From e10328737db3f0e6a1a668495e4971185705d61d Mon Sep 17 00:00:00 2001 From: Nadav Rotem Date: Wed, 10 Oct 2012 22:04:55 +0000 Subject: [PATCH] Add a new interface to allow IR-level passes to access codegen-specific information. llvm-svn: 165665 --- .../llvm/Analysis/ScalarEvolutionExpander.h | 4 +- llvm/include/llvm/CodeGen/CommandFlags.h | 228 +++++++++++++++++++++ llvm/include/llvm/InitializePasses.h | 1 + llvm/include/llvm/Target/TargetMachine.h | 6 + llvm/include/llvm/Target/TargetTransformImpl.h | 54 +++++ llvm/include/llvm/TargetTransformInfo.h | 126 ++++++++++++ llvm/include/llvm/Transforms/Scalar.h | 7 +- llvm/lib/Analysis/ScalarEvolutionExpander.cpp | 12 +- llvm/lib/CodeGen/Passes.cpp | 4 +- llvm/lib/Target/ARM/ARMTargetMachine.cpp | 6 +- llvm/lib/Target/ARM/ARMTargetMachine.h | 18 +- llvm/lib/Target/CMakeLists.txt | 1 + llvm/lib/Target/CellSPU/SPUTargetMachine.cpp | 3 +- llvm/lib/Target/CellSPU/SPUTargetMachine.h | 9 + llvm/lib/Target/Hexagon/HexagonTargetMachine.cpp | 5 +- llvm/lib/Target/Hexagon/HexagonTargetMachine.h | 11 + llvm/lib/Target/MBlaze/MBlazeTargetMachine.cpp | 2 +- llvm/lib/Target/MBlaze/MBlazeTargetMachine.h | 7 + llvm/lib/Target/MSP430/MSP430TargetMachine.cpp | 2 +- llvm/lib/Target/MSP430/MSP430TargetMachine.h | 10 +- llvm/lib/Target/Mips/MipsTargetMachine.cpp | 2 +- llvm/lib/Target/Mips/MipsTargetMachine.h | 3 + llvm/lib/Target/NVPTX/NVPTXTargetMachine.cpp | 3 +- llvm/lib/Target/NVPTX/NVPTXTargetMachine.h | 10 + llvm/lib/Target/PowerPC/PPCTargetMachine.cpp | 3 +- llvm/lib/Target/PowerPC/PPCTargetMachine.h | 9 + llvm/lib/Target/Sparc/SparcTargetMachine.cpp | 2 +- llvm/lib/Target/Sparc/SparcTargetMachine.h | 9 + llvm/lib/Target/Target.cpp | 1 + llvm/lib/Target/TargetTransformImpl.cpp | 43 ++++ llvm/lib/Target/X86/X86TargetMachine.cpp | 6 +- llvm/lib/Target/X86/X86TargetMachine.h | 17 ++ llvm/lib/Target/XCore/XCoreTargetMachine.cpp | 2 +- llvm/lib/Target/XCore/XCoreTargetMachine.h | 9 + llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp | 134 ++++++------ llvm/lib/Transforms/Utils/LowerInvoke.cpp | 32 +-- llvm/lib/VMCore/CMakeLists.txt | 1 + llvm/lib/VMCore/TargetTransformInfo.cpp | 27 +++ llvm/tools/llc/llc.cpp | 215 +------------------ llvm/tools/opt/CMakeLists.txt | 2 +- llvm/tools/opt/LLVMBuild.txt | 2 +- llvm/tools/opt/Makefile | 2 +- llvm/tools/opt/opt.cpp | 78 +++++++ 43 files changed, 811 insertions(+), 317 deletions(-) create mode 100644 llvm/include/llvm/CodeGen/CommandFlags.h create mode 100644 llvm/include/llvm/Target/TargetTransformImpl.h create mode 100644 llvm/include/llvm/TargetTransformInfo.h create mode 100644 llvm/lib/Target/TargetTransformImpl.cpp create mode 100644 llvm/lib/VMCore/TargetTransformInfo.cpp diff --git a/llvm/include/llvm/Analysis/ScalarEvolutionExpander.h b/llvm/include/llvm/Analysis/ScalarEvolutionExpander.h index 3f8f149..3ab9c82 100644 --- a/llvm/include/llvm/Analysis/ScalarEvolutionExpander.h +++ b/llvm/include/llvm/Analysis/ScalarEvolutionExpander.h @@ -22,7 +22,7 @@ #include namespace llvm { - class TargetLowering; + class ScalarTargetTransformInfo; /// Return true if the given expression is safe to expand in the sense that /// all materialized values are safe to speculate. @@ -129,7 +129,7 @@ namespace llvm { /// representative. Return the number of phis eliminated. unsigned replaceCongruentIVs(Loop *L, const DominatorTree *DT, SmallVectorImpl &DeadInsts, - const TargetLowering *TLI = NULL); + const ScalarTargetTransformInfo *STTI = NULL); /// expandCodeFor - Insert code to directly compute the specified SCEV /// expression into the program. The inserted code is inserted into the diff --git a/llvm/include/llvm/CodeGen/CommandFlags.h b/llvm/include/llvm/CodeGen/CommandFlags.h new file mode 100644 index 0000000..90ee234 --- /dev/null +++ b/llvm/include/llvm/CodeGen/CommandFlags.h @@ -0,0 +1,228 @@ +//===-- CommandFlags.h - Register Coalescing Interface ----------*- C++ -*-===// +// +// The LLVM Compiler Infrastructure +// +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// +// +// This file contains codegen-specific flags that are shared between different +// command line tools. The tools "llc" and "opt" both use this file to prevent +// flag duplication. +// +//===----------------------------------------------------------------------===// + +#ifndef LLVM_CODEGEN_COMMAND_LINE_FLAGS_H +#define LLVM_CODEGEN_COMMAND_LINE_FLAGS_H + +#include "llvm/Support/CommandLine.h" +#include "llvm/Support/CodeGen.h" +#include "llvm/Target/TargetMachine.h" + +#include +using namespace llvm; + +cl::opt +MArch("march", cl::desc("Architecture to generate code for (see --version)")); + +cl::opt +MCPU("mcpu", + cl::desc("Target a specific cpu type (-mcpu=help for details)"), + cl::value_desc("cpu-name"), + cl::init("")); + +cl::list +MAttrs("mattr", + cl::CommaSeparated, + cl::desc("Target specific attributes (-mattr=help for details)"), + cl::value_desc("a1,+a2,-a3,...")); + +cl::opt +RelocModel("relocation-model", + cl::desc("Choose relocation model"), + cl::init(Reloc::Default), + cl::values( + clEnumValN(Reloc::Default, "default", + "Target default relocation model"), + clEnumValN(Reloc::Static, "static", + "Non-relocatable code"), + clEnumValN(Reloc::PIC_, "pic", + "Fully relocatable, position independent code"), + clEnumValN(Reloc::DynamicNoPIC, "dynamic-no-pic", + "Relocatable external references, non-relocatable code"), + clEnumValEnd)); + +cl::opt +CMModel("code-model", + cl::desc("Choose code model"), + cl::init(CodeModel::Default), + cl::values(clEnumValN(CodeModel::Default, "default", + "Target default code model"), + clEnumValN(CodeModel::Small, "small", + "Small code model"), + clEnumValN(CodeModel::Kernel, "kernel", + "Kernel code model"), + clEnumValN(CodeModel::Medium, "medium", + "Medium code model"), + clEnumValN(CodeModel::Large, "large", + "Large code model"), + clEnumValEnd)); + +cl::opt +RelaxAll("mc-relax-all", + cl::desc("When used with filetype=obj, " + "relax all fixups in the emitted object file")); + +cl::opt +FileType("filetype", cl::init(TargetMachine::CGFT_AssemblyFile), + cl::desc("Choose a file type (not all types are supported by all targets):"), + cl::values( + clEnumValN(TargetMachine::CGFT_AssemblyFile, "asm", + "Emit an assembly ('.s') file"), + clEnumValN(TargetMachine::CGFT_ObjectFile, "obj", + "Emit a native object ('.o') file"), + clEnumValN(TargetMachine::CGFT_Null, "null", + "Emit nothing, for performance testing"), + clEnumValEnd)); + +cl::opt DisableDotLoc("disable-dot-loc", cl::Hidden, + cl::desc("Do not use .loc entries")); + +cl::opt DisableCFI("disable-cfi", cl::Hidden, + cl::desc("Do not use .cfi_* directives")); + +cl::opt EnableDwarfDirectory("enable-dwarf-directory", cl::Hidden, + cl::desc("Use .file directives with an explicit directory.")); + +cl::opt +DisableRedZone("disable-red-zone", + cl::desc("Do not emit code that uses the red zone."), + cl::init(false)); + +cl::opt +EnableFPMAD("enable-fp-mad", + cl::desc("Enable less precise MAD instructions to be generated"), + cl::init(false)); + +cl::opt +DisableFPElim("disable-fp-elim", + cl::desc("Disable frame pointer elimination optimization"), + cl::init(false)); + +cl::opt +DisableFPElimNonLeaf("disable-non-leaf-fp-elim", + cl::desc("Disable frame pointer elimination optimization for non-leaf funcs"), + cl::init(false)); + +cl::opt +EnableUnsafeFPMath("enable-unsafe-fp-math", + cl::desc("Enable optimizations that may decrease FP precision"), + cl::init(false)); + +cl::opt +EnableNoInfsFPMath("enable-no-infs-fp-math", + cl::desc("Enable FP math optimizations that assume no +-Infs"), + cl::init(false)); + +cl::opt +EnableNoNaNsFPMath("enable-no-nans-fp-math", + cl::desc("Enable FP math optimizations that assume no NaNs"), + cl::init(false)); + +cl::opt +EnableHonorSignDependentRoundingFPMath("enable-sign-dependent-rounding-fp-math", + cl::Hidden, + cl::desc("Force codegen to assume rounding mode can change dynamically"), + cl::init(false)); + +cl::opt +GenerateSoftFloatCalls("soft-float", + cl::desc("Generate software floating point library calls"), + cl::init(false)); + +cl::opt +FloatABIForCalls("float-abi", + cl::desc("Choose float ABI type"), + cl::init(FloatABI::Default), + cl::values( + clEnumValN(FloatABI::Default, "default", + "Target default float ABI type"), + clEnumValN(FloatABI::Soft, "soft", + "Soft float ABI (implied by -soft-float)"), + clEnumValN(FloatABI::Hard, "hard", + "Hard float ABI (uses FP registers)"), + clEnumValEnd)); + +cl::opt +FuseFPOps("fp-contract", + cl::desc("Enable aggresive formation of fused FP ops"), + cl::init(FPOpFusion::Standard), + cl::values( + clEnumValN(FPOpFusion::Fast, "fast", + "Fuse FP ops whenever profitable"), + clEnumValN(FPOpFusion::Standard, "on", + "Only fuse 'blessed' FP ops."), + clEnumValN(FPOpFusion::Strict, "off", + "Only fuse FP ops when the result won't be effected."), + clEnumValEnd)); + +cl::opt +DontPlaceZerosInBSS("nozero-initialized-in-bss", + cl::desc("Don't place zero-initialized symbols into bss section"), + cl::init(false)); + +cl::opt +EnableGuaranteedTailCallOpt("tailcallopt", + cl::desc("Turn fastcc calls into tail calls by (potentially) changing ABI."), + cl::init(false)); + +cl::opt +DisableTailCalls("disable-tail-calls", + cl::desc("Never emit tail calls"), + cl::init(false)); + +cl::opt +OverrideStackAlignment("stack-alignment", + cl::desc("Override default stack alignment"), + cl::init(0)); + +cl::opt +EnableRealignStack("realign-stack", + cl::desc("Realign stack if needed"), + cl::init(true)); + +cl::opt +TrapFuncName("trap-func", cl::Hidden, + cl::desc("Emit a call to trap function rather than a trap instruction"), + cl::init("")); + +cl::opt +EnablePIE("enable-pie", + cl::desc("Assume the creation of a position independent executable."), + cl::init(false)); + +cl::opt +SegmentedStacks("segmented-stacks", + cl::desc("Use segmented stacks if possible."), + cl::init(false)); + +cl::opt +UseInitArray("use-init-array", + cl::desc("Use .init_array instead of .ctors."), + cl::init(false)); + +cl::opt StopAfter("stop-after", + cl::desc("Stop compilation after a specific pass"), + cl::value_desc("pass-name"), + cl::init("")); +cl::opt StartAfter("start-after", + cl::desc("Resume compilation after a specific pass"), + cl::value_desc("pass-name"), + cl::init("")); + +cl::opt +SSPBufferSize("stack-protector-buffer-size", cl::init(8), + cl::desc("Lower bound for a buffer to be considered for " + "stack protection")); +#endif diff --git a/llvm/include/llvm/InitializePasses.h b/llvm/include/llvm/InitializePasses.h index 409246c..067d8da 100644 --- a/llvm/include/llvm/InitializePasses.h +++ b/llvm/include/llvm/InitializePasses.h @@ -247,6 +247,7 @@ void initializeTailCallElimPass(PassRegistry&); void initializeTailDuplicatePassPass(PassRegistry&); void initializeTargetPassConfigPass(PassRegistry&); void initializeDataLayoutPass(PassRegistry&); +void initializeTargetTransformInfoPass(PassRegistry&); void initializeTargetLibraryInfoPass(PassRegistry&); void initializeTwoAddressInstructionPassPass(PassRegistry&); void initializeTypeBasedAliasAnalysisPass(PassRegistry&); diff --git a/llvm/include/llvm/Target/TargetMachine.h b/llvm/include/llvm/Target/TargetMachine.h index 988916f..18e589e 100644 --- a/llvm/include/llvm/Target/TargetMachine.h +++ b/llvm/include/llvm/Target/TargetMachine.h @@ -17,6 +17,8 @@ #include "llvm/Pass.h" #include "llvm/Support/CodeGen.h" #include "llvm/Target/TargetOptions.h" +#include "llvm/TargetTransformInfo.h" +#include "llvm/Target/TargetTransformImpl.h" #include "llvm/ADT/StringRef.h" #include #include @@ -107,6 +109,10 @@ public: virtual const TargetLowering *getTargetLowering() const { return 0; } virtual const TargetSelectionDAGInfo *getSelectionDAGInfo() const{ return 0; } virtual const DataLayout *getDataLayout() const { return 0; } + virtual const ScalarTargetTransformInfo* + getScalarTargetTransformInfo() const { return 0; } + virtual const VectorTargetTransformInfo* + getVectorTargetTransformInfo() const { return 0; } /// getMCAsmInfo - Return target specific asm information. /// diff --git a/llvm/include/llvm/Target/TargetTransformImpl.h b/llvm/include/llvm/Target/TargetTransformImpl.h new file mode 100644 index 0000000..e240cdf --- /dev/null +++ b/llvm/include/llvm/Target/TargetTransformImpl.h @@ -0,0 +1,54 @@ +//=- llvm/Target/TargetTransformImpl.h - Target Loop Trans Info----*- C++ -*-=// +// +// The LLVM Compiler Infrastructure +// +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// +// +// This file contains the target-specific implementations of the +// TargetTransform interfaces. +// +//===----------------------------------------------------------------------===// + +#ifndef LLVM_TARGET_TARGET_TRANSFORMATION_IMPL_H +#define LLVM_TARGET_TARGET_TRANSFORMATION_IMPL_H + +#include "llvm/TargetTransformInfo.h" + +namespace llvm { + +class TargetLowering; + +/// ScalarTargetTransformInfo - This is a default implementation for the +/// ScalarTargetTransformInfo interface. Different targets can implement +/// this interface differently. +class ScalarTargetTransformImpl : public ScalarTargetTransformInfo { +private: + const TargetLowering *TLI; + +public: + /// Ctor + explicit ScalarTargetTransformImpl(const TargetLowering *TL): TLI(TL) {} + + virtual bool isLegalAddImmediate(int64_t imm) const; + + virtual bool isLegalICmpImmediate(int64_t imm) const; + + virtual bool isLegalAddressingMode(const AddrMode &AM, Type *Ty) const; + + virtual bool isTruncateFree(Type *Ty1, Type *Ty2) const; + + virtual bool isTypeLegal(Type *Ty) const; + + virtual unsigned getJumpBufAlignment() const; + + virtual unsigned getJumpBufSize() const; +}; + +class VectorTargetTransformImpl : public VectorTargetTransformInfo { }; + +} // end llvm namespace + +#endif diff --git a/llvm/include/llvm/TargetTransformInfo.h b/llvm/include/llvm/TargetTransformInfo.h new file mode 100644 index 0000000..e1be3ba --- /dev/null +++ b/llvm/include/llvm/TargetTransformInfo.h @@ -0,0 +1,126 @@ +//===- llvm/Transforms/TargetTransformInfo.h --------------------*- C++ -*-===// +// +// The LLVM Compiler Infrastructure +// +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// +// +// This pass exposes codegen information to IR-level passes. Every +// transformation that uses codegen information is broken into three parts: +// 1. The IR-level analysis pass. +// 2. The IR-level transformation interface which provides the needed +// information. +// 3. Codegen-level implementation which uses target-specific hooks. +// +// This file defines #2, which is the interface that IR-level transformations +// use for querying the codegen. +// +//===----------------------------------------------------------------------===// + +#ifndef LLVM_TRANSFORMS_TARGET_TRANSFORM_INTERFACE +#define LLVM_TRANSFORMS_TARGET_TRANSFORM_INTERFACE + +#include "llvm/Pass.h" +#include "llvm/AddressingMode.h" +#include "llvm/Support/DataTypes.h" +#include "llvm/Type.h" + +namespace llvm { + +class ScalarTargetTransformInfo; +class VectorTargetTransformInfo; + +/// TargetTransformInfo - This pass provides access to the codegen +/// interfaces that are needed for IR-level transformations. +class TargetTransformInfo : public ImmutablePass { +private: + const ScalarTargetTransformInfo *STTI; + const VectorTargetTransformInfo *VTTI; +public: + /// Default ctor. + /// + /// @note This has to exist, because this is a pass, but it should never be + /// used. + TargetTransformInfo(); + + explicit TargetTransformInfo(const ScalarTargetTransformInfo* S, + const VectorTargetTransformInfo *V) + : ImmutablePass(ID), STTI(S), VTTI(V) { + initializeTargetTransformInfoPass(*PassRegistry::getPassRegistry()); + } + + TargetTransformInfo(const TargetTransformInfo &T) : + ImmutablePass(ID), STTI(T.STTI), VTTI(T.VTTI) { } + + const ScalarTargetTransformInfo* getScalarTargetTransformInfo() { + return STTI; + } + const VectorTargetTransformInfo* getVectorTargetTransformInfo() { + return VTTI; + } + + /// Pass identification, replacement for typeid. + static char ID; +}; + +// ---------------------------------------------------------------------------// +// The classes below are inherited and implemented by target-specific classes +// in the codegen. +// ---------------------------------------------------------------------------// + +/// ScalarTargetTransformInfo - This interface is used by IR-level passes +/// that need target-dependent information for generic scalar transformations. +/// LSR, and LowerInvoke use this interface. +class ScalarTargetTransformInfo { +public: + /// isLegalAddImmediate - Return true if the specified immediate is legal + /// add immediate, that is the target has add instructions which can add + /// a register with the immediate without having to materialize the + /// immediate into a register. + virtual bool isLegalAddImmediate(int64_t) const { + return false; + } + /// isLegalICmpImmediate - Return true if the specified immediate is legal + /// icmp immediate, that is the target has icmp instructions which can compare + /// a register against the immediate without having to materialize the + /// immediate into a register. + virtual bool isLegalICmpImmediate(int64_t) const { + return false; + } + /// isLegalAddressingMode - Return true if the addressing mode represented by + /// AM is legal for this target, for a load/store of the specified type. + /// The type may be VoidTy, in which case only return true if the addressing + /// mode is legal for a load/store of any legal type. + /// TODO: Handle pre/postinc as well. + virtual bool isLegalAddressingMode(const AddrMode &AM, Type *Ty) const { + return false; + } + /// isTruncateFree - Return true if it's free to truncate a value of + /// type Ty1 to type Ty2. e.g. On x86 it's free to truncate a i32 value in + /// register EAX to i16 by referencing its sub-register AX. + virtual bool isTruncateFree(Type * /*Ty1*/, Type * /*Ty2*/) const { + return false; + } + /// Is this type legal. + virtual bool isTypeLegal(Type *Ty) const { + return false; + } + /// getJumpBufAlignment - returns the target's jmp_buf alignment in bytes + virtual unsigned getJumpBufAlignment() const { + return 0; + } + /// getJumpBufSize - returns the target's jmp_buf size in bytes. + virtual unsigned getJumpBufSize() const { + return 0; + } +}; + +class VectorTargetTransformInfo { + // TODO: define an interface for VectorTargetTransformInfo. +}; + +} // End llvm namespace + +#endif diff --git a/llvm/include/llvm/Transforms/Scalar.h b/llvm/include/llvm/Transforms/Scalar.h index a5d8eed..3b665bf 100644 --- a/llvm/include/llvm/Transforms/Scalar.h +++ b/llvm/include/llvm/Transforms/Scalar.h @@ -119,7 +119,7 @@ Pass *createLICMPass(); // optional parameter used to consult the target machine whether certain // transformations are profitable. // -Pass *createLoopStrengthReducePass(const TargetLowering *TLI = 0); +Pass *createLoopStrengthReducePass(); Pass *createGlobalMergePass(const TargetLowering *TLI = 0); @@ -249,9 +249,8 @@ extern char &LowerSwitchID; // purpose "my LLVM-to-LLVM pass doesn't support the invoke instruction yet" // lowering pass. // -FunctionPass *createLowerInvokePass(const TargetLowering *TLI = 0); -FunctionPass *createLowerInvokePass(const TargetLowering *TLI, - bool useExpensiveEHSupport); +FunctionPass *createLowerInvokePass(); +FunctionPass *createLowerInvokePass(bool useExpensiveEHSupport); extern char &LowerInvokePassID; //===----------------------------------------------------------------------===// diff --git a/llvm/lib/Analysis/ScalarEvolutionExpander.cpp b/llvm/lib/Analysis/ScalarEvolutionExpander.cpp index 5e05f4c..5c2a49e 100644 --- a/llvm/lib/Analysis/ScalarEvolutionExpander.cpp +++ b/llvm/lib/Analysis/ScalarEvolutionExpander.cpp @@ -19,8 +19,8 @@ #include "llvm/LLVMContext.h" #include "llvm/Support/Debug.h" #include "llvm/DataLayout.h" -#include "llvm/Target/TargetLowering.h" #include "llvm/ADT/STLExtras.h" +#include "llvm/TargetTransformInfo.h" using namespace llvm; @@ -1599,15 +1599,15 @@ static bool width_descending(Value *lhs, Value *rhs) { /// This does not depend on any SCEVExpander state but should be used in /// the same context that SCEVExpander is used. unsigned SCEVExpander::replaceCongruentIVs(Loop *L, const DominatorTree *DT, - SmallVectorImpl &DeadInsts, - const TargetLowering *TLI) { + SmallVectorImpl &DeadInsts, + const ScalarTargetTransformInfo *STTI) { // Find integer phis in order of increasing width. SmallVector Phis; for (BasicBlock::iterator I = L->getHeader()->begin(); PHINode *Phi = dyn_cast(I); ++I) { Phis.push_back(Phi); } - if (TLI) + if (STTI) std::sort(Phis.begin(), Phis.end(), width_descending); unsigned NumElim = 0; @@ -1624,8 +1624,8 @@ unsigned SCEVExpander::replaceCongruentIVs(Loop *L, const DominatorTree *DT, PHINode *&OrigPhiRef = ExprToIVMap[SE.getSCEV(Phi)]; if (!OrigPhiRef) { OrigPhiRef = Phi; - if (Phi->getType()->isIntegerTy() && TLI - && TLI->isTruncateFree(Phi->getType(), Phis.back()->getType())) { + if (Phi->getType()->isIntegerTy() && STTI && + STTI->isTruncateFree(Phi->getType(), Phis.back()->getType())) { // This phi can be freely truncated to the narrowest phi type. Map the // truncated expression to it so it will be reused for narrow types. const SCEV *TruncExpr = diff --git a/llvm/lib/CodeGen/Passes.cpp b/llvm/lib/CodeGen/Passes.cpp index 4ea21d4..abd62ef 100644 --- a/llvm/lib/CodeGen/Passes.cpp +++ b/llvm/lib/CodeGen/Passes.cpp @@ -359,7 +359,7 @@ void TargetPassConfig::addIRPasses() { // Run loop strength reduction before anything else. if (getOptLevel() != CodeGenOpt::None && !DisableLSR) { - addPass(createLoopStrengthReducePass(getTargetLowering())); + addPass(createLoopStrengthReducePass()); if (PrintLSR) addPass(createPrintFunctionPass("\n\n*** Code after LSR ***\n", &dbgs())); } @@ -389,7 +389,7 @@ void TargetPassConfig::addPassesToHandleExceptions() { addPass(createDwarfEHPass(TM)); break; case ExceptionHandling::None: - addPass(createLowerInvokePass(TM->getTargetLowering())); + addPass(createLowerInvokePass()); // The lower invoke pass may create unreachable code. Remove it. addPass(createUnreachableBlockEliminationPass()); diff --git a/llvm/lib/Target/ARM/ARMTargetMachine.cpp b/llvm/lib/Target/ARM/ARMTargetMachine.cpp index 6fdf873..c51ae24 100644 --- a/llvm/lib/Target/ARM/ARMTargetMachine.cpp +++ b/llvm/lib/Target/ARM/ARMTargetMachine.cpp @@ -71,7 +71,8 @@ ARMTargetMachine::ARMTargetMachine(const Target &T, StringRef TT, ELFWriterInfo(*this), TLInfo(*this), TSInfo(*this), - FrameLowering(Subtarget) { + FrameLowering(Subtarget), + STTI(&TLInfo) { if (!Subtarget.hasARMOps()) report_fatal_error("CPU: '" + Subtarget.getCPUString() + "' does not " "support ARM mode execution!"); @@ -104,7 +105,8 @@ ThumbTargetMachine::ThumbTargetMachine(const Target &T, StringRef TT, TSInfo(*this), FrameLowering(Subtarget.hasThumb2() ? new ARMFrameLowering(Subtarget) - : (ARMFrameLowering*)new Thumb1FrameLowering(Subtarget)) { + : (ARMFrameLowering*)new Thumb1FrameLowering(Subtarget)), + STTI(&TLInfo){ } namespace { diff --git a/llvm/lib/Target/ARM/ARMTargetMachine.h b/llvm/lib/Target/ARM/ARMTargetMachine.h index f91e5bb..7a65a7f 100644 --- a/llvm/lib/Target/ARM/ARMTargetMachine.h +++ b/llvm/lib/Target/ARM/ARMTargetMachine.h @@ -25,6 +25,7 @@ #include "Thumb1FrameLowering.h" #include "Thumb2InstrInfo.h" #include "llvm/Target/TargetMachine.h" +#include "llvm/Target/TargetTransformImpl.h" #include "llvm/DataLayout.h" #include "llvm/MC/MCStreamer.h" #include "llvm/ADT/OwningPtr.h" @@ -67,6 +68,8 @@ class ARMTargetMachine : public ARMBaseTargetMachine { ARMTargetLowering TLInfo; ARMSelectionDAGInfo TSInfo; ARMFrameLowering FrameLowering; + ScalarTargetTransformImpl STTI; + VectorTargetTransformImpl VTTI; public: ARMTargetMachine(const Target &T, StringRef TT, StringRef CPU, StringRef FS, @@ -88,7 +91,12 @@ class ARMTargetMachine : public ARMBaseTargetMachine { virtual const ARMFrameLowering *getFrameLowering() const { return &FrameLowering; } - + virtual const ScalarTargetTransformInfo *getScalarTargetTransformInfo()const { + return &STTI; + } + virtual const VectorTargetTransformInfo *getVectorTargetTransformInfo()const { + return &VTTI; + } virtual const ARMInstrInfo *getInstrInfo() const { return &InstrInfo; } virtual const DataLayout *getDataLayout() const { return &DL; } virtual const ARMELFWriterInfo *getELFWriterInfo() const { @@ -110,6 +118,8 @@ class ThumbTargetMachine : public ARMBaseTargetMachine { ARMSelectionDAGInfo TSInfo; // Either Thumb1FrameLowering or ARMFrameLowering. OwningPtr FrameLowering; + ScalarTargetTransformImpl STTI; + VectorTargetTransformImpl VTTI; public: ThumbTargetMachine(const Target &T, StringRef TT, StringRef CPU, StringRef FS, @@ -138,6 +148,12 @@ public: virtual const ARMFrameLowering *getFrameLowering() const { return FrameLowering.get(); } + virtual const ScalarTargetTransformInfo *getScalarTargetTransformInfo()const { + return &STTI; + } + virtual const VectorTargetTransformInfo *getVectorTargetTransformInfo()const { + return &VTTI; + } virtual const DataLayout *getDataLayout() const { return &DL; } virtual const ARMELFWriterInfo *getELFWriterInfo() const { return Subtarget.isTargetELF() ? &ELFWriterInfo : 0; diff --git a/llvm/lib/Target/CMakeLists.txt b/llvm/lib/Target/CMakeLists.txt index 096ef00..48df199 100644 --- a/llvm/lib/Target/CMakeLists.txt +++ b/llvm/lib/Target/CMakeLists.txt @@ -11,6 +11,7 @@ add_llvm_library(LLVMTarget TargetMachineC.cpp TargetRegisterInfo.cpp TargetSubtargetInfo.cpp + TargetTransformImpl.cpp ) foreach(t ${LLVM_TARGETS_TO_BUILD}) diff --git a/llvm/lib/Target/CellSPU/SPUTargetMachine.cpp b/llvm/lib/Target/CellSPU/SPUTargetMachine.cpp index a37ad7f..e92ad01 100644 --- a/llvm/lib/Target/CellSPU/SPUTargetMachine.cpp +++ b/llvm/lib/Target/CellSPU/SPUTargetMachine.cpp @@ -43,7 +43,8 @@ SPUTargetMachine::SPUTargetMachine(const Target &T, StringRef TT, FrameLowering(Subtarget), TLInfo(*this), TSInfo(*this), - InstrItins(Subtarget.getInstrItineraryData()) { + InstrItins(Subtarget.getInstrItineraryData()), + STTI(&TLInfo){ } //===----------------------------------------------------------------------===// diff --git a/llvm/lib/Target/CellSPU/SPUTargetMachine.h b/llvm/lib/Target/CellSPU/SPUTargetMachine.h index 58699a3..7f53ea6f 100644 --- a/llvm/lib/Target/CellSPU/SPUTargetMachine.h +++ b/llvm/lib/Target/CellSPU/SPUTargetMachine.h @@ -20,6 +20,7 @@ #include "SPUSelectionDAGInfo.h" #include "SPUFrameLowering.h" #include "llvm/Target/TargetMachine.h" +#include "llvm/Target/TargetTransformImpl.h" #include "llvm/DataLayout.h" namespace llvm { @@ -34,6 +35,8 @@ class SPUTargetMachine : public LLVMTargetMachine { SPUTargetLowering TLInfo; SPUSelectionDAGInfo TSInfo; InstrItineraryData InstrItins; + ScalarTargetTransformImpl STTI; + VectorTargetTransformImpl VTTI; public: SPUTargetMachine(const Target &T, StringRef TT, StringRef CPU, StringRef FS, const TargetOptions &Options, @@ -77,6 +80,12 @@ public: virtual const InstrItineraryData *getInstrItineraryData() const { return &InstrItins; } + virtual const ScalarTargetTransformInfo *getScalarTargetTransformInfo()const { + return &STTI; + } + virtual const VectorTargetTransformInfo *getVectorTargetTransformInfo()const { + return &VTTI; + } // Pass Pipeline Configuration virtual TargetPassConfig *createPassConfig(PassManagerBase &PM); diff --git a/llvm/lib/Target/Hexagon/HexagonTargetMachine.cpp b/llvm/lib/Target/Hexagon/HexagonTargetMachine.cpp index d198a3f..353542a 100644 --- a/llvm/lib/Target/Hexagon/HexagonTargetMachine.cpp +++ b/llvm/lib/Target/Hexagon/HexagonTargetMachine.cpp @@ -74,7 +74,8 @@ HexagonTargetMachine::HexagonTargetMachine(const Target &T, StringRef TT, Subtarget(TT, CPU, FS), InstrInfo(Subtarget), TLInfo(*this), TSInfo(*this), FrameLowering(Subtarget), - InstrItins(&Subtarget.getInstrItineraryData()) { + InstrItins(&Subtarget.getInstrItineraryData()), + STTI(&TLInfo) { setMCUseCFI(false); } @@ -87,7 +88,7 @@ bool HexagonTargetMachine::addPassesForOptimizations(PassManagerBase &PM) { PM.add(createDeadCodeEliminationPass()); PM.add(createConstantPropagationPass()); PM.add(createLoopUnrollPass()); - PM.add(createLoopStrengthReducePass(getTargetLowering())); + PM.add(createLoopStrengthReducePass()); return true; } diff --git a/llvm/lib/Target/Hexagon/HexagonTargetMachine.h b/llvm/lib/Target/Hexagon/HexagonTargetMachine.h index ade5b3e..7a4215c 100644 --- a/llvm/lib/Target/Hexagon/HexagonTargetMachine.h +++ b/llvm/lib/Target/Hexagon/HexagonTargetMachine.h @@ -21,6 +21,7 @@ #include "HexagonFrameLowering.h" #include "llvm/Target/TargetMachine.h" #include "llvm/DataLayout.h" +#include "llvm/Target/TargetTransformImpl.h" namespace llvm { @@ -34,6 +35,8 @@ class HexagonTargetMachine : public LLVMTargetMachine { HexagonSelectionDAGInfo TSInfo; HexagonFrameLowering FrameLowering; const InstrItineraryData* InstrItins; + ScalarTargetTransformImpl STTI; + VectorTargetTransformImpl VTTI; public: HexagonTargetMachine(const Target &T, StringRef TT,StringRef CPU, @@ -68,6 +71,14 @@ public: return &TSInfo; } + virtual const ScalarTargetTransformInfo *getScalarTargetTransformInfo()const { + return &STTI; + } + + virtual const VectorTargetTransformInfo *getVectorTargetTransformInfo()const { + return &VTTI; + } + virtual const DataLayout *getDataLayout() const { return &DL; } static unsigned getModuleMatchQuality(const Module &M); diff --git a/llvm/lib/Target/MBlaze/MBlazeTargetMachine.cpp b/llvm/lib/Target/MBlaze/MBlazeTargetMachine.cpp index 1f2cf6d..cb5f460 100644 --- a/llvm/lib/Target/MBlaze/MBlazeTargetMachine.cpp +++ b/llvm/lib/Target/MBlaze/MBlazeTargetMachine.cpp @@ -42,7 +42,7 @@ MBlazeTargetMachine(const Target &T, StringRef TT, InstrInfo(*this), FrameLowering(Subtarget), TLInfo(*this), TSInfo(*this), ELFWriterInfo(*this), - InstrItins(Subtarget.getInstrItineraryData()) { + InstrItins(Subtarget.getInstrItineraryData()), STTI(&TLInfo) { } namespace { diff --git a/llvm/lib/Target/MBlaze/MBlazeTargetMachine.h b/llvm/lib/Target/MBlaze/MBlazeTargetMachine.h index d949e54..34648b9 100644 --- a/llvm/lib/Target/MBlaze/MBlazeTargetMachine.h +++ b/llvm/lib/Target/MBlaze/MBlazeTargetMachine.h @@ -25,6 +25,7 @@ #include "llvm/Target/TargetMachine.h" #include "llvm/DataLayout.h" #include "llvm/Target/TargetFrameLowering.h" +#include "llvm/Target/TargetTransformImpl.h" namespace llvm { class formatted_raw_ostream; @@ -39,6 +40,8 @@ namespace llvm { MBlazeIntrinsicInfo IntrinsicInfo; MBlazeELFWriterInfo ELFWriterInfo; InstrItineraryData InstrItins; + ScalarTargetTransformImpl STTI; + VectorTargetTransformImpl VTTI; public: MBlazeTargetMachine(const Target &T, StringRef TT, @@ -77,6 +80,10 @@ namespace llvm { virtual const MBlazeELFWriterInfo *getELFWriterInfo() const { return &ELFWriterInfo; } + virtual const ScalarTargetTransformInfo *getScalarTargetTransformInfo()const + { return &STTI; } + virtual const VectorTargetTransformInfo *getVectorTargetTransformInfo()const + { return &VTTI; } // Pass Pipeline Configuration virtual TargetPassConfig *createPassConfig(PassManagerBase &PM); diff --git a/llvm/lib/Target/MSP430/MSP430TargetMachine.cpp b/llvm/lib/Target/MSP430/MSP430TargetMachine.cpp index da5899b..29ea681 100644 --- a/llvm/lib/Target/MSP430/MSP430TargetMachine.cpp +++ b/llvm/lib/Target/MSP430/MSP430TargetMachine.cpp @@ -36,7 +36,7 @@ MSP430TargetMachine::MSP430TargetMachine(const Target &T, // FIXME: Check DataLayout string. DL("e-p:16:16:16-i8:8:8-i16:16:16-i32:16:32-n8:16"), InstrInfo(*this), TLInfo(*this), TSInfo(*this), - FrameLowering(Subtarget) { } + FrameLowering(Subtarget), STTI(&TLInfo) { } namespace { /// MSP430 Code Generator Pass Configuration Options. diff --git a/llvm/lib/Target/MSP430/MSP430TargetMachine.h b/llvm/lib/Target/MSP430/MSP430TargetMachine.h index ba3cef1..186172e 100644 --- a/llvm/lib/Target/MSP430/MSP430TargetMachine.h +++ b/llvm/lib/Target/MSP430/MSP430TargetMachine.h @@ -24,6 +24,7 @@ #include "llvm/DataLayout.h" #include "llvm/Target/TargetFrameLowering.h" #include "llvm/Target/TargetMachine.h" +#include "llvm/Target/TargetTransformImpl.h" namespace llvm { @@ -36,6 +37,8 @@ class MSP430TargetMachine : public LLVMTargetMachine { MSP430TargetLowering TLInfo; MSP430SelectionDAGInfo TSInfo; MSP430FrameLowering FrameLowering; + ScalarTargetTransformImpl STTI; + VectorTargetTransformImpl VTTI; public: MSP430TargetMachine(const Target &T, StringRef TT, @@ -61,7 +64,12 @@ public: virtual const MSP430SelectionDAGInfo* getSelectionDAGInfo() const { return &TSInfo; } - + virtual const ScalarTargetTransformInfo *getScalarTargetTransformInfo()const { + return &STTI; + } + virtual const VectorTargetTransformInfo *getVectorTargetTransformInfo()const { + return &VTTI; + } virtual TargetPassConfig *createPassConfig(PassManagerBase &PM); }; // MSP430TargetMachine. diff --git a/llvm/lib/Target/Mips/MipsTargetMachine.cpp b/llvm/lib/Target/Mips/MipsTargetMachine.cpp index 9c196dd..4c3981d 100644 --- a/llvm/lib/Target/Mips/MipsTargetMachine.cpp +++ b/llvm/lib/Target/Mips/MipsTargetMachine.cpp @@ -53,7 +53,7 @@ MipsTargetMachine(const Target &T, StringRef TT, InstrInfo(MipsInstrInfo::create(*this)), FrameLowering(MipsFrameLowering::create(*this, Subtarget)), TLInfo(*this), TSInfo(*this), JITInfo(), - ELFWriterInfo(false, isLittle) { + ELFWriterInfo(false, isLittle), STTI(&TLInfo) { } void MipsebTargetMachine::anchor() { } diff --git a/llvm/lib/Target/Mips/MipsTargetMachine.h b/llvm/lib/Target/Mips/MipsTargetMachine.h index 3a01828..55524e8 100644 --- a/llvm/lib/Target/Mips/MipsTargetMachine.h +++ b/llvm/lib/Target/Mips/MipsTargetMachine.h @@ -24,6 +24,7 @@ #include "llvm/Target/TargetMachine.h" #include "llvm/DataLayout.h" #include "llvm/Target/TargetFrameLowering.h" +#include "llvm/Target/TargetTransformImpl.h" namespace llvm { class formatted_raw_ostream; @@ -38,6 +39,8 @@ class MipsTargetMachine : public LLVMTargetMachine { MipsSelectionDAGInfo TSInfo; MipsJITInfo JITInfo; MipsELFWriterInfo ELFWriterInfo; + ScalarTargetTransformImpl STTI; + VectorTargetTransformImpl VTTI; public: MipsTargetMachine(const Target &T, StringRef TT, diff --git a/llvm/lib/Target/NVPTX/NVPTXTargetMachine.cpp b/llvm/lib/Target/NVPTX/NVPTXTargetMachine.cpp index dbfc660..7519b4a 100644 --- a/llvm/lib/Target/NVPTX/NVPTXTargetMachine.cpp +++ b/llvm/lib/Target/NVPTX/NVPTXTargetMachine.cpp @@ -72,7 +72,8 @@ NVPTXTargetMachine::NVPTXTargetMachine(const Target &T, : LLVMTargetMachine(T, TT, CPU, FS, Options, RM, CM, OL), Subtarget(TT, CPU, FS, is64bit), DL(Subtarget.getDataLayout()), - InstrInfo(*this), TLInfo(*this), TSInfo(*this), FrameLowering(*this,is64bit) + InstrInfo(*this), TLInfo(*this), TSInfo(*this), FrameLowering(*this,is64bit), + STTI(&TLInfo) /*FrameInfo(TargetFrameInfo::StackGrowsUp, 8, 0)*/ { } diff --git a/llvm/lib/Target/NVPTX/NVPTXTargetMachine.h b/llvm/lib/Target/NVPTX/NVPTXTargetMachine.h index d58a076..11bc9d4 100644 --- a/llvm/lib/Target/NVPTX/NVPTXTargetMachine.h +++ b/llvm/lib/Target/NVPTX/NVPTXTargetMachine.h @@ -25,6 +25,7 @@ #include "llvm/Target/TargetFrameLowering.h" #include "llvm/Target/TargetMachine.h" #include "llvm/Target/TargetSelectionDAGInfo.h" +#include "llvm/Target/TargetTransformImpl.h" namespace llvm { @@ -44,6 +45,9 @@ class NVPTXTargetMachine : public LLVMTargetMachine { // Hold Strings that can be free'd all together with NVPTXTargetMachine ManagedStringPool ManagedStrPool; + ScalarTargetTransformImpl STTI; + VectorTargetTransformImpl VTTI; + //bool addCommonCodeGenPasses(PassManagerBase &, CodeGenOpt::Level, // bool DisableVerify, MCContext *&OutCtx); @@ -72,6 +76,12 @@ public: virtual const TargetSelectionDAGInfo *getSelectionDAGInfo() const { return &TSInfo; } + virtual const ScalarTargetTransformInfo *getScalarTargetTransformInfo()const { + return &STTI; + } + virtual const VectorTargetTransformInfo *getVectorTargetTransformInfo()const { + return &VTTI; + } //virtual bool addInstSelector(PassManagerBase &PM, // CodeGenOpt::Level OptLevel); diff --git a/llvm/lib/Target/PowerPC/PPCTargetMachine.cpp b/llvm/lib/Target/PowerPC/PPCTargetMachine.cpp index 5f39b8d..b861383 100644 --- a/llvm/lib/Target/PowerPC/PPCTargetMachine.cpp +++ b/llvm/lib/Target/PowerPC/PPCTargetMachine.cpp @@ -43,7 +43,8 @@ PPCTargetMachine::PPCTargetMachine(const Target &T, StringRef TT, DL(Subtarget.getDataLayoutString()), InstrInfo(*this), FrameLowering(Subtarget), JITInfo(*this, is64Bit), TLInfo(*this), TSInfo(*this), - InstrItins(Subtarget.getInstrItineraryData()) { + InstrItins(Subtarget.getInstrItineraryData()), + STTI(&TLInfo){ // The binutils for the BG/P are too old for CFI. if (Subtarget.isBGP()) diff --git a/llvm/lib/Target/PowerPC/PPCTargetMachine.h b/llvm/lib/Target/PowerPC/PPCTargetMachine.h index 02d69fd..c168433 100644 --- a/llvm/lib/Target/PowerPC/PPCTargetMachine.h +++ b/llvm/lib/Target/PowerPC/PPCTargetMachine.h @@ -21,6 +21,7 @@ #include "PPCISelLowering.h" #include "PPCSelectionDAGInfo.h" #include "llvm/Target/TargetMachine.h" +#include "llvm/Target/TargetTransformImpl.h" #include "llvm/DataLayout.h" namespace llvm { @@ -36,6 +37,8 @@ class PPCTargetMachine : public LLVMTargetMachine { PPCTargetLowering TLInfo; PPCSelectionDAGInfo TSInfo; InstrItineraryData InstrItins; + ScalarTargetTransformImpl STTI; + VectorTargetTransformImpl VTTI; public: PPCTargetMachine(const Target &T, StringRef TT, @@ -63,6 +66,12 @@ public: virtual const InstrItineraryData *getInstrItineraryData() const { return &InstrItins; } + virtual const ScalarTargetTransformInfo *getScalarTargetTransformInfo()const { + return &STTI; + } + virtual const VectorTargetTransformInfo *getVectorTargetTransformInfo()const { + return &VTTI; + } // Pass Pipeline Configuration virtual TargetPassConfig *createPassConfig(PassManagerBase &PM); diff --git a/llvm/lib/Target/Sparc/SparcTargetMachine.cpp b/llvm/lib/Target/Sparc/SparcTargetMachine.cpp index 8b7559c..1d8cc771 100644 --- a/llvm/lib/Target/Sparc/SparcTargetMachine.cpp +++ b/llvm/lib/Target/Sparc/SparcTargetMachine.cpp @@ -36,7 +36,7 @@ SparcTargetMachine::SparcTargetMachine(const Target &T, StringRef TT, DL(Subtarget.getDataLayout()), InstrInfo(Subtarget), TLInfo(*this), TSInfo(*this), - FrameLowering(Subtarget) { + FrameLowering(Subtarget),STTI(&TLInfo) { } namespace { diff --git a/llvm/lib/Target/Sparc/SparcTargetMachine.h b/llvm/lib/Target/Sparc/SparcTargetMachine.h index c9f2d68..0fbe2d7 100644 --- a/llvm/lib/Target/Sparc/SparcTargetMachine.h +++ b/llvm/lib/Target/Sparc/SparcTargetMachine.h @@ -22,6 +22,7 @@ #include "llvm/Target/TargetMachine.h" #include "llvm/DataLayout.h" #include "llvm/Target/TargetFrameLowering.h" +#include "llvm/Target/TargetTransformImpl.h" namespace llvm { @@ -32,6 +33,8 @@ class SparcTargetMachine : public LLVMTargetMachine { SparcTargetLowering TLInfo; SparcSelectionDAGInfo TSInfo; SparcFrameLowering FrameLowering; + ScalarTargetTransformImpl STTI; + VectorTargetTransformImpl VTTI; public: SparcTargetMachine(const Target &T, StringRef TT, StringRef CPU, StringRef FS, const TargetOptions &Options, @@ -52,6 +55,12 @@ public: virtual const SparcSelectionDAGInfo* getSelectionDAGInfo() const { return &TSInfo; } + virtual const ScalarTargetTransformInfo *getScalarTargetTransformInfo()const { + return &STTI; + } + virtual const VectorTargetTransformInfo *getVectorTargetTransformInfo()const { + return &VTTI; + } virtual const DataLayout *getDataLayout() const { return &DL; } // Pass Pipeline Configuration diff --git a/llvm/lib/Target/Target.cpp b/llvm/lib/Target/Target.cpp index b0b5c87..219cbf1 100644 --- a/llvm/lib/Target/Target.cpp +++ b/llvm/lib/Target/Target.cpp @@ -26,6 +26,7 @@ using namespace llvm; void llvm::initializeTarget(PassRegistry &Registry) { initializeDataLayoutPass(Registry); initializeTargetLibraryInfoPass(Registry); + initializeTargetTransformInfoPass(Registry); } void LLVMInitializeTarget(LLVMPassRegistryRef R) { diff --git a/llvm/lib/Target/TargetTransformImpl.cpp b/llvm/lib/Target/TargetTransformImpl.cpp new file mode 100644 index 0000000..1cb5eda --- /dev/null +++ b/llvm/lib/Target/TargetTransformImpl.cpp @@ -0,0 +1,43 @@ +// llvm/Target/TargetTransformImpl.cpp - Target Loop Trans Info ---*- C++ -*-=// +// +// The LLVM Compiler Infrastructure +// +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// + +#include "llvm/Target/TargetTransformImpl.h" +#include "llvm/Target/TargetLowering.h" + +using namespace llvm; + +bool ScalarTargetTransformImpl::isLegalAddImmediate(int64_t imm) const { + return TLI->isLegalAddImmediate(imm); +} + +bool ScalarTargetTransformImpl::isLegalICmpImmediate(int64_t imm) const { + return TLI->isLegalICmpImmediate(imm); +} + +bool ScalarTargetTransformImpl::isLegalAddressingMode(const AddrMode &AM, + Type *Ty) const { + return TLI->isLegalAddressingMode(AM, Ty); +} + +bool ScalarTargetTransformImpl::isTruncateFree(Type *Ty1, Type *Ty2) const { + return TLI->isTruncateFree(Ty1, Ty2); +} + +bool ScalarTargetTransformImpl::isTypeLegal(Type *Ty) const { + EVT T = TLI->getValueType(Ty); + return TLI->isTypeLegal(T); +} + +unsigned ScalarTargetTransformImpl::getJumpBufAlignment() const { + return TLI->getJumpBufAlignment(); +} + +unsigned ScalarTargetTransformImpl::getJumpBufSize() const { + return TLI->getJumpBufSize(); +} diff --git a/llvm/lib/Target/X86/X86TargetMachine.cpp b/llvm/lib/Target/X86/X86TargetMachine.cpp index f8cced8..655ede7 100644 --- a/llvm/lib/Target/X86/X86TargetMachine.cpp +++ b/llvm/lib/Target/X86/X86TargetMachine.cpp @@ -48,7 +48,8 @@ X86_32TargetMachine::X86_32TargetMachine(const Target &T, StringRef TT, InstrInfo(*this), TSInfo(*this), TLInfo(*this), - JITInfo(*this) { + JITInfo(*this), + STTI(&TLInfo) { } void X86_64TargetMachine::anchor() { } @@ -64,7 +65,8 @@ X86_64TargetMachine::X86_64TargetMachine(const Target &T, StringRef TT, InstrInfo(*this), TSInfo(*this), TLInfo(*this), - JITInfo(*this) { + JITInfo(*this), + STTI(&TLInfo) { } /// X86TargetMachine ctor - Create an X86 target. diff --git a/llvm/lib/Target/X86/X86TargetMachine.h b/llvm/lib/Target/X86/X86TargetMachine.h index 5301299..4bad695 100644 --- a/llvm/lib/Target/X86/X86TargetMachine.h +++ b/llvm/lib/Target/X86/X86TargetMachine.h @@ -25,6 +25,7 @@ #include "llvm/Target/TargetMachine.h" #include "llvm/DataLayout.h" #include "llvm/Target/TargetFrameLowering.h" +#include "llvm/Target/TargetTransformImpl.h" namespace llvm { @@ -85,6 +86,8 @@ class X86_32TargetMachine : public X86TargetMachine { X86SelectionDAGInfo TSInfo; X86TargetLowering TLInfo; X86JITInfo JITInfo; + ScalarTargetTransformImpl STTI; + VectorTargetTransformImpl VTTI; public: X86_32TargetMachine(const Target &T, StringRef TT, StringRef CPU, StringRef FS, const TargetOptions &Options, @@ -103,6 +106,12 @@ public: virtual X86JITInfo *getJITInfo() { return &JITInfo; } + virtual const ScalarTargetTransformInfo *getScalarTargetTransformInfo()const { + return &STTI; + } + virtual const VectorTargetTransformInfo *getVectorTargetTransformInfo()const { + return &VTTI; + } }; /// X86_64TargetMachine - X86 64-bit target machine. @@ -114,6 +123,8 @@ class X86_64TargetMachine : public X86TargetMachine { X86SelectionDAGInfo TSInfo; X86TargetLowering TLInfo; X86JITInfo JITInfo; + ScalarTargetTransformImpl STTI; + VectorTargetTransformImpl VTTI; public: X86_64TargetMachine(const Target &T, StringRef TT, StringRef CPU, StringRef FS, const TargetOptions &Options, @@ -132,6 +143,12 @@ public: virtual X86JITInfo *getJITInfo() { return &JITInfo; } + virtual const ScalarTargetTransformInfo *getScalarTargetTransformInfo()const { + return &STTI; + } + virtual const VectorTargetTransformInfo *getVectorTargetTransformInfo()const { + return &VTTI; + } }; } // End llvm namespace diff --git a/llvm/lib/Target/XCore/XCoreTargetMachine.cpp b/llvm/lib/Target/XCore/XCoreTargetMachine.cpp index c71d978..0b7e3e1 100644 --- a/llvm/lib/Target/XCore/XCoreTargetMachine.cpp +++ b/llvm/lib/Target/XCore/XCoreTargetMachine.cpp @@ -32,7 +32,7 @@ XCoreTargetMachine::XCoreTargetMachine(const Target &T, StringRef TT, InstrInfo(), FrameLowering(Subtarget), TLInfo(*this), - TSInfo(*this) { + TSInfo(*this), STTI(&TLInfo) { } namespace { diff --git a/llvm/lib/Target/XCore/XCoreTargetMachine.h b/llvm/lib/Target/XCore/XCoreTargetMachine.h index f7fec29..c60c6a3 100644 --- a/llvm/lib/Target/XCore/XCoreTargetMachine.h +++ b/llvm/lib/Target/XCore/XCoreTargetMachine.h @@ -20,6 +20,7 @@ #include "XCoreISelLowering.h" #include "XCoreSelectionDAGInfo.h" #include "llvm/Target/TargetMachine.h" +#include "llvm/Target/TargetTransformImpl.h" #include "llvm/DataLayout.h" namespace llvm { @@ -31,6 +32,8 @@ class XCoreTargetMachine : public LLVMTargetMachine { XCoreFrameLowering FrameLowering; XCoreTargetLowering TLInfo; XCoreSelectionDAGInfo TSInfo; + ScalarTargetTransformImpl STTI; + VectorTargetTransformImpl VTTI; public: XCoreTargetMachine(const Target &T, StringRef TT, StringRef CPU, StringRef FS, const TargetOptions &Options, @@ -53,6 +56,12 @@ public: virtual const TargetRegisterInfo *getRegisterInfo() const { return &InstrInfo.getRegisterInfo(); } + virtual const ScalarTargetTransformInfo *getScalarTargetTransformInfo()const { + return &STTI; + } + virtual const VectorTargetTransformInfo *getVectorTargetTransformInfo()const { + return &VTTI; + } virtual const DataLayout *getDataLayout() const { return &DL; } // Pass Pipeline Configuration diff --git a/llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp b/llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp index 958348d..99a62db 100644 --- a/llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp +++ b/llvm/lib/Transforms/Scalar/LoopStrengthReduce.cpp @@ -37,7 +37,7 @@ // // TODO: Handle multiple loops at a time. // -// TODO: Should TargetLowering::AddrMode::BaseGV be changed to a ConstantExpr +// TODO: Should AddrMode::BaseGV be changed to a ConstantExpr // instead of a GlobalValue? // // TODO: When truncation is free, truncate ICmp users' operands to make it a @@ -67,6 +67,7 @@ #include "llvm/Transforms/Scalar.h" #include "llvm/Transforms/Utils/BasicBlockUtils.h" #include "llvm/Transforms/Utils/Local.h" +#include "llvm/TargetTransformInfo.h" #include "llvm/ADT/SmallBitVector.h" #include "llvm/ADT/SetVector.h" #include "llvm/ADT/DenseSet.h" @@ -74,7 +75,6 @@ #include "llvm/Support/CommandLine.h" #include "llvm/Support/ValueHandle.h" #include "llvm/Support/raw_ostream.h" -#include "llvm/Target/TargetLowering.h" #include using namespace llvm; @@ -1118,7 +1118,7 @@ public: enum KindType { Basic, ///< A normal use, with no folding. Special, ///< A special case of basic, allowing -1 scales. - Address, ///< An address use; folding according to TargetLowering + Address, ///< An address use; folding according to ScalarTargetTransformInfo. ICmpZero ///< An equality icmp with both operands folded into one. // TODO: Add a generic icmp too? }; @@ -1272,12 +1272,12 @@ void LSRUse::dump() const { /// address-mode folding and special icmp tricks. static bool isLegalUse(const AddrMode &AM, LSRUse::KindType Kind, Type *AccessTy, - const TargetLowering *TLI) { + const ScalarTargetTransformInfo *STTI) { switch (Kind) { case LSRUse::Address: // If we have low-level target information, ask the target if it can // completely fold this address. - if (TLI) return TLI->isLegalAddressingMode(AM, AccessTy); + if (STTI) return STTI->isLegalAddressingMode(AM, AccessTy); // Otherwise, just guess that reg+reg addressing is legal. return !AM.BaseGV && AM.BaseOffs == 0 && AM.Scale <= 1; @@ -1300,7 +1300,7 @@ static bool isLegalUse(const AddrMode &AM, // If we have low-level target information, ask the target if it can fold an // integer immediate on an icmp. if (AM.BaseOffs != 0) { - if (!TLI) + if (!STTI) return false; // We have one of: // ICmpZero BaseReg + Offset => ICmp BaseReg, -Offset @@ -1309,7 +1309,7 @@ static bool isLegalUse(const AddrMode &AM, int64_t Offs = AM.BaseOffs; if (AM.Scale == 0) Offs = -(uint64_t)Offs; // The cast does the right thing with INT64_MIN. - return TLI->isLegalICmpImmediate(Offs); + return STTI->isLegalICmpImmediate(Offs); } // ICmpZero BaseReg + -1*ScaleReg => ICmp BaseReg, ScaleReg @@ -1330,20 +1330,20 @@ static bool isLegalUse(const AddrMode &AM, static bool isLegalUse(AddrMode AM, int64_t MinOffset, int64_t MaxOffset, LSRUse::KindType Kind, Type *AccessTy, - const TargetLowering *TLI) { + const ScalarTargetTransformInfo *LTTI) { // Check for overflow. if (((int64_t)((uint64_t)AM.BaseOffs + MinOffset) > AM.BaseOffs) != (MinOffset > 0)) return false; AM.BaseOffs = (uint64_t)AM.BaseOffs + MinOffset; - if (isLegalUse(AM, Kind, AccessTy, TLI)) { + if (isLegalUse(AM, Kind, AccessTy, LTTI)) { AM.BaseOffs = (uint64_t)AM.BaseOffs - MinOffset; // Check for overflow. if (((int64_t)((uint64_t)AM.BaseOffs + MaxOffset) > AM.BaseOffs) != (MaxOffset > 0)) return false; AM.BaseOffs = (uint64_t)AM.BaseOffs + MaxOffset; - return isLegalUse(AM, Kind, AccessTy, TLI); + return isLegalUse(AM, Kind, AccessTy, LTTI); } return false; } @@ -1352,7 +1352,7 @@ static bool isAlwaysFoldable(int64_t BaseOffs, GlobalValue *BaseGV, bool HasBaseReg, LSRUse::KindType Kind, Type *AccessTy, - const TargetLowering *TLI) { + const ScalarTargetTransformInfo *LTTI) { // Fast-path: zero is always foldable. if (BaseOffs == 0 && !BaseGV) return true; @@ -1371,14 +1371,14 @@ static bool isAlwaysFoldable(int64_t BaseOffs, AM.HasBaseReg = true; } - return isLegalUse(AM, Kind, AccessTy, TLI); + return isLegalUse(AM, Kind, AccessTy, LTTI); } static bool isAlwaysFoldable(const SCEV *S, int64_t MinOffset, int64_t MaxOffset, bool HasBaseReg, LSRUse::KindType Kind, Type *AccessTy, - const TargetLowering *TLI, + const ScalarTargetTransformInfo *LTTI, ScalarEvolution &SE) { // Fast-path: zero is always foldable. if (S->isZero()) return true; @@ -1402,7 +1402,7 @@ static bool isAlwaysFoldable(const SCEV *S, AM.HasBaseReg = HasBaseReg; AM.Scale = Kind == LSRUse::ICmpZero ? -1 : 1; - return isLegalUse(AM, MinOffset, MaxOffset, Kind, AccessTy, TLI); + return isLegalUse(AM, MinOffset, MaxOffset, Kind, AccessTy, LTTI); } namespace { @@ -1502,7 +1502,7 @@ class LSRInstance { ScalarEvolution &SE; DominatorTree &DT; LoopInfo &LI; - const TargetLowering *const TLI; + const ScalarTargetTransformInfo *const STTI; Loop *const L; bool Changed; @@ -1638,7 +1638,7 @@ class LSRInstance { Pass *P); public: - LSRInstance(const TargetLowering *tli, Loop *l, Pass *P); + LSRInstance(const ScalarTargetTransformInfo *ltti, Loop *l, Pass *P); bool getChanged() const { return Changed; } @@ -1688,11 +1688,10 @@ void LSRInstance::OptimizeShadowIV() { } if (!DestTy) continue; - if (TLI) { + if (STTI) { // If target does not support DestTy natively then do not apply // this transformation. - EVT DVT = TLI->getValueType(DestTy); - if (!TLI->isTypeLegal(DVT)) continue; + if (!STTI->isTypeLegal(DestTy)) continue; } PHINode *PH = dyn_cast(ShadowUse->getOperand(0)); @@ -2015,18 +2014,18 @@ LSRInstance::OptimizeLoopTermCond() { if (C->getValue().getMinSignedBits() >= 64 || C->getValue().isMinSignedValue()) goto decline_post_inc; - // Without TLI, assume that any stride might be valid, and so any + // Without STTI, assume that any stride might be valid, and so any // use might be shared. - if (!TLI) + if (!STTI) goto decline_post_inc; // Check for possible scaled-address reuse. Type *AccessTy = getAccessType(UI->getUser()); AddrMode AM; AM.Scale = C->getSExtValue(); - if (TLI->isLegalAddressingMode(AM, AccessTy)) + if (STTI->isLegalAddressingMode(AM, AccessTy)) goto decline_post_inc; AM.Scale = -AM.Scale; - if (TLI->isLegalAddressingMode(AM, AccessTy)) + if (STTI->isLegalAddressingMode(AM, AccessTy)) goto decline_post_inc; } } @@ -2097,12 +2096,12 @@ LSRInstance::reconcileNewOffset(LSRUse &LU, int64_t NewOffset, bool HasBaseReg, // Conservatively assume HasBaseReg is true for now. if (NewOffset < LU.MinOffset) { if (!isAlwaysFoldable(LU.MaxOffset - NewOffset, 0, HasBaseReg, - Kind, AccessTy, TLI)) + Kind, AccessTy, STTI)) return false; NewMinOffset = NewOffset; } else if (NewOffset > LU.MaxOffset) { if (!isAlwaysFoldable(NewOffset - LU.MinOffset, 0, HasBaseReg, - Kind, AccessTy, TLI)) + Kind, AccessTy, STTI)) return false; NewMaxOffset = NewOffset; } @@ -2131,7 +2130,7 @@ LSRInstance::getUse(const SCEV *&Expr, int64_t Offset = ExtractImmediate(Expr, SE); // Basic uses can't accept any offset, for example. - if (!isAlwaysFoldable(Offset, 0, /*HasBaseReg=*/true, Kind, AccessTy, TLI)) { + if (!isAlwaysFoldable(Offset, 0, /*HasBaseReg=*/true, Kind, AccessTy, STTI)) { Expr = Copy; Offset = 0; } @@ -2396,7 +2395,7 @@ bool IVChain::isProfitableIncrement(const SCEV *OperExpr, /// TODO: Consider IVInc free if it's already used in another chains. static bool isProfitableChain(IVChain &Chain, SmallPtrSet &Users, - ScalarEvolution &SE, const TargetLowering *TLI) { + ScalarEvolution &SE, const ScalarTargetTransformInfo *STTI) { if (StressIVChain) return true; @@ -2654,7 +2653,7 @@ void LSRInstance::CollectChains() { for (unsigned UsersIdx = 0, NChains = IVChainVec.size(); UsersIdx < NChains; ++UsersIdx) { if (!isProfitableChain(IVChainVec[UsersIdx], - ChainUsersVec[UsersIdx].FarUsers, SE, TLI)) + ChainUsersVec[UsersIdx].FarUsers, SE, STTI)) continue; // Preserve the chain at UsesIdx. if (ChainIdx != UsersIdx) @@ -2681,7 +2680,8 @@ void LSRInstance::FinalizeChain(IVChain &Chain) { /// Return true if the IVInc can be folded into an addressing mode. static bool canFoldIVIncExpr(const SCEV *IncExpr, Instruction *UserInst, - Value *Operand, const TargetLowering *TLI) { + Value *Operand, + const ScalarTargetTransformInfo *STTI) { const SCEVConstant *IncConst = dyn_cast(IncExpr); if (!IncConst || !isAddressUse(UserInst, Operand)) return false; @@ -2691,7 +2691,7 @@ static bool canFoldIVIncExpr(const SCEV *IncExpr, Instruction *UserInst, int64_t IncOffset = IncConst->getValue()->getSExtValue(); if (!isAlwaysFoldable(IncOffset, /*BaseGV=*/0, /*HaseBaseReg=*/false, - LSRUse::Address, getAccessType(UserInst), TLI)) + LSRUse::Address, getAccessType(UserInst), STTI)) return false; return true; @@ -2762,7 +2762,7 @@ void LSRInstance::GenerateIVChain(const IVChain &Chain, SCEVExpander &Rewriter, // If an IV increment can't be folded, use it as the next IV value. if (!canFoldIVIncExpr(LeftOverExpr, IncI->UserInst, IncI->IVOperand, - TLI)) { + STTI)) { assert(IVTy == IVOper->getType() && "inconsistent IV increment type"); IVSrc = IVOper; LeftOverExpr = 0; @@ -3108,7 +3108,7 @@ void LSRInstance::GenerateReassociations(LSRUse &LU, unsigned LUIdx, // into an immediate field. if (isAlwaysFoldable(*J, LU.MinOffset, LU.MaxOffset, Base.getNumRegs() > 1, - LU.Kind, LU.AccessTy, TLI, SE)) + LU.Kind, LU.AccessTy, STTI, SE)) continue; // Collect all operands except *J. @@ -3122,7 +3122,7 @@ void LSRInstance::GenerateReassociations(LSRUse &LU, unsigned LUIdx, if (InnerAddOps.size() == 1 && isAlwaysFoldable(InnerAddOps[0], LU.MinOffset, LU.MaxOffset, Base.getNumRegs() > 1, - LU.Kind, LU.AccessTy, TLI, SE)) + LU.Kind, LU.AccessTy, STTI, SE)) continue; const SCEV *InnerSum = SE.getAddExpr(InnerAddOps); @@ -3132,9 +3132,9 @@ void LSRInstance::GenerateReassociations(LSRUse &LU, unsigned LUIdx, // Add the remaining pieces of the add back into the new formula. const SCEVConstant *InnerSumSC = dyn_cast(InnerSum); - if (TLI && InnerSumSC && + if (STTI && InnerSumSC && SE.getTypeSizeInBits(InnerSumSC->getType()) <= 64 && - TLI->isLegalAddImmediate((uint64_t)F.UnfoldedOffset + + STTI->isLegalAddImmediate((uint64_t)F.UnfoldedOffset + InnerSumSC->getValue()->getZExtValue())) { F.UnfoldedOffset = (uint64_t)F.UnfoldedOffset + InnerSumSC->getValue()->getZExtValue(); @@ -3144,8 +3144,8 @@ void LSRInstance::GenerateReassociations(LSRUse &LU, unsigned LUIdx, // Add J as its own register, or an unfolded immediate. const SCEVConstant *SC = dyn_cast(*J); - if (TLI && SC && SE.getTypeSizeInBits(SC->getType()) <= 64 && - TLI->isLegalAddImmediate((uint64_t)F.UnfoldedOffset + + if (STTI && SC && SE.getTypeSizeInBits(SC->getType()) <= 64 && + STTI->isLegalAddImmediate((uint64_t)F.UnfoldedOffset + SC->getValue()->getZExtValue())) F.UnfoldedOffset = (uint64_t)F.UnfoldedOffset + SC->getValue()->getZExtValue(); @@ -3205,7 +3205,7 @@ void LSRInstance::GenerateSymbolicOffsets(LSRUse &LU, unsigned LUIdx, Formula F = Base; F.AM.BaseGV = GV; if (!isLegalUse(F.AM, LU.MinOffset, LU.MaxOffset, - LU.Kind, LU.AccessTy, TLI)) + LU.Kind, LU.AccessTy, STTI)) continue; F.BaseRegs[i] = G; (void)InsertFormula(LU, LUIdx, F); @@ -3230,7 +3230,7 @@ void LSRInstance::GenerateConstantOffsets(LSRUse &LU, unsigned LUIdx, Formula F = Base; F.AM.BaseOffs = (uint64_t)Base.AM.BaseOffs - *I; if (isLegalUse(F.AM, LU.MinOffset - *I, LU.MaxOffset - *I, - LU.Kind, LU.AccessTy, TLI)) { + LU.Kind, LU.AccessTy, STTI)) { // Add the offset to the base register. const SCEV *NewG = SE.getAddExpr(SE.getConstant(G->getType(), *I), G); // If it cancelled out, drop the base register, otherwise update it. @@ -3250,7 +3250,7 @@ void LSRInstance::GenerateConstantOffsets(LSRUse &LU, unsigned LUIdx, Formula F = Base; F.AM.BaseOffs = (uint64_t)F.AM.BaseOffs + Imm; if (!isLegalUse(F.AM, LU.MinOffset, LU.MaxOffset, - LU.Kind, LU.AccessTy, TLI)) + LU.Kind, LU.AccessTy, STTI)) continue; F.BaseRegs[i] = G; (void)InsertFormula(LU, LUIdx, F); @@ -3297,7 +3297,7 @@ void LSRInstance::GenerateICmpZeroScales(LSRUse &LU, unsigned LUIdx, F.AM.BaseOffs = NewBaseOffs; // Check that this scale is legal. - if (!isLegalUse(F.AM, Offset, Offset, LU.Kind, LU.AccessTy, TLI)) + if (!isLegalUse(F.AM, Offset, Offset, LU.Kind, LU.AccessTy, STTI)) continue; // Compensate for the use having MinOffset built into it. @@ -3353,12 +3353,12 @@ void LSRInstance::GenerateScales(LSRUse &LU, unsigned LUIdx, Formula Base) { Base.AM.HasBaseReg = Base.BaseRegs.size() > 1; // Check whether this scale is going to be legal. if (!isLegalUse(Base.AM, LU.MinOffset, LU.MaxOffset, - LU.Kind, LU.AccessTy, TLI)) { + LU.Kind, LU.AccessTy, STTI)) { // As a special-case, handle special out-of-loop Basic users specially. // TODO: Reconsider this special case. if (LU.Kind == LSRUse::Basic && isLegalUse(Base.AM, LU.MinOffset, LU.MaxOffset, - LSRUse::Special, LU.AccessTy, TLI) && + LSRUse::Special, LU.AccessTy, STTI) && LU.AllFixupsOutsideLoop) LU.Kind = LSRUse::Special; else @@ -3391,8 +3391,8 @@ void LSRInstance::GenerateScales(LSRUse &LU, unsigned LUIdx, Formula Base) { /// GenerateTruncates - Generate reuse formulae from different IV types. void LSRInstance::GenerateTruncates(LSRUse &LU, unsigned LUIdx, Formula Base) { - // This requires TargetLowering to tell us which truncates are free. - if (!TLI) return; + // This requires ScalarTargetTransformInfo to tell us which truncates are free. + if (!STTI) return; // Don't bother truncating symbolic values. if (Base.AM.BaseGV) return; @@ -3405,7 +3405,7 @@ void LSRInstance::GenerateTruncates(LSRUse &LU, unsigned LUIdx, Formula Base) { for (SmallSetVector::const_iterator I = Types.begin(), E = Types.end(); I != E; ++I) { Type *SrcTy = *I; - if (SrcTy != DstTy && TLI->isTruncateFree(SrcTy, DstTy)) { + if (SrcTy != DstTy && STTI->isTruncateFree(SrcTy, DstTy)) { Formula F = Base; if (F.ScaledReg) F.ScaledReg = SE.getAnyExtendExpr(F.ScaledReg, *I); @@ -3561,7 +3561,7 @@ void LSRInstance::GenerateCrossUseConstantOffsets() { Formula NewF = F; NewF.AM.BaseOffs = Offs; if (!isLegalUse(NewF.AM, LU.MinOffset, LU.MaxOffset, - LU.Kind, LU.AccessTy, TLI)) + LU.Kind, LU.AccessTy, STTI)) continue; NewF.ScaledReg = SE.getAddExpr(NegImmS, NewF.ScaledReg); @@ -3586,9 +3586,9 @@ void LSRInstance::GenerateCrossUseConstantOffsets() { Formula NewF = F; NewF.AM.BaseOffs = (uint64_t)NewF.AM.BaseOffs + Imm; if (!isLegalUse(NewF.AM, LU.MinOffset, LU.MaxOffset, - LU.Kind, LU.AccessTy, TLI)) { - if (!TLI || - !TLI->isLegalAddImmediate((uint64_t)NewF.UnfoldedOffset + Imm)) + LU.Kind, LU.AccessTy, STTI)) { + if (!STTI || + !STTI->isLegalAddImmediate((uint64_t)NewF.UnfoldedOffset + Imm)) continue; NewF = F; NewF.UnfoldedOffset = (uint64_t)NewF.UnfoldedOffset + Imm; @@ -3900,7 +3900,7 @@ void LSRInstance::NarrowSearchSpaceByCollapsingUnrolledCode() { Formula &F = LUThatHas->Formulae[i]; if (!isLegalUse(F.AM, LUThatHas->MinOffset, LUThatHas->MaxOffset, - LUThatHas->Kind, LUThatHas->AccessTy, TLI)) { + LUThatHas->Kind, LUThatHas->AccessTy, STTI)) { DEBUG(dbgs() << " Deleting "; F.print(dbgs()); dbgs() << '\n'); LUThatHas->DeleteFormula(F); @@ -4589,12 +4589,12 @@ LSRInstance::ImplementSolution(const SmallVectorImpl &Solution, Changed |= DeleteTriviallyDeadInstructions(DeadInsts); } -LSRInstance::LSRInstance(const TargetLowering *tli, Loop *l, Pass *P) +LSRInstance::LSRInstance(const ScalarTargetTransformInfo *stti, Loop *l, Pass *P) : IU(P->getAnalysis()), SE(P->getAnalysis()), DT(P->getAnalysis()), LI(P->getAnalysis()), - TLI(tli), L(l), Changed(false), IVIncInsertPos(0) { + STTI(stti), L(l), Changed(false), IVIncInsertPos(0) { // If LoopSimplify form is not available, stay out of trouble. if (!L->isLoopSimplifyForm()) @@ -4684,7 +4684,7 @@ LSRInstance::LSRInstance(const TargetLowering *tli, Loop *l, Pass *P) for (SmallVectorImpl::const_iterator J = LU.Formulae.begin(), JE = LU.Formulae.end(); J != JE; ++J) assert(isLegalUse(J->AM, LU.MinOffset, LU.MaxOffset, - LU.Kind, LU.AccessTy, TLI) && + LU.Kind, LU.AccessTy, STTI) && "Illegal formula generated!"); }; #endif @@ -4757,13 +4757,13 @@ void LSRInstance::dump() const { namespace { class LoopStrengthReduce : public LoopPass { - /// TLI - Keep a pointer of a TargetLowering to consult for determining - /// transformation profitability. - const TargetLowering *const TLI; + /// ScalarTargetTransformInfo provides target information that is needed + /// for strength reducing loops. + const ScalarTargetTransformInfo *STTI; public: static char ID; // Pass ID, replacement for typeid - explicit LoopStrengthReduce(const TargetLowering *tli = 0); + LoopStrengthReduce(); private: bool runOnLoop(Loop *L, LPPassManager &LPM); @@ -4783,13 +4783,12 @@ INITIALIZE_PASS_DEPENDENCY(LoopSimplify) INITIALIZE_PASS_END(LoopStrengthReduce, "loop-reduce", "Loop Strength Reduction", false, false) - -Pass *llvm::createLoopStrengthReducePass(const TargetLowering *TLI) { - return new LoopStrengthReduce(TLI); +Pass *llvm::createLoopStrengthReducePass() { + return new LoopStrengthReduce(); } -LoopStrengthReduce::LoopStrengthReduce(const TargetLowering *tli) - : LoopPass(ID), TLI(tli) { +LoopStrengthReduce::LoopStrengthReduce() + : LoopPass(ID), STTI(0) { initializeLoopStrengthReducePass(*PassRegistry::getPassRegistry()); } @@ -4815,8 +4814,13 @@ void LoopStrengthReduce::getAnalysisUsage(AnalysisUsage &AU) const { bool LoopStrengthReduce::runOnLoop(Loop *L, LPPassManager & /*LPM*/) { bool Changed = false; + TargetTransformInfo *TTI = getAnalysisIfAvailable(); + + if (TTI) + STTI = TTI->getScalarTargetTransformInfo(); + // Run the main LSR transformation. - Changed |= LSRInstance(TLI, L, this).getChanged(); + Changed |= LSRInstance(STTI, L, this).getChanged(); // Remove any extra phis created by processing inner loops. Changed |= DeleteDeadPHIs(L->getHeader()); @@ -4827,7 +4831,7 @@ bool LoopStrengthReduce::runOnLoop(Loop *L, LPPassManager & /*LPM*/) { Rewriter.setDebugType(DEBUG_TYPE); #endif unsigned numFolded = Rewriter. - replaceCongruentIVs(L, &getAnalysis(), DeadInsts, TLI); + replaceCongruentIVs(L, &getAnalysis(), DeadInsts, STTI); if (numFolded) { Changed = true; DeleteTriviallyDeadInstructions(DeadInsts); diff --git a/llvm/lib/Transforms/Utils/LowerInvoke.cpp b/llvm/lib/Transforms/Utils/LowerInvoke.cpp index 9305554..f35cbbd 100644 --- a/llvm/lib/Transforms/Utils/LowerInvoke.cpp +++ b/llvm/lib/Transforms/Utils/LowerInvoke.cpp @@ -45,10 +45,10 @@ #include "llvm/Pass.h" #include "llvm/Transforms/Utils/BasicBlockUtils.h" #include "llvm/Transforms/Utils/Local.h" +#include "llvm/TargetTransformInfo.h" #include "llvm/ADT/SmallVector.h" #include "llvm/ADT/Statistic.h" #include "llvm/Support/CommandLine.h" -#include "llvm/Target/TargetLowering.h" #include #include using namespace llvm; @@ -70,15 +70,14 @@ namespace { Constant *SetJmpFn, *LongJmpFn, *StackSaveFn, *StackRestoreFn; bool useExpensiveEHSupport; - // We peek in TLI to grab the target's jmp_buf size and alignment - const TargetLowering *TLI; + // We peek in STTI to grab the target's jmp_buf size and alignment + const ScalarTargetTransformInfo *STTI; public: static char ID; // Pass identification, replacement for typeid - explicit LowerInvoke(const TargetLowering *tli = NULL, - bool useExpensiveEHSupport = ExpensiveEHSupport) + explicit LowerInvoke(bool useExpensiveEHSupport = ExpensiveEHSupport) : FunctionPass(ID), useExpensiveEHSupport(useExpensiveEHSupport), - TLI(tli) { + STTI(0) { initializeLowerInvokePass(*PassRegistry::getPassRegistry()); } bool doInitialization(Module &M); @@ -108,21 +107,24 @@ INITIALIZE_PASS(LowerInvoke, "lowerinvoke", char &llvm::LowerInvokePassID = LowerInvoke::ID; // Public Interface To the LowerInvoke pass. -FunctionPass *llvm::createLowerInvokePass(const TargetLowering *TLI) { - return new LowerInvoke(TLI, ExpensiveEHSupport); +FunctionPass *llvm::createLowerInvokePass() { + return new LowerInvoke(ExpensiveEHSupport); } -FunctionPass *llvm::createLowerInvokePass(const TargetLowering *TLI, - bool useExpensiveEHSupport) { - return new LowerInvoke(TLI, useExpensiveEHSupport); +FunctionPass *llvm::createLowerInvokePass(bool useExpensiveEHSupport) { + return new LowerInvoke(useExpensiveEHSupport); } // doInitialization - Make sure that there is a prototype for abort in the // current module. bool LowerInvoke::doInitialization(Module &M) { + TargetTransformInfo *TTI = getAnalysisIfAvailable(); + if (TTI) + STTI = TTI->getScalarTargetTransformInfo(); + Type *VoidPtrTy = Type::getInt8PtrTy(M.getContext()); if (useExpensiveEHSupport) { // Insert a type for the linked list of jump buffers. - unsigned JBSize = TLI ? TLI->getJumpBufSize() : 0; + unsigned JBSize = STTI ? STTI->getJumpBufSize() : 0; JBSize = JBSize ? JBSize : 200; Type *JmpBufTy = ArrayType::get(VoidPtrTy, JBSize); @@ -430,7 +432,7 @@ bool LowerInvoke::insertExpensiveEHSupport(Function &F) { // Create an alloca for the incoming jump buffer ptr and the new jump buffer // that needs to be restored on all exits from the function. This is an // alloca because the value needs to be live across invokes. - unsigned Align = TLI ? TLI->getJumpBufAlignment() : 0; + unsigned Align = STTI ? STTI->getJumpBufAlignment() : 0; AllocaInst *JmpBuf = new AllocaInst(JBLinkTy, 0, Align, "jblink", F.begin()->begin()); @@ -575,6 +577,10 @@ bool LowerInvoke::insertExpensiveEHSupport(Function &F) { } bool LowerInvoke::runOnFunction(Function &F) { + TargetTransformInfo *TTI = getAnalysisIfAvailable(); + if (TTI) + STTI = TTI->getScalarTargetTransformInfo(); + if (useExpensiveEHSupport) return insertExpensiveEHSupport(F); else diff --git a/llvm/lib/VMCore/CMakeLists.txt b/llvm/lib/VMCore/CMakeLists.txt index 6c30967..ba807fc 100644 --- a/llvm/lib/VMCore/CMakeLists.txt +++ b/llvm/lib/VMCore/CMakeLists.txt @@ -33,6 +33,7 @@ add_llvm_library(LLVMCore PrintModulePass.cpp Type.cpp TypeFinder.cpp + TargetTransformInfo.cpp Use.cpp User.cpp Value.cpp diff --git a/llvm/lib/VMCore/TargetTransformInfo.cpp b/llvm/lib/VMCore/TargetTransformInfo.cpp new file mode 100644 index 0000000..3af0222 --- /dev/null +++ b/llvm/lib/VMCore/TargetTransformInfo.cpp @@ -0,0 +1,27 @@ +//===- llvm/VMCore/TargetTransformInfo.cpp ----------------------*- C++ -*-===// +// +// The LLVM Compiler Infrastructure +// +// This file is distributed under the University of Illinois Open Source +// License. See LICENSE.TXT for details. +// +//===----------------------------------------------------------------------===// + +#include "llvm/TargetTransformInfo.h" +#include "llvm/Support/ErrorHandling.h" + +using namespace llvm; + +/// Default ctor. +/// +/// @note This has to exist, because this is a pass, but it should never be +/// used. +TargetTransformInfo::TargetTransformInfo() : ImmutablePass(ID) { + report_fatal_error("Bad TargetTransformInfo ctor used. " + "Tool did not specify a TargetTransformInfo to use?"); +} + +INITIALIZE_PASS(TargetTransformInfo, "TargetTransformInfo", + "Target Transform Info", false, true) +char TargetTransformInfo::ID = 0; + diff --git a/llvm/tools/llc/llc.cpp b/llvm/tools/llc/llc.cpp index 04e5bca..4d4a74c 100644 --- a/llvm/tools/llc/llc.cpp +++ b/llvm/tools/llc/llc.cpp @@ -21,6 +21,7 @@ #include "llvm/ADT/Triple.h" #include "llvm/Assembly/PrintModulePass.h" #include "llvm/Support/IRReader.h" +#include "llvm/CodeGen/CommandFlags.h" #include "llvm/CodeGen/LinkAllAsmWriterComponents.h" #include "llvm/CodeGen/LinkAllCodegenComponents.h" #include "llvm/MC/SubtargetFeature.h" @@ -62,216 +63,13 @@ OptLevel("O", static cl::opt TargetTriple("mtriple", cl::desc("Override target triple for module")); -static cl::opt -MArch("march", cl::desc("Architecture to generate code for (see --version)")); - -static cl::opt -MCPU("mcpu", - cl::desc("Target a specific cpu type (-mcpu=help for details)"), - cl::value_desc("cpu-name"), - cl::init("")); - -static cl::list -MAttrs("mattr", - cl::CommaSeparated, - cl::desc("Target specific attributes (-mattr=help for details)"), - cl::value_desc("a1,+a2,-a3,...")); - -static cl::opt -RelocModel("relocation-model", - cl::desc("Choose relocation model"), - cl::init(Reloc::Default), - cl::values( - clEnumValN(Reloc::Default, "default", - "Target default relocation model"), - clEnumValN(Reloc::Static, "static", - "Non-relocatable code"), - clEnumValN(Reloc::PIC_, "pic", - "Fully relocatable, position independent code"), - clEnumValN(Reloc::DynamicNoPIC, "dynamic-no-pic", - "Relocatable external references, non-relocatable code"), - clEnumValEnd)); - -static cl::opt -CMModel("code-model", - cl::desc("Choose code model"), - cl::init(CodeModel::Default), - cl::values(clEnumValN(CodeModel::Default, "default", - "Target default code model"), - clEnumValN(CodeModel::Small, "small", - "Small code model"), - clEnumValN(CodeModel::Kernel, "kernel", - "Kernel code model"), - clEnumValN(CodeModel::Medium, "medium", - "Medium code model"), - clEnumValN(CodeModel::Large, "large", - "Large code model"), - clEnumValEnd)); - -static cl::opt -RelaxAll("mc-relax-all", - cl::desc("When used with filetype=obj, " - "relax all fixups in the emitted object file")); - -cl::opt -FileType("filetype", cl::init(TargetMachine::CGFT_AssemblyFile), - cl::desc("Choose a file type (not all types are supported by all targets):"), - cl::values( - clEnumValN(TargetMachine::CGFT_AssemblyFile, "asm", - "Emit an assembly ('.s') file"), - clEnumValN(TargetMachine::CGFT_ObjectFile, "obj", - "Emit a native object ('.o') file"), - clEnumValN(TargetMachine::CGFT_Null, "null", - "Emit nothing, for performance testing"), - clEnumValEnd)); - cl::opt NoVerify("disable-verify", cl::Hidden, cl::desc("Do not verify input module")); -cl::opt DisableDotLoc("disable-dot-loc", cl::Hidden, - cl::desc("Do not use .loc entries")); - -cl::opt DisableCFI("disable-cfi", cl::Hidden, - cl::desc("Do not use .cfi_* directives")); - -cl::opt EnableDwarfDirectory("enable-dwarf-directory", cl::Hidden, - cl::desc("Use .file directives with an explicit directory.")); - -static cl::opt -DisableRedZone("disable-red-zone", - cl::desc("Do not emit code that uses the red zone."), - cl::init(false)); - -static cl::opt -EnableFPMAD("enable-fp-mad", - cl::desc("Enable less precise MAD instructions to be generated"), - cl::init(false)); - -static cl::opt -DisableFPElim("disable-fp-elim", - cl::desc("Disable frame pointer elimination optimization"), - cl::init(false)); - -static cl::opt -DisableFPElimNonLeaf("disable-non-leaf-fp-elim", - cl::desc("Disable frame pointer elimination optimization for non-leaf funcs"), - cl::init(false)); - -static cl::opt -EnableUnsafeFPMath("enable-unsafe-fp-math", - cl::desc("Enable optimizations that may decrease FP precision"), - cl::init(false)); - -static cl::opt -EnableNoInfsFPMath("enable-no-infs-fp-math", - cl::desc("Enable FP math optimizations that assume no +-Infs"), - cl::init(false)); - -static cl::opt -EnableNoNaNsFPMath("enable-no-nans-fp-math", - cl::desc("Enable FP math optimizations that assume no NaNs"), - cl::init(false)); - -static cl::opt -EnableHonorSignDependentRoundingFPMath("enable-sign-dependent-rounding-fp-math", - cl::Hidden, - cl::desc("Force codegen to assume rounding mode can change dynamically"), - cl::init(false)); - -static cl::opt -GenerateSoftFloatCalls("soft-float", - cl::desc("Generate software floating point library calls"), - cl::init(false)); - -static cl::opt -FloatABIForCalls("float-abi", - cl::desc("Choose float ABI type"), - cl::init(FloatABI::Default), - cl::values( - clEnumValN(FloatABI::Default, "default", - "Target default float ABI type"), - clEnumValN(FloatABI::Soft, "soft", - "Soft float ABI (implied by -soft-float)"), - clEnumValN(FloatABI::Hard, "hard", - "Hard float ABI (uses FP registers)"), - clEnumValEnd)); - -static cl::opt -FuseFPOps("fp-contract", - cl::desc("Enable aggresive formation of fused FP ops"), - cl::init(FPOpFusion::Standard), - cl::values( - clEnumValN(FPOpFusion::Fast, "fast", - "Fuse FP ops whenever profitable"), - clEnumValN(FPOpFusion::Standard, "on", - "Only fuse 'blessed' FP ops."), - clEnumValN(FPOpFusion::Strict, "off", - "Only fuse FP ops when the result won't be effected."), - clEnumValEnd)); - -static cl::opt -DontPlaceZerosInBSS("nozero-initialized-in-bss", - cl::desc("Don't place zero-initialized symbols into bss section"), - cl::init(false)); - -static cl::opt +cl::opt DisableSimplifyLibCalls("disable-simplify-libcalls", - cl::desc("Disable simplify-libcalls"), - cl::init(false)); - -static cl::opt -EnableGuaranteedTailCallOpt("tailcallopt", - cl::desc("Turn fastcc calls into tail calls by (potentially) changing ABI."), - cl::init(false)); - -static cl::opt -DisableTailCalls("disable-tail-calls", - cl::desc("Never emit tail calls"), - cl::init(false)); - -static cl::opt -OverrideStackAlignment("stack-alignment", - cl::desc("Override default stack alignment"), - cl::init(0)); - -static cl::opt -EnableRealignStack("realign-stack", - cl::desc("Realign stack if needed"), - cl::init(true)); - -static cl::opt -TrapFuncName("trap-func", cl::Hidden, - cl::desc("Emit a call to trap function rather than a trap instruction"), - cl::init("")); - -static cl::opt -EnablePIE("enable-pie", - cl::desc("Assume the creation of a position independent executable."), - cl::init(false)); - -static cl::opt -SegmentedStacks("segmented-stacks", - cl::desc("Use segmented stacks if possible."), - cl::init(false)); - -static cl::opt -UseInitArray("use-init-array", - cl::desc("Use .init_array instead of .ctors."), - cl::init(false)); - -static cl::opt StopAfter("stop-after", - cl::desc("Stop compilation after a specific pass"), - cl::value_desc("pass-name"), - cl::init("")); -static cl::opt StartAfter("start-after", - cl::desc("Resume compilation after a specific pass"), - cl::value_desc("pass-name"), - cl::init("")); - -static cl::opt -SSPBufferSize("stack-protector-buffer-size", cl::init(8), - cl::desc("Lower bound for a buffer to be considered for " - "stack protection")); + cl::desc("Disable simplify-libcalls"), + cl::init(false)); // GetFileNameRoot - Helper function to get the basename of a filename. static inline std::string @@ -505,6 +303,11 @@ int main(int argc, char **argv) { TLI->disableAllFunctions(); PM.add(TLI); + if (target.get()) { + PM.add(new TargetTransformInfo(target->getScalarTargetTransformInfo(), + target->getVectorTargetTransformInfo())); + } + // Add the target data from the target machine, if it exists, or the module. if (const DataLayout *TD = Target.getDataLayout()) PM.add(new DataLayout(*TD)); diff --git a/llvm/tools/opt/CMakeLists.txt b/llvm/tools/opt/CMakeLists.txt index 7daf22a..32de6d4 100644 --- a/llvm/tools/opt/CMakeLists.txt +++ b/llvm/tools/opt/CMakeLists.txt @@ -1,4 +1,4 @@ -set(LLVM_LINK_COMPONENTS bitreader asmparser bitwriter instrumentation scalaropts ipo vectorize) +set(LLVM_LINK_COMPONENTS ${LLVM_TARGETS_TO_BUILD} bitreader asmparser bitwriter instrumentation scalaropts ipo vectorize) add_llvm_tool(opt AnalysisWrappers.cpp diff --git a/llvm/tools/opt/LLVMBuild.txt b/llvm/tools/opt/LLVMBuild.txt index 4de99f5..b174431 100644 --- a/llvm/tools/opt/LLVMBuild.txt +++ b/llvm/tools/opt/LLVMBuild.txt @@ -19,4 +19,4 @@ type = Tool name = opt parent = Tools -required_libraries = AsmParser BitReader BitWriter IPO Instrumentation Scalar +required_libraries = AsmParser BitReader BitWriter IPO Instrumentation Scalar all-targets diff --git a/llvm/tools/opt/Makefile b/llvm/tools/opt/Makefile index 16d116d..ee7e1cf 100644 --- a/llvm/tools/opt/Makefile +++ b/llvm/tools/opt/Makefile @@ -9,6 +9,6 @@ LEVEL := ../.. TOOLNAME := opt -LINK_COMPONENTS := bitreader bitwriter asmparser instrumentation scalaropts ipo vectorize +LINK_COMPONENTS := bitreader bitwriter asmparser instrumentation scalaropts ipo vectorize all-targets include $(LEVEL)/Makefile.common diff --git a/llvm/tools/opt/opt.cpp b/llvm/tools/opt/opt.cpp index 706a7d5..8d8d731 100644 --- a/llvm/tools/opt/opt.cpp +++ b/llvm/tools/opt/opt.cpp @@ -18,6 +18,7 @@ #include "llvm/Module.h" #include "llvm/PassManager.h" #include "llvm/CallGraphSCCPass.h" +#include "llvm/CodeGen/CommandFlags.h" #include "llvm/Bitcode/ReaderWriter.h" #include "llvm/Assembly/PrintModulePass.h" #include "llvm/Analysis/Verifier.h" @@ -36,7 +37,9 @@ #include "llvm/Support/PluginLoader.h" #include "llvm/Support/PrettyStackTrace.h" #include "llvm/Support/SystemUtils.h" +#include "llvm/Support/TargetRegistry.h" #include "llvm/Support/ToolOutputFile.h" +#include "llvm/MC/SubtargetFeature.h" #include "llvm/LinkAllPasses.h" #include "llvm/LinkAllVMCore.h" #include "llvm/Transforms/IPO/PassManagerBuilder.h" @@ -478,6 +481,75 @@ static void AddStandardLinkPasses(PassManagerBase &PM) { /*RunInliner=*/ !DisableInline); } +//===----------------------------------------------------------------------===// +// CodeGen-related helper functions. +// +static TargetOptions GetTargetOptions() { + TargetOptions Options; + Options.LessPreciseFPMADOption = EnableFPMAD; + Options.NoFramePointerElim = DisableFPElim; + Options.NoFramePointerElimNonLeaf = DisableFPElimNonLeaf; + Options.AllowFPOpFusion = FuseFPOps; + Options.UnsafeFPMath = EnableUnsafeFPMath; + Options.NoInfsFPMath = EnableNoInfsFPMath; + Options.NoNaNsFPMath = EnableNoNaNsFPMath; + Options.HonorSignDependentRoundingFPMathOption = + EnableHonorSignDependentRoundingFPMath; + Options.UseSoftFloat = GenerateSoftFloatCalls; + if (FloatABIForCalls != FloatABI::Default) + Options.FloatABIType = FloatABIForCalls; + Options.NoZerosInBSS = DontPlaceZerosInBSS; + Options.GuaranteedTailCallOpt = EnableGuaranteedTailCallOpt; + Options.DisableTailCalls = DisableTailCalls; + Options.StackAlignmentOverride = OverrideStackAlignment; + Options.RealignStack = EnableRealignStack; + Options.TrapFuncName = TrapFuncName; + Options.PositionIndependentExecutable = EnablePIE; + Options.EnableSegmentedStacks = SegmentedStacks; + Options.UseInitArray = UseInitArray; + Options.SSPBufferSize = SSPBufferSize; + return Options; +} + +CodeGenOpt::Level GetCodeGenOptLevel() { + if (OptLevelO1) + return CodeGenOpt::Less; + if (OptLevelO2) + return CodeGenOpt::Default; + if (OptLevelO3) + return CodeGenOpt::Aggressive; + return CodeGenOpt::None; +} + +// Returns the TargetMachine instance or zero if no triple is provided. +static TargetMachine* GetTargetMachine(std::string TripleStr) { + if (TripleStr.empty()) + return 0; + + // Get the target specific parser. + std::string Error; + Triple TheTriple(Triple::normalize(TargetTriple)); + + const Target *TheTarget = TargetRegistry::lookupTarget(MArch, TheTriple, + Error); + if (!TheTarget) { + return 0; + } + + // Package up features to be passed to target/subtarget + std::string FeaturesStr; + if (MAttrs.size()) { + SubtargetFeatures Features; + for (unsigned i = 0; i != MAttrs.size(); ++i) + Features.AddFeature(MAttrs[i]); + FeaturesStr = Features.getString(); + } + + return TheTarget->createTargetMachine(TheTriple.getTriple(), + MCPU, FeaturesStr, GetTargetOptions(), + RelocModel, CMModel, + GetCodeGenOptLevel()); +} //===----------------------------------------------------------------------===// // main for opt @@ -579,6 +651,12 @@ int main(int argc, char **argv) { if (TD) Passes.add(TD); + std::auto_ptr TM(GetTargetMachine(TargetTriple)); + if (TM.get()) { + Passes.add(new TargetTransformInfo(TM->getScalarTargetTransformInfo(), + TM->getVectorTargetTransformInfo())); + } + OwningPtr FPasses; if (OptLevelO1 || OptLevelO2 || OptLevelOs || OptLevelOz || OptLevelO3) { FPasses.reset(new FunctionPassManager(M.get())); -- 2.7.4