RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2025-01-31 20:51:52 +01:00
Code Issues Projects Releases Wiki Activity

[X86] Do something sensible when an expand load intrinsic is passed a 0 mask.

Browse Source 
Previously we just returned undef, but really we should be returning the pass thru input. We also need to make sure we preserve the chain output that the original intrinsic node had to maintain connectivity in the DAG. So we should just return the incoming chain as the output chain.

llvm-svn: 333804
This commit is contained in:
Craig Topper 2018-06-01 22:59:07 +00:00
parent ad60f90629
commit a31261c52e
2 changed files with 10 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
lib/Target/X86/X86ISelLowering.cpp
View File

@ -21641,7 +21641,7 @@ static SDValue LowerINTRINSIC_W_CHAIN(SDValue Op, const X86Subtarget &Subtarget,
if (isAllOnesConstant(Mask)) // Return a regular (unmasked) vector load.
return DAG.getLoad(VT, dl, Chain, Addr, MemIntr->getMemOperand());
if (X86::isZeroNode(Mask))
return DAG.getUNDEF(VT);
return DAG.getMergeValues({PassThru, Chain}, dl);
MVT MaskVT = MVT::getVectorVT(MVT::i1, VT.getVectorNumElements());
SDValue VMask = getMaskNode(Mask, MaskVT, Subtarget, DAG, dl);

9
test/CodeGen/X86/avx512-intrinsics.ll
View File

@ -280,6 +280,15 @@ define <8 x double> @test_expand_load_pd_512(i8* %addr, <8 x double> %data) {
ret <8 x double> %res
}
; Make sure we don't crash if you pass 0 to the mask.
define <8 x double> @test_zero_mask_expand_load_pd_512(i8* %addr, <8 x double> %data, i8 %mask) {
; CHECK-LABEL: test_zero_mask_expand_load_pd_512:
; CHECK: ## %bb.0:
; CHECK-NEXT: retq
%res = call <8 x double> @llvm.x86.avx512.mask.expand.load.pd.512(i8* %addr, <8 x double> %data, i8 0)
ret <8 x double> %res
}
define <16 x float> @test_mask_expand_load_ps_512(i8* %addr, <16 x float> %data, i16 %mask) {
; CHECK-LABEL: test_mask_expand_load_ps_512:
; CHECK: ## %bb.0: