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

Implement a trivial form of store->load forwarding where the store and the 

load are exactly consequtive.  This is picked up by other passes, but this
triggers thousands of times in fortran programs that use static locals
(and is thus a compile-time speedup).

llvm-svn: 23320
This commit is contained in:
Chris Lattner 2005-09-12 22:00:15 +00:00
parent 4cd474ebbd
commit b990f7d8ed
1 changed files with 9 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
llvm/lib/Transforms/Scalar/InstructionCombining.cpp
View File

@ -4926,6 +4926,15 @@ Instruction *InstCombiner::visitLoadInst(LoadInst &LI) {
// None of the following transforms are legal for volatile loads.
if (LI.isVolatile()) return 0;
// If the instruction immediately before this is a store to the same address,
// do a simple form of store->load forwarding.
if (&LI.getParent()->front() != &LI) {
BasicBlock::iterator BBI = &LI; --BBI;
if (StoreInst *SI = dyn_cast<StoreInst>(BBI))
if (SI->getOperand(1) == LI.getOperand(0))
return ReplaceInstUsesWith(LI, SI->getOperand(0));
}
if (GetElementPtrInst *GEPI = dyn_cast<GetElementPtrInst>(Op))
if (isa<ConstantPointerNull>(GEPI->getOperand(0)) ||
isa<UndefValue>(GEPI->getOperand(0))) {