1//===- NaryReassociate.h - Reassociate n-ary expressions --------*- C++ -*-===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8//
9// This pass reassociates n-ary add expressions and eliminates the redundancy
10// exposed by the reassociation.
11//
12// A motivating example:
13//
14// void foo(int a, int b) {
15// bar(a + b);
16// bar((a + 2) + b);
17// }
18//
19// An ideal compiler should reassociate (a + 2) + b to (a + b) + 2 and simplify
20// the above code to
21//
22// int t = a + b;
23// bar(t);
24// bar(t + 2);
25//
26// However, the Reassociate pass is unable to do that because it processes each
27// instruction individually and believes (a + 2) + b is the best form according
28// to its rank system.
29//
30// To address this limitation, NaryReassociate reassociates an expression in a
31// form that reuses existing instructions. As a result, NaryReassociate can
32// reassociate (a + 2) + b in the example to (a + b) + 2 because it detects that
33// (a + b) is computed before.
34//
35// NaryReassociate works as follows. For every instruction in the form of (a +
36// b) + c, it checks whether a + c or b + c is already computed by a dominating
37// instruction. If so, it then reassociates (a + b) + c into (a + c) + b or (b +
38// c) + a and removes the redundancy accordingly. To efficiently look up whether
39// an expression is computed before, we store each instruction seen and its SCEV
40// into an SCEV-to-instruction map.
41//
42// Although the algorithm pattern-matches only ternary additions, it
43// automatically handles many >3-ary expressions by walking through the function
44// in the depth-first order. For example, given
45//
46// (a + c) + d
47// ((a + b) + c) + d
48//
49// NaryReassociate first rewrites (a + b) + c to (a + c) + b, and then rewrites
50// ((a + c) + b) + d into ((a + c) + d) + b.
51//
52// Finally, the above dominator-based algorithm may need to be run multiple
53// iterations before emitting optimal code. One source of this need is that we
54// only split an operand when it is used only once. The above algorithm can
55// eliminate an instruction and decrease the usage count of its operands. As a
56// result, an instruction that previously had multiple uses may become a
57// single-use instruction and thus eligible for split consideration. For
58// example,
59//
60// ac = a + c
61// ab = a + b
62// abc = ab + c
63// ab2 = ab + b
64// ab2c = ab2 + c
65//
66// In the first iteration, we cannot reassociate abc to ac+b because ab is used
67// twice. However, we can reassociate ab2c to abc+b in the first iteration. As a
68// result, ab2 becomes dead and ab will be used only once in the second
69// iteration.
70//
71// Limitations and TODO items:
72//
73// 1) We only considers n-ary adds and muls for now. This should be extended
74// and generalized.
75//
76//===----------------------------------------------------------------------===//
77
78#ifndef LLVM_TRANSFORMS_SCALAR_NARYREASSOCIATE_H
79#define LLVM_TRANSFORMS_SCALAR_NARYREASSOCIATE_H
80
81#include "llvm/ADT/DenseMap.h"
82#include "llvm/ADT/SmallVector.h"
83#include "llvm/Analysis/ScalarEvolution.h"
84#include "llvm/IR/PassManager.h"
85#include "llvm/IR/ValueHandle.h"
86
87namespace llvm {
88
89class AssumptionCache;
90class BinaryOperator;
91class DataLayout;
92class DominatorTree;
93class Function;
94class GetElementPtrInst;
95class Instruction;
96class TargetLibraryInfo;
97class TargetTransformInfo;
98class Type;
99class Value;
100
101class NaryReassociatePass : public PassInfoMixin<NaryReassociatePass> {
102public:
103 PreservedAnalyses run(Function &F, FunctionAnalysisManager &AM);
104
105 // Glue for old PM.
106 bool runImpl(Function &F, AssumptionCache *AC_, DominatorTree *DT_,
107 ScalarEvolution *SE_, TargetLibraryInfo *TLI_,
108 TargetTransformInfo *TTI_);
109
110private:
111 // Runs only one iteration of the dominator-based algorithm. See the header
112 // comments for why we need multiple iterations.
113 bool doOneIteration(Function &F);
114
115 // Reassociates I for better CSE.
116 Instruction *tryReassociate(Instruction *I, SCEVUse &OrigSCEV);
117
118 // Reassociate GEP for better CSE.
119 Instruction *tryReassociateGEP(GetElementPtrInst *GEP);
120
121 // Try splitting GEP at the I-th index and see whether either part can be
122 // CSE'ed. This is a helper function for tryReassociateGEP.
123 //
124 // \p IndexedType The element type indexed by GEP's I-th index. This is
125 // equivalent to
126 // GEP->getIndexedType(GEP->getPointerOperand(), 0-th index,
127 // ..., i-th index).
128 GetElementPtrInst *tryReassociateGEPAtIndex(GetElementPtrInst *GEP,
129 unsigned I, Type *IndexedType);
130
131 // Given GEP's I-th index = LHS + RHS, see whether &Base[..][LHS][..] or
132 // &Base[..][RHS][..] can be CSE'ed and rewrite GEP accordingly.
133 GetElementPtrInst *tryReassociateGEPAtIndex(GetElementPtrInst *GEP,
134 unsigned I, Value *LHS,
135 Value *RHS, Type *IndexedType);
136
137 // Reassociate binary operators for better CSE.
138 Instruction *tryReassociateBinaryOp(BinaryOperator *I);
139
140 // A helper function for tryReassociateBinaryOp. LHS and RHS are explicitly
141 // passed.
142 Instruction *tryReassociateBinaryOp(Value *LHS, Value *RHS,
143 BinaryOperator *I);
144 // Rewrites I to (LHS op RHS) if LHS is computed already.
145 Instruction *tryReassociatedBinaryOp(SCEVUse LHS, Value *RHS,
146 BinaryOperator *I);
147
148 // Tries to match Op1 and Op2 by using V.
149 bool matchTernaryOp(BinaryOperator *I, Value *V, Value *&Op1, Value *&Op2);
150
151 // Gets SCEV for (LHS op RHS).
152 SCEVUse getBinarySCEV(BinaryOperator *I, SCEVUse LHS, SCEVUse RHS);
153
154 // Returns the closest dominator of \c Dominatee that computes
155 // \c CandidateExpr. Returns null if not found.
156 Instruction *findClosestMatchingDominator(SCEVUse CandidateExpr,
157 Instruction *Dominatee);
158
159 // Try to match \p I as signed/unsigned Min/Max and reassociate it. \p
160 // OrigSCEV is set if \I matches Min/Max regardless whether resassociation is
161 // done or not. If reassociation was successful newly generated instruction is
162 // returned, otherwise nullptr.
163 template <typename PredT>
164 Instruction *matchAndReassociateMinOrMax(Instruction *I, SCEVUse &OrigSCEV);
165
166 // Reassociate Min/Max.
167 template <typename MaxMinT>
168 Value *tryReassociateMinOrMax(Instruction *I, MaxMinT MaxMinMatch, Value *LHS,
169 Value *RHS);
170
171 // GetElementPtrInst implicitly sign-extends an index if the index is shorter
172 // than the pointer size. This function returns whether Index is shorter than
173 // GEP's pointer size, i.e., whether Index needs to be sign-extended in order
174 // to be an index of GEP.
175 bool requiresSignExtension(Value *Index, GetElementPtrInst *GEP);
176
177 AssumptionCache *AC;
178 const DataLayout *DL;
179 DominatorTree *DT;
180 ScalarEvolution *SE;
181 TargetLibraryInfo *TLI;
182 TargetTransformInfo *TTI;
183
184 // A lookup table quickly telling which instructions compute the given SCEV.
185 // Note that there can be multiple instructions at different locations
186 // computing to the same SCEV, so we map a SCEV to an instruction list. For
187 // example,
188 //
189 // if (p1)
190 // foo(a + b);
191 // if (p2)
192 // bar(a + b);
193 DenseMap<const SCEV *, SmallVector<WeakTrackingVH, 2>> SeenExprs;
194};
195
196} // end namespace llvm
197
198#endif // LLVM_TRANSFORMS_SCALAR_NARYREASSOCIATE_H
199