[Polly] Identify and hoist definitively invariant loads

[jdoerfert]

Summary

As a first step in the direction of assumed invariant loads (loads that are
not written in some context) we now detect and hoist definitively invariant
loads. Invariant loads will be separated from their statements and preloaded
in the code generation. The preloaded value is then used in the new SCoP. If
the load is only conditionally executed the preloaded version will also only
be executed under the same condition, hence we will never access memory that
wouldn't have been accessed otherwise. This is also the most distinguishing
feature compared to licm.

[jdoerfert]

Some comments to indicate things I will address.

include/polly/ScopInfo.h

875

replace "the whole SCoP." by "this statement."

1151

Will add something here.

1156

This should be only in the next commit, the same for the "required ones" part in the brief desc.

lib/Analysis/ScopInfo.cpp

2297

This is only needed for the next patch. D13195

lib/CodeGen/BlockGenerators.cpp

1

My bad.

lib/CodeGen/IslNodeBuilder.cpp

822

left over.

[Meinersbur]

What's the motivational use case that cannot be handled by LICM?

Could you create a separate pass that modifies the ScopInfo, i.e. lib/Transform/HoistInvariantLoads.cpp

Michael

include/polly/CodeGen/IslNodeBuilder.h

209

Grammar? I don't understand this sentence

lib/Analysis/ScopInfo.cpp

1299

list::remove_if

2285

Could you refactor removing a ScopStmt from a Scop into another function so it might be used by other code as well?

lib/CodeGen/IslNodeBuilder.cpp

830

Can you write a comment about how the CFG will look like?

[jdoerfert]

In D13194#254243, @Meinersbur wrote:

What's the motivational use case that cannot be handled by LICM?

I forgot to split the test cases, sorry. Here is one:

• http://reviews.llvm.org/D13195#2eebfa1a

also note the commit message (it is stated there).

Could you create a separate pass that modifies the ScopInfo, i.e. lib/Transform/HoistInvariantLoads.cpp

I am unsure (not if I could but if it makes sense). This patch performs "licm" only as an optimization, true, but from D13195 onwards we will need it as a crucial part of the SCoP generation. Hence, I think it belongs (at least in the long run) in the ScoP class.

include/polly/CodeGen/IslNodeBuilder.h

209

I'll simplify the sentence and add a comma.

lib/Analysis/ScopInfo.cpp

1299

I am unsure. What should I replace by list::remove_if and why?

2285

We can do that easily as soon as another usecase presents itself. If ppl insist we can do it now but I don't see the benefit at the moment.

lib/CodeGen/IslNodeBuilder.cpp

830

I could, but it would just be the CFG from here a couple of times in sequence.

[jdoerfert]

lnt shows no problems, no gains and <10 regressions that are more than noise. However, these regressions sum up to less than 0.5s in total. Hence the detection of invariant loads costs a little bit but not much.

[Meinersbur]

In D13194#254248, @jdoerfert wrote:

In D13194#254243, @Meinersbur wrote:

Could you create a separate pass that modifies the ScopInfo, i.e. lib/Transform/HoistInvariantLoads.cpp

I am unsure (not if I could but if it makes sense). This patch performs "licm" only as an optimization, true, but from D13195 onwards we will need it as a crucial part of the SCoP generation. Hence, I think it belongs (at least in the long run) in the ScoP class.

OK

lib/Analysis/ScopInfo.cpp

1299

Does this just remove all elements of MAL from MemAccs? Maybe if not by remove_if (std::std::set_difference), this could be probably rewritten using STL functions. It also looks fragile because it depends on the order of elements in MAL and MemAccs.

Please add at least a comment about what the code is supposed to do.

2238

These do not initialize the Scop, but transform it.

lib/CodeGen/IslNodeBuilder.cpp

830

Linking into files in Phabricator reviews doesn't work for me. It just jumps to a location before it lazily loads the diffs.

You could just refer to that code location in the comment.

include/polly/CodeGen/IslNodeBuilder.h

IslNodeBuilder(PollyIRBuilder &Builder, ScopAnnotator &Annotator, Pass *P,
                  &ExprBuilder),
         RegionGen(BlockGen), P(P), DL(DL), LI(LI), SE(SE), DT(DT) {}
 
   virtual ~IslNodeBuilder() {}
 
   void addParameters(__isl_take isl_set *Context);
   void create(__isl_take isl_ast_node *Node);
 
+  /// @brief Preload all memory loads that are invariant.
+  void preloadInvariantLoads();
+
   /// @brief Finalize code generation for the SCoP @p S.
   ///
   /// @see BlockGenerator::finalizeSCoP(Scop &S)
   void finalizeSCoP(Scop &S) { BlockGen.finalizeSCoP(S); }
 
   IslExprBuilder &getExprBuilder() { return ExprBuilder; }
 
   /// @brief Get the associated block generator.
protected:
   ///
   /// For mark nodes with an unknown name, we just forward the code generation
   /// to its child. This is currently the only behavior implemented, as there is
   /// currently not special handling for marker nodes implemented.
   ///
   /// @param Mark The node we generate code for.
   virtual void createMark(__isl_take isl_ast_node *Marker);
   virtual void createFor(__isl_take isl_ast_node *For);
+
+  /// @brief Preload the memory load access @p MA.
+  ///
+  /// If @p MA is not always executed it will be conditionally loaded and
+  /// merged with undef from the same type. Hence, if @p MA is executed only
+  /// under condition C then the preload code will look like this:
+  ///
+  /// MA_preload = undef;
+  /// if (C)
+  ///   MA_preload = load MA;
+  /// use MA_preload
+  ///
+  /// Note that if the runtime context implies the condition C conditional
+  /// execution is not necessary.
+  Value *preloadInvariantLoad(const MemoryAccess &MA,
+                              __isl_keep isl_ast_build *Build,
+                              __isl_keep isl_set *RuntimeContext);
+
   void createForVector(__isl_take isl_ast_node *For, int VectorWidth);
   void createForSequential(__isl_take isl_ast_node *For);
 
   /// Create LLVM-IR that executes a for node thread parallel.
   ///
   /// @param For The FOR isl_ast_node for which code is generated.
   void createForParallel(__isl_take isl_ast_node *For);
 

include/polly/ScopInfo.h

public:
   ///
   ///  @param A vector of array sizes where the rightmost array sizes need to
   ///         match the innermost array sizes already defined in SAI.
   void updateSizes(ArrayRef<const SCEV *> Sizes);
 
   /// @brief Destructor to free the isl id of the base pointer.
   ~ScopArrayInfo();
 
+  /// @brief Set the base pointer to @p BP.
+  void setBasePtr(Value *BP) { BasePtr = BP; }
+
   /// @brief Return the base pointer.
   Value *getBasePtr() const { return BasePtr; }
 
   /// @brief For indirect accesses return the origin SAI of the BP, else null.
   const ScopArrayInfo *getBasePtrOriginSAI() const { return BasePtrOriginSAI; }
 
   /// @brief The set of derived indirect SAIs for this origin SAI.
   const SmallPtrSetImpl<ScopArrayInfo *> &getDerivedSAIs() const {
public:
 
   /// @brief Print the MemoryAccess to stderr.
   void dump() const;
 };
 
 llvm::raw_ostream &operator<<(llvm::raw_ostream &OS,
                               MemoryAccess::ReductionType RT);
 
+/// @brief List to hold all (scalar) memory accesses mapped to an instruction.
+using MemoryAccessList = std::forward_list<MemoryAccess *>;
+
 ///===----------------------------------------------------------------------===//
 /// @brief Statement of the Scop
 ///
 /// A Scop statement represents an instruction in the Scop.
 ///
 /// It is further described by its iteration domain, its schedule and its data
 /// accesses.
 /// At the moment every statement represents a single basic block of LLVM-IR.
 class ScopStmt {
 public:
-  /// @brief List to hold all (scalar) memory accesses mapped to an instruction.
-  using MemoryAccessList = std::forward_list<MemoryAccess *>;
-
   ScopStmt(const ScopStmt &) = delete;
   const ScopStmt &operator=(const ScopStmt &) = delete;
 
   /// Create the ScopStmt from a BasicBlock.
   ScopStmt(Scop &parent, BasicBlock &bb);
 
   /// Create an overapproximating ScopStmt for the region @p R.
   ScopStmt(Scop &parent, Region &R);
public:
   ///
   /// @return The region represented by this ScopStmt, or null if the statement
   ///         represents a basic block.
   Region *getRegion() const { return R; }
 
   /// @brief Return true if this statement represents a whole region.
   bool isRegionStmt() const { return R != nullptr; }
 
+  /// @brief Return true if this statement does not contain any accesses.
+  bool isEmpty() const { return MemAccs.empty(); }
+
   /// @brief Return the (scalar) memory accesses for @p Inst.
   const MemoryAccessList &getAccessesFor(const Instruction *Inst) const {
     MemoryAccessList *MAL = lookupAccessesFor(Inst);
     assert(MAL && "Cannot get memory accesses because they do not exist!");
     return *MAL;
   }
 
   /// @brief Return the (scalar) memory accesses for @p Inst if any.
public:
 
   void setBasicBlock(BasicBlock *Block) {
     // TODO: Handle the case where the statement is a region statement, thus
     //       the entry block was split and needs to be changed in the region R.
     assert(BB && "Cannot set a block for a region statement");
     BB = Block;
   }
 
+  /// @brief Move the memory access @p MA from this statement to @p TargetList.
+  ///
+  /// Note that scalar accesses that are caused by @p MA will be eliminated in
+  /// the whole SCoP.
+  void hoistMemoryAccess(MemoryAccess &MA, MemoryAccessList &TargetList);
+
   typedef MemoryAccessVec::iterator iterator;
   typedef MemoryAccessVec::const_iterator const_iterator;
 
   iterator begin() { return MemAccs.begin(); }
   iterator end() { return MemAccs.end(); }
   const_iterator begin() const { return MemAccs.begin(); }
   const_iterator end() const { return MemAccs.end(); }
 
private:
   bool IsOptimized;
 
   /// @brief True if the underlying region has a single exiting block.
   bool HasSingleExitEdge;
 
   /// Max loop depth.
   unsigned MaxLoopDepth;
 
-  typedef std::deque<ScopStmt> StmtSet;
+  typedef std::list<ScopStmt> StmtSet;
   /// The statements in this Scop.
   StmtSet Stmts;
 
   /// Parameters of this Scop
   typedef SmallVector<const SCEV *, 8> ParamVecType;
   ParamVecType Parameters;
 
   /// The isl_ids that are used to represent the parameters
private:
   /// In a program with int and float pointers annotated with tbaa information
   /// we would probably generate two alias groups, one for the int pointers and
   /// one for the float pointers.
   ///
   /// During code generation we will create a runtime alias check for each alias
   /// group to ensure the SCoP is executed in an alias free environment.
   MinMaxVectorPairVectorTy MinMaxAliasGroups;
 
+  /// @brief List of invariant accesses.
+  MemoryAccessList InvariantAccesses;
+
   /// @brief Scop constructor; invoked from ScopInfo::buildScop.
   Scop(Region &R, AccFuncMapType &AccFuncMap, ScalarEvolution &SE,
        DominatorTree &DT, isl_ctx *ctx, unsigned MaxLoopDepth);
 
   /// @brief Initialize this ScopInfo .
   void init(LoopInfo &LI, ScopDetection &SD, AliasAnalysis &AA);
 
   /// @brief Add loop carried constraints to the header block of the loop @p L.
private:
   /// @param BB The basic block to check
   ///
   /// @return True if the basic block is trivial, otherwise false.
   bool isTrivialBB(BasicBlock *BB);
 
   /// @brief Add parameter constraints to @p C that imply a non-empty domain.
   __isl_give isl_set *addNonEmptyDomainConstraints(__isl_take isl_set *C) const;
 
+  /// @brief TODO
+  void simplifySCoP();
+
+  /// @brief Mark all invariant memory loads and check for required ones.
+  ///
+  /// During SCoP detection loads can be tagged as "required invariant". In such
+  /// a case we need to verify here that we actually tagged them as invariant,
+  /// otherwise the SCoP is not valid and has to be dropped.
+  void collectAndCheckInvariantLoads(ScopDetection &SD);
+
   /// @brief Build the Context of the Scop.
   void buildContext();
 
   /// @brief Build the BoundaryContext based on the wrapping of expressions.
   void buildBoundaryContext();
 
   /// @brief Add user provided parameter constraints to context.
   void addUserContext();
public:
   inline const Region &getRegion() const { return R; }
   inline Region &getRegion() { return R; }
 
   /// @brief Get the maximum depth of the loop.
   ///
   /// @return The maximum depth of the loop.
   inline unsigned getMaxLoopDepth() const { return MaxLoopDepth; }
 
+  /// @brief Return the set of invariant accesses.
+  const MemoryAccessList &getInvariantAccesses() const {
+    return InvariantAccesses;
+  }
+
   /// @brief Mark the SCoP as optimized by the scheduler.
   void markAsOptimized() { IsOptimized = true; }
 
   /// @brief Check if the SCoP has been optimized by the scheduler.
   bool isOptimized() const { return IsOptimized; }
 
   /// @brief Get the name of this Scop.
   std::string getNameStr() const;

lib/Analysis/ScopInfo.cpp

if (Domain) {
     OS.indent(16) << "n/a\n";
 
   for (MemoryAccess *Access : MemAccs)
     Access->print(OS);
 }
 
 void ScopStmt::dump() const { print(dbgs()); }
 
+void ScopStmt::hoistMemoryAccess(MemoryAccess &MA,
+                                 MemoryAccessList &TargetList) {
+  auto &MAL = *lookupAccessesFor(MA.getAccessInstruction());
+  MAL.reverse();
+
+  auto MALIt = MAL.begin();
+  auto MALEnd = MAL.end();
+  auto MemAccsIt = MemAccs.begin();
+  while (MALIt != MALEnd) {
+    while (*MemAccsIt != *MALIt)
+      MemAccsIt++;
+
+    MALIt++;
+    MemAccs.erase(MemAccsIt);
+  }
+
+  TargetList.push_front(&MA);
+  InstructionToAccess.erase(MA.getAccessInstruction());
+
+  // Encode the statement domain in the memory access function of MA as we
+  // might delete Stmt later on but need to access the domain again during
+  // code generation. Though, we only need the parameter constraints as the
+  // access is invariant in all surrounding loops anyway.
+  ScopStmt &Stmt = *MA.getStatement();
+  isl_map *AR = MA.getAccessRelation();
+  AR = isl_map_remove_dims(AR, isl_dim_in, 0, isl_map_n_in(AR));
+  AR = isl_map_reset_tuple_id(AR, isl_dim_in);
+  AR = isl_map_intersect_params(AR, isl_set_params(Stmt.getDomain()));
+  AR = isl_map_detect_equalities(AR);
+  MA.setNewAccessRelation(isl_map_coalesce(AR));
+
+  delete &MAL;
+}
+
 //===----------------------------------------------------------------------===//
 /// Scop class implement
 
 void Scop::setContext(__isl_take isl_set *NewContext) {
   NewContext = isl_set_align_params(NewContext, isl_set_get_space(Context));
   isl_set_free(Context);
   Context = NewContext;
 }
void Scop::init(LoopInfo &LI, ScopDetection &SD, AliasAnalysis &AA) {
   Schedule = LoopSchedules[L].first;
 
   realignParams();
   addParameterBounds();
   addUserContext();
   buildBoundaryContext();
   simplifyContexts();
   buildAliasChecks(AA);
+
+  collectAndCheckInvariantLoads(SD);
+  simplifySCoP();
 }
 
 Scop::~Scop() {
   isl_set_free(Context);
   isl_set_free(AssumedContext);
   isl_set_free(BoundaryContext);
   isl_schedule_free(Schedule);
 

 }
 
 void Scop::updateAccessDimensionality() {
   for (auto &Stmt : *this)
     for (auto &Access : Stmt)
       Access->updateDimensionality();
 }
 
+void Scop::simplifySCoP() {
+
+  for (auto StmtIt = Stmts.begin(), StmtEnd = Stmts.end(); StmtIt != StmtEnd;) {
+    ScopStmt &Stmt = *StmtIt;
+
+    if (!StmtIt->isEmpty()) {
+      StmtIt++;
+      continue;
+    }
+
+    if (Stmt.isRegionStmt())
+      for (BasicBlock *BB : Stmt.getRegion()->blocks())
+        StmtMap.erase(BB);
+    else
+      StmtMap.erase(Stmt.getBasicBlock());
+
+    StmtIt = Stmts.erase(StmtIt);
+  }
+}
+
+void Scop::collectAndCheckInvariantLoads(ScopDetection &SD) {
+  isl_union_map *Writes = getWrites();
+  for (ScopStmt &Stmt : *this) {
+
+    // TODO: Loads that are not loop carried, hence are in a statement with
+    //       zero iterators, are by construction invariant, though we
+    //       currently "hoist" them anyway. This is necessary because we allow
+    //       them to be treated as parameters (e.g., in conditions) and our code
+    //       generation would otherwise use the old value.
+
+    isl_set *Domain = Stmt.getDomain();
+    SmallVector<MemoryAccess *, 4> InvMAs;
+
+    for (MemoryAccess *MA : Stmt) {
+      if (MA->isImplicit() || MA->isWrite() || !MA->isAffine())
+        continue;
+
+      isl_map *AccessRelation = MA->getAccessRelation();
+      if (isl_map_involves_dims(AccessRelation, isl_dim_in, 0,
+                                Stmt.getNumIterators())) {
+        isl_map_free(AccessRelation);
+        continue;
+      }
+
+      AccessRelation =
+          isl_map_intersect_domain(AccessRelation, isl_set_copy(Domain));
+      isl_set *AccessRange = isl_map_range(AccessRelation);
+
+      isl_union_map *Written = isl_union_map_intersect_range(
+          isl_union_map_copy(Writes), isl_union_set_from_set(AccessRange));
+      bool IsWritten = !isl_union_map_is_empty(Written);
+      isl_union_map_free(Written);
+
+      if (IsWritten)
+        continue;
+
+      InvMAs.push_back(MA);
+    }
+
+    // Transfer the memory access from the statement to the SCoP.
+    for (MemoryAccess *MA : InvMAs)
+      Stmt.hoistMemoryAccess(*MA, InvariantAccesses);
+
+    isl_set_free(Domain);
+  }
+  isl_union_map_free(Writes);
+
+  if (!InvariantAccesses.empty())
+    IsOptimized = true;
+
+  // We inserted invariant accesses always in the front but need them to be
+  // sorted in a "natural order". The statements are already sorted in reverse
+  // post order and that suffices for the accesses too. The reason we require
+  // an order in the first place is the dependences between invariant loads
+  // that can be caused by indirect loads but also conditionally executed onces
+  // where the condition is also an invariant load.
+  InvariantAccesses.reverse();
+}
+
 const ScopArrayInfo *
 Scop::getOrCreateScopArrayInfo(Value *BasePtr, Type *AccessType,
                                ArrayRef<const SCEV *> Sizes, bool IsPHI) {
   auto &SAI = ScopArrayInfoMap[std::make_pair(BasePtr, IsPHI)];
   if (!SAI) {
     SAI.reset(new ScopArrayInfo(BasePtr, AccessType, getIslCtx(), Sizes, IsPHI,
                                 this));
   } else {
void Scop::printArrayInfo(raw_ostream &OS) const {
   OS.indent(4) << "}\n";
 }
 
 void Scop::print(raw_ostream &OS) const {
   OS.indent(4) << "Function: " << getRegion().getEntry()->getParent()->getName()
                << "\n";
   OS.indent(4) << "Region: " << getNameStr() << "\n";
   OS.indent(4) << "Max Loop Depth:  " << getMaxLoopDepth() << "\n";
+  OS.indent(4) << "Invariant Accesses: {\n";
+  for (const MemoryAccess *MA : InvariantAccesses)
+    MA->print(OS);
+  OS.indent(4) << "}\n";
   printContext(OS.indent(4));
   printArrayInfo(OS.indent(4));
   printAliasAssumptions(OS);
   printStatements(OS.indent(4));
 }
 
 void Scop::dump() const { print(dbgs()); }
 

lib/CodeGen/BlockGenerators.cpp

-//===--- BlockGenerators.cpp - Generate code for statements -----*- C++ -*-===//
+
 //
 //                     The LLVM Compiler Infrastructure
 //
 // This file is distributed under the University of Illinois Open Source
 // License. See LICENSE.TXT for details.
 //
 //===----------------------------------------------------------------------===//
 //
Value *BlockGenerator::getNewValue(ScopStmt &Stmt, const Value *Old,
                                    ValueMapT &BBMap, LoopToScevMapT &LTS,
                                    Loop *L) const {
   // We assume constants never change.
   // This avoids map lookups for many calls to this function.
   if (isa<Constant>(Old))
     return const_cast<Value *>(Old);
 
   if (Value *New = GlobalMap.lookup(Old)) {
+    if (Value *NewRemapped = GlobalMap.lookup(New))
+      New = NewRemapped;
     if (Old->getType()->getScalarSizeInBits() <
         New->getType()->getScalarSizeInBits())
       New = Builder.CreateTruncOrBitCast(New, Old->getType());
 
     return New;
   }
 
   if (Value *New = BBMap.lookup(Old))

 
 Loop *BlockGenerator::getLoopForInst(const llvm::Instruction *Inst) {
   return LI.getLoopFor(Inst->getParent());
 }
 
 Value *BlockGenerator::generateScalarLoad(ScopStmt &Stmt, const LoadInst *Load,
                                           ValueMapT &BBMap, LoopToScevMapT &LTS,
                                           isl_id_to_ast_expr *NewAccesses) {
+  if (Value *PreloadLoad = GlobalMap.lookup(Load))
+    return PreloadLoad;
+
   const Value *Pointer = Load->getPointerOperand();
   Value *NewPointer =
       generateLocationAccessed(Stmt, Load, Pointer, BBMap, LTS, NewAccesses);
   Value *ScalarLoad = Builder.CreateAlignedLoad(
       NewPointer, Load->getAlignment(), Load->getName() + "_p_scalar_");
 
   if (DebugPrinting)
     RuntimeDebugBuilder::createCPUPrinter(Builder, "Load from ", NewPointer,
Value *VectorBlockGenerator::generateUnknownStrideLoad(
   }
 
   return Vector;
 }
 
 void VectorBlockGenerator::generateLoad(
     ScopStmt &Stmt, const LoadInst *Load, ValueMapT &VectorMap,
     VectorValueMapT &ScalarMaps, __isl_keep isl_id_to_ast_expr *NewAccesses) {
+  if (Value *PreloadLoad = GlobalMap.lookup(Load)) {
+    VectorMap[Load] = Builder.CreateVectorSplat(getVectorWidth(), PreloadLoad,
+                                                Load->getName() + "_p");
+    return;
+  }
+
   if (!VectorType::isValidElementType(Load->getType())) {
     for (int i = 0; i < getVectorWidth(); i++)
       ScalarMaps[i][Load] =
           generateScalarLoad(Stmt, Load, ScalarMaps[i], VLTS[i], NewAccesses);
     return;
   }
 
   const MemoryAccess &Access = Stmt.getAccessFor(Load);

lib/CodeGen/CodeGeneration.cpp

bool runOnScop(Scop &S) override {
     // the SCEVExpander may introduce while code generating the parameters and
     // which may introduce scalar dependences that prevent us from correctly
     // code generating this scop.
     BasicBlock *StartBlock =
         executeScopConditionally(S, this, Builder.getTrue());
     auto SplitBlock = StartBlock->getSinglePredecessor();
     Builder.SetInsertPoint(SplitBlock->getTerminator());
     NodeBuilder.addParameters(S.getContext());
+    NodeBuilder.preloadInvariantLoads();
     Value *RTC = buildRTC(Builder, NodeBuilder.getExprBuilder());
-    SplitBlock->getTerminator()->setOperand(0, RTC);
+    Builder.GetInsertBlock()->getTerminator()->setOperand(0, RTC);
     Builder.SetInsertPoint(StartBlock->begin());
 
     NodeBuilder.create(AstRoot);
 
     NodeBuilder.finalizeSCoP(S);
     fixRegionInfo(EnteringBB->getParent(), R->getParent());
 
     assert(!verifyGeneratedFunction(S, *EnteringBB->getParent()) &&

lib/CodeGen/IslNodeBuilder.cpp

void IslNodeBuilder::create(__isl_take isl_ast_node *Node) {
   case isl_ast_node_block:
     createBlock(Node);
     return;
   }
 
   llvm_unreachable("Unknown isl_ast_node type");
 }
 
+/// @brief Create the actual preload memory access for @p MA.
+static inline Value *createPreloadLoad(Scop &S, const MemoryAccess &MA,
+                                       isl_ast_build *Build,
+                                       IslExprBuilder &ExprBuilder) {
+  isl_set *AccessRange = isl_map_range(MA.getAccessRelation());
+  isl_set_dump(AccessRange);
+  isl_pw_multi_aff *PWAccRel = isl_pw_multi_aff_from_set(AccessRange);
+  PWAccRel = isl_pw_multi_aff_gist_params(PWAccRel, S.getContext());
+  isl_ast_expr *Access =
+      isl_ast_build_access_from_pw_multi_aff(Build, PWAccRel);
+  return ExprBuilder.create(Access);
+}
+
+Value *IslNodeBuilder::preloadInvariantLoad(const MemoryAccess &MA,
+                                            isl_ast_build *Build,
+                                            isl_set *RuntimeContext) {
+
+  isl_set *Domain = isl_map_params(MA.getAccessRelation());
+
+  if (isl_set_is_subset(RuntimeContext, Domain)) {
+    isl_set_free(Domain);
+    return createPreloadLoad(S, MA, Build, ExprBuilder);
+  } else {
+
+    isl_ast_expr *DomainCond = isl_ast_build_expr_from_set(Build, Domain);
+
+    Value *Cond = ExprBuilder.create(DomainCond);
+    if (!Cond->getType()->isIntegerTy(1))
+      Cond = Builder.CreateIsNotNull(Cond);
+
+    BasicBlock *CondBB = SplitBlock(Builder.GetInsertBlock(),
+                                    Builder.GetInsertPoint(), &DT, &LI);
+    CondBB->setName("polly.preload.cond");
+
+    BasicBlock *MergeBB = SplitBlock(CondBB, CondBB->begin(), &DT, &LI);
+    MergeBB->setName("polly.preload.merge");
+
+    Function *F = Builder.GetInsertBlock()->getParent();
+    LLVMContext &Context = F->getContext();
+    BasicBlock *ExecBB = BasicBlock::Create(Context, "polly.preload.exec", F);
+
+    DT.addNewBlock(ExecBB, CondBB);
+    if (Loop *L = LI.getLoopFor(CondBB))
+      L->addBasicBlockToLoop(ExecBB, LI);
+
+    auto *CondBBTerminator = CondBB->getTerminator();
+    Builder.SetInsertPoint(CondBBTerminator);
+    Builder.CreateCondBr(Cond, ExecBB, MergeBB);
+    CondBBTerminator->eraseFromParent();
+
+    Builder.SetInsertPoint(ExecBB);
+    Builder.CreateBr(MergeBB);
+
+    Builder.SetInsertPoint(ExecBB->getTerminator());
+    Instruction *AccInst = MA.getAccessInstruction();
+    Type *AccInstTy = AccInst->getType();
+    Value *PreAccInst = createPreloadLoad(S, MA, Build, ExprBuilder);
+
+    Builder.SetInsertPoint(MergeBB->getTerminator());
+    auto *MergePHI = Builder.CreatePHI(
+        AccInstTy, 2, "polly.preload." + AccInst->getName() + ".merge");
+    MergePHI->addIncoming(PreAccInst, ExecBB);
+    MergePHI->addIncoming(UndefValue::get(AccInstTy), CondBB);
+
+    return MergePHI;
+  }
+}
+
+void IslNodeBuilder::preloadInvariantLoads() {
+
+  const auto &InvAccList = S.getInvariantAccesses();
+  if (InvAccList.empty())
+    return;
+
+  const Region &R = S.getRegion();
+
+  BasicBlock *PreLoadBB =
+      SplitBlock(Builder.GetInsertBlock(), Builder.GetInsertPoint(), &DT, &LI);
+  PreLoadBB->setName("polly.preload.begin");
+  Builder.SetInsertPoint(PreLoadBB->begin());
+
+  isl_ast_build *Build =
+      isl_ast_build_from_context(isl_set_universe(S.getParamSpace()));
+  isl_set *RuntimeContext = S.getRuntimeCheckContext();
+
+  for (auto *MA : InvAccList) {
+    if (MA->isImplicit())
+      continue;
+    Instruction *AccInst = MA->getAccessInstruction();
+    Value *PreloadVal = preloadInvariantLoad(*MA, Build, RuntimeContext);
+    ValueMap[AccInst] = PreloadVal;
+
+    if (SE.isSCEVable(AccInst->getType())) {
+      isl_id *ParamId = S.getIdForParam(SE.getSCEV(AccInst));
+      if (ParamId)
+        IDToValue[ParamId] = PreloadVal;
+      isl_id_free(ParamId);
+    }
+
+    SmallVector<Instruction *, 4> Users;
+    for (auto *U : AccInst->users())
+      if (Instruction *UI = dyn_cast<Instruction>(U))
+        if (!R.contains(UI))
+          Users.push_back(UI);
+    for (auto *U : Users)
+      U->replaceUsesOfWith(AccInst, PreloadVal);
+
+    auto *SAI = S.getScopArrayInfo(MA->getBaseAddr());
+    for (auto *DerivedSAI : SAI->getDerivedSAIs())
+      DerivedSAI->setBasePtr(PreloadVal);
+  }
+
+  isl_set_free(RuntimeContext);
+  isl_ast_build_free(Build);
+}
+
 void IslNodeBuilder::addParameters(__isl_take isl_set *Context) {
 
   for (unsigned i = 0; i < isl_set_dim(Context, isl_dim_param); ++i) {
     isl_id *Id;
 
     Id = isl_set_get_dim_id(Context, isl_dim_param, i);
     IDToValue[Id] = generateSCEV((const SCEV *)isl_id_get_user(Id));
 

test/Isl/CodeGen/aliasing_parametric_simple_2.ll

 ; RUN: opt %loadPolly -polly-detect-unprofitable -polly-code-generator=isl -polly-codegen -S < %s | FileCheck %s
 ;
 ;    void jd(int *A, int *B, int c) {
 ;      for (int i = 0; i < 1024; i++)
 ;        A[i] = B[c - 10] + B[5];
 ;    }
 ;
 ; CHECK:  sext i32 %c to i64
+; CHECK:  sext i32 %c to i64
 ; CHECK:  %[[M0:[._a-zA-Z0-9]*]] = sext i32 %c to i64
 ; CHECK:  %[[M1:[._a-zA-Z0-9]*]] = icmp sle i64 %[[M0]], 15
 ; CHECK:  %[[M2:[._a-zA-Z0-9]*]] = sext i32 %c to i64
 ; CHECK:  %[[M3:[._a-zA-Z0-9]*]] = sub nsw i64 %[[M2]], 9
 ; CHECK:  %[[M4:[._a-zA-Z0-9]*]] = select i1 %[[M1]], i64 6, i64 %[[M3]]
 ; CHECK:  %[[BMax:[._a-zA-Z0-9]*]] = getelementptr i32, i32* %B, i64 %[[M4]]
 ; CHECK:  %[[AMin:[._a-zA-Z0-9]*]] = getelementptr i32, i32* %A, i64 0
 ; CHECK:  %[[BltA:[._a-zA-Z0-9]*]] = icmp ule i32* %[[BMax]], %[[AMin]]
 ; CHECK:  %[[AMax:[._a-zA-Z0-9]*]] = getelementptr i32, i32* %A, i64 1024
 ; CHECK:  %[[m0:[._a-zA-Z0-9]*]] = sext i32 %c to i64
 ; CHECK:  %[[m1:[._a-zA-Z0-9]*]] = icmp sge i64 %[[m0]], 15
 ; CHECK:  %[[m2:[._a-zA-Z0-9]*]] = sext i32 %c to i64
 ; CHECK:  %[[m3:[._a-zA-Z0-9]*]] = sub nsw i64 %[[m2]], 10
 ; CHECK:  %[[m4:[._a-zA-Z0-9]*]] = select i1 %[[m1]], i64 5, i64 %[[m3]]
 ; CHECK:  %[[BMin:[._a-zA-Z0-9]*]] = getelementptr i32, i32* %B, i64 %[[m4]]
 ; CHECK:  %[[AltB:[._a-zA-Z0-9]*]] = icmp ule i32* %[[AMax]], %[[BMin]]
 ; CHECK:  %[[NoAlias:[._a-zA-Z0-9]*]] = or i1 %[[BltA]], %[[AltB]]
-; CHECK:  %[[RTC:[._a-zA-Z0-9]*]] = and i1 %1, %[[NoAlias]]
+; CHECK:  %[[RTC:[._a-zA-Z0-9]*]] = and i1 %3, %[[NoAlias]]
 ; CHECK:  br i1 %[[RTC]], label %polly.start, label %for.cond
 ;
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 
 define void @jd(i32* %A, i32* %B, i32 %c) {
 entry:
   br label %for.cond
 

test/Isl/CodeGen/exprModDiv.ll

 ; RUN: opt %loadPolly -polly-import-jscop -polly-import-jscop-dir=%S \
 ; RUN:     -polly-codegen -polly-detect-unprofitable -S < %s | FileCheck %s
 ; RUN: opt %loadPolly -polly-import-jscop -polly-import-jscop-dir=%S \
 ; RUN:     -polly-codegen -polly-import-jscop-postfix=pow2 \
 ; RUN:     -polly-detect-unprofitable -S < %s | FileCheck %s -check-prefix=POW2
 ;
 ;    void exprModDiv(float *A, float *B, float *C, long N, long p) {
 ;      for (long i = 0; i < N; i++)
-;        C[i] += A[i] + B[i] + A[p] + B[p];
+;        C[i] += A[i] + B[i] + A[i] + B[i + p];
 ;    }
 ;
 ;
 ; This test case changes the access functions such that the resulting index
 ; expressions are modulo or division operations. We test that the code we
 ; generate takes advantage of knowledge about unsigned numerators. This is
 ; useful as LLVM will translate urem and udiv operations with power-of-two
 ; denominators to fast bitwise and or shift operations.

 ;       be mapped to any value. However, to generate correct code we require
 ;       each value of i to indeed be mapped to a value.
 ;
 ; CHECK:  %pexp.p_div_q = udiv i64 %polly.indvar, 127
 ; CHECK:  %polly.access.B8 = getelementptr float, float* %B, i64 %pexp.p_div_q
 
 ; #define floord(n,d) ((n < 0) ? (n - d + 1) : n) / d
 ; A[p + 127 * floord(-p - 1, 127) + 127]
-; CHECK:  %20 = sub nsw i64 0, %p
-; CHECK:  %21 = sub nsw i64 %20, 1
-; CHECK:  %pexp.fdiv_q.0 = sub i64 %21, 127
+; CHECK:  %17 = sub nsw i64 0, %p
+; CHECK:  %18 = sub nsw i64 %17, 1
+; CHECK:  %pexp.fdiv_q.0 = sub i64 %18, 127
 ; CHECK:  %pexp.fdiv_q.1 = add i64 %pexp.fdiv_q.0, 1
-; CHECK:  %pexp.fdiv_q.2 = icmp slt i64 %21, 0
-; CHECK:  %pexp.fdiv_q.3 = select i1 %pexp.fdiv_q.2, i64 %pexp.fdiv_q.1, i64 %21
+; CHECK:  %pexp.fdiv_q.2 = icmp slt i64 %18, 0
+; CHECK:  %pexp.fdiv_q.3 = select i1 %pexp.fdiv_q.2, i64 %pexp.fdiv_q.1, i64 %18
 ; CHECK:  %pexp.fdiv_q.4 = sdiv i64 %pexp.fdiv_q.3, 127
-; CHECK:  %22 = mul nsw i64 127, %pexp.fdiv_q.4
-; CHECK:  %23 = add nsw i64 %p, %22
-; CHECK:  %24 = add nsw i64 %23, 127
-; CHECK:  %polly.access.A10 = getelementptr float, float* %A, i64 %24
+; CHECK:  %19 = mul nsw i64 127, %pexp.fdiv_q.4
+; CHECK:  %20 = add nsw i64 %p, %19
+; CHECK:  %21 = add nsw i64 %20, 127
+; CHECK:  %polly.access.A10 = getelementptr float, float* %A, i64 %21
 
 ; A[p / 127]
 ; CHECK:  %pexp.div = sdiv exact i64 %p, 127
-; CHECK:  %polly.access.B12 = getelementptr float, float* %B, i64 %pexp.div
+; CHECK:  %polly.access.B13 = getelementptr float, float* %B, i64 %pexp.div
 
 ; A[i % 128]
 ; POW2:  %pexp.pdiv_r = urem i64 %polly.indvar, 128
 ; POW2:  %polly.access.A6 = getelementptr float, float* %A, i64 %pexp.pdiv_r
 
 ; A[floor(i / 128)]
 ; POW2:  %pexp.p_div_q = udiv i64 %polly.indvar, 128
 ; POW2:  %polly.access.B8 = getelementptr float, float* %B, i64 %pexp.p_div_q
 
 ; #define floord(n,d) ((n < 0) ? (n - d + 1) : n) / d
 ; A[p + 128 * floord(-p - 1, 128) + 128]
-; POW2:  %20 = sub nsw i64 0, %p
-; POW2:  %21 = sub nsw i64 %20, 1
-; POW2:  %polly.fdiv_q.shr = ashr i64 %21, 7
-; POW2:  %22 = mul nsw i64 128, %polly.fdiv_q.shr
-; POW2:  %23 = add nsw i64 %p, %22
-; POW2:  %24 = add nsw i64 %23, 128
-; POW2:  %polly.access.A10 = getelementptr float, float* %A, i64 %24
+; POW2:  %17 = sub nsw i64 0, %p
+; POW2:  %18 = sub nsw i64 %17, 1
+; POW2:  %polly.fdiv_q.shr = ashr i64 %18, 7
+; POW2:  %19 = mul nsw i64 128, %polly.fdiv_q.shr
+; POW2:  %20 = add nsw i64 %p, %19
+; POW2:  %21 = add nsw i64 %20, 128
+; POW2:  %polly.access.A10 = getelementptr float, float* %A, i64 %21
 
 ; A[p / 128]
 ; POW2:  %pexp.div = sdiv exact i64 %p, 128
-; POW2:  %polly.access.B12 = getelementptr float, float* %B, i64 %pexp.div
+; POW2:  %polly.access.B13 = getelementptr float, float* %B, i64 %pexp.div
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 
 define void @exprModDiv(float* %A, float* %B, float* %C, i64 %N, i64 %p) {
 entry:
   br label %for.cond
 
 for.cond:                                         ; preds = %for.inc, %entry
   %i.0 = phi i64 [ 0, %entry ], [ %inc, %for.inc ]
   %cmp = icmp slt i64 %i.0, %N
   br i1 %cmp, label %for.body, label %for.end
 
 for.body:                                         ; preds = %for.cond
   %arrayidx = getelementptr inbounds float, float* %A, i64 %i.0
   %tmp = load float, float* %arrayidx, align 4
   %arrayidx1 = getelementptr inbounds float, float* %B, i64 %i.0
   %tmp1 = load float, float* %arrayidx1, align 4
   %add = fadd float %tmp, %tmp1
-  %arrayidx2 = getelementptr inbounds float, float* %A, i64 %p
+  %arrayidx2 = getelementptr inbounds float, float* %A, i64 %i.0
   %tmp2 = load float, float* %arrayidx2, align 4
   %add3 = fadd float %add, %tmp2
-  %arrayidx4 = getelementptr inbounds float, float* %B, i64 %p
+  %padd = add nsw i64 %p, %i.0
+  %arrayidx4 = getelementptr inbounds float, float* %B, i64 %padd
   %tmp3 = load float, float* %arrayidx4, align 4
   %add5 = fadd float %add3, %tmp3
   %arrayidx6 = getelementptr inbounds float, float* %C, i64 %i.0
   %tmp4 = load float, float* %arrayidx6, align 4
   %add7 = fadd float %tmp4, %add5
   store float %add7, float* %arrayidx6, align 4
   br label %for.inc
 
 for.inc:                                          ; preds = %for.body
   %inc = add nuw nsw i64 %i.0, 1
   br label %for.cond
 
 for.end:                                          ; preds = %for.cond
   ret void
 }

test/Isl/CodeGen/non-affine-phi-node-expansion.ll

 ; RUN: opt %loadPolly -polly-codegen -polly-detect-unprofitable \
 ; RUN:     -S < %s | FileCheck %s
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 
 %struct.wombat = type {[4 x i32]}
 
+; CHECK:      polly.preload.begin:
+; CHECK-NEXT:   %polly.access.B = getelementptr i32, i32* %B, i64 0
+; CHECK-NEXT:   %polly.access.B.load = load i32, i32* %polly.access.B
+; CHECK-NOT:    %polly.access.B.load = load i32, i32* %polly.access.B
+
 ; CHECK: polly.stmt.bb3.entry:                             ; preds = %polly.start
 ; CHECK:   br label %polly.stmt.bb3
 
 ; CHECK: polly.stmt.bb3:                                   ; preds = %polly.stmt.bb3.entry
 ; CHECK:   %polly.subregion.iv = phi i32 [ 0, %polly.stmt.bb3.entry ]
 ; CHECK:   %polly.subregion.iv.inc = add i32 %polly.subregion.iv, 1
 ; CHECK:   br i1 true, label %polly.stmt.bb4, label %polly.stmt.bb5
 
 ; CHECK: polly.stmt.bb4:                                   ; preds = %polly.stmt.bb3
 ; CHECK:   br label %polly.stmt.bb13.exit
 
 ; CHECK: polly.stmt.bb5:                                   ; preds = %polly.stmt.bb3
-; CHECK:   %tmp7_p_scalar_ = load i32, i32* %B, !alias.scope !0, !noalias !2
-; CHECK:   store i32 %tmp7_p_scalar_, i32* %polly.access.cast.arg1, !alias.scope !3, !noalias !4
+; CHECK:   store i32 %polly.access.B.load, i32* %polly.access.cast.arg2
 ; CHECK:   br label %polly.stmt.bb13.exit
 
 ; Function Attrs: nounwind uwtable
 define void @quux(%struct.wombat* %arg, i32* %B) {
 bb:
   br i1 undef, label %bb2, label %bb1
 
 bb1:                                              ; preds = %bb

test/Isl/CodeGen/phi_in_exit_early_lnt_failure_4.ll

-; RUN: opt %loadPolly -disable-basicaa -polly-detect-unprofitable -polly-codegen -polly-no-early-exit -S < %s | FileCheck %s
-;
-; This caused an lnt crash at some point, just verify it will run through and
-; produce the PHI node in the exit we are looking for.
-;
-; CHECK-LABEL:  polly.merge_new_and_old:
-; CHECK-NEXT:     %.merge = phi %struct.ImageParameters.11.35.59.83.107.323.539.755.1019.1043.1187.1235.1355.1379.1403.1427.1499.1571.1667.1739.1835.2051.2123.2339.2387.2843.2867.2891.2915.3587.3803.3826* [ %.final_reload, %polly.stmt.for.end.298 ], [ %13, %for.end.298 ]
-;
-%struct.ImageParameters.11.35.59.83.107.323.539.755.1019.1043.1187.1235.1355.1379.1403.1427.1499.1571.1667.1739.1835.2051.2123.2339.2387.2843.2867.2891.2915.3587.3803.3826 = type { i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, float, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i8**, i8**, i32, i32***, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, [9 x [16 x [16 x i16]]], [5 x [16 x [16 x i16]]], [9 x [8 x [8 x i16]]], [2 x [4 x [16 x [16 x i16]]]], [16 x [16 x i16]], [16 x [16 x i32]], i32****, i32***, i32***, i32***, i32****, i32****, %struct.Picture.8.32.56.80.104.320.536.752.1016.1040.1184.1232.1352.1376.1400.1424.1496.1568.1664.1736.1832.2048.2120.2336.2384.2840.2864.2888.2912.3584.3800.3823*, %struct.Slice.7.31.55.79.103.319.535.751.1015.1039.1183.1231.1351.1375.1399.1423.1495.1567.1663.1735.1831.2047.2119.2335.2383.2839.2863.2887.2911.3583.3799.3822*, %struct.macroblock.9.33.57.81.105.321.537.753.1017.1041.1185.1233.1353.1377.1401.1425.1497.1569.1665.1737.1833.2049.2121.2337.2385.2841.2865.2889.2913.3585.3801.3824*, i32*, i32*, i32, i32, i32, i32, [4 x [4 x i32]], i32, i32, i32, i32, i32, double, i32, i32, i32, i32, i16******, i16******, i16******, i16******, [15 x i16], i32, i32, i32, i32, i32, i32, i32, i32, [6 x [32 x i32]], i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, [1 x i32], i32, i32, [2 x i32], i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, %struct.DecRefPicMarking_s.10.34.58.82.106.322.538.754.1018.1042.1186.1234.1354.1378.1402.1426.1498.1570.1666.1738.1834.2050.2122.2338.2386.2842.2866.2890.2914.3586.3802.3825*, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, double**, double***, i32***, double**, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, [3 x [2 x i32]], [2 x i32], i32, i32, i16, i32, i32, i32, i32, i32 }
-%struct.Picture.8.32.56.80.104.320.536.752.1016.1040.1184.1232.1352.1376.1400.1424.1496.1568.1664.1736.1832.2048.2120.2336.2384.2840.2864.2888.2912.3584.3800.3823 = type { i32, i32, [100 x %struct.Slice.7.31.55.79.103.319.535.751.1015.1039.1183.1231.1351.1375.1399.1423.1495.1567.1663.1735.1831.2047.2119.2335.2383.2839.2863.2887.2911.3583.3799.3822*], i32, float, float, float }
-%struct.Slice.7.31.55.79.103.319.535.751.1015.1039.1183.1231.1351.1375.1399.1423.1495.1567.1663.1735.1831.2047.2119.2335.2383.2839.2863.2887.2911.3583.3799.3822 = type { i32, i32, i32, i32, i32, i32, %struct.datapartition.3.27.51.75.99.315.531.747.1011.1035.1179.1227.1347.1371.1395.1419.1491.1563.1659.1731.1827.2043.2115.2331.2379.2835.2859.2883.2907.3579.3795.3818*, %struct.MotionInfoContexts.5.29.53.77.101.317.533.749.1013.1037.1181.1229.1349.1373.1397.1421.1493.1565.1661.1733.1829.2045.2117.2333.2381.2837.2861.2885.2909.3581.3797.3820*, %struct.TextureInfoContexts.6.30.54.78.102.318.534.750.1014.1038.1182.1230.1350.1374.1398.1422.1494.1566.1662.1734.1830.2046.2118.2334.2382.2838.2862.2886.2910.3582.3798.3821*, i32, i32*, i32*, i32*, i32, i32*, i32*, i32*, i32 (i32)*, [3 x [2 x i32]] }
-%struct.datapartition.3.27.51.75.99.315.531.747.1011.1035.1179.1227.1347.1371.1395.1419.1491.1563.1659.1731.1827.2043.2115.2331.2379.2835.2859.2883.2907.3579.3795.3818 = type { %struct.Bitstream.1.25.49.73.97.313.529.745.1009.1033.1177.1225.1345.1369.1393.1417.1489.1561.1657.1729.1825.2041.2113.2329.2377.2833.2857.2881.2905.3577.3793.3816*, %struct.EncodingEnvironment.2.26.50.74.98.314.530.746.1010.1034.1178.1226.1346.1370.1394.1418.1490.1562.1658.1730.1826.2042.2114.2330.2378.2834.2858.2882.2906.3578.3794.3817, %struct.EncodingEnvironment.2.26.50.74.98.314.530.746.1010.1034.1178.1226.1346.1370.1394.1418.1490.1562.1658.1730.1826.2042.2114.2330.2378.2834.2858.2882.2906.3578.3794.3817 }
-%struct.Bitstream.1.25.49.73.97.313.529.745.1009.1033.1177.1225.1345.1369.1393.1417.1489.1561.1657.1729.1825.2041.2113.2329.2377.2833.2857.2881.2905.3577.3793.3816 = type { i32, i32, i8, i32, i32, i8, i8, i32, i32, i8*, i32 }
-%struct.EncodingEnvironment.2.26.50.74.98.314.530.746.1010.1034.1178.1226.1346.1370.1394.1418.1490.1562.1658.1730.1826.2042.2114.2330.2378.2834.2858.2882.2906.3578.3794.3817 = type { i32, i32, i32, i32, i32, i8*, i32*, i32, i32 }
-%struct.MotionInfoContexts.5.29.53.77.101.317.533.749.1013.1037.1181.1229.1349.1373.1397.1421.1493.1565.1661.1733.1829.2045.2117.2333.2381.2837.2861.2885.2909.3581.3797.3820 = type { [3 x [11 x %struct.BiContextType.4.28.52.76.100.316.532.748.1012.1036.1180.1228.1348.1372.1396.1420.1492.1564.1660.1732.1828.2044.2116.2332.2380.2836.2860.2884.2908.3580.3796.3819]], [2 x [9 x %struct.BiContextType.4.28.52.76.100.316.532.748.1012.1036.1180.1228.1348.1372.1396.1420.1492.1564.1660.1732.1828.2044.2116.2332.2380.2836.2860.2884.2908.3580.3796.3819]], [2 x [10 x %struct.BiContextType.4.28.52.76.100.316.532.748.1012.1036.1180.1228.1348.1372.1396.1420.1492.1564.1660.1732.1828.2044.2116.2332.2380.2836.2860.2884.2908.3580.3796.3819]], [2 x [6 x %struct.BiContextType.4.28.52.76.100.316.532.748.1012.1036.1180.1228.1348.1372.1396.1420.1492.1564.1660.1732.1828.2044.2116.2332.2380.2836.2860.2884.2908.3580.3796.3819]], [4 x %struct.BiContextType.4.28.52.76.100.316.532.748.1012.1036.1180.1228.1348.1372.1396.1420.1492.1564.1660.1732.1828.2044.2116.2332.2380.2836.2860.2884.2908.3580.3796.3819], [4 x %struct.BiContextType.4.28.52.76.100.316.532.748.1012.1036.1180.1228.1348.1372.1396.1420.1492.1564.1660.1732.1828.2044.2116.2332.2380.2836.2860.2884.2908.3580.3796.3819], [3 x %struct.BiContextType.4.28.52.76.100.316.532.748.1012.1036.1180.1228.1348.1372.1396.1420.1492.1564.1660.1732.1828.2044.2116.2332.2380.2836.2860.2884.2908.3580.3796.3819] }
-%struct.BiContextType.4.28.52.76.100.316.532.748.1012.1036.1180.1228.1348.1372.1396.1420.1492.1564.1660.1732.1828.2044.2116.2332.2380.2836.2860.2884.2908.3580.3796.3819 = type { i16, i8, i64 }
-%struct.TextureInfoContexts.6.30.54.78.102.318.534.750.1014.1038.1182.1230.1350.1374.1398.1422.1494.1566.1662.1734.1830.2046.2118.2334.2382.2838.2862.2886.2910.3582.3798.3821 = type { [2 x %struct.BiContextType.4.28.52.76.100.316.532.748.1012.1036.1180.1228.1348.1372.1396.1420.1492.1564.1660.1732.1828.2044.2116.2332.2380.2836.2860.2884.2908.3580.3796.3819], [4 x %struct.BiContextType.4.28.52.76.100.316.532.748.1012.1036.1180.1228.1348.1372.1396.1420.1492.1564.1660.1732.1828.2044.2116.2332.2380.2836.2860.2884.2908.3580.3796.3819], [3 x [4 x %struct.BiContextType.4.28.52.76.100.316.532.748.1012.1036.1180.1228.1348.1372.1396.1420.1492.1564.1660.1732.1828.2044.2116.2332.2380.2836.2860.2884.2908.3580.3796.3819]], [10 x [4 x %struct.BiContextType.4.28.52.76.100.316.532.748.1012.1036.1180.1228.1348.1372.1396.1420.1492.1564.1660.1732.1828.2044.2116.2332.2380.2836.2860.2884.2908.3580.3796.3819]], [10 x [15 x %struct.BiContextType.4.28.52.76.100.316.532.748.1012.1036.1180.1228.1348.1372.1396.1420.1492.1564.1660.1732.1828.2044.2116.2332.2380.2836.2860.2884.2908.3580.3796.3819]], [10 x [15 x %struct.BiContextType.4.28.52.76.100.316.532.748.1012.1036.1180.1228.1348.1372.1396.1420.1492.1564.1660.1732.1828.2044.2116.2332.2380.2836.2860.2884.2908.3580.3796.3819]], [10 x [5 x %struct.BiContextType.4.28.52.76.100.316.532.748.1012.1036.1180.1228.1348.1372.1396.1420.1492.1564.1660.1732.1828.2044.2116.2332.2380.2836.2860.2884.2908.3580.3796.3819]], [10 x [5 x %struct.BiContextType.4.28.52.76.100.316.532.748.1012.1036.1180.1228.1348.1372.1396.1420.1492.1564.1660.1732.1828.2044.2116.2332.2380.2836.2860.2884.2908.3580.3796.3819]], [10 x [15 x %struct.BiContextType.4.28.52.76.100.316.532.748.1012.1036.1180.1228.1348.1372.1396.1420.1492.1564.1660.1732.1828.2044.2116.2332.2380.2836.2860.2884.2908.3580.3796.3819]], [10 x [15 x %struct.BiContextType.4.28.52.76.100.316.532.748.1012.1036.1180.1228.1348.1372.1396.1420.1492.1564.1660.1732.1828.2044.2116.2332.2380.2836.2860.2884.2908.3580.3796.3819]] }
-%struct.macroblock.9.33.57.81.105.321.537.753.1017.1041.1185.1233.1353.1377.1401.1425.1497.1569.1665.1737.1833.2049.2121.2337.2385.2841.2865.2889.2913.3585.3801.3824 = type { i32, i32, i32, [2 x i32], i32, [8 x i32], %struct.macroblock.9.33.57.81.105.321.537.753.1017.1041.1185.1233.1353.1377.1401.1425.1497.1569.1665.1737.1833.2049.2121.2337.2385.2841.2865.2889.2913.3585.3801.3824*, %struct.macroblock.9.33.57.81.105.321.537.753.1017.1041.1185.1233.1353.1377.1401.1425.1497.1569.1665.1737.1833.2049.2121.2337.2385.2841.2865.2889.2913.3585.3801.3824*, i32, [2 x [4 x [4 x [2 x i32]]]], [16 x i8], [16 x i8], i32, i64, [4 x i32], [4 x i32], i64, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i32, i16, double, i32, i32, i32, i32, i32, i32, i32, i32, i32 }
-%struct.DecRefPicMarking_s.10.34.58.82.106.322.538.754.1018.1042.1186.1234.1354.1378.1402.1426.1498.1570.1666.1738.1834.2050.2122.2338.2386.2842.2866.2890.2914.3586.3802.3825 = type { i32, i32, i32, i32, i32, %struct.DecRefPicMarking_s.10.34.58.82.106.322.538.754.1018.1042.1186.1234.1354.1378.1402.1426.1498.1570.1666.1738.1834.2050.2122.2338.2386.2842.2866.2890.2914.3586.3802.3825* }
-
-@img = external global %struct.ImageParameters.11.35.59.83.107.323.539.755.1019.1043.1187.1235.1355.1379.1403.1427.1499.1571.1667.1739.1835.2051.2123.2339.2387.2843.2867.2891.2915.3587.3803.3826*, align 8
-
-; Function Attrs: nounwind uwtable
-define void @intrapred_luma() #0 {
-entry:
-  %PredPel = alloca [13 x i16], align 16
-  br label %for.body
-
-for.body:                                         ; preds = %for.body, %entry
-  br i1 undef, label %for.body, label %for.body.262
-
-for.body.262:                                     ; preds = %for.body
-  %0 = load %struct.ImageParameters.11.35.59.83.107.323.539.755.1019.1043.1187.1235.1355.1379.1403.1427.1499.1571.1667.1739.1835.2051.2123.2339.2387.2843.2867.2891.2915.3587.3803.3826*, %struct.ImageParameters.11.35.59.83.107.323.539.755.1019.1043.1187.1235.1355.1379.1403.1427.1499.1571.1667.1739.1835.2051.2123.2339.2387.2843.2867.2891.2915.3587.3803.3826** @img, align 8
-  br label %for.body.280
-
-for.body.280:                                     ; preds = %for.body.280, %for.body.262
-  %indvars.iv66 = phi i64 [ 0, %for.body.262 ], [ %indvars.iv.next67, %for.body.280 ]
-  %arrayidx282 = getelementptr inbounds [13 x i16], [13 x i16]* %PredPel, i64 0, i64 1
-  %arrayidx283 = getelementptr inbounds i16, i16* %arrayidx282, i64 %indvars.iv66
-  %1 = load i16, i16* %arrayidx283, align 2
-  %arrayidx289 = getelementptr inbounds %struct.ImageParameters.11.35.59.83.107.323.539.755.1019.1043.1187.1235.1355.1379.1403.1427.1499.1571.1667.1739.1835.2051.2123.2339.2387.2843.2867.2891.2915.3587.3803.3826, %struct.ImageParameters.11.35.59.83.107.323.539.755.1019.1043.1187.1235.1355.1379.1403.1427.1499.1571.1667.1739.1835.2051.2123.2339.2387.2843.2867.2891.2915.3587.3803.3826* %0, i64 0, i32 47, i64 0, i64 2, i64 %indvars.iv66
-  store i16 %1, i16* %arrayidx289, align 2
-  %indvars.iv.next67 = add nuw nsw i64 %indvars.iv66, 1
-  br i1 false, label %for.body.280, label %for.end.298
-
-for.end.298:                                      ; preds = %for.body.280
-  %2 = load %struct.ImageParameters.11.35.59.83.107.323.539.755.1019.1043.1187.1235.1355.1379.1403.1427.1499.1571.1667.1739.1835.2051.2123.2339.2387.2843.2867.2891.2915.3587.3803.3826*, %struct.ImageParameters.11.35.59.83.107.323.539.755.1019.1043.1187.1235.1355.1379.1403.1427.1499.1571.1667.1739.1835.2051.2123.2339.2387.2843.2867.2891.2915.3587.3803.3826** @img, align 8
-  br label %for.body.310
-
-for.body.310:                                     ; preds = %for.body.310, %for.end.298
-  %indvars.iv = phi i64 [ 0, %for.end.298 ], [ %indvars.iv.next, %for.body.310 ]
-  %arrayidx312 = getelementptr inbounds [13 x i16], [13 x i16]* %PredPel, i64 0, i64 9
-  %arrayidx313 = getelementptr inbounds i16, i16* %arrayidx312, i64 %indvars.iv
-  %3 = load i16, i16* %arrayidx313, align 2
-  %arrayidx322 = getelementptr inbounds %struct.ImageParameters.11.35.59.83.107.323.539.755.1019.1043.1187.1235.1355.1379.1403.1427.1499.1571.1667.1739.1835.2051.2123.2339.2387.2843.2867.2891.2915.3587.3803.3826, %struct.ImageParameters.11.35.59.83.107.323.539.755.1019.1043.1187.1235.1355.1379.1403.1427.1499.1571.1667.1739.1835.2051.2123.2339.2387.2843.2867.2891.2915.3587.3803.3826* %2, i64 0, i32 47, i64 1, i64 %indvars.iv, i64 1
-  store i16 %3, i16* %arrayidx322, align 2
-  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
-  br i1 false, label %for.body.310, label %for.end.328
-
-for.end.328:                                      ; preds = %for.body.310
-  ret void
-}

test/Isl/CodeGen/simple_vec_call.ll

body:
   %indvar_next = add i64 %indvar, 1
   %exitcond = icmp eq i64 %indvar_next, 4
   br i1 %exitcond, label %return, label %body
 
 return:
   ret void
 }
 
-; CHECK: %value_p_splat_one = load <1 x float>, <1 x float>* bitcast ([1024 x float]* @A to <1 x float>*), align 8
-; CHECK: %value_p_splat = shufflevector <1 x float> %value_p_splat_one, <1 x float> %value_p_splat_one, <4 x i32> zeroinitializer
-; CHECK: %0 = extractelement <4 x float> %value_p_splat, i32 0
-; CHECK: %1 = extractelement <4 x float> %value_p_splat, i32 1
-; CHECK: %2 = extractelement <4 x float> %value_p_splat, i32 2
-; CHECK: %3 = extractelement <4 x float> %value_p_splat, i32 3
-; CHECK: [[RES1:%[a-zA-Z0-9_]+]] = tail call float @foo(float %0) [[NUW:#[0-9]+]]
-; CHECK: [[RES2:%[a-zA-Z0-9_]+]] = tail call float @foo(float %1) [[NUW]]
-; CHECK: [[RES3:%[a-zA-Z0-9_]+]] = tail call float @foo(float %2) [[NUW]]
-; CHECK: [[RES4:%[a-zA-Z0-9_]+]] = tail call float @foo(float %3) [[NUW]]
+; CHECK: [[RES1:%[a-zA-Z0-9_]+]] = tail call float @foo(float %.load) [[NUW:#[0-9]+]]
+; CHECK: [[RES2:%[a-zA-Z0-9_]+]] = tail call float @foo(float %.load) [[NUW]]
+; CHECK: [[RES3:%[a-zA-Z0-9_]+]] = tail call float @foo(float %.load) [[NUW]]
+; CHECK: [[RES4:%[a-zA-Z0-9_]+]] = tail call float @foo(float %.load) [[NUW]]
 ; CHECK: %4 = insertelement <4 x float> undef, float [[RES1]], i32 0
 ; CHECK: %5 = insertelement <4 x float> %4, float [[RES2]], i32 1
 ; CHECK: %6 = insertelement <4 x float> %5, float [[RES3]], i32 2
 ; CHECK: %7 = insertelement <4 x float> %6, float [[RES4]], i32 3
 ; CHECK:  store <4 x float> %7
 ; CHECK: attributes [[NUW]] = { nounwind }

test/Isl/CodeGen/simple_vec_call_2.ll

body:
   %indvar_next = add i64 %indvar, 1
   %exitcond = icmp eq i64 %indvar_next, 4
   br i1 %exitcond, label %return, label %body
 
 return:
   ret void
 }
 
-; CHECK: %value_p_splat_one = load <1 x float>, <1 x float>* bitcast ([1024 x float]* @A to <1 x float>*), align 8
-; CHECK: %value_p_splat = shufflevector <1 x float> %value_p_splat_one, <1 x float> %value_p_splat_one, <4 x i32> zeroinitializer
-; CHECK: %0 = extractelement <4 x float> %value_p_splat, i32 0
-; CHECK: %1 = extractelement <4 x float> %value_p_splat, i32 1
-; CHECK: %2 = extractelement <4 x float> %value_p_splat, i32 2
-; CHECK: %3 = extractelement <4 x float> %value_p_splat, i32 3
-; CHECK: [[RES1:%[a-zA-Z0-9_]+]] = tail call float** @foo(float %0) [[NUW:#[0-9]+]]
-; CHECK: [[RES2:%[a-zA-Z0-9_]+]] = tail call float** @foo(float %1) [[NUW]]
-; CHECK: [[RES3:%[a-zA-Z0-9_]+]] = tail call float** @foo(float %2) [[NUW]]
-; CHECK: [[RES4:%[a-zA-Z0-9_]+]] = tail call float** @foo(float %3) [[NUW]]
-; CHECK: %4 = insertelement <4 x float**> undef, float** %p_result, i32 0
-; CHECK: %5 = insertelement <4 x float**> %4, float** %p_result1, i32 1
-; CHECK: %6 = insertelement <4 x float**> %5, float** %p_result2, i32 2
-; CHECK: %7 = insertelement <4 x float**> %6, float** %p_result3, i32 3
-; CHECK: store <4 x float**> %7, <4 x float**>* bitcast ([1024 x float**]* @B to <4 x float**>*), align
+; CHECK: [[RES1:%[a-zA-Z0-9_]+]] = tail call float** @foo(float %.load) [[NUW:#[0-9]+]]
+; CHECK: [[RES2:%[a-zA-Z0-9_]+]] = tail call float** @foo(float %.load) [[NUW]]
+; CHECK: [[RES3:%[a-zA-Z0-9_]+]] = tail call float** @foo(float %.load) [[NUW]]
+; CHECK: [[RES4:%[a-zA-Z0-9_]+]] = tail call float** @foo(float %.load) [[NUW]]
+; CHECK: %0 = insertelement <4 x float**> undef, float** %p_result, i32 0
+; CHECK: %1 = insertelement <4 x float**> %0, float** %p_result1, i32 1
+; CHECK: %2 = insertelement <4 x float**> %1, float** %p_result2, i32 2
+; CHECK: %3 = insertelement <4 x float**> %2, float** %p_result3, i32 3
+; CHECK: store <4 x float**> %3, <4 x float**>* bitcast ([1024 x float**]* @B to <4 x float**>*), align
 ; CHECK: attributes [[NUW]] = { nounwind }

test/Isl/CodeGen/simple_vec_cast.ll

 bb3:                                              ; preds = %bb2
   %indvar.next = add i64 %indvar, 1
   br label %bb1
 
 bb4:                                              ; preds = %bb1
   ret void
 }
 
-; CHECK: %tmp_p_splat_one = load <1 x float>, <1 x float>* bitcast ([1024 x float]* @A to <1 x float>*), align 8, !alias.scope !0, !noalias !2
-; CHECK: %tmp_p_splat = shufflevector <1 x float> %tmp_p_splat_one, <1 x float> %tmp_p_splat_one, <4 x i32> zeroinitializer
-; CHECK: %0 = fpext <4 x float> %tmp_p_splat to <4 x double>
-; CHECK: store <4 x double> %0, <4 x double>* bitcast ([1024 x double]* @B to <4 x double>*), align 8, !alias.scope !3, !noalias !4
+; CHECK:   %.load = load float, float* getelementptr inbounds ([1024 x float], [1024 x float]* @A, i32 0, i32 0)
 
+; CHECK: polly.stmt.bb2:                                   ; preds = %polly.start
+; CHECK:   %tmp_p.splatinsert = insertelement <4 x float> undef, float %.load, i32 0
+; CHECK:   %tmp_p.splat = shufflevector <4 x float> %tmp_p.splatinsert, <4 x float> undef, <4 x i32> zeroinitializer
+; CHECK:   %0 = fpext <4 x float> %tmp_p.splat to <4 x double>
+; CHECK:   store <4 x double> %0, <4 x double>*

test/Isl/CodeGen/simple_vec_const.ll

 define i32 @main() nounwind {
   call void @simple_vec_const()
   %1 = load float, float* getelementptr inbounds ([1024 x float], [1024 x float]* @A, i64 0, i64 42), align 8
   %2 = fptosi float %1 to i32
   ret i32 %2
 }
 
 
-; CHECK: load <1 x float>, <1 x float>* bitcast ([1024 x float]* @A to <1 x float>*)
-; CHECK: shufflevector <1 x float> {{.*}}, <1 x float> {{.*}} <4 x i32> zeroinitializer
+; CHECK:   %.load = load float, float* getelementptr inbounds ([1024 x float], [1024 x float]* @A, i32 0, i32 0)
+
+; CHECK: polly.stmt.:                                      ; preds = %polly.start
+; CHECK:   %_p.splatinsert = insertelement <4 x float> undef, float %.load, i32 0
+; CHECK:   %_p.splat = shufflevector <4 x float> %_p.splatinsert, <4 x float> undef, <4 x i32> zeroinitializer

test/Isl/CodeGen/simple_vec_ptr_ptr_ty.ll

body:
   store float** %value, float*** %scevgep, align 4
   %indvar_next = add i64 %indvar, 1
   %exitcond = icmp eq i64 %indvar_next, 4
   br i1 %exitcond, label %return, label %body
 
 return:
   ret void
 }
-; CHECK: %value_p_splat_one = load <1 x float**>, <1 x float**>* bitcast ([1024 x float**]* @A to <1 x float**>*), align 8
-; CHECK: %value_p_splat = shufflevector <1 x float**> %value_p_splat_one, <1 x float**> %value_p_splat_one, <4 x i32> zeroinitializer
-; CHECK: store <4 x float**> %value_p_splat, <4 x float**>* bitcast ([1024 x float**]* @B to <4 x float**>*), align 8
+; CHECK:   %.load = load float**, float*** getelementptr inbounds ([1024 x float**], [1024 x float**]* @A, i32 0, i32 0)
+
+; CHECK-NOT: load <1 x float**>
+; CHECK: %value_p.splatinsert = insertelement <4 x float**> undef, float** %.load, i32 0
+; CHECK: %value_p.splat = shufflevector <4 x float**> %value_p.splatinsert, <4 x float**> undef, <4 x i32> zeroinitializer
+; CHECK: store <4 x float**> %value_p.splat, <4 x float**>* bitcast ([1024 x float**]* @B to <4 x float**>*), align 8

test/Isl/CodeGen/two-scops-in-row.ll

 
 define void @foo(i32* %A) {
 entry:
   %Scalar0 = alloca i32
   br label %for.0
 
 for.0:
   %Scalar0.val = load i32, i32* %Scalar0
+  store i32 1, i32* %Scalar0
   br i1 false, label %for.0, label %for.1.preheader
 
 for.1.preheader:
   fence seq_cst
   br label %for.1
 
 for.1:
   %indvar.1 = phi i32 [ %Scalar0.val, %for.1.preheader ], [ %indvar.1.next, %for.1]

test/ScopInfo/inter_bb_scalar_dep.ll

 ; RUN: opt %loadPolly -polly-detect-unprofitable -basicaa -polly-scops -analyze < %s | FileCheck %s
 
 ; void f(long A[], int N, int *init_ptr) {
 ;   long i, j;
 ;
 ;   for (i = 0; i < N; ++i) {
 ;     init = *init_ptr;
 ;     for (i = 0; i < N; ++i) {
 ;       A[i] = init + 2;
 ;     }
 ;   }
 ; }
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128"
 
 ; Function Attrs: nounwind
+; CHECK: Invariant
+; CHECK:           ReadAccess :=       [Reduction Type: NONE] [Scalar: 0]
+; CHECK-NEXT:          MemRef_init_ptr[0]
+
 define void @f(i64* noalias %A, i64 %N, i64* noalias %init_ptr) #0 {
 entry:
   br label %for.i
 
 for.i:                                            ; preds = %for.i.end, %entry
   %indvar.i = phi i64 [ 0, %entry ], [ %indvar.i.next, %for.i.end ]
   %indvar.i.next = add nsw i64 %indvar.i, 1
   br label %entry.next
 
 entry.next:                                       ; preds = %for.i
   %init = load i64, i64* %init_ptr
-; CHECK-LABEL: Stmt_entry_next
-; CHECK:           ReadAccess :=       [Reduction Type: NONE] [Scalar: 0]
-; CHECK-NEXT:          [N] -> { Stmt_entry_next[i0] -> MemRef_init_ptr[0] };
-; CHECK:           MustWriteAccess :=  [Reduction Type: NONE] [Scalar: 1]
-; CHECK-NEXT:          [N] -> { Stmt_entry_next[i0] -> MemRef_init[] };
+; CHECK-NOT: Stmt_entry_next
   br label %for.j
 
 for.j:                                            ; preds = %for.j, %entry.next
   %indvar.j = phi i64 [ 0, %entry.next ], [ %indvar.j.next, %for.j ]
   %init_plus_two = add i64 %init, 2
 ; CHECK-LABEL: Stmt_for_j
 ; CHECK:           ReadAccess :=       [Reduction Type: NONE] [Scalar: 1]
 ; CHECK-NEXT:          [N] -> { Stmt_for_j[i0, i1] -> MemRef_init[] };

test/ScopInfo/intra_and_inter_bb_scalar_dep.ll

 ; RUN: opt %loadPolly -polly-detect-unprofitable -basicaa -polly-scops -analyze < %s | FileCheck %s
 
 ; void f(long A[], int N, int *init_ptr) {
 ;   long i, j;
 ;
 ;   for (i = 0; i < N; ++i) {
 ;     init = *init_ptr;
 ;     for (i = 0; i < N; ++i) {
 ;       init2 = *init_ptr;
 ;       A[i] = init + init2;
 ;     }
 ;   }
 ; }
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128"
 
-; Function Attrs: nounwind
+; CHECK:       Invariant Accesses: {
+; CHECK:           ReadAccess :=       [Reduction Type: NONE] [Scalar: 0]
+; CHECK:              MemRef_init_ptr[0]
+; CHECK:           ReadAccess :=       [Reduction Type: NONE] [Scalar: 0]
+; CHECK:              MemRef_init_ptr[0]
+; CHECK:       }
 define void @f(i64* noalias %A, i64 %N, i64* noalias %init_ptr) #0 {
 entry:
   br label %for.i
 
 for.i:                                            ; preds = %for.i.end, %entry
   %indvar.i = phi i64 [ 0, %entry ], [ %indvar.i.next, %for.i.end ]
   %indvar.i.next = add nsw i64 %indvar.i, 1
   br label %entry.next
 
 entry.next:                                       ; preds = %for.i
   %init = load i64, i64* %init_ptr
-; CHECK-LABEL: Stmt_entry_next
-; CHECK:           ReadAccess :=       [Reduction Type: NONE] [Scalar: 0]
-; CHECK-NEXT:          [N] -> { Stmt_entry_next[i0] -> MemRef_init_ptr[0] };
-; CHECK:           MustWriteAccess :=  [Reduction Type: NONE] [Scalar: 1]
-; CHECK-NEXT:          [N] -> { Stmt_entry_next[i0] -> MemRef_init[] };
+; CHECK-NOT: Stmt_entry_next
   br label %for.j
 
 for.j:                                            ; preds = %for.j, %entry.next
   %indvar.j = phi i64 [ 0, %entry.next ], [ %indvar.j.next, %for.j ]
   %init_2 = load i64, i64* %init_ptr
   %init_sum = add i64 %init, %init_2
-; CHECK-LABEL: Stmt_for_j
+; CHECK:      Stmt_for_j
 ; CHECK:           ReadAccess :=       [Reduction Type: NONE] [Scalar: 1]
 ; CHECK-NEXT:          [N] -> { Stmt_for_j[i0, i1] -> MemRef_init[] };
-; CHECK:           ReadAccess :=       [Reduction Type: NONE] [Scalar: 0]
-; CHECK-NEXT:          [N] -> { Stmt_for_j[i0, i1] -> MemRef_init_ptr[0] };
-; CHECK:           MustWriteAccess :=  [Reduction Type: NONE] [Scalar: 0]
+; CHECK-NEXT:      MustWriteAccess :=  [Reduction Type: NONE] [Scalar: 0]
 ; CHECK-NEXT:          [N] -> { Stmt_for_j[i0, i1] -> MemRef_A[i1] };
   %scevgep = getelementptr i64, i64* %A, i64 %indvar.j
   store i64 %init_sum, i64* %scevgep
   %indvar.j.next = add nsw i64 %indvar.j, 1
   %exitcond.j = icmp eq i64 %indvar.j.next, %N
   br i1 %exitcond.j, label %for.i.end, label %for.j
 
 for.i.end:                                        ; preds = %for.j
   %exitcond.i = icmp eq i64 %indvar.i.next, %N
   br i1 %exitcond.i, label %return, label %for.i
 
 return:                                           ; preds = %for.i.end
   ret void
 }
 
 attributes #0 = { nounwind }

test/ScopInfo/intra_bb_scalar_dep.ll

 ; RUN: opt %loadPolly -polly-detect-unprofitable -basicaa -polly-scops -analyze < %s | FileCheck %s
 
 ; void f(long A[], int N, int *init_ptr) {
 ;   long i, j;
 ;
 ;   for (i = 0; i < N; ++i) {
 ;     for (i = 0; i < N; ++i) {
 ;       init = *init_ptr;
 ;       A[i] = init + 2;
 ;     }
 ;   }
 ; }
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128"
 
 ; Function Attrs: nounwind
+; CHECK:      Invariant Accesses:
+; CHECK-NEXT:     ReadAccess :=  [Reduction Type: NONE] [Scalar: 0]
+; CHECK-NEXT:          [N] -> { Stmt_for_j[i0, i1] -> MemRef_init_ptr[0] };
 define void @f(i64* noalias %A, i64 %N, i64* noalias %init_ptr) #0 {
 entry:
   br label %for.i
 
 for.i:                                            ; preds = %for.i.end, %entry
   %indvar.i = phi i64 [ 0, %entry ], [ %indvar.i.next, %for.i.end ]
   %indvar.i.next = add nsw i64 %indvar.i, 1
   br label %entry.next
 
 entry.next:                                       ; preds = %for.i
   br label %for.j
 
 for.j:                                            ; preds = %for.j, %entry.next
   %indvar.j = phi i64 [ 0, %entry.next ], [ %indvar.j.next, %for.j ]
   %init = load i64, i64* %init_ptr
   %init_plus_two = add i64 %init, 2
   %scevgep = getelementptr i64, i64* %A, i64 %indvar.j
   store i64 %init_plus_two, i64* %scevgep
-; CHECK-LABEL: Stmt_for_j
-; CHECK:           ReadAccess :=       [Reduction Type: NONE] [Scalar: 0]
-; CHECK-NEXT:          [N] -> { Stmt_for_j[i0, i1] -> MemRef_init_ptr[0] };
-; CHECK:           MustWriteAccess :=  [Reduction Type: NONE] [Scalar: 0]
-; CHECK-NEXT:          [N] -> { Stmt_for_j[i0, i1] -> MemRef_A[i1] };
+; CHECK:      Statements {
+; CHECK-NEXT:   Stmt_for_j
+; CHECK-NOT:     ReadAccess
+; CHECK:         MustWriteAccess := [Reduction Type: NONE] [Scalar: 0]
+; CHECK-NEXT:            [N] -> { Stmt_for_j[i0, i1] -> MemRef_A[i1] };
+; CHECK-NEXT:  }
   %indvar.j.next = add nsw i64 %indvar.j, 1
   %exitcond.j = icmp eq i64 %indvar.j.next, %N
   br i1 %exitcond.j, label %for.i.end, label %for.j
 
 for.i.end:                                        ; preds = %for.j
   %exitcond.i = icmp eq i64 %indvar.i.next, %N
   br i1 %exitcond.i, label %return, label %for.i
 
 return:                                           ; preds = %for.i.end
   ret void
 }
 
 attributes #0 = { nounwind }

test/ScopInfo/tempscop-printing.ll

 ; RUN: opt %loadPolly -polly-detect-unprofitable -basicaa -polly-scops -analyze < %s | FileCheck %s
 
 ; void f(long A[], int N, int *init_ptr) {
 ;   long i, j;
 ;
 ;   for (i = 0; i < N; ++i) {
 ;     init = *init_ptr;
 ;     for (i = 0; i < N; ++i) {
 ;       A[i] = init + 2;
 ;     }
 ;   }
 ; }
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128"
 
 ; CHECK-LABEL: Function: f
+; CHECK:          ReadAccess :=       [Reduction Type: NONE] [Scalar: 0]
+; CHECK-NEXT:         MemRef_init_ptr[0]
+
 define void @f(i64* noalias %A, i64 %N, i64* noalias %init_ptr) nounwind {
 entry:
   br label %for.i
 
 for.i:
   %indvar.i = phi i64 [ 0, %entry ], [ %indvar.i.next, %for.i.end ]
   %indvar.i.next = add nsw i64 %indvar.i, 1
   br label %entry.next
 
 entry.next:
-; CHECK:      Stmt_entry_next
+; CHECK-NOT:      Stmt_entry_next
   %init = load i64, i64* %init_ptr
-; CHECK:          ReadAccess :=       [Reduction Type: NONE] [Scalar: 0]
-; CHECK-NEXT:         [N] -> { Stmt_entry_next[i0] -> MemRef_init_ptr[0] };
-; CHECK:          MustWriteAccess :=  [Reduction Type: NONE] [Scalar: 1]
-; CHECK-NEXT:         [N] -> { Stmt_entry_next[i0] -> MemRef_init[] };
   br label %for.j
 
 for.j:
   %indvar.j = phi i64 [ 0, %entry.next ], [ %indvar.j.next, %for.j ]
 ; CHECK:      Stmt_for_j
 ; CHECK:          ReadAccess :=       [Reduction Type: NONE] [Scalar: 1]
 ; CHECK-NEXT:         [N] -> { Stmt_for_j[i0, i1] -> MemRef_init[] };
 ; CHECK:          MustWriteAccess :=  [Reduction Type: NONE] [Scalar: 0]
for.i.end:
   %exitcond.i = icmp eq i64 %indvar.i.next, %N
   br i1 %exitcond.i, label %return, label %for.i
 
 return:
   ret void
 }
 
 ; CHECK-LABEL: Function: g
+; CHECK:          ReadAccess :=       [Reduction Type: NONE] [Scalar: 0]
+; CHECK-NEXT:         MemRef_init_ptr[0]
 define void @g(i64* noalias %A, i64 %N, i64* noalias %init_ptr) nounwind {
 entry:
   br label %for.i
 
 for.i:
   %indvar.i = phi i64 [ 0, %entry ], [ %indvar.i.next, %for.i.end ]
   %indvar.i.next = add nsw i64 %indvar.i, 1
   br label %entry.next
 
 entry.next:
-; CHECK:      Stmt_entry_next
+; CHECK-NOT:      Stmt_entry_next
   %init = load i64, i64* %init_ptr
-; CHECK:          ReadAccess :=       [Reduction Type: NONE] [Scalar: 0]
-; CHECK-NEXT:         [N] -> { Stmt_entry_next[i0] -> MemRef_init_ptr[0] };
-; CHECK:          MustWriteAccess :=  [Reduction Type: NONE] [Scalar: 1]
-; CHECK-NEXT:         [N] -> { Stmt_entry_next[i0] -> MemRef_init[] };
   br label %for.j
 
 for.j:
 ; CHECK:      Stmt_for_j
   %indvar.j = phi i64 [ 0, %entry.next ], [ %indvar.j.next, %for.j ]
   %scevgep = getelementptr i64, i64* %A, i64 %indvar.j
   store i64 %init, i64* %scevgep
 ; CHECK:          ReadAccess :=       [Reduction Type: NONE] [Scalar: 1]