[RISCV] Merge the pipeline models for Rocket

[evandro]

Summary

Merge the 32 and 64 bit pipeline models for Rocket into a single file.

[HsiangKai]

Why not delete RISCVSchedRocket32.td?

[evandro]

In D87873#2284396, @HsiangKai wrote:

Why not delete RISCVSchedRocket32.td?

Of course!

[HsiangKai]

LGTM.

llvm/lib/Target/RISCV/RISCV.td

 // Registers, calling conventions, instruction descriptions.
 //===----------------------------------------------------------------------===//
 
 include "RISCVSchedule.td"
 include "RISCVRegisterInfo.td"
 include "RISCVCallingConv.td"
 include "RISCVInstrInfo.td"
 include "RISCVRegisterBanks.td"
-include "RISCVSchedRocket32.td"
-include "RISCVSchedRocket64.td"
+include "RISCVSchedRocket.td"
 
 //===----------------------------------------------------------------------===//
 // RISC-V processors supported.
 //===----------------------------------------------------------------------===//
 
 def : ProcessorModel<"generic-rv32", NoSchedModel, []>;
 
 def : ProcessorModel<"generic-rv64", NoSchedModel, [Feature64Bit]>;
 
-def : ProcessorModel<"rocket-rv32", Rocket32Model, []>;
+def : ProcessorModel<"rocket-rv32", RocketModel, []>;
 
-def : ProcessorModel<"rocket-rv64", Rocket64Model, [Feature64Bit]>;
+def : ProcessorModel<"rocket-rv64", RocketModel, [Feature64Bit]>;
 
-def : ProcessorModel<"sifive-e31", Rocket32Model, [FeatureStdExtM,
-                                                   FeatureStdExtA,
-                                                   FeatureStdExtC]>;
+def : ProcessorModel<"sifive-e31", RocketModel, [FeatureStdExtA,
+                                                 FeatureStdExtC,
+                                                 FeatureStdExtM]>;
 
-def : ProcessorModel<"sifive-u54", Rocket64Model, [Feature64Bit,
-                                                   FeatureStdExtM,
-                                                   FeatureStdExtA,
-                                                   FeatureStdExtF,
-                                                   FeatureStdExtD,
-                                                   FeatureStdExtC]>;
+def : ProcessorModel<"sifive-u54", RocketModel, [Feature64Bit,
+                                                 FeatureStdExtC,
+                                                 FeatureStdExtF,
+                                                 FeatureStdExtM]>;
 
 //===----------------------------------------------------------------------===//
 // Define the RISC-V target.
 //===----------------------------------------------------------------------===//
 
 def RISCVInstrInfo : InstrInfo {
   let guessInstructionProperties = 0;
 }

llvm/lib/Target/RISCV/RISCVSchedRocket.td

+//==- RISCVSchedRocket.td - Rocket Scheduling Definitions -*- 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
+//
+//===----------------------------------------------------------------------===//
+
+// ===---------------------------------------------------------------------===//
+// The following definitions describe the simpler per-operand machine model.
+// This works with MachineScheduler. See MCSchedule.h for details.
+
+// Rocket machine model for scheduling and other instruction cost heuristics.
+def RocketModel : SchedMachineModel {
+  let MicroOpBufferSize = 0; // Rocket is in-order.
+  let IssueWidth = 1;        // 1 micro-op is dispatched per cycle.
+  let LoadLatency = 3;
+  let MispredictPenalty = 3;
+  let UnsupportedFeatures = [HasStdExtV, HasStdExtZvamo, HasStdExtZvlsseg];
+}
+
+//===----------------------------------------------------------------------===//
+// Define each kind of processor resource and number available.
+
+// Modeling each pipeline as a ProcResource using the BufferSize = 0 since
+// Rocket is in-order.
+
+let BufferSize = 0 in {
+def RocketUnitALU        : ProcResource<1>; // Int ALU
+def RocketUnitIMul       : ProcResource<1>; // Int Multiply
+def RocketUnitMem        : ProcResource<1>; // Load/Store
+def RocketUnitB          : ProcResource<1>; // Branch
+
+def RocketUnitFPALU      : ProcResource<1>; // FP ALU
+}
+
+let BufferSize = 1 in {
+def RocketUnitIDiv       : ProcResource<1>; // Int Division
+def RocketUnitFPDivSqrt  : ProcResource<1>; // FP Divide/Sqrt
+}
+
+//===----------------------------------------------------------------------===//
+
+let SchedModel = RocketModel in {
+
+// Branching
+def : WriteRes<WriteJmp, [RocketUnitB]>;
+def : WriteRes<WriteJal, [RocketUnitB]>;
+def : WriteRes<WriteJalr, [RocketUnitB]>;
+def : WriteRes<WriteJmpReg, [RocketUnitB]>;
+
+// Integer arithmetic and logic
+def : WriteRes<WriteIALU32, [RocketUnitALU]>;
+def : WriteRes<WriteIALU, [RocketUnitALU]>;
+def : WriteRes<WriteShift32, [RocketUnitALU]>;
+def : WriteRes<WriteShift, [RocketUnitALU]>;
+
+// Integer multiplication
+let Latency = 4 in {
+def : WriteRes<WriteIMul, [RocketUnitIMul]>;
+def : WriteRes<WriteIMul32, [RocketUnitIMul]>;
+}
+
+// Integer division
+// Worst case latency is used.
+def : WriteRes<WriteIDiv32, [RocketUnitIDiv]> {
+  let Latency = 34;
+  let ResourceCycles = [34];
+}
+def : WriteRes<WriteIDiv, [RocketUnitIDiv]> {
+  let Latency = 33;
+  let ResourceCycles = [33];
+}
+
+// Memory
+def : WriteRes<WriteSTB, [RocketUnitMem]>;
+def : WriteRes<WriteSTH, [RocketUnitMem]>;
+def : WriteRes<WriteSTW, [RocketUnitMem]>;
+def : WriteRes<WriteSTD, [RocketUnitMem]>;
+def : WriteRes<WriteFST32, [RocketUnitMem]>;
+def : WriteRes<WriteFST64, [RocketUnitMem]>;
+
+let Latency = 3 in {
+def : WriteRes<WriteLDB, [RocketUnitMem]>;
+def : WriteRes<WriteLDH, [RocketUnitMem]>;
+}
+
+let Latency = 2 in {
+def : WriteRes<WriteLDW, [RocketUnitMem]>;
+def : WriteRes<WriteLDWU, [RocketUnitMem]>;
+def : WriteRes<WriteLDD, [RocketUnitMem]>;
+def : WriteRes<WriteFLD32, [RocketUnitMem]>;
+def : WriteRes<WriteFLD64, [RocketUnitMem]>;
+
+// Atomic memory
+def : WriteRes<WriteAtomicW, [RocketUnitMem]>;
+def : WriteRes<WriteAtomicD, [RocketUnitMem]>;
+
+def : WriteRes<WriteAtomicLDW, [RocketUnitMem]>;
+def : WriteRes<WriteAtomicLDD, [RocketUnitMem]>;
+}
+
+def : WriteRes<WriteAtomicSTW, [RocketUnitMem]>;
+def : WriteRes<WriteAtomicSTD, [RocketUnitMem]>;
+
+// Single precision.
+let Latency = 4 in {
+def : WriteRes<WriteFALU32, [RocketUnitFPALU]>;
+def : WriteRes<WriteFSGNJ32, [RocketUnitFPALU]>;
+def : WriteRes<WriteFMinMax32, [RocketUnitFPALU]>;
+}
+
+// Double precision
+let Latency = 6 in {
+def : WriteRes<WriteFALU64, [RocketUnitFPALU]>;
+def : WriteRes<WriteFSGNJ64, [RocketUnitFPALU]>;
+def : WriteRes<WriteFMinMax64, [RocketUnitFPALU]>;
+}
+
+// Conversions
+let Latency = 2 in {
+def : WriteRes<WriteFCvtI32ToF32, [RocketUnitFPALU]>;
+def : WriteRes<WriteFCvtI32ToF64, [RocketUnitFPALU]>;
+def : WriteRes<WriteFCvtI64ToF32, [RocketUnitFPALU]>;
+def : WriteRes<WriteFCvtI64ToF64, [RocketUnitFPALU]>;
+def : WriteRes<WriteFCvtF32ToI32, [RocketUnitFPALU]>;
+def : WriteRes<WriteFCvtF32ToI64, [RocketUnitFPALU]>;
+def : WriteRes<WriteFCvtF64ToI32, [RocketUnitFPALU]>;
+def : WriteRes<WriteFCvtF64ToI64, [RocketUnitFPALU]>;
+def : WriteRes<WriteFCvtF32ToF64, [RocketUnitFPALU]>;
+def : WriteRes<WriteFCvtF64ToF32, [RocketUnitFPALU]>;
+
+def : WriteRes<WriteFClass32, [RocketUnitFPALU]>;
+def : WriteRes<WriteFClass64, [RocketUnitFPALU]>;
+def : WriteRes<WriteFCmp32, [RocketUnitFPALU]>;
+def : WriteRes<WriteFCmp64, [RocketUnitFPALU]>;
+def : WriteRes<WriteFMovF32ToI32, [RocketUnitFPALU]>;
+def : WriteRes<WriteFMovI32ToF32, [RocketUnitFPALU]>;
+def : WriteRes<WriteFMovF64ToI64, [RocketUnitFPALU]>;
+def : WriteRes<WriteFMovI64ToF64, [RocketUnitFPALU]>;
+}
+
+// FP multiplication
+let Latency = 5 in {
+def : WriteRes<WriteFMul32, [RocketUnitFPALU]>;
+def : WriteRes<WriteFMulAdd32, [RocketUnitFPALU]>;
+def : WriteRes<WriteFMulSub32, [RocketUnitFPALU]>;
+}
+
+let Latency = 7 in {
+def : WriteRes<WriteFMul64, [RocketUnitFPALU]>;
+def : WriteRes<WriteFMulAdd64, [RocketUnitFPALU]>;
+def : WriteRes<WriteFMulSub64, [RocketUnitFPALU]>;
+}
+
+// FP division
+// FP division unit on Rocket is not pipelined, so set resource cycles to latency.
+let Latency = 20, ResourceCycles = [20] in {
+def : WriteRes<WriteFDiv32, [RocketUnitFPDivSqrt]>;
+def : WriteRes<WriteFDiv64, [RocketUnitFPDivSqrt]>;
+}
+
+// FP square root unit on Rocket is not pipelined, so set resource cycles to latency.
+def : WriteRes<WriteFSqrt32, [RocketUnitFPDivSqrt]> { let Latency = 20;
+                                                      let ResourceCycles = [20]; }
+def : WriteRes<WriteFSqrt64, [RocketUnitFPDivSqrt]> { let Latency = 25;
+                                                      let ResourceCycles = [25]; }
+
+// Others
+def : WriteRes<WriteCSR, []>;
+def : WriteRes<WriteNop, []>;
+
+def : InstRW<[WriteIALU], (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<ReadShift, 0>;
+def : ReadAdvance<ReadShift32, 0>;
+def : ReadAdvance<ReadIDiv, 0>;
+def : ReadAdvance<ReadIDiv32, 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<ReadFMemBase, 0>;
+def : ReadAdvance<ReadFALU32, 0>;
+def : ReadAdvance<ReadFALU64, 0>;
+def : ReadAdvance<ReadFMul32, 0>;
+def : ReadAdvance<ReadFMulAdd32, 0>;
+def : ReadAdvance<ReadFMulSub32, 0>;
+def : ReadAdvance<ReadFMul64, 0>;
+def : ReadAdvance<ReadFMulAdd64, 0>;
+def : ReadAdvance<ReadFMulSub64, 0>;
+def : ReadAdvance<ReadFDiv32, 0>;
+def : ReadAdvance<ReadFDiv64, 0>;
+def : ReadAdvance<ReadFSqrt32, 0>;
+def : ReadAdvance<ReadFSqrt64, 0>;
+def : ReadAdvance<ReadFCmp32, 0>;
+def : ReadAdvance<ReadFCmp64, 0>;
+def : ReadAdvance<ReadFSGNJ32, 0>;
+def : ReadAdvance<ReadFSGNJ64, 0>;
+def : ReadAdvance<ReadFMinMax32, 0>;
+def : ReadAdvance<ReadFMinMax64, 0>;
+def : ReadAdvance<ReadFCvtF32ToI32, 0>;
+def : ReadAdvance<ReadFCvtF32ToI64, 0>;
+def : ReadAdvance<ReadFCvtF64ToI32, 0>;
+def : ReadAdvance<ReadFCvtF64ToI64, 0>;
+def : ReadAdvance<ReadFCvtI32ToF32, 0>;
+def : ReadAdvance<ReadFCvtI32ToF64, 0>;
+def : ReadAdvance<ReadFCvtI64ToF32, 0>;
+def : ReadAdvance<ReadFCvtI64ToF64, 0>;
+def : ReadAdvance<ReadFCvtF32ToF64, 0>;
+def : ReadAdvance<ReadFCvtF64ToF32, 0>;
+def : ReadAdvance<ReadFMovF32ToI32, 0>;
+def : ReadAdvance<ReadFMovI32ToF32, 0>;
+def : ReadAdvance<ReadFMovF64ToI64, 0>;
+def : ReadAdvance<ReadFMovI64ToF64, 0>;
+def : ReadAdvance<ReadFClass32, 0>;
+def : ReadAdvance<ReadFClass64, 0>;
+}

llvm/lib/Target/RISCV/RISCVSchedRocket32.td

-//==- RISCVSchedRocket32.td - Rocket Scheduling Definitions -*- 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
-//
-//===----------------------------------------------------------------------===//
-
-// ===---------------------------------------------------------------------===//
-// The following definitions describe the simpler per-operand machine model.
-// This works with MachineScheduler. See MCSchedule.h for details.
-
-// Rocket machine model for scheduling and other instruction cost heuristics.
-def Rocket32Model : SchedMachineModel {
-  let MicroOpBufferSize = 0; // Explicitly set to zero since Rocket is in-order.
-  let IssueWidth = 1;        // 1 micro-ops are dispatched per cycle.
-  let LoadLatency = 3;
-  let MispredictPenalty = 3;
-  let CompleteModel = 1;
-  let UnsupportedFeatures = [HasStdExtV, HasStdExtZvlsseg, HasStdExtZvamo];
-}
-
-//===----------------------------------------------------------------------===//
-// Define each kind of processor resource and number available.
-
-// Modeling each pipeline as a ProcResource using the BufferSize = 0 since
-// Rocket is in-order.
-
-let BufferSize = 0 in {
-def Rocket32UnitALU        : ProcResource<1>; // Int ALU
-def Rocket32UnitIMul       : ProcResource<1>; // Int Multiply
-def Rocket32UnitMem        : ProcResource<1>; // Load/Store
-def Rocket32UnitB          : ProcResource<1>; // Branch
-
-def Rocket32UnitFPALU      : ProcResource<1>; // FP ALU
-}
-
-let BufferSize = 1 in {
-def Rocket32UnitIDiv       : ProcResource<1>; // Int Division
-def Rocket32UnitFPDivSqrt  : ProcResource<1>; // FP Divide/Sqrt'
-}
-
-//===----------------------------------------------------------------------===//
-// Subtarget-specific SchedWrite types which both map the ProcResources and
-// set the latency.
-
-let SchedModel = Rocket32Model in {
-
-def : WriteRes<WriteJmp, [Rocket32UnitB]>;
-def : WriteRes<WriteJal, [Rocket32UnitB]>;
-def : WriteRes<WriteJalr, [Rocket32UnitB]>;
-def : WriteRes<WriteJmpReg, [Rocket32UnitB]>;
-
-def : WriteRes<WriteIALU, [Rocket32UnitALU]>;
-def : WriteRes<WriteShift, [Rocket32UnitALU]>;
-
-// Multiplies on Rocket differ by implementation; placeholder until
-// we can determine how to read from command line
-def : WriteRes<WriteIMul, [Rocket32UnitIMul]> { let Latency = 4; }
-
-// 32-bit divides have worse case latency of 34 cycle
-def : WriteRes<WriteIDiv, [Rocket32UnitIDiv]> {
-  let Latency = 34;
-  let ResourceCycles = [34];
-}
-
-// Memory
-def : WriteRes<WriteSTB, [Rocket32UnitMem]>;
-def : WriteRes<WriteSTH, [Rocket32UnitMem]>;
-def : WriteRes<WriteSTW, [Rocket32UnitMem]>;
-def : WriteRes<WriteFST32, [Rocket32UnitMem]>;
-def : WriteRes<WriteFST64, [Rocket32UnitMem]>;
-
-let Latency = 3 in {
-def : WriteRes<WriteLDB, [Rocket32UnitMem]>;
-def : WriteRes<WriteLDH, [Rocket32UnitMem]>;
-def : WriteRes<WriteCSR, [Rocket32UnitALU]>;
-}
-
-let Latency = 2 in {
-def : WriteRes<WriteLDW, [Rocket32UnitMem]>;
-def : WriteRes<WriteFLD32, [Rocket32UnitMem]>;
-def : WriteRes<WriteFLD64, [Rocket32UnitMem]>;
-
-def : WriteRes<WriteAtomicW, [Rocket32UnitMem]>;
-def : WriteRes<WriteAtomicLDW, [Rocket32UnitMem]>;
-}
-
-def : WriteRes<WriteAtomicSTW, [Rocket32UnitMem]>;
-
-// Most FP single precision operations are 4 cycles
-let Latency = 4 in {
-def : WriteRes<WriteFALU32, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFSGNJ32, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFMinMax32, [Rocket32UnitFPALU]>;
-}
-
-// Most FP double precision operations are 6 cycles
-let Latency = 6 in {
-def : WriteRes<WriteFALU64, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFSGNJ64, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFMinMax64, [Rocket32UnitFPALU]>;
-}
-
-let Latency = 2 in {
-def : WriteRes<WriteFCvtI32ToF32, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFCvtI32ToF64, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFCvtF32ToI32, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFCvtF64ToI32, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFCvtF32ToF64, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFCvtF64ToF32, [Rocket32UnitFPALU]>;
-
-def : WriteRes<WriteFClass32, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFClass64, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFCmp32, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFCmp64, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFMovF32ToI32, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFMovI32ToF32, [Rocket32UnitFPALU]>;
-}
-
-let Latency = 5 in {
-def : WriteRes<WriteFMul32, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFMulAdd32, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFMulSub32, [Rocket32UnitFPALU]>;
-}
-
-let Latency = 7 in {
-def : WriteRes<WriteFMul64, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFMulAdd64, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFMulSub64, [Rocket32UnitFPALU]>;
-}
-
-// FP Divide unit on Rocket is not pipelined, so set resource cycles to latency
-let Latency = 20, ResourceCycles = [20] in {
-def : WriteRes<WriteFDiv32, [Rocket32UnitFPDivSqrt]>;
-def : WriteRes<WriteFDiv64, [Rocket32UnitFPDivSqrt]>;
-}
-
-// FP Sqrt unit on Rocket is not pipelined, so set resource cycles to latency
-def : WriteRes<WriteFSqrt32, [Rocket32UnitFPDivSqrt]> { let Latency = 20;
-                                                        let ResourceCycles = [20];}
-def : WriteRes<WriteFSqrt64, [Rocket32UnitFPDivSqrt]> { let Latency = 25;
-                                                        let ResourceCycles = [25];}
-
-def : WriteRes<WriteNop, []>;
-
-def : InstRW<[WriteIALU], (instrs COPY)>;
-
-let Unsupported = 1 in {
-def : WriteRes<WriteIALU32, []>;
-def : WriteRes<WriteShift32, []>;
-def : WriteRes<WriteIMul32, []>;
-def : WriteRes<WriteIDiv32, []>;
-def : WriteRes<WriteSTD, []>;
-def : WriteRes<WriteLDWU, []>;
-def : WriteRes<WriteLDD, []>;
-def : WriteRes<WriteAtomicD, []>;
-def : WriteRes<WriteAtomicLDD, []>;
-def : WriteRes<WriteAtomicSTD, []>;
-def : WriteRes<WriteFCvtI64ToF32, []>;
-def : WriteRes<WriteFCvtI64ToF64, []>;
-def : WriteRes<WriteFCvtF64ToI64, []>;
-def : WriteRes<WriteFCvtF32ToI64, []>;
-def : WriteRes<WriteFMovI64ToF64, []>;
-def : WriteRes<WriteFMovF64ToI64, []>;
-}
-
-//===----------------------------------------------------------------------===//
-// Subtarget-specific SchedRead types with cycles.
-// Dummy definitions for RocketCore.
-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<ReadShift, 0>;
-def : ReadAdvance<ReadShift32, 0>;
-def : ReadAdvance<ReadIDiv, 0>;
-def : ReadAdvance<ReadIDiv32, 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<ReadFMemBase, 0>;
-def : ReadAdvance<ReadFALU32, 0>;
-def : ReadAdvance<ReadFALU64, 0>;
-def : ReadAdvance<ReadFMul32, 0>;
-def : ReadAdvance<ReadFMulAdd32, 0>;
-def : ReadAdvance<ReadFMulSub32, 0>;
-def : ReadAdvance<ReadFMul64, 0>;
-def : ReadAdvance<ReadFMulAdd64, 0>;
-def : ReadAdvance<ReadFMulSub64, 0>;
-def : ReadAdvance<ReadFDiv32, 0>;
-def : ReadAdvance<ReadFDiv64, 0>;
-def : ReadAdvance<ReadFSqrt32, 0>;
-def : ReadAdvance<ReadFSqrt64, 0>;
-def : ReadAdvance<ReadFCmp32, 0>;
-def : ReadAdvance<ReadFCmp64, 0>;
-def : ReadAdvance<ReadFSGNJ32, 0>;
-def : ReadAdvance<ReadFSGNJ64, 0>;
-def : ReadAdvance<ReadFMinMax32, 0>;
-def : ReadAdvance<ReadFMinMax64, 0>;
-def : ReadAdvance<ReadFCvtF32ToI32, 0>;
-def : ReadAdvance<ReadFCvtF32ToI64, 0>;
-def : ReadAdvance<ReadFCvtF64ToI32, 0>;
-def : ReadAdvance<ReadFCvtF64ToI64, 0>;
-def : ReadAdvance<ReadFCvtI32ToF32, 0>;
-def : ReadAdvance<ReadFCvtI32ToF64, 0>;
-def : ReadAdvance<ReadFCvtI64ToF32, 0>;
-def : ReadAdvance<ReadFCvtI64ToF64, 0>;
-def : ReadAdvance<ReadFCvtF32ToF64, 0>;
-def : ReadAdvance<ReadFCvtF64ToF32, 0>;
-def : ReadAdvance<ReadFMovF32ToI32, 0>;
-def : ReadAdvance<ReadFMovI32ToF32, 0>;
-def : ReadAdvance<ReadFMovF64ToI64, 0>;
-def : ReadAdvance<ReadFMovI64ToF64, 0>;
-def : ReadAdvance<ReadFClass32, 0>;
-def : ReadAdvance<ReadFClass64, 0>;
-}

llvm/lib/Target/RISCV/RISCVSchedRocket64.td

-//==- RISCVSchedRocket64.td - Rocket Scheduling Definitions -*- 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
-//
-//===----------------------------------------------------------------------===//
-
-// ===---------------------------------------------------------------------===//
-// The following definitions describe the simpler per-operand machine model.
-// This works with MachineScheduler. See MCSchedule.h for details.
-
-// Rocket machine model for scheduling and other instruction cost heuristics.
-def Rocket64Model : SchedMachineModel {
-  let MicroOpBufferSize = 0; // Explicitly set to zero since Rocket is in-order.
-  let IssueWidth = 1;        // 1 micro-ops are dispatched per cycle.
-  let LoadLatency = 3;
-  let MispredictPenalty = 3;
-  let UnsupportedFeatures = [HasStdExtV, HasStdExtZvlsseg, HasStdExtZvamo];
-}
-
-//===----------------------------------------------------------------------===//
-// Define each kind of processor resource and number available.
-
-// Modeling each pipeline as a ProcResource using the BufferSize = 0 since
-// Rocket is in-order.
-
-let BufferSize = 0 in {
-def Rocket64UnitALU        : ProcResource<1>; // Int ALU
-def Rocket64UnitIMul       : ProcResource<1>; // Int Multiply
-def Rocket64UnitMem        : ProcResource<1>; // Load/Store
-def Rocket64UnitB          : ProcResource<1>; // Branch
-
-def Rocket64UnitFPALU      : ProcResource<1>; // FP ALU
-}
-
-let BufferSize = 1 in {
-def Rocket64UnitIDiv       : ProcResource<1>; // Int Division
-def Rocket64UnitFPDivSqrt  : ProcResource<1>; // FP Divide/Sqrt
-}
-
-//===----------------------------------------------------------------------===//
-// Subtarget-specific SchedWrite types which both map the ProcResources and
-// set the latency.
-
-let SchedModel = Rocket64Model in {
-
-def : WriteRes<WriteJmp, [Rocket64UnitB]>;
-def : WriteRes<WriteJal, [Rocket64UnitB]>;
-def : WriteRes<WriteJalr, [Rocket64UnitB]>;
-def : WriteRes<WriteJmpReg, [Rocket64UnitB]>;
-
-def : WriteRes<WriteIALU32, [Rocket64UnitALU]>;
-def : WriteRes<WriteIALU, [Rocket64UnitALU]>;
-def : WriteRes<WriteShift32, [Rocket64UnitALU]>;
-def : WriteRes<WriteShift, [Rocket64UnitALU]>;
-
-let Latency = 4 in {
-def : WriteRes<WriteIMul, [Rocket64UnitIMul]>;
-def : WriteRes<WriteIMul32, [Rocket64UnitIMul]>;
-}
-
-// Integer divide varies based on operand magnitude and sign; worse case latency is 34.
-def : WriteRes<WriteIDiv32, [Rocket64UnitIDiv]> {
-  let Latency = 34;
-  let ResourceCycles = [34];
-}
-def : WriteRes<WriteIDiv, [Rocket64UnitIDiv]> {
-  let Latency = 33;
-  let ResourceCycles = [33];
-}
-
-// Memory
-def : WriteRes<WriteSTB, [Rocket64UnitMem]>;
-def : WriteRes<WriteSTH, [Rocket64UnitMem]>;
-def : WriteRes<WriteSTW, [Rocket64UnitMem]>;
-def : WriteRes<WriteSTD, [Rocket64UnitMem]>;
-def : WriteRes<WriteFST32, [Rocket64UnitMem]>;
-def : WriteRes<WriteFST64, [Rocket64UnitMem]>;
-
-let Latency = 3 in {
-def : WriteRes<WriteLDB, [Rocket64UnitMem]>;
-def : WriteRes<WriteLDH, [Rocket64UnitMem]>;
-def : WriteRes<WriteCSR, [Rocket64UnitALU]>;
-}
-
-let Latency = 2 in {
-def : WriteRes<WriteLDW, [Rocket64UnitMem]>;
-def : WriteRes<WriteLDWU, [Rocket64UnitMem]>;
-def : WriteRes<WriteLDD, [Rocket64UnitMem]>;
-def : WriteRes<WriteFLD32, [Rocket64UnitMem]>;
-def : WriteRes<WriteFLD64, [Rocket64UnitMem]>;
-
-def : WriteRes<WriteAtomicW, [Rocket64UnitMem]>;
-def : WriteRes<WriteAtomicD, [Rocket64UnitMem]>;
-
-def : WriteRes<WriteAtomicLDW, [Rocket64UnitMem]>;
-def : WriteRes<WriteAtomicLDD, [Rocket64UnitMem]>;
-}
-
-def : WriteRes<WriteAtomicSTW, [Rocket64UnitMem]>;
-def : WriteRes<WriteAtomicSTD, [Rocket64UnitMem]>;
-
-// Most FP single precision operations are 4 cycles
-let Latency = 4 in {
-def : WriteRes<WriteFALU32, [Rocket64UnitFPALU]>;
-def : WriteRes<WriteFSGNJ32, [Rocket64UnitFPALU]>;
-def : WriteRes<WriteFMinMax32, [Rocket64UnitFPALU]>;
-}
-
-let Latency = 6 in {
-// Most FP double precision operations are 6 cycles
-def : WriteRes<WriteFALU64, [Rocket64UnitFPALU]>;
-def : WriteRes<WriteFSGNJ64, [Rocket64UnitFPALU]>;
-def : WriteRes<WriteFMinMax64, [Rocket64UnitFPALU]>;
-}
-
-// Conversion instructions
-let Latency = 2 in {
-def : WriteRes<WriteFCvtI32ToF32, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFCvtI32ToF64, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFCvtI64ToF32, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFCvtI64ToF64, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFCvtF32ToI32, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFCvtF32ToI64, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFCvtF64ToI32, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFCvtF64ToI64, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFCvtF32ToF64, [Rocket32UnitFPALU]>;
-def : WriteRes<WriteFCvtF64ToF32, [Rocket32UnitFPALU]>;
-
-def : WriteRes<WriteFClass32, [Rocket64UnitFPALU]>;
-def : WriteRes<WriteFClass64, [Rocket64UnitFPALU]>;
-def : WriteRes<WriteFCmp32, [Rocket64UnitFPALU]>;
-def : WriteRes<WriteFCmp64, [Rocket64UnitFPALU]>;
-def : WriteRes<WriteFMovF32ToI32, [Rocket64UnitFPALU]>;
-def : WriteRes<WriteFMovI32ToF32, [Rocket64UnitFPALU]>;
-def : WriteRes<WriteFMovF64ToI64, [Rocket64UnitFPALU]>;
-def : WriteRes<WriteFMovI64ToF64, [Rocket64UnitFPALU]>;
-}
-
-let Latency = 5 in {
-def : WriteRes<WriteFMul32, [Rocket64UnitFPALU]>;
-def : WriteRes<WriteFMulAdd32, [Rocket64UnitFPALU]>;
-def : WriteRes<WriteFMulSub32, [Rocket64UnitFPALU]>;
-}
-
-let Latency = 7 in {
-def : WriteRes<WriteFMul64, [Rocket64UnitFPALU]>;
-def : WriteRes<WriteFMulAdd64, [Rocket64UnitFPALU]>;
-def : WriteRes<WriteFMulSub64, [Rocket64UnitFPALU]>;
-}
-
-// FP Divide unit on Rocket is not pipelined, so set resource cycles to latency
-let Latency = 20, ResourceCycles = [20] in {
-def : WriteRes<WriteFDiv32, [Rocket64UnitFPDivSqrt]>;
-def : WriteRes<WriteFDiv64, [Rocket64UnitFPDivSqrt]>;
-}
-
-// FP Sqrt unit on Rocket is not pipelined, so set resource cycles to latency
-def : WriteRes<WriteFSqrt32, [Rocket64UnitFPDivSqrt]> { let Latency = 20;
-                                                        let ResourceCycles = [20]; }
-def : WriteRes<WriteFSqrt64, [Rocket64UnitFPDivSqrt]> { let Latency = 25;
-                                                        let ResourceCycles = [25]; }
-
-def : WriteRes<WriteNop, []>;
-
-def : InstRW<[WriteIALU], (instrs COPY)>;
-
-//===----------------------------------------------------------------------===//
-// Subtarget-specific SchedRead types with cycles.
-// Dummy definitions for RocketCore.
-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<ReadShift, 0>;
-def : ReadAdvance<ReadShift32, 0>;
-def : ReadAdvance<ReadIDiv, 0>;
-def : ReadAdvance<ReadIDiv32, 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<ReadFMemBase, 0>;
-def : ReadAdvance<ReadFALU32, 0>;
-def : ReadAdvance<ReadFALU64, 0>;
-def : ReadAdvance<ReadFMul32, 0>;
-def : ReadAdvance<ReadFMulAdd32, 0>;
-def : ReadAdvance<ReadFMulSub32, 0>;
-def : ReadAdvance<ReadFMul64, 0>;
-def : ReadAdvance<ReadFMulAdd64, 0>;
-def : ReadAdvance<ReadFMulSub64, 0>;
-def : ReadAdvance<ReadFDiv32, 0>;
-def : ReadAdvance<ReadFDiv64, 0>;
-def : ReadAdvance<ReadFSqrt32, 0>;
-def : ReadAdvance<ReadFSqrt64, 0>;
-def : ReadAdvance<ReadFCmp32, 0>;
-def : ReadAdvance<ReadFCmp64, 0>;
-def : ReadAdvance<ReadFSGNJ32, 0>;
-def : ReadAdvance<ReadFSGNJ64, 0>;
-def : ReadAdvance<ReadFMinMax32, 0>;
-def : ReadAdvance<ReadFMinMax64, 0>;
-def : ReadAdvance<ReadFCvtF32ToI32, 0>;
-def : ReadAdvance<ReadFCvtF32ToI64, 0>;
-def : ReadAdvance<ReadFCvtF64ToI32, 0>;
-def : ReadAdvance<ReadFCvtF64ToI64, 0>;
-def : ReadAdvance<ReadFCvtI32ToF32, 0>;
-def : ReadAdvance<ReadFCvtI32ToF64, 0>;
-def : ReadAdvance<ReadFCvtI64ToF32, 0>;
-def : ReadAdvance<ReadFCvtI64ToF64, 0>;
-def : ReadAdvance<ReadFCvtF32ToF64, 0>;
-def : ReadAdvance<ReadFCvtF64ToF32, 0>;
-def : ReadAdvance<ReadFMovF32ToI32, 0>;
-def : ReadAdvance<ReadFMovI32ToF32, 0>;
-def : ReadAdvance<ReadFMovF64ToI64, 0>;
-def : ReadAdvance<ReadFMovI64ToF64, 0>;
-def : ReadAdvance<ReadFClass32, 0>;
-def : ReadAdvance<ReadFClass64, 0>;
-}