From 63974b2144c87c962effdc0508c27643c8ad98b6 Mon Sep 17 00:00:00 2001 From: Chandler Carruth Date: Tue, 13 Dec 2011 01:56:10 +0000 Subject: [PATCH] Initial CodeGen support for CTTZ/CTLZ where a zero input produces an undefined result. This adds new ISD nodes for the new semantics, selecting them when the LLVM intrinsic indicates that the undef behavior is desired. The new nodes expand trivially to the old nodes, so targets don't actually need to do anything to support these new nodes besides indicating that they should be expanded. I've done this for all the operand types that I could figure out for all the targets. Owners of various targets, please review and let me know if any of these are incorrect. Note that the expand behavior is *conservatively correct*, and exactly matches LLVM's current behavior with these operations. Ideally this patch will not change behavior in any way. For example the regtest suite finds the exact same instruction sequences coming out of the code generator. That's why there are no new tests here -- all of this is being exercised by the existing test suite. Thanks to Duncan Sands for reviewing the various bits of this patch and helping me get the wrinkles ironed out with expanding for each target. Also thanks to Chris for clarifying through all the discussions that this is indeed the approach he was looking for. That said, there are likely still rough spots. Further review much appreciated. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@146466 91177308-0d34-0410-b5e6-96231b3b80d8 --- include/llvm/CodeGen/ISDOpcodes.h | 3 ++ include/llvm/Target/TargetSelectionDAG.td | 2 ++ lib/CodeGen/SelectionDAG/DAGCombiner.cpp | 24 +++++++++++++++ lib/CodeGen/SelectionDAG/LegalizeDAG.cpp | 18 +++++++++-- .../SelectionDAG/LegalizeIntegerTypes.cpp | 30 +++++++++++-------- .../SelectionDAG/LegalizeVectorOps.cpp | 4 ++- .../SelectionDAG/LegalizeVectorTypes.cpp | 4 ++- lib/CodeGen/SelectionDAG/SelectionDAG.cpp | 14 ++++++--- .../SelectionDAG/SelectionDAGBuilder.cpp | 8 +++-- lib/Target/ARM/ARMISelLowering.cpp | 4 +++ lib/Target/CellSPU/SPUISelLowering.cpp | 10 +++++++ lib/Target/Hexagon/HexagonISelLowering.cpp | 2 ++ lib/Target/MBlaze/MBlazeISelLowering.cpp | 2 ++ lib/Target/MSP430/MSP430ISelLowering.cpp | 4 +++ lib/Target/Mips/MipsISelLowering.cpp | 4 +++ lib/Target/PowerPC/PPCISelLowering.cpp | 6 ++++ lib/Target/Sparc/SparcISelLowering.cpp | 2 ++ lib/Target/X86/X86ISelLowering.cpp | 10 +++++++ lib/Target/XCore/XCoreISelLowering.cpp | 2 ++ 19 files changed, 130 insertions(+), 23 deletions(-) diff --git a/include/llvm/CodeGen/ISDOpcodes.h b/include/llvm/CodeGen/ISDOpcodes.h index f90b07e5dfb..cb990edac18 100644 --- a/include/llvm/CodeGen/ISDOpcodes.h +++ b/include/llvm/CodeGen/ISDOpcodes.h @@ -312,6 +312,9 @@ namespace ISD { /// Byte Swap and Counting operators. BSWAP, CTTZ, CTLZ, CTPOP, + /// Bit counting operators with an undefined result for zero inputs. + CTTZ_ZERO_UNDEF, CTLZ_ZERO_UNDEF, + // Select(COND, TRUEVAL, FALSEVAL). If the type of the boolean COND is not // i1 then the high bits must conform to getBooleanContents. SELECT, diff --git a/include/llvm/Target/TargetSelectionDAG.td b/include/llvm/Target/TargetSelectionDAG.td index 612635ea746..3288dd43ab7 100644 --- a/include/llvm/Target/TargetSelectionDAG.td +++ b/include/llvm/Target/TargetSelectionDAG.td @@ -352,6 +352,8 @@ def bswap : SDNode<"ISD::BSWAP" , SDTIntUnaryOp>; def ctlz : SDNode<"ISD::CTLZ" , SDTIntUnaryOp>; def cttz : SDNode<"ISD::CTTZ" , SDTIntUnaryOp>; def ctpop : SDNode<"ISD::CTPOP" , SDTIntUnaryOp>; +def ctlz_zero_undef : SDNode<"ISD::CTLZ_ZERO_UNDEF", SDTIntUnaryOp>; +def cttz_zero_undef : SDNode<"ISD::CTTZ_ZERO_UNDEF", SDTIntUnaryOp>; def sext : SDNode<"ISD::SIGN_EXTEND", SDTIntExtendOp>; def zext : SDNode<"ISD::ZERO_EXTEND", SDTIntExtendOp>; def anyext : SDNode<"ISD::ANY_EXTEND" , SDTIntExtendOp>; diff --git a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp index 9f628e95403..80cf0a8ff1d 100644 --- a/lib/CodeGen/SelectionDAG/DAGCombiner.cpp +++ b/lib/CodeGen/SelectionDAG/DAGCombiner.cpp @@ -180,7 +180,9 @@ namespace { SDValue visitSRA(SDNode *N); SDValue visitSRL(SDNode *N); SDValue visitCTLZ(SDNode *N); + SDValue visitCTLZ_ZERO_UNDEF(SDNode *N); SDValue visitCTTZ(SDNode *N); + SDValue visitCTTZ_ZERO_UNDEF(SDNode *N); SDValue visitCTPOP(SDNode *N); SDValue visitSELECT(SDNode *N); SDValue visitSELECT_CC(SDNode *N); @@ -1078,7 +1080,9 @@ SDValue DAGCombiner::visit(SDNode *N) { case ISD::SRA: return visitSRA(N); case ISD::SRL: return visitSRL(N); case ISD::CTLZ: return visitCTLZ(N); + case ISD::CTLZ_ZERO_UNDEF: return visitCTLZ_ZERO_UNDEF(N); case ISD::CTTZ: return visitCTTZ(N); + case ISD::CTTZ_ZERO_UNDEF: return visitCTTZ_ZERO_UNDEF(N); case ISD::CTPOP: return visitCTPOP(N); case ISD::SELECT: return visitSELECT(N); case ISD::SELECT_CC: return visitSELECT_CC(N); @@ -3717,6 +3721,16 @@ SDValue DAGCombiner::visitCTLZ(SDNode *N) { return SDValue(); } +SDValue DAGCombiner::visitCTLZ_ZERO_UNDEF(SDNode *N) { + SDValue N0 = N->getOperand(0); + EVT VT = N->getValueType(0); + + // fold (ctlz_zero_undef c1) -> c2 + if (isa(N0)) + return DAG.getNode(ISD::CTLZ_ZERO_UNDEF, N->getDebugLoc(), VT, N0); + return SDValue(); +} + SDValue DAGCombiner::visitCTTZ(SDNode *N) { SDValue N0 = N->getOperand(0); EVT VT = N->getValueType(0); @@ -3727,6 +3741,16 @@ SDValue DAGCombiner::visitCTTZ(SDNode *N) { return SDValue(); } +SDValue DAGCombiner::visitCTTZ_ZERO_UNDEF(SDNode *N) { + SDValue N0 = N->getOperand(0); + EVT VT = N->getValueType(0); + + // fold (cttz_zero_undef c1) -> c2 + if (isa(N0)) + return DAG.getNode(ISD::CTTZ_ZERO_UNDEF, N->getDebugLoc(), VT, N0); + return SDValue(); +} + SDValue DAGCombiner::visitCTPOP(SDNode *N) { SDValue N0 = N->getOperand(0); EVT VT = N->getValueType(0); diff --git a/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp b/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp index 0230a85bd52..75f57611126 100644 --- a/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp +++ b/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp @@ -2382,6 +2382,9 @@ SDValue SelectionDAGLegalize::ExpandBitCount(unsigned Opc, SDValue Op, return Op; } + case ISD::CTLZ_ZERO_UNDEF: + // This trivially expands to CTLZ. + return DAG.getNode(ISD::CTLZ, dl, Op.getValueType(), Op); case ISD::CTLZ: { // for now, we do this: // x = x | (x >> 1); @@ -2403,6 +2406,9 @@ SDValue SelectionDAGLegalize::ExpandBitCount(unsigned Opc, SDValue Op, Op = DAG.getNOT(dl, Op, VT); return DAG.getNode(ISD::CTPOP, dl, VT, Op); } + case ISD::CTTZ_ZERO_UNDEF: + // This trivially expands to CTTZ. + return DAG.getNode(ISD::CTTZ, dl, Op.getValueType(), Op); case ISD::CTTZ: { // for now, we use: { return popcount(~x & (x - 1)); } // unless the target has ctlz but not ctpop, in which case we use: @@ -2517,7 +2523,9 @@ void SelectionDAGLegalize::ExpandNode(SDNode *Node) { switch (Node->getOpcode()) { case ISD::CTPOP: case ISD::CTLZ: + case ISD::CTLZ_ZERO_UNDEF: case ISD::CTTZ: + case ISD::CTTZ_ZERO_UNDEF: Tmp1 = ExpandBitCount(Node->getOpcode(), Node->getOperand(0), dl); Results.push_back(Tmp1); break; @@ -3419,20 +3427,24 @@ void SelectionDAGLegalize::PromoteNode(SDNode *Node) { SDValue Tmp1, Tmp2, Tmp3; switch (Node->getOpcode()) { case ISD::CTTZ: + case ISD::CTTZ_ZERO_UNDEF: case ISD::CTLZ: + case ISD::CTLZ_ZERO_UNDEF: case ISD::CTPOP: // Zero extend the argument. Tmp1 = DAG.getNode(ISD::ZERO_EXTEND, dl, NVT, Node->getOperand(0)); - // Perform the larger operation. + // Perform the larger operation. For CTPOP and CTTZ_ZERO_UNDEF, this is + // already the correct result. Tmp1 = DAG.getNode(Node->getOpcode(), dl, NVT, Tmp1); if (Node->getOpcode() == ISD::CTTZ) { - //if Tmp1 == sizeinbits(NVT) then Tmp1 = sizeinbits(Old VT) + // FIXME: This should set a bit in the zero extended value instead. Tmp2 = DAG.getSetCC(dl, TLI.getSetCCResultType(NVT), Tmp1, DAG.getConstant(NVT.getSizeInBits(), NVT), ISD::SETEQ); Tmp1 = DAG.getNode(ISD::SELECT, dl, NVT, Tmp2, DAG.getConstant(OVT.getSizeInBits(), NVT), Tmp1); - } else if (Node->getOpcode() == ISD::CTLZ) { + } else if (Node->getOpcode() == ISD::CTLZ || + Node->getOpcode() == ISD::CTLZ_ZERO_UNDEF) { // Tmp1 = Tmp1 - (sizeinbits(NVT) - sizeinbits(Old VT)) Tmp1 = DAG.getNode(ISD::SUB, dl, NVT, Tmp1, DAG.getConstant(NVT.getSizeInBits() - diff --git a/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp b/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp index a48ee7efdc8..1c02c4f21c8 100644 --- a/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp +++ b/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp @@ -56,8 +56,10 @@ void DAGTypeLegalizer::PromoteIntegerResult(SDNode *N, unsigned ResNo) { case ISD::Constant: Res = PromoteIntRes_Constant(N); break; case ISD::CONVERT_RNDSAT: Res = PromoteIntRes_CONVERT_RNDSAT(N); break; + case ISD::CTLZ_ZERO_UNDEF: case ISD::CTLZ: Res = PromoteIntRes_CTLZ(N); break; case ISD::CTPOP: Res = PromoteIntRes_CTPOP(N); break; + case ISD::CTTZ_ZERO_UNDEF: case ISD::CTTZ: Res = PromoteIntRes_CTTZ(N); break; case ISD::EXTRACT_VECTOR_ELT: Res = PromoteIntRes_EXTRACT_VECTOR_ELT(N); break; @@ -311,7 +313,7 @@ SDValue DAGTypeLegalizer::PromoteIntRes_CTLZ(SDNode *N) { DebugLoc dl = N->getDebugLoc(); EVT OVT = N->getValueType(0); EVT NVT = Op.getValueType(); - Op = DAG.getNode(ISD::CTLZ, dl, NVT, Op); + Op = DAG.getNode(N->getOpcode(), dl, NVT, Op); // Subtract off the extra leading bits in the bigger type. return DAG.getNode(ISD::SUB, dl, NVT, Op, DAG.getConstant(NVT.getSizeInBits() - @@ -329,13 +331,15 @@ SDValue DAGTypeLegalizer::PromoteIntRes_CTTZ(SDNode *N) { EVT OVT = N->getValueType(0); EVT NVT = Op.getValueType(); DebugLoc dl = N->getDebugLoc(); - // The count is the same in the promoted type except if the original - // value was zero. This can be handled by setting the bit just off - // the top of the original type. - APInt TopBit(NVT.getSizeInBits(), 0); - TopBit.setBit(OVT.getSizeInBits()); - Op = DAG.getNode(ISD::OR, dl, NVT, Op, DAG.getConstant(TopBit, NVT)); - return DAG.getNode(ISD::CTTZ, dl, NVT, Op); + if (N->getOpcode() == ISD::CTTZ) { + // The count is the same in the promoted type except if the original + // value was zero. This can be handled by setting the bit just off + // the top of the original type. + APInt TopBit(NVT.getSizeInBits(), 0); + TopBit.setBit(OVT.getSizeInBits()); + Op = DAG.getNode(ISD::OR, dl, NVT, Op, DAG.getConstant(TopBit, NVT)); + } + return DAG.getNode(N->getOpcode(), dl, NVT, Op); } SDValue DAGTypeLegalizer::PromoteIntRes_EXTRACT_VECTOR_ELT(SDNode *N) { @@ -1097,8 +1101,10 @@ void DAGTypeLegalizer::ExpandIntegerResult(SDNode *N, unsigned ResNo) { case ISD::AssertZext: ExpandIntRes_AssertZext(N, Lo, Hi); break; case ISD::BSWAP: ExpandIntRes_BSWAP(N, Lo, Hi); break; case ISD::Constant: ExpandIntRes_Constant(N, Lo, Hi); break; + case ISD::CTLZ_ZERO_UNDEF: case ISD::CTLZ: ExpandIntRes_CTLZ(N, Lo, Hi); break; case ISD::CTPOP: ExpandIntRes_CTPOP(N, Lo, Hi); break; + case ISD::CTTZ_ZERO_UNDEF: case ISD::CTTZ: ExpandIntRes_CTTZ(N, Lo, Hi); break; case ISD::FP_TO_SINT: ExpandIntRes_FP_TO_SINT(N, Lo, Hi); break; case ISD::FP_TO_UINT: ExpandIntRes_FP_TO_UINT(N, Lo, Hi); break; @@ -1701,8 +1707,8 @@ void DAGTypeLegalizer::ExpandIntRes_CTLZ(SDNode *N, SDValue HiNotZero = DAG.getSetCC(dl, TLI.getSetCCResultType(NVT), Hi, DAG.getConstant(0, NVT), ISD::SETNE); - SDValue LoLZ = DAG.getNode(ISD::CTLZ, dl, NVT, Lo); - SDValue HiLZ = DAG.getNode(ISD::CTLZ, dl, NVT, Hi); + SDValue LoLZ = DAG.getNode(N->getOpcode(), dl, NVT, Lo); + SDValue HiLZ = DAG.getNode(ISD::CTLZ_ZERO_UNDEF, dl, NVT, Hi); Lo = DAG.getNode(ISD::SELECT, dl, NVT, HiNotZero, HiLZ, DAG.getNode(ISD::ADD, dl, NVT, LoLZ, @@ -1731,8 +1737,8 @@ void DAGTypeLegalizer::ExpandIntRes_CTTZ(SDNode *N, SDValue LoNotZero = DAG.getSetCC(dl, TLI.getSetCCResultType(NVT), Lo, DAG.getConstant(0, NVT), ISD::SETNE); - SDValue LoLZ = DAG.getNode(ISD::CTTZ, dl, NVT, Lo); - SDValue HiLZ = DAG.getNode(ISD::CTTZ, dl, NVT, Hi); + SDValue LoLZ = DAG.getNode(ISD::CTTZ_ZERO_UNDEF, dl, NVT, Lo); + SDValue HiLZ = DAG.getNode(N->getOpcode(), dl, NVT, Hi); Lo = DAG.getNode(ISD::SELECT, dl, NVT, LoNotZero, LoLZ, DAG.getNode(ISD::ADD, dl, NVT, HiLZ, diff --git a/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp b/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp index 4e02b90aad3..4696c0d7546 100644 --- a/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp +++ b/lib/CodeGen/SelectionDAG/LegalizeVectorOps.cpp @@ -185,8 +185,10 @@ SDValue VectorLegalizer::LegalizeOp(SDValue Op) { case ISD::SRL: case ISD::ROTL: case ISD::ROTR: - case ISD::CTTZ: case ISD::CTLZ: + case ISD::CTTZ: + case ISD::CTLZ_ZERO_UNDEF: + case ISD::CTTZ_ZERO_UNDEF: case ISD::CTPOP: case ISD::SELECT: case ISD::VSELECT: diff --git a/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp b/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp index ad83565e15e..7ca0d1e5c0f 100644 --- a/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp +++ b/lib/CodeGen/SelectionDAG/LegalizeVectorTypes.cpp @@ -441,8 +441,10 @@ void DAGTypeLegalizer::SplitVectorResult(SDNode *N, unsigned ResNo) { case ISD::ANY_EXTEND: case ISD::CONVERT_RNDSAT: case ISD::CTLZ: - case ISD::CTPOP: case ISD::CTTZ: + case ISD::CTLZ_ZERO_UNDEF: + case ISD::CTTZ_ZERO_UNDEF: + case ISD::CTPOP: case ISD::FABS: case ISD::FCEIL: case ISD::FCOS: diff --git a/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/lib/CodeGen/SelectionDAG/SelectionDAG.cpp index 4487a9a5d26..0b507b42719 100644 --- a/lib/CodeGen/SelectionDAG/SelectionDAG.cpp +++ b/lib/CodeGen/SelectionDAG/SelectionDAG.cpp @@ -1856,7 +1856,9 @@ void SelectionDAG::ComputeMaskedBits(SDValue Op, const APInt &Mask, return; } case ISD::CTTZ: + case ISD::CTTZ_ZERO_UNDEF: case ISD::CTLZ: + case ISD::CTLZ_ZERO_UNDEF: case ISD::CTPOP: { unsigned LowBits = Log2_32(BitWidth)+1; KnownZero = APInt::getHighBitsSet(BitWidth, BitWidth - LowBits); @@ -2429,8 +2431,10 @@ SDValue SelectionDAG::getNode(unsigned Opcode, DebugLoc DL, case ISD::CTPOP: return getConstant(Val.countPopulation(), VT); case ISD::CTLZ: + case ISD::CTLZ_ZERO_UNDEF: return getConstant(Val.countLeadingZeros(), VT); case ISD::CTTZ: + case ISD::CTTZ_ZERO_UNDEF: return getConstant(Val.countTrailingZeros(), VT); } } @@ -6111,10 +6115,12 @@ std::string SDNode::getOperationName(const SelectionDAG *G) const { case ISD::TRAP: return "trap"; // Bit manipulation - case ISD::BSWAP: return "bswap"; - case ISD::CTPOP: return "ctpop"; - case ISD::CTTZ: return "cttz"; - case ISD::CTLZ: return "ctlz"; + case ISD::BSWAP: return "bswap"; + case ISD::CTPOP: return "ctpop"; + case ISD::CTTZ: return "cttz"; + case ISD::CTTZ_ZERO_UNDEF: return "cttz_zero_undef"; + case ISD::CTLZ: return "ctlz"; + case ISD::CTLZ_ZERO_UNDEF: return "ctlz_zero_undef"; // Trampolines case ISD::INIT_TRAMPOLINE: return "init_trampoline"; diff --git a/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp b/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp index 760c4e5b07d..d10fa4686cd 100644 --- a/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp +++ b/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp @@ -4948,14 +4948,18 @@ SelectionDAGBuilder::visitIntrinsicCall(const CallInst &I, unsigned Intrinsic) { return 0; case Intrinsic::cttz: { SDValue Arg = getValue(I.getArgOperand(0)); + ConstantInt *CI = cast(I.getArgOperand(1)); EVT Ty = Arg.getValueType(); - setValue(&I, DAG.getNode(ISD::CTTZ, dl, Ty, Arg)); + setValue(&I, DAG.getNode(CI->isZero() ? ISD::CTTZ : ISD::CTTZ_ZERO_UNDEF, + dl, Ty, Arg)); return 0; } case Intrinsic::ctlz: { SDValue Arg = getValue(I.getArgOperand(0)); + ConstantInt *CI = cast(I.getArgOperand(1)); EVT Ty = Arg.getValueType(); - setValue(&I, DAG.getNode(ISD::CTLZ, dl, Ty, Arg)); + setValue(&I, DAG.getNode(CI->isZero() ? ISD::CTLZ : ISD::CTLZ_ZERO_UNDEF, + dl, Ty, Arg)); return 0; } case Intrinsic::ctpop: { diff --git a/lib/Target/ARM/ARMISelLowering.cpp b/lib/Target/ARM/ARMISelLowering.cpp index 229d8cf1389..857eb7ade28 100644 --- a/lib/Target/ARM/ARMISelLowering.cpp +++ b/lib/Target/ARM/ARMISelLowering.cpp @@ -609,6 +609,10 @@ ARMTargetLowering::ARMTargetLowering(TargetMachine &TM) if (!Subtarget->hasV5TOps() || Subtarget->isThumb1Only()) setOperationAction(ISD::CTLZ, MVT::i32, Expand); + // These just redirect to CTTZ and CTLZ on ARM. + setOperationAction(ISD::CTTZ_ZERO_UNDEF , MVT::i32 , Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF , MVT::i32 , Expand); + // Only ARMv6 has BSWAP. if (!Subtarget->hasV6Ops()) setOperationAction(ISD::BSWAP, MVT::i32, Expand); diff --git a/lib/Target/CellSPU/SPUISelLowering.cpp b/lib/Target/CellSPU/SPUISelLowering.cpp index d58e49b8683..dc0d5a60254 100644 --- a/lib/Target/CellSPU/SPUISelLowering.cpp +++ b/lib/Target/CellSPU/SPUISelLowering.cpp @@ -296,12 +296,22 @@ SPUTargetLowering::SPUTargetLowering(SPUTargetMachine &TM) setOperationAction(ISD::CTTZ , MVT::i32, Expand); setOperationAction(ISD::CTTZ , MVT::i64, Expand); setOperationAction(ISD::CTTZ , MVT::i128, Expand); + setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::i8, Expand); + setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::i16, Expand); + setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::i32, Expand); + setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::i64, Expand); + setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::i128, Expand); setOperationAction(ISD::CTLZ , MVT::i8, Promote); setOperationAction(ISD::CTLZ , MVT::i16, Promote); setOperationAction(ISD::CTLZ , MVT::i32, Legal); setOperationAction(ISD::CTLZ , MVT::i64, Expand); setOperationAction(ISD::CTLZ , MVT::i128, Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i8, Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i16, Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i32, Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i64, Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i128, Expand); // SPU has a version of select that implements (a&~c)|(b&c), just like // select ought to work: diff --git a/lib/Target/Hexagon/HexagonISelLowering.cpp b/lib/Target/Hexagon/HexagonISelLowering.cpp index 8d2d3fd8f6a..0ac3cf02d1a 100644 --- a/lib/Target/Hexagon/HexagonISelLowering.cpp +++ b/lib/Target/Hexagon/HexagonISelLowering.cpp @@ -1250,7 +1250,9 @@ HexagonTargetLowering::HexagonTargetLowering(HexagonTargetMachine setOperationAction(ISD::FREM , MVT::f32, Expand); setOperationAction(ISD::CTPOP, MVT::i32, Expand); setOperationAction(ISD::CTTZ , MVT::i32, Expand); + setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::i32, Expand); setOperationAction(ISD::CTLZ , MVT::i32, Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i32, Expand); setOperationAction(ISD::ROTL , MVT::i32, Expand); setOperationAction(ISD::ROTR , MVT::i32, Expand); setOperationAction(ISD::BSWAP, MVT::i32, Expand); diff --git a/lib/Target/MBlaze/MBlazeISelLowering.cpp b/lib/Target/MBlaze/MBlazeISelLowering.cpp index 148d906a24c..0002174f2ba 100644 --- a/lib/Target/MBlaze/MBlazeISelLowering.cpp +++ b/lib/Target/MBlaze/MBlazeISelLowering.cpp @@ -167,7 +167,9 @@ MBlazeTargetLowering::MBlazeTargetLowering(MBlazeTargetMachine &TM) setOperationAction(ISD::SRA_PARTS, MVT::i32, Expand); setOperationAction(ISD::SRL_PARTS, MVT::i32, Expand); setOperationAction(ISD::CTLZ, MVT::i32, Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i32, Expand); setOperationAction(ISD::CTTZ, MVT::i32, Expand); + setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::i32, Expand); setOperationAction(ISD::CTPOP, MVT::i32, Expand); setOperationAction(ISD::BSWAP, MVT::i32, Expand); diff --git a/lib/Target/MSP430/MSP430ISelLowering.cpp b/lib/Target/MSP430/MSP430ISelLowering.cpp index 5c9413767e2..884d69b50b4 100644 --- a/lib/Target/MSP430/MSP430ISelLowering.cpp +++ b/lib/Target/MSP430/MSP430ISelLowering.cpp @@ -122,8 +122,12 @@ MSP430TargetLowering::MSP430TargetLowering(MSP430TargetMachine &tm) : setOperationAction(ISD::CTTZ, MVT::i8, Expand); setOperationAction(ISD::CTTZ, MVT::i16, Expand); + setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::i8, Expand); + setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::i16, Expand); setOperationAction(ISD::CTLZ, MVT::i8, Expand); setOperationAction(ISD::CTLZ, MVT::i16, Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i8, Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i16, Expand); setOperationAction(ISD::CTPOP, MVT::i8, Expand); setOperationAction(ISD::CTPOP, MVT::i16, Expand); diff --git a/lib/Target/Mips/MipsISelLowering.cpp b/lib/Target/Mips/MipsISelLowering.cpp index b5f64403df6..5da9a10fd4f 100644 --- a/lib/Target/Mips/MipsISelLowering.cpp +++ b/lib/Target/Mips/MipsISelLowering.cpp @@ -156,6 +156,10 @@ MipsTargetLowering(MipsTargetMachine &TM) setOperationAction(ISD::SIGN_EXTEND_INREG, MVT::i1, Expand); setOperationAction(ISD::CTPOP, MVT::i32, Expand); setOperationAction(ISD::CTTZ, MVT::i32, Expand); + setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::i32, Expand); + setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::i64, Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i32, Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i64, Expand); setOperationAction(ISD::ROTL, MVT::i32, Expand); setOperationAction(ISD::ROTL, MVT::i64, Expand); diff --git a/lib/Target/PowerPC/PPCISelLowering.cpp b/lib/Target/PowerPC/PPCISelLowering.cpp index 28fed5ad24a..f3a3d178af6 100644 --- a/lib/Target/PowerPC/PPCISelLowering.cpp +++ b/lib/Target/PowerPC/PPCISelLowering.cpp @@ -153,9 +153,13 @@ PPCTargetLowering::PPCTargetLowering(PPCTargetMachine &TM) setOperationAction(ISD::BSWAP, MVT::i32 , Expand); setOperationAction(ISD::CTPOP, MVT::i32 , Expand); setOperationAction(ISD::CTTZ , MVT::i32 , Expand); + setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::i32, Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i32, Expand); setOperationAction(ISD::BSWAP, MVT::i64 , Expand); setOperationAction(ISD::CTPOP, MVT::i64 , Expand); setOperationAction(ISD::CTTZ , MVT::i64 , Expand); + setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::i64, Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i64, Expand); // PowerPC does not have ROTR setOperationAction(ISD::ROTR, MVT::i32 , Expand); @@ -339,7 +343,9 @@ PPCTargetLowering::PPCTargetLowering(PPCTargetMachine &TM) setOperationAction(ISD::FPOW, VT, Expand); setOperationAction(ISD::CTPOP, VT, Expand); setOperationAction(ISD::CTLZ, VT, Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF, VT, Expand); setOperationAction(ISD::CTTZ, VT, Expand); + setOperationAction(ISD::CTTZ_ZERO_UNDEF, VT, Expand); } // We can custom expand all VECTOR_SHUFFLEs to VPERM, others we can handle diff --git a/lib/Target/Sparc/SparcISelLowering.cpp b/lib/Target/Sparc/SparcISelLowering.cpp index 25104d1130e..3608d3b5428 100644 --- a/lib/Target/Sparc/SparcISelLowering.cpp +++ b/lib/Target/Sparc/SparcISelLowering.cpp @@ -763,7 +763,9 @@ SparcTargetLowering::SparcTargetLowering(TargetMachine &TM) setOperationAction(ISD::FMA , MVT::f32, Expand); setOperationAction(ISD::CTPOP, MVT::i32, Expand); setOperationAction(ISD::CTTZ , MVT::i32, Expand); + setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::i32, Expand); setOperationAction(ISD::CTLZ , MVT::i32, Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i32, Expand); setOperationAction(ISD::ROTL , MVT::i32, Expand); setOperationAction(ISD::ROTR , MVT::i32, Expand); setOperationAction(ISD::BSWAP, MVT::i32, Expand); diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp index a7bdb999336..e38de9c0d22 100644 --- a/lib/Target/X86/X86ISelLowering.cpp +++ b/lib/Target/X86/X86ISelLowering.cpp @@ -378,6 +378,10 @@ X86TargetLowering::X86TargetLowering(X86TargetMachine &TM) setOperationAction(ISD::FREM , MVT::f80 , Expand); setOperationAction(ISD::FLT_ROUNDS_ , MVT::i32 , Custom); + setOperationAction(ISD::CTTZ_ZERO_UNDEF , MVT::i8 , Expand); + setOperationAction(ISD::CTTZ_ZERO_UNDEF , MVT::i16 , Expand); + setOperationAction(ISD::CTTZ_ZERO_UNDEF , MVT::i32 , Expand); + setOperationAction(ISD::CTTZ_ZERO_UNDEF , MVT::i64 , Expand); if (Subtarget->hasBMI()) { setOperationAction(ISD::CTTZ , MVT::i8 , Promote); } else { @@ -388,6 +392,10 @@ X86TargetLowering::X86TargetLowering(X86TargetMachine &TM) setOperationAction(ISD::CTTZ , MVT::i64 , Custom); } + setOperationAction(ISD::CTLZ_ZERO_UNDEF , MVT::i8 , Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF , MVT::i16 , Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF , MVT::i32 , Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF , MVT::i64 , Expand); if (Subtarget->hasLZCNT()) { setOperationAction(ISD::CTLZ , MVT::i8 , Promote); } else { @@ -719,7 +727,9 @@ X86TargetLowering::X86TargetLowering(X86TargetMachine &TM) setOperationAction(ISD::FPOW, (MVT::SimpleValueType)VT, Expand); setOperationAction(ISD::CTPOP, (MVT::SimpleValueType)VT, Expand); setOperationAction(ISD::CTTZ, (MVT::SimpleValueType)VT, Expand); + setOperationAction(ISD::CTTZ_ZERO_UNDEF, (MVT::SimpleValueType)VT, Expand); setOperationAction(ISD::CTLZ, (MVT::SimpleValueType)VT, Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF, (MVT::SimpleValueType)VT, Expand); setOperationAction(ISD::SHL, (MVT::SimpleValueType)VT, Expand); setOperationAction(ISD::SRA, (MVT::SimpleValueType)VT, Expand); setOperationAction(ISD::SRL, (MVT::SimpleValueType)VT, Expand); diff --git a/lib/Target/XCore/XCoreISelLowering.cpp b/lib/Target/XCore/XCoreISelLowering.cpp index d791daae226..c5c668e66c1 100644 --- a/lib/Target/XCore/XCoreISelLowering.cpp +++ b/lib/Target/XCore/XCoreISelLowering.cpp @@ -109,6 +109,8 @@ XCoreTargetLowering::XCoreTargetLowering(XCoreTargetMachine &XTM) setOperationAction(ISD::CTPOP, MVT::i32, Expand); setOperationAction(ISD::ROTL , MVT::i32, Expand); setOperationAction(ISD::ROTR , MVT::i32, Expand); + setOperationAction(ISD::CTTZ_ZERO_UNDEF, MVT::i32, Expand); + setOperationAction(ISD::CTLZ_ZERO_UNDEF, MVT::i32, Expand); setOperationAction(ISD::TRAP, MVT::Other, Legal);