From 8a97fddbc2c064e295ad5fb6269795bd904e1ed2 Mon Sep 17 00:00:00 2001 From: Owen Anderson Date: Thu, 13 Mar 2008 22:07:10 +0000 Subject: [PATCH] Fix a bug in GVN that Duncan noticed, where we potentially need to insert a pointer bitcast when performing return slot optimization. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@48343 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/Transforms/Scalar/GVN.cpp | 6 +++++- .../GVN/2008-03-13-ReturnSlotBitcast.ll | 19 +++++++++++++++++++ 2 files changed, 24 insertions(+), 1 deletion(-) create mode 100644 test/Transforms/GVN/2008-03-13-ReturnSlotBitcast.ll diff --git a/lib/Transforms/Scalar/GVN.cpp b/lib/Transforms/Scalar/GVN.cpp index 545f7091cc1..1d764af6888 100644 --- a/lib/Transforms/Scalar/GVN.cpp +++ b/lib/Transforms/Scalar/GVN.cpp @@ -1171,8 +1171,12 @@ bool GVN::performCallSlotOptzn(MemCpyInst* cpy, CallInst* C, // All the checks have passed, so do the transformation. for (unsigned i = 0; i < CS.arg_size(); ++i) - if (CS.getArgument(i) == cpySrc) + if (CS.getArgument(i) == cpySrc) { + if (cpySrc->getType() != cpyDest->getType()) + cpyDest = CastInst::createPointerCast(cpyDest, cpySrc->getType(), + cpyDest->getName(), C); CS.setArgument(i, cpyDest); + } // Drop any cached information about the call, because we may have changed // its dependence information by changing its parameter. diff --git a/test/Transforms/GVN/2008-03-13-ReturnSlotBitcast.ll b/test/Transforms/GVN/2008-03-13-ReturnSlotBitcast.ll new file mode 100644 index 00000000000..7757a92b240 --- /dev/null +++ b/test/Transforms/GVN/2008-03-13-ReturnSlotBitcast.ll @@ -0,0 +1,19 @@ +; RUN: llvm-as < %s | opt -gvn | llvm-dis | not grep {call.*memcpy.} + %a = type { i32 } + %b = type { float } + +declare void @llvm.memcpy.i32(i8*, i8*, i32, i32) nounwind +declare void @g(%a*) + +define float @f() { +entry: + %a_var = alloca %a + %b_var = alloca %b + call void @g(%a *%a_var) + %a_i8 = bitcast %a* %a_var to i8* + %b_i8 = bitcast %b* %b_var to i8* + call void @llvm.memcpy.i32(i8* %b_i8, i8* %a_i8, i32 4, i32 4) + %tmp1 = getelementptr %b* %b_var, i32 0, i32 0 + %tmp2 = load float* %tmp1 + ret float %tmp2 +} -- 2.34.1