maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

GlobalISel: Fix narrowing of G_CTPOP 

The result type is separate from the source type. Tests will be
included in a future AMDGPU patch which uses this from
RegBankSelect/applyMappingImpl.
This commit is contained in:
Matt Arsenault 2020-02-06 22:29:23 -05:00
parent 8de2dad9e0
commit 3b198518ad
2 changed files with 6 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
llvm/lib/CodeGen/GlobalISel/LegalizerHelper.cpp
View File

@ -4040,16 +4040,17 @@ LegalizerHelper::narrowScalarCTPOP(MachineInstr &MI, unsigned TypeIdx,
if (TypeIdx != 1)
return UnableToLegalize;
Register DstReg = MI.getOperand(0).getReg();
LLT DstTy = MRI.getType(DstReg);
LLT SrcTy = MRI.getType(MI.getOperand(1).getReg());
unsigned NarrowSize = NarrowTy.getSizeInBits();
if (SrcTy.isScalar() && SrcTy.getSizeInBits() == 2 * NarrowSize) {
auto UnmergeSrc = MIRBuilder.buildUnmerge(NarrowTy, MI.getOperand(1));
auto LoCTPOP = MIRBuilder.buildCTPOP(NarrowTy, UnmergeSrc.getReg(0));
auto HiCTPOP = MIRBuilder.buildCTPOP(NarrowTy, UnmergeSrc.getReg(1));
auto Out = MIRBuilder.buildAdd(NarrowTy, HiCTPOP, LoCTPOP);
MIRBuilder.buildZExt(MI.getOperand(0), Out);
auto LoCTPOP = MIRBuilder.buildCTPOP(DstTy, UnmergeSrc.getReg(0));
auto HiCTPOP = MIRBuilder.buildCTPOP(DstTy, UnmergeSrc.getReg(1));
MIRBuilder.buildAdd(DstReg, HiCTPOP, LoCTPOP);
MI.eraseFromParent();
return Legalized;

1
llvm/lib/Target/Mips/MipsLegalizerInfo.cpp
View File

@ -216,6 +216,7 @@ MipsLegalizerInfo::MipsLegalizerInfo(const MipsSubtarget &ST) {
getActionDefinitionsBuilder(G_CTPOP)
.lowerFor({{s32, s32}})
.clampScalar(0, s32, s32)
.clampScalar(1, s32, s32);
// FP instructions