LLVM 18.0.0git
|
This transformation implements the well known scalar replacement of aggregates transformation. More...
#include "llvm/Transforms/Scalar/SROA.h"
#include "llvm/ADT/APInt.h"
#include "llvm/ADT/ArrayRef.h"
#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/PointerIntPair.h"
#include "llvm/ADT/STLExtras.h"
#include "llvm/ADT/SetVector.h"
#include "llvm/ADT/SmallBitVector.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/ADT/SmallVector.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/ADT/StringRef.h"
#include "llvm/ADT/Twine.h"
#include "llvm/ADT/iterator.h"
#include "llvm/ADT/iterator_range.h"
#include "llvm/Analysis/AssumptionCache.h"
#include "llvm/Analysis/DomTreeUpdater.h"
#include "llvm/Analysis/GlobalsModRef.h"
#include "llvm/Analysis/Loads.h"
#include "llvm/Analysis/PtrUseVisitor.h"
#include "llvm/Config/llvm-config.h"
#include "llvm/IR/BasicBlock.h"
#include "llvm/IR/Constant.h"
#include "llvm/IR/ConstantFolder.h"
#include "llvm/IR/Constants.h"
#include "llvm/IR/DIBuilder.h"
#include "llvm/IR/DataLayout.h"
#include "llvm/IR/DebugInfo.h"
#include "llvm/IR/DebugInfoMetadata.h"
#include "llvm/IR/DerivedTypes.h"
#include "llvm/IR/Dominators.h"
#include "llvm/IR/Function.h"
#include "llvm/IR/GetElementPtrTypeIterator.h"
#include "llvm/IR/GlobalAlias.h"
#include "llvm/IR/IRBuilder.h"
#include "llvm/IR/InstVisitor.h"
#include "llvm/IR/Instruction.h"
#include "llvm/IR/Instructions.h"
#include "llvm/IR/IntrinsicInst.h"
#include "llvm/IR/LLVMContext.h"
#include "llvm/IR/Metadata.h"
#include "llvm/IR/Module.h"
#include "llvm/IR/Operator.h"
#include "llvm/IR/PassManager.h"
#include "llvm/IR/Type.h"
#include "llvm/IR/Use.h"
#include "llvm/IR/User.h"
#include "llvm/IR/Value.h"
#include "llvm/InitializePasses.h"
#include "llvm/Pass.h"
#include "llvm/Support/Casting.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Compiler.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Transforms/Scalar.h"
#include "llvm/Transforms/Utils/BasicBlockUtils.h"
#include "llvm/Transforms/Utils/Local.h"
#include "llvm/Transforms/Utils/PromoteMemToReg.h"
#include <algorithm>
#include <cassert>
#include <cstddef>
#include <cstdint>
#include <cstring>
#include <iterator>
#include <string>
#include <tuple>
#include <utility>
#include <vector>
Go to the source code of this file.
Classes | |
class | llvm::sroa::AllocaSlices |
Representation of the alloca slices. More... | |
class | llvm::sroa::Partition |
A partition of the slices. More... | |
class | llvm::sroa::AllocaSlices::partition_iterator |
An iterator over partitions of the alloca's slices. More... | |
class | llvm::sroa::AllocaSlices::SliceBuilder |
Builder for the alloca slices. More... | |
class | llvm::sroa::AllocaSliceRewriter |
Visitor to rewrite instructions using p particular slice of an alloca to use a new alloca. More... | |
class | llvm::sroa::SROALegacyPass |
A legacy pass for the legacy pass manager that wraps the SROA pass. More... | |
Macros | |
#define | DEBUG_TYPE "sroa" |
Functions | |
STATISTIC (NumAllocasAnalyzed, "Number of allocas analyzed for replacement") | |
STATISTIC (NumAllocaPartitions, "Number of alloca partitions formed") | |
STATISTIC (MaxPartitionsPerAlloca, "Maximum number of partitions per alloca") | |
STATISTIC (NumAllocaPartitionUses, "Number of alloca partition uses rewritten") | |
STATISTIC (MaxUsesPerAllocaPartition, "Maximum number of uses of a partition") | |
STATISTIC (NumNewAllocas, "Number of new, smaller allocas introduced") | |
STATISTIC (NumPromoted, "Number of allocas promoted to SSA values") | |
STATISTIC (NumLoadsSpeculated, "Number of loads speculated to allow promotion") | |
STATISTIC (NumLoadsPredicated, "Number of loads rewritten into predicated loads to allow promotion") | |
STATISTIC (NumStoresPredicated, "Number of stores rewritten into predicated loads to allow promotion") | |
STATISTIC (NumDeleted, "Number of instructions deleted") | |
STATISTIC (NumVectorized, "Number of vectorized aggregates") | |
static Value * | foldSelectInst (SelectInst &SI) |
static Value * | foldPHINodeOrSelectInst (Instruction &I) |
A helper that folds a PHI node or a select. | |
static std::pair< Type *, IntegerType * > | findCommonType (AllocaSlices::const_iterator B, AllocaSlices::const_iterator E, uint64_t EndOffset) |
Walk the range of a partitioning looking for a common type to cover this sequence of slices. | |
static bool | isSafePHIToSpeculate (PHINode &PN) |
PHI instructions that use an alloca and are subsequently loaded can be rewritten to load both input pointers in the pred blocks and then PHI the results, allowing the load of the alloca to be promoted. | |
static void | speculatePHINodeLoads (IRBuilderTy &IRB, PHINode &PN) |
static sroa::SelectHandSpeculativity | isSafeLoadOfSelectToSpeculate (LoadInst &LI, SelectInst &SI, bool PreserveCFG) |
static void | speculateSelectInstLoads (SelectInst &SI, LoadInst &LI, IRBuilderTy &IRB) |
template<typename T > | |
static void | rewriteMemOpOfSelect (SelectInst &SI, T &I, sroa::SelectHandSpeculativity Spec, DomTreeUpdater &DTU) |
static void | rewriteMemOpOfSelect (SelectInst &SelInst, Instruction &I, sroa::SelectHandSpeculativity Spec, DomTreeUpdater &DTU) |
static bool | rewriteSelectInstMemOps (SelectInst &SI, const sroa::RewriteableMemOps &Ops, IRBuilderTy &IRB, DomTreeUpdater *DTU) |
static Value * | getAdjustedPtr (IRBuilderTy &IRB, const DataLayout &DL, Value *Ptr, APInt Offset, Type *PointerTy, const Twine &NamePrefix) |
Compute an adjusted pointer from Ptr by Offset bytes where the resulting pointer has PointerTy. | |
static Align | getAdjustedAlignment (Instruction *I, uint64_t Offset) |
Compute the adjusted alignment for a load or store from an offset. | |
static bool | canConvertValue (const DataLayout &DL, Type *OldTy, Type *NewTy) |
Test whether we can convert a value from the old to the new type. | |
static Value * | convertValue (const DataLayout &DL, IRBuilderTy &IRB, Value *V, Type *NewTy) |
Generic routine to convert an SSA value to a value of a different type. | |
static bool | isVectorPromotionViableForSlice (Partition &P, const Slice &S, VectorType *Ty, uint64_t ElementSize, const DataLayout &DL) |
Test whether the given slice use can be promoted to a vector. | |
static bool | checkVectorTypeForPromotion (Partition &P, VectorType *VTy, const DataLayout &DL) |
Test whether a vector type is viable for promotion. | |
static VectorType * | isVectorPromotionViable (Partition &P, const DataLayout &DL) |
Test whether the given alloca partitioning and range of slices can be promoted to a vector. | |
static bool | isIntegerWideningViableForSlice (const Slice &S, uint64_t AllocBeginOffset, Type *AllocaTy, const DataLayout &DL, bool &WholeAllocaOp) |
Test whether a slice of an alloca is valid for integer widening. | |
static bool | isIntegerWideningViable (Partition &P, Type *AllocaTy, const DataLayout &DL) |
Test whether the given alloca partition's integer operations can be widened to promotable ones. | |
static Value * | extractInteger (const DataLayout &DL, IRBuilderTy &IRB, Value *V, IntegerType *Ty, uint64_t Offset, const Twine &Name) |
static Value * | insertInteger (const DataLayout &DL, IRBuilderTy &IRB, Value *Old, Value *V, uint64_t Offset, const Twine &Name) |
static Value * | extractVector (IRBuilderTy &IRB, Value *V, unsigned BeginIndex, unsigned EndIndex, const Twine &Name) |
static Value * | insertVector (IRBuilderTy &IRB, Value *Old, Value *V, unsigned BeginIndex, const Twine &Name) |
static Type * | stripAggregateTypeWrapping (const DataLayout &DL, Type *Ty) |
Strip aggregate type wrapping. | |
static Type * | getTypePartition (const DataLayout &DL, Type *Ty, uint64_t Offset, uint64_t Size) |
Try to find a partition of the aggregate type passed in for a given offset and size. | |
INITIALIZE_PASS_BEGIN (SROALegacyPass, "sroa", "Scalar Replacement Of Aggregates", false, false) INITIALIZE_PASS_END(SROALegacyPass | |
Variables | |
static cl::opt< bool > | SROAStrictInbounds ("sroa-strict-inbounds", cl::init(false), cl::Hidden) |
Hidden option to experiment with completely strict handling of inbounds GEPs. | |
static cl::opt< bool > | SROASkipMem2Reg ("sroa-skip-mem2reg", cl::init(false), cl::Hidden) |
Disable running mem2reg during SROA in order to test or debug SROA. | |
static cl::opt< int > | SROAMaxAllocaSlices ("sroa-max-alloca-slices", cl::init(1024), cl::desc("Maximum number of alloca slices allowed " "after which splitting is not attempted"), cl::Hidden) |
The maximum number of alloca slices allowed when splitting. | |
sroa | |
Scalar Replacement Of | Aggregates |
Scalar Replacement Of | false |
This transformation implements the well known scalar replacement of aggregates transformation.
It tries to identify promotable elements of an aggregate alloca, and promote them to registers. It will also try to convert uses of an element (or set of elements) of an alloca into a vector or bitfield-style integer scalar if appropriate.
It works to do this with minimal slicing of the alloca so that regions which are merely transferred in and out of external memory remain unchanged and are not decomposed to scalar code.
Because this also performs alloca promotion, it can be thought of as also serving the purpose of SSA formation. The algorithm iterates on the function until all opportunities for promotion have been realized.
Definition in file SROA.cpp.
|
static |
Test whether we can convert a value from the old to the new type.
This predicate should be used to guard calls to convertValue in order to ensure that we only try to convert viable values. The strategy is that we will peel off single element struct and array wrappings to get to an underlying value, and convert that value.
Definition at line 1795 of file SROA.cpp.
References assert(), DL, getBitWidth(), llvm::Type::getPointerAddressSpace(), llvm::Type::getScalarType(), llvm::Type::isIntegerTy(), llvm::Type::isPointerTy(), llvm::Type::isSingleValueType(), and llvm::Type::isTargetExtTy().
Referenced by convertValue(), isIntegerWideningViable(), isIntegerWideningViableForSlice(), and isVectorPromotionViableForSlice().
|
static |
Test whether a vector type is viable for promotion.
This implements the necessary checking for isVectorPromotionViable
over all slices of the alloca for the given VectorType.
Definition at line 1985 of file SROA.cpp.
References assert(), DL, isVectorPromotionViableForSlice(), and P.
Referenced by isVectorPromotionViable().
|
static |
Generic routine to convert an SSA value to a value of a different type.
This will try various different casting techniques, such as bitcasts, inttoptr, and ptrtoint casts. Use the canConvertValue
predicate to test two types for viability with this routine.
Definition at line 1857 of file SROA.cpp.
References assert(), canConvertValue(), DL, llvm::Type::getPointerAddressSpace(), llvm::Type::isIntOrIntVectorTy(), and llvm::Type::isPtrOrPtrVectorTy().
|
static |
Definition at line 2311 of file SROA.cpp.
References assert(), llvm::dbgs(), DL, llvm::IntegerType::getBitWidth(), LLVM_DEBUG, Name, and llvm::Offset.
|
static |
Definition at line 2369 of file SROA.cpp.
References assert(), llvm::dbgs(), LLVM_DEBUG, and Name.
|
static |
Walk the range of a partitioning looking for a common type to cover this sequence of slices.
Definition at line 1351 of file SROA.cpp.
References B, E, llvm::IntegerType::getBitWidth(), llvm::Value::getType(), and I.
|
static |
A helper that folds a PHI node or a select.
Definition at line 850 of file SROA.cpp.
References foldSelectInst(), and I.
|
static |
Definition at line 837 of file SROA.cpp.
Referenced by foldPHINodeOrSelectInst().
|
static |
Compute the adjusted alignment for a load or store from an offset.
Definition at line 1785 of file SROA.cpp.
References llvm::commonAlignment(), llvm::getLoadStoreAlignment(), I, and llvm::Offset.
|
static |
Compute an adjusted pointer from Ptr by Offset bytes where the resulting pointer has PointerTy.
Definition at line 1774 of file SROA.cpp.
References llvm::Offset, and Ptr.
|
static |
Try to find a partition of the aggregate type passed in for a given offset and size.
This recurses through the aggregate type and tries to compute a subtype based on the offset and size. When the offset and size span a sub-section of an array, it will even compute a new array type for that sub-section, and the same for structs.
Note that this routine is very strict and tries to find a partition of the type which produces the exact right offset and size. It is not forgiving when the size or offset cause either end of type-based partition to be off. Also, this is a best-effort routine. It is reasonable to give up and not return a type if necessary.
Definition at line 3916 of file SROA.cpp.
References assert(), DL, llvm::StructType::element_begin(), llvm::StructType::element_end(), llvm::StructType::get(), llvm::ArrayType::get(), llvm::Type::getContext(), llvm::StructLayout::getElementContainingOffset(), llvm::StructLayout::getElementOffset(), llvm::StructType::getElementType(), llvm::StructLayout::getSizeInBits(), llvm::StructLayout::getSizeInBytes(), getSizeInBytes(), getTypePartition(), llvm::StructType::isPacked(), llvm::details::FixedOrScalableQuantity< LeafTy, ValueTy >::isScalable(), llvm::Offset, Size, and stripAggregateTypeWrapping().
Referenced by getTypePartition().
INITIALIZE_PASS_BEGIN | ( | SROALegacyPass | , |
"sroa" | , | ||
"Scalar Replacement Of Aggregates" | , | ||
false | , | ||
false | |||
) |
|
static |
Definition at line 2336 of file SROA.cpp.
References assert(), llvm::dbgs(), DL, llvm::IntegerType::getBitWidth(), llvm::Value::getType(), LLVM_DEBUG, Name, and llvm::Offset.
|
static |
Definition at line 2391 of file SROA.cpp.
References assert(), llvm::dbgs(), llvm::ConstantVector::get(), llvm::Value::getType(), LLVM_DEBUG, Name, llvm::SmallVectorTemplateBase< T, bool >::push_back(), and llvm::SmallVectorImpl< T >::reserve().
|
static |
Test whether the given alloca partition's integer operations can be widened to promotable ones.
This is a quick test to check whether we can rewrite the integer loads and stores to a particular alloca into wider loads and stores and be able to promote the resulting alloca.
Definition at line 2270 of file SROA.cpp.
References canConvertValue(), DL, llvm::Type::getContext(), llvm::Type::getIntNTy(), isIntegerWideningViableForSlice(), llvm::IntegerType::MAX_INT_BITS, and P.
|
static |
Test whether a slice of an alloca is valid for integer widening.
This implements the necessary checking for the isIntegerWideningViable
test below on a single slice of the alloca.
Definition at line 2177 of file SROA.cpp.
References canConvertValue(), DL, llvm::Value::getType(), llvm::User::isDroppable(), llvm::Instruction::isLifetimeStartOrEnd(), llvm::LoadInst::isVolatile(), MI, and Size.
Referenced by isIntegerWideningViable().
|
static |
Definition at line 1573 of file SROA.cpp.
References assert(), DL, llvm::LoadInst::getAlign(), llvm::Value::getType(), llvm::isSafeToLoadUnconditionally(), llvm::LoadInst::isSimple(), and llvm::PreserveCFG.
PHI instructions that use an alloca and are subsequently loaded can be rewritten to load both input pointers in the pred blocks and then PHI the results, allowing the load of the alloca to be promoted.
From this: P2 = phi [i32* Alloca, i32* Other] V = load i32* P2 to: V1 = load i32* Alloca -> will be mem2reg'd ... V2 = load i32* Other ... V = phi [i32 V1, i32 V2]
We can do this to a select if its only uses are loads and if the operands to the select can be loaded unconditionally.
FIXME: This should be hoisted into a generic utility, likely in Transforms/Util/Local.h
Definition at line 1417 of file SROA.cpp.
References DL, llvm::LoadInst::getAlign(), llvm::Module::getDataLayout(), llvm::PHINode::getIncomingBlock(), llvm::PHINode::getIncomingValue(), llvm::Instruction::getModule(), llvm::PHINode::getNumIncomingValues(), llvm::Instruction::getNumSuccessors(), llvm::Instruction::getParent(), llvm::BasicBlock::getTerminator(), llvm::Value::getType(), Idx, llvm::isSafeToLoadUnconditionally(), llvm::LoadInst::isSimple(), llvm::Instruction::mayHaveSideEffects(), and llvm::Value::users().
|
static |
Test whether the given alloca partitioning and range of slices can be promoted to a vector.
This is a quick test to check whether we can rewrite a particular alloca partition (and its newly formed alloca) into a vector alloca with only whole-vector loads and stores such that it could be promoted to a vector SSA value. We only can ensure this for a limited set of operations, and we don't want to do the rewrites unless we are confident that the result will be promotable, so we have an early test here.
Definition at line 2018 of file SROA.cpp.
References assert(), checkVectorTypeForPromotion(), llvm::SmallVectorImpl< T >::clear(), DL, llvm::SmallVectorBase< Size_T >::empty(), llvm::erase_if(), llvm::VectorType::get(), llvm::Type::getIntNTy(), llvm::Value::getType(), llvm::SetVector< T, Vector, Set, N >::insert(), llvm::Type::isPointerTy(), llvm::VectorType::isValidElementType(), P, llvm::SmallVectorTemplateBase< T, bool >::push_back(), and llvm::sort().
|
static |
Test whether the given slice use can be promoted to a vector.
This function is called to test each entry in a partition which is slated for a single slice.
Definition at line 1911 of file SROA.cpp.
References assert(), canConvertValue(), DL, llvm::FixedVectorType::get(), llvm::Type::getIntNTy(), llvm::Value::getType(), llvm::User::isDroppable(), llvm::Type::isIntegerTy(), llvm::Instruction::isLifetimeStartOrEnd(), llvm::Type::isStructTy(), llvm::LoadInst::isVolatile(), MI, and P.
Referenced by checkVectorTypeForPromotion().
|
static |
Definition at line 1729 of file SROA.cpp.
References I, llvm::llvm_unreachable_internal(), and rewriteMemOpOfSelect().
|
static |
Definition at line 1673 of file SROA.cpp.
References llvm::PHINode::addIncoming(), assert(), llvm::PHINode::Create(), llvm::dbgs(), llvm::Value::getName(), llvm::BasicBlock::getTerminator(), I, LLVM_DEBUG, Ptr, llvm::SplitBlockAndInsertIfThen(), llvm::SplitBlockAndInsertIfThenElse(), llvm::successors(), llvm::CallingConv::Tail, and llvm::Value::takeName().
Referenced by rewriteMemOpOfSelect(), and rewriteSelectInstMemOps().
|
static |
Definition at line 1740 of file SROA.cpp.
References assert(), llvm::dbgs(), I, LLVM_DEBUG, llvm::make_early_inc_range(), rewriteMemOpOfSelect(), and speculateSelectInstLoads().
|
static |
Definition at line 1491 of file SROA.cpp.
References llvm::PHINode::addIncoming(), llvm::dbgs(), llvm::Instruction::eraseFromParent(), llvm::Instruction::getAAMetadata(), llvm::LoadInst::getAlign(), llvm::PHINode::getIncomingBlock(), llvm::PHINode::getIncomingValue(), llvm::Value::getName(), llvm::PHINode::getNumIncomingValues(), llvm::BasicBlock::getTerminator(), llvm::Value::getType(), Idx, LLVM_DEBUG, llvm::DenseMapBase< DerivedT, KeyT, ValueT, KeyInfoT, BucketT >::lookup(), llvm::Value::replaceAllUsesWith(), llvm::Value::use_empty(), and llvm::Instruction::user_back().
|
static |
Definition at line 1635 of file SROA.cpp.
References assert(), llvm::dbgs(), llvm::Instruction::getAAMetadata(), llvm::LoadInst::getAlign(), llvm::Value::getName(), llvm::Value::getType(), llvm::LoadInst::isSimple(), LLVM_DEBUG, llvm::Value::replaceAllUsesWith(), llvm::Instruction::setAAMetadata(), and llvm::LoadInst::setAlignment().
Referenced by rewriteSelectInstMemOps().
STATISTIC | ( | MaxPartitionsPerAlloca | , |
"Maximum number of partitions per alloca" | |||
) |
STATISTIC | ( | MaxUsesPerAllocaPartition | , |
"Maximum number of uses of a partition" | |||
) |
STATISTIC | ( | NumAllocaPartitions | , |
"Number of alloca partitions formed" | |||
) |
STATISTIC | ( | NumAllocaPartitionUses | , |
"Number of alloca partition uses rewritten" | |||
) |
STATISTIC | ( | NumAllocasAnalyzed | , |
"Number of allocas analyzed for replacement" | |||
) |
STATISTIC | ( | NumDeleted | , |
"Number of instructions deleted" | |||
) |
STATISTIC | ( | NumLoadsPredicated | , |
"Number of loads rewritten into predicated loads to allow promotion" | |||
) |
STATISTIC | ( | NumLoadsSpeculated | , |
"Number of loads speculated to allow promotion" | |||
) |
STATISTIC | ( | NumNewAllocas | , |
"Number of | new, | ||
smaller allocas introduced" | |||
) |
STATISTIC | ( | NumPromoted | , |
"Number of allocas promoted to SSA values" | |||
) |
STATISTIC | ( | NumStoresPredicated | , |
"Number of stores rewritten into predicated loads to allow promotion" | |||
) |
STATISTIC | ( | NumVectorized | , |
"Number of vectorized aggregates" | |||
) |
|
static |
Strip aggregate type wrapping.
This removes no-op aggregate types wrapping an underlying type. It will strip as many layers of types as it can without changing either the type size or the allocated size.
Definition at line 3878 of file SROA.cpp.
References DL, llvm::StructLayout::getElementContainingOffset(), llvm::Type::isSingleValueType(), and stripAggregateTypeWrapping().
Referenced by getTypePartition(), and stripAggregateTypeWrapping().
|
static |
The maximum number of alloca slices allowed when splitting.
|
static |
Disable running mem2reg during SROA in order to test or debug SROA.