forked from OSchip/llvm-project
fe34c1d53e
have a "coerce to" type which often matches the default lowering of Clang
type to LLVM IR type, but the coerce case can be handled by making them
not be the same.
This simplifies things and fixes issues where X86-64 abi lowering would
return coerce after making preferred types exactly match up. This caused
us to compile:
typedef float v4f32 __attribute__((__vector_size__(16)));
v4f32 foo(v4f32 X) {
return X+X;
}
into this code at -O0:
define <4 x float> @foo(<4 x float> %X.coerce) nounwind {
entry:
%retval = alloca <4 x float>, align 16 ; <<4 x float>*> [#uses=2]
%coerce = alloca <4 x float>, align 16 ; <<4 x float>*> [#uses=2]
%X.addr = alloca <4 x float>, align 16 ; <<4 x float>*> [#uses=3]
store <4 x float> %X.coerce, <4 x float>* %coerce
%X = load <4 x float>* %coerce ; <<4 x float>> [#uses=1]
store <4 x float> %X, <4 x float>* %X.addr
%tmp = load <4 x float>* %X.addr ; <<4 x float>> [#uses=1]
%tmp1 = load <4 x float>* %X.addr ; <<4 x float>> [#uses=1]
%add = fadd <4 x float> %tmp, %tmp1 ; <<4 x float>> [#uses=1]
store <4 x float> %add, <4 x float>* %retval
%0 = load <4 x float>* %retval ; <<4 x float>> [#uses=1]
ret <4 x float> %0
}
Now we get:
define <4 x float> @foo(<4 x float> %X) nounwind {
entry:
%X.addr = alloca <4 x float>, align 16 ; <<4 x float>*> [#uses=3]
store <4 x float> %X, <4 x float>* %X.addr
%tmp = load <4 x float>* %X.addr ; <<4 x float>> [#uses=1]
%tmp1 = load <4 x float>* %X.addr ; <<4 x float>> [#uses=1]
%add = fadd <4 x float> %tmp, %tmp1 ; <<4 x float>> [#uses=1]
ret <4 x float> %add
}
This implements rdar://8248065
llvm-svn: 109733
|
||
|---|---|---|
| .. | ||
| ABIInfo.h | ||
| BackendUtil.cpp | ||
| CGBlocks.cpp | ||
| CGBlocks.h | ||
| CGBuilder.h | ||
| CGBuiltin.cpp | ||
| CGCXX.cpp | ||
| CGCXX.h | ||
| CGCXXABI.h | ||
| CGCall.cpp | ||
| CGCall.h | ||
| CGClass.cpp | ||
| CGDebugInfo.cpp | ||
| CGDebugInfo.h | ||
| CGDecl.cpp | ||
| CGDeclCXX.cpp | ||
| CGException.cpp | ||
| CGException.h | ||
| CGExpr.cpp | ||
| CGExprAgg.cpp | ||
| CGExprCXX.cpp | ||
| CGExprComplex.cpp | ||
| CGExprConstant.cpp | ||
| CGExprScalar.cpp | ||
| CGObjC.cpp | ||
| CGObjCGNU.cpp | ||
| CGObjCMac.cpp | ||
| CGObjCRuntime.h | ||
| CGRTTI.cpp | ||
| CGRecordLayout.h | ||
| CGRecordLayoutBuilder.cpp | ||
| CGStmt.cpp | ||
| CGTemporaries.cpp | ||
| CGVTT.cpp | ||
| CGVTables.cpp | ||
| CGVTables.h | ||
| CGValue.h | ||
| CMakeLists.txt | ||
| CodeGenAction.cpp | ||
| CodeGenFunction.cpp | ||
| CodeGenFunction.h | ||
| CodeGenModule.cpp | ||
| CodeGenModule.h | ||
| CodeGenTypes.cpp | ||
| CodeGenTypes.h | ||
| GlobalDecl.h | ||
| ItaniumCXXABI.cpp | ||
| Makefile | ||
| Mangle.cpp | ||
| Mangle.h | ||
| MicrosoftCXXABI.cpp | ||
| ModuleBuilder.cpp | ||
| README.txt | ||
| TargetInfo.cpp | ||
| TargetInfo.h | ||
README.txt
IRgen optimization opportunities. //===---------------------------------------------------------------------===// The common pattern of -- short x; // or char, etc (x == 10) -- generates an zext/sext of x which can easily be avoided. //===---------------------------------------------------------------------===// Bitfields accesses can be shifted to simplify masking and sign extension. For example, if the bitfield width is 8 and it is appropriately aligned then is is a lot shorter to just load the char directly. //===---------------------------------------------------------------------===// It may be worth avoiding creation of alloca's for formal arguments for the common situation where the argument is never written to or has its address taken. The idea would be to begin generating code by using the argument directly and if its address is taken or it is stored to then generate the alloca and patch up the existing code. In theory, the same optimization could be a win for block local variables as long as the declaration dominates all statements in the block. NOTE: The main case we care about this for is for -O0 -g compile time performance, and in that scenario we will need to emit the alloca anyway currently to emit proper debug info. So this is blocked by being able to emit debug information which refers to an LLVM temporary, not an alloca. //===---------------------------------------------------------------------===// We should try and avoid generating basic blocks which only contain jumps. At -O0, this penalizes us all the way from IRgen (malloc & instruction overhead), all the way down through code generation and assembly time. On 176.gcc:expr.ll, it looks like over 12% of basic blocks are just direct branches! //===---------------------------------------------------------------------===//