#include "ARMFrameLowering.h"
#include "ARMBaseInstrInfo.h"
#include "ARMBaseRegisterInfo.h"
+#include "ARMConstantPoolValue.h"
#include "ARMMachineFunctionInfo.h"
#include "MCTargetDesc/ARMAddressingModes.h"
-#include "llvm/Function.h"
#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
+#include "llvm/CodeGen/MachineModuleInfo.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/RegisterScavenging.h"
-#include "llvm/Target/TargetOptions.h"
+#include "llvm/MC/MCAsmInfo.h"
+#include "llvm/IR/CallingConv.h"
+#include "llvm/IR/Function.h"
+#include "llvm/MC/MCContext.h"
#include "llvm/Support/CommandLine.h"
+#include "llvm/Target/TargetOptions.h"
using namespace llvm;
skipAlignedDPRCS2Spills(MachineBasicBlock::iterator MI,
unsigned NumAlignedDPRCS2Regs);
+ARMFrameLowering::ARMFrameLowering(const ARMSubtarget &sti)
+ : TargetFrameLowering(StackGrowsDown, sti.getStackAlignment(), 0, 4),
+ STI(sti) {}
+
+bool ARMFrameLowering::noFramePointerElim(const MachineFunction &MF) const {
+ // iOS always has a FP for backtracking, force other targets to keep their FP
+ // when doing FastISel. The emitted code is currently superior, and in cases
+ // like test-suite's lencod FastISel isn't quite correct when FP is eliminated.
+ return TargetFrameLowering::noFramePointerElim(MF) ||
+ MF.getSubtarget<ARMSubtarget>().useFastISel();
+}
+
/// 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 ARMFrameLowering::hasFP(const MachineFunction &MF) const {
- const TargetRegisterInfo *RegInfo = MF.getTarget().getRegisterInfo();
+ const TargetRegisterInfo *RegInfo = MF.getSubtarget().getRegisterInfo();
// iOS requires FP not to be clobbered for backtracing purpose.
- if (STI.isTargetIOS())
+ if (STI.isTargetIOS() || STI.isTargetWatchOS())
return true;
const MachineFrameInfo *MFI = MF.getFrameInfo();
return hasReservedCallFrame(MF) || MF.getFrameInfo()->hasVarSizedObjects();
}
-static bool isCalleeSavedRegister(unsigned Reg, const unsigned *CSRegs) {
- for (unsigned i = 0; CSRegs[i]; ++i)
- if (Reg == CSRegs[i])
- return true;
- return false;
-}
-
static bool isCSRestore(MachineInstr *MI,
const ARMBaseInstrInfo &TII,
- const unsigned *CSRegs) {
+ const MCPhysReg *CSRegs) {
// Integer spill area is handled with "pop".
- if (MI->getOpcode() == ARM::LDMIA_RET ||
- MI->getOpcode() == ARM::t2LDMIA_RET ||
- MI->getOpcode() == ARM::LDMIA_UPD ||
- MI->getOpcode() == ARM::t2LDMIA_UPD ||
- MI->getOpcode() == ARM::VLDMDIA_UPD) {
+ if (isPopOpcode(MI->getOpcode())) {
// The first two operands are predicates. The last two are
// imp-def and imp-use of SP. Check everything in between.
for (int i = 5, e = MI->getNumOperands(); i != e; ++i)
return false;
}
-static void
-emitSPUpdate(bool isARM,
- MachineBasicBlock &MBB, MachineBasicBlock::iterator &MBBI,
- DebugLoc dl, const ARMBaseInstrInfo &TII,
- int NumBytes, unsigned MIFlags = MachineInstr::NoFlags) {
+static void emitRegPlusImmediate(bool isARM, MachineBasicBlock &MBB,
+ MachineBasicBlock::iterator &MBBI, DebugLoc dl,
+ const ARMBaseInstrInfo &TII, unsigned DestReg,
+ unsigned SrcReg, int NumBytes,
+ unsigned MIFlags = MachineInstr::NoFlags,
+ ARMCC::CondCodes Pred = ARMCC::AL,
+ unsigned PredReg = 0) {
if (isARM)
- emitARMRegPlusImmediate(MBB, MBBI, dl, ARM::SP, ARM::SP, NumBytes,
- ARMCC::AL, 0, TII, MIFlags);
+ emitARMRegPlusImmediate(MBB, MBBI, dl, DestReg, SrcReg, NumBytes,
+ Pred, PredReg, TII, MIFlags);
else
- emitT2RegPlusImmediate(MBB, MBBI, dl, ARM::SP, ARM::SP, NumBytes,
- ARMCC::AL, 0, TII, MIFlags);
+ emitT2RegPlusImmediate(MBB, MBBI, dl, DestReg, SrcReg, NumBytes,
+ Pred, PredReg, TII, MIFlags);
}
-void ARMFrameLowering::emitPrologue(MachineFunction &MF) const {
- MachineBasicBlock &MBB = MF.front();
+static void emitSPUpdate(bool isARM, MachineBasicBlock &MBB,
+ MachineBasicBlock::iterator &MBBI, DebugLoc dl,
+ const ARMBaseInstrInfo &TII, int NumBytes,
+ unsigned MIFlags = MachineInstr::NoFlags,
+ ARMCC::CondCodes Pred = ARMCC::AL,
+ unsigned PredReg = 0) {
+ emitRegPlusImmediate(isARM, MBB, MBBI, dl, TII, ARM::SP, ARM::SP, NumBytes,
+ MIFlags, Pred, PredReg);
+}
+
+static int sizeOfSPAdjustment(const MachineInstr *MI) {
+ int RegSize;
+ switch (MI->getOpcode()) {
+ case ARM::VSTMDDB_UPD:
+ RegSize = 8;
+ break;
+ case ARM::STMDB_UPD:
+ case ARM::t2STMDB_UPD:
+ RegSize = 4;
+ break;
+ case ARM::t2STR_PRE:
+ case ARM::STR_PRE_IMM:
+ return 4;
+ default:
+ llvm_unreachable("Unknown push or pop like instruction");
+ }
+
+ int count = 0;
+ // ARM and Thumb2 push/pop insts have explicit "sp, sp" operands (+
+ // pred) so the list starts at 4.
+ for (int i = MI->getNumOperands() - 1; i >= 4; --i)
+ count += RegSize;
+ return count;
+}
+
+static bool WindowsRequiresStackProbe(const MachineFunction &MF,
+ size_t StackSizeInBytes) {
+ const MachineFrameInfo *MFI = MF.getFrameInfo();
+ const Function *F = MF.getFunction();
+ unsigned StackProbeSize = (MFI->getStackProtectorIndex() > 0) ? 4080 : 4096;
+ if (F->hasFnAttribute("stack-probe-size"))
+ F->getFnAttribute("stack-probe-size")
+ .getValueAsString()
+ .getAsInteger(0, StackProbeSize);
+ return StackSizeInBytes >= StackProbeSize;
+}
+
+namespace {
+struct StackAdjustingInsts {
+ struct InstInfo {
+ MachineBasicBlock::iterator I;
+ unsigned SPAdjust;
+ bool BeforeFPSet;
+ };
+
+ SmallVector<InstInfo, 4> Insts;
+
+ void addInst(MachineBasicBlock::iterator I, unsigned SPAdjust,
+ bool BeforeFPSet = false) {
+ InstInfo Info = {I, SPAdjust, BeforeFPSet};
+ Insts.push_back(Info);
+ }
+
+ void addExtraBytes(const MachineBasicBlock::iterator I, unsigned ExtraBytes) {
+ auto Info = std::find_if(Insts.begin(), Insts.end(),
+ [&](InstInfo &Info) { return Info.I == I; });
+ assert(Info != Insts.end() && "invalid sp adjusting instruction");
+ Info->SPAdjust += ExtraBytes;
+ }
+
+ void emitDefCFAOffsets(MachineModuleInfo &MMI, MachineBasicBlock &MBB,
+ DebugLoc dl, const ARMBaseInstrInfo &TII, bool HasFP) {
+ unsigned CFAOffset = 0;
+ for (auto &Info : Insts) {
+ if (HasFP && !Info.BeforeFPSet)
+ return;
+
+ CFAOffset -= Info.SPAdjust;
+ unsigned CFIIndex = MMI.addFrameInst(
+ MCCFIInstruction::createDefCfaOffset(nullptr, CFAOffset));
+ BuildMI(MBB, std::next(Info.I), dl,
+ TII.get(TargetOpcode::CFI_INSTRUCTION))
+ .addCFIIndex(CFIIndex)
+ .setMIFlags(MachineInstr::FrameSetup);
+ }
+ }
+};
+}
+
+/// Emit an instruction sequence that will align the address in
+/// register Reg by zero-ing out the lower bits. For versions of the
+/// architecture that support Neon, this must be done in a single
+/// instruction, since skipAlignedDPRCS2Spills assumes it is done in a
+/// single instruction. That function only gets called when optimizing
+/// spilling of D registers on a core with the Neon instruction set
+/// present.
+static void emitAligningInstructions(MachineFunction &MF, ARMFunctionInfo *AFI,
+ const TargetInstrInfo &TII,
+ MachineBasicBlock &MBB,
+ MachineBasicBlock::iterator MBBI,
+ DebugLoc DL, const unsigned Reg,
+ const unsigned Alignment,
+ const bool MustBeSingleInstruction) {
+ const ARMSubtarget &AST =
+ static_cast<const ARMSubtarget &>(MF.getSubtarget());
+ const bool CanUseBFC = AST.hasV6T2Ops() || AST.hasV7Ops();
+ const unsigned AlignMask = Alignment - 1;
+ const unsigned NrBitsToZero = countTrailingZeros(Alignment);
+ assert(!AFI->isThumb1OnlyFunction() && "Thumb1 not supported");
+ if (!AFI->isThumbFunction()) {
+ // if the BFC instruction is available, use that to zero the lower
+ // bits:
+ // bfc Reg, #0, log2(Alignment)
+ // otherwise use BIC, if the mask to zero the required number of bits
+ // can be encoded in the bic immediate field
+ // bic Reg, Reg, Alignment-1
+ // otherwise, emit
+ // lsr Reg, Reg, log2(Alignment)
+ // lsl Reg, Reg, log2(Alignment)
+ if (CanUseBFC) {
+ AddDefaultPred(BuildMI(MBB, MBBI, DL, TII.get(ARM::BFC), Reg)
+ .addReg(Reg, RegState::Kill)
+ .addImm(~AlignMask));
+ } else if (AlignMask <= 255) {
+ AddDefaultCC(
+ AddDefaultPred(BuildMI(MBB, MBBI, DL, TII.get(ARM::BICri), Reg)
+ .addReg(Reg, RegState::Kill)
+ .addImm(AlignMask)));
+ } else {
+ assert(!MustBeSingleInstruction &&
+ "Shouldn't call emitAligningInstructions demanding a single "
+ "instruction to be emitted for large stack alignment for a target "
+ "without BFC.");
+ AddDefaultCC(AddDefaultPred(
+ BuildMI(MBB, MBBI, DL, TII.get(ARM::MOVsi), Reg)
+ .addReg(Reg, RegState::Kill)
+ .addImm(ARM_AM::getSORegOpc(ARM_AM::lsr, NrBitsToZero))));
+ AddDefaultCC(AddDefaultPred(
+ BuildMI(MBB, MBBI, DL, TII.get(ARM::MOVsi), Reg)
+ .addReg(Reg, RegState::Kill)
+ .addImm(ARM_AM::getSORegOpc(ARM_AM::lsl, NrBitsToZero))));
+ }
+ } else {
+ // Since this is only reached for Thumb-2 targets, the BFC instruction
+ // should always be available.
+ assert(CanUseBFC);
+ AddDefaultPred(BuildMI(MBB, MBBI, DL, TII.get(ARM::t2BFC), Reg)
+ .addReg(Reg, RegState::Kill)
+ .addImm(~AlignMask));
+ }
+}
+
+void ARMFrameLowering::emitPrologue(MachineFunction &MF,
+ MachineBasicBlock &MBB) const {
MachineBasicBlock::iterator MBBI = MBB.begin();
MachineFrameInfo *MFI = MF.getFrameInfo();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
- const ARMBaseRegisterInfo *RegInfo =
- static_cast<const ARMBaseRegisterInfo*>(MF.getTarget().getRegisterInfo());
- const ARMBaseInstrInfo &TII =
- *static_cast<const ARMBaseInstrInfo*>(MF.getTarget().getInstrInfo());
+ MachineModuleInfo &MMI = MF.getMMI();
+ MCContext &Context = MMI.getContext();
+ const TargetMachine &TM = MF.getTarget();
+ const MCRegisterInfo *MRI = Context.getRegisterInfo();
+ const ARMBaseRegisterInfo *RegInfo = STI.getRegisterInfo();
+ const ARMBaseInstrInfo &TII = *STI.getInstrInfo();
assert(!AFI->isThumb1OnlyFunction() &&
"This emitPrologue does not support Thumb1!");
bool isARM = !AFI->isThumbFunction();
- unsigned VARegSaveSize = AFI->getVarArgsRegSaveSize();
+ unsigned Align = STI.getFrameLowering()->getStackAlignment();
+ unsigned ArgRegsSaveSize = AFI->getArgRegsSaveSize();
unsigned NumBytes = MFI->getStackSize();
const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo();
- DebugLoc dl = MBBI != MBB.end() ? MBBI->getDebugLoc() : DebugLoc();
+
+ // Debug location must be unknown since the first debug location is used
+ // to determine the end of the prologue.
+ DebugLoc dl;
+
unsigned FramePtr = RegInfo->getFrameRegister(MF);
// Determine the sizes of each callee-save spill areas and record which frame
int FramePtrSpillFI = 0;
int D8SpillFI = 0;
- // Allocate the vararg register save area. This is not counted in NumBytes.
- if (VARegSaveSize)
- emitSPUpdate(isARM, MBB, MBBI, dl, TII, -VARegSaveSize,
+ // All calls are tail calls in GHC calling conv, and functions have no
+ // prologue/epilogue.
+ if (MF.getFunction()->getCallingConv() == CallingConv::GHC)
+ return;
+
+ StackAdjustingInsts DefCFAOffsetCandidates;
+ bool HasFP = hasFP(MF);
+
+ // Allocate the vararg register save area.
+ if (ArgRegsSaveSize) {
+ emitSPUpdate(isARM, MBB, MBBI, dl, TII, -ArgRegsSaveSize,
MachineInstr::FrameSetup);
+ DefCFAOffsetCandidates.addInst(std::prev(MBBI), ArgRegsSaveSize, true);
+ }
- if (!AFI->hasStackFrame()) {
- if (NumBytes != 0)
- emitSPUpdate(isARM, MBB, MBBI, dl, TII, -NumBytes,
+ if (!AFI->hasStackFrame() &&
+ (!STI.isTargetWindows() || !WindowsRequiresStackProbe(MF, NumBytes))) {
+ if (NumBytes - ArgRegsSaveSize != 0) {
+ emitSPUpdate(isARM, MBB, MBBI, dl, TII, -(NumBytes - ArgRegsSaveSize),
MachineInstr::FrameSetup);
+ DefCFAOffsetCandidates.addInst(std::prev(MBBI),
+ NumBytes - ArgRegsSaveSize, true);
+ }
+ DefCFAOffsetCandidates.emitDefCFAOffsets(MMI, MBB, dl, TII, HasFP);
return;
}
+ // Determine spill area sizes.
for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
unsigned Reg = CSI[i].getReg();
int FI = CSI[i].getFrameIdx();
switch (Reg) {
+ case ARM::R8:
+ case ARM::R9:
+ case ARM::R10:
+ case ARM::R11:
+ case ARM::R12:
+ if (STI.isTargetDarwin()) {
+ GPRCS2Size += 4;
+ break;
+ }
+ // fallthrough
+ case ARM::R0:
+ case ARM::R1:
+ case ARM::R2:
+ case ARM::R3:
case ARM::R4:
case ARM::R5:
case ARM::R6:
case ARM::LR:
if (Reg == FramePtr)
FramePtrSpillFI = FI;
- AFI->addGPRCalleeSavedArea1Frame(FI);
GPRCS1Size += 4;
break;
- case ARM::R8:
- case ARM::R9:
- case ARM::R10:
- case ARM::R11:
- if (Reg == FramePtr)
- FramePtrSpillFI = FI;
- if (STI.isTargetIOS()) {
- AFI->addGPRCalleeSavedArea2Frame(FI);
- GPRCS2Size += 4;
- } else {
- AFI->addGPRCalleeSavedArea1Frame(FI);
- GPRCS1Size += 4;
- }
- break;
default:
// This is a DPR. Exclude the aligned DPRCS2 spills.
if (Reg == ARM::D8)
D8SpillFI = FI;
- if (Reg < ARM::D8 || Reg >= ARM::D8 + AFI->getNumAlignedDPRCS2Regs()) {
- AFI->addDPRCalleeSavedAreaFrame(FI);
+ if (Reg < ARM::D8 || Reg >= ARM::D8 + AFI->getNumAlignedDPRCS2Regs())
DPRCSSize += 8;
- }
}
}
// Move past area 1.
- if (GPRCS1Size > 0) MBBI++;
-
- // Set FP to point to the stack slot that contains the previous FP.
- // For iOS, FP is R7, which has now been stored in spill area 1.
- // Otherwise, if this is not iOS, all the callee-saved registers go
- // into spill area 1, including the FP in R11. In either case, it is
- // now safe to emit this assignment.
- bool HasFP = hasFP(MF);
- if (HasFP) {
- unsigned ADDriOpc = !AFI->isThumbFunction() ? ARM::ADDri : ARM::t2ADDri;
- MachineInstrBuilder MIB =
- BuildMI(MBB, MBBI, dl, TII.get(ADDriOpc), FramePtr)
- .addFrameIndex(FramePtrSpillFI).addImm(0)
- .setMIFlag(MachineInstr::FrameSetup);
- AddDefaultCC(AddDefaultPred(MIB));
+ MachineBasicBlock::iterator LastPush = MBB.end(), GPRCS1Push, GPRCS2Push;
+ if (GPRCS1Size > 0) {
+ GPRCS1Push = LastPush = MBBI++;
+ DefCFAOffsetCandidates.addInst(LastPush, GPRCS1Size, true);
}
- // Move past area 2.
- if (GPRCS2Size > 0) MBBI++;
-
// Determine starting offsets of spill areas.
- unsigned DPRCSOffset = NumBytes - (GPRCS1Size + GPRCS2Size + DPRCSSize);
- unsigned GPRCS2Offset = DPRCSOffset + DPRCSSize;
- unsigned GPRCS1Offset = GPRCS2Offset + GPRCS2Size;
- if (HasFP)
+ unsigned GPRCS1Offset = NumBytes - ArgRegsSaveSize - GPRCS1Size;
+ unsigned GPRCS2Offset = GPRCS1Offset - GPRCS2Size;
+ unsigned DPRAlign = DPRCSSize ? std::min(8U, Align) : 4U;
+ unsigned DPRGapSize = (GPRCS1Size + GPRCS2Size + ArgRegsSaveSize) % DPRAlign;
+ unsigned DPRCSOffset = GPRCS2Offset - DPRGapSize - DPRCSSize;
+ int FramePtrOffsetInPush = 0;
+ if (HasFP) {
+ FramePtrOffsetInPush =
+ MFI->getObjectOffset(FramePtrSpillFI) + ArgRegsSaveSize;
AFI->setFramePtrSpillOffset(MFI->getObjectOffset(FramePtrSpillFI) +
NumBytes);
+ }
AFI->setGPRCalleeSavedArea1Offset(GPRCS1Offset);
AFI->setGPRCalleeSavedArea2Offset(GPRCS2Offset);
AFI->setDPRCalleeSavedAreaOffset(DPRCSOffset);
+ // Move past area 2.
+ if (GPRCS2Size > 0) {
+ GPRCS2Push = LastPush = MBBI++;
+ DefCFAOffsetCandidates.addInst(LastPush, GPRCS2Size);
+ }
+
+ // Prolog/epilog inserter assumes we correctly align DPRs on the stack, so our
+ // .cfi_offset operations will reflect that.
+ if (DPRGapSize) {
+ assert(DPRGapSize == 4 && "unexpected alignment requirements for DPRs");
+ if (tryFoldSPUpdateIntoPushPop(STI, MF, LastPush, DPRGapSize))
+ DefCFAOffsetCandidates.addExtraBytes(LastPush, DPRGapSize);
+ else {
+ emitSPUpdate(isARM, MBB, MBBI, dl, TII, -DPRGapSize,
+ MachineInstr::FrameSetup);
+ DefCFAOffsetCandidates.addInst(std::prev(MBBI), DPRGapSize);
+ }
+ }
+
// Move past area 3.
if (DPRCSSize > 0) {
- MBBI++;
// Since vpush register list cannot have gaps, there may be multiple vpush
// instructions in the prologue.
- while (MBBI->getOpcode() == ARM::VSTMDDB_UPD)
- MBBI++;
+ while (MBBI->getOpcode() == ARM::VSTMDDB_UPD) {
+ DefCFAOffsetCandidates.addInst(MBBI, sizeOfSPAdjustment(MBBI));
+ LastPush = MBBI++;
+ }
}
// Move past the aligned DPRCS2 area.
} else
NumBytes = DPRCSOffset;
+ if (STI.isTargetWindows() && WindowsRequiresStackProbe(MF, NumBytes)) {
+ uint32_t NumWords = NumBytes >> 2;
+
+ if (NumWords < 65536)
+ AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::t2MOVi16), ARM::R4)
+ .addImm(NumWords)
+ .setMIFlags(MachineInstr::FrameSetup));
+ else
+ BuildMI(MBB, MBBI, dl, TII.get(ARM::t2MOVi32imm), ARM::R4)
+ .addImm(NumWords)
+ .setMIFlags(MachineInstr::FrameSetup);
+
+ switch (TM.getCodeModel()) {
+ case CodeModel::Small:
+ case CodeModel::Medium:
+ case CodeModel::Default:
+ case CodeModel::Kernel:
+ BuildMI(MBB, MBBI, dl, TII.get(ARM::tBL))
+ .addImm((unsigned)ARMCC::AL).addReg(0)
+ .addExternalSymbol("__chkstk")
+ .addReg(ARM::R4, RegState::Implicit)
+ .setMIFlags(MachineInstr::FrameSetup);
+ break;
+ case CodeModel::Large:
+ case CodeModel::JITDefault:
+ BuildMI(MBB, MBBI, dl, TII.get(ARM::t2MOVi32imm), ARM::R12)
+ .addExternalSymbol("__chkstk")
+ .setMIFlags(MachineInstr::FrameSetup);
+
+ BuildMI(MBB, MBBI, dl, TII.get(ARM::tBLXr))
+ .addImm((unsigned)ARMCC::AL).addReg(0)
+ .addReg(ARM::R12, RegState::Kill)
+ .addReg(ARM::R4, RegState::Implicit)
+ .setMIFlags(MachineInstr::FrameSetup);
+ break;
+ }
+
+ AddDefaultCC(AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::t2SUBrr),
+ ARM::SP)
+ .addReg(ARM::SP, RegState::Define)
+ .addReg(ARM::R4, RegState::Kill)
+ .setMIFlags(MachineInstr::FrameSetup)));
+ NumBytes = 0;
+ }
+
if (NumBytes) {
// Adjust SP after all the callee-save spills.
- emitSPUpdate(isARM, MBB, MBBI, dl, TII, -NumBytes,
- MachineInstr::FrameSetup);
+ if (AFI->getNumAlignedDPRCS2Regs() == 0 &&
+ tryFoldSPUpdateIntoPushPop(STI, MF, LastPush, NumBytes))
+ DefCFAOffsetCandidates.addExtraBytes(LastPush, NumBytes);
+ else {
+ emitSPUpdate(isARM, MBB, MBBI, dl, TII, -NumBytes,
+ MachineInstr::FrameSetup);
+ DefCFAOffsetCandidates.addInst(std::prev(MBBI), NumBytes);
+ }
+
if (HasFP && isARM)
// Restore from fp only in ARM mode: e.g. sub sp, r7, #24
// Note it's not safe to do this in Thumb2 mode because it would have
AFI->setShouldRestoreSPFromFP(true);
}
+ // Set FP to point to the stack slot that contains the previous FP.
+ // For iOS, FP is R7, which has now been stored in spill area 1.
+ // Otherwise, if this is not iOS, all the callee-saved registers go
+ // into spill area 1, including the FP in R11. In either case, it
+ // is in area one and the adjustment needs to take place just after
+ // that push.
+ if (HasFP) {
+ MachineBasicBlock::iterator AfterPush = std::next(GPRCS1Push);
+ unsigned PushSize = sizeOfSPAdjustment(GPRCS1Push);
+ emitRegPlusImmediate(!AFI->isThumbFunction(), MBB, AfterPush,
+ dl, TII, FramePtr, ARM::SP,
+ PushSize + FramePtrOffsetInPush,
+ MachineInstr::FrameSetup);
+ if (FramePtrOffsetInPush + PushSize != 0) {
+ unsigned CFIIndex = MMI.addFrameInst(MCCFIInstruction::createDefCfa(
+ nullptr, MRI->getDwarfRegNum(FramePtr, true),
+ -(ArgRegsSaveSize - FramePtrOffsetInPush)));
+ BuildMI(MBB, AfterPush, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
+ .addCFIIndex(CFIIndex)
+ .setMIFlags(MachineInstr::FrameSetup);
+ } else {
+ unsigned CFIIndex =
+ MMI.addFrameInst(MCCFIInstruction::createDefCfaRegister(
+ nullptr, MRI->getDwarfRegNum(FramePtr, true)));
+ BuildMI(MBB, AfterPush, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
+ .addCFIIndex(CFIIndex)
+ .setMIFlags(MachineInstr::FrameSetup);
+ }
+ }
+
+ // Now that the prologue's actual instructions are finalised, we can insert
+ // the necessary DWARF cf instructions to describe the situation. Start by
+ // recording where each register ended up:
+ if (GPRCS1Size > 0) {
+ MachineBasicBlock::iterator Pos = std::next(GPRCS1Push);
+ int CFIIndex;
+ for (const auto &Entry : CSI) {
+ unsigned Reg = Entry.getReg();
+ int FI = Entry.getFrameIdx();
+ switch (Reg) {
+ case ARM::R8:
+ case ARM::R9:
+ case ARM::R10:
+ case ARM::R11:
+ case ARM::R12:
+ if (STI.isTargetDarwin())
+ break;
+ // fallthrough
+ case ARM::R0:
+ case ARM::R1:
+ case ARM::R2:
+ case ARM::R3:
+ case ARM::R4:
+ case ARM::R5:
+ case ARM::R6:
+ case ARM::R7:
+ case ARM::LR:
+ CFIIndex = MMI.addFrameInst(MCCFIInstruction::createOffset(
+ nullptr, MRI->getDwarfRegNum(Reg, true), MFI->getObjectOffset(FI)));
+ BuildMI(MBB, Pos, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
+ .addCFIIndex(CFIIndex)
+ .setMIFlags(MachineInstr::FrameSetup);
+ break;
+ }
+ }
+ }
+
+ if (GPRCS2Size > 0) {
+ MachineBasicBlock::iterator Pos = std::next(GPRCS2Push);
+ for (const auto &Entry : CSI) {
+ unsigned Reg = Entry.getReg();
+ int FI = Entry.getFrameIdx();
+ switch (Reg) {
+ case ARM::R8:
+ case ARM::R9:
+ case ARM::R10:
+ case ARM::R11:
+ case ARM::R12:
+ if (STI.isTargetDarwin()) {
+ unsigned DwarfReg = MRI->getDwarfRegNum(Reg, true);
+ unsigned Offset = MFI->getObjectOffset(FI);
+ unsigned CFIIndex = MMI.addFrameInst(
+ MCCFIInstruction::createOffset(nullptr, DwarfReg, Offset));
+ BuildMI(MBB, Pos, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
+ .addCFIIndex(CFIIndex)
+ .setMIFlags(MachineInstr::FrameSetup);
+ }
+ break;
+ }
+ }
+ }
+
+ if (DPRCSSize > 0) {
+ // Since vpush register list cannot have gaps, there may be multiple vpush
+ // instructions in the prologue.
+ MachineBasicBlock::iterator Pos = std::next(LastPush);
+ for (const auto &Entry : CSI) {
+ unsigned Reg = Entry.getReg();
+ int FI = Entry.getFrameIdx();
+ if ((Reg >= ARM::D0 && Reg <= ARM::D31) &&
+ (Reg < ARM::D8 || Reg >= ARM::D8 + AFI->getNumAlignedDPRCS2Regs())) {
+ unsigned DwarfReg = MRI->getDwarfRegNum(Reg, true);
+ unsigned Offset = MFI->getObjectOffset(FI);
+ unsigned CFIIndex = MMI.addFrameInst(
+ MCCFIInstruction::createOffset(nullptr, DwarfReg, Offset));
+ BuildMI(MBB, Pos, dl, TII.get(TargetOpcode::CFI_INSTRUCTION))
+ .addCFIIndex(CFIIndex)
+ .setMIFlags(MachineInstr::FrameSetup);
+ }
+ }
+ }
+
+ // Now we can emit descriptions of where the canonical frame address was
+ // throughout the process. If we have a frame pointer, it takes over the job
+ // half-way through, so only the first few .cfi_def_cfa_offset instructions
+ // actually get emitted.
+ DefCFAOffsetCandidates.emitDefCFAOffsets(MMI, MBB, dl, TII, HasFP);
+
if (STI.isTargetELF() && hasFP(MF))
MFI->setOffsetAdjustment(MFI->getOffsetAdjustment() -
AFI->getFramePtrSpillOffset());
AFI->setGPRCalleeSavedArea1Size(GPRCS1Size);
AFI->setGPRCalleeSavedArea2Size(GPRCS2Size);
+ AFI->setDPRCalleeSavedGapSize(DPRGapSize);
AFI->setDPRCalleeSavedAreaSize(DPRCSSize);
// If we need dynamic stack realignment, do it here. Be paranoid and make
// realigned.
if (!AFI->getNumAlignedDPRCS2Regs() && RegInfo->needsStackRealignment(MF)) {
unsigned MaxAlign = MFI->getMaxAlignment();
- assert (!AFI->isThumb1OnlyFunction());
+ assert(!AFI->isThumb1OnlyFunction());
if (!AFI->isThumbFunction()) {
- // Emit bic sp, sp, MaxAlign
- AddDefaultCC(AddDefaultPred(BuildMI(MBB, MBBI, dl,
- TII.get(ARM::BICri), ARM::SP)
- .addReg(ARM::SP, RegState::Kill)
- .addImm(MaxAlign-1)));
+ emitAligningInstructions(MF, AFI, TII, MBB, MBBI, dl, ARM::SP, MaxAlign,
+ false);
} else {
- // We cannot use sp as source/dest register here, thus we're emitting the
- // following sequence:
+ // We cannot use sp as source/dest register here, thus we're using r4 to
+ // perform the calculations. We're emitting the following sequence:
// mov r4, sp
- // bic r4, r4, MaxAlign
+ // -- use emitAligningInstructions to produce best sequence to zero
+ // -- out lower bits in r4
// mov sp, r4
// FIXME: It will be better just to find spare register here.
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr), ARM::R4)
- .addReg(ARM::SP, RegState::Kill));
- AddDefaultCC(AddDefaultPred(BuildMI(MBB, MBBI, dl,
- TII.get(ARM::t2BICri), ARM::R4)
- .addReg(ARM::R4, RegState::Kill)
- .addImm(MaxAlign-1)));
+ .addReg(ARM::SP, RegState::Kill));
+ emitAligningInstructions(MF, AFI, TII, MBB, MBBI, dl, ARM::R4, MaxAlign,
+ false);
AddDefaultPred(BuildMI(MBB, MBBI, dl, TII.get(ARM::tMOVr), ARM::SP)
- .addReg(ARM::R4, RegState::Kill));
+ .addReg(ARM::R4, RegState::Kill));
}
AFI->setShouldRestoreSPFromFP(true);
void ARMFrameLowering::emitEpilogue(MachineFunction &MF,
MachineBasicBlock &MBB) const {
- MachineBasicBlock::iterator MBBI = MBB.getLastNonDebugInstr();
- assert(MBBI->isReturn() && "Can only insert epilog into returning blocks");
- unsigned RetOpcode = MBBI->getOpcode();
- DebugLoc dl = MBBI->getDebugLoc();
MachineFrameInfo *MFI = MF.getFrameInfo();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
- const TargetRegisterInfo *RegInfo = MF.getTarget().getRegisterInfo();
+ const TargetRegisterInfo *RegInfo = MF.getSubtarget().getRegisterInfo();
const ARMBaseInstrInfo &TII =
- *static_cast<const ARMBaseInstrInfo*>(MF.getTarget().getInstrInfo());
+ *static_cast<const ARMBaseInstrInfo *>(MF.getSubtarget().getInstrInfo());
assert(!AFI->isThumb1OnlyFunction() &&
"This emitEpilogue does not support Thumb1!");
bool isARM = !AFI->isThumbFunction();
- unsigned VARegSaveSize = AFI->getVarArgsRegSaveSize();
+ unsigned ArgRegsSaveSize = AFI->getArgRegsSaveSize();
int NumBytes = (int)MFI->getStackSize();
unsigned FramePtr = RegInfo->getFrameRegister(MF);
+ // All calls are tail calls in GHC calling conv, and functions have no
+ // prologue/epilogue.
+ if (MF.getFunction()->getCallingConv() == CallingConv::GHC)
+ return;
+
+ // First put ourselves on the first (from top) terminator instructions.
+ MachineBasicBlock::iterator MBBI = MBB.getFirstTerminator();
+ DebugLoc dl = MBBI != MBB.end() ? MBBI->getDebugLoc() : DebugLoc();
+
if (!AFI->hasStackFrame()) {
- if (NumBytes != 0)
- emitSPUpdate(isARM, MBB, MBBI, dl, TII, NumBytes);
+ if (NumBytes - ArgRegsSaveSize != 0)
+ emitSPUpdate(isARM, MBB, MBBI, dl, TII, NumBytes - ArgRegsSaveSize);
} else {
// Unwind MBBI to point to first LDR / VLDRD.
- const unsigned *CSRegs = RegInfo->getCalleeSavedRegs();
+ const MCPhysReg *CSRegs = RegInfo->getCalleeSavedRegs(&MF);
if (MBBI != MBB.begin()) {
- do
+ do {
--MBBI;
- while (MBBI != MBB.begin() && isCSRestore(MBBI, TII, CSRegs));
+ } while (MBBI != MBB.begin() && isCSRestore(MBBI, TII, CSRegs));
if (!isCSRestore(MBBI, TII, CSRegs))
++MBBI;
}
// Move SP to start of FP callee save spill area.
- NumBytes -= (AFI->getGPRCalleeSavedArea1Size() +
+ NumBytes -= (ArgRegsSaveSize +
+ AFI->getGPRCalleeSavedArea1Size() +
AFI->getGPRCalleeSavedArea2Size() +
+ AFI->getDPRCalleeSavedGapSize() +
AFI->getDPRCalleeSavedAreaSize());
// Reset SP based on frame pointer only if the stack frame extends beyond
// This is bad, if an interrupt is taken after the mov, sp is in an
// inconsistent state.
// Use the first callee-saved register as a scratch register.
- assert(MF.getRegInfo().isPhysRegUsed(ARM::R4) &&
+ assert(!MFI->getPristineRegs(MF).test(ARM::R4) &&
"No scratch register to restore SP from FP!");
emitT2RegPlusImmediate(MBB, MBBI, dl, ARM::R4, FramePtr, -NumBytes,
ARMCC::AL, 0, TII);
ARM::SP)
.addReg(FramePtr));
}
- } else if (NumBytes)
- emitSPUpdate(isARM, MBB, MBBI, dl, TII, NumBytes);
+ } else if (NumBytes &&
+ !tryFoldSPUpdateIntoPushPop(STI, MF, MBBI, NumBytes))
+ emitSPUpdate(isARM, MBB, MBBI, dl, TII, NumBytes);
// Increment past our save areas.
if (AFI->getDPRCalleeSavedAreaSize()) {
while (MBBI->getOpcode() == ARM::VLDMDIA_UPD)
MBBI++;
}
- if (AFI->getGPRCalleeSavedArea2Size()) MBBI++;
- if (AFI->getGPRCalleeSavedArea1Size()) MBBI++;
- }
-
- if (RetOpcode == ARM::TCRETURNdi || RetOpcode == ARM::TCRETURNdiND ||
- RetOpcode == ARM::TCRETURNri || RetOpcode == ARM::TCRETURNriND) {
- // Tail call return: adjust the stack pointer and jump to callee.
- MBBI = MBB.getLastNonDebugInstr();
- MachineOperand &JumpTarget = MBBI->getOperand(0);
-
- // Jump to label or value in register.
- if (RetOpcode == ARM::TCRETURNdi || RetOpcode == ARM::TCRETURNdiND) {
- unsigned TCOpcode = (RetOpcode == ARM::TCRETURNdi)
- ? (STI.isThumb() ? ARM::tTAILJMPd : ARM::TAILJMPd)
- : (STI.isThumb() ? ARM::tTAILJMPdND : ARM::TAILJMPdND);
- MachineInstrBuilder MIB = BuildMI(MBB, MBBI, dl, TII.get(TCOpcode));
- if (JumpTarget.isGlobal())
- MIB.addGlobalAddress(JumpTarget.getGlobal(), JumpTarget.getOffset(),
- JumpTarget.getTargetFlags());
- else {
- assert(JumpTarget.isSymbol());
- MIB.addExternalSymbol(JumpTarget.getSymbolName(),
- JumpTarget.getTargetFlags());
- }
-
- // Add the default predicate in Thumb mode.
- if (STI.isThumb()) MIB.addImm(ARMCC::AL).addReg(0);
- } else if (RetOpcode == ARM::TCRETURNri) {
- BuildMI(MBB, MBBI, dl,
- TII.get(STI.isThumb() ? ARM::tTAILJMPr : ARM::TAILJMPr)).
- addReg(JumpTarget.getReg(), RegState::Kill);
- } else if (RetOpcode == ARM::TCRETURNriND) {
- BuildMI(MBB, MBBI, dl,
- TII.get(STI.isThumb() ? ARM::tTAILJMPrND : ARM::TAILJMPrND)).
- addReg(JumpTarget.getReg(), RegState::Kill);
+ if (AFI->getDPRCalleeSavedGapSize()) {
+ assert(AFI->getDPRCalleeSavedGapSize() == 4 &&
+ "unexpected DPR alignment gap");
+ emitSPUpdate(isARM, MBB, MBBI, dl, TII, AFI->getDPRCalleeSavedGapSize());
}
- MachineInstr *NewMI = prior(MBBI);
- for (unsigned i = 1, e = MBBI->getNumOperands(); i != e; ++i)
- NewMI->addOperand(MBBI->getOperand(i));
-
- // Delete the pseudo instruction TCRETURN.
- MBB.erase(MBBI);
- MBBI = NewMI;
+ if (AFI->getGPRCalleeSavedArea2Size()) MBBI++;
+ if (AFI->getGPRCalleeSavedArea1Size()) MBBI++;
}
- if (VARegSaveSize)
- emitSPUpdate(isARM, MBB, MBBI, dl, TII, VARegSaveSize);
+ if (ArgRegsSaveSize)
+ emitSPUpdate(isARM, MBB, MBBI, dl, TII, ArgRegsSaveSize);
}
/// getFrameIndexReference - Provide a base+offset reference to an FI slot for
int FI, unsigned &FrameReg,
int SPAdj) const {
const MachineFrameInfo *MFI = MF.getFrameInfo();
- const ARMBaseRegisterInfo *RegInfo =
- static_cast<const ARMBaseRegisterInfo*>(MF.getTarget().getRegisterInfo());
+ const ARMBaseRegisterInfo *RegInfo = static_cast<const ARMBaseRegisterInfo *>(
+ MF.getSubtarget().getRegisterInfo());
const ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
int Offset = MFI->getObjectOffset(FI) + MFI->getStackSize();
int FPOffset = Offset - AFI->getFramePtrSpillOffset();
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();
// SP can move around if there are allocas. We may also lose track of SP
// when emergency spilling inside a non-reserved call frame setup.
- bool hasMovingSP = MFI->hasVarSizedObjects() || !hasReservedCallFrame(MF);
+ bool hasMovingSP = !hasReservedCallFrame(MF);
// When dynamically realigning the stack, use the frame pointer for
// parameters, and the stack/base pointer for locals.
return Offset;
}
-int ARMFrameLowering::getFrameIndexOffset(const MachineFunction &MF,
- int FI) const {
- unsigned FrameReg;
- return getFrameIndexReference(MF, FI, FrameReg);
-}
-
void ARMFrameLowering::emitPushInst(MachineBasicBlock &MBB,
MachineBasicBlock::iterator MI,
const std::vector<CalleeSavedInfo> &CSI,
unsigned NumAlignedDPRCS2Regs,
unsigned MIFlags) const {
MachineFunction &MF = *MBB.getParent();
- const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();
+ const TargetInstrInfo &TII = *MF.getSubtarget().getInstrInfo();
DebugLoc DL;
- if (MI != MBB.end()) DL = MI->getDebugLoc();
SmallVector<std::pair<unsigned,bool>, 4> Regs;
unsigned i = CSI.size();
unsigned LastReg = 0;
for (; i != 0; --i) {
unsigned Reg = CSI[i-1].getReg();
- if (!(Func)(Reg, STI.isTargetIOS())) continue;
+ if (!(Func)(Reg, STI.isTargetDarwin())) continue;
// D-registers in the aligned area DPRCS2 are NOT spilled here.
if (Reg >= ARM::D8 && Reg < ARM::D8 + NumAlignedDPRCS2Regs)
AddDefaultPred(MIB);
}
Regs.clear();
+
+ // Put any subsequent vpush instructions before this one: they will refer to
+ // higher register numbers so need to be pushed first in order to preserve
+ // monotonicity.
+ if (MI != MBB.begin())
+ --MI;
}
}
bool(*Func)(unsigned, bool),
unsigned NumAlignedDPRCS2Regs) const {
MachineFunction &MF = *MBB.getParent();
- const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();
+ const TargetInstrInfo &TII = *MF.getSubtarget().getInstrInfo();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
- DebugLoc DL = MI->getDebugLoc();
- unsigned RetOpcode = MI->getOpcode();
- bool isTailCall = (RetOpcode == ARM::TCRETURNdi ||
- RetOpcode == ARM::TCRETURNdiND ||
- RetOpcode == ARM::TCRETURNri ||
- RetOpcode == ARM::TCRETURNriND);
+ DebugLoc DL;
+ bool isTailCall = false;
+ bool isInterrupt = false;
+ bool isTrap = false;
+ if (MBB.end() != MI) {
+ DL = MI->getDebugLoc();
+ unsigned RetOpcode = MI->getOpcode();
+ isTailCall = (RetOpcode == ARM::TCRETURNdi || RetOpcode == ARM::TCRETURNri);
+ isInterrupt =
+ RetOpcode == ARM::SUBS_PC_LR || RetOpcode == ARM::t2SUBS_PC_LR;
+ isTrap =
+ RetOpcode == ARM::TRAP || RetOpcode == ARM::TRAPNaCl ||
+ RetOpcode == ARM::tTRAP;
+ }
SmallVector<unsigned, 4> Regs;
unsigned i = CSI.size();
bool DeleteRet = false;
for (; i != 0; --i) {
unsigned Reg = CSI[i-1].getReg();
- if (!(Func)(Reg, STI.isTargetIOS())) continue;
+ if (!(Func)(Reg, STI.isTargetDarwin())) continue;
// The aligned reloads from area DPRCS2 are not inserted here.
if (Reg >= ARM::D8 && Reg < ARM::D8 + NumAlignedDPRCS2Regs)
continue;
- if (Reg == ARM::LR && !isTailCall && !isVarArg && STI.hasV5TOps()) {
- Reg = ARM::PC;
- LdmOpc = AFI->isThumbFunction() ? ARM::t2LDMIA_RET : ARM::LDMIA_RET;
+ if (Reg == ARM::LR && !isTailCall && !isVarArg && !isInterrupt &&
+ !isTrap && STI.hasV5TOps()) {
+ if (MBB.succ_empty()) {
+ Reg = ARM::PC;
+ DeleteRet = true;
+ LdmOpc = AFI->isThumbFunction() ? ARM::t2LDMIA_RET : ARM::LDMIA_RET;
+ } else
+ LdmOpc = AFI->isThumbFunction() ? ARM::t2LDMIA_UPD : ARM::LDMIA_UPD;
// Fold the return instruction into the LDM.
- DeleteRet = true;
}
// If NoGap is true, pop consecutive registers and then leave the rest
.addReg(ARM::SP));
for (unsigned i = 0, e = Regs.size(); i < e; ++i)
MIB.addReg(Regs[i], getDefRegState(true));
- if (DeleteRet) {
- MIB->copyImplicitOps(&*MI);
+ if (DeleteRet && MI != MBB.end()) {
+ MIB.copyImplicitOps(&*MI);
MI->eraseFromParent();
}
MI = MIB;
AddDefaultPred(MIB);
}
Regs.clear();
+
+ // Put any subsequent vpop instructions after this one: they will refer to
+ // higher register numbers so need to be popped afterwards.
+ if (MI != MBB.end())
+ ++MI;
}
}
const TargetRegisterInfo *TRI) {
MachineFunction &MF = *MBB.getParent();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
- DebugLoc DL = MI->getDebugLoc();
- const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();
+ DebugLoc DL = MI != MBB.end() ? MI->getDebugLoc() : DebugLoc();
+ const TargetInstrInfo &TII = *MF.getSubtarget().getInstrInfo();
MachineFrameInfo &MFI = *MF.getFrameInfo();
// Mark the D-register spill slots as properly aligned. Since MFI computes
// slot offsets can be wrong. The offset for d8 will always be correct.
for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
unsigned DNum = CSI[i].getReg() - ARM::D8;
- if (DNum >= 8)
+ if (DNum > NumAlignedDPRCS2Regs - 1)
continue;
int FI = CSI[i].getFrameIdx();
// The even-numbered registers will be 16-byte aligned, the odd-numbered
// The immediate is <= 64, so it doesn't need any special encoding.
unsigned Opc = isThumb ? ARM::t2SUBri : ARM::SUBri;
AddDefaultCC(AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(Opc), ARM::R4)
- .addReg(ARM::SP)
- .addImm(8 * NumAlignedDPRCS2Regs)));
+ .addReg(ARM::SP)
+ .addImm(8 * NumAlignedDPRCS2Regs)));
- // bic r4, r4, #align-1
- Opc = isThumb ? ARM::t2BICri : ARM::BICri;
unsigned MaxAlign = MF.getFrameInfo()->getMaxAlignment();
- AddDefaultCC(AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(Opc), ARM::R4)
- .addReg(ARM::R4, RegState::Kill)
- .addImm(MaxAlign - 1)));
+ // We must set parameter MustBeSingleInstruction to true, since
+ // skipAlignedDPRCS2Spills expects exactly 3 instructions to perform
+ // stack alignment. Luckily, this can always be done since all ARM
+ // architecture versions that support Neon also support the BFC
+ // instruction.
+ emitAligningInstructions(MF, AFI, TII, MBB, MI, DL, ARM::R4, MaxAlign, true);
// mov sp, r4
// The stack pointer must be adjusted before spilling anything, otherwise
// The writeback is only needed when emitting two vst1.64 instructions.
if (NumAlignedDPRCS2Regs >= 6) {
unsigned SupReg = TRI->getMatchingSuperReg(NextReg, ARM::dsub_0,
- ARM::QQPRRegisterClass);
+ &ARM::QQPRRegClass);
MBB.addLiveIn(SupReg);
AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(ARM::VST1d64Qwb_fixed),
ARM::R4)
// 16-byte aligned vst1.64 with 4 d-regs, no writeback.
if (NumAlignedDPRCS2Regs >= 4) {
unsigned SupReg = TRI->getMatchingSuperReg(NextReg, ARM::dsub_0,
- ARM::QQPRRegisterClass);
+ &ARM::QQPRRegClass);
MBB.addLiveIn(SupReg);
AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(ARM::VST1d64Q))
.addReg(ARM::R4).addImm(16).addReg(NextReg)
// 16-byte aligned vst1.64 with 2 d-regs.
if (NumAlignedDPRCS2Regs >= 2) {
unsigned SupReg = TRI->getMatchingSuperReg(NextReg, ARM::dsub_0,
- ARM::QPRRegisterClass);
+ &ARM::QPRRegClass);
MBB.addLiveIn(SupReg);
AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(ARM::VST1q64))
- .addReg(ARM::R4).addImm(16).addReg(NextReg)
- .addReg(SupReg, RegState::ImplicitKill));
+ .addReg(ARM::R4).addImm(16).addReg(SupReg));
NextReg += 2;
NumAlignedDPRCS2Regs -= 2;
}
}
// The last spill instruction inserted should kill the scratch register r4.
- llvm::prior(MI)->addRegisterKilled(ARM::R4, TRI);
+ std::prev(MI)->addRegisterKilled(ARM::R4, TRI);
}
/// Skip past the code inserted by emitAlignedDPRCS2Spills, and return an
const TargetRegisterInfo *TRI) {
MachineFunction &MF = *MBB.getParent();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
- DebugLoc DL = MI->getDebugLoc();
- const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();
+ DebugLoc DL = MI != MBB.end() ? MI->getDebugLoc() : DebugLoc();
+ const TargetInstrInfo &TII = *MF.getSubtarget().getInstrInfo();
// Find the frame index assigned to d8.
int D8SpillFI = 0;
// 16-byte aligned vld1.64 with 4 d-regs and writeback.
if (NumAlignedDPRCS2Regs >= 6) {
unsigned SupReg = TRI->getMatchingSuperReg(NextReg, ARM::dsub_0,
- ARM::QQPRRegisterClass);
+ &ARM::QQPRRegClass);
AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(ARM::VLD1d64Qwb_fixed), NextReg)
.addReg(ARM::R4, RegState::Define)
.addReg(ARM::R4, RegState::Kill).addImm(16)
// 16-byte aligned vld1.64 with 4 d-regs, no writeback.
if (NumAlignedDPRCS2Regs >= 4) {
unsigned SupReg = TRI->getMatchingSuperReg(NextReg, ARM::dsub_0,
- ARM::QQPRRegisterClass);
+ &ARM::QQPRRegClass);
AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(ARM::VLD1d64Q), NextReg)
.addReg(ARM::R4).addImm(16)
.addReg(SupReg, RegState::ImplicitDefine));
// 16-byte aligned vld1.64 with 2 d-regs.
if (NumAlignedDPRCS2Regs >= 2) {
unsigned SupReg = TRI->getMatchingSuperReg(NextReg, ARM::dsub_0,
- ARM::QPRRegisterClass);
- AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(ARM::VLD1q64), NextReg)
- .addReg(ARM::R4).addImm(16)
- .addReg(SupReg, RegState::ImplicitDefine));
+ &ARM::QPRRegClass);
+ AddDefaultPred(BuildMI(MBB, MI, DL, TII.get(ARM::VLD1q64), SupReg)
+ .addReg(ARM::R4).addImm(16));
NextReg += 2;
NumAlignedDPRCS2Regs -= 2;
}
.addReg(ARM::R4).addImm(2*(NextReg-R4BaseReg)));
// Last store kills r4.
- llvm::prior(MI)->addRegisterKilled(ARM::R4, TRI);
+ std::prev(MI)->addRegisterKilled(ARM::R4, TRI);
}
bool ARMFrameLowering::spillCalleeSavedRegisters(MachineBasicBlock &MBB,
MachineFunction &MF = *MBB.getParent();
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
- bool isVarArg = AFI->getVarArgsRegSaveSize() > 0;
+ bool isVarArg = AFI->getArgRegsSaveSize() > 0;
unsigned NumAlignedDPRCS2Regs = AFI->getNumAlignedDPRCS2Regs();
// The emitPopInst calls below do not insert reloads for the aligned DPRCS2
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);
+ for (auto &MBB : MF) {
+ for (auto &MI : MBB)
+ FnSize += TII.GetInstSizeInBytes(&MI);
}
return FnSize;
}
-/// estimateStackSize - Estimate and return the size of the frame.
-/// FIXME: Make generic?
-static unsigned estimateStackSize(MachineFunction &MF) {
- const MachineFrameInfo *MFI = MF.getFrameInfo();
- const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering();
- const TargetRegisterInfo *RegInfo = MF.getTarget().getRegisterInfo();
- unsigned MaxAlign = MFI->getMaxAlignment();
- int Offset = 0;
-
- // This code is very, very similar to PEI::calculateFrameObjectOffsets().
- // It really should be refactored to share code. Until then, changes
- // should keep in mind that there's tight coupling between the two.
-
- for (int i = MFI->getObjectIndexBegin(); i != 0; ++i) {
- int FixedOff = -MFI->getObjectOffset(i);
- if (FixedOff > Offset) Offset = FixedOff;
- }
- for (unsigned i = 0, e = MFI->getObjectIndexEnd(); i != e; ++i) {
- if (MFI->isDeadObjectIndex(i))
- continue;
- Offset += MFI->getObjectSize(i);
- unsigned Align = MFI->getObjectAlignment(i);
- // Adjust to alignment boundary
- Offset = (Offset+Align-1)/Align*Align;
-
- MaxAlign = std::max(Align, MaxAlign);
- }
-
- if (MFI->adjustsStack() && TFI->hasReservedCallFrame(MF))
- Offset += MFI->getMaxCallFrameSize();
-
- // Round up the size to a multiple of the alignment. If the function has
- // any calls or alloca's, align to the target's StackAlignment value to
- // ensure that the callee's frame or the alloca data is suitably aligned;
- // otherwise, for leaf functions, align to the TransientStackAlignment
- // value.
- unsigned StackAlign;
- if (MFI->adjustsStack() || MFI->hasVarSizedObjects() ||
- (RegInfo->needsStackRealignment(MF) && MFI->getObjectIndexEnd() != 0))
- StackAlign = TFI->getStackAlignment();
- else
- StackAlign = TFI->getTransientStackAlignment();
-
- // If the frame pointer is eliminated, all frame offsets will be relative to
- // SP not FP. Align to MaxAlign so this works.
- StackAlign = std::max(StackAlign, MaxAlign);
- unsigned AlignMask = StackAlign - 1;
- Offset = (Offset + AlignMask) & ~uint64_t(AlignMask);
-
- 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.
const TargetFrameLowering *TFI) {
const ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
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;
+ for (auto &MBB : MF) {
+ for (auto &MI : MBB) {
+ for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
+ if (!MI.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) {
+ if (MI.getOpcode() == ARM::ADDri) {
Limit = std::min(Limit, (1U << 8) - 1);
break;
}
// Otherwise check the addressing mode.
- switch (I->getDesc().TSFlags & ARMII::AddrModeMask) {
+ switch (MI.getDesc().TSFlags & ARMII::AddrModeMask) {
case ARMII::AddrMode3:
case ARMII::AddrModeT2_i8:
Limit = std::min(Limit, (1U << 8) - 1);
// callee-saved vector registers after realigning the stack. The vst1 and vld1
// instructions take alignment hints that can improve performance.
//
-static void checkNumAlignedDPRCS2Regs(MachineFunction &MF) {
+static void
+checkNumAlignedDPRCS2Regs(MachineFunction &MF, BitVector &SavedRegs) {
MF.getInfo<ARMFunctionInfo>()->setNumAlignedDPRCS2Regs(0);
if (!SpillAlignedNEONRegs)
return;
// Naked functions don't spill callee-saved registers.
- if (MF.getFunction()->hasFnAttr(Attribute::Naked))
+ if (MF.getFunction()->hasFnAttribute(Attribute::Naked))
return;
// We are planning to use NEON instructions vst1 / vld1.
- if (!MF.getTarget().getSubtarget<ARMSubtarget>().hasNEON())
+ if (!static_cast<const ARMSubtarget &>(MF.getSubtarget()).hasNEON())
return;
// Don't bother if the default stack alignment is sufficiently high.
- if (MF.getTarget().getFrameLowering()->getStackAlignment() >= 8)
+ if (MF.getSubtarget().getFrameLowering()->getStackAlignment() >= 8)
return;
// Aligned spills require stack realignment.
- const ARMBaseRegisterInfo *RegInfo =
- static_cast<const ARMBaseRegisterInfo*>(MF.getTarget().getRegisterInfo());
- if (!RegInfo->canRealignStack(MF))
+ if (!static_cast<const ARMBaseRegisterInfo *>(
+ MF.getSubtarget().getRegisterInfo())->canRealignStack(MF))
return;
// We always spill contiguous d-registers starting from d8. Count how many
// callee-saved registers in order, but it can happen that there are holes in
// the range. Registers above the hole will be spilled to the standard DPRCS
// area.
- MachineRegisterInfo &MRI = MF.getRegInfo();
unsigned NumSpills = 0;
for (; NumSpills < 8; ++NumSpills)
- if (!MRI.isPhysRegOrOverlapUsed(ARM::D8 + NumSpills))
+ if (!SavedRegs.test(ARM::D8 + NumSpills))
break;
// Don't do this for just one d-register. It's not worth it.
MF.getInfo<ARMFunctionInfo>()->setNumAlignedDPRCS2Regs(NumSpills);
// A scratch register is required for the vst1 / vld1 instructions.
- MF.getRegInfo().setPhysRegUsed(ARM::R4);
+ SavedRegs.set(ARM::R4);
}
-void
-ARMFrameLowering::processFunctionBeforeCalleeSavedScan(MachineFunction &MF,
- RegScavenger *RS) const {
+void ARMFrameLowering::determineCalleeSaves(MachineFunction &MF,
+ BitVector &SavedRegs,
+ RegScavenger *RS) const {
+ TargetFrameLowering::determineCalleeSaves(MF, SavedRegs, RS);
// 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
unsigned NumGPRSpills = 0;
SmallVector<unsigned, 4> UnspilledCS1GPRs;
SmallVector<unsigned, 4> UnspilledCS2GPRs;
- const ARMBaseRegisterInfo *RegInfo =
- static_cast<const ARMBaseRegisterInfo*>(MF.getTarget().getRegisterInfo());
+ const ARMBaseRegisterInfo *RegInfo = static_cast<const ARMBaseRegisterInfo *>(
+ MF.getSubtarget().getRegisterInfo());
const ARMBaseInstrInfo &TII =
- *static_cast<const ARMBaseInstrInfo*>(MF.getTarget().getInstrInfo());
+ *static_cast<const ARMBaseInstrInfo *>(MF.getSubtarget().getInstrInfo());
ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
MachineFrameInfo *MFI = MF.getFrameInfo();
+ MachineRegisterInfo &MRI = MF.getRegInfo();
unsigned FramePtr = RegInfo->getFrameRegister(MF);
// Spill R4 if Thumb2 function requires stack realignment - it will be used as
// FIXME: It will be better just to find spare register here.
if (AFI->isThumb2Function() &&
(MFI->hasVarSizedObjects() || RegInfo->needsStackRealignment(MF)))
- MF.getRegInfo().setPhysRegUsed(ARM::R4);
+ SavedRegs.set(ARM::R4);
if (AFI->isThumb1OnlyFunction()) {
// Spill LR if Thumb1 function uses variable length argument lists.
- if (AFI->getVarArgsRegSaveSize() > 0)
- MF.getRegInfo().setPhysRegUsed(ARM::LR);
+ if (AFI->getArgRegsSaveSize() > 0)
+ SavedRegs.set(ARM::LR);
// Spill R4 if Thumb1 epilogue has to restore SP from FP. We don't know
// for sure what the stack size will be, but for this, an estimate is good
// we've used all the registers and so R4 is already used, so not marking
// it here will be OK.
// FIXME: It will be better just to find spare register here.
- unsigned StackSize = estimateStackSize(MF);
+ unsigned StackSize = MFI->estimateStackSize(MF);
if (MFI->hasVarSizedObjects() || StackSize > 508)
- MF.getRegInfo().setPhysRegUsed(ARM::R4);
+ SavedRegs.set(ARM::R4);
}
// See if we can spill vector registers to aligned stack.
- checkNumAlignedDPRCS2Regs(MF);
+ checkNumAlignedDPRCS2Regs(MF, SavedRegs);
// Spill the BasePtr if it's used.
if (RegInfo->hasBasePointer(MF))
- MF.getRegInfo().setPhysRegUsed(RegInfo->getBaseRegister());
+ SavedRegs.set(RegInfo->getBaseRegister());
// 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 = RegInfo->getCalleeSavedRegs();
+ // by scanning the callee-save registers to see if any is modified.
+ const MCPhysReg *CSRegs = RegInfo->getCalleeSavedRegs(&MF);
for (unsigned i = 0; CSRegs[i]; ++i) {
unsigned Reg = CSRegs[i];
bool Spilled = false;
- if (MF.getRegInfo().isPhysRegOrOverlapUsed(Reg)) {
+ if (SavedRegs.test(Reg)) {
Spilled = true;
CanEliminateFrame = false;
}
- if (!ARM::GPRRegisterClass->contains(Reg))
+ if (!ARM::GPRRegClass.contains(Reg))
continue;
if (Spilled) {
NumGPRSpills++;
- if (!STI.isTargetIOS()) {
+ if (!STI.isTargetDarwin()) {
if (Reg == ARM::LR)
LRSpilled = true;
CS1Spilled = true;
case ARM::LR:
LRSpilled = true;
// Fallthrough
+ case ARM::R0: case ARM::R1:
+ case ARM::R2: case ARM::R3:
case ARM::R4: case ARM::R5:
case ARM::R6: case ARM::R7:
CS1Spilled = true;
break;
}
} else {
- if (!STI.isTargetIOS()) {
+ if (!STI.isTargetDarwin()) {
UnspilledCS1GPRs.push_back(Reg);
continue;
}
switch (Reg) {
+ case ARM::R0: case ARM::R1:
+ case ARM::R2: case ARM::R3:
case ARM::R4: case ARM::R5:
case ARM::R6: case ARM::R7:
case ARM::LR:
// 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, this)))
- || MFI->hasVarSizedObjects()
- || (MFI->adjustsStack() && !canSimplifyCallFramePseudos(MF));
+ bool BigStack = (RS && (MFI->estimateStackSize(MF) +
+ ((hasFP(MF) && AFI->hasStackFrame()) ? 4 : 0) >=
+ estimateRSStackSizeLimit(MF, this))) ||
+ MFI->hasVarSizedObjects() ||
+ (MFI->adjustsStack() && !canSimplifyCallFramePseudos(MF));
bool ExtraCSSpill = false;
if (BigStack || !CanEliminateFrame || RegInfo->cannotEliminateFrame(MF)) {
// 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);
+ SavedRegs.set(ARM::LR);
NumGPRSpills++;
- UnspilledCS1GPRs.erase(std::find(UnspilledCS1GPRs.begin(),
- UnspilledCS1GPRs.end(), (unsigned)ARM::LR));
+ SmallVectorImpl<unsigned>::iterator LRPos;
+ LRPos = std::find(UnspilledCS1GPRs.begin(), UnspilledCS1GPRs.end(),
+ (unsigned)ARM::LR);
+ if (LRPos != UnspilledCS1GPRs.end())
+ UnspilledCS1GPRs.erase(LRPos);
+
ForceLRSpill = false;
ExtraCSSpill = true;
}
if (hasFP(MF)) {
- MF.getRegInfo().setPhysRegUsed(FramePtr);
+ SavedRegs.set(FramePtr);
+ auto FPPos = std::find(UnspilledCS1GPRs.begin(), UnspilledCS1GPRs.end(),
+ FramePtr);
+ if (FPPos != UnspilledCS1GPRs.end())
+ UnspilledCS1GPRs.erase(FPPos);
NumGPRSpills++;
}
// 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 = getStackAlignment();
- if (TargetAlign == 8 && (NumGPRSpills & 1)) {
+ 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() ||
+ // Don't spill high register if the function is thumb. In the case of
+ // Windows on ARM, accept R11 (frame pointer)
+ if (!AFI->isThumbFunction() ||
+ (STI.isTargetWindows() && Reg == ARM::R11) ||
isARMLowRegister(Reg) || Reg == ARM::LR) {
- MF.getRegInfo().setPhysRegUsed(Reg);
- if (!RegInfo->isReservedReg(MF, Reg))
+ SavedRegs.set(Reg);
+ if (!MRI.isReserved(Reg))
ExtraCSSpill = true;
break;
}
}
} else if (!UnspilledCS2GPRs.empty() && !AFI->isThumb1OnlyFunction()) {
unsigned Reg = UnspilledCS2GPRs.front();
- MF.getRegInfo().setPhysRegUsed(Reg);
- if (!RegInfo->isReservedReg(MF, Reg))
+ SavedRegs.set(Reg);
+ if (!MRI.isReserved(Reg))
ExtraCSSpill = true;
}
}
while (NumExtras && !UnspilledCS1GPRs.empty()) {
unsigned Reg = UnspilledCS1GPRs.back();
UnspilledCS1GPRs.pop_back();
- if (!RegInfo->isReservedReg(MF, Reg) &&
+ if (!MRI.isReserved(Reg) &&
(!AFI->isThumb1OnlyFunction() || isARMLowRegister(Reg) ||
Reg == ARM::LR)) {
Extras.push_back(Reg);
while (NumExtras && !UnspilledCS2GPRs.empty()) {
unsigned Reg = UnspilledCS2GPRs.back();
UnspilledCS2GPRs.pop_back();
- if (!RegInfo->isReservedReg(MF, Reg)) {
+ if (!MRI.isReserved(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]);
+ SavedRegs.set(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(),
+ const TargetRegisterClass *RC = &ARM::GPRRegClass;
+ RS->addScavengingFrameIndex(MFI->CreateStackObject(RC->getSize(),
RC->getAlignment(),
false));
}
}
if (ForceLRSpill) {
- MF.getRegInfo().setPhysRegUsed(ARM::LR);
+ SavedRegs.set(ARM::LR);
AFI->setLRIsSpilledForFarJump(true);
}
}
+
+
+void ARMFrameLowering::
+eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB,
+ MachineBasicBlock::iterator I) const {
+ const ARMBaseInstrInfo &TII =
+ *static_cast<const ARMBaseInstrInfo *>(MF.getSubtarget().getInstrInfo());
+ if (!hasReservedCallFrame(MF)) {
+ // If we have alloca, convert as follows:
+ // ADJCALLSTACKDOWN -> sub, sp, sp, amount
+ // ADJCALLSTACKUP -> add, sp, sp, amount
+ MachineInstr *Old = I;
+ DebugLoc dl = Old->getDebugLoc();
+ unsigned Amount = Old->getOperand(0).getImm();
+ if (Amount != 0) {
+ // 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.
+ Amount = alignSPAdjust(Amount);
+
+ ARMFunctionInfo *AFI = MF.getInfo<ARMFunctionInfo>();
+ assert(!AFI->isThumb1OnlyFunction() &&
+ "This eliminateCallFramePseudoInstr does not support Thumb1!");
+ bool isARM = !AFI->isThumbFunction();
+
+ // Replace the pseudo instruction with a new instruction...
+ unsigned Opc = Old->getOpcode();
+ int PIdx = Old->findFirstPredOperandIdx();
+ ARMCC::CondCodes Pred = (PIdx == -1)
+ ? ARMCC::AL : (ARMCC::CondCodes)Old->getOperand(PIdx).getImm();
+ if (Opc == ARM::ADJCALLSTACKDOWN || Opc == ARM::tADJCALLSTACKDOWN) {
+ // Note: PredReg is operand 2 for ADJCALLSTACKDOWN.
+ unsigned PredReg = Old->getOperand(2).getReg();
+ emitSPUpdate(isARM, MBB, I, dl, TII, -Amount, MachineInstr::NoFlags,
+ Pred, PredReg);
+ } else {
+ // Note: PredReg is operand 3 for ADJCALLSTACKUP.
+ unsigned PredReg = Old->getOperand(3).getReg();
+ assert(Opc == ARM::ADJCALLSTACKUP || Opc == ARM::tADJCALLSTACKUP);
+ emitSPUpdate(isARM, MBB, I, dl, TII, Amount, MachineInstr::NoFlags,
+ Pred, PredReg);
+ }
+ }
+ }
+ MBB.erase(I);
+}
+
+/// Get the minimum constant for ARM that is greater than or equal to the
+/// argument. In ARM, constants can have any value that can be produced by
+/// rotating an 8-bit value to the right by an even number of bits within a
+/// 32-bit word.
+static uint32_t alignToARMConstant(uint32_t Value) {
+ unsigned Shifted = 0;
+
+ if (Value == 0)
+ return 0;
+
+ while (!(Value & 0xC0000000)) {
+ Value = Value << 2;
+ Shifted += 2;
+ }
+
+ bool Carry = (Value & 0x00FFFFFF);
+ Value = ((Value & 0xFF000000) >> 24) + Carry;
+
+ if (Value & 0x0000100)
+ Value = Value & 0x000001FC;
+
+ if (Shifted > 24)
+ Value = Value >> (Shifted - 24);
+ else
+ Value = Value << (24 - Shifted);
+
+ return Value;
+}
+
+// The stack limit in the TCB is set to this many bytes above the actual
+// stack limit.
+static const uint64_t kSplitStackAvailable = 256;
+
+// Adjust the function prologue to enable split stacks. This currently only
+// supports android and linux.
+//
+// The ABI of the segmented stack prologue is a little arbitrarily chosen, but
+// must be well defined in order to allow for consistent implementations of the
+// __morestack helper function. The ABI is also not a normal ABI in that it
+// doesn't follow the normal calling conventions because this allows the
+// prologue of each function to be optimized further.
+//
+// Currently, the ABI looks like (when calling __morestack)
+//
+// * r4 holds the minimum stack size requested for this function call
+// * r5 holds the stack size of the arguments to the function
+// * the beginning of the function is 3 instructions after the call to
+// __morestack
+//
+// Implementations of __morestack should use r4 to allocate a new stack, r5 to
+// place the arguments on to the new stack, and the 3-instruction knowledge to
+// jump directly to the body of the function when working on the new stack.
+//
+// An old (and possibly no longer compatible) implementation of __morestack for
+// ARM can be found at [1].
+//
+// [1] - https://github.com/mozilla/rust/blob/86efd9/src/rt/arch/arm/morestack.S
+void ARMFrameLowering::adjustForSegmentedStacks(
+ MachineFunction &MF, MachineBasicBlock &PrologueMBB) const {
+ unsigned Opcode;
+ unsigned CFIIndex;
+ const ARMSubtarget *ST = &MF.getSubtarget<ARMSubtarget>();
+ bool Thumb = ST->isThumb();
+
+ // Sadly, this currently doesn't support varargs, platforms other than
+ // android/linux. Note that thumb1/thumb2 are support for android/linux.
+ if (MF.getFunction()->isVarArg())
+ report_fatal_error("Segmented stacks do not support vararg functions.");
+ if (!ST->isTargetAndroid() && !ST->isTargetLinux())
+ report_fatal_error("Segmented stacks not supported on this platform.");
+
+ MachineFrameInfo *MFI = MF.getFrameInfo();
+ MachineModuleInfo &MMI = MF.getMMI();
+ MCContext &Context = MMI.getContext();
+ const MCRegisterInfo *MRI = Context.getRegisterInfo();
+ const ARMBaseInstrInfo &TII =
+ *static_cast<const ARMBaseInstrInfo *>(MF.getSubtarget().getInstrInfo());
+ ARMFunctionInfo *ARMFI = MF.getInfo<ARMFunctionInfo>();
+ DebugLoc DL;
+
+ uint64_t StackSize = MFI->getStackSize();
+
+ // Do not generate a prologue for functions with a stack of size zero
+ if (StackSize == 0)
+ return;
+
+ // Use R4 and R5 as scratch registers.
+ // We save R4 and R5 before use and restore them before leaving the function.
+ unsigned ScratchReg0 = ARM::R4;
+ unsigned ScratchReg1 = ARM::R5;
+ uint64_t AlignedStackSize;
+
+ MachineBasicBlock *PrevStackMBB = MF.CreateMachineBasicBlock();
+ MachineBasicBlock *PostStackMBB = MF.CreateMachineBasicBlock();
+ MachineBasicBlock *AllocMBB = MF.CreateMachineBasicBlock();
+ MachineBasicBlock *GetMBB = MF.CreateMachineBasicBlock();
+ MachineBasicBlock *McrMBB = MF.CreateMachineBasicBlock();
+
+ // Grab everything that reaches PrologueMBB to update there liveness as well.
+ SmallPtrSet<MachineBasicBlock *, 8> BeforePrologueRegion;
+ SmallVector<MachineBasicBlock *, 2> WalkList;
+ WalkList.push_back(&PrologueMBB);
+
+ do {
+ MachineBasicBlock *CurMBB = WalkList.pop_back_val();
+ for (MachineBasicBlock *PredBB : CurMBB->predecessors()) {
+ if (BeforePrologueRegion.insert(PredBB).second)
+ WalkList.push_back(PredBB);
+ }
+ } while (!WalkList.empty());
+
+ // The order in that list is important.
+ // The blocks will all be inserted before PrologueMBB using that order.
+ // Therefore the block that should appear first in the CFG should appear
+ // first in the list.
+ MachineBasicBlock *AddedBlocks[] = {PrevStackMBB, McrMBB, GetMBB, AllocMBB,
+ PostStackMBB};
+
+ for (MachineBasicBlock *B : AddedBlocks)
+ BeforePrologueRegion.insert(B);
+
+ for (const auto &LI : PrologueMBB.liveins()) {
+ for (MachineBasicBlock *PredBB : BeforePrologueRegion)
+ PredBB->addLiveIn(LI);
+ }
+
+ // Remove the newly added blocks from the list, since we know
+ // we do not have to do the following updates for them.
+ for (MachineBasicBlock *B : AddedBlocks) {
+ BeforePrologueRegion.erase(B);
+ MF.insert(PrologueMBB.getIterator(), B);
+ }
+
+ for (MachineBasicBlock *MBB : BeforePrologueRegion) {
+ // Make sure the LiveIns are still sorted and unique.
+ MBB->sortUniqueLiveIns();
+ // Replace the edges to PrologueMBB by edges to the sequences
+ // we are about to add.
+ MBB->ReplaceUsesOfBlockWith(&PrologueMBB, AddedBlocks[0]);
+ }
+
+ // The required stack size that is aligned to ARM constant criterion.
+ AlignedStackSize = alignToARMConstant(StackSize);
+
+ // When the frame size is less than 256 we just compare the stack
+ // boundary directly to the value of the stack pointer, per gcc.
+ bool CompareStackPointer = AlignedStackSize < kSplitStackAvailable;
+
+ // We will use two of the callee save registers as scratch registers so we
+ // need to save those registers onto the stack.
+ // We will use SR0 to hold stack limit and SR1 to hold the stack size
+ // requested and arguments for __morestack().
+ // SR0: Scratch Register #0
+ // SR1: Scratch Register #1
+ // push {SR0, SR1}
+ if (Thumb) {
+ AddDefaultPred(BuildMI(PrevStackMBB, DL, TII.get(ARM::tPUSH)))
+ .addReg(ScratchReg0).addReg(ScratchReg1);
+ } else {
+ AddDefaultPred(BuildMI(PrevStackMBB, DL, TII.get(ARM::STMDB_UPD))
+ .addReg(ARM::SP, RegState::Define).addReg(ARM::SP))
+ .addReg(ScratchReg0).addReg(ScratchReg1);
+ }
+
+ // Emit the relevant DWARF information about the change in stack pointer as
+ // well as where to find both r4 and r5 (the callee-save registers)
+ CFIIndex =
+ MMI.addFrameInst(MCCFIInstruction::createDefCfaOffset(nullptr, -8));
+ BuildMI(PrevStackMBB, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
+ .addCFIIndex(CFIIndex);
+ CFIIndex = MMI.addFrameInst(MCCFIInstruction::createOffset(
+ nullptr, MRI->getDwarfRegNum(ScratchReg1, true), -4));
+ BuildMI(PrevStackMBB, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
+ .addCFIIndex(CFIIndex);
+ CFIIndex = MMI.addFrameInst(MCCFIInstruction::createOffset(
+ nullptr, MRI->getDwarfRegNum(ScratchReg0, true), -8));
+ BuildMI(PrevStackMBB, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
+ .addCFIIndex(CFIIndex);
+
+ // mov SR1, sp
+ if (Thumb) {
+ AddDefaultPred(BuildMI(McrMBB, DL, TII.get(ARM::tMOVr), ScratchReg1)
+ .addReg(ARM::SP));
+ } else if (CompareStackPointer) {
+ AddDefaultPred(BuildMI(McrMBB, DL, TII.get(ARM::MOVr), ScratchReg1)
+ .addReg(ARM::SP)).addReg(0);
+ }
+
+ // sub SR1, sp, #StackSize
+ if (!CompareStackPointer && Thumb) {
+ AddDefaultPred(
+ AddDefaultCC(BuildMI(McrMBB, DL, TII.get(ARM::tSUBi8), ScratchReg1))
+ .addReg(ScratchReg1).addImm(AlignedStackSize));
+ } else if (!CompareStackPointer) {
+ AddDefaultPred(BuildMI(McrMBB, DL, TII.get(ARM::SUBri), ScratchReg1)
+ .addReg(ARM::SP).addImm(AlignedStackSize)).addReg(0);
+ }
+
+ if (Thumb && ST->isThumb1Only()) {
+ unsigned PCLabelId = ARMFI->createPICLabelUId();
+ ARMConstantPoolValue *NewCPV = ARMConstantPoolSymbol::Create(
+ MF.getFunction()->getContext(), "__STACK_LIMIT", PCLabelId, 0);
+ MachineConstantPool *MCP = MF.getConstantPool();
+ unsigned CPI = MCP->getConstantPoolIndex(NewCPV, 4);
+
+ // ldr SR0, [pc, offset(STACK_LIMIT)]
+ AddDefaultPred(BuildMI(GetMBB, DL, TII.get(ARM::tLDRpci), ScratchReg0)
+ .addConstantPoolIndex(CPI));
+
+ // ldr SR0, [SR0]
+ AddDefaultPred(BuildMI(GetMBB, DL, TII.get(ARM::tLDRi), ScratchReg0)
+ .addReg(ScratchReg0).addImm(0));
+ } else {
+ // Get TLS base address from the coprocessor
+ // mrc p15, #0, SR0, c13, c0, #3
+ AddDefaultPred(BuildMI(McrMBB, DL, TII.get(ARM::MRC), ScratchReg0)
+ .addImm(15)
+ .addImm(0)
+ .addImm(13)
+ .addImm(0)
+ .addImm(3));
+
+ // Use the last tls slot on android and a private field of the TCP on linux.
+ assert(ST->isTargetAndroid() || ST->isTargetLinux());
+ unsigned TlsOffset = ST->isTargetAndroid() ? 63 : 1;
+
+ // Get the stack limit from the right offset
+ // ldr SR0, [sr0, #4 * TlsOffset]
+ AddDefaultPred(BuildMI(GetMBB, DL, TII.get(ARM::LDRi12), ScratchReg0)
+ .addReg(ScratchReg0).addImm(4 * TlsOffset));
+ }
+
+ // Compare stack limit with stack size requested.
+ // cmp SR0, SR1
+ Opcode = Thumb ? ARM::tCMPr : ARM::CMPrr;
+ AddDefaultPred(BuildMI(GetMBB, DL, TII.get(Opcode))
+ .addReg(ScratchReg0)
+ .addReg(ScratchReg1));
+
+ // This jump is taken if StackLimit < SP - stack required.
+ Opcode = Thumb ? ARM::tBcc : ARM::Bcc;
+ BuildMI(GetMBB, DL, TII.get(Opcode)).addMBB(PostStackMBB)
+ .addImm(ARMCC::LO)
+ .addReg(ARM::CPSR);
+
+
+ // Calling __morestack(StackSize, Size of stack arguments).
+ // __morestack knows that the stack size requested is in SR0(r4)
+ // and amount size of stack arguments is in SR1(r5).
+
+ // Pass first argument for the __morestack by Scratch Register #0.
+ // The amount size of stack required
+ if (Thumb) {
+ AddDefaultPred(AddDefaultCC(BuildMI(AllocMBB, DL, TII.get(ARM::tMOVi8),
+ ScratchReg0)).addImm(AlignedStackSize));
+ } else {
+ AddDefaultPred(BuildMI(AllocMBB, DL, TII.get(ARM::MOVi), ScratchReg0)
+ .addImm(AlignedStackSize)).addReg(0);
+ }
+ // Pass second argument for the __morestack by Scratch Register #1.
+ // The amount size of stack consumed to save function arguments.
+ if (Thumb) {
+ AddDefaultPred(
+ AddDefaultCC(BuildMI(AllocMBB, DL, TII.get(ARM::tMOVi8), ScratchReg1))
+ .addImm(alignToARMConstant(ARMFI->getArgumentStackSize())));
+ } else {
+ AddDefaultPred(BuildMI(AllocMBB, DL, TII.get(ARM::MOVi), ScratchReg1)
+ .addImm(alignToARMConstant(ARMFI->getArgumentStackSize())))
+ .addReg(0);
+ }
+
+ // push {lr} - Save return address of this function.
+ if (Thumb) {
+ AddDefaultPred(BuildMI(AllocMBB, DL, TII.get(ARM::tPUSH)))
+ .addReg(ARM::LR);
+ } else {
+ AddDefaultPred(BuildMI(AllocMBB, DL, TII.get(ARM::STMDB_UPD))
+ .addReg(ARM::SP, RegState::Define)
+ .addReg(ARM::SP))
+ .addReg(ARM::LR);
+ }
+
+ // Emit the DWARF info about the change in stack as well as where to find the
+ // previous link register
+ CFIIndex =
+ MMI.addFrameInst(MCCFIInstruction::createDefCfaOffset(nullptr, -12));
+ BuildMI(AllocMBB, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
+ .addCFIIndex(CFIIndex);
+ CFIIndex = MMI.addFrameInst(MCCFIInstruction::createOffset(
+ nullptr, MRI->getDwarfRegNum(ARM::LR, true), -12));
+ BuildMI(AllocMBB, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
+ .addCFIIndex(CFIIndex);
+
+ // Call __morestack().
+ if (Thumb) {
+ AddDefaultPred(BuildMI(AllocMBB, DL, TII.get(ARM::tBL)))
+ .addExternalSymbol("__morestack");
+ } else {
+ BuildMI(AllocMBB, DL, TII.get(ARM::BL))
+ .addExternalSymbol("__morestack");
+ }
+
+ // pop {lr} - Restore return address of this original function.
+ if (Thumb) {
+ if (ST->isThumb1Only()) {
+ AddDefaultPred(BuildMI(AllocMBB, DL, TII.get(ARM::tPOP)))
+ .addReg(ScratchReg0);
+ AddDefaultPred(BuildMI(AllocMBB, DL, TII.get(ARM::tMOVr), ARM::LR)
+ .addReg(ScratchReg0));
+ } else {
+ AddDefaultPred(BuildMI(AllocMBB, DL, TII.get(ARM::t2LDR_POST))
+ .addReg(ARM::LR, RegState::Define)
+ .addReg(ARM::SP, RegState::Define)
+ .addReg(ARM::SP)
+ .addImm(4));
+ }
+ } else {
+ AddDefaultPred(BuildMI(AllocMBB, DL, TII.get(ARM::LDMIA_UPD))
+ .addReg(ARM::SP, RegState::Define)
+ .addReg(ARM::SP))
+ .addReg(ARM::LR);
+ }
+
+ // Restore SR0 and SR1 in case of __morestack() was called.
+ // __morestack() will skip PostStackMBB block so we need to restore
+ // scratch registers from here.
+ // pop {SR0, SR1}
+ if (Thumb) {
+ AddDefaultPred(BuildMI(AllocMBB, DL, TII.get(ARM::tPOP)))
+ .addReg(ScratchReg0)
+ .addReg(ScratchReg1);
+ } else {
+ AddDefaultPred(BuildMI(AllocMBB, DL, TII.get(ARM::LDMIA_UPD))
+ .addReg(ARM::SP, RegState::Define)
+ .addReg(ARM::SP))
+ .addReg(ScratchReg0)
+ .addReg(ScratchReg1);
+ }
+
+ // Update the CFA offset now that we've popped
+ CFIIndex = MMI.addFrameInst(MCCFIInstruction::createDefCfaOffset(nullptr, 0));
+ BuildMI(AllocMBB, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
+ .addCFIIndex(CFIIndex);
+
+ // bx lr - Return from this function.
+ Opcode = Thumb ? ARM::tBX_RET : ARM::BX_RET;
+ AddDefaultPred(BuildMI(AllocMBB, DL, TII.get(Opcode)));
+
+ // Restore SR0 and SR1 in case of __morestack() was not called.
+ // pop {SR0, SR1}
+ if (Thumb) {
+ AddDefaultPred(BuildMI(PostStackMBB, DL, TII.get(ARM::tPOP)))
+ .addReg(ScratchReg0)
+ .addReg(ScratchReg1);
+ } else {
+ AddDefaultPred(BuildMI(PostStackMBB, DL, TII.get(ARM::LDMIA_UPD))
+ .addReg(ARM::SP, RegState::Define)
+ .addReg(ARM::SP))
+ .addReg(ScratchReg0)
+ .addReg(ScratchReg1);
+ }
+
+ // Update the CFA offset now that we've popped
+ CFIIndex = MMI.addFrameInst(MCCFIInstruction::createDefCfaOffset(nullptr, 0));
+ BuildMI(PostStackMBB, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
+ .addCFIIndex(CFIIndex);
+
+ // Tell debuggers that r4 and r5 are now the same as they were in the
+ // previous function, that they're the "Same Value".
+ CFIIndex = MMI.addFrameInst(MCCFIInstruction::createSameValue(
+ nullptr, MRI->getDwarfRegNum(ScratchReg0, true)));
+ BuildMI(PostStackMBB, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
+ .addCFIIndex(CFIIndex);
+ CFIIndex = MMI.addFrameInst(MCCFIInstruction::createSameValue(
+ nullptr, MRI->getDwarfRegNum(ScratchReg1, true)));
+ BuildMI(PostStackMBB, DL, TII.get(TargetOpcode::CFI_INSTRUCTION))
+ .addCFIIndex(CFIIndex);
+
+ // Organizing MBB lists
+ PostStackMBB->addSuccessor(&PrologueMBB);
+
+ AllocMBB->addSuccessor(PostStackMBB);
+
+ GetMBB->addSuccessor(PostStackMBB);
+ GetMBB->addSuccessor(AllocMBB);
+
+ McrMBB->addSuccessor(GetMBB);
+
+ PrevStackMBB->addSuccessor(McrMBB);
+
+#ifdef XDEBUG
+ MF.verify();
+#endif
+}