Fix two issues that Eli Friedman pointed out, where would misoptimized code like:

char a[200];
init(a, a+200);

OR

int a[200];
char* b = (char*)a;
char* c = (char*)a;
foo(b, c);


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51850 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Owen Anderson 2008-06-01 22:26:26 +00:00
parent 19d4d12ee0
commit 009e4f7609
2 changed files with 15 additions and 4 deletions

View File

@ -554,10 +554,17 @@ bool MemCpyOpt::performCallSlotOptzn(MemCpyInst *cpy, CallInst *C) {
User* UI = srcUseList.back();
srcUseList.pop_back();
if (isa<GetElementPtrInst>(UI) || isa<BitCastInst>(UI)) {
if (isa<BitCastInst>(UI)) {
for (User::use_iterator I = UI->use_begin(), E = UI->use_end();
I != E; ++I)
srcUseList.push_back(*I);
} else if (GetElementPtrInst* G = dyn_cast<GetElementPtrInst>(UI)) {
if (G->hasAllZeroIndices())
for (User::use_iterator I = UI->use_begin(), E = UI->use_end();
I != E; ++I)
srcUseList.push_back(*I);
else
return false;
} else if (UI != C && UI != cpy) {
return false;
}
@ -582,12 +589,16 @@ bool MemCpyOpt::performCallSlotOptzn(MemCpyInst *cpy, CallInst *C) {
// All the checks have passed, so do the transformation.
bool changedArgument = false;
for (unsigned i = 0; i < CS.arg_size(); ++i)
if (CS.getArgument(i) == cpySrc) {
if (CS.getArgument(i)->stripPointerCasts() == cpySrc) {
if (cpySrc->getType() != cpyDest->getType())
cpyDest = CastInst::CreatePointerCast(cpyDest, cpySrc->getType(),
cpyDest->getName(), C);
changedArgument = true;
CS.setArgument(i, cpyDest);
if (CS.getArgument(i)->getType() != cpyDest->getType())
CS.setArgument(i, CastInst::CreatePointerCast(cpyDest,
CS.getArgument(i)->getType(), cpyDest->getName(), C));
else
CS.setArgument(i, cpyDest);
}
if (!changedArgument)

View File

@ -1,4 +1,4 @@
; RUN: llvm-as < %s | opt -memcpyopt | llvm-dis | grep {call.*memcpy.*}
; RUN: llvm-as < %s | opt -memcpyopt | llvm-dis | grep {call.*memmove.*arg1.*}
; PR2401
target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:32:32"