RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-12-15 07:39:31 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Fix a bug in memcpyopt where the memcpy-memcpy transform was never being applied because

Browse Source 
we were checking for it in the wrong order.  This caused a miscompilation because the
return slot optimization assumes that the call it is dealing with is NOT a memcpy.

llvm-svn: 50444
This commit is contained in:
Owen Anderson 2008-04-29 21:26:06 +00:00
parent 8150660ba3
commit 2caa79ae70
2 changed files with 23 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
lib/Transforms/Scalar/MemCpyOptimizer.cpp
View File

@ -615,10 +615,12 @@ bool MemCpyOpt::processMemCpy(MemCpyInst* M) {
if (dep == MemoryDependenceAnalysis::None ||
dep == MemoryDependenceAnalysis::NonLocal)
return false;
else if (CallInst* C = dyn_cast<CallInst>(dep))
return performCallSlotOptzn(M, C);
else if (!isa<MemCpyInst>(dep))
return false;
else if (!isa<MemCpyInst>(dep)) {
if (CallInst* C = dyn_cast<CallInst>(dep))
return performCallSlotOptzn(M, C);
else
return false;
}
MemCpyInst* MDep = cast<MemCpyInst>(dep);

17
test/Transforms/MemCpyOpt/2008-04-29-SRetRemoval.ll Normal file
View File

@ -0,0 +1,17 @@
; RUN: llvm-as < %s | opt -memcpyopt | llvm-dis | grep {call.*memcpy.*agg.result}
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-f80:128:128"
target triple = "i386-apple-darwin8"
@x = external global { x86_fp80, x86_fp80 } ; <{ x86_fp80, x86_fp80 }*> [#uses=1]
define void @foo({ x86_fp80, x86_fp80 }* noalias sret %agg.result) nounwind {
entry:
%x.0 = alloca { x86_fp80, x86_fp80 } ; <{ x86_fp80, x86_fp80 }*> [#uses=1]
%x.01 = bitcast { x86_fp80, x86_fp80 }* %x.0 to i8* ; <i8*> [#uses=2]
call void @llvm.memcpy.i32( i8* %x.01, i8* bitcast ({ x86_fp80, x86_fp80 }* @x to i8*), i32 32, i32 16 )
%agg.result2 = bitcast { x86_fp80, x86_fp80 }* %agg.result to i8* ; <i8*> [#uses=1]
call void @llvm.memcpy.i32( i8* %agg.result2, i8* %x.01, i32 32, i32 16 )
ret void
}
declare void @llvm.memcpy.i32(i8*, i8*, i32, i32) nounwind