//
// This file promotes memory references to be register references. It promotes
// alloca instructions which only have loads and stores as uses. An alloca is
-// transformed by using dominator frontiers to place PHI nodes, then traversing
-// the function in depth-first order to rewrite loads and stores as appropriate.
-// This is just the standard SSA construction algorithm to construct "pruned"
-// SSA form.
+// transformed by using iterated dominator frontiers to place PHI nodes, then
+// traversing the function in depth-first order to rewrite loads and stores as
+// appropriate.
+//
+// The algorithm used here is based on:
+//
+// Sreedhar and Gao. A linear time algorithm for placing phi-nodes.
+// In Proceedings of the 22nd ACM SIGPLAN-SIGACT Symposium on Principles of
+// Programming Languages
+// POPL '95. ACM, New York, NY, 62-73.
+//
+// It has been modified to not explicitly use the DJ graph data structure and to
+// directly compute pruned SSA using per-variable liveness information.
//
//===----------------------------------------------------------------------===//
#define DEBUG_TYPE "mem2reg"
#include "llvm/Transforms/Utils/PromoteMemToReg.h"
-#include "llvm/Constants.h"
-#include "llvm/DerivedTypes.h"
-#include "llvm/Function.h"
-#include "llvm/Instructions.h"
-#include "llvm/IntrinsicInst.h"
-#include "llvm/LLVMContext.h"
-#include "llvm/Analysis/Dominators.h"
-#include "llvm/Analysis/AliasSetTracker.h"
#include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/Hashing.h"
+#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/Statistic.h"
-#include "llvm/ADT/STLExtras.h"
+#include "llvm/Analysis/AliasSetTracker.h"
+#include "llvm/Analysis/Dominators.h"
+#include "llvm/Analysis/InstructionSimplify.h"
+#include "llvm/Analysis/ValueTracking.h"
+#include "llvm/DIBuilder.h"
+#include "llvm/DebugInfo.h"
+#include "llvm/IR/Constants.h"
+#include "llvm/IR/DerivedTypes.h"
+#include "llvm/IR/Function.h"
+#include "llvm/IR/Instructions.h"
+#include "llvm/IR/IntrinsicInst.h"
+#include "llvm/IR/Metadata.h"
#include "llvm/Support/CFG.h"
-#include "llvm/Support/Compiler.h"
+#include "llvm/Transforms/Utils/Local.h"
#include <algorithm>
+#include <queue>
using namespace llvm;
STATISTIC(NumLocalPromoted, "Number of alloca's promoted within one block");
STATISTIC(NumDeadAlloca, "Number of dead alloca's removed");
STATISTIC(NumPHIInsert, "Number of PHI nodes inserted");
-// Provide DenseMapInfo for all pointers.
namespace llvm {
template<>
struct DenseMapInfo<std::pair<BasicBlock*, unsigned> > {
return EltTy(reinterpret_cast<BasicBlock*>(-2), 0U);
}
static unsigned getHashValue(const std::pair<BasicBlock*, unsigned> &Val) {
- return DenseMapInfo<void*>::getHashValue(Val.first) + Val.second*2;
+ using llvm::hash_value;
+ return static_cast<unsigned>(hash_value(Val));
}
static bool isEqual(const EltTy &LHS, const EltTy &RHS) {
return LHS == RHS;
}
- static bool isPod() { return true; }
};
}
// assignments to subsections of the memory unit.
// Only allow direct and non-volatile loads and stores...
- for (Value::use_const_iterator UI = AI->use_begin(), UE = AI->use_end();
- UI != UE; ++UI) // Loop over all of the uses of the alloca
- if (const LoadInst *LI = dyn_cast<LoadInst>(*UI)) {
+ for (Value::const_use_iterator UI = AI->use_begin(), UE = AI->use_end();
+ UI != UE; ++UI) { // Loop over all of the uses of the alloca
+ const User *U = *UI;
+ if (const LoadInst *LI = dyn_cast<LoadInst>(U)) {
+ // Note that atomic loads can be transformed; atomic semantics do
+ // not have any meaning for a local alloca.
if (LI->isVolatile())
return false;
- } else if (const StoreInst *SI = dyn_cast<StoreInst>(*UI)) {
+ } else if (const StoreInst *SI = dyn_cast<StoreInst>(U)) {
if (SI->getOperand(0) == AI)
return false; // Don't allow a store OF the AI, only INTO the AI.
+ // Note that atomic stores can be transformed; atomic semantics do
+ // not have any meaning for a local alloca.
if (SI->isVolatile())
return false;
- } else if (const BitCastInst *BC = dyn_cast<BitCastInst>(*UI)) {
- // A bitcast that does not feed into debug info inhibits promotion.
- if (!BC->hasOneUse() || !isa<DbgInfoIntrinsic>(*BC->use_begin()))
+ } else if (const IntrinsicInst *II = dyn_cast<IntrinsicInst>(U)) {
+ if (II->getIntrinsicID() != Intrinsic::lifetime_start &&
+ II->getIntrinsicID() != Intrinsic::lifetime_end)
+ return false;
+ } else if (const BitCastInst *BCI = dyn_cast<BitCastInst>(U)) {
+ if (BCI->getType() != Type::getInt8PtrTy(U->getContext()))
+ return false;
+ if (!onlyUsedByLifetimeMarkers(BCI))
return false;
- // If the only use is by debug info, this alloca will not exist in
- // non-debug code, so don't try to promote; this ensures the same
- // codegen with debug info. Otherwise, debug info should not
- // inhibit promotion (but we must examine other uses).
- if (AI->hasOneUse())
+ } else if (const GetElementPtrInst *GEPI = dyn_cast<GetElementPtrInst>(U)) {
+ if (GEPI->getType() != Type::getInt8PtrTy(U->getContext()))
+ return false;
+ if (!GEPI->hasAllZeroIndices())
+ return false;
+ if (!onlyUsedByLifetimeMarkers(GEPI))
return false;
} else {
return false;
}
+ }
return true;
}
struct AllocaInfo;
// Data package used by RenamePass()
- class VISIBILITY_HIDDEN RenamePassData {
+ class RenamePassData {
public:
typedef std::vector<Value *> ValVector;
- RenamePassData() {}
+ RenamePassData() : BB(NULL), Pred(NULL), Values() {}
RenamePassData(BasicBlock *B, BasicBlock *P,
const ValVector &V) : BB(B), Pred(P), Values(V) {}
BasicBlock *BB;
///
/// This functionality is important because it avoids scanning large basic
/// blocks multiple times when promoting many allocas in the same block.
- class VISIBILITY_HIDDEN LargeBlockInfo {
+ class LargeBlockInfo {
/// InstNumbers - For each instruction that we track, keep the index of the
/// instruction. The index starts out as the number of the instruction from
/// the start of the block.
}
};
- struct VISIBILITY_HIDDEN PromoteMem2Reg {
+ struct PromoteMem2Reg {
/// Allocas - The alloca instructions being promoted.
///
std::vector<AllocaInst*> Allocas;
DominatorTree &DT;
- DominanceFrontier &DF;
+ DIBuilder *DIB;
/// AST - An AliasSetTracker object to update. If null, don't update it.
///
AliasSetTracker *AST;
- LLVMContext &Context;
-
/// AllocaLookup - Reverse mapping of Allocas.
///
-
std::map<AllocaInst*, unsigned> AllocaLookup;
+
DenseMap<AllocaInst*, unsigned> AllocaLookup;
- /// NewPhiNodes - The PhiNodes we're adding.
+ /// NewPhiNodes - The PhiNodes we're adding. That map is used to simplify
+ /// some Phi nodes as we iterate over it, so it should have deterministic
+ /// iterators. We could use a MapVector, but since we already maintain a
+ /// map from BasicBlock* to a stable numbering (BBNumbers), the DenseMap is
+ /// more efficient (also supports removal).
///
- DenseMap<std::pair<BasicBlock*, unsigned>, PHINode*> NewPhiNodes;
+ DenseMap<std::pair<unsigned, unsigned>, PHINode*> NewPhiNodes;
/// PhiToAllocaMap - For each PHI node, keep track of which entry in Allocas
/// it corresponds to.
///
std::vector<Value*> PointerAllocaValues;
+ /// AllocaDbgDeclares - For each alloca, we keep track of the dbg.declare
+ /// intrinsic that describes it, if any, so that we can convert it to a
+ /// dbg.value intrinsic if the alloca gets promoted.
+ SmallVector<DbgDeclareInst*, 8> AllocaDbgDeclares;
+
/// Visited - The set of basic blocks the renamer has already visited.
///
SmallPtrSet<BasicBlock*, 16> Visited;
/// non-determinstic behavior.
DenseMap<BasicBlock*, unsigned> BBNumbers;
+ /// DomLevels - Maps DomTreeNodes to their level in the dominator tree.
+ DenseMap<DomTreeNode*, unsigned> DomLevels;
+
/// BBNumPreds - Lazily compute the number of predecessors a block has.
DenseMap<const BasicBlock*, unsigned> BBNumPreds;
public:
PromoteMem2Reg(const std::vector<AllocaInst*> &A, DominatorTree &dt,
- DominanceFrontier &df, AliasSetTracker *ast,
- LLVMContext &C)
- : Allocas(A), DT(dt), DF(df), AST(ast), Context(C) {}
+ AliasSetTracker *ast)
+ : Allocas(A), DT(dt), DIB(0), AST(ast) {}
+ ~PromoteMem2Reg() {
+ delete DIB;
+ }
void run();
- /// properlyDominates - Return true if I1 properly dominates I2.
- ///
- bool properlyDominates(Instruction *I1, Instruction *I2) const {
- if (InvokeInst *II = dyn_cast<InvokeInst>(I1))
- I1 = II->getNormalDest()->begin();
- return DT.properlyDominates(I1->getParent(), I2->getParent());
- }
-
/// dominates - Return true if BB1 dominates BB2 using the DominatorTree.
///
bool dominates(BasicBlock *BB1, BasicBlock *BB2) const {
LargeBlockInfo &LBI);
void PromoteSingleBlockAlloca(AllocaInst *AI, AllocaInfo &Info,
LargeBlockInfo &LBI);
-
void RenamePass(BasicBlock *BB, BasicBlock *Pred,
RenamePassData::ValVector &IncVals,
std::vector<RenamePassData> &Worklist);
- bool QueuePhiNode(BasicBlock *BB, unsigned AllocaIdx, unsigned &Version,
- SmallPtrSet<PHINode*, 16> &InsertedPHINodes);
+ bool QueuePhiNode(BasicBlock *BB, unsigned AllocaIdx, unsigned &Version);
};
struct AllocaInfo {
- std::vector<BasicBlock*> DefiningBlocks;
- std::vector<BasicBlock*> UsingBlocks;
+ SmallVector<BasicBlock*, 32> DefiningBlocks;
+ SmallVector<BasicBlock*, 32> UsingBlocks;
StoreInst *OnlyStore;
BasicBlock *OnlyBlock;
bool OnlyUsedInOneBlock;
Value *AllocaPointerVal;
+ DbgDeclareInst *DbgDeclare;
void clear() {
DefiningBlocks.clear();
OnlyBlock = 0;
OnlyUsedInOneBlock = true;
AllocaPointerVal = 0;
+ DbgDeclare = 0;
}
/// AnalyzeAlloca - Scan the uses of the specified alloca, filling in our
// As we scan the uses of the alloca instruction, keep track of stores,
// and decide whether all of the loads and stores to the alloca are within
// the same basic block.
- for (Value::use_iterator U = AI->use_begin(), E = AI->use_end();
- U != E;) {
- Instruction *User = cast<Instruction>(*U);
- ++U;
- if (BitCastInst *BC = dyn_cast<BitCastInst>(User)) {
- // Remove any uses of this alloca in DbgInfoInstrinsics.
- assert(BC->hasOneUse() && "Unexpected alloca uses!");
- DbgInfoIntrinsic *DI = cast<DbgInfoIntrinsic>(*BC->use_begin());
- DI->eraseFromParent();
- BC->eraseFromParent();
- continue;
- }
- else if (StoreInst *SI = dyn_cast<StoreInst>(User)) {
+ for (Value::use_iterator UI = AI->use_begin(), E = AI->use_end();
+ UI != E;) {
+ Instruction *User = cast<Instruction>(*UI++);
+
+ if (StoreInst *SI = dyn_cast<StoreInst>(User)) {
// Remember the basic blocks which define new values for the alloca
DefiningBlocks.push_back(SI->getParent());
AllocaPointerVal = SI->getOperand(0);
OnlyUsedInOneBlock = false;
}
}
+
+ DbgDeclare = FindAllocaDbgDeclare(AI);
+ }
+ };
+
+ typedef std::pair<DomTreeNode*, unsigned> DomTreeNodePair;
+
+ struct DomTreeNodeCompare {
+ bool operator()(const DomTreeNodePair &LHS, const DomTreeNodePair &RHS) {
+ return LHS.second < RHS.second;
}
};
} // end of anonymous namespace
+static void removeLifetimeIntrinsicUsers(AllocaInst *AI) {
+ // Knowing that this alloca is promotable, we know that it's safe to kill all
+ // instructions except for load and store.
+
+ for (Value::use_iterator UI = AI->use_begin(), UE = AI->use_end();
+ UI != UE;) {
+ Instruction *I = cast<Instruction>(*UI);
+ ++UI;
+ if (isa<LoadInst>(I) || isa<StoreInst>(I))
+ continue;
+
+ if (!I->getType()->isVoidTy()) {
+ // The only users of this bitcast/GEP instruction are lifetime intrinsics.
+ // Follow the use/def chain to erase them now instead of leaving it for
+ // dead code elimination later.
+ for (Value::use_iterator UI = I->use_begin(), UE = I->use_end();
+ UI != UE;) {
+ Instruction *Inst = cast<Instruction>(*UI);
+ ++UI;
+ Inst->eraseFromParent();
+ }
+ }
+ I->eraseFromParent();
+ }
+}
void PromoteMem2Reg::run() {
- Function &F = *DF.getRoot()->getParent();
+ Function &F = *DT.getRoot()->getParent();
if (AST) PointerAllocaValues.resize(Allocas.size());
+ AllocaDbgDeclares.resize(Allocas.size());
AllocaInfo Info;
LargeBlockInfo LBI;
assert(AI->getParent()->getParent() == &F &&
"All allocas should be in the same function, which is same as DF!");
+ removeLifetimeIntrinsicUsers(AI);
+
if (AI->use_empty()) {
// If there are no uses of the alloca, just delete it now.
if (AST) AST->deleteValue(AI);
// Finally, after the scan, check to see if the store is all that is left.
if (Info.UsingBlocks.empty()) {
+ // Record debuginfo for the store and remove the declaration's
+ // debuginfo.
+ if (DbgDeclareInst *DDI = Info.DbgDeclare) {
+ if (!DIB)
+ DIB = new DIBuilder(*DDI->getParent()->getParent()->getParent());
+ ConvertDebugDeclareToDebugValue(DDI, Info.OnlyStore, *DIB);
+ DDI->eraseFromParent();
+ }
// Remove the (now dead) store and alloca.
Info.OnlyStore->eraseFromParent();
LBI.deleteValue(Info.OnlyStore);
// Remove the (now dead) stores and alloca.
while (!AI->use_empty()) {
StoreInst *SI = cast<StoreInst>(AI->use_back());
+ // Record debuginfo for the store before removing it.
+ if (DbgDeclareInst *DDI = Info.DbgDeclare) {
+ if (!DIB)
+ DIB = new DIBuilder(*SI->getParent()->getParent()->getParent());
+ ConvertDebugDeclareToDebugValue(DDI, SI, *DIB);
+ }
SI->eraseFromParent();
LBI.deleteValue(SI);
}
// The alloca has been processed, move on.
RemoveFromAllocasList(AllocaNum);
+ // The alloca's debuginfo can be removed as well.
+ if (DbgDeclareInst *DDI = Info.DbgDeclare)
+ DDI->eraseFromParent();
+
++NumLocalPromoted;
continue;
}
}
-
+
+ // If we haven't computed dominator tree levels, do so now.
+ if (DomLevels.empty()) {
+ SmallVector<DomTreeNode*, 32> Worklist;
+
+ DomTreeNode *Root = DT.getRootNode();
+ DomLevels[Root] = 0;
+ Worklist.push_back(Root);
+
+ while (!Worklist.empty()) {
+ DomTreeNode *Node = Worklist.pop_back_val();
+ unsigned ChildLevel = DomLevels[Node] + 1;
+ for (DomTreeNode::iterator CI = Node->begin(), CE = Node->end();
+ CI != CE; ++CI) {
+ DomLevels[*CI] = ChildLevel;
+ Worklist.push_back(*CI);
+ }
+ }
+ }
+
// If we haven't computed a numbering for the BB's in the function, do so
// now.
if (BBNumbers.empty()) {
// stored into the alloca.
if (AST)
PointerAllocaValues[AllocaNum] = Info.AllocaPointerVal;
+
+ // Remember the dbg.declare intrinsic describing this alloca, if any.
+ if (Info.DbgDeclare) AllocaDbgDeclares[AllocaNum] = Info.DbgDeclare;
// Keep the reverse mapping of the 'Allocas' array for the rename pass.
AllocaLookup[Allocas[AllocaNum]] = AllocaNum;
//
std::vector<RenamePassData> RenamePassWorkList;
RenamePassWorkList.push_back(RenamePassData(F.begin(), 0, Values));
- while (!RenamePassWorkList.empty()) {
+ do {
RenamePassData RPD;
RPD.swap(RenamePassWorkList.back());
RenamePassWorkList.pop_back();
// RenamePass may add new worklist entries.
RenamePass(RPD.BB, RPD.Pred, RPD.Values, RenamePassWorkList);
- }
+ } while (!RenamePassWorkList.empty());
// The renamer uses the Visited set to avoid infinite loops. Clear it now.
Visited.clear();
Instruction *A = Allocas[i];
// If there are any uses of the alloca instructions left, they must be in
- // sections of dead code that were not processed on the dominance frontier.
- // Just delete the users now.
- //
+ // unreachable basic blocks that were not processed by walking the dominator
+ // tree. Just delete the users now.
if (!A->use_empty())
A->replaceAllUsesWith(UndefValue::get(A->getType()));
if (AST) AST->deleteValue(A);
A->eraseFromParent();
}
-
+ // Remove alloca's dbg.declare instrinsics from the function.
+ for (unsigned i = 0, e = AllocaDbgDeclares.size(); i != e; ++i)
+ if (DbgDeclareInst *DDI = AllocaDbgDeclares[i])
+ DDI->eraseFromParent();
+
// Loop over all of the PHI nodes and see if there are any that we can get
// rid of because they merge all of the same incoming values. This can
// happen due to undef values coming into the PHI nodes. This process is
while (EliminatedAPHI) {
EliminatedAPHI = false;
- for (DenseMap<std::pair<BasicBlock*, unsigned>, PHINode*>::iterator I =
+ // Iterating over NewPhiNodes is deterministic, so it is safe to try to
+ // simplify and RAUW them as we go. If it was not, we could add uses to
+ // the values we replace with in a non deterministic order, thus creating
+ // non deterministic def->use chains.
+ for (DenseMap<std::pair<unsigned, unsigned>, PHINode*>::iterator I =
NewPhiNodes.begin(), E = NewPhiNodes.end(); I != E;) {
PHINode *PN = I->second;
-
+
// If this PHI node merges one value and/or undefs, get the value.
- if (Value *V = PN->hasConstantValue(true)) {
- if (!isa<Instruction>(V) ||
- properlyDominates(cast<Instruction>(V), PN)) {
- if (AST && isa<PointerType>(PN->getType()))
- AST->deleteValue(PN);
- PN->replaceAllUsesWith(V);
- PN->eraseFromParent();
- NewPhiNodes.erase(I++);
- EliminatedAPHI = true;
- continue;
- }
+ if (Value *V = SimplifyInstruction(PN, 0, 0, &DT)) {
+ if (AST && PN->getType()->isPointerTy())
+ AST->deleteValue(PN);
+ PN->replaceAllUsesWith(V);
+ PN->eraseFromParent();
+ NewPhiNodes.erase(I++);
+ EliminatedAPHI = true;
+ continue;
}
++I;
}
// have incoming values for all predecessors. Loop over all PHI nodes we have
// created, inserting undef values if they are missing any incoming values.
//
- for (DenseMap<std::pair<BasicBlock*, unsigned>, PHINode*>::iterator I =
+ for (DenseMap<std::pair<unsigned, unsigned>, PHINode*>::iterator I =
NewPhiNodes.begin(), E = NewPhiNodes.end(); I != E; ++I) {
// We want to do this once per basic block. As such, only process a block
// when we find the PHI that is the first entry in the block.
// To determine liveness, we must iterate through the predecessors of blocks
// where the def is live. Blocks are added to the worklist if we need to
// check their predecessors. Start with all the using blocks.
- SmallVector<BasicBlock*, 64> LiveInBlockWorklist;
- LiveInBlockWorklist.insert(LiveInBlockWorklist.end(),
- Info.UsingBlocks.begin(), Info.UsingBlocks.end());
+ SmallVector<BasicBlock*, 64> LiveInBlockWorklist(Info.UsingBlocks.begin(),
+ Info.UsingBlocks.end());
// If any of the using blocks is also a definition block, check to see if the
// definition occurs before or after the use. If it happens before the use,
LiveInBlockWorklist.pop_back();
--i, --e;
break;
- } else if (LoadInst *LI = dyn_cast<LoadInst>(I)) {
+ }
+
+ if (LoadInst *LI = dyn_cast<LoadInst>(I)) {
if (LI->getOperand(0) != AI) continue;
// Okay, we found a load before a store to the alloca. It is actually
/// avoiding insertion of dead phi nodes.
void PromoteMem2Reg::DetermineInsertionPoint(AllocaInst *AI, unsigned AllocaNum,
AllocaInfo &Info) {
-
// Unique the set of defining blocks for efficient lookup.
SmallPtrSet<BasicBlock*, 32> DefBlocks;
DefBlocks.insert(Info.DefiningBlocks.begin(), Info.DefiningBlocks.end());
SmallPtrSet<BasicBlock*, 32> LiveInBlocks;
ComputeLiveInBlocks(AI, Info, DefBlocks, LiveInBlocks);
- // Compute the locations where PhiNodes need to be inserted. Look at the
- // dominance frontier of EACH basic-block we have a write in.
- unsigned CurrentVersion = 0;
- SmallPtrSet<PHINode*, 16> InsertedPHINodes;
- std::vector<std::pair<unsigned, BasicBlock*> > DFBlocks;
- while (!Info.DefiningBlocks.empty()) {
- BasicBlock *BB = Info.DefiningBlocks.back();
- Info.DefiningBlocks.pop_back();
-
- // Look up the DF for this write, add it to defining blocks.
- DominanceFrontier::const_iterator it = DF.find(BB);
- if (it == DF.end()) continue;
-
- const DominanceFrontier::DomSetType &S = it->second;
-
- // In theory we don't need the indirection through the DFBlocks vector.
- // In practice, the order of calling QueuePhiNode would depend on the
- // (unspecified) ordering of basic blocks in the dominance frontier,
- // which would give PHI nodes non-determinstic subscripts. Fix this by
- // processing blocks in order of the occurance in the function.
- for (DominanceFrontier::DomSetType::const_iterator P = S.begin(),
- PE = S.end(); P != PE; ++P) {
- // If the frontier block is not in the live-in set for the alloca, don't
- // bother processing it.
- if (!LiveInBlocks.count(*P))
- continue;
-
- DFBlocks.push_back(std::make_pair(BBNumbers[*P], *P));
- }
-
- // Sort by which the block ordering in the function.
- if (DFBlocks.size() > 1)
- std::sort(DFBlocks.begin(), DFBlocks.end());
-
- for (unsigned i = 0, e = DFBlocks.size(); i != e; ++i) {
- BasicBlock *BB = DFBlocks[i].second;
- if (QueuePhiNode(BB, AllocaNum, CurrentVersion, InsertedPHINodes))
- Info.DefiningBlocks.push_back(BB);
+ // Use a priority queue keyed on dominator tree level so that inserted nodes
+ // are handled from the bottom of the dominator tree upwards.
+ typedef std::priority_queue<DomTreeNodePair, SmallVector<DomTreeNodePair, 32>,
+ DomTreeNodeCompare> IDFPriorityQueue;
+ IDFPriorityQueue PQ;
+
+ for (SmallPtrSet<BasicBlock*, 32>::const_iterator I = DefBlocks.begin(),
+ E = DefBlocks.end(); I != E; ++I) {
+ if (DomTreeNode *Node = DT.getNode(*I))
+ PQ.push(std::make_pair(Node, DomLevels[Node]));
+ }
+
+ SmallVector<std::pair<unsigned, BasicBlock*>, 32> DFBlocks;
+ SmallPtrSet<DomTreeNode*, 32> Visited;
+ SmallVector<DomTreeNode*, 32> Worklist;
+ while (!PQ.empty()) {
+ DomTreeNodePair RootPair = PQ.top();
+ PQ.pop();
+ DomTreeNode *Root = RootPair.first;
+ unsigned RootLevel = RootPair.second;
+
+ // Walk all dominator tree children of Root, inspecting their CFG edges with
+ // targets elsewhere on the dominator tree. Only targets whose level is at
+ // most Root's level are added to the iterated dominance frontier of the
+ // definition set.
+
+ Worklist.clear();
+ Worklist.push_back(Root);
+
+ while (!Worklist.empty()) {
+ DomTreeNode *Node = Worklist.pop_back_val();
+ BasicBlock *BB = Node->getBlock();
+
+ for (succ_iterator SI = succ_begin(BB), SE = succ_end(BB); SI != SE;
+ ++SI) {
+ DomTreeNode *SuccNode = DT.getNode(*SI);
+
+ // Quickly skip all CFG edges that are also dominator tree edges instead
+ // of catching them below.
+ if (SuccNode->getIDom() == Node)
+ continue;
+
+ unsigned SuccLevel = DomLevels[SuccNode];
+ if (SuccLevel > RootLevel)
+ continue;
+
+ if (!Visited.insert(SuccNode))
+ continue;
+
+ BasicBlock *SuccBB = SuccNode->getBlock();
+ if (!LiveInBlocks.count(SuccBB))
+ continue;
+
+ DFBlocks.push_back(std::make_pair(BBNumbers[SuccBB], SuccBB));
+ if (!DefBlocks.count(SuccBB))
+ PQ.push(std::make_pair(SuccNode, SuccLevel));
+ }
+
+ for (DomTreeNode::iterator CI = Node->begin(), CE = Node->end(); CI != CE;
+ ++CI) {
+ if (!Visited.count(*CI))
+ Worklist.push_back(*CI);
+ }
}
- DFBlocks.clear();
}
+
+ if (DFBlocks.size() > 1)
+ std::sort(DFBlocks.begin(), DFBlocks.end());
+
+ unsigned CurrentVersion = 0;
+ for (unsigned i = 0, e = DFBlocks.size(); i != e; ++i)
+ QueuePhiNode(DFBlocks[i].second, AllocaNum, CurrentVersion);
}
/// RewriteSingleStoreAlloca - If there is only a single store to this value,
}
// Otherwise, we *can* safely rewrite this load.
- LI->replaceAllUsesWith(OnlyStore->getOperand(0));
- if (AST && isa<PointerType>(LI->getType()))
+ Value *ReplVal = OnlyStore->getOperand(0);
+ // If the replacement value is the load, this must occur in unreachable
+ // code.
+ if (ReplVal == LI)
+ ReplVal = UndefValue::get(LI->getType());
+ LI->replaceAllUsesWith(ReplVal);
+ if (AST && LI->getType()->isPointerTy())
AST->deleteValue(LI);
LI->eraseFromParent();
LBI.deleteValue(LI);
for (Value::use_iterator UI = AI->use_begin(), E = AI->use_end(); UI != E;)
if (LoadInst *LI = dyn_cast<LoadInst>(*UI++)) {
LI->replaceAllUsesWith(UndefValue::get(LI->getType()));
- if (AST && isa<PointerType>(LI->getType()))
+ if (AST && LI->getType()->isPointerTy())
AST->deleteValue(LI);
LBI.deleteValue(LI);
LI->eraseFromParent();
// Find the nearest store that has a lower than this load.
StoresByIndexTy::iterator I =
std::lower_bound(StoresByIndex.begin(), StoresByIndex.end(),
- std::pair<unsigned, StoreInst*>(LoadIdx, 0),
+ std::pair<unsigned, StoreInst*>(LoadIdx, static_cast<StoreInst*>(0)),
StoreIndexSearchPredicate());
// If there is no store before this load, then we can't promote this load.
// Otherwise, there was a store before this load, the load takes its value.
--I;
LI->replaceAllUsesWith(I->second->getOperand(0));
- if (AST && isa<PointerType>(LI->getType()))
+ if (AST && LI->getType()->isPointerTy())
AST->deleteValue(LI);
LI->eraseFromParent();
LBI.deleteValue(LI);
}
}
-
// QueuePhiNode - queues a phi-node to be added to a basic-block for a specific
// Alloca returns true if there wasn't already a phi-node for that variable
//
bool PromoteMem2Reg::QueuePhiNode(BasicBlock *BB, unsigned AllocaNo,
- unsigned &Version,
- SmallPtrSet<PHINode*, 16> &InsertedPHINodes) {
+ unsigned &Version) {
// Look up the basic-block in question.
- PHINode *&PN = NewPhiNodes[std::make_pair(BB, AllocaNo)];
+ PHINode *&PN = NewPhiNodes[std::make_pair(BBNumbers[BB], AllocaNo)];
// If the BB already has a phi node added for the i'th alloca then we're done!
if (PN) return false;
// Create a PhiNode using the dereferenced type... and add the phi-node to the
// BasicBlock.
- PN = PHINode::Create(Allocas[AllocaNo]->getAllocatedType(),
+ PN = PHINode::Create(Allocas[AllocaNo]->getAllocatedType(), getNumPreds(BB),
Allocas[AllocaNo]->getName() + "." + Twine(Version++),
BB->begin());
++NumPHIInsert;
PhiToAllocaMap[PN] = AllocaNo;
- PN->reserveOperandSpace(getNumPreds(BB));
-
- InsertedPHINodes.insert(PN);
- if (AST && isa<PointerType>(PN->getType()))
+ if (AST && PN->getType()->isPointerTy())
AST->copyValue(PointerAllocaValues[AllocaNo], PN);
return true;
AllocaInst *Src = dyn_cast<AllocaInst>(LI->getPointerOperand());
if (!Src) continue;
-
std::map<AllocaInst*, unsigned>::iterator AI = AllocaLookup.find(Src);
+
DenseMap<AllocaInst*, unsigned>::iterator AI = AllocaLookup.find(Src);
if (AI == AllocaLookup.end()) continue;
Value *V = IncomingVals[AI->second];
// Anything using the load now uses the current value.
LI->replaceAllUsesWith(V);
- if (AST && isa<PointerType>(LI->getType()))
+ if (AST && LI->getType()->isPointerTy())
AST->deleteValue(LI);
BB->getInstList().erase(LI);
} else if (StoreInst *SI = dyn_cast<StoreInst>(I)) {
AllocaInst *Dest = dyn_cast<AllocaInst>(SI->getPointerOperand());
if (!Dest) continue;
-
std::map<AllocaInst *, unsigned>::iterator ai = AllocaLookup.find(Dest);
+
DenseMap<AllocaInst *, unsigned>::iterator ai = AllocaLookup.find(Dest);
if (ai == AllocaLookup.end())
continue;
// what value were we writing?
IncomingVals[ai->second] = SI->getOperand(0);
+ // Record debuginfo for the store before removing it.
+ if (DbgDeclareInst *DDI = AllocaDbgDeclares[ai->second]) {
+ if (!DIB)
+ DIB = new DIBuilder(*SI->getParent()->getParent()->getParent());
+ ConvertDebugDeclareToDebugValue(DDI, SI, *DIB);
+ }
BB->getInstList().erase(SI);
}
}
}
/// PromoteMemToReg - Promote the specified list of alloca instructions into
-/// scalar registers, inserting PHI nodes as appropriate. This function makes
-/// use of DominanceFrontier information. This function does not modify the CFG
-/// of the function at all. All allocas must be from the same function.
+/// scalar registers, inserting PHI nodes as appropriate. This function does
+/// not modify the CFG of the function at all. All allocas must be from the
+/// same function.
///
/// If AST is specified, the specified tracker is updated to reflect changes
/// made to the IR.
///
void llvm::PromoteMemToReg(const std::vector<AllocaInst*> &Allocas,
- DominatorTree &DT, DominanceFrontier &DF,
- LLVMContext &Context, AliasSetTracker *AST) {
+ DominatorTree &DT, AliasSetTracker *AST) {
// If there is nothing to do, bail out...
if (Allocas.empty()) return;
- PromoteMem2Reg(Allocas, DT, DF, AST, Context).run();
+ PromoteMem2Reg(Allocas, DT, AST).run();
}
projects / oota-llvm.git / blobdiff