#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/AliasAnalysis.h"
-#include "llvm/Analysis/AssumptionTracker.h"
+#include "llvm/Analysis/AssumptionCache.h"
#include "llvm/Analysis/InstructionSimplify.h"
#include "llvm/Analysis/MemoryBuiltins.h"
#include "llvm/Analysis/PHITransAddr.h"
// Register this pass...
INITIALIZE_PASS_BEGIN(MemoryDependenceAnalysis, "memdep",
"Memory Dependence Analysis", false, true)
-INITIALIZE_PASS_DEPENDENCY(AssumptionTracker)
+INITIALIZE_PASS_DEPENDENCY(AssumptionCacheTracker)
INITIALIZE_AG_DEPENDENCY(AliasAnalysis)
INITIALIZE_PASS_END(MemoryDependenceAnalysis, "memdep",
"Memory Dependence Analysis", false, true)
MemoryDependenceAnalysis::MemoryDependenceAnalysis()
- : FunctionPass(ID), PredCache() {
+ : FunctionPass(ID) {
initializeMemoryDependenceAnalysisPass(*PassRegistry::getPassRegistry());
}
MemoryDependenceAnalysis::~MemoryDependenceAnalysis() {
ReverseLocalDeps.clear();
ReverseNonLocalDeps.clear();
ReverseNonLocalPtrDeps.clear();
- PredCache->clear();
+ PredCache.clear();
}
-
-
/// getAnalysisUsage - Does not modify anything. It uses Alias Analysis.
///
void MemoryDependenceAnalysis::getAnalysisUsage(AnalysisUsage &AU) const {
AU.setPreservesAll();
- AU.addRequired<AssumptionTracker>();
+ AU.addRequired<AssumptionCacheTracker>();
AU.addRequiredTransitive<AliasAnalysis>();
}
-bool MemoryDependenceAnalysis::runOnFunction(Function &) {
+bool MemoryDependenceAnalysis::runOnFunction(Function &F) {
AA = &getAnalysis<AliasAnalysis>();
- AT = &getAnalysis<AssumptionTracker>();
- DataLayoutPass *DLP = getAnalysisIfAvailable<DataLayoutPass>();
- DL = DLP ? &DLP->getDataLayout() : nullptr;
+ AC = &getAnalysis<AssumptionCacheTracker>().getAssumptionCache(F);
DominatorTreeWrapperPass *DTWP =
getAnalysisIfAvailable<DominatorTreeWrapperPass>();
DT = DTWP ? &DTWP->getDomTree() : nullptr;
- if (!PredCache)
- PredCache.reset(new PredIteratorCache());
return false;
}
continue;
}
- if (CallSite InstCS = cast<Value>(Inst)) {
+ if (auto InstCS = CallSite(Inst)) {
// Debug intrinsics don't cause dependences.
if (isa<DbgInfoIntrinsic>(Inst)) continue;
// If these two calls do not interfere, look past it.
///
/// MemLocBase, MemLocOffset are lazily computed here the first time the
/// base/offs of memloc is needed.
-static bool
-isLoadLoadClobberIfExtendedToFullWidth(const AliasAnalysis::Location &MemLoc,
- const Value *&MemLocBase,
- int64_t &MemLocOffs,
- const LoadInst *LI,
- const DataLayout *DL) {
- // If we have no target data, we can't do this.
- if (!DL) return false;
+static bool isLoadLoadClobberIfExtendedToFullWidth(
+ const AliasAnalysis::Location &MemLoc, const Value *&MemLocBase,
+ int64_t &MemLocOffs, const LoadInst *LI) {
+ const DataLayout &DL = LI->getModule()->getDataLayout();
// If we haven't already computed the base/offset of MemLoc, do so now.
if (!MemLocBase)
MemLocBase = GetPointerBaseWithConstantOffset(MemLoc.Ptr, MemLocOffs, DL);
- unsigned Size = MemoryDependenceAnalysis::
- getLoadLoadClobberFullWidthSize(MemLocBase, MemLocOffs, MemLoc.Size,
- LI, *DL);
+ unsigned Size = MemoryDependenceAnalysis::getLoadLoadClobberFullWidthSize(
+ MemLocBase, MemLocOffs, MemLoc.Size, LI);
return Size != 0;
}
/// 2) safe for the target, and 3) would provide the specified memory
/// location value, then this function returns the size in bytes of the
/// load width to use. If not, this returns zero.
-unsigned MemoryDependenceAnalysis::
-getLoadLoadClobberFullWidthSize(const Value *MemLocBase, int64_t MemLocOffs,
- unsigned MemLocSize, const LoadInst *LI,
- const DataLayout &DL) {
+unsigned MemoryDependenceAnalysis::getLoadLoadClobberFullWidthSize(
+ const Value *MemLocBase, int64_t MemLocOffs, unsigned MemLocSize,
+ const LoadInst *LI) {
// We can only extend simple integer loads.
if (!isa<IntegerType>(LI->getType()) || !LI->isSimple()) return 0;
// Load widening is hostile to ThreadSanitizer: it may cause false positives
// or make the reports more cryptic (access sizes are wrong).
- if (LI->getParent()->getParent()->getAttributes().
- hasAttribute(AttributeSet::FunctionIndex, Attribute::SanitizeThread))
+ if (LI->getParent()->getParent()->hasFnAttribute(Attribute::SanitizeThread))
return 0;
+ const DataLayout &DL = LI->getModule()->getDataLayout();
+
// Get the base of this load.
int64_t LIOffs = 0;
const Value *LIBase =
- GetPointerBaseWithConstantOffset(LI->getPointerOperand(), LIOffs, &DL);
+ GetPointerBaseWithConstantOffset(LI->getPointerOperand(), LIOffs, DL);
// If the two pointers are not based on the same pointer, we can't tell that
// they are related.
!DL.fitsInLegalInteger(NewLoadByteSize*8))
return 0;
- if (LIOffs+NewLoadByteSize > MemLocEnd &&
- LI->getParent()->getParent()->getAttributes().
- hasAttribute(AttributeSet::FunctionIndex, Attribute::SanitizeAddress))
+ if (LIOffs + NewLoadByteSize > MemLocEnd &&
+ LI->getParent()->getParent()->hasFnAttribute(
+ Attribute::SanitizeAddress))
// We will be reading past the location accessed by the original program.
// While this is safe in a regular build, Address Safety analysis tools
// may start reporting false warnings. So, don't do widening.
}
}
+static bool isVolatile(Instruction *Inst) {
+ if (LoadInst *LI = dyn_cast<LoadInst>(Inst))
+ return LI->isVolatile();
+ else if (StoreInst *SI = dyn_cast<StoreInst>(Inst))
+ return SI->isVolatile();
+ else if (AtomicCmpXchgInst *AI = dyn_cast<AtomicCmpXchgInst>(Inst))
+ return AI->isVolatile();
+ return false;
+}
+
+
/// getPointerDependencyFrom - Return the instruction on which a memory
/// location depends. If isLoad is true, this routine ignores may-aliases with
/// read-only operations. If isLoad is false, this routine ignores may-aliases
// by every program that can detect any optimisation of that kind: either
// it is racy (undefined) or there is a release followed by an acquire
// between the pair of accesses under consideration.
- bool HasSeenAcquire = false;
+ // If the load is invariant, we "know" that it doesn't alias *any* write. We
+ // do want to respect mustalias results since defs are useful for value
+ // forwarding, but any mayalias write can be assumed to be noalias.
+ // Arguably, this logic should be pushed inside AliasAnalysis itself.
if (isLoad && QueryInst) {
LoadInst *LI = dyn_cast<LoadInst>(QueryInst);
if (LI && LI->getMetadata(LLVMContext::MD_invariant_load) != nullptr)
isInvariantLoad = true;
}
+ const DataLayout &DL = BB->getModule()->getDataLayout();
+
// Walk backwards through the basic block, looking for dependencies.
while (ScanIt != BB->begin()) {
Instruction *Inst = --ScanIt;
// does not alias with when this atomic load indicates that another thread may
// be accessing the location.
if (LoadInst *LI = dyn_cast<LoadInst>(Inst)) {
+
+ // While volatile access cannot be eliminated, they do not have to clobber
+ // non-aliasing locations, as normal accesses, for example, can be safely
+ // reordered with volatile accesses.
+ if (LI->isVolatile()) {
+ if (!QueryInst)
+ // Original QueryInst *may* be volatile
+ return MemDepResult::getClobber(LI);
+ if (isVolatile(QueryInst))
+ // Ordering required if QueryInst is itself volatile
+ return MemDepResult::getClobber(LI);
+ // Otherwise, volatile doesn't imply any special ordering
+ }
+
// Atomic loads have complications involved.
// A Monotonic (or higher) load is OK if the query inst is itself not atomic.
- // An Acquire (or higher) load sets the HasSeenAcquire flag, so that any
- // release store will know to return getClobber.
// FIXME: This is overly conservative.
- if (!LI->isUnordered()) {
+ if (LI->isAtomic() && LI->getOrdering() > Unordered) {
if (!QueryInst)
return MemDepResult::getClobber(LI);
+ if (LI->getOrdering() != Monotonic)
+ return MemDepResult::getClobber(LI);
if (auto *QueryLI = dyn_cast<LoadInst>(QueryInst)) {
if (!QueryLI->isSimple())
return MemDepResult::getClobber(LI);
} else if (QueryInst->mayReadOrWriteMemory()) {
return MemDepResult::getClobber(LI);
}
-
- if (isAtLeastAcquire(LI->getOrdering()))
- HasSeenAcquire = true;
}
- // FIXME: this is overly conservative.
- // While volatile access cannot be eliminated, they do not have to clobber
- // non-aliasing locations, as normal accesses can for example be reordered
- // with volatile accesses.
- if (LI->isVolatile())
- return MemDepResult::getClobber(LI);
-
AliasAnalysis::Location LoadLoc = AA->getLocation(LI);
// If we found a pointer, check if it could be the same as our pointer.
// location is 1 byte at P+1). If so, return it as a load/load
// clobber result, allowing the client to decide to widen the load if
// it wants to.
- if (IntegerType *ITy = dyn_cast<IntegerType>(LI->getType()))
- if (LI->getAlignment()*8 > ITy->getPrimitiveSizeInBits() &&
+ if (IntegerType *ITy = dyn_cast<IntegerType>(LI->getType())) {
+ if (LI->getAlignment() * 8 > ITy->getPrimitiveSizeInBits() &&
isLoadLoadClobberIfExtendedToFullWidth(MemLoc, MemLocBase,
- MemLocOffset, LI, DL))
+ MemLocOffset, LI))
return MemDepResult::getClobber(Inst);
-
+ }
continue;
}
if (StoreInst *SI = dyn_cast<StoreInst>(Inst)) {
// Atomic stores have complications involved.
// A Monotonic store is OK if the query inst is itself not atomic.
- // A Release (or higher) store further requires that no acquire load
- // has been seen.
// FIXME: This is overly conservative.
if (!SI->isUnordered()) {
if (!QueryInst)
return MemDepResult::getClobber(SI);
+ if (SI->getOrdering() != Monotonic)
+ return MemDepResult::getClobber(SI);
if (auto *QueryLI = dyn_cast<LoadInst>(QueryInst)) {
if (!QueryLI->isSimple())
return MemDepResult::getClobber(SI);
} else if (QueryInst->mayReadOrWriteMemory()) {
return MemDepResult::getClobber(SI);
}
-
- if (HasSeenAcquire && isAtLeastRelease(SI->getOrdering()))
- return MemDepResult::getClobber(SI);
}
// FIXME: this is overly conservative.
if (AccessPtr == Inst || AA->isMustAlias(Inst, AccessPtr))
return MemDepResult::getDef(Inst);
+ if (isInvariantLoad)
+ continue;
// Be conservative if the accessed pointer may alias the allocation.
if (AA->alias(Inst, AccessPtr) != AliasAnalysis::NoAlias)
return MemDepResult::getClobber(Inst);
continue;
}
+ if (isInvariantLoad)
+ continue;
+
// See if this instruction (e.g. a call or vaarg) mod/ref's the pointer.
AliasAnalysis::ModRefResult MR = AA->getModRefInfo(Inst, MemLoc);
// If necessary, perform additional analysis.
} else {
// Seed DirtyBlocks with each of the preds of QueryInst's block.
BasicBlock *QueryBB = QueryCS.getInstruction()->getParent();
- for (BasicBlock **PI = PredCache->GetPreds(QueryBB); *PI; ++PI)
- DirtyBlocks.push_back(*PI);
+ for (BasicBlock *Pred : PredCache.get(QueryBB))
+ DirtyBlocks.push_back(Pred);
++NumUncacheNonLocal;
}
// If the block *is* completely transparent to the load, we need to check
// the predecessors of this block. Add them to our worklist.
- for (BasicBlock **PI = PredCache->GetPreds(DirtyBB); *PI; ++PI)
- DirtyBlocks.push_back(*PI);
+ for (BasicBlock *Pred : PredCache.get(DirtyBB))
+ DirtyBlocks.push_back(Pred);
}
}
/// own block.
///
void MemoryDependenceAnalysis::
-getNonLocalPointerDependency(const AliasAnalysis::Location &Loc, bool isLoad,
- BasicBlock *FromBB,
+getNonLocalPointerDependency(Instruction *QueryInst,
SmallVectorImpl<NonLocalDepResult> &Result) {
+ const AliasAnalysis::Location Loc = AA->getLocation(QueryInst);
+ bool isLoad = isa<LoadInst>(QueryInst);
+ BasicBlock *FromBB = QueryInst->getParent();
+ assert(FromBB);
+
assert(Loc.Ptr->getType()->isPointerTy() &&
"Can't get pointer deps of a non-pointer!");
Result.clear();
-
- PHITransAddr Address(const_cast<Value *>(Loc.Ptr), DL, AT);
+
+ // This routine does not expect to deal with volatile instructions.
+ // Doing so would require piping through the QueryInst all the way through.
+ // TODO: volatiles can't be elided, but they can be reordered with other
+ // non-volatile accesses.
+
+ // We currently give up on any instruction which is ordered, but we do handle
+ // atomic instructions which are unordered.
+ // TODO: Handle ordered instructions
+ auto isOrdered = [](Instruction *Inst) {
+ if (LoadInst *LI = dyn_cast<LoadInst>(Inst)) {
+ return !LI->isUnordered();
+ } else if (StoreInst *SI = dyn_cast<StoreInst>(Inst)) {
+ return !SI->isUnordered();
+ }
+ return false;
+ };
+ if (isVolatile(QueryInst) || isOrdered(QueryInst)) {
+ Result.push_back(NonLocalDepResult(FromBB,
+ MemDepResult::getUnknown(),
+ const_cast<Value *>(Loc.Ptr)));
+ return;
+ }
+ const DataLayout &DL = FromBB->getModule()->getDataLayout();
+ PHITransAddr Address(const_cast<Value *>(Loc.Ptr), DL, AC);
// This is the set of blocks we've inspected, and the pointer we consider in
// each block. Because of critical edges, we currently bail out if querying
// a block with multiple different pointers. This can happen during PHI
// translation.
DenseMap<BasicBlock*, Value*> Visited;
- if (!getNonLocalPointerDepFromBB(Address, Loc, isLoad, FromBB,
+ if (!getNonLocalPointerDepFromBB(QueryInst, Address, Loc, isLoad, FromBB,
Result, Visited, true))
return;
Result.clear();
/// lookup (which may use dirty cache info if available). If we do a lookup,
/// add the result to the cache.
MemDepResult MemoryDependenceAnalysis::
-GetNonLocalInfoForBlock(const AliasAnalysis::Location &Loc,
+GetNonLocalInfoForBlock(Instruction *QueryInst,
+ const AliasAnalysis::Location &Loc,
bool isLoad, BasicBlock *BB,
NonLocalDepInfo *Cache, unsigned NumSortedEntries) {
}
// Scan the block for the dependency.
- MemDepResult Dep = getPointerDependencyFrom(Loc, isLoad, ScanPos, BB);
+ MemDepResult Dep = getPointerDependencyFrom(Loc, isLoad, ScanPos, BB,
+ QueryInst);
// If we had a dirty entry for the block, update it. Otherwise, just add
// a new entry.
/// not compute dependence information for some reason. This should be treated
/// as a clobber dependence on the first instruction in the predecessor block.
bool MemoryDependenceAnalysis::
-getNonLocalPointerDepFromBB(const PHITransAddr &Pointer,
+getNonLocalPointerDepFromBB(Instruction *QueryInst,
+ const PHITransAddr &Pointer,
const AliasAnalysis::Location &Loc,
bool isLoad, BasicBlock *StartBB,
SmallVectorImpl<NonLocalDepResult> &Result,
} else if (CacheInfo->Size > Loc.Size) {
// This query's Size is less than the cached one. Conservatively restart
// the query using the greater size.
- return getNonLocalPointerDepFromBB(Pointer,
+ return getNonLocalPointerDepFromBB(QueryInst, Pointer,
Loc.getWithNewSize(CacheInfo->Size),
isLoad, StartBB, Result, Visited,
SkipFirstBlock);
CacheInfo->NonLocalDeps.clear();
}
if (Loc.AATags)
- return getNonLocalPointerDepFromBB(Pointer, Loc.getWithoutAATags(),
+ return getNonLocalPointerDepFromBB(QueryInst,
+ Pointer, Loc.getWithoutAATags(),
isLoad, StartBB, Result, Visited,
SkipFirstBlock);
}
// cache value will only see properly sorted cache arrays.
if (Cache && NumSortedEntries != Cache->size()) {
SortNonLocalDepInfoCache(*Cache, NumSortedEntries);
- NumSortedEntries = Cache->size();
}
// Since we bail out, the "Cache" set won't contain all of the
// results for the query. This is ok (we can still use it to accelerate
// Get the dependency info for Pointer in BB. If we have cached
// information, we will use it, otherwise we compute it.
DEBUG(AssertSorted(*Cache, NumSortedEntries));
- MemDepResult Dep = GetNonLocalInfoForBlock(Loc, isLoad, BB, Cache,
+ MemDepResult Dep = GetNonLocalInfoForBlock(QueryInst,
+ Loc, isLoad, BB, Cache,
NumSortedEntries);
// If we got a Def or Clobber, add this to the list of results.
if (!Pointer.NeedsPHITranslationFromBlock(BB)) {
SkipFirstBlock = false;
SmallVector<BasicBlock*, 16> NewBlocks;
- for (BasicBlock **PI = PredCache->GetPreds(BB); *PI; ++PI) {
+ for (BasicBlock *Pred : PredCache.get(BB)) {
// Verify that we haven't looked at this block yet.
std::pair<DenseMap<BasicBlock*,Value*>::iterator, bool>
- InsertRes = Visited.insert(std::make_pair(*PI, Pointer.getAddr()));
+ InsertRes = Visited.insert(std::make_pair(Pred, Pointer.getAddr()));
if (InsertRes.second) {
// First time we've looked at *PI.
- NewBlocks.push_back(*PI);
+ NewBlocks.push_back(Pred);
continue;
}
Cache = nullptr;
PredList.clear();
- for (BasicBlock **PI = PredCache->GetPreds(BB); *PI; ++PI) {
- BasicBlock *Pred = *PI;
+ for (BasicBlock *Pred : PredCache.get(BB)) {
PredList.push_back(std::make_pair(Pred, Pointer));
// Get the PHI translated pointer in this predecessor. This can fail if
// not translatable, in which case the getAddr() returns null.
PHITransAddr &PredPointer = PredList.back().second;
- PredPointer.PHITranslateValue(BB, Pred, nullptr);
-
+ PredPointer.PHITranslateValue(BB, Pred, DT, /*MustDominate=*/false);
Value *PredPtrVal = PredPointer.getAddr();
// Check to see if we have already visited this pred block with another
// result conflicted with the Visited list; we have to conservatively
// assume it is unknown, but this also does not block PRE of the load.
if (!CanTranslate ||
- getNonLocalPointerDepFromBB(PredPointer,
+ getNonLocalPointerDepFromBB(QueryInst, PredPointer,
Loc.getWithNewPtr(PredPtrVal),
isLoad, Pred,
Result, Visited)) {
if (I->getBB() != BB)
continue;
- assert(I->getResult().isNonLocal() &&
+ assert((I->getResult().isNonLocal() || !DT->isReachableFromEntry(BB)) &&
"Should only be here with transparent block");
I->setResult(MemDepResult::getUnknown());
Result.push_back(NonLocalDepResult(I->getBB(), I->getResult(),
/// This needs to be done when the CFG changes, e.g., due to splitting
/// critical edges.
void MemoryDependenceAnalysis::invalidateCachedPredecessors() {
- PredCache->clear();
+ PredCache.clear();
}
/// removeInstruction - Remove an instruction from the dependence analysis,
projects / oota-llvm.git / blobdiff