maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

[X86][SSE] Target shuffle combine can try to combine up to 16 vectors 

Noticed while profiling PR32037, the target shuffle ops were being stored in SmallVector<*,8> types but the combiner could store as many as 16 ops at maximum depth (2 per depth).

llvm-svn: 296130
This commit is contained in:
Simon Pilgrim 2017-02-24 15:35:52 +00:00
parent ec9a8de0e6
commit 7f6a7c97a7
1 changed files with 6 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
llvm/lib/Target/X86/X86ISelLowering.cpp
View File

@ -5826,7 +5826,7 @@ static bool getFauxShuffleMask(SDValue N, SmallVectorImpl<int> &Mask,
static void resolveTargetShuffleInputsAndMask(SmallVectorImpl<SDValue> &Inputs,
SmallVectorImpl<int> &Mask) {
int MaskWidth = Mask.size();
SmallVector<SDValue, 8> UsedInputs;
SmallVector<SDValue, 16> UsedInputs;
for (int i = 0, e = Inputs.size(); i < e; ++i) {
int lo = UsedInputs.size() * MaskWidth;
int hi = lo + MaskWidth;
@ -27394,8 +27394,8 @@ static bool combineX86ShufflesConstants(const SmallVectorImpl<SDValue> &Ops,
// Extract constant bits from each source op.
bool OneUseConstantOp = false;
SmallVector<SmallBitVector, 4> UndefEltsOps(NumOps);
SmallVector<SmallVector<APInt, 8>, 4> RawBitsOps(NumOps);
SmallVector<SmallBitVector, 16> UndefEltsOps(NumOps);
SmallVector<SmallVector<APInt, 16>, 16> RawBitsOps(NumOps);
for (unsigned i = 0; i != NumOps; ++i) {
SDValue SrcOp = Ops[i];
OneUseConstantOp |= SrcOp.hasOneUse();
@ -27530,7 +27530,7 @@ static bool combineX86ShufflesRecursively(ArrayRef<SDValue> SrcOps,
SDValue Input1 = (OpInputs.size() > 1 ? OpInputs[1] : SDValue());
// Add the inputs to the Ops list, avoiding duplicates.
SmallVector<SDValue, 8> Ops(SrcOps.begin(), SrcOps.end());
SmallVector<SDValue, 16> Ops(SrcOps.begin(), SrcOps.end());
int InputIdx0 = -1, InputIdx1 = -1;
for (int i = 0, e = Ops.size(); i < e; ++i) {
@ -27634,8 +27634,8 @@ static bool combineX86ShufflesRecursively(ArrayRef<SDValue> SrcOps,
HasVariableMask |= isTargetShuffleVariableMask(Op.getOpcode());
// Update the list of shuffle nodes that have been combined so far.
SmallVector<const SDNode *, 8> CombinedNodes(SrcNodes.begin(),
SrcNodes.end());
SmallVector<const SDNode *, 16> CombinedNodes(SrcNodes.begin(),
SrcNodes.end());
CombinedNodes.push_back(Op.getNode());
// See if we can recurse into each shuffle source op (if it's a target