[llvm] [GlobalISel] Fold G_ICMP if possible (PR #86357)
Shilei Tian via llvm-commits
llvm-commits at lists.llvm.org
Mon Mar 25 14:08:07 PDT 2024
https://github.com/shiltian updated https://github.com/llvm/llvm-project/pull/86357
>From 19c86d90b28536f1665068ec21020b580e00f92c Mon Sep 17 00:00:00 2001
From: Shilei Tian <i at tianshilei.me>
Date: Mon, 25 Mar 2024 17:06:46 -0400
Subject: [PATCH] [GlobalISel] Fold G_ICMP if possible
This patch tries to fold `G_ICMP` if possible.
---
llvm/include/llvm/CodeGen/GlobalISel/Utils.h | 4 ++
llvm/lib/CodeGen/GlobalISel/CSEMIRBuilder.cpp | 14 ++++
.../CodeGen/GlobalISel/LegalizerHelper.cpp | 16 +++--
llvm/lib/CodeGen/GlobalISel/Utils.cpp | 68 +++++++++++++++++++
.../legalize-atomic-cmpxchg-with-success.mir | 15 ++--
.../AMDGPU/GlobalISel/legalize-saddo.mir | 31 +++++----
.../AMDGPU/GlobalISel/legalize-saddsat.mir | 45 +++++++-----
.../AMDGPU/GlobalISel/legalize-ssubo.mir | 31 +++++----
.../AMDGPU/GlobalISel/legalize-ssubsat.mir | 45 +++++++-----
9 files changed, 197 insertions(+), 72 deletions(-)
diff --git a/llvm/include/llvm/CodeGen/GlobalISel/Utils.h b/llvm/include/llvm/CodeGen/GlobalISel/Utils.h
index f8900f3434ccaa..fc4ebf795a334e 100644
--- a/llvm/include/llvm/CodeGen/GlobalISel/Utils.h
+++ b/llvm/include/llvm/CodeGen/GlobalISel/Utils.h
@@ -313,6 +313,10 @@ std::optional<APFloat> ConstantFoldIntToFloat(unsigned Opcode, LLT DstTy,
std::optional<SmallVector<unsigned>>
ConstantFoldCTLZ(Register Src, const MachineRegisterInfo &MRI);
+std::optional<SmallVector<APInt>>
+ConstantFoldICmp(unsigned Pred, const Register Op1, const Register Op2,
+ const MachineRegisterInfo &MRI);
+
/// Test if the given value is known to have exactly one bit set. This differs
/// from computeKnownBits in that it doesn't necessarily determine which bit is
/// set.
diff --git a/llvm/lib/CodeGen/GlobalISel/CSEMIRBuilder.cpp b/llvm/lib/CodeGen/GlobalISel/CSEMIRBuilder.cpp
index 1869e0d41a51f6..44e1be2a5e080d 100644
--- a/llvm/lib/CodeGen/GlobalISel/CSEMIRBuilder.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/CSEMIRBuilder.cpp
@@ -174,6 +174,20 @@ MachineInstrBuilder CSEMIRBuilder::buildInstr(unsigned Opc,
switch (Opc) {
default:
break;
+ case TargetOpcode::G_ICMP: {
+ assert(SrcOps.size() == 3 && "Invalid sources");
+ assert(DstOps.size() == 1 && "Invalid dsts");
+ LLT SrcTy = SrcOps[1].getLLTTy(*getMRI());
+
+ if (std::optional<SmallVector<APInt>> Cst =
+ ConstantFoldICmp(SrcOps[0].getPredicate(), SrcOps[1].getReg(),
+ SrcOps[2].getReg(), *getMRI())) {
+ if (SrcTy.isVector())
+ return buildBuildVectorConstant(DstOps[0], *Cst);
+ return buildConstant(DstOps[0], Cst->front());
+ }
+ break;
+ }
case TargetOpcode::G_ADD:
case TargetOpcode::G_PTR_ADD:
case TargetOpcode::G_AND:
diff --git a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
index 1b25da8833e4fb..852cffda82ddef 100644
--- a/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
@@ -3741,9 +3741,11 @@ LegalizerHelper::lower(MachineInstr &MI, unsigned TypeIdx, LLT LowerHintTy) {
}
case TargetOpcode::G_ATOMIC_CMPXCHG_WITH_SUCCESS: {
auto [OldValRes, SuccessRes, Addr, CmpVal, NewVal] = MI.getFirst5Regs();
- MIRBuilder.buildAtomicCmpXchg(OldValRes, Addr, CmpVal, NewVal,
+ auto Tmp = MRI.createGenericVirtualRegister(MRI.getType(OldValRes));
+ MIRBuilder.buildAtomicCmpXchg(Tmp, Addr, CmpVal, NewVal,
**MI.memoperands_begin());
- MIRBuilder.buildICmp(CmpInst::ICMP_EQ, SuccessRes, OldValRes, CmpVal);
+ MIRBuilder.buildCopy(OldValRes, Tmp);
+ MIRBuilder.buildICmp(CmpInst::ICMP_EQ, SuccessRes, Tmp, CmpVal);
MI.eraseFromParent();
return Legalized;
}
@@ -7622,10 +7624,14 @@ LegalizerHelper::lowerSADDO_SSUBO(MachineInstr &MI) {
LLT Ty = Dst0Ty;
LLT BoolTy = Dst1Ty;
+ auto Tmp = MRI.createGenericVirtualRegister(MRI.getType(Dst0));
+
if (IsAdd)
- MIRBuilder.buildAdd(Dst0, LHS, RHS);
+ MIRBuilder.buildAdd(Tmp, LHS, RHS);
else
- MIRBuilder.buildSub(Dst0, LHS, RHS);
+ MIRBuilder.buildSub(Tmp, LHS, RHS);
+
+ MIRBuilder.buildCopy(Dst0, Tmp);
// TODO: If SADDSAT/SSUBSAT is legal, compare results to detect overflow.
@@ -7638,7 +7644,7 @@ LegalizerHelper::lowerSADDO_SSUBO(MachineInstr &MI) {
// (LHS) if and only if the other operand (RHS) is (non-zero) positive,
// otherwise there will be overflow.
auto ResultLowerThanLHS =
- MIRBuilder.buildICmp(CmpInst::ICMP_SLT, BoolTy, Dst0, LHS);
+ MIRBuilder.buildICmp(CmpInst::ICMP_SLT, BoolTy, Tmp, LHS);
auto ConditionRHS = MIRBuilder.buildICmp(
IsAdd ? CmpInst::ICMP_SLT : CmpInst::ICMP_SGT, BoolTy, RHS, Zero);
diff --git a/llvm/lib/CodeGen/GlobalISel/Utils.cpp b/llvm/lib/CodeGen/GlobalISel/Utils.cpp
index a9fa73b60a097f..ce3062270f8e6d 100644
--- a/llvm/lib/CodeGen/GlobalISel/Utils.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/Utils.cpp
@@ -996,6 +996,74 @@ llvm::ConstantFoldCTLZ(Register Src, const MachineRegisterInfo &MRI) {
return std::nullopt;
}
+std::optional<SmallVector<APInt>>
+llvm::ConstantFoldICmp(unsigned Pred, const Register Op1, const Register Op2,
+ const MachineRegisterInfo &MRI) {
+ LLT Ty = MRI.getType(Op1);
+ if (Ty != MRI.getType(Op2))
+ return std::nullopt;
+
+ auto TryFoldScalar = [&MRI, Pred](Register LHS,
+ Register RHS) -> std::optional<APInt> {
+ auto LHSCst = getIConstantVRegVal(LHS, MRI);
+ auto RHSCst = getIConstantVRegVal(RHS, MRI);
+ if (!LHSCst || !RHSCst)
+ return std::nullopt;
+
+ switch (Pred) {
+ case CmpInst::Predicate::ICMP_EQ:
+ return APInt(/*numBits=*/1, LHSCst->eq(*RHSCst));
+ case CmpInst::Predicate::ICMP_NE:
+ return APInt(/*numBits=*/1, LHSCst->ne(*RHSCst));
+ case CmpInst::Predicate::ICMP_UGT:
+ return APInt(/*numBits=*/1, LHSCst->ugt(*RHSCst));
+ case CmpInst::Predicate::ICMP_UGE:
+ return APInt(/*numBits=*/1, LHSCst->uge(*RHSCst));
+ case CmpInst::Predicate::ICMP_ULT:
+ return APInt(/*numBits=*/1, LHSCst->ult(*RHSCst));
+ case CmpInst::Predicate::ICMP_ULE:
+ return APInt(/*numBits=*/1, LHSCst->ule(*RHSCst));
+ case CmpInst::Predicate::ICMP_SGT:
+ return APInt(/*numBits=*/1, LHSCst->sgt(*RHSCst));
+ case CmpInst::Predicate::ICMP_SGE:
+ return APInt(/*numBits=*/1, LHSCst->sge(*RHSCst));
+ case CmpInst::Predicate::ICMP_SLT:
+ return APInt(/*numBits=*/1, LHSCst->slt(*RHSCst));
+ case CmpInst::Predicate::ICMP_SLE:
+ return APInt(/*numBits=*/1, LHSCst->sle(*RHSCst));
+ default:
+ return std::nullopt;
+ }
+ };
+
+ SmallVector<APInt> FoldedICmps;
+
+ if (Ty.isVector()) {
+ // Try to constant fold each element.
+ auto *BV1 = getOpcodeDef<GBuildVector>(Op1, MRI);
+ auto *BV2 = getOpcodeDef<GBuildVector>(Op2, MRI);
+ if (!BV1 || !BV2)
+ return std::nullopt;
+ assert(BV1->getNumSources() == BV2->getNumSources() && "Invalid vectors");
+ for (unsigned I = 0; I < BV1->getNumSources(); ++I) {
+ if (auto MaybeFold =
+ TryFoldScalar(BV1->getSourceReg(I), BV2->getSourceReg(I))) {
+ FoldedICmps.emplace_back(*MaybeFold);
+ continue;
+ }
+ return std::nullopt;
+ }
+ return FoldedICmps;
+ }
+
+ if (auto MaybeCst = TryFoldScalar(Op1, Op2)) {
+ FoldedICmps.emplace_back(*MaybeCst);
+ return FoldedICmps;
+ }
+
+ return std::nullopt;
+}
+
bool llvm::isKnownToBeAPowerOfTwo(Register Reg, const MachineRegisterInfo &MRI,
GISelKnownBits *KB) {
std::optional<DefinitionAndSourceRegister> DefSrcReg =
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-atomic-cmpxchg-with-success.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-atomic-cmpxchg-with-success.mir
index e288d9d5ab3c09..171a8551ae69a9 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-atomic-cmpxchg-with-success.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-atomic-cmpxchg-with-success.mir
@@ -15,8 +15,9 @@ body: |
; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr3
; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY2]](s32), [[COPY1]](s32)
; CHECK-NEXT: [[AMDGPU_ATOMIC_CMPXCHG:%[0-9]+]]:_(s32) = G_AMDGPU_ATOMIC_CMPXCHG [[COPY]](p1), [[BUILD_VECTOR]] :: (load store syncscope("agent-one-as") monotonic monotonic (s32), addrspace 1)
+ ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[AMDGPU_ATOMIC_CMPXCHG]](s32)
; CHECK-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(eq), [[AMDGPU_ATOMIC_CMPXCHG]](s32), [[COPY1]]
- ; CHECK-NEXT: S_ENDPGM 0, implicit [[AMDGPU_ATOMIC_CMPXCHG]](s32), implicit [[ICMP]](s1)
+ ; CHECK-NEXT: S_ENDPGM 0, implicit [[COPY3]](s32), implicit [[ICMP]](s1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s32) = COPY $vgpr2
%2:_(s32) = COPY $vgpr3
@@ -39,8 +40,9 @@ body: |
; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr3
; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY2]](s32), [[COPY1]](s32)
; CHECK-NEXT: [[AMDGPU_ATOMIC_CMPXCHG:%[0-9]+]]:_(s32) = G_AMDGPU_ATOMIC_CMPXCHG [[COPY]](p0), [[BUILD_VECTOR]] :: (load store syncscope("agent-one-as") monotonic monotonic (s32))
+ ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[AMDGPU_ATOMIC_CMPXCHG]](s32)
; CHECK-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(eq), [[AMDGPU_ATOMIC_CMPXCHG]](s32), [[COPY1]]
- ; CHECK-NEXT: S_ENDPGM 0, implicit [[AMDGPU_ATOMIC_CMPXCHG]](s32), implicit [[ICMP]](s1)
+ ; CHECK-NEXT: S_ENDPGM 0, implicit [[COPY3]](s32), implicit [[ICMP]](s1)
%0:_(p0) = COPY $vgpr0_vgpr1
%1:_(s32) = COPY $vgpr2
%2:_(s32) = COPY $vgpr3
@@ -62,8 +64,9 @@ body: |
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY $vgpr2
; CHECK-NEXT: [[ATOMIC_CMPXCHG:%[0-9]+]]:_(s32) = G_ATOMIC_CMPXCHG [[COPY]](p3), [[COPY1]], [[COPY2]] :: (load store syncscope("agent-one-as") monotonic monotonic (s32), addrspace 3)
+ ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[ATOMIC_CMPXCHG]](s32)
; CHECK-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(eq), [[ATOMIC_CMPXCHG]](s32), [[COPY1]]
- ; CHECK-NEXT: S_ENDPGM 0, implicit [[ATOMIC_CMPXCHG]](s32), implicit [[ICMP]](s1)
+ ; CHECK-NEXT: S_ENDPGM 0, implicit [[COPY3]](s32), implicit [[ICMP]](s1)
%0:_(p3) = COPY $vgpr0
%1:_(s32) = COPY $vgpr1
%2:_(s32) = COPY $vgpr2
@@ -86,8 +89,9 @@ body: |
; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY $vgpr4_vgpr5
; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[COPY2]](s64), [[COPY1]](s64)
; CHECK-NEXT: [[AMDGPU_ATOMIC_CMPXCHG:%[0-9]+]]:_(s64) = G_AMDGPU_ATOMIC_CMPXCHG [[COPY]](p1), [[BUILD_VECTOR]] :: (load store syncscope("agent-one-as") monotonic monotonic (s64), addrspace 1)
+ ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s64) = COPY [[AMDGPU_ATOMIC_CMPXCHG]](s64)
; CHECK-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(eq), [[AMDGPU_ATOMIC_CMPXCHG]](s64), [[COPY1]]
- ; CHECK-NEXT: S_ENDPGM 0, implicit [[AMDGPU_ATOMIC_CMPXCHG]](s64), implicit [[ICMP]](s1)
+ ; CHECK-NEXT: S_ENDPGM 0, implicit [[COPY3]](s64), implicit [[ICMP]](s1)
%0:_(p1) = COPY $vgpr0_vgpr1
%1:_(s64) = COPY $vgpr2_vgpr3
%2:_(s64) = COPY $vgpr4_vgpr5
@@ -109,8 +113,9 @@ body: |
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s64) = COPY $vgpr1_vgpr2
; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY $vgpr3_vgpr4
; CHECK-NEXT: [[ATOMIC_CMPXCHG:%[0-9]+]]:_(s64) = G_ATOMIC_CMPXCHG [[COPY]](p3), [[COPY1]], [[COPY2]] :: (load store syncscope("agent-one-as") monotonic monotonic (s64), addrspace 3)
+ ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s64) = COPY [[ATOMIC_CMPXCHG]](s64)
; CHECK-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(eq), [[ATOMIC_CMPXCHG]](s64), [[COPY1]]
- ; CHECK-NEXT: S_ENDPGM 0, implicit [[ATOMIC_CMPXCHG]](s64), implicit [[ICMP]](s1)
+ ; CHECK-NEXT: S_ENDPGM 0, implicit [[COPY3]](s64), implicit [[ICMP]](s1)
%0:_(p3) = COPY $vgpr0
%1:_(s64) = COPY $vgpr1_vgpr2
%2:_(s64) = COPY $vgpr3_vgpr4
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-saddo.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-saddo.mir
index dba20e128237cd..bd932e08f4a7a9 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-saddo.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-saddo.mir
@@ -82,12 +82,13 @@ body: |
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s32) = G_ADD [[COPY]], [[COPY1]]
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[ADD]](s32)
; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; CHECK-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[ADD]](s32), [[COPY]]
; CHECK-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[COPY1]](s32), [[C]]
; CHECK-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP1]], [[ICMP]]
; CHECK-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[XOR]](s1)
- ; CHECK-NEXT: $vgpr0 = COPY [[ADD]](s32)
+ ; CHECK-NEXT: $vgpr0 = COPY [[COPY2]](s32)
; CHECK-NEXT: $vgpr1 = COPY [[ZEXT]](s32)
%0:_(s32) = COPY $vgpr0
%1:_(s32) = COPY $vgpr1
@@ -113,12 +114,13 @@ body: |
; CHECK-NEXT: [[UADDO:%[0-9]+]]:_(s32), [[UADDO1:%[0-9]+]]:_(s1) = G_UADDO [[UV]], [[UV2]]
; CHECK-NEXT: [[UADDE:%[0-9]+]]:_(s32), [[UADDE1:%[0-9]+]]:_(s1) = G_UADDE [[UV1]], [[UV3]], [[UADDO1]]
; CHECK-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO]](s32), [[UADDE]](s32)
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[MV]](s64)
; CHECK-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
; CHECK-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[MV]](s64), [[COPY]]
; CHECK-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[COPY1]](s64), [[C]]
; CHECK-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP1]], [[ICMP]]
; CHECK-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[XOR]](s1)
- ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[MV]](s64)
+ ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[COPY2]](s64)
; CHECK-NEXT: $vgpr2 = COPY [[ZEXT]](s32)
%0:_(s64) = COPY $vgpr0_vgpr1
%1:_(s64) = COPY $vgpr2_vgpr3
@@ -152,6 +154,7 @@ body: |
; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY [[BITCAST2]](<2 x s16>)
; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
; CHECK-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
; CHECK-NEXT: [[SEXT_INREG:%[0-9]+]]:_(s32) = G_SEXT_INREG [[ADD]], 16
@@ -164,8 +167,8 @@ body: |
; CHECK-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
; CHECK-NEXT: [[SEXT_INREG4:%[0-9]+]]:_(s32) = G_SEXT_INREG [[BITCAST4]], 16
; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
- ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
- ; CHECK-NEXT: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SEXT_INREG4]](s32), [[COPY2]]
+ ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+ ; CHECK-NEXT: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SEXT_INREG4]](s32), [[COPY3]]
; CHECK-NEXT: [[SEXT_INREG5:%[0-9]+]]:_(s32) = G_SEXT_INREG [[LSHR3]], 16
; CHECK-NEXT: [[ICMP3:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SEXT_INREG5]](s32), [[C2]]
; CHECK-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP2]], [[ICMP]]
@@ -176,7 +179,7 @@ body: |
; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[ANYEXT]], [[C3]]
; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[ANYEXT1]], [[C3]]
; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[AND2]](s32), [[AND3]](s32)
- ; CHECK-NEXT: $vgpr0 = COPY [[BITCAST2]](<2 x s16>)
+ ; CHECK-NEXT: $vgpr0 = COPY [[COPY2]](<2 x s16>)
; CHECK-NEXT: $vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<2 x s32>)
%0:_(<2 x s16>) = COPY $vgpr0
%1:_(<2 x s16>) = COPY $vgpr1
@@ -318,6 +321,7 @@ body: |
; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>)
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(<4 x s16>) = COPY [[CONCAT_VECTORS]](<4 x s16>)
; CHECK-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
; CHECK-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
; CHECK-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32)
@@ -342,14 +346,14 @@ body: |
; CHECK-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32)
; CHECK-NEXT: [[SEXT_INREG8:%[0-9]+]]:_(s32) = G_SEXT_INREG [[BITCAST8]], 16
; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
- ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
- ; CHECK-NEXT: [[ICMP4:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SEXT_INREG8]](s32), [[COPY2]]
- ; CHECK-NEXT: [[SEXT_INREG9:%[0-9]+]]:_(s32) = G_SEXT_INREG [[LSHR6]], 16
; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
- ; CHECK-NEXT: [[ICMP5:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SEXT_INREG9]](s32), [[COPY3]]
- ; CHECK-NEXT: [[SEXT_INREG10:%[0-9]+]]:_(s32) = G_SEXT_INREG [[BITCAST9]], 16
+ ; CHECK-NEXT: [[ICMP4:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SEXT_INREG8]](s32), [[COPY3]]
+ ; CHECK-NEXT: [[SEXT_INREG9:%[0-9]+]]:_(s32) = G_SEXT_INREG [[LSHR6]], 16
; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
- ; CHECK-NEXT: [[ICMP6:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SEXT_INREG10]](s32), [[COPY4]]
+ ; CHECK-NEXT: [[ICMP5:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SEXT_INREG9]](s32), [[COPY4]]
+ ; CHECK-NEXT: [[SEXT_INREG10:%[0-9]+]]:_(s32) = G_SEXT_INREG [[BITCAST9]], 16
+ ; CHECK-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+ ; CHECK-NEXT: [[ICMP6:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SEXT_INREG10]](s32), [[COPY5]]
; CHECK-NEXT: [[SEXT_INREG11:%[0-9]+]]:_(s32) = G_SEXT_INREG [[LSHR7]], 16
; CHECK-NEXT: [[ICMP7:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SEXT_INREG11]](s32), [[C2]]
; CHECK-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP4]], [[ICMP]]
@@ -366,7 +370,7 @@ body: |
; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[ANYEXT2]], [[C3]]
; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[ANYEXT3]], [[C3]]
; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[AND4]](s32), [[AND5]](s32), [[AND6]](s32), [[AND7]](s32)
- ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[COPY2]](<4 x s16>)
; CHECK-NEXT: $vgpr2_vgpr3_vgpr4_vgpr5 = COPY [[BUILD_VECTOR]](<4 x s32>)
%0:_(<4 x s16>) = COPY $vgpr0_vgpr1
%1:_(<4 x s16>) = COPY $vgpr1_vgpr2
@@ -392,6 +396,7 @@ body: |
; CHECK-NEXT: [[ADD:%[0-9]+]]:_(s32) = G_ADD [[UV]], [[UV2]]
; CHECK-NEXT: [[ADD1:%[0-9]+]]:_(s32) = G_ADD [[UV1]], [[UV3]]
; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[ADD]](s32), [[ADD1]](s32)
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s32>) = COPY [[BUILD_VECTOR]](<2 x s32>)
; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>)
; CHECK-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[ADD]](s32), [[UV4]]
@@ -407,7 +412,7 @@ body: |
; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[ANYEXT]], [[C1]]
; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[ANYEXT1]], [[C1]]
; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[AND]](s32), [[AND1]](s32)
- ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[COPY2]](<2 x s32>)
; CHECK-NEXT: $vgpr2_vgpr3 = COPY [[BUILD_VECTOR1]](<2 x s32>)
%0:_(<2 x s32>) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = COPY $vgpr2_vgpr3
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-saddsat.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-saddsat.mir
index 93d00714158be5..eabfcc447697a5 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-saddsat.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-saddsat.mir
@@ -951,19 +951,20 @@ body: |
; GFX6-NEXT: [[UADDO:%[0-9]+]]:_(s32), [[UADDO1:%[0-9]+]]:_(s1) = G_UADDO [[UV]], [[UV2]]
; GFX6-NEXT: [[UADDE:%[0-9]+]]:_(s32), [[UADDE1:%[0-9]+]]:_(s1) = G_UADDE [[UV1]], [[UV3]], [[UADDO1]]
; GFX6-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO]](s32), [[UADDE]](s32)
+ ; GFX6-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[MV]](s64)
; GFX6-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
; GFX6-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[MV]](s64), [[COPY]]
; GFX6-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[COPY1]](s64), [[C]]
; GFX6-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP1]], [[ICMP]]
; GFX6-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 63
- ; GFX6-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[MV]], [[C1]](s32)
+ ; GFX6-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[COPY2]], [[C1]](s32)
; GFX6-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 -9223372036854775808
; GFX6-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[ASHR]](s64)
; GFX6-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[C2]](s64)
; GFX6-NEXT: [[UADDO2:%[0-9]+]]:_(s32), [[UADDO3:%[0-9]+]]:_(s1) = G_UADDO [[UV4]], [[UV6]]
; GFX6-NEXT: [[UADDE2:%[0-9]+]]:_(s32), [[UADDE3:%[0-9]+]]:_(s1) = G_UADDE [[UV5]], [[UV7]], [[UADDO3]]
; GFX6-NEXT: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO2]](s32), [[UADDE2]](s32)
- ; GFX6-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[MV]]
+ ; GFX6-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[COPY2]]
; GFX6-NEXT: $vgpr0_vgpr1 = COPY [[SELECT]](s64)
;
; GFX8-LABEL: name: saddsat_s64
@@ -976,19 +977,20 @@ body: |
; GFX8-NEXT: [[UADDO:%[0-9]+]]:_(s32), [[UADDO1:%[0-9]+]]:_(s1) = G_UADDO [[UV]], [[UV2]]
; GFX8-NEXT: [[UADDE:%[0-9]+]]:_(s32), [[UADDE1:%[0-9]+]]:_(s1) = G_UADDE [[UV1]], [[UV3]], [[UADDO1]]
; GFX8-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO]](s32), [[UADDE]](s32)
+ ; GFX8-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[MV]](s64)
; GFX8-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
; GFX8-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[MV]](s64), [[COPY]]
; GFX8-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[COPY1]](s64), [[C]]
; GFX8-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP1]], [[ICMP]]
; GFX8-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 63
- ; GFX8-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[MV]], [[C1]](s32)
+ ; GFX8-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[COPY2]], [[C1]](s32)
; GFX8-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 -9223372036854775808
; GFX8-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[ASHR]](s64)
; GFX8-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[C2]](s64)
; GFX8-NEXT: [[UADDO2:%[0-9]+]]:_(s32), [[UADDO3:%[0-9]+]]:_(s1) = G_UADDO [[UV4]], [[UV6]]
; GFX8-NEXT: [[UADDE2:%[0-9]+]]:_(s32), [[UADDE3:%[0-9]+]]:_(s1) = G_UADDE [[UV5]], [[UV7]], [[UADDO3]]
; GFX8-NEXT: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO2]](s32), [[UADDE2]](s32)
- ; GFX8-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[MV]]
+ ; GFX8-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[COPY2]]
; GFX8-NEXT: $vgpr0_vgpr1 = COPY [[SELECT]](s64)
;
; GFX9-LABEL: name: saddsat_s64
@@ -1001,19 +1003,20 @@ body: |
; GFX9-NEXT: [[UADDO:%[0-9]+]]:_(s32), [[UADDO1:%[0-9]+]]:_(s1) = G_UADDO [[UV]], [[UV2]]
; GFX9-NEXT: [[UADDE:%[0-9]+]]:_(s32), [[UADDE1:%[0-9]+]]:_(s1) = G_UADDE [[UV1]], [[UV3]], [[UADDO1]]
; GFX9-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO]](s32), [[UADDE]](s32)
+ ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[MV]](s64)
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
; GFX9-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[MV]](s64), [[COPY]]
; GFX9-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[COPY1]](s64), [[C]]
; GFX9-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP1]], [[ICMP]]
; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 63
- ; GFX9-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[MV]], [[C1]](s32)
+ ; GFX9-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[COPY2]], [[C1]](s32)
; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 -9223372036854775808
; GFX9-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[ASHR]](s64)
; GFX9-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[C2]](s64)
; GFX9-NEXT: [[UADDO2:%[0-9]+]]:_(s32), [[UADDO3:%[0-9]+]]:_(s1) = G_UADDO [[UV4]], [[UV6]]
; GFX9-NEXT: [[UADDE2:%[0-9]+]]:_(s32), [[UADDE3:%[0-9]+]]:_(s1) = G_UADDE [[UV5]], [[UV7]], [[UADDO3]]
; GFX9-NEXT: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO2]](s32), [[UADDE2]](s32)
- ; GFX9-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[MV]]
+ ; GFX9-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[COPY2]]
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[SELECT]](s64)
%0:_(s64) = COPY $vgpr0_vgpr1
%1:_(s64) = COPY $vgpr2_vgpr3
@@ -1039,34 +1042,36 @@ body: |
; GFX6-NEXT: [[UADDO:%[0-9]+]]:_(s32), [[UADDO1:%[0-9]+]]:_(s1) = G_UADDO [[UV4]], [[UV6]]
; GFX6-NEXT: [[UADDE:%[0-9]+]]:_(s32), [[UADDE1:%[0-9]+]]:_(s1) = G_UADDE [[UV5]], [[UV7]], [[UADDO1]]
; GFX6-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO]](s32), [[UADDE]](s32)
+ ; GFX6-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[MV]](s64)
; GFX6-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
; GFX6-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[MV]](s64), [[UV]]
; GFX6-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[UV2]](s64), [[C]]
; GFX6-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP1]], [[ICMP]]
; GFX6-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 63
- ; GFX6-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[MV]], [[C1]](s32)
+ ; GFX6-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[COPY2]], [[C1]](s32)
; GFX6-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 -9223372036854775808
; GFX6-NEXT: [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[ASHR]](s64)
; GFX6-NEXT: [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[C2]](s64)
; GFX6-NEXT: [[UADDO2:%[0-9]+]]:_(s32), [[UADDO3:%[0-9]+]]:_(s1) = G_UADDO [[UV8]], [[UV10]]
; GFX6-NEXT: [[UADDE2:%[0-9]+]]:_(s32), [[UADDE3:%[0-9]+]]:_(s1) = G_UADDE [[UV9]], [[UV11]], [[UADDO3]]
; GFX6-NEXT: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO2]](s32), [[UADDE2]](s32)
- ; GFX6-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[MV]]
+ ; GFX6-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[COPY2]]
; GFX6-NEXT: [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV1]](s64)
; GFX6-NEXT: [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV3]](s64)
; GFX6-NEXT: [[UADDO4:%[0-9]+]]:_(s32), [[UADDO5:%[0-9]+]]:_(s1) = G_UADDO [[UV12]], [[UV14]]
; GFX6-NEXT: [[UADDE4:%[0-9]+]]:_(s32), [[UADDE5:%[0-9]+]]:_(s1) = G_UADDE [[UV13]], [[UV15]], [[UADDO5]]
; GFX6-NEXT: [[MV2:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO4]](s32), [[UADDE4]](s32)
+ ; GFX6-NEXT: [[COPY3:%[0-9]+]]:_(s64) = COPY [[MV2]](s64)
; GFX6-NEXT: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[MV2]](s64), [[UV1]]
; GFX6-NEXT: [[ICMP3:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[UV3]](s64), [[C]]
; GFX6-NEXT: [[XOR1:%[0-9]+]]:_(s1) = G_XOR [[ICMP3]], [[ICMP2]]
- ; GFX6-NEXT: [[ASHR1:%[0-9]+]]:_(s64) = G_ASHR [[MV2]], [[C1]](s32)
+ ; GFX6-NEXT: [[ASHR1:%[0-9]+]]:_(s64) = G_ASHR [[COPY3]], [[C1]](s32)
; GFX6-NEXT: [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[ASHR1]](s64)
; GFX6-NEXT: [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[C2]](s64)
; GFX6-NEXT: [[UADDO6:%[0-9]+]]:_(s32), [[UADDO7:%[0-9]+]]:_(s1) = G_UADDO [[UV16]], [[UV18]]
; GFX6-NEXT: [[UADDE6:%[0-9]+]]:_(s32), [[UADDE7:%[0-9]+]]:_(s1) = G_UADDE [[UV17]], [[UV19]], [[UADDO7]]
; GFX6-NEXT: [[MV3:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO6]](s32), [[UADDE6]](s32)
- ; GFX6-NEXT: [[SELECT1:%[0-9]+]]:_(s64) = G_SELECT [[XOR1]](s1), [[MV3]], [[MV2]]
+ ; GFX6-NEXT: [[SELECT1:%[0-9]+]]:_(s64) = G_SELECT [[XOR1]](s1), [[MV3]], [[COPY3]]
; GFX6-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[SELECT]](s64), [[SELECT1]](s64)
; GFX6-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
;
@@ -1082,34 +1087,36 @@ body: |
; GFX8-NEXT: [[UADDO:%[0-9]+]]:_(s32), [[UADDO1:%[0-9]+]]:_(s1) = G_UADDO [[UV4]], [[UV6]]
; GFX8-NEXT: [[UADDE:%[0-9]+]]:_(s32), [[UADDE1:%[0-9]+]]:_(s1) = G_UADDE [[UV5]], [[UV7]], [[UADDO1]]
; GFX8-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO]](s32), [[UADDE]](s32)
+ ; GFX8-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[MV]](s64)
; GFX8-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
; GFX8-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[MV]](s64), [[UV]]
; GFX8-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[UV2]](s64), [[C]]
; GFX8-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP1]], [[ICMP]]
; GFX8-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 63
- ; GFX8-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[MV]], [[C1]](s32)
+ ; GFX8-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[COPY2]], [[C1]](s32)
; GFX8-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 -9223372036854775808
; GFX8-NEXT: [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[ASHR]](s64)
; GFX8-NEXT: [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[C2]](s64)
; GFX8-NEXT: [[UADDO2:%[0-9]+]]:_(s32), [[UADDO3:%[0-9]+]]:_(s1) = G_UADDO [[UV8]], [[UV10]]
; GFX8-NEXT: [[UADDE2:%[0-9]+]]:_(s32), [[UADDE3:%[0-9]+]]:_(s1) = G_UADDE [[UV9]], [[UV11]], [[UADDO3]]
; GFX8-NEXT: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO2]](s32), [[UADDE2]](s32)
- ; GFX8-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[MV]]
+ ; GFX8-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[COPY2]]
; GFX8-NEXT: [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV1]](s64)
; GFX8-NEXT: [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV3]](s64)
; GFX8-NEXT: [[UADDO4:%[0-9]+]]:_(s32), [[UADDO5:%[0-9]+]]:_(s1) = G_UADDO [[UV12]], [[UV14]]
; GFX8-NEXT: [[UADDE4:%[0-9]+]]:_(s32), [[UADDE5:%[0-9]+]]:_(s1) = G_UADDE [[UV13]], [[UV15]], [[UADDO5]]
; GFX8-NEXT: [[MV2:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO4]](s32), [[UADDE4]](s32)
+ ; GFX8-NEXT: [[COPY3:%[0-9]+]]:_(s64) = COPY [[MV2]](s64)
; GFX8-NEXT: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[MV2]](s64), [[UV1]]
; GFX8-NEXT: [[ICMP3:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[UV3]](s64), [[C]]
; GFX8-NEXT: [[XOR1:%[0-9]+]]:_(s1) = G_XOR [[ICMP3]], [[ICMP2]]
- ; GFX8-NEXT: [[ASHR1:%[0-9]+]]:_(s64) = G_ASHR [[MV2]], [[C1]](s32)
+ ; GFX8-NEXT: [[ASHR1:%[0-9]+]]:_(s64) = G_ASHR [[COPY3]], [[C1]](s32)
; GFX8-NEXT: [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[ASHR1]](s64)
; GFX8-NEXT: [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[C2]](s64)
; GFX8-NEXT: [[UADDO6:%[0-9]+]]:_(s32), [[UADDO7:%[0-9]+]]:_(s1) = G_UADDO [[UV16]], [[UV18]]
; GFX8-NEXT: [[UADDE6:%[0-9]+]]:_(s32), [[UADDE7:%[0-9]+]]:_(s1) = G_UADDE [[UV17]], [[UV19]], [[UADDO7]]
; GFX8-NEXT: [[MV3:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO6]](s32), [[UADDE6]](s32)
- ; GFX8-NEXT: [[SELECT1:%[0-9]+]]:_(s64) = G_SELECT [[XOR1]](s1), [[MV3]], [[MV2]]
+ ; GFX8-NEXT: [[SELECT1:%[0-9]+]]:_(s64) = G_SELECT [[XOR1]](s1), [[MV3]], [[COPY3]]
; GFX8-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[SELECT]](s64), [[SELECT1]](s64)
; GFX8-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
;
@@ -1125,34 +1132,36 @@ body: |
; GFX9-NEXT: [[UADDO:%[0-9]+]]:_(s32), [[UADDO1:%[0-9]+]]:_(s1) = G_UADDO [[UV4]], [[UV6]]
; GFX9-NEXT: [[UADDE:%[0-9]+]]:_(s32), [[UADDE1:%[0-9]+]]:_(s1) = G_UADDE [[UV5]], [[UV7]], [[UADDO1]]
; GFX9-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO]](s32), [[UADDE]](s32)
+ ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[MV]](s64)
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
; GFX9-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[MV]](s64), [[UV]]
; GFX9-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[UV2]](s64), [[C]]
; GFX9-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP1]], [[ICMP]]
; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 63
- ; GFX9-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[MV]], [[C1]](s32)
+ ; GFX9-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[COPY2]], [[C1]](s32)
; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 -9223372036854775808
; GFX9-NEXT: [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[ASHR]](s64)
; GFX9-NEXT: [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[C2]](s64)
; GFX9-NEXT: [[UADDO2:%[0-9]+]]:_(s32), [[UADDO3:%[0-9]+]]:_(s1) = G_UADDO [[UV8]], [[UV10]]
; GFX9-NEXT: [[UADDE2:%[0-9]+]]:_(s32), [[UADDE3:%[0-9]+]]:_(s1) = G_UADDE [[UV9]], [[UV11]], [[UADDO3]]
; GFX9-NEXT: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO2]](s32), [[UADDE2]](s32)
- ; GFX9-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[MV]]
+ ; GFX9-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[COPY2]]
; GFX9-NEXT: [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV1]](s64)
; GFX9-NEXT: [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV3]](s64)
; GFX9-NEXT: [[UADDO4:%[0-9]+]]:_(s32), [[UADDO5:%[0-9]+]]:_(s1) = G_UADDO [[UV12]], [[UV14]]
; GFX9-NEXT: [[UADDE4:%[0-9]+]]:_(s32), [[UADDE5:%[0-9]+]]:_(s1) = G_UADDE [[UV13]], [[UV15]], [[UADDO5]]
; GFX9-NEXT: [[MV2:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO4]](s32), [[UADDE4]](s32)
+ ; GFX9-NEXT: [[COPY3:%[0-9]+]]:_(s64) = COPY [[MV2]](s64)
; GFX9-NEXT: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[MV2]](s64), [[UV1]]
; GFX9-NEXT: [[ICMP3:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[UV3]](s64), [[C]]
; GFX9-NEXT: [[XOR1:%[0-9]+]]:_(s1) = G_XOR [[ICMP3]], [[ICMP2]]
- ; GFX9-NEXT: [[ASHR1:%[0-9]+]]:_(s64) = G_ASHR [[MV2]], [[C1]](s32)
+ ; GFX9-NEXT: [[ASHR1:%[0-9]+]]:_(s64) = G_ASHR [[COPY3]], [[C1]](s32)
; GFX9-NEXT: [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[ASHR1]](s64)
; GFX9-NEXT: [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[C2]](s64)
; GFX9-NEXT: [[UADDO6:%[0-9]+]]:_(s32), [[UADDO7:%[0-9]+]]:_(s1) = G_UADDO [[UV16]], [[UV18]]
; GFX9-NEXT: [[UADDE6:%[0-9]+]]:_(s32), [[UADDE7:%[0-9]+]]:_(s1) = G_UADDE [[UV17]], [[UV19]], [[UADDO7]]
; GFX9-NEXT: [[MV3:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO6]](s32), [[UADDE6]](s32)
- ; GFX9-NEXT: [[SELECT1:%[0-9]+]]:_(s64) = G_SELECT [[XOR1]](s1), [[MV3]], [[MV2]]
+ ; GFX9-NEXT: [[SELECT1:%[0-9]+]]:_(s64) = G_SELECT [[XOR1]](s1), [[MV3]], [[COPY3]]
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[SELECT]](s64), [[SELECT1]](s64)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
%0:_(<2 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ssubo.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ssubo.mir
index 57b1ab9b194ec5..7bc12dd6dd9769 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ssubo.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ssubo.mir
@@ -82,12 +82,13 @@ body: |
; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s32) = COPY $vgpr0
; CHECK-NEXT: [[COPY1:%[0-9]+]]:_(s32) = COPY $vgpr1
; CHECK-NEXT: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[COPY]], [[COPY1]]
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[SUB]](s32)
; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; CHECK-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SUB]](s32), [[COPY]]
; CHECK-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[COPY1]](s32), [[C]]
; CHECK-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP1]], [[ICMP]]
; CHECK-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[XOR]](s1)
- ; CHECK-NEXT: $vgpr0 = COPY [[SUB]](s32)
+ ; CHECK-NEXT: $vgpr0 = COPY [[COPY2]](s32)
; CHECK-NEXT: $vgpr1 = COPY [[ZEXT]](s32)
%0:_(s32) = COPY $vgpr0
%1:_(s32) = COPY $vgpr1
@@ -113,12 +114,13 @@ body: |
; CHECK-NEXT: [[USUBO:%[0-9]+]]:_(s32), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[UV]], [[UV2]]
; CHECK-NEXT: [[USUBE:%[0-9]+]]:_(s32), [[USUBE1:%[0-9]+]]:_(s1) = G_USUBE [[UV1]], [[UV3]], [[USUBO1]]
; CHECK-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO]](s32), [[USUBE]](s32)
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[MV]](s64)
; CHECK-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
; CHECK-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[MV]](s64), [[COPY]]
; CHECK-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[COPY1]](s64), [[C]]
; CHECK-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP1]], [[ICMP]]
; CHECK-NEXT: [[ZEXT:%[0-9]+]]:_(s32) = G_ZEXT [[XOR]](s1)
- ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[MV]](s64)
+ ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[COPY2]](s64)
; CHECK-NEXT: $vgpr2 = COPY [[ZEXT]](s32)
%0:_(s64) = COPY $vgpr0_vgpr1
%1:_(s64) = COPY $vgpr2_vgpr3
@@ -152,6 +154,7 @@ body: |
; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s32) = G_SHL [[AND1]], [[C]](s32)
; CHECK-NEXT: [[OR:%[0-9]+]]:_(s32) = G_OR [[AND]], [[SHL]]
; CHECK-NEXT: [[BITCAST2:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR]](s32)
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s16>) = COPY [[BITCAST2]](<2 x s16>)
; CHECK-NEXT: [[BITCAST3:%[0-9]+]]:_(s32) = G_BITCAST [[COPY]](<2 x s16>)
; CHECK-NEXT: [[LSHR2:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST3]], [[C]](s32)
; CHECK-NEXT: [[SEXT_INREG:%[0-9]+]]:_(s32) = G_SEXT_INREG [[SUB]], 16
@@ -164,8 +167,8 @@ body: |
; CHECK-NEXT: [[LSHR3:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST4]], [[C]](s32)
; CHECK-NEXT: [[SEXT_INREG4:%[0-9]+]]:_(s32) = G_SEXT_INREG [[BITCAST4]], 16
; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
- ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
- ; CHECK-NEXT: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[SEXT_INREG4]](s32), [[COPY2]]
+ ; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+ ; CHECK-NEXT: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[SEXT_INREG4]](s32), [[COPY3]]
; CHECK-NEXT: [[SEXT_INREG5:%[0-9]+]]:_(s32) = G_SEXT_INREG [[LSHR3]], 16
; CHECK-NEXT: [[ICMP3:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[SEXT_INREG5]](s32), [[C2]]
; CHECK-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP2]], [[ICMP]]
@@ -176,7 +179,7 @@ body: |
; CHECK-NEXT: [[AND2:%[0-9]+]]:_(s32) = G_AND [[ANYEXT]], [[C3]]
; CHECK-NEXT: [[AND3:%[0-9]+]]:_(s32) = G_AND [[ANYEXT1]], [[C3]]
; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[AND2]](s32), [[AND3]](s32)
- ; CHECK-NEXT: $vgpr0 = COPY [[BITCAST2]](<2 x s16>)
+ ; CHECK-NEXT: $vgpr0 = COPY [[COPY2]](<2 x s16>)
; CHECK-NEXT: $vgpr1_vgpr2 = COPY [[BUILD_VECTOR]](<2 x s32>)
%0:_(<2 x s16>) = COPY $vgpr0
%1:_(<2 x s16>) = COPY $vgpr1
@@ -318,6 +321,7 @@ body: |
; CHECK-NEXT: [[OR1:%[0-9]+]]:_(s32) = G_OR [[AND2]], [[SHL1]]
; CHECK-NEXT: [[BITCAST5:%[0-9]+]]:_(<2 x s16>) = G_BITCAST [[OR1]](s32)
; CHECK-NEXT: [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[BITCAST4]](<2 x s16>), [[BITCAST5]](<2 x s16>)
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(<4 x s16>) = COPY [[CONCAT_VECTORS]](<4 x s16>)
; CHECK-NEXT: [[UV4:%[0-9]+]]:_(<2 x s16>), [[UV5:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
; CHECK-NEXT: [[BITCAST6:%[0-9]+]]:_(s32) = G_BITCAST [[UV4]](<2 x s16>)
; CHECK-NEXT: [[LSHR4:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST6]], [[C]](s32)
@@ -342,14 +346,14 @@ body: |
; CHECK-NEXT: [[LSHR7:%[0-9]+]]:_(s32) = G_LSHR [[BITCAST9]], [[C]](s32)
; CHECK-NEXT: [[SEXT_INREG8:%[0-9]+]]:_(s32) = G_SEXT_INREG [[BITCAST8]], 16
; CHECK-NEXT: [[C2:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
- ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
- ; CHECK-NEXT: [[ICMP4:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[SEXT_INREG8]](s32), [[COPY2]]
- ; CHECK-NEXT: [[SEXT_INREG9:%[0-9]+]]:_(s32) = G_SEXT_INREG [[LSHR6]], 16
; CHECK-NEXT: [[COPY3:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
- ; CHECK-NEXT: [[ICMP5:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[SEXT_INREG9]](s32), [[COPY3]]
- ; CHECK-NEXT: [[SEXT_INREG10:%[0-9]+]]:_(s32) = G_SEXT_INREG [[BITCAST9]], 16
+ ; CHECK-NEXT: [[ICMP4:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[SEXT_INREG8]](s32), [[COPY3]]
+ ; CHECK-NEXT: [[SEXT_INREG9:%[0-9]+]]:_(s32) = G_SEXT_INREG [[LSHR6]], 16
; CHECK-NEXT: [[COPY4:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
- ; CHECK-NEXT: [[ICMP6:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[SEXT_INREG10]](s32), [[COPY4]]
+ ; CHECK-NEXT: [[ICMP5:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[SEXT_INREG9]](s32), [[COPY4]]
+ ; CHECK-NEXT: [[SEXT_INREG10:%[0-9]+]]:_(s32) = G_SEXT_INREG [[BITCAST9]], 16
+ ; CHECK-NEXT: [[COPY5:%[0-9]+]]:_(s32) = COPY [[C2]](s32)
+ ; CHECK-NEXT: [[ICMP6:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[SEXT_INREG10]](s32), [[COPY5]]
; CHECK-NEXT: [[SEXT_INREG11:%[0-9]+]]:_(s32) = G_SEXT_INREG [[LSHR7]], 16
; CHECK-NEXT: [[ICMP7:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[SEXT_INREG11]](s32), [[C2]]
; CHECK-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP4]], [[ICMP]]
@@ -366,7 +370,7 @@ body: |
; CHECK-NEXT: [[AND6:%[0-9]+]]:_(s32) = G_AND [[ANYEXT2]], [[C3]]
; CHECK-NEXT: [[AND7:%[0-9]+]]:_(s32) = G_AND [[ANYEXT3]], [[C3]]
; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<4 x s32>) = G_BUILD_VECTOR [[AND4]](s32), [[AND5]](s32), [[AND6]](s32), [[AND7]](s32)
- ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[CONCAT_VECTORS]](<4 x s16>)
+ ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[COPY2]](<4 x s16>)
; CHECK-NEXT: $vgpr2_vgpr3_vgpr4_vgpr5 = COPY [[BUILD_VECTOR]](<4 x s32>)
%0:_(<4 x s16>) = COPY $vgpr0_vgpr1
%1:_(<4 x s16>) = COPY $vgpr1_vgpr2
@@ -392,6 +396,7 @@ body: |
; CHECK-NEXT: [[SUB:%[0-9]+]]:_(s32) = G_SUB [[UV]], [[UV2]]
; CHECK-NEXT: [[SUB1:%[0-9]+]]:_(s32) = G_SUB [[UV1]], [[UV3]]
; CHECK-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[SUB]](s32), [[SUB1]](s32)
+ ; CHECK-NEXT: [[COPY2:%[0-9]+]]:_(<2 x s32>) = COPY [[BUILD_VECTOR]](<2 x s32>)
; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 0
; CHECK-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>)
; CHECK-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[SUB]](s32), [[UV4]]
@@ -407,7 +412,7 @@ body: |
; CHECK-NEXT: [[AND:%[0-9]+]]:_(s32) = G_AND [[ANYEXT]], [[C1]]
; CHECK-NEXT: [[AND1:%[0-9]+]]:_(s32) = G_AND [[ANYEXT1]], [[C1]]
; CHECK-NEXT: [[BUILD_VECTOR1:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[AND]](s32), [[AND1]](s32)
- ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[BUILD_VECTOR]](<2 x s32>)
+ ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[COPY2]](<2 x s32>)
; CHECK-NEXT: $vgpr2_vgpr3 = COPY [[BUILD_VECTOR1]](<2 x s32>)
%0:_(<2 x s32>) = COPY $vgpr0_vgpr1
%1:_(<2 x s32>) = COPY $vgpr2_vgpr3
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ssubsat.mir b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ssubsat.mir
index 33a8cda8e84b3e..033f98b6e3d36c 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ssubsat.mir
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-ssubsat.mir
@@ -951,19 +951,20 @@ body: |
; GFX6-NEXT: [[USUBO:%[0-9]+]]:_(s32), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[UV]], [[UV2]]
; GFX6-NEXT: [[USUBE:%[0-9]+]]:_(s32), [[USUBE1:%[0-9]+]]:_(s1) = G_USUBE [[UV1]], [[UV3]], [[USUBO1]]
; GFX6-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO]](s32), [[USUBE]](s32)
+ ; GFX6-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[MV]](s64)
; GFX6-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
; GFX6-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[MV]](s64), [[COPY]]
; GFX6-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[COPY1]](s64), [[C]]
; GFX6-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP1]], [[ICMP]]
; GFX6-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 63
- ; GFX6-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[MV]], [[C1]](s32)
+ ; GFX6-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[COPY2]], [[C1]](s32)
; GFX6-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 -9223372036854775808
; GFX6-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[ASHR]](s64)
; GFX6-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[C2]](s64)
; GFX6-NEXT: [[UADDO:%[0-9]+]]:_(s32), [[UADDO1:%[0-9]+]]:_(s1) = G_UADDO [[UV4]], [[UV6]]
; GFX6-NEXT: [[UADDE:%[0-9]+]]:_(s32), [[UADDE1:%[0-9]+]]:_(s1) = G_UADDE [[UV5]], [[UV7]], [[UADDO1]]
; GFX6-NEXT: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO]](s32), [[UADDE]](s32)
- ; GFX6-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[MV]]
+ ; GFX6-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[COPY2]]
; GFX6-NEXT: $vgpr0_vgpr1 = COPY [[SELECT]](s64)
;
; GFX8-LABEL: name: ssubsat_s64
@@ -976,19 +977,20 @@ body: |
; GFX8-NEXT: [[USUBO:%[0-9]+]]:_(s32), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[UV]], [[UV2]]
; GFX8-NEXT: [[USUBE:%[0-9]+]]:_(s32), [[USUBE1:%[0-9]+]]:_(s1) = G_USUBE [[UV1]], [[UV3]], [[USUBO1]]
; GFX8-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO]](s32), [[USUBE]](s32)
+ ; GFX8-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[MV]](s64)
; GFX8-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
; GFX8-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[MV]](s64), [[COPY]]
; GFX8-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[COPY1]](s64), [[C]]
; GFX8-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP1]], [[ICMP]]
; GFX8-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 63
- ; GFX8-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[MV]], [[C1]](s32)
+ ; GFX8-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[COPY2]], [[C1]](s32)
; GFX8-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 -9223372036854775808
; GFX8-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[ASHR]](s64)
; GFX8-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[C2]](s64)
; GFX8-NEXT: [[UADDO:%[0-9]+]]:_(s32), [[UADDO1:%[0-9]+]]:_(s1) = G_UADDO [[UV4]], [[UV6]]
; GFX8-NEXT: [[UADDE:%[0-9]+]]:_(s32), [[UADDE1:%[0-9]+]]:_(s1) = G_UADDE [[UV5]], [[UV7]], [[UADDO1]]
; GFX8-NEXT: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO]](s32), [[UADDE]](s32)
- ; GFX8-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[MV]]
+ ; GFX8-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[COPY2]]
; GFX8-NEXT: $vgpr0_vgpr1 = COPY [[SELECT]](s64)
;
; GFX9-LABEL: name: ssubsat_s64
@@ -1001,19 +1003,20 @@ body: |
; GFX9-NEXT: [[USUBO:%[0-9]+]]:_(s32), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[UV]], [[UV2]]
; GFX9-NEXT: [[USUBE:%[0-9]+]]:_(s32), [[USUBE1:%[0-9]+]]:_(s1) = G_USUBE [[UV1]], [[UV3]], [[USUBO1]]
; GFX9-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO]](s32), [[USUBE]](s32)
+ ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[MV]](s64)
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
; GFX9-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[MV]](s64), [[COPY]]
; GFX9-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[COPY1]](s64), [[C]]
; GFX9-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP1]], [[ICMP]]
; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 63
- ; GFX9-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[MV]], [[C1]](s32)
+ ; GFX9-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[COPY2]], [[C1]](s32)
; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 -9223372036854775808
; GFX9-NEXT: [[UV4:%[0-9]+]]:_(s32), [[UV5:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[ASHR]](s64)
; GFX9-NEXT: [[UV6:%[0-9]+]]:_(s32), [[UV7:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[C2]](s64)
; GFX9-NEXT: [[UADDO:%[0-9]+]]:_(s32), [[UADDO1:%[0-9]+]]:_(s1) = G_UADDO [[UV4]], [[UV6]]
; GFX9-NEXT: [[UADDE:%[0-9]+]]:_(s32), [[UADDE1:%[0-9]+]]:_(s1) = G_UADDE [[UV5]], [[UV7]], [[UADDO1]]
; GFX9-NEXT: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO]](s32), [[UADDE]](s32)
- ; GFX9-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[MV]]
+ ; GFX9-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[COPY2]]
; GFX9-NEXT: $vgpr0_vgpr1 = COPY [[SELECT]](s64)
%0:_(s64) = COPY $vgpr0_vgpr1
%1:_(s64) = COPY $vgpr2_vgpr3
@@ -1039,34 +1042,36 @@ body: |
; GFX6-NEXT: [[USUBO:%[0-9]+]]:_(s32), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[UV4]], [[UV6]]
; GFX6-NEXT: [[USUBE:%[0-9]+]]:_(s32), [[USUBE1:%[0-9]+]]:_(s1) = G_USUBE [[UV5]], [[UV7]], [[USUBO1]]
; GFX6-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO]](s32), [[USUBE]](s32)
+ ; GFX6-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[MV]](s64)
; GFX6-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
; GFX6-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[MV]](s64), [[UV]]
; GFX6-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[UV2]](s64), [[C]]
; GFX6-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP1]], [[ICMP]]
; GFX6-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 63
- ; GFX6-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[MV]], [[C1]](s32)
+ ; GFX6-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[COPY2]], [[C1]](s32)
; GFX6-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 -9223372036854775808
; GFX6-NEXT: [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[ASHR]](s64)
; GFX6-NEXT: [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[C2]](s64)
; GFX6-NEXT: [[UADDO:%[0-9]+]]:_(s32), [[UADDO1:%[0-9]+]]:_(s1) = G_UADDO [[UV8]], [[UV10]]
; GFX6-NEXT: [[UADDE:%[0-9]+]]:_(s32), [[UADDE1:%[0-9]+]]:_(s1) = G_UADDE [[UV9]], [[UV11]], [[UADDO1]]
; GFX6-NEXT: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO]](s32), [[UADDE]](s32)
- ; GFX6-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[MV]]
+ ; GFX6-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[COPY2]]
; GFX6-NEXT: [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV1]](s64)
; GFX6-NEXT: [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV3]](s64)
; GFX6-NEXT: [[USUBO2:%[0-9]+]]:_(s32), [[USUBO3:%[0-9]+]]:_(s1) = G_USUBO [[UV12]], [[UV14]]
; GFX6-NEXT: [[USUBE2:%[0-9]+]]:_(s32), [[USUBE3:%[0-9]+]]:_(s1) = G_USUBE [[UV13]], [[UV15]], [[USUBO3]]
; GFX6-NEXT: [[MV2:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO2]](s32), [[USUBE2]](s32)
+ ; GFX6-NEXT: [[COPY3:%[0-9]+]]:_(s64) = COPY [[MV2]](s64)
; GFX6-NEXT: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[MV2]](s64), [[UV1]]
; GFX6-NEXT: [[ICMP3:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[UV3]](s64), [[C]]
; GFX6-NEXT: [[XOR1:%[0-9]+]]:_(s1) = G_XOR [[ICMP3]], [[ICMP2]]
- ; GFX6-NEXT: [[ASHR1:%[0-9]+]]:_(s64) = G_ASHR [[MV2]], [[C1]](s32)
+ ; GFX6-NEXT: [[ASHR1:%[0-9]+]]:_(s64) = G_ASHR [[COPY3]], [[C1]](s32)
; GFX6-NEXT: [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[ASHR1]](s64)
; GFX6-NEXT: [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[C2]](s64)
; GFX6-NEXT: [[UADDO2:%[0-9]+]]:_(s32), [[UADDO3:%[0-9]+]]:_(s1) = G_UADDO [[UV16]], [[UV18]]
; GFX6-NEXT: [[UADDE2:%[0-9]+]]:_(s32), [[UADDE3:%[0-9]+]]:_(s1) = G_UADDE [[UV17]], [[UV19]], [[UADDO3]]
; GFX6-NEXT: [[MV3:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO2]](s32), [[UADDE2]](s32)
- ; GFX6-NEXT: [[SELECT1:%[0-9]+]]:_(s64) = G_SELECT [[XOR1]](s1), [[MV3]], [[MV2]]
+ ; GFX6-NEXT: [[SELECT1:%[0-9]+]]:_(s64) = G_SELECT [[XOR1]](s1), [[MV3]], [[COPY3]]
; GFX6-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[SELECT]](s64), [[SELECT1]](s64)
; GFX6-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
;
@@ -1082,34 +1087,36 @@ body: |
; GFX8-NEXT: [[USUBO:%[0-9]+]]:_(s32), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[UV4]], [[UV6]]
; GFX8-NEXT: [[USUBE:%[0-9]+]]:_(s32), [[USUBE1:%[0-9]+]]:_(s1) = G_USUBE [[UV5]], [[UV7]], [[USUBO1]]
; GFX8-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO]](s32), [[USUBE]](s32)
+ ; GFX8-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[MV]](s64)
; GFX8-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
; GFX8-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[MV]](s64), [[UV]]
; GFX8-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[UV2]](s64), [[C]]
; GFX8-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP1]], [[ICMP]]
; GFX8-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 63
- ; GFX8-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[MV]], [[C1]](s32)
+ ; GFX8-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[COPY2]], [[C1]](s32)
; GFX8-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 -9223372036854775808
; GFX8-NEXT: [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[ASHR]](s64)
; GFX8-NEXT: [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[C2]](s64)
; GFX8-NEXT: [[UADDO:%[0-9]+]]:_(s32), [[UADDO1:%[0-9]+]]:_(s1) = G_UADDO [[UV8]], [[UV10]]
; GFX8-NEXT: [[UADDE:%[0-9]+]]:_(s32), [[UADDE1:%[0-9]+]]:_(s1) = G_UADDE [[UV9]], [[UV11]], [[UADDO1]]
; GFX8-NEXT: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO]](s32), [[UADDE]](s32)
- ; GFX8-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[MV]]
+ ; GFX8-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[COPY2]]
; GFX8-NEXT: [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV1]](s64)
; GFX8-NEXT: [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV3]](s64)
; GFX8-NEXT: [[USUBO2:%[0-9]+]]:_(s32), [[USUBO3:%[0-9]+]]:_(s1) = G_USUBO [[UV12]], [[UV14]]
; GFX8-NEXT: [[USUBE2:%[0-9]+]]:_(s32), [[USUBE3:%[0-9]+]]:_(s1) = G_USUBE [[UV13]], [[UV15]], [[USUBO3]]
; GFX8-NEXT: [[MV2:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO2]](s32), [[USUBE2]](s32)
+ ; GFX8-NEXT: [[COPY3:%[0-9]+]]:_(s64) = COPY [[MV2]](s64)
; GFX8-NEXT: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[MV2]](s64), [[UV1]]
; GFX8-NEXT: [[ICMP3:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[UV3]](s64), [[C]]
; GFX8-NEXT: [[XOR1:%[0-9]+]]:_(s1) = G_XOR [[ICMP3]], [[ICMP2]]
- ; GFX8-NEXT: [[ASHR1:%[0-9]+]]:_(s64) = G_ASHR [[MV2]], [[C1]](s32)
+ ; GFX8-NEXT: [[ASHR1:%[0-9]+]]:_(s64) = G_ASHR [[COPY3]], [[C1]](s32)
; GFX8-NEXT: [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[ASHR1]](s64)
; GFX8-NEXT: [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[C2]](s64)
; GFX8-NEXT: [[UADDO2:%[0-9]+]]:_(s32), [[UADDO3:%[0-9]+]]:_(s1) = G_UADDO [[UV16]], [[UV18]]
; GFX8-NEXT: [[UADDE2:%[0-9]+]]:_(s32), [[UADDE3:%[0-9]+]]:_(s1) = G_UADDE [[UV17]], [[UV19]], [[UADDO3]]
; GFX8-NEXT: [[MV3:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO2]](s32), [[UADDE2]](s32)
- ; GFX8-NEXT: [[SELECT1:%[0-9]+]]:_(s64) = G_SELECT [[XOR1]](s1), [[MV3]], [[MV2]]
+ ; GFX8-NEXT: [[SELECT1:%[0-9]+]]:_(s64) = G_SELECT [[XOR1]](s1), [[MV3]], [[COPY3]]
; GFX8-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[SELECT]](s64), [[SELECT1]](s64)
; GFX8-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
;
@@ -1125,34 +1132,36 @@ body: |
; GFX9-NEXT: [[USUBO:%[0-9]+]]:_(s32), [[USUBO1:%[0-9]+]]:_(s1) = G_USUBO [[UV4]], [[UV6]]
; GFX9-NEXT: [[USUBE:%[0-9]+]]:_(s32), [[USUBE1:%[0-9]+]]:_(s1) = G_USUBE [[UV5]], [[UV7]], [[USUBO1]]
; GFX9-NEXT: [[MV:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO]](s32), [[USUBE]](s32)
+ ; GFX9-NEXT: [[COPY2:%[0-9]+]]:_(s64) = COPY [[MV]](s64)
; GFX9-NEXT: [[C:%[0-9]+]]:_(s64) = G_CONSTANT i64 0
; GFX9-NEXT: [[ICMP:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[MV]](s64), [[UV]]
; GFX9-NEXT: [[ICMP1:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[UV2]](s64), [[C]]
; GFX9-NEXT: [[XOR:%[0-9]+]]:_(s1) = G_XOR [[ICMP1]], [[ICMP]]
; GFX9-NEXT: [[C1:%[0-9]+]]:_(s32) = G_CONSTANT i32 63
- ; GFX9-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[MV]], [[C1]](s32)
+ ; GFX9-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[COPY2]], [[C1]](s32)
; GFX9-NEXT: [[C2:%[0-9]+]]:_(s64) = G_CONSTANT i64 -9223372036854775808
; GFX9-NEXT: [[UV8:%[0-9]+]]:_(s32), [[UV9:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[ASHR]](s64)
; GFX9-NEXT: [[UV10:%[0-9]+]]:_(s32), [[UV11:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[C2]](s64)
; GFX9-NEXT: [[UADDO:%[0-9]+]]:_(s32), [[UADDO1:%[0-9]+]]:_(s1) = G_UADDO [[UV8]], [[UV10]]
; GFX9-NEXT: [[UADDE:%[0-9]+]]:_(s32), [[UADDE1:%[0-9]+]]:_(s1) = G_UADDE [[UV9]], [[UV11]], [[UADDO1]]
; GFX9-NEXT: [[MV1:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO]](s32), [[UADDE]](s32)
- ; GFX9-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[MV]]
+ ; GFX9-NEXT: [[SELECT:%[0-9]+]]:_(s64) = G_SELECT [[XOR]](s1), [[MV1]], [[COPY2]]
; GFX9-NEXT: [[UV12:%[0-9]+]]:_(s32), [[UV13:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV1]](s64)
; GFX9-NEXT: [[UV14:%[0-9]+]]:_(s32), [[UV15:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[UV3]](s64)
; GFX9-NEXT: [[USUBO2:%[0-9]+]]:_(s32), [[USUBO3:%[0-9]+]]:_(s1) = G_USUBO [[UV12]], [[UV14]]
; GFX9-NEXT: [[USUBE2:%[0-9]+]]:_(s32), [[USUBE3:%[0-9]+]]:_(s1) = G_USUBE [[UV13]], [[UV15]], [[USUBO3]]
; GFX9-NEXT: [[MV2:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[USUBO2]](s32), [[USUBE2]](s32)
+ ; GFX9-NEXT: [[COPY3:%[0-9]+]]:_(s64) = COPY [[MV2]](s64)
; GFX9-NEXT: [[ICMP2:%[0-9]+]]:_(s1) = G_ICMP intpred(slt), [[MV2]](s64), [[UV1]]
; GFX9-NEXT: [[ICMP3:%[0-9]+]]:_(s1) = G_ICMP intpred(sgt), [[UV3]](s64), [[C]]
; GFX9-NEXT: [[XOR1:%[0-9]+]]:_(s1) = G_XOR [[ICMP3]], [[ICMP2]]
- ; GFX9-NEXT: [[ASHR1:%[0-9]+]]:_(s64) = G_ASHR [[MV2]], [[C1]](s32)
+ ; GFX9-NEXT: [[ASHR1:%[0-9]+]]:_(s64) = G_ASHR [[COPY3]], [[C1]](s32)
; GFX9-NEXT: [[UV16:%[0-9]+]]:_(s32), [[UV17:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[ASHR1]](s64)
; GFX9-NEXT: [[UV18:%[0-9]+]]:_(s32), [[UV19:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[C2]](s64)
; GFX9-NEXT: [[UADDO2:%[0-9]+]]:_(s32), [[UADDO3:%[0-9]+]]:_(s1) = G_UADDO [[UV16]], [[UV18]]
; GFX9-NEXT: [[UADDE2:%[0-9]+]]:_(s32), [[UADDE3:%[0-9]+]]:_(s1) = G_UADDE [[UV17]], [[UV19]], [[UADDO3]]
; GFX9-NEXT: [[MV3:%[0-9]+]]:_(s64) = G_MERGE_VALUES [[UADDO2]](s32), [[UADDE2]](s32)
- ; GFX9-NEXT: [[SELECT1:%[0-9]+]]:_(s64) = G_SELECT [[XOR1]](s1), [[MV3]], [[MV2]]
+ ; GFX9-NEXT: [[SELECT1:%[0-9]+]]:_(s64) = G_SELECT [[XOR1]](s1), [[MV3]], [[COPY3]]
; GFX9-NEXT: [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s64>) = G_BUILD_VECTOR [[SELECT]](s64), [[SELECT1]](s64)
; GFX9-NEXT: $vgpr0_vgpr1_vgpr2_vgpr3 = COPY [[BUILD_VECTOR]](<2 x s64>)
%0:_(<2 x s64>) = COPY $vgpr0_vgpr1_vgpr2_vgpr3
More information about the llvm-commits
mailing list