From fdea5e02ce598be1cdda0259381039e8b5d6cba5 Mon Sep 17 00:00:00 2001 From: Matt Arsenault Date: Tue, 1 Oct 2019 02:23:20 +0000 Subject: [PATCH] AMDGPU/GlobalISel: Select s1 src G_SITOFP/G_UITOFP llvm-svn: 373298 --- .../AMDGPU/AMDGPUInstructionSelector.cpp | 46 +++ .../Target/AMDGPU/AMDGPUInstructionSelector.h | 1 + .../lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp | 4 +- llvm/lib/Target/AMDGPU/SIInstructions.td | 12 +- .../AMDGPU/GlobalISel/inst-select-sitofp.mir | 276 +++++++++++++++-- .../AMDGPU/GlobalISel/inst-select-uitofp.mir | 283 ++++++++++++++++-- 6 files changed, 570 insertions(+), 52 deletions(-) diff --git a/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp b/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp index 4b186c2c8ab0..a2bb2b092e8f 100644 --- a/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp +++ b/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.cpp @@ -1328,6 +1328,49 @@ bool AMDGPUInstructionSelector::selectG_SZA_EXT(MachineInstr &I) const { return false; } +static int64_t getFPTrueImmVal(unsigned Size, bool Signed) { + switch (Size) { + case 16: + return Signed ? 0xBC00 : 0x3C00; + case 32: + return Signed ? 0xbf800000 : 0x3f800000; + case 64: + return Signed ? 0xbff0000000000000 : 0x3ff0000000000000; + default: + llvm_unreachable("Invalid FP type size"); + } +} + +bool AMDGPUInstructionSelector::selectG_SITOFP_UITOFP(MachineInstr &I) const { + MachineBasicBlock *MBB = I.getParent(); + MachineFunction *MF = MBB->getParent(); + MachineRegisterInfo &MRI = MF->getRegInfo(); + Register Src = I.getOperand(1).getReg(); + if (!isSCC(Src, MRI)) + return selectImpl(I, *CoverageInfo); + + bool Signed = I.getOpcode() == AMDGPU::G_SITOFP; + Register DstReg = I.getOperand(0).getReg(); + const LLT DstTy = MRI.getType(DstReg); + const unsigned DstSize = DstTy.getSizeInBits(); + const DebugLoc &DL = I.getDebugLoc(); + + BuildMI(*MBB, I, DL, TII.get(AMDGPU::COPY), AMDGPU::SCC) + .addReg(Src); + + unsigned NewOpc = + DstSize > 32 ? AMDGPU::S_CSELECT_B64 : AMDGPU::S_CSELECT_B32; + auto MIB = BuildMI(*MBB, I, DL, TII.get(NewOpc), DstReg) + .addImm(0) + .addImm(getFPTrueImmVal(DstSize, Signed)); + + if (!constrainSelectedInstRegOperands(*MIB, TII, TRI, RBI)) + return false; + + I.eraseFromParent(); + return true; +} + bool AMDGPUInstructionSelector::selectG_CONSTANT(MachineInstr &I) const { MachineBasicBlock *BB = I.getParent(); MachineOperand &ImmOp = I.getOperand(1); @@ -1672,6 +1715,9 @@ bool AMDGPUInstructionSelector::select(MachineInstr &I) { case TargetOpcode::G_ZEXT: case TargetOpcode::G_ANYEXT: return selectG_SZA_EXT(I); + case TargetOpcode::G_SITOFP: + case TargetOpcode::G_UITOFP: + return selectG_SITOFP_UITOFP(I); case TargetOpcode::G_BRCOND: return selectG_BRCOND(I); case TargetOpcode::G_FRAME_INDEX: diff --git a/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.h b/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.h index 4aadcd9dc204..d3c83a6a872a 100644 --- a/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.h +++ b/llvm/lib/Target/AMDGPU/AMDGPUInstructionSelector.h @@ -79,6 +79,7 @@ private: bool selectPHI(MachineInstr &I) const; bool selectG_TRUNC(MachineInstr &I) const; bool selectG_SZA_EXT(MachineInstr &I) const; + bool selectG_SITOFP_UITOFP(MachineInstr &I) const; bool selectG_CONSTANT(MachineInstr &I) const; bool selectG_AND_OR_XOR(MachineInstr &I) const; bool selectG_ADD_SUB(MachineInstr &I) const; diff --git a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp index e289e8e689a5..10420d6379d2 100644 --- a/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp +++ b/llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp @@ -417,9 +417,9 @@ AMDGPULegalizerInfo::AMDGPULegalizerInfo(const GCNSubtarget &ST_, {S32, S8}, {S128, S32}, {S128, S64}, {S32, LLT::scalar(24)}}) .scalarize(0); - // TODO: Legal for s1->s64, requires split for VALU. + // TODO: Split s1->s64 during regbankselect for VALU. getActionDefinitionsBuilder({G_SITOFP, G_UITOFP}) - .legalFor({{S32, S32}, {S64, S32}, {S16, S32}, {S32, S1}, {S16, S1}}) + .legalFor({{S32, S32}, {S64, S32}, {S16, S32}, {S32, S1}, {S16, S1}, {S64, S1}}) .lowerFor({{S32, S64}}) .customFor({{S64, S64}}) .scalarize(0); diff --git a/llvm/lib/Target/AMDGPU/SIInstructions.td b/llvm/lib/Target/AMDGPU/SIInstructions.td index d56af2977521..15b9fce5341b 100644 --- a/llvm/lib/Target/AMDGPU/SIInstructions.td +++ b/llvm/lib/Target/AMDGPU/SIInstructions.td @@ -1611,7 +1611,7 @@ def : GCNPat < (V_CVT_F16_F32_e32 ( V_CNDMASK_B32_e64 /*src0mod*/(i32 0), /*src0*/(i32 0), /*src1mod*/(i32 0), /*src1*/(i32 CONST.FP32_NEG_ONE), - $src)) + SSrc_i1:$src)) >; def : GCNPat < @@ -1619,35 +1619,35 @@ def : GCNPat < (V_CVT_F16_F32_e32 ( V_CNDMASK_B32_e64 /*src0mod*/(i32 0), /*src0*/(i32 0), /*src1mod*/(i32 0), /*src1*/(i32 CONST.FP32_ONE), - $src)) + SSrc_i1:$src)) >; def : GCNPat < (f32 (sint_to_fp i1:$src)), (V_CNDMASK_B32_e64 /*src0mod*/(i32 0), /*src0*/(i32 0), /*src1mod*/(i32 0), /*src1*/(i32 CONST.FP32_NEG_ONE), - $src) + SSrc_i1:$src) >; def : GCNPat < (f32 (uint_to_fp i1:$src)), (V_CNDMASK_B32_e64 /*src0mod*/(i32 0), /*src0*/(i32 0), /*src1mod*/(i32 0), /*src1*/(i32 CONST.FP32_ONE), - $src) + SSrc_i1:$src) >; def : GCNPat < (f64 (sint_to_fp i1:$src)), (V_CVT_F64_I32_e32 (V_CNDMASK_B32_e64 /*src0mod*/(i32 0), /*src0*/(i32 0), /*src1mod*/(i32 0), /*src1*/(i32 -1), - $src)) + SSrc_i1:$src)) >; def : GCNPat < (f64 (uint_to_fp i1:$src)), (V_CVT_F64_U32_e32 (V_CNDMASK_B32_e64 /*src0mod*/(i32 0), /*src0*/(i32 0), /*src1mod*/(i32 0), /*src1*/(i32 1), - $src)) + SSrc_i1:$src)) >; //===----------------------------------------------------------------------===// diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-sitofp.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-sitofp.mir index 61608cfe86fc..63d6427052c9 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-sitofp.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-sitofp.mir @@ -1,5 +1,6 @@ # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py -# RUN: llc -march=amdgcn -mcpu=hawaii -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s -check-prefixes=GCN +# RUN: llc -march=amdgcn -mcpu=hawaii -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck -check-prefix=WAVE64 %s +# RUN: llc -march=amdgcn -mcpu=gfx1010 -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck -check-prefix=WAVE32 %s --- @@ -11,14 +12,23 @@ body: | bb.0: liveins: $sgpr0, $vgpr0, $vgpr3_vgpr4 - ; GCN-LABEL: name: sitofp - ; GCN: [[COPY:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0 - ; GCN: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0 - ; GCN: [[COPY2:%[0-9]+]]:vreg_64 = COPY $vgpr3_vgpr4 - ; GCN: [[V_CVT_F32_I32_e64_:%[0-9]+]]:vgpr_32 = V_CVT_F32_I32_e64 [[COPY]], 0, 0, implicit $exec - ; GCN: [[V_CVT_F32_I32_e64_1:%[0-9]+]]:vgpr_32 = V_CVT_F32_I32_e64 [[COPY1]], 0, 0, implicit $exec - ; GCN: FLAT_STORE_DWORD [[COPY2]], [[V_CVT_F32_I32_e64_]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4, addrspace 1) - ; GCN: FLAT_STORE_DWORD [[COPY2]], [[V_CVT_F32_I32_e64_1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4, addrspace 1) + ; WAVE64-LABEL: name: sitofp + ; WAVE64: [[COPY:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0 + ; WAVE64: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; WAVE64: [[COPY2:%[0-9]+]]:vreg_64 = COPY $vgpr3_vgpr4 + ; WAVE64: [[V_CVT_F32_I32_e64_:%[0-9]+]]:vgpr_32 = V_CVT_F32_I32_e64 [[COPY]], 0, 0, implicit $exec + ; WAVE64: [[V_CVT_F32_I32_e64_1:%[0-9]+]]:vgpr_32 = V_CVT_F32_I32_e64 [[COPY1]], 0, 0, implicit $exec + ; WAVE64: FLAT_STORE_DWORD [[COPY2]], [[V_CVT_F32_I32_e64_]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4, addrspace 1) + ; WAVE64: FLAT_STORE_DWORD [[COPY2]], [[V_CVT_F32_I32_e64_1]], 0, 0, 0, 0, implicit $exec, implicit $flat_scr :: (store 4, addrspace 1) + ; WAVE32-LABEL: name: sitofp + ; WAVE32: $vcc_hi = IMPLICIT_DEF + ; WAVE32: [[COPY:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0 + ; WAVE32: [[COPY1:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; WAVE32: [[COPY2:%[0-9]+]]:vreg_64 = COPY $vgpr3_vgpr4 + ; WAVE32: [[V_CVT_F32_I32_e64_:%[0-9]+]]:vgpr_32 = V_CVT_F32_I32_e64 [[COPY]], 0, 0, implicit $exec + ; WAVE32: [[V_CVT_F32_I32_e64_1:%[0-9]+]]:vgpr_32 = V_CVT_F32_I32_e64 [[COPY1]], 0, 0, implicit $exec + ; WAVE32: GLOBAL_STORE_DWORD [[COPY2]], [[V_CVT_F32_I32_e64_]], 0, 0, 0, 0, implicit $exec :: (store 4, addrspace 1) + ; WAVE32: GLOBAL_STORE_DWORD [[COPY2]], [[V_CVT_F32_I32_e64_1]], 0, 0, 0, 0, implicit $exec :: (store 4, addrspace 1) %0:sgpr(s32) = COPY $sgpr0 %1:vgpr(s32) = COPY $vgpr0 @@ -45,12 +55,19 @@ body: | bb.0: liveins: $vgpr0 - ; GCN-LABEL: name: sitofp_s32_to_s16_vv - ; GCN: liveins: $vgpr0 - ; GCN: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 - ; GCN: [[V_CVT_F32_I32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F32_I32_e32 [[COPY]], implicit $exec - ; GCN: [[V_CVT_F16_F32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F16_F32_e32 [[V_CVT_F32_I32_e32_]], implicit $exec - ; GCN: $vgpr0 = COPY [[V_CVT_F16_F32_e32_]] + ; WAVE64-LABEL: name: sitofp_s32_to_s16_vv + ; WAVE64: liveins: $vgpr0 + ; WAVE64: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; WAVE64: [[V_CVT_F32_I32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F32_I32_e32 [[COPY]], implicit $exec + ; WAVE64: [[V_CVT_F16_F32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F16_F32_e32 [[V_CVT_F32_I32_e32_]], implicit $exec + ; WAVE64: $vgpr0 = COPY [[V_CVT_F16_F32_e32_]] + ; WAVE32-LABEL: name: sitofp_s32_to_s16_vv + ; WAVE32: liveins: $vgpr0 + ; WAVE32: $vcc_hi = IMPLICIT_DEF + ; WAVE32: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; WAVE32: [[V_CVT_F32_I32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F32_I32_e32 [[COPY]], implicit $exec + ; WAVE32: [[V_CVT_F16_F32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F16_F32_e32 [[V_CVT_F32_I32_e32_]], implicit $exec + ; WAVE32: $vgpr0 = COPY [[V_CVT_F16_F32_e32_]] %0:vgpr(s32) = COPY $vgpr0 %1:vgpr(s16) = G_SITOFP %0 %2:vgpr(s32) = G_ANYEXT %1 @@ -67,14 +84,231 @@ body: | bb.0: liveins: $sgpr0 - ; GCN-LABEL: name: sitofp_s32_to_s16_vs - ; GCN: liveins: $sgpr0 - ; GCN: [[COPY:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0 - ; GCN: [[V_CVT_F32_I32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F32_I32_e32 [[COPY]], implicit $exec - ; GCN: [[V_CVT_F16_F32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F16_F32_e32 [[V_CVT_F32_I32_e32_]], implicit $exec - ; GCN: $vgpr0 = COPY [[V_CVT_F16_F32_e32_]] + ; WAVE64-LABEL: name: sitofp_s32_to_s16_vs + ; WAVE64: liveins: $sgpr0 + ; WAVE64: [[COPY:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0 + ; WAVE64: [[V_CVT_F32_I32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F32_I32_e32 [[COPY]], implicit $exec + ; WAVE64: [[V_CVT_F16_F32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F16_F32_e32 [[V_CVT_F32_I32_e32_]], implicit $exec + ; WAVE64: $vgpr0 = COPY [[V_CVT_F16_F32_e32_]] + ; WAVE32-LABEL: name: sitofp_s32_to_s16_vs + ; WAVE32: liveins: $sgpr0 + ; WAVE32: $vcc_hi = IMPLICIT_DEF + ; WAVE32: [[COPY:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0 + ; WAVE32: [[V_CVT_F32_I32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F32_I32_e32 [[COPY]], implicit $exec + ; WAVE32: [[V_CVT_F16_F32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F16_F32_e32 [[V_CVT_F32_I32_e32_]], implicit $exec + ; WAVE32: $vgpr0 = COPY [[V_CVT_F16_F32_e32_]] %0:sgpr(s32) = COPY $sgpr0 %1:vgpr(s16) = G_SITOFP %0 %2:vgpr(s32) = G_ANYEXT %1 $vgpr0 = COPY %2 ... + +--- +name: sitofp_s1_to_s32_s_scc +legalized: true +regBankSelected: true +tracksRegLiveness: true + +body: | + bb.0: + liveins: $sgpr0 + + ; WAVE64-LABEL: name: sitofp_s1_to_s32_s_scc + ; WAVE64: liveins: $sgpr0 + ; WAVE64: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0 + ; WAVE64: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0 + ; WAVE64: S_CMP_EQ_U32 [[COPY]], [[S_MOV_B32_]], implicit-def $scc + ; WAVE64: [[COPY1:%[0-9]+]]:sreg_32 = COPY $scc + ; WAVE64: $scc = COPY [[COPY1]] + ; WAVE64: [[S_CSELECT_B32_:%[0-9]+]]:sreg_32_xm0 = S_CSELECT_B32 0, 3212836864, implicit $scc + ; WAVE64: $sgpr0 = COPY [[S_CSELECT_B32_]] + ; WAVE32-LABEL: name: sitofp_s1_to_s32_s_scc + ; WAVE32: liveins: $sgpr0 + ; WAVE32: $vcc_hi = IMPLICIT_DEF + ; WAVE32: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0 + ; WAVE32: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0 + ; WAVE32: S_CMP_EQ_U32 [[COPY]], [[S_MOV_B32_]], implicit-def $scc + ; WAVE32: [[COPY1:%[0-9]+]]:sreg_32 = COPY $scc + ; WAVE32: $scc = COPY [[COPY1]] + ; WAVE32: [[S_CSELECT_B32_:%[0-9]+]]:sreg_32_xm0 = S_CSELECT_B32 0, 3212836864, implicit $scc + ; WAVE32: $sgpr0 = COPY [[S_CSELECT_B32_]] + %0:sgpr(s32) = COPY $sgpr0 + %1:sgpr(s32) = G_CONSTANT i32 0 + %2:scc(s1) = G_ICMP intpred(eq), %0, %1 + %3:sgpr(s32) = G_SITOFP %2 + $sgpr0 = COPY %3 +... + +--- +name: sitofp_s1_to_s16_to_s_scc +legalized: true +regBankSelected: true +tracksRegLiveness: true + +body: | + bb.0: + liveins: $sgpr0 + + ; WAVE64-LABEL: name: sitofp_s1_to_s16_to_s_scc + ; WAVE64: liveins: $sgpr0 + ; WAVE64: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0 + ; WAVE64: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0 + ; WAVE64: S_CMP_EQ_U32 [[COPY]], [[S_MOV_B32_]], implicit-def $scc + ; WAVE64: [[COPY1:%[0-9]+]]:sreg_32 = COPY $scc + ; WAVE64: $scc = COPY [[COPY1]] + ; WAVE64: [[S_CSELECT_B32_:%[0-9]+]]:sreg_32_xm0 = S_CSELECT_B32 0, 48128, implicit $scc + ; WAVE64: $sgpr0 = COPY [[S_CSELECT_B32_]] + ; WAVE32-LABEL: name: sitofp_s1_to_s16_to_s_scc + ; WAVE32: liveins: $sgpr0 + ; WAVE32: $vcc_hi = IMPLICIT_DEF + ; WAVE32: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0 + ; WAVE32: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0 + ; WAVE32: S_CMP_EQ_U32 [[COPY]], [[S_MOV_B32_]], implicit-def $scc + ; WAVE32: [[COPY1:%[0-9]+]]:sreg_32 = COPY $scc + ; WAVE32: $scc = COPY [[COPY1]] + ; WAVE32: [[S_CSELECT_B32_:%[0-9]+]]:sreg_32_xm0 = S_CSELECT_B32 0, 48128, implicit $scc + ; WAVE32: $sgpr0 = COPY [[S_CSELECT_B32_]] + %0:sgpr(s32) = COPY $sgpr0 + %1:sgpr(s32) = G_CONSTANT i32 0 + %2:scc(s1) = G_ICMP intpred(eq), %0, %1 + %3:sgpr(s16) = G_SITOFP %2 + %4:sgpr(s32) = G_ANYEXT %3 + $sgpr0 = COPY %4 +... + +--- +name: sitofp_s1_to_s64_s_scc +legalized: true +regBankSelected: true +tracksRegLiveness: true + +body: | + bb.0: + liveins: $sgpr0 + + ; WAVE64-LABEL: name: sitofp_s1_to_s64_s_scc + ; WAVE64: liveins: $sgpr0 + ; WAVE64: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0 + ; WAVE64: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0 + ; WAVE64: S_CMP_EQ_U32 [[COPY]], [[S_MOV_B32_]], implicit-def $scc + ; WAVE64: [[COPY1:%[0-9]+]]:sreg_32 = COPY $scc + ; WAVE64: $scc = COPY [[COPY1]] + ; WAVE64: [[S_CSELECT_B64_:%[0-9]+]]:sreg_64_xexec = S_CSELECT_B64 0, -4616189618054758400, implicit $scc + ; WAVE64: $sgpr0_sgpr1 = COPY [[S_CSELECT_B64_]] + ; WAVE32-LABEL: name: sitofp_s1_to_s64_s_scc + ; WAVE32: liveins: $sgpr0 + ; WAVE32: $vcc_hi = IMPLICIT_DEF + ; WAVE32: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0 + ; WAVE32: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0 + ; WAVE32: S_CMP_EQ_U32 [[COPY]], [[S_MOV_B32_]], implicit-def $scc + ; WAVE32: [[COPY1:%[0-9]+]]:sreg_32 = COPY $scc + ; WAVE32: $scc = COPY [[COPY1]] + ; WAVE32: [[S_CSELECT_B64_:%[0-9]+]]:sreg_64_xexec = S_CSELECT_B64 0, -4616189618054758400, implicit $scc + ; WAVE32: $sgpr0_sgpr1 = COPY [[S_CSELECT_B64_]] + %0:sgpr(s32) = COPY $sgpr0 + %1:sgpr(s32) = G_CONSTANT i32 0 + %2:scc(s1) = G_ICMP intpred(eq), %0, %1 + %3:sgpr(s64) = G_SITOFP %2 + $sgpr0_sgpr1 = COPY %3 +... + +--- +name: sitofp_s1_to_s32_v_vcc +legalized: true +regBankSelected: true +tracksRegLiveness: true + +body: | + bb.0: + liveins: $vgpr0 + + ; WAVE64-LABEL: name: sitofp_s1_to_s32_v_vcc + ; WAVE64: liveins: $vgpr0 + ; WAVE64: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; WAVE64: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec + ; WAVE64: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[COPY]], [[V_MOV_B32_e32_]], implicit $exec + ; WAVE64: [[V_CNDMASK_B32_e64_:%[0-9]+]]:vgpr_32 = V_CNDMASK_B32_e64 0, 0, 0, 3212836864, [[V_CMP_EQ_U32_e64_]], implicit $exec + ; WAVE64: $vgpr0 = COPY [[V_CNDMASK_B32_e64_]] + ; WAVE32-LABEL: name: sitofp_s1_to_s32_v_vcc + ; WAVE32: liveins: $vgpr0 + ; WAVE32: $vcc_hi = IMPLICIT_DEF + ; WAVE32: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; WAVE32: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec + ; WAVE32: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U32_e64 [[COPY]], [[V_MOV_B32_e32_]], implicit $exec + ; WAVE32: [[V_CNDMASK_B32_e64_:%[0-9]+]]:vgpr_32 = V_CNDMASK_B32_e64 0, 0, 0, 3212836864, [[V_CMP_EQ_U32_e64_]], implicit $exec + ; WAVE32: $vgpr0 = COPY [[V_CNDMASK_B32_e64_]] + %0:vgpr(s32) = COPY $vgpr0 + %1:vgpr(s32) = G_CONSTANT i32 0 + %2:vcc(s1) = G_ICMP intpred(eq), %0, %1 + %3:vgpr(s32) = G_SITOFP %2 + $vgpr0 = COPY %3 +... + +--- +name: sitofp_s1_to_s16_to_v_vcc +legalized: true +regBankSelected: true +tracksRegLiveness: true + +body: | + bb.0: + liveins: $vgpr0 + + ; WAVE64-LABEL: name: sitofp_s1_to_s16_to_v_vcc + ; WAVE64: liveins: $vgpr0 + ; WAVE64: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; WAVE64: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec + ; WAVE64: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[COPY]], [[V_MOV_B32_e32_]], implicit $exec + ; WAVE64: [[V_CNDMASK_B32_e64_:%[0-9]+]]:vgpr_32 = V_CNDMASK_B32_e64 0, 0, 0, 3212836864, [[V_CMP_EQ_U32_e64_]], implicit $exec + ; WAVE64: [[V_CVT_F16_F32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F16_F32_e32 [[V_CNDMASK_B32_e64_]], implicit $exec + ; WAVE64: $vgpr0 = COPY [[V_CVT_F16_F32_e32_]] + ; WAVE32-LABEL: name: sitofp_s1_to_s16_to_v_vcc + ; WAVE32: liveins: $vgpr0 + ; WAVE32: $vcc_hi = IMPLICIT_DEF + ; WAVE32: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; WAVE32: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec + ; WAVE32: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U32_e64 [[COPY]], [[V_MOV_B32_e32_]], implicit $exec + ; WAVE32: [[V_CNDMASK_B32_e64_:%[0-9]+]]:vgpr_32 = V_CNDMASK_B32_e64 0, 0, 0, 3212836864, [[V_CMP_EQ_U32_e64_]], implicit $exec + ; WAVE32: [[V_CVT_F16_F32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F16_F32_e32 [[V_CNDMASK_B32_e64_]], implicit $exec + ; WAVE32: $vgpr0 = COPY [[V_CVT_F16_F32_e32_]] + %0:vgpr(s32) = COPY $vgpr0 + %1:vgpr(s32) = G_CONSTANT i32 0 + %2:vcc(s1) = G_ICMP intpred(eq), %0, %1 + %3:vgpr(s16) = G_SITOFP %2 + %4:vgpr(s32) = G_ANYEXT %3 + $vgpr0 = COPY %4 +... + +--- +name: sitofp_s1_to_s64_v_vcc +legalized: true +regBankSelected: true +tracksRegLiveness: true + +body: | + bb.0: + liveins: $vgpr0 + + ; WAVE64-LABEL: name: sitofp_s1_to_s64_v_vcc + ; WAVE64: liveins: $vgpr0 + ; WAVE64: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; WAVE64: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec + ; WAVE64: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[COPY]], [[V_MOV_B32_e32_]], implicit $exec + ; WAVE64: [[V_CNDMASK_B32_e64_:%[0-9]+]]:vgpr_32 = V_CNDMASK_B32_e64 0, 0, 0, -1, [[V_CMP_EQ_U32_e64_]], implicit $exec + ; WAVE64: [[V_CVT_F64_I32_e32_:%[0-9]+]]:vreg_64 = V_CVT_F64_I32_e32 [[V_CNDMASK_B32_e64_]], implicit $exec + ; WAVE64: $vgpr0_vgpr1 = COPY [[V_CVT_F64_I32_e32_]] + ; WAVE32-LABEL: name: sitofp_s1_to_s64_v_vcc + ; WAVE32: liveins: $vgpr0 + ; WAVE32: $vcc_hi = IMPLICIT_DEF + ; WAVE32: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; WAVE32: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec + ; WAVE32: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U32_e64 [[COPY]], [[V_MOV_B32_e32_]], implicit $exec + ; WAVE32: [[V_CNDMASK_B32_e64_:%[0-9]+]]:vgpr_32 = V_CNDMASK_B32_e64 0, 0, 0, -1, [[V_CMP_EQ_U32_e64_]], implicit $exec + ; WAVE32: [[V_CVT_F64_I32_e32_:%[0-9]+]]:vreg_64 = V_CVT_F64_I32_e32 [[V_CNDMASK_B32_e64_]], implicit $exec + ; WAVE32: $vgpr0_vgpr1 = COPY [[V_CVT_F64_I32_e32_]] + %0:vgpr(s32) = COPY $vgpr0 + %1:vgpr(s32) = G_CONSTANT i32 0 + %2:vcc(s1) = G_ICMP intpred(eq), %0, %1 + %3:vgpr(s64) = G_SITOFP %2 + $vgpr0_vgpr1 = COPY %3 +... diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-uitofp.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-uitofp.mir index 7e3dbaf04155..ee2377a66a20 100644 --- a/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-uitofp.mir +++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/inst-select-uitofp.mir @@ -1,5 +1,6 @@ # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py -# RUN: llc -march=amdgcn -mcpu=hawaii -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck %s -check-prefixes=GCN +# RUN: llc -march=amdgcn -mcpu=hawaii -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck -check-prefix=WAVE64 %s +# RUN: llc -march=amdgcn -mcpu=gfx1010 -run-pass=instruction-select -verify-machineinstrs %s -o - | FileCheck -check-prefix=WAVE32 %s --- name: uitofp_s32_to_s32_vv @@ -11,11 +12,17 @@ body: | bb.0: liveins: $vgpr0 - ; GCN-LABEL: name: uitofp_s32_to_s32_vv - ; GCN: liveins: $vgpr0 - ; GCN: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 - ; GCN: [[V_CVT_F32_U32_e64_:%[0-9]+]]:vgpr_32 = V_CVT_F32_U32_e64 [[COPY]], 0, 0, implicit $exec - ; GCN: $vgpr0 = COPY [[V_CVT_F32_U32_e64_]] + ; WAVE64-LABEL: name: uitofp_s32_to_s32_vv + ; WAVE64: liveins: $vgpr0 + ; WAVE64: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; WAVE64: [[V_CVT_F32_U32_e64_:%[0-9]+]]:vgpr_32 = V_CVT_F32_U32_e64 [[COPY]], 0, 0, implicit $exec + ; WAVE64: $vgpr0 = COPY [[V_CVT_F32_U32_e64_]] + ; WAVE32-LABEL: name: uitofp_s32_to_s32_vv + ; WAVE32: liveins: $vgpr0 + ; WAVE32: $vcc_hi = IMPLICIT_DEF + ; WAVE32: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; WAVE32: [[V_CVT_F32_U32_e64_:%[0-9]+]]:vgpr_32 = V_CVT_F32_U32_e64 [[COPY]], 0, 0, implicit $exec + ; WAVE32: $vgpr0 = COPY [[V_CVT_F32_U32_e64_]] %0:vgpr(s32) = COPY $vgpr0 %1:vgpr(s32) = G_UITOFP %0 $vgpr0 = COPY %1 @@ -31,11 +38,17 @@ body: | bb.0: liveins: $sgpr0 - ; GCN-LABEL: name: uitofp_s32_to_s32_vs - ; GCN: liveins: $sgpr0 - ; GCN: [[COPY:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0 - ; GCN: [[V_CVT_F32_U32_e64_:%[0-9]+]]:vgpr_32 = V_CVT_F32_U32_e64 [[COPY]], 0, 0, implicit $exec - ; GCN: $vgpr0 = COPY [[V_CVT_F32_U32_e64_]] + ; WAVE64-LABEL: name: uitofp_s32_to_s32_vs + ; WAVE64: liveins: $sgpr0 + ; WAVE64: [[COPY:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0 + ; WAVE64: [[V_CVT_F32_U32_e64_:%[0-9]+]]:vgpr_32 = V_CVT_F32_U32_e64 [[COPY]], 0, 0, implicit $exec + ; WAVE64: $vgpr0 = COPY [[V_CVT_F32_U32_e64_]] + ; WAVE32-LABEL: name: uitofp_s32_to_s32_vs + ; WAVE32: liveins: $sgpr0 + ; WAVE32: $vcc_hi = IMPLICIT_DEF + ; WAVE32: [[COPY:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0 + ; WAVE32: [[V_CVT_F32_U32_e64_:%[0-9]+]]:vgpr_32 = V_CVT_F32_U32_e64 [[COPY]], 0, 0, implicit $exec + ; WAVE32: $vgpr0 = COPY [[V_CVT_F32_U32_e64_]] %0:sgpr(s32) = COPY $sgpr0 %1:vgpr(s32) = G_UITOFP %0 $vgpr0 = COPY %1 @@ -51,12 +64,19 @@ body: | bb.0: liveins: $vgpr0 - ; GCN-LABEL: name: uitofp_s32_to_s16_vv - ; GCN: liveins: $vgpr0 - ; GCN: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 - ; GCN: [[V_CVT_F32_U32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F32_U32_e32 [[COPY]], implicit $exec - ; GCN: [[V_CVT_F16_F32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F16_F32_e32 [[V_CVT_F32_U32_e32_]], implicit $exec - ; GCN: $vgpr0 = COPY [[V_CVT_F16_F32_e32_]] + ; WAVE64-LABEL: name: uitofp_s32_to_s16_vv + ; WAVE64: liveins: $vgpr0 + ; WAVE64: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; WAVE64: [[V_CVT_F32_U32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F32_U32_e32 [[COPY]], implicit $exec + ; WAVE64: [[V_CVT_F16_F32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F16_F32_e32 [[V_CVT_F32_U32_e32_]], implicit $exec + ; WAVE64: $vgpr0 = COPY [[V_CVT_F16_F32_e32_]] + ; WAVE32-LABEL: name: uitofp_s32_to_s16_vv + ; WAVE32: liveins: $vgpr0 + ; WAVE32: $vcc_hi = IMPLICIT_DEF + ; WAVE32: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; WAVE32: [[V_CVT_F32_U32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F32_U32_e32 [[COPY]], implicit $exec + ; WAVE32: [[V_CVT_F16_F32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F16_F32_e32 [[V_CVT_F32_U32_e32_]], implicit $exec + ; WAVE32: $vgpr0 = COPY [[V_CVT_F16_F32_e32_]] %0:vgpr(s32) = COPY $vgpr0 %1:vgpr(s16) = G_UITOFP %0 %2:vgpr(s32) = G_ANYEXT %1 @@ -73,14 +93,231 @@ body: | bb.0: liveins: $sgpr0 - ; GCN-LABEL: name: uitofp_s32_to_s16_vs - ; GCN: liveins: $sgpr0 - ; GCN: [[COPY:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0 - ; GCN: [[V_CVT_F32_U32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F32_U32_e32 [[COPY]], implicit $exec - ; GCN: [[V_CVT_F16_F32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F16_F32_e32 [[V_CVT_F32_U32_e32_]], implicit $exec - ; GCN: $vgpr0 = COPY [[V_CVT_F16_F32_e32_]] + ; WAVE64-LABEL: name: uitofp_s32_to_s16_vs + ; WAVE64: liveins: $sgpr0 + ; WAVE64: [[COPY:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0 + ; WAVE64: [[V_CVT_F32_U32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F32_U32_e32 [[COPY]], implicit $exec + ; WAVE64: [[V_CVT_F16_F32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F16_F32_e32 [[V_CVT_F32_U32_e32_]], implicit $exec + ; WAVE64: $vgpr0 = COPY [[V_CVT_F16_F32_e32_]] + ; WAVE32-LABEL: name: uitofp_s32_to_s16_vs + ; WAVE32: liveins: $sgpr0 + ; WAVE32: $vcc_hi = IMPLICIT_DEF + ; WAVE32: [[COPY:%[0-9]+]]:sreg_32_xm0 = COPY $sgpr0 + ; WAVE32: [[V_CVT_F32_U32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F32_U32_e32 [[COPY]], implicit $exec + ; WAVE32: [[V_CVT_F16_F32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F16_F32_e32 [[V_CVT_F32_U32_e32_]], implicit $exec + ; WAVE32: $vgpr0 = COPY [[V_CVT_F16_F32_e32_]] %0:sgpr(s32) = COPY $sgpr0 %1:vgpr(s16) = G_UITOFP %0 %2:vgpr(s32) = G_ANYEXT %1 $vgpr0 = COPY %2 ... + +--- +name: uitofp_s1_to_s32_s_scc +legalized: true +regBankSelected: true +tracksRegLiveness: true + +body: | + bb.0: + liveins: $sgpr0 + + ; WAVE64-LABEL: name: uitofp_s1_to_s32_s_scc + ; WAVE64: liveins: $sgpr0 + ; WAVE64: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0 + ; WAVE64: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0 + ; WAVE64: S_CMP_EQ_U32 [[COPY]], [[S_MOV_B32_]], implicit-def $scc + ; WAVE64: [[COPY1:%[0-9]+]]:sreg_32 = COPY $scc + ; WAVE64: $scc = COPY [[COPY1]] + ; WAVE64: [[S_CSELECT_B32_:%[0-9]+]]:sreg_32_xm0 = S_CSELECT_B32 0, 1065353216, implicit $scc + ; WAVE64: $sgpr0 = COPY [[S_CSELECT_B32_]] + ; WAVE32-LABEL: name: uitofp_s1_to_s32_s_scc + ; WAVE32: liveins: $sgpr0 + ; WAVE32: $vcc_hi = IMPLICIT_DEF + ; WAVE32: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0 + ; WAVE32: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0 + ; WAVE32: S_CMP_EQ_U32 [[COPY]], [[S_MOV_B32_]], implicit-def $scc + ; WAVE32: [[COPY1:%[0-9]+]]:sreg_32 = COPY $scc + ; WAVE32: $scc = COPY [[COPY1]] + ; WAVE32: [[S_CSELECT_B32_:%[0-9]+]]:sreg_32_xm0 = S_CSELECT_B32 0, 1065353216, implicit $scc + ; WAVE32: $sgpr0 = COPY [[S_CSELECT_B32_]] + %0:sgpr(s32) = COPY $sgpr0 + %1:sgpr(s32) = G_CONSTANT i32 0 + %2:scc(s1) = G_ICMP intpred(eq), %0, %1 + %3:sgpr(s32) = G_UITOFP %2 + $sgpr0 = COPY %3 +... + +--- +name: uitofp_s1_to_s16_to_s_scc +legalized: true +regBankSelected: true +tracksRegLiveness: true + +body: | + bb.0: + liveins: $sgpr0 + + ; WAVE64-LABEL: name: uitofp_s1_to_s16_to_s_scc + ; WAVE64: liveins: $sgpr0 + ; WAVE64: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0 + ; WAVE64: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0 + ; WAVE64: S_CMP_EQ_U32 [[COPY]], [[S_MOV_B32_]], implicit-def $scc + ; WAVE64: [[COPY1:%[0-9]+]]:sreg_32 = COPY $scc + ; WAVE64: $scc = COPY [[COPY1]] + ; WAVE64: [[S_CSELECT_B32_:%[0-9]+]]:sreg_32_xm0 = S_CSELECT_B32 0, 15360, implicit $scc + ; WAVE64: $sgpr0 = COPY [[S_CSELECT_B32_]] + ; WAVE32-LABEL: name: uitofp_s1_to_s16_to_s_scc + ; WAVE32: liveins: $sgpr0 + ; WAVE32: $vcc_hi = IMPLICIT_DEF + ; WAVE32: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0 + ; WAVE32: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0 + ; WAVE32: S_CMP_EQ_U32 [[COPY]], [[S_MOV_B32_]], implicit-def $scc + ; WAVE32: [[COPY1:%[0-9]+]]:sreg_32 = COPY $scc + ; WAVE32: $scc = COPY [[COPY1]] + ; WAVE32: [[S_CSELECT_B32_:%[0-9]+]]:sreg_32_xm0 = S_CSELECT_B32 0, 15360, implicit $scc + ; WAVE32: $sgpr0 = COPY [[S_CSELECT_B32_]] + %0:sgpr(s32) = COPY $sgpr0 + %1:sgpr(s32) = G_CONSTANT i32 0 + %2:scc(s1) = G_ICMP intpred(eq), %0, %1 + %3:sgpr(s16) = G_UITOFP %2 + %4:sgpr(s32) = G_ANYEXT %3 + $sgpr0 = COPY %4 +... + +--- +name: uitofp_s1_to_s64_s_scc +legalized: true +regBankSelected: true +tracksRegLiveness: true + +body: | + bb.0: + liveins: $sgpr0 + + ; WAVE64-LABEL: name: uitofp_s1_to_s64_s_scc + ; WAVE64: liveins: $sgpr0 + ; WAVE64: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0 + ; WAVE64: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0 + ; WAVE64: S_CMP_EQ_U32 [[COPY]], [[S_MOV_B32_]], implicit-def $scc + ; WAVE64: [[COPY1:%[0-9]+]]:sreg_32 = COPY $scc + ; WAVE64: $scc = COPY [[COPY1]] + ; WAVE64: [[S_CSELECT_B64_:%[0-9]+]]:sreg_64_xexec = S_CSELECT_B64 0, 4607182418800017408, implicit $scc + ; WAVE64: $sgpr0_sgpr1 = COPY [[S_CSELECT_B64_]] + ; WAVE32-LABEL: name: uitofp_s1_to_s64_s_scc + ; WAVE32: liveins: $sgpr0 + ; WAVE32: $vcc_hi = IMPLICIT_DEF + ; WAVE32: [[COPY:%[0-9]+]]:sreg_32 = COPY $sgpr0 + ; WAVE32: [[S_MOV_B32_:%[0-9]+]]:sreg_32 = S_MOV_B32 0 + ; WAVE32: S_CMP_EQ_U32 [[COPY]], [[S_MOV_B32_]], implicit-def $scc + ; WAVE32: [[COPY1:%[0-9]+]]:sreg_32 = COPY $scc + ; WAVE32: $scc = COPY [[COPY1]] + ; WAVE32: [[S_CSELECT_B64_:%[0-9]+]]:sreg_64_xexec = S_CSELECT_B64 0, 4607182418800017408, implicit $scc + ; WAVE32: $sgpr0_sgpr1 = COPY [[S_CSELECT_B64_]] + %0:sgpr(s32) = COPY $sgpr0 + %1:sgpr(s32) = G_CONSTANT i32 0 + %2:scc(s1) = G_ICMP intpred(eq), %0, %1 + %3:sgpr(s64) = G_UITOFP %2 + $sgpr0_sgpr1 = COPY %3 +... + +--- +name: uitofp_s1_to_s32_v_vcc +legalized: true +regBankSelected: true +tracksRegLiveness: true + +body: | + bb.0: + liveins: $vgpr0 + + ; WAVE64-LABEL: name: uitofp_s1_to_s32_v_vcc + ; WAVE64: liveins: $vgpr0 + ; WAVE64: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; WAVE64: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec + ; WAVE64: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[COPY]], [[V_MOV_B32_e32_]], implicit $exec + ; WAVE64: [[V_CNDMASK_B32_e64_:%[0-9]+]]:vgpr_32 = V_CNDMASK_B32_e64 0, 0, 0, 1065353216, [[V_CMP_EQ_U32_e64_]], implicit $exec + ; WAVE64: $vgpr0 = COPY [[V_CNDMASK_B32_e64_]] + ; WAVE32-LABEL: name: uitofp_s1_to_s32_v_vcc + ; WAVE32: liveins: $vgpr0 + ; WAVE32: $vcc_hi = IMPLICIT_DEF + ; WAVE32: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; WAVE32: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec + ; WAVE32: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U32_e64 [[COPY]], [[V_MOV_B32_e32_]], implicit $exec + ; WAVE32: [[V_CNDMASK_B32_e64_:%[0-9]+]]:vgpr_32 = V_CNDMASK_B32_e64 0, 0, 0, 1065353216, [[V_CMP_EQ_U32_e64_]], implicit $exec + ; WAVE32: $vgpr0 = COPY [[V_CNDMASK_B32_e64_]] + %0:vgpr(s32) = COPY $vgpr0 + %1:vgpr(s32) = G_CONSTANT i32 0 + %2:vcc(s1) = G_ICMP intpred(eq), %0, %1 + %3:vgpr(s32) = G_UITOFP %2 + $vgpr0 = COPY %3 +... + +--- +name: uitofp_s1_to_s16_to_v_vcc +legalized: true +regBankSelected: true +tracksRegLiveness: true + +body: | + bb.0: + liveins: $vgpr0 + + ; WAVE64-LABEL: name: uitofp_s1_to_s16_to_v_vcc + ; WAVE64: liveins: $vgpr0 + ; WAVE64: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; WAVE64: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec + ; WAVE64: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[COPY]], [[V_MOV_B32_e32_]], implicit $exec + ; WAVE64: [[V_CNDMASK_B32_e64_:%[0-9]+]]:vgpr_32 = V_CNDMASK_B32_e64 0, 0, 0, 1065353216, [[V_CMP_EQ_U32_e64_]], implicit $exec + ; WAVE64: [[V_CVT_F16_F32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F16_F32_e32 [[V_CNDMASK_B32_e64_]], implicit $exec + ; WAVE64: $vgpr0 = COPY [[V_CVT_F16_F32_e32_]] + ; WAVE32-LABEL: name: uitofp_s1_to_s16_to_v_vcc + ; WAVE32: liveins: $vgpr0 + ; WAVE32: $vcc_hi = IMPLICIT_DEF + ; WAVE32: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; WAVE32: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec + ; WAVE32: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U32_e64 [[COPY]], [[V_MOV_B32_e32_]], implicit $exec + ; WAVE32: [[V_CNDMASK_B32_e64_:%[0-9]+]]:vgpr_32 = V_CNDMASK_B32_e64 0, 0, 0, 1065353216, [[V_CMP_EQ_U32_e64_]], implicit $exec + ; WAVE32: [[V_CVT_F16_F32_e32_:%[0-9]+]]:vgpr_32 = V_CVT_F16_F32_e32 [[V_CNDMASK_B32_e64_]], implicit $exec + ; WAVE32: $vgpr0 = COPY [[V_CVT_F16_F32_e32_]] + %0:vgpr(s32) = COPY $vgpr0 + %1:vgpr(s32) = G_CONSTANT i32 0 + %2:vcc(s1) = G_ICMP intpred(eq), %0, %1 + %3:vgpr(s16) = G_UITOFP %2 + %4:vgpr(s32) = G_ANYEXT %3 + $vgpr0 = COPY %4 +... + +--- +name: uitofp_s1_to_s64_v_vcc +legalized: true +regBankSelected: true +tracksRegLiveness: true + +body: | + bb.0: + liveins: $vgpr0 + + ; WAVE64-LABEL: name: uitofp_s1_to_s64_v_vcc + ; WAVE64: liveins: $vgpr0 + ; WAVE64: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; WAVE64: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec + ; WAVE64: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_64_xexec = V_CMP_EQ_U32_e64 [[COPY]], [[V_MOV_B32_e32_]], implicit $exec + ; WAVE64: [[V_CNDMASK_B32_e64_:%[0-9]+]]:vgpr_32 = V_CNDMASK_B32_e64 0, 0, 0, 1, [[V_CMP_EQ_U32_e64_]], implicit $exec + ; WAVE64: [[V_CVT_F64_U32_e32_:%[0-9]+]]:vreg_64 = V_CVT_F64_U32_e32 [[V_CNDMASK_B32_e64_]], implicit $exec + ; WAVE64: $vgpr0_vgpr1 = COPY [[V_CVT_F64_U32_e32_]] + ; WAVE32-LABEL: name: uitofp_s1_to_s64_v_vcc + ; WAVE32: liveins: $vgpr0 + ; WAVE32: $vcc_hi = IMPLICIT_DEF + ; WAVE32: [[COPY:%[0-9]+]]:vgpr_32 = COPY $vgpr0 + ; WAVE32: [[V_MOV_B32_e32_:%[0-9]+]]:vgpr_32 = V_MOV_B32_e32 0, implicit $exec + ; WAVE32: [[V_CMP_EQ_U32_e64_:%[0-9]+]]:sreg_32_xm0_xexec = V_CMP_EQ_U32_e64 [[COPY]], [[V_MOV_B32_e32_]], implicit $exec + ; WAVE32: [[V_CNDMASK_B32_e64_:%[0-9]+]]:vgpr_32 = V_CNDMASK_B32_e64 0, 0, 0, 1, [[V_CMP_EQ_U32_e64_]], implicit $exec + ; WAVE32: [[V_CVT_F64_U32_e32_:%[0-9]+]]:vreg_64 = V_CVT_F64_U32_e32 [[V_CNDMASK_B32_e64_]], implicit $exec + ; WAVE32: $vgpr0_vgpr1 = COPY [[V_CVT_F64_U32_e32_]] + %0:vgpr(s32) = COPY $vgpr0 + %1:vgpr(s32) = G_CONSTANT i32 0 + %2:vcc(s1) = G_ICMP intpred(eq), %0, %1 + %3:vgpr(s64) = G_UITOFP %2 + $vgpr0_vgpr1 = COPY %3 +...