-//===- MemoryDependenceAnalysis.cpp - Mem Deps Implementation --*- C++ -*-===//
+//===- MemoryDependenceAnalysis.cpp - Mem Deps Implementation -------------===//
//
// The LLVM Compiler Infrastructure
//
//===----------------------------------------------------------------------===//
//
// This file implements an analysis that determines, for a given memory
-// operation, what preceding memory operations it depends on. It builds on
+// operation, what preceding memory operations it depends on. It builds on
// alias analysis information, and tries to provide a lazy, caching interface to
// a common kind of alias information query.
//
//===----------------------------------------------------------------------===//
-#define DEBUG_TYPE "memdep"
#include "llvm/Analysis/MemoryDependenceAnalysis.h"
-#include "llvm/Instructions.h"
-#include "llvm/IntrinsicInst.h"
-#include "llvm/Function.h"
-#include "llvm/LLVMContext.h"
+#include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/AliasAnalysis.h"
-#include "llvm/Analysis/Dominators.h"
#include "llvm/Analysis/InstructionSimplify.h"
#include "llvm/Analysis/MemoryBuiltins.h"
#include "llvm/Analysis/PHITransAddr.h"
#include "llvm/Analysis/ValueTracking.h"
-#include "llvm/ADT/Statistic.h"
-#include "llvm/ADT/STLExtras.h"
-#include "llvm/Support/PredIteratorCache.h"
+#include "llvm/IR/DataLayout.h"
+#include "llvm/IR/Dominators.h"
+#include "llvm/IR/Function.h"
+#include "llvm/IR/Instructions.h"
+#include "llvm/IR/IntrinsicInst.h"
+#include "llvm/IR/LLVMContext.h"
+#include "llvm/IR/PredIteratorCache.h"
#include "llvm/Support/Debug.h"
-#include "llvm/Target/TargetData.h"
using namespace llvm;
+#define DEBUG_TYPE "memdep"
+
STATISTIC(NumCacheNonLocal, "Number of fully cached non-local responses");
STATISTIC(NumCacheDirtyNonLocal, "Number of dirty cached non-local responses");
STATISTIC(NumUncacheNonLocal, "Number of uncached non-local responses");
STATISTIC(NumCacheCompleteNonLocalPtr,
"Number of block queries that were completely cached");
+// Limit for the number of instructions to scan in a block.
+static const int BlockScanLimit = 100;
+
char MemoryDependenceAnalysis::ID = 0;
-
+
// Register this pass...
INITIALIZE_PASS_BEGIN(MemoryDependenceAnalysis, "memdep",
"Memory Dependence Analysis", false, true)
"Memory Dependence Analysis", false, true)
MemoryDependenceAnalysis::MemoryDependenceAnalysis()
-: FunctionPass(ID), PredCache(0) {
+ : FunctionPass(ID), PredCache() {
initializeMemoryDependenceAnalysisPass(*PassRegistry::getPassRegistry());
}
MemoryDependenceAnalysis::~MemoryDependenceAnalysis() {
bool MemoryDependenceAnalysis::runOnFunction(Function &) {
AA = &getAnalysis<AliasAnalysis>();
- TD = getAnalysisIfAvailable<TargetData>();
- if (PredCache == 0)
+ DataLayoutPass *DLP = getAnalysisIfAvailable<DataLayoutPass>();
+ DL = DLP ? &DLP->getDataLayout() : nullptr;
+ DominatorTreeWrapperPass *DTWP =
+ getAnalysisIfAvailable<DominatorTreeWrapperPass>();
+ DT = DTWP ? &DTWP->getDomTree() : nullptr;
+ if (!PredCache)
PredCache.reset(new PredIteratorCache());
return false;
}
/// RemoveFromReverseMap - This is a helper function that removes Val from
/// 'Inst's set in ReverseMap. If the set becomes empty, remove Inst's entry.
template <typename KeyTy>
-static void RemoveFromReverseMap(DenseMap<Instruction*,
+static void RemoveFromReverseMap(DenseMap<Instruction*,
SmallPtrSet<KeyTy, 4> > &ReverseMap,
Instruction *Inst, KeyTy Val) {
typename DenseMap<Instruction*, SmallPtrSet<KeyTy, 4> >::iterator
AliasAnalysis::Location &Loc,
AliasAnalysis *AA) {
if (const LoadInst *LI = dyn_cast<LoadInst>(Inst)) {
- if (LI->isVolatile()) {
- Loc = AliasAnalysis::Location();
+ if (LI->isUnordered()) {
+ Loc = AA->getLocation(LI);
+ return AliasAnalysis::Ref;
+ }
+ if (LI->getOrdering() == Monotonic) {
+ Loc = AA->getLocation(LI);
return AliasAnalysis::ModRef;
}
- Loc = AA->getLocation(LI);
- return AliasAnalysis::Ref;
+ Loc = AliasAnalysis::Location();
+ return AliasAnalysis::ModRef;
}
if (const StoreInst *SI = dyn_cast<StoreInst>(Inst)) {
- if (SI->isVolatile()) {
- Loc = AliasAnalysis::Location();
+ if (SI->isUnordered()) {
+ Loc = AA->getLocation(SI);
+ return AliasAnalysis::Mod;
+ }
+ if (SI->getOrdering() == Monotonic) {
+ Loc = AA->getLocation(SI);
return AliasAnalysis::ModRef;
}
- Loc = AA->getLocation(SI);
- return AliasAnalysis::Mod;
+ Loc = AliasAnalysis::Location();
+ return AliasAnalysis::ModRef;
}
if (const VAArgInst *V = dyn_cast<VAArgInst>(Inst)) {
return AliasAnalysis::ModRef;
}
- if (const CallInst *CI = isFreeCall(Inst)) {
+ if (const CallInst *CI = isFreeCall(Inst, AA->getTargetLibraryInfo())) {
// calls to free() deallocate the entire structure
Loc = AliasAnalysis::Location(CI->getArgOperand(0));
return AliasAnalysis::Mod;
MemDepResult MemoryDependenceAnalysis::
getCallSiteDependencyFrom(CallSite CS, bool isReadOnlyCall,
BasicBlock::iterator ScanIt, BasicBlock *BB) {
+ unsigned Limit = BlockScanLimit;
+
// Walk backwards through the block, looking for dependencies
while (ScanIt != BB->begin()) {
+ // Limit the amount of scanning we do so we don't end up with quadratic
+ // running time on extreme testcases.
+ --Limit;
+ if (!Limit)
+ return MemDepResult::getUnknown();
+
Instruction *Inst = --ScanIt;
-
+
// If this inst is a memory op, get the pointer it accessed
AliasAnalysis::Location Loc;
AliasAnalysis::ModRefResult MR = GetLocation(Inst, Loc, AA);
// Otherwise if the two calls don't interact (e.g. InstCS is readnone)
// keep scanning.
- break;
+ continue;
default:
return MemDepResult::getClobber(Inst);
}
}
+
+ // If we could not obtain a pointer for the instruction and the instruction
+ // touches memory then assume that this is a dependency.
+ if (MR != AliasAnalysis::NoModRef)
+ return MemDepResult::getClobber(Inst);
}
-
- // No dependence found. If this is the entry block of the function, it is a
- // clobber, otherwise it is non-local.
+
+ // No dependence found. If this is the entry block of the function, it is
+ // unknown, otherwise it is non-local.
if (BB != &BB->getParent()->getEntryBlock())
return MemDepResult::getNonLocal();
- return MemDepResult::getClobber(ScanIt);
+ return MemDepResult::getNonFuncLocal();
+}
+
+/// isLoadLoadClobberIfExtendedToFullWidth - Return true if LI is a load that
+/// would fully overlap MemLoc if done as a wider legal integer load.
+///
+/// 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;
+
+ // 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);
+ return Size != 0;
+}
+
+/// getLoadLoadClobberFullWidthSize - This is a little bit of analysis that
+/// looks at a memory location for a load (specified by MemLocBase, Offs,
+/// and Size) and compares it against a load. If the specified load could
+/// be safely widened to a larger integer load that is 1) still efficient,
+/// 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) {
+ // 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))
+ return 0;
+
+ // Get the base of this load.
+ int64_t LIOffs = 0;
+ const Value *LIBase =
+ GetPointerBaseWithConstantOffset(LI->getPointerOperand(), LIOffs, &DL);
+
+ // If the two pointers are not based on the same pointer, we can't tell that
+ // they are related.
+ if (LIBase != MemLocBase) return 0;
+
+ // Okay, the two values are based on the same pointer, but returned as
+ // no-alias. This happens when we have things like two byte loads at "P+1"
+ // and "P+3". Check to see if increasing the size of the "LI" load up to its
+ // alignment (or the largest native integer type) will allow us to load all
+ // the bits required by MemLoc.
+
+ // If MemLoc is before LI, then no widening of LI will help us out.
+ if (MemLocOffs < LIOffs) return 0;
+
+ // Get the alignment of the load in bytes. We assume that it is safe to load
+ // any legal integer up to this size without a problem. For example, if we're
+ // looking at an i8 load on x86-32 that is known 1024 byte aligned, we can
+ // widen it up to an i32 load. If it is known 2-byte aligned, we can widen it
+ // to i16.
+ unsigned LoadAlign = LI->getAlignment();
+
+ int64_t MemLocEnd = MemLocOffs+MemLocSize;
+
+ // If no amount of rounding up will let MemLoc fit into LI, then bail out.
+ if (LIOffs+LoadAlign < MemLocEnd) return 0;
+
+ // This is the size of the load to try. Start with the next larger power of
+ // two.
+ unsigned NewLoadByteSize = LI->getType()->getPrimitiveSizeInBits()/8U;
+ NewLoadByteSize = NextPowerOf2(NewLoadByteSize);
+
+ while (1) {
+ // If this load size is bigger than our known alignment or would not fit
+ // into a native integer register, then we fail.
+ if (NewLoadByteSize > LoadAlign ||
+ !DL.fitsInLegalInteger(NewLoadByteSize*8))
+ return 0;
+
+ if (LIOffs+NewLoadByteSize > MemLocEnd &&
+ LI->getParent()->getParent()->getAttributes().
+ hasAttribute(AttributeSet::FunctionIndex, 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.
+ return 0;
+
+ // If a load of this width would include all of MemLoc, then we succeed.
+ if (LIOffs+NewLoadByteSize >= MemLocEnd)
+ return NewLoadByteSize;
+
+ NewLoadByteSize <<= 1;
+ }
}
/// 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
-/// with reads from read-only locations.
+/// with reads from read-only locations. If possible, pass the query
+/// instruction as well; this function may take advantage of the metadata
+/// annotated to the query instruction to refine the result.
MemDepResult MemoryDependenceAnalysis::
-getPointerDependencyFrom(const AliasAnalysis::Location &MemLoc, bool isLoad,
- BasicBlock::iterator ScanIt, BasicBlock *BB) {
-
- Value *InvariantTag = 0;
+getPointerDependencyFrom(const AliasAnalysis::Location &MemLoc, bool isLoad,
+ BasicBlock::iterator ScanIt, BasicBlock *BB,
+ Instruction *QueryInst) {
+
+ const Value *MemLocBase = nullptr;
+ int64_t MemLocOffset = 0;
+ unsigned Limit = BlockScanLimit;
+ bool isInvariantLoad = false;
+ if (isLoad && QueryInst) {
+ LoadInst *LI = dyn_cast<LoadInst>(QueryInst);
+ if (LI && LI->getMetadata(LLVMContext::MD_invariant_load) != nullptr)
+ isInvariantLoad = true;
+ }
// Walk backwards through the basic block, looking for dependencies.
while (ScanIt != BB->begin()) {
Instruction *Inst = --ScanIt;
- // If we're in an invariant region, no dependencies can be found before
- // we pass an invariant-begin marker.
- if (InvariantTag == Inst) {
- InvariantTag = 0;
- continue;
- }
-
- if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(Inst)) {
- // Debug intrinsics don't (and can't) cause dependences.
+ if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(Inst))
+ // Debug intrinsics don't (and can't) cause dependencies.
if (isa<DbgInfoIntrinsic>(II)) continue;
-
- // If we pass an invariant-end marker, then we've just entered an
- // invariant region and can start ignoring dependencies.
- if (II->getIntrinsicID() == Intrinsic::invariant_end) {
- // FIXME: This only considers queries directly on the invariant-tagged
- // pointer, not on query pointers that are indexed off of them. It'd
- // be nice to handle that at some point.
- AliasAnalysis::AliasResult R =
- AA->alias(AliasAnalysis::Location(II->getArgOperand(2)), MemLoc);
- if (R == AliasAnalysis::MustAlias)
- InvariantTag = II->getArgOperand(0);
- continue;
- }
+ // Limit the amount of scanning we do so we don't end up with quadratic
+ // running time on extreme testcases.
+ --Limit;
+ if (!Limit)
+ return MemDepResult::getUnknown();
+ if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(Inst)) {
// If we reach a lifetime begin or end marker, then the query ends here
// because the value is undefined.
if (II->getIntrinsicID() == Intrinsic::lifetime_start) {
// FIXME: This only considers queries directly on the invariant-tagged
// pointer, not on query pointers that are indexed off of them. It'd
- // be nice to handle that at some point.
- AliasAnalysis::AliasResult R =
- AA->alias(AliasAnalysis::Location(II->getArgOperand(1)), MemLoc);
- if (R == AliasAnalysis::MustAlias)
+ // be nice to handle that at some point (the right approach is to use
+ // GetPointerBaseWithConstantOffset).
+ if (AA->isMustAlias(AliasAnalysis::Location(II->getArgOperand(1)),
+ MemLoc))
return MemDepResult::getDef(II);
continue;
}
}
- // If we're querying on a load and we're in an invariant region, we're done
- // at this point. Nothing a load depends on can live in an invariant region.
- //
- // FIXME: this will prevent us from returning load/load must-aliases, so GVN
- // won't remove redundant loads.
- if (isLoad && InvariantTag) continue;
-
// Values depend on loads if the pointers are must aliased. This means that
// a load depends on another must aliased load from the same value.
if (LoadInst *LI = dyn_cast<LoadInst>(Inst)) {
+ // Atomic loads have complications involved.
+ // FIXME: This is overly conservative.
+ if (!LI->isUnordered())
+ 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.
AliasAnalysis::AliasResult R = AA->alias(LoadLoc, MemLoc);
- if (R == AliasAnalysis::NoAlias)
+
+ if (isLoad) {
+ if (R == AliasAnalysis::NoAlias) {
+ // If this is an over-aligned integer load (for example,
+ // "load i8* %P, align 4") see if it would obviously overlap with the
+ // queried location if widened to a larger load (e.g. if the queried
+ // 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() &&
+ isLoadLoadClobberIfExtendedToFullWidth(MemLoc, MemLocBase,
+ MemLocOffset, LI, DL))
+ return MemDepResult::getClobber(Inst);
+
+ continue;
+ }
+
+ // Must aliased loads are defs of each other.
+ if (R == AliasAnalysis::MustAlias)
+ return MemDepResult::getDef(Inst);
+
+#if 0 // FIXME: Temporarily disabled. GVN is cleverly rewriting loads
+ // in terms of clobbering loads, but since it does this by looking
+ // at the clobbering load directly, it doesn't know about any
+ // phi translation that may have happened along the way.
+
+ // If we have a partial alias, then return this as a clobber for the
+ // client to handle.
+ if (R == AliasAnalysis::PartialAlias)
+ return MemDepResult::getClobber(Inst);
+#endif
+
+ // Random may-alias loads don't depend on each other without a
+ // dependence.
continue;
-
- // May-alias loads don't depend on each other without a dependence.
- if (isLoad && R != AliasAnalysis::MustAlias)
+ }
+
+ // Stores don't depend on other no-aliased accesses.
+ if (R == AliasAnalysis::NoAlias)
continue;
// Stores don't alias loads from read-only memory.
- if (!isLoad && AA->pointsToConstantMemory(LoadLoc))
+ if (AA->pointsToConstantMemory(LoadLoc))
continue;
- // Stores depend on may and must aliased loads, loads depend on must-alias
- // loads.
+ // Stores depend on may/must aliased loads.
return MemDepResult::getDef(Inst);
}
-
+
if (StoreInst *SI = dyn_cast<StoreInst>(Inst)) {
- // There can't be stores to the value we care about inside an
- // invariant region.
- if (InvariantTag) continue;
-
+ // Atomic stores have complications involved.
+ // FIXME: This is overly conservative.
+ if (!SI->isUnordered())
+ return MemDepResult::getClobber(SI);
+
// If alias analysis can tell that this store is guaranteed to not modify
// the query pointer, ignore it. Use getModRefInfo to handle cases where
// the query pointer points to constant memory etc.
// Ok, this store might clobber the query pointer. Check to see if it is
// a must alias: in this case, we want to return this as a def.
AliasAnalysis::Location StoreLoc = AA->getLocation(SI);
-
+
// If we found a pointer, check if it could be the same as our pointer.
AliasAnalysis::AliasResult R = AA->alias(StoreLoc, MemLoc);
-
+
if (R == AliasAnalysis::NoAlias)
continue;
if (R == AliasAnalysis::MustAlias)
return MemDepResult::getDef(Inst);
+ if (isInvariantLoad)
+ continue;
return MemDepResult::getClobber(Inst);
}
// a subsequent bitcast of the malloc call result. There can be stores to
// the malloced memory between the malloc call and its bitcast uses, and we
// need to continue scanning until the malloc call.
- if (isa<AllocaInst>(Inst) ||
- (isa<CallInst>(Inst) && extractMallocCall(Inst))) {
- const Value *AccessPtr = GetUnderlyingObject(MemLoc.Ptr, TD);
-
- if (AccessPtr == Inst ||
- AA->alias(Inst, 1, AccessPtr, 1) == AliasAnalysis::MustAlias)
+ const TargetLibraryInfo *TLI = AA->getTargetLibraryInfo();
+ if (isa<AllocaInst>(Inst) || isNoAliasFn(Inst, TLI)) {
+ const Value *AccessPtr = GetUnderlyingObject(MemLoc.Ptr, DL);
+
+ if (AccessPtr == Inst || AA->isMustAlias(Inst, AccessPtr))
return MemDepResult::getDef(Inst);
- continue;
+ // Be conservative if the accessed pointer may alias the allocation.
+ if (AA->alias(Inst, AccessPtr) != AliasAnalysis::NoAlias)
+ return MemDepResult::getClobber(Inst);
+ // If the allocation is not aliased and does not read memory (like
+ // strdup), it is safe to ignore.
+ if (isa<AllocaInst>(Inst) ||
+ isMallocLikeFn(Inst, TLI) || isCallocLikeFn(Inst, TLI))
+ continue;
}
// See if this instruction (e.g. a call or vaarg) mod/ref's the pointer.
- switch (AA->getModRefInfo(Inst, MemLoc)) {
+ AliasAnalysis::ModRefResult MR = AA->getModRefInfo(Inst, MemLoc);
+ // If necessary, perform additional analysis.
+ if (MR == AliasAnalysis::ModRef)
+ MR = AA->callCapturesBefore(Inst, MemLoc, DT);
+ switch (MR) {
case AliasAnalysis::NoModRef:
// If the call has no effect on the queried pointer, just ignore it.
continue;
case AliasAnalysis::Mod:
- // If we're in an invariant region, we can ignore calls that ONLY
- // modify the pointer.
- if (InvariantTag) continue;
return MemDepResult::getClobber(Inst);
case AliasAnalysis::Ref:
// If the call is known to never store to the pointer, and if this is a
return MemDepResult::getClobber(Inst);
}
}
-
- // No dependence found. If this is the entry block of the function, it is a
- // clobber, otherwise it is non-local.
+
+ // No dependence found. If this is the entry block of the function, it is
+ // unknown, otherwise it is non-local.
if (BB != &BB->getParent()->getEntryBlock())
return MemDepResult::getNonLocal();
- return MemDepResult::getClobber(ScanIt);
+ return MemDepResult::getNonFuncLocal();
}
/// getDependency - Return the instruction on which a memory operation
/// depends.
MemDepResult MemoryDependenceAnalysis::getDependency(Instruction *QueryInst) {
Instruction *ScanPos = QueryInst;
-
+
// Check for a cached result
MemDepResult &LocalCache = LocalDeps[QueryInst];
-
+
// If the cached entry is non-dirty, just return it. Note that this depends
// on MemDepResult's default constructing to 'dirty'.
if (!LocalCache.isDirty())
return LocalCache;
-
+
// Otherwise, if we have a dirty entry, we know we can start the scan at that
// instruction, which may save us some work.
if (Instruction *Inst = LocalCache.getInst()) {
ScanPos = Inst;
-
+
RemoveFromReverseMap(ReverseLocalDeps, Inst, QueryInst);
}
-
+
BasicBlock *QueryParent = QueryInst->getParent();
-
+
// Do the scan.
if (BasicBlock::iterator(QueryInst) == QueryParent->begin()) {
- // No dependence found. If this is the entry block of the function, it is a
- // clobber, otherwise it is non-local.
+ // No dependence found. If this is the entry block of the function, it is
+ // unknown, otherwise it is non-local.
if (QueryParent != &QueryParent->getParent()->getEntryBlock())
LocalCache = MemDepResult::getNonLocal();
else
- LocalCache = MemDepResult::getClobber(QueryInst);
+ LocalCache = MemDepResult::getNonFuncLocal();
} else {
AliasAnalysis::Location MemLoc;
AliasAnalysis::ModRefResult MR = GetLocation(QueryInst, MemLoc, AA);
// If we can do a pointer scan, make it happen.
bool isLoad = !(MR & AliasAnalysis::Mod);
if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(QueryInst))
- isLoad |= II->getIntrinsicID() == Intrinsic::lifetime_end;
+ isLoad |= II->getIntrinsicID() == Intrinsic::lifetime_start;
LocalCache = getPointerDependencyFrom(MemLoc, isLoad, ScanPos,
- QueryParent);
+ QueryParent, QueryInst);
} else if (isa<CallInst>(QueryInst) || isa<InvokeInst>(QueryInst)) {
CallSite QueryCS(QueryInst);
bool isReadOnly = AA->onlyReadsMemory(QueryCS);
QueryParent);
} else
// Non-memory instruction.
- LocalCache = MemDepResult::getClobber(--BasicBlock::iterator(ScanPos));
+ LocalCache = MemDepResult::getUnknown();
}
-
+
// Remember the result!
if (Instruction *I = LocalCache.getInst())
ReverseLocalDeps[I].insert(QueryInst);
-
+
return LocalCache;
}
/// the uncached case, this starts out as the set of predecessors we care
/// about.
SmallVector<BasicBlock*, 32> DirtyBlocks;
-
+
if (!Cache.empty()) {
// Okay, we have a cache entry. If we know it is not dirty, just return it
// with no computation.
++NumCacheNonLocal;
return Cache;
}
-
+
// If we already have a partially computed set of results, scan them to
// determine what is dirty, seeding our initial DirtyBlocks worklist.
for (NonLocalDepInfo::iterator I = Cache.begin(), E = Cache.end();
I != E; ++I)
if (I->getResult().isDirty())
DirtyBlocks.push_back(I->getBB());
-
+
// Sort the cache so that we can do fast binary search lookups below.
std::sort(Cache.begin(), Cache.end());
-
+
++NumCacheDirtyNonLocal;
//cerr << "CACHED CASE: " << DirtyBlocks.size() << " dirty: "
// << Cache.size() << " cached: " << *QueryInst;
DirtyBlocks.push_back(*PI);
++NumUncacheNonLocal;
}
-
+
// isReadonlyCall - If this is a read-only call, we can be more aggressive.
bool isReadonlyCall = AA->onlyReadsMemory(QueryCS);
SmallPtrSet<BasicBlock*, 64> Visited;
-
+
unsigned NumSortedEntries = Cache.size();
DEBUG(AssertSorted(Cache));
-
+
// Iterate while we still have blocks to update.
while (!DirtyBlocks.empty()) {
BasicBlock *DirtyBB = DirtyBlocks.back();
DirtyBlocks.pop_back();
-
+
// Already processed this block?
if (!Visited.insert(DirtyBB))
continue;
-
+
// Do a binary search to see if we already have an entry for this block in
// the cache set. If so, find it.
DEBUG(AssertSorted(Cache, NumSortedEntries));
- NonLocalDepInfo::iterator Entry =
+ NonLocalDepInfo::iterator Entry =
std::upper_bound(Cache.begin(), Cache.begin()+NumSortedEntries,
NonLocalDepEntry(DirtyBB));
- if (Entry != Cache.begin() && prior(Entry)->getBB() == DirtyBB)
+ if (Entry != Cache.begin() && std::prev(Entry)->getBB() == DirtyBB)
--Entry;
-
- NonLocalDepEntry *ExistingResult = 0;
- if (Entry != Cache.begin()+NumSortedEntries &&
+
+ NonLocalDepEntry *ExistingResult = nullptr;
+ if (Entry != Cache.begin()+NumSortedEntries &&
Entry->getBB() == DirtyBB) {
// If we already have an entry, and if it isn't already dirty, the block
// is done.
if (!Entry->getResult().isDirty())
continue;
-
+
// Otherwise, remember this slot so we can update the value.
ExistingResult = &*Entry;
}
-
+
// If the dirty entry has a pointer, start scanning from it so we don't have
// to rescan the entire block.
BasicBlock::iterator ScanPos = DirtyBB->end();
QueryCS.getInstruction());
}
}
-
+
// Find out if this block has a local dependency for QueryInst.
MemDepResult Dep;
-
+
if (ScanPos != DirtyBB->begin()) {
Dep = getCallSiteDependencyFrom(QueryCS, isReadonlyCall,ScanPos, DirtyBB);
} else if (DirtyBB != &DirtyBB->getParent()->getEntryBlock()) {
// No dependence found. If this is the entry block of the function, it is
- // a clobber, otherwise it is non-local.
+ // a clobber, otherwise it is unknown.
Dep = MemDepResult::getNonLocal();
} else {
- Dep = MemDepResult::getClobber(ScanPos);
+ Dep = MemDepResult::getNonFuncLocal();
}
-
+
// If we had a dirty entry for the block, update it. Otherwise, just add
// a new entry.
if (ExistingResult)
ExistingResult->setResult(Dep);
else
Cache.push_back(NonLocalDepEntry(DirtyBB, Dep));
-
+
// If the block has a dependency (i.e. it isn't completely transparent to
// the value), remember the association!
if (!Dep.isNonLocal()) {
if (Instruction *Inst = Dep.getInst())
ReverseNonLocalDeps[Inst].insert(QueryCS.getInstruction());
} else {
-
+
// 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);
}
}
-
+
return Cache;
}
assert(Loc.Ptr->getType()->isPointerTy() &&
"Can't get pointer deps of a non-pointer!");
Result.clear();
-
- PHITransAddr Address(const_cast<Value *>(Loc.Ptr), TD);
-
+
+ PHITransAddr Address(const_cast<Value *>(Loc.Ptr), DL);
+
// 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
return;
Result.clear();
Result.push_back(NonLocalDepResult(FromBB,
- MemDepResult::getClobber(FromBB->begin()),
+ MemDepResult::getUnknown(),
const_cast<Value *>(Loc.Ptr)));
}
GetNonLocalInfoForBlock(const AliasAnalysis::Location &Loc,
bool isLoad, BasicBlock *BB,
NonLocalDepInfo *Cache, unsigned NumSortedEntries) {
-
+
// Do a binary search to see if we already have an entry for this block in
// the cache set. If so, find it.
NonLocalDepInfo::iterator Entry =
NonLocalDepEntry(BB));
if (Entry != Cache->begin() && (Entry-1)->getBB() == BB)
--Entry;
-
- NonLocalDepEntry *ExistingResult = 0;
+
+ NonLocalDepEntry *ExistingResult = nullptr;
if (Entry != Cache->begin()+NumSortedEntries && Entry->getBB() == BB)
ExistingResult = &*Entry;
-
+
// If we have a cached entry, and it is non-dirty, use it as the value for
// this dependency.
if (ExistingResult && !ExistingResult->getResult().isDirty()) {
++NumCacheNonLocalPtr;
return ExistingResult->getResult();
- }
-
+ }
+
// Otherwise, we have to scan for the value. If we have a dirty cache
// entry, start scanning from its position, otherwise we scan from the end
// of the block.
"Instruction invalidated?");
++NumCacheDirtyNonLocalPtr;
ScanPos = ExistingResult->getResult().getInst();
-
+
// Eliminating the dirty entry from 'Cache', so update the reverse info.
ValueIsLoadPair CacheKey(Loc.Ptr, isLoad);
RemoveFromReverseMap(ReverseNonLocalPtrDeps, ScanPos, CacheKey);
} else {
++NumUncacheNonLocalPtr;
}
-
+
// Scan the block for the dependency.
MemDepResult Dep = getPointerDependencyFrom(Loc, isLoad, ScanPos, BB);
-
+
// If we had a dirty entry for the block, update it. Otherwise, just add
// a new entry.
if (ExistingResult)
ExistingResult->setResult(Dep);
else
Cache->push_back(NonLocalDepEntry(BB, Dep));
-
+
// If the block has a dependency (i.e. it isn't completely transparent to
// the value), remember the reverse association because we just added it
// to Cache!
- if (Dep.isNonLocal())
+ if (!Dep.isDef() && !Dep.isClobber())
return Dep;
-
+
// Keep the ReverseNonLocalPtrDeps map up to date so we can efficiently
// update MemDep when we remove instructions.
Instruction *Inst = Dep.getInst();
/// SortNonLocalDepInfoCache - Sort the a NonLocalDepInfo cache, given a certain
/// number of elements in the array that are already properly ordered. This is
/// optimized for the case when only a few entries are added.
-static void
+static void
SortNonLocalDepInfoCache(MemoryDependenceAnalysis::NonLocalDepInfo &Cache,
unsigned NumSortedEntries) {
switch (Cache.size() - NumSortedEntries) {
SmallVectorImpl<NonLocalDepResult> &Result,
DenseMap<BasicBlock*, Value*> &Visited,
bool SkipFirstBlock) {
-
// Look up the cached info for Pointer.
ValueIsLoadPair CacheKey(Pointer.getAddr(), isLoad);
// Get the NLPI for CacheKey, inserting one into the map if it doesn't
// already have one.
- std::pair<CachedNonLocalPointerInfo::iterator, bool> Pair =
+ std::pair<CachedNonLocalPointerInfo::iterator, bool> Pair =
NonLocalPointerDeps.insert(std::make_pair(CacheKey, InitialNLPI));
NonLocalPointerInfo *CacheInfo = &Pair.first->second;
if (!Pair.second) {
if (CacheInfo->Size < Loc.Size) {
// The query's Size is greater than the cached one. Throw out the
- // cached data and procede with the query at the greater size.
+ // cached data and proceed with the query at the greater size.
CacheInfo->Pair = BBSkipFirstBlockPair();
CacheInfo->Size = Loc.Size;
for (NonLocalDepInfo::iterator DI = CacheInfo->NonLocalDeps.begin(),
if (CacheInfo->TBAATag != Loc.TBAATag) {
if (CacheInfo->TBAATag) {
CacheInfo->Pair = BBSkipFirstBlockPair();
- CacheInfo->TBAATag = 0;
+ CacheInfo->TBAATag = nullptr;
for (NonLocalDepInfo::iterator DI = CacheInfo->NonLocalDeps.begin(),
DE = CacheInfo->NonLocalDeps.end(); DI != DE; ++DI)
if (Instruction *Inst = DI->getResult().getInst())
DenseMap<BasicBlock*, Value*>::iterator VI = Visited.find(I->getBB());
if (VI == Visited.end() || VI->second == Pointer.getAddr())
continue;
-
+
// We have a pointer mismatch in a block. Just return clobber, saying
// that something was clobbered in this result. We could also do a
// non-fully cached query, but there is little point in doing this.
return true;
}
}
-
+
Value *Addr = Pointer.getAddr();
for (NonLocalDepInfo::iterator I = Cache->begin(), E = Cache->end();
I != E; ++I) {
Visited.insert(std::make_pair(I->getBB(), Addr));
- if (!I->getResult().isNonLocal())
+ if (I->getResult().isNonLocal()) {
+ continue;
+ }
+
+ if (!DT) {
+ Result.push_back(NonLocalDepResult(I->getBB(),
+ MemDepResult::getUnknown(),
+ Addr));
+ } else if (DT->isReachableFromEntry(I->getBB())) {
Result.push_back(NonLocalDepResult(I->getBB(), I->getResult(), Addr));
+ }
}
++NumCacheCompleteNonLocalPtr;
return false;
}
-
+
// Otherwise, either this is a new block, a block with an invalid cache
// pointer or one that we're about to invalidate by putting more info into it
// than its valid cache info. If empty, the result will be valid cache info,
CacheInfo->Pair = BBSkipFirstBlockPair(StartBB, SkipFirstBlock);
else
CacheInfo->Pair = BBSkipFirstBlockPair();
-
+
SmallVector<BasicBlock*, 32> Worklist;
Worklist.push_back(StartBB);
-
+
+ // PredList used inside loop.
+ SmallVector<std::pair<BasicBlock*, PHITransAddr>, 16> PredList;
+
// Keep track of the entries that we know are sorted. Previously cached
// entries will all be sorted. The entries we add we only sort on demand (we
// don't insert every element into its sorted position). We know that we
// revisit blocks after we insert info for them.
unsigned NumSortedEntries = Cache->size();
DEBUG(AssertSorted(*Cache));
-
+
while (!Worklist.empty()) {
BasicBlock *BB = Worklist.pop_back_val();
-
+
// Skip the first block if we have it.
if (!SkipFirstBlock) {
// Analyze the dependency of *Pointer in FromBB. See if we already have
DEBUG(AssertSorted(*Cache, NumSortedEntries));
MemDepResult Dep = GetNonLocalInfoForBlock(Loc, isLoad, BB, Cache,
NumSortedEntries);
-
+
// If we got a Def or Clobber, add this to the list of results.
if (!Dep.isNonLocal()) {
- Result.push_back(NonLocalDepResult(BB, Dep, Pointer.getAddr()));
- continue;
+ if (!DT) {
+ Result.push_back(NonLocalDepResult(BB,
+ MemDepResult::getUnknown(),
+ Pointer.getAddr()));
+ continue;
+ } else if (DT->isReachableFromEntry(BB)) {
+ Result.push_back(NonLocalDepResult(BB, Dep, Pointer.getAddr()));
+ continue;
+ }
}
}
-
+
// If 'Pointer' is an instruction defined in this block, then we need to do
// phi translation to change it into a value live in the predecessor block.
// If not, we just add the predecessors to the worklist and scan them with
// the same Pointer.
if (!Pointer.NeedsPHITranslationFromBlock(BB)) {
SkipFirstBlock = false;
+ SmallVector<BasicBlock*, 16> NewBlocks;
for (BasicBlock **PI = PredCache->GetPreds(BB); *PI; ++PI) {
// 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()));
if (InsertRes.second) {
// First time we've looked at *PI.
- Worklist.push_back(*PI);
+ NewBlocks.push_back(*PI);
continue;
}
-
+
// If we have seen this block before, but it was with a different
// pointer then we have a phi translation failure and we have to treat
// this as a clobber.
- if (InsertRes.first->second != Pointer.getAddr())
+ if (InsertRes.first->second != Pointer.getAddr()) {
+ // Make sure to clean up the Visited map before continuing on to
+ // PredTranslationFailure.
+ for (unsigned i = 0; i < NewBlocks.size(); i++)
+ Visited.erase(NewBlocks[i]);
goto PredTranslationFailure;
+ }
}
+ Worklist.append(NewBlocks.begin(), NewBlocks.end());
continue;
}
-
+
// We do need to do phi translation, if we know ahead of time we can't phi
// translate this value, don't even try.
if (!Pointer.IsPotentiallyPHITranslatable())
goto PredTranslationFailure;
-
+
// We may have added values to the cache list before this PHI translation.
// If so, we haven't done anything to ensure that the cache remains sorted.
// Sort it now (if needed) so that recursive invocations of
SortNonLocalDepInfoCache(*Cache, NumSortedEntries);
NumSortedEntries = Cache->size();
}
- Cache = 0;
-
+ Cache = nullptr;
+
+ PredList.clear();
for (BasicBlock **PI = PredCache->GetPreds(BB); *PI; ++PI) {
BasicBlock *Pred = *PI;
-
+ 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(Pointer);
- PredPointer.PHITranslateValue(BB, Pred, 0);
+ PHITransAddr &PredPointer = PredList.back().second;
+ PredPointer.PHITranslateValue(BB, Pred, nullptr);
Value *PredPtrVal = PredPointer.getAddr();
-
+
// Check to see if we have already visited this pred block with another
// pointer. If so, we can't do this lookup. This failure can occur
// with PHI translation when a critical edge exists and the PHI node in
InsertRes = Visited.insert(std::make_pair(Pred, PredPtrVal));
if (!InsertRes.second) {
+ // We found the pred; take it off the list of preds to visit.
+ PredList.pop_back();
+
// If the predecessor was visited with PredPtr, then we already did
// the analysis and can ignore it.
if (InsertRes.first->second == PredPtrVal)
continue;
-
+
// Otherwise, the block was previously analyzed with a different
// pointer. We can't represent the result of this case, so we just
// treat this as a phi translation failure.
+
+ // Make sure to clean up the Visited map before continuing on to
+ // PredTranslationFailure.
+ for (unsigned i = 0, n = PredList.size(); i < n; ++i)
+ Visited.erase(PredList[i].first);
+
goto PredTranslationFailure;
}
-
+ }
+
+ // Actually process results here; this need to be a separate loop to avoid
+ // calling getNonLocalPointerDepFromBB for blocks we don't want to return
+ // any results for. (getNonLocalPointerDepFromBB will modify our
+ // datastructures in ways the code after the PredTranslationFailure label
+ // doesn't expect.)
+ for (unsigned i = 0, n = PredList.size(); i < n; ++i) {
+ BasicBlock *Pred = PredList[i].first;
+ PHITransAddr &PredPointer = PredList[i].second;
+ Value *PredPtrVal = PredPointer.getAddr();
+
+ bool CanTranslate = true;
// If PHI translation was unable to find an available pointer in this
// predecessor, then we have to assume that the pointer is clobbered in
// that predecessor. We can still do PRE of the load, which would insert
// a computation of the pointer in this predecessor.
- if (PredPtrVal == 0) {
+ if (!PredPtrVal)
+ CanTranslate = false;
+
+ // FIXME: it is entirely possible that PHI translating will end up with
+ // the same value. Consider PHI translating something like:
+ // X = phi [x, bb1], [y, bb2]. PHI translating for bb1 doesn't *need*
+ // to recurse here, pedantically speaking.
+
+ // If getNonLocalPointerDepFromBB fails here, that means the cached
+ // 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,
+ Loc.getWithNewPtr(PredPtrVal),
+ isLoad, Pred,
+ Result, Visited)) {
// Add the entry to the Result list.
- NonLocalDepResult Entry(Pred,
- MemDepResult::getClobber(Pred->getTerminator()),
- PredPtrVal);
+ NonLocalDepResult Entry(Pred, MemDepResult::getUnknown(), PredPtrVal);
Result.push_back(Entry);
// Since we had a phi translation failure, the cache for CacheKey won't
NLPI.Pair = BBSkipFirstBlockPair();
continue;
}
-
- // FIXME: it is entirely possible that PHI translating will end up with
- // the same value. Consider PHI translating something like:
- // X = phi [x, bb1], [y, bb2]. PHI translating for bb1 doesn't *need*
- // to recurse here, pedantically speaking.
-
- // If we have a problem phi translating, fall through to the code below
- // to handle the failure condition.
- if (getNonLocalPointerDepFromBB(PredPointer,
- Loc.getWithNewPtr(PredPointer.getAddr()),
- isLoad, Pred,
- Result, Visited))
- goto PredTranslationFailure;
}
-
+
// Refresh the CacheInfo/Cache pointer so that it isn't invalidated.
CacheInfo = &NonLocalPointerDeps[CacheKey];
Cache = &CacheInfo->NonLocalDeps;
NumSortedEntries = Cache->size();
-
+
// Since we did phi translation, the "Cache" set won't contain all of the
// results for the query. This is ok (we can still use it to accelerate
// specific block queries) but we can't do the fastpath "return all
continue;
PredTranslationFailure:
-
- if (Cache == 0) {
+ // The following code is "failure"; we can't produce a sane translation
+ // for the given block. It assumes that we haven't modified any of
+ // our datastructures while processing the current block.
+
+ if (!Cache) {
// Refresh the CacheInfo/Cache pointer if it got invalidated.
CacheInfo = &NonLocalPointerDeps[CacheKey];
Cache = &CacheInfo->NonLocalDeps;
NumSortedEntries = Cache->size();
}
-
+
// Since we failed phi translation, the "Cache" set won't contain all of the
// results for the query. This is ok (we can still use it to accelerate
// specific block queries) but we can't do the fastpath "return all
// results from the set". Clear out the indicator for this.
CacheInfo->Pair = BBSkipFirstBlockPair();
-
- // If *nothing* works, mark the pointer as being clobbered by the first
- // instruction in this block.
+
+ // If *nothing* works, mark the pointer as unknown.
//
// If this is the magic first block, return this as a clobber of the whole
// incoming value. Since we can't phi translate to one of the predecessors,
// we have to bail out.
if (SkipFirstBlock)
return true;
-
+
for (NonLocalDepInfo::reverse_iterator I = Cache->rbegin(); ; ++I) {
assert(I != Cache->rend() && "Didn't find current block??");
if (I->getBB() != BB)
continue;
-
+
assert(I->getResult().isNonLocal() &&
"Should only be here with transparent block");
- I->setResult(MemDepResult::getClobber(BB->begin()));
- ReverseNonLocalPtrDeps[BB->begin()].insert(CacheKey);
+ I->setResult(MemDepResult::getUnknown());
Result.push_back(NonLocalDepResult(I->getBB(), I->getResult(),
Pointer.getAddr()));
break;
/// CachedNonLocalPointerInfo, remove it.
void MemoryDependenceAnalysis::
RemoveCachedNonLocalPointerDependencies(ValueIsLoadPair P) {
- CachedNonLocalPointerInfo::iterator It =
+ CachedNonLocalPointerInfo::iterator It =
NonLocalPointerDeps.find(P);
if (It == NonLocalPointerDeps.end()) return;
-
+
// Remove all of the entries in the BB->val map. This involves removing
// instructions from the reverse map.
NonLocalDepInfo &PInfo = It->second.NonLocalDeps;
-
+
for (unsigned i = 0, e = PInfo.size(); i != e; ++i) {
Instruction *Target = PInfo[i].getResult().getInst();
- if (Target == 0) continue; // Ignore non-local dep results.
+ if (!Target) continue; // Ignore non-local dep results.
assert(Target->getParent() == PInfo[i].getBB());
-
+
// Eliminating the dirty entry from 'Cache', so update the reverse info.
RemoveFromReverseMap(ReverseNonLocalPtrDeps, Target, P);
}
-
+
// Remove P from NonLocalPointerDeps (which deletes NonLocalDepInfo).
NonLocalPointerDeps.erase(It);
}
// Remove this local dependency info.
LocalDeps.erase(LocalDepEntry);
}
-
+
// If we have any cached pointer dependencies on this instruction, remove
// them. If the instruction has non-pointer type, then it can't be a pointer
// base.
-
+
// Remove it from both the load info and the store info. The instruction
// can't be in either of these maps if it is non-pointer.
if (RemInst->getType()->isPointerTy()) {
RemoveCachedNonLocalPointerDependencies(ValueIsLoadPair(RemInst, false));
RemoveCachedNonLocalPointerDependencies(ValueIsLoadPair(RemInst, true));
}
-
+
// Loop over all of the things that depend on the instruction we're removing.
- //
+ //
SmallVector<std::pair<Instruction*, Instruction*>, 8> ReverseDepsToAdd;
// If we find RemInst as a clobber or Def in any of the maps for other values,
MemDepResult NewDirtyVal;
if (!RemInst->isTerminator())
NewDirtyVal = MemDepResult::getDirty(++BasicBlock::iterator(RemInst));
-
+
ReverseDepMapType::iterator ReverseDepIt = ReverseLocalDeps.find(RemInst);
if (ReverseDepIt != ReverseLocalDeps.end()) {
SmallPtrSet<Instruction*, 4> &ReverseDeps = ReverseDepIt->second;
// RemInst can't be the terminator if it has local stuff depending on it.
assert(!ReverseDeps.empty() && !isa<TerminatorInst>(RemInst) &&
"Nothing can locally depend on a terminator");
-
+
for (SmallPtrSet<Instruction*, 4>::iterator I = ReverseDeps.begin(),
E = ReverseDeps.end(); I != E; ++I) {
Instruction *InstDependingOnRemInst = *I;
assert(InstDependingOnRemInst != RemInst &&
"Already removed our local dep info");
-
+
LocalDeps[InstDependingOnRemInst] = NewDirtyVal;
-
+
// Make sure to remember that new things depend on NewDepInst.
assert(NewDirtyVal.getInst() && "There is no way something else can have "
"a local dep on this if it is a terminator!");
- ReverseDepsToAdd.push_back(std::make_pair(NewDirtyVal.getInst(),
+ ReverseDepsToAdd.push_back(std::make_pair(NewDirtyVal.getInst(),
InstDependingOnRemInst));
}
-
+
ReverseLocalDeps.erase(ReverseDepIt);
// Add new reverse deps after scanning the set, to avoid invalidating the
ReverseDepsToAdd.pop_back();
}
}
-
+
ReverseDepIt = ReverseNonLocalDeps.find(RemInst);
if (ReverseDepIt != ReverseNonLocalDeps.end()) {
SmallPtrSet<Instruction*, 4> &Set = ReverseDepIt->second;
for (SmallPtrSet<Instruction*, 4>::iterator I = Set.begin(), E = Set.end();
I != E; ++I) {
assert(*I != RemInst && "Already removed NonLocalDep info for RemInst");
-
+
PerInstNLInfo &INLD = NonLocalDeps[*I];
// The information is now dirty!
INLD.second = true;
-
- for (NonLocalDepInfo::iterator DI = INLD.first.begin(),
+
+ for (NonLocalDepInfo::iterator DI = INLD.first.begin(),
DE = INLD.first.end(); DI != DE; ++DI) {
if (DI->getResult().getInst() != RemInst) continue;
-
+
// Convert to a dirty entry for the subsequent instruction.
DI->setResult(NewDirtyVal);
-
+
if (Instruction *NextI = NewDirtyVal.getInst())
ReverseDepsToAdd.push_back(std::make_pair(NextI, *I));
}
ReverseDepsToAdd.pop_back();
}
}
-
+
// If the instruction is in ReverseNonLocalPtrDeps then it appears as a
// value in the NonLocalPointerDeps info.
ReverseNonLocalPtrDepTy::iterator ReversePtrDepIt =
if (ReversePtrDepIt != ReverseNonLocalPtrDeps.end()) {
SmallPtrSet<ValueIsLoadPair, 4> &Set = ReversePtrDepIt->second;
SmallVector<std::pair<Instruction*, ValueIsLoadPair>,8> ReversePtrDepsToAdd;
-
+
for (SmallPtrSet<ValueIsLoadPair, 4>::iterator I = Set.begin(),
E = Set.end(); I != E; ++I) {
ValueIsLoadPair P = *I;
assert(P.getPointer() != RemInst &&
"Already removed NonLocalPointerDeps info for RemInst");
-
+
NonLocalDepInfo &NLPDI = NonLocalPointerDeps[P].NonLocalDeps;
-
+
// The cache is not valid for any specific block anymore.
NonLocalPointerDeps[P].Pair = BBSkipFirstBlockPair();
-
+
// Update any entries for RemInst to use the instruction after it.
for (NonLocalDepInfo::iterator DI = NLPDI.begin(), DE = NLPDI.end();
DI != DE; ++DI) {
if (DI->getResult().getInst() != RemInst) continue;
-
+
// Convert to a dirty entry for the subsequent instruction.
DI->setResult(NewDirtyVal);
-
+
if (Instruction *NewDirtyInst = NewDirtyVal.getInst())
ReversePtrDepsToAdd.push_back(std::make_pair(NewDirtyInst, P));
}
-
+
// Re-sort the NonLocalDepInfo. Changing the dirty entry to its
// subsequent value may invalidate the sortedness.
std::sort(NLPDI.begin(), NLPDI.end());
}
-
+
ReverseNonLocalPtrDeps.erase(ReversePtrDepIt);
-
+
while (!ReversePtrDepsToAdd.empty()) {
ReverseNonLocalPtrDeps[ReversePtrDepsToAdd.back().first]
.insert(ReversePtrDepsToAdd.back().second);
ReversePtrDepsToAdd.pop_back();
}
}
-
-
+
+
assert(!NonLocalDeps.count(RemInst) && "RemInst got reinserted?");
AA->deleteValue(RemInst);
DEBUG(verifyRemoved(RemInst));
assert(I->second.getInst() != D &&
"Inst occurs in data structures");
}
-
+
for (CachedNonLocalPointerInfo::const_iterator I =NonLocalPointerDeps.begin(),
E = NonLocalPointerDeps.end(); I != E; ++I) {
assert(I->first.getPointer() != D && "Inst occurs in NLPD map key");
II != E; ++II)
assert(II->getResult().getInst() != D && "Inst occurs as NLPD value");
}
-
+
for (NonLocalDepMapType::const_iterator I = NonLocalDeps.begin(),
E = NonLocalDeps.end(); I != E; ++I) {
assert(I->first != D && "Inst occurs in data structures");
EE = INLD.first.end(); II != EE; ++II)
assert(II->getResult().getInst() != D && "Inst occurs in data structures");
}
-
+
for (ReverseDepMapType::const_iterator I = ReverseLocalDeps.begin(),
E = ReverseLocalDeps.end(); I != E; ++I) {
assert(I->first != D && "Inst occurs in data structures");
EE = I->second.end(); II != EE; ++II)
assert(*II != D && "Inst occurs in data structures");
}
-
+
for (ReverseDepMapType::const_iterator I = ReverseNonLocalDeps.begin(),
E = ReverseNonLocalDeps.end();
I != E; ++I) {
EE = I->second.end(); II != EE; ++II)
assert(*II != D && "Inst occurs in data structures");
}
-
+
for (ReverseNonLocalPtrDepTy::const_iterator
I = ReverseNonLocalPtrDeps.begin(),
E = ReverseNonLocalPtrDeps.end(); I != E; ++I) {
assert(I->first != D && "Inst occurs in rev NLPD map");
-
+
for (SmallPtrSet<ValueIsLoadPair, 4>::const_iterator II = I->second.begin(),
E = I->second.end(); II != E; ++II)
assert(*II != ValueIsLoadPair(D, false) &&
*II != ValueIsLoadPair(D, true) &&
"Inst occurs in ReverseNonLocalPtrDeps map");
}
-
+
}