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

Fix to handle all non-power-of-2 vector sizes in the mask form of _builtin_shuffle_vector. 

Previously a 2-bit mask was used to mask each element of a vec6 mask before doing the extracts instead of 3-bit mask necessary to cover 0-5. vec3 was the only non-power-of-2 that worked correctly because a +1 conditionally added before calculating floor(log2(elements)).

llvm-svn: 187560
This commit is contained in:
Craig Topper 2013-08-01 06:42:40 +00:00
parent a4f7c7a600
commit b9b7ea697b
1 changed files with 2 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
clang/lib/CodeGen/CGExprScalar.cpp
View File

@ -922,16 +922,8 @@ Value *ScalarExprEmitter::VisitShuffleVectorExpr(ShuffleVectorExpr *E) {
llvm::VectorType *MTy = cast<llvm::VectorType>(Mask->getType());
llvm::Constant* EltMask;
// Treat vec3 like vec4.
if ((LHSElts == 6) && (E->getNumSubExprs() == 3))
EltMask = llvm::ConstantInt::get(MTy->getElementType(),
(1 << llvm::Log2_32(LHSElts+2))-1);
else if ((LHSElts == 3) && (E->getNumSubExprs() == 2))
EltMask = llvm::ConstantInt::get(MTy->getElementType(),
(1 << llvm::Log2_32(LHSElts+1))-1);
else
EltMask = llvm::ConstantInt::get(MTy->getElementType(),
(1 << llvm::Log2_32(LHSElts))-1);
EltMask = llvm::ConstantInt::get(MTy->getElementType(),
llvm::NextPowerOf2(LHSElts-1)-1);
// Mask off the high bits of each shuffle index.
Value *MaskBits = llvm::ConstantVector::getSplat(MTy->getNumElements(),