_readfirstlane_b32 s13, v0 ; GFX6-NEXT: s_add_i32 s12, s13, s12 -; GFX6-NEXT: s_mul_i32 s13, s9, s18 -; GFX6-NEXT: s_add_i32 s20, s12, s13 -; GFX6-NEXT: s_sub_i32 s14, s11, s20 -; GFX6-NEXT: s_mul_i32 s12, s8, s18 +; GFX6-NEXT: s_mul_i32 s13, s9, s16 +; GFX6-NEXT: s_add_i32 s18, s12, s13 +; GFX6-NEXT: s_sub_i32 s19, s11, s18 +; GFX6-NEXT: s_mul_i32 s12, s8, s16 ; GFX6-NEXT: s_sub_u32 s10, s10, s12 ; GFX6-NEXT: s_cselect_b64 s[12:13], -1, 0 -; GFX6-NEXT: s_or_b32 s15, s12, s13 -; GFX6-NEXT: s_subb_u32 s21, s14, s9 -; GFX6-NEXT: s_sub_u32 s22, s10, s8 -; GFX6-NEXT: s_cselect_b64 s[14:15], -1, 0 -; GFX6-NEXT: s_or_b32 s14, s14, s15 -; GFX6-NEXT: s_subb_u32 s14, s21, 0 -; GFX6-NEXT: s_cmp_ge_u32 s14, s9 -; GFX6-NEXT: s_cselect_b32 s15, -1, 0 -; GFX6-NEXT: s_cmp_ge_u32 s22, s8 +; GFX6-NEXT: s_subb_u32 s19, s19, s9 +; GFX6-NEXT: s_sub_u32 s20, s10, s8 +; GFX6-NEXT: s_subb_u32 s19, s19, 0 +; GFX6-NEXT: s_cmp_ge_u32 s19, s9 ; GFX6-NEXT: s_cselect_b32 s21, -1, 0 -; GFX6-NEXT: s_cmp_eq_u32 s14, s9 -; GFX6-NEXT: s_cselect_b32 s14, s21, s15 -; GFX6-NEXT: s_add_u32 s15, s18, 1 -; GFX6-NEXT: s_addc_u32 s21, s19, 0 -; GFX6-NEXT: s_add_u32 s22, s18, 2 -; GFX6-NEXT: s_addc_u32 s23, s19, 0 -; GFX6-NEXT: s_cmp_lg_u32 s14, 0 -; GFX6-NEXT: s_cselect_b32 s14, s22, s15 -; GFX6-NEXT: s_cselect_b32 s15, s23, s21 +; GFX6-NEXT: s_cmp_ge_u32 s20, s8 +; GFX6-NEXT: s_cselect_b32 s20, -1, 0 +; GFX6-NEXT: s_cmp_eq_u32 s19, s9 +; GFX6-NEXT: s_cselect_b32 s19, s20, s21 +; GFX6-NEXT: s_add_u32 s20, s16, 1 +; GFX6-NEXT: s_addc_u32 s21, s17, 0 +; GFX6-NEXT: s_add_u32 s22, s16, 2 +; GFX6-NEXT: s_addc_u32 s23, s17, 0 +; GFX6-NEXT: s_cmp_lg_u32 s19, 0 +; GFX6-NEXT: s_cselect_b32 s19, s22, s20 +; GFX6-NEXT: s_cselect_b32 s20, s23, s21 ; GFX6-NEXT: s_or_b32 s12, s12, s13 -; GFX6-NEXT: s_subb_u32 s11, s11, s20 +; GFX6-NEXT: s_subb_u32 s11, s11, s18 ; GFX6-NEXT: s_cmp_ge_u32 s11, s9 ; GFX6-NEXT: s_cselect_b32 s12, -1, 0 ; GFX6-NEXT: s_cmp_ge_u32 s10, s8 @@ -8567,15 +8546,15 @@ define amdgpu_kernel void @sdiv_v2i64_pow2_shl_denom(ptr addrspace(1) %out, <2 x ; GFX6-NEXT: s_cmp_eq_u32 s11, s9 ; GFX6-NEXT: s_cselect_b32 s8, s8, s12 ; GFX6-NEXT: s_cmp_lg_u32 s8, 0 -; GFX6-NEXT: s_cselect_b32 s9, s15, s19 -; GFX6-NEXT: s_cselect_b32 s8, s14, s18 +; GFX6-NEXT: s_cselect_b32 s9, s20, s17 +; GFX6-NEXT: s_cselect_b32 s8, s19, s16 ; GFX6-NEXT: s_xor_b64 s[4:5], s[4:5], s[6:7] ; GFX6-NEXT: s_xor_b64 s[6:7], s[8:9], s[4:5] ; GFX6-NEXT: s_sub_u32 s4, s6, s4 ; GFX6-NEXT: s_subb_u32 s5, s7, s5 ; GFX6-NEXT: s_mov_b32 s2, -1 -; GFX6-NEXT: v_mov_b32_e32 v0, s16 -; GFX6-NEXT: v_mov_b32_e32 v1, s17 +; GFX6-NEXT: v_mov_b32_e32 v0, s14 +; GFX6-NEXT: v_mov_b32_e32 v1, s15 ; GFX6-NEXT: v_mov_b32_e32 v2, s4 ; GFX6-NEXT: v_mov_b32_e32 v3, s5 ; GFX6-NEXT: s_waitcnt lgkmcnt(0) @@ -9015,105 +8994,100 @@ define amdgpu_kernel void @srem_i64_pow2_shl_denom(ptr addrspace(1) %out, i64 %x ; GFX6-NEXT: s_xor_b64 s[8:9], s[0:1], s[2:3] ; GFX6-NEXT: v_cvt_f32_u32_e32 v0, s8 ; GFX6-NEXT: v_cvt_f32_u32_e32 v1, s9 -; GFX6-NEXT: s_sub_u32 s10, 0, s8 -; GFX6-NEXT: s_subb_u32 s11, 0, s9 +; GFX6-NEXT: s_sub_u32 s0, 0, s8 +; GFX6-NEXT: s_subb_u32 s1, 0, s9 ; GFX6-NEXT: s_mov_b32 s3, 0xf000 ; GFX6-NEXT: v_madmk_f32 v0, v1, 0x4f800000, v0 ; GFX6-NEXT: v_rcp_f32_e32 v0, v0 -; GFX6-NEXT: s_mov_b32 s2, -1 ; GFX6-NEXT: v_mul_f32_e32 v0, 0x5f7ffffc, v0 ; GFX6-NEXT: v_mul_f32_e32 v1, 0x2f800000, v0 ; GFX6-NEXT: v_trunc_f32_e32 v1, v1 ; GFX6-NEXT: v_madmk_f32 v0, v1, 0xcf800000, v0 ; GFX6-NEXT: v_cvt_u32_f32_e32 v0, v0 ; GFX6-NEXT: v_cvt_u32_f32_e32 v1, v1 -; GFX6-NEXT: v_mul_hi_u32 v2, s10, v0 -; GFX6-NEXT: v_readfirstlane_b32 s12, v1 -; GFX6-NEXT: v_readfirstlane_b32 s0, v0 -; GFX6-NEXT: s_mul_i32 s1, s10, s12 -; GFX6-NEXT: v_readfirstlane_b32 s15, v2 -; GFX6-NEXT: s_mul_i32 s13, s11, s0 -; GFX6-NEXT: s_mul_i32 s14, s10, s0 -; GFX6-NEXT: s_add_i32 s1, s15, s1 -; GFX6-NEXT: v_mul_hi_u32 v3, v0, s14 -; GFX6-NEXT: s_add_i32 s1, s1, s13 -; GFX6-NEXT: v_mul_hi_u32 v0, v0, s1 -; GFX6-NEXT: v_mul_hi_u32 v4, v1, s14 -; GFX6-NEXT: v_readfirstlane_b32 s13, v3 -; GFX6-NEXT: s_mul_i32 s15, s0, s1 -; GFX6-NEXT: v_mul_hi_u32 v1, v1, s1 -; GFX6-NEXT: s_add_u32 s13, s13, s15 -; GFX6-NEXT: v_readfirstlane_b32 s15, v0 -; GFX6-NEXT: s_addc_u32 s15, 0, s15 -; GFX6-NEXT: s_mul_i32 s14, s12, s14 -; GFX6-NEXT: v_readfirstlane_b32 s16, v4 -; GFX6-NEXT: s_add_u32 s13, s13, s14 -; GFX6-NEXT: s_addc_u32 s13, s15, s16 -; GFX6-NEXT: v_readfirstlane_b32 s14, v1 -; GFX6-NEXT: s_addc_u32 s14, s14, 0 -; GFX6-NEXT: s_mul_i32 s1, s12, s1 -; GFX6-NEXT: s_add_u32 s1, s13, s1 -; GFX6-NEXT: s_addc_u32 s13, 0, s14 -; GFX6-NEXT: s_add_u32 s14, s0, s1 -; GFX6-NEXT: v_mov_b32_e32 v0, s14 -; GFX6-NEXT: v_mul_hi_u32 v0, s10, v0 -; GFX6-NEXT: s_cselect_b64 s[0:1], -1, 0 -; GFX6-NEXT: s_or_b32 s0, s0, s1 -; GFX6-NEXT: s_addc_u32 s12, s12, s13 -; GFX6-NEXT: s_mul_i32 s0, s10, s12 -; GFX6-NEXT: v_readfirstlane_b32 s1, v0 -; GFX6-NEXT: s_add_i32 s0, s1, s0 -; GFX6-NEXT: s_mul_i32 s11, s11, s14 -; GFX6-NEXT: s_mul_i32 s1, s10, s14 -; GFX6-NEXT: s_add_i32 s0, s0, s11 -; GFX6-NEXT: v_mov_b32_e32 v2, s1 -; GFX6-NEXT: v_mov_b32_e32 v0, s0 -; GFX6-NEXT: v_mul_hi_u32 v3, s12, v2 -; GFX6-NEXT: v_mul_hi_u32 v2, s14, v2 -; GFX6-NEXT: v_mul_hi_u32 v1, s12, v0 -; GFX6-NEXT: v_mul_hi_u32 v0, s14, v0 -; GFX6-NEXT: s_mul_i32 s11, s14, s0 -; GFX6-NEXT: v_readfirstlane_b32 s15, v2 -; GFX6-NEXT: s_add_u32 s11, s15, s11 -; GFX6-NEXT: v_readfirstlane_b32 s13, v0 -; GFX6-NEXT: s_mul_i32 s1, s12, s1 -; GFX6-NEXT: s_addc_u32 s13, 0, s13 -; GFX6-NEXT: v_readfirstlane_b32 s10, v3 -; GFX6-NEXT: s_add_u32 s1, s11, s1 -; GFX6-NEXT: s_addc_u32 s1, s13, s10 +; GFX6-NEXT: v_mul_hi_u32 v2, s0, v0 ; GFX6-NEXT: v_readfirstlane_b32 s10, v1 -; GFX6-NEXT: s_addc_u32 s10, s10, 0 -; GFX6-NEXT: s_mul_i32 s0, s12, s0 -; GFX6-NEXT: s_add_u32 s0, s1, s0 -; GFX6-NEXT: s_addc_u32 s10, 0, s10 -; GFX6-NEXT: s_add_u32 s13, s14, s0 -; GFX6-NEXT: s_cselect_b64 s[0:1], -1, 0 -; GFX6-NEXT: s_or_b32 s0, s0, s1 -; GFX6-NEXT: s_addc_u32 s12, s12, s10 +; GFX6-NEXT: v_readfirstlane_b32 s2, v0 +; GFX6-NEXT: s_mul_i32 s11, s0, s10 +; GFX6-NEXT: v_readfirstlane_b32 s14, v2 +; GFX6-NEXT: s_mul_i32 s12, s1, s2 +; GFX6-NEXT: s_mul_i32 s13, s0, s2 +; GFX6-NEXT: s_add_i32 s11, s14, s11 +; GFX6-NEXT: v_mul_hi_u32 v3, v0, s13 +; GFX6-NEXT: s_add_i32 s11, s11, s12 +; GFX6-NEXT: v_mul_hi_u32 v0, v0, s11 +; GFX6-NEXT: v_mul_hi_u32 v4, v1, s13 +; GFX6-NEXT: v_readfirstlane_b32 s12, v3 +; GFX6-NEXT: s_mul_i32 s14, s2, s11 +; GFX6-NEXT: v_mul_hi_u32 v1, v1, s11 +; GFX6-NEXT: s_add_u32 s12, s12, s14 +; GFX6-NEXT: v_readfirstlane_b32 s14, v0 +; GFX6-NEXT: s_mul_i32 s13, s10, s13 +; GFX6-NEXT: s_addc_u32 s14, 0, s14 +; GFX6-NEXT: v_readfirstlane_b32 s15, v4 +; GFX6-NEXT: s_add_u32 s12, s12, s13 +; GFX6-NEXT: s_addc_u32 s12, s14, s15 +; GFX6-NEXT: v_readfirstlane_b32 s13, v1 +; GFX6-NEXT: s_addc_u32 s13, s13, 0 +; GFX6-NEXT: s_mul_i32 s11, s10, s11 +; GFX6-NEXT: s_add_u32 s11, s12, s11 +; GFX6-NEXT: s_addc_u32 s12, 0, s13 +; GFX6-NEXT: s_add_u32 s11, s2, s11 +; GFX6-NEXT: v_mov_b32_e32 v0, s11 +; GFX6-NEXT: v_mul_hi_u32 v0, s0, v0 +; GFX6-NEXT: s_addc_u32 s10, s10, s12 +; GFX6-NEXT: s_mul_i32 s12, s0, s10 +; GFX6-NEXT: s_mul_i32 s1, s1, s11 +; GFX6-NEXT: v_readfirstlane_b32 s13, v0 +; GFX6-NEXT: s_add_i32 s12, s13, s12 +; GFX6-NEXT: s_mul_i32 s0, s0, s11 +; GFX6-NEXT: s_add_i32 s1, s12, s1 +; GFX6-NEXT: v_mov_b32_e32 v2, s0 +; GFX6-NEXT: v_mov_b32_e32 v0, s1 +; GFX6-NEXT: v_mul_hi_u32 v3, s10, v2 +; GFX6-NEXT: v_mul_hi_u32 v2, s11, v2 +; GFX6-NEXT: v_mul_hi_u32 v1, s10, v0 +; GFX6-NEXT: v_mul_hi_u32 v0, s11, v0 +; GFX6-NEXT: s_mul_i32 s13, s11, s1 +; GFX6-NEXT: v_readfirstlane_b32 s15, v2 +; GFX6-NEXT: s_add_u32 s13, s15, s13 +; GFX6-NEXT: v_readfirstlane_b32 s14, v0 +; GFX6-NEXT: s_mul_i32 s0, s10, s0 +; GFX6-NEXT: s_addc_u32 s14, 0, s14 +; GFX6-NEXT: v_readfirstlane_b32 s12, v3 +; GFX6-NEXT: s_add_u32 s0, s13, s0 +; GFX6-NEXT: s_addc_u32 s0, s14, s12 +; GFX6-NEXT: v_readfirstlane_b32 s12, v1 +; GFX6-NEXT: s_addc_u32 s12, s12, 0 +; GFX6-NEXT: s_mul_i32 s1, s10, s1 +; GFX6-NEXT: s_add_u32 s0, s0, s1 +; GFX6-NEXT: s_addc_u32 s1, 0, s12 +; GFX6-NEXT: s_add_u32 s12, s11, s0 +; GFX6-NEXT: s_addc_u32 s13, s10, s1 ; GFX6-NEXT: s_ashr_i32 s10, s7, 31 ; GFX6-NEXT: s_add_u32 s0, s6, s10 ; GFX6-NEXT: s_mov_b32 s11, s10 ; GFX6-NEXT: s_addc_u32 s1, s7, s10 ; GFX6-NEXT: s_xor_b64 s[6:7], s[0:1], s[10:11] -; GFX6-NEXT: v_mov_b32_e32 v0, s12 +; GFX6-NEXT: v_mov_b32_e32 v0, s13 ; GFX6-NEXT: v_mul_hi_u32 v1, s6, v0 -; GFX6-NEXT: v_mov_b32_e32 v2, s13 +; GFX6-NEXT: v_mov_b32_e32 v2, s12 ; GFX6-NEXT: v_mul_hi_u32 v3, s6, v2 ; GFX6-NEXT: s_mov_b32 s0, s4 ; GFX6-NEXT: v_readfirstlane_b32 s4, v1 ; GFX6-NEXT: v_mul_hi_u32 v1, s7, v2 -; GFX6-NEXT: s_mul_i32 s1, s6, s12 +; GFX6-NEXT: s_mul_i32 s1, s6, s13 ; GFX6-NEXT: v_readfirstlane_b32 s14, v3 ; GFX6-NEXT: v_mul_hi_u32 v0, s7, v0 ; GFX6-NEXT: s_add_u32 s1, s14, s1 ; GFX6-NEXT: s_addc_u32 s4, 0, s4 -; GFX6-NEXT: s_mul_i32 s13, s7, s13 +; GFX6-NEXT: s_mul_i32 s12, s7, s12 ; GFX6-NEXT: v_readfirstlane_b32 s14, v1 -; GFX6-NEXT: s_add_u32 s1, s1, s13 +; GFX6-NEXT: s_add_u32 s1, s1, s12 ; GFX6-NEXT: s_addc_u32 s1, s4, s14 ; GFX6-NEXT: v_readfirstlane_b32 s4, v0 ; GFX6-NEXT: s_addc_u32 s4, s4, 0 -; GFX6-NEXT: s_mul_i32 s12, s7, s12 +; GFX6-NEXT: s_mul_i32 s12, s7, s13 ; GFX6-NEXT: s_add_u32 s12, s1, s12 ; GFX6-NEXT: v_mov_b32_e32 v0, s12 ; GFX6-NEXT: v_mul_hi_u32 v0, s8, v0 @@ -9128,11 +9102,9 @@ define amdgpu_kernel void @srem_i64_pow2_shl_denom(ptr addrspace(1) %out, i64 %x ; GFX6-NEXT: s_mul_i32 s4, s8, s12 ; GFX6-NEXT: s_sub_u32 s6, s6, s4 ; GFX6-NEXT: s_cselect_b64 s[4:5], -1, 0 -; GFX6-NEXT: s_or_b32 s12, s4, s5 ; GFX6-NEXT: s_subb_u32 s15, s13, s9 ; GFX6-NEXT: s_sub_u32 s16, s6, s8 ; GFX6-NEXT: s_cselect_b64 s[12:13], -1, 0 -; GFX6-NEXT: s_or_b32 s17, s12, s13 ; GFX6-NEXT: s_subb_u32 s17, s15, 0 ; GFX6-NEXT: s_cmp_ge_u32 s17, s9 ; GFX6-NEXT: s_cselect_b32 s18, -1, 0 @@ -9141,13 +9113,11 @@ define amdgpu_kernel void @srem_i64_pow2_shl_denom(ptr addrspace(1) %out, i64 %x ; GFX6-NEXT: s_cmp_eq_u32 s17, s9 ; GFX6-NEXT: s_cselect_b32 s18, s19, s18 ; GFX6-NEXT: s_or_b32 s12, s12, s13 -; GFX6-NEXT: s_subb_u32 s15, s15, s9 -; GFX6-NEXT: s_sub_u32 s19, s16, s8 -; GFX6-NEXT: s_cselect_b64 s[12:13], -1, 0 -; GFX6-NEXT: s_or_b32 s12, s12, s13 -; GFX6-NEXT: s_subb_u32 s12, s15, 0 +; GFX6-NEXT: s_subb_u32 s12, s15, s9 +; GFX6-NEXT: s_sub_u32 s13, s16, s8 +; GFX6-NEXT: s_subb_u32 s12, s12, 0 ; GFX6-NEXT: s_cmp_lg_u32 s18, 0 -; GFX6-NEXT: s_cselect_b32 s13, s19, s16 +; GFX6-NEXT: s_cselect_b32 s13, s13, s16 ; GFX6-NEXT: s_cselect_b32 s12, s12, s17 ; GFX6-NEXT: s_or_b32 s4, s4, s5 ; GFX6-NEXT: s_subb_u32 s4, s7, s14 @@ -9164,6 +9134,7 @@ define amdgpu_kernel void @srem_i64_pow2_shl_denom(ptr addrspace(1) %out, i64 %x ; GFX6-NEXT: s_sub_u32 s4, s4, s10 ; GFX6-NEXT: s_subb_u32 s5, s5, s10 ; GFX6-NEXT: v_mov_b32_e32 v0, s4 +; GFX6-NEXT: s_mov_b32 s2, -1 ; GFX6-NEXT: v_mov_b32_e32 v1, s5 ; GFX6-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 ; GFX6-NEXT: s_endpgm @@ -9405,8 +9376,8 @@ define amdgpu_kernel void @srem_v2i64_pow2_shl_denom(ptr addrspace(1) %out, <2 x ; GFX6-NEXT: s_xor_b64 s[2:3], s[2:3], s[6:7] ; GFX6-NEXT: v_cvt_f32_u32_e32 v0, s2 ; GFX6-NEXT: v_cvt_f32_u32_e32 v1, s3 -; GFX6-NEXT: s_sub_u32 s12, 0, s2 -; GFX6-NEXT: s_subb_u32 s13, 0, s3 +; GFX6-NEXT: s_sub_u32 s6, 0, s2 +; GFX6-NEXT: s_subb_u32 s7, 0, s3 ; GFX6-NEXT: v_mac_f32_e32 v0, 0x4f800000, v1 ; GFX6-NEXT: v_rcp_f32_e32 v0, v0 ; GFX6-NEXT: v_mul_f32_e32 v0, 0x5f7ffffc, v0 @@ -9415,69 +9386,65 @@ define amdgpu_kernel void @srem_v2i64_pow2_shl_denom(ptr addrspace(1) %out, <2 x ; GFX6-NEXT: v_mac_f32_e32 v0, 0xcf800000, v1 ; GFX6-NEXT: v_cvt_u32_f32_e32 v0, v0 ; GFX6-NEXT: v_cvt_u32_f32_e32 v1, v1 -; GFX6-NEXT: v_mul_hi_u32 v2, s12, v0 -; GFX6-NEXT: v_readfirstlane_b32 s14, v1 -; GFX6-NEXT: v_readfirstlane_b32 s6, v0 -; GFX6-NEXT: s_mul_i32 s7, s12, s14 +; GFX6-NEXT: v_mul_hi_u32 v2, s6, v0 +; GFX6-NEXT: v_readfirstlane_b32 s12, v1 +; GFX6-NEXT: v_readfirstlane_b32 s13, v0 +; GFX6-NEXT: s_mul_i32 s14, s6, s12 ; GFX6-NEXT: v_readfirstlane_b32 s17, v2 -; GFX6-NEXT: s_mul_i32 s15, s13, s6 -; GFX6-NEXT: s_mul_i32 s16, s12, s6 -; GFX6-NEXT: s_add_i32 s7, s17, s7 +; GFX6-NEXT: s_mul_i32 s15, s7, s13 +; GFX6-NEXT: s_mul_i32 s16, s6, s13 +; GFX6-NEXT: s_add_i32 s14, s17, s14 ; GFX6-NEXT: v_mul_hi_u32 v3, v0, s16 -; GFX6-NEXT: s_add_i32 s7, s7, s15 -; GFX6-NEXT: v_mul_hi_u32 v0, v0, s7 +; GFX6-NEXT: s_add_i32 s14, s14, s15 +; GFX6-NEXT: v_mul_hi_u32 v0, v0, s14 ; GFX6-NEXT: v_mul_hi_u32 v4, v1, s16 ; GFX6-NEXT: v_readfirstlane_b32 s15, v3 -; GFX6-NEXT: s_mul_i32 s18, s6, s7 -; GFX6-NEXT: v_mul_hi_u32 v1, v1, s7 +; GFX6-NEXT: s_mul_i32 s18, s13, s14 +; GFX6-NEXT: v_mul_hi_u32 v1, v1, s14 ; GFX6-NEXT: s_add_u32 s15, s15, s18 ; GFX6-NEXT: v_readfirstlane_b32 s18, v0 -; GFX6-NEXT: s_mul_i32 s16, s14, s16 +; GFX6-NEXT: s_mul_i32 s16, s12, s16 ; GFX6-NEXT: s_addc_u32 s18, 0, s18 ; GFX6-NEXT: v_readfirstlane_b32 s17, v4 ; GFX6-NEXT: s_add_u32 s15, s15, s16 ; GFX6-NEXT: s_addc_u32 s15, s18, s17 ; GFX6-NEXT: v_readfirstlane_b32 s16, v1 ; GFX6-NEXT: s_addc_u32 s16, s16, 0 -; GFX6-NEXT: s_mul_i32 s7, s14, s7 -; GFX6-NEXT: s_add_u32 s7, s15, s7 +; GFX6-NEXT: s_mul_i32 s14, s12, s14 +; GFX6-NEXT: s_add_u32 s14, s15, s14 ; GFX6-NEXT: s_addc_u32 s15, 0, s16 -; GFX6-NEXT: s_add_u32 s16, s6, s7 -; GFX6-NEXT: v_mov_b32_e32 v0, s16 -; GFX6-NEXT: v_mul_hi_u32 v0, s12, v0 -; GFX6-NEXT: s_cselect_b64 s[6:7], -1, 0 -; GFX6-NEXT: s_or_b32 s6, s6, s7 -; GFX6-NEXT: s_addc_u32 s14, s14, s15 -; GFX6-NEXT: s_mul_i32 s6, s12, s14 -; GFX6-NEXT: v_readfirstlane_b32 s7, v0 -; GFX6-NEXT: s_add_i32 s6, s7, s6 -; GFX6-NEXT: s_mul_i32 s13, s13, s16 -; GFX6-NEXT: s_mul_i32 s7, s12, s16 -; GFX6-NEXT: s_add_i32 s6, s6, s13 -; GFX6-NEXT: v_mov_b32_e32 v2, s7 -; GFX6-NEXT: v_mov_b32_e32 v0, s6 -; GFX6-NEXT: v_mul_hi_u32 v3, s14, v2 -; GFX6-NEXT: v_mul_hi_u32 v2, s16, v2 -; GFX6-NEXT: v_mul_hi_u32 v1, s14, v0 -; GFX6-NEXT: v_mul_hi_u32 v0, s16, v0 -; GFX6-NEXT: s_mul_i32 s13, s16, s6 -; GFX6-NEXT: v_readfirstlane_b32 s17, v2 -; GFX6-NEXT: s_add_u32 s13, s17, s13 +; GFX6-NEXT: s_add_u32 s13, s13, s14 +; GFX6-NEXT: v_mov_b32_e32 v0, s13 +; GFX6-NEXT: v_mul_hi_u32 v0, s6, v0 +; GFX6-NEXT: s_addc_u32 s12, s12, s15 +; GFX6-NEXT: s_mul_i32 s14, s6, s12 +; GFX6-NEXT: s_mul_i32 s7, s7, s13 ; GFX6-NEXT: v_readfirstlane_b32 s15, v0 -; GFX6-NEXT: s_mul_i32 s7, s14, s7 -; GFX6-NEXT: s_addc_u32 s15, 0, s15 -; GFX6-NEXT: v_readfirstlane_b32 s12, v3 -; GFX6-NEXT: s_add_u32 s7, s13, s7 -; GFX6-NEXT: s_addc_u32 s7, s15, s12 -; GFX6-NEXT: v_readfirstlane_b32 s12, v1 -; GFX6-NEXT: s_addc_u32 s12, s12, 0 -; GFX6-NEXT: s_mul_i32 s6, s14, s6 -; GFX6-NEXT: s_add_u32 s6, s7, s6 -; GFX6-NEXT: s_addc_u32 s12, 0, s12 -; GFX6-NEXT: s_add_u32 s13, s16, s6 -; GFX6-NEXT: s_cselect_b64 s[6:7], -1, 0 -; GFX6-NEXT: s_or_b32 s6, s6, s7 -; GFX6-NEXT: s_addc_u32 s12, s14, s12 +; GFX6-NEXT: s_add_i32 s14, s15, s14 +; GFX6-NEXT: s_mul_i32 s6, s6, s13 +; GFX6-NEXT: s_add_i32 s7, s14, s7 +; GFX6-NEXT: v_mov_b32_e32 v2, s6 +; GFX6-NEXT: v_mov_b32_e32 v0, s7 +; GFX6-NEXT: v_mul_hi_u32 v3, s12, v2 +; GFX6-NEXT: v_mul_hi_u32 v2, s13, v2 +; GFX6-NEXT: v_mul_hi_u32 v1, s12, v0 +; GFX6-NEXT: v_mul_hi_u32 v0, s13, v0 +; GFX6-NEXT: s_mul_i32 s15, s13, s7 +; GFX6-NEXT: v_readfirstlane_b32 s17, v2 +; GFX6-NEXT: s_add_u32 s15, s17, s15 +; GFX6-NEXT: v_readfirstlane_b32 s16, v0 +; GFX6-NEXT: s_mul_i32 s6, s12, s6 +; GFX6-NEXT: s_addc_u32 s16, 0, s16 +; GFX6-NEXT: v_readfirstlane_b32 s14, v3 +; GFX6-NEXT: s_add_u32 s6, s15, s6 +; GFX6-NEXT: s_addc_u32 s6, s16, s14 +; GFX6-NEXT: v_readfirstlane_b32 s14, v1 +; GFX6-NEXT: s_addc_u32 s14, s14, 0 +; GFX6-NEXT: s_mul_i32 s7, s12, s7 +; GFX6-NEXT: s_add_u32 s6, s6, s7 +; GFX6-NEXT: s_addc_u32 s7, 0, s14 +; GFX6-NEXT: s_add_u32 s13, s13, s6 +; GFX6-NEXT: s_addc_u32 s12, s12, s7 ; GFX6-NEXT: s_ashr_i32 s6, s9, 31 ; GFX6-NEXT: s_add_u32 s8, s8, s6 ; GFX6-NEXT: s_mov_b32 s7, s6 @@ -9514,11 +9481,9 @@ define amdgpu_kernel void @srem_v2i64_pow2_shl_denom(ptr addrspace(1) %out, <2 x ; GFX6-NEXT: s_mul_i32 s12, s2, s12 ; GFX6-NEXT: s_sub_u32 s8, s8, s12 ; GFX6-NEXT: s_cselect_b64 s[12:13], -1, 0 -; GFX6-NEXT: s_or_b32 s15, s12, s13 ; GFX6-NEXT: s_subb_u32 s17, s14, s3 ; GFX6-NEXT: s_sub_u32 s18, s8, s2 ; GFX6-NEXT: s_cselect_b64 s[14:15], -1, 0 -; GFX6-NEXT: s_or_b32 s19, s14, s15 ; GFX6-NEXT: s_subb_u32 s19, s17, 0 ; GFX6-NEXT: s_cmp_ge_u32 s19, s3 ; GFX6-NEXT: s_cselect_b32 s20, -1, 0 @@ -9527,13 +9492,11 @@ define amdgpu_kernel void @srem_v2i64_pow2_shl_denom(ptr addrspace(1) %out, <2 x ; GFX6-NEXT: s_cmp_eq_u32 s19, s3 ; GFX6-NEXT: s_cselect_b32 s20, s21, s20 ; GFX6-NEXT: s_or_b32 s14, s14, s15 -; GFX6-NEXT: s_subb_u32 s17, s17, s3 -; GFX6-NEXT: s_sub_u32 s21, s18, s2 -; GFX6-NEXT: s_cselect_b64 s[14:15], -1, 0 -; GFX6-NEXT: s_or_b32 s14, s14, s15 -; GFX6-NEXT: s_subb_u32 s14, s17, 0 +; GFX6-NEXT: s_subb_u32 s14, s17, s3 +; GFX6-NEXT: s_sub_u32 s15, s18, s2 +; GFX6-NEXT: s_subb_u32 s14, s14, 0 ; GFX6-NEXT: s_cmp_lg_u32 s20, 0 -; GFX6-NEXT: s_cselect_b32 s15, s21, s18 +; GFX6-NEXT: s_cselect_b32 s15, s15, s18 ; GFX6-NEXT: s_cselect_b32 s14, s14, s19 ; GFX6-NEXT: s_or_b32 s12, s12, s13 ; GFX6-NEXT: s_subb_u32 s9, s9, s16 @@ -9556,8 +9519,8 @@ define amdgpu_kernel void @srem_v2i64_pow2_shl_denom(ptr addrspace(1) %out, <2 x ; GFX6-NEXT: s_xor_b64 s[6:7], s[0:1], s[2:3] ; GFX6-NEXT: v_cvt_f32_u32_e32 v0, s6 ; GFX6-NEXT: v_cvt_f32_u32_e32 v1, s7 -; GFX6-NEXT: s_sub_u32 s8, 0, s6 -; GFX6-NEXT: s_subb_u32 s9, 0, s7 +; GFX6-NEXT: s_sub_u32 s2, 0, s6 +; GFX6-NEXT: s_subb_u32 s3, 0, s7 ; GFX6-NEXT: v_mac_f32_e32 v0, 0x4f800000, v1 ; GFX6-NEXT: v_rcp_f32_e32 v0, v0 ; GFX6-NEXT: v_mul_f32_e32 v0, 0x5f7ffffc, v0 @@ -9566,70 +9529,66 @@ define amdgpu_kernel void @srem_v2i64_pow2_shl_denom(ptr addrspace(1) %out, <2 x ; GFX6-NEXT: v_mac_f32_e32 v0, 0xcf800000, v1 ; GFX6-NEXT: v_cvt_u32_f32_e32 v0, v0 ; GFX6-NEXT: v_cvt_u32_f32_e32 v1, v1 -; GFX6-NEXT: v_mul_hi_u32 v2, s8, v0 -; GFX6-NEXT: v_readfirstlane_b32 s12, v1 -; GFX6-NEXT: v_readfirstlane_b32 s2, v0 -; GFX6-NEXT: s_mul_i32 s1, s8, s12 -; GFX6-NEXT: v_readfirstlane_b32 s3, v2 -; GFX6-NEXT: s_mul_i32 s0, s9, s2 -; GFX6-NEXT: s_add_i32 s1, s3, s1 -; GFX6-NEXT: s_add_i32 s3, s1, s0 -; GFX6-NEXT: s_mul_i32 s13, s8, s2 -; GFX6-NEXT: v_mul_hi_u32 v2, v0, s3 -; GFX6-NEXT: v_mul_hi_u32 v0, v0, s13 -; GFX6-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9 -; GFX6-NEXT: s_mul_i32 s4, s2, s3 -; GFX6-NEXT: v_readfirstlane_b32 s5, v2 +; GFX6-NEXT: v_mul_hi_u32 v2, s2, v0 +; GFX6-NEXT: v_readfirstlane_b32 s8, v1 +; GFX6-NEXT: v_readfirstlane_b32 s0, v0 +; GFX6-NEXT: s_mul_i32 s9, s2, s8 +; GFX6-NEXT: v_readfirstlane_b32 s12, v2 +; GFX6-NEXT: s_mul_i32 s1, s3, s0 +; GFX6-NEXT: s_add_i32 s9, s12, s9 +; GFX6-NEXT: s_add_i32 s9, s9, s1 +; GFX6-NEXT: s_mul_i32 s1, s2, s0 +; GFX6-NEXT: v_mul_hi_u32 v2, v0, s9 +; GFX6-NEXT: v_mul_hi_u32 v0, v0, s1 +; GFX6-NEXT: s_mul_i32 s12, s0, s9 +; GFX6-NEXT: v_readfirstlane_b32 s13, v2 ; GFX6-NEXT: v_readfirstlane_b32 s16, v0 -; GFX6-NEXT: v_mul_hi_u32 v0, v1, s13 -; GFX6-NEXT: v_mul_hi_u32 v1, v1, s3 -; GFX6-NEXT: s_add_u32 s4, s16, s4 -; GFX6-NEXT: s_addc_u32 s5, 0, s5 -; GFX6-NEXT: s_mul_i32 s13, s12, s13 +; GFX6-NEXT: v_mul_hi_u32 v0, v1, s1 +; GFX6-NEXT: v_mul_hi_u32 v1, v1, s9 +; GFX6-NEXT: s_add_u32 s12, s16, s12 +; GFX6-NEXT: s_addc_u32 s13, 0, s13 +; GFX6-NEXT: s_mul_i32 s1, s8, s1 ; GFX6-NEXT: v_readfirstlane_b32 s16, v0 -; GFX6-NEXT: s_add_u32 s4, s4, s13 -; GFX6-NEXT: s_addc_u32 s4, s5, s16 -; GFX6-NEXT: v_readfirstlane_b32 s5, v1 -; GFX6-NEXT: s_addc_u32 s5, s5, 0 -; GFX6-NEXT: s_mul_i32 s3, s12, s3 -; GFX6-NEXT: s_add_u32 s3, s4, s3 -; GFX6-NEXT: s_addc_u32 s4, 0, s5 -; GFX6-NEXT: s_add_u32 s5, s2, s3 -; GFX6-NEXT: v_mov_b32_e32 v0, s5 -; GFX6-NEXT: v_mul_hi_u32 v0, s8, v0 -; GFX6-NEXT: s_cselect_b64 s[2:3], -1, 0 -; GFX6-NEXT: s_or_b32 s2, s2, s3 -; GFX6-NEXT: s_addc_u32 s4, s12, s4 -; GFX6-NEXT: s_mul_i32 s2, s8, s4 -; GFX6-NEXT: v_readfirstlane_b32 s3, v0 -; GFX6-NEXT: s_add_i32 s2, s3, s2 -; GFX6-NEXT: s_mul_i32 s9, s9, s5 -; GFX6-NEXT: s_mul_i32 s3, s8, s5 -; GFX6-NEXT: s_add_i32 s2, s2, s9 -; GFX6-NEXT: v_mov_b32_e32 v2, s3 -; GFX6-NEXT: v_mov_b32_e32 v0, s2 +; GFX6-NEXT: s_add_u32 s1, s12, s1 +; GFX6-NEXT: s_addc_u32 s1, s13, s16 +; GFX6-NEXT: v_readfirstlane_b32 s12, v1 +; GFX6-NEXT: s_addc_u32 s12, s12, 0 +; GFX6-NEXT: s_mul_i32 s9, s8, s9 +; GFX6-NEXT: s_add_u32 s1, s1, s9 +; GFX6-NEXT: s_addc_u32 s9, 0, s12 +; GFX6-NEXT: s_add_u32 s12, s0, s1 +; GFX6-NEXT: v_mov_b32_e32 v0, s12 +; GFX6-NEXT: v_mul_hi_u32 v0, s2, v0 +; GFX6-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x9 +; GFX6-NEXT: s_addc_u32 s4, s8, s9 +; GFX6-NEXT: s_mul_i32 s5, s2, s4 +; GFX6-NEXT: v_readfirstlane_b32 s8, v0 +; GFX6-NEXT: s_add_i32 s5, s8, s5 +; GFX6-NEXT: s_mul_i32 s3, s3, s12 +; GFX6-NEXT: s_mul_i32 s2, s2, s12 +; GFX6-NEXT: s_add_i32 s3, s5, s3 +; GFX6-NEXT: v_mov_b32_e32 v2, s2 +; GFX6-NEXT: v_mov_b32_e32 v0, s3 ; GFX6-NEXT: v_mul_hi_u32 v3, s4, v2 -; GFX6-NEXT: v_mul_hi_u32 v2, s5, v2 +; GFX6-NEXT: v_mul_hi_u32 v2, s12, v2 ; GFX6-NEXT: v_mul_hi_u32 v1, s4, v0 -; GFX6-NEXT: v_mul_hi_u32 v0, s5, v0 -; GFX6-NEXT: s_mul_i32 s9, s5, s2 +; GFX6-NEXT: v_mul_hi_u32 v0, s12, v0 +; GFX6-NEXT: s_mul_i32 s8, s12, s3 ; GFX6-NEXT: v_readfirstlane_b32 s13, v2 -; GFX6-NEXT: s_add_u32 s9, s13, s9 -; GFX6-NEXT: v_readfirstlane_b32 s12, v0 -; GFX6-NEXT: s_mul_i32 s3, s4, s3 -; GFX6-NEXT: s_addc_u32 s12, 0, s12 -; GFX6-NEXT: v_readfirstlane_b32 s8, v3 -; GFX6-NEXT: s_add_u32 s3, s9, s3 -; GFX6-NEXT: s_addc_u32 s3, s12, s8 -; GFX6-NEXT: v_readfirstlane_b32 s8, v1 -; GFX6-NEXT: s_addc_u32 s8, s8, 0 +; GFX6-NEXT: s_add_u32 s8, s13, s8 +; GFX6-NEXT: v_readfirstlane_b32 s9, v0 ; GFX6-NEXT: s_mul_i32 s2, s4, s2 -; GFX6-NEXT: s_add_u32 s2, s3, s2 -; GFX6-NEXT: s_addc_u32 s8, 0, s8 -; GFX6-NEXT: s_add_u32 s12, s5, s2 -; GFX6-NEXT: s_cselect_b64 s[2:3], -1, 0 -; GFX6-NEXT: s_or_b32 s2, s2, s3 -; GFX6-NEXT: s_addc_u32 s13, s4, s8 +; GFX6-NEXT: s_addc_u32 s9, 0, s9 +; GFX6-NEXT: v_readfirstlane_b32 s5, v3 +; GFX6-NEXT: s_add_u32 s2, s8, s2 +; GFX6-NEXT: s_addc_u32 s2, s9, s5 +; GFX6-NEXT: v_readfirstlane_b32 s5, v1 +; GFX6-NEXT: s_addc_u32 s5, s5, 0 +; GFX6-NEXT: s_mul_i32 s3, s4, s3 +; GFX6-NEXT: s_add_u32 s2, s2, s3 +; GFX6-NEXT: s_addc_u32 s3, 0, s5 +; GFX6-NEXT: s_add_u32 s12, s12, s2 +; GFX6-NEXT: s_addc_u32 s13, s4, s3 ; GFX6-NEXT: s_ashr_i32 s4, s11, 31 ; GFX6-NEXT: s_add_u32 s2, s10, s4 ; GFX6-NEXT: s_mov_b32 s5, s4 @@ -9667,11 +9626,9 @@ define amdgpu_kernel void @srem_v2i64_pow2_shl_denom(ptr addrspace(1) %out, <2 x ; GFX6-NEXT: s_mul_i32 s10, s6, s11 ; GFX6-NEXT: s_sub_u32 s8, s8, s10 ; GFX6-NEXT: s_cselect_b64 s[10:11], -1, 0 -; GFX6-NEXT: s_or_b32 s13, s10, s11 ; GFX6-NEXT: s_subb_u32 s17, s12, s7 ; GFX6-NEXT: s_sub_u32 s18, s8, s6 ; GFX6-NEXT: s_cselect_b64 s[12:13], -1, 0 -; GFX6-NEXT: s_or_b32 s19, s12, s13 ; GFX6-NEXT: s_subb_u32 s19, s17, 0 ; GFX6-NEXT: s_cmp_ge_u32 s19, s7 ; GFX6-NEXT: s_cselect_b32 s20, -1, 0 @@ -9680,13 +9637,11 @@ define amdgpu_kernel void @srem_v2i64_pow2_shl_denom(ptr addrspace(1) %out, <2 x ; GFX6-NEXT: s_cmp_eq_u32 s19, s7 ; GFX6-NEXT: s_cselect_b32 s20, s21, s20 ; GFX6-NEXT: s_or_b32 s12, s12, s13 -; GFX6-NEXT: s_subb_u32 s17, s17, s7 -; GFX6-NEXT: s_sub_u32 s21, s18, s6 -; GFX6-NEXT: s_cselect_b64 s[12:13], -1, 0 -; GFX6-NEXT: s_or_b32 s12, s12, s13 -; GFX6-NEXT: s_subb_u32 s12, s17, 0 +; GFX6-NEXT: s_subb_u32 s12, s17, s7 +; GFX6-NEXT: s_sub_u32 s13, s18, s6 +; GFX6-NEXT: s_subb_u32 s12, s12, 0 ; GFX6-NEXT: s_cmp_lg_u32 s20, 0 -; GFX6-NEXT: s_cselect_b32 s13, s21, s18 +; GFX6-NEXT: s_cselect_b32 s13, s13, s18 ; GFX6-NEXT: s_cselect_b32 s12, s12, s19 ; GFX6-NEXT: s_or_b32 s10, s10, s11 ; GFX6-NEXT: s_subb_u32 s9, s9, s16 diff --git a/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-uniform-waterfall.ll b/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-uniform-waterfall.ll index c962c05..5d79696 100644 --- a/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-uniform-waterfall.ll +++ b/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-uniform-waterfall.ll @@ -239,7 +239,8 @@ define protected amdgpu_kernel void @trivial_uniform_waterfall(ptr addrspace(1) ; PASS-CHECK-NEXT: [[IS_DONE:%.*]] = icmp eq i64 [[BALLOT]], 0 ; PASS-CHECK-NEXT: br i1 [[TMP0]], label %[[EXIT:.*]], label %[[IF:.*]] ; PASS-CHECK: [[IF]]: -; PASS-CHECK-NEXT: [[IS_FIRST_ACTIVE_ID:%.*]] = icmp eq i32 0, 0 +; PASS-CHECK-NEXT: [[FIRST_ACTIVE_ID:%.*]] = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 0) +; PASS-CHECK-NEXT: [[IS_FIRST_ACTIVE_ID:%.*]] = icmp eq i32 0, [[FIRST_ACTIVE_ID]] ; PASS-CHECK-NEXT: br i1 [[IS_FIRST_ACTIVE_ID]], label %[[WORK:.*]], label %[[TAIL]] ; PASS-CHECK: [[WORK]]: ; PASS-CHECK-NEXT: store i32 5, ptr addrspace(1) [[OUT]], align 4 @@ -308,7 +309,8 @@ define protected amdgpu_kernel void @uniform_waterfall(ptr addrspace(1) %out, i3 ; PASS-CHECK-NEXT: [[IS_DONE:%.*]] = icmp eq i64 [[BALLOT]], 0 ; PASS-CHECK-NEXT: br i1 [[TMP0]], label %[[EXIT:.*]], label %[[IF:.*]] ; PASS-CHECK: [[IF]]: -; PASS-CHECK-NEXT: [[IS_FIRST_ACTIVE_ID:%.*]] = icmp eq i32 [[MYMASK]], [[MYMASK]] +; PASS-CHECK-NEXT: [[FIRST_ACTIVE_ID:%.*]] = tail call noundef i32 @llvm.amdgcn.readfirstlane.i32(i32 [[MYMASK]]) +; PASS-CHECK-NEXT: [[IS_FIRST_ACTIVE_ID:%.*]] = icmp eq i32 [[MYMASK]], [[FIRST_ACTIVE_ID]] ; PASS-CHECK-NEXT: br i1 [[IS_FIRST_ACTIVE_ID]], label %[[WORK:.*]], label %[[TAIL]] ; PASS-CHECK: [[WORK]]: ; PASS-CHECK-NEXT: store i32 5, ptr addrspace(1) [[OUT]], align 4 diff --git a/llvm/test/CodeGen/AMDGPU/amdgpu-uniform-intrinsic-combine.ll b/llvm/test/CodeGen/AMDGPU/amdgpu-uniform-intrinsic-combine.ll index a7e828c..402ccd9 100644 --- a/llvm/test/CodeGen/AMDGPU/amdgpu-uniform-intrinsic-combine.ll +++ b/llvm/test/CodeGen/AMDGPU/amdgpu-uniform-intrinsic-combine.ll @@ -248,12 +248,14 @@ define amdgpu_kernel void @readfirstlane_constant(ptr addrspace(1) %out) { ; ; PASS-CHECK-LABEL: define amdgpu_kernel void @readfirstlane_constant( ; PASS-CHECK-SAME: ptr addrspace(1) [[OUT:%.*]]) #[[ATTR0]] { -; PASS-CHECK-NEXT: store i32 7, ptr addrspace(1) [[OUT]], align 4 +; PASS-CHECK-NEXT: [[V:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 7) +; PASS-CHECK-NEXT: store i32 [[V]], ptr addrspace(1) [[OUT]], align 4 ; PASS-CHECK-NEXT: ret void ; ; DCE-CHECK-LABEL: define amdgpu_kernel void @readfirstlane_constant( ; DCE-CHECK-SAME: ptr addrspace(1) [[OUT:%.*]]) #[[ATTR0]] { -; DCE-CHECK-NEXT: store i32 7, ptr addrspace(1) [[OUT]], align 4 +; DCE-CHECK-NEXT: [[V:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 7) +; DCE-CHECK-NEXT: store i32 [[V]], ptr addrspace(1) [[OUT]], align 4 ; DCE-CHECK-NEXT: ret void ; %v = call i32 @llvm.amdgcn.readfirstlane(i32 7) @@ -269,12 +271,14 @@ define amdgpu_kernel void @readfirstlane_with_argument(ptr addrspace(1) %out, i3 ; ; PASS-CHECK-LABEL: define amdgpu_kernel void @readfirstlane_with_argument( ; PASS-CHECK-SAME: ptr addrspace(1) [[OUT:%.*]], i32 [[SRC0:%.*]]) #[[ATTR0]] { -; PASS-CHECK-NEXT: store i32 [[SRC0]], ptr addrspace(1) [[OUT]], align 4 +; PASS-CHECK-NEXT: [[V:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[SRC0]]) +; PASS-CHECK-NEXT: store i32 [[V]], ptr addrspace(1) [[OUT]], align 4 ; PASS-CHECK-NEXT: ret void ; ; DCE-CHECK-LABEL: define amdgpu_kernel void @readfirstlane_with_argument( ; DCE-CHECK-SAME: ptr addrspace(1) [[OUT:%.*]], i32 [[SRC0:%.*]]) #[[ATTR0]] { -; DCE-CHECK-NEXT: store i32 [[SRC0]], ptr addrspace(1) [[OUT]], align 4 +; DCE-CHECK-NEXT: [[V:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[SRC0]]) +; DCE-CHECK-NEXT: store i32 [[V]], ptr addrspace(1) [[OUT]], align 4 ; DCE-CHECK-NEXT: ret void ; %v = call i32 @llvm.amdgcn.readfirstlane(i32 %src0) @@ -360,12 +364,16 @@ define amdgpu_kernel void @readfirstlane_with_readfirstlane(ptr addrspace(1) %ou ; ; PASS-CHECK-LABEL: define amdgpu_kernel void @readfirstlane_with_readfirstlane( ; PASS-CHECK-SAME: ptr addrspace(1) [[OUT:%.*]]) #[[ATTR0]] { -; PASS-CHECK-NEXT: store i32 5, ptr addrspace(1) [[OUT]], align 4 +; PASS-CHECK-NEXT: [[V1:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 5) +; PASS-CHECK-NEXT: [[V2:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[V1]]) +; PASS-CHECK-NEXT: store i32 [[V2]], ptr addrspace(1) [[OUT]], align 4 ; PASS-CHECK-NEXT: ret void ; ; DCE-CHECK-LABEL: define amdgpu_kernel void @readfirstlane_with_readfirstlane( ; DCE-CHECK-SAME: ptr addrspace(1) [[OUT:%.*]]) #[[ATTR0]] { -; DCE-CHECK-NEXT: store i32 5, ptr addrspace(1) [[OUT]], align 4 +; DCE-CHECK-NEXT: [[V1:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 5) +; DCE-CHECK-NEXT: [[V2:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[V1]]) +; DCE-CHECK-NEXT: store i32 [[V2]], ptr addrspace(1) [[OUT]], align 4 ; DCE-CHECK-NEXT: ret void ; %v1 = call i32 @llvm.amdgcn.readfirstlane(i32 5) @@ -388,7 +396,8 @@ define amdgpu_kernel void @readfirstlane_with_readlane(ptr addrspace(1) %out) { ; PASS-CHECK-NEXT: [[TIDX:%.*]] = call i32 @llvm.amdgcn.workitem.id.x() ; PASS-CHECK-NEXT: [[TIDY:%.*]] = call i32 @llvm.amdgcn.workitem.id.y() ; PASS-CHECK-NEXT: [[V1:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TIDX]], i32 [[TIDY]]) -; PASS-CHECK-NEXT: store i32 [[V1]], ptr addrspace(1) [[OUT]], align 4 +; PASS-CHECK-NEXT: [[V2:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[V1]]) +; PASS-CHECK-NEXT: store i32 [[V2]], ptr addrspace(1) [[OUT]], align 4 ; PASS-CHECK-NEXT: ret void ; ; DCE-CHECK-LABEL: define amdgpu_kernel void @readfirstlane_with_readlane( @@ -396,7 +405,8 @@ define amdgpu_kernel void @readfirstlane_with_readlane(ptr addrspace(1) %out) { ; DCE-CHECK-NEXT: [[TIDX:%.*]] = call i32 @llvm.amdgcn.workitem.id.x() ; DCE-CHECK-NEXT: [[TIDY:%.*]] = call i32 @llvm.amdgcn.workitem.id.y() ; DCE-CHECK-NEXT: [[V1:%.*]] = call i32 @llvm.amdgcn.readlane.i32(i32 [[TIDX]], i32 [[TIDY]]) -; DCE-CHECK-NEXT: store i32 [[V1]], ptr addrspace(1) [[OUT]], align 4 +; DCE-CHECK-NEXT: [[V2:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[V1]]) +; DCE-CHECK-NEXT: store i32 [[V2]], ptr addrspace(1) [[OUT]], align 4 ; DCE-CHECK-NEXT: ret void ; %tidx = call i32 @llvm.amdgcn.workitem.id.x() @@ -537,13 +547,15 @@ define amdgpu_kernel void @readfirstlane_random(ptr addrspace(1) %out) { ; PASS-CHECK-LABEL: define amdgpu_kernel void @readfirstlane_random( ; PASS-CHECK-SAME: ptr addrspace(1) [[OUT:%.*]]) #[[ATTR0]] { ; PASS-CHECK-NEXT: [[RANDOM:%.*]] = xor i32 123, 456 -; PASS-CHECK-NEXT: store i32 [[RANDOM]], ptr addrspace(1) [[OUT]], align 4 +; PASS-CHECK-NEXT: [[V:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[RANDOM]]) +; PASS-CHECK-NEXT: store i32 [[V]], ptr addrspace(1) [[OUT]], align 4 ; PASS-CHECK-NEXT: ret void ; ; DCE-CHECK-LABEL: define amdgpu_kernel void @readfirstlane_random( ; DCE-CHECK-SAME: ptr addrspace(1) [[OUT:%.*]]) #[[ATTR0]] { ; DCE-CHECK-NEXT: [[RANDOM:%.*]] = xor i32 123, 456 -; DCE-CHECK-NEXT: store i32 [[RANDOM]], ptr addrspace(1) [[OUT]], align 4 +; DCE-CHECK-NEXT: [[V:%.*]] = call i32 @llvm.amdgcn.readfirstlane.i32(i32 [[RANDOM]]) +; DCE-CHECK-NEXT: store i32 [[V]], ptr addrspace(1) [[OUT]], align 4 ; DCE-CHECK-NEXT: ret void ; %random = xor i32 123, 456 diff --git a/llvm/test/CodeGen/AMDGPU/call-args-inreg-bfloat.ll b/llvm/test/CodeGen/AMDGPU/call-args-inreg-bfloat.ll new file mode 100644 index 0000000..04e4724 --- /dev/null +++ b/llvm/test/CodeGen/AMDGPU/call-args-inreg-bfloat.ll @@ -0,0 +1,130 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck -enable-var-scope -check-prefix=GFX9 %s +; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 < %s | FileCheck -enable-var-scope -check-prefix=GFX11 %s + +; We've separated this file from call-args-inreg.ll since GlobalISel does not support the bfloat type. +; Ideally, we should merge the two files once that support lands. + +declare hidden void @external_void_func_bf16_inreg(bfloat inreg) #0 +declare hidden void @external_void_func_v2bf16_inreg(<2 x bfloat> inreg) #0 + +define void @test_call_external_void_func_bf16_inreg(bfloat inreg %arg) #0 { +; GFX9-LABEL: test_call_external_void_func_bf16_inreg: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: s_mov_b32 s17, s33 +; GFX9-NEXT: s_mov_b32 s33, s32 +; GFX9-NEXT: s_or_saveexec_b64 s[18:19], -1 +; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill +; GFX9-NEXT: s_mov_b64 exec, s[18:19] +; GFX9-NEXT: v_writelane_b32 v40, s17, 2 +; GFX9-NEXT: s_addk_i32 s32, 0x400 +; GFX9-NEXT: v_writelane_b32 v40, s30, 0 +; GFX9-NEXT: s_getpc_b64 s[18:19] +; GFX9-NEXT: s_add_u32 s18, s18, external_void_func_bf16_inreg@rel32@lo+4 +; GFX9-NEXT: s_addc_u32 s19, s19, external_void_func_bf16_inreg@rel32@hi+12 +; GFX9-NEXT: s_mov_b32 s0, s16 +; GFX9-NEXT: v_writelane_b32 v40, s31, 1 +; GFX9-NEXT: s_swappc_b64 s[30:31], s[18:19] +; GFX9-NEXT: v_readlane_b32 s31, v40, 1 +; GFX9-NEXT: v_readlane_b32 s30, v40, 0 +; GFX9-NEXT: s_mov_b32 s32, s33 +; GFX9-NEXT: v_readlane_b32 s4, v40, 2 +; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1 +; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload +; GFX9-NEXT: s_mov_b64 exec, s[6:7] +; GFX9-NEXT: s_mov_b32 s33, s4 +; GFX9-NEXT: s_waitcnt vmcnt(0) +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: test_call_external_void_func_bf16_inreg: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_mov_b32 s1, s33 +; GFX11-NEXT: s_mov_b32 s33, s32 +; GFX11-NEXT: s_or_saveexec_b32 s2, -1 +; GFX11-NEXT: scratch_store_b32 off, v40, s33 ; 4-byte Folded Spill +; GFX11-NEXT: s_mov_b32 exec_lo, s2 +; GFX11-NEXT: v_writelane_b32 v40, s1, 2 +; GFX11-NEXT: s_add_i32 s32, s32, 16 +; GFX11-NEXT: s_getpc_b64 s[2:3] +; GFX11-NEXT: s_add_u32 s2, s2, external_void_func_bf16_inreg@rel32@lo+4 +; GFX11-NEXT: s_addc_u32 s3, s3, external_void_func_bf16_inreg@rel32@hi+12 +; GFX11-NEXT: v_writelane_b32 v40, s30, 0 +; GFX11-NEXT: v_writelane_b32 v40, s31, 1 +; GFX11-NEXT: s_swappc_b64 s[30:31], s[2:3] +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_readlane_b32 s31, v40, 1 +; GFX11-NEXT: v_readlane_b32 s30, v40, 0 +; GFX11-NEXT: s_mov_b32 s32, s33 +; GFX11-NEXT: v_readlane_b32 s0, v40, 2 +; GFX11-NEXT: s_or_saveexec_b32 s1, -1 +; GFX11-NEXT: scratch_load_b32 v40, off, s33 ; 4-byte Folded Reload +; GFX11-NEXT: s_mov_b32 exec_lo, s1 +; GFX11-NEXT: s_mov_b32 s33, s0 +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: s_setpc_b64 s[30:31] + call void @external_void_func_bf16_inreg(bfloat inreg %arg) + ret void +} + +define void @test_call_external_void_func_v2bf16_inreg(<2 x bfloat> inreg %arg) #0 { +; GFX9-LABEL: test_call_external_void_func_v2bf16_inreg: +; GFX9: ; %bb.0: +; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-NEXT: s_mov_b32 s17, s33 +; GFX9-NEXT: s_mov_b32 s33, s32 +; GFX9-NEXT: s_or_saveexec_b64 s[18:19], -1 +; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill +; GFX9-NEXT: s_mov_b64 exec, s[18:19] +; GFX9-NEXT: v_writelane_b32 v40, s17, 2 +; GFX9-NEXT: s_addk_i32 s32, 0x400 +; GFX9-NEXT: v_writelane_b32 v40, s30, 0 +; GFX9-NEXT: s_getpc_b64 s[18:19] +; GFX9-NEXT: s_add_u32 s18, s18, external_void_func_v2bf16_inreg@rel32@lo+4 +; GFX9-NEXT: s_addc_u32 s19, s19, external_void_func_v2bf16_inreg@rel32@hi+12 +; GFX9-NEXT: s_mov_b32 s0, s16 +; GFX9-NEXT: v_writelane_b32 v40, s31, 1 +; GFX9-NEXT: s_swappc_b64 s[30:31], s[18:19] +; GFX9-NEXT: v_readlane_b32 s31, v40, 1 +; GFX9-NEXT: v_readlane_b32 s30, v40, 0 +; GFX9-NEXT: s_mov_b32 s32, s33 +; GFX9-NEXT: v_readlane_b32 s4, v40, 2 +; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1 +; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload +; GFX9-NEXT: s_mov_b64 exec, s[6:7] +; GFX9-NEXT: s_mov_b32 s33, s4 +; GFX9-NEXT: s_waitcnt vmcnt(0) +; GFX9-NEXT: s_setpc_b64 s[30:31] +; +; GFX11-LABEL: test_call_external_void_func_v2bf16_inreg: +; GFX11: ; %bb.0: +; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX11-NEXT: s_mov_b32 s1, s33 +; GFX11-NEXT: s_mov_b32 s33, s32 +; GFX11-NEXT: s_or_saveexec_b32 s2, -1 +; GFX11-NEXT: scratch_store_b32 off, v40, s33 ; 4-byte Folded Spill +; GFX11-NEXT: s_mov_b32 exec_lo, s2 +; GFX11-NEXT: v_writelane_b32 v40, s1, 2 +; GFX11-NEXT: s_add_i32 s32, s32, 16 +; GFX11-NEXT: s_getpc_b64 s[2:3] +; GFX11-NEXT: s_add_u32 s2, s2, external_void_func_v2bf16_inreg@rel32@lo+4 +; GFX11-NEXT: s_addc_u32 s3, s3, external_void_func_v2bf16_inreg@rel32@hi+12 +; GFX11-NEXT: v_writelane_b32 v40, s30, 0 +; GFX11-NEXT: v_writelane_b32 v40, s31, 1 +; GFX11-NEXT: s_swappc_b64 s[30:31], s[2:3] +; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) +; GFX11-NEXT: v_readlane_b32 s31, v40, 1 +; GFX11-NEXT: v_readlane_b32 s30, v40, 0 +; GFX11-NEXT: s_mov_b32 s32, s33 +; GFX11-NEXT: v_readlane_b32 s0, v40, 2 +; GFX11-NEXT: s_or_saveexec_b32 s1, -1 +; GFX11-NEXT: scratch_load_b32 v40, off, s33 ; 4-byte Folded Reload +; GFX11-NEXT: s_mov_b32 exec_lo, s1 +; GFX11-NEXT: s_mov_b32 s33, s0 +; GFX11-NEXT: s_waitcnt vmcnt(0) +; GFX11-NEXT: s_setpc_b64 s[30:31] + call void @external_void_func_v2bf16_inreg(<2 x bfloat> inreg %arg) + ret void +} + diff --git a/llvm/test/CodeGen/AMDGPU/call-args-inreg.ll b/llvm/test/CodeGen/AMDGPU/call-args-inreg.ll index d1cede6..f96007a 100644 --- a/llvm/test/CodeGen/AMDGPU/call-args-inreg.ll +++ b/llvm/test/CodeGen/AMDGPU/call-args-inreg.ll @@ -1,6 +1,8 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 -; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck -enable-var-scope -check-prefix=GFX9 %s -; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 < %s | FileCheck -enable-var-scope -check-prefix=GFX11 %s +; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck -enable-var-scope -check-prefixes=GFX9,SDAG %s +; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 < %s | FileCheck -enable-var-scope -check-prefixes=GFX11,SDAG %s +; RUN: llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck -enable-var-scope -check-prefixes=GFX9,GISEL %s +; RUN: llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 < %s | FileCheck -enable-var-scope -check-prefixes=GFX11,GISEL %s declare hidden void @external_void_func_i8_inreg(i8 inreg) #0 declare hidden void @external_void_func_i16_inreg(i32 inreg) #0 @@ -12,11 +14,9 @@ declare hidden void @external_void_func_v4i32_inreg(<4 x i32> inreg) #0 declare hidden void @external_void_func_v8i32_inreg(<8 x i32> inreg) #0 declare hidden void @external_void_func_v16i32_inreg(<16 x i32> inreg) #0 declare hidden void @external_void_func_f16_inreg(half inreg) #0 -declare hidden void @external_void_func_bf16_inreg(bfloat inreg) #0 declare hidden void @external_void_func_f32_inreg(float inreg) #0 declare hidden void @external_void_func_f64_inreg(double inreg) #0 declare hidden void @external_void_func_v2f16_inreg(<2 x half> inreg) #0 -declare hidden void @external_void_func_v2bf16_inreg(<2 x bfloat> inreg) #0 declare hidden void @external_void_func_v3f16_inreg(<3 x half> inreg) #0 declare hidden void @external_void_func_v4f16_inreg(<4 x half> inreg) #0 @@ -212,35 +212,6 @@ define void @test_call_external_void_func_i32_inreg(i32 inreg %arg) #0 { } define void @test_call_external_void_func_i64_inreg(i64 inreg %arg) #0 { -; GFX9-LABEL: test_call_external_void_func_i64_inreg: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX9-NEXT: s_mov_b32 s18, s33 -; GFX9-NEXT: s_mov_b32 s33, s32 -; GFX9-NEXT: s_or_saveexec_b64 s[20:21], -1 -; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill -; GFX9-NEXT: s_mov_b64 exec, s[20:21] -; GFX9-NEXT: v_writelane_b32 v40, s18, 2 -; GFX9-NEXT: s_addk_i32 s32, 0x400 -; GFX9-NEXT: v_writelane_b32 v40, s30, 0 -; GFX9-NEXT: s_getpc_b64 s[18:19] -; GFX9-NEXT: s_add_u32 s18, s18, external_void_func_i64_inreg@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s19, s19, external_void_func_i64_inreg@rel32@hi+12 -; GFX9-NEXT: s_mov_b32 s1, s17 -; GFX9-NEXT: s_mov_b32 s0, s16 -; GFX9-NEXT: v_writelane_b32 v40, s31, 1 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[18:19] -; GFX9-NEXT: v_readlane_b32 s31, v40, 1 -; GFX9-NEXT: v_readlane_b32 s30, v40, 0 -; GFX9-NEXT: s_mov_b32 s32, s33 -; GFX9-NEXT: v_readlane_b32 s4, v40, 2 -; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1 -; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload -; GFX9-NEXT: s_mov_b64 exec, s[6:7] -; GFX9-NEXT: s_mov_b32 s33, s4 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_setpc_b64 s[30:31] -; ; GFX11-LABEL: test_call_external_void_func_i64_inreg: ; GFX11: ; %bb.0: ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) @@ -273,35 +244,6 @@ define void @test_call_external_void_func_i64_inreg(i64 inreg %arg) #0 { } define void @test_call_external_void_func_v2i32_inreg(<2 x i32> inreg %arg) #0 { -; GFX9-LABEL: test_call_external_void_func_v2i32_inreg: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX9-NEXT: s_mov_b32 s18, s33 -; GFX9-NEXT: s_mov_b32 s33, s32 -; GFX9-NEXT: s_or_saveexec_b64 s[20:21], -1 -; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill -; GFX9-NEXT: s_mov_b64 exec, s[20:21] -; GFX9-NEXT: v_writelane_b32 v40, s18, 2 -; GFX9-NEXT: s_addk_i32 s32, 0x400 -; GFX9-NEXT: v_writelane_b32 v40, s30, 0 -; GFX9-NEXT: s_getpc_b64 s[18:19] -; GFX9-NEXT: s_add_u32 s18, s18, external_void_func_v2i32_inreg@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s19, s19, external_void_func_v2i32_inreg@rel32@hi+12 -; GFX9-NEXT: s_mov_b32 s1, s17 -; GFX9-NEXT: s_mov_b32 s0, s16 -; GFX9-NEXT: v_writelane_b32 v40, s31, 1 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[18:19] -; GFX9-NEXT: v_readlane_b32 s31, v40, 1 -; GFX9-NEXT: v_readlane_b32 s30, v40, 0 -; GFX9-NEXT: s_mov_b32 s32, s33 -; GFX9-NEXT: v_readlane_b32 s4, v40, 2 -; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1 -; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload -; GFX9-NEXT: s_mov_b64 exec, s[6:7] -; GFX9-NEXT: s_mov_b32 s33, s4 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_setpc_b64 s[30:31] -; ; GFX11-LABEL: test_call_external_void_func_v2i32_inreg: ; GFX11: ; %bb.0: ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) @@ -334,36 +276,6 @@ define void @test_call_external_void_func_v2i32_inreg(<2 x i32> inreg %arg) #0 { } define void @test_call_external_void_func_v3i32_inreg(<3 x i32> inreg %arg) #0 { -; GFX9-LABEL: test_call_external_void_func_v3i32_inreg: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX9-NEXT: s_mov_b32 s19, s33 -; GFX9-NEXT: s_mov_b32 s33, s32 -; GFX9-NEXT: s_or_saveexec_b64 s[20:21], -1 -; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill -; GFX9-NEXT: s_mov_b64 exec, s[20:21] -; GFX9-NEXT: v_writelane_b32 v40, s19, 2 -; GFX9-NEXT: s_addk_i32 s32, 0x400 -; GFX9-NEXT: v_writelane_b32 v40, s30, 0 -; GFX9-NEXT: s_getpc_b64 s[20:21] -; GFX9-NEXT: s_add_u32 s20, s20, external_void_func_v3i32_inreg@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s21, s21, external_void_func_v3i32_inreg@rel32@hi+12 -; GFX9-NEXT: s_mov_b32 s2, s18 -; GFX9-NEXT: s_mov_b32 s1, s17 -; GFX9-NEXT: s_mov_b32 s0, s16 -; GFX9-NEXT: v_writelane_b32 v40, s31, 1 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[20:21] -; GFX9-NEXT: v_readlane_b32 s31, v40, 1 -; GFX9-NEXT: v_readlane_b32 s30, v40, 0 -; GFX9-NEXT: s_mov_b32 s32, s33 -; GFX9-NEXT: v_readlane_b32 s4, v40, 2 -; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1 -; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload -; GFX9-NEXT: s_mov_b64 exec, s[6:7] -; GFX9-NEXT: s_mov_b32 s33, s4 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_setpc_b64 s[30:31] -; ; GFX11-LABEL: test_call_external_void_func_v3i32_inreg: ; GFX11: ; %bb.0: ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) @@ -396,37 +308,6 @@ define void @test_call_external_void_func_v3i32_inreg(<3 x i32> inreg %arg) #0 { } define void @test_call_external_void_func_v4i32_inreg(<4 x i32> inreg %arg) #0 { -; GFX9-LABEL: test_call_external_void_func_v4i32_inreg: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX9-NEXT: s_mov_b32 s20, s33 -; GFX9-NEXT: s_mov_b32 s33, s32 -; GFX9-NEXT: s_or_saveexec_b64 s[22:23], -1 -; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill -; GFX9-NEXT: s_mov_b64 exec, s[22:23] -; GFX9-NEXT: v_writelane_b32 v40, s20, 2 -; GFX9-NEXT: s_addk_i32 s32, 0x400 -; GFX9-NEXT: v_writelane_b32 v40, s30, 0 -; GFX9-NEXT: s_getpc_b64 s[20:21] -; GFX9-NEXT: s_add_u32 s20, s20, external_void_func_v4i32_inreg@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s21, s21, external_void_func_v4i32_inreg@rel32@hi+12 -; GFX9-NEXT: s_mov_b32 s3, s19 -; GFX9-NEXT: s_mov_b32 s2, s18 -; GFX9-NEXT: s_mov_b32 s1, s17 -; GFX9-NEXT: s_mov_b32 s0, s16 -; GFX9-NEXT: v_writelane_b32 v40, s31, 1 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[20:21] -; GFX9-NEXT: v_readlane_b32 s31, v40, 1 -; GFX9-NEXT: v_readlane_b32 s30, v40, 0 -; GFX9-NEXT: s_mov_b32 s32, s33 -; GFX9-NEXT: v_readlane_b32 s4, v40, 2 -; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1 -; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload -; GFX9-NEXT: s_mov_b64 exec, s[6:7] -; GFX9-NEXT: s_mov_b32 s33, s4 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_setpc_b64 s[30:31] -; ; GFX11-LABEL: test_call_external_void_func_v4i32_inreg: ; GFX11: ; %bb.0: ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) @@ -459,41 +340,6 @@ define void @test_call_external_void_func_v4i32_inreg(<4 x i32> inreg %arg) #0 { } define void @test_call_external_void_func_v8i32_inreg(<8 x i32> inreg %arg) #0 { -; GFX9-LABEL: test_call_external_void_func_v8i32_inreg: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX9-NEXT: s_mov_b32 s24, s33 -; GFX9-NEXT: s_mov_b32 s33, s32 -; GFX9-NEXT: s_or_saveexec_b64 s[26:27], -1 -; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill -; GFX9-NEXT: s_mov_b64 exec, s[26:27] -; GFX9-NEXT: v_writelane_b32 v40, s24, 2 -; GFX9-NEXT: s_addk_i32 s32, 0x400 -; GFX9-NEXT: v_writelane_b32 v40, s30, 0 -; GFX9-NEXT: s_getpc_b64 s[24:25] -; GFX9-NEXT: s_add_u32 s24, s24, external_void_func_v8i32_inreg@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s25, s25, external_void_func_v8i32_inreg@rel32@hi+12 -; GFX9-NEXT: s_mov_b32 s3, s19 -; GFX9-NEXT: s_mov_b32 s2, s18 -; GFX9-NEXT: s_mov_b32 s1, s17 -; GFX9-NEXT: s_mov_b32 s0, s16 -; GFX9-NEXT: s_mov_b32 s16, s20 -; GFX9-NEXT: s_mov_b32 s17, s21 -; GFX9-NEXT: s_mov_b32 s18, s22 -; GFX9-NEXT: s_mov_b32 s19, s23 -; GFX9-NEXT: v_writelane_b32 v40, s31, 1 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[24:25] -; GFX9-NEXT: v_readlane_b32 s31, v40, 1 -; GFX9-NEXT: v_readlane_b32 s30, v40, 0 -; GFX9-NEXT: s_mov_b32 s32, s33 -; GFX9-NEXT: v_readlane_b32 s4, v40, 2 -; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1 -; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload -; GFX9-NEXT: s_mov_b64 exec, s[6:7] -; GFX9-NEXT: s_mov_b32 s33, s4 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_setpc_b64 s[30:31] -; ; GFX11-LABEL: test_call_external_void_func_v8i32_inreg: ; GFX11: ; %bb.0: ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) @@ -585,66 +431,6 @@ define void @test_call_external_void_func_f16_inreg(half inreg %arg) #0 { ret void } -define void @test_call_external_void_func_bf16_inreg(bfloat inreg %arg) #0 { -; GFX9-LABEL: test_call_external_void_func_bf16_inreg: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX9-NEXT: s_mov_b32 s17, s33 -; GFX9-NEXT: s_mov_b32 s33, s32 -; GFX9-NEXT: s_or_saveexec_b64 s[18:19], -1 -; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill -; GFX9-NEXT: s_mov_b64 exec, s[18:19] -; GFX9-NEXT: v_writelane_b32 v40, s17, 2 -; GFX9-NEXT: s_addk_i32 s32, 0x400 -; GFX9-NEXT: v_writelane_b32 v40, s30, 0 -; GFX9-NEXT: s_getpc_b64 s[18:19] -; GFX9-NEXT: s_add_u32 s18, s18, external_void_func_bf16_inreg@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s19, s19, external_void_func_bf16_inreg@rel32@hi+12 -; GFX9-NEXT: s_mov_b32 s0, s16 -; GFX9-NEXT: v_writelane_b32 v40, s31, 1 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[18:19] -; GFX9-NEXT: v_readlane_b32 s31, v40, 1 -; GFX9-NEXT: v_readlane_b32 s30, v40, 0 -; GFX9-NEXT: s_mov_b32 s32, s33 -; GFX9-NEXT: v_readlane_b32 s4, v40, 2 -; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1 -; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload -; GFX9-NEXT: s_mov_b64 exec, s[6:7] -; GFX9-NEXT: s_mov_b32 s33, s4 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_setpc_b64 s[30:31] -; -; GFX11-LABEL: test_call_external_void_func_bf16_inreg: -; GFX11: ; %bb.0: -; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX11-NEXT: s_mov_b32 s1, s33 -; GFX11-NEXT: s_mov_b32 s33, s32 -; GFX11-NEXT: s_or_saveexec_b32 s2, -1 -; GFX11-NEXT: scratch_store_b32 off, v40, s33 ; 4-byte Folded Spill -; GFX11-NEXT: s_mov_b32 exec_lo, s2 -; GFX11-NEXT: v_writelane_b32 v40, s1, 2 -; GFX11-NEXT: s_add_i32 s32, s32, 16 -; GFX11-NEXT: s_getpc_b64 s[2:3] -; GFX11-NEXT: s_add_u32 s2, s2, external_void_func_bf16_inreg@rel32@lo+4 -; GFX11-NEXT: s_addc_u32 s3, s3, external_void_func_bf16_inreg@rel32@hi+12 -; GFX11-NEXT: v_writelane_b32 v40, s30, 0 -; GFX11-NEXT: v_writelane_b32 v40, s31, 1 -; GFX11-NEXT: s_swappc_b64 s[30:31], s[2:3] -; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-NEXT: v_readlane_b32 s31, v40, 1 -; GFX11-NEXT: v_readlane_b32 s30, v40, 0 -; GFX11-NEXT: s_mov_b32 s32, s33 -; GFX11-NEXT: v_readlane_b32 s0, v40, 2 -; GFX11-NEXT: s_or_saveexec_b32 s1, -1 -; GFX11-NEXT: scratch_load_b32 v40, off, s33 ; 4-byte Folded Reload -; GFX11-NEXT: s_mov_b32 exec_lo, s1 -; GFX11-NEXT: s_mov_b32 s33, s0 -; GFX11-NEXT: s_waitcnt vmcnt(0) -; GFX11-NEXT: s_setpc_b64 s[30:31] - call void @external_void_func_bf16_inreg(bfloat inreg %arg) - ret void -} - define void @test_call_external_void_func_f32_inreg(float inreg %arg) #0 { ; GFX9-LABEL: test_call_external_void_func_f32_inreg: ; GFX9: ; %bb.0: @@ -706,35 +492,6 @@ define void @test_call_external_void_func_f32_inreg(float inreg %arg) #0 { } define void @test_call_external_void_func_f64_inreg(double inreg %arg) #0 { -; GFX9-LABEL: test_call_external_void_func_f64_inreg: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX9-NEXT: s_mov_b32 s18, s33 -; GFX9-NEXT: s_mov_b32 s33, s32 -; GFX9-NEXT: s_or_saveexec_b64 s[20:21], -1 -; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill -; GFX9-NEXT: s_mov_b64 exec, s[20:21] -; GFX9-NEXT: v_writelane_b32 v40, s18, 2 -; GFX9-NEXT: s_addk_i32 s32, 0x400 -; GFX9-NEXT: v_writelane_b32 v40, s30, 0 -; GFX9-NEXT: s_getpc_b64 s[18:19] -; GFX9-NEXT: s_add_u32 s18, s18, external_void_func_f64_inreg@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s19, s19, external_void_func_f64_inreg@rel32@hi+12 -; GFX9-NEXT: s_mov_b32 s1, s17 -; GFX9-NEXT: s_mov_b32 s0, s16 -; GFX9-NEXT: v_writelane_b32 v40, s31, 1 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[18:19] -; GFX9-NEXT: v_readlane_b32 s31, v40, 1 -; GFX9-NEXT: v_readlane_b32 s30, v40, 0 -; GFX9-NEXT: s_mov_b32 s32, s33 -; GFX9-NEXT: v_readlane_b32 s4, v40, 2 -; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1 -; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload -; GFX9-NEXT: s_mov_b64 exec, s[6:7] -; GFX9-NEXT: s_mov_b32 s33, s4 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_setpc_b64 s[30:31] -; ; GFX11-LABEL: test_call_external_void_func_f64_inreg: ; GFX11: ; %bb.0: ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) @@ -826,97 +583,7 @@ define void @test_call_external_void_func_v2f16_inreg(<2 x half> inreg %arg) #0 ret void } - -define void @test_call_external_void_func_v2bf16_inreg(<2 x bfloat> inreg %arg) #0 { -; GFX9-LABEL: test_call_external_void_func_v2bf16_inreg: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX9-NEXT: s_mov_b32 s17, s33 -; GFX9-NEXT: s_mov_b32 s33, s32 -; GFX9-NEXT: s_or_saveexec_b64 s[18:19], -1 -; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill -; GFX9-NEXT: s_mov_b64 exec, s[18:19] -; GFX9-NEXT: v_writelane_b32 v40, s17, 2 -; GFX9-NEXT: s_addk_i32 s32, 0x400 -; GFX9-NEXT: v_writelane_b32 v40, s30, 0 -; GFX9-NEXT: s_getpc_b64 s[18:19] -; GFX9-NEXT: s_add_u32 s18, s18, external_void_func_v2bf16_inreg@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s19, s19, external_void_func_v2bf16_inreg@rel32@hi+12 -; GFX9-NEXT: s_mov_b32 s0, s16 -; GFX9-NEXT: v_writelane_b32 v40, s31, 1 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[18:19] -; GFX9-NEXT: v_readlane_b32 s31, v40, 1 -; GFX9-NEXT: v_readlane_b32 s30, v40, 0 -; GFX9-NEXT: s_mov_b32 s32, s33 -; GFX9-NEXT: v_readlane_b32 s4, v40, 2 -; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1 -; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload -; GFX9-NEXT: s_mov_b64 exec, s[6:7] -; GFX9-NEXT: s_mov_b32 s33, s4 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_setpc_b64 s[30:31] -; -; GFX11-LABEL: test_call_external_void_func_v2bf16_inreg: -; GFX11: ; %bb.0: -; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX11-NEXT: s_mov_b32 s1, s33 -; GFX11-NEXT: s_mov_b32 s33, s32 -; GFX11-NEXT: s_or_saveexec_b32 s2, -1 -; GFX11-NEXT: scratch_store_b32 off, v40, s33 ; 4-byte Folded Spill -; GFX11-NEXT: s_mov_b32 exec_lo, s2 -; GFX11-NEXT: v_writelane_b32 v40, s1, 2 -; GFX11-NEXT: s_add_i32 s32, s32, 16 -; GFX11-NEXT: s_getpc_b64 s[2:3] -; GFX11-NEXT: s_add_u32 s2, s2, external_void_func_v2bf16_inreg@rel32@lo+4 -; GFX11-NEXT: s_addc_u32 s3, s3, external_void_func_v2bf16_inreg@rel32@hi+12 -; GFX11-NEXT: v_writelane_b32 v40, s30, 0 -; GFX11-NEXT: v_writelane_b32 v40, s31, 1 -; GFX11-NEXT: s_swappc_b64 s[30:31], s[2:3] -; GFX11-NEXT: s_delay_alu instid0(VALU_DEP_1) -; GFX11-NEXT: v_readlane_b32 s31, v40, 1 -; GFX11-NEXT: v_readlane_b32 s30, v40, 0 -; GFX11-NEXT: s_mov_b32 s32, s33 -; GFX11-NEXT: v_readlane_b32 s0, v40, 2 -; GFX11-NEXT: s_or_saveexec_b32 s1, -1 -; GFX11-NEXT: scratch_load_b32 v40, off, s33 ; 4-byte Folded Reload -; GFX11-NEXT: s_mov_b32 exec_lo, s1 -; GFX11-NEXT: s_mov_b32 s33, s0 -; GFX11-NEXT: s_waitcnt vmcnt(0) -; GFX11-NEXT: s_setpc_b64 s[30:31] - call void @external_void_func_v2bf16_inreg(<2 x bfloat> inreg %arg) - ret void -} - define void @test_call_external_void_func_v3f16_inreg(<3 x half> inreg %arg) #0 { -; GFX9-LABEL: test_call_external_void_func_v3f16_inreg: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX9-NEXT: s_mov_b32 s18, s33 -; GFX9-NEXT: s_mov_b32 s33, s32 -; GFX9-NEXT: s_or_saveexec_b64 s[20:21], -1 -; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill -; GFX9-NEXT: s_mov_b64 exec, s[20:21] -; GFX9-NEXT: v_writelane_b32 v40, s18, 2 -; GFX9-NEXT: s_addk_i32 s32, 0x400 -; GFX9-NEXT: v_writelane_b32 v40, s30, 0 -; GFX9-NEXT: s_getpc_b64 s[18:19] -; GFX9-NEXT: s_add_u32 s18, s18, external_void_func_v3f16_inreg@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s19, s19, external_void_func_v3f16_inreg@rel32@hi+12 -; GFX9-NEXT: s_mov_b32 s1, s17 -; GFX9-NEXT: s_mov_b32 s0, s16 -; GFX9-NEXT: v_writelane_b32 v40, s31, 1 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[18:19] -; GFX9-NEXT: v_readlane_b32 s31, v40, 1 -; GFX9-NEXT: v_readlane_b32 s30, v40, 0 -; GFX9-NEXT: s_mov_b32 s32, s33 -; GFX9-NEXT: v_readlane_b32 s4, v40, 2 -; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1 -; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload -; GFX9-NEXT: s_mov_b64 exec, s[6:7] -; GFX9-NEXT: s_mov_b32 s33, s4 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_setpc_b64 s[30:31] -; ; GFX11-LABEL: test_call_external_void_func_v3f16_inreg: ; GFX11: ; %bb.0: ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) @@ -949,35 +616,6 @@ define void @test_call_external_void_func_v3f16_inreg(<3 x half> inreg %arg) #0 } define void @test_call_external_void_func_v4f16_inreg(<4 x half> inreg %arg) #0 { -; GFX9-LABEL: test_call_external_void_func_v4f16_inreg: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX9-NEXT: s_mov_b32 s18, s33 -; GFX9-NEXT: s_mov_b32 s33, s32 -; GFX9-NEXT: s_or_saveexec_b64 s[20:21], -1 -; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill -; GFX9-NEXT: s_mov_b64 exec, s[20:21] -; GFX9-NEXT: v_writelane_b32 v40, s18, 2 -; GFX9-NEXT: s_addk_i32 s32, 0x400 -; GFX9-NEXT: v_writelane_b32 v40, s30, 0 -; GFX9-NEXT: s_getpc_b64 s[18:19] -; GFX9-NEXT: s_add_u32 s18, s18, external_void_func_v4f16_inreg@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s19, s19, external_void_func_v4f16_inreg@rel32@hi+12 -; GFX9-NEXT: s_mov_b32 s1, s17 -; GFX9-NEXT: s_mov_b32 s0, s16 -; GFX9-NEXT: v_writelane_b32 v40, s31, 1 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[18:19] -; GFX9-NEXT: v_readlane_b32 s31, v40, 1 -; GFX9-NEXT: v_readlane_b32 s30, v40, 0 -; GFX9-NEXT: s_mov_b32 s32, s33 -; GFX9-NEXT: v_readlane_b32 s4, v40, 2 -; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1 -; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload -; GFX9-NEXT: s_mov_b64 exec, s[6:7] -; GFX9-NEXT: s_mov_b32 s33, s4 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_setpc_b64 s[30:31] -; ; GFX11-LABEL: test_call_external_void_func_v4f16_inreg: ; GFX11: ; %bb.0: ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) @@ -1010,35 +648,6 @@ define void @test_call_external_void_func_v4f16_inreg(<4 x half> inreg %arg) #0 } define void @test_call_external_void_func_p0_inreg(ptr inreg %arg) #0 { -; GFX9-LABEL: test_call_external_void_func_p0_inreg: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX9-NEXT: s_mov_b32 s18, s33 -; GFX9-NEXT: s_mov_b32 s33, s32 -; GFX9-NEXT: s_or_saveexec_b64 s[20:21], -1 -; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill -; GFX9-NEXT: s_mov_b64 exec, s[20:21] -; GFX9-NEXT: v_writelane_b32 v40, s18, 2 -; GFX9-NEXT: s_addk_i32 s32, 0x400 -; GFX9-NEXT: v_writelane_b32 v40, s30, 0 -; GFX9-NEXT: s_getpc_b64 s[18:19] -; GFX9-NEXT: s_add_u32 s18, s18, external_void_func_p0_inreg@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s19, s19, external_void_func_p0_inreg@rel32@hi+12 -; GFX9-NEXT: s_mov_b32 s1, s17 -; GFX9-NEXT: s_mov_b32 s0, s16 -; GFX9-NEXT: v_writelane_b32 v40, s31, 1 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[18:19] -; GFX9-NEXT: v_readlane_b32 s31, v40, 1 -; GFX9-NEXT: v_readlane_b32 s30, v40, 0 -; GFX9-NEXT: s_mov_b32 s32, s33 -; GFX9-NEXT: v_readlane_b32 s4, v40, 2 -; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1 -; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload -; GFX9-NEXT: s_mov_b64 exec, s[6:7] -; GFX9-NEXT: s_mov_b32 s33, s4 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_setpc_b64 s[30:31] -; ; GFX11-LABEL: test_call_external_void_func_p0_inreg: ; GFX11: ; %bb.0: ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) @@ -1071,35 +680,6 @@ define void @test_call_external_void_func_p0_inreg(ptr inreg %arg) #0 { } define void @test_call_external_void_func_p1_inreg(ptr addrspace(1) inreg %arg) #0 { -; GFX9-LABEL: test_call_external_void_func_p1_inreg: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX9-NEXT: s_mov_b32 s18, s33 -; GFX9-NEXT: s_mov_b32 s33, s32 -; GFX9-NEXT: s_or_saveexec_b64 s[20:21], -1 -; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill -; GFX9-NEXT: s_mov_b64 exec, s[20:21] -; GFX9-NEXT: v_writelane_b32 v40, s18, 2 -; GFX9-NEXT: s_addk_i32 s32, 0x400 -; GFX9-NEXT: v_writelane_b32 v40, s30, 0 -; GFX9-NEXT: s_getpc_b64 s[18:19] -; GFX9-NEXT: s_add_u32 s18, s18, external_void_func_p1_inreg@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s19, s19, external_void_func_p1_inreg@rel32@hi+12 -; GFX9-NEXT: s_mov_b32 s1, s17 -; GFX9-NEXT: s_mov_b32 s0, s16 -; GFX9-NEXT: v_writelane_b32 v40, s31, 1 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[18:19] -; GFX9-NEXT: v_readlane_b32 s31, v40, 1 -; GFX9-NEXT: v_readlane_b32 s30, v40, 0 -; GFX9-NEXT: s_mov_b32 s32, s33 -; GFX9-NEXT: v_readlane_b32 s4, v40, 2 -; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1 -; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload -; GFX9-NEXT: s_mov_b64 exec, s[6:7] -; GFX9-NEXT: s_mov_b32 s33, s4 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_setpc_b64 s[30:31] -; ; GFX11-LABEL: test_call_external_void_func_p1_inreg: ; GFX11: ; %bb.0: ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) @@ -1192,37 +772,6 @@ define void @test_call_external_void_func_p3_inreg(ptr addrspace(3) inreg %arg) } define void @test_call_external_void_func_v2p1_inreg(<2 x ptr addrspace(1)> inreg %arg) #0 { -; GFX9-LABEL: test_call_external_void_func_v2p1_inreg: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX9-NEXT: s_mov_b32 s20, s33 -; GFX9-NEXT: s_mov_b32 s33, s32 -; GFX9-NEXT: s_or_saveexec_b64 s[22:23], -1 -; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill -; GFX9-NEXT: s_mov_b64 exec, s[22:23] -; GFX9-NEXT: v_writelane_b32 v40, s20, 2 -; GFX9-NEXT: s_addk_i32 s32, 0x400 -; GFX9-NEXT: v_writelane_b32 v40, s30, 0 -; GFX9-NEXT: s_getpc_b64 s[20:21] -; GFX9-NEXT: s_add_u32 s20, s20, external_void_func_v2p1_inreg@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s21, s21, external_void_func_v2p1_inreg@rel32@hi+12 -; GFX9-NEXT: s_mov_b32 s3, s19 -; GFX9-NEXT: s_mov_b32 s2, s18 -; GFX9-NEXT: s_mov_b32 s1, s17 -; GFX9-NEXT: s_mov_b32 s0, s16 -; GFX9-NEXT: v_writelane_b32 v40, s31, 1 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[20:21] -; GFX9-NEXT: v_readlane_b32 s31, v40, 1 -; GFX9-NEXT: v_readlane_b32 s30, v40, 0 -; GFX9-NEXT: s_mov_b32 s32, s33 -; GFX9-NEXT: v_readlane_b32 s4, v40, 2 -; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1 -; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload -; GFX9-NEXT: s_mov_b64 exec, s[6:7] -; GFX9-NEXT: s_mov_b32 s33, s4 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_setpc_b64 s[30:31] -; ; GFX11-LABEL: test_call_external_void_func_v2p1_inreg: ; GFX11: ; %bb.0: ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) @@ -1255,35 +804,6 @@ define void @test_call_external_void_func_v2p1_inreg(<2 x ptr addrspace(1)> inre } define void @test_call_external_void_func_v2p5_inreg(<2 x ptr addrspace(5)> inreg %arg) #0 { -; GFX9-LABEL: test_call_external_void_func_v2p5_inreg: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX9-NEXT: s_mov_b32 s18, s33 -; GFX9-NEXT: s_mov_b32 s33, s32 -; GFX9-NEXT: s_or_saveexec_b64 s[20:21], -1 -; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill -; GFX9-NEXT: s_mov_b64 exec, s[20:21] -; GFX9-NEXT: v_writelane_b32 v40, s18, 2 -; GFX9-NEXT: s_addk_i32 s32, 0x400 -; GFX9-NEXT: v_writelane_b32 v40, s30, 0 -; GFX9-NEXT: s_getpc_b64 s[18:19] -; GFX9-NEXT: s_add_u32 s18, s18, external_void_func_v2p5_inreg@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s19, s19, external_void_func_v2p5_inreg@rel32@hi+12 -; GFX9-NEXT: s_mov_b32 s1, s17 -; GFX9-NEXT: s_mov_b32 s0, s16 -; GFX9-NEXT: v_writelane_b32 v40, s31, 1 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[18:19] -; GFX9-NEXT: v_readlane_b32 s31, v40, 1 -; GFX9-NEXT: v_readlane_b32 s30, v40, 0 -; GFX9-NEXT: s_mov_b32 s32, s33 -; GFX9-NEXT: v_readlane_b32 s4, v40, 2 -; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1 -; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload -; GFX9-NEXT: s_mov_b64 exec, s[6:7] -; GFX9-NEXT: s_mov_b32 s33, s4 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_setpc_b64 s[30:31] -; ; GFX11-LABEL: test_call_external_void_func_v2p5_inreg: ; GFX11: ; %bb.0: ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) @@ -1316,38 +836,6 @@ define void @test_call_external_void_func_v2p5_inreg(<2 x ptr addrspace(5)> inre } define void @test_call_external_void_func_i64_inreg_i32_inreg_i64_inreg(i64 inreg %arg0, i32 inreg %arg1, i64 inreg %arg2) #0 { -; GFX9-LABEL: test_call_external_void_func_i64_inreg_i32_inreg_i64_inreg: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX9-NEXT: s_mov_b32 s21, s33 -; GFX9-NEXT: s_mov_b32 s33, s32 -; GFX9-NEXT: s_or_saveexec_b64 s[22:23], -1 -; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill -; GFX9-NEXT: s_mov_b64 exec, s[22:23] -; GFX9-NEXT: v_writelane_b32 v40, s21, 2 -; GFX9-NEXT: s_addk_i32 s32, 0x400 -; GFX9-NEXT: v_writelane_b32 v40, s30, 0 -; GFX9-NEXT: s_getpc_b64 s[22:23] -; GFX9-NEXT: s_add_u32 s22, s22, external_void_func_i64_inreg_i32_inreg_i64_inreg@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s23, s23, external_void_func_i64_inreg_i32_inreg_i64_inreg@rel32@hi+12 -; GFX9-NEXT: s_mov_b32 s3, s19 -; GFX9-NEXT: s_mov_b32 s2, s18 -; GFX9-NEXT: s_mov_b32 s1, s17 -; GFX9-NEXT: s_mov_b32 s0, s16 -; GFX9-NEXT: s_mov_b32 s16, s20 -; GFX9-NEXT: v_writelane_b32 v40, s31, 1 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[22:23] -; GFX9-NEXT: v_readlane_b32 s31, v40, 1 -; GFX9-NEXT: v_readlane_b32 s30, v40, 0 -; GFX9-NEXT: s_mov_b32 s32, s33 -; GFX9-NEXT: v_readlane_b32 s4, v40, 2 -; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1 -; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload -; GFX9-NEXT: s_mov_b64 exec, s[6:7] -; GFX9-NEXT: s_mov_b32 s33, s4 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_setpc_b64 s[30:31] -; ; GFX11-LABEL: test_call_external_void_func_i64_inreg_i32_inreg_i64_inreg: ; GFX11: ; %bb.0: ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) @@ -1380,46 +868,6 @@ define void @test_call_external_void_func_i64_inreg_i32_inreg_i64_inreg(i64 inre } define void @test_call_external_void_func_a15i32_inreg([13 x i32] inreg %arg0) #0 { -; GFX9-LABEL: test_call_external_void_func_a15i32_inreg: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX9-NEXT: s_mov_b32 s29, s33 -; GFX9-NEXT: s_mov_b32 s33, s32 -; GFX9-NEXT: s_or_saveexec_b64 s[40:41], -1 -; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill -; GFX9-NEXT: s_mov_b64 exec, s[40:41] -; GFX9-NEXT: v_writelane_b32 v40, s29, 2 -; GFX9-NEXT: s_addk_i32 s32, 0x400 -; GFX9-NEXT: v_writelane_b32 v40, s30, 0 -; GFX9-NEXT: s_getpc_b64 s[40:41] -; GFX9-NEXT: s_add_u32 s40, s40, external_void_func_a15i32_inreg@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s41, s41, external_void_func_a15i32_inreg@rel32@hi+12 -; GFX9-NEXT: s_mov_b32 s3, s19 -; GFX9-NEXT: s_mov_b32 s2, s18 -; GFX9-NEXT: s_mov_b32 s1, s17 -; GFX9-NEXT: s_mov_b32 s0, s16 -; GFX9-NEXT: s_mov_b32 s16, s20 -; GFX9-NEXT: s_mov_b32 s17, s21 -; GFX9-NEXT: s_mov_b32 s18, s22 -; GFX9-NEXT: s_mov_b32 s19, s23 -; GFX9-NEXT: s_mov_b32 s20, s24 -; GFX9-NEXT: s_mov_b32 s21, s25 -; GFX9-NEXT: s_mov_b32 s22, s26 -; GFX9-NEXT: s_mov_b32 s23, s27 -; GFX9-NEXT: s_mov_b32 s24, s28 -; GFX9-NEXT: v_writelane_b32 v40, s31, 1 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[40:41] -; GFX9-NEXT: v_readlane_b32 s31, v40, 1 -; GFX9-NEXT: v_readlane_b32 s30, v40, 0 -; GFX9-NEXT: s_mov_b32 s32, s33 -; GFX9-NEXT: v_readlane_b32 s4, v40, 2 -; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1 -; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload -; GFX9-NEXT: s_mov_b64 exec, s[6:7] -; GFX9-NEXT: s_mov_b32 s33, s4 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_setpc_b64 s[30:31] -; ; GFX11-LABEL: test_call_external_void_func_a15i32_inreg: ; GFX11: ; %bb.0: ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) @@ -1454,47 +902,6 @@ define void @test_call_external_void_func_a15i32_inreg([13 x i32] inreg %arg0) # ; FIXME: This should also fail define void @test_call_external_void_func_a15i32_inreg_i32_inreg([13 x i32] inreg %arg0, i32 inreg %arg1) #1 { -; GFX9-LABEL: test_call_external_void_func_a15i32_inreg_i32_inreg: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX9-NEXT: s_mov_b32 s21, s33 -; GFX9-NEXT: s_mov_b32 s33, s32 -; GFX9-NEXT: s_or_saveexec_b64 s[22:23], -1 -; GFX9-NEXT: buffer_store_dword v40, off, s[0:3], s33 ; 4-byte Folded Spill -; GFX9-NEXT: s_mov_b64 exec, s[22:23] -; GFX9-NEXT: v_writelane_b32 v40, s21, 2 -; GFX9-NEXT: s_addk_i32 s32, 0x400 -; GFX9-NEXT: v_writelane_b32 v40, s30, 0 -; GFX9-NEXT: s_getpc_b64 s[22:23] -; GFX9-NEXT: s_add_u32 s22, s22, external_void_func_a15i32_inreg_i32_inreg__noimplicit@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s23, s23, external_void_func_a15i32_inreg_i32_inreg__noimplicit@rel32@hi+12 -; GFX9-NEXT: s_mov_b32 s3, s7 -; GFX9-NEXT: s_mov_b32 s2, s6 -; GFX9-NEXT: s_mov_b32 s1, s5 -; GFX9-NEXT: s_mov_b32 s0, s4 -; GFX9-NEXT: s_mov_b32 s4, s8 -; GFX9-NEXT: s_mov_b32 s5, s9 -; GFX9-NEXT: s_mov_b32 s6, s10 -; GFX9-NEXT: s_mov_b32 s7, s11 -; GFX9-NEXT: s_mov_b32 s8, s15 -; GFX9-NEXT: s_mov_b32 s9, s16 -; GFX9-NEXT: s_mov_b32 s10, s17 -; GFX9-NEXT: s_mov_b32 s11, s18 -; GFX9-NEXT: s_mov_b32 s15, s19 -; GFX9-NEXT: s_mov_b32 s16, s20 -; GFX9-NEXT: v_writelane_b32 v40, s31, 1 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[22:23] -; GFX9-NEXT: v_readlane_b32 s31, v40, 1 -; GFX9-NEXT: v_readlane_b32 s30, v40, 0 -; GFX9-NEXT: s_mov_b32 s32, s33 -; GFX9-NEXT: v_readlane_b32 s4, v40, 2 -; GFX9-NEXT: s_or_saveexec_b64 s[6:7], -1 -; GFX9-NEXT: buffer_load_dword v40, off, s[0:3], s33 ; 4-byte Folded Reload -; GFX9-NEXT: s_mov_b64 exec, s[6:7] -; GFX9-NEXT: s_mov_b32 s33, s4 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_setpc_b64 s[30:31] -; ; GFX11-LABEL: test_call_external_void_func_a15i32_inreg_i32_inreg: ; GFX11: ; %bb.0: ; GFX11-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) @@ -1529,3 +936,6 @@ define void @test_call_external_void_func_a15i32_inreg_i32_inreg([13 x i32] inre attributes #0 = { nounwind } attributes #1 = { nounwind "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-queue-ptr" "amdgpu-no-work-group-id-x" "amdgpu-no-work-group-id-y" "amdgpu-no-work-group-id-z" "amdgpu-no-work-item-id-x" "amdgpu-no-work-item-id-y" "amdgpu-no-work-item-id-z" } +;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line: +; GISEL: {{.*}} +; SDAG: {{.*}} diff --git a/llvm/test/CodeGen/AMDGPU/call-argument-types.ll b/llvm/test/CodeGen/AMDGPU/call-argument-types.ll index 832e43f..c407f76 100644 --- a/llvm/test/CodeGen/AMDGPU/call-argument-types.ll +++ b/llvm/test/CodeGen/AMDGPU/call-argument-types.ll @@ -1,10 +1,11 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2 ; RUN: llc -mtriple=amdgcn -mcpu=fiji -mattr=-flat-for-global -amdgpu-scalarize-global-loads=0 < %s | FileCheck -enable-var-scope -check-prefixes=VI %s ; RUN: llc -mtriple=amdgcn -mcpu=hawaii -amdgpu-scalarize-global-loads=0 < %s | FileCheck -enable-var-scope -check-prefixes=CI %s -; RUN: llc -mtriple=amdgcn -mcpu=gfx900 -mattr=-flat-for-global -amdgpu-scalarize-global-loads=0 < %s | FileCheck -enable-var-scope -check-prefixes=GFX9 %s +; RUN: llc -mtriple=amdgcn -global-isel=0 -mcpu=gfx900 -mattr=-flat-for-global -amdgpu-scalarize-global-loads=0 < %s | FileCheck -enable-var-scope -check-prefixes=GFX9,SDAG %s ; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=+real-true16 -mattr=-flat-for-global -amdgpu-scalarize-global-loads=0 < %s | FileCheck -enable-var-scope -check-prefixes=GFX11,GFX11-TRUE16 %s ; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=-real-true16 -mattr=-flat-for-global -amdgpu-scalarize-global-loads=0 < %s | FileCheck -enable-var-scope -check-prefixes=GFX11,GFX11-FAKE16 %s ; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -mattr=-flat-for-global -amdgpu-scalarize-global-loads=0 < %s | FileCheck -enable-var-scope -check-prefixes=HSA %s +; RUN: llc -mtriple=amdgcn -global-isel=1 -mcpu=gfx900 -mattr=-flat-for-global -amdgpu-scalarize-global-loads=0 < %s | FileCheck -enable-var-scope -check-prefixes=GFX9,GISEL %s declare hidden void @external_void_func_i1(i1) #0 declare hidden void @external_void_func_i1_signext(i1 signext) #0 @@ -100,24 +101,24 @@ define amdgpu_kernel void @test_call_external_void_func_i1_imm() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_i1_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_i1@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_i1@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 1 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_i1_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_i1@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_i1@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 1 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_i1_imm: ; GFX11: ; %bb.0: @@ -145,6 +146,25 @@ define amdgpu_kernel void @test_call_external_void_func_i1_imm() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_i1_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_i1@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_i1@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 1 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_i1(i1 true) ret void } @@ -196,28 +216,28 @@ define amdgpu_kernel void @test_call_external_void_func_i1_signext(i32) #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_i1_signext: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: buffer_load_ubyte v0, off, s[0:3], 0 glc -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s5 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_i1_signext@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_i1_signext@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: v_bfe_i32 v0, v0, 0, 1 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_i1_signext: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: buffer_load_ubyte v0, off, s[0:3], 0 glc +; SDAG-NEXT: s_waitcnt vmcnt(0) +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s5 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_i1_signext@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_i1_signext@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: v_bfe_i32 v0, v0, 0, 1 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_i1_signext: ; GFX11: ; %bb.0: @@ -253,6 +273,29 @@ define amdgpu_kernel void @test_call_external_void_func_i1_signext(i32) #0 { ; HSA-NEXT: v_bfe_i32 v0, v0, 0, 1 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_i1_signext: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s2, -1 +; GISEL-NEXT: s_mov_b32 s3, 0xf000 +; GISEL-NEXT: buffer_load_ubyte v0, off, s[0:3], 0 glc +; GISEL-NEXT: s_waitcnt vmcnt(0) +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s5 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_i1_signext@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_i1_signext@rel32@hi+12 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: v_bfe_i32 v0, v0, 0, 1 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %var = load volatile i1, ptr addrspace(1) poison call void @external_void_func_i1_signext(i1 signext %var) ret void @@ -306,28 +349,28 @@ define amdgpu_kernel void @test_call_external_void_func_i1_zeroext(i32) #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_i1_zeroext: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: buffer_load_ubyte v0, off, s[0:3], 0 glc -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s5 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_i1_zeroext@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_i1_zeroext@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: v_and_b32_e32 v0, 1, v0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_i1_zeroext: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: buffer_load_ubyte v0, off, s[0:3], 0 glc +; SDAG-NEXT: s_waitcnt vmcnt(0) +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s5 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_i1_zeroext@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_i1_zeroext@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: v_and_b32_e32 v0, 1, v0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_i1_zeroext: ; GFX11: ; %bb.0: @@ -363,6 +406,29 @@ define amdgpu_kernel void @test_call_external_void_func_i1_zeroext(i32) #0 { ; HSA-NEXT: v_and_b32_e32 v0, 1, v0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_i1_zeroext: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s2, -1 +; GISEL-NEXT: s_mov_b32 s3, 0xf000 +; GISEL-NEXT: buffer_load_ubyte v0, off, s[0:3], 0 glc +; GISEL-NEXT: s_waitcnt vmcnt(0) +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s5 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_i1_zeroext@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_i1_zeroext@rel32@hi+12 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: v_and_b32_e32 v0, 1, v0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %var = load volatile i1, ptr addrspace(1) poison call void @external_void_func_i1_zeroext(i1 zeroext %var) ret void @@ -407,24 +473,24 @@ define amdgpu_kernel void @test_call_external_void_func_i8_imm(i32) #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_i8_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s5 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_i8@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_i8@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 0x7b -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_i8_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s5 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_i8@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_i8@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 0x7b +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-TRUE16-LABEL: test_call_external_void_func_i8_imm: ; GFX11-TRUE16: ; %bb.0: @@ -463,6 +529,25 @@ define amdgpu_kernel void @test_call_external_void_func_i8_imm(i32) #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_i8_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s5 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_i8@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_i8@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 0x7b +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_i8(i8 123) ret void } @@ -513,27 +598,27 @@ define amdgpu_kernel void @test_call_external_void_func_i8_signext(i32) #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_i8_signext: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: buffer_load_sbyte v0, off, s[0:3], 0 glc -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s5 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_i8_signext@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_i8_signext@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_i8_signext: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: buffer_load_sbyte v0, off, s[0:3], 0 glc +; SDAG-NEXT: s_waitcnt vmcnt(0) +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s5 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_i8_signext@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_i8_signext@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_i8_signext: ; GFX11: ; %bb.0: @@ -567,6 +652,28 @@ define amdgpu_kernel void @test_call_external_void_func_i8_signext(i32) #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_i8_signext: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s2, -1 +; GISEL-NEXT: s_mov_b32 s3, 0xf000 +; GISEL-NEXT: buffer_load_sbyte v0, off, s[0:3], 0 glc +; GISEL-NEXT: s_waitcnt vmcnt(0) +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s5 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_i8_signext@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_i8_signext@rel32@hi+12 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %var = load volatile i8, ptr addrspace(1) poison call void @external_void_func_i8_signext(i8 signext %var) ret void @@ -617,27 +724,27 @@ define amdgpu_kernel void @test_call_external_void_func_i8_zeroext(i32) #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_i8_zeroext: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: buffer_load_ubyte v0, off, s[0:3], 0 glc -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s5 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_i8_zeroext@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_i8_zeroext@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_i8_zeroext: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: buffer_load_ubyte v0, off, s[0:3], 0 glc +; SDAG-NEXT: s_waitcnt vmcnt(0) +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s5 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_i8_zeroext@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_i8_zeroext@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_i8_zeroext: ; GFX11: ; %bb.0: @@ -671,6 +778,28 @@ define amdgpu_kernel void @test_call_external_void_func_i8_zeroext(i32) #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_i8_zeroext: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s2, -1 +; GISEL-NEXT: s_mov_b32 s3, 0xf000 +; GISEL-NEXT: buffer_load_ubyte v0, off, s[0:3], 0 glc +; GISEL-NEXT: s_waitcnt vmcnt(0) +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s5 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_i8_zeroext@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_i8_zeroext@rel32@hi+12 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %var = load volatile i8, ptr addrspace(1) poison call void @external_void_func_i8_zeroext(i8 zeroext %var) ret void @@ -715,24 +844,24 @@ define amdgpu_kernel void @test_call_external_void_func_i16_imm() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_i16_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_i16@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_i16@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 0x7b -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_i16_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_i16@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_i16@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 0x7b +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-TRUE16-LABEL: test_call_external_void_func_i16_imm: ; GFX11-TRUE16: ; %bb.0: @@ -771,6 +900,25 @@ define amdgpu_kernel void @test_call_external_void_func_i16_imm() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_i16_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_i16@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_i16@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 0x7b +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_i16(i16 123) ret void } @@ -820,27 +968,27 @@ define amdgpu_kernel void @test_call_external_void_func_i16_signext(i32) #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_i16_signext: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: buffer_load_sshort v0, off, s[0:3], 0 glc -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s5 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_i16_signext@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_i16_signext@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_i16_signext: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: buffer_load_sshort v0, off, s[0:3], 0 glc +; SDAG-NEXT: s_waitcnt vmcnt(0) +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s5 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_i16_signext@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_i16_signext@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_i16_signext: ; GFX11: ; %bb.0: @@ -874,6 +1022,28 @@ define amdgpu_kernel void @test_call_external_void_func_i16_signext(i32) #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_i16_signext: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s2, -1 +; GISEL-NEXT: s_mov_b32 s3, 0xf000 +; GISEL-NEXT: buffer_load_sshort v0, off, s[0:3], 0 glc +; GISEL-NEXT: s_waitcnt vmcnt(0) +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s5 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_i16_signext@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_i16_signext@rel32@hi+12 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %var = load volatile i16, ptr addrspace(1) poison call void @external_void_func_i16_signext(i16 signext %var) ret void @@ -924,27 +1094,27 @@ define amdgpu_kernel void @test_call_external_void_func_i16_zeroext(i32) #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_i16_zeroext: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: buffer_load_ushort v0, off, s[0:3], 0 glc -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s5 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_i16_zeroext@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_i16_zeroext@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_i16_zeroext: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: buffer_load_ushort v0, off, s[0:3], 0 glc +; SDAG-NEXT: s_waitcnt vmcnt(0) +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s5 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_i16_zeroext@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_i16_zeroext@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_i16_zeroext: ; GFX11: ; %bb.0: @@ -978,6 +1148,28 @@ define amdgpu_kernel void @test_call_external_void_func_i16_zeroext(i32) #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_i16_zeroext: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s2, -1 +; GISEL-NEXT: s_mov_b32 s3, 0xf000 +; GISEL-NEXT: buffer_load_ushort v0, off, s[0:3], 0 glc +; GISEL-NEXT: s_waitcnt vmcnt(0) +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s5 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_i16_zeroext@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_i16_zeroext@rel32@hi+12 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %var = load volatile i16, ptr addrspace(1) poison call void @external_void_func_i16_zeroext(i16 zeroext %var) ret void @@ -1022,24 +1214,24 @@ define amdgpu_kernel void @test_call_external_void_func_i32_imm(i32) #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_i32_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s5 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_i32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_i32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 42 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_i32_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s5 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_i32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_i32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 42 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_i32_imm: ; GFX11: ; %bb.0: @@ -1067,6 +1259,25 @@ define amdgpu_kernel void @test_call_external_void_func_i32_imm(i32) #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_i32_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s5 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_i32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_i32@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 42 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_i32(i32 42) ret void } @@ -1112,25 +1323,25 @@ define amdgpu_kernel void @test_call_external_void_func_i64_imm() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_i64_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_i64@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_i64@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 0x7b -; GFX9-NEXT: v_mov_b32_e32 v1, 0 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_i64_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_i64@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_i64@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 0x7b +; SDAG-NEXT: v_mov_b32_e32 v1, 0 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_i64_imm: ; GFX11: ; %bb.0: @@ -1159,6 +1370,26 @@ define amdgpu_kernel void @test_call_external_void_func_i64_imm() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_i64_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_i64@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_i64@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 0x7b +; GISEL-NEXT: v_mov_b32_e32 v1, 0 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_i64(i64 123) ret void } @@ -1208,27 +1439,27 @@ define amdgpu_kernel void @test_call_external_void_func_v2i64() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v2i64: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], 0 -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: buffer_load_dwordx4 v[0:3], off, s[0:3], 0 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v2i64@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v2i64@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v2i64: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], 0 +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: buffer_load_dwordx4 v[0:3], off, s[0:3], 0 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v2i64@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v2i64@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v2i64: ; GFX11: ; %bb.0: @@ -1262,6 +1493,31 @@ define amdgpu_kernel void @test_call_external_void_func_v2i64() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v2i64: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], 0 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x0 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v2i64@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v2i64@rel32@hi+12 +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: v_mov_b32_e32 v0, s0 +; GISEL-NEXT: v_mov_b32_e32 v1, s1 +; GISEL-NEXT: v_mov_b32_e32 v2, s2 +; GISEL-NEXT: v_mov_b32_e32 v3, s3 +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %val = load <2 x i64>, ptr addrspace(1) null call void @external_void_func_v2i64(<2 x i64> %val) ret void @@ -1312,27 +1568,27 @@ define amdgpu_kernel void @test_call_external_void_func_v2i64_imm() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v2i64_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v2i64@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v2i64@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 1 -; GFX9-NEXT: v_mov_b32_e32 v1, 2 -; GFX9-NEXT: v_mov_b32_e32 v2, 3 -; GFX9-NEXT: v_mov_b32_e32 v3, 4 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v2i64_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v2i64@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v2i64@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 1 +; SDAG-NEXT: v_mov_b32_e32 v1, 2 +; SDAG-NEXT: v_mov_b32_e32 v2, 3 +; SDAG-NEXT: v_mov_b32_e32 v3, 4 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v2i64_imm: ; GFX11: ; %bb.0: @@ -1364,6 +1620,28 @@ define amdgpu_kernel void @test_call_external_void_func_v2i64_imm() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v2i64_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v2i64@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v2i64@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 1 +; GISEL-NEXT: v_mov_b32_e32 v1, 2 +; GISEL-NEXT: v_mov_b32_e32 v2, 3 +; GISEL-NEXT: v_mov_b32_e32 v3, 4 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_v2i64(<2 x i64> <i64 8589934593, i64 17179869187>) ret void } @@ -1417,29 +1695,29 @@ define amdgpu_kernel void @test_call_external_void_func_v3i64() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v3i64: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], 0 -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: buffer_load_dwordx4 v[0:3], off, s[0:3], 0 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v3i64@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v3i64@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v4, 1 -; GFX9-NEXT: v_mov_b32_e32 v5, 2 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v3i64: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], 0 +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: buffer_load_dwordx4 v[0:3], off, s[0:3], 0 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v3i64@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v3i64@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v4, 1 +; SDAG-NEXT: v_mov_b32_e32 v5, 2 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v3i64: ; GFX11: ; %bb.0: @@ -1476,6 +1754,33 @@ define amdgpu_kernel void @test_call_external_void_func_v3i64() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v3i64: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], 0 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x0 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v3i64@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v3i64@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v4, 1 +; GISEL-NEXT: v_mov_b32_e32 v5, 2 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: v_mov_b32_e32 v0, s0 +; GISEL-NEXT: v_mov_b32_e32 v1, s1 +; GISEL-NEXT: v_mov_b32_e32 v2, s2 +; GISEL-NEXT: v_mov_b32_e32 v3, s3 +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %load = load <2 x i64>, ptr addrspace(1) null %val = shufflevector <2 x i64> %load, <2 x i64> <i64 8589934593, i64 poison>, <3 x i32> <i32 0, i32 1, i32 2> @@ -1536,31 +1841,31 @@ define amdgpu_kernel void @test_call_external_void_func_v4i64() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v4i64: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], 0 -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: buffer_load_dwordx4 v[0:3], off, s[0:3], 0 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v4i64@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v4i64@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v4, 1 -; GFX9-NEXT: v_mov_b32_e32 v5, 2 -; GFX9-NEXT: v_mov_b32_e32 v6, 3 -; GFX9-NEXT: v_mov_b32_e32 v7, 4 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v4i64: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], 0 +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: buffer_load_dwordx4 v[0:3], off, s[0:3], 0 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v4i64@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v4i64@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v4, 1 +; SDAG-NEXT: v_mov_b32_e32 v5, 2 +; SDAG-NEXT: v_mov_b32_e32 v6, 3 +; SDAG-NEXT: v_mov_b32_e32 v7, 4 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v4i64: ; GFX11: ; %bb.0: @@ -1600,6 +1905,35 @@ define amdgpu_kernel void @test_call_external_void_func_v4i64() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v4i64: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], 0 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x0 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v4i64@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v4i64@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v4, 1 +; GISEL-NEXT: v_mov_b32_e32 v5, 2 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: v_mov_b32_e32 v0, s0 +; GISEL-NEXT: v_mov_b32_e32 v1, s1 +; GISEL-NEXT: v_mov_b32_e32 v2, s2 +; GISEL-NEXT: v_mov_b32_e32 v3, s3 +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: v_mov_b32_e32 v6, 3 +; GISEL-NEXT: v_mov_b32_e32 v7, 4 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %load = load <2 x i64>, ptr addrspace(1) null %val = shufflevector <2 x i64> %load, <2 x i64> <i64 8589934593, i64 17179869187>, <4 x i32> <i32 0, i32 1, i32 2, i32 3> call void @external_void_func_v4i64(<4 x i64> %val) @@ -1645,24 +1979,24 @@ define amdgpu_kernel void @test_call_external_void_func_f16_imm() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_f16_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_f16@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_f16@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 0x4400 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_f16_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_f16@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_f16@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 0x4400 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-TRUE16-LABEL: test_call_external_void_func_f16_imm: ; GFX11-TRUE16: ; %bb.0: @@ -1701,6 +2035,25 @@ define amdgpu_kernel void @test_call_external_void_func_f16_imm() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_f16_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_f16@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_f16@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 0x4400 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_f16(half 4.0) ret void } @@ -1744,24 +2097,24 @@ define amdgpu_kernel void @test_call_external_void_func_f32_imm() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_f32_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_f32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_f32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 4.0 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_f32_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_f32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_f32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 4.0 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_f32_imm: ; GFX11: ; %bb.0: @@ -1789,6 +2142,25 @@ define amdgpu_kernel void @test_call_external_void_func_f32_imm() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_f32_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_f32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_f32@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 4.0 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_f32(float 4.0) ret void } @@ -1834,25 +2206,25 @@ define amdgpu_kernel void @test_call_external_void_func_v2f32_imm() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v2f32_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v2f32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v2f32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 1.0 -; GFX9-NEXT: v_mov_b32_e32 v1, 2.0 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v2f32_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v2f32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v2f32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 1.0 +; SDAG-NEXT: v_mov_b32_e32 v1, 2.0 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v2f32_imm: ; GFX11: ; %bb.0: @@ -1881,6 +2253,26 @@ define amdgpu_kernel void @test_call_external_void_func_v2f32_imm() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v2f32_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v2f32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v2f32@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 1.0 +; GISEL-NEXT: v_mov_b32_e32 v1, 2.0 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_v2f32(<2 x float> <float 1.0, float 2.0>) ret void } @@ -1928,26 +2320,26 @@ define amdgpu_kernel void @test_call_external_void_func_v3f32_imm() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v3f32_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v3f32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v3f32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 1.0 -; GFX9-NEXT: v_mov_b32_e32 v1, 2.0 -; GFX9-NEXT: v_mov_b32_e32 v2, 4.0 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v3f32_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v3f32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v3f32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 1.0 +; SDAG-NEXT: v_mov_b32_e32 v1, 2.0 +; SDAG-NEXT: v_mov_b32_e32 v2, 4.0 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v3f32_imm: ; GFX11: ; %bb.0: @@ -1978,6 +2370,27 @@ define amdgpu_kernel void @test_call_external_void_func_v3f32_imm() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v3f32_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v3f32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v3f32@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 1.0 +; GISEL-NEXT: v_mov_b32_e32 v1, 2.0 +; GISEL-NEXT: v_mov_b32_e32 v2, 4.0 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_v3f32(<3 x float> <float 1.0, float 2.0, float 4.0>) ret void } @@ -2029,28 +2442,28 @@ define amdgpu_kernel void @test_call_external_void_func_v5f32_imm() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v5f32_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v5f32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v5f32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 1.0 -; GFX9-NEXT: v_mov_b32_e32 v1, 2.0 -; GFX9-NEXT: v_mov_b32_e32 v2, 4.0 -; GFX9-NEXT: v_mov_b32_e32 v3, -1.0 -; GFX9-NEXT: v_mov_b32_e32 v4, 0.5 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v5f32_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v5f32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v5f32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 1.0 +; SDAG-NEXT: v_mov_b32_e32 v1, 2.0 +; SDAG-NEXT: v_mov_b32_e32 v2, 4.0 +; SDAG-NEXT: v_mov_b32_e32 v3, -1.0 +; SDAG-NEXT: v_mov_b32_e32 v4, 0.5 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v5f32_imm: ; GFX11: ; %bb.0: @@ -2084,6 +2497,29 @@ define amdgpu_kernel void @test_call_external_void_func_v5f32_imm() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v5f32_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v5f32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v5f32@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 1.0 +; GISEL-NEXT: v_mov_b32_e32 v1, 2.0 +; GISEL-NEXT: v_mov_b32_e32 v2, 4.0 +; GISEL-NEXT: v_mov_b32_e32 v3, -1.0 +; GISEL-NEXT: v_mov_b32_e32 v4, 0.5 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_v5f32(<5 x float> <float 1.0, float 2.0, float 4.0, float -1.0, float 0.5>) ret void } @@ -2129,25 +2565,25 @@ define amdgpu_kernel void @test_call_external_void_func_f64_imm() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_f64_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_f64@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_f64@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 0 -; GFX9-NEXT: v_mov_b32_e32 v1, 0x40100000 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_f64_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_f64@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_f64@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: v_mov_b32_e32 v1, 0x40100000 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_f64_imm: ; GFX11: ; %bb.0: @@ -2176,6 +2612,26 @@ define amdgpu_kernel void @test_call_external_void_func_f64_imm() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_f64_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_f64@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_f64@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: v_mov_b32_e32 v1, 0x40100000 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_f64(double 4.0) ret void } @@ -2225,27 +2681,27 @@ define amdgpu_kernel void @test_call_external_void_func_v2f64_imm() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v2f64_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v2f64@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v2f64@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 0 -; GFX9-NEXT: v_mov_b32_e32 v1, 2.0 -; GFX9-NEXT: v_mov_b32_e32 v2, 0 -; GFX9-NEXT: v_mov_b32_e32 v3, 0x40100000 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v2f64_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v2f64@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v2f64@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: v_mov_b32_e32 v1, 2.0 +; SDAG-NEXT: v_mov_b32_e32 v2, 0 +; SDAG-NEXT: v_mov_b32_e32 v3, 0x40100000 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v2f64_imm: ; GFX11: ; %bb.0: @@ -2277,6 +2733,28 @@ define amdgpu_kernel void @test_call_external_void_func_v2f64_imm() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v2f64_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v2f64@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v2f64@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: v_mov_b32_e32 v1, 2.0 +; GISEL-NEXT: v_mov_b32_e32 v2, 0 +; GISEL-NEXT: v_mov_b32_e32 v3, 0x40100000 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_v2f64(<2 x double> <double 2.0, double 4.0>) ret void } @@ -2330,29 +2808,29 @@ define amdgpu_kernel void @test_call_external_void_func_v3f64_imm() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v3f64_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v3f64@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v3f64@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 0 -; GFX9-NEXT: v_mov_b32_e32 v1, 2.0 -; GFX9-NEXT: v_mov_b32_e32 v2, 0 -; GFX9-NEXT: v_mov_b32_e32 v3, 0x40100000 -; GFX9-NEXT: v_mov_b32_e32 v4, 0 -; GFX9-NEXT: v_mov_b32_e32 v5, 0x40200000 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v3f64_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v3f64@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v3f64@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: v_mov_b32_e32 v1, 2.0 +; SDAG-NEXT: v_mov_b32_e32 v2, 0 +; SDAG-NEXT: v_mov_b32_e32 v3, 0x40100000 +; SDAG-NEXT: v_mov_b32_e32 v4, 0 +; SDAG-NEXT: v_mov_b32_e32 v5, 0x40200000 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v3f64_imm: ; GFX11: ; %bb.0: @@ -2387,6 +2865,30 @@ define amdgpu_kernel void @test_call_external_void_func_v3f64_imm() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v3f64_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v3f64@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v3f64@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: v_mov_b32_e32 v1, 2.0 +; GISEL-NEXT: v_mov_b32_e32 v2, 0 +; GISEL-NEXT: v_mov_b32_e32 v3, 0x40100000 +; GISEL-NEXT: v_mov_b32_e32 v4, 0 +; GISEL-NEXT: v_mov_b32_e32 v5, 0x40200000 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_v3f64(<3 x double> <double 2.0, double 4.0, double 8.0>) ret void } @@ -2436,26 +2938,26 @@ define amdgpu_kernel void @test_call_external_void_func_v2i16() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v2i16: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], 0 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v2i16@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v2i16@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v2i16: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: buffer_load_dword v0, off, s[0:3], 0 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v2i16@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v2i16@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v2i16: ; GFX11: ; %bb.0: @@ -2487,6 +2989,27 @@ define amdgpu_kernel void @test_call_external_void_func_v2i16() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v2i16: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_load_dword s8, s[0:1], 0x0 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v2i16@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v2i16@rel32@hi+12 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: v_mov_b32_e32 v0, s8 +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %val = load <2 x i16>, ptr addrspace(1) poison call void @external_void_func_v2i16(<2 x i16> %val) ret void @@ -2539,26 +3062,26 @@ define amdgpu_kernel void @test_call_external_void_func_v3i16() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v3i16: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: buffer_load_dwordx2 v[0:1], off, s[0:3], 0 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v3i16@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v3i16@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v3i16: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: buffer_load_dwordx2 v[0:1], off, s[0:3], 0 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v3i16@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v3i16@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v3i16: ; GFX11: ; %bb.0: @@ -2590,6 +3113,28 @@ define amdgpu_kernel void @test_call_external_void_func_v3i16() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v3i16: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x0 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: v_mov_b32_e32 v0, s0 +; GISEL-NEXT: v_mov_b32_e32 v1, s1 +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v3i16@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v3i16@rel32@hi+12 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %val = load <3 x i16>, ptr addrspace(1) poison call void @external_void_func_v3i16(<3 x i16> %val) ret void @@ -2643,26 +3188,26 @@ define amdgpu_kernel void @test_call_external_void_func_v3f16() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v3f16: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: buffer_load_dwordx2 v[0:1], off, s[0:3], 0 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v3f16@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v3f16@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v3f16: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: buffer_load_dwordx2 v[0:1], off, s[0:3], 0 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v3f16@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v3f16@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v3f16: ; GFX11: ; %bb.0: @@ -2694,6 +3239,28 @@ define amdgpu_kernel void @test_call_external_void_func_v3f16() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v3f16: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x0 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: v_mov_b32_e32 v0, s0 +; GISEL-NEXT: v_mov_b32_e32 v1, s1 +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v3f16@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v3f16@rel32@hi+12 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %val = load <3 x half>, ptr addrspace(1) poison call void @external_void_func_v3f16(<3 x half> %val) ret void @@ -2741,25 +3308,25 @@ define amdgpu_kernel void @test_call_external_void_func_v3i16_imm() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v3i16_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v3i16@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v3i16@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 0x20001 -; GFX9-NEXT: v_mov_b32_e32 v1, 3 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v3i16_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v3i16@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v3i16@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 0x20001 +; SDAG-NEXT: v_mov_b32_e32 v1, 3 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v3i16_imm: ; GFX11: ; %bb.0: @@ -2788,6 +3355,26 @@ define amdgpu_kernel void @test_call_external_void_func_v3i16_imm() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v3i16_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v3i16@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v3i16@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 0x20001 +; GISEL-NEXT: v_mov_b32_e32 v1, 3 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_v3i16(<3 x i16> <i16 1, i16 2, i16 3>) ret void } @@ -2834,25 +3421,25 @@ define amdgpu_kernel void @test_call_external_void_func_v3f16_imm() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v3f16_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v3f16@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v3f16@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 0x40003c00 -; GFX9-NEXT: v_mov_b32_e32 v1, 0x4400 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v3f16_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v3f16@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v3f16@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 0x40003c00 +; SDAG-NEXT: v_mov_b32_e32 v1, 0x4400 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v3f16_imm: ; GFX11: ; %bb.0: @@ -2882,6 +3469,26 @@ define amdgpu_kernel void @test_call_external_void_func_v3f16_imm() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v3f16_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v3f16@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v3f16@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 0x40003c00 +; GISEL-NEXT: v_mov_b32_e32 v1, 0x4400 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_v3f16(<3 x half> <half 1.0, half 2.0, half 4.0>) ret void } @@ -2934,26 +3541,26 @@ define amdgpu_kernel void @test_call_external_void_func_v4i16() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v4i16: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: buffer_load_dwordx2 v[0:1], off, s[0:3], 0 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v4i16@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v4i16@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v4i16: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: buffer_load_dwordx2 v[0:1], off, s[0:3], 0 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v4i16@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v4i16@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v4i16: ; GFX11: ; %bb.0: @@ -2985,6 +3592,28 @@ define amdgpu_kernel void @test_call_external_void_func_v4i16() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v4i16: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x0 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: v_mov_b32_e32 v0, s0 +; GISEL-NEXT: v_mov_b32_e32 v1, s1 +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v4i16@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v4i16@rel32@hi+12 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %val = load <4 x i16>, ptr addrspace(1) poison call void @external_void_func_v4i16(<4 x i16> %val) ret void @@ -3033,25 +3662,25 @@ define amdgpu_kernel void @test_call_external_void_func_v4i16_imm() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v4i16_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v4i16@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v4i16@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 0x20001 -; GFX9-NEXT: v_mov_b32_e32 v1, 0x40003 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v4i16_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v4i16@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v4i16@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 0x20001 +; SDAG-NEXT: v_mov_b32_e32 v1, 0x40003 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v4i16_imm: ; GFX11: ; %bb.0: @@ -3081,6 +3710,26 @@ define amdgpu_kernel void @test_call_external_void_func_v4i16_imm() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v4i16_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v4i16@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v4i16@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 0x20001 +; GISEL-NEXT: v_mov_b32_e32 v1, 0x40003 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_v4i16(<4 x i16> <i16 1, i16 2, i16 3, i16 4>) ret void } @@ -3132,26 +3781,26 @@ define amdgpu_kernel void @test_call_external_void_func_v2f16() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v2f16: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: buffer_load_dword v0, off, s[0:3], 0 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v2f16@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v2f16@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v2f16: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: buffer_load_dword v0, off, s[0:3], 0 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v2f16@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v2f16@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v2f16: ; GFX11: ; %bb.0: @@ -3183,6 +3832,27 @@ define amdgpu_kernel void @test_call_external_void_func_v2f16() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v2f16: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_load_dword s8, s[0:1], 0x0 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v2f16@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v2f16@rel32@hi+12 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: v_mov_b32_e32 v0, s8 +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %val = load <2 x half>, ptr addrspace(1) poison call void @external_void_func_v2f16(<2 x half> %val) ret void @@ -3231,26 +3901,26 @@ define amdgpu_kernel void @test_call_external_void_func_v2i32() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v2i32: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: buffer_load_dwordx2 v[0:1], off, s[0:3], 0 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v2i32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v2i32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v2i32: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: buffer_load_dwordx2 v[0:1], off, s[0:3], 0 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v2i32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v2i32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v2i32: ; GFX11: ; %bb.0: @@ -3282,6 +3952,28 @@ define amdgpu_kernel void @test_call_external_void_func_v2i32() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v2i32: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x0 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: v_mov_b32_e32 v0, s0 +; GISEL-NEXT: v_mov_b32_e32 v1, s1 +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v2i32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v2i32@rel32@hi+12 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %val = load <2 x i32>, ptr addrspace(1) poison call void @external_void_func_v2i32(<2 x i32> %val) ret void @@ -3328,25 +4020,25 @@ define amdgpu_kernel void @test_call_external_void_func_v2i32_imm() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v2i32_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v2i32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v2i32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 1 -; GFX9-NEXT: v_mov_b32_e32 v1, 2 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v2i32_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v2i32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v2i32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 1 +; SDAG-NEXT: v_mov_b32_e32 v1, 2 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v2i32_imm: ; GFX11: ; %bb.0: @@ -3375,6 +4067,26 @@ define amdgpu_kernel void @test_call_external_void_func_v2i32_imm() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v2i32_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v2i32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v2i32@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 1 +; GISEL-NEXT: v_mov_b32_e32 v1, 2 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_v2i32(<2 x i32> <i32 1, i32 2>) ret void } @@ -3422,26 +4134,26 @@ define amdgpu_kernel void @test_call_external_void_func_v3i32_imm(i32) #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v3i32_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s5 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v3i32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v3i32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 3 -; GFX9-NEXT: v_mov_b32_e32 v1, 4 -; GFX9-NEXT: v_mov_b32_e32 v2, 5 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v3i32_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s5 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v3i32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v3i32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 3 +; SDAG-NEXT: v_mov_b32_e32 v1, 4 +; SDAG-NEXT: v_mov_b32_e32 v2, 5 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v3i32_imm: ; GFX11: ; %bb.0: @@ -3472,6 +4184,27 @@ define amdgpu_kernel void @test_call_external_void_func_v3i32_imm(i32) #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v3i32_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s5 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v3i32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v3i32@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 3 +; GISEL-NEXT: v_mov_b32_e32 v1, 4 +; GISEL-NEXT: v_mov_b32_e32 v2, 5 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_v3i32(<3 x i32> <i32 3, i32 4, i32 5>) ret void } @@ -3521,27 +4254,27 @@ define amdgpu_kernel void @test_call_external_void_func_v3i32_i32(i32) #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v3i32_i32: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s5 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v3i32_i32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v3i32_i32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 3 -; GFX9-NEXT: v_mov_b32_e32 v1, 4 -; GFX9-NEXT: v_mov_b32_e32 v2, 5 -; GFX9-NEXT: v_mov_b32_e32 v3, 6 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v3i32_i32: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s5 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v3i32_i32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v3i32_i32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 3 +; SDAG-NEXT: v_mov_b32_e32 v1, 4 +; SDAG-NEXT: v_mov_b32_e32 v2, 5 +; SDAG-NEXT: v_mov_b32_e32 v3, 6 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v3i32_i32: ; GFX11: ; %bb.0: @@ -3573,6 +4306,28 @@ define amdgpu_kernel void @test_call_external_void_func_v3i32_i32(i32) #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v3i32_i32: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s5 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v3i32_i32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v3i32_i32@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 3 +; GISEL-NEXT: v_mov_b32_e32 v1, 4 +; GISEL-NEXT: v_mov_b32_e32 v2, 5 +; GISEL-NEXT: v_mov_b32_e32 v3, 6 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_v3i32_i32(<3 x i32> <i32 3, i32 4, i32 5>, i32 6) ret void } @@ -3620,26 +4375,26 @@ define amdgpu_kernel void @test_call_external_void_func_v4i32() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v4i32: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: buffer_load_dwordx4 v[0:3], off, s[0:3], 0 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v4i32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v4i32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v4i32: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: buffer_load_dwordx4 v[0:3], off, s[0:3], 0 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v4i32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v4i32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v4i32: ; GFX11: ; %bb.0: @@ -3671,6 +4426,30 @@ define amdgpu_kernel void @test_call_external_void_func_v4i32() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v4i32: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x0 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v4i32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v4i32@rel32@hi+12 +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: v_mov_b32_e32 v0, s0 +; GISEL-NEXT: v_mov_b32_e32 v1, s1 +; GISEL-NEXT: v_mov_b32_e32 v2, s2 +; GISEL-NEXT: v_mov_b32_e32 v3, s3 +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %val = load <4 x i32>, ptr addrspace(1) poison call void @external_void_func_v4i32(<4 x i32> %val) ret void @@ -3721,27 +4500,27 @@ define amdgpu_kernel void @test_call_external_void_func_v4i32_imm() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v4i32_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v4i32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v4i32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 1 -; GFX9-NEXT: v_mov_b32_e32 v1, 2 -; GFX9-NEXT: v_mov_b32_e32 v2, 3 -; GFX9-NEXT: v_mov_b32_e32 v3, 4 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v4i32_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v4i32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v4i32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 1 +; SDAG-NEXT: v_mov_b32_e32 v1, 2 +; SDAG-NEXT: v_mov_b32_e32 v2, 3 +; SDAG-NEXT: v_mov_b32_e32 v3, 4 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v4i32_imm: ; GFX11: ; %bb.0: @@ -3773,6 +4552,28 @@ define amdgpu_kernel void @test_call_external_void_func_v4i32_imm() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v4i32_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v4i32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v4i32@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 1 +; GISEL-NEXT: v_mov_b32_e32 v1, 2 +; GISEL-NEXT: v_mov_b32_e32 v2, 3 +; GISEL-NEXT: v_mov_b32_e32 v3, 4 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_v4i32(<4 x i32> <i32 1, i32 2, i32 3, i32 4>) ret void } @@ -3824,28 +4625,28 @@ define amdgpu_kernel void @test_call_external_void_func_v5i32_imm() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v5i32_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v5i32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v5i32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 1 -; GFX9-NEXT: v_mov_b32_e32 v1, 2 -; GFX9-NEXT: v_mov_b32_e32 v2, 3 -; GFX9-NEXT: v_mov_b32_e32 v3, 4 -; GFX9-NEXT: v_mov_b32_e32 v4, 5 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v5i32_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v5i32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v5i32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 1 +; SDAG-NEXT: v_mov_b32_e32 v1, 2 +; SDAG-NEXT: v_mov_b32_e32 v2, 3 +; SDAG-NEXT: v_mov_b32_e32 v3, 4 +; SDAG-NEXT: v_mov_b32_e32 v4, 5 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v5i32_imm: ; GFX11: ; %bb.0: @@ -3879,6 +4680,29 @@ define amdgpu_kernel void @test_call_external_void_func_v5i32_imm() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v5i32_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v5i32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v5i32@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 1 +; GISEL-NEXT: v_mov_b32_e32 v1, 2 +; GISEL-NEXT: v_mov_b32_e32 v2, 3 +; GISEL-NEXT: v_mov_b32_e32 v3, 4 +; GISEL-NEXT: v_mov_b32_e32 v4, 5 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_v5i32(<5 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5>) ret void } @@ -3932,29 +4756,29 @@ define amdgpu_kernel void @test_call_external_void_func_v8i32() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v8i32: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: s_waitcnt lgkmcnt(0) -; GFX9-NEXT: buffer_load_dwordx4 v[0:3], off, s[0:3], 0 -; GFX9-NEXT: buffer_load_dwordx4 v[4:7], off, s[0:3], 0 offset:16 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v8i32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v8i32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v8i32: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: s_waitcnt lgkmcnt(0) +; SDAG-NEXT: buffer_load_dwordx4 v[0:3], off, s[0:3], 0 +; SDAG-NEXT: buffer_load_dwordx4 v[4:7], off, s[0:3], 0 offset:16 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v8i32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v8i32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v8i32: ; GFX11: ; %bb.0: @@ -3993,6 +4817,36 @@ define amdgpu_kernel void @test_call_external_void_func_v8i32() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v8i32: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x0 +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: s_load_dwordx8 s[8:15], s[0:1], 0x0 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v8i32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v8i32@rel32@hi+12 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: v_mov_b32_e32 v0, s8 +; GISEL-NEXT: v_mov_b32_e32 v1, s9 +; GISEL-NEXT: v_mov_b32_e32 v2, s10 +; GISEL-NEXT: v_mov_b32_e32 v3, s11 +; GISEL-NEXT: v_mov_b32_e32 v4, s12 +; GISEL-NEXT: v_mov_b32_e32 v5, s13 +; GISEL-NEXT: v_mov_b32_e32 v6, s14 +; GISEL-NEXT: v_mov_b32_e32 v7, s15 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %ptr = load ptr addrspace(1), ptr addrspace(4) poison %val = load <8 x i32>, ptr addrspace(1) %ptr call void @external_void_func_v8i32(<8 x i32> %val) @@ -4052,31 +4906,31 @@ define amdgpu_kernel void @test_call_external_void_func_v8i32_imm() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v8i32_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v8i32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v8i32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: v_mov_b32_e32 v0, 1 -; GFX9-NEXT: v_mov_b32_e32 v1, 2 -; GFX9-NEXT: v_mov_b32_e32 v2, 3 -; GFX9-NEXT: v_mov_b32_e32 v3, 4 -; GFX9-NEXT: v_mov_b32_e32 v4, 5 -; GFX9-NEXT: v_mov_b32_e32 v5, 6 -; GFX9-NEXT: v_mov_b32_e32 v6, 7 -; GFX9-NEXT: v_mov_b32_e32 v7, 8 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v8i32_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v8i32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v8i32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: v_mov_b32_e32 v0, 1 +; SDAG-NEXT: v_mov_b32_e32 v1, 2 +; SDAG-NEXT: v_mov_b32_e32 v2, 3 +; SDAG-NEXT: v_mov_b32_e32 v3, 4 +; SDAG-NEXT: v_mov_b32_e32 v4, 5 +; SDAG-NEXT: v_mov_b32_e32 v5, 6 +; SDAG-NEXT: v_mov_b32_e32 v6, 7 +; SDAG-NEXT: v_mov_b32_e32 v7, 8 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v8i32_imm: ; GFX11: ; %bb.0: @@ -4114,6 +4968,32 @@ define amdgpu_kernel void @test_call_external_void_func_v8i32_imm() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v8i32_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v8i32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v8i32@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 1 +; GISEL-NEXT: v_mov_b32_e32 v1, 2 +; GISEL-NEXT: v_mov_b32_e32 v2, 3 +; GISEL-NEXT: v_mov_b32_e32 v3, 4 +; GISEL-NEXT: v_mov_b32_e32 v4, 5 +; GISEL-NEXT: v_mov_b32_e32 v5, 6 +; GISEL-NEXT: v_mov_b32_e32 v6, 7 +; GISEL-NEXT: v_mov_b32_e32 v7, 8 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm call void @external_void_func_v8i32(<8 x i32> <i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>) ret void } @@ -4171,31 +5051,31 @@ define amdgpu_kernel void @test_call_external_void_func_v16i32() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v16i32: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: s_waitcnt lgkmcnt(0) -; GFX9-NEXT: buffer_load_dwordx4 v[0:3], off, s[0:3], 0 -; GFX9-NEXT: buffer_load_dwordx4 v[4:7], off, s[0:3], 0 offset:16 -; GFX9-NEXT: buffer_load_dwordx4 v[8:11], off, s[0:3], 0 offset:32 -; GFX9-NEXT: buffer_load_dwordx4 v[12:15], off, s[0:3], 0 offset:48 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v16i32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v16i32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v16i32: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: s_waitcnt lgkmcnt(0) +; SDAG-NEXT: buffer_load_dwordx4 v[0:3], off, s[0:3], 0 +; SDAG-NEXT: buffer_load_dwordx4 v[4:7], off, s[0:3], 0 offset:16 +; SDAG-NEXT: buffer_load_dwordx4 v[8:11], off, s[0:3], 0 offset:32 +; SDAG-NEXT: buffer_load_dwordx4 v[12:15], off, s[0:3], 0 offset:48 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v16i32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v16i32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v16i32: ; GFX11: ; %bb.0: @@ -4238,6 +5118,44 @@ define amdgpu_kernel void @test_call_external_void_func_v16i32() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v16i32: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x0 +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: s_load_dwordx16 s[8:23], s[0:1], 0x0 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v16i32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v16i32@rel32@hi+12 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: v_mov_b32_e32 v0, s8 +; GISEL-NEXT: v_mov_b32_e32 v1, s9 +; GISEL-NEXT: v_mov_b32_e32 v2, s10 +; GISEL-NEXT: v_mov_b32_e32 v3, s11 +; GISEL-NEXT: v_mov_b32_e32 v4, s12 +; GISEL-NEXT: v_mov_b32_e32 v5, s13 +; GISEL-NEXT: v_mov_b32_e32 v6, s14 +; GISEL-NEXT: v_mov_b32_e32 v7, s15 +; GISEL-NEXT: v_mov_b32_e32 v8, s16 +; GISEL-NEXT: v_mov_b32_e32 v9, s17 +; GISEL-NEXT: v_mov_b32_e32 v10, s18 +; GISEL-NEXT: v_mov_b32_e32 v11, s19 +; GISEL-NEXT: v_mov_b32_e32 v12, s20 +; GISEL-NEXT: v_mov_b32_e32 v13, s21 +; GISEL-NEXT: v_mov_b32_e32 v14, s22 +; GISEL-NEXT: v_mov_b32_e32 v15, s23 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %ptr = load ptr addrspace(1), ptr addrspace(4) poison %val = load <16 x i32>, ptr addrspace(1) %ptr call void @external_void_func_v16i32(<16 x i32> %val) @@ -4309,37 +5227,37 @@ define amdgpu_kernel void @test_call_external_void_func_v32i32() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[8:9] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v32i32: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x0 -; GFX9-NEXT: s_mov_b32 s7, 0xf000 -; GFX9-NEXT: s_mov_b32 s6, -1 -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_waitcnt lgkmcnt(0) -; GFX9-NEXT: buffer_load_dwordx4 v[24:27], off, s[4:7], 0 offset:96 -; GFX9-NEXT: buffer_load_dwordx4 v[28:31], off, s[4:7], 0 offset:112 -; GFX9-NEXT: buffer_load_dwordx4 v[0:3], off, s[4:7], 0 -; GFX9-NEXT: buffer_load_dwordx4 v[4:7], off, s[4:7], 0 offset:16 -; GFX9-NEXT: buffer_load_dwordx4 v[8:11], off, s[4:7], 0 offset:32 -; GFX9-NEXT: buffer_load_dwordx4 v[12:15], off, s[4:7], 0 offset:48 -; GFX9-NEXT: buffer_load_dwordx4 v[16:19], off, s[4:7], 0 offset:64 -; GFX9-NEXT: buffer_load_dwordx4 v[20:23], off, s[4:7], 0 offset:80 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_getpc_b64 s[8:9] -; GFX9-NEXT: s_add_u32 s8, s8, external_void_func_v32i32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s9, s9, external_void_func_v32i32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_waitcnt vmcnt(6) -; GFX9-NEXT: buffer_store_dword v31, off, s[36:39], s32 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[8:9] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v32i32: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x0 +; SDAG-NEXT: s_mov_b32 s7, 0xf000 +; SDAG-NEXT: s_mov_b32 s6, -1 +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_waitcnt lgkmcnt(0) +; SDAG-NEXT: buffer_load_dwordx4 v[24:27], off, s[4:7], 0 offset:96 +; SDAG-NEXT: buffer_load_dwordx4 v[28:31], off, s[4:7], 0 offset:112 +; SDAG-NEXT: buffer_load_dwordx4 v[0:3], off, s[4:7], 0 +; SDAG-NEXT: buffer_load_dwordx4 v[4:7], off, s[4:7], 0 offset:16 +; SDAG-NEXT: buffer_load_dwordx4 v[8:11], off, s[4:7], 0 offset:32 +; SDAG-NEXT: buffer_load_dwordx4 v[12:15], off, s[4:7], 0 offset:48 +; SDAG-NEXT: buffer_load_dwordx4 v[16:19], off, s[4:7], 0 offset:64 +; SDAG-NEXT: buffer_load_dwordx4 v[20:23], off, s[4:7], 0 offset:80 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_getpc_b64 s[8:9] +; SDAG-NEXT: s_add_u32 s8, s8, external_void_func_v32i32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s9, s9, external_void_func_v32i32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_waitcnt vmcnt(6) +; SDAG-NEXT: buffer_store_dword v31, off, s[36:39], s32 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[8:9] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v32i32: ; GFX11: ; %bb.0: @@ -4394,6 +5312,62 @@ define amdgpu_kernel void @test_call_external_void_func_v32i32() #0 { ; HSA-NEXT: buffer_store_dword v31, off, s[0:3], s32 ; HSA-NEXT: s_swappc_b64 s[30:31], s[12:13] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v32i32: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x0 +; GISEL-NEXT: s_mov_b32 s52, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s53, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s54, -1 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: s_load_dwordx16 s[8:23], s[0:1], 0x40 +; GISEL-NEXT: s_load_dwordx16 s[36:51], s[0:1], 0x0 +; GISEL-NEXT: s_mov_b32 s55, 0xe00000 +; GISEL-NEXT: s_add_u32 s52, s52, s3 +; GISEL-NEXT: s_addc_u32 s53, s53, 0 +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: v_mov_b32_e32 v0, s23 +; GISEL-NEXT: s_mov_b64 s[0:1], s[52:53] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v32i32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v32i32@rel32@hi+12 +; GISEL-NEXT: buffer_store_dword v0, off, s[52:55], s32 +; GISEL-NEXT: v_mov_b32_e32 v0, s36 +; GISEL-NEXT: v_mov_b32_e32 v1, s37 +; GISEL-NEXT: v_mov_b32_e32 v2, s38 +; GISEL-NEXT: v_mov_b32_e32 v3, s39 +; GISEL-NEXT: v_mov_b32_e32 v4, s40 +; GISEL-NEXT: v_mov_b32_e32 v5, s41 +; GISEL-NEXT: v_mov_b32_e32 v6, s42 +; GISEL-NEXT: v_mov_b32_e32 v7, s43 +; GISEL-NEXT: v_mov_b32_e32 v8, s44 +; GISEL-NEXT: v_mov_b32_e32 v9, s45 +; GISEL-NEXT: v_mov_b32_e32 v10, s46 +; GISEL-NEXT: v_mov_b32_e32 v11, s47 +; GISEL-NEXT: v_mov_b32_e32 v12, s48 +; GISEL-NEXT: v_mov_b32_e32 v13, s49 +; GISEL-NEXT: v_mov_b32_e32 v14, s50 +; GISEL-NEXT: v_mov_b32_e32 v15, s51 +; GISEL-NEXT: v_mov_b32_e32 v16, s8 +; GISEL-NEXT: v_mov_b32_e32 v17, s9 +; GISEL-NEXT: v_mov_b32_e32 v18, s10 +; GISEL-NEXT: v_mov_b32_e32 v19, s11 +; GISEL-NEXT: s_mov_b64 s[2:3], s[54:55] +; GISEL-NEXT: v_mov_b32_e32 v20, s12 +; GISEL-NEXT: v_mov_b32_e32 v21, s13 +; GISEL-NEXT: v_mov_b32_e32 v22, s14 +; GISEL-NEXT: v_mov_b32_e32 v23, s15 +; GISEL-NEXT: v_mov_b32_e32 v24, s16 +; GISEL-NEXT: v_mov_b32_e32 v25, s17 +; GISEL-NEXT: v_mov_b32_e32 v26, s18 +; GISEL-NEXT: v_mov_b32_e32 v27, s19 +; GISEL-NEXT: v_mov_b32_e32 v28, s20 +; GISEL-NEXT: v_mov_b32_e32 v29, s21 +; GISEL-NEXT: v_mov_b32_e32 v30, s22 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %ptr = load ptr addrspace(1), ptr addrspace(4) poison %val = load <32 x i32>, ptr addrspace(1) %ptr call void @external_void_func_v32i32(<32 x i32> %val) @@ -4471,40 +5445,40 @@ define amdgpu_kernel void @test_call_external_void_func_v32i32_i32(i32) #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v32i32_i32: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s5 -; GFX9-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x0 -; GFX9-NEXT: s_mov_b32 s7, 0xf000 -; GFX9-NEXT: s_mov_b32 s6, -1 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_waitcnt lgkmcnt(0) -; GFX9-NEXT: buffer_load_dword v32, off, s[4:7], 0 -; GFX9-NEXT: buffer_load_dwordx4 v[28:31], off, s[4:7], 0 offset:112 -; GFX9-NEXT: buffer_load_dwordx4 v[0:3], off, s[4:7], 0 -; GFX9-NEXT: buffer_load_dwordx4 v[4:7], off, s[4:7], 0 offset:16 -; GFX9-NEXT: buffer_load_dwordx4 v[8:11], off, s[4:7], 0 offset:32 -; GFX9-NEXT: buffer_load_dwordx4 v[12:15], off, s[4:7], 0 offset:48 -; GFX9-NEXT: buffer_load_dwordx4 v[16:19], off, s[4:7], 0 offset:64 -; GFX9-NEXT: buffer_load_dwordx4 v[20:23], off, s[4:7], 0 offset:80 -; GFX9-NEXT: buffer_load_dwordx4 v[24:27], off, s[4:7], 0 offset:96 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v32i32_i32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v32i32_i32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_waitcnt vmcnt(8) -; GFX9-NEXT: buffer_store_dword v32, off, s[36:39], s32 offset:4 -; GFX9-NEXT: s_waitcnt vmcnt(8) -; GFX9-NEXT: buffer_store_dword v31, off, s[36:39], s32 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v32i32_i32: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s5 +; SDAG-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x0 +; SDAG-NEXT: s_mov_b32 s7, 0xf000 +; SDAG-NEXT: s_mov_b32 s6, -1 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_waitcnt lgkmcnt(0) +; SDAG-NEXT: buffer_load_dword v32, off, s[4:7], 0 +; SDAG-NEXT: buffer_load_dwordx4 v[28:31], off, s[4:7], 0 offset:112 +; SDAG-NEXT: buffer_load_dwordx4 v[0:3], off, s[4:7], 0 +; SDAG-NEXT: buffer_load_dwordx4 v[4:7], off, s[4:7], 0 offset:16 +; SDAG-NEXT: buffer_load_dwordx4 v[8:11], off, s[4:7], 0 offset:32 +; SDAG-NEXT: buffer_load_dwordx4 v[12:15], off, s[4:7], 0 offset:48 +; SDAG-NEXT: buffer_load_dwordx4 v[16:19], off, s[4:7], 0 offset:64 +; SDAG-NEXT: buffer_load_dwordx4 v[20:23], off, s[4:7], 0 offset:80 +; SDAG-NEXT: buffer_load_dwordx4 v[24:27], off, s[4:7], 0 offset:96 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v32i32_i32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v32i32_i32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_waitcnt vmcnt(8) +; SDAG-NEXT: buffer_store_dword v32, off, s[36:39], s32 offset:4 +; SDAG-NEXT: s_waitcnt vmcnt(8) +; SDAG-NEXT: buffer_store_dword v31, off, s[36:39], s32 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v32i32_i32: ; GFX11: ; %bb.0: @@ -4566,6 +5540,67 @@ define amdgpu_kernel void @test_call_external_void_func_v32i32_i32(i32) #0 { ; HSA-NEXT: buffer_store_dword v31, off, s[0:3], s32 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v32i32_i32: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x0 +; GISEL-NEXT: s_mov_b32 s52, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s53, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s54, -1 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: s_load_dwordx16 s[8:23], s[0:1], 0x40 +; GISEL-NEXT: s_load_dwordx16 s[36:51], s[0:1], 0x0 +; GISEL-NEXT: s_load_dword s2, s[0:1], 0x0 +; GISEL-NEXT: s_mov_b32 s55, 0xe00000 +; GISEL-NEXT: s_add_u32 s52, s52, s5 +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_addc_u32 s53, s53, 0 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: v_mov_b32_e32 v0, s2 +; GISEL-NEXT: ; kill: killed $sgpr0_sgpr1 +; GISEL-NEXT: ; kill: killed $sgpr0_sgpr1 +; GISEL-NEXT: buffer_store_dword v0, off, s[52:55], s32 offset:4 +; GISEL-NEXT: v_mov_b32_e32 v0, s23 +; GISEL-NEXT: s_mov_b64 s[0:1], s[52:53] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v32i32_i32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v32i32_i32@rel32@hi+12 +; GISEL-NEXT: buffer_store_dword v0, off, s[52:55], s32 +; GISEL-NEXT: v_mov_b32_e32 v0, s36 +; GISEL-NEXT: v_mov_b32_e32 v1, s37 +; GISEL-NEXT: v_mov_b32_e32 v2, s38 +; GISEL-NEXT: v_mov_b32_e32 v3, s39 +; GISEL-NEXT: v_mov_b32_e32 v4, s40 +; GISEL-NEXT: v_mov_b32_e32 v5, s41 +; GISEL-NEXT: v_mov_b32_e32 v6, s42 +; GISEL-NEXT: v_mov_b32_e32 v7, s43 +; GISEL-NEXT: v_mov_b32_e32 v8, s44 +; GISEL-NEXT: v_mov_b32_e32 v9, s45 +; GISEL-NEXT: v_mov_b32_e32 v10, s46 +; GISEL-NEXT: v_mov_b32_e32 v11, s47 +; GISEL-NEXT: v_mov_b32_e32 v12, s48 +; GISEL-NEXT: v_mov_b32_e32 v13, s49 +; GISEL-NEXT: v_mov_b32_e32 v14, s50 +; GISEL-NEXT: v_mov_b32_e32 v15, s51 +; GISEL-NEXT: v_mov_b32_e32 v16, s8 +; GISEL-NEXT: v_mov_b32_e32 v17, s9 +; GISEL-NEXT: v_mov_b32_e32 v18, s10 +; GISEL-NEXT: v_mov_b32_e32 v19, s11 +; GISEL-NEXT: s_mov_b64 s[2:3], s[54:55] +; GISEL-NEXT: v_mov_b32_e32 v20, s12 +; GISEL-NEXT: v_mov_b32_e32 v21, s13 +; GISEL-NEXT: v_mov_b32_e32 v22, s14 +; GISEL-NEXT: v_mov_b32_e32 v23, s15 +; GISEL-NEXT: v_mov_b32_e32 v24, s16 +; GISEL-NEXT: v_mov_b32_e32 v25, s17 +; GISEL-NEXT: v_mov_b32_e32 v26, s18 +; GISEL-NEXT: v_mov_b32_e32 v27, s19 +; GISEL-NEXT: v_mov_b32_e32 v28, s20 +; GISEL-NEXT: v_mov_b32_e32 v29, s21 +; GISEL-NEXT: v_mov_b32_e32 v30, s22 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %ptr0 = load ptr addrspace(1), ptr addrspace(4) poison %val0 = load <32 x i32>, ptr addrspace(1) %ptr0 %val1 = load i32, ptr addrspace(1) poison @@ -4622,29 +5657,29 @@ define amdgpu_kernel void @test_call_external_i32_func_i32_imm(ptr addrspace(1) ; CI-NEXT: s_waitcnt vmcnt(0) ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_i32_func_i32_imm: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s48, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s49, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s50, -1 -; GFX9-NEXT: s_mov_b32 s51, 0xe00000 -; GFX9-NEXT: s_add_u32 s48, s48, s5 -; GFX9-NEXT: s_load_dwordx2 s[36:37], s[2:3], 0x24 -; GFX9-NEXT: s_addc_u32 s49, s49, 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[48:49] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_i32_func_i32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_i32_func_i32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[50:51] -; GFX9-NEXT: v_mov_b32_e32 v0, 42 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_mov_b32 s39, 0xf000 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: buffer_store_dword v0, off, s[36:39], 0 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_i32_func_i32_imm: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s48, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s49, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s50, -1 +; SDAG-NEXT: s_mov_b32 s51, 0xe00000 +; SDAG-NEXT: s_add_u32 s48, s48, s5 +; SDAG-NEXT: s_load_dwordx2 s[36:37], s[2:3], 0x24 +; SDAG-NEXT: s_addc_u32 s49, s49, 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[48:49] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_i32_func_i32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_i32_func_i32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[50:51] +; SDAG-NEXT: v_mov_b32_e32 v0, 42 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_mov_b32 s39, 0xf000 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: buffer_store_dword v0, off, s[36:39], 0 +; SDAG-NEXT: s_waitcnt vmcnt(0) +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_i32_func_i32_imm: ; GFX11: ; %bb.0: @@ -4682,6 +5717,30 @@ define amdgpu_kernel void @test_call_external_i32_func_i32_imm(ptr addrspace(1) ; HSA-NEXT: buffer_store_dword v0, off, s[36:39], 0 ; HSA-NEXT: s_waitcnt vmcnt(0) ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_i32_func_i32_imm: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s48, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s49, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s50, -1 +; GISEL-NEXT: s_mov_b32 s51, 0xe00000 +; GISEL-NEXT: s_add_u32 s48, s48, s5 +; GISEL-NEXT: s_load_dwordx2 s[36:37], s[2:3], 0x24 +; GISEL-NEXT: s_addc_u32 s49, s49, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[48:49] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_i32_func_i32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_i32_func_i32@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 42 +; GISEL-NEXT: s_mov_b64 s[2:3], s[50:51] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xf000 +; GISEL-NEXT: buffer_store_dword v0, off, s[36:39], 0 +; GISEL-NEXT: s_waitcnt vmcnt(0) +; GISEL-NEXT: s_endpgm %val = call i32 @external_i32_func_i32(i32 42) store volatile i32 %val, ptr addrspace(1) %out ret void @@ -4736,29 +5795,29 @@ define amdgpu_kernel void @test_call_external_void_func_struct_i8_i32() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_struct_i8_i32: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: s_waitcnt lgkmcnt(0) -; GFX9-NEXT: buffer_load_ubyte v0, off, s[0:3], 0 -; GFX9-NEXT: buffer_load_dword v1, off, s[0:3], 0 offset:4 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_struct_i8_i32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_struct_i8_i32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_struct_i8_i32: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: s_waitcnt lgkmcnt(0) +; SDAG-NEXT: buffer_load_ubyte v0, off, s[0:3], 0 +; SDAG-NEXT: buffer_load_dword v1, off, s[0:3], 0 offset:4 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_struct_i8_i32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_struct_i8_i32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_struct_i8_i32: ; GFX11: ; %bb.0: @@ -4797,6 +5856,30 @@ define amdgpu_kernel void @test_call_external_void_func_struct_i8_i32() #0 { ; HSA-NEXT: s_mov_b32 s32, 0 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_struct_i8_i32: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x0 +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: s_load_dwordx2 s[4:5], s[0:1], 0x0 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[8:9] +; GISEL-NEXT: s_add_u32 s8, s8, external_void_func_struct_i8_i32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s9, s9, external_void_func_struct_i8_i32@rel32@hi+12 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: v_mov_b32_e32 v0, s4 +; GISEL-NEXT: v_mov_b32_e32 v1, s5 +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[8:9] +; GISEL-NEXT: s_endpgm %ptr0 = load ptr addrspace(1), ptr addrspace(4) poison %val = load { i8, i32 }, ptr addrspace(1) %ptr0 call void @external_void_func_struct_i8_i32({ i8, i32 } %val) @@ -4860,34 +5943,34 @@ define amdgpu_kernel void @test_call_external_void_func_byval_struct_i8_i32() #0 ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_byval_struct_i8_i32: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: v_mov_b32_e32 v0, 3 -; GFX9-NEXT: buffer_store_byte v0, off, s[36:39], 0 -; GFX9-NEXT: v_mov_b32_e32 v0, 8 -; GFX9-NEXT: buffer_store_dword v0, off, s[36:39], 0 offset:4 -; GFX9-NEXT: buffer_load_dword v0, off, s[36:39], 0 offset:4 -; GFX9-NEXT: s_nop 0 -; GFX9-NEXT: buffer_load_dword v1, off, s[36:39], 0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_movk_i32 s32, 0x400 -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_byval_struct_i8_i32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_byval_struct_i8_i32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_waitcnt vmcnt(1) -; GFX9-NEXT: buffer_store_dword v0, off, s[36:39], s32 offset:4 -; GFX9-NEXT: s_waitcnt vmcnt(1) -; GFX9-NEXT: buffer_store_dword v1, off, s[36:39], s32 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_byval_struct_i8_i32: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: v_mov_b32_e32 v0, 3 +; SDAG-NEXT: buffer_store_byte v0, off, s[36:39], 0 +; SDAG-NEXT: v_mov_b32_e32 v0, 8 +; SDAG-NEXT: buffer_store_dword v0, off, s[36:39], 0 offset:4 +; SDAG-NEXT: buffer_load_dword v0, off, s[36:39], 0 offset:4 +; SDAG-NEXT: s_nop 0 +; SDAG-NEXT: buffer_load_dword v1, off, s[36:39], 0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_movk_i32 s32, 0x400 +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_byval_struct_i8_i32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_byval_struct_i8_i32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_waitcnt vmcnt(1) +; SDAG-NEXT: buffer_store_dword v0, off, s[36:39], s32 offset:4 +; SDAG-NEXT: s_waitcnt vmcnt(1) +; SDAG-NEXT: buffer_store_dword v1, off, s[36:39], s32 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-TRUE16-LABEL: test_call_external_void_func_byval_struct_i8_i32: ; GFX11-TRUE16: ; %bb.0: @@ -4948,6 +6031,35 @@ define amdgpu_kernel void @test_call_external_void_func_byval_struct_i8_i32() #0 ; HSA-NEXT: buffer_store_dword v1, off, s[0:3], s32 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_byval_struct_i8_i32: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: v_mov_b32_e32 v0, 3 +; GISEL-NEXT: buffer_store_byte v0, off, s[36:39], 0 +; GISEL-NEXT: v_mov_b32_e32 v0, 8 +; GISEL-NEXT: buffer_store_dword v0, off, s[36:39], 0 offset:4 +; GISEL-NEXT: buffer_load_dword v0, off, s[36:39], 0 +; GISEL-NEXT: s_nop 0 +; GISEL-NEXT: buffer_load_dword v1, off, s[36:39], 0 offset:4 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_movk_i32 s32, 0x400 +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_byval_struct_i8_i32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_byval_struct_i8_i32@rel32@hi+12 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_waitcnt vmcnt(1) +; GISEL-NEXT: buffer_store_dword v0, off, s[36:39], s32 +; GISEL-NEXT: s_waitcnt vmcnt(1) +; GISEL-NEXT: buffer_store_dword v1, off, s[36:39], s32 offset:4 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %val = alloca { i8, i32 }, align 8, addrspace(5) %gep0 = getelementptr inbounds { i8, i32 }, ptr addrspace(5) %val, i32 0, i32 0 %gep1 = getelementptr inbounds { i8, i32 }, ptr addrspace(5) %val, i32 0, i32 1 @@ -5034,44 +6146,44 @@ define amdgpu_kernel void @test_call_external_void_func_sret_struct_i8_i32_byval ; CI-NEXT: s_waitcnt vmcnt(0) ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_sret_struct_i8_i32_byval_struct_i8_i32: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s5 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: v_mov_b32_e32 v0, 3 -; GFX9-NEXT: buffer_store_byte v0, off, s[36:39], 0 -; GFX9-NEXT: v_mov_b32_e32 v0, 8 -; GFX9-NEXT: buffer_store_dword v0, off, s[36:39], 0 offset:4 -; GFX9-NEXT: buffer_load_dword v0, off, s[36:39], 0 offset:4 -; GFX9-NEXT: s_nop 0 -; GFX9-NEXT: buffer_load_dword v1, off, s[36:39], 0 -; GFX9-NEXT: s_movk_i32 s32, 0x800 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_sret_struct_i8_i32_byval_struct_i8_i32@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_sret_struct_i8_i32_byval_struct_i8_i32@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_waitcnt vmcnt(1) -; GFX9-NEXT: buffer_store_dword v0, off, s[36:39], s32 offset:4 -; GFX9-NEXT: s_waitcnt vmcnt(1) -; GFX9-NEXT: buffer_store_dword v1, off, s[36:39], s32 -; GFX9-NEXT: v_mov_b32_e32 v0, 8 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: buffer_load_ubyte v0, off, s[36:39], 0 offset:8 -; GFX9-NEXT: buffer_load_dword v1, off, s[36:39], 0 offset:12 -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: s_waitcnt vmcnt(1) -; GFX9-NEXT: buffer_store_byte v0, off, s[0:3], 0 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: buffer_store_dword v1, off, s[0:3], 0 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_sret_struct_i8_i32_byval_struct_i8_i32: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s5 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: v_mov_b32_e32 v0, 3 +; SDAG-NEXT: buffer_store_byte v0, off, s[36:39], 0 +; SDAG-NEXT: v_mov_b32_e32 v0, 8 +; SDAG-NEXT: buffer_store_dword v0, off, s[36:39], 0 offset:4 +; SDAG-NEXT: buffer_load_dword v0, off, s[36:39], 0 offset:4 +; SDAG-NEXT: s_nop 0 +; SDAG-NEXT: buffer_load_dword v1, off, s[36:39], 0 +; SDAG-NEXT: s_movk_i32 s32, 0x800 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_sret_struct_i8_i32_byval_struct_i8_i32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_sret_struct_i8_i32_byval_struct_i8_i32@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_waitcnt vmcnt(1) +; SDAG-NEXT: buffer_store_dword v0, off, s[36:39], s32 offset:4 +; SDAG-NEXT: s_waitcnt vmcnt(1) +; SDAG-NEXT: buffer_store_dword v1, off, s[36:39], s32 +; SDAG-NEXT: v_mov_b32_e32 v0, 8 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: buffer_load_ubyte v0, off, s[36:39], 0 offset:8 +; SDAG-NEXT: buffer_load_dword v1, off, s[36:39], 0 offset:12 +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: s_waitcnt vmcnt(1) +; SDAG-NEXT: buffer_store_byte v0, off, s[0:3], 0 +; SDAG-NEXT: s_waitcnt vmcnt(0) +; SDAG-NEXT: buffer_store_dword v1, off, s[0:3], 0 +; SDAG-NEXT: s_waitcnt vmcnt(0) +; SDAG-NEXT: s_endpgm ; ; GFX11-TRUE16-LABEL: test_call_external_void_func_sret_struct_i8_i32_byval_struct_i8_i32: ; GFX11-TRUE16: ; %bb.0: @@ -5170,6 +6282,45 @@ define amdgpu_kernel void @test_call_external_void_func_sret_struct_i8_i32_byval ; HSA-NEXT: buffer_store_dword v1, off, s[4:7], 0 ; HSA-NEXT: s_waitcnt vmcnt(0) ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_sret_struct_i8_i32_byval_struct_i8_i32: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s5 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: v_mov_b32_e32 v0, 3 +; GISEL-NEXT: buffer_store_byte v0, off, s[36:39], 0 +; GISEL-NEXT: v_mov_b32_e32 v0, 8 +; GISEL-NEXT: buffer_store_dword v0, off, s[36:39], 0 offset:4 +; GISEL-NEXT: buffer_load_dword v0, off, s[36:39], 0 +; GISEL-NEXT: s_nop 0 +; GISEL-NEXT: buffer_load_dword v1, off, s[36:39], 0 offset:4 +; GISEL-NEXT: s_movk_i32 s32, 0x800 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_sret_struct_i8_i32_byval_struct_i8_i32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_sret_struct_i8_i32_byval_struct_i8_i32@rel32@hi+12 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_waitcnt vmcnt(1) +; GISEL-NEXT: buffer_store_dword v0, off, s[36:39], s32 +; GISEL-NEXT: s_waitcnt vmcnt(1) +; GISEL-NEXT: buffer_store_dword v1, off, s[36:39], s32 offset:4 +; GISEL-NEXT: v_mov_b32_e32 v0, 8 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: buffer_load_ubyte v0, off, s[36:39], 0 offset:8 +; GISEL-NEXT: buffer_load_dword v1, off, s[36:39], 0 offset:12 +; GISEL-NEXT: s_mov_b32 s2, -1 +; GISEL-NEXT: s_mov_b32 s3, 0xf000 +; GISEL-NEXT: s_waitcnt vmcnt(1) +; GISEL-NEXT: buffer_store_byte v0, off, s[0:3], 0 +; GISEL-NEXT: s_waitcnt vmcnt(0) +; GISEL-NEXT: buffer_store_dword v1, off, s[0:3], 0 +; GISEL-NEXT: s_waitcnt vmcnt(0) +; GISEL-NEXT: s_endpgm %in.val = alloca { i8, i32 }, align 8, addrspace(5) %out.val = alloca { i8, i32 }, align 8, addrspace(5) %in.gep0 = getelementptr inbounds { i8, i32 }, ptr addrspace(5) %in.val, i32 0, i32 0 @@ -5272,47 +6423,47 @@ define amdgpu_kernel void @test_call_external_void_func_v16i8() #0 { ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: test_call_external_void_func_v16i8: -; GFX9: ; %bb.0: -; GFX9-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x0 -; GFX9-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s38, -1 -; GFX9-NEXT: s_mov_b32 s39, 0xe00000 -; GFX9-NEXT: s_add_u32 s36, s36, s3 -; GFX9-NEXT: s_mov_b32 s3, 0xf000 -; GFX9-NEXT: s_mov_b32 s2, -1 -; GFX9-NEXT: s_waitcnt lgkmcnt(0) -; GFX9-NEXT: buffer_load_dwordx4 v[0:3], off, s[0:3], 0 -; GFX9-NEXT: s_addc_u32 s37, s37, 0 -; GFX9-NEXT: s_mov_b64 s[0:1], s[36:37] -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, external_void_func_v16i8@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, external_void_func_v16i8@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[38:39] -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_waitcnt vmcnt(0) -; GFX9-NEXT: v_lshrrev_b32_e32 v16, 8, v0 -; GFX9-NEXT: v_lshrrev_b32_e32 v17, 16, v0 -; GFX9-NEXT: v_lshrrev_b32_e32 v18, 24, v0 -; GFX9-NEXT: v_lshrrev_b32_e32 v5, 8, v1 -; GFX9-NEXT: v_lshrrev_b32_e32 v6, 16, v1 -; GFX9-NEXT: v_lshrrev_b32_e32 v7, 24, v1 -; GFX9-NEXT: v_lshrrev_b32_e32 v9, 8, v2 -; GFX9-NEXT: v_lshrrev_b32_e32 v10, 16, v2 -; GFX9-NEXT: v_lshrrev_b32_e32 v11, 24, v2 -; GFX9-NEXT: v_lshrrev_b32_e32 v13, 8, v3 -; GFX9-NEXT: v_lshrrev_b32_e32 v14, 16, v3 -; GFX9-NEXT: v_lshrrev_b32_e32 v15, 24, v3 -; GFX9-NEXT: v_mov_b32_e32 v4, v1 -; GFX9-NEXT: v_mov_b32_e32 v8, v2 -; GFX9-NEXT: v_mov_b32_e32 v12, v3 -; GFX9-NEXT: v_mov_b32_e32 v1, v16 -; GFX9-NEXT: v_mov_b32_e32 v2, v17 -; GFX9-NEXT: v_mov_b32_e32 v3, v18 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: test_call_external_void_func_v16i8: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x0 +; SDAG-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s38, -1 +; SDAG-NEXT: s_mov_b32 s39, 0xe00000 +; SDAG-NEXT: s_add_u32 s36, s36, s3 +; SDAG-NEXT: s_mov_b32 s3, 0xf000 +; SDAG-NEXT: s_mov_b32 s2, -1 +; SDAG-NEXT: s_waitcnt lgkmcnt(0) +; SDAG-NEXT: buffer_load_dwordx4 v[0:3], off, s[0:3], 0 +; SDAG-NEXT: s_addc_u32 s37, s37, 0 +; SDAG-NEXT: s_mov_b64 s[0:1], s[36:37] +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, external_void_func_v16i8@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, external_void_func_v16i8@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[38:39] +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_waitcnt vmcnt(0) +; SDAG-NEXT: v_lshrrev_b32_e32 v16, 8, v0 +; SDAG-NEXT: v_lshrrev_b32_e32 v17, 16, v0 +; SDAG-NEXT: v_lshrrev_b32_e32 v18, 24, v0 +; SDAG-NEXT: v_lshrrev_b32_e32 v5, 8, v1 +; SDAG-NEXT: v_lshrrev_b32_e32 v6, 16, v1 +; SDAG-NEXT: v_lshrrev_b32_e32 v7, 24, v1 +; SDAG-NEXT: v_lshrrev_b32_e32 v9, 8, v2 +; SDAG-NEXT: v_lshrrev_b32_e32 v10, 16, v2 +; SDAG-NEXT: v_lshrrev_b32_e32 v11, 24, v2 +; SDAG-NEXT: v_lshrrev_b32_e32 v13, 8, v3 +; SDAG-NEXT: v_lshrrev_b32_e32 v14, 16, v3 +; SDAG-NEXT: v_lshrrev_b32_e32 v15, 24, v3 +; SDAG-NEXT: v_mov_b32_e32 v4, v1 +; SDAG-NEXT: v_mov_b32_e32 v8, v2 +; SDAG-NEXT: v_mov_b32_e32 v12, v3 +; SDAG-NEXT: v_mov_b32_e32 v1, v16 +; SDAG-NEXT: v_mov_b32_e32 v2, v17 +; SDAG-NEXT: v_mov_b32_e32 v3, v18 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: test_call_external_void_func_v16i8: ; GFX11: ; %bb.0: @@ -5384,6 +6535,56 @@ define amdgpu_kernel void @test_call_external_void_func_v16i8() #0 { ; HSA-NEXT: v_mov_b32_e32 v3, v18 ; HSA-NEXT: s_swappc_b64 s[30:31], s[8:9] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: test_call_external_void_func_v16i8: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_load_dwordx2 s[0:1], s[0:1], 0x0 +; GISEL-NEXT: s_mov_b32 s36, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s37, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s38, -1 +; GISEL-NEXT: s_mov_b32 s39, 0xe00000 +; GISEL-NEXT: s_add_u32 s36, s36, s3 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: s_load_dwordx4 s[0:3], s[0:1], 0x0 +; GISEL-NEXT: s_addc_u32 s37, s37, 0 +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: s_lshr_b32 s8, s0, 8 +; GISEL-NEXT: s_lshr_b32 s9, s0, 16 +; GISEL-NEXT: s_lshr_b32 s10, s0, 24 +; GISEL-NEXT: s_lshr_b32 s11, s1, 8 +; GISEL-NEXT: s_lshr_b32 s12, s1, 16 +; GISEL-NEXT: s_lshr_b32 s13, s1, 24 +; GISEL-NEXT: s_lshr_b32 s14, s2, 8 +; GISEL-NEXT: s_lshr_b32 s15, s2, 16 +; GISEL-NEXT: s_lshr_b32 s16, s2, 24 +; GISEL-NEXT: s_lshr_b32 s17, s3, 8 +; GISEL-NEXT: s_lshr_b32 s18, s3, 16 +; GISEL-NEXT: s_lshr_b32 s19, s3, 24 +; GISEL-NEXT: v_mov_b32_e32 v0, s0 +; GISEL-NEXT: v_mov_b32_e32 v4, s1 +; GISEL-NEXT: v_mov_b32_e32 v8, s2 +; GISEL-NEXT: v_mov_b32_e32 v12, s3 +; GISEL-NEXT: s_mov_b64 s[0:1], s[36:37] +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, external_void_func_v16i8@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, external_void_func_v16i8@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v1, s8 +; GISEL-NEXT: v_mov_b32_e32 v2, s9 +; GISEL-NEXT: v_mov_b32_e32 v3, s10 +; GISEL-NEXT: v_mov_b32_e32 v5, s11 +; GISEL-NEXT: v_mov_b32_e32 v6, s12 +; GISEL-NEXT: v_mov_b32_e32 v7, s13 +; GISEL-NEXT: v_mov_b32_e32 v9, s14 +; GISEL-NEXT: v_mov_b32_e32 v10, s15 +; GISEL-NEXT: v_mov_b32_e32 v11, s16 +; GISEL-NEXT: v_mov_b32_e32 v13, s17 +; GISEL-NEXT: v_mov_b32_e32 v14, s18 +; GISEL-NEXT: v_mov_b32_e32 v15, s19 +; GISEL-NEXT: s_mov_b64 s[2:3], s[38:39] +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm %ptr = load ptr addrspace(1), ptr addrspace(4) poison %val = load <16 x i8>, ptr addrspace(1) %ptr call void @external_void_func_v16i8(<16 x i8> %val) @@ -5509,64 +6710,64 @@ define amdgpu_kernel void @stack_passed_arg_alignment_v32i32_f64(<32 x i32> %val ; CI-NEXT: s_swappc_b64 s[30:31], s[4:5] ; CI-NEXT: s_endpgm ; -; GFX9-LABEL: stack_passed_arg_alignment_v32i32_f64: -; GFX9: ; %bb.0: ; %entry -; GFX9-NEXT: s_mov_b32 s52, SCRATCH_RSRC_DWORD0 -; GFX9-NEXT: s_mov_b32 s53, SCRATCH_RSRC_DWORD1 -; GFX9-NEXT: s_mov_b32 s54, -1 -; GFX9-NEXT: s_mov_b32 s55, 0xe00000 -; GFX9-NEXT: s_add_u32 s52, s52, s5 -; GFX9-NEXT: s_load_dwordx16 s[8:23], s[2:3], 0x64 -; GFX9-NEXT: s_load_dwordx2 s[4:5], s[2:3], 0xa4 -; GFX9-NEXT: s_load_dwordx16 s[36:51], s[2:3], 0x24 -; GFX9-NEXT: s_mov_b32 s32, 0 -; GFX9-NEXT: s_addc_u32 s53, s53, 0 -; GFX9-NEXT: s_waitcnt lgkmcnt(0) -; GFX9-NEXT: v_mov_b32_e32 v0, s23 -; GFX9-NEXT: buffer_store_dword v0, off, s[52:55], s32 -; GFX9-NEXT: v_mov_b32_e32 v0, s4 -; GFX9-NEXT: buffer_store_dword v0, off, s[52:55], s32 offset:4 -; GFX9-NEXT: v_mov_b32_e32 v0, s5 -; GFX9-NEXT: s_mov_b64 s[6:7], s[0:1] -; GFX9-NEXT: s_mov_b64 s[0:1], s[52:53] -; GFX9-NEXT: buffer_store_dword v0, off, s[52:55], s32 offset:8 -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, stack_passed_f64_arg@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, stack_passed_f64_arg@rel32@hi+12 -; GFX9-NEXT: s_mov_b64 s[2:3], s[54:55] -; GFX9-NEXT: v_mov_b32_e32 v0, s36 -; GFX9-NEXT: v_mov_b32_e32 v1, s37 -; GFX9-NEXT: v_mov_b32_e32 v2, s38 -; GFX9-NEXT: v_mov_b32_e32 v3, s39 -; GFX9-NEXT: v_mov_b32_e32 v4, s40 -; GFX9-NEXT: v_mov_b32_e32 v5, s41 -; GFX9-NEXT: v_mov_b32_e32 v6, s42 -; GFX9-NEXT: v_mov_b32_e32 v7, s43 -; GFX9-NEXT: v_mov_b32_e32 v8, s44 -; GFX9-NEXT: v_mov_b32_e32 v9, s45 -; GFX9-NEXT: v_mov_b32_e32 v10, s46 -; GFX9-NEXT: v_mov_b32_e32 v11, s47 -; GFX9-NEXT: v_mov_b32_e32 v12, s48 -; GFX9-NEXT: v_mov_b32_e32 v13, s49 -; GFX9-NEXT: v_mov_b32_e32 v14, s50 -; GFX9-NEXT: v_mov_b32_e32 v15, s51 -; GFX9-NEXT: v_mov_b32_e32 v16, s8 -; GFX9-NEXT: v_mov_b32_e32 v17, s9 -; GFX9-NEXT: v_mov_b32_e32 v18, s10 -; GFX9-NEXT: v_mov_b32_e32 v19, s11 -; GFX9-NEXT: v_mov_b32_e32 v20, s12 -; GFX9-NEXT: v_mov_b32_e32 v21, s13 -; GFX9-NEXT: v_mov_b32_e32 v22, s14 -; GFX9-NEXT: v_mov_b32_e32 v23, s15 -; GFX9-NEXT: v_mov_b32_e32 v24, s16 -; GFX9-NEXT: v_mov_b32_e32 v25, s17 -; GFX9-NEXT: v_mov_b32_e32 v26, s18 -; GFX9-NEXT: v_mov_b32_e32 v27, s19 -; GFX9-NEXT: v_mov_b32_e32 v28, s20 -; GFX9-NEXT: v_mov_b32_e32 v29, s21 -; GFX9-NEXT: v_mov_b32_e32 v30, s22 -; GFX9-NEXT: s_swappc_b64 s[30:31], s[4:5] -; GFX9-NEXT: s_endpgm +; SDAG-LABEL: stack_passed_arg_alignment_v32i32_f64: +; SDAG: ; %bb.0: ; %entry +; SDAG-NEXT: s_mov_b32 s52, SCRATCH_RSRC_DWORD0 +; SDAG-NEXT: s_mov_b32 s53, SCRATCH_RSRC_DWORD1 +; SDAG-NEXT: s_mov_b32 s54, -1 +; SDAG-NEXT: s_mov_b32 s55, 0xe00000 +; SDAG-NEXT: s_add_u32 s52, s52, s5 +; SDAG-NEXT: s_load_dwordx16 s[8:23], s[2:3], 0x64 +; SDAG-NEXT: s_load_dwordx2 s[4:5], s[2:3], 0xa4 +; SDAG-NEXT: s_load_dwordx16 s[36:51], s[2:3], 0x24 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_addc_u32 s53, s53, 0 +; SDAG-NEXT: s_waitcnt lgkmcnt(0) +; SDAG-NEXT: v_mov_b32_e32 v0, s23 +; SDAG-NEXT: buffer_store_dword v0, off, s[52:55], s32 +; SDAG-NEXT: v_mov_b32_e32 v0, s4 +; SDAG-NEXT: buffer_store_dword v0, off, s[52:55], s32 offset:4 +; SDAG-NEXT: v_mov_b32_e32 v0, s5 +; SDAG-NEXT: s_mov_b64 s[6:7], s[0:1] +; SDAG-NEXT: s_mov_b64 s[0:1], s[52:53] +; SDAG-NEXT: buffer_store_dword v0, off, s[52:55], s32 offset:8 +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, stack_passed_f64_arg@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, stack_passed_f64_arg@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[2:3], s[54:55] +; SDAG-NEXT: v_mov_b32_e32 v0, s36 +; SDAG-NEXT: v_mov_b32_e32 v1, s37 +; SDAG-NEXT: v_mov_b32_e32 v2, s38 +; SDAG-NEXT: v_mov_b32_e32 v3, s39 +; SDAG-NEXT: v_mov_b32_e32 v4, s40 +; SDAG-NEXT: v_mov_b32_e32 v5, s41 +; SDAG-NEXT: v_mov_b32_e32 v6, s42 +; SDAG-NEXT: v_mov_b32_e32 v7, s43 +; SDAG-NEXT: v_mov_b32_e32 v8, s44 +; SDAG-NEXT: v_mov_b32_e32 v9, s45 +; SDAG-NEXT: v_mov_b32_e32 v10, s46 +; SDAG-NEXT: v_mov_b32_e32 v11, s47 +; SDAG-NEXT: v_mov_b32_e32 v12, s48 +; SDAG-NEXT: v_mov_b32_e32 v13, s49 +; SDAG-NEXT: v_mov_b32_e32 v14, s50 +; SDAG-NEXT: v_mov_b32_e32 v15, 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_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: s_swappc_b64 s[30:31], s[4:5] +; SDAG-NEXT: s_endpgm ; ; GFX11-LABEL: stack_passed_arg_alignment_v32i32_f64: ; GFX11: ; %bb.0: ; %entry @@ -5662,6 +6863,65 @@ define amdgpu_kernel void @stack_passed_arg_alignment_v32i32_f64(<32 x i32> %val ; HSA-NEXT: v_mov_b32_e32 v30, s22 ; HSA-NEXT: s_swappc_b64 s[30:31], s[24:25] ; HSA-NEXT: s_endpgm +; +; GISEL-LABEL: stack_passed_arg_alignment_v32i32_f64: +; GISEL: ; %bb.0: ; %entry +; GISEL-NEXT: s_mov_b32 s52, SCRATCH_RSRC_DWORD0 +; GISEL-NEXT: s_mov_b32 s53, SCRATCH_RSRC_DWORD1 +; GISEL-NEXT: s_mov_b32 s54, -1 +; GISEL-NEXT: s_mov_b64 s[6:7], s[0:1] +; GISEL-NEXT: s_load_dwordx16 s[8:23], s[2:3], 0x64 +; GISEL-NEXT: s_load_dwordx2 s[0:1], s[2:3], 0xa4 +; GISEL-NEXT: s_load_dwordx16 s[36:51], s[2:3], 0x24 +; GISEL-NEXT: s_mov_b32 s55, 0xe00000 +; GISEL-NEXT: s_add_u32 s52, s52, s5 +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_addc_u32 s53, s53, 0 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: v_mov_b32_e32 v0, s23 +; GISEL-NEXT: buffer_store_dword v0, off, s[52:55], s32 +; GISEL-NEXT: v_mov_b32_e32 v0, s0 +; GISEL-NEXT: buffer_store_dword v0, off, s[52:55], s32 offset:4 +; GISEL-NEXT: v_mov_b32_e32 v0, s1 +; GISEL-NEXT: s_mov_b64 s[0:1], s[52:53] +; GISEL-NEXT: buffer_store_dword v0, off, s[52:55], s32 offset:8 +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, stack_passed_f64_arg@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, stack_passed_f64_arg@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, s36 +; GISEL-NEXT: v_mov_b32_e32 v1, s37 +; GISEL-NEXT: v_mov_b32_e32 v2, s38 +; GISEL-NEXT: v_mov_b32_e32 v3, s39 +; GISEL-NEXT: v_mov_b32_e32 v4, s40 +; GISEL-NEXT: v_mov_b32_e32 v5, s41 +; GISEL-NEXT: v_mov_b32_e32 v6, s42 +; GISEL-NEXT: v_mov_b32_e32 v7, s43 +; GISEL-NEXT: v_mov_b32_e32 v8, s44 +; GISEL-NEXT: v_mov_b32_e32 v9, s45 +; GISEL-NEXT: v_mov_b32_e32 v10, s46 +; GISEL-NEXT: v_mov_b32_e32 v11, s47 +; GISEL-NEXT: v_mov_b32_e32 v12, s48 +; GISEL-NEXT: v_mov_b32_e32 v13, s49 +; GISEL-NEXT: v_mov_b32_e32 v14, s50 +; GISEL-NEXT: v_mov_b32_e32 v15, s51 +; GISEL-NEXT: v_mov_b32_e32 v16, s8 +; GISEL-NEXT: v_mov_b32_e32 v17, s9 +; GISEL-NEXT: v_mov_b32_e32 v18, s10 +; GISEL-NEXT: v_mov_b32_e32 v19, s11 +; GISEL-NEXT: s_mov_b64 s[2:3], s[54:55] +; GISEL-NEXT: v_mov_b32_e32 v20, s12 +; GISEL-NEXT: v_mov_b32_e32 v21, s13 +; GISEL-NEXT: v_mov_b32_e32 v22, s14 +; GISEL-NEXT: v_mov_b32_e32 v23, s15 +; GISEL-NEXT: v_mov_b32_e32 v24, s16 +; GISEL-NEXT: v_mov_b32_e32 v25, s17 +; GISEL-NEXT: v_mov_b32_e32 v26, s18 +; GISEL-NEXT: v_mov_b32_e32 v27, s19 +; GISEL-NEXT: v_mov_b32_e32 v28, s20 +; GISEL-NEXT: v_mov_b32_e32 v29, s21 +; GISEL-NEXT: v_mov_b32_e32 v30, s22 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[4:5] +; GISEL-NEXT: s_endpgm entry: call void @stack_passed_f64_arg(<32 x i32> %val, double %tmp) ret void @@ -5702,22 +6962,22 @@ define void @tail_call_byval_align16(<32 x i32> %val, double %tmp) #0 { ; CI-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:16 ; CI-NEXT: s_setpc_b64 s[4:5] ; -; GFX9-LABEL: tail_call_byval_align16: -; GFX9: ; %bb.0: ; %entry -; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) -; GFX9-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:28 -; GFX9-NEXT: buffer_load_dword v32, off, s[0:3], s32 -; GFX9-NEXT: s_getpc_b64 s[4:5] -; GFX9-NEXT: s_add_u32 s4, s4, byval_align16_f64_arg@rel32@lo+4 -; GFX9-NEXT: s_addc_u32 s5, s5, byval_align16_f64_arg@rel32@hi+12 -; GFX9-NEXT: s_waitcnt vmcnt(1) -; GFX9-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:20 -; GFX9-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:24 -; GFX9-NEXT: s_waitcnt vmcnt(2) -; GFX9-NEXT: buffer_store_dword v32, off, s[0:3], s32 -; GFX9-NEXT: s_waitcnt vmcnt(1) -; GFX9-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:16 -; GFX9-NEXT: s_setpc_b64 s[4:5] +; SDAG-LABEL: tail_call_byval_align16: +; SDAG: ; %bb.0: ; %entry +; SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; SDAG-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:28 +; SDAG-NEXT: buffer_load_dword v32, off, s[0:3], s32 +; SDAG-NEXT: s_getpc_b64 s[4:5] +; SDAG-NEXT: s_add_u32 s4, s4, byval_align16_f64_arg@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s5, s5, byval_align16_f64_arg@rel32@hi+12 +; SDAG-NEXT: s_waitcnt vmcnt(1) +; SDAG-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:20 +; SDAG-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:24 +; SDAG-NEXT: s_waitcnt vmcnt(2) +; SDAG-NEXT: buffer_store_dword v32, off, s[0:3], s32 +; SDAG-NEXT: s_waitcnt vmcnt(1) +; SDAG-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:16 +; SDAG-NEXT: s_setpc_b64 s[4:5] ; ; GFX11-LABEL: tail_call_byval_align16: ; GFX11: ; %bb.0: ; %entry @@ -5749,6 +7009,23 @@ define void @tail_call_byval_align16(<32 x i32> %val, double %tmp) #0 { ; HSA-NEXT: s_waitcnt vmcnt(1) ; HSA-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:16 ; HSA-NEXT: s_setpc_b64 s[4:5] +; +; GISEL-LABEL: tail_call_byval_align16: +; GISEL: ; %bb.0: ; %entry +; GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GISEL-NEXT: buffer_load_dword v31, off, s[0:3], s32 +; GISEL-NEXT: s_getpc_b64 s[4:5] +; GISEL-NEXT: s_add_u32 s4, s4, byval_align16_f64_arg@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s5, s5, byval_align16_f64_arg@rel32@hi+12 +; GISEL-NEXT: s_waitcnt vmcnt(0) +; GISEL-NEXT: buffer_store_dword v31, off, s[0:3], s32 +; GISEL-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:28 +; GISEL-NEXT: s_waitcnt vmcnt(0) +; GISEL-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:20 +; GISEL-NEXT: buffer_load_dword v31, off, s[0:3], s32 offset:24 +; GISEL-NEXT: s_waitcnt vmcnt(0) +; GISEL-NEXT: buffer_store_dword v31, off, s[0:3], s32 offset:16 +; GISEL-NEXT: s_setpc_b64 s[4:5] entry: %alloca = alloca double, align 8, addrspace(5) tail call void @byval_align16_f64_arg(<32 x i32> %val, ptr addrspace(5) byval(double) align 16 %alloca) diff --git a/llvm/test/CodeGen/AMDGPU/call-c-function.ll b/llvm/test/CodeGen/AMDGPU/call-c-function.ll index e1bb3ea..4fbc727 100644 --- a/llvm/test/CodeGen/AMDGPU/call-c-function.ll +++ b/llvm/test/CodeGen/AMDGPU/call-c-function.ll @@ -1,21 +1,68 @@ ; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py -; RUN: llc -global-isel=0 -stop-after=finalize-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -o - %s | FileCheck -enable-var-scope %s +; RUN: llc -global-isel=0 -stop-after=finalize-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -o - %s | FileCheck -check-prefix=SDAG -enable-var-scope %s +; RUN: llc -global-isel=1 -stop-after=finalize-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -o - %s | FileCheck -check-prefix=GISEL -enable-var-scope %s ; Test that we don't explode on calls from shaders to functions with the C calling convention. define amdgpu_ps void @amdgpu_ps_call_default_cc() { - ; CHECK-LABEL: name: amdgpu_ps_call_default_cc - ; CHECK: bb.0.main_body: - ; CHECK-NEXT: S_ENDPGM 0 + ; SDAG-LABEL: name: amdgpu_ps_call_default_cc + ; SDAG: bb.0.main_body: + ; SDAG-NEXT: S_ENDPGM 0 + ; + ; GISEL-LABEL: name: amdgpu_ps_call_default_cc + ; GISEL: bb.1.main_body: + ; GISEL-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $scc, implicit-def $sgpr32, implicit $sgpr32 + ; GISEL-NEXT: [[DEF:%[0-9]+]]:sreg_64 = IMPLICIT_DEF + ; GISEL-NEXT: [[DEF1:%[0-9]+]]:sreg_64 = IMPLICIT_DEF + ; GISEL-NEXT: [[DEF2:%[0-9]+]]:sreg_32 = IMPLICIT_DEF + ; GISEL-NEXT: [[COPY:%[0-9]+]]:sgpr_128 = COPY $sgpr96_sgpr97_sgpr98_sgpr99 + ; GISEL-NEXT: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]] + ; GISEL-NEXT: $sgpr4_sgpr5 = COPY [[DEF]] + ; GISEL-NEXT: $sgpr6_sgpr7 = COPY [[DEF]] + ; GISEL-NEXT: [[S_MOV_B:%[0-9]+]]:sreg_64 = S_MOV_B64_IMM_PSEUDO 0 + ; GISEL-NEXT: $sgpr8_sgpr9 = COPY [[S_MOV_B]] + ; GISEL-NEXT: $sgpr10_sgpr11 = COPY [[DEF1]] + ; GISEL-NEXT: $sgpr12 = COPY [[DEF2]] + ; GISEL-NEXT: $sgpr13 = COPY [[DEF2]] + ; GISEL-NEXT: $sgpr14 = COPY [[DEF2]] + ; GISEL-NEXT: $sgpr15 = COPY [[DEF2]] + ; GISEL-NEXT: $vgpr31 = COPY [[DEF2]] + ; GISEL-NEXT: [[S_MOV_B1:%[0-9]+]]:sreg_64 = S_MOV_B64_IMM_PSEUDO 0 + ; GISEL-NEXT: $sgpr30_sgpr31 = noconvergent SI_CALL [[S_MOV_B1]], 0, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31 + ; GISEL-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $scc, implicit-def $sgpr32, implicit $sgpr32 + ; GISEL-NEXT: S_ENDPGM 0 main_body: call void null() ret void } define amdgpu_gfx void @amdgpu_gfx_call_default_cc() { - ; CHECK-LABEL: name: amdgpu_gfx_call_default_cc - ; CHECK: bb.0.main_body: - ; CHECK-NEXT: SI_RETURN + ; SDAG-LABEL: name: amdgpu_gfx_call_default_cc + ; SDAG: bb.0.main_body: + ; SDAG-NEXT: SI_RETURN + ; + ; GISEL-LABEL: name: amdgpu_gfx_call_default_cc + ; GISEL: bb.1.main_body: + ; GISEL-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $scc, implicit-def $sgpr32, implicit $sgpr32 + ; GISEL-NEXT: [[DEF:%[0-9]+]]:sreg_64 = IMPLICIT_DEF + ; GISEL-NEXT: [[DEF1:%[0-9]+]]:sreg_64 = IMPLICIT_DEF + ; GISEL-NEXT: [[DEF2:%[0-9]+]]:sreg_32 = IMPLICIT_DEF + ; GISEL-NEXT: [[COPY:%[0-9]+]]:sgpr_128 = COPY $sgpr0_sgpr1_sgpr2_sgpr3 + ; GISEL-NEXT: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]] + ; GISEL-NEXT: $sgpr4_sgpr5 = COPY [[DEF]] + ; GISEL-NEXT: $sgpr6_sgpr7 = COPY [[DEF]] + ; GISEL-NEXT: [[S_MOV_B:%[0-9]+]]:sreg_64 = S_MOV_B64_IMM_PSEUDO 0 + ; GISEL-NEXT: $sgpr8_sgpr9 = COPY [[S_MOV_B]] + ; GISEL-NEXT: $sgpr10_sgpr11 = COPY [[DEF1]] + ; GISEL-NEXT: $sgpr12 = COPY [[DEF2]] + ; GISEL-NEXT: $sgpr13 = COPY [[DEF2]] + ; GISEL-NEXT: $sgpr14 = COPY [[DEF2]] + ; GISEL-NEXT: $sgpr15 = COPY [[DEF2]] + ; GISEL-NEXT: $vgpr31 = COPY [[DEF2]] + ; GISEL-NEXT: [[S_MOV_B1:%[0-9]+]]:sreg_64 = S_MOV_B64_IMM_PSEUDO 0 + ; GISEL-NEXT: $sgpr30_sgpr31 = noconvergent SI_CALL [[S_MOV_B1]], 0, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit $sgpr4_sgpr5, implicit $sgpr6_sgpr7, implicit $sgpr8_sgpr9, implicit $sgpr10_sgpr11, implicit $sgpr12, implicit $sgpr13, implicit $sgpr14, implicit $sgpr15, implicit $vgpr31 + ; GISEL-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $scc, implicit-def $sgpr32, implicit $sgpr32 + ; GISEL-NEXT: SI_RETURN main_body: call void null() ret void diff --git a/llvm/test/CodeGen/AMDGPU/call-constexpr.ll b/llvm/test/CodeGen/AMDGPU/call-constexpr.ll index 5f324df..fe0b018 100644 --- a/llvm/test/CodeGen/AMDGPU/call-constexpr.ll +++ b/llvm/test/CodeGen/AMDGPU/call-constexpr.ll @@ -1,84 +1,341 @@ -; RUN: llc -mtriple=amdgcn-amd-amdhsa < %s | FileCheck -check-prefix=GCN %s +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4 +; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa < %s | FileCheck -check-prefixes=GCN,SDAG %s +; RUN: llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa < %s | FileCheck -check-prefixes=GCN,GISEL %s -; GCN-LABEL: {{^}}test_bitcast_return_type_noinline: -; GCN: s_getpc_b64 -; GCN: s_add_u32 s{{[0-9]+}}, s{{[0-9]+}}, ret_i32_noinline@rel32@lo+4 -; GCN: s_addc_u32 s{{[0-9]+}}, s{{[0-9]+}}, ret_i32_noinline@rel32@hi+12 -; GCN: s_swappc_b64 define amdgpu_kernel void @test_bitcast_return_type_noinline() #0 { +; SDAG-LABEL: test_bitcast_return_type_noinline: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 +; SDAG-NEXT: s_add_i32 s12, s12, s17 +; SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; SDAG-NEXT: s_add_u32 s0, s0, s17 +; SDAG-NEXT: s_addc_u32 s1, s1, 0 +; SDAG-NEXT: s_mov_b32 s13, s15 +; SDAG-NEXT: s_mov_b32 s12, s14 +; SDAG-NEXT: s_getpc_b64 s[18:19] +; SDAG-NEXT: s_add_u32 s18, s18, ret_i32_noinline@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s19, s19, ret_i32_noinline@rel32@hi+12 +; SDAG-NEXT: v_lshlrev_b32_e32 v2, 20, v2 +; SDAG-NEXT: v_lshlrev_b32_e32 v1, 10, v1 +; SDAG-NEXT: v_or_b32_e32 v0, v0, v1 +; SDAG-NEXT: v_or_b32_e32 v31, v0, v2 +; SDAG-NEXT: s_mov_b32 s14, s16 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[18:19] +; SDAG-NEXT: v_add_f32_e32 v0, 1.0, v0 +; SDAG-NEXT: flat_store_dword v[0:1], v0 +; SDAG-NEXT: s_waitcnt vmcnt(0) +; SDAG-NEXT: s_endpgm +; +; GISEL-LABEL: test_bitcast_return_type_noinline: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13 +; GISEL-NEXT: s_add_i32 s12, s12, s17 +; GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; GISEL-NEXT: s_add_u32 s0, s0, s17 +; GISEL-NEXT: s_addc_u32 s1, s1, 0 +; GISEL-NEXT: s_mov_b32 s13, s15 +; GISEL-NEXT: s_mov_b32 s12, s14 +; GISEL-NEXT: v_lshlrev_b32_e32 v1, 10, v1 +; GISEL-NEXT: v_lshlrev_b32_e32 v2, 20, v2 +; GISEL-NEXT: v_or_b32_e32 v0, v0, v1 +; GISEL-NEXT: v_or_b32_e32 v31, v0, v2 +; GISEL-NEXT: s_getpc_b64 s[18:19] +; GISEL-NEXT: s_add_u32 s18, s18, ret_i32_noinline@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s19, s19, ret_i32_noinline@rel32@hi+12 +; GISEL-NEXT: s_mov_b32 s14, s16 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[18:19] +; GISEL-NEXT: v_add_f32_e32 v0, 1.0, v0 +; GISEL-NEXT: flat_store_dword v[0:1], v0 +; GISEL-NEXT: s_waitcnt vmcnt(0) +; GISEL-NEXT: s_endpgm %val = call float @ret_i32_noinline() %op = fadd float %val, 1.0 store volatile float %op, ptr addrspace(1) poison ret void } -; GCN-LABEL: {{^}}test_bitcast_return_type_alwaysinline: -; GCN: s_swappc_b64 define amdgpu_kernel void @test_bitcast_return_type_alwaysinline() #0 { +; SDAG-LABEL: test_bitcast_return_type_alwaysinline: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 +; SDAG-NEXT: s_add_i32 s12, s12, s17 +; SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; SDAG-NEXT: s_add_u32 s0, s0, s17 +; SDAG-NEXT: s_addc_u32 s1, s1, 0 +; SDAG-NEXT: s_mov_b32 s13, s15 +; SDAG-NEXT: s_mov_b32 s12, s14 +; SDAG-NEXT: s_getpc_b64 s[18:19] +; SDAG-NEXT: s_add_u32 s18, s18, ret_i32_alwaysinline@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s19, s19, ret_i32_alwaysinline@rel32@hi+12 +; SDAG-NEXT: v_lshlrev_b32_e32 v2, 20, v2 +; SDAG-NEXT: v_lshlrev_b32_e32 v1, 10, v1 +; SDAG-NEXT: v_or_b32_e32 v0, v0, v1 +; SDAG-NEXT: v_or_b32_e32 v31, v0, v2 +; SDAG-NEXT: s_mov_b32 s14, s16 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[18:19] +; SDAG-NEXT: v_add_f32_e32 v0, 1.0, v0 +; SDAG-NEXT: flat_store_dword v[0:1], v0 +; SDAG-NEXT: s_waitcnt vmcnt(0) +; SDAG-NEXT: s_endpgm +; +; GISEL-LABEL: test_bitcast_return_type_alwaysinline: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13 +; GISEL-NEXT: s_add_i32 s12, s12, s17 +; GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; GISEL-NEXT: s_add_u32 s0, s0, s17 +; GISEL-NEXT: s_addc_u32 s1, s1, 0 +; GISEL-NEXT: s_mov_b32 s13, s15 +; GISEL-NEXT: s_mov_b32 s12, s14 +; GISEL-NEXT: v_lshlrev_b32_e32 v1, 10, v1 +; GISEL-NEXT: v_lshlrev_b32_e32 v2, 20, v2 +; GISEL-NEXT: v_or_b32_e32 v0, v0, v1 +; GISEL-NEXT: v_or_b32_e32 v31, v0, v2 +; GISEL-NEXT: s_getpc_b64 s[18:19] +; GISEL-NEXT: s_add_u32 s18, s18, ret_i32_alwaysinline@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s19, s19, ret_i32_alwaysinline@rel32@hi+12 +; GISEL-NEXT: s_mov_b32 s14, s16 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[18:19] +; GISEL-NEXT: v_add_f32_e32 v0, 1.0, v0 +; GISEL-NEXT: flat_store_dword v[0:1], v0 +; GISEL-NEXT: s_waitcnt vmcnt(0) +; GISEL-NEXT: s_endpgm %val = call float @ret_i32_alwaysinline() %op = fadd float %val, 1.0 store volatile float %op, ptr addrspace(1) poison ret void } -; GCN-LABEL: {{^}}test_bitcast_argument_type: -; GCN: s_getpc_b64 -; GCN: s_add_u32 s{{[0-9]+}}, s{{[0-9]+}}, ident_i32@rel32@lo+4 -; GCN: s_addc_u32 s{{[0-9]+}}, s{{[0-9]+}}, ident_i32@rel32@hi+12 -; GCN: s_swappc_b64 define amdgpu_kernel void @test_bitcast_argument_type() #0 { +; SDAG-LABEL: test_bitcast_argument_type: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 +; SDAG-NEXT: s_add_i32 s12, s12, s17 +; SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; SDAG-NEXT: s_add_u32 s0, s0, s17 +; SDAG-NEXT: s_addc_u32 s1, s1, 0 +; SDAG-NEXT: s_mov_b32 s13, s15 +; SDAG-NEXT: s_mov_b32 s12, s14 +; SDAG-NEXT: s_getpc_b64 s[18:19] +; SDAG-NEXT: s_add_u32 s18, s18, ident_i32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s19, s19, ident_i32@rel32@hi+12 +; SDAG-NEXT: v_lshlrev_b32_e32 v2, 20, v2 +; SDAG-NEXT: v_lshlrev_b32_e32 v1, 10, v1 +; SDAG-NEXT: v_or_b32_e32 v0, v0, v1 +; SDAG-NEXT: v_or_b32_e32 v31, v0, v2 +; SDAG-NEXT: v_mov_b32_e32 v0, 2.0 +; SDAG-NEXT: s_mov_b32 s14, s16 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[18:19] +; SDAG-NEXT: v_add_i32_e32 v0, vcc, 1, v0 +; SDAG-NEXT: flat_store_dword v[0:1], v0 +; SDAG-NEXT: s_waitcnt vmcnt(0) +; SDAG-NEXT: s_endpgm +; +; GISEL-LABEL: test_bitcast_argument_type: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13 +; GISEL-NEXT: s_add_i32 s12, s12, s17 +; GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; GISEL-NEXT: s_add_u32 s0, s0, s17 +; GISEL-NEXT: s_addc_u32 s1, s1, 0 +; GISEL-NEXT: s_mov_b32 s13, s15 +; GISEL-NEXT: s_mov_b32 s12, s14 +; GISEL-NEXT: v_lshlrev_b32_e32 v1, 10, v1 +; GISEL-NEXT: v_lshlrev_b32_e32 v2, 20, v2 +; GISEL-NEXT: s_getpc_b64 s[18:19] +; GISEL-NEXT: s_add_u32 s18, s18, ident_i32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s19, s19, ident_i32@rel32@hi+12 +; GISEL-NEXT: v_or_b32_e32 v0, v0, v1 +; GISEL-NEXT: v_or_b32_e32 v31, v0, v2 +; GISEL-NEXT: v_mov_b32_e32 v0, 2.0 +; GISEL-NEXT: s_mov_b32 s14, s16 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[18:19] +; GISEL-NEXT: v_add_i32_e32 v0, vcc, 1, v0 +; GISEL-NEXT: flat_store_dword v[0:1], v0 +; GISEL-NEXT: s_waitcnt vmcnt(0) +; GISEL-NEXT: s_endpgm %val = call i32 @ident_i32(float 2.0) %op = add i32 %val, 1 store volatile i32 %op, ptr addrspace(1) poison ret void } -; GCN-LABEL: {{^}}test_bitcast_argument_and_return_types: -; GCN: s_getpc_b64 -; GCN: s_add_u32 s{{[0-9]+}}, s{{[0-9]+}}, ident_i32@rel32@lo+4 -; GCN: s_addc_u32 s{{[0-9]+}}, s{{[0-9]+}}, ident_i32@rel32@hi+12 -; GCN: s_swappc_b64 define amdgpu_kernel void @test_bitcast_argument_and_return_types() #0 { +; SDAG-LABEL: test_bitcast_argument_and_return_types: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 +; SDAG-NEXT: s_add_i32 s12, s12, s17 +; SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; SDAG-NEXT: s_add_u32 s0, s0, s17 +; SDAG-NEXT: s_addc_u32 s1, s1, 0 +; SDAG-NEXT: s_mov_b32 s13, s15 +; SDAG-NEXT: s_mov_b32 s12, s14 +; SDAG-NEXT: s_getpc_b64 s[18:19] +; SDAG-NEXT: s_add_u32 s18, s18, ident_i32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s19, s19, ident_i32@rel32@hi+12 +; SDAG-NEXT: v_lshlrev_b32_e32 v2, 20, v2 +; SDAG-NEXT: v_lshlrev_b32_e32 v1, 10, v1 +; SDAG-NEXT: v_or_b32_e32 v0, v0, v1 +; SDAG-NEXT: v_or_b32_e32 v31, v0, v2 +; SDAG-NEXT: v_mov_b32_e32 v0, 2.0 +; SDAG-NEXT: s_mov_b32 s14, s16 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[18:19] +; SDAG-NEXT: v_add_f32_e32 v0, 1.0, v0 +; SDAG-NEXT: flat_store_dword v[0:1], v0 +; SDAG-NEXT: s_waitcnt vmcnt(0) +; SDAG-NEXT: s_endpgm +; +; GISEL-LABEL: test_bitcast_argument_and_return_types: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13 +; GISEL-NEXT: s_add_i32 s12, s12, s17 +; GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; GISEL-NEXT: s_add_u32 s0, s0, s17 +; GISEL-NEXT: s_addc_u32 s1, s1, 0 +; GISEL-NEXT: s_mov_b32 s13, s15 +; GISEL-NEXT: s_mov_b32 s12, s14 +; GISEL-NEXT: v_lshlrev_b32_e32 v1, 10, v1 +; GISEL-NEXT: v_lshlrev_b32_e32 v2, 20, v2 +; GISEL-NEXT: s_getpc_b64 s[18:19] +; GISEL-NEXT: s_add_u32 s18, s18, ident_i32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s19, s19, ident_i32@rel32@hi+12 +; GISEL-NEXT: v_or_b32_e32 v0, v0, v1 +; GISEL-NEXT: v_or_b32_e32 v31, v0, v2 +; GISEL-NEXT: v_mov_b32_e32 v0, 2.0 +; GISEL-NEXT: s_mov_b32 s14, s16 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[18:19] +; GISEL-NEXT: v_add_f32_e32 v0, 1.0, v0 +; GISEL-NEXT: flat_store_dword v[0:1], v0 +; GISEL-NEXT: s_waitcnt vmcnt(0) +; GISEL-NEXT: s_endpgm %val = call float @ident_i32(float 2.0) %op = fadd float %val, 1.0 store volatile float %op, ptr addrspace(1) poison ret void } -; GCN-LABEL: {{^}}use_workitem_id_x: -; GCN: s_waitcnt -; GCN-NEXT: v_and_b32_e32 [[TMP:v[0-9]+]], 0x3ff, v31 -; GCN-NEXT: v_add_i32_e32 v0, vcc, [[TMP]], v0 -; GCN-NEXT: s_setpc_b64 define hidden i32 @use_workitem_id_x(i32 %arg0) #3 { +; GCN-LABEL: use_workitem_id_x: +; GCN: ; %bb.0: +; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GCN-NEXT: v_and_b32_e32 v1, 0x3ff, v31 +; GCN-NEXT: v_add_i32_e32 v0, vcc, v1, v0 +; GCN-NEXT: s_setpc_b64 s[30:31] %id = call i32 @llvm.amdgcn.workitem.id.x() %op = add i32 %id, %arg0 ret i32 %op } -; GCN-LABEL: {{^}}test_bitcast_use_workitem_id_x: -; GCN: v_mov_b32_e32 v31, v0 -; GCN: s_getpc_b64 -; GCN: s_add_u32 s{{[0-9]+}}, s{{[0-9]+}}, use_workitem_id_x@rel32@lo+4 -; GCN: s_addc_u32 s{{[0-9]+}}, s{{[0-9]+}}, use_workitem_id_x@rel32@hi+12 -; GCN: v_mov_b32_e32 v0, 9 -; GCN: s_swappc_b64 -; GCN: v_add_f32_e32 define amdgpu_kernel void @test_bitcast_use_workitem_id_x() #3 { +; SDAG-LABEL: test_bitcast_use_workitem_id_x: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 +; SDAG-NEXT: s_add_i32 s12, s12, s17 +; SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; SDAG-NEXT: s_add_u32 s0, s0, s17 +; SDAG-NEXT: s_addc_u32 s1, s1, 0 +; SDAG-NEXT: s_mov_b32 s13, s15 +; SDAG-NEXT: s_mov_b32 s12, s14 +; SDAG-NEXT: v_mov_b32_e32 v31, v0 +; SDAG-NEXT: s_getpc_b64 s[18:19] +; SDAG-NEXT: s_add_u32 s18, s18, use_workitem_id_x@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s19, s19, use_workitem_id_x@rel32@hi+12 +; SDAG-NEXT: v_mov_b32_e32 v0, 9 +; SDAG-NEXT: s_mov_b32 s14, s16 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[18:19] +; SDAG-NEXT: v_add_f32_e32 v0, 1.0, v0 +; SDAG-NEXT: flat_store_dword v[0:1], v0 +; SDAG-NEXT: s_waitcnt vmcnt(0) +; SDAG-NEXT: s_endpgm +; +; GISEL-LABEL: test_bitcast_use_workitem_id_x: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13 +; GISEL-NEXT: s_add_i32 s12, s12, s17 +; GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; GISEL-NEXT: s_add_u32 s0, s0, s17 +; GISEL-NEXT: s_addc_u32 s1, s1, 0 +; GISEL-NEXT: v_mov_b32_e32 v31, v0 +; GISEL-NEXT: s_mov_b32 s13, s15 +; GISEL-NEXT: s_mov_b32 s12, s14 +; GISEL-NEXT: s_getpc_b64 s[18:19] +; GISEL-NEXT: s_add_u32 s18, s18, use_workitem_id_x@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s19, s19, use_workitem_id_x@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 9 +; GISEL-NEXT: s_mov_b32 s14, s16 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[18:19] +; GISEL-NEXT: v_add_f32_e32 v0, 1.0, v0 +; GISEL-NEXT: flat_store_dword v[0:1], v0 +; GISEL-NEXT: s_waitcnt vmcnt(0) +; GISEL-NEXT: s_endpgm %val = call float @use_workitem_id_x(i32 9) %op = fadd float %val, 1.0 store volatile float %op, ptr addrspace(1) poison ret void } -; GCN-LABEL: {{^}}test_invoke: -; GCN: s_getpc_b64 -; GCN: s_add_u32 s{{[0-9]+}}, s{{[0-9]+}}, ident_i32@rel32@lo+4 -; GCN: s_addc_u32 s{{[0-9]+}}, s{{[0-9]+}}, ident_i32@rel32@hi+12 -; GCN: s_swappc_b64 @_ZTIi = external global ptr declare i32 @__gxx_personality_v0(...) define amdgpu_kernel void @test_invoke() #0 personality ptr @__gxx_personality_v0 { +; SDAG-LABEL: test_invoke: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 +; SDAG-NEXT: s_add_i32 s12, s12, s17 +; SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; SDAG-NEXT: s_add_u32 s0, s0, s17 +; SDAG-NEXT: s_addc_u32 s1, s1, 0 +; SDAG-NEXT: s_mov_b32 s13, s15 +; SDAG-NEXT: s_mov_b32 s12, s14 +; SDAG-NEXT: s_getpc_b64 s[18:19] +; SDAG-NEXT: s_add_u32 s18, s18, ident_i32@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s19, s19, ident_i32@rel32@hi+12 +; SDAG-NEXT: v_lshlrev_b32_e32 v2, 20, v2 +; SDAG-NEXT: v_lshlrev_b32_e32 v1, 10, v1 +; SDAG-NEXT: v_or_b32_e32 v0, v0, v1 +; SDAG-NEXT: v_or_b32_e32 v31, v0, v2 +; SDAG-NEXT: v_mov_b32_e32 v0, 2.0 +; SDAG-NEXT: s_mov_b32 s14, s16 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[18:19] +; SDAG-NEXT: v_add_f32_e32 v0, 1.0, v0 +; SDAG-NEXT: flat_store_dword v[0:1], v0 +; SDAG-NEXT: s_waitcnt vmcnt(0) +; SDAG-NEXT: s_endpgm +; +; GISEL-LABEL: test_invoke: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13 +; GISEL-NEXT: s_add_i32 s12, s12, s17 +; GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; GISEL-NEXT: s_add_u32 s0, s0, s17 +; GISEL-NEXT: s_addc_u32 s1, s1, 0 +; GISEL-NEXT: s_mov_b32 s13, s15 +; GISEL-NEXT: s_mov_b32 s12, s14 +; GISEL-NEXT: v_lshlrev_b32_e32 v1, 10, v1 +; GISEL-NEXT: v_lshlrev_b32_e32 v2, 20, v2 +; GISEL-NEXT: s_getpc_b64 s[18:19] +; GISEL-NEXT: s_add_u32 s18, s18, ident_i32@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s19, s19, ident_i32@rel32@hi+12 +; GISEL-NEXT: v_or_b32_e32 v0, v0, v1 +; GISEL-NEXT: v_or_b32_e32 v31, v0, v2 +; GISEL-NEXT: v_mov_b32_e32 v0, 2.0 +; GISEL-NEXT: s_mov_b32 s14, s16 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[18:19] +; GISEL-NEXT: v_add_f32_e32 v0, 1.0, v0 +; GISEL-NEXT: flat_store_dword v[0:1], v0 +; GISEL-NEXT: s_waitcnt vmcnt(0) +; GISEL-NEXT: s_endpgm %val = invoke float @ident_i32(float 2.0) to label %continue unwind label %broken @@ -96,14 +353,28 @@ continue: ; arguments before we lower any calls to them. define hidden i32 @ret_i32_noinline() #0 { +; GCN-LABEL: ret_i32_noinline: +; GCN: ; %bb.0: +; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GCN-NEXT: v_mov_b32_e32 v0, 4 +; GCN-NEXT: s_setpc_b64 s[30:31] ret i32 4 } define hidden i32 @ret_i32_alwaysinline() #1 { +; GCN-LABEL: ret_i32_alwaysinline: +; GCN: ; %bb.0: +; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GCN-NEXT: v_mov_b32_e32 v0, 4 +; GCN-NEXT: s_setpc_b64 s[30:31] ret i32 4 } define hidden i32 @ident_i32(i32 %i) #0 { +; GCN-LABEL: ident_i32: +; GCN: ; %bb.0: +; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GCN-NEXT: s_setpc_b64 s[30:31] ret i32 %i } diff --git a/llvm/test/CodeGen/AMDGPU/call-defs-mode-register.ll b/llvm/test/CodeGen/AMDGPU/call-defs-mode-register.ll index ffe536d..4b5a49f 100644 --- a/llvm/test/CodeGen/AMDGPU/call-defs-mode-register.ll +++ b/llvm/test/CodeGen/AMDGPU/call-defs-mode-register.ll @@ -1,5 +1,6 @@ ; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 5 -; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -simplify-mir -stop-after=finalize-isel < %s | FileCheck %s +; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -simplify-mir -stop-after=finalize-isel < %s | FileCheck -check-prefixes=SDAG %s +; RUN: llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -simplify-mir -stop-after=finalize-isel < %s | FileCheck -check-prefixes=GISEL %s ; Check that call / asm get an implicit-def $mode added to them in ; strictfp functions. @@ -7,46 +8,80 @@ declare protected void @maybe_defs_mode() #0 define float @call_changes_mode(float %x, float %y) #0 { - ; CHECK-LABEL: name: call_changes_mode - ; CHECK: bb.0 (%ir-block.0): - ; CHECK-NEXT: liveins: $vgpr0, $vgpr1 - ; CHECK-NEXT: {{ $}} - ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr1 - ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0 - ; CHECK-NEXT: ADJCALLSTACKUP 0, 0, implicit-def dead $scc, implicit-def $sgpr32, implicit $sgpr32 - ; CHECK-NEXT: [[SI_PC_ADD_REL_OFFSET:%[0-9]+]]:sreg_64 = SI_PC_ADD_REL_OFFSET target-flags(amdgpu-rel32-lo) @maybe_defs_mode, target-flags(amdgpu-rel32-hi) @maybe_defs_mode, implicit-def dead $scc - ; CHECK-NEXT: [[COPY2:%[0-9]+]]:sgpr_128 = COPY $sgpr0_sgpr1_sgpr2_sgpr3 - ; CHECK-NEXT: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY2]] - ; CHECK-NEXT: $sgpr30_sgpr31 = SI_CALL killed [[SI_PC_ADD_REL_OFFSET]], @maybe_defs_mode, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $mode - ; CHECK-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def dead $scc, implicit-def $sgpr32, implicit $sgpr32 - ; CHECK-NEXT: [[V_ADD_F32_e64_:%[0-9]+]]:vgpr_32 = nofpexcept V_ADD_F32_e64 0, [[COPY1]], 0, [[COPY]], 0, 0, implicit $mode, implicit $exec - ; CHECK-NEXT: $vgpr0 = COPY [[V_ADD_F32_e64_]] - ; CHECK-NEXT: SI_RETURN implicit $vgpr0 + ; SDAG-LABEL: name: call_changes_mode + ; SDAG: bb.0 (%ir-block.0): + ; SDAG-NEXT: liveins: $vgpr0, $vgpr1 + ; SDAG-NEXT: {{ $}} + ; SDAG-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr1 + ; SDAG-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; SDAG-NEXT: ADJCALLSTACKUP 0, 0, implicit-def dead $scc, implicit-def $sgpr32, implicit $sgpr32 + ; SDAG-NEXT: [[SI_PC_ADD_REL_OFFSET:%[0-9]+]]:sreg_64 = SI_PC_ADD_REL_OFFSET target-flags(amdgpu-rel32-lo) @maybe_defs_mode, target-flags(amdgpu-rel32-hi) @maybe_defs_mode, implicit-def dead $scc + ; SDAG-NEXT: [[COPY2:%[0-9]+]]:sgpr_128 = COPY $sgpr0_sgpr1_sgpr2_sgpr3 + ; SDAG-NEXT: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY2]] + ; SDAG-NEXT: $sgpr30_sgpr31 = SI_CALL killed [[SI_PC_ADD_REL_OFFSET]], @maybe_defs_mode, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3, implicit-def $mode + ; SDAG-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def dead $scc, implicit-def $sgpr32, implicit $sgpr32 + ; SDAG-NEXT: [[V_ADD_F32_e64_:%[0-9]+]]:vgpr_32 = nofpexcept V_ADD_F32_e64 0, [[COPY1]], 0, [[COPY]], 0, 0, implicit $mode, implicit $exec + ; SDAG-NEXT: $vgpr0 = COPY [[V_ADD_F32_e64_]] + ; SDAG-NEXT: SI_RETURN implicit $vgpr0 + ; + ; GISEL-LABEL: name: call_changes_mode + ; GISEL: bb.1 (%ir-block.0): + ; GISEL-NEXT: liveins: $vgpr0, $vgpr1 + ; GISEL-NEXT: {{ $}} + ; GISEL-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GISEL-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1 + ; GISEL-NEXT: ADJCALLSTACKUP 0, 0, implicit-def $scc, implicit-def $sgpr32, implicit $sgpr32 + ; GISEL-NEXT: [[COPY2:%[0-9]+]]:sgpr_128 = COPY $sgpr0_sgpr1_sgpr2_sgpr3 + ; GISEL-NEXT: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY2]] + ; GISEL-NEXT: [[SI_PC_ADD_REL_OFFSET:%[0-9]+]]:sreg_64 = SI_PC_ADD_REL_OFFSET target-flags(amdgpu-rel32-lo) @maybe_defs_mode, target-flags(amdgpu-rel32-hi) @maybe_defs_mode, implicit-def $scc + ; GISEL-NEXT: $sgpr30_sgpr31 = noconvergent SI_CALL [[SI_PC_ADD_REL_OFFSET]], @maybe_defs_mode, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3 + ; GISEL-NEXT: ADJCALLSTACKDOWN 0, 0, implicit-def $scc, implicit-def $sgpr32, implicit $sgpr32 + ; GISEL-NEXT: [[V_ADD_F32_e64_:%[0-9]+]]:vgpr_32 = nofpexcept V_ADD_F32_e64 0, [[COPY]], 0, [[COPY1]], 0, 0, implicit $mode, implicit $exec + ; GISEL-NEXT: $vgpr0 = COPY [[V_ADD_F32_e64_]] + ; GISEL-NEXT: SI_RETURN implicit $vgpr0 call void @maybe_defs_mode() %val = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.dynamic", metadata !"fpexcept.ignore") ret float %val } define void @tail_call_changes_mode() #0 { - ; CHECK-LABEL: name: tail_call_changes_mode - ; CHECK: bb.0 (%ir-block.0): - ; CHECK-NEXT: [[SI_PC_ADD_REL_OFFSET:%[0-9]+]]:ccr_sgpr_64 = SI_PC_ADD_REL_OFFSET target-flags(amdgpu-rel32-lo) @maybe_defs_mode, target-flags(amdgpu-rel32-hi) @maybe_defs_mode, implicit-def dead $scc - ; CHECK-NEXT: SI_TCRETURN killed [[SI_PC_ADD_REL_OFFSET]], @maybe_defs_mode, 0, csr_amdgpu, implicit-def $mode + ; SDAG-LABEL: name: tail_call_changes_mode + ; SDAG: bb.0 (%ir-block.0): + ; SDAG-NEXT: [[SI_PC_ADD_REL_OFFSET:%[0-9]+]]:ccr_sgpr_64 = SI_PC_ADD_REL_OFFSET target-flags(amdgpu-rel32-lo) @maybe_defs_mode, target-flags(amdgpu-rel32-hi) @maybe_defs_mode, implicit-def dead $scc + ; SDAG-NEXT: SI_TCRETURN killed [[SI_PC_ADD_REL_OFFSET]], @maybe_defs_mode, 0, csr_amdgpu, implicit-def $mode + ; + ; GISEL-LABEL: name: tail_call_changes_mode + ; GISEL: bb.1 (%ir-block.0): + ; GISEL-NEXT: [[COPY:%[0-9]+]]:sgpr_128 = COPY $sgpr0_sgpr1_sgpr2_sgpr3 + ; GISEL-NEXT: $sgpr0_sgpr1_sgpr2_sgpr3 = COPY [[COPY]] + ; GISEL-NEXT: [[SI_PC_ADD_REL_OFFSET:%[0-9]+]]:ccr_sgpr_64 = SI_PC_ADD_REL_OFFSET target-flags(amdgpu-rel32-lo) @maybe_defs_mode, target-flags(amdgpu-rel32-hi) @maybe_defs_mode, implicit-def $scc + ; GISEL-NEXT: SI_TCRETURN [[SI_PC_ADD_REL_OFFSET]], @maybe_defs_mode, 0, csr_amdgpu, implicit $sgpr0_sgpr1_sgpr2_sgpr3 tail call void @maybe_defs_mode() ret void } define float @asm_changes_mode(float %x, float %y) #0 { - ; CHECK-LABEL: name: asm_changes_mode - ; CHECK: bb.0 (%ir-block.0): - ; CHECK-NEXT: liveins: $vgpr0, $vgpr1 - ; CHECK-NEXT: {{ $}} - ; CHECK-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr1 - ; CHECK-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0 - ; CHECK-NEXT: INLINEASM &"; maybe defs mode", 1 /* sideeffect attdialect */, implicit-def $mode - ; CHECK-NEXT: [[V_ADD_F32_e64_:%[0-9]+]]:vgpr_32 = nofpexcept V_ADD_F32_e64 0, [[COPY1]], 0, [[COPY]], 0, 0, implicit $mode, implicit $exec - ; CHECK-NEXT: $vgpr0 = COPY [[V_ADD_F32_e64_]] - ; CHECK-NEXT: SI_RETURN implicit $vgpr0 + ; SDAG-LABEL: name: asm_changes_mode + ; SDAG: bb.0 (%ir-block.0): + ; SDAG-NEXT: liveins: $vgpr0, $vgpr1 + ; SDAG-NEXT: {{ $}} + ; SDAG-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr1 + ; SDAG-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; SDAG-NEXT: INLINEASM &"; maybe defs mode", 1 /* sideeffect attdialect */, implicit-def $mode + ; SDAG-NEXT: [[V_ADD_F32_e64_:%[0-9]+]]:vgpr_32 = nofpexcept V_ADD_F32_e64 0, [[COPY1]], 0, [[COPY]], 0, 0, implicit $mode, implicit $exec + ; SDAG-NEXT: $vgpr0 = COPY [[V_ADD_F32_e64_]] + ; SDAG-NEXT: SI_RETURN implicit $vgpr0 + ; + ; GISEL-LABEL: name: asm_changes_mode + ; GISEL: bb.1 (%ir-block.0): + ; GISEL-NEXT: liveins: $vgpr0, $vgpr1 + ; GISEL-NEXT: {{ $}} + ; GISEL-NEXT: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; GISEL-NEXT: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr1 + ; GISEL-NEXT: INLINEASM &"; maybe defs mode", 1 /* sideeffect attdialect */, implicit-def $mode + ; GISEL-NEXT: [[V_ADD_F32_e64_:%[0-9]+]]:vgpr_32 = nofpexcept V_ADD_F32_e64 0, [[COPY]], 0, [[COPY1]], 0, 0, implicit $mode, implicit $exec + ; GISEL-NEXT: $vgpr0 = COPY [[V_ADD_F32_e64_]] + ; GISEL-NEXT: SI_RETURN implicit $vgpr0 call void asm sideeffect "; maybe defs mode", ""() %val = call float @llvm.experimental.constrained.fadd.f32(float %x, float %y, metadata !"round.dynamic", metadata !"fpexcept.ignore") ret float %val diff --git a/llvm/test/CodeGen/AMDGPU/call-encoding.ll b/llvm/test/CodeGen/AMDGPU/call-encoding.ll index 6954c34..6c36c242 100644 --- a/llvm/test/CodeGen/AMDGPU/call-encoding.ll +++ b/llvm/test/CodeGen/AMDGPU/call-encoding.ll @@ -1,5 +1,7 @@ -; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -filetype=obj < %s | llvm-objdump --triple=amdgcn--amdhsa --mcpu=fiji -d - | FileCheck --check-prefix=GCN %s -; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -filetype=obj < %s | llvm-objdump --triple=amdgcn--amdhsa --mcpu=gfx900 -d - | FileCheck --check-prefix=GCN %s +; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -filetype=obj < %s | llvm-objdump --triple=amdgcn--amdhsa --mcpu=fiji -d - | FileCheck --check-prefix=GCN %s +; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -filetype=obj < %s | llvm-objdump --triple=amdgcn--amdhsa --mcpu=gfx900 -d - | FileCheck --check-prefix=GCN %s +; RUN: llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -filetype=obj < %s | llvm-objdump --triple=amdgcn--amdhsa --mcpu=fiji -d - | FileCheck --check-prefix=GCN %s +; RUN: llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -filetype=obj < %s | llvm-objdump --triple=amdgcn--amdhsa --mcpu=gfx900 -d - | FileCheck --check-prefix=GCN %s ; XUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=hawaii -filetype=obj < %s | llvm-objdump --triple=amdgcn--amdhsa --mcpu=hawaii -d - | FileCheck --check-prefixes=GCN,CI %s ; GCN: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) diff --git a/llvm/test/CodeGen/AMDGPU/call-graph-register-usage.ll b/llvm/test/CodeGen/AMDGPU/call-graph-register-usage.ll index 4df1049..b250227 100644 --- a/llvm/test/CodeGen/AMDGPU/call-graph-register-usage.ll +++ b/llvm/test/CodeGen/AMDGPU/call-graph-register-usage.ll @@ -1,8 +1,13 @@ -; RUN: sed 's/CODE_OBJECT_VERSION/400/g' %s | llc -mtriple=amdgcn-amd-amdhsa -enable-ipra=0 | FileCheck -check-prefixes=GCN,CI %s -; RUN: sed 's/CODE_OBJECT_VERSION/500/g' %s | llc -mtriple=amdgcn-amd-amdhsa -enable-ipra=0 | FileCheck -check-prefixes=GCN-V5 %s -; RUN: sed 's/CODE_OBJECT_VERSION/600/g' %s | llc -mtriple=amdgcn-amd-amdhsa -enable-ipra=0 | FileCheck -check-prefixes=GCN-V5 %s -; RUN: sed 's/CODE_OBJECT_VERSION/400/g' %s | llc -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -enable-ipra=0 | FileCheck -check-prefixes=GCN,VI,VI-NOBUG %s -; RUN: sed 's/CODE_OBJECT_VERSION/400/g' %s | llc -mtriple=amdgcn-amd-amdhsa -mcpu=iceland -enable-ipra=0 | FileCheck -check-prefixes=GCN,VI,VI-BUG %s +; RUN: sed 's/CODE_OBJECT_VERSION/400/g' %s | llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -enable-ipra=0 | FileCheck -check-prefixes=GCN,CI %s +; RUN: sed 's/CODE_OBJECT_VERSION/500/g' %s | llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -enable-ipra=0 | FileCheck -check-prefixes=GCN-V5 %s +; RUN: sed 's/CODE_OBJECT_VERSION/600/g' %s | llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -enable-ipra=0 | FileCheck -check-prefixes=GCN-V5 %s +; RUN: sed 's/CODE_OBJECT_VERSION/400/g' %s | llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -enable-ipra=0 | FileCheck -check-prefixes=GCN,VI,VI-NOBUG %s +; RUN: sed 's/CODE_OBJECT_VERSION/400/g' %s | llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=iceland -enable-ipra=0 | FileCheck -check-prefixes=GCN,VI,VI-BUG %s +; RUN: sed 's/CODE_OBJECT_VERSION/400/g' %s | llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -enable-ipra=0 | FileCheck -check-prefixes=GCN,CI %s +; RUN: sed 's/CODE_OBJECT_VERSION/500/g' %s | llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -enable-ipra=0 | FileCheck -check-prefixes=GCN-V5 %s +; RUN: sed 's/CODE_OBJECT_VERSION/600/g' %s | llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -enable-ipra=0 | FileCheck -check-prefixes=GCN-V5 %s +; RUN: sed 's/CODE_OBJECT_VERSION/400/g' %s | llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -enable-ipra=0 | FileCheck -check-prefixes=GCN,VI,VI-NOBUG %s +; RUN: sed 's/CODE_OBJECT_VERSION/400/g' %s | llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -mcpu=iceland -enable-ipra=0 | FileCheck -check-prefixes=GCN,VI,VI-BUG %s ; Make sure to run a GPU with the SGPR allocation bug. diff --git a/llvm/test/CodeGen/AMDGPU/call-preserved-registers.ll b/llvm/test/CodeGen/AMDGPU/call-preserved-registers.ll index 61a195f..aed1079 100644 --- a/llvm/test/CodeGen/AMDGPU/call-preserved-registers.ll +++ b/llvm/test/CodeGen/AMDGPU/call-preserved-registers.ll @@ -1,8 +1,12 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 -; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -enable-ipra=0 < %s | FileCheck -check-prefixes=GCN,MUBUF %s -; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=hawaii -enable-ipra=0 < %s | FileCheck -check-prefixes=GCN,MUBUF %s -; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -enable-ipra=0 < %s | FileCheck -check-prefixes=GCN,MUBUF %s -; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -enable-ipra=0 -mattr=+enable-flat-scratch < %s | FileCheck -check-prefixes=GCN,FLATSCR %s +; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -enable-ipra=0 < %s | FileCheck -check-prefixes=GCN,MUBUF,SDAG %s +; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=hawaii -enable-ipra=0 < %s | FileCheck -check-prefixes=GCN,MUBUF,SDAG %s +; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -enable-ipra=0 < %s | FileCheck -check-prefixes=GCN,MUBUF,SDAG %s +; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -enable-ipra=0 -mattr=+enable-flat-scratch < %s | FileCheck -check-prefixes=GCN,FLATSCR,SDAG %s +; RUN: llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -enable-ipra=0 < %s | FileCheck -check-prefixes=GCN,MUBUF,GISEL %s +; RUN: llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -mcpu=hawaii -enable-ipra=0 < %s | FileCheck -check-prefixes=GCN,MUBUF,GISEL %s +; RUN: llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -enable-ipra=0 < %s | FileCheck -check-prefixes=GCN,MUBUF,GISEL %s +; RUN: llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -enable-ipra=0 -mattr=+enable-flat-scratch < %s | FileCheck -check-prefixes=GCN,FLATSCR,GISEL %s declare hidden void @external_void_func_void() #3 @@ -223,41 +227,6 @@ define hidden void @void_func_void_clobber_vcc() #2 { } define amdgpu_kernel void @test_call_void_func_void_clobber_vcc(ptr addrspace(1) %out) #0 { -; FLATSCR-LABEL: test_call_void_func_void_clobber_vcc: -; FLATSCR: ; %bb.0: -; FLATSCR-NEXT: s_add_u32 flat_scratch_lo, s8, s13 -; FLATSCR-NEXT: s_addc_u32 flat_scratch_hi, s9, 0 -; FLATSCR-NEXT: s_add_u32 s8, s4, 8 -; FLATSCR-NEXT: s_addc_u32 s9, s5, 0 -; FLATSCR-NEXT: v_lshlrev_b32_e32 v2, 20, v2 -; FLATSCR-NEXT: v_lshlrev_b32_e32 v1, 10, v1 -; FLATSCR-NEXT: s_mov_b32 s14, s12 -; FLATSCR-NEXT: s_mov_b32 s13, s11 -; FLATSCR-NEXT: s_mov_b32 s12, s10 -; FLATSCR-NEXT: s_mov_b64 s[10:11], s[6:7] -; FLATSCR-NEXT: s_getpc_b64 s[16:17] -; FLATSCR-NEXT: s_add_u32 s16, s16, void_func_void_clobber_vcc@rel32@lo+4 -; FLATSCR-NEXT: s_addc_u32 s17, s17, void_func_void_clobber_vcc@rel32@hi+12 -; FLATSCR-NEXT: v_or3_b32 v31, v0, v1, v2 -; FLATSCR-NEXT: s_mov_b64 s[4:5], s[0:1] -; FLATSCR-NEXT: s_mov_b64 s[6:7], s[2:3] -; FLATSCR-NEXT: s_mov_b32 s32, 0 -; FLATSCR-NEXT: ;;#ASMSTART -; FLATSCR-NEXT: ; def vcc -; FLATSCR-NEXT: ;;#ASMEND -; FLATSCR-NEXT: s_mov_b64 s[34:35], vcc -; FLATSCR-NEXT: s_swappc_b64 s[30:31], s[16:17] -; FLATSCR-NEXT: global_load_dword v0, v[0:1], off glc -; FLATSCR-NEXT: s_waitcnt vmcnt(0) -; FLATSCR-NEXT: s_mov_b64 vcc, s[34:35] -; FLATSCR-NEXT: global_load_dword v0, v[0:1], off glc -; FLATSCR-NEXT: s_waitcnt vmcnt(0) -; FLATSCR-NEXT: ; kill: killed $vgpr0_vgpr1 -; FLATSCR-NEXT: ; kill: killed $vgpr0_vgpr1 -; FLATSCR-NEXT: ;;#ASMSTART -; FLATSCR-NEXT: ; use vcc -; FLATSCR-NEXT: ;;#ASMEND -; FLATSCR-NEXT: s_endpgm %vcc = call i64 asm sideeffect "; def $0", "={vcc}"() call void @void_func_void_clobber_vcc() %val0 = load volatile i32, ptr addrspace(1) poison @@ -463,51 +432,11 @@ define hidden void @void_func_void_clobber_s34() #2 { } define amdgpu_kernel void @test_call_void_func_void_clobber_s33() #0 { -; FLATSCR-LABEL: test_call_void_func_void_clobber_s33: -; FLATSCR: ; %bb.0: -; FLATSCR-NEXT: s_add_u32 flat_scratch_lo, s8, s13 -; FLATSCR-NEXT: s_addc_u32 flat_scratch_hi, s9, 0 -; FLATSCR-NEXT: v_lshlrev_b32_e32 v2, 20, v2 -; FLATSCR-NEXT: v_lshlrev_b32_e32 v1, 10, v1 -; FLATSCR-NEXT: s_mov_b32 s14, s12 -; FLATSCR-NEXT: s_mov_b32 s13, s11 -; FLATSCR-NEXT: s_mov_b32 s12, s10 -; FLATSCR-NEXT: s_mov_b64 s[10:11], s[6:7] -; FLATSCR-NEXT: s_mov_b64 s[8:9], s[4:5] -; FLATSCR-NEXT: s_getpc_b64 s[16:17] -; FLATSCR-NEXT: s_add_u32 s16, s16, void_func_void_clobber_s33@rel32@lo+4 -; FLATSCR-NEXT: s_addc_u32 s17, s17, void_func_void_clobber_s33@rel32@hi+12 -; FLATSCR-NEXT: v_or3_b32 v31, v0, v1, v2 -; FLATSCR-NEXT: s_mov_b64 s[4:5], s[0:1] -; FLATSCR-NEXT: s_mov_b64 s[6:7], s[2:3] -; FLATSCR-NEXT: s_mov_b32 s32, 0 -; FLATSCR-NEXT: s_swappc_b64 s[30:31], s[16:17] -; FLATSCR-NEXT: s_endpgm call void @void_func_void_clobber_s33() ret void } define amdgpu_kernel void @test_call_void_func_void_clobber_s34() #0 { -; FLATSCR-LABEL: test_call_void_func_void_clobber_s34: -; FLATSCR: ; %bb.0: -; FLATSCR-NEXT: s_add_u32 flat_scratch_lo, s8, s13 -; FLATSCR-NEXT: s_addc_u32 flat_scratch_hi, s9, 0 -; FLATSCR-NEXT: v_lshlrev_b32_e32 v2, 20, v2 -; FLATSCR-NEXT: v_lshlrev_b32_e32 v1, 10, v1 -; FLATSCR-NEXT: s_mov_b32 s14, s12 -; FLATSCR-NEXT: s_mov_b32 s13, s11 -; FLATSCR-NEXT: s_mov_b32 s12, s10 -; FLATSCR-NEXT: s_mov_b64 s[10:11], s[6:7] -; FLATSCR-NEXT: s_mov_b64 s[8:9], s[4:5] -; FLATSCR-NEXT: s_getpc_b64 s[16:17] -; FLATSCR-NEXT: s_add_u32 s16, s16, void_func_void_clobber_s34@rel32@lo+4 -; FLATSCR-NEXT: s_addc_u32 s17, s17, void_func_void_clobber_s34@rel32@hi+12 -; FLATSCR-NEXT: v_or3_b32 v31, v0, v1, v2 -; FLATSCR-NEXT: s_mov_b64 s[4:5], s[0:1] -; FLATSCR-NEXT: s_mov_b64 s[6:7], s[2:3] -; FLATSCR-NEXT: s_mov_b32 s32, 0 -; FLATSCR-NEXT: s_swappc_b64 s[30:31], s[16:17] -; FLATSCR-NEXT: s_endpgm call void @void_func_void_clobber_s34() ret void } @@ -748,3 +677,6 @@ attributes #0 = { nounwind } attributes #1 = { nounwind readnone } attributes #2 = { nounwind noinline } attributes #3 = { nounwind "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-cluster-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-cluster-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-cluster-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" } +;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line: +; GISEL: {{.*}} +; SDAG: {{.*}} diff --git a/llvm/test/CodeGen/AMDGPU/call-return-types.ll b/llvm/test/CodeGen/AMDGPU/call-return-types.ll index c0f74fd..21c3696 100644 --- a/llvm/test/CodeGen/AMDGPU/call-return-types.ll +++ b/llvm/test/CodeGen/AMDGPU/call-return-types.ll @@ -1,7 +1,12 @@ -; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -mattr=-flat-for-global < %s | FileCheck -check-prefixes=GCN,GFX89 %s -; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=hawaii < %s | FileCheck -check-prefixes=GCN,GFX7 %s -; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -mattr=-flat-for-global < %s | FileCheck -check-prefixes=GCN,GFX89 %s -; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-flat-for-global < %s | FileCheck -check-prefixes=GCN,GFX11 %s +; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -mattr=-flat-for-global < %s | FileCheck -check-prefixes=GCN,GFX89 %s +; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=hawaii < %s | FileCheck -check-prefixes=GCN,GFX7 %s +; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -mattr=-flat-for-global < %s | FileCheck -check-prefixes=GCN,GFX89 %s +; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1100 -mattr=-flat-for-global < %s | FileCheck -check-prefixes=GCN,GFX11 %s +; RUN: llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -mcpu=fiji -mattr=-flat-for-global < %s | FileCheck -check-prefixes=GCN,GFX89 %s +; RUN: llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -mcpu=hawaii < %s | FileCheck -check-prefixes=GCN,GFX7 %s +; RUN: llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -mattr=-flat-for-global < %s | FileCheck -check-prefixes=GCN,GFX89 %s + +; Ideally, we would also like to test GlobalISel with gfx11 but we are currently blocked on llvm-project#166501. declare void @external_void_func_void() #0 diff --git a/llvm/test/CodeGen/AMDGPU/call-skip.ll b/llvm/test/CodeGen/AMDGPU/call-skip.ll index ea2bba1..e2ca278 100644 --- a/llvm/test/CodeGen/AMDGPU/call-skip.ll +++ b/llvm/test/CodeGen/AMDGPU/call-skip.ll @@ -1,4 +1,6 @@ -; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=hawaii < %s | FileCheck -enable-var-scope -check-prefix=GCN %s +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py +; RUN: llc -mtriple=amdgcn-amd-amdhsa -global-isel=0 -mcpu=hawaii < %s | FileCheck -enable-var-scope -check-prefixes=GCN,SDAG %s +; RUN: llc -mtriple=amdgcn-amd-amdhsa -global-isel=1 -mcpu=hawaii < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GISEL %s ; A call should be skipped if all lanes are zero, since we don't know ; what side effects should be avoided inside the call. @@ -6,12 +8,37 @@ define hidden void @func() #1 { ret void } -; GCN-LABEL: {{^}}if_call: -; GCN: s_and_saveexec_b64 -; GCN-NEXT: s_cbranch_execz [[END:.LBB[0-9]+_[0-9]+]] -; GCN: s_swappc_b64 -; GCN: [[END]]: define void @if_call(i32 %flag) #0 { +; GCN-LABEL: if_call: +; GCN: ; %bb.0: +; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GCN-NEXT: s_mov_b32 s20, s33 +; GCN-NEXT: s_mov_b32 s33, s32 +; GCN-NEXT: s_xor_saveexec_b64 s[16:17], -1 +; GCN-NEXT: buffer_store_dword v1, off, s[0:3], s33 ; 4-byte Folded Spill +; GCN-NEXT: s_mov_b64 exec, s[16:17] +; GCN-NEXT: v_writelane_b32 v1, s30, 0 +; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GCN-NEXT: s_addk_i32 s32, 0x400 +; GCN-NEXT: v_writelane_b32 v1, s31, 1 +; GCN-NEXT: s_and_saveexec_b64 s[16:17], vcc +; GCN-NEXT: s_cbranch_execz .LBB1_2 +; GCN-NEXT: ; %bb.1: ; %call +; GCN-NEXT: s_getpc_b64 s[18:19] +; GCN-NEXT: s_add_u32 s18, s18, func@rel32@lo+4 +; GCN-NEXT: s_addc_u32 s19, s19, func@rel32@hi+12 +; GCN-NEXT: s_swappc_b64 s[30:31], s[18:19] +; GCN-NEXT: .LBB1_2: ; %end +; GCN-NEXT: s_or_b64 exec, exec, s[16:17] +; GCN-NEXT: v_readlane_b32 s31, v1, 1 +; GCN-NEXT: v_readlane_b32 s30, v1, 0 +; GCN-NEXT: s_mov_b32 s32, s33 +; GCN-NEXT: s_xor_saveexec_b64 s[4:5], -1 +; GCN-NEXT: buffer_load_dword v1, off, s[0:3], s33 ; 4-byte Folded Reload +; GCN-NEXT: s_mov_b64 exec, s[4:5] +; GCN-NEXT: s_mov_b32 s33, s20 +; GCN-NEXT: s_waitcnt vmcnt(0) +; GCN-NEXT: s_setpc_b64 s[30:31] %cc = icmp eq i32 %flag, 0 br i1 %cc, label %call, label %end @@ -23,12 +50,20 @@ end: ret void } -; GCN-LABEL: {{^}}if_asm: -; GCN: s_and_saveexec_b64 -; GCN-NEXT: s_cbranch_execz [[END:.LBB[0-9]+_[0-9]+]] -; GCN: ; sample asm -; GCN: [[END]]: define void @if_asm(i32 %flag) #0 { +; GCN-LABEL: if_asm: +; GCN: ; %bb.0: +; GCN-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GCN-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GCN-NEXT: s_and_saveexec_b64 s[4:5], vcc +; GCN-NEXT: s_cbranch_execz .LBB2_2 +; GCN-NEXT: ; %bb.1: ; %call +; GCN-NEXT: ;;#ASMSTART +; GCN-NEXT: ; sample asm +; GCN-NEXT: ;;#ASMEND +; GCN-NEXT: .LBB2_2: ; %end +; GCN-NEXT: s_or_b64 exec, exec, s[4:5] +; GCN-NEXT: s_setpc_b64 s[30:31] %cc = icmp eq i32 %flag, 0 br i1 %cc, label %call, label %end @@ -40,11 +75,58 @@ end: ret void } -; GCN-LABEL: {{^}}if_call_kernel: -; GCN: s_and_saveexec_b64 -; GCN-NEXT: s_cbranch_execz .LBB3_2 -; GCN: s_swappc_b64 define amdgpu_kernel void @if_call_kernel() #0 { +; SDAG-LABEL: if_call_kernel: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_add_i32 s12, s12, s17 +; SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; SDAG-NEXT: s_add_u32 s0, s0, s17 +; SDAG-NEXT: s_addc_u32 s1, s1, 0 +; SDAG-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 +; SDAG-NEXT: s_and_saveexec_b64 s[12:13], vcc +; SDAG-NEXT: s_cbranch_execz .LBB3_2 +; SDAG-NEXT: ; %bb.1: ; %call +; SDAG-NEXT: v_lshlrev_b32_e32 v1, 10, v1 +; SDAG-NEXT: v_lshlrev_b32_e32 v2, 20, v2 +; SDAG-NEXT: v_or_b32_e32 v0, v0, v1 +; SDAG-NEXT: s_getpc_b64 s[18:19] +; SDAG-NEXT: s_add_u32 s18, s18, func@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s19, s19, func@rel32@hi+12 +; SDAG-NEXT: v_or_b32_e32 v31, v0, v2 +; SDAG-NEXT: s_mov_b32 s12, s14 +; SDAG-NEXT: s_mov_b32 s13, s15 +; SDAG-NEXT: s_mov_b32 s14, s16 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[18:19] +; SDAG-NEXT: .LBB3_2: ; %end +; SDAG-NEXT: s_endpgm +; +; GISEL-LABEL: if_call_kernel: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_add_i32 s12, s12, s17 +; GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; GISEL-NEXT: s_add_u32 s0, s0, s17 +; GISEL-NEXT: s_addc_u32 s1, s1, 0 +; GISEL-NEXT: v_cmp_eq_u32_e32 vcc, 0, v0 +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13 +; GISEL-NEXT: s_and_saveexec_b64 s[12:13], vcc +; GISEL-NEXT: s_cbranch_execz .LBB3_2 +; GISEL-NEXT: ; %bb.1: ; %call +; GISEL-NEXT: v_lshlrev_b32_e32 v1, 10, v1 +; GISEL-NEXT: v_or_b32_e32 v0, v0, v1 +; GISEL-NEXT: v_lshlrev_b32_e32 v1, 20, v2 +; GISEL-NEXT: s_getpc_b64 s[18:19] +; GISEL-NEXT: s_add_u32 s18, s18, func@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s19, s19, func@rel32@hi+12 +; GISEL-NEXT: v_or_b32_e32 v31, v0, v1 +; GISEL-NEXT: s_mov_b32 s12, s14 +; GISEL-NEXT: s_mov_b32 s13, s15 +; GISEL-NEXT: s_mov_b32 s14, s16 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[18:19] +; GISEL-NEXT: .LBB3_2: ; %end +; GISEL-NEXT: s_endpgm %id = call i32 @llvm.amdgcn.workitem.id.x() %cc = icmp eq i32 %id, 0 br i1 %cc, label %call, label %end diff --git a/llvm/test/CodeGen/AMDGPU/call-waitcnt.ll b/llvm/test/CodeGen/AMDGPU/call-waitcnt.ll index 675acd0..a52942c 100644 --- a/llvm/test/CodeGen/AMDGPU/call-waitcnt.ll +++ b/llvm/test/CodeGen/AMDGPU/call-waitcnt.ll @@ -1,5 +1,6 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py -; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck -enable-var-scope -check-prefix=GCN %s +; RUN: llc -global-isel=0 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck -enable-var-scope -check-prefixes=GCN,SDAG %s +; RUN: llc -global-isel=1 -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GISEL %s ; Load argument depends on waitcnt which should be skipped. define amdgpu_kernel void @call_memory_arg_load(ptr addrspace(3) %ptr, i32) #0 { @@ -27,24 +28,43 @@ define amdgpu_kernel void @call_memory_arg_load(ptr addrspace(3) %ptr, i32) #0 { ; Memory waitcnt with no register dependence on the call define amdgpu_kernel void @call_memory_no_dep(ptr addrspace(1) %ptr, i32) #0 { -; GCN-LABEL: call_memory_no_dep: -; GCN: ; %bb.0: -; GCN-NEXT: s_load_dwordx2 s[6:7], s[6:7], 0x0 -; GCN-NEXT: s_add_u32 flat_scratch_lo, s8, s11 -; GCN-NEXT: s_addc_u32 flat_scratch_hi, s9, 0 -; GCN-NEXT: s_add_u32 s0, s0, s11 -; GCN-NEXT: s_addc_u32 s1, s1, 0 -; GCN-NEXT: v_mov_b32_e32 v0, 0 -; GCN-NEXT: s_getpc_b64 s[8:9] -; GCN-NEXT: s_add_u32 s8, s8, func@rel32@lo+4 -; GCN-NEXT: s_addc_u32 s9, s9, func@rel32@hi+12 -; GCN-NEXT: s_waitcnt lgkmcnt(0) -; GCN-NEXT: global_store_dword v0, v0, s[6:7] -; GCN-NEXT: s_mov_b64 s[6:7], s[4:5] -; GCN-NEXT: v_mov_b32_e32 v0, 0 -; GCN-NEXT: s_mov_b32 s32, 0 -; GCN-NEXT: s_swappc_b64 s[30:31], s[8:9] -; GCN-NEXT: s_endpgm +; SDAG-LABEL: call_memory_no_dep: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_load_dwordx2 s[6:7], s[6:7], 0x0 +; SDAG-NEXT: s_add_u32 flat_scratch_lo, s8, s11 +; SDAG-NEXT: s_addc_u32 flat_scratch_hi, s9, 0 +; SDAG-NEXT: s_add_u32 s0, s0, s11 +; SDAG-NEXT: s_addc_u32 s1, s1, 0 +; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: s_getpc_b64 s[8:9] +; SDAG-NEXT: s_add_u32 s8, s8, func@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s9, s9, func@rel32@hi+12 +; SDAG-NEXT: s_waitcnt lgkmcnt(0) +; SDAG-NEXT: global_store_dword v0, v0, s[6:7] +; SDAG-NEXT: s_mov_b64 s[6:7], s[4:5] +; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[8:9] +; SDAG-NEXT: s_endpgm +; +; GISEL-LABEL: call_memory_no_dep: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_load_dwordx2 s[6:7], s[6:7], 0x0 +; GISEL-NEXT: s_add_u32 flat_scratch_lo, s8, s11 +; GISEL-NEXT: s_addc_u32 flat_scratch_hi, s9, 0 +; GISEL-NEXT: s_add_u32 s0, s0, s11 +; GISEL-NEXT: s_addc_u32 s1, s1, 0 +; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: s_getpc_b64 s[8:9] +; GISEL-NEXT: s_add_u32 s8, s8, func@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s9, s9, func@rel32@hi+12 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: global_store_dword v0, v0, s[6:7] +; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[4:5] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[8:9] +; GISEL-NEXT: s_endpgm store i32 0, ptr addrspace(1) %ptr call void @func(i32 0) ret void @@ -52,46 +72,82 @@ define amdgpu_kernel void @call_memory_no_dep(ptr addrspace(1) %ptr, i32) #0 { ; Should not wait after the call before memory define amdgpu_kernel void @call_no_wait_after_call(ptr addrspace(1) %ptr, i32) #0 { -; GCN-LABEL: call_no_wait_after_call: -; GCN: ; %bb.0: -; GCN-NEXT: s_add_u32 flat_scratch_lo, s8, s11 -; GCN-NEXT: s_addc_u32 flat_scratch_hi, s9, 0 -; GCN-NEXT: s_load_dwordx2 s[34:35], s[6:7], 0x0 -; GCN-NEXT: s_add_u32 s0, s0, s11 -; GCN-NEXT: s_addc_u32 s1, s1, 0 -; GCN-NEXT: s_getpc_b64 s[8:9] -; GCN-NEXT: s_add_u32 s8, s8, func@rel32@lo+4 -; GCN-NEXT: s_addc_u32 s9, s9, func@rel32@hi+12 -; GCN-NEXT: s_mov_b64 s[6:7], s[4:5] -; GCN-NEXT: v_mov_b32_e32 v0, 0 -; GCN-NEXT: s_mov_b32 s32, 0 -; GCN-NEXT: v_mov_b32_e32 v40, 0 -; GCN-NEXT: s_swappc_b64 s[30:31], s[8:9] -; GCN-NEXT: global_store_dword v40, v40, s[34:35] -; GCN-NEXT: s_endpgm +; SDAG-LABEL: call_no_wait_after_call: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_add_u32 flat_scratch_lo, s8, s11 +; SDAG-NEXT: s_addc_u32 flat_scratch_hi, s9, 0 +; SDAG-NEXT: s_load_dwordx2 s[34:35], s[6:7], 0x0 +; SDAG-NEXT: s_add_u32 s0, s0, s11 +; SDAG-NEXT: s_addc_u32 s1, s1, 0 +; SDAG-NEXT: s_getpc_b64 s[8:9] +; SDAG-NEXT: s_add_u32 s8, s8, func@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s9, s9, func@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[6:7], s[4:5] +; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: v_mov_b32_e32 v40, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[8:9] +; SDAG-NEXT: global_store_dword v40, v40, s[34:35] +; SDAG-NEXT: s_endpgm +; +; GISEL-LABEL: call_no_wait_after_call: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_add_u32 flat_scratch_lo, s8, s11 +; GISEL-NEXT: s_addc_u32 flat_scratch_hi, s9, 0 +; GISEL-NEXT: s_load_dwordx2 s[34:35], s[6:7], 0x0 +; GISEL-NEXT: s_add_u32 s0, s0, s11 +; GISEL-NEXT: s_addc_u32 s1, s1, 0 +; GISEL-NEXT: s_getpc_b64 s[8:9] +; GISEL-NEXT: s_add_u32 s8, s8, func@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s9, s9, func@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[4:5] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[8:9] +; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: global_store_dword v0, v0, s[34:35] +; GISEL-NEXT: s_endpgm call void @func(i32 0) store i32 0, ptr addrspace(1) %ptr ret void } define amdgpu_kernel void @call_no_wait_after_call_return_val(ptr addrspace(1) %ptr, i32) #0 { -; GCN-LABEL: call_no_wait_after_call_return_val: -; GCN: ; %bb.0: -; GCN-NEXT: s_add_u32 flat_scratch_lo, s8, s11 -; GCN-NEXT: s_addc_u32 flat_scratch_hi, s9, 0 -; GCN-NEXT: s_load_dwordx2 s[34:35], s[6:7], 0x0 -; GCN-NEXT: s_add_u32 s0, s0, s11 -; GCN-NEXT: s_addc_u32 s1, s1, 0 -; GCN-NEXT: s_getpc_b64 s[8:9] -; GCN-NEXT: s_add_u32 s8, s8, func.return@rel32@lo+4 -; GCN-NEXT: s_addc_u32 s9, s9, func.return@rel32@hi+12 -; GCN-NEXT: s_mov_b64 s[6:7], s[4:5] -; GCN-NEXT: v_mov_b32_e32 v0, 0 -; GCN-NEXT: s_mov_b32 s32, 0 -; GCN-NEXT: v_mov_b32_e32 v40, 0 -; GCN-NEXT: s_swappc_b64 s[30:31], s[8:9] -; GCN-NEXT: global_store_dword v40, v0, s[34:35] -; GCN-NEXT: s_endpgm +; SDAG-LABEL: call_no_wait_after_call_return_val: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_add_u32 flat_scratch_lo, s8, s11 +; SDAG-NEXT: s_addc_u32 flat_scratch_hi, s9, 0 +; SDAG-NEXT: s_load_dwordx2 s[34:35], s[6:7], 0x0 +; SDAG-NEXT: s_add_u32 s0, s0, s11 +; SDAG-NEXT: s_addc_u32 s1, s1, 0 +; SDAG-NEXT: s_getpc_b64 s[8:9] +; SDAG-NEXT: s_add_u32 s8, s8, func.return@rel32@lo+4 +; SDAG-NEXT: s_addc_u32 s9, s9, func.return@rel32@hi+12 +; SDAG-NEXT: s_mov_b64 s[6:7], s[4:5] +; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: v_mov_b32_e32 v40, 0 +; SDAG-NEXT: s_swappc_b64 s[30:31], s[8:9] +; SDAG-NEXT: global_store_dword v40, v0, s[34:35] +; SDAG-NEXT: s_endpgm +; +; GISEL-LABEL: call_no_wait_after_call_return_val: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_add_u32 flat_scratch_lo, s8, s11 +; GISEL-NEXT: s_addc_u32 flat_scratch_hi, s9, 0 +; GISEL-NEXT: s_load_dwordx2 s[34:35], s[6:7], 0x0 +; GISEL-NEXT: s_add_u32 s0, s0, s11 +; GISEL-NEXT: s_addc_u32 s1, s1, 0 +; GISEL-NEXT: s_getpc_b64 s[8:9] +; GISEL-NEXT: s_add_u32 s8, s8, func.return@rel32@lo+4 +; GISEL-NEXT: s_addc_u32 s9, s9, func.return@rel32@hi+12 +; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[4:5] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_swappc_b64 s[30:31], s[8:9] +; GISEL-NEXT: v_mov_b32_e32 v1, 0 +; GISEL-NEXT: global_store_dword v1, v0, s[34:35] +; GISEL-NEXT: s_endpgm %rv = call i32 @func.return(i32 0) store i32 %rv, ptr addrspace(1) %ptr ret void @@ -99,22 +155,39 @@ define amdgpu_kernel void @call_no_wait_after_call_return_val(ptr addrspace(1) % ; Need to wait for the address dependency define amdgpu_kernel void @call_got_load(ptr addrspace(1) %ptr, i32) #0 { -; GCN-LABEL: call_got_load: -; GCN: ; %bb.0: -; GCN-NEXT: s_add_u32 flat_scratch_lo, s8, s11 -; GCN-NEXT: s_addc_u32 flat_scratch_hi, s9, 0 -; GCN-NEXT: s_add_u32 s0, s0, s11 -; GCN-NEXT: s_addc_u32 s1, s1, 0 -; GCN-NEXT: s_getpc_b64 s[6:7] -; GCN-NEXT: s_add_u32 s6, s6, got.func@gotpcrel32@lo+4 -; GCN-NEXT: s_addc_u32 s7, s7, got.func@gotpcrel32@hi+12 -; GCN-NEXT: s_load_dwordx2 s[8:9], s[6:7], 0x0 -; GCN-NEXT: s_mov_b64 s[6:7], s[4:5] -; GCN-NEXT: v_mov_b32_e32 v0, 0 -; GCN-NEXT: s_mov_b32 s32, 0 -; GCN-NEXT: s_waitcnt lgkmcnt(0) -; GCN-NEXT: s_swappc_b64 s[30:31], s[8:9] -; GCN-NEXT: s_endpgm +; SDAG-LABEL: call_got_load: +; SDAG: ; %bb.0: +; SDAG-NEXT: s_add_u32 flat_scratch_lo, s8, s11 +; SDAG-NEXT: s_addc_u32 flat_scratch_hi, s9, 0 +; SDAG-NEXT: s_add_u32 s0, s0, s11 +; SDAG-NEXT: s_addc_u32 s1, s1, 0 +; SDAG-NEXT: s_getpc_b64 s[6:7] +; SDAG-NEXT: s_add_u32 s6, s6, got.func@gotpcrel32@lo+4 +; SDAG-NEXT: s_addc_u32 s7, s7, got.func@gotpcrel32@hi+12 +; SDAG-NEXT: s_load_dwordx2 s[8:9], s[6:7], 0x0 +; SDAG-NEXT: s_mov_b64 s[6:7], s[4:5] +; SDAG-NEXT: v_mov_b32_e32 v0, 0 +; SDAG-NEXT: s_mov_b32 s32, 0 +; SDAG-NEXT: s_waitcnt lgkmcnt(0) +; SDAG-NEXT: s_swappc_b64 s[30:31], s[8:9] +; SDAG-NEXT: s_endpgm +; +; GISEL-LABEL: call_got_load: +; GISEL: ; %bb.0: +; GISEL-NEXT: s_add_u32 flat_scratch_lo, s8, s11 +; GISEL-NEXT: s_addc_u32 flat_scratch_hi, s9, 0 +; GISEL-NEXT: s_add_u32 s0, s0, s11 +; GISEL-NEXT: s_addc_u32 s1, s1, 0 +; GISEL-NEXT: s_getpc_b64 s[6:7] +; GISEL-NEXT: s_add_u32 s6, s6, got.func@gotpcrel32@lo+4 +; GISEL-NEXT: s_addc_u32 s7, s7, got.func@gotpcrel32@hi+12 +; GISEL-NEXT: s_load_dwordx2 s[8:9], s[6:7], 0x0 +; GISEL-NEXT: v_mov_b32_e32 v0, 0 +; GISEL-NEXT: s_mov_b64 s[6:7], s[4:5] +; GISEL-NEXT: s_mov_b32 s32, 0 +; GISEL-NEXT: s_waitcnt lgkmcnt(0) +; GISEL-NEXT: s_swappc_b64 s[30:31], s[8:9] +; GISEL-NEXT: s_endpgm call void @got.func(i32 0) ret void } diff --git a/llvm/test/CodeGen/AMDGPU/carryout-selection.ll b/llvm/test/CodeGen/AMDGPU/carryout-selection.ll index b96de17..8d05317 100644 --- a/llvm/test/CodeGen/AMDGPU/carryout-selection.ll +++ b/llvm/test/CodeGen/AMDGPU/carryout-selection.ll @@ -702,8 +702,6 @@ define amdgpu_kernel void @suaddo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; CISI-NEXT: s_mov_b32 s10, -1 ; CISI-NEXT: s_waitcnt lgkmcnt(0) ; CISI-NEXT: s_add_u32 s4, s4, s6 -; CISI-NEXT: s_cselect_b64 s[12:13], -1, 0 -; CISI-NEXT: s_or_b32 s6, s12, s13 ; CISI-NEXT: s_addc_u32 s5, s5, s7 ; CISI-NEXT: s_mov_b32 s8, s0 ; CISI-NEXT: s_mov_b32 s9, s1 @@ -1674,8 +1672,6 @@ define amdgpu_kernel void @susubo64(ptr addrspace(1) %out, ptr addrspace(1) %car ; CISI-NEXT: s_mov_b32 s10, -1 ; CISI-NEXT: s_waitcnt lgkmcnt(0) ; CISI-NEXT: s_sub_u32 s4, s4, s6 -; CISI-NEXT: s_cselect_b64 s[12:13], -1, 0 -; CISI-NEXT: s_or_b32 s6, s12, s13 ; CISI-NEXT: s_subb_u32 s5, s5, s7 ; CISI-NEXT: s_mov_b32 s8, s0 ; CISI-NEXT: s_mov_b32 s9, s1 diff --git a/llvm/test/CodeGen/AMDGPU/expand-scalar-carry-out-select-user.ll b/llvm/test/CodeGen/AMDGPU/expand-scalar-carry-out-select-user.ll index dbdea8e..71af21a1 100644 --- a/llvm/test/CodeGen/AMDGPU/expand-scalar-carry-out-select-user.ll +++ b/llvm/test/CodeGen/AMDGPU/expand-scalar-carry-out-select-user.ll @@ -12,8 +12,6 @@ define i32 @s_add_co_select_user() { ; GFX7-NEXT: s_load_dword s6, s[4:5], 0x0 ; GFX7-NEXT: s_waitcnt lgkmcnt(0) ; GFX7-NEXT: s_add_u32 s7, s6, s6 -; GFX7-NEXT: s_cselect_b64 s[4:5], -1, 0 -; GFX7-NEXT: s_or_b32 s4, s4, s5 ; GFX7-NEXT: s_addc_u32 s8, s6, 0 ; GFX7-NEXT: s_cselect_b64 s[4:5], -1, 0 ; GFX7-NEXT: s_and_b64 s[4:5], s[4:5], exec @@ -88,15 +86,13 @@ bb: define amdgpu_kernel void @s_add_co_br_user(i32 %i) { ; GFX7-LABEL: s_add_co_br_user: ; GFX7: ; %bb.0: ; %bb -; GFX7-NEXT: s_load_dword s2, s[8:9], 0x0 +; GFX7-NEXT: s_load_dword s0, s[8:9], 0x0 ; GFX7-NEXT: s_add_i32 s12, s12, s17 ; GFX7-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 ; GFX7-NEXT: s_mov_b32 flat_scratch_lo, s13 ; GFX7-NEXT: s_waitcnt lgkmcnt(0) -; GFX7-NEXT: s_add_u32 s0, s2, s2 -; GFX7-NEXT: s_cselect_b64 s[0:1], -1, 0 -; GFX7-NEXT: s_or_b32 s0, s0, s1 -; GFX7-NEXT: s_addc_u32 s0, s2, 0 +; GFX7-NEXT: s_add_u32 s1, s0, s0 +; GFX7-NEXT: s_addc_u32 s0, s0, 0 ; GFX7-NEXT: s_cselect_b64 s[0:1], -1, 0 ; GFX7-NEXT: s_andn2_b64 vcc, exec, s[0:1] ; GFX7-NEXT: s_cbranch_vccnz .LBB1_2 diff --git a/llvm/test/CodeGen/AMDGPU/fix-sgpr-copies-wwm.ll b/llvm/test/CodeGen/AMDGPU/fix-sgpr-copies-wwm.ll index b8f084d..db32135 100644 --- a/llvm/test/CodeGen/AMDGPU/fix-sgpr-copies-wwm.ll +++ b/llvm/test/CodeGen/AMDGPU/fix-sgpr-copies-wwm.ll @@ -4,14 +4,24 @@ define amdgpu_gs i32 @main() { ; CHECK-LABEL: main: ; CHECK: ; %bb.0: ; %bb +; CHECK-NEXT: s_bitcmp1_b32 0, 0 ; CHECK-NEXT: s_mov_b32 s0, 0 -; CHECK-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1) +; CHECK-NEXT: s_cselect_b32 s1, -1, 0 +; CHECK-NEXT: s_or_saveexec_b32 s2, -1 +; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, 1, s1 +; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_1) +; CHECK-NEXT: v_readfirstlane_b32 s1, v0 +; CHECK-NEXT: s_mov_b32 exec_lo, s2 +; CHECK-NEXT: s_or_b32 s0, s0, s1 +; CHECK-NEXT: s_wait_alu 0xfffe ; CHECK-NEXT: s_bitcmp1_b32 s0, 0 ; CHECK-NEXT: s_cselect_b32 s0, -1, 0 +; CHECK-NEXT: s_wait_alu 0xfffe ; CHECK-NEXT: s_xor_b32 s0, s0, -1 -; CHECK-NEXT: s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1) -; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, 1, s0 -; CHECK-NEXT: v_readfirstlane_b32 s0, v0 +; CHECK-NEXT: s_wait_alu 0xfffe +; CHECK-NEXT: v_cndmask_b32_e64 v1, 0, 1, s0 +; CHECK-NEXT: s_delay_alu instid0(VALU_DEP_1) +; CHECK-NEXT: v_readfirstlane_b32 s0, v1 ; CHECK-NEXT: s_wait_alu 0xf1ff ; CHECK-NEXT: ; return to shader part epilog bb: diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx90a.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx90a.ll index 22bc62ac..679b289 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx90a.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx90a.ll @@ -1,8 +1,8 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 -; RUN: llc -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX90A %s -; RUN: llc -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX942 %s -; RUN: llc -mtriple=amdgcn -mcpu=gfx90a -amdgpu-mfma-vgpr-form < %s | FileCheck -enable-var-scope --check-prefixes=VGPR,GFX90A-VGPR %s -; RUN: llc -mtriple=amdgcn -mcpu=gfx942 -amdgpu-mfma-vgpr-form < %s | FileCheck -enable-var-scope --check-prefixes=VGPR,GFX942-VGPR %s +; RUN: llc -verify-machineinstrs -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX90A %s +; RUN: llc -verify-machineinstrs -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX942 %s +; RUN: llc -verify-machineinstrs -mtriple=amdgcn -mcpu=gfx90a -amdgpu-mfma-vgpr-form < %s | FileCheck -enable-var-scope --check-prefixes=VGPR,GFX90A-VGPR %s +; RUN: llc -verify-machineinstrs -mtriple=amdgcn -mcpu=gfx942 -amdgpu-mfma-vgpr-form < %s | FileCheck -enable-var-scope --check-prefixes=VGPR,GFX942-VGPR %s declare <32 x float> @llvm.amdgcn.mfma.f32.32x32x4bf16.1k(<4 x i16>, <4 x i16>, <32 x float>, i32, i32, i32) declare <16 x float> @llvm.amdgcn.mfma.f32.16x16x4bf16.1k(<4 x i16>, <4 x i16>, <16 x float>, i32, i32, i32) diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll index 7e30af9..e7d7f87 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll @@ -1,9 +1,9 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 -; RUN: llc -mtriple=amdgcn -mcpu=gfx908 < %s | FileCheck -enable-var-scope --check-prefixes=GCN,NOLIT-SRCC,GFX908,GFX908_A %s -; RUN: llc -mtriple=amdgcn -mcpu=gfx908 -mattr=-mfma-inline-literal-bug < %s | FileCheck -enable-var-scope --check-prefixes=GCN,LIT-SRCC,GFX908,GFX908_A %s -; RUN: llc -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX90A,GFX908_A,GFX90A_42 %s -; RUN: llc -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX942,GFX90A_42 %s -; RUN: llc -mtriple=amdgcn -mcpu=gfx942 -amdgpu-mfma-vgpr-form < %s | FileCheck -enable-var-scope --check-prefix=GFX942-VGPR %s +; RUN: llc -verify-machineinstrs -mtriple=amdgcn -mcpu=gfx908 < %s | FileCheck -enable-var-scope --check-prefixes=GCN,NOLIT-SRCC,GFX908,GFX908_A %s +; RUN: llc -verify-machineinstrs -mtriple=amdgcn -mcpu=gfx908 -mattr=-mfma-inline-literal-bug < %s | FileCheck -enable-var-scope --check-prefixes=GCN,LIT-SRCC,GFX908,GFX908_A %s +; RUN: llc -verify-machineinstrs -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX90A,GFX908_A,GFX90A_42 %s +; RUN: llc -verify-machineinstrs -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX942,GFX90A_42 %s +; RUN: llc -verify-machineinstrs -mtriple=amdgcn -mcpu=gfx942 -amdgpu-mfma-vgpr-form < %s | FileCheck -enable-var-scope --check-prefix=GFX942-VGPR %s declare <32 x float> @llvm.amdgcn.mfma.f32.32x32x1f32(float, float, <32 x float>, i32, i32, i32) declare <16 x float> @llvm.amdgcn.mfma.f32.16x16x1f32(float, float, <16 x float>, i32, i32, i32) @@ -3186,13 +3186,14 @@ define amdgpu_kernel void @test_mfma_i32_16x16x4i8_splatimm_src2_64(ptr addrspac ; ; GFX942-VGPR-LABEL: test_mfma_i32_16x16x4i8_splatimm_src2_64: ; GFX942-VGPR: ; %bb.0: ; %bb -; GFX942-VGPR-NEXT: v_mov_b32_e32 v0, 1 -; GFX942-VGPR-NEXT: v_mov_b32_e32 v1, 2 +; GFX942-VGPR-NEXT: v_mov_b32_e32 v16, 1 +; GFX942-VGPR-NEXT: v_mov_b32_e32 v17, 2 ; GFX942-VGPR-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24 +; GFX942-VGPR-NEXT: s_nop 0 +; GFX942-VGPR-NEXT: v_mfma_i32_16x16x4_4b_i8 v[0:15], v16, v17, 64 cbsz:1 abid:2 blgp:3 ; GFX942-VGPR-NEXT: v_mov_b32_e32 v16, 0 -; GFX942-VGPR-NEXT: v_mfma_i32_16x16x4_4b_i8 v[0:15], v0, v1, 64 cbsz:1 abid:2 blgp:3 ; GFX942-VGPR-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-VGPR-NEXT: s_nop 9 +; GFX942-VGPR-NEXT: s_nop 8 ; GFX942-VGPR-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1] offset:48 ; GFX942-VGPR-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1] offset:32 ; GFX942-VGPR-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1] offset:16 @@ -4538,13 +4539,14 @@ define amdgpu_kernel void @test_mfma_f32_16x16x1f32_imm_splat(ptr addrspace(1) % ; ; GFX942-VGPR-LABEL: test_mfma_f32_16x16x1f32_imm_splat: ; GFX942-VGPR: ; %bb.0: ; %bb -; GFX942-VGPR-NEXT: v_mov_b32_e32 v0, 1.0 -; GFX942-VGPR-NEXT: v_mov_b32_e32 v1, 2.0 +; GFX942-VGPR-NEXT: v_mov_b32_e32 v16, 1.0 +; GFX942-VGPR-NEXT: v_mov_b32_e32 v17, 2.0 ; GFX942-VGPR-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24 +; GFX942-VGPR-NEXT: s_nop 0 +; GFX942-VGPR-NEXT: v_mfma_f32_16x16x1_4b_f32 v[0:15], v16, v17, 1.0 ; GFX942-VGPR-NEXT: v_mov_b32_e32 v16, 0 -; GFX942-VGPR-NEXT: v_mfma_f32_16x16x1_4b_f32 v[0:15], v0, v1, 1.0 ; GFX942-VGPR-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-VGPR-NEXT: s_nop 8 +; GFX942-VGPR-NEXT: s_nop 7 ; GFX942-VGPR-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1] offset:48 ; GFX942-VGPR-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1] offset:32 ; GFX942-VGPR-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1] offset:16 @@ -4689,15 +4691,16 @@ define amdgpu_kernel void @test_mfma_f32_32x32x8f16_imm_splat(ptr addrspace(1) % ; ; GFX942-VGPR-LABEL: test_mfma_f32_32x32x8f16_imm_splat: ; GFX942-VGPR: ; %bb.0: ; %bb -; GFX942-VGPR-NEXT: v_mov_b32_e32 v0, 0x3c003c00 -; GFX942-VGPR-NEXT: v_mov_b32_e32 v1, v0 -; GFX942-VGPR-NEXT: v_mov_b32_e32 v2, 0x40004000 -; GFX942-VGPR-NEXT: v_mov_b32_e32 v3, v2 +; GFX942-VGPR-NEXT: v_mov_b32_e32 v16, 0x3c003c00 +; GFX942-VGPR-NEXT: v_mov_b32_e32 v17, v16 +; GFX942-VGPR-NEXT: v_mov_b32_e32 v18, 0x40004000 +; GFX942-VGPR-NEXT: v_mov_b32_e32 v19, v18 ; GFX942-VGPR-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24 +; GFX942-VGPR-NEXT: s_nop 0 +; GFX942-VGPR-NEXT: v_mfma_f32_32x32x8_f16 v[0:15], v[16:17], v[18:19], 1.0 ; GFX942-VGPR-NEXT: v_mov_b32_e32 v16, 0 -; GFX942-VGPR-NEXT: v_mfma_f32_32x32x8_f16 v[0:15], v[0:1], v[2:3], 1.0 ; GFX942-VGPR-NEXT: s_waitcnt lgkmcnt(0) -; GFX942-VGPR-NEXT: s_nop 9 +; GFX942-VGPR-NEXT: s_nop 8 ; GFX942-VGPR-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1] offset:48 ; GFX942-VGPR-NEXT: global_store_dwordx4 v16, v[8:11], s[0:1] offset:32 ; GFX942-VGPR-NEXT: global_store_dwordx4 v16, v[4:7], s[0:1] offset:16 @@ -4908,14 +4911,14 @@ define amdgpu_kernel void @test_mfma_f32_32x32x1f32_imm_splat(ptr addrspace(1) % ; ; GFX942-VGPR-LABEL: test_mfma_f32_32x32x1f32_imm_splat: ; GFX942-VGPR: ; %bb.0: ; %bb -; GFX942-VGPR-NEXT: v_mov_b32_e32 v0, 1.0 -; GFX942-VGPR-NEXT: v_mov_b32_e32 v1, 2.0 +; GFX942-VGPR-NEXT: v_mov_b32_e32 v32, 1.0 +; GFX942-VGPR-NEXT: v_mov_b32_e32 v33, 2.0 ; GFX942-VGPR-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0x24 +; GFX942-VGPR-NEXT: s_nop 0 +; GFX942-VGPR-NEXT: v_mfma_f32_32x32x1_2b_f32 v[0:31], v32, v33, 0 ; GFX942-VGPR-NEXT: v_mov_b32_e32 v32, 0 -; GFX942-VGPR-NEXT: v_mfma_f32_32x32x1_2b_f32 v[0:31], v0, v1, 0 ; GFX942-VGPR-NEXT: s_waitcnt lgkmcnt(0) ; GFX942-VGPR-NEXT: s_nop 15 -; GFX942-VGPR-NEXT: s_nop 0 ; GFX942-VGPR-NEXT: global_store_dwordx4 v32, v[28:31], s[0:1] offset:112 ; GFX942-VGPR-NEXT: global_store_dwordx4 v32, v[24:27], s[0:1] offset:96 ; GFX942-VGPR-NEXT: global_store_dwordx4 v32, v[20:23], s[0:1] offset:80 diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll index 02d2990..d1ba892 100644 --- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll +++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.readfirstlane.ll @@ -396,7 +396,8 @@ define amdgpu_kernel void @test_readfirstlane_imm_f64(ptr addrspace(1) %out) { ; ; CHECK-GISEL-LABEL: test_readfirstlane_imm_f64: ; CHECK-GISEL: ; %bb.0: -; CHECK-GISEL-NEXT: s_mov_b64 s[0:1], 0x4040000000000000 +; CHECK-GISEL-NEXT: s_mov_b32 s0, 0 +; CHECK-GISEL-NEXT: s_mov_b32 s1, 0x40400000 ; CHECK-GISEL-NEXT: ;;#ASMSTART ; CHECK-GISEL-NEXT: ; use s[0:1] ; CHECK-GISEL-NEXT: ;;#ASMEND @@ -455,13 +456,14 @@ define amdgpu_kernel void @test_readfirstlane_imm_fold_i64(ptr addrspace(1) %out ; CHECK-GISEL-LABEL: test_readfirstlane_imm_fold_i64: ; CHECK-GISEL: ; %bb.0: ; CHECK-GISEL-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 +; CHECK-GISEL-NEXT: s_mov_b64 s[2:3], 32 ; CHECK-GISEL-NEXT: s_add_i32 s12, s12, s17 -; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, 32 +; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, s2 ; CHECK-GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13 -; CHECK-GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 ; CHECK-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; CHECK-GISEL-NEXT: v_mov_b32_e32 v3, s1 -; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, 0 +; CHECK-GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, s3 ; CHECK-GISEL-NEXT: v_mov_b32_e32 v2, s0 ; CHECK-GISEL-NEXT: flat_store_dwordx2 v[2:3], v[0:1] ; CHECK-GISEL-NEXT: s_endpgm @@ -488,13 +490,15 @@ define amdgpu_kernel void @test_readfirstlane_imm_fold_f64(ptr addrspace(1) %out ; CHECK-GISEL-LABEL: test_readfirstlane_imm_fold_f64: ; CHECK-GISEL: ; %bb.0: ; CHECK-GISEL-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 +; CHECK-GISEL-NEXT: s_mov_b32 s2, 0 ; CHECK-GISEL-NEXT: s_add_i32 s12, s12, s17 -; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, 0 -; CHECK-GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13 -; CHECK-GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; CHECK-GISEL-NEXT: s_mov_b32 s3, 0x40400000 +; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, s2 ; CHECK-GISEL-NEXT: s_waitcnt lgkmcnt(0) ; CHECK-GISEL-NEXT: v_mov_b32_e32 v3, s1 -; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, 0x40400000 +; CHECK-GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13 +; CHECK-GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 +; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, s3 ; CHECK-GISEL-NEXT: v_mov_b32_e32 v2, s0 ; CHECK-GISEL-NEXT: flat_store_dwordx2 v[2:3], v[0:1] ; CHECK-GISEL-NEXT: s_endpgm @@ -584,17 +588,17 @@ define amdgpu_kernel void @test_readfirstlane_copy_from_sgpr_i64(ptr addrspace(1 ; CHECK-SDAG: ; %bb.0: ; CHECK-SDAG-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 ; CHECK-SDAG-NEXT: s_add_i32 s12, s12, s17 +; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 +; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 ; CHECK-SDAG-NEXT: ;;#ASMSTART ; CHECK-SDAG-NEXT: s_mov_b64 s[2:3], 0 ; CHECK-SDAG-NEXT: ;;#ASMEND -; CHECK-SDAG-NEXT: v_mov_b32_e32 v2, s2 -; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 ; CHECK-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s0 -; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s1 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v3, s3 -; CHECK-SDAG-NEXT: flat_store_dwordx2 v[0:1], v[2:3] +; CHECK-SDAG-NEXT: v_mov_b32_e32 v3, s1 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s2 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s3 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v2, s0 +; CHECK-SDAG-NEXT: flat_store_dwordx2 v[2:3], v[0:1] ; CHECK-SDAG-NEXT: s_endpgm ; ; CHECK-GISEL-LABEL: test_readfirstlane_copy_from_sgpr_i64: @@ -624,17 +628,17 @@ define amdgpu_kernel void @test_readfirstlane_copy_from_sgpr_f64(ptr addrspace(1 ; CHECK-SDAG: ; %bb.0: ; CHECK-SDAG-NEXT: s_load_dwordx2 s[0:1], s[8:9], 0x0 ; CHECK-SDAG-NEXT: s_add_i32 s12, s12, s17 +; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 +; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 ; CHECK-SDAG-NEXT: ;;#ASMSTART ; CHECK-SDAG-NEXT: s_mov_b64 s[2:3], 0 ; CHECK-SDAG-NEXT: ;;#ASMEND -; CHECK-SDAG-NEXT: v_mov_b32_e32 v2, s2 -; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13 ; CHECK-SDAG-NEXT: s_waitcnt lgkmcnt(0) -; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s0 -; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s1 -; CHECK-SDAG-NEXT: v_mov_b32_e32 v3, s3 -; CHECK-SDAG-NEXT: flat_store_dwordx2 v[0:1], v[2:3] +; CHECK-SDAG-NEXT: v_mov_b32_e32 v3, s1 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s2 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s3 +; CHECK-SDAG-NEXT: v_mov_b32_e32 v2, s0 +; CHECK-SDAG-NEXT: flat_store_dwordx2 v[2:3], v[0:1] ; CHECK-SDAG-NEXT: s_endpgm ; ; CHECK-GISEL-LABEL: test_readfirstlane_copy_from_sgpr_f64: diff --git a/llvm/test/CodeGen/AMDGPU/mfma-loop.ll b/llvm/test/CodeGen/AMDGPU/mfma-loop.ll index 4bb6538..e330c72 100644 --- a/llvm/test/CodeGen/AMDGPU/mfma-loop.ll +++ b/llvm/test/CodeGen/AMDGPU/mfma-loop.ll @@ -1,7 +1,7 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 -; RUN: llc -mtriple=amdgcn -mcpu=gfx908 < %s | FileCheck -enable-var-scope -check-prefixes=GFX908 %s -; RUN: llc -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck -enable-var-scope -check-prefixes=GFX90A %s -; RUN: llc -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck -enable-var-scope -check-prefixes=GFX942 %s +; RUN: llc -verify-machineinstrs -mtriple=amdgcn -mcpu=gfx908 < %s | FileCheck -enable-var-scope -check-prefixes=GFX908 %s +; RUN: llc -verify-machineinstrs -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck -enable-var-scope -check-prefixes=GFX90A %s +; RUN: llc -verify-machineinstrs -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck -enable-var-scope -check-prefixes=GFX942 %s ; Check that we do not copy agprs to vgprs and back inside the loop. diff --git a/llvm/test/CodeGen/AMDGPU/optimize-compare.mir b/llvm/test/CodeGen/AMDGPU/optimize-compare.mir index fba42c4..fa452f3 100644 --- a/llvm/test/CodeGen/AMDGPU/optimize-compare.mir +++ b/llvm/test/CodeGen/AMDGPU/optimize-compare.mir @@ -2277,3 +2277,181 @@ body: | S_ENDPGM 0 ... + +--- +name: s_cselect_b64_s_or_b32_s_cmp_lg_u32_0x00000000 +body: | + ; GCN-LABEL: name: s_cselect_b64_s_or_b32_s_cmp_lg_u32_0x00000000 + ; GCN: bb.0: + ; GCN-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000) + ; GCN-NEXT: liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: [[DEF:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF + ; GCN-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY [[DEF]] + ; GCN-NEXT: S_CMP_LG_U32 [[COPY]], 0, implicit-def $scc + ; GCN-NEXT: [[S_CSELECT_B64_:%[0-9]+]]:sreg_64_xexec = S_CSELECT_B64 -1, 0, implicit $scc + ; GCN-NEXT: [[COPY1:%[0-9]+]]:sreg_32_xm0_xexec = COPY [[S_CSELECT_B64_]].sub0 + ; GCN-NEXT: [[COPY2:%[0-9]+]]:sreg_32_xm0_xexec = COPY [[S_CSELECT_B64_]].sub1 + ; GCN-NEXT: S_CBRANCH_SCC0 %bb.2, implicit $scc + ; GCN-NEXT: S_BRANCH %bb.1 + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: bb.1: + ; GCN-NEXT: successors: %bb.2(0x80000000) + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: bb.2: + ; GCN-NEXT: S_ENDPGM 0 + bb.0: + successors: %bb.1(0x40000000), %bb.2(0x40000000) + liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + %0:vgpr_32 = IMPLICIT_DEF + %2:sreg_32 = COPY %0 + S_CMP_LG_U32 %2, 0, implicit-def $scc + %31:sreg_64_xexec = S_CSELECT_B64 -1, 0, implicit $scc + %40:sreg_32_xm0_xexec = COPY %31.sub0:sreg_64_xexec + %41:sreg_32_xm0_xexec = COPY %31.sub1:sreg_64_xexec + %sgpr4:sreg_32 = S_OR_B32 %40:sreg_32_xm0_xexec, %41:sreg_32_xm0_xexec, implicit-def $scc + S_CMP_LG_U32 %sgpr4, 0, implicit-def $scc + S_CBRANCH_SCC0 %bb.2, implicit $scc + S_BRANCH %bb.1 + + bb.1: + successors: %bb.2(0x80000000) + + bb.2: + S_ENDPGM 0 + +... +--- +# Do not delete s_or_b32 because of intervening def of scc +name: s_cselect_b64_s_or_b32_s_cmp_lg_u32_0x00000000_cant_optimize_intervening +body: | + ; GCN-LABEL: name: s_cselect_b64_s_or_b32_s_cmp_lg_u32_0x00000000_cant_optimize_intervening + ; GCN: bb.0: + ; GCN-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000) + ; GCN-NEXT: liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: [[DEF:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF + ; GCN-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY [[DEF]] + ; GCN-NEXT: S_CMP_LG_U32 [[COPY]], 0, implicit-def $scc + ; GCN-NEXT: [[S_CSELECT_B64_:%[0-9]+]]:sreg_64_xexec = S_CSELECT_B64 -1, 0, implicit $scc + ; GCN-NEXT: S_CMP_LG_U32 [[COPY]], 0, implicit-def $scc + ; GCN-NEXT: [[COPY1:%[0-9]+]]:sreg_32_xm0_xexec = COPY [[S_CSELECT_B64_]].sub0 + ; GCN-NEXT: [[COPY2:%[0-9]+]]:sreg_32_xm0_xexec = COPY [[S_CSELECT_B64_]].sub1 + ; GCN-NEXT: %sgpr4:sreg_32 = S_OR_B32 [[COPY1]], [[COPY2]], implicit-def $scc + ; GCN-NEXT: S_CBRANCH_SCC0 %bb.2, implicit $scc + ; GCN-NEXT: S_BRANCH %bb.1 + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: bb.1: + ; GCN-NEXT: successors: %bb.2(0x80000000) + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: bb.2: + ; GCN-NEXT: S_ENDPGM 0 + bb.0: + successors: %bb.1(0x40000000), %bb.2(0x40000000) + liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + %0:vgpr_32 = IMPLICIT_DEF + %2:sreg_32 = COPY %0 + S_CMP_LG_U32 %2, 0, implicit-def $scc + %31:sreg_64_xexec = S_CSELECT_B64 -1, 0, implicit $scc + S_CMP_LG_U32 %2, 0, implicit-def $scc + %40:sreg_32_xm0_xexec = COPY %31.sub0:sreg_64_xexec + %41:sreg_32_xm0_xexec = COPY %31.sub1:sreg_64_xexec + %sgpr4:sreg_32 = S_OR_B32 %40:sreg_32_xm0_xexec, %41:sreg_32_xm0_xexec, implicit-def $scc + S_CMP_LG_U32 %sgpr4, 0, implicit-def $scc + S_CBRANCH_SCC0 %bb.2, implicit $scc + S_BRANCH %bb.1 + + bb.1: + successors: %bb.2(0x80000000) + + bb.2: + S_ENDPGM 0 + +... + +--- +# Do not delete s_or_b32 since both operands are sub1. +name: s_cselect_b64_s_or_b32_s_cmp_lg_u32_0x00000000_cant_optimize +body: | + ; GCN-LABEL: name: s_cselect_b64_s_or_b32_s_cmp_lg_u32_0x00000000_cant_optimize + ; GCN: bb.0: + ; GCN-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000) + ; GCN-NEXT: liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: [[DEF:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF + ; GCN-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY [[DEF]] + ; GCN-NEXT: S_CMP_LG_U32 [[COPY]], 0, implicit-def $scc + ; GCN-NEXT: [[S_CSELECT_B64_:%[0-9]+]]:sreg_64_xexec = S_CSELECT_B64 1, 0, implicit $scc + ; GCN-NEXT: [[COPY1:%[0-9]+]]:sreg_32_xm0_xexec = COPY [[S_CSELECT_B64_]].sub1 + ; GCN-NEXT: [[COPY2:%[0-9]+]]:sreg_32 = COPY [[S_CSELECT_B64_]].sub1 + ; GCN-NEXT: %sgpr4:sreg_32 = S_OR_B32 [[COPY1]], [[COPY2]], implicit-def $scc + ; GCN-NEXT: S_CBRANCH_SCC0 %bb.2, implicit $scc + ; GCN-NEXT: S_BRANCH %bb.1 + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: bb.1: + ; GCN-NEXT: successors: %bb.2(0x80000000) + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: bb.2: + ; GCN-NEXT: S_ENDPGM 0 + bb.0: + successors: %bb.1(0x40000000), %bb.2(0x40000000) + liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + %0:vgpr_32 = IMPLICIT_DEF + %2:sreg_32 = COPY %0 + S_CMP_LG_U32 %2, 0, implicit-def $scc + %31:sreg_64_xexec = S_CSELECT_B64 1, 0, implicit $scc + %40:sreg_32_xm0_xexec = COPY %31.sub1:sreg_64_xexec + %41:sreg_32 = COPY %31.sub1:sreg_64_xexec + %sgpr4:sreg_32 = S_OR_B32 %40:sreg_32_xm0_xexec, %41:sreg_32, implicit-def $scc + S_CMP_LG_U32 %sgpr4, 0, implicit-def $scc + S_CBRANCH_SCC0 %bb.2, implicit $scc + S_BRANCH %bb.1 + + bb.1: + successors: %bb.2(0x80000000) + + bb.2: + S_ENDPGM 0 + +... + +--- +name: s_cselect_b64_undef_s_or_b32_s_cmp_lg_u32_0x00000000 +body: | + ; GCN-LABEL: name: s_cselect_b64_undef_s_or_b32_s_cmp_lg_u32_0x00000000 + ; GCN: bb.0: + ; GCN-NEXT: successors: %bb.1(0x40000000), %bb.2(0x40000000) + ; GCN-NEXT: liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: [[DEF:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF + ; GCN-NEXT: [[COPY:%[0-9]+]]:sreg_32 = COPY [[DEF]] + ; GCN-NEXT: S_CMP_LG_U32 [[COPY]], 0, implicit-def $scc + ; GCN-NEXT: [[S_CSELECT_B64_:%[0-9]+]]:sreg_64_xexec = S_CSELECT_B64 -1, 0, implicit $scc + ; GCN-NEXT: %sgpr4:sreg_32 = S_OR_B32 undef %4:sreg_32_xm0_xexec, undef %5:sreg_32_xm0_xexec, implicit-def $scc + ; GCN-NEXT: S_CBRANCH_SCC0 %bb.2, implicit $scc + ; GCN-NEXT: S_BRANCH %bb.1 + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: bb.1: + ; GCN-NEXT: successors: %bb.2(0x80000000) + ; GCN-NEXT: {{ $}} + ; GCN-NEXT: bb.2: + ; GCN-NEXT: S_ENDPGM 0 + bb.0: + successors: %bb.1(0x40000000), %bb.2(0x40000000) + liveins: $sgpr0_sgpr1, $vgpr0_vgpr1 + %0:vgpr_32 = IMPLICIT_DEF + %2:sreg_32 = COPY %0 + S_CMP_LG_U32 %2, 0, implicit-def $scc + %31:sreg_64_xexec = S_CSELECT_B64 -1, 0, implicit $scc + %sgpr4:sreg_32 = S_OR_B32 undef %40:sreg_32_xm0_xexec, undef %41:sreg_32_xm0_xexec, implicit-def $scc + S_CMP_LG_U32 %sgpr4, 0, implicit-def $scc + S_CBRANCH_SCC0 %bb.2, implicit $scc + S_BRANCH %bb.1 + + bb.1: + successors: %bb.2(0x80000000) + + bb.2: + S_ENDPGM 0 + +... diff --git a/llvm/test/CodeGen/AMDGPU/rewrite-vgpr-mfma-to-agpr.ll b/llvm/test/CodeGen/AMDGPU/rewrite-vgpr-mfma-to-agpr.ll index 8803f3a..fc79916 100644 --- a/llvm/test/CodeGen/AMDGPU/rewrite-vgpr-mfma-to-agpr.ll +++ b/llvm/test/CodeGen/AMDGPU/rewrite-vgpr-mfma-to-agpr.ll @@ -1,5 +1,5 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5 -; RUN: llc -mcpu=gfx942 -amdgpu-mfma-vgpr-form < %s | FileCheck %s +; RUN: llc -verify-machineinstrs -mcpu=gfx942 -amdgpu-mfma-vgpr-form < %s | FileCheck %s target triple = "amdgcn-amd-amdhsa" diff --git a/llvm/test/CodeGen/AMDGPU/sdiv64.ll b/llvm/test/CodeGen/AMDGPU/sdiv64.ll index 71f5a94..74a6d7f 100644 --- a/llvm/test/CodeGen/AMDGPU/sdiv64.ll +++ b/llvm/test/CodeGen/AMDGPU/sdiv64.ll @@ -8,7 +8,6 @@ define amdgpu_kernel void @s_test_sdiv(ptr addrspace(1) %out, i64 %x, i64 %y) { ; GCN-NEXT: s_load_dwordx2 s[0:1], s[4:5], 0xd ; GCN-NEXT: s_load_dwordx4 s[4:7], s[4:5], 0x9 ; GCN-NEXT: s_mov_b32 s3, 0xf000 -; GCN-NEXT: s_mov_b32 s2, -1 ; GCN-NEXT: s_waitcnt lgkmcnt(0) ; GCN-NEXT: s_ashr_i32 s8, s1, 31 ; GCN-NEXT: s_add_u32 s0, s0, s8 @@ -17,8 +16,8 @@ define amdgpu_kernel void @s_test_sdiv(ptr addrspace(1) %out, i64 %x, i64 %y) { ; GCN-NEXT: s_xor_b64 s[10:11], s[0:1], s[8:9] ; GCN-NEXT: v_cvt_f32_u32_e32 v0, s10 ; GCN-NEXT: v_cvt_f32_u32_e32 v1, s11 -; GCN-NEXT: s_sub_u32 s12, 0, s10 -; GCN-NEXT: s_subb_u32 s13, 0, s11 +; GCN-NEXT: s_sub_u32 s0, 0, s10 +; GCN-NEXT: s_subb_u32 s1, 0, s11 ; GCN-NEXT: v_madmk_f32 v0, v1, 0x4f800000, v0 ; GCN-NEXT: v_rcp_f32_e32 v0, v0 ; GCN-NEXT: v_mul_f32_e32 v0, 0x5f7ffffc, v0 @@ -27,128 +26,121 @@ define amdgpu_kernel void @s_test_sdiv(ptr addrspace(1) %out, i64 %x, i64 %y) { ; GCN-NEXT: v_madmk_f32 v0, v1, 0xcf800000, v0 ; GCN-NEXT: v_cvt_u32_f32_e32 v0, v0 ; GCN-NEXT: v_cvt_u32_f32_e32 v1, v1 -; GCN-NEXT: v_mul_hi_u32 v2, s12, v0 -; GCN-NEXT: v_readfirstlane_b32 s14, v1 -; GCN-NEXT: v_readfirstlane_b32 s0, v0 -; GCN-NEXT: s_mul_i32 s1, s12, s14 -; GCN-NEXT: v_readfirstlane_b32 s17, v2 -; GCN-NEXT: s_mul_i32 s15, s13, s0 -; GCN-NEXT: s_mul_i32 s16, s12, s0 -; GCN-NEXT: s_add_i32 s1, s17, s1 -; GCN-NEXT: v_mul_hi_u32 v3, v0, s16 -; GCN-NEXT: s_add_i32 s1, s1, s15 -; GCN-NEXT: v_mul_hi_u32 v0, v0, s1 -; GCN-NEXT: v_mul_hi_u32 v4, v1, s16 -; GCN-NEXT: v_readfirstlane_b32 s15, v3 -; GCN-NEXT: s_mul_i32 s17, s0, s1 -; GCN-NEXT: v_mul_hi_u32 v1, v1, s1 -; GCN-NEXT: s_add_u32 s15, s15, s17 -; GCN-NEXT: v_readfirstlane_b32 s17, v0 -; GCN-NEXT: s_addc_u32 s17, 0, s17 -; GCN-NEXT: s_mul_i32 s16, s14, s16 -; GCN-NEXT: v_readfirstlane_b32 s18, v4 -; GCN-NEXT: s_add_u32 s15, s15, s16 -; GCN-NEXT: s_addc_u32 s15, s17, s18 -; GCN-NEXT: v_readfirstlane_b32 s16, v1 -; GCN-NEXT: s_addc_u32 s16, s16, 0 -; GCN-NEXT: s_mul_i32 s1, s14, s1 -; GCN-NEXT: s_add_u32 s1, s15, s1 -; GCN-NEXT: s_addc_u32 s15, 0, s16 -; GCN-NEXT: s_add_u32 s16, s0, s1 -; GCN-NEXT: v_mov_b32_e32 v0, s16 -; GCN-NEXT: v_mul_hi_u32 v0, s12, v0 -; GCN-NEXT: s_cselect_b64 s[0:1], -1, 0 -; GCN-NEXT: s_or_b32 s0, s0, s1 -; GCN-NEXT: s_addc_u32 s14, s14, s15 -; GCN-NEXT: s_mul_i32 s0, s12, s14 -; GCN-NEXT: v_readfirstlane_b32 s1, v0 -; GCN-NEXT: s_add_i32 s0, s1, s0 -; GCN-NEXT: s_mul_i32 s13, s13, s16 -; GCN-NEXT: s_mul_i32 s1, s12, s16 -; GCN-NEXT: s_add_i32 s0, s0, s13 -; GCN-NEXT: v_mov_b32_e32 v2, s1 -; GCN-NEXT: v_mov_b32_e32 v0, s0 -; GCN-NEXT: v_mul_hi_u32 v3, s14, v2 -; GCN-NEXT: v_mul_hi_u32 v2, s16, v2 -; GCN-NEXT: v_mul_hi_u32 v1, s14, v0 -; GCN-NEXT: v_mul_hi_u32 v0, s16, v0 -; GCN-NEXT: s_mul_i32 s13, s16, s0 -; GCN-NEXT: v_readfirstlane_b32 s17, v2 -; GCN-NEXT: s_add_u32 s13, s17, s13 -; GCN-NEXT: v_readfirstlane_b32 s15, v0 -; GCN-NEXT: s_mul_i32 s1, s14, s1 -; GCN-NEXT: s_addc_u32 s15, 0, s15 -; GCN-NEXT: v_readfirstlane_b32 s12, v3 -; GCN-NEXT: s_add_u32 s1, s13, s1 -; GCN-NEXT: s_addc_u32 s1, s15, s12 +; GCN-NEXT: v_mul_hi_u32 v2, s0, v0 ; GCN-NEXT: v_readfirstlane_b32 s12, v1 -; GCN-NEXT: s_addc_u32 s12, s12, 0 -; GCN-NEXT: s_mul_i32 s0, s14, s0 -; GCN-NEXT: s_add_u32 s0, s1, s0 -; GCN-NEXT: s_addc_u32 s12, 0, s12 -; GCN-NEXT: s_add_u32 s15, s16, s0 -; GCN-NEXT: s_cselect_b64 s[0:1], -1, 0 -; GCN-NEXT: s_or_b32 s0, s0, s1 -; GCN-NEXT: s_addc_u32 s14, s14, s12 +; GCN-NEXT: v_readfirstlane_b32 s2, v0 +; GCN-NEXT: s_mul_i32 s13, s0, s12 +; GCN-NEXT: v_readfirstlane_b32 s16, v2 +; GCN-NEXT: s_mul_i32 s14, s1, s2 +; GCN-NEXT: s_mul_i32 s15, s0, s2 +; GCN-NEXT: s_add_i32 s13, s16, s13 +; GCN-NEXT: v_mul_hi_u32 v3, v0, s15 +; GCN-NEXT: s_add_i32 s13, s13, s14 +; GCN-NEXT: v_mul_hi_u32 v0, v0, s13 +; GCN-NEXT: v_mul_hi_u32 v4, v1, s15 +; GCN-NEXT: v_readfirstlane_b32 s14, v3 +; GCN-NEXT: s_mul_i32 s16, s2, s13 +; GCN-NEXT: v_mul_hi_u32 v1, v1, s13 +; GCN-NEXT: s_add_u32 s14, s14, s16 +; GCN-NEXT: v_readfirstlane_b32 s16, v0 +; GCN-NEXT: s_mul_i32 s15, s12, s15 +; GCN-NEXT: s_addc_u32 s16, 0, s16 +; GCN-NEXT: v_readfirstlane_b32 s17, v4 +; GCN-NEXT: s_add_u32 s14, s14, s15 +; GCN-NEXT: s_addc_u32 s14, s16, s17 +; GCN-NEXT: v_readfirstlane_b32 s15, v1 +; GCN-NEXT: s_addc_u32 s15, s15, 0 +; GCN-NEXT: s_mul_i32 s13, s12, s13 +; GCN-NEXT: s_add_u32 s13, s14, s13 +; GCN-NEXT: s_addc_u32 s14, 0, s15 +; GCN-NEXT: s_add_u32 s13, s2, s13 +; GCN-NEXT: v_mov_b32_e32 v0, s13 +; GCN-NEXT: v_mul_hi_u32 v0, s0, v0 +; GCN-NEXT: s_addc_u32 s12, s12, s14 +; GCN-NEXT: s_mul_i32 s14, s0, s12 +; GCN-NEXT: s_mul_i32 s1, s1, s13 +; GCN-NEXT: v_readfirstlane_b32 s15, v0 +; GCN-NEXT: s_add_i32 s14, s15, s14 +; GCN-NEXT: s_mul_i32 s0, s0, s13 +; GCN-NEXT: s_add_i32 s1, s14, s1 +; GCN-NEXT: v_mov_b32_e32 v2, s0 +; GCN-NEXT: v_mov_b32_e32 v0, s1 +; GCN-NEXT: v_mul_hi_u32 v3, s12, v2 +; GCN-NEXT: v_mul_hi_u32 v2, s13, v2 +; GCN-NEXT: v_mul_hi_u32 v1, s12, v0 +; GCN-NEXT: v_mul_hi_u32 v0, s13, v0 +; GCN-NEXT: s_mul_i32 s15, s13, s1 +; GCN-NEXT: v_readfirstlane_b32 s17, v2 +; GCN-NEXT: s_add_u32 s15, s17, s15 +; GCN-NEXT: v_readfirstlane_b32 s16, v0 +; GCN-NEXT: s_mul_i32 s0, s12, s0 +; GCN-NEXT: s_addc_u32 s16, 0, s16 +; GCN-NEXT: v_readfirstlane_b32 s14, v3 +; GCN-NEXT: s_add_u32 s0, s15, s0 +; GCN-NEXT: s_addc_u32 s0, s16, s14 +; GCN-NEXT: v_readfirstlane_b32 s14, v1 +; GCN-NEXT: s_addc_u32 s14, s14, 0 +; GCN-NEXT: s_mul_i32 s1, s12, s1 +; GCN-NEXT: s_add_u32 s0, s0, s1 +; GCN-NEXT: s_addc_u32 s1, 0, s14 +; GCN-NEXT: s_add_u32 s14, s13, s0 +; GCN-NEXT: s_addc_u32 s15, s12, s1 ; GCN-NEXT: s_ashr_i32 s12, s7, 31 ; GCN-NEXT: s_add_u32 s0, s6, s12 ; GCN-NEXT: s_mov_b32 s13, s12 ; GCN-NEXT: s_addc_u32 s1, s7, s12 ; GCN-NEXT: s_xor_b64 s[6:7], s[0:1], s[12:13] -; GCN-NEXT: v_mov_b32_e32 v0, s14 +; GCN-NEXT: v_mov_b32_e32 v0, s15 ; GCN-NEXT: v_mul_hi_u32 v1, s6, v0 -; GCN-NEXT: v_mov_b32_e32 v2, s15 +; GCN-NEXT: v_mov_b32_e32 v2, s14 ; GCN-NEXT: v_mul_hi_u32 v3, s6, v2 ; GCN-NEXT: s_mov_b32 s0, s4 ; GCN-NEXT: v_readfirstlane_b32 s4, v1 ; GCN-NEXT: v_mul_hi_u32 v1, s7, v2 -; GCN-NEXT: s_mul_i32 s1, s6, s14 +; GCN-NEXT: s_mul_i32 s1, s6, s15 ; GCN-NEXT: v_readfirstlane_b32 s16, v3 ; GCN-NEXT: v_mul_hi_u32 v0, s7, v0 ; GCN-NEXT: s_add_u32 s1, s16, s1 ; GCN-NEXT: s_addc_u32 s4, 0, s4 -; GCN-NEXT: s_mul_i32 s15, s7, s15 +; GCN-NEXT: s_mul_i32 s14, s7, s14 ; GCN-NEXT: v_readfirstlane_b32 s16, v1 -; GCN-NEXT: s_add_u32 s1, s1, s15 +; GCN-NEXT: s_add_u32 s1, s1, s14 ; GCN-NEXT: s_addc_u32 s1, s4, s16 ; GCN-NEXT: v_readfirstlane_b32 s4, v0 ; GCN-NEXT: s_addc_u32 s4, s4, 0 -; GCN-NEXT: s_mul_i32 s14, s7, s14 -; GCN-NEXT: s_add_u32 s16, s1, s14 -; GCN-NEXT: v_mov_b32_e32 v0, s16 +; GCN-NEXT: s_mul_i32 s14, s7, s15 +; GCN-NEXT: s_add_u32 s14, s1, s14 +; GCN-NEXT: v_mov_b32_e32 v0, s14 ; GCN-NEXT: v_mul_hi_u32 v0, s10, v0 -; GCN-NEXT: s_addc_u32 s17, 0, s4 +; GCN-NEXT: s_addc_u32 s15, 0, s4 ; GCN-NEXT: s_mov_b32 s1, s5 -; GCN-NEXT: s_mul_i32 s4, s10, s17 +; GCN-NEXT: s_mul_i32 s4, s10, s15 ; GCN-NEXT: v_readfirstlane_b32 s5, v0 ; GCN-NEXT: s_add_i32 s4, s5, s4 -; GCN-NEXT: s_mul_i32 s5, s11, s16 -; GCN-NEXT: s_add_i32 s18, s4, s5 -; GCN-NEXT: s_sub_i32 s14, s7, s18 -; GCN-NEXT: s_mul_i32 s4, s10, s16 +; GCN-NEXT: s_mul_i32 s5, s11, s14 +; GCN-NEXT: s_add_i32 s16, s4, s5 +; GCN-NEXT: s_sub_i32 s17, s7, s16 +; GCN-NEXT: s_mul_i32 s4, s10, s14 ; GCN-NEXT: s_sub_u32 s6, s6, s4 ; GCN-NEXT: s_cselect_b64 s[4:5], -1, 0 -; GCN-NEXT: s_or_b32 s15, s4, s5 -; GCN-NEXT: s_subb_u32 s19, s14, s11 -; GCN-NEXT: s_sub_u32 s20, s6, s10 -; GCN-NEXT: s_cselect_b64 s[14:15], -1, 0 -; GCN-NEXT: s_or_b32 s14, s14, s15 -; GCN-NEXT: s_subb_u32 s14, s19, 0 -; GCN-NEXT: s_cmp_ge_u32 s14, s11 -; GCN-NEXT: s_cselect_b32 s15, -1, 0 -; GCN-NEXT: s_cmp_ge_u32 s20, s10 +; GCN-NEXT: s_subb_u32 s17, s17, s11 +; GCN-NEXT: s_sub_u32 s18, s6, s10 +; GCN-NEXT: s_subb_u32 s17, s17, 0 +; GCN-NEXT: s_cmp_ge_u32 s17, s11 ; GCN-NEXT: s_cselect_b32 s19, -1, 0 -; GCN-NEXT: s_cmp_eq_u32 s14, s11 -; GCN-NEXT: s_cselect_b32 s14, s19, s15 -; GCN-NEXT: s_add_u32 s15, s16, 1 -; GCN-NEXT: s_addc_u32 s19, s17, 0 -; GCN-NEXT: s_add_u32 s20, s16, 2 -; GCN-NEXT: s_addc_u32 s21, s17, 0 -; GCN-NEXT: s_cmp_lg_u32 s14, 0 -; GCN-NEXT: s_cselect_b32 s14, s20, s15 -; GCN-NEXT: s_cselect_b32 s15, s21, s19 +; GCN-NEXT: s_cmp_ge_u32 s18, s10 +; GCN-NEXT: s_cselect_b32 s18, -1, 0 +; GCN-NEXT: s_cmp_eq_u32 s17, s11 +; GCN-NEXT: s_cselect_b32 s17, s18, s19 +; GCN-NEXT: s_add_u32 s18, s14, 1 +; GCN-NEXT: s_addc_u32 s19, s15, 0 +; GCN-NEXT: s_add_u32 s20, s14, 2 +; GCN-NEXT: s_addc_u32 s21, s15, 0 +; GCN-NEXT: s_cmp_lg_u32 s17, 0 +; GCN-NEXT: s_cselect_b32 s17, s20, s18 +; GCN-NEXT: s_cselect_b32 s18, s21, s19 ; GCN-NEXT: s_or_b32 s4, s4, s5 -; GCN-NEXT: s_subb_u32 s4, s7, s18 +; GCN-NEXT: s_subb_u32 s4, s7, s16 ; GCN-NEXT: s_cmp_ge_u32 s4, s11 ; GCN-NEXT: s_cselect_b32 s5, -1, 0 ; GCN-NEXT: s_cmp_ge_u32 s6, s10 @@ -156,13 +148,14 @@ define amdgpu_kernel void @s_test_sdiv(ptr addrspace(1) %out, i64 %x, i64 %y) { ; GCN-NEXT: s_cmp_eq_u32 s4, s11 ; GCN-NEXT: s_cselect_b32 s4, s6, s5 ; GCN-NEXT: s_cmp_lg_u32 s4, 0 -; GCN-NEXT: s_cselect_b32 s5, s15, s17 -; GCN-NEXT: s_cselect_b32 s4, s14, s16 +; GCN-NEXT: s_cselect_b32 s5, s18, s15 +; GCN-NEXT: s_cselect_b32 s4, s17, s14 ; GCN-NEXT: s_xor_b64 s[6:7], s[12:13], s[8:9] ; GCN-NEXT: s_xor_b64 s[4:5], s[4:5], s[6:7] ; GCN-NEXT: s_sub_u32 s4, s4, s6 ; GCN-NEXT: s_subb_u32 s5, s5, s7 ; GCN-NEXT: v_mov_b32_e32 v0, s4 +; GCN-NEXT: s_mov_b32 s2, -1 ; GCN-NEXT: v_mov_b32_e32 v1, s5 ; GCN-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 ; GCN-NEXT: s_endpgm @@ -202,8 +195,6 @@ define amdgpu_kernel void @s_test_sdiv(ptr addrspace(1) %out, i64 %x, i64 %y) { ; GCN-IR-NEXT: s_cbranch_vccz .LBB0_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: s_add_u32 s18, s16, 1 -; GCN-IR-NEXT: s_cselect_b64 s[10:11], -1, 0 -; GCN-IR-NEXT: s_or_b32 s10, s10, s11 ; GCN-IR-NEXT: s_addc_u32 s10, s17, 0 ; GCN-IR-NEXT: s_cselect_b64 s[10:11], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s16, 63, s16 @@ -235,8 +226,6 @@ define amdgpu_kernel void @s_test_sdiv(ptr addrspace(1) %out, i64 %x, i64 %y) { ; GCN-IR-NEXT: s_sub_u32 s16, s16, s20 ; GCN-IR-NEXT: s_subb_u32 s17, s17, s21 ; GCN-IR-NEXT: s_add_u32 s14, s14, 1 -; GCN-IR-NEXT: s_cselect_b64 s[20:21], -1, 0 -; GCN-IR-NEXT: s_or_b32 s20, s20, s21 ; GCN-IR-NEXT: s_addc_u32 s15, s15, 0 ; GCN-IR-NEXT: s_cselect_b64 s[20:21], -1, 0 ; GCN-IR-NEXT: s_mov_b64 s[12:13], s[8:9] @@ -1150,8 +1139,7 @@ define amdgpu_kernel void @s_test_sdiv_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-NEXT: v_cvt_f32_u32_e32 v0, s6 ; GCN-NEXT: v_cvt_f32_u32_e32 v1, s7 ; GCN-NEXT: s_sub_u32 s2, 0, s6 -; GCN-NEXT: s_subb_u32 s10, 0, s7 -; GCN-NEXT: s_mov_b32 s3, 0xf000 +; GCN-NEXT: s_subb_u32 s8, 0, s7 ; GCN-NEXT: v_madmk_f32 v0, v1, 0x4f800000, v0 ; GCN-NEXT: v_rcp_f32_e32 v0, v0 ; GCN-NEXT: v_mul_f32_e32 v0, 0x5f7ffffc, v0 @@ -1161,115 +1149,109 @@ define amdgpu_kernel void @s_test_sdiv_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-NEXT: v_cvt_u32_f32_e32 v0, v0 ; GCN-NEXT: v_cvt_u32_f32_e32 v1, v1 ; GCN-NEXT: v_mul_hi_u32 v2, s2, v0 -; GCN-NEXT: v_readfirstlane_b32 s11, v1 -; GCN-NEXT: v_readfirstlane_b32 s8, v0 -; GCN-NEXT: s_mul_i32 s9, s2, s11 -; GCN-NEXT: v_readfirstlane_b32 s14, v2 -; GCN-NEXT: s_mul_i32 s12, s10, s8 -; GCN-NEXT: s_mul_i32 s13, s2, s8 -; GCN-NEXT: s_add_i32 s9, s14, s9 -; GCN-NEXT: v_mul_hi_u32 v3, v0, s13 -; GCN-NEXT: s_add_i32 s9, s9, s12 -; GCN-NEXT: v_mul_hi_u32 v0, v0, s9 -; GCN-NEXT: v_mul_hi_u32 v4, v1, s13 -; GCN-NEXT: v_readfirstlane_b32 s12, v3 -; GCN-NEXT: s_mul_i32 s15, s8, s9 -; GCN-NEXT: v_mul_hi_u32 v1, v1, s9 -; GCN-NEXT: s_add_u32 s12, s12, s15 -; GCN-NEXT: v_readfirstlane_b32 s15, v0 -; GCN-NEXT: s_mul_i32 s13, s11, s13 -; GCN-NEXT: s_addc_u32 s15, 0, s15 -; GCN-NEXT: v_readfirstlane_b32 s14, v4 -; GCN-NEXT: s_add_u32 s12, s12, s13 -; GCN-NEXT: s_addc_u32 s12, s15, s14 -; GCN-NEXT: v_readfirstlane_b32 s13, v1 -; GCN-NEXT: s_addc_u32 s13, s13, 0 -; GCN-NEXT: s_mul_i32 s9, s11, s9 -; GCN-NEXT: s_add_u32 s9, s12, s9 -; GCN-NEXT: s_addc_u32 s12, 0, s13 -; GCN-NEXT: s_add_u32 s13, s8, s9 -; GCN-NEXT: v_mov_b32_e32 v0, s13 +; GCN-NEXT: v_readfirstlane_b32 s9, v1 +; GCN-NEXT: v_readfirstlane_b32 s3, v0 +; GCN-NEXT: s_mul_i32 s10, s2, s9 +; GCN-NEXT: v_readfirstlane_b32 s13, v2 +; GCN-NEXT: s_mul_i32 s11, s8, s3 +; GCN-NEXT: s_mul_i32 s12, s2, s3 +; GCN-NEXT: s_add_i32 s10, s13, s10 +; GCN-NEXT: v_mul_hi_u32 v3, v0, s12 +; GCN-NEXT: s_add_i32 s10, s10, s11 +; GCN-NEXT: v_mul_hi_u32 v0, v0, s10 +; GCN-NEXT: v_mul_hi_u32 v4, v1, s12 +; GCN-NEXT: v_readfirstlane_b32 s11, v3 +; GCN-NEXT: v_mul_hi_u32 v1, v1, s10 +; GCN-NEXT: s_mul_i32 s14, s3, s10 +; GCN-NEXT: s_add_u32 s11, s11, s14 +; GCN-NEXT: v_readfirstlane_b32 s14, v0 +; GCN-NEXT: s_mul_i32 s12, s9, s12 +; GCN-NEXT: s_addc_u32 s14, 0, s14 +; GCN-NEXT: v_readfirstlane_b32 s13, v4 +; GCN-NEXT: s_add_u32 s11, s11, s12 +; GCN-NEXT: v_readfirstlane_b32 s15, v1 +; GCN-NEXT: s_addc_u32 s11, s14, s13 +; GCN-NEXT: s_addc_u32 s12, s15, 0 +; GCN-NEXT: s_mul_i32 s10, s9, s10 +; GCN-NEXT: s_add_u32 s10, s11, s10 +; GCN-NEXT: s_addc_u32 s11, 0, s12 +; GCN-NEXT: s_add_u32 s10, s3, s10 +; GCN-NEXT: v_mov_b32_e32 v0, s10 ; GCN-NEXT: v_mul_hi_u32 v0, s2, v0 -; GCN-NEXT: s_cselect_b64 s[8:9], -1, 0 -; GCN-NEXT: s_or_b32 s8, s8, s9 -; GCN-NEXT: s_addc_u32 s11, s11, s12 -; GCN-NEXT: s_mul_i32 s8, s2, s11 -; GCN-NEXT: v_readfirstlane_b32 s9, v0 -; GCN-NEXT: s_add_i32 s8, s9, s8 -; GCN-NEXT: s_mul_i32 s10, s10, s13 -; GCN-NEXT: s_mul_i32 s2, s2, s13 -; GCN-NEXT: s_add_i32 s8, s8, s10 +; GCN-NEXT: s_addc_u32 s9, s9, s11 +; GCN-NEXT: s_mul_i32 s11, s2, s9 +; GCN-NEXT: s_mul_i32 s8, s8, s10 +; GCN-NEXT: v_readfirstlane_b32 s12, v0 +; GCN-NEXT: s_add_i32 s11, s12, s11 +; GCN-NEXT: s_mul_i32 s2, s2, s10 +; GCN-NEXT: s_add_i32 s8, s11, s8 ; GCN-NEXT: v_mov_b32_e32 v2, s2 ; GCN-NEXT: v_mov_b32_e32 v0, s8 -; GCN-NEXT: v_mul_hi_u32 v3, s11, v2 -; GCN-NEXT: v_mul_hi_u32 v2, s13, v2 -; GCN-NEXT: v_mul_hi_u32 v1, s11, v0 -; GCN-NEXT: v_mul_hi_u32 v0, s13, v0 -; GCN-NEXT: s_mul_i32 s10, s13, s8 +; GCN-NEXT: v_mul_hi_u32 v3, s9, v2 +; GCN-NEXT: v_mul_hi_u32 v2, s10, v2 +; GCN-NEXT: v_mul_hi_u32 v1, s9, v0 +; GCN-NEXT: v_mul_hi_u32 v0, s10, v0 +; GCN-NEXT: s_mul_i32 s12, s10, s8 ; GCN-NEXT: v_readfirstlane_b32 s14, v2 -; GCN-NEXT: s_add_u32 s10, s14, s10 -; GCN-NEXT: v_readfirstlane_b32 s12, v0 -; GCN-NEXT: s_mul_i32 s2, s11, s2 -; GCN-NEXT: s_addc_u32 s12, 0, s12 -; GCN-NEXT: v_readfirstlane_b32 s9, v3 -; GCN-NEXT: s_add_u32 s2, s10, s2 -; GCN-NEXT: s_addc_u32 s2, s12, s9 -; GCN-NEXT: v_readfirstlane_b32 s9, v1 -; GCN-NEXT: s_addc_u32 s9, s9, 0 -; GCN-NEXT: s_mul_i32 s8, s11, s8 +; GCN-NEXT: s_add_u32 s12, s14, s12 +; GCN-NEXT: v_readfirstlane_b32 s13, v0 +; GCN-NEXT: s_mul_i32 s2, s9, s2 +; GCN-NEXT: s_addc_u32 s13, 0, s13 +; GCN-NEXT: v_readfirstlane_b32 s11, v3 +; GCN-NEXT: s_add_u32 s2, s12, s2 +; GCN-NEXT: s_addc_u32 s2, s13, s11 +; GCN-NEXT: v_readfirstlane_b32 s11, v1 +; GCN-NEXT: s_addc_u32 s11, s11, 0 +; GCN-NEXT: s_mul_i32 s8, s9, s8 ; GCN-NEXT: s_add_u32 s2, s2, s8 -; GCN-NEXT: s_addc_u32 s10, 0, s9 -; GCN-NEXT: s_add_u32 s2, s13, s2 -; GCN-NEXT: s_cselect_b64 s[8:9], -1, 0 -; GCN-NEXT: s_or_b32 s8, s8, s9 -; GCN-NEXT: s_addc_u32 s8, s11, s10 +; GCN-NEXT: s_addc_u32 s8, 0, s11 +; GCN-NEXT: s_add_u32 s2, s10, s2 +; GCN-NEXT: s_addc_u32 s8, s9, s8 ; GCN-NEXT: v_mul_hi_u32 v1, s2, 24 ; GCN-NEXT: v_mul_hi_u32 v0, s8, 24 ; GCN-NEXT: s_mul_i32 s8, s8, 24 -; GCN-NEXT: s_mov_b32 s2, -1 +; GCN-NEXT: s_mov_b32 s3, 0xf000 ; GCN-NEXT: v_readfirstlane_b32 s10, v1 ; GCN-NEXT: v_readfirstlane_b32 s9, v0 ; GCN-NEXT: s_add_u32 s8, s10, s8 -; GCN-NEXT: s_addc_u32 s12, 0, s9 -; GCN-NEXT: v_mov_b32_e32 v0, s12 +; GCN-NEXT: s_addc_u32 s10, 0, s9 +; GCN-NEXT: v_mov_b32_e32 v0, s10 ; GCN-NEXT: v_mul_hi_u32 v0, s6, v0 -; GCN-NEXT: s_mul_i32 s8, s7, s12 +; GCN-NEXT: s_mul_i32 s8, s7, s10 +; GCN-NEXT: s_mov_b32 s2, -1 ; GCN-NEXT: v_readfirstlane_b32 s9, v0 -; GCN-NEXT: s_add_i32 s13, s9, s8 -; GCN-NEXT: s_sub_i32 s10, 0, s13 -; GCN-NEXT: s_mul_i32 s8, s6, s12 -; GCN-NEXT: s_sub_u32 s14, 24, s8 +; GCN-NEXT: s_add_i32 s11, s9, s8 +; GCN-NEXT: s_sub_i32 s12, 0, s11 +; GCN-NEXT: s_mul_i32 s8, s6, s10 +; GCN-NEXT: s_sub_u32 s13, 24, s8 ; GCN-NEXT: s_cselect_b64 s[8:9], -1, 0 -; GCN-NEXT: s_or_b32 s11, s8, s9 -; GCN-NEXT: s_subb_u32 s15, s10, s7 -; GCN-NEXT: s_sub_u32 s16, s14, s6 -; GCN-NEXT: s_cselect_b64 s[10:11], -1, 0 -; GCN-NEXT: s_or_b32 s10, s10, s11 -; GCN-NEXT: s_subb_u32 s10, s15, 0 -; GCN-NEXT: s_cmp_ge_u32 s10, s7 -; GCN-NEXT: s_cselect_b32 s11, -1, 0 -; GCN-NEXT: s_cmp_ge_u32 s16, s6 +; GCN-NEXT: s_subb_u32 s12, s12, s7 +; GCN-NEXT: s_sub_u32 s14, s13, s6 +; GCN-NEXT: s_subb_u32 s12, s12, 0 +; GCN-NEXT: s_cmp_ge_u32 s12, s7 ; GCN-NEXT: s_cselect_b32 s15, -1, 0 -; GCN-NEXT: s_cmp_eq_u32 s10, s7 -; GCN-NEXT: s_cselect_b32 s10, s15, s11 -; GCN-NEXT: s_add_u32 s11, s12, 1 +; GCN-NEXT: s_cmp_ge_u32 s14, s6 +; GCN-NEXT: s_cselect_b32 s14, -1, 0 +; GCN-NEXT: s_cmp_eq_u32 s12, s7 +; GCN-NEXT: s_cselect_b32 s12, s14, s15 +; GCN-NEXT: s_add_u32 s14, s10, 1 ; GCN-NEXT: s_addc_u32 s15, 0, 0 -; GCN-NEXT: s_add_u32 s16, s12, 2 +; GCN-NEXT: s_add_u32 s16, s10, 2 ; GCN-NEXT: s_addc_u32 s17, 0, 0 -; GCN-NEXT: s_cmp_lg_u32 s10, 0 -; GCN-NEXT: s_cselect_b32 s10, s16, s11 -; GCN-NEXT: s_cselect_b32 s11, s17, s15 +; GCN-NEXT: s_cmp_lg_u32 s12, 0 +; GCN-NEXT: s_cselect_b32 s12, s16, s14 +; GCN-NEXT: s_cselect_b32 s14, s17, s15 ; GCN-NEXT: s_or_b32 s8, s8, s9 -; GCN-NEXT: s_subb_u32 s8, 0, s13 +; GCN-NEXT: s_subb_u32 s8, 0, s11 ; GCN-NEXT: s_cmp_ge_u32 s8, s7 ; GCN-NEXT: s_cselect_b32 s9, -1, 0 -; GCN-NEXT: s_cmp_ge_u32 s14, s6 +; GCN-NEXT: s_cmp_ge_u32 s13, s6 ; GCN-NEXT: s_cselect_b32 s6, -1, 0 ; GCN-NEXT: s_cmp_eq_u32 s8, s7 ; GCN-NEXT: s_cselect_b32 s6, s6, s9 ; GCN-NEXT: s_cmp_lg_u32 s6, 0 -; GCN-NEXT: s_cselect_b32 s7, s11, 0 -; GCN-NEXT: s_cselect_b32 s6, s10, s12 +; GCN-NEXT: s_cselect_b32 s7, s14, 0 +; GCN-NEXT: s_cselect_b32 s6, s12, s10 ; GCN-NEXT: s_xor_b64 s[6:7], s[6:7], s[4:5] ; GCN-NEXT: s_sub_u32 s6, s6, s4 ; GCN-NEXT: s_subb_u32 s7, s7, s4 @@ -1303,8 +1285,6 @@ define amdgpu_kernel void @s_test_sdiv_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR-NEXT: s_cbranch_vccz .LBB10_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: s_add_u32 s12, s10, 1 -; GCN-IR-NEXT: s_cselect_b64 s[8:9], -1, 0 -; GCN-IR-NEXT: s_or_b32 s8, s8, s9 ; GCN-IR-NEXT: s_addc_u32 s8, s11, 0 ; GCN-IR-NEXT: s_cselect_b64 s[8:9], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s10, 63, s10 @@ -1335,8 +1315,6 @@ define amdgpu_kernel void @s_test_sdiv_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR-NEXT: s_sub_u32 s12, s12, s18 ; GCN-IR-NEXT: s_subb_u32 s13, s13, s19 ; GCN-IR-NEXT: s_add_u32 s16, s16, 1 -; GCN-IR-NEXT: s_cselect_b64 s[18:19], -1, 0 -; GCN-IR-NEXT: s_or_b32 s18, s18, s19 ; GCN-IR-NEXT: s_addc_u32 s17, s17, 0 ; GCN-IR-NEXT: s_cselect_b64 s[18:19], -1, 0 ; GCN-IR-NEXT: s_mov_b64 s[10:11], s[6:7] diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/shlN_add.ll b/llvm/test/CodeGen/AMDGPU/shlN_add.ll index 9f4a6f2..3e507a0 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/shlN_add.ll +++ b/llvm/test/CodeGen/AMDGPU/shlN_add.ll @@ -1,4 +1,9 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py +; RUN: llc -mtriple=amdgcn -mcpu=gfx900 < %s | FileCheck -check-prefixes=GCN,GFX9-SDAG %s +; RUN: llc -mtriple=amdgcn -mcpu=fiji < %s | FileCheck -check-prefixes=GCN,GFX8-SDAG %s +; RUN: llc -mtriple=amdgcn -mcpu=gfx1010 < %s | FileCheck -check-prefixes=GCN,GFX10-SDAG %s +; RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -amdgpu-enable-delay-alu=0 < %s | FileCheck -check-prefixes=GCN,GFX10-SDAG %s + ; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx900 < %s | FileCheck -check-prefixes=GCN,GFX9 %s ; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn -mcpu=fiji < %s | FileCheck -check-prefixes=GCN,GFX8 %s ; RUN: llc -global-isel -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx1010 < %s | FileCheck -check-prefixes=GCN,GFX10 %s @@ -7,6 +12,24 @@ ; Test gfx9+ s_shl[1-4]_add_u32 pattern matching define amdgpu_ps i32 @s_shl1_add_u32(i32 inreg %src0, i32 inreg %src1) { +; GFX9-SDAG-LABEL: s_shl1_add_u32: +; GFX9-SDAG: ; %bb.0: +; GFX9-SDAG-NEXT: s_lshl_b32 s0, s0, 1 +; GFX9-SDAG-NEXT: s_add_i32 s0, s0, s1 +; GFX9-SDAG-NEXT: ; return to shader part epilog +; +; GFX8-SDAG-LABEL: s_shl1_add_u32: +; GFX8-SDAG: ; %bb.0: +; GFX8-SDAG-NEXT: s_lshl_b32 s0, s0, 1 +; GFX8-SDAG-NEXT: s_add_i32 s0, s0, s1 +; GFX8-SDAG-NEXT: ; return to shader part epilog +; +; GFX10-SDAG-LABEL: s_shl1_add_u32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_lshl_b32 s0, s0, 1 +; GFX10-SDAG-NEXT: s_add_i32 s0, s0, s1 +; GFX10-SDAG-NEXT: ; return to shader part epilog +; ; GFX9-LABEL: s_shl1_add_u32: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_lshl1_add_u32 s0, s0, s1 @@ -28,6 +51,24 @@ define amdgpu_ps i32 @s_shl1_add_u32(i32 inreg %src0, i32 inreg %src1) { } define amdgpu_ps i32 @s_shl2_add_u32(i32 inreg %src0, i32 inreg %src1) { +; GFX9-SDAG-LABEL: s_shl2_add_u32: +; GFX9-SDAG: ; %bb.0: +; GFX9-SDAG-NEXT: s_lshl_b32 s0, s0, 2 +; GFX9-SDAG-NEXT: s_add_i32 s0, s0, s1 +; GFX9-SDAG-NEXT: ; return to shader part epilog +; +; GFX8-SDAG-LABEL: s_shl2_add_u32: +; GFX8-SDAG: ; %bb.0: +; GFX8-SDAG-NEXT: s_lshl_b32 s0, s0, 2 +; GFX8-SDAG-NEXT: s_add_i32 s0, s0, s1 +; GFX8-SDAG-NEXT: ; return to shader part epilog +; +; GFX10-SDAG-LABEL: s_shl2_add_u32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_lshl_b32 s0, s0, 2 +; GFX10-SDAG-NEXT: s_add_i32 s0, s0, s1 +; GFX10-SDAG-NEXT: ; return to shader part epilog +; ; GFX9-LABEL: s_shl2_add_u32: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_lshl2_add_u32 s0, s0, s1 @@ -49,6 +90,24 @@ define amdgpu_ps i32 @s_shl2_add_u32(i32 inreg %src0, i32 inreg %src1) { } define amdgpu_ps i32 @s_shl3_add_u32(i32 inreg %src0, i32 inreg %src1) { +; GFX9-SDAG-LABEL: s_shl3_add_u32: +; GFX9-SDAG: ; %bb.0: +; GFX9-SDAG-NEXT: s_lshl_b32 s0, s0, 3 +; GFX9-SDAG-NEXT: s_add_i32 s0, s0, s1 +; GFX9-SDAG-NEXT: ; return to shader part epilog +; +; GFX8-SDAG-LABEL: s_shl3_add_u32: +; GFX8-SDAG: ; %bb.0: +; GFX8-SDAG-NEXT: s_lshl_b32 s0, s0, 3 +; GFX8-SDAG-NEXT: s_add_i32 s0, s0, s1 +; GFX8-SDAG-NEXT: ; return to shader part epilog +; +; GFX10-SDAG-LABEL: s_shl3_add_u32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_lshl_b32 s0, s0, 3 +; GFX10-SDAG-NEXT: s_add_i32 s0, s0, s1 +; GFX10-SDAG-NEXT: ; return to shader part epilog +; ; GFX9-LABEL: s_shl3_add_u32: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_lshl3_add_u32 s0, s0, s1 @@ -70,6 +129,24 @@ define amdgpu_ps i32 @s_shl3_add_u32(i32 inreg %src0, i32 inreg %src1) { } define amdgpu_ps i32 @s_shl4_add_u32(i32 inreg %src0, i32 inreg %src1) { +; GFX9-SDAG-LABEL: s_shl4_add_u32: +; GFX9-SDAG: ; %bb.0: +; GFX9-SDAG-NEXT: s_lshl_b32 s0, s0, 4 +; GFX9-SDAG-NEXT: s_add_i32 s0, s0, s1 +; GFX9-SDAG-NEXT: ; return to shader part epilog +; +; GFX8-SDAG-LABEL: s_shl4_add_u32: +; GFX8-SDAG: ; %bb.0: +; GFX8-SDAG-NEXT: s_lshl_b32 s0, s0, 4 +; GFX8-SDAG-NEXT: s_add_i32 s0, s0, s1 +; GFX8-SDAG-NEXT: ; return to shader part epilog +; +; GFX10-SDAG-LABEL: s_shl4_add_u32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_lshl_b32 s0, s0, 4 +; GFX10-SDAG-NEXT: s_add_i32 s0, s0, s1 +; GFX10-SDAG-NEXT: ; return to shader part epilog +; ; GFX9-LABEL: s_shl4_add_u32: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_lshl4_add_u32 s0, s0, s1 @@ -102,6 +179,25 @@ define amdgpu_ps i32 @s_shl5_add_u32(i32 inreg %src0, i32 inreg %src1) { } define i32 @v_shl1_add_u32(i32 %src0, i32 %src1) { +; GFX9-SDAG-LABEL: v_shl1_add_u32: +; GFX9-SDAG: ; %bb.0: +; GFX9-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-SDAG-NEXT: v_lshl_add_u32 v0, v0, 1, v1 +; GFX9-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-SDAG-LABEL: v_shl1_add_u32: +; GFX8-SDAG: ; %bb.0: +; GFX8-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-SDAG-NEXT: v_lshlrev_b32_e32 v0, 1, v0 +; GFX8-SDAG-NEXT: v_add_u32_e32 v0, vcc, v0, v1 +; GFX8-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-SDAG-LABEL: v_shl1_add_u32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: v_lshl_add_u32 v0, v0, 1, v1 +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; ; GFX9-LABEL: v_shl1_add_u32: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) @@ -126,6 +222,25 @@ define i32 @v_shl1_add_u32(i32 %src0, i32 %src1) { } define i32 @v_shl2_add_u32(i32 %src0, i32 %src1) { +; GFX9-SDAG-LABEL: v_shl2_add_u32: +; GFX9-SDAG: ; %bb.0: +; GFX9-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-SDAG-NEXT: v_lshl_add_u32 v0, v0, 2, v1 +; GFX9-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-SDAG-LABEL: v_shl2_add_u32: +; GFX8-SDAG: ; %bb.0: +; GFX8-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-SDAG-NEXT: v_lshlrev_b32_e32 v0, 2, v0 +; GFX8-SDAG-NEXT: v_add_u32_e32 v0, vcc, v0, v1 +; GFX8-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-SDAG-LABEL: v_shl2_add_u32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: v_lshl_add_u32 v0, v0, 2, v1 +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; ; GFX9-LABEL: v_shl2_add_u32: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) @@ -150,6 +265,25 @@ define i32 @v_shl2_add_u32(i32 %src0, i32 %src1) { } define i32 @v_shl3_add_u32(i32 %src0, i32 %src1) { +; GFX9-SDAG-LABEL: v_shl3_add_u32: +; GFX9-SDAG: ; %bb.0: +; GFX9-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-SDAG-NEXT: v_lshl_add_u32 v0, v0, 3, v1 +; GFX9-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-SDAG-LABEL: v_shl3_add_u32: +; GFX8-SDAG: ; %bb.0: +; GFX8-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-SDAG-NEXT: v_lshlrev_b32_e32 v0, 3, v0 +; GFX8-SDAG-NEXT: v_add_u32_e32 v0, vcc, v0, v1 +; GFX8-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-SDAG-LABEL: v_shl3_add_u32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: v_lshl_add_u32 v0, v0, 3, v1 +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; ; GFX9-LABEL: v_shl3_add_u32: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) @@ -174,6 +308,25 @@ define i32 @v_shl3_add_u32(i32 %src0, i32 %src1) { } define i32 @v_shl4_add_u32(i32 %src0, i32 %src1) { +; GFX9-SDAG-LABEL: v_shl4_add_u32: +; GFX9-SDAG: ; %bb.0: +; GFX9-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-SDAG-NEXT: v_lshl_add_u32 v0, v0, 4, v1 +; GFX9-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-SDAG-LABEL: v_shl4_add_u32: +; GFX8-SDAG: ; %bb.0: +; GFX8-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-SDAG-NEXT: v_lshlrev_b32_e32 v0, 4, v0 +; GFX8-SDAG-NEXT: v_add_u32_e32 v0, vcc, v0, v1 +; GFX8-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-SDAG-LABEL: v_shl4_add_u32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: v_lshl_add_u32 v0, v0, 4, v1 +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; ; GFX9-LABEL: v_shl4_add_u32: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) @@ -198,6 +351,25 @@ define i32 @v_shl4_add_u32(i32 %src0, i32 %src1) { } define i32 @v_shl5_add_u32(i32 %src0, i32 %src1) { +; GFX9-SDAG-LABEL: v_shl5_add_u32: +; GFX9-SDAG: ; %bb.0: +; GFX9-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX9-SDAG-NEXT: v_lshl_add_u32 v0, v0, 5, v1 +; GFX9-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX8-SDAG-LABEL: v_shl5_add_u32: +; GFX8-SDAG: ; %bb.0: +; GFX8-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX8-SDAG-NEXT: v_lshlrev_b32_e32 v0, 5, v0 +; GFX8-SDAG-NEXT: v_add_u32_e32 v0, vcc, v0, v1 +; GFX8-SDAG-NEXT: s_setpc_b64 s[30:31] +; +; GFX10-SDAG-LABEL: v_shl5_add_u32: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) +; GFX10-SDAG-NEXT: v_lshl_add_u32 v0, v0, 5, v1 +; GFX10-SDAG-NEXT: s_setpc_b64 s[30:31] +; ; GFX9-LABEL: v_shl5_add_u32: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0) @@ -224,6 +396,22 @@ define i32 @v_shl5_add_u32(i32 %src0, i32 %src1) { ; FIXME: Use v_lshl_add_u32 ; shift is scalar, but add is vector. define amdgpu_ps float @shl1_add_u32_vgpr1(i32 inreg %src0, i32 %src1) { +; GFX9-SDAG-LABEL: shl1_add_u32_vgpr1: +; GFX9-SDAG: ; %bb.0: +; GFX9-SDAG-NEXT: v_lshl_add_u32 v0, s0, 1, v0 +; GFX9-SDAG-NEXT: ; return to shader part epilog +; +; GFX8-SDAG-LABEL: shl1_add_u32_vgpr1: +; GFX8-SDAG: ; %bb.0: +; GFX8-SDAG-NEXT: s_lshl_b32 s0, s0, 1 +; GFX8-SDAG-NEXT: v_add_u32_e32 v0, vcc, s0, v0 +; GFX8-SDAG-NEXT: ; return to shader part epilog +; +; GFX10-SDAG-LABEL: shl1_add_u32_vgpr1: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: v_lshl_add_u32 v0, s0, 1, v0 +; GFX10-SDAG-NEXT: ; return to shader part epilog +; ; GFX9-LABEL: shl1_add_u32_vgpr1: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_lshl_b32 s0, s0, 1 @@ -248,6 +436,22 @@ define amdgpu_ps float @shl1_add_u32_vgpr1(i32 inreg %src0, i32 %src1) { } define amdgpu_ps float @shl2_add_u32_vgpr1(i32 inreg %src0, i32 %src1) { +; GFX9-SDAG-LABEL: shl2_add_u32_vgpr1: +; GFX9-SDAG: ; %bb.0: +; GFX9-SDAG-NEXT: v_lshl_add_u32 v0, s0, 2, v0 +; GFX9-SDAG-NEXT: ; return to shader part epilog +; +; GFX8-SDAG-LABEL: shl2_add_u32_vgpr1: +; GFX8-SDAG: ; %bb.0: +; GFX8-SDAG-NEXT: s_lshl_b32 s0, s0, 2 +; GFX8-SDAG-NEXT: v_add_u32_e32 v0, vcc, s0, v0 +; GFX8-SDAG-NEXT: ; return to shader part epilog +; +; GFX10-SDAG-LABEL: shl2_add_u32_vgpr1: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: v_lshl_add_u32 v0, s0, 2, v0 +; GFX10-SDAG-NEXT: ; return to shader part epilog +; ; GFX9-LABEL: shl2_add_u32_vgpr1: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_lshl_b32 s0, s0, 2 @@ -272,6 +476,22 @@ define amdgpu_ps float @shl2_add_u32_vgpr1(i32 inreg %src0, i32 %src1) { } define amdgpu_ps float @shl3_add_u32_vgpr1(i32 inreg %src0, i32 %src1) { +; GFX9-SDAG-LABEL: shl3_add_u32_vgpr1: +; GFX9-SDAG: ; %bb.0: +; GFX9-SDAG-NEXT: v_lshl_add_u32 v0, s0, 3, v0 +; GFX9-SDAG-NEXT: ; return to shader part epilog +; +; GFX8-SDAG-LABEL: shl3_add_u32_vgpr1: +; GFX8-SDAG: ; %bb.0: +; GFX8-SDAG-NEXT: s_lshl_b32 s0, s0, 3 +; GFX8-SDAG-NEXT: v_add_u32_e32 v0, vcc, s0, v0 +; GFX8-SDAG-NEXT: ; return to shader part epilog +; +; GFX10-SDAG-LABEL: shl3_add_u32_vgpr1: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: v_lshl_add_u32 v0, s0, 3, v0 +; GFX10-SDAG-NEXT: ; return to shader part epilog +; ; GFX9-LABEL: shl3_add_u32_vgpr1: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_lshl_b32 s0, s0, 3 @@ -296,6 +516,22 @@ define amdgpu_ps float @shl3_add_u32_vgpr1(i32 inreg %src0, i32 %src1) { } define amdgpu_ps float @shl4_add_u32_vgpr1(i32 inreg %src0, i32 %src1) { +; GFX9-SDAG-LABEL: shl4_add_u32_vgpr1: +; GFX9-SDAG: ; %bb.0: +; GFX9-SDAG-NEXT: v_lshl_add_u32 v0, s0, 4, v0 +; GFX9-SDAG-NEXT: ; return to shader part epilog +; +; GFX8-SDAG-LABEL: shl4_add_u32_vgpr1: +; GFX8-SDAG: ; %bb.0: +; GFX8-SDAG-NEXT: s_lshl_b32 s0, s0, 4 +; GFX8-SDAG-NEXT: v_add_u32_e32 v0, vcc, s0, v0 +; GFX8-SDAG-NEXT: ; return to shader part epilog +; +; GFX10-SDAG-LABEL: shl4_add_u32_vgpr1: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: v_lshl_add_u32 v0, s0, 4, v0 +; GFX10-SDAG-NEXT: ; return to shader part epilog +; ; GFX9-LABEL: shl4_add_u32_vgpr1: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_lshl_b32 s0, s0, 4 @@ -320,6 +556,22 @@ define amdgpu_ps float @shl4_add_u32_vgpr1(i32 inreg %src0, i32 %src1) { } define amdgpu_ps float @shl5_add_u32_vgpr1(i32 inreg %src0, i32 %src1) { +; GFX9-SDAG-LABEL: shl5_add_u32_vgpr1: +; GFX9-SDAG: ; %bb.0: +; GFX9-SDAG-NEXT: v_lshl_add_u32 v0, s0, 5, v0 +; GFX9-SDAG-NEXT: ; return to shader part epilog +; +; GFX8-SDAG-LABEL: shl5_add_u32_vgpr1: +; GFX8-SDAG: ; %bb.0: +; GFX8-SDAG-NEXT: s_lshl_b32 s0, s0, 5 +; GFX8-SDAG-NEXT: v_add_u32_e32 v0, vcc, s0, v0 +; GFX8-SDAG-NEXT: ; return to shader part epilog +; +; GFX10-SDAG-LABEL: shl5_add_u32_vgpr1: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: v_lshl_add_u32 v0, s0, 5, v0 +; GFX10-SDAG-NEXT: ; return to shader part epilog +; ; GFX9-LABEL: shl5_add_u32_vgpr1: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_lshl_b32 s0, s0, 5 @@ -344,6 +596,30 @@ define amdgpu_ps float @shl5_add_u32_vgpr1(i32 inreg %src0, i32 %src1) { } define amdgpu_ps <2 x i32> @s_shl1_add_u32_v2(<2 x i32> inreg %src0, <2 x i32> inreg %src1) { +; GFX9-SDAG-LABEL: s_shl1_add_u32_v2: +; GFX9-SDAG: ; %bb.0: +; GFX9-SDAG-NEXT: s_lshl_b32 s0, s0, 1 +; GFX9-SDAG-NEXT: s_lshl_b32 s1, s1, 1 +; GFX9-SDAG-NEXT: s_add_i32 s1, s1, s3 +; GFX9-SDAG-NEXT: s_add_i32 s0, s0, s2 +; GFX9-SDAG-NEXT: ; return to shader part epilog +; +; GFX8-SDAG-LABEL: s_shl1_add_u32_v2: +; GFX8-SDAG: ; %bb.0: +; GFX8-SDAG-NEXT: s_lshl_b32 s0, s0, 1 +; GFX8-SDAG-NEXT: s_lshl_b32 s1, s1, 1 +; GFX8-SDAG-NEXT: s_add_i32 s1, s1, s3 +; GFX8-SDAG-NEXT: s_add_i32 s0, s0, s2 +; GFX8-SDAG-NEXT: ; return to shader part epilog +; +; GFX10-SDAG-LABEL: s_shl1_add_u32_v2: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_lshl_b32 s0, s0, 1 +; GFX10-SDAG-NEXT: s_lshl_b32 s1, s1, 1 +; GFX10-SDAG-NEXT: s_add_i32 s0, s0, s2 +; GFX10-SDAG-NEXT: s_add_i32 s1, s1, s3 +; GFX10-SDAG-NEXT: ; return to shader part epilog +; ; GFX9-LABEL: s_shl1_add_u32_v2: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_lshl1_add_u32 s0, s0, s2 @@ -369,6 +645,30 @@ define amdgpu_ps <2 x i32> @s_shl1_add_u32_v2(<2 x i32> inreg %src0, <2 x i32> i } define amdgpu_ps <2 x i32> @s_shl2_add_u32_v2(<2 x i32> inreg %src0, <2 x i32> inreg %src1) { +; GFX9-SDAG-LABEL: s_shl2_add_u32_v2: +; GFX9-SDAG: ; %bb.0: +; GFX9-SDAG-NEXT: s_lshl_b32 s0, s0, 2 +; GFX9-SDAG-NEXT: s_lshl_b32 s1, s1, 2 +; GFX9-SDAG-NEXT: s_add_i32 s1, s1, s3 +; GFX9-SDAG-NEXT: s_add_i32 s0, s0, s2 +; GFX9-SDAG-NEXT: ; return to shader part epilog +; +; GFX8-SDAG-LABEL: s_shl2_add_u32_v2: +; GFX8-SDAG: ; %bb.0: +; GFX8-SDAG-NEXT: s_lshl_b32 s0, s0, 2 +; GFX8-SDAG-NEXT: s_lshl_b32 s1, s1, 2 +; GFX8-SDAG-NEXT: s_add_i32 s1, s1, s3 +; GFX8-SDAG-NEXT: s_add_i32 s0, s0, s2 +; GFX8-SDAG-NEXT: ; return to shader part epilog +; +; GFX10-SDAG-LABEL: s_shl2_add_u32_v2: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_lshl_b32 s0, s0, 2 +; GFX10-SDAG-NEXT: s_lshl_b32 s1, s1, 2 +; GFX10-SDAG-NEXT: s_add_i32 s0, s0, s2 +; GFX10-SDAG-NEXT: s_add_i32 s1, s1, s3 +; GFX10-SDAG-NEXT: ; return to shader part epilog +; ; GFX9-LABEL: s_shl2_add_u32_v2: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_lshl2_add_u32 s0, s0, s2 @@ -394,6 +694,30 @@ define amdgpu_ps <2 x i32> @s_shl2_add_u32_v2(<2 x i32> inreg %src0, <2 x i32> i } define amdgpu_ps <2 x i32> @s_shl3_add_u32_v2(<2 x i32> inreg %src0, <2 x i32> inreg %src1) { +; GFX9-SDAG-LABEL: s_shl3_add_u32_v2: +; GFX9-SDAG: ; %bb.0: +; GFX9-SDAG-NEXT: s_lshl_b32 s0, s0, 3 +; GFX9-SDAG-NEXT: s_lshl_b32 s1, s1, 3 +; GFX9-SDAG-NEXT: s_add_i32 s1, s1, s3 +; GFX9-SDAG-NEXT: s_add_i32 s0, s0, s2 +; GFX9-SDAG-NEXT: ; return to shader part epilog +; +; GFX8-SDAG-LABEL: s_shl3_add_u32_v2: +; GFX8-SDAG: ; %bb.0: +; GFX8-SDAG-NEXT: s_lshl_b32 s0, s0, 3 +; GFX8-SDAG-NEXT: s_lshl_b32 s1, s1, 3 +; GFX8-SDAG-NEXT: s_add_i32 s1, s1, s3 +; GFX8-SDAG-NEXT: s_add_i32 s0, s0, s2 +; GFX8-SDAG-NEXT: ; return to shader part epilog +; +; GFX10-SDAG-LABEL: s_shl3_add_u32_v2: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_lshl_b32 s0, s0, 3 +; GFX10-SDAG-NEXT: s_lshl_b32 s1, s1, 3 +; GFX10-SDAG-NEXT: s_add_i32 s0, s0, s2 +; GFX10-SDAG-NEXT: s_add_i32 s1, s1, s3 +; GFX10-SDAG-NEXT: ; return to shader part epilog +; ; GFX9-LABEL: s_shl3_add_u32_v2: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_lshl3_add_u32 s0, s0, s2 @@ -419,6 +743,30 @@ define amdgpu_ps <2 x i32> @s_shl3_add_u32_v2(<2 x i32> inreg %src0, <2 x i32> i } define amdgpu_ps <2 x i32> @s_shl4_add_u32_v2(<2 x i32> inreg %src0, <2 x i32> inreg %src1) { +; GFX9-SDAG-LABEL: s_shl4_add_u32_v2: +; GFX9-SDAG: ; %bb.0: +; GFX9-SDAG-NEXT: s_lshl_b32 s0, s0, 4 +; GFX9-SDAG-NEXT: s_lshl_b32 s1, s1, 4 +; GFX9-SDAG-NEXT: s_add_i32 s1, s1, s3 +; GFX9-SDAG-NEXT: s_add_i32 s0, s0, s2 +; GFX9-SDAG-NEXT: ; return to shader part epilog +; +; GFX8-SDAG-LABEL: s_shl4_add_u32_v2: +; GFX8-SDAG: ; %bb.0: +; GFX8-SDAG-NEXT: s_lshl_b32 s0, s0, 4 +; GFX8-SDAG-NEXT: s_lshl_b32 s1, s1, 4 +; GFX8-SDAG-NEXT: s_add_i32 s1, s1, s3 +; GFX8-SDAG-NEXT: s_add_i32 s0, s0, s2 +; GFX8-SDAG-NEXT: ; return to shader part epilog +; +; GFX10-SDAG-LABEL: s_shl4_add_u32_v2: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_lshl_b32 s0, s0, 4 +; GFX10-SDAG-NEXT: s_lshl_b32 s1, s1, 4 +; GFX10-SDAG-NEXT: s_add_i32 s0, s0, s2 +; GFX10-SDAG-NEXT: s_add_i32 s1, s1, s3 +; GFX10-SDAG-NEXT: ; return to shader part epilog +; ; GFX9-LABEL: s_shl4_add_u32_v2: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_lshl4_add_u32 s0, s0, s2 @@ -444,6 +792,30 @@ define amdgpu_ps <2 x i32> @s_shl4_add_u32_v2(<2 x i32> inreg %src0, <2 x i32> i } define amdgpu_ps <2 x i32> @s_shl_2_4_add_u32_v2(<2 x i32> inreg %src0, <2 x i32> inreg %src1) { +; GFX9-SDAG-LABEL: s_shl_2_4_add_u32_v2: +; GFX9-SDAG: ; %bb.0: +; GFX9-SDAG-NEXT: s_lshl_b32 s0, s0, 2 +; GFX9-SDAG-NEXT: s_lshl_b32 s1, s1, 4 +; GFX9-SDAG-NEXT: s_add_i32 s1, s1, s3 +; GFX9-SDAG-NEXT: s_add_i32 s0, s0, s2 +; GFX9-SDAG-NEXT: ; return to shader part epilog +; +; GFX8-SDAG-LABEL: s_shl_2_4_add_u32_v2: +; GFX8-SDAG: ; %bb.0: +; GFX8-SDAG-NEXT: s_lshl_b32 s0, s0, 2 +; GFX8-SDAG-NEXT: s_lshl_b32 s1, s1, 4 +; GFX8-SDAG-NEXT: s_add_i32 s1, s1, s3 +; GFX8-SDAG-NEXT: s_add_i32 s0, s0, s2 +; GFX8-SDAG-NEXT: ; return to shader part epilog +; +; GFX10-SDAG-LABEL: s_shl_2_4_add_u32_v2: +; GFX10-SDAG: ; %bb.0: +; GFX10-SDAG-NEXT: s_lshl_b32 s0, s0, 2 +; GFX10-SDAG-NEXT: s_lshl_b32 s1, s1, 4 +; GFX10-SDAG-NEXT: s_add_i32 s0, s0, s2 +; GFX10-SDAG-NEXT: s_add_i32 s1, s1, s3 +; GFX10-SDAG-NEXT: ; return to shader part epilog +; ; GFX9-LABEL: s_shl_2_4_add_u32_v2: ; GFX9: ; %bb.0: ; GFX9-NEXT: s_lshl2_add_u32 s0, s0, s2 diff --git a/llvm/test/CodeGen/AMDGPU/spill-vgpr-to-agpr-update-regscavenger.ll b/llvm/test/CodeGen/AMDGPU/spill-vgpr-to-agpr-update-regscavenger.ll index ef96944..586579f 100644 --- a/llvm/test/CodeGen/AMDGPU/spill-vgpr-to-agpr-update-regscavenger.ll +++ b/llvm/test/CodeGen/AMDGPU/spill-vgpr-to-agpr-update-regscavenger.ll @@ -20,33 +20,38 @@ define void @test() { ; CHECK-NEXT: ; in Loop: Header=BB0_1 Depth=1 ; CHECK-NEXT: .LBB0_3: ; %bb.3 ; CHECK-NEXT: ; in Loop: Header=BB0_1 Depth=1 +; CHECK-NEXT: ; implicit-def: $sgpr4 +; CHECK-NEXT: v_mov_b32_e32 v0, s4 +; CHECK-NEXT: v_readfirstlane_b32 s6, v0 ; CHECK-NEXT: s_mov_b64 s[4:5], -1 +; CHECK-NEXT: s_mov_b32 s7, 0 +; CHECK-NEXT: s_cmp_eq_u32 s6, s7 ; CHECK-NEXT: ; implicit-def: $vgpr1 : SGPR spill to VGPR lane ; CHECK-NEXT: v_writelane_b32 v1, s4, 0 ; CHECK-NEXT: v_writelane_b32 v1, s5, 1 -; CHECK-NEXT: s_or_saveexec_b64 s[8:9], -1 -; CHECK-NEXT: s_nop 0 +; CHECK-NEXT: s_mov_b64 s[10:11], exec +; CHECK-NEXT: s_mov_b64 exec, -1 ; CHECK-NEXT: v_accvgpr_write_b32 a0, v1 ; Reload Reuse -; CHECK-NEXT: s_mov_b64 exec, s[8:9] +; CHECK-NEXT: s_mov_b64 exec, s[10:11] ; CHECK-NEXT: s_cbranch_scc1 .LBB0_5 ; CHECK-NEXT: ; %bb.4: ; %bb.4 ; CHECK-NEXT: ; in Loop: Header=BB0_1 Depth=1 -; CHECK-NEXT: s_or_saveexec_b64 s[8:9], -1 +; CHECK-NEXT: s_or_saveexec_b64 s[10:11], -1 ; CHECK-NEXT: v_accvgpr_read_b32 v1, a0 ; Reload Reuse -; CHECK-NEXT: s_mov_b64 exec, s[8:9] +; CHECK-NEXT: s_mov_b64 exec, s[10:11] ; CHECK-NEXT: s_mov_b64 s[4:5], 0 ; CHECK-NEXT: v_writelane_b32 v1, s4, 0 ; CHECK-NEXT: v_writelane_b32 v1, s5, 1 -; CHECK-NEXT: s_or_saveexec_b64 s[8:9], -1 +; CHECK-NEXT: s_or_saveexec_b64 s[10:11], -1 ; CHECK-NEXT: s_nop 0 ; CHECK-NEXT: v_accvgpr_write_b32 a0, v1 ; Reload Reuse -; CHECK-NEXT: s_mov_b64 exec, s[8:9] +; CHECK-NEXT: s_mov_b64 exec, s[10:11] ; CHECK-NEXT: .LBB0_5: ; %Flow ; CHECK-NEXT: ; in Loop: Header=BB0_1 Depth=1 -; CHECK-NEXT: s_or_saveexec_b64 s[8:9], -1 +; CHECK-NEXT: s_or_saveexec_b64 s[10:11], -1 ; CHECK-NEXT: s_nop 0 ; CHECK-NEXT: v_accvgpr_read_b32 v1, a0 ; Reload Reuse -; CHECK-NEXT: s_mov_b64 exec, s[8:9] +; CHECK-NEXT: s_mov_b64 exec, s[10:11] ; CHECK-NEXT: v_readlane_b32 s4, v1, 0 ; CHECK-NEXT: v_readlane_b32 s5, v1, 1 ; CHECK-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[4:5] diff --git a/llvm/test/CodeGen/AMDGPU/splitkit-getsubrangeformask.ll b/llvm/test/CodeGen/AMDGPU/splitkit-getsubrangeformask.ll index 364598f..5aafb0f 100644 --- a/llvm/test/CodeGen/AMDGPU/splitkit-getsubrangeformask.ll +++ b/llvm/test/CodeGen/AMDGPU/splitkit-getsubrangeformask.ll @@ -31,8 +31,8 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x ; CHECK-NEXT: [[COPY13:%[0-9]+]]:sgpr_32 = COPY $sgpr10 ; CHECK-NEXT: [[COPY14:%[0-9]+]]:sgpr_32 = COPY $sgpr8 ; CHECK-NEXT: undef [[S_LOAD_DWORDX2_IMM:%[0-9]+]].sub0_sub1:sgpr_128 = S_LOAD_DWORDX2_IMM [[COPY]], 232, 0 :: (invariant load (s64) from %ir.39, addrspace 4) - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM undef %117:sgpr_128, 0, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: KILL undef %117:sgpr_128 + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM undef %125:sgpr_128, 0, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: KILL undef %125:sgpr_128 ; CHECK-NEXT: [[S_LSHL_B32_:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY5]], 4, implicit-def dead $scc ; CHECK-NEXT: [[S_LSHL_B32_1:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY4]], 4, implicit-def dead $scc ; CHECK-NEXT: [[S_LSHL_B32_2:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY3]], 4, implicit-def dead $scc @@ -44,85 +44,87 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x ; CHECK-NEXT: [[S_SUB_I32_1:%[0-9]+]]:sreg_32 = S_SUB_I32 [[S_BUFFER_LOAD_DWORD_IMM]], 30, implicit-def dead $scc ; CHECK-NEXT: undef [[S_ADD_U32_:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], [[S_LSHL_B32_2]], implicit-def $scc ; CHECK-NEXT: [[S_ADD_U32_:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_]], 16, 0 :: (invariant load (s128) from %ir.71, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_]], 16, 0 :: (invariant load (s128) from %ir.81, addrspace 4) ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM1:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM undef %74:sreg_64, 0, 0 :: (invariant load (s128) from `ptr addrspace(4) poison`, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM2:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_]], 64, 0 :: (invariant load (s128) from %ir.88, addrspace 4) ; CHECK-NEXT: KILL undef %74:sreg_64 ; CHECK-NEXT: KILL [[S_ADD_U32_]].sub0, [[S_ADD_U32_]].sub1 ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM1:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM [[S_LOAD_DWORDX4_IMM]], 0, 0 :: (dereferenceable invariant load (s32)) ; CHECK-NEXT: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec ; CHECK-NEXT: undef [[S_MOV_B32_:%[0-9]+]].sub1:sgpr_128 = S_MOV_B32 0 - ; CHECK-NEXT: [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET undef %112:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], undef %87:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) + ; CHECK-NEXT: [[BUFFER_LOAD_DWORD_OFFSET:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET undef %118:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) + ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], undef %89:sgpr_128, 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN1:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: KILL undef %112:sgpr_128 - ; CHECK-NEXT: KILL undef %87:sgpr_128 + ; CHECK-NEXT: KILL undef %89:sgpr_128 + ; CHECK-NEXT: KILL undef %118:sgpr_128 ; CHECK-NEXT: [[S_SUB_I32_2:%[0-9]+]]:sreg_32 = S_SUB_I32 [[S_BUFFER_LOAD_DWORD_IMM1]], 31, implicit-def dead $scc ; CHECK-NEXT: undef [[S_ADD_U32_1:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], [[S_LSHL_B32_]], implicit-def $scc ; CHECK-NEXT: [[S_ADD_U32_1:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc ; CHECK-NEXT: undef [[S_ADD_U32_2:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], [[S_LSHL_B32_1]], implicit-def $scc ; CHECK-NEXT: [[S_ADD_U32_2:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_ASHR_I32_3:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 undef %148:sreg_32, 31, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_3:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], undef %148:sreg_32, implicit-def $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM3:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_1]], 64, 0 :: (invariant load (s128) from %ir.77, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM4:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_2]], 64, 0 :: (invariant load (s128) from %ir.83, addrspace 4) - ; CHECK-NEXT: KILL [[S_ADD_U32_2]].sub0, [[S_ADD_U32_2]].sub1 + ; CHECK-NEXT: undef [[S_ADD_U32_3:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], [[S_LSHL_B32_2]], implicit-def $scc + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM2:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_1]], 64, 0 :: (invariant load (s128) from %ir.87, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM3:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_2]], 64, 0 :: (invariant load (s128) from %ir.93, addrspace 4) ; CHECK-NEXT: KILL [[S_ADD_U32_1]].sub0, [[S_ADD_U32_1]].sub1 - ; CHECK-NEXT: [[S_ADD_U32_3:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_3]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_4:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_4:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_5:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_1]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_5:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_6:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, undef %148:sreg_32, implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_6:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_3]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_7:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_2]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_7:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_8:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY8]], [[S_LSHL_B32_]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_8:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %48:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_9:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY9]], [[S_LSHL_B32_1]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_9:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %45:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_10:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY9]], [[S_LSHL_B32_2]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_10:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %45:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: KILL [[S_ADD_U32_2]].sub0, [[S_ADD_U32_2]].sub1 + ; CHECK-NEXT: [[S_ADD_U32_3:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_ASHR_I32_3:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 undef %169:sreg_32, 31, implicit-def dead $scc + ; CHECK-NEXT: undef [[S_ADD_U32_4:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY6]], undef %169:sreg_32, implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_4:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %54:sreg_32, [[S_ASHR_I32_3]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_5:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_5:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_6:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_1]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_6:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_7:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, undef %169:sreg_32, implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_7:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_3]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_8:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY7]].sub0, [[S_LSHL_B32_2]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_8:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %51:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_9:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY8]], [[S_LSHL_B32_]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_9:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %48:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_10:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY9]], [[S_LSHL_B32_1]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_10:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %45:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_11:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY9]], [[S_LSHL_B32_2]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_11:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %45:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc ; CHECK-NEXT: [[S_ADD_I32_:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_LSHL_B32_]], 16, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_1:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_LSHL_B32_2]], 16, implicit-def dead $scc ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM [[S_MOV_B32_]], [[S_ADD_I32_]], 0, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM1:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM [[S_MOV_B32_]], undef %279:sreg_32, 0, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM1:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM [[S_MOV_B32_]], undef %302:sreg_32, 0, 0 :: (dereferenceable invariant load (s32)) ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM2:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM [[S_MOV_B32_]], [[S_ADD_I32_1]], 0, 0 :: (dereferenceable invariant load (s32)) ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM2:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM [[S_MOV_B32_]], 16, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM3:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %334:sgpr_128, undef %335:sreg_32, 0, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM3:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM undef %345:sgpr_128, 16, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM5:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_3]], 64, 0 :: (invariant load (s128) from %ir.95, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM6:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_4]], 0, 0 :: (invariant load (s128) from %ir.100, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM7:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_5]], 0, 0 :: (invariant load (s128) from %ir.105, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM8:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_6]], 0, 0 :: (invariant load (s128) from %ir.112, addrspace 4) - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM4:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %329:sgpr_128, [[S_ADD_I32_]], 0, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM5:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %340:sgpr_128, [[S_ADD_I32_1]], 0, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN2:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM3]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN3:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM3:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %357:sgpr_128, undef %358:sreg_32, 0, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM3:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM undef %368:sgpr_128, 16, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM4:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_3]], 64, 0 :: (invariant load (s128) from %ir.99, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM5:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_4]], 64, 0 :: (invariant load (s128) from %ir.107, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM6:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_5]], 0, 0 :: (invariant load (s128) from %ir.112, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM7:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_6]], 0, 0 :: (invariant load (s128) from %ir.117, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM8:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_7]], 0, 0 :: (invariant load (s128) from %ir.124, addrspace 4) + ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN2:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM2]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM4:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %352:sgpr_128, [[S_ADD_I32_]], 0, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM5:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %363:sgpr_128, [[S_ADD_I32_1]], 0, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN3:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM3]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ADD_I32_2:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM]], -98, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_3:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM1]], -114, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_4:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM2]], -130, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_5:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM2]], -178, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_11:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY10]], [[S_LSHL_B32_]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_11:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %42:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_12:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_12:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_13:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_1]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_13:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: undef [[S_ADD_U32_14:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_2]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_14:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_12:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY10]], [[S_LSHL_B32_]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_12:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %42:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_13:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_13:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_14:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_1]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_14:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_15:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY11]], [[S_LSHL_B32_2]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_15:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %39:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc ; CHECK-NEXT: [[S_LSHL_B32_3:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY12]], 4, implicit-def dead $scc - ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN4:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM2]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) + ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN4:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM4]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ADD_I32_6:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_LSHL_B32_3]], 16, implicit-def dead $scc - ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM6:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %361:sgpr_128, [[S_ADD_I32_6]], 0, 0 :: (dereferenceable invariant load (s32)) + ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_SGPR_IMM6:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_SGPR_IMM undef %384:sgpr_128, [[S_ADD_I32_6]], 0, 0 :: (dereferenceable invariant load (s32)) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN5:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM5]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM9:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_4]], 224, 0 :: (invariant load (s128) from %ir.117, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM10:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[COPY7]], 224, 0 :: (invariant load (s128) from %ir.133, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM11:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_4]], 576, 0 :: (invariant load (s128) from %ir.138, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM9:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_5]], 224, 0 :: (invariant load (s128) from %ir.129, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM10:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[COPY7]], 224, 0 :: (invariant load (s128) from %ir.145, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM11:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_5]], 576, 0 :: (invariant load (s128) from %ir.150, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN6:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM6]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM12:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_5]], 224, 0 :: (invariant load (s128) from %ir.122, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM13:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_6]], 576, 0 :: (invariant load (s128) from %ir.150, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM14:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_7]], 224, 0 :: (invariant load (s128) from %ir.128, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM12:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_6]], 224, 0 :: (invariant load (s128) from %ir.134, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM13:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_7]], 576, 0 :: (invariant load (s128) from %ir.162, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM14:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_8]], 224, 0 :: (invariant load (s128) from %ir.140, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN7:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM7]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN8:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM8]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ADD_I32_7:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM4]], -217, implicit-def dead $scc @@ -133,49 +135,49 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x ; CHECK-NEXT: [[S_ADD_I32_12:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM3]], -329, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_13:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM3]], -345, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_14:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM6]], -441, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_15:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY2]], [[S_LSHL_B32_2]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_15:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %36:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_16:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY2]], [[S_LSHL_B32_2]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_16:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %36:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc ; CHECK-NEXT: [[S_LSHL_B32_4:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY13]], 4, implicit-def dead $scc ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN9:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM9]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ASHR_I32_4:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_4]], 31, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_16:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY2]], [[S_LSHL_B32_4]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_16:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %36:sreg_32, [[S_ASHR_I32_4]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: undef [[S_ADD_U32_17:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY2]], [[S_LSHL_B32_4]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_17:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %36:sreg_32, [[S_ASHR_I32_4]], implicit-def dead $scc, implicit $scc ; CHECK-NEXT: [[S_LSHL_B32_5:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY5]], 3, implicit-def dead $scc ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN10:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM12]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ASHR_I32_5:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_5]], 31, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_17:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_5]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_17:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_5]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORD_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM [[S_ADD_U32_17]], 168, 0 :: (invariant load (s32) from %ir.260, align 8, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM15:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_7]], 576, 0 :: (invariant load (s128) from %ir.145, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_18:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_5]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_18:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_5]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORD_IMM:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM [[S_ADD_U32_18]], 168, 0 :: (invariant load (s32) from %ir.273, align 8, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM15:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_8]], 576, 0 :: (invariant load (s128) from %ir.157, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN11:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM14]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN12:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM10]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN13:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM11]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_LOAD_DWORDX2_IMM:%[0-9]+]].sub3:sgpr_128 = S_MOV_B32 553734060 ; CHECK-NEXT: [[S_LOAD_DWORDX2_IMM:%[0-9]+]].sub2:sgpr_128 = S_MOV_B32 -1 ; CHECK-NEXT: [[COPY15:%[0-9]+]]:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM]] - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM16:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_8]], 0, 0 :: (invariant load (s128) from %ir.158, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM16:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_9]], 0, 0 :: (invariant load (s128) from %ir.170, addrspace 4) ; CHECK-NEXT: [[COPY15:%[0-9]+]].sub1:sgpr_128 = COPY [[S_MOV_B32_]].sub1 ; CHECK-NEXT: [[COPY15:%[0-9]+]].sub0:sgpr_128 = COPY [[S_LOAD_DWORD_IMM]] ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM4:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM [[COPY15]], 0, 0 :: (dereferenceable invariant load (s32)) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN14:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM15]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN15:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM13]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM17:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_9]], 0, 0 :: (invariant load (s128) from %ir.166, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM18:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_10]], 0, 0 :: (invariant load (s128) from %ir.171, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM17:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_10]], 0, 0 :: (invariant load (s128) from %ir.178, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM18:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_11]], 0, 0 :: (invariant load (s128) from %ir.183, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN16:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM16]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_LSHL_B32_6:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY4]], 3, implicit-def dead $scc ; CHECK-NEXT: [[BUFFER_LOAD_DWORD_OFFSET1:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[S_LOAD_DWORDX4_IMM1]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ASHR_I32_6:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_6]], 31, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_15:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM4]], -467, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_18:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_6]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_18:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_6]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX2_IMM1:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[S_ADD_U32_18]], 168, 0 :: (invariant load (s64) from %ir.269, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_19:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_6]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_19:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_6]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORDX2_IMM1:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[S_ADD_U32_19]], 168, 0 :: (invariant load (s64) from %ir.282, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_DWORD_OFFSET2:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[S_LOAD_DWORDX4_IMM17]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_DWORD_OFFSET3:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_DWORD_OFFSET [[S_LOAD_DWORDX4_IMM18]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM19:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_11]], 0, 0 :: (invariant load (s128) from %ir.193, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM20:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_12]], 0, 0 :: (invariant load (s128) from %ir.199, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM19:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_12]], 0, 0 :: (invariant load (s128) from %ir.205, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM20:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_13]], 0, 0 :: (invariant load (s128) from %ir.211, addrspace 4) ; CHECK-NEXT: [[COPY16:%[0-9]+]]:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM]] - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM21:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_13]], 0, 0 :: (invariant load (s128) from %ir.204, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM22:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_14]], 0, 0 :: (invariant load (s128) from %ir.209, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM21:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_14]], 0, 0 :: (invariant load (s128) from %ir.216, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM22:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_15]], 0, 0 :: (invariant load (s128) from %ir.221, addrspace 4) ; CHECK-NEXT: [[S_AND_B32_:%[0-9]+]]:sreg_32 = S_AND_B32 [[S_LOAD_DWORDX2_IMM1]].sub1, 65535, implicit-def dead $scc ; CHECK-NEXT: [[COPY16:%[0-9]+]].sub0:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM1]].sub0 ; CHECK-NEXT: [[COPY16:%[0-9]+]].sub1:sgpr_128 = COPY [[S_AND_B32_]] @@ -187,30 +189,30 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN20:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM22]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[S_ASHR_I32_7:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_7]], 31, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_16:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM5]], -468, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_19:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_7]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_19:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_7]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX2_IMM2:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[S_ADD_U32_19]], 168, 0 :: (invariant load (s64) from %ir.280, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_20:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_7]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_20:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_7]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORDX2_IMM2:%[0-9]+]]:sreg_64_xexec = S_LOAD_DWORDX2_IMM [[S_ADD_U32_20]], 168, 0 :: (invariant load (s64) from %ir.293, addrspace 4) ; CHECK-NEXT: [[COPY17:%[0-9]+]]:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM]] ; CHECK-NEXT: [[S_AND_B32_1:%[0-9]+]]:sreg_32 = S_AND_B32 [[S_LOAD_DWORDX2_IMM2]].sub1, 65535, implicit-def dead $scc ; CHECK-NEXT: [[COPY17:%[0-9]+]].sub0:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM2]].sub0 ; CHECK-NEXT: [[COPY17:%[0-9]+]].sub1:sgpr_128 = COPY [[S_AND_B32_1]] ; CHECK-NEXT: [[S_BUFFER_LOAD_DWORD_IMM6:%[0-9]+]]:sreg_32_xm0_xexec = S_BUFFER_LOAD_DWORD_IMM [[COPY17]], 0, 0 :: (dereferenceable invariant load (s32)) - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM23:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_15]], 160, 0 :: (invariant load (s128) from %ir.244, addrspace 4) - ; CHECK-NEXT: [[S_LOAD_DWORD_IMM1:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM undef %443:sreg_64, 0, 0 :: (invariant load (s32) from `ptr addrspace(4) poison`, addrspace 4) - ; CHECK-NEXT: KILL [[S_ADD_U32_15]].sub0, [[S_ADD_U32_15]].sub1 + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM23:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_16]], 160, 0 :: (invariant load (s128) from %ir.256, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORD_IMM1:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM undef %470:sreg_64, 0, 0 :: (invariant load (s32) from `ptr addrspace(4) poison`, addrspace 4) + ; CHECK-NEXT: KILL [[S_ADD_U32_16]].sub0, [[S_ADD_U32_16]].sub1 + ; CHECK-NEXT: KILL undef %470:sreg_64 ; CHECK-NEXT: KILL [[COPY17]].sub0_sub1_sub2, [[COPY17]].sub3 - ; CHECK-NEXT: KILL undef %443:sreg_64 ; CHECK-NEXT: [[S_LSHL_B32_8:%[0-9]+]]:sreg_32 = S_LSHL_B32 [[COPY14]], 3, implicit-def dead $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM24:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_16]], 160, 0 :: (invariant load (s128) from %ir.252, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM24:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_17]], 160, 0 :: (invariant load (s128) from %ir.265, addrspace 4) ; CHECK-NEXT: [[S_ASHR_I32_8:%[0-9]+]]:sreg_32_xm0 = S_ASHR_I32 [[S_LSHL_B32_8]], 31, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_17:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM6]], -469, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_20:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_8]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_20:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_8]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORD_IMM2:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM [[S_ADD_U32_20]], 168, 0 :: (invariant load (s32) from %ir.291, align 8, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_21:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY]].sub0, [[S_LSHL_B32_8]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_21:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %57:sreg_32, [[S_ASHR_I32_8]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORD_IMM2:%[0-9]+]]:sreg_32_xm0_xexec = S_LOAD_DWORD_IMM [[S_ADD_U32_21]], 168, 0 :: (invariant load (s32) from %ir.305, align 8, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN21:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM23]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN22:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM24]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) - ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM23]] ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM24]] + ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM23]] ; CHECK-NEXT: [[S_AND_B32_2:%[0-9]+]]:sreg_32 = S_AND_B32 [[S_LOAD_DWORD_IMM1]], 65535, implicit-def dead $scc ; CHECK-NEXT: [[COPY18:%[0-9]+]]:sgpr_128 = COPY [[S_LOAD_DWORDX2_IMM]] ; CHECK-NEXT: [[COPY18:%[0-9]+]].sub1:sgpr_128 = COPY [[S_AND_B32_2]] @@ -222,22 +224,22 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x ; CHECK-NEXT: [[S_ADD_I32_21:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM3]], -507, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_22:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_SGPR_IMM3]], -539, implicit-def dead $scc ; CHECK-NEXT: [[S_ADD_I32_23:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM7]], -473, implicit-def dead $scc - ; CHECK-NEXT: undef [[S_ADD_U32_21:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_21:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM25:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_21]], 96, 0 :: (invariant load (s128) from %ir.309, addrspace 4) - ; CHECK-NEXT: undef [[S_ADD_U32_22:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_1]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_22:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM26:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_22]], 96, 0 :: (invariant load (s128) from %ir.315, addrspace 4) - ; CHECK-NEXT: undef [[S_ADD_U32_23:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_2]], implicit-def $scc - ; CHECK-NEXT: [[S_ADD_U32_23:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc - ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM27:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_23]], 96, 0 :: (invariant load (s128) from %ir.321, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_22:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_22:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM25:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_22]], 96, 0 :: (invariant load (s128) from %ir.323, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_23:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_1]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_23:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_1]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM26:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_23]], 96, 0 :: (invariant load (s128) from %ir.329, addrspace 4) + ; CHECK-NEXT: undef [[S_ADD_U32_24:%[0-9]+]].sub0:sreg_64 = S_ADD_U32 [[COPY1]], [[S_LSHL_B32_2]], implicit-def $scc + ; CHECK-NEXT: [[S_ADD_U32_24:%[0-9]+]].sub1:sreg_64 = S_ADDC_U32 undef %33:sreg_32, [[S_ASHR_I32_2]], implicit-def dead $scc, implicit $scc + ; CHECK-NEXT: [[S_LOAD_DWORDX4_IMM27:%[0-9]+]]:sgpr_128 = S_LOAD_DWORDX4_IMM [[S_ADD_U32_24]], 96, 0 :: (invariant load (s128) from %ir.335, addrspace 4) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN23:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM25]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN24:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM26]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) ; CHECK-NEXT: [[BUFFER_LOAD_FORMAT_X_IDXEN25:%[0-9]+]]:vgpr_32 = BUFFER_LOAD_FORMAT_X_IDXEN [[V_MOV_B32_e32_]], [[S_LOAD_DWORDX4_IMM27]], 0, 0, 0, 0, implicit $exec :: (dereferenceable load (s32), align 1, addrspace 8) + ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM27]] ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM25]] - ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM26]] ; CHECK-NEXT: KILL [[V_MOV_B32_e32_]] - ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM27]] + ; CHECK-NEXT: KILL [[S_LOAD_DWORDX4_IMM26]] ; CHECK-NEXT: [[V_ADD_U32_e64_:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 -2, [[BUFFER_LOAD_FORMAT_X_IDXEN]], 0, implicit $exec ; CHECK-NEXT: [[V_ADD_U32_e64_1:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 -1, [[BUFFER_LOAD_FORMAT_X_IDXEN1]], 0, implicit $exec ; CHECK-NEXT: [[V_ADD_U32_e64_2:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 -3, [[BUFFER_LOAD_FORMAT_X_IDXEN]], 0, implicit $exec @@ -349,13 +351,13 @@ define amdgpu_gs void @_amdgpu_gs_main(i32 inreg %primShaderTableAddrLow, <31 x ; CHECK-NEXT: [[V_OR_B32_e64_64:%[0-9]+]]:vgpr_32 = V_OR_B32_e64 [[V_OR_B32_e64_63]], [[V_ADD_U32_e64_28]], implicit $exec ; CHECK-NEXT: [[V_ADD_U32_e64_30:%[0-9]+]]:vgpr_32 = V_ADD_U32_e64 -593, [[BUFFER_LOAD_FORMAT_X_IDXEN]], 0, implicit $exec ; CHECK-NEXT: [[V_OR_B32_e64_65:%[0-9]+]]:vgpr_32 = V_OR_B32_e64 [[V_OR_B32_e64_64]], [[V_ADD_U32_e64_29]], implicit $exec - ; CHECK-NEXT: [[S_LOAD_DWORDX8_IMM:%[0-9]+]]:sgpr_256 = S_LOAD_DWORDX8_IMM undef %516:sreg_64, 0, 0 :: (invariant load (s256) from `ptr addrspace(4) poison`, addrspace 4) + ; CHECK-NEXT: [[S_LOAD_DWORDX8_IMM:%[0-9]+]]:sgpr_256 = S_LOAD_DWORDX8_IMM undef %543:sreg_64, 0, 0 :: (invariant load (s256) from `ptr addrspace(4) poison`, addrspace 4) ; CHECK-NEXT: [[V_OR_B32_e64_66:%[0-9]+]]:vgpr_32 = V_OR_B32_e64 [[V_OR_B32_e64_65]], [[V_ADD_U32_e64_30]], implicit $exec ; CHECK-NEXT: [[S_ADD_I32_24:%[0-9]+]]:sreg_32 = S_ADD_I32 [[S_BUFFER_LOAD_DWORD_IMM8]], -594, implicit-def dead $scc ; CHECK-NEXT: [[V_OR_B32_e64_67:%[0-9]+]]:vgpr_32 = V_OR_B32_e64 [[S_ADD_I32_24]], [[V_OR_B32_e64_66]], implicit $exec ; CHECK-NEXT: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U32_e64 0, [[V_OR_B32_e64_67]], implicit $exec ; CHECK-NEXT: undef [[V_CNDMASK_B32_e64_:%[0-9]+]].sub3:vreg_128 = V_CNDMASK_B32_e64 0, 0, 0, 1, [[V_CMP_EQ_U32_e64_]], implicit $exec - ; CHECK-NEXT: IMAGE_STORE_V4_V2_nsa_gfx10 [[V_CNDMASK_B32_e64_]], undef %530:vgpr_32, undef %532:vgpr_32, [[S_LOAD_DWORDX8_IMM]], 15, 1, -1, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), addrspace 8) + ; CHECK-NEXT: IMAGE_STORE_V4_V2_nsa_gfx10 [[V_CNDMASK_B32_e64_]], undef %557:vgpr_32, undef %559:vgpr_32, [[S_LOAD_DWORDX8_IMM]], 15, 1, -1, 0, 0, 0, 0, 0, 0, implicit $exec :: (dereferenceable store (s128), addrspace 8) ; CHECK-NEXT: S_ENDPGM 0 .expVert: %0 = extractelement <31 x i32> %userData, i64 2 diff --git a/llvm/test/CodeGen/AMDGPU/srem64.ll b/llvm/test/CodeGen/AMDGPU/srem64.ll index ea9bb04..862e2dd 100644 --- a/llvm/test/CodeGen/AMDGPU/srem64.ll +++ b/llvm/test/CodeGen/AMDGPU/srem64.ll @@ -8,12 +8,11 @@ define amdgpu_kernel void @s_test_srem(ptr addrspace(1) %out, i64 %x, i64 %y) { ; GCN-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd ; GCN-NEXT: s_load_dwordx4 s[4:7], s[4:5], 0x9 ; GCN-NEXT: s_mov_b32 s3, 0xf000 -; GCN-NEXT: s_mov_b32 s2, -1 ; GCN-NEXT: s_waitcnt lgkmcnt(0) ; GCN-NEXT: v_cvt_f32_u32_e32 v0, s8 ; GCN-NEXT: v_cvt_f32_u32_e32 v1, s9 -; GCN-NEXT: s_sub_u32 s10, 0, s8 -; GCN-NEXT: s_subb_u32 s11, 0, s9 +; GCN-NEXT: s_sub_u32 s0, 0, s8 +; GCN-NEXT: s_subb_u32 s1, 0, s9 ; GCN-NEXT: v_madmk_f32 v0, v1, 0x4f800000, v0 ; GCN-NEXT: v_rcp_f32_e32 v0, v0 ; GCN-NEXT: v_mul_f32_e32 v0, 0x5f7ffffc, v0 @@ -22,69 +21,65 @@ define amdgpu_kernel void @s_test_srem(ptr addrspace(1) %out, i64 %x, i64 %y) { ; GCN-NEXT: v_madmk_f32 v0, v1, 0xcf800000, v0 ; GCN-NEXT: v_cvt_u32_f32_e32 v0, v0 ; GCN-NEXT: v_cvt_u32_f32_e32 v1, v1 -; GCN-NEXT: v_mul_hi_u32 v2, s10, v0 -; GCN-NEXT: v_readfirstlane_b32 s12, v1 -; GCN-NEXT: v_readfirstlane_b32 s0, v0 -; GCN-NEXT: s_mul_i32 s1, s10, s12 -; GCN-NEXT: v_readfirstlane_b32 s15, v2 -; GCN-NEXT: s_mul_i32 s13, s11, s0 -; GCN-NEXT: s_mul_i32 s14, s10, s0 -; GCN-NEXT: s_add_i32 s1, s15, s1 -; GCN-NEXT: v_mul_hi_u32 v3, v0, s14 -; GCN-NEXT: s_add_i32 s1, s1, s13 -; GCN-NEXT: v_mul_hi_u32 v0, v0, s1 -; GCN-NEXT: v_mul_hi_u32 v4, v1, s14 -; GCN-NEXT: v_readfirstlane_b32 s13, v3 -; GCN-NEXT: s_mul_i32 s15, s0, s1 -; GCN-NEXT: v_mul_hi_u32 v1, v1, s1 -; GCN-NEXT: s_add_u32 s13, s13, s15 +; GCN-NEXT: v_mul_hi_u32 v2, s0, v0 +; GCN-NEXT: v_readfirstlane_b32 s10, v1 +; GCN-NEXT: v_readfirstlane_b32 s2, v0 +; GCN-NEXT: s_mul_i32 s11, s0, s10 +; GCN-NEXT: v_readfirstlane_b32 s14, v2 +; GCN-NEXT: s_mul_i32 s12, s1, s2 +; GCN-NEXT: s_mul_i32 s13, s0, s2 +; GCN-NEXT: s_add_i32 s11, s14, s11 +; GCN-NEXT: v_mul_hi_u32 v3, v0, s13 +; GCN-NEXT: s_add_i32 s11, s11, s12 +; GCN-NEXT: v_mul_hi_u32 v0, v0, s11 +; GCN-NEXT: v_mul_hi_u32 v4, v1, s13 +; GCN-NEXT: v_readfirstlane_b32 s12, v3 +; GCN-NEXT: s_mul_i32 s15, s2, s11 +; GCN-NEXT: v_mul_hi_u32 v1, v1, s11 +; GCN-NEXT: s_add_u32 s12, s12, s15 ; GCN-NEXT: v_readfirstlane_b32 s15, v0 -; GCN-NEXT: s_mul_i32 s14, s12, s14 +; GCN-NEXT: s_mul_i32 s13, s10, s13 ; GCN-NEXT: s_addc_u32 s15, 0, s15 -; GCN-NEXT: v_readfirstlane_b32 s16, v4 -; GCN-NEXT: s_add_u32 s13, s13, s14 -; GCN-NEXT: s_addc_u32 s13, s15, s16 -; GCN-NEXT: v_readfirstlane_b32 s14, v1 -; GCN-NEXT: s_addc_u32 s14, s14, 0 -; GCN-NEXT: s_mul_i32 s1, s12, s1 -; GCN-NEXT: s_add_u32 s1, s13, s1 -; GCN-NEXT: s_addc_u32 s13, 0, s14 -; GCN-NEXT: s_add_u32 s14, s0, s1 -; GCN-NEXT: v_mov_b32_e32 v0, s14 -; GCN-NEXT: v_mul_hi_u32 v0, s10, v0 -; GCN-NEXT: s_cselect_b64 s[0:1], -1, 0 -; GCN-NEXT: s_or_b32 s0, s0, s1 -; GCN-NEXT: s_addc_u32 s12, s12, s13 -; GCN-NEXT: s_mul_i32 s0, s10, s12 -; GCN-NEXT: v_readfirstlane_b32 s1, v0 -; GCN-NEXT: s_add_i32 s0, s1, s0 -; GCN-NEXT: s_mul_i32 s11, s11, s14 -; GCN-NEXT: s_mul_i32 s1, s10, s14 -; GCN-NEXT: s_add_i32 s0, s0, s11 -; GCN-NEXT: v_mov_b32_e32 v2, s1 -; GCN-NEXT: v_mov_b32_e32 v0, s0 -; GCN-NEXT: v_mul_hi_u32 v3, s12, v2 -; GCN-NEXT: v_mul_hi_u32 v2, s14, v2 -; GCN-NEXT: v_mul_hi_u32 v1, s12, v0 -; GCN-NEXT: v_mul_hi_u32 v0, s14, v0 -; GCN-NEXT: s_mul_i32 s11, s14, s0 -; GCN-NEXT: v_readfirstlane_b32 s15, v2 -; GCN-NEXT: s_add_u32 s11, s15, s11 +; GCN-NEXT: v_readfirstlane_b32 s14, v4 +; GCN-NEXT: s_add_u32 s12, s12, s13 +; GCN-NEXT: s_addc_u32 s12, s15, s14 +; GCN-NEXT: v_readfirstlane_b32 s13, v1 +; GCN-NEXT: s_addc_u32 s13, s13, 0 +; GCN-NEXT: s_mul_i32 s11, s10, s11 +; GCN-NEXT: s_add_u32 s11, s12, s11 +; GCN-NEXT: s_addc_u32 s12, 0, s13 +; GCN-NEXT: s_add_u32 s11, s2, s11 +; GCN-NEXT: v_mov_b32_e32 v0, s11 +; GCN-NEXT: v_mul_hi_u32 v0, s0, v0 +; GCN-NEXT: s_addc_u32 s10, s10, s12 +; GCN-NEXT: s_mul_i32 s12, s0, s10 +; GCN-NEXT: s_mul_i32 s1, s1, s11 ; GCN-NEXT: v_readfirstlane_b32 s13, v0 -; GCN-NEXT: s_mul_i32 s1, s12, s1 -; GCN-NEXT: s_addc_u32 s13, 0, s13 -; GCN-NEXT: v_readfirstlane_b32 s10, v3 -; GCN-NEXT: s_add_u32 s1, s11, s1 -; GCN-NEXT: s_addc_u32 s1, s13, s10 -; GCN-NEXT: v_readfirstlane_b32 s10, v1 -; GCN-NEXT: s_addc_u32 s10, s10, 0 -; GCN-NEXT: s_mul_i32 s0, s12, s0 -; GCN-NEXT: s_add_u32 s0, s1, s0 -; GCN-NEXT: s_addc_u32 s10, 0, s10 -; GCN-NEXT: s_add_u32 s11, s14, s0 -; GCN-NEXT: s_cselect_b64 s[0:1], -1, 0 -; GCN-NEXT: s_or_b32 s0, s0, s1 -; GCN-NEXT: s_addc_u32 s1, s12, s10 +; GCN-NEXT: s_add_i32 s12, s13, s12 +; GCN-NEXT: s_mul_i32 s0, s0, s11 +; GCN-NEXT: s_add_i32 s1, s12, s1 +; GCN-NEXT: v_mov_b32_e32 v2, s0 +; GCN-NEXT: v_mov_b32_e32 v0, s1 +; GCN-NEXT: v_mul_hi_u32 v3, s10, v2 +; GCN-NEXT: v_mul_hi_u32 v2, s11, v2 +; GCN-NEXT: v_mul_hi_u32 v1, s10, v0 +; GCN-NEXT: v_mul_hi_u32 v0, s11, v0 +; GCN-NEXT: s_mul_i32 s13, s11, s1 +; GCN-NEXT: v_readfirstlane_b32 s15, v2 +; GCN-NEXT: s_add_u32 s13, s15, s13 +; GCN-NEXT: v_readfirstlane_b32 s14, v0 +; GCN-NEXT: s_mul_i32 s0, s10, s0 +; GCN-NEXT: s_addc_u32 s14, 0, s14 +; GCN-NEXT: v_readfirstlane_b32 s12, v3 +; GCN-NEXT: s_add_u32 s0, s13, s0 +; GCN-NEXT: s_addc_u32 s0, s14, s12 +; GCN-NEXT: v_readfirstlane_b32 s12, v1 +; GCN-NEXT: s_addc_u32 s12, s12, 0 +; GCN-NEXT: s_mul_i32 s1, s10, s1 +; GCN-NEXT: s_add_u32 s0, s0, s1 +; GCN-NEXT: s_addc_u32 s1, 0, s12 +; GCN-NEXT: s_add_u32 s11, s11, s0 +; GCN-NEXT: s_addc_u32 s1, s10, s1 ; GCN-NEXT: v_mov_b32_e32 v0, s1 ; GCN-NEXT: v_mul_hi_u32 v1, s6, v0 ; GCN-NEXT: v_mov_b32_e32 v2, s11 @@ -118,11 +113,9 @@ define amdgpu_kernel void @s_test_srem(ptr addrspace(1) %out, i64 %x, i64 %y) { ; GCN-NEXT: s_mul_i32 s4, s8, s4 ; GCN-NEXT: s_sub_u32 s6, s6, s4 ; GCN-NEXT: s_cselect_b64 s[4:5], -1, 0 -; GCN-NEXT: s_or_b32 s11, s4, s5 ; GCN-NEXT: s_subb_u32 s13, s10, s9 ; GCN-NEXT: s_sub_u32 s14, s6, s8 ; GCN-NEXT: s_cselect_b64 s[10:11], -1, 0 -; GCN-NEXT: s_or_b32 s15, s10, s11 ; GCN-NEXT: s_subb_u32 s15, s13, 0 ; GCN-NEXT: s_cmp_ge_u32 s15, s9 ; GCN-NEXT: s_cselect_b32 s16, -1, 0 @@ -131,13 +124,11 @@ define amdgpu_kernel void @s_test_srem(ptr addrspace(1) %out, i64 %x, i64 %y) { ; GCN-NEXT: s_cmp_eq_u32 s15, s9 ; GCN-NEXT: s_cselect_b32 s16, s17, s16 ; GCN-NEXT: s_or_b32 s10, s10, s11 -; GCN-NEXT: s_subb_u32 s13, s13, s9 -; GCN-NEXT: s_sub_u32 s17, s14, s8 -; GCN-NEXT: s_cselect_b64 s[10:11], -1, 0 -; GCN-NEXT: s_or_b32 s10, s10, s11 -; GCN-NEXT: s_subb_u32 s10, s13, 0 +; GCN-NEXT: s_subb_u32 s10, s13, s9 +; GCN-NEXT: s_sub_u32 s11, s14, s8 +; GCN-NEXT: s_subb_u32 s10, s10, 0 ; GCN-NEXT: s_cmp_lg_u32 s16, 0 -; GCN-NEXT: s_cselect_b32 s11, s17, s14 +; GCN-NEXT: s_cselect_b32 s11, s11, s14 ; GCN-NEXT: s_cselect_b32 s10, s10, s15 ; GCN-NEXT: s_or_b32 s4, s4, s5 ; GCN-NEXT: s_subb_u32 s4, s7, s12 @@ -150,6 +141,7 @@ define amdgpu_kernel void @s_test_srem(ptr addrspace(1) %out, i64 %x, i64 %y) { ; GCN-NEXT: s_cmp_lg_u32 s5, 0 ; GCN-NEXT: s_cselect_b32 s4, s10, s4 ; GCN-NEXT: s_cselect_b32 s5, s11, s6 +; GCN-NEXT: s_mov_b32 s2, -1 ; GCN-NEXT: v_mov_b32_e32 v0, s5 ; GCN-NEXT: v_mov_b32_e32 v1, s4 ; GCN-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 @@ -180,8 +172,6 @@ define amdgpu_kernel void @s_test_srem(ptr addrspace(1) %out, i64 %x, i64 %y) { ; GCN-IR-NEXT: s_cbranch_vccz .LBB0_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: s_add_u32 s14, s12, 1 -; GCN-IR-NEXT: s_cselect_b64 s[8:9], -1, 0 -; GCN-IR-NEXT: s_or_b32 s8, s8, s9 ; GCN-IR-NEXT: s_addc_u32 s8, s13, 0 ; GCN-IR-NEXT: s_cselect_b64 s[8:9], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s12, 63, s12 @@ -213,8 +203,6 @@ define amdgpu_kernel void @s_test_srem(ptr addrspace(1) %out, i64 %x, i64 %y) { ; GCN-IR-NEXT: s_sub_u32 s12, s12, s18 ; GCN-IR-NEXT: s_subb_u32 s13, s13, s19 ; GCN-IR-NEXT: s_add_u32 s16, s16, 1 -; GCN-IR-NEXT: s_cselect_b64 s[18:19], -1, 0 -; GCN-IR-NEXT: s_or_b32 s18, s18, s19 ; GCN-IR-NEXT: s_addc_u32 s17, s17, 0 ; GCN-IR-NEXT: s_cselect_b64 s[18:19], -1, 0 ; GCN-IR-NEXT: s_mov_b64 s[10:11], s[4:5] @@ -968,81 +956,76 @@ define amdgpu_kernel void @s_test_srem33_64(ptr addrspace(1) %out, i64 %x, i64 % ; GCN-NEXT: s_xor_b64 s[4:5], s[2:3], s[4:5] ; GCN-NEXT: v_cvt_f32_u32_e32 v0, s4 ; GCN-NEXT: v_cvt_f32_u32_e32 v1, s5 -; GCN-NEXT: s_sub_u32 s10, 0, s4 -; GCN-NEXT: s_subb_u32 s11, 0, s5 +; GCN-NEXT: s_sub_u32 s8, 0, s4 +; GCN-NEXT: s_subb_u32 s9, 0, s5 ; GCN-NEXT: s_mov_b32 s3, 0xf000 ; GCN-NEXT: v_madmk_f32 v0, v1, 0x4f800000, v0 ; GCN-NEXT: v_rcp_f32_e32 v0, v0 -; GCN-NEXT: s_mov_b32 s2, -1 ; GCN-NEXT: v_mul_f32_e32 v0, 0x5f7ffffc, v0 ; GCN-NEXT: v_mul_f32_e32 v1, 0x2f800000, v0 ; GCN-NEXT: v_trunc_f32_e32 v1, v1 ; GCN-NEXT: v_madmk_f32 v0, v1, 0xcf800000, v0 ; GCN-NEXT: v_cvt_u32_f32_e32 v0, v0 ; GCN-NEXT: v_cvt_u32_f32_e32 v1, v1 -; GCN-NEXT: v_mul_hi_u32 v2, s10, v0 -; GCN-NEXT: v_readfirstlane_b32 s12, v1 -; GCN-NEXT: v_readfirstlane_b32 s8, v0 -; GCN-NEXT: s_mul_i32 s9, s10, s12 -; GCN-NEXT: v_readfirstlane_b32 s15, v2 -; GCN-NEXT: s_mul_i32 s13, s11, s8 -; GCN-NEXT: s_mul_i32 s14, s10, s8 -; GCN-NEXT: s_add_i32 s9, s15, s9 -; GCN-NEXT: v_mul_hi_u32 v3, v0, s14 -; GCN-NEXT: s_add_i32 s9, s9, s13 -; GCN-NEXT: v_mul_hi_u32 v0, v0, s9 -; GCN-NEXT: v_mul_hi_u32 v4, v1, s14 -; GCN-NEXT: v_readfirstlane_b32 s13, v3 -; GCN-NEXT: s_mul_i32 s15, s8, s9 -; GCN-NEXT: s_add_u32 s13, s13, s15 -; GCN-NEXT: v_readfirstlane_b32 s15, v0 -; GCN-NEXT: v_mul_hi_u32 v0, v1, s9 -; GCN-NEXT: s_addc_u32 s15, 0, s15 -; GCN-NEXT: s_mul_i32 s14, s12, s14 -; GCN-NEXT: v_readfirstlane_b32 s16, v4 -; GCN-NEXT: s_add_u32 s13, s13, s14 -; GCN-NEXT: s_addc_u32 s13, s15, s16 +; GCN-NEXT: v_mul_hi_u32 v2, s8, v0 +; GCN-NEXT: v_readfirstlane_b32 s10, v1 +; GCN-NEXT: v_readfirstlane_b32 s2, v0 +; GCN-NEXT: s_mul_i32 s11, s8, s10 +; GCN-NEXT: v_readfirstlane_b32 s14, v2 +; GCN-NEXT: s_mul_i32 s12, s9, s2 +; GCN-NEXT: s_mul_i32 s13, s8, s2 +; GCN-NEXT: s_add_i32 s11, s14, s11 +; GCN-NEXT: v_mul_hi_u32 v3, v0, s13 +; GCN-NEXT: s_add_i32 s11, s11, s12 +; GCN-NEXT: v_mul_hi_u32 v0, v0, s11 +; GCN-NEXT: v_mul_hi_u32 v4, v1, s13 +; GCN-NEXT: v_readfirstlane_b32 s12, v3 +; GCN-NEXT: s_mul_i32 s14, s2, s11 +; GCN-NEXT: v_mul_hi_u32 v1, v1, s11 +; GCN-NEXT: s_add_u32 s12, s12, s14 ; GCN-NEXT: v_readfirstlane_b32 s14, v0 -; GCN-NEXT: s_addc_u32 s14, s14, 0 -; GCN-NEXT: s_mul_i32 s9, s12, s9 -; GCN-NEXT: s_add_u32 s9, s13, s9 -; GCN-NEXT: s_addc_u32 s13, 0, s14 -; GCN-NEXT: s_add_u32 s14, s8, s9 -; GCN-NEXT: v_mov_b32_e32 v0, s14 -; GCN-NEXT: v_mul_hi_u32 v0, s10, v0 -; GCN-NEXT: s_cselect_b64 s[8:9], -1, 0 -; GCN-NEXT: s_or_b32 s8, s8, s9 -; GCN-NEXT: s_addc_u32 s12, s12, s13 -; GCN-NEXT: s_mul_i32 s8, s10, s12 -; GCN-NEXT: v_readfirstlane_b32 s9, v0 -; GCN-NEXT: s_add_i32 s8, s9, s8 -; GCN-NEXT: s_mul_i32 s11, s11, s14 -; GCN-NEXT: s_mul_i32 s9, s10, s14 -; GCN-NEXT: s_add_i32 s8, s8, s11 -; GCN-NEXT: v_mov_b32_e32 v2, s9 -; GCN-NEXT: v_mov_b32_e32 v0, s8 -; GCN-NEXT: v_mul_hi_u32 v3, s12, v2 -; GCN-NEXT: v_mul_hi_u32 v2, s14, v2 -; GCN-NEXT: v_mul_hi_u32 v1, s12, v0 -; GCN-NEXT: v_mul_hi_u32 v0, s14, v0 -; GCN-NEXT: s_mul_i32 s11, s14, s8 -; GCN-NEXT: v_readfirstlane_b32 s15, v2 -; GCN-NEXT: s_add_u32 s11, s15, s11 +; GCN-NEXT: s_addc_u32 s14, 0, s14 +; GCN-NEXT: s_mul_i32 s13, s10, s13 +; GCN-NEXT: v_readfirstlane_b32 s15, v4 +; GCN-NEXT: s_add_u32 s12, s12, s13 +; GCN-NEXT: s_addc_u32 s12, s14, s15 +; GCN-NEXT: v_readfirstlane_b32 s13, v1 +; GCN-NEXT: s_addc_u32 s13, s13, 0 +; GCN-NEXT: s_mul_i32 s11, s10, s11 +; GCN-NEXT: s_add_u32 s11, s12, s11 +; GCN-NEXT: s_addc_u32 s12, 0, s13 +; GCN-NEXT: s_add_u32 s11, s2, s11 +; GCN-NEXT: v_mov_b32_e32 v0, s11 +; GCN-NEXT: v_mul_hi_u32 v0, s8, v0 +; GCN-NEXT: s_addc_u32 s10, s10, s12 +; GCN-NEXT: s_mul_i32 s12, s8, s10 +; GCN-NEXT: s_mul_i32 s9, s9, s11 ; GCN-NEXT: v_readfirstlane_b32 s13, v0 -; GCN-NEXT: s_mul_i32 s9, s12, s9 -; GCN-NEXT: s_addc_u32 s13, 0, s13 -; GCN-NEXT: v_readfirstlane_b32 s10, v3 -; GCN-NEXT: s_add_u32 s9, s11, s9 -; GCN-NEXT: s_addc_u32 s9, s13, s10 -; GCN-NEXT: v_readfirstlane_b32 s10, v1 -; GCN-NEXT: s_addc_u32 s10, s10, 0 -; GCN-NEXT: s_mul_i32 s8, s12, s8 -; GCN-NEXT: s_add_u32 s8, s9, s8 -; GCN-NEXT: s_addc_u32 s10, 0, s10 -; GCN-NEXT: s_add_u32 s11, s14, s8 -; GCN-NEXT: s_cselect_b64 s[8:9], -1, 0 -; GCN-NEXT: s_or_b32 s8, s8, s9 -; GCN-NEXT: s_addc_u32 s10, s12, s10 +; GCN-NEXT: s_add_i32 s12, s13, s12 +; GCN-NEXT: s_mul_i32 s8, s8, s11 +; GCN-NEXT: s_add_i32 s9, s12, s9 +; GCN-NEXT: v_mov_b32_e32 v2, s8 +; GCN-NEXT: v_mov_b32_e32 v0, s9 +; GCN-NEXT: v_mul_hi_u32 v3, s10, v2 +; GCN-NEXT: v_mul_hi_u32 v2, s11, v2 +; GCN-NEXT: v_mul_hi_u32 v1, s10, v0 +; GCN-NEXT: v_mul_hi_u32 v0, s11, v0 +; GCN-NEXT: s_mul_i32 s13, s11, s9 +; GCN-NEXT: v_readfirstlane_b32 s15, v2 +; GCN-NEXT: s_add_u32 s13, s15, s13 +; GCN-NEXT: v_readfirstlane_b32 s14, v0 +; GCN-NEXT: s_mul_i32 s8, s10, s8 +; GCN-NEXT: s_addc_u32 s14, 0, s14 +; GCN-NEXT: v_readfirstlane_b32 s12, v3 +; GCN-NEXT: s_add_u32 s8, s13, s8 +; GCN-NEXT: s_addc_u32 s8, s14, s12 +; GCN-NEXT: v_readfirstlane_b32 s12, v1 +; GCN-NEXT: s_addc_u32 s12, s12, 0 +; GCN-NEXT: s_mul_i32 s9, s10, s9 +; GCN-NEXT: s_add_u32 s8, s8, s9 +; GCN-NEXT: s_addc_u32 s9, 0, s12 +; GCN-NEXT: s_add_u32 s11, s11, s8 +; GCN-NEXT: s_addc_u32 s10, s10, s9 ; GCN-NEXT: s_ashr_i32 s8, s7, 31 ; GCN-NEXT: s_add_u32 s6, s6, s8 ; GCN-NEXT: s_mov_b32 s9, s8 @@ -1071,6 +1054,7 @@ define amdgpu_kernel void @s_test_srem33_64(ptr addrspace(1) %out, i64 %x, i64 % ; GCN-NEXT: v_mul_hi_u32 v0, s4, v0 ; GCN-NEXT: s_addc_u32 s11, 0, s12 ; GCN-NEXT: s_mul_i32 s11, s4, s11 +; GCN-NEXT: s_mov_b32 s2, -1 ; GCN-NEXT: v_readfirstlane_b32 s12, v0 ; GCN-NEXT: s_add_i32 s11, s12, s11 ; GCN-NEXT: s_mul_i32 s12, s5, s10 @@ -1079,11 +1063,9 @@ define amdgpu_kernel void @s_test_srem33_64(ptr addrspace(1) %out, i64 %x, i64 % ; GCN-NEXT: s_mul_i32 s10, s4, s10 ; GCN-NEXT: s_sub_u32 s6, s6, s10 ; GCN-NEXT: s_cselect_b64 s[10:11], -1, 0 -; GCN-NEXT: s_or_b32 s13, s10, s11 ; GCN-NEXT: s_subb_u32 s15, s12, s5 ; GCN-NEXT: s_sub_u32 s16, s6, s4 ; GCN-NEXT: s_cselect_b64 s[12:13], -1, 0 -; GCN-NEXT: s_or_b32 s17, s12, s13 ; GCN-NEXT: s_subb_u32 s17, s15, 0 ; GCN-NEXT: s_cmp_ge_u32 s17, s5 ; GCN-NEXT: s_cselect_b32 s18, -1, 0 @@ -1092,13 +1074,11 @@ define amdgpu_kernel void @s_test_srem33_64(ptr addrspace(1) %out, i64 %x, i64 % ; GCN-NEXT: s_cmp_eq_u32 s17, s5 ; GCN-NEXT: s_cselect_b32 s18, s19, s18 ; GCN-NEXT: s_or_b32 s12, s12, s13 -; GCN-NEXT: s_subb_u32 s15, s15, s5 -; GCN-NEXT: s_sub_u32 s19, s16, s4 -; GCN-NEXT: s_cselect_b64 s[12:13], -1, 0 -; GCN-NEXT: s_or_b32 s12, s12, s13 -; GCN-NEXT: s_subb_u32 s12, s15, 0 +; GCN-NEXT: s_subb_u32 s12, s15, s5 +; GCN-NEXT: s_sub_u32 s13, s16, s4 +; GCN-NEXT: s_subb_u32 s12, s12, 0 ; GCN-NEXT: s_cmp_lg_u32 s18, 0 -; GCN-NEXT: s_cselect_b32 s13, s19, s16 +; GCN-NEXT: s_cselect_b32 s13, s13, s16 ; GCN-NEXT: s_cselect_b32 s12, s12, s17 ; GCN-NEXT: s_or_b32 s10, s10, s11 ; GCN-NEXT: s_subb_u32 s7, s7, s14 @@ -1156,8 +1136,6 @@ define amdgpu_kernel void @s_test_srem33_64(ptr addrspace(1) %out, i64 %x, i64 % ; GCN-IR-NEXT: s_cbranch_vccz .LBB8_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: s_add_u32 s16, s14, 1 -; GCN-IR-NEXT: s_cselect_b64 s[10:11], -1, 0 -; GCN-IR-NEXT: s_or_b32 s10, s10, s11 ; GCN-IR-NEXT: s_addc_u32 s10, s15, 0 ; GCN-IR-NEXT: s_cselect_b64 s[10:11], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s14, 63, s14 @@ -1189,8 +1167,6 @@ define amdgpu_kernel void @s_test_srem33_64(ptr addrspace(1) %out, i64 %x, i64 % ; GCN-IR-NEXT: s_sub_u32 s14, s14, s20 ; GCN-IR-NEXT: s_subb_u32 s15, s15, s21 ; GCN-IR-NEXT: s_add_u32 s18, s18, 1 -; GCN-IR-NEXT: s_cselect_b64 s[20:21], -1, 0 -; GCN-IR-NEXT: s_or_b32 s20, s20, s21 ; GCN-IR-NEXT: s_addc_u32 s19, s19, 0 ; GCN-IR-NEXT: s_cselect_b64 s[20:21], -1, 0 ; GCN-IR-NEXT: s_mov_b64 s[12:13], s[2:3] @@ -1316,8 +1292,7 @@ define amdgpu_kernel void @s_test_srem_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-NEXT: v_cvt_f32_u32_e32 v0, s4 ; GCN-NEXT: v_cvt_f32_u32_e32 v1, s5 ; GCN-NEXT: s_sub_u32 s2, 0, s4 -; GCN-NEXT: s_subb_u32 s8, 0, s5 -; GCN-NEXT: s_mov_b32 s3, 0xf000 +; GCN-NEXT: s_subb_u32 s6, 0, s5 ; GCN-NEXT: v_madmk_f32 v0, v1, 0x4f800000, v0 ; GCN-NEXT: v_rcp_f32_e32 v0, v0 ; GCN-NEXT: v_mul_f32_e32 v0, 0x5f7ffffc, v0 @@ -1327,72 +1302,68 @@ define amdgpu_kernel void @s_test_srem_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-NEXT: v_cvt_u32_f32_e32 v0, v0 ; GCN-NEXT: v_cvt_u32_f32_e32 v1, v1 ; GCN-NEXT: v_mul_hi_u32 v2, s2, v0 -; GCN-NEXT: v_readfirstlane_b32 s9, v1 -; GCN-NEXT: v_readfirstlane_b32 s6, v0 -; GCN-NEXT: s_mul_i32 s7, s2, s9 -; GCN-NEXT: v_readfirstlane_b32 s12, v2 -; GCN-NEXT: s_mul_i32 s10, s8, s6 -; GCN-NEXT: s_mul_i32 s11, s2, s6 -; GCN-NEXT: s_add_i32 s7, s12, s7 -; GCN-NEXT: v_mul_hi_u32 v3, v0, s11 -; GCN-NEXT: s_add_i32 s7, s7, s10 -; GCN-NEXT: v_mul_hi_u32 v0, v0, s7 -; GCN-NEXT: v_mul_hi_u32 v4, v1, s11 -; GCN-NEXT: v_readfirstlane_b32 s10, v3 -; GCN-NEXT: s_mul_i32 s13, s6, s7 -; GCN-NEXT: v_mul_hi_u32 v1, v1, s7 -; GCN-NEXT: s_add_u32 s10, s10, s13 -; GCN-NEXT: v_readfirstlane_b32 s13, v0 -; GCN-NEXT: s_mul_i32 s11, s9, s11 -; GCN-NEXT: s_addc_u32 s13, 0, s13 -; GCN-NEXT: v_readfirstlane_b32 s12, v4 -; GCN-NEXT: s_add_u32 s10, s10, s11 -; GCN-NEXT: s_addc_u32 s10, s13, s12 -; GCN-NEXT: v_readfirstlane_b32 s11, v1 -; GCN-NEXT: s_addc_u32 s11, s11, 0 -; GCN-NEXT: s_mul_i32 s7, s9, s7 -; GCN-NEXT: s_add_u32 s7, s10, s7 -; GCN-NEXT: s_addc_u32 s10, 0, s11 -; GCN-NEXT: s_add_u32 s11, s6, s7 -; GCN-NEXT: v_mov_b32_e32 v0, s11 +; GCN-NEXT: v_readfirstlane_b32 s7, v1 +; GCN-NEXT: v_readfirstlane_b32 s3, v0 +; GCN-NEXT: s_mul_i32 s8, s2, s7 +; GCN-NEXT: v_readfirstlane_b32 s11, v2 +; GCN-NEXT: s_mul_i32 s9, s6, s3 +; GCN-NEXT: s_mul_i32 s10, s2, s3 +; GCN-NEXT: s_add_i32 s8, s11, s8 +; GCN-NEXT: v_mul_hi_u32 v3, v0, s10 +; GCN-NEXT: s_add_i32 s8, s8, s9 +; GCN-NEXT: v_mul_hi_u32 v0, v0, s8 +; GCN-NEXT: v_mul_hi_u32 v4, v1, s10 +; GCN-NEXT: v_readfirstlane_b32 s9, v3 +; GCN-NEXT: v_mul_hi_u32 v1, v1, s8 +; GCN-NEXT: s_mul_i32 s12, s3, s8 +; GCN-NEXT: s_add_u32 s9, s9, s12 +; GCN-NEXT: v_readfirstlane_b32 s12, v0 +; GCN-NEXT: s_mul_i32 s10, s7, s10 +; GCN-NEXT: s_addc_u32 s12, 0, s12 +; GCN-NEXT: v_readfirstlane_b32 s11, v4 +; GCN-NEXT: s_add_u32 s9, s9, s10 +; GCN-NEXT: v_readfirstlane_b32 s13, v1 +; GCN-NEXT: s_addc_u32 s9, s12, s11 +; GCN-NEXT: s_addc_u32 s10, s13, 0 +; GCN-NEXT: s_mul_i32 s8, s7, s8 +; GCN-NEXT: s_add_u32 s8, s9, s8 +; GCN-NEXT: s_addc_u32 s9, 0, s10 +; GCN-NEXT: s_add_u32 s8, s3, s8 +; GCN-NEXT: v_mov_b32_e32 v0, s8 ; GCN-NEXT: v_mul_hi_u32 v0, s2, v0 -; GCN-NEXT: s_cselect_b64 s[6:7], -1, 0 -; GCN-NEXT: s_or_b32 s6, s6, s7 -; GCN-NEXT: s_addc_u32 s9, s9, s10 -; GCN-NEXT: s_mul_i32 s6, s2, s9 -; GCN-NEXT: v_readfirstlane_b32 s7, v0 -; GCN-NEXT: s_add_i32 s6, s7, s6 -; GCN-NEXT: s_mul_i32 s8, s8, s11 -; GCN-NEXT: s_mul_i32 s2, s2, s11 -; GCN-NEXT: s_add_i32 s6, s6, s8 +; GCN-NEXT: s_addc_u32 s7, s7, s9 +; GCN-NEXT: s_mul_i32 s9, s2, s7 +; GCN-NEXT: s_mul_i32 s6, s6, s8 +; GCN-NEXT: v_readfirstlane_b32 s10, v0 +; GCN-NEXT: s_add_i32 s9, s10, s9 +; GCN-NEXT: s_mul_i32 s2, s2, s8 +; GCN-NEXT: s_add_i32 s6, s9, s6 ; GCN-NEXT: v_mov_b32_e32 v2, s2 ; GCN-NEXT: v_mov_b32_e32 v0, s6 -; GCN-NEXT: v_mul_hi_u32 v3, s9, v2 -; GCN-NEXT: v_mul_hi_u32 v2, s11, v2 -; GCN-NEXT: v_mul_hi_u32 v1, s9, v0 -; GCN-NEXT: v_mul_hi_u32 v0, s11, v0 -; GCN-NEXT: s_mul_i32 s8, s11, s6 +; GCN-NEXT: v_mul_hi_u32 v3, s7, v2 +; GCN-NEXT: v_mul_hi_u32 v2, s8, v2 +; GCN-NEXT: v_mul_hi_u32 v1, s7, v0 +; GCN-NEXT: v_mul_hi_u32 v0, s8, v0 +; GCN-NEXT: s_mul_i32 s10, s8, s6 ; GCN-NEXT: v_readfirstlane_b32 s12, v2 -; GCN-NEXT: s_add_u32 s8, s12, s8 -; GCN-NEXT: v_readfirstlane_b32 s10, v0 -; GCN-NEXT: s_mul_i32 s2, s9, s2 -; GCN-NEXT: s_addc_u32 s10, 0, s10 -; GCN-NEXT: v_readfirstlane_b32 s7, v3 -; GCN-NEXT: s_add_u32 s2, s8, s2 -; GCN-NEXT: s_addc_u32 s2, s10, s7 -; GCN-NEXT: v_readfirstlane_b32 s7, v1 -; GCN-NEXT: s_addc_u32 s7, s7, 0 -; GCN-NEXT: s_mul_i32 s6, s9, s6 +; GCN-NEXT: s_add_u32 s10, s12, s10 +; GCN-NEXT: v_readfirstlane_b32 s11, v0 +; GCN-NEXT: s_mul_i32 s2, s7, s2 +; GCN-NEXT: s_addc_u32 s11, 0, s11 +; GCN-NEXT: v_readfirstlane_b32 s9, v3 +; GCN-NEXT: s_add_u32 s2, s10, s2 +; GCN-NEXT: s_addc_u32 s2, s11, s9 +; GCN-NEXT: v_readfirstlane_b32 s9, v1 +; GCN-NEXT: s_addc_u32 s9, s9, 0 +; GCN-NEXT: s_mul_i32 s6, s7, s6 ; GCN-NEXT: s_add_u32 s2, s2, s6 -; GCN-NEXT: s_addc_u32 s8, 0, s7 -; GCN-NEXT: s_add_u32 s2, s11, s2 -; GCN-NEXT: s_cselect_b64 s[6:7], -1, 0 -; GCN-NEXT: s_or_b32 s6, s6, s7 -; GCN-NEXT: s_addc_u32 s6, s9, s8 +; GCN-NEXT: s_addc_u32 s6, 0, s9 +; GCN-NEXT: s_add_u32 s2, s8, s2 +; GCN-NEXT: s_addc_u32 s6, s7, s6 ; GCN-NEXT: v_mul_hi_u32 v1, s2, 24 ; GCN-NEXT: v_mul_hi_u32 v0, s6, 24 ; GCN-NEXT: s_mul_i32 s6, s6, 24 -; GCN-NEXT: s_mov_b32 s2, -1 +; GCN-NEXT: s_mov_b32 s3, 0xf000 ; GCN-NEXT: v_readfirstlane_b32 s8, v1 ; GCN-NEXT: v_readfirstlane_b32 s7, v0 ; GCN-NEXT: s_add_u32 s6, s8, s6 @@ -1401,16 +1372,15 @@ define amdgpu_kernel void @s_test_srem_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-NEXT: v_mul_hi_u32 v0, s4, v0 ; GCN-NEXT: s_mul_i32 s7, s5, s6 ; GCN-NEXT: s_mul_i32 s6, s4, s6 +; GCN-NEXT: s_mov_b32 s2, -1 ; GCN-NEXT: v_readfirstlane_b32 s8, v0 ; GCN-NEXT: s_add_i32 s10, s8, s7 ; GCN-NEXT: s_sub_i32 s8, 0, s10 ; GCN-NEXT: s_sub_u32 s11, 24, s6 ; GCN-NEXT: s_cselect_b64 s[6:7], -1, 0 -; GCN-NEXT: s_or_b32 s9, s6, s7 ; GCN-NEXT: s_subb_u32 s12, s8, s5 ; GCN-NEXT: s_sub_u32 s13, s11, s4 ; GCN-NEXT: s_cselect_b64 s[8:9], -1, 0 -; GCN-NEXT: s_or_b32 s14, s8, s9 ; GCN-NEXT: s_subb_u32 s14, s12, 0 ; GCN-NEXT: s_cmp_ge_u32 s14, s5 ; GCN-NEXT: s_cselect_b32 s15, -1, 0 @@ -1419,13 +1389,11 @@ define amdgpu_kernel void @s_test_srem_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-NEXT: s_cmp_eq_u32 s14, s5 ; GCN-NEXT: s_cselect_b32 s15, s16, s15 ; GCN-NEXT: s_or_b32 s8, s8, s9 -; GCN-NEXT: s_subb_u32 s12, s12, s5 -; GCN-NEXT: s_sub_u32 s16, s13, s4 -; GCN-NEXT: s_cselect_b64 s[8:9], -1, 0 -; GCN-NEXT: s_or_b32 s8, s8, s9 -; GCN-NEXT: s_subb_u32 s8, s12, 0 +; GCN-NEXT: s_subb_u32 s8, s12, s5 +; GCN-NEXT: s_sub_u32 s9, s13, s4 +; GCN-NEXT: s_subb_u32 s8, s8, 0 ; GCN-NEXT: s_cmp_lg_u32 s15, 0 -; GCN-NEXT: s_cselect_b32 s9, s16, s13 +; GCN-NEXT: s_cselect_b32 s9, s9, s13 ; GCN-NEXT: s_cselect_b32 s8, s8, s14 ; GCN-NEXT: s_or_b32 s6, s6, s7 ; GCN-NEXT: s_subb_u32 s6, 0, s10 @@ -1468,8 +1436,6 @@ define amdgpu_kernel void @s_test_srem_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR-NEXT: s_cbranch_vccz .LBB10_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: s_add_u32 s8, s2, 1 -; GCN-IR-NEXT: s_cselect_b64 s[10:11], -1, 0 -; GCN-IR-NEXT: s_or_b32 s9, s10, s11 ; GCN-IR-NEXT: s_addc_u32 s3, s3, 0 ; GCN-IR-NEXT: s_cselect_b64 s[10:11], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s2, 63, s2 @@ -1500,8 +1466,6 @@ define amdgpu_kernel void @s_test_srem_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR-NEXT: s_sub_u32 s10, s10, s16 ; GCN-IR-NEXT: s_subb_u32 s11, s11, s17 ; GCN-IR-NEXT: s_add_u32 s14, s14, 1 -; GCN-IR-NEXT: s_cselect_b64 s[16:17], -1, 0 -; GCN-IR-NEXT: s_or_b32 s16, s16, s17 ; GCN-IR-NEXT: s_addc_u32 s15, s15, 0 ; GCN-IR-NEXT: s_cselect_b64 s[16:17], -1, 0 ; GCN-IR-NEXT: s_mov_b64 s[8:9], s[6:7] diff --git a/llvm/test/CodeGen/AMDGPU/uaddo.ll b/llvm/test/CodeGen/AMDGPU/uaddo.ll index bdd22f25..b000fae 100644 --- a/llvm/test/CodeGen/AMDGPU/uaddo.ll +++ b/llvm/test/CodeGen/AMDGPU/uaddo.ll @@ -15,10 +15,8 @@ define amdgpu_kernel void @s_uaddo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 % ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_mov_b32 s4, s0 ; SI-NEXT: s_add_u32 s2, s2, s8 -; SI-NEXT: s_mov_b32 s5, s1 -; SI-NEXT: s_cselect_b64 s[0:1], -1, 0 -; SI-NEXT: s_or_b32 s0, s0, s1 ; SI-NEXT: s_addc_u32 s3, s3, s9 +; SI-NEXT: s_mov_b32 s5, s1 ; SI-NEXT: s_cselect_b64 s[0:1], -1, 0 ; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] ; SI-NEXT: v_mov_b32_e32 v1, s3 @@ -433,8 +431,6 @@ 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 s4, s4, s6 -; SI-NEXT: s_cselect_b64 s[12:13], -1, 0 -; SI-NEXT: s_or_b32 s6, s12, s13 ; SI-NEXT: s_addc_u32 s5, s5, s7 ; SI-NEXT: s_mov_b32 s8, s0 ; SI-NEXT: s_mov_b32 s9, s1 diff --git a/llvm/test/CodeGen/AMDGPU/udiv64.ll b/llvm/test/CodeGen/AMDGPU/udiv64.ll index fd461ac..775483c 100644 --- a/llvm/test/CodeGen/AMDGPU/udiv64.ll +++ b/llvm/test/CodeGen/AMDGPU/udiv64.ll @@ -146,8 +146,6 @@ define amdgpu_kernel void @s_test_udiv_i64(ptr addrspace(1) %out, i64 %x, i64 %y ; GCN-IR-NEXT: s_cbranch_vccz .LBB0_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: s_add_u32 s14, s12, 1 -; GCN-IR-NEXT: s_cselect_b64 s[8:9], -1, 0 -; GCN-IR-NEXT: s_or_b32 s8, s8, s9 ; GCN-IR-NEXT: s_addc_u32 s8, s13, 0 ; GCN-IR-NEXT: s_cselect_b64 s[8:9], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s12, 63, s12 @@ -179,8 +177,6 @@ define amdgpu_kernel void @s_test_udiv_i64(ptr addrspace(1) %out, i64 %x, i64 %y ; GCN-IR-NEXT: s_sub_u32 s12, s12, s16 ; GCN-IR-NEXT: s_subb_u32 s13, s13, s17 ; GCN-IR-NEXT: s_add_u32 s10, s10, 1 -; GCN-IR-NEXT: s_cselect_b64 s[16:17], -1, 0 -; GCN-IR-NEXT: s_or_b32 s16, s16, s17 ; GCN-IR-NEXT: s_addc_u32 s11, s11, 0 ; GCN-IR-NEXT: s_cselect_b64 s[16:17], -1, 0 ; GCN-IR-NEXT: s_mov_b64 s[2:3], s[4:5] @@ -786,12 +782,11 @@ define amdgpu_kernel void @s_test_udiv_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-LABEL: s_test_udiv_k_num_i64: ; GCN: ; %bb.0: ; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 -; GCN-NEXT: s_mov_b32 s7, 0xf000 ; GCN-NEXT: s_waitcnt lgkmcnt(0) ; GCN-NEXT: v_cvt_f32_u32_e32 v0, s2 ; GCN-NEXT: v_cvt_f32_u32_e32 v1, s3 -; GCN-NEXT: s_sub_u32 s6, 0, s2 -; GCN-NEXT: s_subb_u32 s8, 0, s3 +; GCN-NEXT: s_sub_u32 s4, 0, s2 +; GCN-NEXT: s_subb_u32 s5, 0, s3 ; GCN-NEXT: v_madmk_f32 v0, v1, 0x4f800000, v0 ; GCN-NEXT: v_rcp_f32_e32 v0, v0 ; GCN-NEXT: v_mul_f32_e32 v0, 0x5f7ffffc, v0 @@ -800,118 +795,112 @@ define amdgpu_kernel void @s_test_udiv_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-NEXT: v_madmk_f32 v0, v1, 0xcf800000, v0 ; GCN-NEXT: v_cvt_u32_f32_e32 v0, v0 ; GCN-NEXT: v_cvt_u32_f32_e32 v1, v1 -; GCN-NEXT: v_mul_hi_u32 v2, s6, v0 +; GCN-NEXT: v_mul_hi_u32 v2, s4, v0 +; GCN-NEXT: v_readfirstlane_b32 s6, v1 +; GCN-NEXT: v_readfirstlane_b32 s7, v0 +; GCN-NEXT: s_mul_i32 s8, s4, s6 +; GCN-NEXT: v_readfirstlane_b32 s11, v2 +; GCN-NEXT: s_mul_i32 s9, s5, s7 +; GCN-NEXT: s_mul_i32 s10, s4, s7 +; GCN-NEXT: s_add_i32 s8, s11, s8 +; GCN-NEXT: v_mul_hi_u32 v3, v0, s10 +; GCN-NEXT: s_add_i32 s8, s8, s9 +; GCN-NEXT: v_mul_hi_u32 v0, v0, s8 +; GCN-NEXT: v_mul_hi_u32 v4, v1, s10 +; GCN-NEXT: v_readfirstlane_b32 s9, v3 +; GCN-NEXT: s_mul_i32 s12, s7, s8 +; GCN-NEXT: v_mul_hi_u32 v1, v1, s8 +; GCN-NEXT: s_add_u32 s9, s9, s12 +; GCN-NEXT: v_readfirstlane_b32 s12, v0 +; GCN-NEXT: s_mul_i32 s10, s6, s10 +; GCN-NEXT: s_addc_u32 s12, 0, s12 +; GCN-NEXT: v_readfirstlane_b32 s11, v4 +; GCN-NEXT: s_add_u32 s9, s9, s10 +; GCN-NEXT: v_readfirstlane_b32 s13, v1 +; GCN-NEXT: s_addc_u32 s9, s12, s11 +; GCN-NEXT: s_mul_i32 s8, s6, s8 +; GCN-NEXT: s_addc_u32 s10, s13, 0 +; GCN-NEXT: s_add_u32 s8, s9, s8 +; GCN-NEXT: s_addc_u32 s9, 0, s10 +; GCN-NEXT: s_add_u32 s8, s7, s8 +; GCN-NEXT: v_mov_b32_e32 v0, s8 +; GCN-NEXT: v_mul_hi_u32 v0, s4, v0 +; GCN-NEXT: s_addc_u32 s6, s6, s9 +; GCN-NEXT: s_mul_i32 s9, s4, s6 +; GCN-NEXT: s_mul_i32 s5, s5, s8 +; GCN-NEXT: v_readfirstlane_b32 s10, v0 +; GCN-NEXT: s_add_i32 s9, s10, s9 +; GCN-NEXT: s_mul_i32 s4, s4, s8 +; GCN-NEXT: s_add_i32 s5, s9, s5 +; GCN-NEXT: v_mov_b32_e32 v2, s4 +; GCN-NEXT: v_mov_b32_e32 v0, s5 +; GCN-NEXT: v_mul_hi_u32 v3, s6, v2 +; GCN-NEXT: v_mul_hi_u32 v2, s8, v2 +; GCN-NEXT: v_mul_hi_u32 v1, s6, v0 +; GCN-NEXT: v_mul_hi_u32 v0, s8, v0 +; GCN-NEXT: s_mul_i32 s10, s8, s5 +; GCN-NEXT: v_readfirstlane_b32 s12, v2 +; GCN-NEXT: s_add_u32 s10, s12, s10 +; GCN-NEXT: v_readfirstlane_b32 s11, v0 +; GCN-NEXT: s_mul_i32 s4, s6, s4 +; GCN-NEXT: s_addc_u32 s11, 0, s11 +; GCN-NEXT: v_readfirstlane_b32 s9, v3 +; GCN-NEXT: s_add_u32 s4, s10, s4 +; GCN-NEXT: s_addc_u32 s4, s11, s9 ; GCN-NEXT: v_readfirstlane_b32 s9, v1 +; GCN-NEXT: s_addc_u32 s9, s9, 0 +; GCN-NEXT: s_mul_i32 s5, s6, s5 +; GCN-NEXT: s_add_u32 s4, s4, s5 +; GCN-NEXT: s_addc_u32 s5, 0, s9 +; GCN-NEXT: s_add_u32 s4, s8, s4 +; GCN-NEXT: s_addc_u32 s5, s6, s5 +; GCN-NEXT: v_mul_hi_u32 v1, s4, 24 +; GCN-NEXT: v_mul_hi_u32 v0, s5, 24 +; GCN-NEXT: s_mul_i32 s5, s5, 24 +; GCN-NEXT: s_mov_b32 s7, 0xf000 +; GCN-NEXT: v_readfirstlane_b32 s8, v1 ; GCN-NEXT: v_readfirstlane_b32 s4, v0 -; GCN-NEXT: s_mul_i32 s5, s6, s9 -; GCN-NEXT: v_readfirstlane_b32 s12, v2 -; GCN-NEXT: s_mul_i32 s10, s8, s4 -; GCN-NEXT: s_mul_i32 s11, s6, s4 -; GCN-NEXT: s_add_i32 s5, s12, s5 -; GCN-NEXT: v_mul_hi_u32 v3, v0, s11 -; GCN-NEXT: s_add_i32 s5, s5, s10 -; GCN-NEXT: v_mul_hi_u32 v0, v0, s5 -; GCN-NEXT: v_mul_hi_u32 v4, v1, s11 -; GCN-NEXT: v_readfirstlane_b32 s10, v3 -; GCN-NEXT: v_mul_hi_u32 v1, v1, s5 -; GCN-NEXT: s_mul_i32 s13, s4, s5 -; GCN-NEXT: s_add_u32 s10, s10, s13 -; GCN-NEXT: v_readfirstlane_b32 s13, v0 -; GCN-NEXT: s_mul_i32 s11, s9, s11 -; GCN-NEXT: s_addc_u32 s13, 0, s13 -; GCN-NEXT: v_readfirstlane_b32 s12, v4 -; GCN-NEXT: s_add_u32 s10, s10, s11 -; GCN-NEXT: v_readfirstlane_b32 s14, v1 -; GCN-NEXT: s_addc_u32 s10, s13, s12 -; GCN-NEXT: s_addc_u32 s11, s14, 0 -; GCN-NEXT: s_mul_i32 s5, s9, s5 -; GCN-NEXT: s_add_u32 s5, s10, s5 -; GCN-NEXT: s_addc_u32 s10, 0, s11 -; GCN-NEXT: s_add_u32 s11, s4, s5 -; GCN-NEXT: v_mov_b32_e32 v0, s11 -; GCN-NEXT: v_mul_hi_u32 v0, s6, v0 -; GCN-NEXT: s_cselect_b64 s[4:5], -1, 0 -; GCN-NEXT: s_or_b32 s4, s4, s5 -; GCN-NEXT: s_addc_u32 s9, s9, s10 -; GCN-NEXT: s_mul_i32 s4, s6, s9 -; GCN-NEXT: v_readfirstlane_b32 s5, v0 -; GCN-NEXT: s_add_i32 s4, s5, s4 -; GCN-NEXT: s_mul_i32 s8, s8, s11 -; GCN-NEXT: s_mul_i32 s5, s6, s11 -; GCN-NEXT: s_add_i32 s4, s4, s8 -; GCN-NEXT: v_mov_b32_e32 v2, s5 -; GCN-NEXT: v_mov_b32_e32 v0, s4 -; GCN-NEXT: v_mul_hi_u32 v3, s9, v2 -; GCN-NEXT: v_mul_hi_u32 v2, s11, v2 -; GCN-NEXT: v_mul_hi_u32 v1, s9, v0 -; GCN-NEXT: v_mul_hi_u32 v0, s11, v0 -; GCN-NEXT: s_mul_i32 s8, s11, s4 -; GCN-NEXT: v_readfirstlane_b32 s12, v2 -; GCN-NEXT: s_add_u32 s8, s12, s8 -; GCN-NEXT: v_readfirstlane_b32 s10, v0 -; GCN-NEXT: s_mul_i32 s5, s9, s5 -; GCN-NEXT: s_addc_u32 s10, 0, s10 -; GCN-NEXT: v_readfirstlane_b32 s6, v3 ; GCN-NEXT: s_add_u32 s5, s8, s5 -; GCN-NEXT: s_addc_u32 s5, s10, s6 -; GCN-NEXT: v_readfirstlane_b32 s6, v1 -; GCN-NEXT: s_addc_u32 s6, s6, 0 -; GCN-NEXT: s_mul_i32 s4, s9, s4 -; GCN-NEXT: s_add_u32 s4, s5, s4 -; GCN-NEXT: s_addc_u32 s6, 0, s6 -; GCN-NEXT: s_add_u32 s8, s11, s4 -; GCN-NEXT: s_cselect_b64 s[4:5], -1, 0 -; GCN-NEXT: s_or_b32 s4, s4, s5 -; GCN-NEXT: s_addc_u32 s4, s9, s6 -; GCN-NEXT: v_mul_hi_u32 v1, s8, 24 -; GCN-NEXT: v_mul_hi_u32 v0, s4, 24 -; GCN-NEXT: s_mul_i32 s4, s4, 24 -; GCN-NEXT: s_mov_b32 s6, -1 -; GCN-NEXT: v_readfirstlane_b32 s8, v1 -; GCN-NEXT: v_readfirstlane_b32 s5, v0 -; GCN-NEXT: s_add_u32 s4, s8, s4 -; GCN-NEXT: s_addc_u32 s10, 0, s5 -; GCN-NEXT: v_mov_b32_e32 v0, s10 +; GCN-NEXT: s_addc_u32 s8, 0, s4 +; GCN-NEXT: v_mov_b32_e32 v0, s8 ; GCN-NEXT: v_mul_hi_u32 v0, s2, v0 ; GCN-NEXT: s_mov_b32 s4, s0 ; GCN-NEXT: s_mov_b32 s5, s1 -; GCN-NEXT: s_mul_i32 s0, s3, s10 +; GCN-NEXT: s_mul_i32 s0, s3, s8 ; GCN-NEXT: v_readfirstlane_b32 s1, v0 -; GCN-NEXT: s_add_i32 s11, s1, s0 -; GCN-NEXT: s_sub_i32 s8, 0, s11 -; GCN-NEXT: s_mul_i32 s0, s2, s10 -; GCN-NEXT: s_sub_u32 s12, 24, s0 +; GCN-NEXT: s_add_i32 s9, s1, s0 +; GCN-NEXT: s_sub_i32 s10, 0, s9 +; GCN-NEXT: s_mul_i32 s0, s2, s8 +; GCN-NEXT: s_sub_u32 s11, 24, s0 ; GCN-NEXT: s_cselect_b64 s[0:1], -1, 0 -; GCN-NEXT: s_or_b32 s9, s0, s1 -; GCN-NEXT: s_subb_u32 s13, s8, s3 -; GCN-NEXT: s_sub_u32 s14, s12, s2 -; GCN-NEXT: s_cselect_b64 s[8:9], -1, 0 -; GCN-NEXT: s_or_b32 s8, s8, s9 -; GCN-NEXT: s_subb_u32 s8, s13, 0 -; GCN-NEXT: s_cmp_ge_u32 s8, s3 -; GCN-NEXT: s_cselect_b32 s9, -1, 0 -; GCN-NEXT: s_cmp_ge_u32 s14, s2 +; GCN-NEXT: s_subb_u32 s10, s10, s3 +; GCN-NEXT: s_sub_u32 s12, s11, s2 +; GCN-NEXT: s_subb_u32 s10, s10, 0 +; GCN-NEXT: s_cmp_ge_u32 s10, s3 ; GCN-NEXT: s_cselect_b32 s13, -1, 0 -; GCN-NEXT: s_cmp_eq_u32 s8, s3 -; GCN-NEXT: s_cselect_b32 s8, s13, s9 -; GCN-NEXT: s_add_u32 s9, s10, 1 +; GCN-NEXT: s_cmp_ge_u32 s12, s2 +; GCN-NEXT: s_cselect_b32 s12, -1, 0 +; GCN-NEXT: s_cmp_eq_u32 s10, s3 +; GCN-NEXT: s_cselect_b32 s10, s12, s13 +; GCN-NEXT: s_add_u32 s12, s8, 1 ; GCN-NEXT: s_addc_u32 s13, 0, 0 -; GCN-NEXT: s_add_u32 s14, s10, 2 +; GCN-NEXT: s_add_u32 s14, s8, 2 ; GCN-NEXT: s_addc_u32 s15, 0, 0 -; GCN-NEXT: s_cmp_lg_u32 s8, 0 -; GCN-NEXT: s_cselect_b32 s8, s14, s9 -; GCN-NEXT: s_cselect_b32 s9, s15, s13 +; GCN-NEXT: s_cmp_lg_u32 s10, 0 +; GCN-NEXT: s_cselect_b32 s10, s14, s12 +; GCN-NEXT: s_cselect_b32 s12, s15, s13 ; GCN-NEXT: s_or_b32 s0, s0, s1 -; GCN-NEXT: s_subb_u32 s0, 0, s11 +; GCN-NEXT: s_subb_u32 s0, 0, s9 ; GCN-NEXT: s_cmp_ge_u32 s0, s3 ; GCN-NEXT: s_cselect_b32 s1, -1, 0 -; GCN-NEXT: s_cmp_ge_u32 s12, s2 +; GCN-NEXT: s_cmp_ge_u32 s11, s2 ; GCN-NEXT: s_cselect_b32 s2, -1, 0 ; GCN-NEXT: s_cmp_eq_u32 s0, s3 ; GCN-NEXT: s_cselect_b32 s0, s2, s1 ; GCN-NEXT: s_cmp_lg_u32 s0, 0 -; GCN-NEXT: s_cselect_b32 s0, s9, 0 -; GCN-NEXT: s_cselect_b32 s1, s8, s10 +; GCN-NEXT: s_cselect_b32 s0, s12, 0 +; GCN-NEXT: s_cselect_b32 s1, s10, s8 +; GCN-NEXT: s_mov_b32 s6, -1 ; GCN-NEXT: v_mov_b32_e32 v0, s1 ; GCN-NEXT: v_mov_b32_e32 v1, s0 ; GCN-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0 @@ -937,8 +926,6 @@ define amdgpu_kernel void @s_test_udiv_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR-NEXT: s_cbranch_vccz .LBB8_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: s_add_u32 s10, s8, 1 -; GCN-IR-NEXT: s_cselect_b64 s[6:7], -1, 0 -; GCN-IR-NEXT: s_or_b32 s6, s6, s7 ; GCN-IR-NEXT: s_addc_u32 s6, s9, 0 ; GCN-IR-NEXT: s_cselect_b64 s[6:7], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s8, 63, s8 @@ -969,8 +956,6 @@ define amdgpu_kernel void @s_test_udiv_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR-NEXT: s_sub_u32 s10, s10, s16 ; GCN-IR-NEXT: s_subb_u32 s11, s11, s17 ; GCN-IR-NEXT: s_add_u32 s14, s14, 1 -; GCN-IR-NEXT: s_cselect_b64 s[16:17], -1, 0 -; GCN-IR-NEXT: s_or_b32 s16, s16, s17 ; GCN-IR-NEXT: s_addc_u32 s15, s15, 0 ; GCN-IR-NEXT: s_cselect_b64 s[16:17], -1, 0 ; GCN-IR-NEXT: s_mov_b64 s[8:9], s[4:5] @@ -1307,8 +1292,6 @@ define amdgpu_kernel void @s_test_udiv_k_den_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR-NEXT: s_cbranch_vccz .LBB11_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: s_add_u32 s11, s8, 1 -; GCN-IR-NEXT: s_cselect_b64 s[6:7], -1, 0 -; GCN-IR-NEXT: s_or_b32 s6, s6, s7 ; GCN-IR-NEXT: s_addc_u32 s6, s9, 0 ; GCN-IR-NEXT: s_cselect_b64 s[6:7], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s8, 63, s8 @@ -1336,8 +1319,6 @@ define amdgpu_kernel void @s_test_udiv_k_den_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR-NEXT: s_sub_u32 s2, s2, s8 ; GCN-IR-NEXT: s_subb_u32 s3, s3, 0 ; GCN-IR-NEXT: s_add_u32 s10, s10, 1 -; GCN-IR-NEXT: s_cselect_b64 s[12:13], -1, 0 -; GCN-IR-NEXT: s_or_b32 s12, s12, s13 ; GCN-IR-NEXT: s_addc_u32 s11, s11, 0 ; GCN-IR-NEXT: s_cselect_b64 s[12:13], -1, 0 ; GCN-IR-NEXT: s_mov_b64 s[8:9], s[4:5] diff --git a/llvm/test/CodeGen/AMDGPU/urem64.ll b/llvm/test/CodeGen/AMDGPU/urem64.ll index 137dc1f..28e6627 100644 --- a/llvm/test/CodeGen/AMDGPU/urem64.ll +++ b/llvm/test/CodeGen/AMDGPU/urem64.ll @@ -8,12 +8,11 @@ define amdgpu_kernel void @s_test_urem_i64(ptr addrspace(1) %out, i64 %x, i64 %y ; GCN-NEXT: s_load_dwordx2 s[8:9], s[4:5], 0xd ; GCN-NEXT: s_load_dwordx4 s[4:7], s[4:5], 0x9 ; GCN-NEXT: s_mov_b32 s3, 0xf000 -; GCN-NEXT: s_mov_b32 s2, -1 ; GCN-NEXT: s_waitcnt lgkmcnt(0) ; GCN-NEXT: v_cvt_f32_u32_e32 v0, s8 ; GCN-NEXT: v_cvt_f32_u32_e32 v1, s9 -; GCN-NEXT: s_sub_u32 s10, 0, s8 -; GCN-NEXT: s_subb_u32 s11, 0, s9 +; GCN-NEXT: s_sub_u32 s0, 0, s8 +; GCN-NEXT: s_subb_u32 s1, 0, s9 ; GCN-NEXT: v_madmk_f32 v0, v1, 0x4f800000, v0 ; GCN-NEXT: v_rcp_f32_e32 v0, v0 ; GCN-NEXT: v_mul_f32_e32 v0, 0x5f7ffffc, v0 @@ -22,69 +21,65 @@ define amdgpu_kernel void @s_test_urem_i64(ptr addrspace(1) %out, i64 %x, i64 %y ; GCN-NEXT: v_madmk_f32 v0, v1, 0xcf800000, v0 ; GCN-NEXT: v_cvt_u32_f32_e32 v0, v0 ; GCN-NEXT: v_cvt_u32_f32_e32 v1, v1 -; GCN-NEXT: v_mul_hi_u32 v2, s10, v0 -; GCN-NEXT: v_readfirstlane_b32 s12, v1 -; GCN-NEXT: v_readfirstlane_b32 s0, v0 -; GCN-NEXT: s_mul_i32 s1, s10, s12 -; GCN-NEXT: v_readfirstlane_b32 s15, v2 -; GCN-NEXT: s_mul_i32 s13, s11, s0 -; GCN-NEXT: s_mul_i32 s14, s10, s0 -; GCN-NEXT: s_add_i32 s1, s15, s1 -; GCN-NEXT: v_mul_hi_u32 v3, v0, s14 -; GCN-NEXT: s_add_i32 s1, s1, s13 -; GCN-NEXT: v_mul_hi_u32 v0, v0, s1 -; GCN-NEXT: v_mul_hi_u32 v4, v1, s14 -; GCN-NEXT: v_readfirstlane_b32 s13, v3 -; GCN-NEXT: s_mul_i32 s15, s0, s1 -; GCN-NEXT: v_mul_hi_u32 v1, v1, s1 -; GCN-NEXT: s_add_u32 s13, s13, s15 +; GCN-NEXT: v_mul_hi_u32 v2, s0, v0 +; GCN-NEXT: v_readfirstlane_b32 s10, v1 +; GCN-NEXT: v_readfirstlane_b32 s2, v0 +; GCN-NEXT: s_mul_i32 s11, s0, s10 +; GCN-NEXT: v_readfirstlane_b32 s14, v2 +; GCN-NEXT: s_mul_i32 s12, s1, s2 +; GCN-NEXT: s_mul_i32 s13, s0, s2 +; GCN-NEXT: s_add_i32 s11, s14, s11 +; GCN-NEXT: v_mul_hi_u32 v3, v0, s13 +; GCN-NEXT: s_add_i32 s11, s11, s12 +; GCN-NEXT: v_mul_hi_u32 v0, v0, s11 +; GCN-NEXT: v_mul_hi_u32 v4, v1, s13 +; GCN-NEXT: v_readfirstlane_b32 s12, v3 +; GCN-NEXT: s_mul_i32 s15, s2, s11 +; GCN-NEXT: v_mul_hi_u32 v1, v1, s11 +; GCN-NEXT: s_add_u32 s12, s12, s15 ; GCN-NEXT: v_readfirstlane_b32 s15, v0 -; GCN-NEXT: s_mul_i32 s14, s12, s14 +; GCN-NEXT: s_mul_i32 s13, s10, s13 ; GCN-NEXT: s_addc_u32 s15, 0, s15 -; GCN-NEXT: v_readfirstlane_b32 s16, v4 -; GCN-NEXT: s_add_u32 s13, s13, s14 -; GCN-NEXT: s_addc_u32 s13, s15, s16 -; GCN-NEXT: v_readfirstlane_b32 s14, v1 -; GCN-NEXT: s_addc_u32 s14, s14, 0 -; GCN-NEXT: s_mul_i32 s1, s12, s1 -; GCN-NEXT: s_add_u32 s1, s13, s1 -; GCN-NEXT: s_addc_u32 s13, 0, s14 -; GCN-NEXT: s_add_u32 s14, s0, s1 -; GCN-NEXT: v_mov_b32_e32 v0, s14 -; GCN-NEXT: v_mul_hi_u32 v0, s10, v0 -; GCN-NEXT: s_cselect_b64 s[0:1], -1, 0 -; GCN-NEXT: s_or_b32 s0, s0, s1 -; GCN-NEXT: s_addc_u32 s12, s12, s13 -; GCN-NEXT: s_mul_i32 s0, s10, s12 -; GCN-NEXT: v_readfirstlane_b32 s1, v0 -; GCN-NEXT: s_add_i32 s0, s1, s0 -; GCN-NEXT: s_mul_i32 s11, s11, s14 -; GCN-NEXT: s_mul_i32 s1, s10, s14 -; GCN-NEXT: s_add_i32 s0, s0, s11 -; GCN-NEXT: v_mov_b32_e32 v2, s1 -; GCN-NEXT: v_mov_b32_e32 v0, s0 -; GCN-NEXT: v_mul_hi_u32 v3, s12, v2 -; GCN-NEXT: v_mul_hi_u32 v2, s14, v2 -; GCN-NEXT: v_mul_hi_u32 v1, s12, v0 -; GCN-NEXT: v_mul_hi_u32 v0, s14, v0 -; GCN-NEXT: s_mul_i32 s11, s14, s0 -; GCN-NEXT: v_readfirstlane_b32 s15, v2 -; GCN-NEXT: s_add_u32 s11, s15, s11 +; GCN-NEXT: v_readfirstlane_b32 s14, v4 +; GCN-NEXT: s_add_u32 s12, s12, s13 +; GCN-NEXT: s_addc_u32 s12, s15, s14 +; GCN-NEXT: v_readfirstlane_b32 s13, v1 +; GCN-NEXT: s_addc_u32 s13, s13, 0 +; GCN-NEXT: s_mul_i32 s11, s10, s11 +; GCN-NEXT: s_add_u32 s11, s12, s11 +; GCN-NEXT: s_addc_u32 s12, 0, s13 +; GCN-NEXT: s_add_u32 s11, s2, s11 +; GCN-NEXT: v_mov_b32_e32 v0, s11 +; GCN-NEXT: v_mul_hi_u32 v0, s0, v0 +; GCN-NEXT: s_addc_u32 s10, s10, s12 +; GCN-NEXT: s_mul_i32 s12, s0, s10 +; GCN-NEXT: s_mul_i32 s1, s1, s11 ; GCN-NEXT: v_readfirstlane_b32 s13, v0 -; GCN-NEXT: s_mul_i32 s1, s12, s1 -; GCN-NEXT: s_addc_u32 s13, 0, s13 -; GCN-NEXT: v_readfirstlane_b32 s10, v3 -; GCN-NEXT: s_add_u32 s1, s11, s1 -; GCN-NEXT: s_addc_u32 s1, s13, s10 -; GCN-NEXT: v_readfirstlane_b32 s10, v1 -; GCN-NEXT: s_addc_u32 s10, s10, 0 -; GCN-NEXT: s_mul_i32 s0, s12, s0 -; GCN-NEXT: s_add_u32 s0, s1, s0 -; GCN-NEXT: s_addc_u32 s10, 0, s10 -; GCN-NEXT: s_add_u32 s11, s14, s0 -; GCN-NEXT: s_cselect_b64 s[0:1], -1, 0 -; GCN-NEXT: s_or_b32 s0, s0, s1 -; GCN-NEXT: s_addc_u32 s1, s12, s10 +; GCN-NEXT: s_add_i32 s12, s13, s12 +; GCN-NEXT: s_mul_i32 s0, s0, s11 +; GCN-NEXT: s_add_i32 s1, s12, s1 +; GCN-NEXT: v_mov_b32_e32 v2, s0 +; GCN-NEXT: v_mov_b32_e32 v0, s1 +; GCN-NEXT: v_mul_hi_u32 v3, s10, v2 +; GCN-NEXT: v_mul_hi_u32 v2, s11, v2 +; GCN-NEXT: v_mul_hi_u32 v1, s10, v0 +; GCN-NEXT: v_mul_hi_u32 v0, s11, v0 +; GCN-NEXT: s_mul_i32 s13, s11, s1 +; GCN-NEXT: v_readfirstlane_b32 s15, v2 +; GCN-NEXT: s_add_u32 s13, s15, s13 +; GCN-NEXT: v_readfirstlane_b32 s14, v0 +; GCN-NEXT: s_mul_i32 s0, s10, s0 +; GCN-NEXT: s_addc_u32 s14, 0, s14 +; GCN-NEXT: v_readfirstlane_b32 s12, v3 +; GCN-NEXT: s_add_u32 s0, s13, s0 +; GCN-NEXT: s_addc_u32 s0, s14, s12 +; GCN-NEXT: v_readfirstlane_b32 s12, v1 +; GCN-NEXT: s_addc_u32 s12, s12, 0 +; GCN-NEXT: s_mul_i32 s1, s10, s1 +; GCN-NEXT: s_add_u32 s0, s0, s1 +; GCN-NEXT: s_addc_u32 s1, 0, s12 +; GCN-NEXT: s_add_u32 s11, s11, s0 +; GCN-NEXT: s_addc_u32 s1, s10, s1 ; GCN-NEXT: v_mov_b32_e32 v0, s1 ; GCN-NEXT: v_mul_hi_u32 v1, s6, v0 ; GCN-NEXT: v_mov_b32_e32 v2, s11 @@ -118,11 +113,9 @@ define amdgpu_kernel void @s_test_urem_i64(ptr addrspace(1) %out, i64 %x, i64 %y ; GCN-NEXT: s_mul_i32 s4, s8, s4 ; GCN-NEXT: s_sub_u32 s6, s6, s4 ; GCN-NEXT: s_cselect_b64 s[4:5], -1, 0 -; GCN-NEXT: s_or_b32 s11, s4, s5 ; GCN-NEXT: s_subb_u32 s13, s10, s9 ; GCN-NEXT: s_sub_u32 s14, s6, s8 ; GCN-NEXT: s_cselect_b64 s[10:11], -1, 0 -; GCN-NEXT: s_or_b32 s15, s10, s11 ; GCN-NEXT: s_subb_u32 s15, s13, 0 ; GCN-NEXT: s_cmp_ge_u32 s15, s9 ; GCN-NEXT: s_cselect_b32 s16, -1, 0 @@ -131,13 +124,11 @@ define amdgpu_kernel void @s_test_urem_i64(ptr addrspace(1) %out, i64 %x, i64 %y ; GCN-NEXT: s_cmp_eq_u32 s15, s9 ; GCN-NEXT: s_cselect_b32 s16, s17, s16 ; GCN-NEXT: s_or_b32 s10, s10, s11 -; GCN-NEXT: s_subb_u32 s13, s13, s9 -; GCN-NEXT: s_sub_u32 s17, s14, s8 -; GCN-NEXT: s_cselect_b64 s[10:11], -1, 0 -; GCN-NEXT: s_or_b32 s10, s10, s11 -; GCN-NEXT: s_subb_u32 s10, s13, 0 +; GCN-NEXT: s_subb_u32 s10, s13, s9 +; GCN-NEXT: s_sub_u32 s11, s14, s8 +; GCN-NEXT: s_subb_u32 s10, s10, 0 ; GCN-NEXT: s_cmp_lg_u32 s16, 0 -; GCN-NEXT: s_cselect_b32 s11, s17, s14 +; GCN-NEXT: s_cselect_b32 s11, s11, s14 ; GCN-NEXT: s_cselect_b32 s10, s10, s15 ; GCN-NEXT: s_or_b32 s4, s4, s5 ; GCN-NEXT: s_subb_u32 s4, s7, s12 @@ -150,6 +141,7 @@ define amdgpu_kernel void @s_test_urem_i64(ptr addrspace(1) %out, i64 %x, i64 %y ; GCN-NEXT: s_cmp_lg_u32 s5, 0 ; GCN-NEXT: s_cselect_b32 s4, s10, s4 ; GCN-NEXT: s_cselect_b32 s5, s11, s6 +; GCN-NEXT: s_mov_b32 s2, -1 ; GCN-NEXT: v_mov_b32_e32 v0, s5 ; GCN-NEXT: v_mov_b32_e32 v1, s4 ; GCN-NEXT: buffer_store_dwordx2 v[0:1], off, s[0:3], 0 @@ -180,8 +172,6 @@ define amdgpu_kernel void @s_test_urem_i64(ptr addrspace(1) %out, i64 %x, i64 %y ; GCN-IR-NEXT: s_cbranch_vccz .LBB0_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: s_add_u32 s14, s12, 1 -; GCN-IR-NEXT: s_cselect_b64 s[8:9], -1, 0 -; GCN-IR-NEXT: s_or_b32 s8, s8, s9 ; GCN-IR-NEXT: s_addc_u32 s8, s13, 0 ; GCN-IR-NEXT: s_cselect_b64 s[8:9], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s12, 63, s12 @@ -213,8 +203,6 @@ define amdgpu_kernel void @s_test_urem_i64(ptr addrspace(1) %out, i64 %x, i64 %y ; GCN-IR-NEXT: s_sub_u32 s12, s12, s18 ; GCN-IR-NEXT: s_subb_u32 s13, s13, s19 ; GCN-IR-NEXT: s_add_u32 s16, s16, 1 -; GCN-IR-NEXT: s_cselect_b64 s[18:19], -1, 0 -; GCN-IR-NEXT: s_or_b32 s18, s18, s19 ; GCN-IR-NEXT: s_addc_u32 s17, s17, 0 ; GCN-IR-NEXT: s_cselect_b64 s[18:19], -1, 0 ; GCN-IR-NEXT: s_mov_b64 s[10:11], s[4:5] @@ -803,12 +791,11 @@ define amdgpu_kernel void @s_test_urem_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-LABEL: s_test_urem_k_num_i64: ; GCN: ; %bb.0: ; GCN-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x9 -; GCN-NEXT: s_mov_b32 s7, 0xf000 ; GCN-NEXT: s_waitcnt lgkmcnt(0) ; GCN-NEXT: v_cvt_f32_u32_e32 v0, s2 ; GCN-NEXT: v_cvt_f32_u32_e32 v1, s3 -; GCN-NEXT: s_sub_u32 s6, 0, s2 -; GCN-NEXT: s_subb_u32 s8, 0, s3 +; GCN-NEXT: s_sub_u32 s4, 0, s2 +; GCN-NEXT: s_subb_u32 s5, 0, s3 ; GCN-NEXT: v_madmk_f32 v0, v1, 0x4f800000, v0 ; GCN-NEXT: v_rcp_f32_e32 v0, v0 ; GCN-NEXT: v_mul_f32_e32 v0, 0x5f7ffffc, v0 @@ -817,77 +804,73 @@ define amdgpu_kernel void @s_test_urem_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-NEXT: v_madmk_f32 v0, v1, 0xcf800000, v0 ; GCN-NEXT: v_cvt_u32_f32_e32 v0, v0 ; GCN-NEXT: v_cvt_u32_f32_e32 v1, v1 -; GCN-NEXT: v_mul_hi_u32 v2, s6, v0 +; GCN-NEXT: v_mul_hi_u32 v2, s4, v0 +; GCN-NEXT: v_readfirstlane_b32 s6, v1 +; GCN-NEXT: v_readfirstlane_b32 s7, v0 +; GCN-NEXT: s_mul_i32 s8, s4, s6 +; GCN-NEXT: v_readfirstlane_b32 s11, v2 +; GCN-NEXT: s_mul_i32 s9, s5, s7 +; GCN-NEXT: s_mul_i32 s10, s4, s7 +; GCN-NEXT: s_add_i32 s8, s11, s8 +; GCN-NEXT: v_mul_hi_u32 v3, v0, s10 +; GCN-NEXT: s_add_i32 s8, s8, s9 +; GCN-NEXT: v_mul_hi_u32 v0, v0, s8 +; GCN-NEXT: v_mul_hi_u32 v4, v1, s10 +; GCN-NEXT: v_readfirstlane_b32 s9, v3 +; GCN-NEXT: s_mul_i32 s12, s7, s8 +; GCN-NEXT: v_mul_hi_u32 v1, v1, s8 +; GCN-NEXT: s_add_u32 s9, s9, s12 +; GCN-NEXT: v_readfirstlane_b32 s12, v0 +; GCN-NEXT: s_mul_i32 s10, s6, s10 +; GCN-NEXT: s_addc_u32 s12, 0, s12 +; GCN-NEXT: v_readfirstlane_b32 s11, v4 +; GCN-NEXT: s_add_u32 s9, s9, s10 +; GCN-NEXT: v_readfirstlane_b32 s13, v1 +; GCN-NEXT: s_addc_u32 s9, s12, s11 +; GCN-NEXT: s_mul_i32 s8, s6, s8 +; GCN-NEXT: s_addc_u32 s10, s13, 0 +; GCN-NEXT: s_add_u32 s8, s9, s8 +; GCN-NEXT: s_addc_u32 s9, 0, s10 +; GCN-NEXT: s_add_u32 s8, s7, s8 +; GCN-NEXT: v_mov_b32_e32 v0, s8 +; GCN-NEXT: v_mul_hi_u32 v0, s4, v0 +; GCN-NEXT: s_addc_u32 s6, s6, s9 +; GCN-NEXT: s_mul_i32 s9, s4, s6 +; GCN-NEXT: s_mul_i32 s5, s5, s8 +; GCN-NEXT: v_readfirstlane_b32 s10, v0 +; GCN-NEXT: s_add_i32 s9, s10, s9 +; GCN-NEXT: s_mul_i32 s4, s4, s8 +; GCN-NEXT: s_add_i32 s5, s9, s5 +; GCN-NEXT: v_mov_b32_e32 v2, s4 +; GCN-NEXT: v_mov_b32_e32 v0, s5 +; GCN-NEXT: v_mul_hi_u32 v3, s6, v2 +; GCN-NEXT: v_mul_hi_u32 v2, s8, v2 +; GCN-NEXT: v_mul_hi_u32 v1, s6, v0 +; GCN-NEXT: v_mul_hi_u32 v0, s8, v0 +; GCN-NEXT: s_mul_i32 s10, s8, s5 +; GCN-NEXT: v_readfirstlane_b32 s12, v2 +; GCN-NEXT: s_add_u32 s10, s12, s10 +; GCN-NEXT: v_readfirstlane_b32 s11, v0 +; GCN-NEXT: s_mul_i32 s4, s6, s4 +; GCN-NEXT: s_addc_u32 s11, 0, s11 +; GCN-NEXT: v_readfirstlane_b32 s9, v3 +; GCN-NEXT: s_add_u32 s4, s10, s4 +; GCN-NEXT: s_addc_u32 s4, s11, s9 ; GCN-NEXT: v_readfirstlane_b32 s9, v1 +; GCN-NEXT: s_addc_u32 s9, s9, 0 +; GCN-NEXT: s_mul_i32 s5, s6, s5 +; GCN-NEXT: s_add_u32 s4, s4, s5 +; GCN-NEXT: s_addc_u32 s5, 0, s9 +; GCN-NEXT: s_add_u32 s4, s8, s4 +; GCN-NEXT: s_addc_u32 s5, s6, s5 +; GCN-NEXT: v_mul_hi_u32 v1, s4, 24 +; GCN-NEXT: v_mul_hi_u32 v0, s5, 24 +; GCN-NEXT: s_mul_i32 s5, s5, 24 +; GCN-NEXT: s_mov_b32 s7, 0xf000 +; GCN-NEXT: v_readfirstlane_b32 s8, v1 ; GCN-NEXT: v_readfirstlane_b32 s4, v0 -; GCN-NEXT: s_mul_i32 s5, s6, s9 -; GCN-NEXT: v_readfirstlane_b32 s12, v2 -; GCN-NEXT: s_mul_i32 s10, s8, s4 -; GCN-NEXT: s_mul_i32 s11, s6, s4 -; GCN-NEXT: s_add_i32 s5, s12, s5 -; GCN-NEXT: v_mul_hi_u32 v3, v0, s11 -; GCN-NEXT: s_add_i32 s5, s5, s10 -; GCN-NEXT: v_mul_hi_u32 v0, v0, s5 -; GCN-NEXT: v_mul_hi_u32 v4, v1, s11 -; GCN-NEXT: v_readfirstlane_b32 s10, v3 -; GCN-NEXT: v_mul_hi_u32 v1, v1, s5 -; GCN-NEXT: s_mul_i32 s13, s4, s5 -; GCN-NEXT: s_add_u32 s10, s10, s13 -; GCN-NEXT: v_readfirstlane_b32 s13, v0 -; GCN-NEXT: s_mul_i32 s11, s9, s11 -; GCN-NEXT: s_addc_u32 s13, 0, s13 -; GCN-NEXT: v_readfirstlane_b32 s12, v4 -; GCN-NEXT: s_add_u32 s10, s10, s11 -; GCN-NEXT: v_readfirstlane_b32 s14, v1 -; GCN-NEXT: s_addc_u32 s10, s13, s12 -; GCN-NEXT: s_addc_u32 s11, s14, 0 -; GCN-NEXT: s_mul_i32 s5, s9, s5 -; GCN-NEXT: s_add_u32 s5, s10, s5 -; GCN-NEXT: s_addc_u32 s10, 0, s11 -; GCN-NEXT: s_add_u32 s11, s4, s5 -; GCN-NEXT: v_mov_b32_e32 v0, s11 -; GCN-NEXT: v_mul_hi_u32 v0, s6, v0 -; GCN-NEXT: s_cselect_b64 s[4:5], -1, 0 -; GCN-NEXT: s_or_b32 s4, s4, s5 -; GCN-NEXT: s_addc_u32 s9, s9, s10 -; GCN-NEXT: s_mul_i32 s4, s6, s9 -; GCN-NEXT: v_readfirstlane_b32 s5, v0 -; GCN-NEXT: s_add_i32 s4, s5, s4 -; GCN-NEXT: s_mul_i32 s8, s8, s11 -; GCN-NEXT: s_mul_i32 s5, s6, s11 -; GCN-NEXT: s_add_i32 s4, s4, s8 -; GCN-NEXT: v_mov_b32_e32 v2, s5 -; GCN-NEXT: v_mov_b32_e32 v0, s4 -; GCN-NEXT: v_mul_hi_u32 v3, s9, v2 -; GCN-NEXT: v_mul_hi_u32 v2, s11, v2 -; GCN-NEXT: v_mul_hi_u32 v1, s9, v0 -; GCN-NEXT: v_mul_hi_u32 v0, s11, v0 -; GCN-NEXT: s_mul_i32 s8, s11, s4 -; GCN-NEXT: v_readfirstlane_b32 s12, v2 -; GCN-NEXT: s_add_u32 s8, s12, s8 -; GCN-NEXT: v_readfirstlane_b32 s10, v0 -; GCN-NEXT: s_mul_i32 s5, s9, s5 -; GCN-NEXT: s_addc_u32 s10, 0, s10 -; GCN-NEXT: v_readfirstlane_b32 s6, v3 ; GCN-NEXT: s_add_u32 s5, s8, s5 -; GCN-NEXT: s_addc_u32 s5, s10, s6 -; GCN-NEXT: v_readfirstlane_b32 s6, v1 -; GCN-NEXT: s_addc_u32 s6, s6, 0 -; GCN-NEXT: s_mul_i32 s4, s9, s4 -; GCN-NEXT: s_add_u32 s4, s5, s4 -; GCN-NEXT: s_addc_u32 s6, 0, s6 -; GCN-NEXT: s_add_u32 s8, s11, s4 -; GCN-NEXT: s_cselect_b64 s[4:5], -1, 0 -; GCN-NEXT: s_or_b32 s4, s4, s5 -; GCN-NEXT: s_addc_u32 s4, s9, s6 -; GCN-NEXT: v_mul_hi_u32 v1, s8, 24 -; GCN-NEXT: v_mul_hi_u32 v0, s4, 24 -; GCN-NEXT: s_mul_i32 s4, s4, 24 -; GCN-NEXT: s_mov_b32 s6, -1 -; GCN-NEXT: v_readfirstlane_b32 s8, v1 -; GCN-NEXT: v_readfirstlane_b32 s5, v0 -; GCN-NEXT: s_add_u32 s4, s8, s4 -; GCN-NEXT: s_addc_u32 s8, 0, s5 +; GCN-NEXT: s_addc_u32 s8, 0, s4 ; GCN-NEXT: v_mov_b32_e32 v0, s8 ; GCN-NEXT: v_mul_hi_u32 v0, s2, v0 ; GCN-NEXT: s_mov_b32 s4, s0 @@ -899,11 +882,9 @@ define amdgpu_kernel void @s_test_urem_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-NEXT: s_mul_i32 s0, s2, s8 ; GCN-NEXT: s_sub_u32 s11, 24, s0 ; GCN-NEXT: s_cselect_b64 s[0:1], -1, 0 -; GCN-NEXT: s_or_b32 s8, s0, s1 ; GCN-NEXT: s_subb_u32 s12, s9, s3 ; GCN-NEXT: s_sub_u32 s13, s11, s2 ; GCN-NEXT: s_cselect_b64 s[8:9], -1, 0 -; GCN-NEXT: s_or_b32 s14, s8, s9 ; GCN-NEXT: s_subb_u32 s14, s12, 0 ; GCN-NEXT: s_cmp_ge_u32 s14, s3 ; GCN-NEXT: s_cselect_b32 s15, -1, 0 @@ -912,13 +893,11 @@ define amdgpu_kernel void @s_test_urem_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-NEXT: s_cmp_eq_u32 s14, s3 ; GCN-NEXT: s_cselect_b32 s15, s16, s15 ; GCN-NEXT: s_or_b32 s8, s8, s9 -; GCN-NEXT: s_subb_u32 s12, s12, s3 -; GCN-NEXT: s_sub_u32 s16, s13, s2 -; GCN-NEXT: s_cselect_b64 s[8:9], -1, 0 -; GCN-NEXT: s_or_b32 s8, s8, s9 -; GCN-NEXT: s_subb_u32 s8, s12, 0 +; GCN-NEXT: s_subb_u32 s8, s12, s3 +; GCN-NEXT: s_sub_u32 s9, s13, s2 +; GCN-NEXT: s_subb_u32 s8, s8, 0 ; GCN-NEXT: s_cmp_lg_u32 s15, 0 -; GCN-NEXT: s_cselect_b32 s9, s16, s13 +; GCN-NEXT: s_cselect_b32 s9, s9, s13 ; GCN-NEXT: s_cselect_b32 s8, s8, s14 ; GCN-NEXT: s_or_b32 s0, s0, s1 ; GCN-NEXT: s_subb_u32 s0, 0, s10 @@ -931,6 +910,7 @@ define amdgpu_kernel void @s_test_urem_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-NEXT: s_cmp_lg_u32 s1, 0 ; GCN-NEXT: s_cselect_b32 s0, s8, s0 ; GCN-NEXT: s_cselect_b32 s1, s9, s11 +; GCN-NEXT: s_mov_b32 s6, -1 ; GCN-NEXT: v_mov_b32_e32 v0, s1 ; GCN-NEXT: v_mov_b32_e32 v1, s0 ; GCN-NEXT: buffer_store_dwordx2 v[0:1], off, s[4:7], 0 @@ -956,8 +936,6 @@ define amdgpu_kernel void @s_test_urem_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR-NEXT: s_cbranch_vccz .LBB6_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: s_add_u32 s10, s8, 1 -; GCN-IR-NEXT: s_cselect_b64 s[6:7], -1, 0 -; GCN-IR-NEXT: s_or_b32 s6, s6, s7 ; GCN-IR-NEXT: s_addc_u32 s6, s9, 0 ; GCN-IR-NEXT: s_cselect_b64 s[6:7], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s8, 63, s8 @@ -988,8 +966,6 @@ define amdgpu_kernel void @s_test_urem_k_num_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR-NEXT: s_sub_u32 s10, s10, s16 ; GCN-IR-NEXT: s_subb_u32 s11, s11, s17 ; GCN-IR-NEXT: s_add_u32 s14, s14, 1 -; GCN-IR-NEXT: s_cselect_b64 s[16:17], -1, 0 -; GCN-IR-NEXT: s_or_b32 s16, s16, s17 ; GCN-IR-NEXT: s_addc_u32 s15, s15, 0 ; GCN-IR-NEXT: s_cselect_b64 s[16:17], -1, 0 ; GCN-IR-NEXT: s_mov_b64 s[8:9], s[4:5] @@ -1077,8 +1053,6 @@ define amdgpu_kernel void @s_test_urem_k_den_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR-NEXT: s_cbranch_vccz .LBB7_5 ; GCN-IR-NEXT: ; %bb.1: ; %udiv-bb1 ; GCN-IR-NEXT: s_add_u32 s11, s8, 1 -; GCN-IR-NEXT: s_cselect_b64 s[6:7], -1, 0 -; GCN-IR-NEXT: s_or_b32 s6, s6, s7 ; GCN-IR-NEXT: s_addc_u32 s6, s9, 0 ; GCN-IR-NEXT: s_cselect_b64 s[6:7], -1, 0 ; GCN-IR-NEXT: s_sub_i32 s8, 63, s8 @@ -1106,8 +1080,6 @@ define amdgpu_kernel void @s_test_urem_k_den_i64(ptr addrspace(1) %out, i64 %x) ; GCN-IR-NEXT: s_sub_u32 s8, s8, s10 ; GCN-IR-NEXT: s_subb_u32 s9, s9, 0 ; GCN-IR-NEXT: s_add_u32 s12, s12, 1 -; GCN-IR-NEXT: s_cselect_b64 s[14:15], -1, 0 -; GCN-IR-NEXT: s_or_b32 s14, s14, s15 ; GCN-IR-NEXT: s_addc_u32 s13, s13, 0 ; GCN-IR-NEXT: s_cselect_b64 s[14:15], -1, 0 ; GCN-IR-NEXT: s_mov_b64 s[10:11], s[4:5] diff --git a/llvm/test/CodeGen/AMDGPU/usubo.ll b/llvm/test/CodeGen/AMDGPU/usubo.ll index e8db647..8a54ad3 100644 --- a/llvm/test/CodeGen/AMDGPU/usubo.ll +++ b/llvm/test/CodeGen/AMDGPU/usubo.ll @@ -15,10 +15,8 @@ define amdgpu_kernel void @s_usubo_i64_zext(ptr addrspace(1) %out, i64 %a, i64 % ; SI-NEXT: s_waitcnt lgkmcnt(0) ; SI-NEXT: s_mov_b32 s4, s0 ; SI-NEXT: s_sub_u32 s2, s2, s8 -; SI-NEXT: s_mov_b32 s5, s1 -; SI-NEXT: s_cselect_b64 s[0:1], -1, 0 -; SI-NEXT: s_or_b32 s0, s0, s1 ; SI-NEXT: s_subb_u32 s3, s3, s9 +; SI-NEXT: s_mov_b32 s5, s1 ; SI-NEXT: s_cselect_b64 s[0:1], -1, 0 ; SI-NEXT: v_cndmask_b32_e64 v0, 0, 1, s[0:1] ; SI-NEXT: v_mov_b32_e32 v1, s3 @@ -432,8 +430,6 @@ 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 s4, s4, s6 -; SI-NEXT: s_cselect_b64 s[12:13], -1, 0 -; SI-NEXT: s_or_b32 s6, s12, s13 ; SI-NEXT: s_subb_u32 s5, s5, s7 ; SI-NEXT: s_mov_b32 s8, s0 ; SI-NEXT: s_mov_b32 s9, s1 diff --git a/llvm/test/CodeGen/PowerPC/vp-ld-st.ll b/llvm/test/CodeGen/PowerPC/vp-ld-st.ll new file mode 100644 index 0000000..f0f9943 --- /dev/null +++ b/llvm/test/CodeGen/PowerPC/vp-ld-st.ll @@ -0,0 +1,160 @@ +; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6 +; RUN: llc -verify-machineinstrs -mcpu=pwr10 \ +; RUN: -mtriple=powerpc64le-unknown-unknown < %s | FileCheck %s +; RUN: llc -verify-machineinstrs -mcpu=future \ +; RUN: -mtriple=powerpc64le-unknown-unknown < %s | FileCheck -check-prefix=FUTURE %s + +; RUN: llc -verify-machineinstrs -mcpu=pwr10 \ +; RUN: -mtriple=powerpc64-unknown-unknown < %s | FileCheck %s +; RUN: llc -verify-machineinstrs -mcpu=future \ +; RUN: -mtriple=powerpc64-unknown-unknown < %s | FileCheck --check-prefix=FUTURE %s + +; Function Attrs: nounwind readnone +define void @stxvl1(<16 x i8> %a, ptr %b, i64 %c) { +; CHECK-LABEL: stxvl1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sldi 3, 6, 56 +; CHECK-NEXT: stxvl 34, 5, 3 +; CHECK-NEXT: blr +; +; FUTURE-LABEL: stxvl1: +; FUTURE: # %bb.0: # %entry +; FUTURE-NEXT: stxvrl 34, 5, 6 +; FUTURE-NEXT: blr +entry: + %cconv = trunc i64 %c to i32 + tail call void @llvm.vp.store.v16i8.p0(<16 x i8> %a, ptr %b, <16 x i1> splat (i1 true), i32 %cconv) + ret void +} + +; Function Attrs: nounwind readnone +define void @stxvl2(<8 x i16> %a, ptr %b, i64 %c) { +; CHECK-LABEL: stxvl2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sldi 3, 6, 57 +; CHECK-NEXT: stxvl 34, 5, 3 +; CHECK-NEXT: blr +; +; FUTURE-LABEL: stxvl2: +; FUTURE: # %bb.0: # %entry +; FUTURE-NEXT: sldi 3, 6, 1 +; FUTURE-NEXT: stxvrl 34, 5, 3 +; FUTURE-NEXT: blr +entry: + %cconv = trunc i64 %c to i32 + tail call void @llvm.vp.store.v8i16.p0(<8 x i16> %a, ptr %b, <8 x i1> splat (i1 true), i32 %cconv) + ret void +} + +; Function Attrs: nounwind readnone +define void @stxvl4(<4 x i32> %a, ptr %b, i64 %c) { +; CHECK-LABEL: stxvl4: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sldi 3, 6, 58 +; CHECK-NEXT: stxvl 34, 5, 3 +; CHECK-NEXT: blr +; +; FUTURE-LABEL: stxvl4: +; FUTURE: # %bb.0: # %entry +; FUTURE-NEXT: sldi 3, 6, 2 +; FUTURE-NEXT: stxvrl 34, 5, 3 +; FUTURE-NEXT: blr +entry: + %cconv = trunc i64 %c to i32 + tail call void @llvm.vp.store.v4i32.p0(<4 x i32> %a, ptr %b, <4 x i1> splat (i1 true), i32 %cconv) + ret void +} + +; Function Attrs: nounwind readnone +define void @stxvl8(<2 x i64> %a, ptr %b, i64 %c) { +; CHECK-LABEL: stxvl8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sldi 3, 6, 59 +; CHECK-NEXT: stxvl 34, 5, 3 +; CHECK-NEXT: blr +; +; FUTURE-LABEL: stxvl8: +; FUTURE: # %bb.0: # %entry +; FUTURE-NEXT: sldi 3, 6, 3 +; FUTURE-NEXT: stxvrl 34, 5, 3 +; FUTURE-NEXT: blr +entry: + %cconv = trunc i64 %c to i32 + tail call void @llvm.vp.store.v2i64.p0(<2 x i64> %a, ptr %b, <2 x i1> splat (i1 true), i32 %cconv) + ret void +} + +; Function Attrs: nounwind readnone +define <16 x i8> @lxvl1(ptr %a, i64 %b) { +; CHECK-LABEL: lxvl1: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sldi 4, 4, 56 +; CHECK-NEXT: lxvl 34, 3, 4 +; CHECK-NEXT: blr +; +; FUTURE-LABEL: lxvl1: +; FUTURE: # %bb.0: # %entry +; FUTURE-NEXT: lxvrl 34, 3, 4 +; FUTURE-NEXT: blr +entry: + %bconv = trunc i64 %b to i32 + %0 = tail call <16 x i8> @llvm.vp.load.v16i8.p0(ptr %a, <16 x i1> splat (i1 true), i32 %bconv) + ret <16 x i8> %0 +} + +; Function Attrs: nounwind readnone +define <8 x i16> @lxvl2(ptr %a, i64 %b) { +; CHECK-LABEL: lxvl2: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sldi 4, 4, 57 +; CHECK-NEXT: lxvl 34, 3, 4 +; CHECK-NEXT: blr +; +; FUTURE-LABEL: lxvl2: +; FUTURE: # %bb.0: # %entry +; FUTURE-NEXT: sldi 4, 4, 1 +; FUTURE-NEXT: lxvrl 34, 3, 4 +; FUTURE-NEXT: blr +entry: + %bconv = trunc i64 %b to i32 + %0 = tail call <8 x i16> @llvm.vp.load.v8i16.p0(ptr %a, <8 x i1> splat (i1 true), i32 %bconv) + ret <8 x i16> %0 +} + +; Function Attrs: nounwind readnone +define <4 x i32> @lxvl4(ptr %a, i64 %b) { +; CHECK-LABEL: lxvl4: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sldi 4, 4, 58 +; CHECK-NEXT: lxvl 34, 3, 4 +; CHECK-NEXT: blr +; +; FUTURE-LABEL: lxvl4: +; FUTURE: # %bb.0: # %entry +; FUTURE-NEXT: sldi 4, 4, 2 +; FUTURE-NEXT: lxvrl 34, 3, 4 +; FUTURE-NEXT: blr +entry: + %bconv = trunc i64 %b to i32 + %0 = tail call <4 x i32> @llvm.vp.load.v4i32.p0(ptr %a, <4 x i1> splat (i1 true), i32 %bconv) + ret <4 x i32> %0 +} + +; Function Attrs: nounwind readnone +define <2 x i64> @lxvl8(ptr %a, i64 %b) { +; CHECK-LABEL: lxvl8: +; CHECK: # %bb.0: # %entry +; CHECK-NEXT: sldi 4, 4, 59 +; CHECK-NEXT: lxvl 34, 3, 4 +; CHECK-NEXT: blr +; +; FUTURE-LABEL: lxvl8: +; FUTURE: # %bb.0: # %entry +; FUTURE-NEXT: sldi 4, 4, 3 +; FUTURE-NEXT: lxvrl 34, 3, 4 +; FUTURE-NEXT: blr +entry: + %bconv = trunc i64 %b to i32 + %0 = tail call <2 x i64> @llvm.vp.load.v2i64.p0(ptr %a, <2 x i1> splat (i1 true), i32 %bconv) + ret <2 x i64> %0 +} |