summaryrefslogtreecommitdiffstats
path: root/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
diff options
context:
space:
mode:
authorJohn Brawn <john.brawn@arm.com>2017-05-26 13:59:12 +0000
committerJohn Brawn <john.brawn@arm.com>2017-05-26 13:59:12 +0000
commit9009d2905debfc210859e5d20a1fb3ec319a9ce7 (patch)
tree93ad3b86753bdbce6d9657acf75739ae0d92f12d /llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
parentba9d8ba82aab927dd8e132c947cd64efff4e09d1 (diff)
downloadbcm5719-llvm-9009d2905debfc210859e5d20a1fb3ec319a9ce7.tar.gz
bcm5719-llvm-9009d2905debfc210859e5d20a1fb3ec319a9ce7.zip
[ARM] Fix lowering of misaligned memcpy/memset
Currently getOptimalMemOpType returns i32 for large enough sizes without checking for alignment, leading to poor code generation when misaligned accesses aren't permitted as we generate a word store then later split it up into byte stores. This means we inadvertantly go over the MaxStoresPerMemcpy limit and for memset we splat the memset value into a word then immediately split it up again. Fix this by leaving it up to FindOptimalMemOpLowering to figure out which type to use, but also fix a bug there where it wasn't correctly checking if misaligned memory accesses are allowed. Differential Revision: https://reviews.llvm.org/D33442 llvm-svn: 303990
Diffstat (limited to 'llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp')
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp24
1 files changed, 12 insertions, 12 deletions
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index b26f1ce8c5e..177898e1e95 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -4779,23 +4779,23 @@ static bool FindOptimalMemOpLowering(std::vector<EVT> &MemOps,
DAG.getMachineFunction());
if (VT == MVT::Other) {
- if (DstAlign >= DAG.getDataLayout().getPointerPrefAlignment(DstAS) ||
- TLI.allowsMisalignedMemoryAccesses(VT, DstAS, DstAlign)) {
- VT = TLI.getPointerTy(DAG.getDataLayout(), DstAS);
- } else {
- switch (DstAlign & 7) {
- case 0: VT = MVT::i64; break;
- case 4: VT = MVT::i32; break;
- case 2: VT = MVT::i16; break;
- default: VT = MVT::i8; break;
- }
- }
-
+ // Use the largest integer type whose alignment constraints are satisfied.
+ // We only need to check DstAlign here as SrcAlign is always greater or
+ // equal to DstAlign (or zero).
+ VT = MVT::i64;
+ while (DstAlign && DstAlign < VT.getSizeInBits() / 8 &&
+ !TLI.allowsMisalignedMemoryAccesses(VT, DstAS, DstAlign))
+ VT = (MVT::SimpleValueType)(VT.getSimpleVT().SimpleTy - 1);
+ assert(VT.isInteger());
+
+ // Find the largest legal integer type.
MVT LVT = MVT::i64;
while (!TLI.isTypeLegal(LVT))
LVT = (MVT::SimpleValueType)(LVT.SimpleTy - 1);
assert(LVT.isInteger());
+ // If the type we've chosen is larger than the largest legal integer type
+ // then use that instead.
if (VT.bitsGT(LVT))
VT = LVT;
}
OpenPOWER on IntegriCloud