//===- LoopUnrollAndJam.cpp - Loop unroll and jam pass --------------------===// // // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. // See https://llvm.org/LICENSE.txt for license information. // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // // This pass implements an unroll and jam pass. Most of the work is done by // Utils/UnrollLoopAndJam.cpp. //===----------------------------------------------------------------------===// #include "llvm/Transforms/Scalar/LoopUnrollAndJamPass.h" #include "llvm/ADT/ArrayRef.h" #include "llvm/ADT/PriorityWorklist.h" #include "llvm/ADT/SmallPtrSet.h" #include "llvm/ADT/StringRef.h" #include "llvm/Analysis/AssumptionCache.h" #include "llvm/Analysis/CodeMetrics.h" #include "llvm/Analysis/DependenceAnalysis.h" #include "llvm/Analysis/LoopAnalysisManager.h" #include "llvm/Analysis/LoopInfo.h" #include "llvm/Analysis/LoopNestAnalysis.h" #include "llvm/Analysis/LoopPass.h" #include "llvm/Analysis/OptimizationRemarkEmitter.h" #include "llvm/Analysis/ScalarEvolution.h" #include "llvm/Analysis/TargetTransformInfo.h" #include "llvm/IR/BasicBlock.h" #include "llvm/IR/Constants.h" #include "llvm/IR/Dominators.h" #include "llvm/IR/Function.h" #include "llvm/IR/Instructions.h" #include "llvm/IR/Metadata.h" #include "llvm/IR/PassManager.h" #include "llvm/Support/Casting.h" #include "llvm/Support/CommandLine.h" #include "llvm/Support/Compiler.h" #include "llvm/Support/Debug.h" #include "llvm/Support/raw_ostream.h" #include "llvm/Transforms/Scalar/LoopPassManager.h" #include "llvm/Transforms/Utils/LoopPeel.h" #include "llvm/Transforms/Utils/LoopUtils.h" #include "llvm/Transforms/Utils/UnrollLoop.h" #include #include namespace llvm { class Instruction; class Value; } // namespace llvm using namespace llvm; #define DEBUG_TYPE "loop-unroll-and-jam" /// @{ /// Metadata attribute names static const char *const LLVMLoopUnrollAndJamFollowupAll = "llvm.loop.unroll_and_jam.followup_all"; static const char *const LLVMLoopUnrollAndJamFollowupInner = "llvm.loop.unroll_and_jam.followup_inner"; static const char *const LLVMLoopUnrollAndJamFollowupOuter = "llvm.loop.unroll_and_jam.followup_outer"; static const char *const LLVMLoopUnrollAndJamFollowupRemainderInner = "llvm.loop.unroll_and_jam.followup_remainder_inner"; static const char *const LLVMLoopUnrollAndJamFollowupRemainderOuter = "llvm.loop.unroll_and_jam.followup_remainder_outer"; /// @} static cl::opt AllowUnrollAndJam("allow-unroll-and-jam", cl::Hidden, cl::desc("Allows loops to be unroll-and-jammed.")); static cl::opt UnrollAndJamCount( "unroll-and-jam-count", cl::Hidden, cl::desc("Use this unroll count for all loops including those with " "unroll_and_jam_count pragma values, for testing purposes")); static cl::opt UnrollAndJamThreshold( "unroll-and-jam-threshold", cl::init(60), cl::Hidden, cl::desc("Threshold to use for inner loop when doing unroll and jam.")); static cl::opt PragmaUnrollAndJamThreshold( "pragma-unroll-and-jam-threshold", cl::init(1024), cl::Hidden, cl::desc("Unrolled size limit for loops with an unroll_and_jam(full) or " "unroll_count pragma.")); // Returns the loop hint metadata node with the given name (for example, // "llvm.loop.unroll.count"). If no such metadata node exists, then nullptr is // returned. static MDNode *getUnrollMetadataForLoop(const Loop *L, StringRef Name) { if (MDNode *LoopID = L->getLoopID()) return GetUnrollMetadata(LoopID, Name); return nullptr; } // Returns true if the loop has any metadata starting with Prefix. For example a // Prefix of "llvm.loop.unroll." returns true if we have any unroll metadata. static bool hasAnyUnrollPragma(const Loop *L, StringRef Prefix) { if (MDNode *LoopID = L->getLoopID()) { // First operand should refer to the loop id itself. assert(LoopID->getNumOperands() > 0 && "requires at least one operand"); assert(LoopID->getOperand(0) == LoopID && "invalid loop id"); for (unsigned I = 1, E = LoopID->getNumOperands(); I < E; ++I) { MDNode *MD = dyn_cast(LoopID->getOperand(I)); if (!MD) continue; MDString *S = dyn_cast(MD->getOperand(0)); if (!S) continue; if (S->getString().starts_with(Prefix)) return true; } } return false; } // Returns true if the loop has an unroll_and_jam(enable) pragma. static bool hasUnrollAndJamEnablePragma(const Loop *L) { return getUnrollMetadataForLoop(L, "llvm.loop.unroll_and_jam.enable"); } // If loop has an unroll_and_jam_count pragma return the (necessarily // positive) value from the pragma. Otherwise return 0. static unsigned unrollAndJamCountPragmaValue(const Loop *L) { MDNode *MD = getUnrollMetadataForLoop(L, "llvm.loop.unroll_and_jam.count"); if (MD) { assert(MD->getNumOperands() == 2 && "Unroll count hint metadata should have two operands."); unsigned Count = mdconst::extract(MD->getOperand(1))->getZExtValue(); assert(Count >= 1 && "Unroll count must be positive."); return Count; } return 0; } // Returns loop size estimation for unrolled loop. static uint64_t getUnrollAndJammedLoopSize(unsigned LoopSize, TargetTransformInfo::UnrollingPreferences &UP) { assert(LoopSize >= UP.BEInsns && "LoopSize should not be less than BEInsns!"); return static_cast(LoopSize - UP.BEInsns) * UP.Count + UP.BEInsns; } // Calculates unroll and jam count and writes it to UP.Count. Returns true if // unroll count was set explicitly. static bool computeUnrollAndJamCount( Loop *L, Loop *SubLoop, const TargetTransformInfo &TTI, DominatorTree &DT, LoopInfo *LI, AssumptionCache *AC, ScalarEvolution &SE, const SmallPtrSetImpl &EphValues, OptimizationRemarkEmitter *ORE, unsigned OuterTripCount, unsigned OuterTripMultiple, const UnrollCostEstimator &OuterUCE, unsigned InnerTripCount, unsigned InnerLoopSize, TargetTransformInfo::UnrollingPreferences &UP, TargetTransformInfo::PeelingPreferences &PP) { unsigned OuterLoopSize = OuterUCE.getRolledLoopSize(); // First up use computeUnrollCount from the loop unroller to get a count // for unrolling the outer loop, plus any loops requiring explicit // unrolling we leave to the unroller. This uses UP.Threshold / // UP.PartialThreshold / UP.MaxCount to come up with sensible loop values. // We have already checked that the loop has no unroll.* pragmas. unsigned MaxTripCount = 0; bool UseUpperBound = false; bool ExplicitUnroll = computeUnrollCount( L, TTI, DT, LI, AC, SE, EphValues, ORE, OuterTripCount, MaxTripCount, /*MaxOrZero*/ false, OuterTripMultiple, OuterUCE, UP, PP, UseUpperBound); if (ExplicitUnroll || UseUpperBound) { // If the user explicitly set the loop as unrolled, dont UnJ it. Leave it // for the unroller instead. LLVM_DEBUG(dbgs() << "Won't unroll-and-jam; explicit count set by " "computeUnrollCount\n"); UP.Count = 0; return false; } // Override with any explicit Count from the "unroll-and-jam-count" option. bool UserUnrollCount = UnrollAndJamCount.getNumOccurrences() > 0; if (UserUnrollCount) { UP.Count = UnrollAndJamCount; UP.Force = true; if (UP.AllowRemainder && getUnrollAndJammedLoopSize(OuterLoopSize, UP) < UP.Threshold && getUnrollAndJammedLoopSize(InnerLoopSize, UP) < UP.UnrollAndJamInnerLoopThreshold) return true; } // Check for unroll_and_jam pragmas unsigned PragmaCount = unrollAndJamCountPragmaValue(L); if (PragmaCount > 0) { UP.Count = PragmaCount; UP.Runtime = true; UP.Force = true; if ((UP.AllowRemainder || (OuterTripMultiple % PragmaCount == 0)) && getUnrollAndJammedLoopSize(OuterLoopSize, UP) < UP.Threshold && getUnrollAndJammedLoopSize(InnerLoopSize, UP) < UP.UnrollAndJamInnerLoopThreshold) return true; } bool PragmaEnableUnroll = hasUnrollAndJamEnablePragma(L); bool ExplicitUnrollAndJamCount = PragmaCount > 0 || UserUnrollCount; bool ExplicitUnrollAndJam = PragmaEnableUnroll || ExplicitUnrollAndJamCount; // If the loop has an unrolling pragma, we want to be more aggressive with // unrolling limits. if (ExplicitUnrollAndJam) UP.UnrollAndJamInnerLoopThreshold = PragmaUnrollAndJamThreshold; if (!UP.AllowRemainder && getUnrollAndJammedLoopSize(InnerLoopSize, UP) >= UP.UnrollAndJamInnerLoopThreshold) { LLVM_DEBUG(dbgs() << "Won't unroll-and-jam; can't create remainder and " "inner loop too large\n"); UP.Count = 0; return false; } // We have a sensible limit for the outer loop, now adjust it for the inner // loop and UP.UnrollAndJamInnerLoopThreshold. If the outer limit was set // explicitly, we want to stick to it. if (!ExplicitUnrollAndJamCount && UP.AllowRemainder) { while (UP.Count != 0 && getUnrollAndJammedLoopSize(InnerLoopSize, UP) >= UP.UnrollAndJamInnerLoopThreshold) UP.Count--; } // If we are explicitly unroll and jamming, we are done. Otherwise there are a // number of extra performance heuristics to check. if (ExplicitUnrollAndJam) return true; // If the inner loop count is known and small, leave the entire loop nest to // be the unroller if (InnerTripCount && InnerLoopSize * InnerTripCount < UP.Threshold) { LLVM_DEBUG(dbgs() << "Won't unroll-and-jam; small inner loop count is " "being left for the unroller\n"); UP.Count = 0; return false; } // Check for situations where UnJ is likely to be unprofitable. Including // subloops with more than 1 block. if (SubLoop->getBlocks().size() != 1) { LLVM_DEBUG( dbgs() << "Won't unroll-and-jam; More than one inner loop block\n"); UP.Count = 0; return false; } // Limit to loops where there is something to gain from unrolling and // jamming the loop. In this case, look for loads that are invariant in the // outer loop and can become shared. unsigned NumInvariant = 0; for (BasicBlock *BB : SubLoop->getBlocks()) { for (Instruction &I : *BB) { if (auto *Ld = dyn_cast(&I)) { Value *V = Ld->getPointerOperand(); const SCEV *LSCEV = SE.getSCEVAtScope(V, L); if (SE.isLoopInvariant(LSCEV, L)) NumInvariant++; } } } if (NumInvariant == 0) { LLVM_DEBUG(dbgs() << "Won't unroll-and-jam; No loop invariant loads\n"); UP.Count = 0; return false; } return false; } static LoopUnrollResult tryToUnrollAndJamLoop(Loop *L, DominatorTree &DT, LoopInfo *LI, ScalarEvolution &SE, const TargetTransformInfo &TTI, AssumptionCache &AC, DependenceInfo &DI, OptimizationRemarkEmitter &ORE, int OptLevel) { TargetTransformInfo::UnrollingPreferences UP = gatherUnrollingPreferences( L, SE, TTI, nullptr, nullptr, ORE, OptLevel, std::nullopt, std::nullopt, std::nullopt, std::nullopt, std::nullopt, std::nullopt); TargetTransformInfo::PeelingPreferences PP = gatherPeelingPreferences(L, SE, TTI, std::nullopt, std::nullopt); TransformationMode EnableMode = hasUnrollAndJamTransformation(L); if (EnableMode & TM_Disable) return LoopUnrollResult::Unmodified; if (EnableMode & TM_ForcedByUser) UP.UnrollAndJam = true; if (AllowUnrollAndJam.getNumOccurrences() > 0) UP.UnrollAndJam = AllowUnrollAndJam; if (UnrollAndJamThreshold.getNumOccurrences() > 0) UP.UnrollAndJamInnerLoopThreshold = UnrollAndJamThreshold; // Exit early if unrolling is disabled. if (!UP.UnrollAndJam || UP.UnrollAndJamInnerLoopThreshold == 0) return LoopUnrollResult::Unmodified; LLVM_DEBUG(dbgs() << "Loop Unroll and Jam: F[" << L->getHeader()->getParent()->getName() << "] Loop %" << L->getHeader()->getName() << "\n"); // A loop with any unroll pragma (enabling/disabling/count/etc) is left for // the unroller, so long as it does not explicitly have unroll_and_jam // metadata. This means #pragma nounroll will disable unroll and jam as well // as unrolling if (hasAnyUnrollPragma(L, "llvm.loop.unroll.") && !hasAnyUnrollPragma(L, "llvm.loop.unroll_and_jam.")) { LLVM_DEBUG(dbgs() << " Disabled due to pragma.\n"); return LoopUnrollResult::Unmodified; } if (!isSafeToUnrollAndJam(L, SE, DT, DI, *LI)) { LLVM_DEBUG(dbgs() << " Disabled due to not being safe.\n"); return LoopUnrollResult::Unmodified; } // Approximate the loop size and collect useful info SmallPtrSet EphValues; CodeMetrics::collectEphemeralValues(L, &AC, EphValues); Loop *SubLoop = L->getSubLoops()[0]; UnrollCostEstimator InnerUCE(SubLoop, TTI, EphValues, UP.BEInsns); UnrollCostEstimator OuterUCE(L, TTI, EphValues, UP.BEInsns); if (!InnerUCE.canUnroll() || !OuterUCE.canUnroll()) { LLVM_DEBUG(dbgs() << " Loop not considered unrollable\n"); return LoopUnrollResult::Unmodified; } unsigned InnerLoopSize = InnerUCE.getRolledLoopSize(); LLVM_DEBUG(dbgs() << " Outer Loop Size: " << OuterUCE.getRolledLoopSize() << "\n"); LLVM_DEBUG(dbgs() << " Inner Loop Size: " << InnerLoopSize << "\n"); if (InnerUCE.NumInlineCandidates != 0 || OuterUCE.NumInlineCandidates != 0) { LLVM_DEBUG(dbgs() << " Not unrolling loop with inlinable calls.\n"); return LoopUnrollResult::Unmodified; } // FIXME: The call to canUnroll() allows some controlled convergent // operations, but we block them here for future changes. if (InnerUCE.Convergence != ConvergenceKind::None || OuterUCE.Convergence != ConvergenceKind::None) { LLVM_DEBUG( dbgs() << " Not unrolling loop with convergent instructions.\n"); return LoopUnrollResult::Unmodified; } // Save original loop IDs for after the transformation. MDNode *OrigOuterLoopID = L->getLoopID(); MDNode *OrigSubLoopID = SubLoop->getLoopID(); // To assign the loop id of the epilogue, assign it before unrolling it so it // is applied to every inner loop of the epilogue. We later apply the loop ID // for the jammed inner loop. std::optional NewInnerEpilogueLoopID = makeFollowupLoopID( OrigOuterLoopID, {LLVMLoopUnrollAndJamFollowupAll, LLVMLoopUnrollAndJamFollowupRemainderInner}); if (NewInnerEpilogueLoopID) SubLoop->setLoopID(*NewInnerEpilogueLoopID); // Find trip count and trip multiple BasicBlock *Latch = L->getLoopLatch(); BasicBlock *SubLoopLatch = SubLoop->getLoopLatch(); unsigned OuterTripCount = SE.getSmallConstantTripCount(L, Latch); unsigned OuterTripMultiple = SE.getSmallConstantTripMultiple(L, Latch); unsigned InnerTripCount = SE.getSmallConstantTripCount(SubLoop, SubLoopLatch); // Decide if, and by how much, to unroll bool IsCountSetExplicitly = computeUnrollAndJamCount( L, SubLoop, TTI, DT, LI, &AC, SE, EphValues, &ORE, OuterTripCount, OuterTripMultiple, OuterUCE, InnerTripCount, InnerLoopSize, UP, PP); if (UP.Count <= 1) return LoopUnrollResult::Unmodified; // Unroll factor (Count) must be less or equal to TripCount. if (OuterTripCount && UP.Count > OuterTripCount) UP.Count = OuterTripCount; Loop *EpilogueOuterLoop = nullptr; LoopUnrollResult UnrollResult = UnrollAndJamLoop( L, UP.Count, OuterTripCount, OuterTripMultiple, UP.UnrollRemainder, LI, &SE, &DT, &AC, &TTI, &ORE, &EpilogueOuterLoop); // Assign new loop attributes. if (EpilogueOuterLoop) { std::optional NewOuterEpilogueLoopID = makeFollowupLoopID( OrigOuterLoopID, {LLVMLoopUnrollAndJamFollowupAll, LLVMLoopUnrollAndJamFollowupRemainderOuter}); if (NewOuterEpilogueLoopID) EpilogueOuterLoop->setLoopID(*NewOuterEpilogueLoopID); } std::optional NewInnerLoopID = makeFollowupLoopID(OrigOuterLoopID, {LLVMLoopUnrollAndJamFollowupAll, LLVMLoopUnrollAndJamFollowupInner}); if (NewInnerLoopID) SubLoop->setLoopID(*NewInnerLoopID); else SubLoop->setLoopID(OrigSubLoopID); if (UnrollResult == LoopUnrollResult::PartiallyUnrolled) { std::optional NewOuterLoopID = makeFollowupLoopID( OrigOuterLoopID, {LLVMLoopUnrollAndJamFollowupAll, LLVMLoopUnrollAndJamFollowupOuter}); if (NewOuterLoopID) { L->setLoopID(*NewOuterLoopID); // Do not setLoopAlreadyUnrolled if a followup was given. return UnrollResult; } } // If loop has an unroll count pragma or unrolled by explicitly set count // mark loop as unrolled to prevent unrolling beyond that requested. if (UnrollResult != LoopUnrollResult::FullyUnrolled && IsCountSetExplicitly) L->setLoopAlreadyUnrolled(); return UnrollResult; } static bool tryToUnrollAndJamLoop(LoopNest &LN, DominatorTree &DT, LoopInfo &LI, ScalarEvolution &SE, const TargetTransformInfo &TTI, AssumptionCache &AC, DependenceInfo &DI, OptimizationRemarkEmitter &ORE, int OptLevel, LPMUpdater &U) { bool DidSomething = false; ArrayRef Loops = LN.getLoops(); Loop *OutmostLoop = &LN.getOutermostLoop(); // Add the loop nests in the reverse order of LN. See method // declaration. SmallPriorityWorklist Worklist; appendLoopsToWorklist(Loops, Worklist); while (!Worklist.empty()) { Loop *L = Worklist.pop_back_val(); std::string LoopName = std::string(L->getName()); LoopUnrollResult Result = tryToUnrollAndJamLoop(L, DT, &LI, SE, TTI, AC, DI, ORE, OptLevel); if (Result != LoopUnrollResult::Unmodified) DidSomething = true; if (L == OutmostLoop && Result == LoopUnrollResult::FullyUnrolled) U.markLoopAsDeleted(*L, LoopName); } return DidSomething; } PreservedAnalyses LoopUnrollAndJamPass::run(LoopNest &LN, LoopAnalysisManager &AM, LoopStandardAnalysisResults &AR, LPMUpdater &U) { Function &F = *LN.getParent(); DependenceInfo DI(&F, &AR.AA, &AR.SE, &AR.LI); OptimizationRemarkEmitter ORE(&F); if (!tryToUnrollAndJamLoop(LN, AR.DT, AR.LI, AR.SE, AR.TTI, AR.AC, DI, ORE, OptLevel, U)) return PreservedAnalyses::all(); auto PA = getLoopPassPreservedAnalyses(); PA.preserve(); return PA; }