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

CodeGenPrepare: Reorder check for cold and shouldOptimizeForSize 

shouldOptimizeForSize is showing up in a profile, spending around 10%
of the pass time in one function. This should probably not be so slow,
but the much cheaper attribute check should be done first anyway.
This commit is contained in:
Matt Arsenault 2020-01-31 14:35:53 -05:00
parent 3c7e9c34b3
commit 23b76096b7
1 changed files with 9 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
llvm/lib/CodeGen/CodeGenPrepare.cpp
View File

@ -1937,8 +1937,8 @@ bool CodeGenPrepare::optimizeCallInst(CallInst *CI, bool &ModifiedDT) {
// cold block. This interacts with our handling for loads and stores to
// ensure that we can fold all uses of a potential addressing computation
// into their uses. TODO: generalize this to work over profiling data
bool OptForSize = OptSize || llvm::shouldOptimizeForSize(BB, PSI, BFI.get());
if (!OptForSize && CI->hasFnAttr(Attribute::Cold))
if (CI->hasFnAttr(Attribute::Cold) &&
!OptSize && !llvm::shouldOptimizeForSize(BB, PSI, BFI.get()))
for (auto &Arg : CI->arg_operands()) {
if (!Arg->getType()->isPointerTy())
continue;
@ -4587,12 +4587,14 @@ static bool FindAllMemoryUses(
}
if (CallInst *CI = dyn_cast<CallInst>(UserI)) {
if (CI->hasFnAttr(Attribute::Cold)) {
// If this is a cold call, we can sink the addressing calculation into
// the cold path. See optimizeCallInst
bool OptForSize = OptSize ||
llvm::shouldOptimizeForSize(CI->getParent(), PSI, BFI);
if (!OptForSize && CI->hasFnAttr(Attribute::Cold))
if (!OptForSize)
continue;
}
InlineAsm *IA = dyn_cast<InlineAsm>(CI->getCalledValue());
if (!IA) return true;