X-Git-Url: http://plrg.eecs.uci.edu/git/?a=blobdiff_plain;f=lib%2FTarget%2FARM%2FARMBaseRegisterInfo.cpp;h=2608685e045f595351e46cc3225cd1d04108bf48;hb=90f20044ade3712c8b0c3f4ebe47d57ad15ae6ce;hp=b2dad50b65d58580239bd631d7c7491c4dae881f;hpb=33464912237efaa0ed7060829e66b59055bdd48b;p=oota-llvm.git diff --git a/lib/Target/ARM/ARMBaseRegisterInfo.cpp b/lib/Target/ARM/ARMBaseRegisterInfo.cpp index b2dad50b65d..2608685e045 100644 --- a/lib/Target/ARM/ARMBaseRegisterInfo.cpp +++ b/lib/Target/ARM/ARMBaseRegisterInfo.cpp @@ -1,4 +1,4 @@ -//===- ARMBaseRegisterInfo.cpp - ARM Register Information -------*- C++ -*-===// +//===-- ARMBaseRegisterInfo.cpp - ARM Register Information ----------------===// // // The LLVM Compiler Infrastructure // @@ -12,12 +12,13 @@ //===----------------------------------------------------------------------===// #include "ARM.h" -#include "ARMAddressingModes.h" #include "ARMBaseInstrInfo.h" #include "ARMBaseRegisterInfo.h" +#include "ARMFrameLowering.h" #include "ARMInstrInfo.h" #include "ARMMachineFunctionInfo.h" #include "ARMSubtarget.h" +#include "MCTargetDesc/ARMAddressingModes.h" #include "llvm/Constants.h" #include "llvm/DerivedTypes.h" #include "llvm/Function.h" @@ -26,86 +27,78 @@ #include "llvm/CodeGen/MachineFrameInfo.h" #include "llvm/CodeGen/MachineFunction.h" #include "llvm/CodeGen/MachineInstrBuilder.h" -#include "llvm/CodeGen/MachineLocation.h" #include "llvm/CodeGen/MachineRegisterInfo.h" #include "llvm/CodeGen/RegisterScavenging.h" #include "llvm/Support/Debug.h" #include "llvm/Support/ErrorHandling.h" #include "llvm/Support/raw_ostream.h" -#include "llvm/Target/TargetFrameInfo.h" +#include "llvm/Target/TargetFrameLowering.h" #include "llvm/Target/TargetMachine.h" #include "llvm/Target/TargetOptions.h" #include "llvm/ADT/BitVector.h" #include "llvm/ADT/SmallVector.h" #include "llvm/Support/CommandLine.h" -namespace llvm { +#define GET_REGINFO_TARGET_DESC +#include "ARMGenRegisterInfo.inc" + +using namespace llvm; + static cl::opt ForceAllBaseRegAlloc("arm-force-base-reg-alloc", cl::Hidden, cl::init(false), cl::desc("Force use of virtual base registers for stack load/store")); static cl::opt EnableLocalStackAlloc("enable-local-stack-alloc", cl::init(true), cl::Hidden, cl::desc("Enable pre-regalloc stack frame index allocation")); -} - -using namespace llvm; - static cl::opt EnableBasePointer("arm-use-base-pointer", cl::Hidden, cl::init(true), cl::desc("Enable use of a base pointer for complex stack frames")); - ARMBaseRegisterInfo::ARMBaseRegisterInfo(const ARMBaseInstrInfo &tii, const ARMSubtarget &sti) - : ARMGenRegisterInfo(ARM::ADJCALLSTACKDOWN, ARM::ADJCALLSTACKUP), - TII(tii), STI(sti), + : ARMGenRegisterInfo(ARM::LR), TII(tii), STI(sti), FramePtr((STI.isTargetDarwin() || STI.isThumb()) ? ARM::R7 : ARM::R11), BasePtr(ARM::R6) { } const unsigned* ARMBaseRegisterInfo::getCalleeSavedRegs(const MachineFunction *MF) const { - static const unsigned CalleeSavedRegs[] = { - ARM::LR, ARM::R11, ARM::R10, ARM::R9, ARM::R8, - ARM::R7, ARM::R6, ARM::R5, ARM::R4, - - ARM::D15, ARM::D14, ARM::D13, ARM::D12, - ARM::D11, ARM::D10, ARM::D9, ARM::D8, - 0 - }; - - static const unsigned DarwinCalleeSavedRegs[] = { - // Darwin ABI deviates from ARM standard ABI. R9 is not a callee-saved - // register. - ARM::LR, ARM::R7, ARM::R6, ARM::R5, ARM::R4, - ARM::R11, ARM::R10, ARM::R8, + return (STI.isTargetIOS()) ? CSR_iOS_SaveList : CSR_AAPCS_SaveList; +} - ARM::D15, ARM::D14, ARM::D13, ARM::D12, - ARM::D11, ARM::D10, ARM::D9, ARM::D8, - 0 - }; - return STI.isTargetDarwin() ? DarwinCalleeSavedRegs : CalleeSavedRegs; +const uint32_t* +ARMBaseRegisterInfo::getCallPreservedMask(CallingConv::ID) const { + return (STI.isTargetIOS()) ? CSR_iOS_RegMask : CSR_AAPCS_RegMask; } BitVector ARMBaseRegisterInfo:: getReservedRegs(const MachineFunction &MF) const { - // FIXME: avoid re-calculating this everytime. + const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering(); + + // FIXME: avoid re-calculating this every time. BitVector Reserved(getNumRegs()); Reserved.set(ARM::SP); Reserved.set(ARM::PC); - Reserved.set(ARM::FPSCR); - if (hasFP(MF)) + if (TFI->hasFP(MF)) Reserved.set(FramePtr); if (hasBasePointer(MF)) Reserved.set(BasePtr); // Some targets reserve R9. if (STI.isR9Reserved()) Reserved.set(ARM::R9); + // Reserve D16-D31 if the subtarget doesn't support them. + if (!STI.hasVFP3() || STI.hasD16()) { + assert(ARM::D31 == ARM::D16 + 15); + for (unsigned i = 0; i != 16; ++i) + Reserved.set(ARM::D16 + i); + } return Reserved; } bool ARMBaseRegisterInfo::isReservedReg(const MachineFunction &MF, unsigned Reg) const { + const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering(); + switch (Reg) { default: break; case ARM::SP: @@ -117,7 +110,7 @@ bool ARMBaseRegisterInfo::isReservedReg(const MachineFunction &MF, break; case ARM::R7: case ARM::R11: - if (FramePtr == Reg && hasFP(MF)) + if (FramePtr == Reg && TFI->hasFP(MF)) return true; break; case ARM::R9: @@ -127,104 +120,6 @@ bool ARMBaseRegisterInfo::isReservedReg(const MachineFunction &MF, return false; } -const TargetRegisterClass * -ARMBaseRegisterInfo::getMatchingSuperRegClass(const TargetRegisterClass *A, - const TargetRegisterClass *B, - unsigned SubIdx) const { - switch (SubIdx) { - default: return 0; - case ARM::ssub_0: - case ARM::ssub_1: - case ARM::ssub_2: - case ARM::ssub_3: { - // S sub-registers. - if (A->getSize() == 8) { - if (B == &ARM::SPR_8RegClass) - return &ARM::DPR_8RegClass; - assert(B == &ARM::SPRRegClass && "Expecting SPR register class!"); - if (A == &ARM::DPR_8RegClass) - return A; - return &ARM::DPR_VFP2RegClass; - } - - if (A->getSize() == 16) { - if (B == &ARM::SPR_8RegClass) - return &ARM::QPR_8RegClass; - return &ARM::QPR_VFP2RegClass; - } - - if (A->getSize() == 32) { - if (B == &ARM::SPR_8RegClass) - return 0; // Do not allow coalescing! - return &ARM::QQPR_VFP2RegClass; - } - - assert(A->getSize() == 64 && "Expecting a QQQQ register class!"); - return 0; // Do not allow coalescing! - } - case ARM::dsub_0: - case ARM::dsub_1: - case ARM::dsub_2: - case ARM::dsub_3: { - // D sub-registers. - if (A->getSize() == 16) { - if (B == &ARM::DPR_VFP2RegClass) - return &ARM::QPR_VFP2RegClass; - if (B == &ARM::DPR_8RegClass) - return 0; // Do not allow coalescing! - return A; - } - - if (A->getSize() == 32) { - if (B == &ARM::DPR_VFP2RegClass) - return &ARM::QQPR_VFP2RegClass; - if (B == &ARM::DPR_8RegClass) - return 0; // Do not allow coalescing! - return A; - } - - assert(A->getSize() == 64 && "Expecting a QQQQ register class!"); - if (B != &ARM::DPRRegClass) - return 0; // Do not allow coalescing! - return A; - } - case ARM::dsub_4: - case ARM::dsub_5: - case ARM::dsub_6: - case ARM::dsub_7: { - // D sub-registers of QQQQ registers. - if (A->getSize() == 64 && B == &ARM::DPRRegClass) - return A; - return 0; // Do not allow coalescing! - } - - case ARM::qsub_0: - case ARM::qsub_1: { - // Q sub-registers. - if (A->getSize() == 32) { - if (B == &ARM::QPR_VFP2RegClass) - return &ARM::QQPR_VFP2RegClass; - if (B == &ARM::QPR_8RegClass) - return 0; // Do not allow coalescing! - return A; - } - - assert(A->getSize() == 64 && "Expecting a QQQQ register class!"); - if (B == &ARM::QPRRegClass) - return A; - return 0; // Do not allow coalescing! - } - case ARM::qsub_2: - case ARM::qsub_3: { - // Q sub-registers of QQQQ registers. - if (A->getSize() == 64 && B == &ARM::QPRRegClass) - return A; - return 0; // Do not allow coalescing! - } - } - return 0; -} - bool ARMBaseRegisterInfo::canCombineSubRegIndices(const TargetRegisterClass *RC, SmallVectorImpl &SubIndices, @@ -341,18 +236,65 @@ ARMBaseRegisterInfo::canCombineSubRegIndices(const TargetRegisterClass *RC, return false; } +const TargetRegisterClass* +ARMBaseRegisterInfo::getLargestLegalSuperClass(const TargetRegisterClass *RC) + const { + const TargetRegisterClass *Super = RC; + TargetRegisterClass::sc_iterator I = RC->getSuperClasses(); + do { + switch (Super->getID()) { + case ARM::GPRRegClassID: + case ARM::SPRRegClassID: + case ARM::DPRRegClassID: + case ARM::QPRRegClassID: + case ARM::QQPRRegClassID: + case ARM::QQQQPRRegClassID: + return Super; + } + Super = *I++; + } while (Super); + return RC; +} const TargetRegisterClass * ARMBaseRegisterInfo::getPointerRegClass(unsigned Kind) const { return ARM::GPRRegisterClass; } -/// getAllocationOrder - Returns the register allocation order for a specified -/// register class in the form of a pair of TargetRegisterClass iterators. -std::pair -ARMBaseRegisterInfo::getAllocationOrder(const TargetRegisterClass *RC, - unsigned HintType, unsigned HintReg, - const MachineFunction &MF) const { +const TargetRegisterClass * +ARMBaseRegisterInfo::getCrossCopyRegClass(const TargetRegisterClass *RC) const { + if (RC == &ARM::CCRRegClass) + return 0; // Can't copy CCR registers. + return RC; +} + +unsigned +ARMBaseRegisterInfo::getRegPressureLimit(const TargetRegisterClass *RC, + MachineFunction &MF) const { + const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering(); + + switch (RC->getID()) { + default: + return 0; + case ARM::tGPRRegClassID: + return TFI->hasFP(MF) ? 4 : 5; + case ARM::GPRRegClassID: { + unsigned FP = TFI->hasFP(MF) ? 1 : 0; + return 10 - FP - (STI.isR9Reserved() ? 1 : 0); + } + case ARM::SPRRegClassID: // Currently not used as 'rep' register class. + case ARM::DPRRegClassID: + return 32 - 10; + } +} + +/// getRawAllocationOrder - Returns the register allocation order for a +/// specified register class with a target-dependent hint. +ArrayRef +ARMBaseRegisterInfo::getRawAllocationOrder(const TargetRegisterClass *RC, + unsigned HintType, unsigned HintReg, + const MachineFunction &MF) const { + const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering(); // Alternative register allocation orders when favoring even / odd registers // of register pairs. @@ -426,68 +368,56 @@ ARMBaseRegisterInfo::getAllocationOrder(const TargetRegisterClass *RC, ARM::R0, ARM::R2, ARM::R10,ARM::R12,ARM::LR, ARM::R4, ARM::R6, ARM::R8 }; + // We only support even/odd hints for GPR and rGPR. + if (RC != ARM::GPRRegisterClass && RC != ARM::rGPRRegisterClass) + return RC->getRawAllocationOrder(MF); if (HintType == ARMRI::RegPairEven) { if (isPhysicalRegister(HintReg) && getRegisterPairEven(HintReg, MF) == 0) // It's no longer possible to fulfill this hint. Return the default // allocation order. - return std::make_pair(RC->allocation_order_begin(MF), - RC->allocation_order_end(MF)); + return RC->getRawAllocationOrder(MF); - if (!hasFP(MF)) { + if (!TFI->hasFP(MF)) { if (!STI.isR9Reserved()) - return std::make_pair(GPREven1, - GPREven1 + (sizeof(GPREven1)/sizeof(unsigned))); + return makeArrayRef(GPREven1); else - return std::make_pair(GPREven4, - GPREven4 + (sizeof(GPREven4)/sizeof(unsigned))); + return makeArrayRef(GPREven4); } else if (FramePtr == ARM::R7) { if (!STI.isR9Reserved()) - return std::make_pair(GPREven2, - GPREven2 + (sizeof(GPREven2)/sizeof(unsigned))); + return makeArrayRef(GPREven2); else - return std::make_pair(GPREven5, - GPREven5 + (sizeof(GPREven5)/sizeof(unsigned))); + return makeArrayRef(GPREven5); } else { // FramePtr == ARM::R11 if (!STI.isR9Reserved()) - return std::make_pair(GPREven3, - GPREven3 + (sizeof(GPREven3)/sizeof(unsigned))); + return makeArrayRef(GPREven3); else - return std::make_pair(GPREven6, - GPREven6 + (sizeof(GPREven6)/sizeof(unsigned))); + return makeArrayRef(GPREven6); } } else if (HintType == ARMRI::RegPairOdd) { if (isPhysicalRegister(HintReg) && getRegisterPairOdd(HintReg, MF) == 0) // It's no longer possible to fulfill this hint. Return the default // allocation order. - return std::make_pair(RC->allocation_order_begin(MF), - RC->allocation_order_end(MF)); + return RC->getRawAllocationOrder(MF); - if (!hasFP(MF)) { + if (!TFI->hasFP(MF)) { if (!STI.isR9Reserved()) - return std::make_pair(GPROdd1, - GPROdd1 + (sizeof(GPROdd1)/sizeof(unsigned))); + return makeArrayRef(GPROdd1); else - return std::make_pair(GPROdd4, - GPROdd4 + (sizeof(GPROdd4)/sizeof(unsigned))); + return makeArrayRef(GPROdd4); } else if (FramePtr == ARM::R7) { if (!STI.isR9Reserved()) - return std::make_pair(GPROdd2, - GPROdd2 + (sizeof(GPROdd2)/sizeof(unsigned))); + return makeArrayRef(GPROdd2); else - return std::make_pair(GPROdd5, - GPROdd5 + (sizeof(GPROdd5)/sizeof(unsigned))); + return makeArrayRef(GPROdd5); } else { // FramePtr == ARM::R11 if (!STI.isR9Reserved()) - return std::make_pair(GPROdd3, - GPROdd3 + (sizeof(GPROdd3)/sizeof(unsigned))); + return makeArrayRef(GPROdd3); else - return std::make_pair(GPROdd6, - GPROdd6 + (sizeof(GPROdd6)/sizeof(unsigned))); + return makeArrayRef(GPROdd6); } } - return std::make_pair(RC->allocation_order_begin(MF), - RC->allocation_order_end(MF)); + return RC->getRawAllocationOrder(MF); } /// ResolveRegAllocHint - Resolves the specified register allocation hint @@ -515,7 +445,7 @@ ARMBaseRegisterInfo::UpdateRegAllocHint(unsigned Reg, unsigned NewReg, std::pair Hint = MRI->getRegAllocationHint(Reg); if ((Hint.first == (unsigned)ARMRI::RegPairOdd || Hint.first == (unsigned)ARMRI::RegPairEven) && - Hint.second && TargetRegisterInfo::isVirtualRegister(Hint.second)) { + TargetRegisterInfo::isVirtualRegister(Hint.second)) { // If 'Reg' is one of the even / odd register pair and it's now changed // (e.g. coalesced) into a different register. The other register of the // pair allocation hint must be updated to reflect the relationship @@ -528,21 +458,27 @@ ARMBaseRegisterInfo::UpdateRegAllocHint(unsigned Reg, unsigned NewReg, } } -/// hasFP - Return true if the specified function should have a dedicated frame -/// pointer register. This is true if the function has variable sized allocas -/// or if frame pointer elimination is disabled. -/// -bool ARMBaseRegisterInfo::hasFP(const MachineFunction &MF) const { - // Mac OS X requires FP not to be clobbered for backtracing purpose. - if (STI.isTargetDarwin()) - return true; +bool +ARMBaseRegisterInfo::avoidWriteAfterWrite(const TargetRegisterClass *RC) const { + // CortexA9 has a Write-after-write hazard for NEON registers. + if (!STI.isCortexA9()) + return false; - const MachineFrameInfo *MFI = MF.getFrameInfo(); - // Always eliminate non-leaf frame pointers. - return ((DisableFramePointerElim(MF) && MFI->hasCalls()) || - needsStackRealignment(MF) || - MFI->hasVarSizedObjects() || - MFI->isFrameAddressTaken()); + switch (RC->getID()) { + case ARM::DPRRegClassID: + case ARM::DPR_8RegClassID: + case ARM::DPR_VFP2RegClassID: + case ARM::QPRRegClassID: + case ARM::QPR_8RegClassID: + case ARM::QPR_VFP2RegClassID: + case ARM::SPRRegClassID: + case ARM::SPR_8RegClassID: + // Avoid reusing S, D, and Q registers. + // Don't increase register pressure for QQ and QQQQ. + return true; + default: + return false; + } } bool ARMBaseRegisterInfo::hasBasePointer(const MachineFunction &MF) const { @@ -577,21 +513,36 @@ bool ARMBaseRegisterInfo::hasBasePointer(const MachineFunction &MF) const { bool ARMBaseRegisterInfo::canRealignStack(const MachineFunction &MF) const { const MachineFrameInfo *MFI = MF.getFrameInfo(); + const MachineRegisterInfo *MRI = &MF.getRegInfo(); const ARMFunctionInfo *AFI = MF.getInfo(); // We can't realign the stack if: // 1. Dynamic stack realignment is explicitly disabled, // 2. This is a Thumb1 function (it's not useful, so we don't bother), or // 3. There are VLAs in the function and the base pointer is disabled. - return (RealignStack && !AFI->isThumb1OnlyFunction() && - (!MFI->hasVarSizedObjects() || EnableBasePointer)); + if (!MF.getTarget().Options.RealignStack) + return false; + if (AFI->isThumb1OnlyFunction()) + return false; + // Stack realignment requires a frame pointer. If we already started + // register allocation with frame pointer elimination, it is too late now. + if (!MRI->canReserveReg(FramePtr)) + return false; + // We may also need a base pointer if there are dynamic allocas. + if (!MFI->hasVarSizedObjects()) + return true; + if (!EnableBasePointer) + return false; + // A base pointer is required and allowed. Check that it isn't too late to + // reserve it. + return MRI->canReserveReg(BasePtr); } bool ARMBaseRegisterInfo:: needsStackRealignment(const MachineFunction &MF) const { const MachineFrameInfo *MFI = MF.getFrameInfo(); const Function *F = MF.getFunction(); - unsigned StackAlign = MF.getTarget().getFrameInfo()->getStackAlignment(); - bool requiresRealignment = ((MFI->getLocalFrameMaxAlign() > StackAlign) || + unsigned StackAlign = MF.getTarget().getFrameLowering()->getStackAlignment(); + bool requiresRealignment = ((MFI->getMaxAlignment() > StackAlign) || F->hasFnAttr(Attribute::StackAlignment)); return requiresRealignment && canRealignStack(MF); @@ -600,444 +551,27 @@ needsStackRealignment(const MachineFunction &MF) const { bool ARMBaseRegisterInfo:: cannotEliminateFrame(const MachineFunction &MF) const { const MachineFrameInfo *MFI = MF.getFrameInfo(); - if (DisableFramePointerElim(MF) && MFI->adjustsStack()) + if (MF.getTarget().Options.DisableFramePointerElim(MF) && MFI->adjustsStack()) return true; return MFI->hasVarSizedObjects() || MFI->isFrameAddressTaken() || needsStackRealignment(MF); } -/// estimateStackSize - Estimate and return the size of the frame. -static unsigned estimateStackSize(MachineFunction &MF) { - const MachineFrameInfo *FFI = MF.getFrameInfo(); - int Offset = 0; - for (int i = FFI->getObjectIndexBegin(); i != 0; ++i) { - int FixedOff = -FFI->getObjectOffset(i); - if (FixedOff > Offset) Offset = FixedOff; - } - for (unsigned i = 0, e = FFI->getObjectIndexEnd(); i != e; ++i) { - if (FFI->isDeadObjectIndex(i)) - continue; - Offset += FFI->getObjectSize(i); - unsigned Align = FFI->getObjectAlignment(i); - // Adjust to alignment boundary - Offset = (Offset+Align-1)/Align*Align; - } - return (unsigned)Offset; -} - -/// estimateRSStackSizeLimit - Look at each instruction that references stack -/// frames and return the stack size limit beyond which some of these -/// instructions will require a scratch register during their expansion later. -unsigned -ARMBaseRegisterInfo::estimateRSStackSizeLimit(MachineFunction &MF) const { - const ARMFunctionInfo *AFI = MF.getInfo(); - unsigned Limit = (1 << 12) - 1; - for (MachineFunction::iterator BB = MF.begin(),E = MF.end(); BB != E; ++BB) { - for (MachineBasicBlock::iterator I = BB->begin(), E = BB->end(); - I != E; ++I) { - for (unsigned i = 0, e = I->getNumOperands(); i != e; ++i) { - if (!I->getOperand(i).isFI()) continue; - - // When using ADDri to get the address of a stack object, 255 is the - // largest offset guaranteed to fit in the immediate offset. - if (I->getOpcode() == ARM::ADDri) { - Limit = std::min(Limit, (1U << 8) - 1); - break; - } - - // Otherwise check the addressing mode. - switch (I->getDesc().TSFlags & ARMII::AddrModeMask) { - case ARMII::AddrMode3: - case ARMII::AddrModeT2_i8: - Limit = std::min(Limit, (1U << 8) - 1); - break; - case ARMII::AddrMode5: - case ARMII::AddrModeT2_i8s4: - Limit = std::min(Limit, ((1U << 8) - 1) * 4); - break; - case ARMII::AddrModeT2_i12: - // i12 supports only positive offset so these will be converted to - // i8 opcodes. See llvm::rewriteT2FrameIndex. - if (hasFP(MF) && AFI->hasStackFrame()) - Limit = std::min(Limit, (1U << 8) - 1); - break; - case ARMII::AddrMode4: - case ARMII::AddrMode6: - // Addressing modes 4 & 6 (load/store) instructions can't encode an - // immediate offset for stack references. - return 0; - default: - break; - } - break; // At most one FI per instruction - } - } - } - - return Limit; -} - -static unsigned GetFunctionSizeInBytes(const MachineFunction &MF, - const ARMBaseInstrInfo &TII) { - unsigned FnSize = 0; - for (MachineFunction::const_iterator MBBI = MF.begin(), E = MF.end(); - MBBI != E; ++MBBI) { - const MachineBasicBlock &MBB = *MBBI; - for (MachineBasicBlock::const_iterator I = MBB.begin(),E = MBB.end(); - I != E; ++I) - FnSize += TII.GetInstSizeInBytes(I); - } - return FnSize; -} - -void -ARMBaseRegisterInfo::processFunctionBeforeCalleeSavedScan(MachineFunction &MF, - RegScavenger *RS) const { - // This tells PEI to spill the FP as if it is any other callee-save register - // to take advantage the eliminateFrameIndex machinery. This also ensures it - // is spilled in the order specified by getCalleeSavedRegs() to make it easier - // to combine multiple loads / stores. - bool CanEliminateFrame = true; - bool CS1Spilled = false; - bool LRSpilled = false; - unsigned NumGPRSpills = 0; - SmallVector UnspilledCS1GPRs; - SmallVector UnspilledCS2GPRs; - ARMFunctionInfo *AFI = MF.getInfo(); - MachineFrameInfo *MFI = MF.getFrameInfo(); - - // Spill R4 if Thumb2 function requires stack realignment - it will be used as - // scratch register. - // FIXME: It will be better just to find spare register here. - if (needsStackRealignment(MF) && - AFI->isThumb2Function()) - MF.getRegInfo().setPhysRegUsed(ARM::R4); - - // Spill LR if Thumb1 function uses variable length argument lists. - if (AFI->isThumb1OnlyFunction() && AFI->getVarArgsRegSaveSize() > 0) - MF.getRegInfo().setPhysRegUsed(ARM::LR); - - // Spill the BasePtr if it's used. - if (hasBasePointer(MF)) - MF.getRegInfo().setPhysRegUsed(BasePtr); - - // Don't spill FP if the frame can be eliminated. This is determined - // by scanning the callee-save registers to see if any is used. - const unsigned *CSRegs = getCalleeSavedRegs(); - for (unsigned i = 0; CSRegs[i]; ++i) { - unsigned Reg = CSRegs[i]; - bool Spilled = false; - if (MF.getRegInfo().isPhysRegUsed(Reg)) { - AFI->setCSRegisterIsSpilled(Reg); - Spilled = true; - CanEliminateFrame = false; - } else { - // Check alias registers too. - for (const unsigned *Aliases = getAliasSet(Reg); *Aliases; ++Aliases) { - if (MF.getRegInfo().isPhysRegUsed(*Aliases)) { - Spilled = true; - CanEliminateFrame = false; - } - } - } - - if (!ARM::GPRRegisterClass->contains(Reg)) - continue; - - if (Spilled) { - NumGPRSpills++; - - if (!STI.isTargetDarwin()) { - if (Reg == ARM::LR) - LRSpilled = true; - CS1Spilled = true; - continue; - } - - // Keep track if LR and any of R4, R5, R6, and R7 is spilled. - switch (Reg) { - case ARM::LR: - LRSpilled = true; - // Fallthrough - case ARM::R4: - case ARM::R5: - case ARM::R6: - case ARM::R7: - CS1Spilled = true; - break; - default: - break; - } - } else { - if (!STI.isTargetDarwin()) { - UnspilledCS1GPRs.push_back(Reg); - continue; - } - - switch (Reg) { - case ARM::R4: - case ARM::R5: - case ARM::R6: - case ARM::R7: - case ARM::LR: - UnspilledCS1GPRs.push_back(Reg); - break; - default: - UnspilledCS2GPRs.push_back(Reg); - break; - } - } - } - - bool ForceLRSpill = false; - if (!LRSpilled && AFI->isThumb1OnlyFunction()) { - unsigned FnSize = GetFunctionSizeInBytes(MF, TII); - // Force LR to be spilled if the Thumb function size is > 2048. This enables - // use of BL to implement far jump. If it turns out that it's not needed - // then the branch fix up path will undo it. - if (FnSize >= (1 << 11)) { - CanEliminateFrame = false; - ForceLRSpill = true; - } - } - - // If any of the stack slot references may be out of range of an immediate - // offset, make sure a register (or a spill slot) is available for the - // register scavenger. Note that if we're indexing off the frame pointer, the - // effective stack size is 4 bytes larger since the FP points to the stack - // slot of the previous FP. Also, if we have variable sized objects in the - // function, stack slot references will often be negative, and some of - // our instructions are positive-offset only, so conservatively consider - // that case to want a spill slot (or register) as well. Similarly, if - // the function adjusts the stack pointer during execution and the - // adjustments aren't already part of our stack size estimate, our offset - // calculations may be off, so be conservative. - // FIXME: We could add logic to be more precise about negative offsets - // and which instructions will need a scratch register for them. Is it - // worth the effort and added fragility? - bool BigStack = - (RS && - (estimateStackSize(MF) + ((hasFP(MF) && AFI->hasStackFrame()) ? 4:0) >= - estimateRSStackSizeLimit(MF))) - || MFI->hasVarSizedObjects() - || (MFI->adjustsStack() && !canSimplifyCallFramePseudos(MF)); - - bool ExtraCSSpill = false; - if (BigStack || !CanEliminateFrame || cannotEliminateFrame(MF)) { - AFI->setHasStackFrame(true); - - // If LR is not spilled, but at least one of R4, R5, R6, and R7 is spilled. - // Spill LR as well so we can fold BX_RET to the registers restore (LDM). - if (!LRSpilled && CS1Spilled) { - MF.getRegInfo().setPhysRegUsed(ARM::LR); - AFI->setCSRegisterIsSpilled(ARM::LR); - NumGPRSpills++; - UnspilledCS1GPRs.erase(std::find(UnspilledCS1GPRs.begin(), - UnspilledCS1GPRs.end(), (unsigned)ARM::LR)); - ForceLRSpill = false; - ExtraCSSpill = true; - } - - if (hasFP(MF)) { - MF.getRegInfo().setPhysRegUsed(FramePtr); - NumGPRSpills++; - } - - // If stack and double are 8-byte aligned and we are spilling an odd number - // of GPRs, spill one extra callee save GPR so we won't have to pad between - // the integer and double callee save areas. - unsigned TargetAlign = MF.getTarget().getFrameInfo()->getStackAlignment(); - if (TargetAlign == 8 && (NumGPRSpills & 1)) { - if (CS1Spilled && !UnspilledCS1GPRs.empty()) { - for (unsigned i = 0, e = UnspilledCS1GPRs.size(); i != e; ++i) { - unsigned Reg = UnspilledCS1GPRs[i]; - // Don't spill high register if the function is thumb1 - if (!AFI->isThumb1OnlyFunction() || - isARMLowRegister(Reg) || Reg == ARM::LR) { - MF.getRegInfo().setPhysRegUsed(Reg); - AFI->setCSRegisterIsSpilled(Reg); - if (!isReservedReg(MF, Reg)) - ExtraCSSpill = true; - break; - } - } - } else if (!UnspilledCS2GPRs.empty() && - !AFI->isThumb1OnlyFunction()) { - unsigned Reg = UnspilledCS2GPRs.front(); - MF.getRegInfo().setPhysRegUsed(Reg); - AFI->setCSRegisterIsSpilled(Reg); - if (!isReservedReg(MF, Reg)) - ExtraCSSpill = true; - } - } - - // Estimate if we might need to scavenge a register at some point in order - // to materialize a stack offset. If so, either spill one additional - // callee-saved register or reserve a special spill slot to facilitate - // register scavenging. Thumb1 needs a spill slot for stack pointer - // adjustments also, even when the frame itself is small. - if (BigStack && !ExtraCSSpill) { - // If any non-reserved CS register isn't spilled, just spill one or two - // extra. That should take care of it! - unsigned NumExtras = TargetAlign / 4; - SmallVector Extras; - while (NumExtras && !UnspilledCS1GPRs.empty()) { - unsigned Reg = UnspilledCS1GPRs.back(); - UnspilledCS1GPRs.pop_back(); - if (!isReservedReg(MF, Reg) && - (!AFI->isThumb1OnlyFunction() || isARMLowRegister(Reg) || - Reg == ARM::LR)) { - Extras.push_back(Reg); - NumExtras--; - } - } - // For non-Thumb1 functions, also check for hi-reg CS registers - if (!AFI->isThumb1OnlyFunction()) { - while (NumExtras && !UnspilledCS2GPRs.empty()) { - unsigned Reg = UnspilledCS2GPRs.back(); - UnspilledCS2GPRs.pop_back(); - if (!isReservedReg(MF, Reg)) { - Extras.push_back(Reg); - NumExtras--; - } - } - } - if (Extras.size() && NumExtras == 0) { - for (unsigned i = 0, e = Extras.size(); i != e; ++i) { - MF.getRegInfo().setPhysRegUsed(Extras[i]); - AFI->setCSRegisterIsSpilled(Extras[i]); - } - } else if (!AFI->isThumb1OnlyFunction()) { - // note: Thumb1 functions spill to R12, not the stack. Reserve a slot - // closest to SP or frame pointer. - const TargetRegisterClass *RC = ARM::GPRRegisterClass; - RS->setScavengingFrameIndex(MFI->CreateStackObject(RC->getSize(), - RC->getAlignment(), - false)); - } - } - } - - if (ForceLRSpill) { - MF.getRegInfo().setPhysRegUsed(ARM::LR); - AFI->setCSRegisterIsSpilled(ARM::LR); - AFI->setLRIsSpilledForFarJump(true); - } -} - -unsigned ARMBaseRegisterInfo::getRARegister() const { - return ARM::LR; -} - unsigned ARMBaseRegisterInfo::getFrameRegister(const MachineFunction &MF) const { - if (hasFP(MF)) + const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering(); + + if (TFI->hasFP(MF)) return FramePtr; return ARM::SP; } -// Provide a base+offset reference to an FI slot for debug info. It's the -// same as what we use for resolving the code-gen references for now. -// FIXME: This can go wrong when references are SP-relative and simple call -// frames aren't used. -int -ARMBaseRegisterInfo::getFrameIndexReference(const MachineFunction &MF, int FI, - unsigned &FrameReg) const { - return ResolveFrameIndexReference(MF, FI, FrameReg, 0); -} - -int -ARMBaseRegisterInfo::ResolveFrameIndexReference(const MachineFunction &MF, - int FI, - unsigned &FrameReg, - int SPAdj) const { - const MachineFrameInfo *MFI = MF.getFrameInfo(); - const ARMFunctionInfo *AFI = MF.getInfo(); - int Offset = MFI->getObjectOffset(FI) + MFI->getStackSize(); - int FPOffset = Offset - AFI->getFramePtrSpillOffset(); - bool isFixed = MFI->isFixedObjectIndex(FI); - - FrameReg = ARM::SP; - Offset += SPAdj; - if (AFI->isGPRCalleeSavedArea1Frame(FI)) - return Offset - AFI->getGPRCalleeSavedArea1Offset(); - else if (AFI->isGPRCalleeSavedArea2Frame(FI)) - return Offset - AFI->getGPRCalleeSavedArea2Offset(); - else if (AFI->isDPRCalleeSavedAreaFrame(FI)) - return Offset - AFI->getDPRCalleeSavedAreaOffset(); - - // When dynamically realigning the stack, use the frame pointer for - // parameters, and the stack/base pointer for locals. - if (needsStackRealignment(MF)) { - assert (hasFP(MF) && "dynamic stack realignment without a FP!"); - if (isFixed) { - FrameReg = getFrameRegister(MF); - Offset = FPOffset; - } else if (MFI->hasVarSizedObjects()) { - assert(hasBasePointer(MF) && - "VLAs and dynamic stack alignment, but missing base pointer!"); - FrameReg = BasePtr; - } - return Offset; - } - - // If there is a frame pointer, use it when we can. - if (hasFP(MF) && AFI->hasStackFrame()) { - // Use frame pointer to reference fixed objects. Use it for locals if - // there are VLAs (and thus the SP isn't reliable as a base). - if (isFixed || (MFI->hasVarSizedObjects() && !hasBasePointer(MF))) { - FrameReg = getFrameRegister(MF); - return FPOffset; - } else if (MFI->hasVarSizedObjects()) { - assert(hasBasePointer(MF) && "missing base pointer!"); - // Try to use the frame pointer if we can, else use the base pointer - // since it's available. This is handy for the emergency spill slot, in - // particular. - if (AFI->isThumb2Function()) { - if (FPOffset >= -255 && FPOffset < 0) { - FrameReg = getFrameRegister(MF); - return FPOffset; - } - } else - FrameReg = BasePtr; - } else if (AFI->isThumb2Function()) { - // In Thumb2 mode, the negative offset is very limited. Try to avoid - // out of range references. - if (FPOffset >= -255 && FPOffset < 0) { - FrameReg = getFrameRegister(MF); - return FPOffset; - } - } else if (Offset > (FPOffset < 0 ? -FPOffset : FPOffset)) { - // Otherwise, use SP or FP, whichever is closer to the stack slot. - FrameReg = getFrameRegister(MF); - return FPOffset; - } - } - // Use the base pointer if we have one. - if (hasBasePointer(MF)) - FrameReg = BasePtr; - return Offset; -} - -int -ARMBaseRegisterInfo::getFrameIndexOffset(const MachineFunction &MF, - int FI) const { - unsigned FrameReg; - return getFrameIndexReference(MF, FI, FrameReg); -} - unsigned ARMBaseRegisterInfo::getEHExceptionRegister() const { llvm_unreachable("What is the exception register"); - return 0; } unsigned ARMBaseRegisterInfo::getEHHandlerRegister() const { llvm_unreachable("What is the exception handler register"); - return 0; -} - -int ARMBaseRegisterInfo::getDwarfRegNum(unsigned RegNum, bool isEH) const { - return ARMGenRegisterInfo::getDwarfRegNumFull(RegNum, 0); } unsigned ARMBaseRegisterInfo::getRegisterPairEven(unsigned Reg, @@ -1046,85 +580,48 @@ unsigned ARMBaseRegisterInfo::getRegisterPairEven(unsigned Reg, default: break; // Return 0 if either register of the pair is a special register. // So no R12, etc. - case ARM::R1: - return ARM::R0; - case ARM::R3: - return ARM::R2; - case ARM::R5: - return ARM::R4; + case ARM::R1: return ARM::R0; + case ARM::R3: return ARM::R2; + case ARM::R5: return ARM::R4; case ARM::R7: return (isReservedReg(MF, ARM::R7) || isReservedReg(MF, ARM::R6)) ? 0 : ARM::R6; - case ARM::R9: - return isReservedReg(MF, ARM::R9) ? 0 :ARM::R8; - case ARM::R11: - return isReservedReg(MF, ARM::R11) ? 0 : ARM::R10; - - case ARM::S1: - return ARM::S0; - case ARM::S3: - return ARM::S2; - case ARM::S5: - return ARM::S4; - case ARM::S7: - return ARM::S6; - case ARM::S9: - return ARM::S8; - case ARM::S11: - return ARM::S10; - case ARM::S13: - return ARM::S12; - case ARM::S15: - return ARM::S14; - case ARM::S17: - return ARM::S16; - case ARM::S19: - return ARM::S18; - case ARM::S21: - return ARM::S20; - case ARM::S23: - return ARM::S22; - case ARM::S25: - return ARM::S24; - case ARM::S27: - return ARM::S26; - case ARM::S29: - return ARM::S28; - case ARM::S31: - return ARM::S30; - - case ARM::D1: - return ARM::D0; - case ARM::D3: - return ARM::D2; - case ARM::D5: - return ARM::D4; - case ARM::D7: - return ARM::D6; - case ARM::D9: - return ARM::D8; - case ARM::D11: - return ARM::D10; - case ARM::D13: - return ARM::D12; - case ARM::D15: - return ARM::D14; - case ARM::D17: - return ARM::D16; - case ARM::D19: - return ARM::D18; - case ARM::D21: - return ARM::D20; - case ARM::D23: - return ARM::D22; - case ARM::D25: - return ARM::D24; - case ARM::D27: - return ARM::D26; - case ARM::D29: - return ARM::D28; - case ARM::D31: - return ARM::D30; + case ARM::R9: return isReservedReg(MF, ARM::R9) ? 0 :ARM::R8; + case ARM::R11: return isReservedReg(MF, ARM::R11) ? 0 : ARM::R10; + + case ARM::S1: return ARM::S0; + case ARM::S3: return ARM::S2; + case ARM::S5: return ARM::S4; + case ARM::S7: return ARM::S6; + case ARM::S9: return ARM::S8; + case ARM::S11: return ARM::S10; + case ARM::S13: return ARM::S12; + case ARM::S15: return ARM::S14; + case ARM::S17: return ARM::S16; + case ARM::S19: return ARM::S18; + case ARM::S21: return ARM::S20; + case ARM::S23: return ARM::S22; + case ARM::S25: return ARM::S24; + case ARM::S27: return ARM::S26; + case ARM::S29: return ARM::S28; + case ARM::S31: return ARM::S30; + + case ARM::D1: return ARM::D0; + case ARM::D3: return ARM::D2; + case ARM::D5: return ARM::D4; + case ARM::D7: return ARM::D6; + case ARM::D9: return ARM::D8; + case ARM::D11: return ARM::D10; + case ARM::D13: return ARM::D12; + case ARM::D15: return ARM::D14; + case ARM::D17: return ARM::D16; + case ARM::D19: return ARM::D18; + case ARM::D21: return ARM::D20; + case ARM::D23: return ARM::D22; + case ARM::D25: return ARM::D24; + case ARM::D27: return ARM::D26; + case ARM::D29: return ARM::D28; + case ARM::D31: return ARM::D30; } return 0; @@ -1136,85 +633,48 @@ unsigned ARMBaseRegisterInfo::getRegisterPairOdd(unsigned Reg, default: break; // Return 0 if either register of the pair is a special register. // So no R12, etc. - case ARM::R0: - return ARM::R1; - case ARM::R2: - return ARM::R3; - case ARM::R4: - return ARM::R5; + case ARM::R0: return ARM::R1; + case ARM::R2: return ARM::R3; + case ARM::R4: return ARM::R5; case ARM::R6: return (isReservedReg(MF, ARM::R7) || isReservedReg(MF, ARM::R6)) ? 0 : ARM::R7; - case ARM::R8: - return isReservedReg(MF, ARM::R9) ? 0 :ARM::R9; - case ARM::R10: - return isReservedReg(MF, ARM::R11) ? 0 : ARM::R11; - - case ARM::S0: - return ARM::S1; - case ARM::S2: - return ARM::S3; - case ARM::S4: - return ARM::S5; - case ARM::S6: - return ARM::S7; - case ARM::S8: - return ARM::S9; - case ARM::S10: - return ARM::S11; - case ARM::S12: - return ARM::S13; - case ARM::S14: - return ARM::S15; - case ARM::S16: - return ARM::S17; - case ARM::S18: - return ARM::S19; - case ARM::S20: - return ARM::S21; - case ARM::S22: - return ARM::S23; - case ARM::S24: - return ARM::S25; - case ARM::S26: - return ARM::S27; - case ARM::S28: - return ARM::S29; - case ARM::S30: - return ARM::S31; - - case ARM::D0: - return ARM::D1; - case ARM::D2: - return ARM::D3; - case ARM::D4: - return ARM::D5; - case ARM::D6: - return ARM::D7; - case ARM::D8: - return ARM::D9; - case ARM::D10: - return ARM::D11; - case ARM::D12: - return ARM::D13; - case ARM::D14: - return ARM::D15; - case ARM::D16: - return ARM::D17; - case ARM::D18: - return ARM::D19; - case ARM::D20: - return ARM::D21; - case ARM::D22: - return ARM::D23; - case ARM::D24: - return ARM::D25; - case ARM::D26: - return ARM::D27; - case ARM::D28: - return ARM::D29; - case ARM::D30: - return ARM::D31; + case ARM::R8: return isReservedReg(MF, ARM::R9) ? 0 :ARM::R9; + case ARM::R10: return isReservedReg(MF, ARM::R11) ? 0 : ARM::R11; + + case ARM::S0: return ARM::S1; + case ARM::S2: return ARM::S3; + case ARM::S4: return ARM::S5; + case ARM::S6: return ARM::S7; + case ARM::S8: return ARM::S9; + case ARM::S10: return ARM::S11; + case ARM::S12: return ARM::S13; + case ARM::S14: return ARM::S15; + case ARM::S16: return ARM::S17; + case ARM::S18: return ARM::S19; + case ARM::S20: return ARM::S21; + case ARM::S22: return ARM::S23; + case ARM::S24: return ARM::S25; + case ARM::S26: return ARM::S27; + case ARM::S28: return ARM::S29; + case ARM::S30: return ARM::S31; + + case ARM::D0: return ARM::D1; + case ARM::D2: return ARM::D3; + case ARM::D4: return ARM::D5; + case ARM::D6: return ARM::D7; + case ARM::D8: return ARM::D9; + case ARM::D10: return ARM::D11; + case ARM::D12: return ARM::D13; + case ARM::D14: return ARM::D15; + case ARM::D16: return ARM::D17; + case ARM::D18: return ARM::D19; + case ARM::D20: return ARM::D21; + case ARM::D22: return ARM::D23; + case ARM::D24: return ARM::D25; + case ARM::D26: return ARM::D27; + case ARM::D28: return ARM::D29; + case ARM::D30: return ARM::D31; } return 0; @@ -1228,7 +688,7 @@ emitLoadConstPool(MachineBasicBlock &MBB, DebugLoc dl, unsigned DestReg, unsigned SubIdx, int Val, ARMCC::CondCodes Pred, - unsigned PredReg) const { + unsigned PredReg, unsigned MIFlags) const { MachineFunction &MF = *MBB.getParent(); MachineConstantPool *ConstantPool = MF.getConstantPool(); const Constant *C = @@ -1238,7 +698,8 @@ emitLoadConstPool(MachineBasicBlock &MBB, BuildMI(MBB, MBBI, dl, TII.get(ARM::LDRcp)) .addReg(DestReg, getDefRegState(true), SubIdx) .addConstantPoolIndex(Idx) - .addImm(0).addImm(Pred).addReg(PredReg); + .addImm(0).addImm(Pred).addReg(PredReg) + .setMIFlags(MIFlags); } bool ARMBaseRegisterInfo:: @@ -1256,34 +717,6 @@ requiresVirtualBaseRegisters(const MachineFunction &MF) const { return EnableLocalStackAlloc; } -// hasReservedCallFrame - Under normal circumstances, when a frame pointer is -// not required, we reserve argument space for call sites in the function -// immediately on entry to the current function. This eliminates the need for -// add/sub sp brackets around call sites. Returns true if the call frame is -// included as part of the stack frame. -bool ARMBaseRegisterInfo:: -hasReservedCallFrame(const MachineFunction &MF) const { - const MachineFrameInfo *FFI = MF.getFrameInfo(); - unsigned CFSize = FFI->getMaxCallFrameSize(); - // It's not always a good idea to include the call frame as part of the - // stack frame. ARM (especially Thumb) has small immediate offset to - // address the stack frame. So a large call frame can cause poor codegen - // and may even makes it impossible to scavenge a register. - if (CFSize >= ((1 << 12) - 1) / 2) // Half of imm12 - return false; - - return !MF.getFrameInfo()->hasVarSizedObjects(); -} - -// canSimplifyCallFramePseudos - If there is a reserved call frame, the -// call frame pseudos can be simplified. Unlike most targets, having a FP -// is not sufficient here since we still may reference some objects via SP -// even when FP is available in Thumb2 mode. -bool ARMBaseRegisterInfo:: -canSimplifyCallFramePseudos(const MachineFunction &MF) const { - return hasReservedCallFrame(MF) || MF.getFrameInfo()->hasVarSizedObjects(); -} - static void emitSPUpdate(bool isARM, MachineBasicBlock &MBB, MachineBasicBlock::iterator &MBBI, @@ -1302,7 +735,8 @@ emitSPUpdate(bool isARM, void ARMBaseRegisterInfo:: eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB, MachineBasicBlock::iterator I) const { - if (!hasReservedCallFrame(MF)) { + const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering(); + if (!TFI->hasReservedCallFrame(MF)) { // If we have alloca, convert as follows: // ADJCALLSTACKDOWN -> sub, sp, sp, amount // ADJCALLSTACKUP -> add, sp, sp, amount @@ -1313,7 +747,7 @@ eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB, // We need to keep the stack aligned properly. To do this, we round the // amount of space needed for the outgoing arguments up to the next // alignment boundary. - unsigned Align = MF.getTarget().getFrameInfo()->getStackAlignment(); + unsigned Align = TFI->getStackAlignment(); Amount = (Amount+Align-1)/Align*Align; ARMFunctionInfo *AFI = MF.getInfo(); @@ -1343,9 +777,9 @@ eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB, int64_t ARMBaseRegisterInfo:: getFrameIndexInstrOffset(const MachineInstr *MI, int Idx) const { - const TargetInstrDesc &Desc = MI->getDesc(); + const MCInstrDesc &Desc = MI->getDesc(); unsigned AddrMode = (Desc.TSFlags & ARMII::AddrModeMask); - int64_t InstrOffs = 0;; + int64_t InstrOffs = 0; int Scale = 1; unsigned ImmIdx = 0; switch (AddrMode) { @@ -1386,7 +820,6 @@ getFrameIndexInstrOffset(const MachineInstr *MI, int Idx) const { } default: llvm_unreachable("Unsupported addressing mode!"); - break; } return InstrOffs * Scale; @@ -1433,6 +866,7 @@ needsFrameBaseReg(MachineInstr *MI, int64_t Offset) const { // Note that the incoming offset is based on the SP value at function entry, // so it'll be negative. MachineFunction &MF = *MI->getParent()->getParent(); + const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering(); MachineFrameInfo *MFI = MF.getFrameInfo(); ARMFunctionInfo *AFI = MF.getInfo(); @@ -1459,8 +893,8 @@ needsFrameBaseReg(MachineInstr *MI, int64_t Offset) const { // The FP is only available if there is no dynamic realignment. We // don't know for sure yet whether we'll need that, so we guess based // on whether there are any local variables that would trigger it. - unsigned StackAlign = MF.getTarget().getFrameInfo()->getStackAlignment(); - if (hasFP(MF) && + unsigned StackAlign = TFI->getStackAlignment(); + if (TFI->hasFP(MF) && !((MFI->getLocalFrameMaxAlign() > StackAlign) && canRealignStack(MF))) { if (isFrameOffsetLegal(MI, FPOffset)) return false; @@ -1477,21 +911,30 @@ needsFrameBaseReg(MachineInstr *MI, int64_t Offset) const { return true; } -/// materializeFrameBaseRegister - Insert defining instruction(s) for -/// BaseReg to be a pointer to FrameIdx before insertion point I. +/// materializeFrameBaseRegister - Insert defining instruction(s) for BaseReg to +/// be a pointer to FrameIdx at the beginning of the basic block. void ARMBaseRegisterInfo:: -materializeFrameBaseRegister(MachineBasicBlock::iterator I, unsigned BaseReg, - int FrameIdx, int64_t Offset) const { - ARMFunctionInfo *AFI = - I->getParent()->getParent()->getInfo(); +materializeFrameBaseRegister(MachineBasicBlock *MBB, + unsigned BaseReg, int FrameIdx, + int64_t Offset) const { + ARMFunctionInfo *AFI = MBB->getParent()->getInfo(); unsigned ADDriOpc = !AFI->isThumbFunction() ? ARM::ADDri : (AFI->isThumb1OnlyFunction() ? ARM::tADDrSPi : ARM::t2ADDri); - MachineInstrBuilder MIB = - BuildMI(*I->getParent(), I, I->getDebugLoc(), TII.get(ADDriOpc), BaseReg) - .addFrameIndex(FrameIdx).addImm(Offset); + MachineBasicBlock::iterator Ins = MBB->begin(); + DebugLoc DL; // Defaults to "unknown" + if (Ins != MBB->end()) + DL = Ins->getDebugLoc(); + + const MCInstrDesc &MCID = TII.get(ADDriOpc); + MachineRegisterInfo &MRI = MBB->getParent()->getRegInfo(); + MRI.constrainRegClass(BaseReg, TII.getRegClass(MCID, 0, this)); + + MachineInstrBuilder MIB = AddDefaultPred(BuildMI(*MBB, Ins, DL, MCID, BaseReg) + .addFrameIndex(FrameIdx).addImm(Offset)); + if (!AFI->isThumb1OnlyFunction()) - AddDefaultCC(AddDefaultPred(MIB)); + AddDefaultCC(MIB); } void @@ -1519,11 +962,12 @@ ARMBaseRegisterInfo::resolveFrameIndex(MachineBasicBlock::iterator I, Done = rewriteT2FrameIndex(MI, i, BaseReg, Off, TII); } assert (Done && "Unable to resolve frame index!"); + (void)Done; } bool ARMBaseRegisterInfo::isFrameOffsetLegal(const MachineInstr *MI, int64_t Offset) const { - const TargetInstrDesc &Desc = MI->getDesc(); + const MCInstrDesc &Desc = MI->getDesc(); unsigned AddrMode = (Desc.TSFlags & ARMII::AddrModeMask); unsigned i = 0; @@ -1571,7 +1015,6 @@ bool ARMBaseRegisterInfo::isFrameOffsetLegal(const MachineInstr *MI, break; default: llvm_unreachable("Unsupported addressing mode!"); - break; } Offset += getFrameIndexInstrOffset(MI, i); @@ -1597,6 +1040,8 @@ ARMBaseRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II, MachineInstr &MI = *II; MachineBasicBlock &MBB = *MI.getParent(); MachineFunction &MF = *MBB.getParent(); + const ARMFrameLowering *TFI = + static_cast(MF.getTarget().getFrameLowering()); ARMFunctionInfo *AFI = MF.getInfo(); assert(!AFI->isThumb1OnlyFunction() && "This eliminateFrameIndex does not support Thumb1!"); @@ -1609,7 +1054,7 @@ ARMBaseRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II, int FrameIndex = MI.getOperand(i).getIndex(); unsigned FrameReg; - int Offset = ResolveFrameIndexReference(MF, FrameIndex, FrameReg, SPAdj); + int Offset = TFI->ResolveFrameIndexReference(MF, FrameIndex, FrameReg, SPAdj); // Special handling of dbg_value instructions. if (MI.isDebugValue()) { @@ -1655,8 +1100,7 @@ ARMBaseRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II, emitT2RegPlusImmediate(MBB, II, MI.getDebugLoc(), ScratchReg, FrameReg, Offset, Pred, PredReg, TII); } + // Update the original instruction to use the scratch register. MI.getOperand(i).ChangeToRegister(ScratchReg, false, false, true); } } - -#include "ARMGenRegisterInfo.inc"