diff options
| author | Evan Cheng <evan.cheng@apple.com> | 2012-12-12 20:43:23 +0000 |
|---|---|---|
| committer | Evan Cheng <evan.cheng@apple.com> | 2012-12-12 20:43:23 +0000 |
| commit | a16e49d56f6349c12da2b561da00c22e13eda09b (patch) | |
| tree | 905a15c1729dca89f596e2ac1ae7f41850b86955 | |
| parent | 94aa3bd0c165c822b682c363fba1a4b5d7960ecd (diff) | |
| download | llvm-a16e49d56f6349c12da2b561da00c22e13eda09b.tar.gz llvm-a16e49d56f6349c12da2b561da00c22e13eda09b.tar.bz2 llvm-a16e49d56f6349c12da2b561da00c22e13eda09b.tar.xz | |
Fix a logic bug in inline expansion of memcpy / memset with an overlapping
load / store pair. It's not legal to use a wider load than the size of
the remaining bytes if it's the first pair of load / store.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@170018 91177308-0d34-0410-b5e6-96231b3b80d8
| -rw-r--r-- | lib/CodeGen/SelectionDAG/SelectionDAG.cpp | 9 | ||||
| -rw-r--r-- | test/CodeGen/Mips/2012-12-12-ExpandMemcpy.ll | 11 |
2 files changed, 16 insertions, 4 deletions
diff --git a/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/lib/CodeGen/SelectionDAG/SelectionDAG.cpp index 269f22150a..2375182167 100644 --- a/lib/CodeGen/SelectionDAG/SelectionDAG.cpp +++ b/lib/CodeGen/SelectionDAG/SelectionDAG.cpp @@ -3469,9 +3469,6 @@ static bool FindOptimalMemOpLowering(std::vector<EVT> &MemOps, unsigned NumMemOps = 0; while (Size != 0) { - if (++NumMemOps > Limit) - return false; - unsigned VTSize = VT.getSizeInBits() / 8; while (VTSize > Size) { // For now, only use non-vector load / store's for the left-over pieces. @@ -3507,7 +3504,8 @@ static bool FindOptimalMemOpLowering(std::vector<EVT> &MemOps, // FIXME: Only does this for 64-bit or more since we don't have proper // cost model for unaligned load / store. bool Fast; - if (AllowOverlap && VTSize >= 8 && NewVTSize < Size && + if (NumMemOps && AllowOverlap && + VTSize >= 8 && NewVTSize < Size && TLI.allowsUnalignedMemoryAccesses(VT, &Fast) && Fast) VTSize = Size; else { @@ -3516,6 +3514,9 @@ static bool FindOptimalMemOpLowering(std::vector<EVT> &MemOps, } } + if (++NumMemOps > Limit) + return false; + MemOps.push_back(VT); Size -= VTSize; } diff --git a/test/CodeGen/Mips/2012-12-12-ExpandMemcpy.ll b/test/CodeGen/Mips/2012-12-12-ExpandMemcpy.ll new file mode 100644 index 0000000000..9d4daee696 --- /dev/null +++ b/test/CodeGen/Mips/2012-12-12-ExpandMemcpy.ll @@ -0,0 +1,11 @@ +; RUN: llc -march=mips64el -mcpu=mips64r2 < %s + +@.str = private unnamed_addr constant [7 x i8] c"hello\0A\00", align 1 + +define void @t(i8* %ptr) { +entry: + tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %ptr, i8* getelementptr inbounds ([7 x i8]* @.str, i64 0, i64 0), i64 7, i32 1, i1 false) + ret void +} + +declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture, i64, i32, i1) nounwind |