1 //===- llvm/Analysis/ScalarEvolution.h - Scalar Evolution -------*- C++ -*-===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // The ScalarEvolution class is an LLVM pass which can be used to analyze and
11 // categorize scalar expressions in loops. It specializes in recognizing
12 // general induction variables, representing them with the abstract and opaque
13 // SCEV class. Given this analysis, trip counts of loops and other important
14 // properties can be obtained.
16 // This analysis is primarily useful for induction variable substitution and
17 // strength reduction.
19 //===----------------------------------------------------------------------===//
21 #ifndef LLVM_ANALYSIS_SCALAREVOLUTION_H
22 #define LLVM_ANALYSIS_SCALAREVOLUTION_H
24 #include "llvm/ADT/DenseSet.h"
25 #include "llvm/ADT/FoldingSet.h"
26 #include "llvm/ADT/SetVector.h"
27 #include "llvm/Analysis/LoopInfo.h"
28 #include "llvm/IR/ConstantRange.h"
29 #include "llvm/IR/Instructions.h"
30 #include "llvm/IR/Operator.h"
31 #include "llvm/IR/PassManager.h"
32 #include "llvm/IR/ValueHandle.h"
33 #include "llvm/IR/ValueMap.h"
34 #include "llvm/Pass.h"
35 #include "llvm/Support/Allocator.h"
36 #include "llvm/Support/DataTypes.h"
40 class AssumptionCache;
45 class ScalarEvolution;
47 class TargetLibraryInfo;
58 template <> struct FoldingSetTrait<SCEV>;
59 template <> struct FoldingSetTrait<SCEVPredicate>;
61 /// This class represents an analyzed expression in the program. These are
62 /// opaque objects that the client is not allowed to do much with directly.
64 class SCEV : public FoldingSetNode {
65 friend struct FoldingSetTrait<SCEV>;
67 /// A reference to an Interned FoldingSetNodeID for this node. The
68 /// ScalarEvolution's BumpPtrAllocator holds the data.
69 FoldingSetNodeIDRef FastID;
71 // The SCEV baseclass this node corresponds to
72 const unsigned short SCEVType;
75 /// This field is initialized to zero and may be used in subclasses to store
76 /// miscellaneous information.
77 unsigned short SubclassData;
80 SCEV(const SCEV &) = delete;
81 void operator=(const SCEV &) = delete;
84 /// NoWrapFlags are bitfield indices into SubclassData.
86 /// Add and Mul expressions may have no-unsigned-wrap <NUW> or
87 /// no-signed-wrap <NSW> properties, which are derived from the IR
88 /// operator. NSW is a misnomer that we use to mean no signed overflow or
91 /// AddRec expressions may have a no-self-wraparound <NW> property if, in
92 /// the integer domain, abs(step) * max-iteration(loop) <=
93 /// unsigned-max(bitwidth). This means that the recurrence will never reach
94 /// its start value if the step is non-zero. Computing the same value on
95 /// each iteration is not considered wrapping, and recurrences with step = 0
96 /// are trivially <NW>. <NW> is independent of the sign of step and the
97 /// value the add recurrence starts with.
99 /// Note that NUW and NSW are also valid properties of a recurrence, and
100 /// either implies NW. For convenience, NW will be set for a recurrence
101 /// whenever either NUW or NSW are set.
102 enum NoWrapFlags { FlagAnyWrap = 0, // No guarantee.
103 FlagNW = (1 << 0), // No self-wrap.
104 FlagNUW = (1 << 1), // No unsigned wrap.
105 FlagNSW = (1 << 2), // No signed wrap.
106 NoWrapMask = (1 << 3) -1 };
108 explicit SCEV(const FoldingSetNodeIDRef ID, unsigned SCEVTy) :
109 FastID(ID), SCEVType(SCEVTy), SubclassData(0) {}
111 unsigned getSCEVType() const { return SCEVType; }
113 /// Return the LLVM type of this SCEV expression.
115 Type *getType() const;
117 /// Return true if the expression is a constant zero.
121 /// Return true if the expression is a constant one.
125 /// Return true if the expression is a constant all-ones value.
127 bool isAllOnesValue() const;
129 /// Return true if the specified scev is negated, but not a constant.
130 bool isNonConstantNegative() const;
132 /// Print out the internal representation of this scalar to the specified
133 /// stream. This should really only be used for debugging purposes.
134 void print(raw_ostream &OS) const;
136 /// This method is used for debugging.
141 // Specialize FoldingSetTrait for SCEV to avoid needing to compute
142 // temporary FoldingSetNodeID values.
143 template<> struct FoldingSetTrait<SCEV> : DefaultFoldingSetTrait<SCEV> {
144 static void Profile(const SCEV &X, FoldingSetNodeID& ID) {
147 static bool Equals(const SCEV &X, const FoldingSetNodeID &ID,
148 unsigned IDHash, FoldingSetNodeID &TempID) {
149 return ID == X.FastID;
151 static unsigned ComputeHash(const SCEV &X, FoldingSetNodeID &TempID) {
152 return X.FastID.ComputeHash();
156 inline raw_ostream &operator<<(raw_ostream &OS, const SCEV &S) {
161 /// An object of this class is returned by queries that could not be answered.
162 /// For example, if you ask for the number of iterations of a linked-list
163 /// traversal loop, you will get one of these. None of the standard SCEV
164 /// operations are valid on this class, it is just a marker.
165 struct SCEVCouldNotCompute : public SCEV {
166 SCEVCouldNotCompute();
168 /// Methods for support type inquiry through isa, cast, and dyn_cast:
169 static bool classof(const SCEV *S);
172 /// This class represents an assumption made using SCEV expressions which can
173 /// be checked at run-time.
174 class SCEVPredicate : public FoldingSetNode {
175 friend struct FoldingSetTrait<SCEVPredicate>;
177 /// A reference to an Interned FoldingSetNodeID for this node. The
178 /// ScalarEvolution's BumpPtrAllocator holds the data.
179 FoldingSetNodeIDRef FastID;
182 enum SCEVPredicateKind { P_Union, P_Equal, P_Wrap };
185 SCEVPredicateKind Kind;
186 ~SCEVPredicate() = default;
187 SCEVPredicate(const SCEVPredicate&) = default;
188 SCEVPredicate &operator=(const SCEVPredicate&) = default;
191 SCEVPredicate(const FoldingSetNodeIDRef ID, SCEVPredicateKind Kind);
193 SCEVPredicateKind getKind() const { return Kind; }
195 /// Returns the estimated complexity of this predicate. This is roughly
196 /// measured in the number of run-time checks required.
197 virtual unsigned getComplexity() const { return 1; }
199 /// Returns true if the predicate is always true. This means that no
200 /// assumptions were made and nothing needs to be checked at run-time.
201 virtual bool isAlwaysTrue() const = 0;
203 /// Returns true if this predicate implies \p N.
204 virtual bool implies(const SCEVPredicate *N) const = 0;
206 /// Prints a textual representation of this predicate with an indentation of
208 virtual void print(raw_ostream &OS, unsigned Depth = 0) const = 0;
210 /// Returns the SCEV to which this predicate applies, or nullptr if this is
211 /// a SCEVUnionPredicate.
212 virtual const SCEV *getExpr() const = 0;
215 inline raw_ostream &operator<<(raw_ostream &OS, const SCEVPredicate &P) {
220 // Specialize FoldingSetTrait for SCEVPredicate to avoid needing to compute
221 // temporary FoldingSetNodeID values.
223 struct FoldingSetTrait<SCEVPredicate>
224 : DefaultFoldingSetTrait<SCEVPredicate> {
226 static void Profile(const SCEVPredicate &X, FoldingSetNodeID &ID) {
230 static bool Equals(const SCEVPredicate &X, const FoldingSetNodeID &ID,
231 unsigned IDHash, FoldingSetNodeID &TempID) {
232 return ID == X.FastID;
234 static unsigned ComputeHash(const SCEVPredicate &X,
235 FoldingSetNodeID &TempID) {
236 return X.FastID.ComputeHash();
240 /// This class represents an assumption that two SCEV expressions are equal,
241 /// and this can be checked at run-time. We assume that the left hand side is
242 /// a SCEVUnknown and the right hand side a constant.
243 class SCEVEqualPredicate final : public SCEVPredicate {
244 /// We assume that LHS == RHS, where LHS is a SCEVUnknown and RHS a
246 const SCEVUnknown *LHS;
247 const SCEVConstant *RHS;
250 SCEVEqualPredicate(const FoldingSetNodeIDRef ID, const SCEVUnknown *LHS,
251 const SCEVConstant *RHS);
253 /// Implementation of the SCEVPredicate interface
254 bool implies(const SCEVPredicate *N) const override;
255 void print(raw_ostream &OS, unsigned Depth = 0) const override;
256 bool isAlwaysTrue() const override;
257 const SCEV *getExpr() const override;
259 /// Returns the left hand side of the equality.
260 const SCEVUnknown *getLHS() const { return LHS; }
262 /// Returns the right hand side of the equality.
263 const SCEVConstant *getRHS() const { return RHS; }
265 /// Methods for support type inquiry through isa, cast, and dyn_cast:
266 static inline bool classof(const SCEVPredicate *P) {
267 return P->getKind() == P_Equal;
271 /// This class represents an assumption made on an AddRec expression. Given an
272 /// affine AddRec expression {a,+,b}, we assume that it has the nssw or nusw
273 /// flags (defined below) in the first X iterations of the loop, where X is a
274 /// SCEV expression returned by getPredicatedBackedgeTakenCount).
276 /// Note that this does not imply that X is equal to the backedge taken
277 /// count. This means that if we have a nusw predicate for i32 {0,+,1} with a
278 /// predicated backedge taken count of X, we only guarantee that {0,+,1} has
279 /// nusw in the first X iterations. {0,+,1} may still wrap in the loop if we
280 /// have more than X iterations.
281 class SCEVWrapPredicate final : public SCEVPredicate {
283 /// Similar to SCEV::NoWrapFlags, but with slightly different semantics
284 /// for FlagNUSW. The increment is considered to be signed, and a + b
285 /// (where b is the increment) is considered to wrap if:
286 /// zext(a + b) != zext(a) + sext(b)
288 /// If Signed is a function that takes an n-bit tuple and maps to the
289 /// integer domain as the tuples value interpreted as twos complement,
290 /// and Unsigned a function that takes an n-bit tuple and maps to the
291 /// integer domain as as the base two value of input tuple, then a + b
292 /// has IncrementNUSW iff:
294 /// 0 <= Unsigned(a) + Signed(b) < 2^n
296 /// The IncrementNSSW flag has identical semantics with SCEV::FlagNSW.
298 /// Note that the IncrementNUSW flag is not commutative: if base + inc
299 /// has IncrementNUSW, then inc + base doesn't neccessarily have this
300 /// property. The reason for this is that this is used for sign/zero
301 /// extending affine AddRec SCEV expressions when a SCEVWrapPredicate is
302 /// assumed. A {base,+,inc} expression is already non-commutative with
303 /// regards to base and inc, since it is interpreted as:
304 /// (((base + inc) + inc) + inc) ...
305 enum IncrementWrapFlags {
306 IncrementAnyWrap = 0, // No guarantee.
307 IncrementNUSW = (1 << 0), // No unsigned with signed increment wrap.
308 IncrementNSSW = (1 << 1), // No signed with signed increment wrap
309 // (equivalent with SCEV::NSW)
310 IncrementNoWrapMask = (1 << 2) - 1
313 /// Convenient IncrementWrapFlags manipulation methods.
314 static SCEVWrapPredicate::IncrementWrapFlags LLVM_ATTRIBUTE_UNUSED_RESULT
315 clearFlags(SCEVWrapPredicate::IncrementWrapFlags Flags,
316 SCEVWrapPredicate::IncrementWrapFlags OffFlags) {
317 assert((Flags & IncrementNoWrapMask) == Flags && "Invalid flags value!");
318 assert((OffFlags & IncrementNoWrapMask) == OffFlags &&
319 "Invalid flags value!");
320 return (SCEVWrapPredicate::IncrementWrapFlags)(Flags & ~OffFlags);
323 static SCEVWrapPredicate::IncrementWrapFlags LLVM_ATTRIBUTE_UNUSED_RESULT
324 maskFlags(SCEVWrapPredicate::IncrementWrapFlags Flags, int Mask) {
325 assert((Flags & IncrementNoWrapMask) == Flags && "Invalid flags value!");
326 assert((Mask & IncrementNoWrapMask) == Mask && "Invalid mask value!");
328 return (SCEVWrapPredicate::IncrementWrapFlags)(Flags & Mask);
331 static SCEVWrapPredicate::IncrementWrapFlags LLVM_ATTRIBUTE_UNUSED_RESULT
332 setFlags(SCEVWrapPredicate::IncrementWrapFlags Flags,
333 SCEVWrapPredicate::IncrementWrapFlags OnFlags) {
334 assert((Flags & IncrementNoWrapMask) == Flags && "Invalid flags value!");
335 assert((OnFlags & IncrementNoWrapMask) == OnFlags &&
336 "Invalid flags value!");
338 return (SCEVWrapPredicate::IncrementWrapFlags)(Flags | OnFlags);
341 /// Returns the set of SCEVWrapPredicate no wrap flags implied by a
343 static SCEVWrapPredicate::IncrementWrapFlags
344 getImpliedFlags(const SCEVAddRecExpr *AR, ScalarEvolution &SE);
347 const SCEVAddRecExpr *AR;
348 IncrementWrapFlags Flags;
351 explicit SCEVWrapPredicate(const FoldingSetNodeIDRef ID,
352 const SCEVAddRecExpr *AR,
353 IncrementWrapFlags Flags);
355 /// Returns the set assumed no overflow flags.
356 IncrementWrapFlags getFlags() const { return Flags; }
357 /// Implementation of the SCEVPredicate interface
358 const SCEV *getExpr() const override;
359 bool implies(const SCEVPredicate *N) const override;
360 void print(raw_ostream &OS, unsigned Depth = 0) const override;
361 bool isAlwaysTrue() const override;
363 /// Methods for support type inquiry through isa, cast, and dyn_cast:
364 static inline bool classof(const SCEVPredicate *P) {
365 return P->getKind() == P_Wrap;
369 /// This class represents a composition of other SCEV predicates, and is the
370 /// class that most clients will interact with. This is equivalent to a
371 /// logical "AND" of all the predicates in the union.
372 class SCEVUnionPredicate final : public SCEVPredicate {
374 typedef DenseMap<const SCEV *, SmallVector<const SCEVPredicate *, 4>>
377 /// Vector with references to all predicates in this union.
378 SmallVector<const SCEVPredicate *, 16> Preds;
379 /// Maps SCEVs to predicates for quick look-ups.
380 PredicateMap SCEVToPreds;
383 SCEVUnionPredicate();
385 const SmallVectorImpl<const SCEVPredicate *> &getPredicates() const {
389 /// Adds a predicate to this union.
390 void add(const SCEVPredicate *N);
392 /// Returns a reference to a vector containing all predicates which apply to
394 ArrayRef<const SCEVPredicate *> getPredicatesForExpr(const SCEV *Expr);
396 /// Implementation of the SCEVPredicate interface
397 bool isAlwaysTrue() const override;
398 bool implies(const SCEVPredicate *N) const override;
399 void print(raw_ostream &OS, unsigned Depth) const override;
400 const SCEV *getExpr() const override;
402 /// We estimate the complexity of a union predicate as the size number of
403 /// predicates in the union.
404 unsigned getComplexity() const override { return Preds.size(); }
406 /// Methods for support type inquiry through isa, cast, and dyn_cast:
407 static inline bool classof(const SCEVPredicate *P) {
408 return P->getKind() == P_Union;
412 /// The main scalar evolution driver. Because client code (intentionally)
413 /// can't do much with the SCEV objects directly, they must ask this class
415 class ScalarEvolution {
417 /// An enum describing the relationship between a SCEV and a loop.
418 enum LoopDisposition {
419 LoopVariant, ///< The SCEV is loop-variant (unknown).
420 LoopInvariant, ///< The SCEV is loop-invariant.
421 LoopComputable ///< The SCEV varies predictably with the loop.
424 /// An enum describing the relationship between a SCEV and a basic block.
425 enum BlockDisposition {
426 DoesNotDominateBlock, ///< The SCEV does not dominate the block.
427 DominatesBlock, ///< The SCEV dominates the block.
428 ProperlyDominatesBlock ///< The SCEV properly dominates the block.
431 /// Convenient NoWrapFlags manipulation that hides enum casts and is
432 /// visible in the ScalarEvolution name space.
433 static SCEV::NoWrapFlags LLVM_ATTRIBUTE_UNUSED_RESULT
434 maskFlags(SCEV::NoWrapFlags Flags, int Mask) {
435 return (SCEV::NoWrapFlags)(Flags & Mask);
437 static SCEV::NoWrapFlags LLVM_ATTRIBUTE_UNUSED_RESULT
438 setFlags(SCEV::NoWrapFlags Flags, SCEV::NoWrapFlags OnFlags) {
439 return (SCEV::NoWrapFlags)(Flags | OnFlags);
441 static SCEV::NoWrapFlags LLVM_ATTRIBUTE_UNUSED_RESULT
442 clearFlags(SCEV::NoWrapFlags Flags, SCEV::NoWrapFlags OffFlags) {
443 return (SCEV::NoWrapFlags)(Flags & ~OffFlags);
447 /// A CallbackVH to arrange for ScalarEvolution to be notified whenever a
448 /// Value is deleted.
449 class SCEVCallbackVH final : public CallbackVH {
451 void deleted() override;
452 void allUsesReplacedWith(Value *New) override;
454 SCEVCallbackVH(Value *V, ScalarEvolution *SE = nullptr);
457 friend class SCEVCallbackVH;
458 friend class SCEVExpander;
459 friend class SCEVUnknown;
461 /// The function we are analyzing.
465 /// Does the module have any calls to the llvm.experimental.guard intrinsic
466 /// at all? If this is false, we avoid doing work that will only help if
467 /// thare are guards present in the IR.
471 /// The target library information for the target we are targeting.
473 TargetLibraryInfo &TLI;
475 /// The tracker for @llvm.assume intrinsics in this function.
478 /// The dominator tree.
482 /// The loop information for the function we are currently analyzing.
486 /// This SCEV is used to represent unknown trip counts and things.
487 std::unique_ptr<SCEVCouldNotCompute> CouldNotCompute;
489 /// The typedef for HasRecMap.
491 typedef DenseMap<const SCEV *, bool> HasRecMapType;
493 /// This is a cache to record whether a SCEV contains any scAddRecExpr.
494 HasRecMapType HasRecMap;
496 /// The typedef for ExprValueMap.
498 typedef std::pair<Value *, ConstantInt *> ValueOffsetPair;
499 typedef DenseMap<const SCEV *, SetVector<ValueOffsetPair>> ExprValueMapType;
501 /// ExprValueMap -- This map records the original values from which
502 /// the SCEV expr is generated from.
504 /// We want to represent the mapping as SCEV -> ValueOffsetPair instead
505 /// of SCEV -> Value:
506 /// Suppose we know S1 expands to V1, and
509 /// where C_a and C_b are different SCEVConstants. Then we'd like to
510 /// expand S3 as V1 - C_a + C_b instead of expanding S2 literally.
511 /// It is helpful when S2 is a complex SCEV expr.
513 /// In order to do that, we represent ExprValueMap as a mapping from
514 /// SCEV to ValueOffsetPair. We will save both S1->{V1, 0} and
515 /// S2->{V1, C_a} into the map when we create SCEV for V1. When S3
516 /// is expanded, it will first expand S2 to V1 - C_a because of
517 /// S2->{V1, C_a} in the map, then expand S3 to V1 - C_a + C_b.
519 /// Note: S->{V, Offset} in the ExprValueMap means S can be expanded
521 ExprValueMapType ExprValueMap;
523 /// The typedef for ValueExprMap.
525 typedef DenseMap<SCEVCallbackVH, const SCEV *, DenseMapInfo<Value *> >
528 /// This is a cache of the values we have analyzed so far.
530 ValueExprMapType ValueExprMap;
532 /// Mark predicate values currently being processed by isImpliedCond.
533 DenseSet<Value*> PendingLoopPredicates;
535 /// Set to true by isLoopBackedgeGuardedByCond when we're walking the set of
536 /// conditions dominating the backedge of a loop.
537 bool WalkingBEDominatingConds;
539 /// Set to true by isKnownPredicateViaSplitting when we're trying to prove a
540 /// predicate by splitting it into a set of independent predicates.
541 bool ProvingSplitPredicate;
543 /// Information about the number of loop iterations for which a loop exit's
544 /// branch condition evaluates to the not-taken path. This is a temporary
545 /// pair of exact and max expressions that are eventually summarized in
546 /// ExitNotTakenInfo and BackedgeTakenInfo.
551 /// A predicate union guard for this ExitLimit. The result is only
552 /// valid if this predicate evaluates to 'true' at run-time.
553 SCEVUnionPredicate Pred;
555 /*implicit*/ ExitLimit(const SCEV *E) : Exact(E), Max(E) {}
557 ExitLimit(const SCEV *E, const SCEV *M, SCEVUnionPredicate &P)
558 : Exact(E), Max(M), Pred(P) {
559 assert((isa<SCEVCouldNotCompute>(Exact) ||
560 !isa<SCEVCouldNotCompute>(Max)) &&
561 "Exact is not allowed to be less precise than Max");
564 /// Test whether this ExitLimit contains any computed information, or
565 /// whether it's all SCEVCouldNotCompute values.
566 bool hasAnyInfo() const {
567 return !isa<SCEVCouldNotCompute>(Exact) ||
568 !isa<SCEVCouldNotCompute>(Max);
571 /// Test whether this ExitLimit contains all information.
572 bool hasFullInfo() const { return !isa<SCEVCouldNotCompute>(Exact); }
575 /// Forward declaration of ExitNotTakenExtras
576 struct ExitNotTakenExtras;
578 /// Information about the number of times a particular loop exit may be
579 /// reached before exiting the loop.
580 struct ExitNotTakenInfo {
581 AssertingVH<BasicBlock> ExitingBlock;
582 const SCEV *ExactNotTaken;
584 ExitNotTakenExtras *ExtraInfo;
588 : ExitingBlock(nullptr), ExactNotTaken(nullptr), ExtraInfo(nullptr),
591 ExitNotTakenInfo(BasicBlock *ExitBlock, const SCEV *Expr,
592 ExitNotTakenExtras *Ptr)
593 : ExitingBlock(ExitBlock), ExactNotTaken(Expr), ExtraInfo(Ptr),
596 /// Return true if all loop exits are computable.
597 bool isCompleteList() const { return Complete; }
599 /// Sets the incomplete property, indicating that one of the loop exits
600 /// doesn't have a corresponding ExitNotTakenInfo entry.
601 void setIncomplete() { Complete = false; }
603 /// Returns a pointer to the predicate associated with this information,
604 /// or nullptr if this doesn't exist (meaning always true).
605 SCEVUnionPredicate *getPred() const {
607 return &ExtraInfo->Pred;
612 /// Return true if the SCEV predicate associated with this information
614 bool hasAlwaysTruePred() const {
615 return !getPred() || getPred()->isAlwaysTrue();
618 /// Defines a simple forward iterator for ExitNotTakenInfo.
619 class ExitNotTakenInfoIterator
620 : public std::iterator<std::forward_iterator_tag, ExitNotTakenInfo> {
621 const ExitNotTakenInfo *Start;
625 ExitNotTakenInfoIterator(const ExitNotTakenInfo *Start,
627 : Start(Start), Position(Position) {}
629 const ExitNotTakenInfo &operator*() const {
633 return Start->ExtraInfo->Exits[Position - 1];
636 const ExitNotTakenInfo *operator->() const {
640 return &Start->ExtraInfo->Exits[Position - 1];
643 bool operator==(const ExitNotTakenInfoIterator &RHS) const {
644 return Start == RHS.Start && Position == RHS.Position;
647 bool operator!=(const ExitNotTakenInfoIterator &RHS) const {
648 return Start != RHS.Start || Position != RHS.Position;
651 ExitNotTakenInfoIterator &operator++() { // Preincrement
656 Start->ExtraInfo ? Start->ExtraInfo->Exits.size() + 1 : 1;
660 // We've run out of elements.
661 if (Position == Elements) {
668 ExitNotTakenInfoIterator operator++(int) { // Postincrement
669 ExitNotTakenInfoIterator Tmp = *this;
676 ExitNotTakenInfoIterator begin() const {
677 return ExitNotTakenInfoIterator(this, 0);
679 ExitNotTakenInfoIterator end() const {
680 return ExitNotTakenInfoIterator(nullptr, 0);
684 /// Describes the extra information that a ExitNotTakenInfo can have.
685 struct ExitNotTakenExtras {
686 /// The predicate associated with the ExitNotTakenInfo struct.
687 SCEVUnionPredicate Pred;
689 /// The extra exits in the loop. Only the ExitNotTakenExtras structure
690 /// pointed to by the first ExitNotTakenInfo struct (associated with the
691 /// first loop exit) will populate this vector to prevent having
692 /// redundant information.
693 SmallVector<ExitNotTakenInfo, 4> Exits;
696 /// A struct containing the information attached to a backedge.
698 EdgeInfo(BasicBlock *Block, const SCEV *Taken, SCEVUnionPredicate &P) :
699 ExitBlock(Block), Taken(Taken), Pred(std::move(P)) {}
701 /// The exit basic block.
702 BasicBlock *ExitBlock;
704 /// The (exact) number of time we take the edge back.
707 /// The SCEV predicated associated with Taken. If Pred doesn't evaluate
708 /// to true, the information in Taken is not valid (or equivalent with
709 /// a CouldNotCompute.
710 SCEVUnionPredicate Pred;
713 /// Information about the backedge-taken count of a loop. This currently
714 /// includes an exact count and a maximum count.
716 class BackedgeTakenInfo {
717 /// A list of computable exits and their not-taken counts. Loops almost
718 /// never have more than one computable exit.
719 ExitNotTakenInfo ExitNotTaken;
721 /// An expression indicating the least maximum backedge-taken count of the
722 /// loop that is known, or a SCEVCouldNotCompute. This expression is only
723 /// valid if the predicates associated with all loop exits are true.
727 BackedgeTakenInfo() : Max(nullptr) {}
729 /// Initialize BackedgeTakenInfo from a list of exact exit counts.
730 BackedgeTakenInfo(SmallVectorImpl<EdgeInfo> &ExitCounts, bool Complete,
731 const SCEV *MaxCount);
733 /// Test whether this BackedgeTakenInfo contains any computed information,
734 /// or whether it's all SCEVCouldNotCompute values.
735 bool hasAnyInfo() const {
736 return ExitNotTaken.ExitingBlock || !isa<SCEVCouldNotCompute>(Max);
739 /// Test whether this BackedgeTakenInfo contains complete information.
740 bool hasFullInfo() const { return ExitNotTaken.isCompleteList(); }
742 /// Return an expression indicating the exact backedge-taken count of the
743 /// loop if it is known or SCEVCouldNotCompute otherwise. This is the
744 /// number of times the loop header can be guaranteed to execute, minus
747 /// If the SCEV predicate associated with the answer can be different
748 /// from AlwaysTrue, we must add a (non null) Predicates argument.
749 /// The SCEV predicate associated with the answer will be added to
750 /// Predicates. A run-time check needs to be emitted for the SCEV
751 /// predicate in order for the answer to be valid.
753 /// Note that we should always know if we need to pass a predicate
754 /// argument or not from the way the ExitCounts vector was computed.
755 /// If we allowed SCEV predicates to be generated when populating this
756 /// vector, this information can contain them and therefore a
757 /// SCEVPredicate argument should be added to getExact.
758 const SCEV *getExact(ScalarEvolution *SE,
759 SCEVUnionPredicate *Predicates = nullptr) const;
761 /// Return the number of times this loop exit may fall through to the back
762 /// edge, or SCEVCouldNotCompute. The loop is guaranteed not to exit via
763 /// this block before this number of iterations, but may exit via another
765 const SCEV *getExact(BasicBlock *ExitingBlock, ScalarEvolution *SE) const;
767 /// Get the max backedge taken count for the loop.
768 const SCEV *getMax(ScalarEvolution *SE) const;
770 /// Return true if any backedge taken count expressions refer to the given
772 bool hasOperand(const SCEV *S, ScalarEvolution *SE) const;
774 /// Invalidate this result and free associated memory.
778 /// Cache the backedge-taken count of the loops for this function as they
780 DenseMap<const Loop *, BackedgeTakenInfo> BackedgeTakenCounts;
782 /// Cache the predicated backedge-taken count of the loops for this
783 /// function as they are computed.
784 DenseMap<const Loop *, BackedgeTakenInfo> PredicatedBackedgeTakenCounts;
786 /// This map contains entries for all of the PHI instructions that we
787 /// attempt to compute constant evolutions for. This allows us to avoid
788 /// potentially expensive recomputation of these properties. An instruction
789 /// maps to null if we are unable to compute its exit value.
790 DenseMap<PHINode*, Constant*> ConstantEvolutionLoopExitValue;
792 /// This map contains entries for all the expressions that we attempt to
793 /// compute getSCEVAtScope information for, which can be expensive in
795 DenseMap<const SCEV *,
796 SmallVector<std::pair<const Loop *, const SCEV *>, 2> > ValuesAtScopes;
798 /// Memoized computeLoopDisposition results.
799 DenseMap<const SCEV *,
800 SmallVector<PointerIntPair<const Loop *, 2, LoopDisposition>, 2>>
803 /// Cache for \c loopHasNoAbnormalExits.
804 DenseMap<const Loop *, bool> LoopHasNoAbnormalExits;
806 /// Returns true if \p L contains no instruction that can abnormally exit
807 /// the loop (i.e. via throwing an exception, by terminating the thread
808 /// cleanly or by infinite looping in a called function). Strictly
809 /// speaking, the last one is not leaving the loop, but is identical to
810 /// leaving the loop for reasoning about undefined behavior.
811 bool loopHasNoAbnormalExits(const Loop *L);
813 /// Compute a LoopDisposition value.
814 LoopDisposition computeLoopDisposition(const SCEV *S, const Loop *L);
816 /// Memoized computeBlockDisposition results.
819 SmallVector<PointerIntPair<const BasicBlock *, 2, BlockDisposition>, 2>>
822 /// Compute a BlockDisposition value.
823 BlockDisposition computeBlockDisposition(const SCEV *S, const BasicBlock *BB);
825 /// Memoized results from getRange
826 DenseMap<const SCEV *, ConstantRange> UnsignedRanges;
828 /// Memoized results from getRange
829 DenseMap<const SCEV *, ConstantRange> SignedRanges;
831 /// Used to parameterize getRange
832 enum RangeSignHint { HINT_RANGE_UNSIGNED, HINT_RANGE_SIGNED };
834 /// Set the memoized range for the given SCEV.
835 const ConstantRange &setRange(const SCEV *S, RangeSignHint Hint,
836 const ConstantRange &CR) {
837 DenseMap<const SCEV *, ConstantRange> &Cache =
838 Hint == HINT_RANGE_UNSIGNED ? UnsignedRanges : SignedRanges;
840 auto Pair = Cache.insert({S, CR});
842 Pair.first->second = CR;
843 return Pair.first->second;
846 /// Determine the range for a particular SCEV.
847 ConstantRange getRange(const SCEV *S, RangeSignHint Hint);
849 /// Determines the range for the affine SCEVAddRecExpr {\p Start,+,\p Stop}.
850 /// Helper for \c getRange.
851 ConstantRange getRangeForAffineAR(const SCEV *Start, const SCEV *Stop,
852 const SCEV *MaxBECount,
855 /// Try to compute a range for the affine SCEVAddRecExpr {\p Start,+,\p
856 /// Stop} by "factoring out" a ternary expression from the add recurrence.
857 /// Helper called by \c getRange.
858 ConstantRange getRangeViaFactoring(const SCEV *Start, const SCEV *Stop,
859 const SCEV *MaxBECount,
862 /// We know that there is no SCEV for the specified value. Analyze the
864 const SCEV *createSCEV(Value *V);
866 /// Provide the special handling we need to analyze PHI SCEVs.
867 const SCEV *createNodeForPHI(PHINode *PN);
869 /// Helper function called from createNodeForPHI.
870 const SCEV *createAddRecFromPHI(PHINode *PN);
872 /// Helper function called from createNodeForPHI.
873 const SCEV *createNodeFromSelectLikePHI(PHINode *PN);
875 /// Provide special handling for a select-like instruction (currently this
876 /// is either a select instruction or a phi node). \p I is the instruction
877 /// being processed, and it is assumed equivalent to "Cond ? TrueVal :
879 const SCEV *createNodeForSelectOrPHI(Instruction *I, Value *Cond,
880 Value *TrueVal, Value *FalseVal);
882 /// Provide the special handling we need to analyze GEP SCEVs.
883 const SCEV *createNodeForGEP(GEPOperator *GEP);
885 /// Implementation code for getSCEVAtScope; called at most once for each
888 const SCEV *computeSCEVAtScope(const SCEV *S, const Loop *L);
890 /// This looks up computed SCEV values for all instructions that depend on
891 /// the given instruction and removes them from the ValueExprMap map if they
892 /// reference SymName. This is used during PHI resolution.
893 void forgetSymbolicName(Instruction *I, const SCEV *SymName);
895 /// Return the BackedgeTakenInfo for the given loop, lazily computing new
896 /// values if the loop hasn't been analyzed yet. The returned result is
897 /// guaranteed not to be predicated.
898 const BackedgeTakenInfo &getBackedgeTakenInfo(const Loop *L);
900 /// Similar to getBackedgeTakenInfo, but will add predicates as required
901 /// with the purpose of returning complete information.
902 const BackedgeTakenInfo &getPredicatedBackedgeTakenInfo(const Loop *L);
904 /// Compute the number of times the specified loop will iterate.
905 /// If AllowPredicates is set, we will create new SCEV predicates as
906 /// necessary in order to return an exact answer.
907 BackedgeTakenInfo computeBackedgeTakenCount(const Loop *L,
908 bool AllowPredicates = false);
910 /// Compute the number of times the backedge of the specified loop will
911 /// execute if it exits via the specified block. If AllowPredicates is set,
912 /// this call will try to use a minimal set of SCEV predicates in order to
913 /// return an exact answer.
914 ExitLimit computeExitLimit(const Loop *L, BasicBlock *ExitingBlock,
915 bool AllowPredicates = false);
917 /// Compute the number of times the backedge of the specified loop will
918 /// execute if its exit condition were a conditional branch of ExitCond,
921 /// \p ControlsExit is true if ExitCond directly controls the exit
922 /// branch. In this case, we can assume that the loop exits only if the
923 /// condition is true and can infer that failing to meet the condition prior
924 /// to integer wraparound results in undefined behavior.
926 /// If \p AllowPredicates is set, this call will try to use a minimal set of
927 /// SCEV predicates in order to return an exact answer.
928 ExitLimit computeExitLimitFromCond(const Loop *L,
933 bool AllowPredicates = false);
935 /// Compute the number of times the backedge of the specified loop will
936 /// execute if its exit condition were a conditional branch of the ICmpInst
937 /// ExitCond, TBB, and FBB. If AllowPredicates is set, this call will try
938 /// to use a minimal set of SCEV predicates in order to return an exact
940 ExitLimit computeExitLimitFromICmp(const Loop *L,
945 bool AllowPredicates = false);
947 /// Compute the number of times the backedge of the specified loop will
948 /// execute if its exit condition were a switch with a single exiting case
951 computeExitLimitFromSingleExitSwitch(const Loop *L, SwitchInst *Switch,
952 BasicBlock *ExitingBB, bool IsSubExpr);
954 /// Given an exit condition of 'icmp op load X, cst', try to see if we can
955 /// compute the backedge-taken count.
956 ExitLimit computeLoadConstantCompareExitLimit(LoadInst *LI,
959 ICmpInst::Predicate p);
961 /// Compute the exit limit of a loop that is controlled by a
962 /// "(IV >> 1) != 0" type comparison. We cannot compute the exact trip
963 /// count in these cases (since SCEV has no way of expressing them), but we
964 /// can still sometimes compute an upper bound.
966 /// Return an ExitLimit for a loop whose backedge is guarded by `LHS Pred
968 ExitLimit computeShiftCompareExitLimit(Value *LHS, Value *RHS,
970 ICmpInst::Predicate Pred);
972 /// If the loop is known to execute a constant number of times (the
973 /// condition evolves only from constants), try to evaluate a few iterations
974 /// of the loop until we get the exit condition gets a value of ExitWhen
975 /// (true or false). If we cannot evaluate the exit count of the loop,
976 /// return CouldNotCompute.
977 const SCEV *computeExitCountExhaustively(const Loop *L,
981 /// Return the number of times an exit condition comparing the specified
982 /// value to zero will execute. If not computable, return CouldNotCompute.
983 /// If AllowPredicates is set, this call will try to use a minimal set of
984 /// SCEV predicates in order to return an exact answer.
985 ExitLimit howFarToZero(const SCEV *V, const Loop *L, bool IsSubExpr,
986 bool AllowPredicates = false);
988 /// Return the number of times an exit condition checking the specified
989 /// value for nonzero will execute. If not computable, return
991 ExitLimit howFarToNonZero(const SCEV *V, const Loop *L);
993 /// Return the number of times an exit condition containing the specified
994 /// less-than comparison will execute. If not computable, return
997 /// \p isSigned specifies whether the less-than is signed.
999 /// \p ControlsExit is true when the LHS < RHS condition directly controls
1000 /// the branch (loops exits only if condition is true). In this case, we can
1001 /// use NoWrapFlags to skip overflow checks.
1003 /// If \p AllowPredicates is set, this call will try to use a minimal set of
1004 /// SCEV predicates in order to return an exact answer.
1005 ExitLimit howManyLessThans(const SCEV *LHS, const SCEV *RHS, const Loop *L,
1006 bool isSigned, bool ControlsExit,
1007 bool AllowPredicates = false);
1009 ExitLimit howManyGreaterThans(const SCEV *LHS, const SCEV *RHS,
1010 const Loop *L, bool isSigned, bool IsSubExpr,
1011 bool AllowPredicates = false);
1013 /// Return a predecessor of BB (which may not be an immediate predecessor)
1014 /// which has exactly one successor from which BB is reachable, or null if
1015 /// no such block is found.
1016 std::pair<BasicBlock *, BasicBlock *>
1017 getPredecessorWithUniqueSuccessorForBB(BasicBlock *BB);
1019 /// Test whether the condition described by Pred, LHS, and RHS is true
1020 /// whenever the given FoundCondValue value evaluates to true.
1021 bool isImpliedCond(ICmpInst::Predicate Pred,
1022 const SCEV *LHS, const SCEV *RHS,
1023 Value *FoundCondValue,
1026 /// Test whether the condition described by Pred, LHS, and RHS is true
1027 /// whenever the condition described by FoundPred, FoundLHS, FoundRHS is
1029 bool isImpliedCond(ICmpInst::Predicate Pred, const SCEV *LHS,
1030 const SCEV *RHS, ICmpInst::Predicate FoundPred,
1031 const SCEV *FoundLHS, const SCEV *FoundRHS);
1033 /// Test whether the condition described by Pred, LHS, and RHS is true
1034 /// whenever the condition described by Pred, FoundLHS, and FoundRHS is
1036 bool isImpliedCondOperands(ICmpInst::Predicate Pred,
1037 const SCEV *LHS, const SCEV *RHS,
1038 const SCEV *FoundLHS, const SCEV *FoundRHS);
1040 /// Test whether the condition described by Pred, LHS, and RHS is true
1041 /// whenever the condition described by Pred, FoundLHS, and FoundRHS is
1043 bool isImpliedCondOperandsHelper(ICmpInst::Predicate Pred,
1044 const SCEV *LHS, const SCEV *RHS,
1045 const SCEV *FoundLHS,
1046 const SCEV *FoundRHS);
1048 /// Test whether the condition described by Pred, LHS, and RHS is true
1049 /// whenever the condition described by Pred, FoundLHS, and FoundRHS is
1050 /// true. Utility function used by isImpliedCondOperands. Tries to get
1051 /// cases like "X `sgt` 0 => X - 1 `sgt` -1".
1052 bool isImpliedCondOperandsViaRanges(ICmpInst::Predicate Pred,
1053 const SCEV *LHS, const SCEV *RHS,
1054 const SCEV *FoundLHS,
1055 const SCEV *FoundRHS);
1057 /// Return true if the condition denoted by \p LHS \p Pred \p RHS is implied
1058 /// by a call to \c @llvm.experimental.guard in \p BB.
1059 bool isImpliedViaGuard(BasicBlock *BB, ICmpInst::Predicate Pred,
1060 const SCEV *LHS, const SCEV *RHS);
1062 /// Test whether the condition described by Pred, LHS, and RHS is true
1063 /// whenever the condition described by Pred, FoundLHS, and FoundRHS is
1066 /// This routine tries to rule out certain kinds of integer overflow, and
1067 /// then tries to reason about arithmetic properties of the predicates.
1068 bool isImpliedCondOperandsViaNoOverflow(ICmpInst::Predicate Pred,
1069 const SCEV *LHS, const SCEV *RHS,
1070 const SCEV *FoundLHS,
1071 const SCEV *FoundRHS);
1073 /// If we know that the specified Phi is in the header of its containing
1074 /// loop, we know the loop executes a constant number of times, and the PHI
1075 /// node is just a recurrence involving constants, fold it.
1076 Constant *getConstantEvolutionLoopExitValue(PHINode *PN, const APInt& BEs,
1079 /// Test if the given expression is known to satisfy the condition described
1080 /// by Pred and the known constant ranges of LHS and RHS.
1082 bool isKnownPredicateViaConstantRanges(ICmpInst::Predicate Pred,
1083 const SCEV *LHS, const SCEV *RHS);
1085 /// Try to prove the condition described by "LHS Pred RHS" by ruling out
1086 /// integer overflow.
1088 /// For instance, this will return true for "A s< (A + C)<nsw>" if C is
1090 bool isKnownPredicateViaNoOverflow(ICmpInst::Predicate Pred,
1091 const SCEV *LHS, const SCEV *RHS);
1093 /// Try to split Pred LHS RHS into logical conjunctions (and's) and try to
1094 /// prove them individually.
1095 bool isKnownPredicateViaSplitting(ICmpInst::Predicate Pred, const SCEV *LHS,
1098 /// Try to match the Expr as "(L + R)<Flags>".
1099 bool splitBinaryAdd(const SCEV *Expr, const SCEV *&L, const SCEV *&R,
1100 SCEV::NoWrapFlags &Flags);
1102 /// Return true if More == (Less + C), where C is a constant. This is
1103 /// intended to be used as a cheaper substitute for full SCEV subtraction.
1104 bool computeConstantDifference(const SCEV *Less, const SCEV *More,
1107 /// Drop memoized information computed for S.
1108 void forgetMemoizedResults(const SCEV *S);
1110 /// Return an existing SCEV for V if there is one, otherwise return nullptr.
1111 const SCEV *getExistingSCEV(Value *V);
1113 /// Return false iff given SCEV contains a SCEVUnknown with NULL value-
1115 bool checkValidity(const SCEV *S) const;
1117 /// Return true if `ExtendOpTy`({`Start`,+,`Step`}) can be proved to be
1118 /// equal to {`ExtendOpTy`(`Start`),+,`ExtendOpTy`(`Step`)}. This is
1119 /// equivalent to proving no signed (resp. unsigned) wrap in
1120 /// {`Start`,+,`Step`} if `ExtendOpTy` is `SCEVSignExtendExpr`
1121 /// (resp. `SCEVZeroExtendExpr`).
1123 template<typename ExtendOpTy>
1124 bool proveNoWrapByVaryingStart(const SCEV *Start, const SCEV *Step,
1127 /// Try to prove NSW or NUW on \p AR relying on ConstantRange manipulation.
1128 SCEV::NoWrapFlags proveNoWrapViaConstantRanges(const SCEVAddRecExpr *AR);
1130 bool isMonotonicPredicateImpl(const SCEVAddRecExpr *LHS,
1131 ICmpInst::Predicate Pred, bool &Increasing);
1133 /// Return true if, for all loop invariant X, the predicate "LHS `Pred` X"
1134 /// is monotonically increasing or decreasing. In the former case set
1135 /// `Increasing` to true and in the latter case set `Increasing` to false.
1137 /// A predicate is said to be monotonically increasing if may go from being
1138 /// false to being true as the loop iterates, but never the other way
1139 /// around. A predicate is said to be monotonically decreasing if may go
1140 /// from being true to being false as the loop iterates, but never the other
1142 bool isMonotonicPredicate(const SCEVAddRecExpr *LHS,
1143 ICmpInst::Predicate Pred, bool &Increasing);
1145 /// Return SCEV no-wrap flags that can be proven based on reasoning about
1146 /// how poison produced from no-wrap flags on this value (e.g. a nuw add)
1147 /// would trigger undefined behavior on overflow.
1148 SCEV::NoWrapFlags getNoWrapFlagsFromUB(const Value *V);
1150 /// Return true if the SCEV corresponding to \p I is never poison. Proving
1151 /// this is more complex than proving that just \p I is never poison, since
1152 /// SCEV commons expressions across control flow, and you can have cases
1156 /// ptr[idx0] = 100;
1157 /// if (<condition>) {
1158 /// idx1 = a +nsw b;
1159 /// ptr[idx1] = 200;
1162 /// where the SCEV expression (+ a b) is guaranteed to not be poison (and
1163 /// hence not sign-overflow) only if "<condition>" is true. Since both
1164 /// `idx0` and `idx1` will be mapped to the same SCEV expression, (+ a b),
1165 /// it is not okay to annotate (+ a b) with <nsw> in the above example.
1166 bool isSCEVExprNeverPoison(const Instruction *I);
1168 /// This is like \c isSCEVExprNeverPoison but it specifically works for
1169 /// instructions that will get mapped to SCEV add recurrences. Return true
1170 /// if \p I will never generate poison under the assumption that \p I is an
1171 /// add recurrence on the loop \p L.
1172 bool isAddRecNeverPoison(const Instruction *I, const Loop *L);
1175 ScalarEvolution(Function &F, TargetLibraryInfo &TLI, AssumptionCache &AC,
1176 DominatorTree &DT, LoopInfo &LI);
1178 ScalarEvolution(ScalarEvolution &&Arg);
1180 LLVMContext &getContext() const { return F.getContext(); }
1182 /// Test if values of the given type are analyzable within the SCEV
1183 /// framework. This primarily includes integer types, and it can optionally
1184 /// include pointer types if the ScalarEvolution class has access to
1185 /// target-specific information.
1186 bool isSCEVable(Type *Ty) const;
1188 /// Return the size in bits of the specified type, for which isSCEVable must
1190 uint64_t getTypeSizeInBits(Type *Ty) const;
1192 /// Return a type with the same bitwidth as the given type and which
1193 /// represents how SCEV will treat the given type, for which isSCEVable must
1194 /// return true. For pointer types, this is the pointer-sized integer type.
1195 Type *getEffectiveSCEVType(Type *Ty) const;
1197 /// Return true if the SCEV is a scAddRecExpr or it contains
1198 /// scAddRecExpr. The result will be cached in HasRecMap.
1200 bool containsAddRecurrence(const SCEV *S);
1202 /// Return the Value set from which the SCEV expr is generated.
1203 SetVector<ValueOffsetPair> *getSCEVValues(const SCEV *S);
1205 /// Erase Value from ValueExprMap and ExprValueMap.
1206 void eraseValueFromMap(Value *V);
1208 /// Return a SCEV expression for the full generality of the specified
1210 const SCEV *getSCEV(Value *V);
1212 const SCEV *getConstant(ConstantInt *V);
1213 const SCEV *getConstant(const APInt& Val);
1214 const SCEV *getConstant(Type *Ty, uint64_t V, bool isSigned = false);
1215 const SCEV *getTruncateExpr(const SCEV *Op, Type *Ty);
1216 const SCEV *getZeroExtendExpr(const SCEV *Op, Type *Ty);
1217 const SCEV *getSignExtendExpr(const SCEV *Op, Type *Ty);
1218 const SCEV *getAnyExtendExpr(const SCEV *Op, Type *Ty);
1219 const SCEV *getAddExpr(SmallVectorImpl<const SCEV *> &Ops,
1220 SCEV::NoWrapFlags Flags = SCEV::FlagAnyWrap);
1221 const SCEV *getAddExpr(const SCEV *LHS, const SCEV *RHS,
1222 SCEV::NoWrapFlags Flags = SCEV::FlagAnyWrap) {
1223 SmallVector<const SCEV *, 2> Ops = {LHS, RHS};
1224 return getAddExpr(Ops, Flags);
1226 const SCEV *getAddExpr(const SCEV *Op0, const SCEV *Op1, const SCEV *Op2,
1227 SCEV::NoWrapFlags Flags = SCEV::FlagAnyWrap) {
1228 SmallVector<const SCEV *, 3> Ops = {Op0, Op1, Op2};
1229 return getAddExpr(Ops, Flags);
1231 const SCEV *getMulExpr(SmallVectorImpl<const SCEV *> &Ops,
1232 SCEV::NoWrapFlags Flags = SCEV::FlagAnyWrap);
1233 const SCEV *getMulExpr(const SCEV *LHS, const SCEV *RHS,
1234 SCEV::NoWrapFlags Flags = SCEV::FlagAnyWrap) {
1235 SmallVector<const SCEV *, 2> Ops = {LHS, RHS};
1236 return getMulExpr(Ops, Flags);
1238 const SCEV *getMulExpr(const SCEV *Op0, const SCEV *Op1, const SCEV *Op2,
1239 SCEV::NoWrapFlags Flags = SCEV::FlagAnyWrap) {
1240 SmallVector<const SCEV *, 3> Ops = {Op0, Op1, Op2};
1241 return getMulExpr(Ops, Flags);
1243 const SCEV *getUDivExpr(const SCEV *LHS, const SCEV *RHS);
1244 const SCEV *getUDivExactExpr(const SCEV *LHS, const SCEV *RHS);
1245 const SCEV *getAddRecExpr(const SCEV *Start, const SCEV *Step,
1246 const Loop *L, SCEV::NoWrapFlags Flags);
1247 const SCEV *getAddRecExpr(SmallVectorImpl<const SCEV *> &Operands,
1248 const Loop *L, SCEV::NoWrapFlags Flags);
1249 const SCEV *getAddRecExpr(const SmallVectorImpl<const SCEV *> &Operands,
1250 const Loop *L, SCEV::NoWrapFlags Flags) {
1251 SmallVector<const SCEV *, 4> NewOp(Operands.begin(), Operands.end());
1252 return getAddRecExpr(NewOp, L, Flags);
1254 /// Returns an expression for a GEP
1256 /// \p PointeeType The type used as the basis for the pointer arithmetics
1257 /// \p BaseExpr The expression for the pointer operand.
1258 /// \p IndexExprs The expressions for the indices.
1259 /// \p InBounds Whether the GEP is in bounds.
1260 const SCEV *getGEPExpr(Type *PointeeType, const SCEV *BaseExpr,
1261 const SmallVectorImpl<const SCEV *> &IndexExprs,
1262 bool InBounds = false);
1263 const SCEV *getSMaxExpr(const SCEV *LHS, const SCEV *RHS);
1264 const SCEV *getSMaxExpr(SmallVectorImpl<const SCEV *> &Operands);
1265 const SCEV *getUMaxExpr(const SCEV *LHS, const SCEV *RHS);
1266 const SCEV *getUMaxExpr(SmallVectorImpl<const SCEV *> &Operands);
1267 const SCEV *getSMinExpr(const SCEV *LHS, const SCEV *RHS);
1268 const SCEV *getUMinExpr(const SCEV *LHS, const SCEV *RHS);
1269 const SCEV *getUnknown(Value *V);
1270 const SCEV *getCouldNotCompute();
1272 /// Return a SCEV for the constant 0 of a specific type.
1273 const SCEV *getZero(Type *Ty) { return getConstant(Ty, 0); }
1275 /// Return a SCEV for the constant 1 of a specific type.
1276 const SCEV *getOne(Type *Ty) { return getConstant(Ty, 1); }
1278 /// Return an expression for sizeof AllocTy that is type IntTy
1280 const SCEV *getSizeOfExpr(Type *IntTy, Type *AllocTy);
1282 /// Return an expression for offsetof on the given field with type IntTy
1284 const SCEV *getOffsetOfExpr(Type *IntTy, StructType *STy, unsigned FieldNo);
1286 /// Return the SCEV object corresponding to -V.
1288 const SCEV *getNegativeSCEV(const SCEV *V,
1289 SCEV::NoWrapFlags Flags = SCEV::FlagAnyWrap);
1291 /// Return the SCEV object corresponding to ~V.
1293 const SCEV *getNotSCEV(const SCEV *V);
1295 /// Return LHS-RHS. Minus is represented in SCEV as A+B*-1.
1296 const SCEV *getMinusSCEV(const SCEV *LHS, const SCEV *RHS,
1297 SCEV::NoWrapFlags Flags = SCEV::FlagAnyWrap);
1299 /// Return a SCEV corresponding to a conversion of the input value to the
1300 /// specified type. If the type must be extended, it is zero extended.
1301 const SCEV *getTruncateOrZeroExtend(const SCEV *V, Type *Ty);
1303 /// Return a SCEV corresponding to a conversion of the input value to the
1304 /// specified type. If the type must be extended, it is sign extended.
1305 const SCEV *getTruncateOrSignExtend(const SCEV *V, Type *Ty);
1307 /// Return a SCEV corresponding to a conversion of the input value to the
1308 /// specified type. If the type must be extended, it is zero extended. The
1309 /// conversion must not be narrowing.
1310 const SCEV *getNoopOrZeroExtend(const SCEV *V, Type *Ty);
1312 /// Return a SCEV corresponding to a conversion of the input value to the
1313 /// specified type. If the type must be extended, it is sign extended. The
1314 /// conversion must not be narrowing.
1315 const SCEV *getNoopOrSignExtend(const SCEV *V, Type *Ty);
1317 /// Return a SCEV corresponding to a conversion of the input value to the
1318 /// specified type. If the type must be extended, it is extended with
1319 /// unspecified bits. The conversion must not be narrowing.
1320 const SCEV *getNoopOrAnyExtend(const SCEV *V, Type *Ty);
1322 /// Return a SCEV corresponding to a conversion of the input value to the
1323 /// specified type. The conversion must not be widening.
1324 const SCEV *getTruncateOrNoop(const SCEV *V, Type *Ty);
1326 /// Promote the operands to the wider of the types using zero-extension, and
1327 /// then perform a umax operation with them.
1328 const SCEV *getUMaxFromMismatchedTypes(const SCEV *LHS,
1331 /// Promote the operands to the wider of the types using zero-extension, and
1332 /// then perform a umin operation with them.
1333 const SCEV *getUMinFromMismatchedTypes(const SCEV *LHS,
1336 /// Transitively follow the chain of pointer-type operands until reaching a
1337 /// SCEV that does not have a single pointer operand. This returns a
1338 /// SCEVUnknown pointer for well-formed pointer-type expressions, but corner
1340 const SCEV *getPointerBase(const SCEV *V);
1342 /// Return a SCEV expression for the specified value at the specified scope
1343 /// in the program. The L value specifies a loop nest to evaluate the
1344 /// expression at, where null is the top-level or a specified loop is
1345 /// immediately inside of the loop.
1347 /// This method can be used to compute the exit value for a variable defined
1348 /// in a loop by querying what the value will hold in the parent loop.
1350 /// In the case that a relevant loop exit value cannot be computed, the
1351 /// original value V is returned.
1352 const SCEV *getSCEVAtScope(const SCEV *S, const Loop *L);
1354 /// This is a convenience function which does getSCEVAtScope(getSCEV(V), L).
1355 const SCEV *getSCEVAtScope(Value *V, const Loop *L);
1357 /// Test whether entry to the loop is protected by a conditional between LHS
1358 /// and RHS. This is used to help avoid max expressions in loop trip
1359 /// counts, and to eliminate casts.
1360 bool isLoopEntryGuardedByCond(const Loop *L, ICmpInst::Predicate Pred,
1361 const SCEV *LHS, const SCEV *RHS);
1363 /// Test whether the backedge of the loop is protected by a conditional
1364 /// between LHS and RHS. This is used to to eliminate casts.
1365 bool isLoopBackedgeGuardedByCond(const Loop *L, ICmpInst::Predicate Pred,
1366 const SCEV *LHS, const SCEV *RHS);
1368 /// Returns the maximum trip count of the loop if it is a single-exit
1369 /// loop and we can compute a small maximum for that loop.
1371 /// Implemented in terms of the \c getSmallConstantTripCount overload with
1372 /// the single exiting block passed to it. See that routine for details.
1373 unsigned getSmallConstantTripCount(Loop *L);
1375 /// Returns the maximum trip count of this loop as a normal unsigned
1376 /// value. Returns 0 if the trip count is unknown or not constant. This
1377 /// "trip count" assumes that control exits via ExitingBlock. More
1378 /// precisely, it is the number of times that control may reach ExitingBlock
1379 /// before taking the branch. For loops with multiple exits, it may not be
1380 /// the number times that the loop header executes if the loop exits
1381 /// prematurely via another branch.
1382 unsigned getSmallConstantTripCount(Loop *L, BasicBlock *ExitingBlock);
1384 /// Returns the largest constant divisor of the trip count of the
1385 /// loop if it is a single-exit loop and we can compute a small maximum for
1388 /// Implemented in terms of the \c getSmallConstantTripMultiple overload with
1389 /// the single exiting block passed to it. See that routine for details.
1390 unsigned getSmallConstantTripMultiple(Loop *L);
1392 /// Returns the largest constant divisor of the trip count of this loop as a
1393 /// normal unsigned value, if possible. This means that the actual trip
1394 /// count is always a multiple of the returned value (don't forget the trip
1395 /// count could very well be zero as well!). As explained in the comments
1396 /// for getSmallConstantTripCount, this assumes that control exits the loop
1397 /// via ExitingBlock.
1398 unsigned getSmallConstantTripMultiple(Loop *L, BasicBlock *ExitingBlock);
1400 /// Get the expression for the number of loop iterations for which this loop
1401 /// is guaranteed not to exit via ExitingBlock. Otherwise return
1402 /// SCEVCouldNotCompute.
1403 const SCEV *getExitCount(Loop *L, BasicBlock *ExitingBlock);
1405 /// If the specified loop has a predictable backedge-taken count, return it,
1406 /// otherwise return a SCEVCouldNotCompute object. The backedge-taken count
1407 /// is the number of times the loop header will be branched to from within
1408 /// the loop. This is one less than the trip count of the loop, since it
1409 /// doesn't count the first iteration, when the header is branched to from
1410 /// outside the loop.
1412 /// Note that it is not valid to call this method on a loop without a
1413 /// loop-invariant backedge-taken count (see
1414 /// hasLoopInvariantBackedgeTakenCount).
1416 const SCEV *getBackedgeTakenCount(const Loop *L);
1418 /// Similar to getBackedgeTakenCount, except it will add a set of
1419 /// SCEV predicates to Predicates that are required to be true in order for
1420 /// the answer to be correct. Predicates can be checked with run-time
1421 /// checks and can be used to perform loop versioning.
1422 const SCEV *getPredicatedBackedgeTakenCount(const Loop *L,
1423 SCEVUnionPredicate &Predicates);
1425 /// Similar to getBackedgeTakenCount, except return the least SCEV value
1426 /// that is known never to be less than the actual backedge taken count.
1427 const SCEV *getMaxBackedgeTakenCount(const Loop *L);
1429 /// Return true if the specified loop has an analyzable loop-invariant
1430 /// backedge-taken count.
1431 bool hasLoopInvariantBackedgeTakenCount(const Loop *L);
1433 /// This method should be called by the client when it has changed a loop in
1434 /// a way that may effect ScalarEvolution's ability to compute a trip count,
1435 /// or if the loop is deleted. This call is potentially expensive for large
1437 void forgetLoop(const Loop *L);
1439 /// This method should be called by the client when it has changed a value
1440 /// in a way that may effect its value, or which may disconnect it from a
1441 /// def-use chain linking it to a loop.
1442 void forgetValue(Value *V);
1444 /// Called when the client has changed the disposition of values in
1447 /// We don't have a way to invalidate per-loop dispositions. Clear and
1448 /// recompute is simpler.
1449 void forgetLoopDispositions(const Loop *L) { LoopDispositions.clear(); }
1451 /// Determine the minimum number of zero bits that S is guaranteed to end in
1452 /// (at every loop iteration). It is, at the same time, the minimum number
1453 /// of times S is divisible by 2. For example, given {4,+,8} it returns 2.
1454 /// If S is guaranteed to be 0, it returns the bitwidth of S.
1455 uint32_t GetMinTrailingZeros(const SCEV *S);
1457 /// Determine the unsigned range for a particular SCEV.
1459 ConstantRange getUnsignedRange(const SCEV *S) {
1460 return getRange(S, HINT_RANGE_UNSIGNED);
1463 /// Determine the signed range for a particular SCEV.
1465 ConstantRange getSignedRange(const SCEV *S) {
1466 return getRange(S, HINT_RANGE_SIGNED);
1469 /// Test if the given expression is known to be negative.
1471 bool isKnownNegative(const SCEV *S);
1473 /// Test if the given expression is known to be positive.
1475 bool isKnownPositive(const SCEV *S);
1477 /// Test if the given expression is known to be non-negative.
1479 bool isKnownNonNegative(const SCEV *S);
1481 /// Test if the given expression is known to be non-positive.
1483 bool isKnownNonPositive(const SCEV *S);
1485 /// Test if the given expression is known to be non-zero.
1487 bool isKnownNonZero(const SCEV *S);
1489 /// Test if the given expression is known to satisfy the condition described
1490 /// by Pred, LHS, and RHS.
1492 bool isKnownPredicate(ICmpInst::Predicate Pred,
1493 const SCEV *LHS, const SCEV *RHS);
1495 /// Return true if the result of the predicate LHS `Pred` RHS is loop
1496 /// invariant with respect to L. Set InvariantPred, InvariantLHS and
1497 /// InvariantLHS so that InvariantLHS `InvariantPred` InvariantRHS is the
1498 /// loop invariant form of LHS `Pred` RHS.
1499 bool isLoopInvariantPredicate(ICmpInst::Predicate Pred, const SCEV *LHS,
1500 const SCEV *RHS, const Loop *L,
1501 ICmpInst::Predicate &InvariantPred,
1502 const SCEV *&InvariantLHS,
1503 const SCEV *&InvariantRHS);
1505 /// Simplify LHS and RHS in a comparison with predicate Pred. Return true
1506 /// iff any changes were made. If the operands are provably equal or
1507 /// unequal, LHS and RHS are set to the same value and Pred is set to either
1508 /// ICMP_EQ or ICMP_NE.
1510 bool SimplifyICmpOperands(ICmpInst::Predicate &Pred,
1513 unsigned Depth = 0);
1515 /// Return the "disposition" of the given SCEV with respect to the given
1517 LoopDisposition getLoopDisposition(const SCEV *S, const Loop *L);
1519 /// Return true if the value of the given SCEV is unchanging in the
1521 bool isLoopInvariant(const SCEV *S, const Loop *L);
1523 /// Return true if the given SCEV changes value in a known way in the
1524 /// specified loop. This property being true implies that the value is
1525 /// variant in the loop AND that we can emit an expression to compute the
1526 /// value of the expression at any particular loop iteration.
1527 bool hasComputableLoopEvolution(const SCEV *S, const Loop *L);
1529 /// Return the "disposition" of the given SCEV with respect to the given
1531 BlockDisposition getBlockDisposition(const SCEV *S, const BasicBlock *BB);
1533 /// Return true if elements that makes up the given SCEV dominate the
1534 /// specified basic block.
1535 bool dominates(const SCEV *S, const BasicBlock *BB);
1537 /// Return true if elements that makes up the given SCEV properly dominate
1538 /// the specified basic block.
1539 bool properlyDominates(const SCEV *S, const BasicBlock *BB);
1541 /// Test whether the given SCEV has Op as a direct or indirect operand.
1542 bool hasOperand(const SCEV *S, const SCEV *Op) const;
1544 /// Return the size of an element read or written by Inst.
1545 const SCEV *getElementSize(Instruction *Inst);
1547 /// Compute the array dimensions Sizes from the set of Terms extracted from
1548 /// the memory access function of this SCEVAddRecExpr (second step of
1549 /// delinearization).
1550 void findArrayDimensions(SmallVectorImpl<const SCEV *> &Terms,
1551 SmallVectorImpl<const SCEV *> &Sizes,
1552 const SCEV *ElementSize) const;
1554 void print(raw_ostream &OS) const;
1555 void verify() const;
1557 /// Collect parametric terms occurring in step expressions (first step of
1558 /// delinearization).
1559 void collectParametricTerms(const SCEV *Expr,
1560 SmallVectorImpl<const SCEV *> &Terms);
1564 /// Return in Subscripts the access functions for each dimension in Sizes
1565 /// (third step of delinearization).
1566 void computeAccessFunctions(const SCEV *Expr,
1567 SmallVectorImpl<const SCEV *> &Subscripts,
1568 SmallVectorImpl<const SCEV *> &Sizes);
1570 /// Split this SCEVAddRecExpr into two vectors of SCEVs representing the
1571 /// subscripts and sizes of an array access.
1573 /// The delinearization is a 3 step process: the first two steps compute the
1574 /// sizes of each subscript and the third step computes the access functions
1575 /// for the delinearized array:
1577 /// 1. Find the terms in the step functions
1578 /// 2. Compute the array size
1579 /// 3. Compute the access function: divide the SCEV by the array size
1580 /// starting with the innermost dimensions found in step 2. The Quotient
1581 /// is the SCEV to be divided in the next step of the recursion. The
1582 /// Remainder is the subscript of the innermost dimension. Loop over all
1583 /// array dimensions computed in step 2.
1585 /// To compute a uniform array size for several memory accesses to the same
1586 /// object, one can collect in step 1 all the step terms for all the memory
1587 /// accesses, and compute in step 2 a unique array shape. This guarantees
1588 /// that the array shape will be the same across all memory accesses.
1590 /// FIXME: We could derive the result of steps 1 and 2 from a description of
1591 /// the array shape given in metadata.
1600 /// A[j+k][2i][5i] =
1602 /// The initial SCEV:
1604 /// A[{{{0,+,2*m+5}_i, +, n*m}_j, +, n*m}_k]
1606 /// 1. Find the different terms in the step functions:
1607 /// -> [2*m, 5, n*m, n*m]
1609 /// 2. Compute the array size: sort and unique them
1610 /// -> [n*m, 2*m, 5]
1611 /// find the GCD of all the terms = 1
1612 /// divide by the GCD and erase constant terms
1615 /// divide by GCD -> [n, 2]
1616 /// remove constant terms
1618 /// size of the array is A[unknown][n][m]
1620 /// 3. Compute the access function
1621 /// a. Divide {{{0,+,2*m+5}_i, +, n*m}_j, +, n*m}_k by the innermost size m
1622 /// Quotient: {{{0,+,2}_i, +, n}_j, +, n}_k
1623 /// Remainder: {{{0,+,5}_i, +, 0}_j, +, 0}_k
1624 /// The remainder is the subscript of the innermost array dimension: [5i].
1626 /// b. Divide Quotient: {{{0,+,2}_i, +, n}_j, +, n}_k by next outer size n
1627 /// Quotient: {{{0,+,0}_i, +, 1}_j, +, 1}_k
1628 /// Remainder: {{{0,+,2}_i, +, 0}_j, +, 0}_k
1629 /// The Remainder is the subscript of the next array dimension: [2i].
1631 /// The subscript of the outermost dimension is the Quotient: [j+k].
1633 /// Overall, we have: A[][n][m], and the access function: A[j+k][2i][5i].
1634 void delinearize(const SCEV *Expr,
1635 SmallVectorImpl<const SCEV *> &Subscripts,
1636 SmallVectorImpl<const SCEV *> &Sizes,
1637 const SCEV *ElementSize);
1639 /// Return the DataLayout associated with the module this SCEV instance is
1641 const DataLayout &getDataLayout() const {
1642 return F.getParent()->getDataLayout();
1645 const SCEVPredicate *getEqualPredicate(const SCEVUnknown *LHS,
1646 const SCEVConstant *RHS);
1648 const SCEVPredicate *
1649 getWrapPredicate(const SCEVAddRecExpr *AR,
1650 SCEVWrapPredicate::IncrementWrapFlags AddedFlags);
1652 /// Re-writes the SCEV according to the Predicates in \p A.
1653 const SCEV *rewriteUsingPredicate(const SCEV *S, const Loop *L,
1654 SCEVUnionPredicate &A);
1655 /// Tries to convert the \p S expression to an AddRec expression,
1656 /// adding additional predicates to \p Preds as required.
1657 const SCEVAddRecExpr *
1658 convertSCEVToAddRecWithPredicates(const SCEV *S, const Loop *L,
1659 SCEVUnionPredicate &Preds);
1662 /// Compute the backedge taken count knowing the interval difference, the
1663 /// stride and presence of the equality in the comparison.
1664 const SCEV *computeBECount(const SCEV *Delta, const SCEV *Stride,
1667 /// Verify if an linear IV with positive stride can overflow when in a
1668 /// less-than comparison, knowing the invariant term of the comparison,
1669 /// the stride and the knowledge of NSW/NUW flags on the recurrence.
1670 bool doesIVOverflowOnLT(const SCEV *RHS, const SCEV *Stride,
1671 bool IsSigned, bool NoWrap);
1673 /// Verify if an linear IV with negative stride can overflow when in a
1674 /// greater-than comparison, knowing the invariant term of the comparison,
1675 /// the stride and the knowledge of NSW/NUW flags on the recurrence.
1676 bool doesIVOverflowOnGT(const SCEV *RHS, const SCEV *Stride,
1677 bool IsSigned, bool NoWrap);
1680 FoldingSet<SCEV> UniqueSCEVs;
1681 FoldingSet<SCEVPredicate> UniquePreds;
1682 BumpPtrAllocator SCEVAllocator;
1684 /// The head of a linked list of all SCEVUnknown values that have been
1685 /// allocated. This is used by releaseMemory to locate them all and call
1686 /// their destructors.
1687 SCEVUnknown *FirstUnknown;
1690 /// Analysis pass that exposes the \c ScalarEvolution for a function.
1691 class ScalarEvolutionAnalysis
1692 : public AnalysisInfoMixin<ScalarEvolutionAnalysis> {
1693 friend AnalysisInfoMixin<ScalarEvolutionAnalysis>;
1697 typedef ScalarEvolution Result;
1699 ScalarEvolution run(Function &F, AnalysisManager<Function> &AM);
1702 /// Printer pass for the \c ScalarEvolutionAnalysis results.
1703 class ScalarEvolutionPrinterPass
1704 : public PassInfoMixin<ScalarEvolutionPrinterPass> {
1708 explicit ScalarEvolutionPrinterPass(raw_ostream &OS) : OS(OS) {}
1709 PreservedAnalyses run(Function &F, AnalysisManager<Function> &AM);
1712 class ScalarEvolutionWrapperPass : public FunctionPass {
1713 std::unique_ptr<ScalarEvolution> SE;
1718 ScalarEvolutionWrapperPass();
1720 ScalarEvolution &getSE() { return *SE; }
1721 const ScalarEvolution &getSE() const { return *SE; }
1723 bool runOnFunction(Function &F) override;
1724 void releaseMemory() override;
1725 void getAnalysisUsage(AnalysisUsage &AU) const override;
1726 void print(raw_ostream &OS, const Module * = nullptr) const override;
1727 void verifyAnalysis() const override;
1730 /// An interface layer with SCEV used to manage how we see SCEV expressions
1731 /// for values in the context of existing predicates. We can add new
1732 /// predicates, but we cannot remove them.
1734 /// This layer has multiple purposes:
1735 /// - provides a simple interface for SCEV versioning.
1736 /// - guarantees that the order of transformations applied on a SCEV
1737 /// expression for a single Value is consistent across two different
1738 /// getSCEV calls. This means that, for example, once we've obtained
1739 /// an AddRec expression for a certain value through expression
1740 /// rewriting, we will continue to get an AddRec expression for that
1742 /// - lowers the number of expression rewrites.
1743 class PredicatedScalarEvolution {
1745 PredicatedScalarEvolution(ScalarEvolution &SE, Loop &L);
1746 const SCEVUnionPredicate &getUnionPredicate() const;
1748 /// Returns the SCEV expression of V, in the context of the current SCEV
1749 /// predicate. The order of transformations applied on the expression of V
1750 /// returned by ScalarEvolution is guaranteed to be preserved, even when
1751 /// adding new predicates.
1752 const SCEV *getSCEV(Value *V);
1754 /// Get the (predicated) backedge count for the analyzed loop.
1755 const SCEV *getBackedgeTakenCount();
1757 /// Adds a new predicate.
1758 void addPredicate(const SCEVPredicate &Pred);
1760 /// Attempts to produce an AddRecExpr for V by adding additional SCEV
1761 /// predicates. If we can't transform the expression into an AddRecExpr we
1762 /// return nullptr and not add additional SCEV predicates to the current
1764 const SCEVAddRecExpr *getAsAddRec(Value *V);
1766 /// Proves that V doesn't overflow by adding SCEV predicate.
1767 void setNoOverflow(Value *V, SCEVWrapPredicate::IncrementWrapFlags Flags);
1769 /// Returns true if we've proved that V doesn't wrap by means of a SCEV
1771 bool hasNoOverflow(Value *V, SCEVWrapPredicate::IncrementWrapFlags Flags);
1773 /// Returns the ScalarEvolution analysis used.
1774 ScalarEvolution *getSE() const { return &SE; }
1776 /// We need to explicitly define the copy constructor because of FlagsMap.
1777 PredicatedScalarEvolution(const PredicatedScalarEvolution&);
1779 /// Print the SCEV mappings done by the Predicated Scalar Evolution.
1780 /// The printed text is indented by \p Depth.
1781 void print(raw_ostream &OS, unsigned Depth) const;
1784 /// Increments the version number of the predicate. This needs to be called
1785 /// every time the SCEV predicate changes.
1786 void updateGeneration();
1788 /// Holds a SCEV and the version number of the SCEV predicate used to
1789 /// perform the rewrite of the expression.
1790 typedef std::pair<unsigned, const SCEV *> RewriteEntry;
1792 /// Maps a SCEV to the rewrite result of that SCEV at a certain version
1793 /// number. If this number doesn't match the current Generation, we will
1794 /// need to do a rewrite. To preserve the transformation order of previous
1795 /// rewrites, we will rewrite the previous result instead of the original
1797 DenseMap<const SCEV *, RewriteEntry> RewriteMap;
1799 /// Records what NoWrap flags we've added to a Value *.
1800 ValueMap<Value *, SCEVWrapPredicate::IncrementWrapFlags> FlagsMap;
1802 /// The ScalarEvolution analysis.
1803 ScalarEvolution &SE;
1805 /// The analyzed Loop.
1808 /// The SCEVPredicate that forms our context. We will rewrite all
1809 /// expressions assuming that this predicate true.
1810 SCEVUnionPredicate Preds;
1812 /// Marks the version of the SCEV predicate used. When rewriting a SCEV
1813 /// expression we mark it with the version of the predicate. We use this to
1814 /// figure out if the predicate has changed from the last rewrite of the
1815 /// SCEV. If so, we need to perform a new rewrite.
1816 unsigned Generation;
1818 /// The backedge taken count.
1819 const SCEV *BackedgeCount;