File Manager

003 File Manager

Current Path: /usr/src/contrib/llvm-project/llvm/lib/Target/NVPTX

usr / src / contrib / llvm-project / llvm / lib / Target / NVPTX /

📁 ..
📁 MCTargetDesc
📄 ManagedStringPool.h(1.41 KB)
📄 NVPTX.h(3.22 KB)
📄 NVPTX.td(4.99 KB)
📄 NVPTXAllocaHoisting.cpp(2.12 KB)
📄 NVPTXAllocaHoisting.h(755 B)
📄 NVPTXAsmPrinter.cpp(71.02 KB)
📄 NVPTXAsmPrinter.h(11.23 KB)
📄 NVPTXAssignValidGlobalNames.cpp(2.74 KB)
📄 NVPTXFrameLowering.cpp(3.41 KB)
📄 NVPTXFrameLowering.h(1.34 KB)
📄 NVPTXGenericToNVVM.cpp(11.59 KB)
📄 NVPTXISelDAGToDAG.cpp(134.03 KB)
📄 NVPTXISelDAGToDAG.h(3.54 KB)
📄 NVPTXISelLowering.cpp(198.59 KB)
📄 NVPTXISelLowering.h(15.65 KB)
📄 NVPTXImageOptimizer.cpp(5.64 KB)
📄 NVPTXInstrFormats.td(1.72 KB)
📄 NVPTXInstrInfo.cpp(7.88 KB)
📄 NVPTXInstrInfo.h(2.87 KB)
📄 NVPTXInstrInfo.td(133.35 KB)
📄 NVPTXIntrinsics.td(329.76 KB)
📄 NVPTXLowerAggrCopies.cpp(4.85 KB)
📄 NVPTXLowerAggrCopies.h(744 B)
📄 NVPTXLowerAlloca.cpp(4.27 KB)
📄 NVPTXLowerArgs.cpp(9.16 KB)
📄 NVPTXMCExpr.cpp(2.04 KB)
📄 NVPTXMCExpr.h(3.92 KB)
📄 NVPTXMachineFunctionInfo.h(1.71 KB)
📄 NVPTXPeephole.cpp(4.92 KB)
📄 NVPTXPrologEpilogPass.cpp(8.91 KB)
📄 NVPTXProxyRegErasure.cpp(3.81 KB)
📄 NVPTXRegisterInfo.cpp(4.41 KB)
📄 NVPTXRegisterInfo.h(2.04 KB)
📄 NVPTXRegisterInfo.td(3.12 KB)
📄 NVPTXReplaceImageHandles.cpp(6.27 KB)
📄 NVPTXSubtarget.cpp(2.15 KB)
📄 NVPTXSubtarget.h(3.02 KB)
📄 NVPTXTargetMachine.cpp(14.25 KB)
📄 NVPTXTargetMachine.h(3.33 KB)
📄 NVPTXTargetObjectFile.h(1.53 KB)
📄 NVPTXTargetTransformInfo.cpp(6.18 KB)
📄 NVPTXTargetTransformInfo.h(4.7 KB)
📄 NVPTXUtilities.cpp(9.37 KB)
📄 NVPTXUtilities.h(2.03 KB)
📄 NVVMIntrRange.cpp(4.88 KB)
📄 NVVMReflect.cpp(6.8 KB)
📁 TargetInfo
📄 cl_common_defines.h(3.94 KB)

Editing: NVPTXLowerAggrCopies.cpp

//===- NVPTXLowerAggrCopies.cpp - ------------------------------*- C++ -*--===//
//
// 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
// Lower aggregate copies, memset, memcpy, memmov intrinsics into loops when
// the size is large or is not a compile-time constant.
//
//===----------------------------------------------------------------------===//

#include "NVPTXLowerAggrCopies.h"
#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/CodeGen/StackProtector.h"
#include "llvm/IR/Constants.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/IRBuilder.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/Intrinsics.h"
#include "llvm/IR/LLVMContext.h"
#include "llvm/IR/Module.h"
#include "llvm/Support/Debug.h"
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
#include "llvm/Transforms/Utils/LowerMemIntrinsics.h"

#define DEBUG_TYPE "nvptx"

using namespace llvm;

namespace {

// actual analysis class, which is a functionpass
struct NVPTXLowerAggrCopies : public FunctionPass {
  static char ID;

NVPTXLowerAggrCopies() : FunctionPass(ID) {}

void getAnalysisUsage(AnalysisUsage &AU) const override {
    AU.addPreserved<StackProtector>();
    AU.addRequired<TargetTransformInfoWrapperPass>();
  }

bool runOnFunction(Function &F) override;

static const unsigned MaxAggrCopySize = 128;

StringRef getPassName() const override {
    return "Lower aggregate copies/intrinsics into loops";
  }
};

char NVPTXLowerAggrCopies::ID = 0;

bool NVPTXLowerAggrCopies::runOnFunction(Function &F) {
  SmallVector<LoadInst *, 4> AggrLoads;
  SmallVector<MemIntrinsic *, 4> MemCalls;

const DataLayout &DL = F.getParent()->getDataLayout();
  LLVMContext &Context = F.getParent()->getContext();
  const TargetTransformInfo &TTI =
      getAnalysis<TargetTransformInfoWrapperPass>().getTTI(F);

// Collect all aggregate loads and mem* calls.
  for (Function::iterator BI = F.begin(), BE = F.end(); BI != BE; ++BI) {
    for (BasicBlock::iterator II = BI->begin(), IE = BI->end(); II != IE;
         ++II) {
      if (LoadInst *LI = dyn_cast<LoadInst>(II)) {
        if (!LI->hasOneUse())
          continue;

if (DL.getTypeStoreSize(LI->getType()) < MaxAggrCopySize)
          continue;

if (StoreInst *SI = dyn_cast<StoreInst>(LI->user_back())) {
          if (SI->getOperand(0) != LI)
            continue;
          AggrLoads.push_back(LI);
        }
      } else if (MemIntrinsic *IntrCall = dyn_cast<MemIntrinsic>(II)) {
        // Convert intrinsic calls with variable size or with constant size
        // larger than the MaxAggrCopySize threshold.
        if (ConstantInt *LenCI = dyn_cast<ConstantInt>(IntrCall->getLength())) {
          if (LenCI->getZExtValue() >= MaxAggrCopySize) {
            MemCalls.push_back(IntrCall);
          }
        } else {
          MemCalls.push_back(IntrCall);
        }
      }
    }
  }

if (AggrLoads.size() == 0 && MemCalls.size() == 0) {
    return false;
  }

//
  // Do the transformation of an aggr load/copy/set to a loop
  //
  for (LoadInst *LI : AggrLoads) {
    auto *SI = cast<StoreInst>(*LI->user_begin());
    Value *SrcAddr = LI->getOperand(0);
    Value *DstAddr = SI->getOperand(1);
    unsigned NumLoads = DL.getTypeStoreSize(LI->getType());
    ConstantInt *CopyLen =
        ConstantInt::get(Type::getInt32Ty(Context), NumLoads);

createMemCpyLoopKnownSize(/* ConvertedInst */ SI,
                              /* SrcAddr */ SrcAddr, /* DstAddr */ DstAddr,
                              /* CopyLen */ CopyLen,
                              /* SrcAlign */ LI->getAlign(),
                              /* DestAlign */ SI->getAlign(),
                              /* SrcIsVolatile */ LI->isVolatile(),
                              /* DstIsVolatile */ SI->isVolatile(), TTI);

SI->eraseFromParent();
    LI->eraseFromParent();
  }

// Transform mem* intrinsic calls.
  for (MemIntrinsic *MemCall : MemCalls) {
    if (MemCpyInst *Memcpy = dyn_cast<MemCpyInst>(MemCall)) {
      expandMemCpyAsLoop(Memcpy, TTI);
    } else if (MemMoveInst *Memmove = dyn_cast<MemMoveInst>(MemCall)) {
      expandMemMoveAsLoop(Memmove);
    } else if (MemSetInst *Memset = dyn_cast<MemSetInst>(MemCall)) {
      expandMemSetAsLoop(Memset);
    }
    MemCall->eraseFromParent();
  }

return true;
}

} // namespace

namespace llvm {
void initializeNVPTXLowerAggrCopiesPass(PassRegistry &);
}

INITIALIZE_PASS(NVPTXLowerAggrCopies, "nvptx-lower-aggr-copies",
                "Lower aggregate copies, and llvm.mem* intrinsics into loops",
                false, false)

FunctionPass *llvm::createLowerAggrCopies() {
  return new NVPTXLowerAggrCopies();
}

003 File Manager

Editing: NVPTXLowerAggrCopies.cpp

Upload File

Create Folder