summaryrefslogtreecommitdiffstats
path: root/llvm/lib/CodeGen
diff options
context:
space:
mode:
authorSanjay Patel <spatel@rotateright.com>2020-01-07 09:47:50 -0500
committerSanjay Patel <spatel@rotateright.com>2020-01-07 09:48:59 -0500
commit58e2e92a57fcc3c628fd03ae33698fcc9aabedb9 (patch)
treea868bd918c4f06b22ab4b9d2d385a6ba55ebc438 /llvm/lib/CodeGen
parent3f2e3dc44b42fab2e991222e74248b7006f1091e (diff)
downloadbcm5719-llvm-58e2e92a57fcc3c628fd03ae33698fcc9aabedb9.tar.gz
bcm5719-llvm-58e2e92a57fcc3c628fd03ae33698fcc9aabedb9.zip
[DAGCombiner] reduce shuffle of concat of same vector
This is possibly a small part towards solving PR42024: https://bugs.llvm.org/show_bug.cgi?id=42024 The vectorizer is creating shuffles of concat like this: %63 = shufflevector <4 x i64> %x, <4 x i64> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3> %64 = shufflevector <8 x i64> %63, <8 x i64> undef, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7> That might be fixable in the vectorizers, but we're not allowed to fold that into a single shuffle in instcombine, so we should have a backend backstop to convert that into the likely simpler form: %64 = shufflevector <4 x i64> %x, <4 x i64> undef, <8 x i32> <i32 0, i32 0, i32 1, i32 1, i32 2, i32 2, i32 3, i32 3> Differential Revision: https://reviews.llvm.org/D72300
Diffstat (limited to 'llvm/lib/CodeGen')
-rw-r--r--llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp24
1 files changed, 24 insertions, 0 deletions
diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index eba7d695c77..fd98b9e7b75 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -19286,6 +19286,30 @@ SDValue DAGCombiner::visitVECTOR_SHUFFLE(SDNode *N) {
return V;
}
+ // A shuffle of a concat of the same narrow vector can be reduced to use
+ // only low-half elements of a concat with undef:
+ // shuf (concat X, X), undef, Mask --> shuf (concat X, undef), undef, Mask'
+ if (N0.getOpcode() == ISD::CONCAT_VECTORS && N1.isUndef() &&
+ N0.getNumOperands() == 2 &&
+ N0.getOperand(0) == N0.getOperand(1)) {
+ int HalfNumElts = (int)NumElts / 2;
+ SmallVector<int, 8> NewMask;
+ for (unsigned i = 0; i != NumElts; ++i) {
+ int Idx = SVN->getMaskElt(i);
+ if (Idx >= HalfNumElts) {
+ assert(Idx < (int)NumElts && "Shuffle mask chooses undef op");
+ Idx -= HalfNumElts;
+ }
+ NewMask.push_back(Idx);
+ }
+ if (TLI.isShuffleMaskLegal(NewMask, VT)) {
+ SDValue UndefVec = DAG.getUNDEF(N0.getOperand(0).getValueType());
+ SDValue NewCat = DAG.getNode(ISD::CONCAT_VECTORS, SDLoc(N), VT,
+ N0.getOperand(0), UndefVec);
+ return DAG.getVectorShuffle(VT, SDLoc(N), NewCat, N1, NewMask);
+ }
+ }
+
// Attempt to combine a shuffle of 2 inputs of 'scalar sources' -
// BUILD_VECTOR or SCALAR_TO_VECTOR into a single BUILD_VECTOR.
if (Level < AfterLegalizeDAG && TLI.isTypeLegal(VT))
OpenPOWER on IntegriCloud