[RISCV] Add P550 scheduler model. (llvm#124639)

P550 falls between P450 and P650. It has 1 additional FEX pipe over
P450. Mul and cpop latency are 3 instead of 2.

I've set the MicroOpBufferSize to 96 instead of 56 based on the ROB size
measurement from
https://chipsandcheese.com/p/inside-sifives-p550-microarchitecture I
believe we set this value too low for P450 and P650 and should update
them in a separate PR.

Author: topperc

llvm/lib/Target/RISCV/RISCV.td

@@ -50,6 +50,7 @@ include "RISCVSchedMIPSP8700.td"
 include "RISCVSchedRocket.td"
 include "RISCVSchedSiFive7.td"
 include "RISCVSchedSiFiveP400.td"
+include "RISCVSchedSiFiveP500.td"
 include "RISCVSchedSiFiveP600.td"
 include "RISCVSchedSyntacoreSCR1.td"
 include "RISCVSchedSyntacoreSCR345.td"

llvm/lib/Target/RISCV/RISCVProcessors.td

@@ -327,7 +327,7 @@ defvar SiFiveP500TuneFeatures = [TuneNoDefaultUnroll,
                                  TuneAUIPCADDIFusion,
                                  TunePostRAScheduler];
 
-def SIFIVE_P550 : RISCVProcessorModel<"sifive-p550", NoSchedModel,
+def SIFIVE_P550 : RISCVProcessorModel<"sifive-p550", SiFiveP500Model,
                                       [Feature64Bit,
                                        FeatureStdExtI,
                                        FeatureStdExtZifencei,

llvm/lib/Target/RISCV/RISCVSchedSiFiveP500.td

@@ -0,0 +1,361 @@
+//==- RISCVSchedSiFiveP500.td - SiFiveP500 Scheduling Defs ---*- tablegen -*-=//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+//===----------------------------------------------------------------------===//
+
+def SiFiveP500Model : SchedMachineModel {
+  let IssueWidth = 3;         // 3 micro-ops are dispatched per cycle.
+  let MicroOpBufferSize = 96; // Max micro-ops that can be buffered.
+  let LoadLatency = 4;        // Cycles for loads to access the cache.
+  let MispredictPenalty = 9;  // Extra cycles for a mispredicted branch.
+  let CompleteModel = false;
+}
+
+// The SiFiveP500 microarchitecure has 7 pipelines:
+// Three pipelines for integer operations.
+// Two pipelines for FPU operations.
+// One pipeline for Load operations.
+// One pipeline for Store operations.
+let SchedModel = SiFiveP500Model in {
+
+def SiFiveP500IEXQ0       : ProcResource<1>;
+def SiFiveP500IEXQ1       : ProcResource<1>;
+def SiFiveP500IEXQ2       : ProcResource<1>;
+def SiFiveP500FEXQ0       : ProcResource<1>;
+def SiFiveP500FEXQ1       : ProcResource<1>;
+def SiFiveP500Load        : ProcResource<1>;
+def SiFiveP500Store       : ProcResource<1>;
+
+def SiFiveP500IntArith    : ProcResGroup<[SiFiveP500IEXQ0, SiFiveP500IEXQ1, SiFiveP500IEXQ2]>;
+defvar SiFiveP500Branch   = SiFiveP500IEXQ0;
+defvar SiFiveP500SYS      = SiFiveP500IEXQ1;
+defvar SiFiveP500CMOV     = SiFiveP500IEXQ1;
+defvar SiFiveP500MulI2F   = SiFiveP500IEXQ2;
+def SiFiveP500Div         : ProcResource<1>;
+
+def SiFiveP500FloatArith  : ProcResGroup<[SiFiveP500FEXQ0, SiFiveP500FEXQ1]>;
+defvar SiFiveP500F2I      = SiFiveP500FEXQ0;
+def SiFiveP500FloatDiv    : ProcResource<1>;
+
+let Latency = 1 in {
+// Integer arithmetic and logic
+def : WriteRes<WriteIALU, [SiFiveP500IntArith]>;
+def : WriteRes<WriteIALU32, [SiFiveP500IntArith]>;
+def : WriteRes<WriteShiftImm, [SiFiveP500IntArith]>;
+def : WriteRes<WriteShiftImm32, [SiFiveP500IntArith]>;
+def : WriteRes<WriteShiftReg, [SiFiveP500IntArith]>;
+def : WriteRes<WriteShiftReg32, [SiFiveP500IntArith]>;
+// Branching
+def : WriteRes<WriteJmp, [SiFiveP500Branch]>;
+def : WriteRes<WriteJal, [SiFiveP500Branch]>;
+def : WriteRes<WriteJalr, [SiFiveP500Branch]>;
+}
+
+// CMOV
+def P500WriteCMOV : SchedWriteRes<[SiFiveP500Branch, SiFiveP500CMOV]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}
+def : InstRW<[P500WriteCMOV], (instrs PseudoCCMOVGPRNoX0)>;
+
+let Latency = 3 in {
+// Integer multiplication
+def : WriteRes<WriteIMul, [SiFiveP500MulI2F]>;
+def : WriteRes<WriteIMul32, [SiFiveP500MulI2F]>;
+// cpop[w] look exactly like multiply.
+def : WriteRes<WriteCPOP, [SiFiveP500MulI2F]>;
+def : WriteRes<WriteCPOP32, [SiFiveP500MulI2F]>;
+}
+
+// Integer division
+def : WriteRes<WriteIDiv, [SiFiveP500MulI2F, SiFiveP500Div]> {
+  let Latency = 35;
+  let ReleaseAtCycles = [1, 34];
+}
+def : WriteRes<WriteIDiv32, [SiFiveP500MulI2F, SiFiveP500Div]> {
+  let Latency = 20;
+  let ReleaseAtCycles = [1, 19];
+}
+
+// Integer remainder
+def : WriteRes<WriteIRem, [SiFiveP500MulI2F, SiFiveP500Div]> {
+  let Latency = 35;
+  let ReleaseAtCycles = [1, 34];
+}
+def : WriteRes<WriteIRem32, [SiFiveP500MulI2F, SiFiveP500Div]> {
+  let Latency = 20;
+  let ReleaseAtCycles = [1, 19];
+}
+
+let Latency = 1 in {
+// Bitmanip
+def : WriteRes<WriteRotateImm, [SiFiveP500IntArith]>;
+def : WriteRes<WriteRotateImm32, [SiFiveP500IntArith]>;
+def : WriteRes<WriteRotateReg, [SiFiveP500IntArith]>;
+def : WriteRes<WriteRotateReg32, [SiFiveP500IntArith]>;
+
+def : WriteRes<WriteCLZ, [SiFiveP500IntArith]>;
+def : WriteRes<WriteCLZ32, [SiFiveP500IntArith]>;
+def : WriteRes<WriteCTZ, [SiFiveP500IntArith]>;
+def : WriteRes<WriteCTZ32, [SiFiveP500IntArith]>;
+
+def : WriteRes<WriteORCB, [SiFiveP500IntArith]>;
+def : WriteRes<WriteIMinMax, [SiFiveP500IntArith]>;
+
+def : WriteRes<WriteREV8, [SiFiveP500IntArith]>;
+
+def : WriteRes<WriteSHXADD, [SiFiveP500IntArith]>;
+def : WriteRes<WriteSHXADD32, [SiFiveP500IntArith]>;
+}
+
+// Memory
+let Latency = 1 in {
+def : WriteRes<WriteSTB, [SiFiveP500Store]>;
+def : WriteRes<WriteSTH, [SiFiveP500Store]>;
+def : WriteRes<WriteSTW, [SiFiveP500Store]>;
+def : WriteRes<WriteSTD, [SiFiveP500Store]>;
+def : WriteRes<WriteFST16, [SiFiveP500Store]>;
+def : WriteRes<WriteFST32, [SiFiveP500Store]>;
+def : WriteRes<WriteFST64, [SiFiveP500Store]>;
+}
+let Latency = 4 in {
+def : WriteRes<WriteLDB, [SiFiveP500Load]>;
+def : WriteRes<WriteLDH, [SiFiveP500Load]>;
+}
+let Latency = 4 in {
+def : WriteRes<WriteLDW, [SiFiveP500Load]>;
+def : WriteRes<WriteLDD, [SiFiveP500Load]>;
+}
+
+let Latency = 6 in {
+def : WriteRes<WriteFLD16, [SiFiveP500Load]>;
+def : WriteRes<WriteFLD32, [SiFiveP500Load]>;
+def : WriteRes<WriteFLD64, [SiFiveP500Load]>;
+}
+
+// Atomic memory
+let Latency = 3 in {
+def : WriteRes<WriteAtomicSTW, [SiFiveP500Store]>;
+def : WriteRes<WriteAtomicSTD, [SiFiveP500Store]>;
+def : WriteRes<WriteAtomicW, [SiFiveP500Load]>;
+def : WriteRes<WriteAtomicD, [SiFiveP500Load]>;
+def : WriteRes<WriteAtomicLDW, [SiFiveP500Load]>;
+def : WriteRes<WriteAtomicLDD, [SiFiveP500Load]>;
+}
+
+// Floating point
+let Latency = 4 in {
+def : WriteRes<WriteFAdd16, [SiFiveP500FloatArith]>;
+def : WriteRes<WriteFAdd32, [SiFiveP500FloatArith]>;
+def : WriteRes<WriteFAdd64, [SiFiveP500FloatArith]>;
+
+def : WriteRes<WriteFMul16, [SiFiveP500FloatArith]>;
+def : WriteRes<WriteFMul32, [SiFiveP500FloatArith]>;
+def : WriteRes<WriteFMul64, [SiFiveP500FloatArith]>;
+
+def : WriteRes<WriteFMA16, [SiFiveP500FloatArith]>;
+def : WriteRes<WriteFMA32, [SiFiveP500FloatArith]>;
+def : WriteRes<WriteFMA64, [SiFiveP500FloatArith]>;
+}
+
+let Latency = 2 in {
+def : WriteRes<WriteFSGNJ16, [SiFiveP500FloatArith]>;
+def : WriteRes<WriteFSGNJ32, [SiFiveP500FloatArith]>;
+def : WriteRes<WriteFSGNJ64, [SiFiveP500FloatArith]>;
+
+def : WriteRes<WriteFMinMax16, [SiFiveP500FloatArith]>;
+def : WriteRes<WriteFMinMax32, [SiFiveP500FloatArith]>;
+def : WriteRes<WriteFMinMax64, [SiFiveP500FloatArith]>;
+}
+
+// Half precision.
+def : WriteRes<WriteFDiv16, [SiFiveP500FEXQ1, SiFiveP500FloatDiv]> {
+  let Latency = 19;
+  let ReleaseAtCycles = [1, 18];
+}
+def : WriteRes<WriteFSqrt16, [SiFiveP500FEXQ1, SiFiveP500FloatDiv]> {
+  let Latency = 18;
+  let ReleaseAtCycles = [1, 17];
+}
+
+// Single precision.
+def : WriteRes<WriteFDiv32, [SiFiveP500FEXQ1, SiFiveP500FloatDiv]> {
+  let Latency = 19;
+  let ReleaseAtCycles = [1, 18];
+}
+def : WriteRes<WriteFSqrt32, [SiFiveP500FEXQ1, SiFiveP500FloatDiv]> {
+  let Latency = 18;
+  let ReleaseAtCycles = [1, 17];
+}
+
+// Double precision
+def : WriteRes<WriteFDiv64, [SiFiveP500FEXQ1, SiFiveP500FloatDiv]> {
+  let Latency = 33;
+  let ReleaseAtCycles = [1, 32];
+}
+def : WriteRes<WriteFSqrt64, [SiFiveP500FEXQ1, SiFiveP500FloatDiv]> {
+  let Latency = 33;
+  let ReleaseAtCycles = [1, 32];
+}
+
+// Conversions
+let Latency = 2 in {
+def : WriteRes<WriteFCvtI32ToF16, [SiFiveP500MulI2F]>;
+def : WriteRes<WriteFCvtI32ToF32, [SiFiveP500MulI2F]>;
+def : WriteRes<WriteFCvtI32ToF64, [SiFiveP500MulI2F]>;
+def : WriteRes<WriteFCvtI64ToF16, [SiFiveP500MulI2F]>;
+def : WriteRes<WriteFCvtI64ToF32, [SiFiveP500MulI2F]>;
+def : WriteRes<WriteFCvtI64ToF64, [SiFiveP500MulI2F]>;
+def : WriteRes<WriteFCvtF16ToI32, [SiFiveP500F2I]>;
+def : WriteRes<WriteFCvtF16ToI64, [SiFiveP500F2I]>;
+def : WriteRes<WriteFCvtF16ToF32, [SiFiveP500FloatArith]>;
+def : WriteRes<WriteFCvtF16ToF64, [SiFiveP500FloatArith]>;
+def : WriteRes<WriteFCvtF32ToI32, [SiFiveP500F2I]>;
+def : WriteRes<WriteFCvtF32ToI64, [SiFiveP500F2I]>;
+def : WriteRes<WriteFCvtF32ToF16, [SiFiveP500FloatArith]>;
+def : WriteRes<WriteFCvtF32ToF64, [SiFiveP500FloatArith]>;
+def : WriteRes<WriteFCvtF64ToI32, [SiFiveP500F2I]>;
+def : WriteRes<WriteFCvtF64ToI64, [SiFiveP500F2I]>;
+def : WriteRes<WriteFCvtF64ToF16, [SiFiveP500FloatArith]>;
+def : WriteRes<WriteFCvtF64ToF32, [SiFiveP500FloatArith]>;
+
+def : WriteRes<WriteFClass16, [SiFiveP500F2I]>;
+def : WriteRes<WriteFClass32, [SiFiveP500F2I]>;
+def : WriteRes<WriteFClass64, [SiFiveP500F2I]>;
+def : WriteRes<WriteFCmp16, [SiFiveP500F2I]>;
+def : WriteRes<WriteFCmp32, [SiFiveP500F2I]>;
+def : WriteRes<WriteFCmp64, [SiFiveP500F2I]>;
+def : WriteRes<WriteFMovI16ToF16, [SiFiveP500MulI2F]>;
+def : WriteRes<WriteFMovF16ToI16, [SiFiveP500F2I]>;
+def : WriteRes<WriteFMovI32ToF32, [SiFiveP500MulI2F]>;
+def : WriteRes<WriteFMovF32ToI32, [SiFiveP500F2I]>;
+def : WriteRes<WriteFMovI64ToF64, [SiFiveP500MulI2F]>;
+def : WriteRes<WriteFMovF64ToI64, [SiFiveP500F2I]>;
+}
+
+// Others
+def : WriteRes<WriteCSR, [SiFiveP500SYS]>;
+def : WriteRes<WriteNop, []>;
+
+// FIXME: This could be better modeled by looking at the regclasses of the operands.
+def : InstRW<[WriteIALU, ReadIALU], (instrs COPY)>;
+
+//===----------------------------------------------------------------------===//
+// Bypass and advance
+def : ReadAdvance<ReadJmp, 0>;
+def : ReadAdvance<ReadJalr, 0>;
+def : ReadAdvance<ReadCSR, 0>;
+def : ReadAdvance<ReadStoreData, 0>;
+def : ReadAdvance<ReadMemBase, 0>;
+def : ReadAdvance<ReadIALU, 0>;
+def : ReadAdvance<ReadIALU32, 0>;
+def : ReadAdvance<ReadShiftImm, 0>;
+def : ReadAdvance<ReadShiftImm32, 0>;
+def : ReadAdvance<ReadShiftReg, 0>;
+def : ReadAdvance<ReadShiftReg32, 0>;
+def : ReadAdvance<ReadIDiv, 0>;
+def : ReadAdvance<ReadIDiv32, 0>;
+def : ReadAdvance<ReadIRem, 0>;
+def : ReadAdvance<ReadIRem32, 0>;
+def : ReadAdvance<ReadIMul, 0>;
+def : ReadAdvance<ReadIMul32, 0>;
+def : ReadAdvance<ReadAtomicWA, 0>;
+def : ReadAdvance<ReadAtomicWD, 0>;
+def : ReadAdvance<ReadAtomicDA, 0>;
+def : ReadAdvance<ReadAtomicDD, 0>;
+def : ReadAdvance<ReadAtomicLDW, 0>;
+def : ReadAdvance<ReadAtomicLDD, 0>;
+def : ReadAdvance<ReadAtomicSTW, 0>;
+def : ReadAdvance<ReadAtomicSTD, 0>;
+def : ReadAdvance<ReadFStoreData, 0>;
+def : ReadAdvance<ReadFMemBase, 0>;
+def : ReadAdvance<ReadFAdd16, 0>;
+def : ReadAdvance<ReadFAdd32, 0>;
+def : ReadAdvance<ReadFAdd64, 0>;
+def : ReadAdvance<ReadFMul16, 0>;
+def : ReadAdvance<ReadFMA16, 0>;
+def : ReadAdvance<ReadFMA16Addend, 0>;
+def : ReadAdvance<ReadFMul32, 0>;
+def : ReadAdvance<ReadFMA32, 0>;
+def : ReadAdvance<ReadFMA32Addend, 0>;
+def : ReadAdvance<ReadFMul64, 0>;
+def : ReadAdvance<ReadFMA64, 0>;
+def : ReadAdvance<ReadFMA64Addend, 0>;
+def : ReadAdvance<ReadFDiv16, 0>;
+def : ReadAdvance<ReadFDiv32, 0>;
+def : ReadAdvance<ReadFDiv64, 0>;
+def : ReadAdvance<ReadFSqrt16, 0>;
+def : ReadAdvance<ReadFSqrt32, 0>;
+def : ReadAdvance<ReadFSqrt64, 0>;
+def : ReadAdvance<ReadFCmp16, 0>;
+def : ReadAdvance<ReadFCmp32, 0>;
+def : ReadAdvance<ReadFCmp64, 0>;
+def : ReadAdvance<ReadFSGNJ16, 0>;
+def : ReadAdvance<ReadFSGNJ32, 0>;
+def : ReadAdvance<ReadFSGNJ64, 0>;
+def : ReadAdvance<ReadFMinMax16, 0>;
+def : ReadAdvance<ReadFMinMax32, 0>;
+def : ReadAdvance<ReadFMinMax64, 0>;
+def : ReadAdvance<ReadFCvtF16ToI32, 0>;
+def : ReadAdvance<ReadFCvtF16ToI64, 0>;
+def : ReadAdvance<ReadFCvtF32ToI32, 0>;
+def : ReadAdvance<ReadFCvtF32ToI64, 0>;
+def : ReadAdvance<ReadFCvtF64ToI32, 0>;
+def : ReadAdvance<ReadFCvtF64ToI64, 0>;
+def : ReadAdvance<ReadFCvtI32ToF16, 0>;
+def : ReadAdvance<ReadFCvtI32ToF32, 0>;
+def : ReadAdvance<ReadFCvtI32ToF64, 0>;
+def : ReadAdvance<ReadFCvtI64ToF16, 0>;
+def : ReadAdvance<ReadFCvtI64ToF32, 0>;
+def : ReadAdvance<ReadFCvtI64ToF64, 0>;
+def : ReadAdvance<ReadFCvtF32ToF64, 0>;
+def : ReadAdvance<ReadFCvtF64ToF32, 0>;
+def : ReadAdvance<ReadFCvtF16ToF32, 0>;
+def : ReadAdvance<ReadFCvtF32ToF16, 0>;
+def : ReadAdvance<ReadFCvtF16ToF64, 0>;
+def : ReadAdvance<ReadFCvtF64ToF16, 0>;
+def : ReadAdvance<ReadFMovF16ToI16, 0>;
+def : ReadAdvance<ReadFMovI16ToF16, 0>;
+def : ReadAdvance<ReadFMovF32ToI32, 0>;
+def : ReadAdvance<ReadFMovI32ToF32, 0>;
+def : ReadAdvance<ReadFMovF64ToI64, 0>;
+def : ReadAdvance<ReadFMovI64ToF64, 0>;
+def : ReadAdvance<ReadFClass16, 0>;
+def : ReadAdvance<ReadFClass32, 0>;
+def : ReadAdvance<ReadFClass64, 0>;
+
+// Bitmanip
+def : ReadAdvance<ReadRotateImm, 0>;
+def : ReadAdvance<ReadRotateImm32, 0>;
+def : ReadAdvance<ReadRotateReg, 0>;
+def : ReadAdvance<ReadRotateReg32, 0>;
+def : ReadAdvance<ReadCLZ, 0>;
+def : ReadAdvance<ReadCLZ32, 0>;
+def : ReadAdvance<ReadCTZ, 0>;
+def : ReadAdvance<ReadCTZ32, 0>;
+def : ReadAdvance<ReadCPOP, 0>;
+def : ReadAdvance<ReadCPOP32, 0>;
+def : ReadAdvance<ReadORCB, 0>;
+def : ReadAdvance<ReadIMinMax, 0>;
+def : ReadAdvance<ReadREV8, 0>;
+def : ReadAdvance<ReadSHXADD, 0>;
+def : ReadAdvance<ReadSHXADD32, 0>;
+
+//===----------------------------------------------------------------------===//
+// Unsupported extensions
+defm : UnsupportedSchedV;
+defm : UnsupportedSchedZabha;
+defm : UnsupportedSchedZbc;
+defm : UnsupportedSchedZbs;
+defm : UnsupportedSchedZbkb;
+defm : UnsupportedSchedZbkx;
+defm : UnsupportedSchedSFB;
+defm : UnsupportedSchedZfa;
+defm : UnsupportedSchedZvk;
+defm : UnsupportedSchedXsfvcp;
+}