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

In my recent change to avoid use of underaligned memory I didn't notice that 

cpyDest can be mutated in some cases, which would then cause a crash later if
indeed the memory was underaligned.  This brought down several buildbots, so
I guess the underaligned case is much more common than I thought!

llvm-svn: 165228
This commit is contained in:
Duncan Sands 2012-10-04 13:53:21 +00:00
parent d2f6b97417
commit a6d20010fe
2 changed files with 9 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp
View File

@ -692,15 +692,15 @@ bool MemCpyOpt::performCallSlotOptzn(Instruction *cpy,
bool changedArgument = false;
for (unsigned i = 0; i < CS.arg_size(); ++i)
if (CS.getArgument(i)->stripPointerCasts() == cpySrc) {
if (cpySrc->getType() != cpyDest->getType())
cpyDest = CastInst::CreatePointerCast(cpyDest, cpySrc->getType(),
cpyDest->getName(), C);
Value *Dest = cpySrc->getType() == cpyDest->getType() ? cpyDest
: CastInst::CreatePointerCast(cpyDest, cpySrc->getType(),
cpyDest->getName(), C);
changedArgument = true;
if (CS.getArgument(i)->getType() == cpyDest->getType())
CS.setArgument(i, cpyDest);
if (CS.getArgument(i)->getType() == Dest->getType())
CS.setArgument(i, Dest);
else
CS.setArgument(i, CastInst::CreatePointerCast(cpyDest,
CS.getArgument(i)->getType(), cpyDest->getName(), C));
CS.setArgument(i, CastInst::CreatePointerCast(Dest,
CS.getArgument(i)->getType(), Dest->getName(), C));
}
if (!changedArgument)

4
llvm/test/Transforms/MemCpyOpt/2008-03-13-ReturnSlotBitcast.ll
View File

@ -9,11 +9,11 @@ declare void @g(%a*)
define float @f() {
entry:
%a_var = alloca %a
%b_var = alloca %b
%b_var = alloca %b, align 1
call void @g(%a* %a_var)
%a_i8 = bitcast %a* %a_var to i8*
%b_i8 = bitcast %b* %b_var to i8*
call void @llvm.memcpy.p0i8.p0i8.i32(i8* %b_i8, i8* %a_i8, i32 4, i32 4, i1 false)
call void @llvm.memcpy.p0i8.p0i8.i32(i8* %b_i8, i8* %a_i8, i32 4, i32 1, i1 false)
%tmp1 = getelementptr %b* %b_var, i32 0, i32 0
%tmp2 = load float* %tmp1
ret float %tmp2