From f8d1d0019094acb3979aee60f98d11abe6cc92cf Mon Sep 17 00:00:00 2001 From: Simon Pilgrim Date: Sat, 31 Aug 2019 16:21:31 +0000 Subject: [X86] EltsFromConsecutiveLoads - Don't confuse elt count with vector element count (PR43170) EltsFromConsecutiveLoads was assuming that the number of input elts was the same as the number of elements in the output vector type when creating a zeroing shuffle, causing an assert when subvectors were being combined instead of just scalars. llvm-svn: 370592 --- llvm/lib/Target/X86/X86ISelLowering.cpp | 27 ++++++++++++++++----------- 1 file changed, 16 insertions(+), 11 deletions(-) (limited to 'llvm/lib/Target') diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp index 5812cc39b2e..c603aa51efc 100644 --- a/llvm/lib/Target/X86/X86ISelLowering.cpp +++ b/llvm/lib/Target/X86/X86ISelLowering.cpp @@ -7832,17 +7832,22 @@ static SDValue EltsFromConsecutiveLoads(EVT VT, ArrayRef Elts, // IsConsecutiveLoadWithZeros - we need to create a shuffle of the loaded // vector and a zero vector to clear out the zero elements. if (!isAfterLegalize && VT.isVector()) { - SmallVector ClearMask(NumElems, -1); - for (unsigned i = 0; i < NumElems; ++i) { - if (ZeroMask[i]) - ClearMask[i] = i + NumElems; - else if (LoadMask[i]) - ClearMask[i] = i; - } - SDValue V = CreateLoad(VT, LDBase); - SDValue Z = VT.isInteger() ? DAG.getConstant(0, DL, VT) - : DAG.getConstantFP(0.0, DL, VT); - return DAG.getVectorShuffle(VT, DL, V, Z, ClearMask); + unsigned NumMaskElts = VT.getVectorNumElements(); + if ((NumMaskElts % NumElems) == 0) { + unsigned Scale = NumMaskElts / NumElems; + SmallVector ClearMask(NumMaskElts, -1); + for (unsigned i = 0; i < NumElems; ++i) { + if (UndefMask[i]) + continue; + int Offset = ZeroMask[i] ? NumMaskElts : 0; + for (unsigned j = 0; j != Scale; ++j) + ClearMask[(i * Scale) + j] = (i * Scale) + j + Offset; + } + SDValue V = CreateLoad(VT, LDBase); + SDValue Z = VT.isInteger() ? DAG.getConstant(0, DL, VT) + : DAG.getConstantFP(0.0, DL, VT); + return DAG.getVectorShuffle(VT, DL, V, Z, ClearMask); + } } } -- cgit v1.2.3