File Manager

003 File Manager

Current Path: /usr/src/contrib/llvm-project/llvm/lib/CodeGen

usr / src / contrib / llvm-project / llvm / lib / CodeGen /

📁 ..
📄 AggressiveAntiDepBreaker.cpp(37.23 KB)
📄 AggressiveAntiDepBreaker.h(6.8 KB)
📄 AllocationOrder.cpp(1.96 KB)
📄 AllocationOrder.h(2.96 KB)
📄 Analysis.cpp(32.62 KB)
📁 AsmPrinter
📄 AtomicExpandPass.cpp(71.86 KB)
📄 BBSectionsPrepare.cpp(18.8 KB)
📄 BasicTargetTransformInfo.cpp(1.53 KB)
📄 BranchFolding.cpp(77.92 KB)
📄 BranchFolding.h(7.36 KB)
📄 BranchRelaxation.cpp(19.45 KB)
📄 BreakFalseDeps.cpp(9.79 KB)
📄 BuiltinGCs.cpp(4.88 KB)
📄 CFGuardLongjmp.cpp(3.73 KB)
📄 CFIInstrInserter.cpp(17.53 KB)
📄 CalcSpillWeights.cpp(10.22 KB)
📄 CallingConvLower.cpp(10.4 KB)
📄 CodeGen.cpp(5.28 KB)
📄 CodeGenPrepare.cpp(295.01 KB)
📄 CommandFlags.cpp(24.89 KB)
📄 CriticalAntiDepBreaker.cpp(27.91 KB)
📄 CriticalAntiDepBreaker.h(4.22 KB)
📄 DFAPacketizer.cpp(10.91 KB)
📄 DeadMachineInstructionElim.cpp(6.52 KB)
📄 DetectDeadLanes.cpp(20.74 KB)
📄 DwarfEHPrepare.cpp(9.01 KB)
📄 EarlyIfConversion.cpp(37.51 KB)
📄 EdgeBundles.cpp(3.21 KB)
📄 ExecutionDomainFix.cpp(14.67 KB)
📄 ExpandMemCmp.cpp(33.66 KB)
📄 ExpandPostRAPseudos.cpp(7.28 KB)
📄 ExpandReductions.cpp(7.23 KB)
📄 FEntryInserter.cpp(1.81 KB)
📄 FaultMaps.cpp(4.99 KB)
📄 FinalizeISel.cpp(2.65 KB)
📄 FixupStatepointCallerSaved.cpp(11.06 KB)
📄 FuncletLayout.cpp(2.21 KB)
📄 GCMetadata.cpp(5.1 KB)
📄 GCMetadataPrinter.cpp(748 B)
📄 GCRootLowering.cpp(11.46 KB)
📄 GCStrategy.cpp(708 B)
📁 GlobalISel
📄 GlobalMerge.cpp(24.52 KB)
📄 HardwareLoops.cpp(18.44 KB)
📄 IfConversion.cpp(89.43 KB)
📄 ImplicitNullChecks.cpp(25.14 KB)
📄 IndirectBrExpandPass.cpp(7.79 KB)
📄 InlineSpiller.cpp(58.24 KB)
📄 InterferenceCache.cpp(8.83 KB)
📄 InterferenceCache.h(7.22 KB)
📄 InterleavedAccessPass.cpp(16.59 KB)
📄 InterleavedLoadCombinePass.cpp(42.35 KB)
📄 IntrinsicLowering.cpp(17.08 KB)
📄 LLVMTargetMachine.cpp(10.25 KB)
📄 LatencyPriorityQueue.cpp(5.64 KB)
📄 LazyMachineBlockFrequencyInfo.cpp(3.36 KB)
📄 LexicalScopes.cpp(12.16 KB)
📄 LiveDebugValues.cpp(78.98 KB)
📄 LiveDebugVariables.cpp(51.79 KB)
📄 LiveDebugVariables.h(2.15 KB)
📄 LiveInterval.cpp(46.67 KB)
📄 LiveIntervalCalc.cpp(7.62 KB)
📄 LiveIntervalUnion.cpp(6.36 KB)
📄 LiveIntervals.cpp(64.59 KB)
📄 LivePhysRegs.cpp(11.08 KB)
📄 LiveRangeCalc.cpp(15.72 KB)
📄 LiveRangeEdit.cpp(17.03 KB)
📄 LiveRangeShrink.cpp(8.69 KB)
📄 LiveRangeUtils.h(2.12 KB)
📄 LiveRegMatrix.cpp(7.47 KB)
📄 LiveRegUnits.cpp(4.72 KB)
📄 LiveStacks.cpp(2.95 KB)
📄 LiveVariables.cpp(30.26 KB)
📄 LocalStackSlotAllocation.cpp(17.26 KB)
📄 LoopTraversal.cpp(2.89 KB)
📄 LowLevelType.cpp(1.93 KB)
📄 LowerEmuTLS.cpp(5.66 KB)
📄 MBFIWrapper.cpp(1.57 KB)
📄 MIRCanonicalizerPass.cpp(12.46 KB)
📄 MIRNamerPass.cpp(2.16 KB)
📁 MIRParser
📄 MIRPrinter.cpp(32.67 KB)
📄 MIRPrintingPass.cpp(1.99 KB)
📄 MIRVRegNamerUtils.cpp(6.04 KB)
📄 MIRVRegNamerUtils.h(3.25 KB)
📄 MachineBasicBlock.cpp(50.47 KB)
📄 MachineBlockFrequencyInfo.cpp(10.13 KB)
📄 MachineBlockPlacement.cpp(137.61 KB)
📄 MachineBranchProbabilityInfo.cpp(3.5 KB)
📄 MachineCSE.cpp(31.82 KB)
📄 MachineCombiner.cpp(28.13 KB)
📄 MachineCopyPropagation.cpp(29.21 KB)
📄 MachineDebugify.cpp(6.47 KB)
📄 MachineDominanceFrontier.cpp(1.83 KB)
📄 MachineDominators.cpp(4.86 KB)
📄 MachineFrameInfo.cpp(9.77 KB)
📄 MachineFunction.cpp(42.97 KB)
📄 MachineFunctionPass.cpp(4.78 KB)
📄 MachineFunctionPrinterPass.cpp(2.3 KB)
📄 MachineInstr.cpp(76.39 KB)
📄 MachineInstrBundle.cpp(11.49 KB)
📄 MachineLICM.cpp(57.05 KB)
📄 MachineLoopInfo.cpp(4.98 KB)
📄 MachineLoopUtils.cpp(5.16 KB)
📄 MachineModuleInfo.cpp(9.9 KB)
📄 MachineModuleInfoImpls.cpp(1.5 KB)
📄 MachineOperand.cpp(39.6 KB)
📄 MachineOptimizationRemarkEmitter.cpp(3.29 KB)
📄 MachineOutliner.cpp(42.13 KB)
📄 MachinePipeliner.cpp(111.33 KB)
📄 MachinePostDominators.cpp(2.42 KB)
📄 MachineRegionInfo.cpp(4.75 KB)
📄 MachineRegisterInfo.cpp(22.97 KB)
📄 MachineSSAUpdater.cpp(12.99 KB)
📄 MachineScheduler.cpp(136.89 KB)
📄 MachineSink.cpp(51.94 KB)
📄 MachineSizeOpts.cpp(8.76 KB)
📄 MachineStripDebug.cpp(3.76 KB)
📄 MachineTraceMetrics.cpp(49.58 KB)
📄 MachineVerifier.cpp(107.98 KB)
📄 MacroFusion.cpp(7.55 KB)
📄 ModuloSchedule.cpp(85.09 KB)
📄 NonRelocatableStringpool.cpp(1.65 KB)
📄 OptimizePHIs.cpp(6.7 KB)
📄 PHIElimination.cpp(27.73 KB)
📄 PHIEliminationUtils.cpp(2.56 KB)
📄 PHIEliminationUtils.h(972 B)
📄 ParallelCG.cpp(3.71 KB)
📄 PatchableFunction.cpp(3.44 KB)
📄 PeepholeOptimizer.cpp(78.41 KB)
📄 PostRAHazardRecognizer.cpp(3.5 KB)
📄 PostRASchedulerList.cpp(24.31 KB)
📄 PreISelIntrinsicLowering.cpp(7.91 KB)
📄 ProcessImplicitDefs.cpp(5.4 KB)
📄 PrologEpilogInserter.cpp(50.45 KB)
📄 PseudoSourceValue.cpp(4.71 KB)
📄 RDFGraph.cpp(58.39 KB)
📄 RDFLiveness.cpp(40.7 KB)
📄 RDFRegisters.cpp(11.29 KB)
📄 ReachingDefAnalysis.cpp(21.74 KB)
📄 RegAllocBase.cpp(6.31 KB)
📄 RegAllocBase.h(4.63 KB)
📄 RegAllocBasic.cpp(11.33 KB)
📄 RegAllocFast.cpp(45.78 KB)
📄 RegAllocGreedy.cpp(123.32 KB)
📄 RegAllocPBQP.cpp(33.14 KB)
📄 RegUsageInfoCollector.cpp(7.39 KB)
📄 RegUsageInfoPropagate.cpp(5.07 KB)
📄 RegisterClassInfo.cpp(6.62 KB)
📄 RegisterCoalescer.cpp(151.71 KB)
📄 RegisterCoalescer.h(4.04 KB)
📄 RegisterPressure.cpp(48.86 KB)
📄 RegisterScavenging.cpp(27.48 KB)
📄 RegisterUsageInfo.cpp(3.18 KB)
📄 RenameIndependentSubregs.cpp(14.79 KB)
📄 ResetMachineFunctionPass.cpp(3.48 KB)
📄 SafeStack.cpp(34.12 KB)
📄 SafeStackLayout.cpp(5.3 KB)
📄 SafeStackLayout.h(2.41 KB)
📄 ScalarizeMaskedMemIntrin.cpp(31.46 KB)
📄 ScheduleDAG.cpp(21.34 KB)
📄 ScheduleDAGInstrs.cpp(54.59 KB)
📄 ScheduleDAGPrinter.cpp(3.21 KB)
📄 ScoreboardHazardRecognizer.cpp(7.96 KB)
📁 SelectionDAG
📄 ShadowStackGCLowering.cpp(14.16 KB)
📄 ShrinkWrap.cpp(23.03 KB)
📄 SjLjEHPrepare.cpp(18.93 KB)
📄 SlotIndexes.cpp(9.35 KB)
📄 SpillPlacement.cpp(12.58 KB)
📄 SpillPlacement.h(6.67 KB)
📄 SplitKit.cpp(66.39 KB)
📄 SplitKit.h(23.7 KB)
📄 StackColoring.cpp(49.03 KB)
📄 StackMapLivenessAnalysis.cpp(6.16 KB)
📄 StackMaps.cpp(19.74 KB)
📄 StackProtector.cpp(22.94 KB)
📄 StackSlotColoring.cpp(17.12 KB)
📄 SwiftErrorValueTracking.cpp(11.37 KB)
📄 SwitchLoweringUtils.cpp(18.33 KB)
📄 TailDuplication.cpp(3.32 KB)
📄 TailDuplicator.cpp(38.29 KB)
📄 TargetFrameLoweringImpl.cpp(6.24 KB)
📄 TargetInstrInfo.cpp(51.1 KB)
📄 TargetLoweringBase.cpp(82.53 KB)
📄 TargetLoweringObjectFileImpl.cpp(80.52 KB)
📄 TargetOptionsImpl.cpp(2 KB)
📄 TargetPassConfig.cpp(48.89 KB)
📄 TargetRegisterInfo.cpp(19.15 KB)
📄 TargetSchedule.cpp(13.16 KB)
📄 TargetSubtargetInfo.cpp(1.89 KB)
📄 TwoAddressInstructionPass.cpp(62.08 KB)
📄 TypePromotion.cpp(32.46 KB)
📄 UnreachableBlockElim.cpp(7.48 KB)
📄 ValueTypes.cpp(19.87 KB)
📄 VirtRegMap.cpp(21.4 KB)
📄 WasmEHPrepare.cpp(17.48 KB)
📄 WinEHPrepare.cpp(51.16 KB)
📄 XRayInstrumentation.cpp(9.66 KB)

Editing: MacroFusion.cpp

//===- MacroFusion.cpp - Macro Fusion -------------------------------------===//
//
// 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
//
//===----------------------------------------------------------------------===//
//
/// \file This file contains the implementation of the DAG scheduling mutation
/// to pair instructions back to back.
//
//===----------------------------------------------------------------------===//

#include "llvm/CodeGen/MacroFusion.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/CodeGen/MachineInstr.h"
#include "llvm/CodeGen/MachineScheduler.h"
#include "llvm/CodeGen/ScheduleDAG.h"
#include "llvm/CodeGen/ScheduleDAGMutation.h"
#include "llvm/CodeGen/TargetInstrInfo.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/raw_ostream.h"

#define DEBUG_TYPE "machine-scheduler"

STATISTIC(NumFused, "Number of instr pairs fused");

using namespace llvm;

static cl::opt<bool> EnableMacroFusion("misched-fusion", cl::Hidden,
  cl::desc("Enable scheduling for macro fusion."), cl::init(true));

static bool isHazard(const SDep &Dep) {
  return Dep.getKind() == SDep::Anti || Dep.getKind() == SDep::Output;
}

static SUnit *getPredClusterSU(const SUnit &SU) {
  for (const SDep &SI : SU.Preds)
    if (SI.isCluster())
      return SI.getSUnit();

return nullptr;
}

static bool hasLessThanNumFused(const SUnit &SU, unsigned FuseLimit) {
  unsigned Num = 1;
  const SUnit *CurrentSU = &SU;
  while ((CurrentSU = getPredClusterSU(*CurrentSU)) && Num < FuseLimit) Num ++;
  return Num < FuseLimit;
}

static bool fuseInstructionPair(ScheduleDAGInstrs &DAG, SUnit &FirstSU,
                                SUnit &SecondSU) {
  // Check that neither instr is already paired with another along the edge
  // between them.
  for (SDep &SI : FirstSU.Succs)
    if (SI.isCluster())
      return false;

for (SDep &SI : SecondSU.Preds)
    if (SI.isCluster())
      return false;
  // Though the reachability checks above could be made more generic,
  // perhaps as part of ScheduleDAGInstrs::addEdge(), since such edges are valid,
  // the extra computation cost makes it less interesting in general cases.

// Create a single weak edge between the adjacent instrs. The only effect is
  // to cause bottom-up scheduling to heavily prioritize the clustered instrs.
  if (!DAG.addEdge(&SecondSU, SDep(&FirstSU, SDep::Cluster)))
    return false;

// TODO - If we want to chain more than two instructions, we need to create
  // artifical edges to make dependencies from the FirstSU also dependent
  // on other chained instructions, and other chained instructions also
  // dependent on the dependencies of the SecondSU, to prevent them from being
  // scheduled into these chained instructions.
  assert(hasLessThanNumFused(FirstSU, 2) &&
         "Currently we only support chaining together two instructions");

// Adjust the latency between both instrs.
  for (SDep &SI : FirstSU.Succs)
    if (SI.getSUnit() == &SecondSU)
      SI.setLatency(0);

for (SDep &SI : SecondSU.Preds)
    if (SI.getSUnit() == &FirstSU)
      SI.setLatency(0);

LLVM_DEBUG(
      dbgs() << "Macro fuse: "; DAG.dumpNodeName(FirstSU); dbgs() << " - ";
      DAG.dumpNodeName(SecondSU); dbgs() << " /  ";
      dbgs() << DAG.TII->getName(FirstSU.getInstr()->getOpcode()) << " - "
             << DAG.TII->getName(SecondSU.getInstr()->getOpcode()) << '\n';);

// Make data dependencies from the FirstSU also dependent on the SecondSU to
  // prevent them from being scheduled between the FirstSU and the SecondSU.
  if (&SecondSU != &DAG.ExitSU)
    for (const SDep &SI : FirstSU.Succs) {
      SUnit *SU = SI.getSUnit();
      if (SI.isWeak() || isHazard(SI) ||
          SU == &DAG.ExitSU || SU == &SecondSU || SU->isPred(&SecondSU))
        continue;
      LLVM_DEBUG(dbgs() << "  Bind "; DAG.dumpNodeName(SecondSU);
                 dbgs() << " - "; DAG.dumpNodeName(*SU); dbgs() << '\n';);
      DAG.addEdge(SU, SDep(&SecondSU, SDep::Artificial));
    }

// Make the FirstSU also dependent on the dependencies of the SecondSU to
  // prevent them from being scheduled between the FirstSU and the SecondSU.
  if (&FirstSU != &DAG.EntrySU) {
    for (const SDep &SI : SecondSU.Preds) {
      SUnit *SU = SI.getSUnit();
      if (SI.isWeak() || isHazard(SI) || &FirstSU == SU || FirstSU.isSucc(SU))
        continue;
      LLVM_DEBUG(dbgs() << "  Bind "; DAG.dumpNodeName(*SU); dbgs() << " - ";
                 DAG.dumpNodeName(FirstSU); dbgs() << '\n';);
      DAG.addEdge(&FirstSU, SDep(SU, SDep::Artificial));
    }
    // ExitSU comes last by design, which acts like an implicit dependency
    // between ExitSU and any bottom root in the graph. We should transfer
    // this to FirstSU as well.
    if (&SecondSU == &DAG.ExitSU) {
      for (SUnit &SU : DAG.SUnits) {
        if (SU.Succs.empty())
          DAG.addEdge(&FirstSU, SDep(&SU, SDep::Artificial));
      }
    }
  }

++NumFused;
  return true;
}

namespace {

/// Post-process the DAG to create cluster edges between instrs that may
/// be fused by the processor into a single operation.
class MacroFusion : public ScheduleDAGMutation {
  ShouldSchedulePredTy shouldScheduleAdjacent;
  bool FuseBlock;
  bool scheduleAdjacentImpl(ScheduleDAGInstrs &DAG, SUnit &AnchorSU);

public:
  MacroFusion(ShouldSchedulePredTy shouldScheduleAdjacent, bool FuseBlock)
    : shouldScheduleAdjacent(shouldScheduleAdjacent), FuseBlock(FuseBlock) {}

void apply(ScheduleDAGInstrs *DAGInstrs) override;
};

} // end anonymous namespace

void MacroFusion::apply(ScheduleDAGInstrs *DAG) {
  if (FuseBlock)
    // For each of the SUnits in the scheduling block, try to fuse the instr in
    // it with one in its predecessors.
    for (SUnit &ISU : DAG->SUnits)
        scheduleAdjacentImpl(*DAG, ISU);

if (DAG->ExitSU.getInstr())
    // Try to fuse the instr in the ExitSU with one in its predecessors.
    scheduleAdjacentImpl(*DAG, DAG->ExitSU);
}

/// Implement the fusion of instr pairs in the scheduling DAG,
/// anchored at the instr in AnchorSU..
bool MacroFusion::scheduleAdjacentImpl(ScheduleDAGInstrs &DAG, SUnit &AnchorSU) {
  const MachineInstr &AnchorMI = *AnchorSU.getInstr();
  const TargetInstrInfo &TII = *DAG.TII;
  const TargetSubtargetInfo &ST = DAG.MF.getSubtarget();

// Check if the anchor instr may be fused.
  if (!shouldScheduleAdjacent(TII, ST, nullptr, AnchorMI))
    return false;

// Explorer for fusion candidates among the dependencies of the anchor instr.
  for (SDep &Dep : AnchorSU.Preds) {
    // Ignore dependencies other than data or strong ordering.
    if (Dep.isWeak() || isHazard(Dep))
      continue;

SUnit &DepSU = *Dep.getSUnit();
    if (DepSU.isBoundaryNode())
      continue;

// Only chain two instructions together at most.
    const MachineInstr *DepMI = DepSU.getInstr();
    if (!hasLessThanNumFused(DepSU, 2) ||
        !shouldScheduleAdjacent(TII, ST, DepMI, AnchorMI))
      continue;

if (fuseInstructionPair(DAG, DepSU, AnchorSU))
      return true;
  }

return false;
}

std::unique_ptr<ScheduleDAGMutation>
llvm::createMacroFusionDAGMutation(
     ShouldSchedulePredTy shouldScheduleAdjacent) {
  if(EnableMacroFusion)
    return std::make_unique<MacroFusion>(shouldScheduleAdjacent, true);
  return nullptr;
}

std::unique_ptr<ScheduleDAGMutation>
llvm::createBranchMacroFusionDAGMutation(
     ShouldSchedulePredTy shouldScheduleAdjacent) {
  if(EnableMacroFusion)
    return std::make_unique<MacroFusion>(shouldScheduleAdjacent, false);
  return nullptr;
}

003 File Manager

Editing: MacroFusion.cpp

Upload File

Create Folder