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

[X86][SSE] Minor regression fix for r225551 

r225551 vector byte shuffle optimization caused an assertion as fully zeroable vectors can be produced under certain circumstances. This fix drops the assert and returns a zero vector where the assert would have failed.

llvm-svn: 225718
This commit is contained in:
Simon Pilgrim 2015-01-12 22:38:08 +00:00
parent 0e6ffb9d0d
commit d88ab87064
1 changed files with 2 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
llvm/lib/Target/X86/X86ISelLowering.cpp
View File

@ -9652,7 +9652,6 @@ static SDValue lowerV16I8VectorShuffle(SDValue Op, SDValue V1, SDValue V2,
V2InUse |= (ZeroMask != V2Idx);
}
}
assert((V1InUse || V2InUse) && "Shuffling to a zeroable vector");
if (V1InUse)
V1 = DAG.getNode(X86ISD::PSHUFB, DL, MVT::v16i8, V1,
@ -9668,6 +9667,8 @@ static SDValue lowerV16I8VectorShuffle(SDValue Op, SDValue V1, SDValue V2,
return V1; // Single inputs are easy.
if (V2InUse)
return V2; // Single inputs are easy.
// Shuffling to a zeroable vector.
return getZeroVector(MVT::v16i8, Subtarget, DAG, DL);
}
// There are special ways we can lower some single-element blends.