1 //===- LoopUnroll.cpp - Loop unroller pass --------------------------------===//
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 // This pass implements a simple loop unroller. It works best when loops have
11 // been canonicalized by the -indvars pass, allowing it to determine the trip
12 // counts of loops easily.
13 //===----------------------------------------------------------------------===//
15 #include "llvm/Transforms/Scalar/LoopUnrollPass.h"
16 #include "llvm/ADT/DenseMap.h"
17 #include "llvm/ADT/DenseMapInfo.h"
18 #include "llvm/ADT/DenseSet.h"
19 #include "llvm/ADT/None.h"
20 #include "llvm/ADT/Optional.h"
21 #include "llvm/ADT/STLExtras.h"
22 #include "llvm/ADT/SetVector.h"
23 #include "llvm/ADT/SmallPtrSet.h"
24 #include "llvm/ADT/SmallVector.h"
25 #include "llvm/ADT/StringRef.h"
26 #include "llvm/Analysis/AssumptionCache.h"
27 #include "llvm/Analysis/CodeMetrics.h"
28 #include "llvm/Analysis/LoopAnalysisManager.h"
29 #include "llvm/Analysis/LoopInfo.h"
30 #include "llvm/Analysis/LoopPass.h"
31 #include "llvm/Analysis/LoopUnrollAnalyzer.h"
32 #include "llvm/Analysis/OptimizationRemarkEmitter.h"
33 #include "llvm/Analysis/ProfileSummaryInfo.h"
34 #include "llvm/Analysis/ScalarEvolution.h"
35 #include "llvm/Analysis/TargetTransformInfo.h"
36 #include "llvm/IR/BasicBlock.h"
37 #include "llvm/IR/CFG.h"
38 #include "llvm/IR/Constant.h"
39 #include "llvm/IR/Constants.h"
40 #include "llvm/IR/DiagnosticInfo.h"
41 #include "llvm/IR/Dominators.h"
42 #include "llvm/IR/Function.h"
43 #include "llvm/IR/Instruction.h"
44 #include "llvm/IR/Instructions.h"
45 #include "llvm/IR/IntrinsicInst.h"
46 #include "llvm/IR/Metadata.h"
47 #include "llvm/IR/PassManager.h"
48 #include "llvm/Pass.h"
49 #include "llvm/Support/Casting.h"
50 #include "llvm/Support/CommandLine.h"
51 #include "llvm/Support/Debug.h"
52 #include "llvm/Support/ErrorHandling.h"
53 #include "llvm/Support/raw_ostream.h"
54 #include "llvm/Transforms/Scalar.h"
55 #include "llvm/Transforms/Scalar/LoopPassManager.h"
56 #include "llvm/Transforms/Utils.h"
57 #include "llvm/Transforms/Utils/LoopSimplify.h"
58 #include "llvm/Transforms/Utils/LoopUtils.h"
59 #include "llvm/Transforms/Utils/UnrollLoop.h"
70 #define DEBUG_TYPE "loop-unroll"
72 static cl::opt<unsigned>
73 UnrollThreshold("unroll-threshold", cl::Hidden,
74 cl::desc("The cost threshold for loop unrolling"));
76 static cl::opt<unsigned> UnrollPartialThreshold(
77 "unroll-partial-threshold", cl::Hidden,
78 cl::desc("The cost threshold for partial loop unrolling"));
80 static cl::opt<unsigned> UnrollMaxPercentThresholdBoost(
81 "unroll-max-percent-threshold-boost", cl::init(400), cl::Hidden,
82 cl::desc("The maximum 'boost' (represented as a percentage >= 100) applied "
83 "to the threshold when aggressively unrolling a loop due to the "
84 "dynamic cost savings. If completely unrolling a loop will reduce "
85 "the total runtime from X to Y, we boost the loop unroll "
86 "threshold to DefaultThreshold*std::min(MaxPercentThresholdBoost, "
87 "X/Y). This limit avoids excessive code bloat."));
89 static cl::opt<unsigned> UnrollMaxIterationsCountToAnalyze(
90 "unroll-max-iteration-count-to-analyze", cl::init(10), cl::Hidden,
91 cl::desc("Don't allow loop unrolling to simulate more than this number of"
92 "iterations when checking full unroll profitability"));
94 static cl::opt<unsigned> UnrollCount(
95 "unroll-count", cl::Hidden,
96 cl::desc("Use this unroll count for all loops including those with "
97 "unroll_count pragma values, for testing purposes"));
99 static cl::opt<unsigned> UnrollMaxCount(
100 "unroll-max-count", cl::Hidden,
101 cl::desc("Set the max unroll count for partial and runtime unrolling, for"
102 "testing purposes"));
104 static cl::opt<unsigned> UnrollFullMaxCount(
105 "unroll-full-max-count", cl::Hidden,
107 "Set the max unroll count for full unrolling, for testing purposes"));
109 static cl::opt<unsigned> UnrollPeelCount(
110 "unroll-peel-count", cl::Hidden,
111 cl::desc("Set the unroll peeling count, for testing purposes"));
114 UnrollAllowPartial("unroll-allow-partial", cl::Hidden,
115 cl::desc("Allows loops to be partially unrolled until "
116 "-unroll-threshold loop size is reached."));
118 static cl::opt<bool> UnrollAllowRemainder(
119 "unroll-allow-remainder", cl::Hidden,
120 cl::desc("Allow generation of a loop remainder (extra iterations) "
121 "when unrolling a loop."));
124 UnrollRuntime("unroll-runtime", cl::ZeroOrMore, cl::Hidden,
125 cl::desc("Unroll loops with run-time trip counts"));
127 static cl::opt<unsigned> UnrollMaxUpperBound(
128 "unroll-max-upperbound", cl::init(8), cl::Hidden,
130 "The max of trip count upper bound that is considered in unrolling"));
132 static cl::opt<unsigned> PragmaUnrollThreshold(
133 "pragma-unroll-threshold", cl::init(16 * 1024), cl::Hidden,
134 cl::desc("Unrolled size limit for loops with an unroll(full) or "
135 "unroll_count pragma."));
137 static cl::opt<unsigned> FlatLoopTripCountThreshold(
138 "flat-loop-tripcount-threshold", cl::init(5), cl::Hidden,
139 cl::desc("If the runtime tripcount for the loop is lower than the "
140 "threshold, the loop is considered as flat and will be less "
141 "aggressively unrolled."));
144 UnrollAllowPeeling("unroll-allow-peeling", cl::init(true), cl::Hidden,
145 cl::desc("Allows loops to be peeled when the dynamic "
146 "trip count is known to be low."));
148 static cl::opt<bool> UnrollUnrollRemainder(
149 "unroll-remainder", cl::Hidden,
150 cl::desc("Allow the loop remainder to be unrolled."));
152 // This option isn't ever intended to be enabled, it serves to allow
153 // experiments to check the assumptions about when this kind of revisit is
155 static cl::opt<bool> UnrollRevisitChildLoops(
156 "unroll-revisit-child-loops", cl::Hidden,
157 cl::desc("Enqueue and re-visit child loops in the loop PM after unrolling. "
158 "This shouldn't typically be needed as child loops (or their "
159 "clones) were already visited."));
161 /// A magic value for use with the Threshold parameter to indicate
162 /// that the loop unroll should be performed regardless of how much
163 /// code expansion would result.
164 static const unsigned NoThreshold = std::numeric_limits<unsigned>::max();
166 /// Gather the various unrolling parameters based on the defaults, compiler
167 /// flags, TTI overrides and user specified parameters.
168 TargetTransformInfo::UnrollingPreferences llvm::gatherUnrollingPreferences(
169 Loop *L, ScalarEvolution &SE, const TargetTransformInfo &TTI, int OptLevel,
170 Optional<unsigned> UserThreshold, Optional<unsigned> UserCount,
171 Optional<bool> UserAllowPartial, Optional<bool> UserRuntime,
172 Optional<bool> UserUpperBound, Optional<bool> UserAllowPeeling) {
173 TargetTransformInfo::UnrollingPreferences UP;
175 // Set up the defaults
176 UP.Threshold = OptLevel > 2 ? 300 : 150;
177 UP.MaxPercentThresholdBoost = 400;
178 UP.OptSizeThreshold = 0;
179 UP.PartialThreshold = 150;
180 UP.PartialOptSizeThreshold = 0;
183 UP.DefaultUnrollRuntimeCount = 8;
184 UP.MaxCount = std::numeric_limits<unsigned>::max();
185 UP.FullUnrollMaxCount = std::numeric_limits<unsigned>::max();
189 UP.AllowRemainder = true;
190 UP.UnrollRemainder = false;
191 UP.AllowExpensiveTripCount = false;
193 UP.UpperBound = false;
194 UP.AllowPeeling = true;
195 UP.UnrollAndJam = false;
196 UP.UnrollAndJamInnerLoopThreshold = 60;
198 // Override with any target specific settings
199 TTI.getUnrollingPreferences(L, SE, UP);
201 // Apply size attributes
202 if (L->getHeader()->getParent()->optForSize()) {
203 UP.Threshold = UP.OptSizeThreshold;
204 UP.PartialThreshold = UP.PartialOptSizeThreshold;
207 // Apply any user values specified by cl::opt
208 if (UnrollThreshold.getNumOccurrences() > 0)
209 UP.Threshold = UnrollThreshold;
210 if (UnrollPartialThreshold.getNumOccurrences() > 0)
211 UP.PartialThreshold = UnrollPartialThreshold;
212 if (UnrollMaxPercentThresholdBoost.getNumOccurrences() > 0)
213 UP.MaxPercentThresholdBoost = UnrollMaxPercentThresholdBoost;
214 if (UnrollMaxCount.getNumOccurrences() > 0)
215 UP.MaxCount = UnrollMaxCount;
216 if (UnrollFullMaxCount.getNumOccurrences() > 0)
217 UP.FullUnrollMaxCount = UnrollFullMaxCount;
218 if (UnrollPeelCount.getNumOccurrences() > 0)
219 UP.PeelCount = UnrollPeelCount;
220 if (UnrollAllowPartial.getNumOccurrences() > 0)
221 UP.Partial = UnrollAllowPartial;
222 if (UnrollAllowRemainder.getNumOccurrences() > 0)
223 UP.AllowRemainder = UnrollAllowRemainder;
224 if (UnrollRuntime.getNumOccurrences() > 0)
225 UP.Runtime = UnrollRuntime;
226 if (UnrollMaxUpperBound == 0)
227 UP.UpperBound = false;
228 if (UnrollAllowPeeling.getNumOccurrences() > 0)
229 UP.AllowPeeling = UnrollAllowPeeling;
230 if (UnrollUnrollRemainder.getNumOccurrences() > 0)
231 UP.UnrollRemainder = UnrollUnrollRemainder;
233 // Apply user values provided by argument
234 if (UserThreshold.hasValue()) {
235 UP.Threshold = *UserThreshold;
236 UP.PartialThreshold = *UserThreshold;
238 if (UserCount.hasValue())
239 UP.Count = *UserCount;
240 if (UserAllowPartial.hasValue())
241 UP.Partial = *UserAllowPartial;
242 if (UserRuntime.hasValue())
243 UP.Runtime = *UserRuntime;
244 if (UserUpperBound.hasValue())
245 UP.UpperBound = *UserUpperBound;
246 if (UserAllowPeeling.hasValue())
247 UP.AllowPeeling = *UserAllowPeeling;
254 /// A struct to densely store the state of an instruction after unrolling at
257 /// This is designed to work like a tuple of <Instruction *, int> for the
258 /// purposes of hashing and lookup, but to be able to associate two boolean
259 /// states with each key.
260 struct UnrolledInstState {
264 unsigned IsCounted : 1;
267 /// Hashing and equality testing for a set of the instruction states.
268 struct UnrolledInstStateKeyInfo {
269 using PtrInfo = DenseMapInfo<Instruction *>;
270 using PairInfo = DenseMapInfo<std::pair<Instruction *, int>>;
272 static inline UnrolledInstState getEmptyKey() {
273 return {PtrInfo::getEmptyKey(), 0, 0, 0};
276 static inline UnrolledInstState getTombstoneKey() {
277 return {PtrInfo::getTombstoneKey(), 0, 0, 0};
280 static inline unsigned getHashValue(const UnrolledInstState &S) {
281 return PairInfo::getHashValue({S.I, S.Iteration});
284 static inline bool isEqual(const UnrolledInstState &LHS,
285 const UnrolledInstState &RHS) {
286 return PairInfo::isEqual({LHS.I, LHS.Iteration}, {RHS.I, RHS.Iteration});
290 struct EstimatedUnrollCost {
291 /// The estimated cost after unrolling.
292 unsigned UnrolledCost;
294 /// The estimated dynamic cost of executing the instructions in the
296 unsigned RolledDynamicCost;
299 } // end anonymous namespace
301 /// Figure out if the loop is worth full unrolling.
303 /// Complete loop unrolling can make some loads constant, and we need to know
304 /// if that would expose any further optimization opportunities. This routine
305 /// estimates this optimization. It computes cost of unrolled loop
306 /// (UnrolledCost) and dynamic cost of the original loop (RolledDynamicCost). By
307 /// dynamic cost we mean that we won't count costs of blocks that are known not
308 /// to be executed (i.e. if we have a branch in the loop and we know that at the
309 /// given iteration its condition would be resolved to true, we won't add up the
310 /// cost of the 'false'-block).
311 /// \returns Optional value, holding the RolledDynamicCost and UnrolledCost. If
312 /// the analysis failed (no benefits expected from the unrolling, or the loop is
313 /// too big to analyze), the returned value is None.
314 static Optional<EstimatedUnrollCost> analyzeLoopUnrollCost(
315 const Loop *L, unsigned TripCount, DominatorTree &DT, ScalarEvolution &SE,
316 const SmallPtrSetImpl<const Value *> &EphValues,
317 const TargetTransformInfo &TTI, unsigned MaxUnrolledLoopSize) {
318 // We want to be able to scale offsets by the trip count and add more offsets
319 // to them without checking for overflows, and we already don't want to
320 // analyze *massive* trip counts, so we force the max to be reasonably small.
321 assert(UnrollMaxIterationsCountToAnalyze <
322 (unsigned)(std::numeric_limits<int>::max() / 2) &&
323 "The unroll iterations max is too large!");
325 // Only analyze inner loops. We can't properly estimate cost of nested loops
326 // and we won't visit inner loops again anyway.
330 // Don't simulate loops with a big or unknown tripcount
331 if (!UnrollMaxIterationsCountToAnalyze || !TripCount ||
332 TripCount > UnrollMaxIterationsCountToAnalyze)
335 SmallSetVector<BasicBlock *, 16> BBWorklist;
336 SmallSetVector<std::pair<BasicBlock *, BasicBlock *>, 4> ExitWorklist;
337 DenseMap<Value *, Constant *> SimplifiedValues;
338 SmallVector<std::pair<Value *, Constant *>, 4> SimplifiedInputValues;
340 // The estimated cost of the unrolled form of the loop. We try to estimate
341 // this by simplifying as much as we can while computing the estimate.
342 unsigned UnrolledCost = 0;
344 // We also track the estimated dynamic (that is, actually executed) cost in
345 // the rolled form. This helps identify cases when the savings from unrolling
346 // aren't just exposing dead control flows, but actual reduced dynamic
347 // instructions due to the simplifications which we expect to occur after
349 unsigned RolledDynamicCost = 0;
351 // We track the simplification of each instruction in each iteration. We use
352 // this to recursively merge costs into the unrolled cost on-demand so that
353 // we don't count the cost of any dead code. This is essentially a map from
354 // <instruction, int> to <bool, bool>, but stored as a densely packed struct.
355 DenseSet<UnrolledInstState, UnrolledInstStateKeyInfo> InstCostMap;
357 // A small worklist used to accumulate cost of instructions from each
358 // observable and reached root in the loop.
359 SmallVector<Instruction *, 16> CostWorklist;
361 // PHI-used worklist used between iterations while accumulating cost.
362 SmallVector<Instruction *, 4> PHIUsedList;
364 // Helper function to accumulate cost for instructions in the loop.
365 auto AddCostRecursively = [&](Instruction &RootI, int Iteration) {
366 assert(Iteration >= 0 && "Cannot have a negative iteration!");
367 assert(CostWorklist.empty() && "Must start with an empty cost list");
368 assert(PHIUsedList.empty() && "Must start with an empty phi used list");
369 CostWorklist.push_back(&RootI);
370 for (;; --Iteration) {
372 Instruction *I = CostWorklist.pop_back_val();
374 // InstCostMap only uses I and Iteration as a key, the other two values
375 // don't matter here.
376 auto CostIter = InstCostMap.find({I, Iteration, 0, 0});
377 if (CostIter == InstCostMap.end())
378 // If an input to a PHI node comes from a dead path through the loop
379 // we may have no cost data for it here. What that actually means is
382 auto &Cost = *CostIter;
384 // Already counted this instruction.
387 // Mark that we are counting the cost of this instruction now.
388 Cost.IsCounted = true;
390 // If this is a PHI node in the loop header, just add it to the PHI set.
391 if (auto *PhiI = dyn_cast<PHINode>(I))
392 if (PhiI->getParent() == L->getHeader()) {
393 assert(Cost.IsFree && "Loop PHIs shouldn't be evaluated as they "
394 "inherently simplify during unrolling.");
398 // Push the incoming value from the backedge into the PHI used list
399 // if it is an in-loop instruction. We'll use this to populate the
400 // cost worklist for the next iteration (as we count backwards).
401 if (auto *OpI = dyn_cast<Instruction>(
402 PhiI->getIncomingValueForBlock(L->getLoopLatch())))
403 if (L->contains(OpI))
404 PHIUsedList.push_back(OpI);
408 // First accumulate the cost of this instruction.
410 UnrolledCost += TTI.getUserCost(I);
411 LLVM_DEBUG(dbgs() << "Adding cost of instruction (iteration "
412 << Iteration << "): ");
413 LLVM_DEBUG(I->dump());
416 // We must count the cost of every operand which is not free,
417 // recursively. If we reach a loop PHI node, simply add it to the set
418 // to be considered on the next iteration (backwards!).
419 for (Value *Op : I->operands()) {
420 // Check whether this operand is free due to being a constant or
422 auto *OpI = dyn_cast<Instruction>(Op);
423 if (!OpI || !L->contains(OpI))
426 // Otherwise accumulate its cost.
427 CostWorklist.push_back(OpI);
429 } while (!CostWorklist.empty());
431 if (PHIUsedList.empty())
432 // We've exhausted the search.
435 assert(Iteration > 0 &&
436 "Cannot track PHI-used values past the first iteration!");
437 CostWorklist.append(PHIUsedList.begin(), PHIUsedList.end());
442 // Ensure that we don't violate the loop structure invariants relied on by
444 assert(L->isLoopSimplifyForm() && "Must put loop into normal form first.");
445 assert(L->isLCSSAForm(DT) &&
446 "Must have loops in LCSSA form to track live-out values.");
448 LLVM_DEBUG(dbgs() << "Starting LoopUnroll profitability analysis...\n");
450 // Simulate execution of each iteration of the loop counting instructions,
451 // which would be simplified.
452 // Since the same load will take different values on different iterations,
453 // we literally have to go through all loop's iterations.
454 for (unsigned Iteration = 0; Iteration < TripCount; ++Iteration) {
455 LLVM_DEBUG(dbgs() << " Analyzing iteration " << Iteration << "\n");
457 // Prepare for the iteration by collecting any simplified entry or backedge
459 for (Instruction &I : *L->getHeader()) {
460 auto *PHI = dyn_cast<PHINode>(&I);
464 // The loop header PHI nodes must have exactly two input: one from the
465 // loop preheader and one from the loop latch.
467 PHI->getNumIncomingValues() == 2 &&
468 "Must have an incoming value only for the preheader and the latch.");
470 Value *V = PHI->getIncomingValueForBlock(
471 Iteration == 0 ? L->getLoopPreheader() : L->getLoopLatch());
472 Constant *C = dyn_cast<Constant>(V);
473 if (Iteration != 0 && !C)
474 C = SimplifiedValues.lookup(V);
476 SimplifiedInputValues.push_back({PHI, C});
479 // Now clear and re-populate the map for the next iteration.
480 SimplifiedValues.clear();
481 while (!SimplifiedInputValues.empty())
482 SimplifiedValues.insert(SimplifiedInputValues.pop_back_val());
484 UnrolledInstAnalyzer Analyzer(Iteration, SimplifiedValues, SE, L);
487 BBWorklist.insert(L->getHeader());
488 // Note that we *must not* cache the size, this loop grows the worklist.
489 for (unsigned Idx = 0; Idx != BBWorklist.size(); ++Idx) {
490 BasicBlock *BB = BBWorklist[Idx];
492 // Visit all instructions in the given basic block and try to simplify
493 // it. We don't change the actual IR, just count optimization
495 for (Instruction &I : *BB) {
496 // These won't get into the final code - don't even try calculating the
498 if (isa<DbgInfoIntrinsic>(I) || EphValues.count(&I))
501 // Track this instruction's expected baseline cost when executing the
503 RolledDynamicCost += TTI.getUserCost(&I);
505 // Visit the instruction to analyze its loop cost after unrolling,
506 // and if the visitor returns true, mark the instruction as free after
507 // unrolling and continue.
508 bool IsFree = Analyzer.visit(I);
509 bool Inserted = InstCostMap.insert({&I, (int)Iteration,
511 /*IsCounted*/ false}).second;
513 assert(Inserted && "Cannot have a state for an unvisited instruction!");
518 // Can't properly model a cost of a call.
519 // FIXME: With a proper cost model we should be able to do it.
520 if (auto *CI = dyn_cast<CallInst>(&I)) {
521 const Function *Callee = CI->getCalledFunction();
522 if (!Callee || TTI.isLoweredToCall(Callee)) {
523 LLVM_DEBUG(dbgs() << "Can't analyze cost of loop with call\n");
528 // If the instruction might have a side-effect recursively account for
529 // the cost of it and all the instructions leading up to it.
530 if (I.mayHaveSideEffects())
531 AddCostRecursively(I, Iteration);
533 // If unrolled body turns out to be too big, bail out.
534 if (UnrolledCost > MaxUnrolledLoopSize) {
535 LLVM_DEBUG(dbgs() << " Exceeded threshold.. exiting.\n"
536 << " UnrolledCost: " << UnrolledCost
537 << ", MaxUnrolledLoopSize: " << MaxUnrolledLoopSize
543 TerminatorInst *TI = BB->getTerminator();
545 // Add in the live successors by first checking whether we have terminator
546 // that may be simplified based on the values simplified by this call.
547 BasicBlock *KnownSucc = nullptr;
548 if (BranchInst *BI = dyn_cast<BranchInst>(TI)) {
549 if (BI->isConditional()) {
550 if (Constant *SimpleCond =
551 SimplifiedValues.lookup(BI->getCondition())) {
552 // Just take the first successor if condition is undef
553 if (isa<UndefValue>(SimpleCond))
554 KnownSucc = BI->getSuccessor(0);
555 else if (ConstantInt *SimpleCondVal =
556 dyn_cast<ConstantInt>(SimpleCond))
557 KnownSucc = BI->getSuccessor(SimpleCondVal->isZero() ? 1 : 0);
560 } else if (SwitchInst *SI = dyn_cast<SwitchInst>(TI)) {
561 if (Constant *SimpleCond =
562 SimplifiedValues.lookup(SI->getCondition())) {
563 // Just take the first successor if condition is undef
564 if (isa<UndefValue>(SimpleCond))
565 KnownSucc = SI->getSuccessor(0);
566 else if (ConstantInt *SimpleCondVal =
567 dyn_cast<ConstantInt>(SimpleCond))
568 KnownSucc = SI->findCaseValue(SimpleCondVal)->getCaseSuccessor();
572 if (L->contains(KnownSucc))
573 BBWorklist.insert(KnownSucc);
575 ExitWorklist.insert({BB, KnownSucc});
579 // Add BB's successors to the worklist.
580 for (BasicBlock *Succ : successors(BB))
581 if (L->contains(Succ))
582 BBWorklist.insert(Succ);
584 ExitWorklist.insert({BB, Succ});
585 AddCostRecursively(*TI, Iteration);
588 // If we found no optimization opportunities on the first iteration, we
589 // won't find them on later ones too.
590 if (UnrolledCost == RolledDynamicCost) {
591 LLVM_DEBUG(dbgs() << " No opportunities found.. exiting.\n"
592 << " UnrolledCost: " << UnrolledCost << "\n");
597 while (!ExitWorklist.empty()) {
598 BasicBlock *ExitingBB, *ExitBB;
599 std::tie(ExitingBB, ExitBB) = ExitWorklist.pop_back_val();
601 for (Instruction &I : *ExitBB) {
602 auto *PN = dyn_cast<PHINode>(&I);
606 Value *Op = PN->getIncomingValueForBlock(ExitingBB);
607 if (auto *OpI = dyn_cast<Instruction>(Op))
608 if (L->contains(OpI))
609 AddCostRecursively(*OpI, TripCount - 1);
613 LLVM_DEBUG(dbgs() << "Analysis finished:\n"
614 << "UnrolledCost: " << UnrolledCost << ", "
615 << "RolledDynamicCost: " << RolledDynamicCost << "\n");
616 return {{UnrolledCost, RolledDynamicCost}};
619 /// ApproximateLoopSize - Approximate the size of the loop.
620 unsigned llvm::ApproximateLoopSize(
621 const Loop *L, unsigned &NumCalls, bool &NotDuplicatable, bool &Convergent,
622 const TargetTransformInfo &TTI,
623 const SmallPtrSetImpl<const Value *> &EphValues, unsigned BEInsns) {
625 for (BasicBlock *BB : L->blocks())
626 Metrics.analyzeBasicBlock(BB, TTI, EphValues);
627 NumCalls = Metrics.NumInlineCandidates;
628 NotDuplicatable = Metrics.notDuplicatable;
629 Convergent = Metrics.convergent;
631 unsigned LoopSize = Metrics.NumInsts;
633 // Don't allow an estimate of size zero. This would allows unrolling of loops
634 // with huge iteration counts, which is a compile time problem even if it's
635 // not a problem for code quality. Also, the code using this size may assume
636 // that each loop has at least three instructions (likely a conditional
637 // branch, a comparison feeding that branch, and some kind of loop increment
638 // feeding that comparison instruction).
639 LoopSize = std::max(LoopSize, BEInsns + 1);
644 // Returns the loop hint metadata node with the given name (for example,
645 // "llvm.loop.unroll.count"). If no such metadata node exists, then nullptr is
647 static MDNode *GetUnrollMetadataForLoop(const Loop *L, StringRef Name) {
648 if (MDNode *LoopID = L->getLoopID())
649 return GetUnrollMetadata(LoopID, Name);
653 // Returns true if the loop has an unroll(full) pragma.
654 static bool HasUnrollFullPragma(const Loop *L) {
655 return GetUnrollMetadataForLoop(L, "llvm.loop.unroll.full");
658 // Returns true if the loop has an unroll(enable) pragma. This metadata is used
659 // for both "#pragma unroll" and "#pragma clang loop unroll(enable)" directives.
660 static bool HasUnrollEnablePragma(const Loop *L) {
661 return GetUnrollMetadataForLoop(L, "llvm.loop.unroll.enable");
664 // Returns true if the loop has an unroll(disable) pragma.
665 static bool HasUnrollDisablePragma(const Loop *L) {
666 return GetUnrollMetadataForLoop(L, "llvm.loop.unroll.disable");
669 // Returns true if the loop has an runtime unroll(disable) pragma.
670 static bool HasRuntimeUnrollDisablePragma(const Loop *L) {
671 return GetUnrollMetadataForLoop(L, "llvm.loop.unroll.runtime.disable");
674 // If loop has an unroll_count pragma return the (necessarily
675 // positive) value from the pragma. Otherwise return 0.
676 static unsigned UnrollCountPragmaValue(const Loop *L) {
677 MDNode *MD = GetUnrollMetadataForLoop(L, "llvm.loop.unroll.count");
679 assert(MD->getNumOperands() == 2 &&
680 "Unroll count hint metadata should have two operands.");
682 mdconst::extract<ConstantInt>(MD->getOperand(1))->getZExtValue();
683 assert(Count >= 1 && "Unroll count must be positive.");
689 // Computes the boosting factor for complete unrolling.
690 // If fully unrolling the loop would save a lot of RolledDynamicCost, it would
691 // be beneficial to fully unroll the loop even if unrolledcost is large. We
692 // use (RolledDynamicCost / UnrolledCost) to model the unroll benefits to adjust
693 // the unroll threshold.
694 static unsigned getFullUnrollBoostingFactor(const EstimatedUnrollCost &Cost,
695 unsigned MaxPercentThresholdBoost) {
696 if (Cost.RolledDynamicCost >= std::numeric_limits<unsigned>::max() / 100)
698 else if (Cost.UnrolledCost != 0)
699 // The boosting factor is RolledDynamicCost / UnrolledCost
700 return std::min(100 * Cost.RolledDynamicCost / Cost.UnrolledCost,
701 MaxPercentThresholdBoost);
703 return MaxPercentThresholdBoost;
706 // Returns loop size estimation for unrolled loop.
707 static uint64_t getUnrolledLoopSize(
709 TargetTransformInfo::UnrollingPreferences &UP) {
710 assert(LoopSize >= UP.BEInsns && "LoopSize should not be less than BEInsns!");
711 return (uint64_t)(LoopSize - UP.BEInsns) * UP.Count + UP.BEInsns;
714 // Returns true if unroll count was set explicitly.
715 // Calculates unroll count and writes it to UP.Count.
716 bool llvm::computeUnrollCount(
717 Loop *L, const TargetTransformInfo &TTI, DominatorTree &DT, LoopInfo *LI,
718 ScalarEvolution &SE, const SmallPtrSetImpl<const Value *> &EphValues,
719 OptimizationRemarkEmitter *ORE, unsigned &TripCount, unsigned MaxTripCount,
720 unsigned &TripMultiple, unsigned LoopSize,
721 TargetTransformInfo::UnrollingPreferences &UP, bool &UseUpperBound) {
722 // Check for explicit Count.
723 // 1st priority is unroll count set by "unroll-count" option.
724 bool UserUnrollCount = UnrollCount.getNumOccurrences() > 0;
725 if (UserUnrollCount) {
726 UP.Count = UnrollCount;
727 UP.AllowExpensiveTripCount = true;
729 if (UP.AllowRemainder && getUnrolledLoopSize(LoopSize, UP) < UP.Threshold)
733 // 2nd priority is unroll count set by pragma.
734 unsigned PragmaCount = UnrollCountPragmaValue(L);
735 if (PragmaCount > 0) {
736 UP.Count = PragmaCount;
738 UP.AllowExpensiveTripCount = true;
740 if ((UP.AllowRemainder || (TripMultiple % PragmaCount == 0)) &&
741 getUnrolledLoopSize(LoopSize, UP) < PragmaUnrollThreshold)
744 bool PragmaFullUnroll = HasUnrollFullPragma(L);
745 if (PragmaFullUnroll && TripCount != 0) {
746 UP.Count = TripCount;
747 if (getUnrolledLoopSize(LoopSize, UP) < PragmaUnrollThreshold)
751 bool PragmaEnableUnroll = HasUnrollEnablePragma(L);
752 bool ExplicitUnroll = PragmaCount > 0 || PragmaFullUnroll ||
753 PragmaEnableUnroll || UserUnrollCount;
755 if (ExplicitUnroll && TripCount != 0) {
756 // If the loop has an unrolling pragma, we want to be more aggressive with
757 // unrolling limits. Set thresholds to at least the PragmaUnrollThreshold
758 // value which is larger than the default limits.
759 UP.Threshold = std::max<unsigned>(UP.Threshold, PragmaUnrollThreshold);
760 UP.PartialThreshold =
761 std::max<unsigned>(UP.PartialThreshold, PragmaUnrollThreshold);
764 // 3rd priority is full unroll count.
765 // Full unroll makes sense only when TripCount or its upper bound could be
766 // statically calculated.
767 // Also we need to check if we exceed FullUnrollMaxCount.
768 // If using the upper bound to unroll, TripMultiple should be set to 1 because
769 // we do not know when loop may exit.
770 // MaxTripCount and ExactTripCount cannot both be non zero since we only
771 // compute the former when the latter is zero.
772 unsigned ExactTripCount = TripCount;
773 assert((ExactTripCount == 0 || MaxTripCount == 0) &&
774 "ExtractTripCount and MaxTripCount cannot both be non zero.");
775 unsigned FullUnrollTripCount = ExactTripCount ? ExactTripCount : MaxTripCount;
776 UP.Count = FullUnrollTripCount;
777 if (FullUnrollTripCount && FullUnrollTripCount <= UP.FullUnrollMaxCount) {
778 // When computing the unrolled size, note that BEInsns are not replicated
779 // like the rest of the loop body.
780 if (getUnrolledLoopSize(LoopSize, UP) < UP.Threshold) {
781 UseUpperBound = (MaxTripCount == FullUnrollTripCount);
782 TripCount = FullUnrollTripCount;
783 TripMultiple = UP.UpperBound ? 1 : TripMultiple;
784 return ExplicitUnroll;
786 // The loop isn't that small, but we still can fully unroll it if that
787 // helps to remove a significant number of instructions.
788 // To check that, run additional analysis on the loop.
789 if (Optional<EstimatedUnrollCost> Cost = analyzeLoopUnrollCost(
790 L, FullUnrollTripCount, DT, SE, EphValues, TTI,
791 UP.Threshold * UP.MaxPercentThresholdBoost / 100)) {
793 getFullUnrollBoostingFactor(*Cost, UP.MaxPercentThresholdBoost);
794 if (Cost->UnrolledCost < UP.Threshold * Boost / 100) {
795 UseUpperBound = (MaxTripCount == FullUnrollTripCount);
796 TripCount = FullUnrollTripCount;
797 TripMultiple = UP.UpperBound ? 1 : TripMultiple;
798 return ExplicitUnroll;
804 // 4th priority is loop peeling
805 computePeelCount(L, LoopSize, UP, TripCount, SE);
809 return ExplicitUnroll;
812 // 5th priority is partial unrolling.
813 // Try partial unroll only when TripCount could be statically calculated.
815 UP.Partial |= ExplicitUnroll;
817 LLVM_DEBUG(dbgs() << " will not try to unroll partially because "
818 << "-unroll-allow-partial not given\n");
823 UP.Count = TripCount;
824 if (UP.PartialThreshold != NoThreshold) {
825 // Reduce unroll count to be modulo of TripCount for partial unrolling.
826 if (getUnrolledLoopSize(LoopSize, UP) > UP.PartialThreshold)
828 (std::max(UP.PartialThreshold, UP.BEInsns + 1) - UP.BEInsns) /
829 (LoopSize - UP.BEInsns);
830 if (UP.Count > UP.MaxCount)
831 UP.Count = UP.MaxCount;
832 while (UP.Count != 0 && TripCount % UP.Count != 0)
834 if (UP.AllowRemainder && UP.Count <= 1) {
835 // If there is no Count that is modulo of TripCount, set Count to
836 // largest power-of-two factor that satisfies the threshold limit.
837 // As we'll create fixup loop, do the type of unrolling only if
838 // remainder loop is allowed.
839 UP.Count = UP.DefaultUnrollRuntimeCount;
840 while (UP.Count != 0 &&
841 getUnrolledLoopSize(LoopSize, UP) > UP.PartialThreshold)
845 if (PragmaEnableUnroll)
847 return OptimizationRemarkMissed(DEBUG_TYPE,
848 "UnrollAsDirectedTooLarge",
849 L->getStartLoc(), L->getHeader())
850 << "Unable to unroll loop as directed by unroll(enable) "
852 "because unrolled size is too large.";
857 UP.Count = TripCount;
859 if (UP.Count > UP.MaxCount)
860 UP.Count = UP.MaxCount;
861 if ((PragmaFullUnroll || PragmaEnableUnroll) && TripCount &&
862 UP.Count != TripCount)
864 return OptimizationRemarkMissed(DEBUG_TYPE,
865 "FullUnrollAsDirectedTooLarge",
866 L->getStartLoc(), L->getHeader())
867 << "Unable to fully unroll loop as directed by unroll pragma "
869 "unrolled size is too large.";
871 return ExplicitUnroll;
873 assert(TripCount == 0 &&
874 "All cases when TripCount is constant should be covered here.");
875 if (PragmaFullUnroll)
877 return OptimizationRemarkMissed(
878 DEBUG_TYPE, "CantFullUnrollAsDirectedRuntimeTripCount",
879 L->getStartLoc(), L->getHeader())
880 << "Unable to fully unroll loop as directed by unroll(full) "
882 "because loop has a runtime trip count.";
885 // 6th priority is runtime unrolling.
886 // Don't unroll a runtime trip count loop when it is disabled.
887 if (HasRuntimeUnrollDisablePragma(L)) {
892 // Check if the runtime trip count is too small when profile is available.
893 if (L->getHeader()->getParent()->hasProfileData()) {
894 if (auto ProfileTripCount = getLoopEstimatedTripCount(L)) {
895 if (*ProfileTripCount < FlatLoopTripCountThreshold)
898 UP.AllowExpensiveTripCount = true;
902 // Reduce count based on the type of unrolling and the threshold values.
903 UP.Runtime |= PragmaEnableUnroll || PragmaCount > 0 || UserUnrollCount;
906 dbgs() << " will not try to unroll loop with runtime trip count "
907 << "-unroll-runtime not given\n");
912 UP.Count = UP.DefaultUnrollRuntimeCount;
914 // Reduce unroll count to be the largest power-of-two factor of
915 // the original count which satisfies the threshold limit.
916 while (UP.Count != 0 &&
917 getUnrolledLoopSize(LoopSize, UP) > UP.PartialThreshold)
921 unsigned OrigCount = UP.Count;
924 if (!UP.AllowRemainder && UP.Count != 0 && (TripMultiple % UP.Count) != 0) {
925 while (UP.Count != 0 && TripMultiple % UP.Count != 0)
928 dbgs() << "Remainder loop is restricted (that could architecture "
929 "specific or because the loop contains a convergent "
930 "instruction), so unroll count must divide the trip "
932 << TripMultiple << ". Reducing unroll count from " << OrigCount
933 << " to " << UP.Count << ".\n");
937 if (PragmaCount > 0 && !UP.AllowRemainder)
939 return OptimizationRemarkMissed(DEBUG_TYPE,
940 "DifferentUnrollCountFromDirected",
941 L->getStartLoc(), L->getHeader())
942 << "Unable to unroll loop the number of times directed by "
943 "unroll_count pragma because remainder loop is restricted "
944 "(that could architecture specific or because the loop "
945 "contains a convergent instruction) and so must have an "
947 "count that divides the loop trip multiple of "
948 << NV("TripMultiple", TripMultiple) << ". Unrolling instead "
949 << NV("UnrollCount", UP.Count) << " time(s).";
953 if (UP.Count > UP.MaxCount)
954 UP.Count = UP.MaxCount;
955 LLVM_DEBUG(dbgs() << " partially unrolling with count: " << UP.Count
959 return ExplicitUnroll;
962 static LoopUnrollResult tryToUnrollLoop(
963 Loop *L, DominatorTree &DT, LoopInfo *LI, ScalarEvolution &SE,
964 const TargetTransformInfo &TTI, AssumptionCache &AC,
965 OptimizationRemarkEmitter &ORE, bool PreserveLCSSA, int OptLevel,
966 Optional<unsigned> ProvidedCount, Optional<unsigned> ProvidedThreshold,
967 Optional<bool> ProvidedAllowPartial, Optional<bool> ProvidedRuntime,
968 Optional<bool> ProvidedUpperBound, Optional<bool> ProvidedAllowPeeling) {
969 LLVM_DEBUG(dbgs() << "Loop Unroll: F["
970 << L->getHeader()->getParent()->getName() << "] Loop %"
971 << L->getHeader()->getName() << "\n");
972 if (HasUnrollDisablePragma(L))
973 return LoopUnrollResult::Unmodified;
974 if (!L->isLoopSimplifyForm()) {
976 dbgs() << " Not unrolling loop which is not in loop-simplify form.\n");
977 return LoopUnrollResult::Unmodified;
980 unsigned NumInlineCandidates;
981 bool NotDuplicatable;
983 TargetTransformInfo::UnrollingPreferences UP = gatherUnrollingPreferences(
984 L, SE, TTI, OptLevel, ProvidedThreshold, ProvidedCount,
985 ProvidedAllowPartial, ProvidedRuntime, ProvidedUpperBound,
986 ProvidedAllowPeeling);
987 // Exit early if unrolling is disabled.
988 if (UP.Threshold == 0 && (!UP.Partial || UP.PartialThreshold == 0))
989 return LoopUnrollResult::Unmodified;
991 SmallPtrSet<const Value *, 32> EphValues;
992 CodeMetrics::collectEphemeralValues(L, &AC, EphValues);
995 ApproximateLoopSize(L, NumInlineCandidates, NotDuplicatable, Convergent,
996 TTI, EphValues, UP.BEInsns);
997 LLVM_DEBUG(dbgs() << " Loop Size = " << LoopSize << "\n");
998 if (NotDuplicatable) {
999 LLVM_DEBUG(dbgs() << " Not unrolling loop which contains non-duplicatable"
1000 << " instructions.\n");
1001 return LoopUnrollResult::Unmodified;
1003 if (NumInlineCandidates != 0) {
1004 LLVM_DEBUG(dbgs() << " Not unrolling loop with inlinable calls.\n");
1005 return LoopUnrollResult::Unmodified;
1008 // Find trip count and trip multiple if count is not available
1009 unsigned TripCount = 0;
1010 unsigned MaxTripCount = 0;
1011 unsigned TripMultiple = 1;
1012 // If there are multiple exiting blocks but one of them is the latch, use the
1013 // latch for the trip count estimation. Otherwise insist on a single exiting
1014 // block for the trip count estimation.
1015 BasicBlock *ExitingBlock = L->getLoopLatch();
1016 if (!ExitingBlock || !L->isLoopExiting(ExitingBlock))
1017 ExitingBlock = L->getExitingBlock();
1019 TripCount = SE.getSmallConstantTripCount(L, ExitingBlock);
1020 TripMultiple = SE.getSmallConstantTripMultiple(L, ExitingBlock);
1023 // If the loop contains a convergent operation, the prelude we'd add
1024 // to do the first few instructions before we hit the unrolled loop
1025 // is unsafe -- it adds a control-flow dependency to the convergent
1026 // operation. Therefore restrict remainder loop (try unrollig without).
1028 // TODO: This is quite conservative. In practice, convergent_op()
1029 // is likely to be called unconditionally in the loop. In this
1030 // case, the program would be ill-formed (on most architectures)
1031 // unless n were the same on all threads in a thread group.
1032 // Assuming n is the same on all threads, any kind of unrolling is
1033 // safe. But currently llvm's notion of convergence isn't powerful
1034 // enough to express this.
1036 UP.AllowRemainder = false;
1038 // Try to find the trip count upper bound if we cannot find the exact trip
1040 bool MaxOrZero = false;
1042 MaxTripCount = SE.getSmallConstantMaxTripCount(L);
1043 MaxOrZero = SE.isBackedgeTakenCountMaxOrZero(L);
1044 // We can unroll by the upper bound amount if it's generally allowed or if
1045 // we know that the loop is executed either the upper bound or zero times.
1046 // (MaxOrZero unrolling keeps only the first loop test, so the number of
1047 // loop tests remains the same compared to the non-unrolled version, whereas
1048 // the generic upper bound unrolling keeps all but the last loop test so the
1049 // number of loop tests goes up which may end up being worse on targets with
1050 // constrained branch predictor resources so is controlled by an option.)
1051 // In addition we only unroll small upper bounds.
1052 if (!(UP.UpperBound || MaxOrZero) || MaxTripCount > UnrollMaxUpperBound) {
1057 // computeUnrollCount() decides whether it is beneficial to use upper bound to
1058 // fully unroll the loop.
1059 bool UseUpperBound = false;
1060 bool IsCountSetExplicitly = computeUnrollCount(
1061 L, TTI, DT, LI, SE, EphValues, &ORE, TripCount, MaxTripCount,
1062 TripMultiple, LoopSize, UP, UseUpperBound);
1064 return LoopUnrollResult::Unmodified;
1065 // Unroll factor (Count) must be less or equal to TripCount.
1066 if (TripCount && UP.Count > TripCount)
1067 UP.Count = TripCount;
1070 LoopUnrollResult UnrollResult = UnrollLoop(
1071 L, UP.Count, TripCount, UP.Force, UP.Runtime, UP.AllowExpensiveTripCount,
1072 UseUpperBound, MaxOrZero, TripMultiple, UP.PeelCount, UP.UnrollRemainder,
1073 LI, &SE, &DT, &AC, &ORE, PreserveLCSSA);
1074 if (UnrollResult == LoopUnrollResult::Unmodified)
1075 return LoopUnrollResult::Unmodified;
1077 // If loop has an unroll count pragma or unrolled by explicitly set count
1078 // mark loop as unrolled to prevent unrolling beyond that requested.
1079 // If the loop was peeled, we already "used up" the profile information
1080 // we had, so we don't want to unroll or peel again.
1081 if (UnrollResult != LoopUnrollResult::FullyUnrolled &&
1082 (IsCountSetExplicitly || UP.PeelCount))
1083 L->setLoopAlreadyUnrolled();
1085 return UnrollResult;
1090 class LoopUnroll : public LoopPass {
1092 static char ID; // Pass ID, replacement for typeid
1095 Optional<unsigned> ProvidedCount;
1096 Optional<unsigned> ProvidedThreshold;
1097 Optional<bool> ProvidedAllowPartial;
1098 Optional<bool> ProvidedRuntime;
1099 Optional<bool> ProvidedUpperBound;
1100 Optional<bool> ProvidedAllowPeeling;
1102 LoopUnroll(int OptLevel = 2, Optional<unsigned> Threshold = None,
1103 Optional<unsigned> Count = None,
1104 Optional<bool> AllowPartial = None, Optional<bool> Runtime = None,
1105 Optional<bool> UpperBound = None,
1106 Optional<bool> AllowPeeling = None)
1107 : LoopPass(ID), OptLevel(OptLevel), ProvidedCount(std::move(Count)),
1108 ProvidedThreshold(Threshold), ProvidedAllowPartial(AllowPartial),
1109 ProvidedRuntime(Runtime), ProvidedUpperBound(UpperBound),
1110 ProvidedAllowPeeling(AllowPeeling) {
1111 initializeLoopUnrollPass(*PassRegistry::getPassRegistry());
1114 bool runOnLoop(Loop *L, LPPassManager &LPM) override {
1118 Function &F = *L->getHeader()->getParent();
1120 auto &DT = getAnalysis<DominatorTreeWrapperPass>().getDomTree();
1121 LoopInfo *LI = &getAnalysis<LoopInfoWrapperPass>().getLoopInfo();
1122 ScalarEvolution &SE = getAnalysis<ScalarEvolutionWrapperPass>().getSE();
1123 const TargetTransformInfo &TTI =
1124 getAnalysis<TargetTransformInfoWrapperPass>().getTTI(F);
1125 auto &AC = getAnalysis<AssumptionCacheTracker>().getAssumptionCache(F);
1126 // For the old PM, we can't use OptimizationRemarkEmitter as an analysis
1127 // pass. Function analyses need to be preserved across loop transformations
1128 // but ORE cannot be preserved (see comment before the pass definition).
1129 OptimizationRemarkEmitter ORE(&F);
1130 bool PreserveLCSSA = mustPreserveAnalysisID(LCSSAID);
1132 LoopUnrollResult Result = tryToUnrollLoop(
1133 L, DT, LI, SE, TTI, AC, ORE, PreserveLCSSA, OptLevel, ProvidedCount,
1134 ProvidedThreshold, ProvidedAllowPartial, ProvidedRuntime,
1135 ProvidedUpperBound, ProvidedAllowPeeling);
1137 if (Result == LoopUnrollResult::FullyUnrolled)
1138 LPM.markLoopAsDeleted(*L);
1140 return Result != LoopUnrollResult::Unmodified;
1143 /// This transformation requires natural loop information & requires that
1144 /// loop preheaders be inserted into the CFG...
1145 void getAnalysisUsage(AnalysisUsage &AU) const override {
1146 AU.addRequired<AssumptionCacheTracker>();
1147 AU.addRequired<TargetTransformInfoWrapperPass>();
1148 // FIXME: Loop passes are required to preserve domtree, and for now we just
1149 // recreate dom info if anything gets unrolled.
1150 getLoopAnalysisUsage(AU);
1154 } // end anonymous namespace
1156 char LoopUnroll::ID = 0;
1158 INITIALIZE_PASS_BEGIN(LoopUnroll, "loop-unroll", "Unroll loops", false, false)
1159 INITIALIZE_PASS_DEPENDENCY(AssumptionCacheTracker)
1160 INITIALIZE_PASS_DEPENDENCY(LoopPass)
1161 INITIALIZE_PASS_DEPENDENCY(TargetTransformInfoWrapperPass)
1162 INITIALIZE_PASS_END(LoopUnroll, "loop-unroll", "Unroll loops", false, false)
1164 Pass *llvm::createLoopUnrollPass(int OptLevel, int Threshold, int Count,
1165 int AllowPartial, int Runtime, int UpperBound,
1167 // TODO: It would make more sense for this function to take the optionals
1168 // directly, but that's dangerous since it would silently break out of tree
1170 return new LoopUnroll(
1171 OptLevel, Threshold == -1 ? None : Optional<unsigned>(Threshold),
1172 Count == -1 ? None : Optional<unsigned>(Count),
1173 AllowPartial == -1 ? None : Optional<bool>(AllowPartial),
1174 Runtime == -1 ? None : Optional<bool>(Runtime),
1175 UpperBound == -1 ? None : Optional<bool>(UpperBound),
1176 AllowPeeling == -1 ? None : Optional<bool>(AllowPeeling));
1179 Pass *llvm::createSimpleLoopUnrollPass(int OptLevel) {
1180 return createLoopUnrollPass(OptLevel, -1, -1, 0, 0, 0, 0);
1183 PreservedAnalyses LoopFullUnrollPass::run(Loop &L, LoopAnalysisManager &AM,
1184 LoopStandardAnalysisResults &AR,
1185 LPMUpdater &Updater) {
1187 AM.getResult<FunctionAnalysisManagerLoopProxy>(L, AR).getManager();
1188 Function *F = L.getHeader()->getParent();
1190 auto *ORE = FAM.getCachedResult<OptimizationRemarkEmitterAnalysis>(*F);
1191 // FIXME: This should probably be optional rather than required.
1194 "LoopFullUnrollPass: OptimizationRemarkEmitterAnalysis not "
1195 "cached at a higher level");
1197 // Keep track of the previous loop structure so we can identify new loops
1198 // created by unrolling.
1199 Loop *ParentL = L.getParentLoop();
1200 SmallPtrSet<Loop *, 4> OldLoops;
1202 OldLoops.insert(ParentL->begin(), ParentL->end());
1204 OldLoops.insert(AR.LI.begin(), AR.LI.end());
1206 std::string LoopName = L.getName();
1209 tryToUnrollLoop(&L, AR.DT, &AR.LI, AR.SE, AR.TTI, AR.AC, *ORE,
1210 /*PreserveLCSSA*/ true, OptLevel, /*Count*/ None,
1211 /*Threshold*/ None, /*AllowPartial*/ false,
1212 /*Runtime*/ false, /*UpperBound*/ false,
1213 /*AllowPeeling*/ false) != LoopUnrollResult::Unmodified;
1215 return PreservedAnalyses::all();
1217 // The parent must not be damaged by unrolling!
1220 ParentL->verifyLoop();
1223 // Unrolling can do several things to introduce new loops into a loop nest:
1224 // - Full unrolling clones child loops within the current loop but then
1225 // removes the current loop making all of the children appear to be new
1228 // When a new loop appears as a sibling loop after fully unrolling,
1229 // its nesting structure has fundamentally changed and we want to revisit
1230 // it to reflect that.
1232 // When unrolling has removed the current loop, we need to tell the
1233 // infrastructure that it is gone.
1235 // Finally, we support a debugging/testing mode where we revisit child loops
1236 // as well. These are not expected to require further optimizations as either
1237 // they or the loop they were cloned from have been directly visited already.
1238 // But the debugging mode allows us to check this assumption.
1239 bool IsCurrentLoopValid = false;
1240 SmallVector<Loop *, 4> SibLoops;
1242 SibLoops.append(ParentL->begin(), ParentL->end());
1244 SibLoops.append(AR.LI.begin(), AR.LI.end());
1245 erase_if(SibLoops, [&](Loop *SibLoop) {
1246 if (SibLoop == &L) {
1247 IsCurrentLoopValid = true;
1251 // Otherwise erase the loop from the list if it was in the old loops.
1252 return OldLoops.count(SibLoop) != 0;
1254 Updater.addSiblingLoops(SibLoops);
1256 if (!IsCurrentLoopValid) {
1257 Updater.markLoopAsDeleted(L, LoopName);
1259 // We can only walk child loops if the current loop remained valid.
1260 if (UnrollRevisitChildLoops) {
1261 // Walk *all* of the child loops.
1262 SmallVector<Loop *, 4> ChildLoops(L.begin(), L.end());
1263 Updater.addChildLoops(ChildLoops);
1267 return getLoopPassPreservedAnalyses();
1270 template <typename RangeT>
1271 static SmallVector<Loop *, 8> appendLoopsToWorklist(RangeT &&Loops) {
1272 SmallVector<Loop *, 8> Worklist;
1273 // We use an internal worklist to build up the preorder traversal without
1275 SmallVector<Loop *, 4> PreOrderLoops, PreOrderWorklist;
1277 for (Loop *RootL : Loops) {
1278 assert(PreOrderLoops.empty() && "Must start with an empty preorder walk.");
1279 assert(PreOrderWorklist.empty() &&
1280 "Must start with an empty preorder walk worklist.");
1281 PreOrderWorklist.push_back(RootL);
1283 Loop *L = PreOrderWorklist.pop_back_val();
1284 PreOrderWorklist.append(L->begin(), L->end());
1285 PreOrderLoops.push_back(L);
1286 } while (!PreOrderWorklist.empty());
1288 Worklist.append(PreOrderLoops.begin(), PreOrderLoops.end());
1289 PreOrderLoops.clear();
1294 PreservedAnalyses LoopUnrollPass::run(Function &F,
1295 FunctionAnalysisManager &AM) {
1296 auto &SE = AM.getResult<ScalarEvolutionAnalysis>(F);
1297 auto &LI = AM.getResult<LoopAnalysis>(F);
1298 auto &TTI = AM.getResult<TargetIRAnalysis>(F);
1299 auto &DT = AM.getResult<DominatorTreeAnalysis>(F);
1300 auto &AC = AM.getResult<AssumptionAnalysis>(F);
1301 auto &ORE = AM.getResult<OptimizationRemarkEmitterAnalysis>(F);
1303 LoopAnalysisManager *LAM = nullptr;
1304 if (auto *LAMProxy = AM.getCachedResult<LoopAnalysisManagerFunctionProxy>(F))
1305 LAM = &LAMProxy->getManager();
1307 const ModuleAnalysisManager &MAM =
1308 AM.getResult<ModuleAnalysisManagerFunctionProxy>(F).getManager();
1309 ProfileSummaryInfo *PSI =
1310 MAM.getCachedResult<ProfileSummaryAnalysis>(*F.getParent());
1312 bool Changed = false;
1314 // The unroller requires loops to be in simplified form, and also needs LCSSA.
1315 // Since simplification may add new inner loops, it has to run before the
1316 // legality and profitability checks. This means running the loop unroller
1317 // will simplify all loops, regardless of whether anything end up being
1319 for (auto &L : LI) {
1320 Changed |= simplifyLoop(L, &DT, &LI, &SE, &AC, false /* PreserveLCSSA */);
1321 Changed |= formLCSSARecursively(*L, DT, &LI, &SE);
1324 SmallVector<Loop *, 8> Worklist = appendLoopsToWorklist(LI);
1326 while (!Worklist.empty()) {
1327 // Because the LoopInfo stores the loops in RPO, we walk the worklist
1328 // from back to front so that we work forward across the CFG, which
1329 // for unrolling is only needed to get optimization remarks emitted in
1331 Loop &L = *Worklist.pop_back_val();
1333 Loop *ParentL = L.getParentLoop();
1336 // The API here is quite complex to call, but there are only two interesting
1337 // states we support: partial and full (or "simple") unrolling. However, to
1338 // enable these things we actually pass "None" in for the optional to avoid
1339 // providing an explicit choice.
1340 Optional<bool> AllowPartialParam, RuntimeParam, UpperBoundParam,
1342 // Check if the profile summary indicates that the profiled application
1343 // has a huge working set size, in which case we disable peeling to avoid
1344 // bloating it further.
1345 if (PSI && PSI->hasHugeWorkingSetSize())
1346 AllowPeeling = false;
1347 std::string LoopName = L.getName();
1348 LoopUnrollResult Result =
1349 tryToUnrollLoop(&L, DT, &LI, SE, TTI, AC, ORE,
1350 /*PreserveLCSSA*/ true, OptLevel, /*Count*/ None,
1351 /*Threshold*/ None, AllowPartialParam, RuntimeParam,
1352 UpperBoundParam, AllowPeeling);
1353 Changed |= Result != LoopUnrollResult::Unmodified;
1355 // The parent must not be damaged by unrolling!
1357 if (Result != LoopUnrollResult::Unmodified && ParentL)
1358 ParentL->verifyLoop();
1361 // Clear any cached analysis results for L if we removed it completely.
1362 if (LAM && Result == LoopUnrollResult::FullyUnrolled)
1363 LAM->clear(L, LoopName);
1367 return PreservedAnalyses::all();
1369 return getLoopPassPreservedAnalyses();