diff options
| author | Simon Pilgrim <llvm-dev@redking.me.uk> | 2018-02-15 13:33:15 +0000 |
|---|---|---|
| committer | Simon Pilgrim <llvm-dev@redking.me.uk> | 2018-02-15 13:33:15 +0000 |
| commit | 908f833e57bfacb310248cd9ef1c0257360a0c4e (patch) | |
| tree | 076c59aea087e44288d007900b2c9dc0372a60c3 /llvm/lib/Target/X86/X86ISelLowering.cpp | |
| parent | 6acc0e555b8a91a07fbeec1ea46cf33bee9c71cb (diff) | |
| download | bcm5719-llvm-908f833e57bfacb310248cd9ef1c0257360a0c4e.tar.gz bcm5719-llvm-908f833e57bfacb310248cd9ef1c0257360a0c4e.zip | |
[X86][SSE] combineTruncateWithSat - use truncateVectorWithPACK to chain PACKSS vXi32-vXi8 saturated truncation
We can use PACKSS to saturate each stage of the chain: PACKSSDW down to [-32768,32767] and then PACKSSWB to [-128,127].
PACKUS is a little trickier and will be handled in a separate patch.
llvm-svn: 325235
Diffstat (limited to 'llvm/lib/Target/X86/X86ISelLowering.cpp')
| -rw-r--r-- | llvm/lib/Target/X86/X86ISelLowering.cpp | 19 |
1 files changed, 12 insertions, 7 deletions
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp index 6a82de1d7fe..d9b7f192b51 100644 --- a/llvm/lib/Target/X86/X86ISelLowering.cpp +++ b/llvm/lib/Target/X86/X86ISelLowering.cpp @@ -34212,7 +34212,9 @@ static SDValue detectAVX512USatPattern(SDValue In, EVT VT, static SDValue combineTruncateWithSat(SDValue In, EVT VT, const SDLoc &DL, SelectionDAG &DAG, const X86Subtarget &Subtarget) { + EVT SVT = VT.getScalarType(); EVT InVT = In.getValueType(); + EVT InSVT = InVT.getScalarType(); const TargetLowering &TLI = DAG.getTargetLoweringInfo(); if (TLI.isTypeLegal(InVT) && TLI.isTypeLegal(VT) && isSATValidOnAVX512Subtarget(InVT, VT, Subtarget)) { @@ -34221,16 +34223,19 @@ static SDValue combineTruncateWithSat(SDValue In, EVT VT, const SDLoc &DL, if (auto USatVal = detectUSatPattern(In, VT)) return DAG.getNode(X86ISD::VTRUNCUS, DL, VT, USatVal); } - if (VT.isVector() && isPowerOf2_32(VT.getVectorNumElements()) && - ((VT.getScalarType() == MVT::i8 && InVT.getScalarType() == MVT::i16) || - (VT.getScalarType() == MVT::i16 && InVT.getScalarType() == MVT::i32))) { - if (auto SSatVal = detectSSatPattern(In, VT)) - return truncateVectorWithPACK(X86ISD::PACKSS, VT, SSatVal, DL, DAG, - Subtarget); - if (Subtarget.hasSSE41() || VT.getScalarType() == MVT::i8) + if (VT.isVector() && isPowerOf2_32(VT.getVectorNumElements())) { + if ((SVT == MVT::i8 || SVT == MVT::i16) && + (InSVT == MVT::i16 || InSVT == MVT::i32)) { + if (auto SSatVal = detectSSatPattern(In, VT)) + return truncateVectorWithPACK(X86ISD::PACKSS, VT, SSatVal, DL, DAG, + Subtarget); + } + if ((SVT == MVT::i8 && InSVT == MVT::i16) || + (SVT == MVT::i16 && InSVT == MVT::i32 && Subtarget.hasSSE41())) { if (auto USatVal = detectSSatPattern(In, VT, true)) return truncateVectorWithPACK(X86ISD::PACKUS, VT, USatVal, DL, DAG, Subtarget); + } } return SDValue(); } |

