1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
|
//===- PromoteMemoryToRegister.cpp - Convert allocas to registers ---------===//
//
// This file promote memory references to be register references. It promotes
// alloca instructions which only have loads and stores as uses. An alloca is
// transformed by using dominator frontiers to place PHI nodes, then traversing
// the function in depth-first order to rewrite loads and stores as appropriate.
// This is just the standard SSA construction algorithm.
//
//===----------------------------------------------------------------------===//
#include "llvm/Transforms/Utils/PromoteMemToReg.h"
#include "llvm/Analysis/Dominators.h"
#include "llvm/iMemory.h"
#include "llvm/iPHINode.h"
#include "llvm/Function.h"
#include "llvm/Constant.h"
#include "llvm/Support/CFG.h"
#include "Support/StringExtras.h"
/// isAllocaPromotable - Return true if this alloca is legal for promotion.
/// This is true if there are only loads and stores to the alloca...
///
bool isAllocaPromotable(const AllocaInst *AI, const TargetData &TD) {
// FIXME: If the memory unit is of pointer or integer type, we can permit
// assignments to subsections of the memory unit.
// Only allow direct loads and stores...
for (Value::use_const_iterator UI = AI->use_begin(), UE = AI->use_end();
UI != UE; ++UI) // Loop over all of the uses of the alloca
if (!isa<LoadInst>(*UI))
if (const StoreInst *SI = dyn_cast<StoreInst>(*UI)) {
if (SI->getOperand(0) == AI)
return false; // Don't allow a store of the AI, only INTO the AI.
} else {
return false; // Not a load or store?
}
return true;
}
namespace {
struct PromoteMem2Reg {
// Allocas - The alloca instructions being promoted
std::vector<AllocaInst*> Allocas;
DominanceFrontier &DF;
const TargetData &TD;
// AllocaLookup - Reverse mapping of Allocas
std::map<AllocaInst*, unsigned> AllocaLookup;
// NewPhiNodes - The PhiNodes we're adding.
std::map<BasicBlock*, std::vector<PHINode*> > NewPhiNodes;
// Visited - The set of basic blocks the renamer has already visited.
std::set<BasicBlock*> Visited;
public:
PromoteMem2Reg(const std::vector<AllocaInst*> &A, DominanceFrontier &df,
const TargetData &td) : Allocas(A), DF(df), TD(td) {}
void run();
private:
void PromoteLocallyUsedAlloca(AllocaInst *AI);
void RenamePass(BasicBlock *BB, BasicBlock *Pred,
std::vector<Value*> &IncVals);
bool QueuePhiNode(BasicBlock *BB, unsigned AllocaIdx, unsigned &Version);
};
} // end of anonymous namespace
void PromoteMem2Reg::run() {
Function &F = *DF.getRoot()->getParent();
for (unsigned i = 0; i != Allocas.size(); ++i) {
AllocaInst *AI = Allocas[i];
assert(isAllocaPromotable(AI, TD) &&
"Cannot promote non-promotable alloca!");
assert(Allocas[i]->getParent()->getParent() == &F &&
"All allocas should be in the same function, which is same as DF!");
if (AI->use_empty()) {
// If there are no uses of the alloca, just delete it now.
AI->getParent()->getInstList().erase(AI);
// Remove the alloca from the Allocas list, since it has been processed
Allocas[i] = Allocas.back();
Allocas.pop_back();
--i;
continue;
}
// Calculate the set of write-locations for each alloca. This is analogous
// to counting the number of 'redefinitions' of each variable.
std::vector<BasicBlock*> DefiningBlocks;
BasicBlock *OnlyBlock = 0;
bool OnlyUsedInOneBlock = true;
// As we scan the uses of the alloca instruction, keep track of stores, and
// decide whether all of the loads and stores to the alloca are within the
// same basic block.
for (Value::use_iterator U =AI->use_begin(), E = AI->use_end(); U != E;++U){
Instruction *User = cast<Instruction>(*U);
if (StoreInst *SI = dyn_cast<StoreInst>(User)) {
// Remember the basic blocks which define new values for the alloca
DefiningBlocks.push_back(SI->getParent());
}
if (OnlyUsedInOneBlock) {
if (OnlyBlock == 0)
OnlyBlock = User->getParent();
else if (OnlyBlock != User->getParent())
OnlyUsedInOneBlock = false;
}
}
// If the alloca is only read and written in one basic block, just perform a
// linear sweep over the block to eliminate it.
if (OnlyUsedInOneBlock) {
PromoteLocallyUsedAlloca(AI);
// Remove the alloca from the Allocas list, since it has been processed
Allocas[i] = Allocas.back();
Allocas.pop_back();
--i;
continue;
}
AllocaLookup[Allocas[i]] = i;
// PhiNodeBlocks - A list of blocks that phi nodes have been inserted for
// this alloca.
std::vector<BasicBlock*> PhiNodeBlocks;
// Compute the locations where PhiNodes need to be inserted. Look at the
// dominance frontier of EACH basic-block we have a write in.
//
unsigned CurrentVersion = 0;
while (!DefiningBlocks.empty()) {
BasicBlock *BB = DefiningBlocks.back();
DefiningBlocks.pop_back();
// Look up the DF for this write, add it to PhiNodes
DominanceFrontier::const_iterator it = DF.find(BB);
if (it != DF.end()) {
const DominanceFrontier::DomSetType &S = it->second;
for (DominanceFrontier::DomSetType::iterator P = S.begin(),PE = S.end();
P != PE; ++P)
if (QueuePhiNode(*P, i, CurrentVersion))
DefiningBlocks.push_back(*P);
}
}
}
if (Allocas.empty())
return; // All of the allocas must have been trivial!
// Set the incoming values for the basic block to be null values for all of
// the alloca's. We do this in case there is a load of a value that has not
// been stored yet. In this case, it will get this null value.
//
std::vector<Value *> Values(Allocas.size());
for (unsigned i = 0, e = Allocas.size(); i != e; ++i)
Values[i] = Constant::getNullValue(Allocas[i]->getAllocatedType());
// Walks all basic blocks in the function performing the SSA rename algorithm
// and inserting the phi nodes we marked as necessary
//
RenamePass(F.begin(), 0, Values);
// The renamer uses the Visited set to avoid infinite loops. Clear it now.
Visited.clear();
// Remove the allocas themselves from the function...
for (unsigned i = 0, e = Allocas.size(); i != e; ++i) {
Instruction *A = Allocas[i];
// If there are any uses of the alloca instructions left, they must be in
// sections of dead code that were not processed on the dominance frontier.
// Just delete the users now.
//
if (!A->use_empty())
A->replaceAllUsesWith(Constant::getNullValue(A->getType()));
A->getParent()->getInstList().erase(A);
}
// At this point, the renamer has added entries to PHI nodes for all reachable
// code. Unfortunately, there may be blocks which are not reachable, which
// the renamer hasn't traversed. If this is the case, the PHI nodes may not
// have incoming values for all predecessors. Loop over all PHI nodes we have
// created, inserting null constants if they are missing any incoming values.
//
for (std::map<BasicBlock*, std::vector<PHINode *> >::iterator I =
NewPhiNodes.begin(), E = NewPhiNodes.end(); I != E; ++I) {
std::vector<BasicBlock*> Preds(pred_begin(I->first), pred_end(I->first));
std::vector<PHINode*> &PNs = I->second;
assert(!PNs.empty() && "Empty PHI node list??");
// Only do work here if there the PHI nodes are missing incoming values. We
// know that all PHI nodes that were inserted in a block will have the same
// number of incoming values, so we can just check any PHI node.
PHINode *FirstPHI;
for (unsigned i = 0; (FirstPHI = PNs[i]) == 0; ++i)
/*empty*/;
if (Preds.size() != FirstPHI->getNumIncomingValues()) {
// Ok, now we know that all of the PHI nodes are missing entries for some
// basic blocks. Start by sorting the incoming predecessors for efficient
// access.
std::sort(Preds.begin(), Preds.end());
// Now we loop through all BB's which have entries in FirstPHI and remove
// them from the Preds list.
for (unsigned i = 0, e = FirstPHI->getNumIncomingValues(); i != e; ++i) {
// Do a log(n) search of teh Preds list for the entry we want.
std::vector<BasicBlock*>::iterator EntIt =
std::lower_bound(Preds.begin(), Preds.end(),
FirstPHI->getIncomingBlock(i));
assert(EntIt != Preds.end() && *EntIt == FirstPHI->getIncomingBlock(i)&&
"PHI node has entry for a block which is not a predecessor!");
|