From: Geoff Berry Date: Wed, 28 Jun 2017 15:53:17 +0000 (+0000) Subject: [LoopUnroll] Pass SCEV to getUnrollingPreferences hook. NFCI. X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=28b3f06e1a44fb24aae4b3477663e39631ee531c;p=llvm [LoopUnroll] Pass SCEV to getUnrollingPreferences hook. NFCI. Reviewers: sanjoy, anna, reames, apilipenko, igor-laevsky, mkuper Subscribers: jholewinski, arsenm, mzolotukhin, nemanjai, nhaehnle, javed.absar, mcrosier, llvm-commits Differential Revision: https://reviews.llvm.org/D34531 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@306554 91177308-0d34-0410-b5e6-96231b3b80d8 --- diff --git a/include/llvm/Analysis/TargetTransformInfo.h b/include/llvm/Analysis/TargetTransformInfo.h index af2ebb7b6b4..f2a07006da4 100644 --- a/include/llvm/Analysis/TargetTransformInfo.h +++ b/include/llvm/Analysis/TargetTransformInfo.h @@ -366,7 +366,8 @@ public: /// \brief Get target-customized preferences for the generic loop unrolling /// transformation. The caller will initialize UP with the current /// target-independent defaults. - void getUnrollingPreferences(Loop *L, UnrollingPreferences &UP) const; + void getUnrollingPreferences(Loop *L, ScalarEvolution &, + UnrollingPreferences &UP) const; /// @} @@ -829,7 +830,8 @@ public: virtual bool isAlwaysUniform(const Value *V) = 0; virtual unsigned getFlatAddressSpace() = 0; virtual bool isLoweredToCall(const Function *F) = 0; - virtual void getUnrollingPreferences(Loop *L, UnrollingPreferences &UP) = 0; + virtual void getUnrollingPreferences(Loop *L, ScalarEvolution &, + UnrollingPreferences &UP) = 0; virtual bool isLegalAddImmediate(int64_t Imm) = 0; virtual bool isLegalICmpImmediate(int64_t Imm) = 0; virtual bool isLegalAddressingMode(Type *Ty, GlobalValue *BaseGV, @@ -1015,8 +1017,9 @@ public: bool isLoweredToCall(const Function *F) override { return Impl.isLoweredToCall(F); } - void getUnrollingPreferences(Loop *L, UnrollingPreferences &UP) override { - return Impl.getUnrollingPreferences(L, UP); + void getUnrollingPreferences(Loop *L, ScalarEvolution &SE, + UnrollingPreferences &UP) override { + return Impl.getUnrollingPreferences(L, SE, UP); } bool isLegalAddImmediate(int64_t Imm) override { return Impl.isLegalAddImmediate(Imm); diff --git a/include/llvm/Analysis/TargetTransformInfoImpl.h b/include/llvm/Analysis/TargetTransformInfoImpl.h index 24ac3b1213e..defd9e73bc0 100644 --- a/include/llvm/Analysis/TargetTransformInfoImpl.h +++ b/include/llvm/Analysis/TargetTransformInfoImpl.h @@ -217,7 +217,8 @@ public: return true; } - void getUnrollingPreferences(Loop *, TTI::UnrollingPreferences &) {} + void getUnrollingPreferences(Loop *, ScalarEvolution &, + TTI::UnrollingPreferences &) {} bool isLegalAddImmediate(int64_t Imm) { return false; } diff --git a/include/llvm/CodeGen/BasicTTIImpl.h b/include/llvm/CodeGen/BasicTTIImpl.h index 5eb7a0f61ee..a740df96899 100644 --- a/include/llvm/CodeGen/BasicTTIImpl.h +++ b/include/llvm/CodeGen/BasicTTIImpl.h @@ -277,7 +277,8 @@ public: unsigned getInliningThresholdMultiplier() { return 1; } - void getUnrollingPreferences(Loop *L, TTI::UnrollingPreferences &UP) { + void getUnrollingPreferences(Loop *L, ScalarEvolution &SE, + TTI::UnrollingPreferences &UP) { // This unrolling functionality is target independent, but to provide some // motivation for its intended use, for x86: diff --git a/lib/Analysis/TargetTransformInfo.cpp b/lib/Analysis/TargetTransformInfo.cpp index 92328f6e5ef..91160954bd9 100644 --- a/lib/Analysis/TargetTransformInfo.cpp +++ b/lib/Analysis/TargetTransformInfo.cpp @@ -116,8 +116,8 @@ bool TargetTransformInfo::isLoweredToCall(const Function *F) const { } void TargetTransformInfo::getUnrollingPreferences( - Loop *L, UnrollingPreferences &UP) const { - return TTIImpl->getUnrollingPreferences(L, UP); + Loop *L, ScalarEvolution &SE, UnrollingPreferences &UP) const { + return TTIImpl->getUnrollingPreferences(L, SE, UP); } bool TargetTransformInfo::isLegalAddImmediate(int64_t Imm) const { diff --git a/lib/Target/AArch64/AArch64TargetTransformInfo.cpp b/lib/Target/AArch64/AArch64TargetTransformInfo.cpp index 1d0f0b98b76..f5d038d52e1 100644 --- a/lib/Target/AArch64/AArch64TargetTransformInfo.cpp +++ b/lib/Target/AArch64/AArch64TargetTransformInfo.cpp @@ -645,10 +645,10 @@ unsigned AArch64TTIImpl::getMaxInterleaveFactor(unsigned VF) { return ST->getMaxInterleaveFactor(); } -void AArch64TTIImpl::getUnrollingPreferences(Loop *L, +void AArch64TTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE, TTI::UnrollingPreferences &UP) { // Enable partial unrolling and runtime unrolling. - BaseT::getUnrollingPreferences(L, UP); + BaseT::getUnrollingPreferences(L, SE, UP); // For inner loop, it is more likely to be a hot one, and the runtime check // can be promoted out from LICM pass, so the overhead is less, let's try diff --git a/lib/Target/AArch64/AArch64TargetTransformInfo.h b/lib/Target/AArch64/AArch64TargetTransformInfo.h index 75c62ea3233..31c03735492 100644 --- a/lib/Target/AArch64/AArch64TargetTransformInfo.h +++ b/lib/Target/AArch64/AArch64TargetTransformInfo.h @@ -122,7 +122,8 @@ public: int getCostOfKeepingLiveOverCall(ArrayRef Tys); - void getUnrollingPreferences(Loop *L, TTI::UnrollingPreferences &UP); + void getUnrollingPreferences(Loop *L, ScalarEvolution &SE, + TTI::UnrollingPreferences &UP); Value *getOrCreateResultFromMemIntrinsic(IntrinsicInst *Inst, Type *ExpectedType); diff --git a/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp b/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp index 88245b01683..89a03902dc6 100644 --- a/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp +++ b/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp @@ -63,7 +63,7 @@ static bool dependsOnLocalPhi(const Loop *L, const Value *Cond, return false; } -void AMDGPUTTIImpl::getUnrollingPreferences(Loop *L, +void AMDGPUTTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE, TTI::UnrollingPreferences &UP) { UP.Threshold = 300; // Twice the default. UP.MaxCount = UINT_MAX; diff --git a/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h b/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h index 485e20411ab..9a320bdfcc3 100644 --- a/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h +++ b/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.h @@ -68,7 +68,8 @@ public: bool hasBranchDivergence() { return true; } - void getUnrollingPreferences(Loop *L, TTI::UnrollingPreferences &UP); + void getUnrollingPreferences(Loop *L, ScalarEvolution &SE, + TTI::UnrollingPreferences &UP); TTI::PopcntSupportKind getPopcntSupport(unsigned TyWidth) { assert(isPowerOf2_32(TyWidth) && "Ty width must be power of 2"); diff --git a/lib/Target/Hexagon/HexagonTargetTransformInfo.cpp b/lib/Target/Hexagon/HexagonTargetTransformInfo.cpp index d578bfab365..d3848104d2b 100644 --- a/lib/Target/Hexagon/HexagonTargetTransformInfo.cpp +++ b/lib/Target/Hexagon/HexagonTargetTransformInfo.cpp @@ -29,7 +29,7 @@ HexagonTTIImpl::getPopcntSupport(unsigned IntTyWidthInBit) const { } // The Hexagon target can unroll loops with run-time trip counts. -void HexagonTTIImpl::getUnrollingPreferences(Loop *L, +void HexagonTTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE, TTI::UnrollingPreferences &UP) { UP.Runtime = UP.Partial = true; } diff --git a/lib/Target/Hexagon/HexagonTargetTransformInfo.h b/lib/Target/Hexagon/HexagonTargetTransformInfo.h index 8414bfc4e19..08faa2acd90 100644 --- a/lib/Target/Hexagon/HexagonTargetTransformInfo.h +++ b/lib/Target/Hexagon/HexagonTargetTransformInfo.h @@ -46,7 +46,8 @@ public: TTI::PopcntSupportKind getPopcntSupport(unsigned IntTyWidthInBit) const; // The Hexagon target can unroll loops with run-time trip counts. - void getUnrollingPreferences(Loop *L, TTI::UnrollingPreferences &UP); + void getUnrollingPreferences(Loop *L, ScalarEvolution &SE, + TTI::UnrollingPreferences &UP); // L1 cache prefetch. unsigned getPrefetchDistance() const; diff --git a/lib/Target/NVPTX/NVPTXTargetTransformInfo.cpp b/lib/Target/NVPTX/NVPTXTargetTransformInfo.cpp index dd770708494..a64d95512a4 100644 --- a/lib/Target/NVPTX/NVPTXTargetTransformInfo.cpp +++ b/lib/Target/NVPTX/NVPTXTargetTransformInfo.cpp @@ -141,9 +141,9 @@ int NVPTXTTIImpl::getArithmeticInstrCost( } } -void NVPTXTTIImpl::getUnrollingPreferences(Loop *L, +void NVPTXTTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE, TTI::UnrollingPreferences &UP) { - BaseT::getUnrollingPreferences(L, UP); + BaseT::getUnrollingPreferences(L, SE, UP); // Enable partial unrolling and runtime unrolling, but reduce the // threshold. This partially unrolls small loops which are often diff --git a/lib/Target/NVPTX/NVPTXTargetTransformInfo.h b/lib/Target/NVPTX/NVPTXTargetTransformInfo.h index 03075b55042..f987892ba67 100644 --- a/lib/Target/NVPTX/NVPTXTargetTransformInfo.h +++ b/lib/Target/NVPTX/NVPTXTargetTransformInfo.h @@ -61,7 +61,8 @@ public: TTI::OperandValueProperties Opd2PropInfo = TTI::OP_None, ArrayRef Args = ArrayRef()); - void getUnrollingPreferences(Loop *L, TTI::UnrollingPreferences &UP); + void getUnrollingPreferences(Loop *L, ScalarEvolution &SE, + TTI::UnrollingPreferences &UP); }; } // end namespace llvm diff --git a/lib/Target/PowerPC/PPCTargetTransformInfo.cpp b/lib/Target/PowerPC/PPCTargetTransformInfo.cpp index 3dbd5f5b9a9..6110706b01b 100644 --- a/lib/Target/PowerPC/PPCTargetTransformInfo.cpp +++ b/lib/Target/PowerPC/PPCTargetTransformInfo.cpp @@ -189,7 +189,7 @@ int PPCTTIImpl::getIntImmCost(unsigned Opcode, unsigned Idx, const APInt &Imm, return PPCTTIImpl::getIntImmCost(Imm, Ty); } -void PPCTTIImpl::getUnrollingPreferences(Loop *L, +void PPCTTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE, TTI::UnrollingPreferences &UP) { if (ST->getDarwinDirective() == PPC::DIR_A2) { // The A2 is in-order with a deep pipeline, and concatenation unrolling @@ -201,7 +201,7 @@ void PPCTTIImpl::getUnrollingPreferences(Loop *L, UP.AllowExpensiveTripCount = true; } - BaseT::getUnrollingPreferences(L, UP); + BaseT::getUnrollingPreferences(L, SE, UP); } bool PPCTTIImpl::enableAggressiveInterleaving(bool LoopHasReductions) { diff --git a/lib/Target/PowerPC/PPCTargetTransformInfo.h b/lib/Target/PowerPC/PPCTargetTransformInfo.h index 758c335def0..99ca6394d1b 100644 --- a/lib/Target/PowerPC/PPCTargetTransformInfo.h +++ b/lib/Target/PowerPC/PPCTargetTransformInfo.h @@ -52,7 +52,8 @@ public: Type *Ty); TTI::PopcntSupportKind getPopcntSupport(unsigned TyWidth); - void getUnrollingPreferences(Loop *L, TTI::UnrollingPreferences &UP); + void getUnrollingPreferences(Loop *L, ScalarEvolution &SE, + TTI::UnrollingPreferences &UP); /// @} diff --git a/lib/Target/SystemZ/SystemZTargetTransformInfo.cpp b/lib/Target/SystemZ/SystemZTargetTransformInfo.cpp index 422c16b8eb6..ce5c57e0f51 100644 --- a/lib/Target/SystemZ/SystemZTargetTransformInfo.cpp +++ b/lib/Target/SystemZ/SystemZTargetTransformInfo.cpp @@ -238,7 +238,7 @@ SystemZTTIImpl::getPopcntSupport(unsigned TyWidth) { return TTI::PSK_Software; } -void SystemZTTIImpl::getUnrollingPreferences(Loop *L, +void SystemZTTIImpl::getUnrollingPreferences(Loop *L, ScalarEvolution &SE, TTI::UnrollingPreferences &UP) { // Find out if L contains a call, what the machine instruction count // estimate is, and how many stores there are. diff --git a/lib/Target/SystemZ/SystemZTargetTransformInfo.h b/lib/Target/SystemZ/SystemZTargetTransformInfo.h index bdba7601eb7..6923fc6fc91 100644 --- a/lib/Target/SystemZ/SystemZTargetTransformInfo.h +++ b/lib/Target/SystemZ/SystemZTargetTransformInfo.h @@ -45,7 +45,8 @@ public: TTI::PopcntSupportKind getPopcntSupport(unsigned TyWidth); - void getUnrollingPreferences(Loop *L, TTI::UnrollingPreferences &UP); + void getUnrollingPreferences(Loop *L, ScalarEvolution &SE, + TTI::UnrollingPreferences &UP); /// @} diff --git a/lib/Transforms/Scalar/LoopUnrollPass.cpp b/lib/Transforms/Scalar/LoopUnrollPass.cpp index 62aa6ee4806..8bdbb3ff01f 100644 --- a/lib/Transforms/Scalar/LoopUnrollPass.cpp +++ b/lib/Transforms/Scalar/LoopUnrollPass.cpp @@ -131,7 +131,7 @@ static const unsigned NoThreshold = UINT_MAX; /// Gather the various unrolling parameters based on the defaults, compiler /// flags, TTI overrides and user specified parameters. static TargetTransformInfo::UnrollingPreferences gatherUnrollingPreferences( - Loop *L, const TargetTransformInfo &TTI, int OptLevel, + Loop *L, ScalarEvolution &SE, const TargetTransformInfo &TTI, int OptLevel, Optional UserThreshold, Optional UserCount, Optional UserAllowPartial, Optional UserRuntime, Optional UserUpperBound) { @@ -158,7 +158,7 @@ static TargetTransformInfo::UnrollingPreferences gatherUnrollingPreferences( UP.AllowPeeling = true; // Override with any target specific settings - TTI.getUnrollingPreferences(L, UP); + TTI.getUnrollingPreferences(L, SE, UP); // Apply size attributes if (L->getHeader()->getParent()->optForSize()) { @@ -699,7 +699,7 @@ static uint64_t getUnrolledLoopSize( // Calculates unroll count and writes it to UP.Count. static bool computeUnrollCount( Loop *L, const TargetTransformInfo &TTI, DominatorTree &DT, LoopInfo *LI, - ScalarEvolution *SE, OptimizationRemarkEmitter *ORE, unsigned &TripCount, + ScalarEvolution &SE, OptimizationRemarkEmitter *ORE, unsigned &TripCount, unsigned MaxTripCount, unsigned &TripMultiple, unsigned LoopSize, TargetTransformInfo::UnrollingPreferences &UP, bool &UseUpperBound) { // Check for explicit Count. @@ -770,7 +770,7 @@ static bool computeUnrollCount( // helps to remove a significant number of instructions. // To check that, run additional analysis on the loop. if (Optional Cost = analyzeLoopUnrollCost( - L, FullUnrollTripCount, DT, *SE, TTI, + L, FullUnrollTripCount, DT, SE, TTI, UP.Threshold * UP.MaxPercentThresholdBoost / 100)) { unsigned Boost = getFullUnrollBoostingFactor(*Cost, UP.MaxPercentThresholdBoost); @@ -926,7 +926,7 @@ static bool computeUnrollCount( } static bool tryToUnrollLoop(Loop *L, DominatorTree &DT, LoopInfo *LI, - ScalarEvolution *SE, const TargetTransformInfo &TTI, + ScalarEvolution &SE, const TargetTransformInfo &TTI, AssumptionCache &AC, OptimizationRemarkEmitter &ORE, bool PreserveLCSSA, int OptLevel, Optional ProvidedCount, @@ -948,8 +948,8 @@ static bool tryToUnrollLoop(Loop *L, DominatorTree &DT, LoopInfo *LI, bool NotDuplicatable; bool Convergent; TargetTransformInfo::UnrollingPreferences UP = gatherUnrollingPreferences( - L, TTI, OptLevel, ProvidedThreshold, ProvidedCount, ProvidedAllowPartial, - ProvidedRuntime, ProvidedUpperBound); + L, SE, TTI, OptLevel, ProvidedThreshold, ProvidedCount, + ProvidedAllowPartial, ProvidedRuntime, ProvidedUpperBound); // Exit early if unrolling is disabled. if (UP.Threshold == 0 && (!UP.Partial || UP.PartialThreshold == 0)) return false; @@ -977,8 +977,8 @@ static bool tryToUnrollLoop(Loop *L, DominatorTree &DT, LoopInfo *LI, if (!ExitingBlock || !L->isLoopExiting(ExitingBlock)) ExitingBlock = L->getExitingBlock(); if (ExitingBlock) { - TripCount = SE->getSmallConstantTripCount(L, ExitingBlock); - TripMultiple = SE->getSmallConstantTripMultiple(L, ExitingBlock); + TripCount = SE.getSmallConstantTripCount(L, ExitingBlock); + TripMultiple = SE.getSmallConstantTripMultiple(L, ExitingBlock); } // If the loop contains a convergent operation, the prelude we'd add @@ -1000,8 +1000,8 @@ static bool tryToUnrollLoop(Loop *L, DominatorTree &DT, LoopInfo *LI, // count. bool MaxOrZero = false; if (!TripCount) { - MaxTripCount = SE->getSmallConstantMaxTripCount(L); - MaxOrZero = SE->isBackedgeTakenCountMaxOrZero(L); + MaxTripCount = SE.getSmallConstantMaxTripCount(L); + MaxOrZero = SE.isBackedgeTakenCountMaxOrZero(L); // We can unroll by the upper bound amount if it's generally allowed or if // we know that the loop is executed either the upper bound or zero times. // (MaxOrZero unrolling keeps only the first loop test, so the number of @@ -1030,7 +1030,7 @@ static bool tryToUnrollLoop(Loop *L, DominatorTree &DT, LoopInfo *LI, // Unroll the loop. if (!UnrollLoop(L, UP.Count, TripCount, UP.Force, UP.Runtime, UP.AllowExpensiveTripCount, UseUpperBound, MaxOrZero, - TripMultiple, UP.PeelCount, LI, SE, &DT, &AC, &ORE, + TripMultiple, UP.PeelCount, LI, &SE, &DT, &AC, &ORE, PreserveLCSSA)) return false; @@ -1073,7 +1073,7 @@ public: auto &DT = getAnalysis().getDomTree(); LoopInfo *LI = &getAnalysis().getLoopInfo(); - ScalarEvolution *SE = &getAnalysis().getSE(); + ScalarEvolution &SE = getAnalysis().getSE(); const TargetTransformInfo &TTI = getAnalysis().getTTI(F); auto &AC = getAnalysis().getAssumptionCache(F); @@ -1157,7 +1157,7 @@ PreservedAnalyses LoopUnrollPass::run(Loop &L, LoopAnalysisManager &AM, if (!AllowPartialUnrolling) AllowPartialParam = RuntimeParam = UpperBoundParam = false; bool Changed = tryToUnrollLoop( - &L, AR.DT, &AR.LI, &AR.SE, AR.TTI, AR.AC, *ORE, + &L, AR.DT, &AR.LI, AR.SE, AR.TTI, AR.AC, *ORE, /*PreserveLCSSA*/ true, OptLevel, /*Count*/ None, /*Threshold*/ None, AllowPartialParam, RuntimeParam, UpperBoundParam); if (!Changed)