LLVM API Documentation

 All Classes Namespaces Files Functions Variables Typedefs Enumerations Enumerator Friends Macros Groups Pages
ScalarEvolutionNormalization.cpp
Go to the documentation of this file.
1 //===- ScalarEvolutionNormalization.cpp - See below -------------*- C++ -*-===//
2 //
3 // The LLVM Compiler Infrastructure
4 //
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
7 //
8 //===----------------------------------------------------------------------===//
9 //
10 // This file implements utilities for working with "normalized" expressions.
11 // See the comments at the top of ScalarEvolutionNormalization.h for details.
12 //
13 //===----------------------------------------------------------------------===//
14 
16 #include "llvm/Analysis/LoopInfo.h"
19 using namespace llvm;
20 
21 /// IVUseShouldUsePostIncValue - We have discovered a "User" of an IV expression
22 /// and now we need to decide whether the user should use the preinc or post-inc
23 /// value. If this user should use the post-inc version of the IV, return true.
24 ///
25 /// Choosing wrong here can break dominance properties (if we choose to use the
26 /// post-inc value when we cannot) or it can end up adding extra live-ranges to
27 /// the loop, resulting in reg-reg copies (if we use the pre-inc value when we
28 /// should use the post-inc value).
30  const Loop *L, DominatorTree *DT) {
31  // If the user is in the loop, use the preinc value.
32  if (L->contains(User)) return false;
33 
34  BasicBlock *LatchBlock = L->getLoopLatch();
35  if (!LatchBlock)
36  return false;
37 
38  // Ok, the user is outside of the loop. If it is dominated by the latch
39  // block, use the post-inc value.
40  if (DT->dominates(LatchBlock, User->getParent()))
41  return true;
42 
43  // There is one case we have to be careful of: PHI nodes. These little guys
44  // can live in blocks that are not dominated by the latch block, but (since
45  // their uses occur in the predecessor block, not the block the PHI lives in)
46  // should still use the post-inc value. Check for this case now.
47  PHINode *PN = dyn_cast<PHINode>(User);
48  if (!PN || !Operand) return false; // not a phi, not dominated by latch block.
49 
50  // Look at all of the uses of Operand by the PHI node. If any use corresponds
51  // to a block that is not dominated by the latch block, give up and use the
52  // preincremented value.
53  for (unsigned i = 0, e = PN->getNumIncomingValues(); i != e; ++i)
54  if (PN->getIncomingValue(i) == Operand &&
55  !DT->dominates(LatchBlock, PN->getIncomingBlock(i)))
56  return false;
57 
58  // Okay, all uses of Operand by PN are in predecessor blocks that really are
59  // dominated by the latch block. Use the post-incremented value.
60  return true;
61 }
62 
63 namespace {
64 
65 /// Hold the state used during post-inc expression transformation, including a
66 /// map of transformed expressions.
67 class PostIncTransform {
70  ScalarEvolution &SE;
71  DominatorTree &DT;
72 
74 
75 public:
76  PostIncTransform(TransformKind kind, PostIncLoopSet &loops,
78  Kind(kind), Loops(loops), SE(se), DT(dt) {}
79 
80  const SCEV *TransformSubExpr(const SCEV *S, Instruction *User,
81  Value *OperandValToReplace);
82 
83 protected:
84  const SCEV *TransformImpl(const SCEV *S, Instruction *User,
85  Value *OperandValToReplace);
86 };
87 
88 } // namespace
89 
90 /// Implement post-inc transformation for all valid expression types.
91 const SCEV *PostIncTransform::
92 TransformImpl(const SCEV *S, Instruction *User, Value *OperandValToReplace) {
93 
94  if (const SCEVCastExpr *X = dyn_cast<SCEVCastExpr>(S)) {
95  const SCEV *O = X->getOperand();
96  const SCEV *N = TransformSubExpr(O, User, OperandValToReplace);
97  if (O != N)
98  switch (S->getSCEVType()) {
99  case scZeroExtend: return SE.getZeroExtendExpr(N, S->getType());
100  case scSignExtend: return SE.getSignExtendExpr(N, S->getType());
101  case scTruncate: return SE.getTruncateExpr(N, S->getType());
102  default: llvm_unreachable("Unexpected SCEVCastExpr kind!");
103  }
104  return S;
105  }
106 
107  if (const SCEVAddRecExpr *AR = dyn_cast<SCEVAddRecExpr>(S)) {
108  // An addrec. This is the interesting part.
110  const Loop *L = AR->getLoop();
111  // The addrec conceptually uses its operands at loop entry.
112  Instruction *LUser = L->getHeader()->begin();
113  // Transform each operand.
114  for (SCEVNAryExpr::op_iterator I = AR->op_begin(), E = AR->op_end();
115  I != E; ++I) {
116  Operands.push_back(TransformSubExpr(*I, LUser, 0));
117  }
118  // Conservatively use AnyWrap until/unless we need FlagNW.
119  const SCEV *Result = SE.getAddRecExpr(Operands, L, SCEV::FlagAnyWrap);
120  switch (Kind) {
121  case NormalizeAutodetect:
122  // Normalize this SCEV by subtracting the expression for the final step.
123  // We only allow affine AddRecs to be normalized, otherwise we would not
124  // be able to correctly denormalize.
125  // e.g. {1,+,3,+,2} == {-2,+,1,+,2} + {3,+,2}
126  // Normalized form: {-2,+,1,+,2}
127  // Denormalized form: {1,+,3,+,2}
128  //
129  // However, denormalization would use the a different step expression than
130  // normalization (see getPostIncExpr), generating the wrong final
131  // expression: {-2,+,1,+,2} + {1,+,2} => {-1,+,3,+,2}
132  if (AR->isAffine() &&
133  IVUseShouldUsePostIncValue(User, OperandValToReplace, L, &DT)) {
134  Result = SE.getMinusSCEV(Result, AR->getStepRecurrence(SE));
135  Loops.insert(L);
136  }
137 #if 0
138  // This assert is conceptually correct, but ScalarEvolution currently
139  // sometimes fails to canonicalize two equal SCEVs to exactly the same
140  // form. It's possibly a pessimization when this happens, but it isn't a
141  // correctness problem, so disable this assert for now.
142  assert(S == TransformSubExpr(Result, User, OperandValToReplace) &&
143  "SCEV normalization is not invertible!");
144 #endif
145  break;
146  case Normalize:
147  if (Loops.count(L)) {
148  const SCEV *TransformedStep =
149  TransformSubExpr(AR->getStepRecurrence(SE),
150  User, OperandValToReplace);
151  Result = SE.getMinusSCEV(Result, TransformedStep);
152  }
153 #if 0
154  // See the comment on the assert above.
155  assert(S == TransformSubExpr(Result, User, OperandValToReplace) &&
156  "SCEV normalization is not invertible!");
157 #endif
158  break;
159  case Denormalize:
160  if (Loops.count(L))
161  Result = cast<SCEVAddRecExpr>(Result)->getPostIncExpr(SE);
162  break;
163  }
164  return Result;
165  }
166 
167  if (const SCEVNAryExpr *X = dyn_cast<SCEVNAryExpr>(S)) {
169  bool Changed = false;
170  // Transform each operand.
171  for (SCEVNAryExpr::op_iterator I = X->op_begin(), E = X->op_end();
172  I != E; ++I) {
173  const SCEV *O = *I;
174  const SCEV *N = TransformSubExpr(O, User, OperandValToReplace);
175  Changed |= N != O;
176  Operands.push_back(N);
177  }
178  // If any operand actually changed, return a transformed result.
179  if (Changed)
180  switch (S->getSCEVType()) {
181  case scAddExpr: return SE.getAddExpr(Operands);
182  case scMulExpr: return SE.getMulExpr(Operands);
183  case scSMaxExpr: return SE.getSMaxExpr(Operands);
184  case scUMaxExpr: return SE.getUMaxExpr(Operands);
185  default: llvm_unreachable("Unexpected SCEVNAryExpr kind!");
186  }
187  return S;
188  }
189 
190  if (const SCEVUDivExpr *X = dyn_cast<SCEVUDivExpr>(S)) {
191  const SCEV *LO = X->getLHS();
192  const SCEV *RO = X->getRHS();
193  const SCEV *LN = TransformSubExpr(LO, User, OperandValToReplace);
194  const SCEV *RN = TransformSubExpr(RO, User, OperandValToReplace);
195  if (LO != LN || RO != RN)
196  return SE.getUDivExpr(LN, RN);
197  return S;
198  }
199 
200  llvm_unreachable("Unexpected SCEV kind!");
201 }
202 
203 /// Manage recursive transformation across an expression DAG. Revisiting
204 /// expressions would lead to exponential recursion.
205 const SCEV *PostIncTransform::
206 TransformSubExpr(const SCEV *S, Instruction *User, Value *OperandValToReplace) {
207 
208  if (isa<SCEVConstant>(S) || isa<SCEVUnknown>(S))
209  return S;
210 
211  const SCEV *Result = Transformed.lookup(S);
212  if (Result)
213  return Result;
214 
215  Result = TransformImpl(S, User, OperandValToReplace);
216  Transformed[S] = Result;
217  return Result;
218 }
219 
220 /// Top level driver for transforming an expression DAG into its requested
221 /// post-inc form (either "Normalized" or "Denormalized".
223  const SCEV *S,
224  Instruction *User,
225  Value *OperandValToReplace,
227  ScalarEvolution &SE,
228  DominatorTree &DT) {
229  PostIncTransform Transform(Kind, Loops, SE, DT);
230  return Transform.TransformSubExpr(S, User, OperandValToReplace);
231 }
const SCEV * TransformForPostIncUse(TransformKind Kind, const SCEV *S, Instruction *User, Value *OperandValToReplace, PostIncLoopSet &Loops, ScalarEvolution &SE, DominatorTree &DT)
enable_if_c<!is_simple_type< Y >::value, typename cast_retty< X, const Y >::ret_type >::type dyn_cast(const Y &Val)
Definition: Casting.h:266
BlockT * getHeader() const
Definition: LoopInfo.h:95
BlockT * getLoopLatch() const
Definition: LoopInfoImpl.h:154
Hexagon Hardware Loops
#define llvm_unreachable(msg)
const SCEV *const * op_iterator
unsigned getNumIncomingValues() const
#define LO(Val)
LLVM Basic Block Representation.
Definition: BasicBlock.h:72
Type * getType() const
Normalize - Normalize according to the given loops.
BasicBlock * getIncomingBlock(unsigned i) const
bool contains(const LoopT *L) const
Definition: LoopInfo.h:104
bool dominates(const DomTreeNode *A, const DomTreeNode *B) const
Definition: Dominators.h:801
Value * getIncomingValue(unsigned i) const
machine loops
#define I(x, y, z)
Definition: MD5.cpp:54
#define N
static bool IVUseShouldUsePostIncValue(Instruction *User, Value *Operand, const Loop *L, DominatorTree *DT)
unsigned getSCEVType() const
LLVM Value Representation.
Definition: Value.h:66
static RegisterPass< NVPTXAllocaHoisting > X("alloca-hoisting","Hoisting alloca instructions in non-entry ""blocks to the entry block")
const BasicBlock * getParent() const
Definition: Instruction.h:52