LLVM 20.0.0git
|
#include "llvm/Analysis/LoopAccessAnalysis.h"
#include "llvm/ADT/APInt.h"
#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/EquivalenceClasses.h"
#include "llvm/ADT/PointerIntPair.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/SetVector.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/ADT/SmallSet.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/Analysis/AliasSetTracker.h"
#include "llvm/Analysis/LoopAnalysisManager.h"
#include "llvm/Analysis/LoopInfo.h"
#include "llvm/Analysis/LoopIterator.h"
#include "llvm/Analysis/MemoryLocation.h"
#include "llvm/Analysis/OptimizationRemarkEmitter.h"
#include "llvm/Analysis/ScalarEvolution.h"
#include "llvm/Analysis/ScalarEvolutionExpressions.h"
#include "llvm/Analysis/TargetLibraryInfo.h"
#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/Analysis/ValueTracking.h"
#include "llvm/Analysis/VectorUtils.h"
#include "llvm/IR/BasicBlock.h"
#include "llvm/IR/Constants.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/DebugLoc.h"
#include "llvm/IR/DerivedTypes.h"
#include "llvm/IR/DiagnosticInfo.h"
#include "llvm/IR/Dominators.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/GetElementPtrTypeIterator.h"
#include "llvm/IR/InstrTypes.h"
#include "llvm/IR/Instruction.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/Operator.h"
#include "llvm/IR/PassManager.h"
#include "llvm/IR/PatternMatch.h"
#include "llvm/IR/Type.h"
#include "llvm/IR/Value.h"
#include "llvm/IR/ValueHandle.h"
#include "llvm/Support/Casting.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/raw_ostream.h"
#include <algorithm>
#include <cassert>
#include <cstdint>
#include <iterator>
#include <utility>
#include <variant>
#include <vector>
Go to the source code of this file.
Macros | |
#define | DEBUG_TYPE "loop-accesses" |
Functions | |
static std::pair< const SCEV *, const SCEV * > | getStartAndEndForAccess (const Loop *Lp, const SCEV *PtrExpr, Type *AccessTy, PredicatedScalarEvolution &PSE, DenseMap< std::pair< const SCEV *, Type * >, std::pair< const SCEV *, const SCEV * > > &PointerBounds) |
Calculate Start and End points of memory access. | |
static const SCEV * | getMinFromExprs (const SCEV *I, const SCEV *J, ScalarEvolution *SE) |
Compare I and J and return the minimum. | |
static bool | hasComputableBounds (PredicatedScalarEvolution &PSE, Value *Ptr, const SCEV *PtrScev, Loop *L, bool Assume) |
Check whether a pointer can participate in a runtime bounds check. | |
static bool | isNoWrap (PredicatedScalarEvolution &PSE, const DenseMap< Value *, const SCEV * > &Strides, Value *Ptr, Type *AccessTy, Loop *L) |
Check whether a pointer address cannot wrap. | |
static void | visitPointers (Value *StartPtr, const Loop &InnermostLoop, function_ref< void(Value *)> AddPointer) |
static void | findForkedSCEVs (ScalarEvolution *SE, const Loop *L, Value *Ptr, SmallVectorImpl< PointerIntPair< const SCEV *, 1, bool > > &ScevList, unsigned Depth) |
static SmallVector< PointerIntPair< const SCEV *, 1, bool > > | findForkedPointer (PredicatedScalarEvolution &PSE, const DenseMap< Value *, const SCEV * > &StridesMap, Value *Ptr, const Loop *L) |
static bool | isNoWrapAddRec (Value *Ptr, const SCEVAddRecExpr *AR, PredicatedScalarEvolution &PSE, const Loop *L) |
Return true if an AddRec pointer Ptr is unsigned non-wrapping, i.e. | |
static bool | isSafeDependenceDistance (const DataLayout &DL, ScalarEvolution &SE, const SCEV &MaxBTC, const SCEV &Dist, uint64_t MaxStride, uint64_t TypeByteSize) |
Given a dependence-distance Dist between two memory accesses, that have strides in the same direction whose absolute value of the maximum stride is given in MaxStride , and that have the same type size TypeByteSize , in a loop whose maximum backedge taken count is MaxBTC , check if it is possible to prove statically that the dependence distance is larger than the range that the accesses will travel through the execution of the loop. | |
static bool | areStridedAccessesIndependent (uint64_t Distance, uint64_t Stride, uint64_t TypeByteSize) |
Check the dependence for two accesses with the same stride Stride . | |
static unsigned | getGEPInductionOperand (const GetElementPtrInst *Gep) |
Find the operand of the GEP that should be checked for consecutive stores. | |
static Value * | stripGetElementPtr (Value *Ptr, ScalarEvolution *SE, Loop *Lp) |
If the argument is a GEP, then returns the operand identified by getGEPInductionOperand. | |
static const SCEV * | getStrideFromPointer (Value *Ptr, ScalarEvolution *SE, Loop *Lp) |
Get the stride of a pointer access in a loop. | |
Variables | |
static cl::opt< unsigned, true > | VectorizationFactor ("force-vector-width", cl::Hidden, cl::desc("Sets the SIMD width. Zero is autoselect."), cl::location(VectorizerParams::VectorizationFactor)) |
static cl::opt< unsigned, true > | VectorizationInterleave ("force-vector-interleave", cl::Hidden, cl::desc("Sets the vectorization interleave count. " "Zero is autoselect."), cl::location(VectorizerParams::VectorizationInterleave)) |
static cl::opt< unsigned, true > | RuntimeMemoryCheckThreshold ("runtime-memory-check-threshold", cl::Hidden, cl::desc("When performing memory disambiguation checks at runtime do not " "generate more than this number of comparisons (default = 8)."), cl::location(VectorizerParams::RuntimeMemoryCheckThreshold), cl::init(8)) |
static cl::opt< unsigned > | MemoryCheckMergeThreshold ("memory-check-merge-threshold", cl::Hidden, cl::desc("Maximum number of comparisons done when trying to merge " "runtime memory checks. (default = 100)"), cl::init(100)) |
The maximum iterations used to merge memory checks. | |
static cl::opt< unsigned > | MaxDependences ("max-dependences", cl::Hidden, cl::desc("Maximum number of dependences collected by " "loop-access analysis (default = 100)"), cl::init(100)) |
We collect dependences up to this threshold. | |
static cl::opt< bool > | EnableMemAccessVersioning ("enable-mem-access-versioning", cl::init(true), cl::Hidden, cl::desc("Enable symbolic stride memory access versioning")) |
This enables versioning on the strides of symbolically striding memory accesses in code like the following. | |
static cl::opt< bool > | EnableForwardingConflictDetection ("store-to-load-forwarding-conflict-detection", cl::Hidden, cl::desc("Enable conflict detection in loop-access analysis"), cl::init(true)) |
Enable store-to-load forwarding conflict detection. | |
static cl::opt< unsigned > | MaxForkedSCEVDepth ("max-forked-scev-depth", cl::Hidden, cl::desc("Maximum recursion depth when finding forked SCEVs (default = 5)"), cl::init(5)) |
static cl::opt< bool > | SpeculateUnitStride ("laa-speculate-unit-stride", cl::Hidden, cl::desc("Speculate that non-constant strides are unit in LAA"), cl::init(true)) |
static cl::opt< bool, true > | HoistRuntimeChecks ("hoist-runtime-checks", cl::Hidden, cl::desc("Hoist inner loop runtime memory checks to outer loop if possible"), cl::location(VectorizerParams::HoistRuntimeChecks), cl::init(true)) |
#define DEBUG_TYPE "loop-accesses" |
Definition at line 71 of file LoopAccessAnalysis.cpp.
|
static |
Check the dependence for two accesses with the same stride Stride
.
Distance
is the positive distance and TypeByteSize
is type size in bytes.
Definition at line 1863 of file LoopAccessAnalysis.cpp.
References assert().
|
static |
Definition at line 1036 of file LoopAccessAnalysis.cpp.
References assert(), llvm::dbgs(), findForkedSCEVs(), llvm::PredicatedScalarEvolution::getSE(), llvm::ScalarEvolution::isLoopInvariant(), llvm::ScalarEvolution::isSCEVable(), LLVM_DEBUG, MaxForkedSCEVDepth, Ptr, llvm::replaceSymbolicStrideSCEV(), and llvm::SmallVectorBase< Size_T >::size().
|
static |
Definition at line 881 of file LoopAccessAnalysis.cpp.
References llvm::any_of(), llvm::dbgs(), llvm::Depth, findForkedSCEVs(), GEP, llvm::ScalarEvolution::getAddExpr(), llvm::ScalarEvolution::getEffectiveSCEVType(), llvm::ScalarEvolution::getMinusSCEV(), llvm::ScalarEvolution::getMulExpr(), llvm::ScalarEvolution::getSCEV(), llvm::ScalarEvolution::getSizeOfExpr(), llvm::ScalarEvolution::getTruncateOrSignExtend(), llvm::SCEV::getType(), I, llvm::isGuaranteedNotToBeUndefOrPoison(), llvm::Type::isVectorTy(), LLVM_DEBUG, llvm_unreachable, Ptr, llvm::SmallVectorTemplateBase< T, bool >::push_back(), llvm::SmallVectorBase< Size_T >::size(), and Size.
Referenced by findForkedPointer(), and findForkedSCEVs().
|
static |
Find the operand of the GEP that should be checked for consecutive stores.
This ignores trailing indices that have no effect on the final pointer.
Definition at line 2806 of file LoopAccessAnalysis.cpp.
References DL, llvm::gep_type_begin(), llvm::Instruction::getDataLayout(), llvm::generic_gep_type_iterator< ItTy >::getIndexedType(), llvm::User::getNumOperands(), llvm::User::getOperand(), llvm::GetElementPtrInst::getResultElementType(), llvm::generic_gep_type_iterator< ItTy >::getSequentialElementStride(), llvm::generic_gep_type_iterator< ItTy >::isStruct(), llvm::PatternMatch::m_Zero(), and llvm::PatternMatch::match().
Referenced by stripGetElementPtr().
Compare I
and J
and return the minimum.
Return nullptr in case we couldn't find an answer.
Definition at line 413 of file LoopAccessAnalysis.cpp.
References llvm::ScalarEvolution::computeConstantDifference(), and I.
Referenced by llvm::RuntimeCheckingPtrGroup::addPointer().
|
static |
Calculate Start and End points of memory access.
Let's assume A is the first access and B is a memory access on N-th loop iteration. Then B is calculated as: B = A + Step*N . Step value may be positive or negative. N is a calculated back-edge taken count: N = (TripCount > 0) ? RoundDown(TripCount -1 , VF) : 0 Start and End points are calculated in the following way: Start = UMIN(A, B) ; End = UMAX(A, B) + SizeOfElt, where SizeOfElt is the size of single memory access in bytes.
There is no conflict when the intervals are disjoint: NoConflict = (P2.Start >= P1.End) || (P1.Start >= P2.End)
Definition at line 206 of file LoopAccessAnalysis.cpp.
References assert(), DL, llvm::ScalarEvolution::getAddExpr(), llvm::ScalarEvolution::getCouldNotCompute(), llvm::BasicBlock::getDataLayout(), llvm::LoopBase< BlockT, LoopT >::getHeader(), llvm::PredicatedScalarEvolution::getSE(), llvm::ScalarEvolution::getStoreSizeOfExpr(), llvm::PredicatedScalarEvolution::getSymbolicMaxBackedgeTakenCount(), llvm::SCEV::getType(), llvm::ScalarEvolution::getUMaxExpr(), llvm::ScalarEvolution::getUMinExpr(), llvm::ScalarEvolution::isLoopInvariant(), and std::swap().
Referenced by llvm::RuntimePointerChecking::insert().
|
static |
Get the stride of a pointer access in a loop.
Looks for symbolic strides "a[i*stride]". Returns the symbolic stride, or null otherwise.
Definition at line 2851 of file LoopAccessAnalysis.cpp.
References llvm::CallingConv::C, llvm::APInt::getBitWidth(), llvm::SCEVAddRecExpr::getLoop(), llvm::ScalarEvolution::getSCEV(), llvm::APInt::getSExtValue(), llvm::SCEVAddRecExpr::getStepRecurrence(), llvm::ScalarEvolution::isLoopInvariant(), Ptr, llvm::scConstant, and stripGetElementPtr().
|
static |
Check whether a pointer can participate in a runtime bounds check.
If Assume
, try harder to prove that we can compute the bounds of Ptr
by adding run-time checks (overflow checks) if necessary.
Definition at line 811 of file LoopAccessAnalysis.cpp.
References llvm::PredicatedScalarEvolution::getAsAddRec(), llvm::PredicatedScalarEvolution::getSE(), llvm::SCEVAddRecExpr::isAffine(), llvm::ScalarEvolution::isLoopInvariant(), and Ptr.
|
static |
Check whether a pointer address cannot wrap.
Definition at line 829 of file LoopAccessAnalysis.cpp.
References llvm::getPtrStride(), llvm::PredicatedScalarEvolution::getSCEV(), llvm::PredicatedScalarEvolution::getSE(), llvm::PredicatedScalarEvolution::hasNoOverflow(), llvm::SCEVWrapPredicate::IncrementNUSW, llvm::ScalarEvolution::isLoopInvariant(), and Ptr.
|
static |
Return true if an AddRec pointer Ptr
is unsigned non-wrapping, i.e.
monotonically increasing/decreasing.
Definition at line 1400 of file LoopAccessAnalysis.cpp.
References llvm::SCEV::FlagNSW, GEP, llvm::SCEVNAryExpr::getNoWrapFlags(), llvm::PredicatedScalarEvolution::getSCEV(), llvm::PredicatedScalarEvolution::hasNoOverflow(), llvm::SCEVWrapPredicate::IncrementNUSW, llvm::SCEV::NoWrapMask, and Ptr.
Referenced by llvm::getPtrStride().
|
static |
Given a dependence-distance Dist
between two memory accesses, that have strides in the same direction whose absolute value of the maximum stride is given in MaxStride
, and that have the same type size TypeByteSize
, in a loop whose maximum backedge taken count is MaxBTC
, check if it is possible to prove statically that the dependence distance is larger than the range that the accesses will travel through the execution of the loop.
If so, return true; false otherwise. This is useful for example in loops such as the following (PR31098): for (i = 0; i < D; ++i) { = out[i]; out[i+D] = }
Definition at line 1806 of file LoopAccessAnalysis.cpp.
References DL, llvm::ScalarEvolution::getConstant(), llvm::ScalarEvolution::getMinusSCEV(), llvm::ScalarEvolution::getMulExpr(), llvm::ScalarEvolution::getNegativeSCEV(), llvm::ScalarEvolution::getNoopOrSignExtend(), llvm::SCEV::getType(), llvm::ScalarEvolution::getZeroExtendExpr(), llvm::ScalarEvolution::isKnownPositive(), and llvm::Minus.
|
static |
If the argument is a GEP, then returns the operand identified by getGEPInductionOperand.
However, if there is some other non-loop-invariant operand, it returns that instead.
Definition at line 2833 of file LoopAccessAnalysis.cpp.
References GEP, getGEPInductionOperand(), llvm::ScalarEvolution::getSCEV(), I, llvm::ScalarEvolution::isLoopInvariant(), and Ptr.
Referenced by getStrideFromPointer().
|
static |
Definition at line 841 of file LoopAccessAnalysis.cpp.
References llvm::LoopBase< BlockT, LoopT >::contains(), llvm::SmallVectorBase< Size_T >::empty(), llvm::LoopBase< BlockT, LoopT >::getHeader(), llvm::SmallPtrSetImpl< PtrType >::insert(), llvm::SmallVectorImpl< T >::pop_back_val(), Ptr, and llvm::SmallVectorTemplateBase< T, bool >::push_back().
Referenced by llvm::MemoryDepChecker::addAccess().
|
static |
Enable store-to-load forwarding conflict detection.
This option can be disabled for correctness testing.
|
static |
This enables versioning on the strides of symbolically striding memory accesses in code like the following.
for (i = 0; i < N; ++i) A[i * Stride1] += B[i * Stride2] ...
Will be roughly translated to if (Stride1 == 1 && Stride2 == 1) { for (i = 0; i < N; i+=4) A[i:i+3] += ... } else ...
|
static |
Referenced by llvm::addRuntimeChecks(), and expandBounds().
|
static |
We collect dependences up to this threshold.
Referenced by llvm::MemoryDepChecker::areDepsSafe().
|
static |
Referenced by findForkedPointer().
|
static |
The maximum iterations used to merge memory checks.
|
static |
|
static |
|
static |
|
static |