1 | //===--- ExpandMemCmp.cpp - Expand memcmp() to load/stores ----------------===// |
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 tries to expand memcmp() calls into optimally-sized loads and |
10 | // compares for the target. |
11 | // |
12 | //===----------------------------------------------------------------------===// |
13 | |
14 | #include "llvm/CodeGen/ExpandMemCmp.h" |
15 | #include "llvm/ADT/Statistic.h" |
16 | #include "llvm/Analysis/ConstantFolding.h" |
17 | #include "llvm/Analysis/DomTreeUpdater.h" |
18 | #include "llvm/Analysis/LazyBlockFrequencyInfo.h" |
19 | #include "llvm/Analysis/ProfileSummaryInfo.h" |
20 | #include "llvm/Analysis/TargetLibraryInfo.h" |
21 | #include "llvm/Analysis/TargetTransformInfo.h" |
22 | #include "llvm/Analysis/ValueTracking.h" |
23 | #include "llvm/CodeGen/TargetPassConfig.h" |
24 | #include "llvm/CodeGen/TargetSubtargetInfo.h" |
25 | #include "llvm/IR/Dominators.h" |
26 | #include "llvm/IR/IRBuilder.h" |
27 | #include "llvm/IR/PatternMatch.h" |
28 | #include "llvm/InitializePasses.h" |
29 | #include "llvm/Target/TargetMachine.h" |
30 | #include "llvm/Transforms/Utils/BasicBlockUtils.h" |
31 | #include "llvm/Transforms/Utils/Local.h" |
32 | #include "llvm/Transforms/Utils/SizeOpts.h" |
33 | #include <optional> |
34 | |
35 | using namespace llvm; |
36 | using namespace llvm::PatternMatch; |
37 | |
38 | namespace llvm { |
39 | class TargetLowering; |
40 | } |
41 | |
42 | #define DEBUG_TYPE "expand-memcmp" |
43 | |
44 | STATISTIC(NumMemCmpCalls, "Number of memcmp calls" ); |
45 | STATISTIC(NumMemCmpNotConstant, "Number of memcmp calls without constant size" ); |
46 | STATISTIC(NumMemCmpGreaterThanMax, |
47 | "Number of memcmp calls with size greater than max size" ); |
48 | STATISTIC(NumMemCmpInlined, "Number of inlined memcmp calls" ); |
49 | |
50 | static cl::opt<unsigned> MemCmpEqZeroNumLoadsPerBlock( |
51 | "memcmp-num-loads-per-block" , cl::Hidden, cl::init(Val: 1), |
52 | cl::desc("The number of loads per basic block for inline expansion of " |
53 | "memcmp that is only being compared against zero." )); |
54 | |
55 | static cl::opt<unsigned> MaxLoadsPerMemcmp( |
56 | "max-loads-per-memcmp" , cl::Hidden, |
57 | cl::desc("Set maximum number of loads used in expanded memcmp" )); |
58 | |
59 | static cl::opt<unsigned> MaxLoadsPerMemcmpOptSize( |
60 | "max-loads-per-memcmp-opt-size" , cl::Hidden, |
61 | cl::desc("Set maximum number of loads used in expanded memcmp for -Os/Oz" )); |
62 | |
63 | namespace { |
64 | |
65 | |
66 | // This class provides helper functions to expand a memcmp library call into an |
67 | // inline expansion. |
68 | class MemCmpExpansion { |
69 | struct ResultBlock { |
70 | BasicBlock *BB = nullptr; |
71 | PHINode *PhiSrc1 = nullptr; |
72 | PHINode *PhiSrc2 = nullptr; |
73 | |
74 | ResultBlock() = default; |
75 | }; |
76 | |
77 | CallInst *const CI = nullptr; |
78 | ResultBlock ResBlock; |
79 | const uint64_t Size; |
80 | unsigned MaxLoadSize = 0; |
81 | uint64_t NumLoadsNonOneByte = 0; |
82 | const uint64_t NumLoadsPerBlockForZeroCmp; |
83 | std::vector<BasicBlock *> LoadCmpBlocks; |
84 | BasicBlock *EndBlock = nullptr; |
85 | PHINode *PhiRes = nullptr; |
86 | const bool IsUsedForZeroCmp; |
87 | const DataLayout &DL; |
88 | DomTreeUpdater *DTU = nullptr; |
89 | IRBuilder<> Builder; |
90 | // Represents the decomposition in blocks of the expansion. For example, |
91 | // comparing 33 bytes on X86+sse can be done with 2x16-byte loads and |
92 | // 1x1-byte load, which would be represented as [{16, 0}, {16, 16}, {1, 32}. |
93 | struct LoadEntry { |
94 | LoadEntry(unsigned LoadSize, uint64_t Offset) |
95 | : LoadSize(LoadSize), Offset(Offset) { |
96 | } |
97 | |
98 | // The size of the load for this block, in bytes. |
99 | unsigned LoadSize; |
100 | // The offset of this load from the base pointer, in bytes. |
101 | uint64_t Offset; |
102 | }; |
103 | using LoadEntryVector = SmallVector<LoadEntry, 8>; |
104 | LoadEntryVector LoadSequence; |
105 | |
106 | void createLoadCmpBlocks(); |
107 | void createResultBlock(); |
108 | void setupResultBlockPHINodes(); |
109 | void setupEndBlockPHINodes(); |
110 | Value *getCompareLoadPairs(unsigned BlockIndex, unsigned &LoadIndex); |
111 | void emitLoadCompareBlock(unsigned BlockIndex); |
112 | void emitLoadCompareBlockMultipleLoads(unsigned BlockIndex, |
113 | unsigned &LoadIndex); |
114 | void emitLoadCompareByteBlock(unsigned BlockIndex, unsigned OffsetBytes); |
115 | void emitMemCmpResultBlock(); |
116 | Value *getMemCmpExpansionZeroCase(); |
117 | Value *getMemCmpEqZeroOneBlock(); |
118 | Value *getMemCmpOneBlock(); |
119 | struct LoadPair { |
120 | Value *Lhs = nullptr; |
121 | Value *Rhs = nullptr; |
122 | }; |
123 | LoadPair getLoadPair(Type *LoadSizeType, Type *BSwapSizeType, |
124 | Type *CmpSizeType, unsigned OffsetBytes); |
125 | |
126 | static LoadEntryVector |
127 | computeGreedyLoadSequence(uint64_t Size, llvm::ArrayRef<unsigned> LoadSizes, |
128 | unsigned MaxNumLoads, unsigned &NumLoadsNonOneByte); |
129 | static LoadEntryVector |
130 | computeOverlappingLoadSequence(uint64_t Size, unsigned MaxLoadSize, |
131 | unsigned MaxNumLoads, |
132 | unsigned &NumLoadsNonOneByte); |
133 | |
134 | static void optimiseLoadSequence( |
135 | LoadEntryVector &LoadSequence, |
136 | const TargetTransformInfo::MemCmpExpansionOptions &Options, |
137 | bool IsUsedForZeroCmp); |
138 | |
139 | public: |
140 | MemCmpExpansion(CallInst *CI, uint64_t Size, |
141 | const TargetTransformInfo::MemCmpExpansionOptions &Options, |
142 | const bool IsUsedForZeroCmp, const DataLayout &TheDataLayout, |
143 | DomTreeUpdater *DTU); |
144 | |
145 | unsigned getNumBlocks(); |
146 | uint64_t getNumLoads() const { return LoadSequence.size(); } |
147 | |
148 | Value *getMemCmpExpansion(); |
149 | }; |
150 | |
151 | MemCmpExpansion::LoadEntryVector MemCmpExpansion::computeGreedyLoadSequence( |
152 | uint64_t Size, llvm::ArrayRef<unsigned> LoadSizes, |
153 | const unsigned MaxNumLoads, unsigned &NumLoadsNonOneByte) { |
154 | NumLoadsNonOneByte = 0; |
155 | LoadEntryVector LoadSequence; |
156 | uint64_t Offset = 0; |
157 | while (Size && !LoadSizes.empty()) { |
158 | const unsigned LoadSize = LoadSizes.front(); |
159 | const uint64_t NumLoadsForThisSize = Size / LoadSize; |
160 | if (LoadSequence.size() + NumLoadsForThisSize > MaxNumLoads) { |
161 | // Do not expand if the total number of loads is larger than what the |
162 | // target allows. Note that it's important that we exit before completing |
163 | // the expansion to avoid using a ton of memory to store the expansion for |
164 | // large sizes. |
165 | return {}; |
166 | } |
167 | if (NumLoadsForThisSize > 0) { |
168 | for (uint64_t I = 0; I < NumLoadsForThisSize; ++I) { |
169 | LoadSequence.push_back(Elt: {LoadSize, Offset}); |
170 | Offset += LoadSize; |
171 | } |
172 | if (LoadSize > 1) |
173 | ++NumLoadsNonOneByte; |
174 | Size = Size % LoadSize; |
175 | } |
176 | LoadSizes = LoadSizes.drop_front(); |
177 | } |
178 | return LoadSequence; |
179 | } |
180 | |
181 | MemCmpExpansion::LoadEntryVector |
182 | MemCmpExpansion::computeOverlappingLoadSequence(uint64_t Size, |
183 | const unsigned MaxLoadSize, |
184 | const unsigned MaxNumLoads, |
185 | unsigned &NumLoadsNonOneByte) { |
186 | // These are already handled by the greedy approach. |
187 | if (Size < 2 || MaxLoadSize < 2) |
188 | return {}; |
189 | |
190 | // We try to do as many non-overlapping loads as possible starting from the |
191 | // beginning. |
192 | const uint64_t NumNonOverlappingLoads = Size / MaxLoadSize; |
193 | assert(NumNonOverlappingLoads && "there must be at least one load" ); |
194 | // There remain 0 to (MaxLoadSize - 1) bytes to load, this will be done with |
195 | // an overlapping load. |
196 | Size = Size - NumNonOverlappingLoads * MaxLoadSize; |
197 | // Bail if we do not need an overloapping store, this is already handled by |
198 | // the greedy approach. |
199 | if (Size == 0) |
200 | return {}; |
201 | // Bail if the number of loads (non-overlapping + potential overlapping one) |
202 | // is larger than the max allowed. |
203 | if ((NumNonOverlappingLoads + 1) > MaxNumLoads) |
204 | return {}; |
205 | |
206 | // Add non-overlapping loads. |
207 | LoadEntryVector LoadSequence; |
208 | uint64_t Offset = 0; |
209 | for (uint64_t I = 0; I < NumNonOverlappingLoads; ++I) { |
210 | LoadSequence.push_back(Elt: {MaxLoadSize, Offset}); |
211 | Offset += MaxLoadSize; |
212 | } |
213 | |
214 | // Add the last overlapping load. |
215 | assert(Size > 0 && Size < MaxLoadSize && "broken invariant" ); |
216 | LoadSequence.push_back(Elt: {MaxLoadSize, Offset - (MaxLoadSize - Size)}); |
217 | NumLoadsNonOneByte = 1; |
218 | return LoadSequence; |
219 | } |
220 | |
221 | void MemCmpExpansion::optimiseLoadSequence( |
222 | LoadEntryVector &LoadSequence, |
223 | const TargetTransformInfo::MemCmpExpansionOptions &Options, |
224 | bool IsUsedForZeroCmp) { |
225 | // This part of code attempts to optimize the LoadSequence by merging allowed |
226 | // subsequences into single loads of allowed sizes from |
227 | // `MemCmpExpansionOptions::AllowedTailExpansions`. If it is for zero |
228 | // comparison or if no allowed tail expansions are specified, we exit early. |
229 | if (IsUsedForZeroCmp || Options.AllowedTailExpansions.empty()) |
230 | return; |
231 | |
232 | while (LoadSequence.size() >= 2) { |
233 | auto Last = LoadSequence[LoadSequence.size() - 1]; |
234 | auto PreLast = LoadSequence[LoadSequence.size() - 2]; |
235 | |
236 | // Exit the loop if the two sequences are not contiguous |
237 | if (PreLast.Offset + PreLast.LoadSize != Last.Offset) |
238 | break; |
239 | |
240 | auto LoadSize = Last.LoadSize + PreLast.LoadSize; |
241 | if (find(Range: Options.AllowedTailExpansions, Val: LoadSize) == |
242 | Options.AllowedTailExpansions.end()) |
243 | break; |
244 | |
245 | // Remove the last two sequences and replace with the combined sequence |
246 | LoadSequence.pop_back(); |
247 | LoadSequence.pop_back(); |
248 | LoadSequence.emplace_back(Args&: PreLast.Offset, Args&: LoadSize); |
249 | } |
250 | } |
251 | |
252 | // Initialize the basic block structure required for expansion of memcmp call |
253 | // with given maximum load size and memcmp size parameter. |
254 | // This structure includes: |
255 | // 1. A list of load compare blocks - LoadCmpBlocks. |
256 | // 2. An EndBlock, split from original instruction point, which is the block to |
257 | // return from. |
258 | // 3. ResultBlock, block to branch to for early exit when a |
259 | // LoadCmpBlock finds a difference. |
260 | MemCmpExpansion::MemCmpExpansion( |
261 | CallInst *const CI, uint64_t Size, |
262 | const TargetTransformInfo::MemCmpExpansionOptions &Options, |
263 | const bool IsUsedForZeroCmp, const DataLayout &TheDataLayout, |
264 | DomTreeUpdater *DTU) |
265 | : CI(CI), Size(Size), NumLoadsPerBlockForZeroCmp(Options.NumLoadsPerBlock), |
266 | IsUsedForZeroCmp(IsUsedForZeroCmp), DL(TheDataLayout), DTU(DTU), |
267 | Builder(CI) { |
268 | assert(Size > 0 && "zero blocks" ); |
269 | // Scale the max size down if the target can load more bytes than we need. |
270 | llvm::ArrayRef<unsigned> LoadSizes(Options.LoadSizes); |
271 | while (!LoadSizes.empty() && LoadSizes.front() > Size) { |
272 | LoadSizes = LoadSizes.drop_front(); |
273 | } |
274 | assert(!LoadSizes.empty() && "cannot load Size bytes" ); |
275 | MaxLoadSize = LoadSizes.front(); |
276 | // Compute the decomposition. |
277 | unsigned GreedyNumLoadsNonOneByte = 0; |
278 | LoadSequence = computeGreedyLoadSequence(Size, LoadSizes, MaxNumLoads: Options.MaxNumLoads, |
279 | NumLoadsNonOneByte&: GreedyNumLoadsNonOneByte); |
280 | NumLoadsNonOneByte = GreedyNumLoadsNonOneByte; |
281 | assert(LoadSequence.size() <= Options.MaxNumLoads && "broken invariant" ); |
282 | // If we allow overlapping loads and the load sequence is not already optimal, |
283 | // use overlapping loads. |
284 | if (Options.AllowOverlappingLoads && |
285 | (LoadSequence.empty() || LoadSequence.size() > 2)) { |
286 | unsigned OverlappingNumLoadsNonOneByte = 0; |
287 | auto OverlappingLoads = computeOverlappingLoadSequence( |
288 | Size, MaxLoadSize, MaxNumLoads: Options.MaxNumLoads, NumLoadsNonOneByte&: OverlappingNumLoadsNonOneByte); |
289 | if (!OverlappingLoads.empty() && |
290 | (LoadSequence.empty() || |
291 | OverlappingLoads.size() < LoadSequence.size())) { |
292 | LoadSequence = OverlappingLoads; |
293 | NumLoadsNonOneByte = OverlappingNumLoadsNonOneByte; |
294 | } |
295 | } |
296 | assert(LoadSequence.size() <= Options.MaxNumLoads && "broken invariant" ); |
297 | optimiseLoadSequence(LoadSequence, Options, IsUsedForZeroCmp); |
298 | } |
299 | |
300 | unsigned MemCmpExpansion::getNumBlocks() { |
301 | if (IsUsedForZeroCmp) |
302 | return getNumLoads() / NumLoadsPerBlockForZeroCmp + |
303 | (getNumLoads() % NumLoadsPerBlockForZeroCmp != 0 ? 1 : 0); |
304 | return getNumLoads(); |
305 | } |
306 | |
307 | void MemCmpExpansion::createLoadCmpBlocks() { |
308 | for (unsigned i = 0; i < getNumBlocks(); i++) { |
309 | BasicBlock *BB = BasicBlock::Create(Context&: CI->getContext(), Name: "loadbb" , |
310 | Parent: EndBlock->getParent(), InsertBefore: EndBlock); |
311 | LoadCmpBlocks.push_back(x: BB); |
312 | } |
313 | } |
314 | |
315 | void MemCmpExpansion::createResultBlock() { |
316 | ResBlock.BB = BasicBlock::Create(Context&: CI->getContext(), Name: "res_block" , |
317 | Parent: EndBlock->getParent(), InsertBefore: EndBlock); |
318 | } |
319 | |
320 | MemCmpExpansion::LoadPair MemCmpExpansion::getLoadPair(Type *LoadSizeType, |
321 | Type *BSwapSizeType, |
322 | Type *CmpSizeType, |
323 | unsigned OffsetBytes) { |
324 | // Get the memory source at offset `OffsetBytes`. |
325 | Value *LhsSource = CI->getArgOperand(i: 0); |
326 | Value *RhsSource = CI->getArgOperand(i: 1); |
327 | Align LhsAlign = LhsSource->getPointerAlignment(DL); |
328 | Align RhsAlign = RhsSource->getPointerAlignment(DL); |
329 | if (OffsetBytes > 0) { |
330 | auto *ByteType = Type::getInt8Ty(C&: CI->getContext()); |
331 | LhsSource = Builder.CreateConstGEP1_64(Ty: ByteType, Ptr: LhsSource, Idx0: OffsetBytes); |
332 | RhsSource = Builder.CreateConstGEP1_64(Ty: ByteType, Ptr: RhsSource, Idx0: OffsetBytes); |
333 | LhsAlign = commonAlignment(A: LhsAlign, Offset: OffsetBytes); |
334 | RhsAlign = commonAlignment(A: RhsAlign, Offset: OffsetBytes); |
335 | } |
336 | |
337 | // Create a constant or a load from the source. |
338 | Value *Lhs = nullptr; |
339 | if (auto *C = dyn_cast<Constant>(Val: LhsSource)) |
340 | Lhs = ConstantFoldLoadFromConstPtr(C, Ty: LoadSizeType, DL); |
341 | if (!Lhs) |
342 | Lhs = Builder.CreateAlignedLoad(Ty: LoadSizeType, Ptr: LhsSource, Align: LhsAlign); |
343 | |
344 | Value *Rhs = nullptr; |
345 | if (auto *C = dyn_cast<Constant>(Val: RhsSource)) |
346 | Rhs = ConstantFoldLoadFromConstPtr(C, Ty: LoadSizeType, DL); |
347 | if (!Rhs) |
348 | Rhs = Builder.CreateAlignedLoad(Ty: LoadSizeType, Ptr: RhsSource, Align: RhsAlign); |
349 | |
350 | // Zero extend if Byte Swap intrinsic has different type |
351 | if (BSwapSizeType && LoadSizeType != BSwapSizeType) { |
352 | Lhs = Builder.CreateZExt(V: Lhs, DestTy: BSwapSizeType); |
353 | Rhs = Builder.CreateZExt(V: Rhs, DestTy: BSwapSizeType); |
354 | } |
355 | |
356 | // Swap bytes if required. |
357 | if (BSwapSizeType) { |
358 | Function *Bswap = Intrinsic::getDeclaration( |
359 | M: CI->getModule(), id: Intrinsic::bswap, Tys: BSwapSizeType); |
360 | Lhs = Builder.CreateCall(Callee: Bswap, Args: Lhs); |
361 | Rhs = Builder.CreateCall(Callee: Bswap, Args: Rhs); |
362 | } |
363 | |
364 | // Zero extend if required. |
365 | if (CmpSizeType != nullptr && CmpSizeType != Lhs->getType()) { |
366 | Lhs = Builder.CreateZExt(V: Lhs, DestTy: CmpSizeType); |
367 | Rhs = Builder.CreateZExt(V: Rhs, DestTy: CmpSizeType); |
368 | } |
369 | return {.Lhs: Lhs, .Rhs: Rhs}; |
370 | } |
371 | |
372 | // This function creates the IR instructions for loading and comparing 1 byte. |
373 | // It loads 1 byte from each source of the memcmp parameters with the given |
374 | // GEPIndex. It then subtracts the two loaded values and adds this result to the |
375 | // final phi node for selecting the memcmp result. |
376 | void MemCmpExpansion::emitLoadCompareByteBlock(unsigned BlockIndex, |
377 | unsigned OffsetBytes) { |
378 | BasicBlock *BB = LoadCmpBlocks[BlockIndex]; |
379 | Builder.SetInsertPoint(BB); |
380 | const LoadPair Loads = |
381 | getLoadPair(LoadSizeType: Type::getInt8Ty(C&: CI->getContext()), BSwapSizeType: nullptr, |
382 | CmpSizeType: Type::getInt32Ty(C&: CI->getContext()), OffsetBytes); |
383 | Value *Diff = Builder.CreateSub(LHS: Loads.Lhs, RHS: Loads.Rhs); |
384 | |
385 | PhiRes->addIncoming(V: Diff, BB); |
386 | |
387 | if (BlockIndex < (LoadCmpBlocks.size() - 1)) { |
388 | // Early exit branch if difference found to EndBlock. Otherwise, continue to |
389 | // next LoadCmpBlock, |
390 | Value *Cmp = Builder.CreateICmp(P: ICmpInst::ICMP_NE, LHS: Diff, |
391 | RHS: ConstantInt::get(Ty: Diff->getType(), V: 0)); |
392 | BranchInst *CmpBr = |
393 | BranchInst::Create(IfTrue: EndBlock, IfFalse: LoadCmpBlocks[BlockIndex + 1], Cond: Cmp); |
394 | Builder.Insert(I: CmpBr); |
395 | if (DTU) |
396 | DTU->applyUpdates( |
397 | Updates: {{DominatorTree::Insert, BB, EndBlock}, |
398 | {DominatorTree::Insert, BB, LoadCmpBlocks[BlockIndex + 1]}}); |
399 | } else { |
400 | // The last block has an unconditional branch to EndBlock. |
401 | BranchInst *CmpBr = BranchInst::Create(IfTrue: EndBlock); |
402 | Builder.Insert(I: CmpBr); |
403 | if (DTU) |
404 | DTU->applyUpdates(Updates: {{DominatorTree::Insert, BB, EndBlock}}); |
405 | } |
406 | } |
407 | |
408 | /// Generate an equality comparison for one or more pairs of loaded values. |
409 | /// This is used in the case where the memcmp() call is compared equal or not |
410 | /// equal to zero. |
411 | Value *MemCmpExpansion::getCompareLoadPairs(unsigned BlockIndex, |
412 | unsigned &LoadIndex) { |
413 | assert(LoadIndex < getNumLoads() && |
414 | "getCompareLoadPairs() called with no remaining loads" ); |
415 | std::vector<Value *> XorList, OrList; |
416 | Value *Diff = nullptr; |
417 | |
418 | const unsigned NumLoads = |
419 | std::min(a: getNumLoads() - LoadIndex, b: NumLoadsPerBlockForZeroCmp); |
420 | |
421 | // For a single-block expansion, start inserting before the memcmp call. |
422 | if (LoadCmpBlocks.empty()) |
423 | Builder.SetInsertPoint(CI); |
424 | else |
425 | Builder.SetInsertPoint(LoadCmpBlocks[BlockIndex]); |
426 | |
427 | Value *Cmp = nullptr; |
428 | // If we have multiple loads per block, we need to generate a composite |
429 | // comparison using xor+or. The type for the combinations is the largest load |
430 | // type. |
431 | IntegerType *const MaxLoadType = |
432 | NumLoads == 1 ? nullptr |
433 | : IntegerType::get(C&: CI->getContext(), NumBits: MaxLoadSize * 8); |
434 | |
435 | for (unsigned i = 0; i < NumLoads; ++i, ++LoadIndex) { |
436 | const LoadEntry &CurLoadEntry = LoadSequence[LoadIndex]; |
437 | const LoadPair Loads = getLoadPair( |
438 | LoadSizeType: IntegerType::get(C&: CI->getContext(), NumBits: CurLoadEntry.LoadSize * 8), BSwapSizeType: nullptr, |
439 | CmpSizeType: MaxLoadType, OffsetBytes: CurLoadEntry.Offset); |
440 | |
441 | if (NumLoads != 1) { |
442 | // If we have multiple loads per block, we need to generate a composite |
443 | // comparison using xor+or. |
444 | Diff = Builder.CreateXor(LHS: Loads.Lhs, RHS: Loads.Rhs); |
445 | Diff = Builder.CreateZExt(V: Diff, DestTy: MaxLoadType); |
446 | XorList.push_back(x: Diff); |
447 | } else { |
448 | // If there's only one load per block, we just compare the loaded values. |
449 | Cmp = Builder.CreateICmpNE(LHS: Loads.Lhs, RHS: Loads.Rhs); |
450 | } |
451 | } |
452 | |
453 | auto pairWiseOr = [&](std::vector<Value *> &InList) -> std::vector<Value *> { |
454 | std::vector<Value *> OutList; |
455 | for (unsigned i = 0; i < InList.size() - 1; i = i + 2) { |
456 | Value *Or = Builder.CreateOr(LHS: InList[i], RHS: InList[i + 1]); |
457 | OutList.push_back(x: Or); |
458 | } |
459 | if (InList.size() % 2 != 0) |
460 | OutList.push_back(x: InList.back()); |
461 | return OutList; |
462 | }; |
463 | |
464 | if (!Cmp) { |
465 | // Pairwise OR the XOR results. |
466 | OrList = pairWiseOr(XorList); |
467 | |
468 | // Pairwise OR the OR results until one result left. |
469 | while (OrList.size() != 1) { |
470 | OrList = pairWiseOr(OrList); |
471 | } |
472 | |
473 | assert(Diff && "Failed to find comparison diff" ); |
474 | Cmp = Builder.CreateICmpNE(LHS: OrList[0], RHS: ConstantInt::get(Ty: Diff->getType(), V: 0)); |
475 | } |
476 | |
477 | return Cmp; |
478 | } |
479 | |
480 | void MemCmpExpansion::emitLoadCompareBlockMultipleLoads(unsigned BlockIndex, |
481 | unsigned &LoadIndex) { |
482 | Value *Cmp = getCompareLoadPairs(BlockIndex, LoadIndex); |
483 | |
484 | BasicBlock *NextBB = (BlockIndex == (LoadCmpBlocks.size() - 1)) |
485 | ? EndBlock |
486 | : LoadCmpBlocks[BlockIndex + 1]; |
487 | // Early exit branch if difference found to ResultBlock. Otherwise, |
488 | // continue to next LoadCmpBlock or EndBlock. |
489 | BasicBlock *BB = Builder.GetInsertBlock(); |
490 | BranchInst *CmpBr = BranchInst::Create(IfTrue: ResBlock.BB, IfFalse: NextBB, Cond: Cmp); |
491 | Builder.Insert(I: CmpBr); |
492 | if (DTU) |
493 | DTU->applyUpdates(Updates: {{DominatorTree::Insert, BB, ResBlock.BB}, |
494 | {DominatorTree::Insert, BB, NextBB}}); |
495 | |
496 | // Add a phi edge for the last LoadCmpBlock to Endblock with a value of 0 |
497 | // since early exit to ResultBlock was not taken (no difference was found in |
498 | // any of the bytes). |
499 | if (BlockIndex == LoadCmpBlocks.size() - 1) { |
500 | Value *Zero = ConstantInt::get(Ty: Type::getInt32Ty(C&: CI->getContext()), V: 0); |
501 | PhiRes->addIncoming(V: Zero, BB: LoadCmpBlocks[BlockIndex]); |
502 | } |
503 | } |
504 | |
505 | // This function creates the IR intructions for loading and comparing using the |
506 | // given LoadSize. It loads the number of bytes specified by LoadSize from each |
507 | // source of the memcmp parameters. It then does a subtract to see if there was |
508 | // a difference in the loaded values. If a difference is found, it branches |
509 | // with an early exit to the ResultBlock for calculating which source was |
510 | // larger. Otherwise, it falls through to the either the next LoadCmpBlock or |
511 | // the EndBlock if this is the last LoadCmpBlock. Loading 1 byte is handled with |
512 | // a special case through emitLoadCompareByteBlock. The special handling can |
513 | // simply subtract the loaded values and add it to the result phi node. |
514 | void MemCmpExpansion::emitLoadCompareBlock(unsigned BlockIndex) { |
515 | // There is one load per block in this case, BlockIndex == LoadIndex. |
516 | const LoadEntry &CurLoadEntry = LoadSequence[BlockIndex]; |
517 | |
518 | if (CurLoadEntry.LoadSize == 1) { |
519 | MemCmpExpansion::emitLoadCompareByteBlock(BlockIndex, OffsetBytes: CurLoadEntry.Offset); |
520 | return; |
521 | } |
522 | |
523 | Type *LoadSizeType = |
524 | IntegerType::get(C&: CI->getContext(), NumBits: CurLoadEntry.LoadSize * 8); |
525 | Type *BSwapSizeType = |
526 | DL.isLittleEndian() |
527 | ? IntegerType::get(C&: CI->getContext(), |
528 | NumBits: PowerOf2Ceil(A: CurLoadEntry.LoadSize * 8)) |
529 | : nullptr; |
530 | Type *MaxLoadType = IntegerType::get( |
531 | C&: CI->getContext(), |
532 | NumBits: std::max(a: MaxLoadSize, b: (unsigned)PowerOf2Ceil(A: CurLoadEntry.LoadSize)) * 8); |
533 | assert(CurLoadEntry.LoadSize <= MaxLoadSize && "Unexpected load type" ); |
534 | |
535 | Builder.SetInsertPoint(LoadCmpBlocks[BlockIndex]); |
536 | |
537 | const LoadPair Loads = getLoadPair(LoadSizeType, BSwapSizeType, CmpSizeType: MaxLoadType, |
538 | OffsetBytes: CurLoadEntry.Offset); |
539 | |
540 | // Add the loaded values to the phi nodes for calculating memcmp result only |
541 | // if result is not used in a zero equality. |
542 | if (!IsUsedForZeroCmp) { |
543 | ResBlock.PhiSrc1->addIncoming(V: Loads.Lhs, BB: LoadCmpBlocks[BlockIndex]); |
544 | ResBlock.PhiSrc2->addIncoming(V: Loads.Rhs, BB: LoadCmpBlocks[BlockIndex]); |
545 | } |
546 | |
547 | Value *Cmp = Builder.CreateICmp(P: ICmpInst::ICMP_EQ, LHS: Loads.Lhs, RHS: Loads.Rhs); |
548 | BasicBlock *NextBB = (BlockIndex == (LoadCmpBlocks.size() - 1)) |
549 | ? EndBlock |
550 | : LoadCmpBlocks[BlockIndex + 1]; |
551 | // Early exit branch if difference found to ResultBlock. Otherwise, continue |
552 | // to next LoadCmpBlock or EndBlock. |
553 | BasicBlock *BB = Builder.GetInsertBlock(); |
554 | BranchInst *CmpBr = BranchInst::Create(IfTrue: NextBB, IfFalse: ResBlock.BB, Cond: Cmp); |
555 | Builder.Insert(I: CmpBr); |
556 | if (DTU) |
557 | DTU->applyUpdates(Updates: {{DominatorTree::Insert, BB, NextBB}, |
558 | {DominatorTree::Insert, BB, ResBlock.BB}}); |
559 | |
560 | // Add a phi edge for the last LoadCmpBlock to Endblock with a value of 0 |
561 | // since early exit to ResultBlock was not taken (no difference was found in |
562 | // any of the bytes). |
563 | if (BlockIndex == LoadCmpBlocks.size() - 1) { |
564 | Value *Zero = ConstantInt::get(Ty: Type::getInt32Ty(C&: CI->getContext()), V: 0); |
565 | PhiRes->addIncoming(V: Zero, BB: LoadCmpBlocks[BlockIndex]); |
566 | } |
567 | } |
568 | |
569 | // This function populates the ResultBlock with a sequence to calculate the |
570 | // memcmp result. It compares the two loaded source values and returns -1 if |
571 | // src1 < src2 and 1 if src1 > src2. |
572 | void MemCmpExpansion::emitMemCmpResultBlock() { |
573 | // Special case: if memcmp result is used in a zero equality, result does not |
574 | // need to be calculated and can simply return 1. |
575 | if (IsUsedForZeroCmp) { |
576 | BasicBlock::iterator InsertPt = ResBlock.BB->getFirstInsertionPt(); |
577 | Builder.SetInsertPoint(TheBB: ResBlock.BB, IP: InsertPt); |
578 | Value *Res = ConstantInt::get(Ty: Type::getInt32Ty(C&: CI->getContext()), V: 1); |
579 | PhiRes->addIncoming(V: Res, BB: ResBlock.BB); |
580 | BranchInst *NewBr = BranchInst::Create(IfTrue: EndBlock); |
581 | Builder.Insert(I: NewBr); |
582 | if (DTU) |
583 | DTU->applyUpdates(Updates: {{DominatorTree::Insert, ResBlock.BB, EndBlock}}); |
584 | return; |
585 | } |
586 | BasicBlock::iterator InsertPt = ResBlock.BB->getFirstInsertionPt(); |
587 | Builder.SetInsertPoint(TheBB: ResBlock.BB, IP: InsertPt); |
588 | |
589 | Value *Cmp = Builder.CreateICmp(P: ICmpInst::ICMP_ULT, LHS: ResBlock.PhiSrc1, |
590 | RHS: ResBlock.PhiSrc2); |
591 | |
592 | Value *Res = |
593 | Builder.CreateSelect(C: Cmp, True: ConstantInt::get(Ty: Builder.getInt32Ty(), V: -1), |
594 | False: ConstantInt::get(Ty: Builder.getInt32Ty(), V: 1)); |
595 | |
596 | PhiRes->addIncoming(V: Res, BB: ResBlock.BB); |
597 | BranchInst *NewBr = BranchInst::Create(IfTrue: EndBlock); |
598 | Builder.Insert(I: NewBr); |
599 | if (DTU) |
600 | DTU->applyUpdates(Updates: {{DominatorTree::Insert, ResBlock.BB, EndBlock}}); |
601 | } |
602 | |
603 | void MemCmpExpansion::setupResultBlockPHINodes() { |
604 | Type *MaxLoadType = IntegerType::get(C&: CI->getContext(), NumBits: MaxLoadSize * 8); |
605 | Builder.SetInsertPoint(ResBlock.BB); |
606 | // Note: this assumes one load per block. |
607 | ResBlock.PhiSrc1 = |
608 | Builder.CreatePHI(Ty: MaxLoadType, NumReservedValues: NumLoadsNonOneByte, Name: "phi.src1" ); |
609 | ResBlock.PhiSrc2 = |
610 | Builder.CreatePHI(Ty: MaxLoadType, NumReservedValues: NumLoadsNonOneByte, Name: "phi.src2" ); |
611 | } |
612 | |
613 | void MemCmpExpansion::setupEndBlockPHINodes() { |
614 | Builder.SetInsertPoint(TheBB: EndBlock, IP: EndBlock->begin()); |
615 | PhiRes = Builder.CreatePHI(Ty: Type::getInt32Ty(C&: CI->getContext()), NumReservedValues: 2, Name: "phi.res" ); |
616 | } |
617 | |
618 | Value *MemCmpExpansion::getMemCmpExpansionZeroCase() { |
619 | unsigned LoadIndex = 0; |
620 | // This loop populates each of the LoadCmpBlocks with the IR sequence to |
621 | // handle multiple loads per block. |
622 | for (unsigned I = 0; I < getNumBlocks(); ++I) { |
623 | emitLoadCompareBlockMultipleLoads(BlockIndex: I, LoadIndex); |
624 | } |
625 | |
626 | emitMemCmpResultBlock(); |
627 | return PhiRes; |
628 | } |
629 | |
630 | /// A memcmp expansion that compares equality with 0 and only has one block of |
631 | /// load and compare can bypass the compare, branch, and phi IR that is required |
632 | /// in the general case. |
633 | Value *MemCmpExpansion::getMemCmpEqZeroOneBlock() { |
634 | unsigned LoadIndex = 0; |
635 | Value *Cmp = getCompareLoadPairs(BlockIndex: 0, LoadIndex); |
636 | assert(LoadIndex == getNumLoads() && "some entries were not consumed" ); |
637 | return Builder.CreateZExt(V: Cmp, DestTy: Type::getInt32Ty(C&: CI->getContext())); |
638 | } |
639 | |
640 | /// A memcmp expansion that only has one block of load and compare can bypass |
641 | /// the compare, branch, and phi IR that is required in the general case. |
642 | /// This function also analyses users of memcmp, and if there is only one user |
643 | /// from which we can conclude that only 2 out of 3 memcmp outcomes really |
644 | /// matter, then it generates more efficient code with only one comparison. |
645 | Value *MemCmpExpansion::getMemCmpOneBlock() { |
646 | bool NeedsBSwap = DL.isLittleEndian() && Size != 1; |
647 | Type *LoadSizeType = IntegerType::get(C&: CI->getContext(), NumBits: Size * 8); |
648 | Type *BSwapSizeType = |
649 | NeedsBSwap ? IntegerType::get(C&: CI->getContext(), NumBits: PowerOf2Ceil(A: Size * 8)) |
650 | : nullptr; |
651 | Type *MaxLoadType = |
652 | IntegerType::get(C&: CI->getContext(), |
653 | NumBits: std::max(a: MaxLoadSize, b: (unsigned)PowerOf2Ceil(A: Size)) * 8); |
654 | |
655 | // The i8 and i16 cases don't need compares. We zext the loaded values and |
656 | // subtract them to get the suitable negative, zero, or positive i32 result. |
657 | if (Size == 1 || Size == 2) { |
658 | const LoadPair Loads = getLoadPair(LoadSizeType, BSwapSizeType, |
659 | CmpSizeType: Builder.getInt32Ty(), /*Offset*/ OffsetBytes: 0); |
660 | return Builder.CreateSub(LHS: Loads.Lhs, RHS: Loads.Rhs); |
661 | } |
662 | |
663 | const LoadPair Loads = getLoadPair(LoadSizeType, BSwapSizeType, CmpSizeType: MaxLoadType, |
664 | /*Offset*/ OffsetBytes: 0); |
665 | |
666 | // If a user of memcmp cares only about two outcomes, for example: |
667 | // bool result = memcmp(a, b, NBYTES) > 0; |
668 | // We can generate more optimal code with a smaller number of operations |
669 | if (CI->hasOneUser()) { |
670 | auto *UI = cast<Instruction>(Val: *CI->user_begin()); |
671 | ICmpInst::Predicate Pred = ICmpInst::Predicate::BAD_ICMP_PREDICATE; |
672 | uint64_t Shift; |
673 | bool NeedsZExt = false; |
674 | // This is a special case because instead of checking if the result is less |
675 | // than zero: |
676 | // bool result = memcmp(a, b, NBYTES) < 0; |
677 | // Compiler is clever enough to generate the following code: |
678 | // bool result = memcmp(a, b, NBYTES) >> 31; |
679 | if (match(V: UI, P: m_LShr(L: m_Value(), R: m_ConstantInt(V&: Shift))) && |
680 | Shift == (CI->getType()->getIntegerBitWidth() - 1)) { |
681 | Pred = ICmpInst::ICMP_SLT; |
682 | NeedsZExt = true; |
683 | } else { |
684 | // In case of a successful match this call will set `Pred` variable |
685 | match(V: UI, P: m_ICmp(Pred, L: m_Specific(V: CI), R: m_Zero())); |
686 | } |
687 | // Generate new code and remove the original memcmp call and the user |
688 | if (ICmpInst::isSigned(predicate: Pred)) { |
689 | Value *Cmp = Builder.CreateICmp(P: CmpInst::getUnsignedPredicate(pred: Pred), |
690 | LHS: Loads.Lhs, RHS: Loads.Rhs); |
691 | auto *Result = NeedsZExt ? Builder.CreateZExt(V: Cmp, DestTy: UI->getType()) : Cmp; |
692 | UI->replaceAllUsesWith(V: Result); |
693 | UI->eraseFromParent(); |
694 | CI->eraseFromParent(); |
695 | return nullptr; |
696 | } |
697 | } |
698 | |
699 | // The result of memcmp is negative, zero, or positive, so produce that by |
700 | // subtracting 2 extended compare bits: sub (ugt, ult). |
701 | // If a target prefers to use selects to get -1/0/1, they should be able |
702 | // to transform this later. The inverse transform (going from selects to math) |
703 | // may not be possible in the DAG because the selects got converted into |
704 | // branches before we got there. |
705 | Value *CmpUGT = Builder.CreateICmpUGT(LHS: Loads.Lhs, RHS: Loads.Rhs); |
706 | Value *CmpULT = Builder.CreateICmpULT(LHS: Loads.Lhs, RHS: Loads.Rhs); |
707 | Value *ZextUGT = Builder.CreateZExt(V: CmpUGT, DestTy: Builder.getInt32Ty()); |
708 | Value *ZextULT = Builder.CreateZExt(V: CmpULT, DestTy: Builder.getInt32Ty()); |
709 | return Builder.CreateSub(LHS: ZextUGT, RHS: ZextULT); |
710 | } |
711 | |
712 | // This function expands the memcmp call into an inline expansion and returns |
713 | // the memcmp result. Returns nullptr if the memcmp is already replaced. |
714 | Value *MemCmpExpansion::getMemCmpExpansion() { |
715 | // Create the basic block framework for a multi-block expansion. |
716 | if (getNumBlocks() != 1) { |
717 | BasicBlock *StartBlock = CI->getParent(); |
718 | EndBlock = SplitBlock(Old: StartBlock, SplitPt: CI, DTU, /*LI=*/nullptr, |
719 | /*MSSAU=*/nullptr, BBName: "endblock" ); |
720 | setupEndBlockPHINodes(); |
721 | createResultBlock(); |
722 | |
723 | // If return value of memcmp is not used in a zero equality, we need to |
724 | // calculate which source was larger. The calculation requires the |
725 | // two loaded source values of each load compare block. |
726 | // These will be saved in the phi nodes created by setupResultBlockPHINodes. |
727 | if (!IsUsedForZeroCmp) setupResultBlockPHINodes(); |
728 | |
729 | // Create the number of required load compare basic blocks. |
730 | createLoadCmpBlocks(); |
731 | |
732 | // Update the terminator added by SplitBlock to branch to the first |
733 | // LoadCmpBlock. |
734 | StartBlock->getTerminator()->setSuccessor(Idx: 0, BB: LoadCmpBlocks[0]); |
735 | if (DTU) |
736 | DTU->applyUpdates(Updates: {{DominatorTree::Insert, StartBlock, LoadCmpBlocks[0]}, |
737 | {DominatorTree::Delete, StartBlock, EndBlock}}); |
738 | } |
739 | |
740 | Builder.SetCurrentDebugLocation(CI->getDebugLoc()); |
741 | |
742 | if (IsUsedForZeroCmp) |
743 | return getNumBlocks() == 1 ? getMemCmpEqZeroOneBlock() |
744 | : getMemCmpExpansionZeroCase(); |
745 | |
746 | if (getNumBlocks() == 1) |
747 | return getMemCmpOneBlock(); |
748 | |
749 | for (unsigned I = 0; I < getNumBlocks(); ++I) { |
750 | emitLoadCompareBlock(BlockIndex: I); |
751 | } |
752 | |
753 | emitMemCmpResultBlock(); |
754 | return PhiRes; |
755 | } |
756 | |
757 | // This function checks to see if an expansion of memcmp can be generated. |
758 | // It checks for constant compare size that is less than the max inline size. |
759 | // If an expansion cannot occur, returns false to leave as a library call. |
760 | // Otherwise, the library call is replaced with a new IR instruction sequence. |
761 | /// We want to transform: |
762 | /// %call = call signext i32 @memcmp(i8* %0, i8* %1, i64 15) |
763 | /// To: |
764 | /// loadbb: |
765 | /// %0 = bitcast i32* %buffer2 to i8* |
766 | /// %1 = bitcast i32* %buffer1 to i8* |
767 | /// %2 = bitcast i8* %1 to i64* |
768 | /// %3 = bitcast i8* %0 to i64* |
769 | /// %4 = load i64, i64* %2 |
770 | /// %5 = load i64, i64* %3 |
771 | /// %6 = call i64 @llvm.bswap.i64(i64 %4) |
772 | /// %7 = call i64 @llvm.bswap.i64(i64 %5) |
773 | /// %8 = sub i64 %6, %7 |
774 | /// %9 = icmp ne i64 %8, 0 |
775 | /// br i1 %9, label %res_block, label %loadbb1 |
776 | /// res_block: ; preds = %loadbb2, |
777 | /// %loadbb1, %loadbb |
778 | /// %phi.src1 = phi i64 [ %6, %loadbb ], [ %22, %loadbb1 ], [ %36, %loadbb2 ] |
779 | /// %phi.src2 = phi i64 [ %7, %loadbb ], [ %23, %loadbb1 ], [ %37, %loadbb2 ] |
780 | /// %10 = icmp ult i64 %phi.src1, %phi.src2 |
781 | /// %11 = select i1 %10, i32 -1, i32 1 |
782 | /// br label %endblock |
783 | /// loadbb1: ; preds = %loadbb |
784 | /// %12 = bitcast i32* %buffer2 to i8* |
785 | /// %13 = bitcast i32* %buffer1 to i8* |
786 | /// %14 = bitcast i8* %13 to i32* |
787 | /// %15 = bitcast i8* %12 to i32* |
788 | /// %16 = getelementptr i32, i32* %14, i32 2 |
789 | /// %17 = getelementptr i32, i32* %15, i32 2 |
790 | /// %18 = load i32, i32* %16 |
791 | /// %19 = load i32, i32* %17 |
792 | /// %20 = call i32 @llvm.bswap.i32(i32 %18) |
793 | /// %21 = call i32 @llvm.bswap.i32(i32 %19) |
794 | /// %22 = zext i32 %20 to i64 |
795 | /// %23 = zext i32 %21 to i64 |
796 | /// %24 = sub i64 %22, %23 |
797 | /// %25 = icmp ne i64 %24, 0 |
798 | /// br i1 %25, label %res_block, label %loadbb2 |
799 | /// loadbb2: ; preds = %loadbb1 |
800 | /// %26 = bitcast i32* %buffer2 to i8* |
801 | /// %27 = bitcast i32* %buffer1 to i8* |
802 | /// %28 = bitcast i8* %27 to i16* |
803 | /// %29 = bitcast i8* %26 to i16* |
804 | /// %30 = getelementptr i16, i16* %28, i16 6 |
805 | /// %31 = getelementptr i16, i16* %29, i16 6 |
806 | /// %32 = load i16, i16* %30 |
807 | /// %33 = load i16, i16* %31 |
808 | /// %34 = call i16 @llvm.bswap.i16(i16 %32) |
809 | /// %35 = call i16 @llvm.bswap.i16(i16 %33) |
810 | /// %36 = zext i16 %34 to i64 |
811 | /// %37 = zext i16 %35 to i64 |
812 | /// %38 = sub i64 %36, %37 |
813 | /// %39 = icmp ne i64 %38, 0 |
814 | /// br i1 %39, label %res_block, label %loadbb3 |
815 | /// loadbb3: ; preds = %loadbb2 |
816 | /// %40 = bitcast i32* %buffer2 to i8* |
817 | /// %41 = bitcast i32* %buffer1 to i8* |
818 | /// %42 = getelementptr i8, i8* %41, i8 14 |
819 | /// %43 = getelementptr i8, i8* %40, i8 14 |
820 | /// %44 = load i8, i8* %42 |
821 | /// %45 = load i8, i8* %43 |
822 | /// %46 = zext i8 %44 to i32 |
823 | /// %47 = zext i8 %45 to i32 |
824 | /// %48 = sub i32 %46, %47 |
825 | /// br label %endblock |
826 | /// endblock: ; preds = %res_block, |
827 | /// %loadbb3 |
828 | /// %phi.res = phi i32 [ %48, %loadbb3 ], [ %11, %res_block ] |
829 | /// ret i32 %phi.res |
830 | static bool expandMemCmp(CallInst *CI, const TargetTransformInfo *TTI, |
831 | const TargetLowering *TLI, const DataLayout *DL, |
832 | ProfileSummaryInfo *PSI, BlockFrequencyInfo *BFI, |
833 | DomTreeUpdater *DTU, const bool IsBCmp) { |
834 | NumMemCmpCalls++; |
835 | |
836 | // Early exit from expansion if -Oz. |
837 | if (CI->getFunction()->hasMinSize()) |
838 | return false; |
839 | |
840 | // Early exit from expansion if size is not a constant. |
841 | ConstantInt *SizeCast = dyn_cast<ConstantInt>(Val: CI->getArgOperand(i: 2)); |
842 | if (!SizeCast) { |
843 | NumMemCmpNotConstant++; |
844 | return false; |
845 | } |
846 | const uint64_t SizeVal = SizeCast->getZExtValue(); |
847 | |
848 | if (SizeVal == 0) { |
849 | return false; |
850 | } |
851 | // TTI call to check if target would like to expand memcmp. Also, get the |
852 | // available load sizes. |
853 | const bool IsUsedForZeroCmp = |
854 | IsBCmp || isOnlyUsedInZeroEqualityComparison(CxtI: CI); |
855 | bool OptForSize = CI->getFunction()->hasOptSize() || |
856 | llvm::shouldOptimizeForSize(BB: CI->getParent(), PSI, BFI); |
857 | auto Options = TTI->enableMemCmpExpansion(OptSize: OptForSize, |
858 | IsZeroCmp: IsUsedForZeroCmp); |
859 | if (!Options) return false; |
860 | |
861 | if (MemCmpEqZeroNumLoadsPerBlock.getNumOccurrences()) |
862 | Options.NumLoadsPerBlock = MemCmpEqZeroNumLoadsPerBlock; |
863 | |
864 | if (OptForSize && |
865 | MaxLoadsPerMemcmpOptSize.getNumOccurrences()) |
866 | Options.MaxNumLoads = MaxLoadsPerMemcmpOptSize; |
867 | |
868 | if (!OptForSize && MaxLoadsPerMemcmp.getNumOccurrences()) |
869 | Options.MaxNumLoads = MaxLoadsPerMemcmp; |
870 | |
871 | MemCmpExpansion Expansion(CI, SizeVal, Options, IsUsedForZeroCmp, *DL, DTU); |
872 | |
873 | // Don't expand if this will require more loads than desired by the target. |
874 | if (Expansion.getNumLoads() == 0) { |
875 | NumMemCmpGreaterThanMax++; |
876 | return false; |
877 | } |
878 | |
879 | NumMemCmpInlined++; |
880 | |
881 | if (Value *Res = Expansion.getMemCmpExpansion()) { |
882 | // Replace call with result of expansion and erase call. |
883 | CI->replaceAllUsesWith(V: Res); |
884 | CI->eraseFromParent(); |
885 | } |
886 | |
887 | return true; |
888 | } |
889 | |
890 | // Returns true if a change was made. |
891 | static bool runOnBlock(BasicBlock &BB, const TargetLibraryInfo *TLI, |
892 | const TargetTransformInfo *TTI, const TargetLowering *TL, |
893 | const DataLayout &DL, ProfileSummaryInfo *PSI, |
894 | BlockFrequencyInfo *BFI, DomTreeUpdater *DTU); |
895 | |
896 | static PreservedAnalyses runImpl(Function &F, const TargetLibraryInfo *TLI, |
897 | const TargetTransformInfo *TTI, |
898 | const TargetLowering *TL, |
899 | ProfileSummaryInfo *PSI, |
900 | BlockFrequencyInfo *BFI, DominatorTree *DT); |
901 | |
902 | class ExpandMemCmpLegacyPass : public FunctionPass { |
903 | public: |
904 | static char ID; |
905 | |
906 | ExpandMemCmpLegacyPass() : FunctionPass(ID) { |
907 | initializeExpandMemCmpLegacyPassPass(*PassRegistry::getPassRegistry()); |
908 | } |
909 | |
910 | bool runOnFunction(Function &F) override { |
911 | if (skipFunction(F)) return false; |
912 | |
913 | auto *TPC = getAnalysisIfAvailable<TargetPassConfig>(); |
914 | if (!TPC) { |
915 | return false; |
916 | } |
917 | const TargetLowering* TL = |
918 | TPC->getTM<TargetMachine>().getSubtargetImpl(F)->getTargetLowering(); |
919 | |
920 | const TargetLibraryInfo *TLI = |
921 | &getAnalysis<TargetLibraryInfoWrapperPass>().getTLI(F); |
922 | const TargetTransformInfo *TTI = |
923 | &getAnalysis<TargetTransformInfoWrapperPass>().getTTI(F); |
924 | auto *PSI = &getAnalysis<ProfileSummaryInfoWrapperPass>().getPSI(); |
925 | auto *BFI = (PSI && PSI->hasProfileSummary()) ? |
926 | &getAnalysis<LazyBlockFrequencyInfoPass>().getBFI() : |
927 | nullptr; |
928 | DominatorTree *DT = nullptr; |
929 | if (auto *DTWP = getAnalysisIfAvailable<DominatorTreeWrapperPass>()) |
930 | DT = &DTWP->getDomTree(); |
931 | auto PA = runImpl(F, TLI, TTI, TL, PSI, BFI, DT); |
932 | return !PA.areAllPreserved(); |
933 | } |
934 | |
935 | private: |
936 | void getAnalysisUsage(AnalysisUsage &AU) const override { |
937 | AU.addRequired<TargetLibraryInfoWrapperPass>(); |
938 | AU.addRequired<TargetTransformInfoWrapperPass>(); |
939 | AU.addRequired<ProfileSummaryInfoWrapperPass>(); |
940 | AU.addPreserved<DominatorTreeWrapperPass>(); |
941 | LazyBlockFrequencyInfoPass::getLazyBFIAnalysisUsage(AU); |
942 | FunctionPass::getAnalysisUsage(AU); |
943 | } |
944 | }; |
945 | |
946 | bool runOnBlock(BasicBlock &BB, const TargetLibraryInfo *TLI, |
947 | const TargetTransformInfo *TTI, const TargetLowering *TL, |
948 | const DataLayout &DL, ProfileSummaryInfo *PSI, |
949 | BlockFrequencyInfo *BFI, DomTreeUpdater *DTU) { |
950 | for (Instruction &I : BB) { |
951 | CallInst *CI = dyn_cast<CallInst>(Val: &I); |
952 | if (!CI) { |
953 | continue; |
954 | } |
955 | LibFunc Func; |
956 | if (TLI->getLibFunc(CB: *CI, F&: Func) && |
957 | (Func == LibFunc_memcmp || Func == LibFunc_bcmp) && |
958 | expandMemCmp(CI, TTI, TLI: TL, DL: &DL, PSI, BFI, DTU, IsBCmp: Func == LibFunc_bcmp)) { |
959 | return true; |
960 | } |
961 | } |
962 | return false; |
963 | } |
964 | |
965 | PreservedAnalyses runImpl(Function &F, const TargetLibraryInfo *TLI, |
966 | const TargetTransformInfo *TTI, |
967 | const TargetLowering *TL, ProfileSummaryInfo *PSI, |
968 | BlockFrequencyInfo *BFI, DominatorTree *DT) { |
969 | std::optional<DomTreeUpdater> DTU; |
970 | if (DT) |
971 | DTU.emplace(args&: DT, args: DomTreeUpdater::UpdateStrategy::Lazy); |
972 | |
973 | const DataLayout& DL = F.getDataLayout(); |
974 | bool MadeChanges = false; |
975 | for (auto BBIt = F.begin(); BBIt != F.end();) { |
976 | if (runOnBlock(BB&: *BBIt, TLI, TTI, TL, DL, PSI, BFI, DTU: DTU ? &*DTU : nullptr)) { |
977 | MadeChanges = true; |
978 | // If changes were made, restart the function from the beginning, since |
979 | // the structure of the function was changed. |
980 | BBIt = F.begin(); |
981 | } else { |
982 | ++BBIt; |
983 | } |
984 | } |
985 | if (MadeChanges) |
986 | for (BasicBlock &BB : F) |
987 | SimplifyInstructionsInBlock(BB: &BB); |
988 | if (!MadeChanges) |
989 | return PreservedAnalyses::all(); |
990 | PreservedAnalyses PA; |
991 | PA.preserve<DominatorTreeAnalysis>(); |
992 | return PA; |
993 | } |
994 | |
995 | } // namespace |
996 | |
997 | PreservedAnalyses ExpandMemCmpPass::run(Function &F, |
998 | FunctionAnalysisManager &FAM) { |
999 | const auto *TL = TM->getSubtargetImpl(F)->getTargetLowering(); |
1000 | const auto &TLI = FAM.getResult<TargetLibraryAnalysis>(IR&: F); |
1001 | const auto &TTI = FAM.getResult<TargetIRAnalysis>(IR&: F); |
1002 | auto *PSI = FAM.getResult<ModuleAnalysisManagerFunctionProxy>(IR&: F) |
1003 | .getCachedResult<ProfileSummaryAnalysis>(IR&: *F.getParent()); |
1004 | BlockFrequencyInfo *BFI = (PSI && PSI->hasProfileSummary()) |
1005 | ? &FAM.getResult<BlockFrequencyAnalysis>(IR&: F) |
1006 | : nullptr; |
1007 | auto *DT = FAM.getCachedResult<DominatorTreeAnalysis>(IR&: F); |
1008 | |
1009 | return runImpl(F, TLI: &TLI, TTI: &TTI, TL, PSI, BFI, DT); |
1010 | } |
1011 | |
1012 | char ExpandMemCmpLegacyPass::ID = 0; |
1013 | INITIALIZE_PASS_BEGIN(ExpandMemCmpLegacyPass, DEBUG_TYPE, |
1014 | "Expand memcmp() to load/stores" , false, false) |
1015 | INITIALIZE_PASS_DEPENDENCY(TargetLibraryInfoWrapperPass) |
1016 | INITIALIZE_PASS_DEPENDENCY(TargetTransformInfoWrapperPass) |
1017 | INITIALIZE_PASS_DEPENDENCY(LazyBlockFrequencyInfoPass) |
1018 | INITIALIZE_PASS_DEPENDENCY(ProfileSummaryInfoWrapperPass) |
1019 | INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass) |
1020 | INITIALIZE_PASS_END(ExpandMemCmpLegacyPass, DEBUG_TYPE, |
1021 | "Expand memcmp() to load/stores" , false, false) |
1022 | |
1023 | FunctionPass *llvm::createExpandMemCmpLegacyPass() { |
1024 | return new ExpandMemCmpLegacyPass(); |
1025 | } |
1026 | |