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

If both operands are loads from stores in memory we can't use movlpd/movlps 

since one needs to be a register operand. Just use movss instead of forcing
an operand into a register.

Fixes PR9239

llvm-svn: 126072
This commit is contained in:
Eric Christopher 2011-02-20 05:04:42 +00:00
parent a32b8d7801
commit ac6b001f56
1 changed files with 4 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
llvm/lib/Target/X86/X86ISelLowering.cpp
View File

@ -5472,6 +5472,10 @@ SDValue getMOVLP(SDValue &Op, DebugLoc &dl, SelectionDAG &DAG, bool HasSSE2) {
if (MayFoldVectorLoad(V1) && MayFoldIntoStore(Op))
CanFoldLoad = true;
// Both of them can't be memory operations though.
if (MayFoldVectorLoad(V1) && MayFoldVectorLoad(V2))
CanFoldLoad = false;
if (CanFoldLoad) {
if (HasSSE2 && NumElems == 2)
return getTargetShuffleNode(X86ISD::MOVLPD, dl, VT, V1, V2, DAG);