--- /dev/null
+//===---- MemCpyOptimizer.h - memcpy optimization ---------------*- C++ -*-===//
+//
+// The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This pass performs various transformations related to eliminating memcpy
+// calls, or transforming sets of stores into memset's.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_TRANSFORMS_SCALAR_MEMCPYOPTIMIZER_H
+#define LLVM_TRANSFORMS_SCALAR_MEMCPYOPTIMIZER_H
+
+#include "llvm/ADT/STLExtras.h"
+#include "llvm/Analysis/AliasAnalysis.h"
+#include "llvm/Analysis/AssumptionCache.h"
+#include "llvm/Analysis/GlobalsModRef.h"
+#include "llvm/Analysis/MemoryDependenceAnalysis.h"
+#include "llvm/Analysis/TargetLibraryInfo.h"
+#include "llvm/IR/Dominators.h"
+#include "llvm/IR/Function.h"
+#include "llvm/IR/Instructions.h"
+#include "llvm/IR/IntrinsicInst.h"
+#include "llvm/IR/PassManager.h"
+
+namespace llvm {
+
+class MemCpyOptPass : public PassInfoMixin<MemCpyOptPass> {
+ MemoryDependenceResults *MD = nullptr;
+ TargetLibraryInfo *TLI = nullptr;
+ std::function<AliasAnalysis &()> LookupAliasAnalysis;
+ std::function<AssumptionCache &()> LookupAssumptionCache;
+ std::function<DominatorTree &()> LookupDomTree;
+
+public:
+ MemCpyOptPass() {}
+ PreservedAnalyses run(Function &F, FunctionAnalysisManager &AM);
+ // Glue for the old PM.
+ bool runImpl(Function &F, MemoryDependenceResults *MD_,
+ TargetLibraryInfo *TLI_,
+ std::function<AliasAnalysis &()> LookupAliasAnalysis_,
+ std::function<AssumptionCache &()> LookupAssumptionCache_,
+ std::function<DominatorTree &()> LookupDomTree_);
+
+private:
+ // Helper functions
+ bool processStore(StoreInst *SI, BasicBlock::iterator &BBI);
+ bool processMemSet(MemSetInst *SI, BasicBlock::iterator &BBI);
+ bool processMemCpy(MemCpyInst *M);
+ bool processMemMove(MemMoveInst *M);
+ bool performCallSlotOptzn(Instruction *cpy, Value *cpyDst, Value *cpySrc,
+ uint64_t cpyLen, unsigned cpyAlign, CallInst *C);
+ bool processMemCpyMemCpyDependence(MemCpyInst *M, MemCpyInst *MDep);
+ bool processMemSetMemCpyDependence(MemCpyInst *M, MemSetInst *MDep);
+ bool performMemCpyToMemSetOptzn(MemCpyInst *M, MemSetInst *MDep);
+ bool processByValArgument(CallSite CS, unsigned ArgNo);
+ Instruction *tryMergingIntoMemset(Instruction *I, Value *StartPtr,
+ Value *ByteVal);
+
+ bool iterateOnFunction(Function &F);
+};
+}
+
+#endif // LLVM_TRANSFORMS_SCALAR_MEMCPYOPTIMIZER_H
//
//===----------------------------------------------------------------------===//
+#include "llvm/Transforms/Scalar/MemCpyOptimizer.h"
#include "llvm/Transforms/Scalar.h"
#include "llvm/ADT/DenseSet.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/Statistic.h"
-#include "llvm/Analysis/AliasAnalysis.h"
-#include "llvm/Analysis/AssumptionCache.h"
-#include "llvm/Analysis/GlobalsModRef.h"
-#include "llvm/Analysis/MemoryDependenceAnalysis.h"
-#include "llvm/Analysis/TargetLibraryInfo.h"
#include "llvm/Analysis/ValueTracking.h"
#include "llvm/IR/DataLayout.h"
-#include "llvm/IR/Dominators.h"
#include "llvm/IR/GetElementPtrTypeIterator.h"
#include "llvm/IR/GlobalVariable.h"
#include "llvm/IR/IRBuilder.h"
-#include "llvm/IR/Instructions.h"
-#include "llvm/IR/IntrinsicInst.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Transforms/Utils/Local.h"
}
//===----------------------------------------------------------------------===//
-// MemCpyOpt Pass
+// MemCpyOptLegacyPass Pass
//===----------------------------------------------------------------------===//
namespace {
- class MemCpyOpt : public FunctionPass {
- MemoryDependenceResults *MD;
- TargetLibraryInfo *TLI;
+ class MemCpyOptLegacyPass : public FunctionPass {
+ MemCpyOptPass Impl;
public:
static char ID; // Pass identification, replacement for typeid
- MemCpyOpt() : FunctionPass(ID) {
- initializeMemCpyOptPass(*PassRegistry::getPassRegistry());
- MD = nullptr;
- TLI = nullptr;
+ MemCpyOptLegacyPass() : FunctionPass(ID) {
+ initializeMemCpyOptLegacyPassPass(*PassRegistry::getPassRegistry());
}
bool runOnFunction(Function &F) override;
bool iterateOnFunction(Function &F);
};
- char MemCpyOpt::ID = 0;
+ char MemCpyOptLegacyPass::ID = 0;
}
/// The public interface to this file...
-FunctionPass *llvm::createMemCpyOptPass() { return new MemCpyOpt(); }
+FunctionPass *llvm::createMemCpyOptPass() { return new MemCpyOptLegacyPass(); }
-INITIALIZE_PASS_BEGIN(MemCpyOpt, "memcpyopt", "MemCpy Optimization",
+INITIALIZE_PASS_BEGIN(MemCpyOptLegacyPass, "memcpyopt", "MemCpy Optimization",
false, false)
INITIALIZE_PASS_DEPENDENCY(AssumptionCacheTracker)
INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass)
INITIALIZE_PASS_DEPENDENCY(TargetLibraryInfoWrapperPass)
INITIALIZE_PASS_DEPENDENCY(AAResultsWrapperPass)
INITIALIZE_PASS_DEPENDENCY(GlobalsAAWrapperPass)
-INITIALIZE_PASS_END(MemCpyOpt, "memcpyopt", "MemCpy Optimization",
+INITIALIZE_PASS_END(MemCpyOptLegacyPass, "memcpyopt", "MemCpy Optimization",
false, false)
/// When scanning forward over instructions, we look for some other patterns to
/// fold away. In particular, this looks for stores to neighboring locations of
/// memory. If it sees enough consecutive ones, it attempts to merge them
/// together into a memcpy/memset.
-Instruction *MemCpyOpt::tryMergingIntoMemset(Instruction *StartInst,
- Value *StartPtr, Value *ByteVal) {
+Instruction *MemCpyOptPass::tryMergingIntoMemset(Instruction *StartInst,
+ Value *StartPtr,
+ Value *ByteVal) {
const DataLayout &DL = StartInst->getModule()->getDataLayout();
// Okay, so we now have a single store that can be splatable. Scan to find
return true;
}
-bool MemCpyOpt::processStore(StoreInst *SI, BasicBlock::iterator &BBI) {
+bool MemCpyOptPass::processStore(StoreInst *SI, BasicBlock::iterator &BBI) {
if (!SI->isSimple()) return false;
// Avoid merging nontemporal stores since the resulting
auto *T = LI->getType();
if (T->isAggregateType()) {
- AliasAnalysis &AA = getAnalysis<AAResultsWrapperPass>().getAAResults();
+ AliasAnalysis &AA = LookupAliasAnalysis();
MemoryLocation LoadLoc = MemoryLocation::get(LI);
// We use alias analysis to check if an instruction may store to
// the call and the store.
Value *CpyDest = SI->getPointerOperand()->stripPointerCasts();
bool CpyDestIsLocal = isa<AllocaInst>(CpyDest);
- AliasAnalysis &AA = getAnalysis<AAResultsWrapperPass>().getAAResults();
+ AliasAnalysis &AA = LookupAliasAnalysis();
MemoryLocation StoreLoc = MemoryLocation::get(SI);
for (BasicBlock::iterator I = --SI->getIterator(), E = C->getIterator();
I != E; --I) {
return false;
}
-bool MemCpyOpt::processMemSet(MemSetInst *MSI, BasicBlock::iterator &BBI) {
+bool MemCpyOptPass::processMemSet(MemSetInst *MSI, BasicBlock::iterator &BBI) {
// See if there is another memset or store neighboring this memset which
// allows us to widen out the memset to do a single larger store.
if (isa<ConstantInt>(MSI->getLength()) && !MSI->isVolatile())
/// Takes a memcpy and a call that it depends on,
/// and checks for the possibility of a call slot optimization by having
/// the call write its result directly into the destination of the memcpy.
-bool MemCpyOpt::performCallSlotOptzn(Instruction *cpy,
- Value *cpyDest, Value *cpySrc,
- uint64_t cpyLen, unsigned cpyAlign,
- CallInst *C) {
+bool MemCpyOptPass::performCallSlotOptzn(Instruction *cpy, Value *cpyDest,
+ Value *cpySrc, uint64_t cpyLen,
+ unsigned cpyAlign, CallInst *C) {
// The general transformation to keep in mind is
//
// call @func(..., src, ...)
// Since we're changing the parameter to the callsite, we need to make sure
// that what would be the new parameter dominates the callsite.
- DominatorTree &DT = getAnalysis<DominatorTreeWrapperPass>().getDomTree();
+ DominatorTree &DT = LookupDomTree();
if (Instruction *cpyDestInst = dyn_cast<Instruction>(cpyDest))
if (!DT.dominates(cpyDestInst, C))
return false;
// unexpected manner, for example via a global, which we deduce from
// the use analysis, we also need to know that it does not sneakily
// access dest. We rely on AA to figure this out for us.
- AliasAnalysis &AA = getAnalysis<AAResultsWrapperPass>().getAAResults();
+ AliasAnalysis &AA = LookupAliasAnalysis();
ModRefInfo MR = AA.getModRefInfo(C, cpyDest, srcSize);
// If necessary, perform additional analysis.
if (MR != MRI_NoModRef)
/// We've found that the (upward scanning) memory dependence of memcpy 'M' is
/// the memcpy 'MDep'. Try to simplify M to copy from MDep's input if we can.
-bool MemCpyOpt::processMemCpyMemCpyDependence(MemCpyInst *M, MemCpyInst *MDep) {
+bool MemCpyOptPass::processMemCpyMemCpyDependence(MemCpyInst *M,
+ MemCpyInst *MDep) {
// We can only transforms memcpy's where the dest of one is the source of the
// other.
if (M->getSource() != MDep->getDest() || MDep->isVolatile())
if (!MDepLen || !MLen || MDepLen->getZExtValue() < MLen->getZExtValue())
return false;
- AliasAnalysis &AA = getAnalysis<AAResultsWrapperPass>().getAAResults();
+ AliasAnalysis &AA = LookupAliasAnalysis();
// Verify that the copied-from memory doesn't change in between the two
// transfers. For example, in:
/// memcpy(dst, src, src_size);
/// memset(dst + src_size, c, dst_size <= src_size ? 0 : dst_size - src_size);
/// \endcode
-bool MemCpyOpt::processMemSetMemCpyDependence(MemCpyInst *MemCpy,
- MemSetInst *MemSet) {
+bool MemCpyOptPass::processMemSetMemCpyDependence(MemCpyInst *MemCpy,
+ MemSetInst *MemSet) {
// We can only transform memset/memcpy with the same destination.
if (MemSet->getDest() != MemCpy->getDest())
return false;
/// When dst2_size <= dst1_size.
///
/// The \p MemCpy must have a Constant length.
-bool MemCpyOpt::performMemCpyToMemSetOptzn(MemCpyInst *MemCpy,
- MemSetInst *MemSet) {
+bool MemCpyOptPass::performMemCpyToMemSetOptzn(MemCpyInst *MemCpy,
+ MemSetInst *MemSet) {
// This only makes sense on memcpy(..., memset(...), ...).
if (MemSet->getRawDest() != MemCpy->getRawSource())
return false;
/// B to be a memcpy from X to Z (or potentially a memmove, depending on
/// circumstances). This allows later passes to remove the first memcpy
/// altogether.
-bool MemCpyOpt::processMemCpy(MemCpyInst *M) {
+bool MemCpyOptPass::processMemCpy(MemCpyInst *M) {
// We can only optimize non-volatile memcpy's.
if (M->isVolatile()) return false;
/// Transforms memmove calls to memcpy calls when the src/dst are guaranteed
/// not to alias.
-bool MemCpyOpt::processMemMove(MemMoveInst *M) {
- AliasAnalysis &AA = getAnalysis<AAResultsWrapperPass>().getAAResults();
+bool MemCpyOptPass::processMemMove(MemMoveInst *M) {
+ AliasAnalysis &AA = LookupAliasAnalysis();
if (!TLI->has(LibFunc::memmove))
return false;
MemoryLocation::getForSource(M)))
return false;
- DEBUG(dbgs() << "MemCpyOpt: Optimizing memmove -> memcpy: " << *M << "\n");
+ DEBUG(dbgs() << "MemCpyOptPass: Optimizing memmove -> memcpy: " << *M
+ << "\n");
// If not, then we know we can transform this.
Type *ArgTys[3] = { M->getRawDest()->getType(),
}
/// This is called on every byval argument in call sites.
-bool MemCpyOpt::processByValArgument(CallSite CS, unsigned ArgNo) {
+bool MemCpyOptPass::processByValArgument(CallSite CS, unsigned ArgNo) {
const DataLayout &DL = CS.getCaller()->getParent()->getDataLayout();
// Find out what feeds this byval argument.
Value *ByValArg = CS.getArgument(ArgNo);
// If it is greater than the memcpy, then we check to see if we can force the
// source of the memcpy to the alignment we need. If we fail, we bail out.
- AssumptionCache &AC =
- getAnalysis<AssumptionCacheTracker>().getAssumptionCache(
- *CS->getParent()->getParent());
- DominatorTree &DT = getAnalysis<DominatorTreeWrapperPass>().getDomTree();
+ AssumptionCache &AC = LookupAssumptionCache();
+ DominatorTree &DT = LookupDomTree();
if (MDep->getAlignment() < ByValAlign &&
getOrEnforceKnownAlignment(MDep->getSource(), ByValAlign, DL,
CS.getInstruction(), &AC, &DT) < ByValAlign)
TmpCast = new BitCastInst(MDep->getSource(), ByValArg->getType(),
"tmpcast", CS.getInstruction());
- DEBUG(dbgs() << "MemCpyOpt: Forwarding memcpy to byval:\n"
+ DEBUG(dbgs() << "MemCpyOptPass: Forwarding memcpy to byval:\n"
<< " " << *MDep << "\n"
<< " " << *CS.getInstruction() << "\n");
return true;
}
-/// Executes one iteration of MemCpyOpt.
-bool MemCpyOpt::iterateOnFunction(Function &F) {
+/// Executes one iteration of MemCpyOptPass.
+bool MemCpyOptPass::iterateOnFunction(Function &F) {
bool MadeChange = false;
// Walk all instruction in the function.
return MadeChange;
}
-/// This is the main transformation entry point for a function.
-bool MemCpyOpt::runOnFunction(Function &F) {
- if (skipFunction(F))
- return false;
+PreservedAnalyses MemCpyOptPass::run(Function &F, FunctionAnalysisManager &AM) {
+ auto &MD = AM.getResult<MemoryDependenceAnalysis>(F);
+ auto &TLI = AM.getResult<TargetLibraryAnalysis>(F);
+
+ auto LookupAliasAnalysis = [&]() -> AliasAnalysis & {
+ return AM.getResult<AAManager>(F);
+ };
+ auto LookupAssumptionCache = [&]() -> AssumptionCache & {
+ return AM.getResult<AssumptionAnalysis>(F);
+ };
+ auto LookupDomTree = [&]() -> DominatorTree & {
+ return AM.getResult<DominatorTreeAnalysis>(F);
+ };
+
+ bool MadeChange = runImpl(F, &MD, &TLI, LookupAliasAnalysis,
+ LookupAssumptionCache, LookupDomTree);
+ if (!MadeChange)
+ return PreservedAnalyses::all();
+ PreservedAnalyses PA;
+ PA.preserve<GlobalsAA>();
+ PA.preserve<MemoryDependenceAnalysis>();
+ return PA;
+}
+
+bool MemCpyOptPass::runImpl(
+ Function &F, MemoryDependenceResults *MD_, TargetLibraryInfo *TLI_,
+ std::function<AliasAnalysis &()> LookupAliasAnalysis_,
+ std::function<AssumptionCache &()> LookupAssumptionCache_,
+ std::function<DominatorTree &()> LookupDomTree_) {
bool MadeChange = false;
- MD = &getAnalysis<MemoryDependenceWrapperPass>().getMemDep();
- TLI = &getAnalysis<TargetLibraryInfoWrapperPass>().getTLI();
+ MD = MD_;
+ TLI = TLI_;
+ LookupAliasAnalysis = LookupAliasAnalysis_;
+ LookupAssumptionCache = LookupAssumptionCache_;
+ LookupDomTree = LookupDomTree_;
// If we don't have at least memset and memcpy, there is little point of doing
// anything here. These are required by a freestanding implementation, so if
MD = nullptr;
return MadeChange;
}
+
+/// This is the main transformation entry point for a function.
+bool MemCpyOptLegacyPass::runOnFunction(Function &F) {
+ if (skipFunction(F))
+ return false;
+
+ auto *MD = &getAnalysis<MemoryDependenceWrapperPass>().getMemDep();
+ auto *TLI = &getAnalysis<TargetLibraryInfoWrapperPass>().getTLI();
+
+ auto LookupAliasAnalysis = [this]() -> AliasAnalysis & {
+ return getAnalysis<AAResultsWrapperPass>().getAAResults();
+ };
+ auto LookupAssumptionCache = [this, &F]() -> AssumptionCache & {
+ return getAnalysis<AssumptionCacheTracker>().getAssumptionCache(F);
+ };
+ auto LookupDomTree = [this]() -> DominatorTree & {
+ return getAnalysis<DominatorTreeWrapperPass>().getDomTree();
+ };
+
+ return Impl.runImpl(F, MD, TLI, LookupAliasAnalysis, LookupAssumptionCache,
+ LookupDomTree);
+}
projects / llvm / commitdiff