[llvm] r296130 - [X86][SSE] Target shuffle combine can try to combine up to 16 vectors

[Simon Pilgrim via llvm-commits]

Author: rksimon
Date: Fri Feb 24 09:35:52 2017
New Revision: 296130

URL: http://llvm.org/viewvc/llvm-project?rev=296130&view=rev
Log:
[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).

Modified:
    llvm/trunk/lib/Target/X86/X86ISelLowering.cpp

Modified: llvm/trunk/lib/Target/X86/X86ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/X86/X86ISelLowering.cpp?rev=296130&r1=296129&r2=296130&view=diff
==============================================================================
--- llvm/trunk/lib/Target/X86/X86ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/X86/X86ISelLowering.cpp Fri Feb 24 09:35:52 2017
@@ -5826,7 +5826,7 @@ static bool getFauxShuffleMask(SDValue N
 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(
 
   // 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 combineX86ShufflesRecursivel
   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 combineX86ShufflesRecursivel
   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