llvm-project/llvm/lib/Target/X86/X86ScheduleSLM.td

491 lines
22 KiB
TableGen
Raw Normal View History

//=- X86ScheduleSLM.td - X86 Silvermont Scheduling -----------*- tablegen -*-=//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file defines the machine model for Intel Silvermont to support
// instruction scheduling and other instruction cost heuristics.
//
//===----------------------------------------------------------------------===//
def SLMModel : SchedMachineModel {
// All x86 instructions are modeled as a single micro-op, and SLM can decode 2
// instructions per cycle.
let IssueWidth = 2;
let MicroOpBufferSize = 32; // Based on the reorder buffer.
let LoadLatency = 3;
let MispredictPenalty = 10;
let PostRAScheduler = 1;
// For small loops, expand by a small factor to hide the backedge cost.
let LoopMicroOpBufferSize = 10;
// FIXME: SSE4 is unimplemented. This flag is set to allow
// the scheduler to assign a default model to unrecognized opcodes.
let CompleteModel = 0;
}
let SchedModel = SLMModel in {
// Silvermont has 5 reservation stations for micro-ops
def SLM_IEC_RSV0 : ProcResource<1>;
def SLM_IEC_RSV1 : ProcResource<1>;
def SLM_FPC_RSV0 : ProcResource<1> { let BufferSize = 1; }
def SLM_FPC_RSV1 : ProcResource<1> { let BufferSize = 1; }
def SLM_MEC_RSV : ProcResource<1>;
// Many micro-ops are capable of issuing on multiple ports.
def SLM_IEC_RSV01 : ProcResGroup<[SLM_IEC_RSV0, SLM_IEC_RSV1]>;
def SLM_FPC_RSV01 : ProcResGroup<[SLM_FPC_RSV0, SLM_FPC_RSV1]>;
def SLMDivider : ProcResource<1>;
def SLMFPMultiplier : ProcResource<1>;
def SLMFPDivider : ProcResource<1>;
// Loads are 3 cycles, so ReadAfterLd registers needn't be available until 3
// cycles after the memory operand.
def : ReadAdvance<ReadAfterLd, 3>;
// Many SchedWrites are defined in pairs with and without a folded load.
// Instructions with folded loads are usually micro-fused, so they only appear
// as two micro-ops when queued in the reservation station.
// This multiclass defines the resource usage for variants with and without
// folded loads.
multiclass SLMWriteResPair<X86FoldableSchedWrite SchedRW,
list<ProcResourceKind> ExePorts,
int Lat, list<int> Res = [1], int UOps = 1,
int LoadLat = 3> {
// Register variant is using a single cycle on ExePort.
def : WriteRes<SchedRW, ExePorts> {
let Latency = Lat;
let ResourceCycles = Res;
let NumMicroOps = UOps;
}
// Memory variant also uses a cycle on MEC_RSV and adds LoadLat cycles to
// the latency (default = 3).
def : WriteRes<SchedRW.Folded, !listconcat([SLM_MEC_RSV], ExePorts)> {
let Latency = !add(Lat, LoadLat);
let ResourceCycles = !listconcat([1], Res);
let NumMicroOps = UOps;
}
}
// A folded store needs a cycle on MEC_RSV for the store data, but it does not
// need an extra port cycle to recompute the address.
def : WriteRes<WriteRMW, [SLM_MEC_RSV]>;
def : WriteRes<WriteStore, [SLM_IEC_RSV01, SLM_MEC_RSV]>;
def : WriteRes<WriteStoreNT, [SLM_IEC_RSV01, SLM_MEC_RSV]>;
def : WriteRes<WriteLoad, [SLM_MEC_RSV]> { let Latency = 3; }
def : WriteRes<WriteMove, [SLM_IEC_RSV01]>;
def : WriteRes<WriteZero, []>;
// Load/store MXCSR.
// FIXME: These are probably wrong. They are copy pasted from WriteStore/Load.
def : WriteRes<WriteSTMXCSR, [SLM_IEC_RSV01, SLM_MEC_RSV]>;
def : WriteRes<WriteLDMXCSR, [SLM_MEC_RSV]> { let Latency = 3; }
// Treat misc copies as a move.
def : InstRW<[WriteMove], (instrs COPY)>;
defm : SLMWriteResPair<WriteALU, [SLM_IEC_RSV01], 1>;
defm : SLMWriteResPair<WriteADC, [SLM_IEC_RSV01], 1>;
defm : SLMWriteResPair<WriteIMul, [SLM_IEC_RSV1], 3>;
defm : SLMWriteResPair<WriteIMul64, [SLM_IEC_RSV1], 3>;
defm : X86WriteRes<WriteBSWAP32, [SLM_IEC_RSV01], 1, [1], 1>;
defm : X86WriteRes<WriteBSWAP64, [SLM_IEC_RSV01], 1, [1], 1>;
defm : X86WriteRes<WriteCMPXCHG, [SLM_IEC_RSV01], 1, [1], 1>;
defm : X86WriteRes<WriteCMPXCHGRMW, [SLM_IEC_RSV01, SLM_MEC_RSV], 4, [1, 2], 2>;
defm : X86WriteRes<WriteXCHG, [SLM_IEC_RSV01], 1, [1], 1>;
defm : SLMWriteResPair<WriteShift, [SLM_IEC_RSV0], 1>;
defm : X86WriteRes<WriteSHDrri, [SLM_IEC_RSV0], 1, [1], 1>;
defm : X86WriteRes<WriteSHDrrcl,[SLM_IEC_RSV0], 1, [1], 1>;
defm : X86WriteRes<WriteSHDmri, [SLM_MEC_RSV, SLM_IEC_RSV0], 4, [2, 1], 2>;
defm : X86WriteRes<WriteSHDmrcl,[SLM_MEC_RSV, SLM_IEC_RSV0], 4, [2, 1], 2>;
defm : SLMWriteResPair<WriteJump, [SLM_IEC_RSV1], 1>;
defm : SLMWriteResPair<WriteCRC32, [SLM_IEC_RSV1], 3>;
defm : SLMWriteResPair<WriteCMOV, [SLM_IEC_RSV01], 2, [2]>;
defm : SLMWriteResPair<WriteCMOV2, [SLM_IEC_RSV01], 2, [2]>;
defm : X86WriteRes<WriteFCMOV, [SLM_FPC_RSV1], 3, [1], 1>; // x87 conditional move.
def : WriteRes<WriteSETCC, [SLM_IEC_RSV01]>;
def : WriteRes<WriteSETCCStore, [SLM_IEC_RSV01, SLM_MEC_RSV]> {
// FIXME Latency and NumMicrOps?
let ResourceCycles = [2,1];
}
def : WriteRes<WriteLAHFSAHF, [SLM_IEC_RSV01]>;
def : WriteRes<WriteBitTest,[SLM_IEC_RSV01]>;
// This is for simple LEAs with one or two input operands.
// The complex ones can only execute on port 1, and they require two cycles on
// the port to read all inputs. We don't model that.
def : WriteRes<WriteLEA, [SLM_IEC_RSV1]>;
// Bit counts.
defm : SLMWriteResPair<WriteBSF, [SLM_IEC_RSV01], 10, [20], 10>;
defm : SLMWriteResPair<WriteBSR, [SLM_IEC_RSV01], 10, [20], 10>;
defm : SLMWriteResPair<WriteLZCNT, [SLM_IEC_RSV0], 3>;
defm : SLMWriteResPair<WriteTZCNT, [SLM_IEC_RSV0], 3>;
defm : SLMWriteResPair<WritePOPCNT, [SLM_IEC_RSV0], 3>;
// BMI1 BEXTR/BLS, BMI2 BZHI
defm : X86WriteResPairUnsupported<WriteBEXTR>;
defm : X86WriteResPairUnsupported<WriteBLS>;
defm : X86WriteResPairUnsupported<WriteBZHI>;
defm : SLMWriteResPair<WriteDiv8, [SLM_IEC_RSV01, SLMDivider], 25, [1,25], 1, 4>;
defm : SLMWriteResPair<WriteDiv16, [SLM_IEC_RSV01, SLMDivider], 25, [1,25], 1, 4>;
defm : SLMWriteResPair<WriteDiv32, [SLM_IEC_RSV01, SLMDivider], 25, [1,25], 1, 4>;
defm : SLMWriteResPair<WriteDiv64, [SLM_IEC_RSV01, SLMDivider], 25, [1,25], 1, 4>;
defm : SLMWriteResPair<WriteIDiv8, [SLM_IEC_RSV01, SLMDivider], 25, [1,25], 1, 4>;
defm : SLMWriteResPair<WriteIDiv16, [SLM_IEC_RSV01, SLMDivider], 25, [1,25], 1, 4>;
defm : SLMWriteResPair<WriteIDiv32, [SLM_IEC_RSV01, SLMDivider], 25, [1,25], 1, 4>;
defm : SLMWriteResPair<WriteIDiv64, [SLM_IEC_RSV01, SLMDivider], 25, [1,25], 1, 4>;
// Scalar and vector floating point.
defm : X86WriteRes<WriteFLD0, [SLM_FPC_RSV01], 1, [1], 1>;
defm : X86WriteRes<WriteFLD1, [SLM_FPC_RSV01], 1, [1], 1>;
defm : X86WriteRes<WriteFLDC, [SLM_FPC_RSV01], 1, [2], 2>;
def : WriteRes<WriteFLoad, [SLM_MEC_RSV]> { let Latency = 3; }
def : WriteRes<WriteFLoadX, [SLM_MEC_RSV]> { let Latency = 3; }
def : WriteRes<WriteFLoadY, [SLM_MEC_RSV]> { let Latency = 3; }
def : WriteRes<WriteFMaskedLoad, [SLM_MEC_RSV]> { let Latency = 3; }
def : WriteRes<WriteFMaskedLoadY, [SLM_MEC_RSV]> { let Latency = 3; }
def : WriteRes<WriteFStore, [SLM_MEC_RSV]>;
def : WriteRes<WriteFStoreX, [SLM_MEC_RSV]>;
def : WriteRes<WriteFStoreY, [SLM_MEC_RSV]>;
def : WriteRes<WriteFStoreNT, [SLM_MEC_RSV]>;
def : WriteRes<WriteFStoreNTX, [SLM_MEC_RSV]>;
def : WriteRes<WriteFStoreNTY, [SLM_MEC_RSV]>;
def : WriteRes<WriteFMaskedStore, [SLM_MEC_RSV]>;
def : WriteRes<WriteFMaskedStoreY, [SLM_MEC_RSV]>;
def : WriteRes<WriteFMove, [SLM_FPC_RSV01]>;
def : WriteRes<WriteFMoveX, [SLM_FPC_RSV01]>;
def : WriteRes<WriteFMoveY, [SLM_FPC_RSV01]>;
defm : X86WriteRes<WriteEMMS, [SLM_FPC_RSV01], 10, [10], 9>;
defm : SLMWriteResPair<WriteFAdd, [SLM_FPC_RSV1], 3>;
defm : SLMWriteResPair<WriteFAddX, [SLM_FPC_RSV1], 3>;
defm : SLMWriteResPair<WriteFAddY, [SLM_FPC_RSV1], 3>;
defm : X86WriteResPairUnsupported<WriteFAddZ>;
defm : SLMWriteResPair<WriteFAdd64, [SLM_FPC_RSV1], 3>;
defm : SLMWriteResPair<WriteFAdd64X, [SLM_FPC_RSV1], 3>;
defm : SLMWriteResPair<WriteFAdd64Y, [SLM_FPC_RSV1], 3>;
defm : X86WriteResPairUnsupported<WriteFAdd64Z>;
defm : SLMWriteResPair<WriteFCmp, [SLM_FPC_RSV1], 3>;
defm : SLMWriteResPair<WriteFCmpX, [SLM_FPC_RSV1], 3>;
defm : SLMWriteResPair<WriteFCmpY, [SLM_FPC_RSV1], 3>;
defm : X86WriteResPairUnsupported<WriteFCmpZ>;
defm : SLMWriteResPair<WriteFCmp64, [SLM_FPC_RSV1], 3>;
defm : SLMWriteResPair<WriteFCmp64X, [SLM_FPC_RSV1], 3>;
defm : SLMWriteResPair<WriteFCmp64Y, [SLM_FPC_RSV1], 3>;
defm : X86WriteResPairUnsupported<WriteFCmp64Z>;
defm : SLMWriteResPair<WriteFCom, [SLM_FPC_RSV1], 3>;
defm : SLMWriteResPair<WriteFMul, [SLM_FPC_RSV0, SLMFPMultiplier], 5, [1,2]>;
defm : SLMWriteResPair<WriteFMulX, [SLM_FPC_RSV0, SLMFPMultiplier], 5, [1,2]>;
defm : SLMWriteResPair<WriteFMulY, [SLM_FPC_RSV0, SLMFPMultiplier], 5, [1,2]>;
defm : X86WriteResPairUnsupported<WriteFMulZ>;
defm : SLMWriteResPair<WriteFMul64, [SLM_FPC_RSV0, SLMFPMultiplier], 5, [1,2]>;
defm : SLMWriteResPair<WriteFMul64X, [SLM_FPC_RSV0, SLMFPMultiplier], 5, [1,2]>;
defm : SLMWriteResPair<WriteFMul64Y, [SLM_FPC_RSV0, SLMFPMultiplier], 5, [1,2]>;
defm : X86WriteResPairUnsupported<WriteFMul64Z>;
defm : SLMWriteResPair<WriteFDiv, [SLM_FPC_RSV0, SLMFPDivider], 19, [1,17]>;
defm : SLMWriteResPair<WriteFDivX, [SLM_FPC_RSV0, SLMFPDivider], 39, [1,39]>;
defm : SLMWriteResPair<WriteFDivY, [SLM_FPC_RSV0, SLMFPDivider], 39, [1,39]>;
defm : X86WriteResPairUnsupported<WriteFDivZ>;
defm : SLMWriteResPair<WriteFDiv64, [SLM_FPC_RSV0, SLMFPDivider], 34, [1,32]>;
defm : SLMWriteResPair<WriteFDiv64X, [SLM_FPC_RSV0, SLMFPDivider], 69, [1,69]>;
defm : SLMWriteResPair<WriteFDiv64Y, [SLM_FPC_RSV0, SLMFPDivider], 69, [1,69]>;
defm : X86WriteResPairUnsupported<WriteFDiv64Z>;
defm : SLMWriteResPair<WriteFRcp, [SLM_FPC_RSV0], 5>;
defm : SLMWriteResPair<WriteFRcpX, [SLM_FPC_RSV0], 5>;
defm : SLMWriteResPair<WriteFRcpY, [SLM_FPC_RSV0], 5>;
defm : X86WriteResPairUnsupported<WriteFRcpZ>;
defm : SLMWriteResPair<WriteFRsqrt, [SLM_FPC_RSV0], 5>;
defm : SLMWriteResPair<WriteFRsqrtX, [SLM_FPC_RSV0], 5>;
defm : SLMWriteResPair<WriteFRsqrtY, [SLM_FPC_RSV0], 5>;
defm : X86WriteResPairUnsupported<WriteFRsqrtZ>;
defm : SLMWriteResPair<WriteFSqrt, [SLM_FPC_RSV0,SLMFPDivider], 20, [1,20], 1, 3>;
defm : SLMWriteResPair<WriteFSqrtX, [SLM_FPC_RSV0,SLMFPDivider], 41, [1,40], 1, 3>;
defm : SLMWriteResPair<WriteFSqrtY, [SLM_FPC_RSV0,SLMFPDivider], 41, [1,40], 1, 3>;
defm : X86WriteResPairUnsupported<WriteFSqrtZ>;
defm : SLMWriteResPair<WriteFSqrt64, [SLM_FPC_RSV0,SLMFPDivider], 35, [1,35], 1, 3>;
defm : SLMWriteResPair<WriteFSqrt64X, [SLM_FPC_RSV0,SLMFPDivider], 71, [1,70], 1, 3>;
defm : SLMWriteResPair<WriteFSqrt64Y, [SLM_FPC_RSV0,SLMFPDivider], 71, [1,70], 1, 3>;
defm : X86WriteResPairUnsupported<WriteFSqrt64Z>;
defm : SLMWriteResPair<WriteFSqrt80, [SLM_FPC_RSV0,SLMFPDivider], 40, [1,40]>;
defm : SLMWriteResPair<WriteDPPD, [SLM_FPC_RSV1], 3>;
defm : SLMWriteResPair<WriteDPPS, [SLM_FPC_RSV1], 3>;
defm : SLMWriteResPair<WriteDPPSY, [SLM_FPC_RSV1], 3>;
defm : X86WriteResPairUnsupported<WriteDPPSZ>;
defm : SLMWriteResPair<WriteFSign, [SLM_FPC_RSV01], 1>;
defm : SLMWriteResPair<WriteFRnd, [SLM_FPC_RSV1], 3>;
defm : SLMWriteResPair<WriteFRndY, [SLM_FPC_RSV1], 3>;
defm : X86WriteResPairUnsupported<WriteFRndZ>;
defm : SLMWriteResPair<WriteFLogic, [SLM_FPC_RSV01], 1>;
defm : SLMWriteResPair<WriteFLogicY, [SLM_FPC_RSV01], 1>;
defm : X86WriteResPairUnsupported<WriteFLogicZ>;
defm : SLMWriteResPair<WriteFTest, [SLM_FPC_RSV01], 1>;
defm : SLMWriteResPair<WriteFTestY, [SLM_FPC_RSV01], 1>;
defm : X86WriteResPairUnsupported<WriteFTestZ>;
defm : SLMWriteResPair<WriteFShuffle, [SLM_FPC_RSV0], 1>;
defm : SLMWriteResPair<WriteFShuffleY, [SLM_FPC_RSV0], 1>;
defm : X86WriteResPairUnsupported<WriteFShuffleZ>;
defm : SLMWriteResPair<WriteFVarShuffle, [SLM_FPC_RSV0], 1>;
defm : SLMWriteResPair<WriteFVarShuffleY,[SLM_FPC_RSV0], 1>;
defm : X86WriteResPairUnsupported<WriteFVarShuffleZ>;
defm : SLMWriteResPair<WriteFBlend, [SLM_FPC_RSV0], 1>;
// Conversion between integer and float.
defm : SLMWriteResPair<WriteCvtSS2I, [SLM_FPC_RSV01], 4>;
defm : SLMWriteResPair<WriteCvtPS2I, [SLM_FPC_RSV01], 4>;
defm : SLMWriteResPair<WriteCvtPS2IY, [SLM_FPC_RSV01], 4>;
defm : X86WriteResPairUnsupported<WriteCvtPS2IZ>;
defm : SLMWriteResPair<WriteCvtSD2I, [SLM_FPC_RSV01], 4>;
defm : SLMWriteResPair<WriteCvtPD2I, [SLM_FPC_RSV01], 4>;
defm : SLMWriteResPair<WriteCvtPD2IY, [SLM_FPC_RSV01], 4>;
defm : X86WriteResPairUnsupported<WriteCvtPD2IZ>;
defm : SLMWriteResPair<WriteCvtI2SS, [SLM_FPC_RSV01], 4>;
defm : SLMWriteResPair<WriteCvtI2PS, [SLM_FPC_RSV01], 4>;
defm : SLMWriteResPair<WriteCvtI2PSY, [SLM_FPC_RSV01], 4>;
defm : X86WriteResPairUnsupported<WriteCvtI2PSZ>;
defm : SLMWriteResPair<WriteCvtI2SD, [SLM_FPC_RSV01], 4>;
defm : SLMWriteResPair<WriteCvtI2PD, [SLM_FPC_RSV01], 4>;
defm : SLMWriteResPair<WriteCvtI2PDY, [SLM_FPC_RSV01], 4>;
defm : X86WriteResPairUnsupported<WriteCvtI2PDZ>;
defm : SLMWriteResPair<WriteCvtSS2SD, [SLM_FPC_RSV01], 4>;
defm : SLMWriteResPair<WriteCvtPS2PD, [SLM_FPC_RSV01], 4>;
defm : SLMWriteResPair<WriteCvtPS2PDY, [SLM_FPC_RSV01], 4>;
defm : X86WriteResPairUnsupported<WriteCvtPS2PDZ>;
defm : SLMWriteResPair<WriteCvtSD2SS, [SLM_FPC_RSV01], 4>;
defm : SLMWriteResPair<WriteCvtPD2PS, [SLM_FPC_RSV01], 4>;
defm : SLMWriteResPair<WriteCvtPD2PSY, [SLM_FPC_RSV01], 4>;
defm : X86WriteResPairUnsupported<WriteCvtPD2PSZ>;
// Vector integer operations.
def : WriteRes<WriteVecLoad, [SLM_MEC_RSV]> { let Latency = 3; }
def : WriteRes<WriteVecLoadX, [SLM_MEC_RSV]> { let Latency = 3; }
def : WriteRes<WriteVecLoadY, [SLM_MEC_RSV]> { let Latency = 3; }
def : WriteRes<WriteVecLoadNT, [SLM_MEC_RSV]> { let Latency = 3; }
def : WriteRes<WriteVecLoadNTY, [SLM_MEC_RSV]> { let Latency = 3; }
def : WriteRes<WriteVecMaskedLoad, [SLM_MEC_RSV]> { let Latency = 3; }
def : WriteRes<WriteVecMaskedLoadY, [SLM_MEC_RSV]> { let Latency = 3; }
def : WriteRes<WriteVecStore, [SLM_MEC_RSV]>;
def : WriteRes<WriteVecStoreX, [SLM_MEC_RSV]>;
def : WriteRes<WriteVecStoreY, [SLM_MEC_RSV]>;
def : WriteRes<WriteVecStoreNT, [SLM_MEC_RSV]>;
def : WriteRes<WriteVecStoreNTY, [SLM_MEC_RSV]>;
def : WriteRes<WriteVecMaskedStore, [SLM_MEC_RSV]>;
def : WriteRes<WriteVecMaskedStoreY, [SLM_MEC_RSV]>;
def : WriteRes<WriteVecMove, [SLM_FPC_RSV01]>;
def : WriteRes<WriteVecMoveX, [SLM_FPC_RSV01]>;
def : WriteRes<WriteVecMoveY, [SLM_FPC_RSV01]>;
def : WriteRes<WriteVecMoveToGpr, [SLM_IEC_RSV01]>;
def : WriteRes<WriteVecMoveFromGpr, [SLM_IEC_RSV01]>;
defm : SLMWriteResPair<WriteVecShift, [SLM_FPC_RSV0], 1>;
defm : SLMWriteResPair<WriteVecShiftX, [SLM_FPC_RSV0], 1>;
defm : SLMWriteResPair<WriteVecShiftY, [SLM_FPC_RSV0], 1>;
defm : X86WriteResPairUnsupported<WriteVecShiftZ>;
defm : SLMWriteResPair<WriteVecShiftImm, [SLM_FPC_RSV0], 1>;
defm : SLMWriteResPair<WriteVecShiftImmX,[SLM_FPC_RSV0], 1>;
defm : SLMWriteResPair<WriteVecShiftImmY,[SLM_FPC_RSV0], 1>;
defm : X86WriteResPairUnsupported<WriteVecShiftImmZ>;
defm : SLMWriteResPair<WriteVecLogic, [SLM_FPC_RSV01], 1>;
defm : SLMWriteResPair<WriteVecLogicX,[SLM_FPC_RSV01], 1>;
defm : SLMWriteResPair<WriteVecLogicY,[SLM_FPC_RSV01], 1>;
defm : X86WriteResPairUnsupported<WriteVecLogicZ>;
defm : SLMWriteResPair<WriteVecTest, [SLM_FPC_RSV01], 1>;
defm : SLMWriteResPair<WriteVecTestY, [SLM_FPC_RSV01], 1>;
defm : X86WriteResPairUnsupported<WriteVecTestZ>;
defm : SLMWriteResPair<WriteVecALU, [SLM_FPC_RSV01], 1>;
defm : SLMWriteResPair<WriteVecALUX, [SLM_FPC_RSV01], 1>;
defm : SLMWriteResPair<WriteVecALUY, [SLM_FPC_RSV01], 1>;
defm : X86WriteResPairUnsupported<WriteVecALUZ>;
defm : SLMWriteResPair<WriteVecIMul, [SLM_FPC_RSV0], 4>;
defm : SLMWriteResPair<WriteVecIMulX, [SLM_FPC_RSV0], 4>;
defm : SLMWriteResPair<WriteVecIMulY, [SLM_FPC_RSV0], 4>;
defm : X86WriteResPairUnsupported<WriteVecIMulZ>;
// FIXME: The below is closer to correct, but caused some perf regressions.
//defm : SLMWriteResPair<WritePMULLD, [SLM_FPC_RSV0], 11, [11], 7>;
defm : SLMWriteResPair<WritePMULLD, [SLM_FPC_RSV0], 4>;
defm : SLMWriteResPair<WritePMULLDY, [SLM_FPC_RSV0], 4>;
defm : X86WriteResPairUnsupported<WritePMULLDZ>;
defm : SLMWriteResPair<WriteShuffle, [SLM_FPC_RSV0], 1>;
defm : SLMWriteResPair<WriteShuffleY, [SLM_FPC_RSV0], 1>;
defm : X86WriteResPairUnsupported<WriteShuffleZ>;
defm : SLMWriteResPair<WriteShuffleX, [SLM_FPC_RSV0], 1>;
defm : SLMWriteResPair<WriteVarShuffle, [SLM_FPC_RSV0], 1>;
defm : SLMWriteResPair<WriteVarShuffleX, [SLM_FPC_RSV0], 1>;
defm : SLMWriteResPair<WriteVarShuffleY, [SLM_FPC_RSV0], 1>;
defm : X86WriteResPairUnsupported<WriteVarShuffleZ>;
defm : SLMWriteResPair<WriteBlend, [SLM_FPC_RSV0], 1>;
defm : SLMWriteResPair<WriteBlendY, [SLM_FPC_RSV0], 1>;
defm : X86WriteResPairUnsupported<WriteBlendZ>;
defm : SLMWriteResPair<WriteMPSAD, [SLM_FPC_RSV0], 7>;
defm : SLMWriteResPair<WriteMPSADY, [SLM_FPC_RSV0], 7>;
defm : X86WriteResPairUnsupported<WriteMPSADZ>;
defm : SLMWriteResPair<WritePSADBW, [SLM_FPC_RSV0], 4>;
defm : SLMWriteResPair<WritePSADBWX, [SLM_FPC_RSV0], 4>;
defm : SLMWriteResPair<WritePSADBWY, [SLM_FPC_RSV0], 4>;
defm : X86WriteResPairUnsupported<WritePSADBWZ>;
defm : SLMWriteResPair<WritePHMINPOS, [SLM_FPC_RSV0], 4>;
// Vector insert/extract operations.
defm : SLMWriteResPair<WriteVecInsert, [SLM_FPC_RSV0], 1>;
def : WriteRes<WriteVecExtract, [SLM_FPC_RSV0]>;
def : WriteRes<WriteVecExtractSt, [SLM_FPC_RSV0, SLM_MEC_RSV]> {
let Latency = 4;
let NumMicroOps = 2;
let ResourceCycles = [1, 2];
}
////////////////////////////////////////////////////////////////////////////////
// Horizontal add/sub instructions.
////////////////////////////////////////////////////////////////////////////////
defm : SLMWriteResPair<WriteFHAdd, [SLM_FPC_RSV01], 3, [2]>;
defm : SLMWriteResPair<WriteFHAddY, [SLM_FPC_RSV01], 3, [2]>;
defm : X86WriteResPairUnsupported<WriteFHAddZ>;
defm : SLMWriteResPair<WritePHAdd, [SLM_FPC_RSV01], 1>;
defm : SLMWriteResPair<WritePHAddX, [SLM_FPC_RSV01], 1>;
defm : SLMWriteResPair<WritePHAddY, [SLM_FPC_RSV01], 1>;
defm : X86WriteResPairUnsupported<WritePHAddZ>;
// String instructions.
// Packed Compare Implicit Length Strings, Return Mask
def : WriteRes<WritePCmpIStrM, [SLM_FPC_RSV0]> {
let Latency = 13;
let ResourceCycles = [13];
}
def : WriteRes<WritePCmpIStrMLd, [SLM_FPC_RSV0, SLM_MEC_RSV]> {
let Latency = 13;
let ResourceCycles = [13, 1];
}
// Packed Compare Explicit Length Strings, Return Mask
def : WriteRes<WritePCmpEStrM, [SLM_FPC_RSV0]> {
let Latency = 17;
let ResourceCycles = [17];
}
def : WriteRes<WritePCmpEStrMLd, [SLM_FPC_RSV0, SLM_MEC_RSV]> {
let Latency = 17;
let ResourceCycles = [17, 1];
}
// Packed Compare Implicit Length Strings, Return Index
def : WriteRes<WritePCmpIStrI, [SLM_FPC_RSV0]> {
let Latency = 17;
let ResourceCycles = [17];
}
def : WriteRes<WritePCmpIStrILd, [SLM_FPC_RSV0, SLM_MEC_RSV]> {
let Latency = 17;
let ResourceCycles = [17, 1];
}
// Packed Compare Explicit Length Strings, Return Index
def : WriteRes<WritePCmpEStrI, [SLM_FPC_RSV0]> {
let Latency = 21;
let ResourceCycles = [21];
}
def : WriteRes<WritePCmpEStrILd, [SLM_FPC_RSV0, SLM_MEC_RSV]> {
let Latency = 21;
let ResourceCycles = [21, 1];
}
// MOVMSK Instructions.
def : WriteRes<WriteFMOVMSK, [SLM_FPC_RSV1]> { let Latency = 4; }
def : WriteRes<WriteVecMOVMSK, [SLM_FPC_RSV1]> { let Latency = 4; }
def : WriteRes<WriteVecMOVMSKY, [SLM_FPC_RSV1]> { let Latency = 4; }
def : WriteRes<WriteMMXMOVMSK, [SLM_FPC_RSV1]> { let Latency = 4; }
// AES Instructions.
def : WriteRes<WriteAESDecEnc, [SLM_FPC_RSV0]> {
let Latency = 8;
let ResourceCycles = [5];
}
def : WriteRes<WriteAESDecEncLd, [SLM_FPC_RSV0, SLM_MEC_RSV]> {
let Latency = 8;
let ResourceCycles = [5, 1];
}
def : WriteRes<WriteAESIMC, [SLM_FPC_RSV0]> {
let Latency = 8;
let ResourceCycles = [5];
}
def : WriteRes<WriteAESIMCLd, [SLM_FPC_RSV0, SLM_MEC_RSV]> {
let Latency = 8;
let ResourceCycles = [5, 1];
}
def : WriteRes<WriteAESKeyGen, [SLM_FPC_RSV0]> {
let Latency = 8;
let ResourceCycles = [5];
}
def : WriteRes<WriteAESKeyGenLd, [SLM_FPC_RSV0, SLM_MEC_RSV]> {
let Latency = 8;
let ResourceCycles = [5, 1];
}
// Carry-less multiplication instructions.
def : WriteRes<WriteCLMul, [SLM_FPC_RSV0]> {
let Latency = 10;
let ResourceCycles = [10];
}
def : WriteRes<WriteCLMulLd, [SLM_FPC_RSV0, SLM_MEC_RSV]> {
let Latency = 10;
let ResourceCycles = [10, 1];
}
def : WriteRes<WriteSystem, [SLM_FPC_RSV0]> { let Latency = 100; }
def : WriteRes<WriteMicrocoded, [SLM_FPC_RSV0]> { let Latency = 100; }
def : WriteRes<WriteFence, [SLM_MEC_RSV]>;
def : WriteRes<WriteNop, []>;
// AVX/FMA is not supported on that architecture, but we should define the basic
// scheduling resources anyway.
def : WriteRes<WriteIMulH, [SLM_FPC_RSV0]>;
defm : X86WriteResPairUnsupported<WriteFBlendY>;
defm : X86WriteResPairUnsupported<WriteFBlendZ>;
defm : SLMWriteResPair<WriteVarBlend, [SLM_FPC_RSV0], 1>;
defm : X86WriteResPairUnsupported<WriteVarBlendY>;
defm : X86WriteResPairUnsupported<WriteVarBlendZ>;
defm : SLMWriteResPair<WriteFVarBlend, [SLM_FPC_RSV0], 1>;
defm : X86WriteResPairUnsupported<WriteFVarBlendY>;
defm : X86WriteResPairUnsupported<WriteFVarBlendZ>;
defm : X86WriteResPairUnsupported<WriteFShuffle256>;
defm : X86WriteResPairUnsupported<WriteFVarShuffle256>;
defm : X86WriteResPairUnsupported<WriteShuffle256>;
defm : X86WriteResPairUnsupported<WriteVarShuffle256>;
defm : SLMWriteResPair<WriteVarVecShift, [SLM_FPC_RSV0], 1>;
defm : X86WriteResPairUnsupported<WriteVarVecShiftY>;
defm : X86WriteResPairUnsupported<WriteVarVecShiftZ>;
defm : X86WriteResPairUnsupported<WriteFMA>;
defm : X86WriteResPairUnsupported<WriteFMAX>;
defm : X86WriteResPairUnsupported<WriteFMAY>;
defm : X86WriteResPairUnsupported<WriteFMAZ>;
defm : X86WriteResPairUnsupported<WriteCvtPH2PS>;
defm : X86WriteResPairUnsupported<WriteCvtPH2PSY>;
defm : X86WriteResPairUnsupported<WriteCvtPH2PSZ>;
defm : X86WriteResUnsupported<WriteCvtPS2PH>;
defm : X86WriteResUnsupported<WriteCvtPS2PHY>;
defm : X86WriteResUnsupported<WriteCvtPS2PHZ>;
defm : X86WriteResUnsupported<WriteCvtPS2PHSt>;
defm : X86WriteResUnsupported<WriteCvtPS2PHYSt>;
defm : X86WriteResUnsupported<WriteCvtPS2PHZSt>;
} // SchedModel