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

Generate fewer first class aggregate values for other 

coerce cases (e.g. {double,int}) which avoids fastisel
bailing out at -O0.

llvm-svn: 107628
This commit is contained in:
Chris Lattner 2010-07-05 20:41:41 +00:00
parent c401de9998
commit ceddafb846
1 changed files with 13 additions and 35 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

48
clang/lib/CodeGen/CGCall.cpp
View File

@ -933,25 +933,14 @@ void CodeGenFunction::EmitFunctionProlog(const CGFunctionInfo &FI,
// and the optimizer generally likes scalar values better than FCAs. // and the optimizer generally likes scalar values better than FCAs.
if (const llvm::StructType *STy = if (const llvm::StructType *STy =
dyn_cast<llvm::StructType>(ArgI.getCoerceToType())) { dyn_cast<llvm::StructType>(ArgI.getCoerceToType())) {
// If the argument and alloca types match up, we don't have to build the llvm::Value *Ptr = V;
// FCA at all, emit a series of GEPs and stores, which is better for Ptr = Builder.CreateBitCast(Ptr, llvm::PointerType::getUnqual(STy));
// fast isel.
if (STy == cast<llvm::PointerType>(V->getType())->getElementType()) { for (unsigned i = 0, e = STy->getNumElements(); i != e; ++i) {
for (unsigned i = 0, e = STy->getNumElements(); i != e; ++i) { assert(AI != Fn->arg_end() && "Argument mismatch!");
assert(AI != Fn->arg_end() && "Argument mismatch!"); AI->setName(Arg->getName() + ".coerce" + llvm::Twine(i));
AI->setName(Arg->getName() + ".coerce" + llvm::Twine(i)); llvm::Value *EltPtr = Builder.CreateConstGEP2_32(Ptr, 0, i);
llvm::Value *EltPtr = Builder.CreateConstGEP2_32(V, 0, i); Builder.CreateStore(AI++, EltPtr);
Builder.CreateStore(AI++, EltPtr);
}
} else {
// Reconstruct the FCA here so we can do a coerced store.
llvm::Value *FormalArg = llvm::UndefValue::get(STy);
for (unsigned i = 0, e = STy->getNumElements(); i != e; ++i) {
assert(AI != Fn->arg_end() && "Argument mismatch!");
AI->setName(Arg->getName() + ".coerce" + llvm::Twine(i));
FormalArg = Builder.CreateInsertValue(FormalArg, AI++, i);
}
CreateCoercedStore(FormalArg, V, /*DestIsVolatile=*/false, *this);
} }
} else { } else {
// Simple case, just do a coerced store of the argument into the alloca. // Simple case, just do a coerced store of the argument into the alloca.
@ -1166,22 +1155,11 @@ RValue CodeGenFunction::EmitCall(const CGFunctionInfo &CallInfo,
// and the optimizer generally likes scalar values better than FCAs. // and the optimizer generally likes scalar values better than FCAs.
if (const llvm::StructType *STy = if (const llvm::StructType *STy =
dyn_cast<llvm::StructType>(ArgInfo.getCoerceToType())) { dyn_cast<llvm::StructType>(ArgInfo.getCoerceToType())) {
// If the argument and alloca types match up, we don't have to build the SrcPtr = Builder.CreateBitCast(SrcPtr,
// FCA at all, emit a series of GEPs and loads, which is better for llvm::PointerType::getUnqual(STy));
// fast isel. for (unsigned i = 0, e = STy->getNumElements(); i != e; ++i) {
if (STy ==cast<llvm::PointerType>(SrcPtr->getType())->getElementType()){ llvm::Value *EltPtr = Builder.CreateConstGEP2_32(SrcPtr, 0, i);
for (unsigned i = 0, e = STy->getNumElements(); i != e; ++i) { Args.push_back(Builder.CreateLoad(EltPtr));
llvm::Value *EltPtr = Builder.CreateConstGEP2_32(SrcPtr, 0, i);
Args.push_back(Builder.CreateLoad(EltPtr));
}
} else {
// Otherwise, do a coerced load the entire FCA and handle the pieces.
llvm::Value *SrcVal =
CreateCoercedLoad(SrcPtr, ArgInfo.getCoerceToType(), *this);
// Extract the elements of the value to pass in.
for (unsigned i = 0, e = STy->getNumElements(); i != e; ++i)
Args.push_back(Builder.CreateExtractValue(SrcVal, i));
} }
} else { } else {
// In the simple case, just pass the coerced loaded value. // In the simple case, just pass the coerced loaded value.