From 444e7f978344d11156b345953acaddd4073d51cb Mon Sep 17 00:00:00 2001 From: Sanjay Patel Date: Thu, 10 Mar 2016 22:35:33 +0000 Subject: [PATCH] [x86] don't use a shuffle when a vselect will do; NFCI Looking at the IR definition of a masked load made me realize there was no reason to use a shuffle here, so we don't need to convert the format of the mask at all. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@263167 91177308-0d34-0410-b5e6-96231b3b80d8 --- lib/Target/X86/X86ISelLowering.cpp | 21 +++++---------------- 1 file changed, 5 insertions(+), 16 deletions(-) diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp index 78907023215..2d2d4745978 100644 --- a/lib/Target/X86/X86ISelLowering.cpp +++ b/lib/Target/X86/X86ISelLowering.cpp @@ -27336,8 +27336,8 @@ reduceMaskedLoadToScalarLoad(MaskedLoadSDNode *ML, SelectionDAG &DAG, return DCI.CombineTo(ML, Insert, Load.getValue(1), true); } -/// Convert a masked load with a constant mask into a masked load and a shuffle. -/// This allows the blend operation to use a faster kind of shuffle instruction +/// Convert a masked load with a constant mask into a masked load and a select. +/// This allows the select operation to use a faster kind of shuffle instruction /// (for example, vblendvps -> vblendps). static SDValue combineMaskedLoadConstantMask(MaskedLoadSDNode *ML, SelectionDAG &DAG, @@ -27348,26 +27348,15 @@ combineMaskedLoadConstantMask(MaskedLoadSDNode *ML, SelectionDAG &DAG, ML->getSrc0().getOpcode() == ISD::UNDEF) return SDValue(); - // Convert the masked load's mask into a blend mask for a vector shuffle node. - EVT VT = ML->getValueType(0); - unsigned NumElts = VT.getVectorNumElements(); - BuildVectorSDNode *MaskBV = cast(ML->getMask()); - SmallVector ShufMask(NumElts, SM_SentinelUndef); - for (unsigned i = 0; i < NumElts; ++i) { - // If this mask bit of the masked load is false, the pass-through vector - // (Src0) element will be selected for that vector lane. - if (MaskBV->getOperand(i).getOpcode() != ISD::UNDEF) - ShufMask[i] = isNullConstant(MaskBV->getOperand(i)) ? i + NumElts : i; - } - - // The new masked load has an undef pass-through operand. The shuffle uses the + // The new masked load has an undef pass-through operand. The select uses the // original pass-through operand. SDLoc DL(ML); + EVT VT = ML->getValueType(0); SDValue NewML = DAG.getMaskedLoad(VT, DL, ML->getChain(), ML->getBasePtr(), ML->getMask(), DAG.getUNDEF(VT), ML->getMemoryVT(), ML->getMemOperand(), ML->getExtensionType()); - SDValue Blend = DAG.getVectorShuffle(VT, DL, NewML, ML->getSrc0(), ShufMask); + SDValue Blend = DAG.getSelect(DL, VT, ML->getMask(), NewML, ML->getSrc0()); return DCI.CombineTo(ML, Blend, NewML.getValue(1), true); } -- 2.11.0