diff options
Diffstat (limited to 'contrib/llvm/lib/Target/ARM/ARMTargetMachine.cpp')
-rw-r--r-- | contrib/llvm/lib/Target/ARM/ARMTargetMachine.cpp | 188 |
1 files changed, 128 insertions, 60 deletions
diff --git a/contrib/llvm/lib/Target/ARM/ARMTargetMachine.cpp b/contrib/llvm/lib/Target/ARM/ARMTargetMachine.cpp index 7a8181b..0aceaed 100644 --- a/contrib/llvm/lib/Target/ARM/ARMTargetMachine.cpp +++ b/contrib/llvm/lib/Target/ARM/ARMTargetMachine.cpp @@ -14,10 +14,11 @@ #include "ARMFrameLowering.h" #include "ARMTargetMachine.h" #include "ARMTargetObjectFile.h" +#include "ARMTargetTransformInfo.h" #include "llvm/CodeGen/Passes.h" #include "llvm/IR/Function.h" +#include "llvm/IR/LegacyPassManager.h" #include "llvm/MC/MCAsmInfo.h" -#include "llvm/PassManager.h" #include "llvm/Support/CommandLine.h" #include "llvm/Support/FormattedStream.h" #include "llvm/Support/TargetRegistry.h" @@ -36,6 +37,16 @@ EnableAtomicTidy("arm-atomic-cfg-tidy", cl::Hidden, " to make use of cmpxchg flow-based information"), cl::init(true)); +static cl::opt<bool> +EnableARMLoadStoreOpt("arm-load-store-opt", cl::Hidden, + cl::desc("Enable ARM load/store optimization pass"), + cl::init(true)); + +// FIXME: Unify control over GlobalMerge. +static cl::opt<cl::boolOrDefault> +EnableGlobalMerge("arm-global-merge", cl::Hidden, + cl::desc("Enable the global merge pass")); + extern "C" void LLVMInitializeARMTarget() { // Register the target. RegisterTargetMachine<ARMLETargetMachine> X(TheARMLETarget); @@ -104,6 +115,60 @@ computeTargetABI(const Triple &TT, StringRef CPU, return TargetABI; } +static std::string computeDataLayout(StringRef TT, StringRef CPU, + const TargetOptions &Options, + bool isLittle) { + const Triple Triple(TT); + auto ABI = computeTargetABI(Triple, CPU, Options); + std::string Ret = ""; + + if (isLittle) + // Little endian. + Ret += "e"; + else + // Big endian. + Ret += "E"; + + Ret += DataLayout::getManglingComponent(Triple); + + // Pointers are 32 bits and aligned to 32 bits. + Ret += "-p:32:32"; + + // ABIs other than APCS have 64 bit integers with natural alignment. + if (ABI != ARMBaseTargetMachine::ARM_ABI_APCS) + Ret += "-i64:64"; + + // We have 64 bits floats. The APCS ABI requires them to be aligned to 32 + // bits, others to 64 bits. We always try to align to 64 bits. + if (ABI == ARMBaseTargetMachine::ARM_ABI_APCS) + Ret += "-f64:32:64"; + + // We have 128 and 64 bit vectors. The APCS ABI aligns them to 32 bits, others + // to 64. We always ty to give them natural alignment. + if (ABI == ARMBaseTargetMachine::ARM_ABI_APCS) + Ret += "-v64:32:64-v128:32:128"; + else + Ret += "-v128:64:128"; + + // Try to align aggregates to 32 bits (the default is 64 bits, which has no + // particular hardware support on 32-bit ARM). + Ret += "-a:0:32"; + + // Integer registers are 32 bits. + Ret += "-n32"; + + // The stack is 128 bit aligned on NaCl, 64 bit aligned on AAPCS and 32 bit + // aligned everywhere else. + if (Triple.isOSNaCl()) + Ret += "-S128"; + else if (ABI == ARMBaseTargetMachine::ARM_ABI_AAPCS) + Ret += "-S64"; + else + Ret += "-S32"; + + return Ret; +} + /// TargetMachine ctor - Create an ARM architecture model. /// ARMBaseTargetMachine::ARMBaseTargetMachine(const Target &T, StringRef TT, @@ -111,7 +176,8 @@ ARMBaseTargetMachine::ARMBaseTargetMachine(const Target &T, StringRef TT, const TargetOptions &Options, Reloc::Model RM, CodeModel::Model CM, CodeGenOpt::Level OL, bool isLittle) - : LLVMTargetMachine(T, TT, CPU, FS, Options, RM, CM, OL), + : LLVMTargetMachine(T, computeDataLayout(TT, CPU, Options, isLittle), TT, + CPU, FS, Options, RM, CM, OL), TargetABI(computeTargetABI(Triple(TT), CPU, Options)), TLOF(createTLOF(Triple(getTargetTriple()))), Subtarget(TT, CPU, FS, *this, isLittle), isLittle(isLittle) { @@ -126,11 +192,8 @@ ARMBaseTargetMachine::~ARMBaseTargetMachine() {} const ARMSubtarget * ARMBaseTargetMachine::getSubtargetImpl(const Function &F) const { - AttributeSet FnAttrs = F.getAttributes(); - Attribute CPUAttr = - FnAttrs.getAttribute(AttributeSet::FunctionIndex, "target-cpu"); - Attribute FSAttr = - FnAttrs.getAttribute(AttributeSet::FunctionIndex, "target-features"); + Attribute CPUAttr = F.getFnAttribute("target-cpu"); + Attribute FSAttr = F.getFnAttribute("target-features"); std::string CPU = !CPUAttr.hasAttribute(Attribute::None) ? CPUAttr.getValueAsString().str() @@ -144,14 +207,15 @@ ARMBaseTargetMachine::getSubtargetImpl(const Function &F) const { // function before we can generate a subtarget. We also need to use // it as a key for the subtarget since that can be the only difference // between two functions. - Attribute SFAttr = - FnAttrs.getAttribute(AttributeSet::FunctionIndex, "use-soft-float"); - bool SoftFloat = !SFAttr.hasAttribute(Attribute::None) - ? SFAttr.getValueAsString() == "true" - : Options.UseSoftFloat; - - auto &I = SubtargetMap[CPU + FS + (SoftFloat ? "use-soft-float=true" - : "use-soft-float=false")]; + bool SoftFloat = + F.hasFnAttribute("use-soft-float") && + F.getFnAttribute("use-soft-float").getValueAsString() == "true"; + // If the soft float attribute is set on the function turn on the soft float + // subtarget feature. + if (SoftFloat) + FS += FS.empty() ? "+soft-float" : ",+soft-float"; + + auto &I = SubtargetMap[CPU + FS]; if (!I) { // This needs to be done before we create a new subtarget since any // creation will depend on the TM and the code generation flags on the @@ -162,12 +226,9 @@ ARMBaseTargetMachine::getSubtargetImpl(const Function &F) const { return I.get(); } -void ARMBaseTargetMachine::addAnalysisPasses(PassManagerBase &PM) { - // Add first the target-independent BasicTTI pass, then our ARM pass. This - // allows the ARM pass to delegate to the target independent layer when - // appropriate. - PM.add(createBasicTargetTransformInfoPass(this)); - PM.add(createARMTargetTransformInfoPass(this)); +TargetIRAnalysis ARMBaseTargetMachine::getTargetIRAnalysis() { + return TargetIRAnalysis( + [this](Function &F) { return TargetTransformInfo(ARMTTIImpl(this, F)); }); } @@ -243,10 +304,6 @@ public: return getTM<ARMBaseTargetMachine>(); } - const ARMSubtarget &getARMSubtarget() const { - return *getARMTargetMachine().getSubtargetImpl(); - } - void addIRPasses() override; bool addPreISel() override; bool addInstSelector() override; @@ -269,17 +326,28 @@ void ARMPassConfig::addIRPasses() { // Cmpxchg instructions are often used with a subsequent comparison to // determine whether it succeeded. We can exploit existing control-flow in // ldrex/strex loops to simplify this, but it needs tidying up. - const ARMSubtarget *Subtarget = &getARMSubtarget(); - if (Subtarget->hasAnyDataBarrier() && !Subtarget->isThumb1Only()) - if (TM->getOptLevel() != CodeGenOpt::None && EnableAtomicTidy) - addPass(createCFGSimplificationPass()); + if (TM->getOptLevel() != CodeGenOpt::None && EnableAtomicTidy) + addPass(createCFGSimplificationPass(-1, [this](const Function &F) { + const auto &ST = this->TM->getSubtarget<ARMSubtarget>(F); + return ST.hasAnyDataBarrier() && !ST.isThumb1Only(); + })); TargetPassConfig::addIRPasses(); } bool ARMPassConfig::addPreISel() { - if (TM->getOptLevel() != CodeGenOpt::None) - addPass(createGlobalMergePass(TM)); + if ((TM->getOptLevel() != CodeGenOpt::None && + EnableGlobalMerge == cl::BOU_UNSET) || + EnableGlobalMerge == cl::BOU_TRUE) { + // FIXME: This is using the thumb1 only constant value for + // maximal global offset for merging globals. We may want + // to look into using the old value for non-thumb1 code of + // 4095 based on the TargetMachine, but this starts to become + // tricky when doing code gen per function. + bool OnlyOptimizeForSize = (TM->getOptLevel() < CodeGenOpt::Aggressive) && + (EnableGlobalMerge == cl::BOU_UNSET); + addPass(createGlobalMergePass(TM, 127, OnlyOptimizeForSize)); + } return false; } @@ -287,32 +355,30 @@ bool ARMPassConfig::addPreISel() { bool ARMPassConfig::addInstSelector() { addPass(createARMISelDag(getARMTargetMachine(), getOptLevel())); - const ARMSubtarget *Subtarget = &getARMSubtarget(); - if (Subtarget->isTargetELF() && !Subtarget->isThumb1Only() && + if (Triple(TM->getTargetTriple()).isOSBinFormatELF() && TM->Options.EnableFastISel) addPass(createARMGlobalBaseRegPass()); return false; } void ARMPassConfig::addPreRegAlloc() { - if (getOptLevel() != CodeGenOpt::None) - addPass(createARMLoadStoreOptimizationPass(true)); - if (getOptLevel() != CodeGenOpt::None && getARMSubtarget().isCortexA9()) + if (getOptLevel() != CodeGenOpt::None) { addPass(createMLxExpansionPass()); - // Since the A15SDOptimizer pass can insert VDUP instructions, it can only be - // enabled when NEON is available. - if (getOptLevel() != CodeGenOpt::None && getARMSubtarget().isCortexA15() && - getARMSubtarget().hasNEON() && !DisableA15SDOptimization) { - addPass(createA15SDOptimizerPass()); + + if (EnableARMLoadStoreOpt) + addPass(createARMLoadStoreOptimizationPass(/* pre-register alloc */ true)); + + if (!DisableA15SDOptimization) + addPass(createA15SDOptimizerPass()); } } void ARMPassConfig::addPreSched2() { if (getOptLevel() != CodeGenOpt::None) { - addPass(createARMLoadStoreOptimizationPass()); + if (EnableARMLoadStoreOpt) + addPass(createARMLoadStoreOptimizationPass()); - if (getARMSubtarget().hasNEON()) - addPass(createExecutionDependencyFixPass(&ARM::DPRRegClass)); + addPass(createExecutionDependencyFixPass(&ARM::DPRRegClass)); } // Expand some pseudo instructions into multiple instructions to allow @@ -320,27 +386,29 @@ void ARMPassConfig::addPreSched2() { addPass(createARMExpandPseudoPass()); if (getOptLevel() != CodeGenOpt::None) { - if (!getARMSubtarget().isThumb1Only()) { - // in v8, IfConversion depends on Thumb instruction widths - if (getARMSubtarget().restrictIT() && - !getARMSubtarget().prefers32BitThumb()) - addPass(createThumb2SizeReductionPass()); - addPass(&IfConverterID); - } + // in v8, IfConversion depends on Thumb instruction widths + addPass(createThumb2SizeReductionPass([this](const Function &F) { + return this->TM->getSubtarget<ARMSubtarget>(F).restrictIT(); + })); + + addPass(createIfConverter([this](const Function &F) { + return !this->TM->getSubtarget<ARMSubtarget>(F).isThumb1Only(); + })); } - if (getARMSubtarget().isThumb2()) - addPass(createThumb2ITBlockPass()); + addPass(createThumb2ITBlockPass()); } void ARMPassConfig::addPreEmitPass() { - if (getARMSubtarget().isThumb2()) { - if (!getARMSubtarget().prefers32BitThumb()) - addPass(createThumb2SizeReductionPass()); + addPass(createThumb2SizeReductionPass()); - // Constant island pass work on unbundled instructions. - addPass(&UnpackMachineBundlesID); - } + // Constant island pass work on unbundled instructions. + addPass(createUnpackMachineBundles([this](const Function &F) { + return this->TM->getSubtarget<ARMSubtarget>(F).isThumb2(); + })); + + // Don't optimize barriers at -O0. + if (getOptLevel() != CodeGenOpt::None) + addPass(createARMOptimizeBarriersPass()); - addPass(createARMOptimizeBarriersPass()); addPass(createARMConstantIslandPass()); } |