-//===- PPCRegisterInfo.cpp - PowerPC Register Information -------*- C++ -*-===//
+//===-- PPCRegisterInfo.cpp - PowerPC Register Information ----------------===//
//
// The LLVM Compiler Infrastructure
//
//===----------------------------------------------------------------------===//
#define DEBUG_TYPE "reginfo"
+#include "PPCRegisterInfo.h"
#include "PPC.h"
+#include "PPCFrameLowering.h"
#include "PPCInstrBuilder.h"
#include "PPCMachineFunctionInfo.h"
-#include "PPCRegisterInfo.h"
-#include "PPCFrameInfo.h"
#include "PPCSubtarget.h"
-#include "llvm/CallingConv.h"
-#include "llvm/Constants.h"
-#include "llvm/Function.h"
-#include "llvm/Type.h"
-#include "llvm/CodeGen/ValueTypes.h"
+#include "llvm/ADT/BitVector.h"
+#include "llvm/ADT/STLExtras.h"
+#include "llvm/CodeGen/MachineFrameInfo.h"
+#include "llvm/CodeGen/MachineFunction.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineModuleInfo.h"
-#include "llvm/CodeGen/MachineFunction.h"
-#include "llvm/CodeGen/MachineFrameInfo.h"
-#include "llvm/CodeGen/MachineLocation.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/CodeGen/RegisterScavenging.h"
-#include "llvm/Target/TargetFrameInfo.h"
-#include "llvm/Target/TargetInstrInfo.h"
-#include "llvm/Target/TargetMachine.h"
-#include "llvm/Target/TargetOptions.h"
+#include "llvm/IR/CallingConv.h"
+#include "llvm/IR/Constants.h"
+#include "llvm/IR/Function.h"
+#include "llvm/IR/Type.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/MathExtras.h"
#include "llvm/Support/raw_ostream.h"
-#include "llvm/ADT/BitVector.h"
-#include "llvm/ADT/STLExtras.h"
+#include "llvm/Target/TargetFrameLowering.h"
+#include "llvm/Target/TargetInstrInfo.h"
+#include "llvm/Target/TargetMachine.h"
+#include "llvm/Target/TargetOptions.h"
#include <cstdlib>
-// FIXME This disables some code that aligns the stack to a boundary
-// bigger than the default (16 bytes on Darwin) when there is a stack local
-// of greater alignment. This does not currently work, because the delta
-// between old and new stack pointers is added to offsets that reference
-// incoming parameters after the prolog is generated, and the code that
-// does that doesn't handle a variable delta. You don't want to do that
-// anyway; a better approach is to reserve another register that retains
-// to the incoming stack pointer, and reference parameters relative to that.
-#define ALIGN_STACK 0
-
-// FIXME (64-bit): Eventually enable by default.
-namespace llvm {
-cl::opt<bool> EnablePPC32RS("enable-ppc32-regscavenger",
- cl::init(false),
- cl::desc("Enable PPC32 register scavenger"),
- cl::Hidden);
-cl::opt<bool> EnablePPC64RS("enable-ppc64-regscavenger",
- cl::init(false),
- cl::desc("Enable PPC64 register scavenger"),
- cl::Hidden);
-}
+#define GET_REGINFO_TARGET_DESC
+#include "PPCGenRegisterInfo.inc"
using namespace llvm;
-#define EnableRegisterScavenging \
- ((EnablePPC32RS && !Subtarget.isPPC64()) || \
- (EnablePPC64RS && Subtarget.isPPC64()))
+static cl::opt<bool>
+EnableBasePointer("ppc-use-base-pointer", cl::Hidden, cl::init(true),
+ cl::desc("Enable use of a base pointer for complex stack frames"));
-// FIXME (64-bit): Should be inlined.
-bool
-PPCRegisterInfo::requiresRegisterScavenging(const MachineFunction &) const {
- return EnableRegisterScavenging;
-}
+static cl::opt<bool>
+AlwaysBasePointer("ppc-always-use-base-pointer", cl::Hidden, cl::init(false),
+ cl::desc("Force the use of a base pointer in every function"));
-/// getRegisterNumbering - Given the enum value for some register, e.g.
-/// PPC::F14, return the number that it corresponds to (e.g. 14).
-unsigned PPCRegisterInfo::getRegisterNumbering(unsigned RegEnum) {
- using namespace PPC;
- switch (RegEnum) {
- case 0: return 0;
- case R0 : case X0 : case F0 : case V0 : case CR0: case CR0LT: return 0;
- case R1 : case X1 : case F1 : case V1 : case CR1: case CR0GT: return 1;
- case R2 : case X2 : case F2 : case V2 : case CR2: case CR0EQ: return 2;
- case R3 : case X3 : case F3 : case V3 : case CR3: case CR0UN: return 3;
- case R4 : case X4 : case F4 : case V4 : case CR4: case CR1LT: return 4;
- case R5 : case X5 : case F5 : case V5 : case CR5: case CR1GT: return 5;
- case R6 : case X6 : case F6 : case V6 : case CR6: case CR1EQ: return 6;
- case R7 : case X7 : case F7 : case V7 : case CR7: case CR1UN: return 7;
- case R8 : case X8 : case F8 : case V8 : case CR2LT: return 8;
- case R9 : case X9 : case F9 : case V9 : case CR2GT: return 9;
- case R10: case X10: case F10: case V10: case CR2EQ: return 10;
- case R11: case X11: case F11: case V11: case CR2UN: return 11;
- case R12: case X12: case F12: case V12: case CR3LT: return 12;
- case R13: case X13: case F13: case V13: case CR3GT: return 13;
- case R14: case X14: case F14: case V14: case CR3EQ: return 14;
- case R15: case X15: case F15: case V15: case CR3UN: return 15;
- case R16: case X16: case F16: case V16: case CR4LT: return 16;
- case R17: case X17: case F17: case V17: case CR4GT: return 17;
- case R18: case X18: case F18: case V18: case CR4EQ: return 18;
- case R19: case X19: case F19: case V19: case CR4UN: return 19;
- case R20: case X20: case F20: case V20: case CR5LT: return 20;
- case R21: case X21: case F21: case V21: case CR5GT: return 21;
- case R22: case X22: case F22: case V22: case CR5EQ: return 22;
- case R23: case X23: case F23: case V23: case CR5UN: return 23;
- case R24: case X24: case F24: case V24: case CR6LT: return 24;
- case R25: case X25: case F25: case V25: case CR6GT: return 25;
- case R26: case X26: case F26: case V26: case CR6EQ: return 26;
- case R27: case X27: case F27: case V27: case CR6UN: return 27;
- case R28: case X28: case F28: case V28: case CR7LT: return 28;
- case R29: case X29: case F29: case V29: case CR7GT: return 29;
- case R30: case X30: case F30: case V30: case CR7EQ: return 30;
- case R31: case X31: case F31: case V31: case CR7UN: return 31;
- default:
- llvm_unreachable("Unhandled reg in PPCRegisterInfo::getRegisterNumbering!");
- }
-}
-
-PPCRegisterInfo::PPCRegisterInfo(const PPCSubtarget &ST,
- const TargetInstrInfo &tii)
- : PPCGenRegisterInfo(PPC::ADJCALLSTACKDOWN, PPC::ADJCALLSTACKUP),
- Subtarget(ST), TII(tii) {
+PPCRegisterInfo::PPCRegisterInfo(const PPCSubtarget &ST)
+ : PPCGenRegisterInfo(ST.isPPC64() ? PPC::LR8 : PPC::LR,
+ ST.isPPC64() ? 0 : 1,
+ ST.isPPC64() ? 0 : 1),
+ Subtarget(ST) {
ImmToIdxMap[PPC::LD] = PPC::LDX; ImmToIdxMap[PPC::STD] = PPC::STDX;
ImmToIdxMap[PPC::LBZ] = PPC::LBZX; ImmToIdxMap[PPC::STB] = PPC::STBX;
ImmToIdxMap[PPC::LHZ] = PPC::LHZX; ImmToIdxMap[PPC::LHA] = PPC::LHAX;
ImmToIdxMap[PPC::STH] = PPC::STHX; ImmToIdxMap[PPC::STW] = PPC::STWX;
ImmToIdxMap[PPC::STFS] = PPC::STFSX; ImmToIdxMap[PPC::STFD] = PPC::STFDX;
ImmToIdxMap[PPC::ADDI] = PPC::ADD4;
+ ImmToIdxMap[PPC::LWA_32] = PPC::LWAX_32;
// 64-bit
ImmToIdxMap[PPC::LHA8] = PPC::LHAX8; ImmToIdxMap[PPC::LBZ8] = PPC::LBZX8;
ImmToIdxMap[PPC::LHZ8] = PPC::LHZX8; ImmToIdxMap[PPC::LWZ8] = PPC::LWZX8;
ImmToIdxMap[PPC::STB8] = PPC::STBX8; ImmToIdxMap[PPC::STH8] = PPC::STHX8;
ImmToIdxMap[PPC::STW8] = PPC::STWX8; ImmToIdxMap[PPC::STDU] = PPC::STDUX;
- ImmToIdxMap[PPC::ADDI8] = PPC::ADD8; ImmToIdxMap[PPC::STD_32] = PPC::STDX_32;
+ ImmToIdxMap[PPC::ADDI8] = PPC::ADD8;
}
/// getPointerRegClass - Return the register class to use to hold pointers.
/// This is used for addressing modes.
const TargetRegisterClass *
-PPCRegisterInfo::getPointerRegClass(unsigned Kind) const {
+PPCRegisterInfo::getPointerRegClass(const MachineFunction &MF, unsigned Kind)
+ const {
+ // Note that PPCInstrInfo::FoldImmediate also directly uses this Kind value
+ // when it checks for ZERO folding.
+ if (Kind == 1) {
+ if (Subtarget.isPPC64())
+ return &PPC::G8RC_NOX0RegClass;
+ return &PPC::GPRC_NOR0RegClass;
+ }
+
if (Subtarget.isPPC64())
return &PPC::G8RCRegClass;
return &PPC::GPRCRegClass;
}
-const unsigned*
+const uint16_t*
PPCRegisterInfo::getCalleeSavedRegs(const MachineFunction *MF) const {
- // 32-bit Darwin calling convention.
- static const unsigned Darwin32_CalleeSavedRegs[] = {
- PPC::R13, PPC::R14, PPC::R15,
- PPC::R16, PPC::R17, PPC::R18, PPC::R19,
- PPC::R20, PPC::R21, PPC::R22, PPC::R23,
- PPC::R24, PPC::R25, PPC::R26, PPC::R27,
- PPC::R28, PPC::R29, PPC::R30, PPC::R31,
-
- PPC::F14, PPC::F15, PPC::F16, PPC::F17,
- PPC::F18, PPC::F19, PPC::F20, PPC::F21,
- PPC::F22, PPC::F23, PPC::F24, PPC::F25,
- PPC::F26, PPC::F27, PPC::F28, PPC::F29,
- PPC::F30, PPC::F31,
-
- PPC::CR2, PPC::CR3, PPC::CR4,
- PPC::V20, PPC::V21, PPC::V22, PPC::V23,
- PPC::V24, PPC::V25, PPC::V26, PPC::V27,
- PPC::V28, PPC::V29, PPC::V30, PPC::V31,
-
- PPC::CR2LT, PPC::CR2GT, PPC::CR2EQ, PPC::CR2UN,
- PPC::CR3LT, PPC::CR3GT, PPC::CR3EQ, PPC::CR3UN,
- PPC::CR4LT, PPC::CR4GT, PPC::CR4EQ, PPC::CR4UN,
-
- PPC::LR, 0
- };
-
- // 32-bit SVR4 calling convention.
- static const unsigned SVR4_CalleeSavedRegs[] = {
- PPC::R14, PPC::R15,
- PPC::R16, PPC::R17, PPC::R18, PPC::R19,
- PPC::R20, PPC::R21, PPC::R22, PPC::R23,
- PPC::R24, PPC::R25, PPC::R26, PPC::R27,
- PPC::R28, PPC::R29, PPC::R30, PPC::R31,
-
- PPC::F14, PPC::F15, PPC::F16, PPC::F17,
- PPC::F18, PPC::F19, PPC::F20, PPC::F21,
- PPC::F22, PPC::F23, PPC::F24, PPC::F25,
- PPC::F26, PPC::F27, PPC::F28, PPC::F29,
- PPC::F30, PPC::F31,
-
- PPC::CR2, PPC::CR3, PPC::CR4,
-
- PPC::VRSAVE,
-
- PPC::V20, PPC::V21, PPC::V22, PPC::V23,
- PPC::V24, PPC::V25, PPC::V26, PPC::V27,
- PPC::V28, PPC::V29, PPC::V30, PPC::V31,
-
- PPC::CR2LT, PPC::CR2GT, PPC::CR2EQ, PPC::CR2UN,
- PPC::CR3LT, PPC::CR3GT, PPC::CR3EQ, PPC::CR3UN,
- PPC::CR4LT, PPC::CR4GT, PPC::CR4EQ, PPC::CR4UN,
-
- 0
- };
- // 64-bit Darwin calling convention.
- static const unsigned Darwin64_CalleeSavedRegs[] = {
- PPC::X14, PPC::X15,
- PPC::X16, PPC::X17, PPC::X18, PPC::X19,
- PPC::X20, PPC::X21, PPC::X22, PPC::X23,
- PPC::X24, PPC::X25, PPC::X26, PPC::X27,
- PPC::X28, PPC::X29, PPC::X30, PPC::X31,
-
- PPC::F14, PPC::F15, PPC::F16, PPC::F17,
- PPC::F18, PPC::F19, PPC::F20, PPC::F21,
- PPC::F22, PPC::F23, PPC::F24, PPC::F25,
- PPC::F26, PPC::F27, PPC::F28, PPC::F29,
- PPC::F30, PPC::F31,
-
- PPC::CR2, PPC::CR3, PPC::CR4,
- PPC::V20, PPC::V21, PPC::V22, PPC::V23,
- PPC::V24, PPC::V25, PPC::V26, PPC::V27,
- PPC::V28, PPC::V29, PPC::V30, PPC::V31,
-
- PPC::CR2LT, PPC::CR2GT, PPC::CR2EQ, PPC::CR2UN,
- PPC::CR3LT, PPC::CR3GT, PPC::CR3EQ, PPC::CR3UN,
- PPC::CR4LT, PPC::CR4GT, PPC::CR4EQ, PPC::CR4UN,
-
- PPC::LR8, 0
- };
-
- // 64-bit SVR4 calling convention.
- static const unsigned SVR4_64_CalleeSavedRegs[] = {
- PPC::X14, PPC::X15,
- PPC::X16, PPC::X17, PPC::X18, PPC::X19,
- PPC::X20, PPC::X21, PPC::X22, PPC::X23,
- PPC::X24, PPC::X25, PPC::X26, PPC::X27,
- PPC::X28, PPC::X29, PPC::X30, PPC::X31,
-
- PPC::F14, PPC::F15, PPC::F16, PPC::F17,
- PPC::F18, PPC::F19, PPC::F20, PPC::F21,
- PPC::F22, PPC::F23, PPC::F24, PPC::F25,
- PPC::F26, PPC::F27, PPC::F28, PPC::F29,
- PPC::F30, PPC::F31,
-
- PPC::CR2, PPC::CR3, PPC::CR4,
-
- PPC::VRSAVE,
-
- PPC::V20, PPC::V21, PPC::V22, PPC::V23,
- PPC::V24, PPC::V25, PPC::V26, PPC::V27,
- PPC::V28, PPC::V29, PPC::V30, PPC::V31,
-
- PPC::CR2LT, PPC::CR2GT, PPC::CR2EQ, PPC::CR2UN,
- PPC::CR3LT, PPC::CR3GT, PPC::CR3EQ, PPC::CR3UN,
- PPC::CR4LT, PPC::CR4GT, PPC::CR4EQ, PPC::CR4UN,
-
- 0
- };
-
if (Subtarget.isDarwinABI())
- return Subtarget.isPPC64() ? Darwin64_CalleeSavedRegs :
- Darwin32_CalleeSavedRegs;
-
- return Subtarget.isPPC64() ? SVR4_64_CalleeSavedRegs : SVR4_CalleeSavedRegs;
+ return Subtarget.isPPC64() ? (Subtarget.hasAltivec() ?
+ CSR_Darwin64_Altivec_SaveList :
+ CSR_Darwin64_SaveList) :
+ (Subtarget.hasAltivec() ?
+ CSR_Darwin32_Altivec_SaveList :
+ CSR_Darwin32_SaveList);
+
+ return Subtarget.isPPC64() ? (Subtarget.hasAltivec() ?
+ CSR_SVR464_Altivec_SaveList :
+ CSR_SVR464_SaveList) :
+ (Subtarget.hasAltivec() ?
+ CSR_SVR432_Altivec_SaveList :
+ CSR_SVR432_SaveList);
}
-// needsFP - 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.
-//
-static bool needsFP(const MachineFunction &MF) {
- const MachineFrameInfo *MFI = MF.getFrameInfo();
- // Naked functions have no stack frame pushed, so we don't have a frame pointer.
- if (MF.getFunction()->hasFnAttr(Attribute::Naked))
- return false;
- return DisableFramePointerElim(MF) || MFI->hasVarSizedObjects() ||
- (GuaranteedTailCallOpt && MF.getInfo<PPCFunctionInfo>()->hasFastCall());
+const uint32_t*
+PPCRegisterInfo::getCallPreservedMask(CallingConv::ID CC) const {
+ if (Subtarget.isDarwinABI())
+ return Subtarget.isPPC64() ? (Subtarget.hasAltivec() ?
+ CSR_Darwin64_Altivec_RegMask :
+ CSR_Darwin64_RegMask) :
+ (Subtarget.hasAltivec() ?
+ CSR_Darwin32_Altivec_RegMask :
+ CSR_Darwin32_RegMask);
+
+ return Subtarget.isPPC64() ? (Subtarget.hasAltivec() ?
+ CSR_SVR464_Altivec_RegMask :
+ CSR_SVR464_RegMask) :
+ (Subtarget.hasAltivec() ?
+ CSR_SVR432_Altivec_RegMask :
+ CSR_SVR432_RegMask);
}
-static bool spillsCR(const MachineFunction &MF) {
- const PPCFunctionInfo *FuncInfo = MF.getInfo<PPCFunctionInfo>();
- return FuncInfo->isCRSpilled();
+const uint32_t*
+PPCRegisterInfo::getNoPreservedMask() const {
+ return CSR_NoRegs_RegMask;
}
BitVector PPCRegisterInfo::getReservedRegs(const MachineFunction &MF) const {
BitVector Reserved(getNumRegs());
- Reserved.set(PPC::R0);
+ const PPCFrameLowering *PPCFI =
+ static_cast<const PPCFrameLowering*>(MF.getTarget().getFrameLowering());
+
+ // The ZERO register is not really a register, but the representation of r0
+ // when used in instructions that treat r0 as the constant 0.
+ Reserved.set(PPC::ZERO);
+ Reserved.set(PPC::ZERO8);
+
+ // The FP register is also not really a register, but is the representation
+ // of the frame pointer register used by ISD::FRAMEADDR.
+ Reserved.set(PPC::FP);
+ Reserved.set(PPC::FP8);
+
+ // The BP register is also not really a register, but is the representation
+ // of the base pointer register used by setjmp.
+ Reserved.set(PPC::BP);
+ Reserved.set(PPC::BP8);
+
+ // The counter registers must be reserved so that counter-based loops can
+ // be correctly formed (and the mtctr instructions are not DCE'd).
+ Reserved.set(PPC::CTR);
+ Reserved.set(PPC::CTR8);
+
Reserved.set(PPC::R1);
Reserved.set(PPC::LR);
Reserved.set(PPC::LR8);
Reserved.set(PPC::RM);
+ if (!Subtarget.isDarwinABI() || !Subtarget.hasAltivec())
+ Reserved.set(PPC::VRSAVE);
+
// The SVR4 ABI reserves r2 and r13
if (Subtarget.isSVR4ABI()) {
Reserved.set(PPC::R2); // System-reserved register
Reserved.set(PPC::R13); // Small Data Area pointer register
}
- // Reserve R2 on Darwin to hack around the problem of save/restore of CR
- // when the stack frame is too big to address directly; we need two regs.
- // This is a hack.
- if (Subtarget.isDarwinABI()) {
- Reserved.set(PPC::R2);
- }
// On PPC64, r13 is the thread pointer. Never allocate this register.
- // Note that this is over conservative, as it also prevents allocation of R31
- // when the FP is not needed.
if (Subtarget.isPPC64()) {
Reserved.set(PPC::R13);
- Reserved.set(PPC::R31);
-
- if (!EnableRegisterScavenging)
- Reserved.set(PPC::R0); // FIXME (64-bit): Remove
- Reserved.set(PPC::X0);
Reserved.set(PPC::X1);
Reserved.set(PPC::X13);
- Reserved.set(PPC::X31);
+
+ if (PPCFI->needsFP(MF))
+ Reserved.set(PPC::X31);
+
+ if (hasBasePointer(MF))
+ Reserved.set(PPC::X30);
// The 64-bit SVR4 ABI reserves r2 for the TOC pointer.
if (Subtarget.isSVR4ABI()) {
Reserved.set(PPC::X2);
}
- // Reserve R2 on Darwin to hack around the problem of save/restore of CR
- // when the stack frame is too big to address directly; we need two regs.
- // This is a hack.
- if (Subtarget.isDarwinABI()) {
- Reserved.set(PPC::X2);
- }
}
- if (needsFP(MF))
+ if (PPCFI->needsFP(MF))
Reserved.set(PPC::R31);
- return Reserved;
-}
-
-//===----------------------------------------------------------------------===//
-// Stack Frame Processing methods
-//===----------------------------------------------------------------------===//
+ if (hasBasePointer(MF))
+ Reserved.set(PPC::R30);
-// hasFP - Return true if the specified function actually has a dedicated frame
-// pointer register. This is true if the function needs a frame pointer and has
-// a non-zero stack size.
-bool PPCRegisterInfo::hasFP(const MachineFunction &MF) const {
- const MachineFrameInfo *MFI = MF.getFrameInfo();
- return MFI->getStackSize() && needsFP(MF);
-}
+ // Reserve Altivec registers when Altivec is unavailable.
+ if (!Subtarget.hasAltivec())
+ for (TargetRegisterClass::iterator I = PPC::VRRCRegClass.begin(),
+ IE = PPC::VRRCRegClass.end(); I != IE; ++I)
+ Reserved.set(*I);
-/// MustSaveLR - Return true if this function requires that we save the LR
-/// register onto the stack in the prolog and restore it in the epilog of the
-/// function.
-static bool MustSaveLR(const MachineFunction &MF, unsigned LR) {
- const PPCFunctionInfo *MFI = MF.getInfo<PPCFunctionInfo>();
-
- // We need a save/restore of LR if there is any def of LR (which is
- // defined by calls, including the PIC setup sequence), or if there is
- // some use of the LR stack slot (e.g. for builtin_return_address).
- // (LR comes in 32 and 64 bit versions.)
- MachineRegisterInfo::def_iterator RI = MF.getRegInfo().def_begin(LR);
- return RI !=MF.getRegInfo().def_end() || MFI->isLRStoreRequired();
+ return Reserved;
}
+unsigned
+PPCRegisterInfo::getRegPressureLimit(const TargetRegisterClass *RC,
+ MachineFunction &MF) const {
+ const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering();
+ const unsigned DefaultSafety = 1;
-
-void PPCRegisterInfo::
-eliminateCallFramePseudoInstr(MachineFunction &MF, MachineBasicBlock &MBB,
- MachineBasicBlock::iterator I) const {
- if (GuaranteedTailCallOpt && I->getOpcode() == PPC::ADJCALLSTACKUP) {
- // Add (actually subtract) back the amount the callee popped on return.
- if (int CalleeAmt = I->getOperand(1).getImm()) {
- bool is64Bit = Subtarget.isPPC64();
- CalleeAmt *= -1;
- unsigned StackReg = is64Bit ? PPC::X1 : PPC::R1;
- unsigned TmpReg = is64Bit ? PPC::X0 : PPC::R0;
- unsigned ADDIInstr = is64Bit ? PPC::ADDI8 : PPC::ADDI;
- unsigned ADDInstr = is64Bit ? PPC::ADD8 : PPC::ADD4;
- unsigned LISInstr = is64Bit ? PPC::LIS8 : PPC::LIS;
- unsigned ORIInstr = is64Bit ? PPC::ORI8 : PPC::ORI;
- MachineInstr *MI = I;
- DebugLoc dl = MI->getDebugLoc();
-
- if (isInt<16>(CalleeAmt)) {
- BuildMI(MBB, I, dl, TII.get(ADDIInstr), StackReg).addReg(StackReg).
- addImm(CalleeAmt);
- } else {
- MachineBasicBlock::iterator MBBI = I;
- BuildMI(MBB, MBBI, dl, TII.get(LISInstr), TmpReg)
- .addImm(CalleeAmt >> 16);
- BuildMI(MBB, MBBI, dl, TII.get(ORIInstr), TmpReg)
- .addReg(TmpReg, RegState::Kill)
- .addImm(CalleeAmt & 0xFFFF);
- BuildMI(MBB, MBBI, dl, TII.get(ADDInstr))
- .addReg(StackReg)
- .addReg(StackReg)
- .addReg(TmpReg);
- }
- }
+ switch (RC->getID()) {
+ default:
+ return 0;
+ case PPC::G8RC_NOX0RegClassID:
+ case PPC::GPRC_NOR0RegClassID:
+ case PPC::G8RCRegClassID:
+ case PPC::GPRCRegClassID: {
+ unsigned FP = TFI->hasFP(MF) ? 1 : 0;
+ return 32 - FP - DefaultSafety;
+ }
+ case PPC::F8RCRegClassID:
+ case PPC::F4RCRegClassID:
+ case PPC::VRRCRegClassID:
+ case PPC::VSLRCRegClassID:
+ case PPC::VSHRCRegClassID:
+ return 32 - DefaultSafety;
+ case PPC::VSRCRegClassID:
+ return 64 - DefaultSafety;
+ case PPC::CRRCRegClassID:
+ return 8 - DefaultSafety;
}
- // Simply discard ADJCALLSTACKDOWN, ADJCALLSTACKUP instructions.
- MBB.erase(I);
}
-/// findScratchRegister - Find a 'free' PPC register. Try for a call-clobbered
-/// register first and then a spilled callee-saved register if that fails.
-static
-unsigned findScratchRegister(MachineBasicBlock::iterator II, RegScavenger *RS,
- const TargetRegisterClass *RC, int SPAdj) {
- assert(RS && "Register scavenging must be on");
- unsigned Reg = RS->FindUnusedReg(RC);
- // FIXME: move ARM callee-saved reg scan to target independent code, then
- // search for already spilled CS register here.
- if (Reg == 0)
- Reg = RS->scavengeRegister(RC, II, SPAdj);
- return Reg;
-}
+//===----------------------------------------------------------------------===//
+// Stack Frame Processing methods
+//===----------------------------------------------------------------------===//
/// lowerDynamicAlloc - Generate the code for allocating an object in the
/// current frame. The sequence of code with be in the general form
/// stwxu R0, SP, Rnegsize ; add and update the SP with the negated size
/// addi Rnew, SP, \#maxCalFrameSize ; get the top of the allocation
///
-void PPCRegisterInfo::lowerDynamicAlloc(MachineBasicBlock::iterator II,
- int SPAdj, RegScavenger *RS) const {
+void PPCRegisterInfo::lowerDynamicAlloc(MachineBasicBlock::iterator II) const {
// Get the instruction.
MachineInstr &MI = *II;
// Get the instruction's basic block.
MachineFunction &MF = *MBB.getParent();
// Get the frame info.
MachineFrameInfo *MFI = MF.getFrameInfo();
+ // Get the instruction info.
+ const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();
// Determine whether 64-bit pointers are used.
bool LP64 = Subtarget.isPPC64();
DebugLoc dl = MI.getDebugLoc();
unsigned FrameSize = MFI->getStackSize();
// Get stack alignments.
- unsigned TargetAlign = MF.getTarget().getFrameInfo()->getStackAlignment();
+ unsigned TargetAlign = MF.getTarget().getFrameLowering()->getStackAlignment();
unsigned MaxAlign = MFI->getMaxAlignment();
- assert(MaxAlign <= TargetAlign &&
- "Dynamic alloca with large aligns not supported");
+ assert((maxCallFrameSize & (MaxAlign-1)) == 0 &&
+ "Maximum call-frame size not sufficiently aligned");
// Determine the previous frame's address. If FrameSize can't be
// represented as 16 bits or we need special alignment, then we load the
// Fortunately, a frame greater than 32K is rare.
const TargetRegisterClass *G8RC = &PPC::G8RCRegClass;
const TargetRegisterClass *GPRC = &PPC::GPRCRegClass;
- const TargetRegisterClass *RC = LP64 ? G8RC : GPRC;
-
- // FIXME (64-bit): Use "findScratchRegister"
- unsigned Reg;
- if (EnableRegisterScavenging)
- Reg = findScratchRegister(II, RS, RC, SPAdj);
- else
- Reg = PPC::R0;
+ unsigned Reg = MF.getRegInfo().createVirtualRegister(LP64 ? G8RC : GPRC);
if (MaxAlign < TargetAlign && isInt<16>(FrameSize)) {
BuildMI(MBB, II, dl, TII.get(PPC::ADDI), Reg)
.addReg(PPC::R31)
.addImm(FrameSize);
} else if (LP64) {
- if (EnableRegisterScavenging) // FIXME (64-bit): Use "true" part.
- BuildMI(MBB, II, dl, TII.get(PPC::LD), Reg)
- .addImm(0)
- .addReg(PPC::X1);
- else
- BuildMI(MBB, II, dl, TII.get(PPC::LD), PPC::X0)
- .addImm(0)
- .addReg(PPC::X1);
+ BuildMI(MBB, II, dl, TII.get(PPC::LD), Reg)
+ .addImm(0)
+ .addReg(PPC::X1);
} else {
BuildMI(MBB, II, dl, TII.get(PPC::LWZ), Reg)
.addImm(0)
.addReg(PPC::R1);
}
-
+
+ bool KillNegSizeReg = MI.getOperand(1).isKill();
+ unsigned NegSizeReg = MI.getOperand(1).getReg();
+
// Grow the stack and update the stack pointer link, then determine the
// address of new allocated space.
if (LP64) {
- if (EnableRegisterScavenging) // FIXME (64-bit): Use "true" part.
- BuildMI(MBB, II, dl, TII.get(PPC::STDUX))
- .addReg(Reg, RegState::Kill)
- .addReg(PPC::X1)
- .addReg(MI.getOperand(1).getReg());
- else
- BuildMI(MBB, II, dl, TII.get(PPC::STDUX))
- .addReg(PPC::X0, RegState::Kill)
- .addReg(PPC::X1)
- .addReg(MI.getOperand(1).getReg());
-
- if (!MI.getOperand(1).isKill())
- BuildMI(MBB, II, dl, TII.get(PPC::ADDI8), MI.getOperand(0).getReg())
- .addReg(PPC::X1)
- .addImm(maxCallFrameSize);
- else
- // Implicitly kill the register.
- BuildMI(MBB, II, dl, TII.get(PPC::ADDI8), MI.getOperand(0).getReg())
- .addReg(PPC::X1)
- .addImm(maxCallFrameSize)
- .addReg(MI.getOperand(1).getReg(), RegState::ImplicitKill);
+ if (MaxAlign > TargetAlign) {
+ unsigned UnalNegSizeReg = NegSizeReg;
+ NegSizeReg = MF.getRegInfo().createVirtualRegister(G8RC);
+
+ // Unfortunately, there is no andi, only andi., and we can't insert that
+ // here because we might clobber cr0 while it is live.
+ BuildMI(MBB, II, dl, TII.get(PPC::LI8), NegSizeReg)
+ .addImm(~(MaxAlign-1));
+
+ unsigned NegSizeReg1 = NegSizeReg;
+ NegSizeReg = MF.getRegInfo().createVirtualRegister(G8RC);
+ BuildMI(MBB, II, dl, TII.get(PPC::AND8), NegSizeReg)
+ .addReg(UnalNegSizeReg, getKillRegState(KillNegSizeReg))
+ .addReg(NegSizeReg1, RegState::Kill);
+ KillNegSizeReg = true;
+ }
+
+ BuildMI(MBB, II, dl, TII.get(PPC::STDUX), PPC::X1)
+ .addReg(Reg, RegState::Kill)
+ .addReg(PPC::X1)
+ .addReg(NegSizeReg, getKillRegState(KillNegSizeReg));
+ BuildMI(MBB, II, dl, TII.get(PPC::ADDI8), MI.getOperand(0).getReg())
+ .addReg(PPC::X1)
+ .addImm(maxCallFrameSize);
} else {
- BuildMI(MBB, II, dl, TII.get(PPC::STWUX))
+ if (MaxAlign > TargetAlign) {
+ unsigned UnalNegSizeReg = NegSizeReg;
+ NegSizeReg = MF.getRegInfo().createVirtualRegister(GPRC);
+
+ // Unfortunately, there is no andi, only andi., and we can't insert that
+ // here because we might clobber cr0 while it is live.
+ BuildMI(MBB, II, dl, TII.get(PPC::LI), NegSizeReg)
+ .addImm(~(MaxAlign-1));
+
+ unsigned NegSizeReg1 = NegSizeReg;
+ NegSizeReg = MF.getRegInfo().createVirtualRegister(GPRC);
+ BuildMI(MBB, II, dl, TII.get(PPC::AND), NegSizeReg)
+ .addReg(UnalNegSizeReg, getKillRegState(KillNegSizeReg))
+ .addReg(NegSizeReg1, RegState::Kill);
+ KillNegSizeReg = true;
+ }
+
+ BuildMI(MBB, II, dl, TII.get(PPC::STWUX), PPC::R1)
.addReg(Reg, RegState::Kill)
.addReg(PPC::R1)
- .addReg(MI.getOperand(1).getReg());
-
- if (!MI.getOperand(1).isKill())
- BuildMI(MBB, II, dl, TII.get(PPC::ADDI), MI.getOperand(0).getReg())
- .addReg(PPC::R1)
- .addImm(maxCallFrameSize);
- else
- // Implicitly kill the register.
- BuildMI(MBB, II, dl, TII.get(PPC::ADDI), MI.getOperand(0).getReg())
- .addReg(PPC::R1)
- .addImm(maxCallFrameSize)
- .addReg(MI.getOperand(1).getReg(), RegState::ImplicitKill);
+ .addReg(NegSizeReg, getKillRegState(KillNegSizeReg));
+ BuildMI(MBB, II, dl, TII.get(PPC::ADDI), MI.getOperand(0).getReg())
+ .addReg(PPC::R1)
+ .addImm(maxCallFrameSize);
}
// Discard the DYNALLOC instruction.
/// stw rA, FI ; Store rA to the frame.
///
void PPCRegisterInfo::lowerCRSpilling(MachineBasicBlock::iterator II,
- unsigned FrameIndex, int SPAdj,
- RegScavenger *RS) const {
+ unsigned FrameIndex) const {
// Get the instruction.
- MachineInstr &MI = *II; // ; SPILL_CR <SrcReg>, <offset>, <FI>
+ MachineInstr &MI = *II; // ; SPILL_CR <SrcReg>, <offset>
// Get the instruction's basic block.
MachineBasicBlock &MBB = *MI.getParent();
+ MachineFunction &MF = *MBB.getParent();
+ const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();
DebugLoc dl = MI.getDebugLoc();
+ bool LP64 = Subtarget.isPPC64();
const TargetRegisterClass *G8RC = &PPC::G8RCRegClass;
const TargetRegisterClass *GPRC = &PPC::GPRCRegClass;
- const TargetRegisterClass *RC = Subtarget.isPPC64() ? G8RC : GPRC;
- unsigned Reg = findScratchRegister(II, RS, RC, SPAdj);
+
+ unsigned Reg = MF.getRegInfo().createVirtualRegister(LP64 ? G8RC : GPRC);
unsigned SrcReg = MI.getOperand(0).getReg();
// We need to store the CR in the low 4-bits of the saved value. First, issue
- // an MFCRpsued to save all of the CRBits and, if needed, kill the SrcReg.
- BuildMI(MBB, II, dl, TII.get(PPC::MFCRpseud), Reg)
+ // an MFOCRF to save all of the CRBits and, if needed, kill the SrcReg.
+ BuildMI(MBB, II, dl, TII.get(LP64 ? PPC::MFOCRF8 : PPC::MFOCRF), Reg)
.addReg(SrcReg, getKillRegState(MI.getOperand(0).isKill()));
// If the saved register wasn't CR0, shift the bits left so that they are in
// CR0's slot.
- if (SrcReg != PPC::CR0)
+ if (SrcReg != PPC::CR0) {
+ unsigned Reg1 = Reg;
+ Reg = MF.getRegInfo().createVirtualRegister(LP64 ? G8RC : GPRC);
+
// rlwinm rA, rA, ShiftBits, 0, 31.
- BuildMI(MBB, II, dl, TII.get(PPC::RLWINM), Reg)
- .addReg(Reg, RegState::Kill)
- .addImm(PPCRegisterInfo::getRegisterNumbering(SrcReg) * 4)
+ BuildMI(MBB, II, dl, TII.get(LP64 ? PPC::RLWINM8 : PPC::RLWINM), Reg)
+ .addReg(Reg1, RegState::Kill)
+ .addImm(getEncodingValue(SrcReg) * 4)
.addImm(0)
.addImm(31);
+ }
+
+ addFrameReference(BuildMI(MBB, II, dl, TII.get(LP64 ? PPC::STW8 : PPC::STW))
+ .addReg(Reg, RegState::Kill),
+ FrameIndex);
+ // Discard the pseudo instruction.
+ MBB.erase(II);
+}
+
+void PPCRegisterInfo::lowerCRRestore(MachineBasicBlock::iterator II,
+ unsigned FrameIndex) const {
+ // Get the instruction.
+ MachineInstr &MI = *II; // ; <DestReg> = RESTORE_CR <offset>
+ // Get the instruction's basic block.
+ MachineBasicBlock &MBB = *MI.getParent();
+ MachineFunction &MF = *MBB.getParent();
+ const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();
+ DebugLoc dl = MI.getDebugLoc();
+
+ bool LP64 = Subtarget.isPPC64();
+ const TargetRegisterClass *G8RC = &PPC::G8RCRegClass;
+ const TargetRegisterClass *GPRC = &PPC::GPRCRegClass;
+
+ unsigned Reg = MF.getRegInfo().createVirtualRegister(LP64 ? G8RC : GPRC);
+ unsigned DestReg = MI.getOperand(0).getReg();
+ assert(MI.definesRegister(DestReg) &&
+ "RESTORE_CR does not define its destination");
+
+ addFrameReference(BuildMI(MBB, II, dl, TII.get(LP64 ? PPC::LWZ8 : PPC::LWZ),
+ Reg), FrameIndex);
+
+ // If the reloaded register isn't CR0, shift the bits right so that they are
+ // in the right CR's slot.
+ if (DestReg != PPC::CR0) {
+ unsigned Reg1 = Reg;
+ Reg = MF.getRegInfo().createVirtualRegister(LP64 ? G8RC : GPRC);
+
+ unsigned ShiftBits = getEncodingValue(DestReg)*4;
+ // rlwinm r11, r11, 32-ShiftBits, 0, 31.
+ BuildMI(MBB, II, dl, TII.get(LP64 ? PPC::RLWINM8 : PPC::RLWINM), Reg)
+ .addReg(Reg1, RegState::Kill).addImm(32-ShiftBits).addImm(0)
+ .addImm(31);
+ }
+
+ BuildMI(MBB, II, dl, TII.get(LP64 ? PPC::MTOCRF8 : PPC::MTOCRF), DestReg)
+ .addReg(Reg, RegState::Kill);
+
+ // Discard the pseudo instruction.
+ MBB.erase(II);
+}
+
+static unsigned getCRFromCRBit(unsigned SrcReg) {
+ unsigned Reg = 0;
+ if (SrcReg == PPC::CR0LT || SrcReg == PPC::CR0GT ||
+ SrcReg == PPC::CR0EQ || SrcReg == PPC::CR0UN)
+ Reg = PPC::CR0;
+ else if (SrcReg == PPC::CR1LT || SrcReg == PPC::CR1GT ||
+ SrcReg == PPC::CR1EQ || SrcReg == PPC::CR1UN)
+ Reg = PPC::CR1;
+ else if (SrcReg == PPC::CR2LT || SrcReg == PPC::CR2GT ||
+ SrcReg == PPC::CR2EQ || SrcReg == PPC::CR2UN)
+ Reg = PPC::CR2;
+ else if (SrcReg == PPC::CR3LT || SrcReg == PPC::CR3GT ||
+ SrcReg == PPC::CR3EQ || SrcReg == PPC::CR3UN)
+ Reg = PPC::CR3;
+ else if (SrcReg == PPC::CR4LT || SrcReg == PPC::CR4GT ||
+ SrcReg == PPC::CR4EQ || SrcReg == PPC::CR4UN)
+ Reg = PPC::CR4;
+ else if (SrcReg == PPC::CR5LT || SrcReg == PPC::CR5GT ||
+ SrcReg == PPC::CR5EQ || SrcReg == PPC::CR5UN)
+ Reg = PPC::CR5;
+ else if (SrcReg == PPC::CR6LT || SrcReg == PPC::CR6GT ||
+ SrcReg == PPC::CR6EQ || SrcReg == PPC::CR6UN)
+ Reg = PPC::CR6;
+ else if (SrcReg == PPC::CR7LT || SrcReg == PPC::CR7GT ||
+ SrcReg == PPC::CR7EQ || SrcReg == PPC::CR7UN)
+ Reg = PPC::CR7;
+
+ assert(Reg != 0 && "Invalid CR bit register");
+ return Reg;
+}
+
+void PPCRegisterInfo::lowerCRBitSpilling(MachineBasicBlock::iterator II,
+ unsigned FrameIndex) const {
+ // Get the instruction.
+ MachineInstr &MI = *II; // ; SPILL_CRBIT <SrcReg>, <offset>
+ // Get the instruction's basic block.
+ MachineBasicBlock &MBB = *MI.getParent();
+ MachineFunction &MF = *MBB.getParent();
+ const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();
+ DebugLoc dl = MI.getDebugLoc();
+
+ bool LP64 = Subtarget.isPPC64();
+ const TargetRegisterClass *G8RC = &PPC::G8RCRegClass;
+ const TargetRegisterClass *GPRC = &PPC::GPRCRegClass;
+
+ unsigned Reg = MF.getRegInfo().createVirtualRegister(LP64 ? G8RC : GPRC);
+ unsigned SrcReg = MI.getOperand(0).getReg();
+
+ BuildMI(MBB, II, dl, TII.get(TargetOpcode::KILL),
+ getCRFromCRBit(SrcReg))
+ .addReg(SrcReg, getKillRegState(MI.getOperand(0).isKill()));
+
+ BuildMI(MBB, II, dl, TII.get(LP64 ? PPC::MFOCRF8 : PPC::MFOCRF), Reg)
+ .addReg(getCRFromCRBit(SrcReg));
+
+ // If the saved register wasn't CR0LT, shift the bits left so that the bit to
+ // store is the first one. Mask all but that bit.
+ unsigned Reg1 = Reg;
+ Reg = MF.getRegInfo().createVirtualRegister(LP64 ? G8RC : GPRC);
+
+ // rlwinm rA, rA, ShiftBits, 0, 0.
+ BuildMI(MBB, II, dl, TII.get(LP64 ? PPC::RLWINM8 : PPC::RLWINM), Reg)
+ .addReg(Reg1, RegState::Kill)
+ .addImm(getEncodingValue(SrcReg))
+ .addImm(0).addImm(0);
+
+ addFrameReference(BuildMI(MBB, II, dl, TII.get(LP64 ? PPC::STW8 : PPC::STW))
+ .addReg(Reg, RegState::Kill),
+ FrameIndex);
+
+ // Discard the pseudo instruction.
+ MBB.erase(II);
+}
+
+void PPCRegisterInfo::lowerCRBitRestore(MachineBasicBlock::iterator II,
+ unsigned FrameIndex) const {
+ // Get the instruction.
+ MachineInstr &MI = *II; // ; <DestReg> = RESTORE_CRBIT <offset>
+ // Get the instruction's basic block.
+ MachineBasicBlock &MBB = *MI.getParent();
+ MachineFunction &MF = *MBB.getParent();
+ const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();
+ DebugLoc dl = MI.getDebugLoc();
+
+ bool LP64 = Subtarget.isPPC64();
+ const TargetRegisterClass *G8RC = &PPC::G8RCRegClass;
+ const TargetRegisterClass *GPRC = &PPC::GPRCRegClass;
+
+ unsigned Reg = MF.getRegInfo().createVirtualRegister(LP64 ? G8RC : GPRC);
+ unsigned DestReg = MI.getOperand(0).getReg();
+ assert(MI.definesRegister(DestReg) &&
+ "RESTORE_CRBIT does not define its destination");
+
+ addFrameReference(BuildMI(MBB, II, dl, TII.get(LP64 ? PPC::LWZ8 : PPC::LWZ),
+ Reg), FrameIndex);
+
+ BuildMI(MBB, II, dl, TII.get(TargetOpcode::IMPLICIT_DEF), DestReg);
+
+ unsigned RegO = MF.getRegInfo().createVirtualRegister(LP64 ? G8RC : GPRC);
+ BuildMI(MBB, II, dl, TII.get(LP64 ? PPC::MFOCRF8 : PPC::MFOCRF), RegO)
+ .addReg(getCRFromCRBit(DestReg));
+
+ unsigned ShiftBits = getEncodingValue(DestReg);
+ // rlwimi r11, r10, 32-ShiftBits, ..., ...
+ BuildMI(MBB, II, dl, TII.get(LP64 ? PPC::RLWIMI8 : PPC::RLWIMI), RegO)
+ .addReg(RegO, RegState::Kill).addReg(Reg, RegState::Kill)
+ .addImm(ShiftBits ? 32-ShiftBits : 0)
+ .addImm(ShiftBits).addImm(ShiftBits);
+
+ BuildMI(MBB, II, dl, TII.get(LP64 ? PPC::MTOCRF8 : PPC::MTOCRF),
+ getCRFromCRBit(DestReg))
+ .addReg(RegO, RegState::Kill)
+ // Make sure we have a use dependency all the way through this
+ // sequence of instructions. We can't have the other bits in the CR
+ // modified in between the mfocrf and the mtocrf.
+ .addReg(getCRFromCRBit(DestReg), RegState::Implicit);
+
+ // Discard the pseudo instruction.
+ MBB.erase(II);
+}
+
+void PPCRegisterInfo::lowerVRSAVESpilling(MachineBasicBlock::iterator II,
+ unsigned FrameIndex) const {
+ // Get the instruction.
+ MachineInstr &MI = *II; // ; SPILL_VRSAVE <SrcReg>, <offset>
+ // Get the instruction's basic block.
+ MachineBasicBlock &MBB = *MI.getParent();
+ MachineFunction &MF = *MBB.getParent();
+ const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();
+ DebugLoc dl = MI.getDebugLoc();
+
+ const TargetRegisterClass *GPRC = &PPC::GPRCRegClass;
+ unsigned Reg = MF.getRegInfo().createVirtualRegister(GPRC);
+ unsigned SrcReg = MI.getOperand(0).getReg();
+
+ BuildMI(MBB, II, dl, TII.get(PPC::MFVRSAVEv), Reg)
+ .addReg(SrcReg, getKillRegState(MI.getOperand(0).isKill()));
+
addFrameReference(BuildMI(MBB, II, dl, TII.get(PPC::STW))
- .addReg(Reg, getKillRegState(MI.getOperand(1).getImm())),
+ .addReg(Reg, RegState::Kill),
FrameIndex);
// Discard the pseudo instruction.
MBB.erase(II);
}
-unsigned
+void PPCRegisterInfo::lowerVRSAVERestore(MachineBasicBlock::iterator II,
+ unsigned FrameIndex) const {
+ // Get the instruction.
+ MachineInstr &MI = *II; // ; <DestReg> = RESTORE_VRSAVE <offset>
+ // Get the instruction's basic block.
+ MachineBasicBlock &MBB = *MI.getParent();
+ MachineFunction &MF = *MBB.getParent();
+ const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();
+ DebugLoc dl = MI.getDebugLoc();
+
+ const TargetRegisterClass *GPRC = &PPC::GPRCRegClass;
+ unsigned Reg = MF.getRegInfo().createVirtualRegister(GPRC);
+ unsigned DestReg = MI.getOperand(0).getReg();
+ assert(MI.definesRegister(DestReg) &&
+ "RESTORE_VRSAVE does not define its destination");
+
+ addFrameReference(BuildMI(MBB, II, dl, TII.get(PPC::LWZ),
+ Reg), FrameIndex);
+
+ BuildMI(MBB, II, dl, TII.get(PPC::MTVRSAVEv), DestReg)
+ .addReg(Reg, RegState::Kill);
+
+ // Discard the pseudo instruction.
+ MBB.erase(II);
+}
+
+bool
+PPCRegisterInfo::hasReservedSpillSlot(const MachineFunction &MF,
+ unsigned Reg, int &FrameIdx) const {
+
+ // For the nonvolatile condition registers (CR2, CR3, CR4) in an SVR4
+ // ABI, return true to prevent allocating an additional frame slot.
+ // For 64-bit, the CR save area is at SP+8; the value of FrameIdx = 0
+ // is arbitrary and will be subsequently ignored. For 32-bit, we have
+ // previously created the stack slot if needed, so return its FrameIdx.
+ if (Subtarget.isSVR4ABI() && PPC::CR2 <= Reg && Reg <= PPC::CR4) {
+ if (Subtarget.isPPC64())
+ FrameIdx = 0;
+ else {
+ const PPCFunctionInfo *FI = MF.getInfo<PPCFunctionInfo>();
+ FrameIdx = FI->getCRSpillFrameIndex();
+ }
+ return true;
+ }
+ return false;
+}
+
+// Figure out if the offset in the instruction must be a multiple of 4.
+// This is true for instructions like "STD".
+static bool usesIXAddr(const MachineInstr &MI) {
+ unsigned OpC = MI.getOpcode();
+
+ switch (OpC) {
+ default:
+ return false;
+ case PPC::LWA:
+ case PPC::LWA_32:
+ case PPC::LD:
+ case PPC::STD:
+ return true;
+ }
+}
+
+// Return the OffsetOperandNo given the FIOperandNum (and the instruction).
+static unsigned getOffsetONFromFION(const MachineInstr &MI,
+ unsigned FIOperandNum) {
+ // Take into account whether it's an add or mem instruction
+ unsigned OffsetOperandNo = (FIOperandNum == 2) ? 1 : 2;
+ if (MI.isInlineAsm())
+ OffsetOperandNo = FIOperandNum-1;
+
+ return OffsetOperandNo;
+}
+
+void
PPCRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator II,
- int SPAdj, FrameIndexValue *Value,
+ int SPAdj, unsigned FIOperandNum,
RegScavenger *RS) const {
assert(SPAdj == 0 && "Unexpected");
MachineBasicBlock &MBB = *MI.getParent();
// Get the basic block's function.
MachineFunction &MF = *MBB.getParent();
+ // Get the instruction info.
+ const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();
// Get the frame info.
MachineFrameInfo *MFI = MF.getFrameInfo();
DebugLoc dl = MI.getDebugLoc();
- // Find out which operand is the frame index.
- unsigned FIOperandNo = 0;
- while (!MI.getOperand(FIOperandNo).isFI()) {
- ++FIOperandNo;
- assert(FIOperandNo != MI.getNumOperands() &&
- "Instr doesn't have FrameIndex operand!");
- }
- // Take into account whether it's an add or mem instruction
- unsigned OffsetOperandNo = (FIOperandNo == 2) ? 1 : 2;
- if (MI.isInlineAsm())
- OffsetOperandNo = FIOperandNo-1;
+ unsigned OffsetOperandNo = getOffsetONFromFION(MI, FIOperandNum);
// Get the frame index.
- int FrameIndex = MI.getOperand(FIOperandNo).getIndex();
+ int FrameIndex = MI.getOperand(FIOperandNum).getIndex();
// Get the frame pointer save index. Users of this index are primarily
// DYNALLOC instructions.
// Special case for dynamic alloca.
if (FPSI && FrameIndex == FPSI &&
(OpC == PPC::DYNALLOC || OpC == PPC::DYNALLOC8)) {
- lowerDynamicAlloc(II, SPAdj, RS);
- return 0;
+ lowerDynamicAlloc(II);
+ return;
}
- // Special case for pseudo-op SPILL_CR.
- if (EnableRegisterScavenging) // FIXME (64-bit): Enable by default.
- if (OpC == PPC::SPILL_CR) {
- lowerCRSpilling(II, FrameIndex, SPAdj, RS);
- return 0;
- }
+ // Special case for pseudo-ops SPILL_CR and RESTORE_CR, etc.
+ if (OpC == PPC::SPILL_CR) {
+ lowerCRSpilling(II, FrameIndex);
+ return;
+ } else if (OpC == PPC::RESTORE_CR) {
+ lowerCRRestore(II, FrameIndex);
+ return;
+ } else if (OpC == PPC::SPILL_CRBIT) {
+ lowerCRBitSpilling(II, FrameIndex);
+ return;
+ } else if (OpC == PPC::RESTORE_CRBIT) {
+ lowerCRBitRestore(II, FrameIndex);
+ return;
+ } else if (OpC == PPC::SPILL_VRSAVE) {
+ lowerVRSAVESpilling(II, FrameIndex);
+ return;
+ } else if (OpC == PPC::RESTORE_VRSAVE) {
+ lowerVRSAVERestore(II, FrameIndex);
+ return;
+ }
// Replace the FrameIndex with base register with GPR1 (SP) or GPR31 (FP).
- MI.getOperand(FIOperandNo).ChangeToRegister(hasFP(MF) ? PPC::R31 : PPC::R1,
- false);
+ MI.getOperand(FIOperandNum).ChangeToRegister(
+ FrameIndex < 0 ? getBaseRegister(MF) : getFrameRegister(MF), false);
+
+ // Figure out if the offset in the instruction is shifted right two bits.
+ bool isIXAddr = usesIXAddr(MI);
+
+ // If the instruction is not present in ImmToIdxMap, then it has no immediate
+ // form (and must be r+r).
+ bool noImmForm = !MI.isInlineAsm() && !ImmToIdxMap.count(OpC);
- // Figure out if the offset in the instruction is shifted right two bits. This
- // is true for instructions like "STD", which the machine implicitly adds two
- // low zeros to.
- bool isIXAddr = false;
- switch (OpC) {
- case PPC::LWA:
- case PPC::LD:
- case PPC::STD:
- case PPC::STD_32:
- isIXAddr = true;
- break;
- }
-
// Now add the frame object offset to the offset from r1.
int Offset = MFI->getObjectOffset(FrameIndex);
- if (!isIXAddr)
- Offset += MI.getOperand(OffsetOperandNo).getImm();
- else
- Offset += MI.getOperand(OffsetOperandNo).getImm() << 2;
+ Offset += MI.getOperand(OffsetOperandNo).getImm();
// If we're not using a Frame Pointer that has been set to the value of the
// SP before having the stack size subtracted from it, then add the stack size
// to Offset to get the correct offset.
// Naked functions have stack size 0, although getStackSize may not reflect that
// because we didn't call all the pieces that compute it for naked functions.
- if (!MF.getFunction()->hasFnAttr(Attribute::Naked))
- Offset += MFI->getStackSize();
+ if (!MF.getFunction()->getAttributes().
+ hasAttribute(AttributeSet::FunctionIndex, Attribute::Naked)) {
+ if (!(hasBasePointer(MF) && FrameIndex < 0))
+ Offset += MFI->getStackSize();
+ }
// If we can, encode the offset directly into the instruction. If this is a
// normal PPC "ri" instruction, any 16-bit value can be safely encoded. If
// clear can be encoded. This is extremely uncommon, because normally you
// only "std" to a stack slot that is at least 4-byte aligned, but it can
// happen in invalid code.
- if (isInt<16>(Offset) && (!isIXAddr || (Offset & 3) == 0)) {
- if (isIXAddr)
- Offset >>= 2; // The actual encoded value has the low two bits zero.
+ assert(OpC != PPC::DBG_VALUE &&
+ "This should be handle in a target independent way");
+ if (!noImmForm && isInt<16>(Offset) && (!isIXAddr || (Offset & 3) == 0)) {
MI.getOperand(OffsetOperandNo).ChangeToImmediate(Offset);
- return 0;
+ return;
}
// The offset doesn't fit into a single register, scavenge one to build the
// offset in.
- // FIXME: figure out what SPAdj is doing here.
- // FIXME (64-bit): Use "findScratchRegister".
- unsigned SReg;
- if (EnableRegisterScavenging)
- SReg = findScratchRegister(II, RS, &PPC::GPRCRegClass, SPAdj);
- else
- SReg = PPC::R0;
+ bool is64Bit = Subtarget.isPPC64();
+ const TargetRegisterClass *G8RC = &PPC::G8RCRegClass;
+ const TargetRegisterClass *GPRC = &PPC::GPRCRegClass;
+ const TargetRegisterClass *RC = is64Bit ? G8RC : GPRC;
+ unsigned SRegHi = MF.getRegInfo().createVirtualRegister(RC),
+ SReg = MF.getRegInfo().createVirtualRegister(RC);
// Insert a set of rA with the full offset value before the ld, st, or add
- BuildMI(MBB, II, dl, TII.get(PPC::LIS), SReg)
+ BuildMI(MBB, II, dl, TII.get(is64Bit ? PPC::LIS8 : PPC::LIS), SRegHi)
.addImm(Offset >> 16);
- BuildMI(MBB, II, dl, TII.get(PPC::ORI), SReg)
- .addReg(SReg, RegState::Kill)
+ BuildMI(MBB, II, dl, TII.get(is64Bit ? PPC::ORI8 : PPC::ORI), SReg)
+ .addReg(SRegHi, RegState::Kill)
.addImm(Offset);
// Convert into indexed form of the instruction:
// addi 0:rA 1:rB, 2, imm ==> add 0:rA, 1:rB, 2:r0
unsigned OperandBase;
- if (OpC != TargetOpcode::INLINEASM) {
+ if (noImmForm)
+ OperandBase = 1;
+ else if (OpC != TargetOpcode::INLINEASM) {
assert(ImmToIdxMap.count(OpC) &&
"No indexed form of load or store available!");
unsigned NewOpcode = ImmToIdxMap.find(OpC)->second;
} else {
OperandBase = OffsetOperandNo;
}
-
- unsigned StackReg = MI.getOperand(FIOperandNo).getReg();
+
+ unsigned StackReg = MI.getOperand(FIOperandNum).getReg();
MI.getOperand(OperandBase).ChangeToRegister(StackReg, false);
- MI.getOperand(OperandBase + 1).ChangeToRegister(SReg, false);
- return 0;
+ MI.getOperand(OperandBase + 1).ChangeToRegister(SReg, false, false, true);
}
-/// VRRegNo - Map from a numbered VR register to its enum value.
-///
-static const unsigned short VRRegNo[] = {
- PPC::V0 , PPC::V1 , PPC::V2 , PPC::V3 , PPC::V4 , PPC::V5 , PPC::V6 , PPC::V7 ,
- PPC::V8 , PPC::V9 , PPC::V10, PPC::V11, PPC::V12, PPC::V13, PPC::V14, PPC::V15,
- PPC::V16, PPC::V17, PPC::V18, PPC::V19, PPC::V20, PPC::V21, PPC::V22, PPC::V23,
- PPC::V24, PPC::V25, PPC::V26, PPC::V27, PPC::V28, PPC::V29, PPC::V30, PPC::V31
-};
-
-/// RemoveVRSaveCode - We have found that this function does not need any code
-/// to manipulate the VRSAVE register, even though it uses vector registers.
-/// This can happen when the only registers used are known to be live in or out
-/// of the function. Remove all of the VRSAVE related code from the function.
-static void RemoveVRSaveCode(MachineInstr *MI) {
- MachineBasicBlock *Entry = MI->getParent();
- MachineFunction *MF = Entry->getParent();
-
- // We know that the MTVRSAVE instruction immediately follows MI. Remove it.
- MachineBasicBlock::iterator MBBI = MI;
- ++MBBI;
- assert(MBBI != Entry->end() && MBBI->getOpcode() == PPC::MTVRSAVE);
- MBBI->eraseFromParent();
-
- bool RemovedAllMTVRSAVEs = true;
- // See if we can find and remove the MTVRSAVE instruction from all of the
- // epilog blocks.
- for (MachineFunction::iterator I = MF->begin(), E = MF->end(); I != E; ++I) {
- // If last instruction is a return instruction, add an epilogue
- if (!I->empty() && I->back().getDesc().isReturn()) {
- bool FoundIt = false;
- for (MBBI = I->end(); MBBI != I->begin(); ) {
- --MBBI;
- if (MBBI->getOpcode() == PPC::MTVRSAVE) {
- MBBI->eraseFromParent(); // remove it.
- FoundIt = true;
- break;
- }
- }
- RemovedAllMTVRSAVEs &= FoundIt;
- }
- }
+unsigned PPCRegisterInfo::getFrameRegister(const MachineFunction &MF) const {
+ const TargetFrameLowering *TFI = MF.getTarget().getFrameLowering();
- // If we found and removed all MTVRSAVE instructions, remove the read of
- // VRSAVE as well.
- if (RemovedAllMTVRSAVEs) {
- MBBI = MI;
- assert(MBBI != Entry->begin() && "UPDATE_VRSAVE is first instr in block?");
- --MBBI;
- assert(MBBI->getOpcode() == PPC::MFVRSAVE && "VRSAVE instrs wandered?");
- MBBI->eraseFromParent();
- }
-
- // Finally, nuke the UPDATE_VRSAVE.
- MI->eraseFromParent();
+ if (!Subtarget.isPPC64())
+ return TFI->hasFP(MF) ? PPC::R31 : PPC::R1;
+ else
+ return TFI->hasFP(MF) ? PPC::X31 : PPC::X1;
}
-// HandleVRSaveUpdate - MI is the UPDATE_VRSAVE instruction introduced by the
-// instruction selector. Based on the vector registers that have been used,
-// transform this into the appropriate ORI instruction.
-static void HandleVRSaveUpdate(MachineInstr *MI, const TargetInstrInfo &TII) {
- MachineFunction *MF = MI->getParent()->getParent();
- DebugLoc dl = MI->getDebugLoc();
-
- unsigned UsedRegMask = 0;
- for (unsigned i = 0; i != 32; ++i)
- if (MF->getRegInfo().isPhysRegUsed(VRRegNo[i]))
- UsedRegMask |= 1 << (31-i);
-
- // Live in and live out values already must be in the mask, so don't bother
- // marking them.
- for (MachineRegisterInfo::livein_iterator
- I = MF->getRegInfo().livein_begin(),
- E = MF->getRegInfo().livein_end(); I != E; ++I) {
- unsigned RegNo = PPCRegisterInfo::getRegisterNumbering(I->first);
- if (VRRegNo[RegNo] == I->first) // If this really is a vector reg.
- UsedRegMask &= ~(1 << (31-RegNo)); // Doesn't need to be marked.
- }
- for (MachineRegisterInfo::liveout_iterator
- I = MF->getRegInfo().liveout_begin(),
- E = MF->getRegInfo().liveout_end(); I != E; ++I) {
- unsigned RegNo = PPCRegisterInfo::getRegisterNumbering(*I);
- if (VRRegNo[RegNo] == *I) // If this really is a vector reg.
- UsedRegMask &= ~(1 << (31-RegNo)); // Doesn't need to be marked.
- }
-
- // If no registers are used, turn this into a copy.
- if (UsedRegMask == 0) {
- // Remove all VRSAVE code.
- RemoveVRSaveCode(MI);
- return;
- }
+unsigned PPCRegisterInfo::getBaseRegister(const MachineFunction &MF) const {
+ if (!hasBasePointer(MF))
+ return getFrameRegister(MF);
- unsigned SrcReg = MI->getOperand(1).getReg();
- unsigned DstReg = MI->getOperand(0).getReg();
-
- if ((UsedRegMask & 0xFFFF) == UsedRegMask) {
- if (DstReg != SrcReg)
- BuildMI(*MI->getParent(), MI, dl, TII.get(PPC::ORI), DstReg)
- .addReg(SrcReg)
- .addImm(UsedRegMask);
- else
- BuildMI(*MI->getParent(), MI, dl, TII.get(PPC::ORI), DstReg)
- .addReg(SrcReg, RegState::Kill)
- .addImm(UsedRegMask);
- } else if ((UsedRegMask & 0xFFFF0000) == UsedRegMask) {
- if (DstReg != SrcReg)
- BuildMI(*MI->getParent(), MI, dl, TII.get(PPC::ORIS), DstReg)
- .addReg(SrcReg)
- .addImm(UsedRegMask >> 16);
- else
- BuildMI(*MI->getParent(), MI, dl, TII.get(PPC::ORIS), DstReg)
- .addReg(SrcReg, RegState::Kill)
- .addImm(UsedRegMask >> 16);
- } else {
- if (DstReg != SrcReg)
- BuildMI(*MI->getParent(), MI, dl, TII.get(PPC::ORIS), DstReg)
- .addReg(SrcReg)
- .addImm(UsedRegMask >> 16);
- else
- BuildMI(*MI->getParent(), MI, dl, TII.get(PPC::ORIS), DstReg)
- .addReg(SrcReg, RegState::Kill)
- .addImm(UsedRegMask >> 16);
-
- BuildMI(*MI->getParent(), MI, dl, TII.get(PPC::ORI), DstReg)
- .addReg(DstReg, RegState::Kill)
- .addImm(UsedRegMask & 0xFFFF);
- }
-
- // Remove the old UPDATE_VRSAVE instruction.
- MI->eraseFromParent();
+ return Subtarget.isPPC64() ? PPC::X30 : PPC::R30;
}
-/// determineFrameLayout - Determine the size of the frame and maximum call
-/// frame size.
-void PPCRegisterInfo::determineFrameLayout(MachineFunction &MF) const {
- MachineFrameInfo *MFI = MF.getFrameInfo();
-
- // Get the number of bytes to allocate from the FrameInfo
- unsigned FrameSize = MFI->getStackSize();
-
- // Get the alignments provided by the target, and the maximum alignment
- // (if any) of the fixed frame objects.
- unsigned MaxAlign = MFI->getMaxAlignment();
- unsigned TargetAlign = MF.getTarget().getFrameInfo()->getStackAlignment();
- unsigned AlignMask = TargetAlign - 1; //
-
- // If we are a leaf function, and use up to 224 bytes of stack space,
- // don't have a frame pointer, calls, or dynamic alloca then we do not need
- // to adjust the stack pointer (we fit in the Red Zone).
- bool DisableRedZone = MF.getFunction()->hasFnAttr(Attribute::NoRedZone);
- // FIXME SVR4 The 32-bit SVR4 ABI has no red zone.
- if (!DisableRedZone &&
- FrameSize <= 224 && // Fits in red zone.
- !MFI->hasVarSizedObjects() && // No dynamic alloca.
- !MFI->adjustsStack() && // No calls.
- (!ALIGN_STACK || MaxAlign <= TargetAlign)) { // No special alignment.
- // No need for frame
- MFI->setStackSize(0);
- return;
- }
-
- // Get the maximum call frame size of all the calls.
- unsigned maxCallFrameSize = MFI->getMaxCallFrameSize();
-
- // Maximum call frame needs to be at least big enough for linkage and 8 args.
- unsigned minCallFrameSize =
- PPCFrameInfo::getMinCallFrameSize(Subtarget.isPPC64(),
- Subtarget.isDarwinABI());
- maxCallFrameSize = std::max(maxCallFrameSize, minCallFrameSize);
-
- // If we have dynamic alloca then maxCallFrameSize needs to be aligned so
- // that allocations will be aligned.
- if (MFI->hasVarSizedObjects())
- maxCallFrameSize = (maxCallFrameSize + AlignMask) & ~AlignMask;
-
- // Update maximum call frame size.
- MFI->setMaxCallFrameSize(maxCallFrameSize);
-
- // Include call frame size in total.
- FrameSize += maxCallFrameSize;
-
- // Make sure the frame is aligned.
- FrameSize = (FrameSize + AlignMask) & ~AlignMask;
+bool PPCRegisterInfo::hasBasePointer(const MachineFunction &MF) const {
+ if (!EnableBasePointer)
+ return false;
+ if (AlwaysBasePointer)
+ return true;
- // Update frame info.
- MFI->setStackSize(FrameSize);
+ // If we need to realign the stack, then the stack pointer can no longer
+ // serve as an offset into the caller's stack space. As a result, we need a
+ // base pointer.
+ return needsStackRealignment(MF);
}
-void
-PPCRegisterInfo::processFunctionBeforeCalleeSavedScan(MachineFunction &MF,
- RegScavenger *RS) const {
- // Save and clear the LR state.
- PPCFunctionInfo *FI = MF.getInfo<PPCFunctionInfo>();
- unsigned LR = getRARegister();
- FI->setMustSaveLR(MustSaveLR(MF, LR));
- MF.getRegInfo().setPhysRegUnused(LR);
-
- // Save R31 if necessary
- int FPSI = FI->getFramePointerSaveIndex();
- bool isPPC64 = Subtarget.isPPC64();
- bool isDarwinABI = Subtarget.isDarwinABI();
- MachineFrameInfo *MFI = MF.getFrameInfo();
-
- // If the frame pointer save index hasn't been defined yet.
- if (!FPSI && needsFP(MF)) {
- // Find out what the fix offset of the frame pointer save area.
- int FPOffset = PPCFrameInfo::getFramePointerSaveOffset(isPPC64,
- isDarwinABI);
- // Allocate the frame index for frame pointer save area.
- FPSI = MF.getFrameInfo()->CreateFixedObject(isPPC64? 8 : 4, FPOffset,
- true, false);
- // Save the result.
- FI->setFramePointerSaveIndex(FPSI);
- }
+bool PPCRegisterInfo::canRealignStack(const MachineFunction &MF) const {
+ if (MF.getFunction()->hasFnAttribute("no-realign-stack"))
+ return false;
- // Reserve stack space to move the linkage area to in case of a tail call.
- int TCSPDelta = 0;
- if (GuaranteedTailCallOpt && (TCSPDelta = FI->getTailCallSPDelta()) < 0) {
- MF.getFrameInfo()->CreateFixedObject(-1 * TCSPDelta, TCSPDelta,
- true, false);
- }
-
- // Reserve a slot closest to SP or frame pointer if we have a dynalloc or
- // a large stack, which will require scavenging a register to materialize a
- // large offset.
- // FIXME: this doesn't actually check stack size, so is a bit pessimistic
- // FIXME: doesn't detect whether or not we need to spill vXX, which requires
- // r0 for now.
-
- if (EnableRegisterScavenging) // FIXME (64-bit): Enable.
- if (needsFP(MF) || spillsCR(MF)) {
- const TargetRegisterClass *GPRC = &PPC::GPRCRegClass;
- const TargetRegisterClass *G8RC = &PPC::G8RCRegClass;
- const TargetRegisterClass *RC = isPPC64 ? G8RC : GPRC;
- RS->setScavengingFrameIndex(MFI->CreateStackObject(RC->getSize(),
- RC->getAlignment(),
- false));
- }
+ return true;
}
-void
-PPCRegisterInfo::processFunctionBeforeFrameFinalized(MachineFunction &MF)
- const {
- // Early exit if not using the SVR4 ABI.
- if (!Subtarget.isSVR4ABI()) {
- return;
- }
-
- // Get callee saved register information.
- MachineFrameInfo *FFI = MF.getFrameInfo();
- const std::vector<CalleeSavedInfo> &CSI = FFI->getCalleeSavedInfo();
-
- // Early exit if no callee saved registers are modified!
- if (CSI.empty() && !needsFP(MF)) {
- return;
- }
-
- unsigned MinGPR = PPC::R31;
- unsigned MinG8R = PPC::X31;
- unsigned MinFPR = PPC::F31;
- unsigned MinVR = PPC::V31;
-
- bool HasGPSaveArea = false;
- bool HasG8SaveArea = false;
- bool HasFPSaveArea = false;
- bool HasCRSaveArea = false;
- bool HasVRSAVESaveArea = false;
- bool HasVRSaveArea = false;
-
- SmallVector<CalleeSavedInfo, 18> GPRegs;
- SmallVector<CalleeSavedInfo, 18> G8Regs;
- SmallVector<CalleeSavedInfo, 18> FPRegs;
- SmallVector<CalleeSavedInfo, 18> VRegs;
-
- for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
- unsigned Reg = CSI[i].getReg();
- if (PPC::GPRCRegisterClass->contains(Reg)) {
- HasGPSaveArea = true;
-
- GPRegs.push_back(CSI[i]);
-
- if (Reg < MinGPR) {
- MinGPR = Reg;
- }
- } else if (PPC::G8RCRegisterClass->contains(Reg)) {
- HasG8SaveArea = true;
-
- G8Regs.push_back(CSI[i]);
-
- if (Reg < MinG8R) {
- MinG8R = Reg;
- }
- } else if (PPC::F8RCRegisterClass->contains(Reg)) {
- HasFPSaveArea = true;
-
- FPRegs.push_back(CSI[i]);
-
- if (Reg < MinFPR) {
- MinFPR = Reg;
- }
-// FIXME SVR4: Disable CR save area for now.
- } else if (PPC::CRBITRCRegisterClass->contains(Reg)
- || PPC::CRRCRegisterClass->contains(Reg)) {
-// HasCRSaveArea = true;
- } else if (PPC::VRSAVERCRegisterClass->contains(Reg)) {
- HasVRSAVESaveArea = true;
- } else if (PPC::VRRCRegisterClass->contains(Reg)) {
- HasVRSaveArea = true;
-
- VRegs.push_back(CSI[i]);
-
- if (Reg < MinVR) {
- MinVR = Reg;
- }
- } else {
- llvm_unreachable("Unknown RegisterClass!");
- }
- }
-
- PPCFunctionInfo *PFI = MF.getInfo<PPCFunctionInfo>();
-
- int64_t LowerBound = 0;
-
- // Take into account stack space reserved for tail calls.
- int TCSPDelta = 0;
- if (GuaranteedTailCallOpt && (TCSPDelta = PFI->getTailCallSPDelta()) < 0) {
- LowerBound = TCSPDelta;
- }
-
- // The Floating-point register save area is right below the back chain word
- // of the previous stack frame.
- if (HasFPSaveArea) {
- for (unsigned i = 0, e = FPRegs.size(); i != e; ++i) {
- int FI = FPRegs[i].getFrameIdx();
-
- FFI->setObjectOffset(FI, LowerBound + FFI->getObjectOffset(FI));
- }
-
- LowerBound -= (31 - getRegisterNumbering(MinFPR) + 1) * 8;
- }
-
- // Check whether the frame pointer register is allocated. If so, make sure it
- // is spilled to the correct offset.
- if (needsFP(MF)) {
- HasGPSaveArea = true;
-
- int FI = PFI->getFramePointerSaveIndex();
- assert(FI && "No Frame Pointer Save Slot!");
-
- FFI->setObjectOffset(FI, LowerBound + FFI->getObjectOffset(FI));
- }
-
- // General register save area starts right below the Floating-point
- // register save area.
- if (HasGPSaveArea || HasG8SaveArea) {
- // Move general register save area spill slots down, taking into account
- // the size of the Floating-point register save area.
- for (unsigned i = 0, e = GPRegs.size(); i != e; ++i) {
- int FI = GPRegs[i].getFrameIdx();
-
- FFI->setObjectOffset(FI, LowerBound + FFI->getObjectOffset(FI));
- }
-
- // Move general register save area spill slots down, taking into account
- // the size of the Floating-point register save area.
- for (unsigned i = 0, e = G8Regs.size(); i != e; ++i) {
- int FI = G8Regs[i].getFrameIdx();
-
- FFI->setObjectOffset(FI, LowerBound + FFI->getObjectOffset(FI));
- }
-
- unsigned MinReg = std::min<unsigned>(getRegisterNumbering(MinGPR),
- getRegisterNumbering(MinG8R));
-
- if (Subtarget.isPPC64()) {
- LowerBound -= (31 - MinReg + 1) * 8;
- } else {
- LowerBound -= (31 - MinReg + 1) * 4;
- }
- }
-
- // The CR save area is below the general register save area.
- if (HasCRSaveArea) {
- // FIXME SVR4: Is it actually possible to have multiple elements in CSI
- // which have the CR/CRBIT register class?
- // Adjust the frame index of the CR spill slot.
- for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
- unsigned Reg = CSI[i].getReg();
-
- if (PPC::CRBITRCRegisterClass->contains(Reg) ||
- PPC::CRRCRegisterClass->contains(Reg)) {
- int FI = CSI[i].getFrameIdx();
-
- FFI->setObjectOffset(FI, LowerBound + FFI->getObjectOffset(FI));
- }
- }
-
- LowerBound -= 4; // The CR save area is always 4 bytes long.
- }
-
- if (HasVRSAVESaveArea) {
- // FIXME SVR4: Is it actually possible to have multiple elements in CSI
- // which have the VRSAVE register class?
- // Adjust the frame index of the VRSAVE spill slot.
- for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
- unsigned Reg = CSI[i].getReg();
-
- if (PPC::VRSAVERCRegisterClass->contains(Reg)) {
- int FI = CSI[i].getFrameIdx();
-
- FFI->setObjectOffset(FI, LowerBound + FFI->getObjectOffset(FI));
- }
- }
-
- LowerBound -= 4; // The VRSAVE save area is always 4 bytes long.
- }
-
- if (HasVRSaveArea) {
- // Insert alignment padding, we need 16-byte alignment.
- LowerBound = (LowerBound - 15) & ~(15);
-
- for (unsigned i = 0, e = VRegs.size(); i != e; ++i) {
- int FI = VRegs[i].getFrameIdx();
-
- FFI->setObjectOffset(FI, LowerBound + FFI->getObjectOffset(FI));
- }
- }
+bool PPCRegisterInfo::needsStackRealignment(const MachineFunction &MF) const {
+ const MachineFrameInfo *MFI = MF.getFrameInfo();
+ const Function *F = MF.getFunction();
+ unsigned StackAlign = MF.getTarget().getFrameLowering()->getStackAlignment();
+ bool requiresRealignment =
+ ((MFI->getMaxAlignment() > StackAlign) ||
+ F->getAttributes().hasAttribute(AttributeSet::FunctionIndex,
+ Attribute::StackAlignment));
+
+ return requiresRealignment && canRealignStack(MF);
}
-void
-PPCRegisterInfo::emitPrologue(MachineFunction &MF) const {
- MachineBasicBlock &MBB = MF.front(); // Prolog goes in entry BB
- MachineBasicBlock::iterator MBBI = MBB.begin();
- MachineFrameInfo *MFI = MF.getFrameInfo();
- MachineModuleInfo &MMI = MF.getMMI();
- DebugLoc dl;
- bool needsFrameMoves = MMI.hasDebugInfo() ||
- !MF.getFunction()->doesNotThrow() ||
- UnwindTablesMandatory;
-
- // Prepare for frame info.
- MCSymbol *FrameLabel = 0;
-
- // Scan the prolog, looking for an UPDATE_VRSAVE instruction. If we find it,
- // process it.
- for (unsigned i = 0; MBBI != MBB.end(); ++i, ++MBBI) {
- if (MBBI->getOpcode() == PPC::UPDATE_VRSAVE) {
- HandleVRSaveUpdate(MBBI, TII);
- break;
- }
- }
-
- // Move MBBI back to the beginning of the function.
- MBBI = MBB.begin();
-
- // Work out frame sizes.
- determineFrameLayout(MF);
- unsigned FrameSize = MFI->getStackSize();
-
- int NegFrameSize = -FrameSize;
-
- // Get processor type.
- bool isPPC64 = Subtarget.isPPC64();
- // Get operating system
- bool isDarwinABI = Subtarget.isDarwinABI();
- // Check if the link register (LR) must be saved.
- PPCFunctionInfo *FI = MF.getInfo<PPCFunctionInfo>();
- bool MustSaveLR = FI->mustSaveLR();
- // Do we have a frame pointer for this function?
- bool HasFP = hasFP(MF) && FrameSize;
-
- int LROffset = PPCFrameInfo::getReturnSaveOffset(isPPC64, isDarwinABI);
-
- int FPOffset = 0;
- if (HasFP) {
- if (Subtarget.isSVR4ABI()) {
- MachineFrameInfo *FFI = MF.getFrameInfo();
- int FPIndex = FI->getFramePointerSaveIndex();
- assert(FPIndex && "No Frame Pointer Save Slot!");
- FPOffset = FFI->getObjectOffset(FPIndex);
- } else {
- FPOffset = PPCFrameInfo::getFramePointerSaveOffset(isPPC64, isDarwinABI);
- }
- }
-
- if (isPPC64) {
- if (MustSaveLR)
- BuildMI(MBB, MBBI, dl, TII.get(PPC::MFLR8), PPC::X0);
-
- if (HasFP)
- BuildMI(MBB, MBBI, dl, TII.get(PPC::STD))
- .addReg(PPC::X31)
- .addImm(FPOffset/4)
- .addReg(PPC::X1);
-
- if (MustSaveLR)
- BuildMI(MBB, MBBI, dl, TII.get(PPC::STD))
- .addReg(PPC::X0)
- .addImm(LROffset / 4)
- .addReg(PPC::X1);
- } else {
- if (MustSaveLR)
- BuildMI(MBB, MBBI, dl, TII.get(PPC::MFLR), PPC::R0);
-
- if (HasFP)
- BuildMI(MBB, MBBI, dl, TII.get(PPC::STW))
- .addReg(PPC::R31)
- .addImm(FPOffset)
- .addReg(PPC::R1);
-
- if (MustSaveLR)
- BuildMI(MBB, MBBI, dl, TII.get(PPC::STW))
- .addReg(PPC::R0)
- .addImm(LROffset)
- .addReg(PPC::R1);
- }
-
- // Skip if a leaf routine.
- if (!FrameSize) return;
-
- // Get stack alignments.
- unsigned TargetAlign = MF.getTarget().getFrameInfo()->getStackAlignment();
- unsigned MaxAlign = MFI->getMaxAlignment();
-
- // Adjust stack pointer: r1 += NegFrameSize.
- // If there is a preferred stack alignment, align R1 now
- if (!isPPC64) {
- // PPC32.
- if (ALIGN_STACK && MaxAlign > TargetAlign) {
- assert(isPowerOf2_32(MaxAlign) && isInt<16>(MaxAlign) &&
- "Invalid alignment!");
- assert(isInt<16>(NegFrameSize) && "Unhandled stack size and alignment!");
-
- BuildMI(MBB, MBBI, dl, TII.get(PPC::RLWINM), PPC::R0)
- .addReg(PPC::R1)
- .addImm(0)
- .addImm(32 - Log2_32(MaxAlign))
- .addImm(31);
- BuildMI(MBB, MBBI, dl, TII.get(PPC::SUBFIC) ,PPC::R0)
- .addReg(PPC::R0, RegState::Kill)
- .addImm(NegFrameSize);
- BuildMI(MBB, MBBI, dl, TII.get(PPC::STWUX))
- .addReg(PPC::R1)
- .addReg(PPC::R1)
- .addReg(PPC::R0);
- } else if (isInt<16>(NegFrameSize)) {
- BuildMI(MBB, MBBI, dl, TII.get(PPC::STWU), PPC::R1)
- .addReg(PPC::R1)
- .addImm(NegFrameSize)
- .addReg(PPC::R1);
- } else {
- BuildMI(MBB, MBBI, dl, TII.get(PPC::LIS), PPC::R0)
- .addImm(NegFrameSize >> 16);
- BuildMI(MBB, MBBI, dl, TII.get(PPC::ORI), PPC::R0)
- .addReg(PPC::R0, RegState::Kill)
- .addImm(NegFrameSize & 0xFFFF);
- BuildMI(MBB, MBBI, dl, TII.get(PPC::STWUX))
- .addReg(PPC::R1)
- .addReg(PPC::R1)
- .addReg(PPC::R0);
- }
- } else { // PPC64.
- if (ALIGN_STACK && MaxAlign > TargetAlign) {
- assert(isPowerOf2_32(MaxAlign) && isInt<16>(MaxAlign) &&
- "Invalid alignment!");
- assert(isInt<16>(NegFrameSize) && "Unhandled stack size and alignment!");
-
- BuildMI(MBB, MBBI, dl, TII.get(PPC::RLDICL), PPC::X0)
- .addReg(PPC::X1)
- .addImm(0)
- .addImm(64 - Log2_32(MaxAlign));
- BuildMI(MBB, MBBI, dl, TII.get(PPC::SUBFIC8), PPC::X0)
- .addReg(PPC::X0)
- .addImm(NegFrameSize);
- BuildMI(MBB, MBBI, dl, TII.get(PPC::STDUX))
- .addReg(PPC::X1)
- .addReg(PPC::X1)
- .addReg(PPC::X0);
- } else if (isInt<16>(NegFrameSize)) {
- BuildMI(MBB, MBBI, dl, TII.get(PPC::STDU), PPC::X1)
- .addReg(PPC::X1)
- .addImm(NegFrameSize / 4)
- .addReg(PPC::X1);
- } else {
- BuildMI(MBB, MBBI, dl, TII.get(PPC::LIS8), PPC::X0)
- .addImm(NegFrameSize >> 16);
- BuildMI(MBB, MBBI, dl, TII.get(PPC::ORI8), PPC::X0)
- .addReg(PPC::X0, RegState::Kill)
- .addImm(NegFrameSize & 0xFFFF);
- BuildMI(MBB, MBBI, dl, TII.get(PPC::STDUX))
- .addReg(PPC::X1)
- .addReg(PPC::X1)
- .addReg(PPC::X0);
- }
- }
-
- std::vector<MachineMove> &Moves = MMI.getFrameMoves();
-
- // Add the "machine moves" for the instructions we generated above, but in
- // reverse order.
- if (needsFrameMoves) {
- // Mark effective beginning of when frame pointer becomes valid.
- FrameLabel = MMI.getContext().CreateTempSymbol();
- BuildMI(MBB, MBBI, dl, TII.get(PPC::DBG_LABEL)).addSym(FrameLabel);
-
- // Show update of SP.
- if (NegFrameSize) {
- MachineLocation SPDst(MachineLocation::VirtualFP);
- MachineLocation SPSrc(MachineLocation::VirtualFP, NegFrameSize);
- Moves.push_back(MachineMove(FrameLabel, SPDst, SPSrc));
- } else {
- MachineLocation SP(isPPC64 ? PPC::X31 : PPC::R31);
- Moves.push_back(MachineMove(FrameLabel, SP, SP));
- }
-
- if (HasFP) {
- MachineLocation FPDst(MachineLocation::VirtualFP, FPOffset);
- MachineLocation FPSrc(isPPC64 ? PPC::X31 : PPC::R31);
- Moves.push_back(MachineMove(FrameLabel, FPDst, FPSrc));
- }
-
- if (MustSaveLR) {
- MachineLocation LRDst(MachineLocation::VirtualFP, LROffset);
- MachineLocation LRSrc(isPPC64 ? PPC::LR8 : PPC::LR);
- Moves.push_back(MachineMove(FrameLabel, LRDst, LRSrc));
- }
- }
-
- MCSymbol *ReadyLabel = 0;
-
- // If there is a frame pointer, copy R1 into R31
- if (HasFP) {
- if (!isPPC64) {
- BuildMI(MBB, MBBI, dl, TII.get(PPC::OR), PPC::R31)
- .addReg(PPC::R1)
- .addReg(PPC::R1);
- } else {
- BuildMI(MBB, MBBI, dl, TII.get(PPC::OR8), PPC::X31)
- .addReg(PPC::X1)
- .addReg(PPC::X1);
- }
-
- if (needsFrameMoves) {
- ReadyLabel = MMI.getContext().CreateTempSymbol();
-
- // Mark effective beginning of when frame pointer is ready.
- BuildMI(MBB, MBBI, dl, TII.get(PPC::DBG_LABEL)).addSym(ReadyLabel);
-
- MachineLocation FPDst(HasFP ? (isPPC64 ? PPC::X31 : PPC::R31) :
- (isPPC64 ? PPC::X1 : PPC::R1));
- MachineLocation FPSrc(MachineLocation::VirtualFP);
- Moves.push_back(MachineMove(ReadyLabel, FPDst, FPSrc));
- }
- }
-
- if (needsFrameMoves) {
- MCSymbol *Label = HasFP ? ReadyLabel : FrameLabel;
-
- // Add callee saved registers to move list.
- const std::vector<CalleeSavedInfo> &CSI = MFI->getCalleeSavedInfo();
- for (unsigned I = 0, E = CSI.size(); I != E; ++I) {
- int Offset = MFI->getObjectOffset(CSI[I].getFrameIdx());
- unsigned Reg = CSI[I].getReg();
- if (Reg == PPC::LR || Reg == PPC::LR8 || Reg == PPC::RM) continue;
- MachineLocation CSDst(MachineLocation::VirtualFP, Offset);
- MachineLocation CSSrc(Reg);
- Moves.push_back(MachineMove(Label, CSDst, CSSrc));
- }
+/// Returns true if the instruction's frame index
+/// reference would be better served by a base register other than FP
+/// or SP. Used by LocalStackFrameAllocation to determine which frame index
+/// references it should create new base registers for.
+bool PPCRegisterInfo::
+needsFrameBaseReg(MachineInstr *MI, int64_t Offset) const {
+ assert(Offset < 0 && "Local offset must be negative");
+
+ unsigned FIOperandNum = 0;
+ while (!MI->getOperand(FIOperandNum).isFI()) {
+ ++FIOperandNum;
+ assert(FIOperandNum < MI->getNumOperands() &&
+ "Instr doesn't have FrameIndex operand!");
}
-}
-void PPCRegisterInfo::emitEpilogue(MachineFunction &MF,
- MachineBasicBlock &MBB) const {
- MachineBasicBlock::iterator MBBI = prior(MBB.end());
- unsigned RetOpcode = MBBI->getOpcode();
- DebugLoc dl;
-
- assert( (RetOpcode == PPC::BLR ||
- RetOpcode == PPC::TCRETURNri ||
- RetOpcode == PPC::TCRETURNdi ||
- RetOpcode == PPC::TCRETURNai ||
- RetOpcode == PPC::TCRETURNri8 ||
- RetOpcode == PPC::TCRETURNdi8 ||
- RetOpcode == PPC::TCRETURNai8) &&
- "Can only insert epilog into returning blocks");
-
- // Get alignment info so we know how to restore r1
- const MachineFrameInfo *MFI = MF.getFrameInfo();
- unsigned TargetAlign = MF.getTarget().getFrameInfo()->getStackAlignment();
- unsigned MaxAlign = MFI->getMaxAlignment();
+ unsigned OffsetOperandNo = getOffsetONFromFION(*MI, FIOperandNum);
+ Offset += MI->getOperand(OffsetOperandNo).getImm();
- // Get the number of bytes allocated from the FrameInfo.
- int FrameSize = MFI->getStackSize();
+ // It's the load/store FI references that cause issues, as it can be difficult
+ // to materialize the offset if it won't fit in the literal field. Estimate
+ // based on the size of the local frame and some conservative assumptions
+ // about the rest of the stack frame (note, this is pre-regalloc, so
+ // we don't know everything for certain yet) whether this offset is likely
+ // to be out of range of the immediate. Return true if so.
- // Get processor type.
- bool isPPC64 = Subtarget.isPPC64();
- // Get operating system
- bool isDarwinABI = Subtarget.isDarwinABI();
- // Check if the link register (LR) has been saved.
- PPCFunctionInfo *FI = MF.getInfo<PPCFunctionInfo>();
- bool MustSaveLR = FI->mustSaveLR();
- // Do we have a frame pointer for this function?
- bool HasFP = hasFP(MF) && FrameSize;
-
- int LROffset = PPCFrameInfo::getReturnSaveOffset(isPPC64, isDarwinABI);
+ // We only generate virtual base registers for loads and stores that have
+ // an r+i form. Return false for everything else.
+ unsigned OpC = MI->getOpcode();
+ if (!ImmToIdxMap.count(OpC))
+ return false;
- int FPOffset = 0;
- if (HasFP) {
- if (Subtarget.isSVR4ABI()) {
- MachineFrameInfo *FFI = MF.getFrameInfo();
- int FPIndex = FI->getFramePointerSaveIndex();
- assert(FPIndex && "No Frame Pointer Save Slot!");
- FPOffset = FFI->getObjectOffset(FPIndex);
- } else {
- FPOffset = PPCFrameInfo::getFramePointerSaveOffset(isPPC64, isDarwinABI);
- }
- }
-
- bool UsesTCRet = RetOpcode == PPC::TCRETURNri ||
- RetOpcode == PPC::TCRETURNdi ||
- RetOpcode == PPC::TCRETURNai ||
- RetOpcode == PPC::TCRETURNri8 ||
- RetOpcode == PPC::TCRETURNdi8 ||
- RetOpcode == PPC::TCRETURNai8;
-
- if (UsesTCRet) {
- int MaxTCRetDelta = FI->getTailCallSPDelta();
- MachineOperand &StackAdjust = MBBI->getOperand(1);
- assert(StackAdjust.isImm() && "Expecting immediate value.");
- // Adjust stack pointer.
- int StackAdj = StackAdjust.getImm();
- int Delta = StackAdj - MaxTCRetDelta;
- assert((Delta >= 0) && "Delta must be positive");
- if (MaxTCRetDelta>0)
- FrameSize += (StackAdj +Delta);
- else
- FrameSize += StackAdj;
- }
+ // Don't generate a new virtual base register just to add zero to it.
+ if ((OpC == PPC::ADDI || OpC == PPC::ADDI8) &&
+ MI->getOperand(2).getImm() == 0)
+ return false;
- if (FrameSize) {
- // The loaded (or persistent) stack pointer value is offset by the 'stwu'
- // on entry to the function. Add this offset back now.
- if (!isPPC64) {
- // If this function contained a fastcc call and GuaranteedTailCallOpt is
- // enabled (=> hasFastCall()==true) the fastcc call might contain a tail
- // call which invalidates the stack pointer value in SP(0). So we use the
- // value of R31 in this case.
- if (FI->hasFastCall() && isInt<16>(FrameSize)) {
- assert(hasFP(MF) && "Expecting a valid the frame pointer.");
- BuildMI(MBB, MBBI, dl, TII.get(PPC::ADDI), PPC::R1)
- .addReg(PPC::R31).addImm(FrameSize);
- } else if(FI->hasFastCall()) {
- BuildMI(MBB, MBBI, dl, TII.get(PPC::LIS), PPC::R0)
- .addImm(FrameSize >> 16);
- BuildMI(MBB, MBBI, dl, TII.get(PPC::ORI), PPC::R0)
- .addReg(PPC::R0, RegState::Kill)
- .addImm(FrameSize & 0xFFFF);
- BuildMI(MBB, MBBI, dl, TII.get(PPC::ADD4))
- .addReg(PPC::R1)
- .addReg(PPC::R31)
- .addReg(PPC::R0);
- } else if (isInt<16>(FrameSize) &&
- (!ALIGN_STACK || TargetAlign >= MaxAlign) &&
- !MFI->hasVarSizedObjects()) {
- BuildMI(MBB, MBBI, dl, TII.get(PPC::ADDI), PPC::R1)
- .addReg(PPC::R1).addImm(FrameSize);
- } else {
- BuildMI(MBB, MBBI, dl, TII.get(PPC::LWZ),PPC::R1)
- .addImm(0).addReg(PPC::R1);
- }
- } else {
- if (FI->hasFastCall() && isInt<16>(FrameSize)) {
- assert(hasFP(MF) && "Expecting a valid the frame pointer.");
- BuildMI(MBB, MBBI, dl, TII.get(PPC::ADDI8), PPC::X1)
- .addReg(PPC::X31).addImm(FrameSize);
- } else if(FI->hasFastCall()) {
- BuildMI(MBB, MBBI, dl, TII.get(PPC::LIS8), PPC::X0)
- .addImm(FrameSize >> 16);
- BuildMI(MBB, MBBI, dl, TII.get(PPC::ORI8), PPC::X0)
- .addReg(PPC::X0, RegState::Kill)
- .addImm(FrameSize & 0xFFFF);
- BuildMI(MBB, MBBI, dl, TII.get(PPC::ADD8))
- .addReg(PPC::X1)
- .addReg(PPC::X31)
- .addReg(PPC::X0);
- } else if (isInt<16>(FrameSize) && TargetAlign >= MaxAlign &&
- !MFI->hasVarSizedObjects()) {
- BuildMI(MBB, MBBI, dl, TII.get(PPC::ADDI8), PPC::X1)
- .addReg(PPC::X1).addImm(FrameSize);
- } else {
- BuildMI(MBB, MBBI, dl, TII.get(PPC::LD), PPC::X1)
- .addImm(0).addReg(PPC::X1);
- }
- }
- }
+ MachineBasicBlock &MBB = *MI->getParent();
+ MachineFunction &MF = *MBB.getParent();
- if (isPPC64) {
- if (MustSaveLR)
- BuildMI(MBB, MBBI, dl, TII.get(PPC::LD), PPC::X0)
- .addImm(LROffset/4).addReg(PPC::X1);
-
- if (HasFP)
- BuildMI(MBB, MBBI, dl, TII.get(PPC::LD), PPC::X31)
- .addImm(FPOffset/4).addReg(PPC::X1);
-
- if (MustSaveLR)
- BuildMI(MBB, MBBI, dl, TII.get(PPC::MTLR8)).addReg(PPC::X0);
- } else {
- if (MustSaveLR)
- BuildMI(MBB, MBBI, dl, TII.get(PPC::LWZ), PPC::R0)
- .addImm(LROffset).addReg(PPC::R1);
-
- if (HasFP)
- BuildMI(MBB, MBBI, dl, TII.get(PPC::LWZ), PPC::R31)
- .addImm(FPOffset).addReg(PPC::R1);
-
- if (MustSaveLR)
- BuildMI(MBB, MBBI, dl, TII.get(PPC::MTLR)).addReg(PPC::R0);
- }
+ const PPCFrameLowering *PPCFI =
+ static_cast<const PPCFrameLowering*>(MF.getTarget().getFrameLowering());
+ unsigned StackEst =
+ PPCFI->determineFrameLayout(MF, false, true);
- // Callee pop calling convention. Pop parameter/linkage area. Used for tail
- // call optimization
- if (GuaranteedTailCallOpt && RetOpcode == PPC::BLR &&
- MF.getFunction()->getCallingConv() == CallingConv::Fast) {
- PPCFunctionInfo *FI = MF.getInfo<PPCFunctionInfo>();
- unsigned CallerAllocatedAmt = FI->getMinReservedArea();
- unsigned StackReg = isPPC64 ? PPC::X1 : PPC::R1;
- unsigned FPReg = isPPC64 ? PPC::X31 : PPC::R31;
- unsigned TmpReg = isPPC64 ? PPC::X0 : PPC::R0;
- unsigned ADDIInstr = isPPC64 ? PPC::ADDI8 : PPC::ADDI;
- unsigned ADDInstr = isPPC64 ? PPC::ADD8 : PPC::ADD4;
- unsigned LISInstr = isPPC64 ? PPC::LIS8 : PPC::LIS;
- unsigned ORIInstr = isPPC64 ? PPC::ORI8 : PPC::ORI;
-
- if (CallerAllocatedAmt && isInt<16>(CallerAllocatedAmt)) {
- BuildMI(MBB, MBBI, dl, TII.get(ADDIInstr), StackReg)
- .addReg(StackReg).addImm(CallerAllocatedAmt);
- } else {
- BuildMI(MBB, MBBI, dl, TII.get(LISInstr), TmpReg)
- .addImm(CallerAllocatedAmt >> 16);
- BuildMI(MBB, MBBI, dl, TII.get(ORIInstr), TmpReg)
- .addReg(TmpReg, RegState::Kill)
- .addImm(CallerAllocatedAmt & 0xFFFF);
- BuildMI(MBB, MBBI, dl, TII.get(ADDInstr))
- .addReg(StackReg)
- .addReg(FPReg)
- .addReg(TmpReg);
- }
- } else if (RetOpcode == PPC::TCRETURNdi) {
- MBBI = prior(MBB.end());
- MachineOperand &JumpTarget = MBBI->getOperand(0);
- BuildMI(MBB, MBBI, dl, TII.get(PPC::TAILB)).
- addGlobalAddress(JumpTarget.getGlobal(), JumpTarget.getOffset());
- } else if (RetOpcode == PPC::TCRETURNri) {
- MBBI = prior(MBB.end());
- assert(MBBI->getOperand(0).isReg() && "Expecting register operand.");
- BuildMI(MBB, MBBI, dl, TII.get(PPC::TAILBCTR));
- } else if (RetOpcode == PPC::TCRETURNai) {
- MBBI = prior(MBB.end());
- MachineOperand &JumpTarget = MBBI->getOperand(0);
- BuildMI(MBB, MBBI, dl, TII.get(PPC::TAILBA)).addImm(JumpTarget.getImm());
- } else if (RetOpcode == PPC::TCRETURNdi8) {
- MBBI = prior(MBB.end());
- MachineOperand &JumpTarget = MBBI->getOperand(0);
- BuildMI(MBB, MBBI, dl, TII.get(PPC::TAILB8)).
- addGlobalAddress(JumpTarget.getGlobal(), JumpTarget.getOffset());
- } else if (RetOpcode == PPC::TCRETURNri8) {
- MBBI = prior(MBB.end());
- assert(MBBI->getOperand(0).isReg() && "Expecting register operand.");
- BuildMI(MBB, MBBI, dl, TII.get(PPC::TAILBCTR8));
- } else if (RetOpcode == PPC::TCRETURNai8) {
- MBBI = prior(MBB.end());
- MachineOperand &JumpTarget = MBBI->getOperand(0);
- BuildMI(MBB, MBBI, dl, TII.get(PPC::TAILBA8)).addImm(JumpTarget.getImm());
- }
-}
+ // If we likely don't need a stack frame, then we probably don't need a
+ // virtual base register either.
+ if (!StackEst)
+ return false;
-unsigned PPCRegisterInfo::getRARegister() const {
- return !Subtarget.isPPC64() ? PPC::LR : PPC::LR8;
-}
+ // Estimate an offset from the stack pointer.
+ // The incoming offset is relating to the SP at the start of the function,
+ // but when we access the local it'll be relative to the SP after local
+ // allocation, so adjust our SP-relative offset by that allocation size.
+ Offset += StackEst;
-unsigned PPCRegisterInfo::getFrameRegister(const MachineFunction &MF) const {
- if (!Subtarget.isPPC64())
- return hasFP(MF) ? PPC::R31 : PPC::R1;
- else
- return hasFP(MF) ? PPC::X31 : PPC::X1;
+ // The frame pointer will point to the end of the stack, so estimate the
+ // offset as the difference between the object offset and the FP location.
+ return !isFrameOffsetLegal(MI, Offset);
}
-void PPCRegisterInfo::getInitialFrameState(std::vector<MachineMove> &Moves)
- const {
- // Initial state of the frame pointer is R1.
- MachineLocation Dst(MachineLocation::VirtualFP);
- MachineLocation Src(PPC::R1, 0);
- Moves.push_back(MachineMove(0, Dst, Src));
+/// Insert defining instruction(s) for BaseReg to
+/// be a pointer to FrameIdx at the beginning of the basic block.
+void PPCRegisterInfo::
+materializeFrameBaseRegister(MachineBasicBlock *MBB,
+ unsigned BaseReg, int FrameIdx,
+ int64_t Offset) const {
+ unsigned ADDriOpc = Subtarget.isPPC64() ? PPC::ADDI8 : PPC::ADDI;
+
+ MachineBasicBlock::iterator Ins = MBB->begin();
+ DebugLoc DL; // Defaults to "unknown"
+ if (Ins != MBB->end())
+ DL = Ins->getDebugLoc();
+
+ const MachineFunction &MF = *MBB->getParent();
+ const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();
+ const MCInstrDesc &MCID = TII.get(ADDriOpc);
+ MachineRegisterInfo &MRI = MBB->getParent()->getRegInfo();
+ MRI.constrainRegClass(BaseReg, TII.getRegClass(MCID, 0, this, MF));
+
+ BuildMI(*MBB, Ins, DL, MCID, BaseReg)
+ .addFrameIndex(FrameIdx).addImm(Offset);
}
-unsigned PPCRegisterInfo::getEHExceptionRegister() const {
- return !Subtarget.isPPC64() ? PPC::R3 : PPC::X3;
-}
+void
+PPCRegisterInfo::resolveFrameIndex(MachineBasicBlock::iterator I,
+ unsigned BaseReg, int64_t Offset) const {
+ MachineInstr &MI = *I;
+
+ unsigned FIOperandNum = 0;
+ while (!MI.getOperand(FIOperandNum).isFI()) {
+ ++FIOperandNum;
+ assert(FIOperandNum < MI.getNumOperands() &&
+ "Instr doesn't have FrameIndex operand!");
+ }
-unsigned PPCRegisterInfo::getEHHandlerRegister() const {
- return !Subtarget.isPPC64() ? PPC::R4 : PPC::X4;
+ MI.getOperand(FIOperandNum).ChangeToRegister(BaseReg, false);
+ unsigned OffsetOperandNo = getOffsetONFromFION(MI, FIOperandNum);
+ Offset += MI.getOperand(OffsetOperandNo).getImm();
+ MI.getOperand(OffsetOperandNo).ChangeToImmediate(Offset);
}
-int PPCRegisterInfo::getDwarfRegNum(unsigned RegNum, bool isEH) const {
- // FIXME: Most probably dwarf numbers differs for Linux and Darwin
- return PPCGenRegisterInfo::getDwarfRegNumFull(RegNum, 0);
+bool PPCRegisterInfo::isFrameOffsetLegal(const MachineInstr *MI,
+ int64_t Offset) const {
+ return MI->getOpcode() == PPC::DBG_VALUE || // DBG_VALUE is always Reg+Imm
+ (isInt<16>(Offset) && (!usesIXAddr(*MI) || (Offset & 3) == 0));
}
-#include "PPCGenRegisterInfo.inc"