forked from OSchip/llvm-project
[DAG] Fix wrong canonicalization performed on shuffle nodes.
This fixes a regression introduced by r226816. When replacing a splat shuffle node with a constant build_vector, make sure that the new build_vector has a valid number of elements. Thanks to Patrik Hagglund for reporting this problem and providing a small reproducible. llvm-svn: 227002
This commit is contained in:
parent
9dea5cdb8e
commit
8381475a75
|
@ -1538,13 +1538,15 @@ SDValue SelectionDAG::getVectorShuffle(EVT VT, SDLoc dl, SDValue N1,
|
||||||
if (Splat && Splat.getOpcode() == ISD::UNDEF)
|
if (Splat && Splat.getOpcode() == ISD::UNDEF)
|
||||||
return getUNDEF(VT);
|
return getUNDEF(VT);
|
||||||
|
|
||||||
|
bool SameNumElts =
|
||||||
|
V.getValueType().getVectorNumElements() == VT.getVectorNumElements();
|
||||||
|
|
||||||
// We only have a splat which can skip shuffles if there is a splatted
|
// We only have a splat which can skip shuffles if there is a splatted
|
||||||
// value and no undef lanes rearranged by the shuffle.
|
// value and no undef lanes rearranged by the shuffle.
|
||||||
if (Splat && UndefElements.none()) {
|
if (Splat && UndefElements.none()) {
|
||||||
// Splat of <x, x, ..., x>, return <x, x, ..., x>, provided that the
|
// Splat of <x, x, ..., x>, return <x, x, ..., x>, provided that the
|
||||||
// number of elements match or the value splatted is a zero constant.
|
// number of elements match or the value splatted is a zero constant.
|
||||||
if (V.getValueType().getVectorNumElements() ==
|
if (SameNumElts)
|
||||||
VT.getVectorNumElements())
|
|
||||||
return N1;
|
return N1;
|
||||||
if (auto *C = dyn_cast<ConstantSDNode>(Splat))
|
if (auto *C = dyn_cast<ConstantSDNode>(Splat))
|
||||||
if (C->isNullValue())
|
if (C->isNullValue())
|
||||||
|
@ -1553,15 +1555,15 @@ SDValue SelectionDAG::getVectorShuffle(EVT VT, SDLoc dl, SDValue N1,
|
||||||
|
|
||||||
// If the shuffle itself creates a constant splat, build the vector
|
// If the shuffle itself creates a constant splat, build the vector
|
||||||
// directly.
|
// directly.
|
||||||
if (AllSame) {
|
if (AllSame && SameNumElts) {
|
||||||
const SDValue &Splatted = BV->getOperand(MaskVec[0]);
|
const SDValue &Splatted = BV->getOperand(MaskVec[0]);
|
||||||
if (isa<ConstantSDNode>(Splatted) || isa<ConstantFPSDNode>(Splatted)) {
|
if (isa<ConstantSDNode>(Splatted) || isa<ConstantFPSDNode>(Splatted)) {
|
||||||
SmallVector<SDValue, 8> Ops;
|
SmallVector<SDValue, 8> Ops;
|
||||||
for (unsigned i = 0; i != NElts; ++i) {
|
for (unsigned i = 0; i != NElts; ++i)
|
||||||
Ops.push_back(Splatted);
|
Ops.push_back(Splatted);
|
||||||
}
|
|
||||||
SDValue NewBV = getNode(ISD::BUILD_VECTOR, dl,
|
SDValue NewBV =
|
||||||
BV->getValueType(0), Ops);
|
getNode(ISD::BUILD_VECTOR, dl, BV->getValueType(0), Ops);
|
||||||
|
|
||||||
// We may have jumped through bitcasts, so the type of the
|
// We may have jumped through bitcasts, so the type of the
|
||||||
// BUILD_VECTOR may not match the type of the shuffle.
|
// BUILD_VECTOR may not match the type of the shuffle.
|
||||||
|
|
|
@ -0,0 +1,15 @@
|
||||||
|
; RUN: llc -mtriple=x86_64-unknown-unknown -mcpu=corei7 < %s
|
||||||
|
|
||||||
|
; Check that llc doesn't crash in the attempt to fold a shuffle with
|
||||||
|
; a splat mask into a constant build_vector.
|
||||||
|
|
||||||
|
define <8 x i8> @autogen_SD26299(i8) {
|
||||||
|
BB:
|
||||||
|
%Shuff = shufflevector <8 x i32> <i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1, i32 -1>, <8 x i32> zeroinitializer, <8 x i32> <i32 2, i32 undef, i32 6, i32 8, i32 undef, i32 12, i32 14, i32 0>
|
||||||
|
%Shuff14 = shufflevector <8 x i32> %Shuff, <8 x i32> %Shuff, <8 x i32> <i32 7, i32 9, i32 11, i32 undef, i32 undef, i32 1, i32 3, i32 5>
|
||||||
|
%Shuff35 = shufflevector <8 x i32> %Shuff14, <8 x i32> %Shuff, <8 x i32> <i32 undef, i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13>
|
||||||
|
%I42 = insertelement <8 x i32> %Shuff35, i32 88608, i32 0
|
||||||
|
%Shuff48 = shufflevector <8 x i32> %Shuff35, <8 x i32> %I42, <8 x i32> <i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 0, i32 2>
|
||||||
|
%Tr59 = trunc <8 x i32> %Shuff48 to <8 x i8>
|
||||||
|
ret <8 x i8> %Tr59
|
||||||
|
}
|
Loading…
Reference in New Issue