1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386
|
//=====- SystemZFrameLowering.cpp - SystemZ Frame Information ------*- C++ -*-====//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file contains the SystemZ implementation of TargetFrameLowering class.
//
//===----------------------------------------------------------------------===//
#include "SystemZFrameLowering.h"
#include "SystemZInstrBuilder.h"
#include "SystemZInstrInfo.h"
#include "SystemZMachineFunctionInfo.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/Target/TargetData.h"
#include "llvm/Target/TargetOptions.h"
#include "llvm/Support/CommandLine.h"
using namespace llvm;
SystemZFrameLowering::SystemZFrameLowering(const SystemZSubtarget &sti)
: TargetFrameLowering(TargetFrameLowering::StackGrowsDown, 8, -160), STI(sti) {
// Fill the spill offsets map
static const unsigned SpillOffsTab[][2] = {
{ SystemZ::R2D, 0x10 },
{ SystemZ::R3D, 0x18 },
{ SystemZ::R4D, 0x20 },
{ SystemZ::R5D, 0x28 },
{ SystemZ::R6D, 0x30 },
{ SystemZ::R7D, 0x38 },
{ SystemZ::R8D, 0x40 },
{ SystemZ::R9D, 0x48 },
{ SystemZ::R10D, 0x50 },
{ SystemZ::R11D, 0x58 },
{ SystemZ::R12D, 0x60 },
{ SystemZ::R13D, 0x68 },
{ SystemZ::R14D, 0x70 },
{ SystemZ::R15D, 0x78 }
};
RegSpillOffsets.grow(SystemZ::NUM_TARGET_REGS);
for (unsigned i = 0, e = array_lengthof(SpillOffsTab); i != e; ++i)
RegSpillOffsets[SpillOffsTab[i][0]] = SpillOffsTab[i][1];
}
/// 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.
bool SystemZFrameLowering::hasFP(const MachineFunction &MF) const {
const MachineFrameInfo *MFI = MF.getFrameInfo();
return DisableFramePointerElim(MF) || MFI->hasVarSizedObjects();
}
/// emitSPUpdate - Emit a series of instructions to increment / decrement the
/// stack pointer by a constant value.
static
void emitSPUpdate(MachineBasicBlock &MBB, MachineBasicBlock::iterator &MBBI,
int64_t NumBytes, const TargetInstrInfo &TII) {
unsigned Opc; uint64_t Chunk;
bool isSub = NumBytes < 0;
uint64_t Offset = isSub ? -NumBytes : NumBytes;
if (Offset >= (1LL << 15) - 1) {
Opc = SystemZ::ADD64ri32;
Chunk = (1LL << 31) - 1;
} else {
Opc = SystemZ::ADD64ri16;
Chunk = (1LL << 15) - 1;
}
DebugLoc DL = MBBI != MBB.end() ? MBBI->getDebugLoc() : DebugLoc();
while (Offset) {
uint64_t ThisVal = (Offset > Chunk) ? Chunk : Offset;
MachineInstr *MI =
BuildMI(MBB, MBBI, DL, TII.get(Opc), SystemZ::R15D)
.addReg(SystemZ::R15D).addImm(isSub ? -ThisVal : ThisVal);
// The PSW implicit def is dead.
MI->getOperand(3).setIsDead();
Offset -= ThisVal;
}
}
void SystemZFrameLowering::emitPrologue(MachineFunction &MF) const {
MachineBasicBlock &MBB = MF.front(); // Prolog goes in entry BB
MachineFrameInfo *MFI = MF.getFrameInfo();
const SystemZInstrInfo &TII =
*static_cast<const SystemZInstrInfo*>(MF.getTarget().getInstrInfo());
SystemZMachineFunctionInfo *SystemZMFI =
MF.getInfo<SystemZMachineFunctionInfo>();
MachineBasicBlock::iterator MBBI = MBB.begin();
DebugLoc DL = MBBI != MBB.end() ? MBBI->getDebugLoc() : DebugLoc();
// Get the number of bytes to allocate from the FrameInfo.
// Note that area for callee-saved stuff is already allocated, thus we need to
// 'undo' the stack movement.
uint64_t StackSize = MFI->getStackSize();
StackSize -= SystemZMFI->getCalleeSavedFrameSize();
uint64_t NumBytes = StackSize - getOffsetOfLocalArea();
// Skip the callee-saved push instructions.
while (MBBI != MBB.end() &&
(MBBI->getOpcode() == SystemZ::MOV64mr ||
MBBI->getOpcode() == SystemZ::MOV64mrm))
++MBBI;
if (MBBI != MBB.end())
DL = MBBI->getDebugLoc();
// adjust stack pointer: R15 -= numbytes
if (StackSize || MFI->hasCalls()) {
assert(MF.getRegInfo().isPhysRegUsed(SystemZ::R15D) &&
"Invalid stack frame calculation!");
emitSPUpdate(MBB, MBBI, -(int64_t)NumBytes, TII);
}
if (hasFP(MF)) {
// Update R11 with the new base value...
BuildMI(MBB, MBBI, DL, TII.get(SystemZ::MOV64rr), SystemZ::R11D)
.addReg(SystemZ::R15D);
// Mark the FramePtr as live-in in every block except the entry.
for (MachineFunction::iterator I = llvm::next(MF.begin()), E = MF.end();
I != E; ++I)
I->addLiveIn(SystemZ::R11D);
}
}
void SystemZFrameLowering::emitEpilogue(MachineFunction &MF,
MachineBasicBlock &MBB) const {
const MachineFrameInfo *MFI = MF.getFrameInfo();
MachineBasicBlock::iterator MBBI = MBB.getLastNonDebugInstr();
const SystemZInstrInfo &TII =
*static_cast<const SystemZInstrInfo*>(MF.getTarget().getInstrInfo());
SystemZMachineFunctionInfo *SystemZMFI =
MF.getInfo<SystemZMachineFunctionInfo>();
unsigned RetOpcode = MBBI->getOpcode();
switch (RetOpcode) {
case SystemZ::RET: break; // These are ok
default:
assert(0 && "Can only insert epilog into returning blocks");
}
// Get the number of bytes to allocate from the FrameInfo
// Note that area for callee-saved stuff is already allocated, thus we need to
// 'undo' the stack movement.
uint64_t StackSize =
MFI->getStackSize() - SystemZMFI->getCalleeSavedFrameSize();
uint64_t NumBytes = StackSize - getOffsetOfLocalArea();
// Skip the final terminator instruction.
while (MBBI != MBB.begin()) {
MachineBasicBlock::iterator PI = prior(MBBI);
--MBBI;
if (!PI->getDesc().isTerminator())
break;
}
// During callee-saved restores emission stack frame was not yet finialized
// (and thus - the stack size was unknown). Tune the offset having full stack
// size in hands.
if (StackSize || MFI->hasCalls()) {
assert((MBBI->getOpcode() == SystemZ::MOV64rmm ||
MBBI->getOpcode() == SystemZ::MOV64rm) &&
"Expected to see callee-save register restore code");
assert(MF.getRegInfo().isPhysRegUsed(SystemZ::R15D) &&
"Invalid stack frame calculation!");
unsigned i = 0;
MachineInstr &MI = *MBBI;
while (!MI.getOperand(i).isImm()) {
++i;
assert(i < MI.getNumOperands() && "Unexpected restore code!");
}
uint64_t Offset = NumBytes + MI.getOperand(i).getImm();
// If Offset does not fit into 20-bit signed displacement field we need to
// emit some additional code...
if (Offset > 524287) {
// Fold the displacement into load instruction as much as possible.
NumBytes = Offset - 524287;
Offset = 524287;
emitSPUpdate(MBB, MBBI, NumBytes, TII);
}
MI.getOperand(i).ChangeToImmediate(Offset);
}
}
int SystemZFrameLowering::getFrameIndexOffset(const MachineFunction &MF,
int FI) const {
const MachineFrameInfo *MFI = MF.getFrameInfo();
const SystemZMachineFunctionInfo *SystemZMFI =
MF.getInfo<SystemZMachineFunctionInfo>();
int Offset = MFI->getObjectOffset(FI) + MFI->getOffsetAdjustment();
uint64_t StackSize = MFI->getStackSize();
// Fixed objects are really located in the "previous" frame.
if (FI < 0)
StackSize -= SystemZMFI->getCalleeSavedFrameSize();
Offset += StackSize - getOffsetOfLocalArea();
// Skip the register save area if we generated the stack frame.
if (StackSize || MFI->hasCalls())
Offset -= getOffsetOfLocalArea();
return Offset;
}
bool
SystemZFrameLowering::spillCalleeSavedRegisters(MachineBasicBlock &MBB,
MachineBasicBlock::iterator MI,
const std::vector<CalleeSavedInfo> &CSI,
const TargetRegisterInfo *TRI) const {
if (CSI.empty())
return false;
DebugLoc DL;
if (MI != MBB.end()) DL = MI->getDebugLoc();
MachineFunction &MF = *MBB.getParent();
const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();
SystemZMachineFunctionInfo *MFI = MF.getInfo<SystemZMachineFunctionInfo>();
unsigned CalleeFrameSize = 0;
// Scan the callee-saved and find the bounds of register spill area.
unsigned LowReg = 0, HighReg = 0, StartOffset = -1U, EndOffset = 0;
for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
unsigned Reg = CSI[i].getReg();
if (!SystemZ::FP64RegClass.contains(Reg)) {
unsigned Offset = RegSpillOffsets[Reg];
CalleeFrameSize += 8;
if (StartOffset > Offset) {
LowReg = Reg; StartOffset = Offset;
}
if (EndOffset < Offset) {
HighReg = Reg; EndOffset = RegSpillOffsets[Reg];
}
}
}
// Save information for epilogue inserter.
MFI->setCalleeSavedFrameSize(CalleeFrameSize);
MFI->setLowReg(LowReg); MFI->setHighReg(HighReg);
// Save GPRs
if (StartOffset) {
// Build a store instruction. Use STORE MULTIPLE instruction if there are many
// registers to store, otherwise - just STORE.
MachineInstrBuilder MIB =
BuildMI(MBB, MI, DL, TII.get((LowReg == HighReg ?
SystemZ::MOV64mr : SystemZ::MOV64mrm)));
// Add store operands.
MIB.addReg(SystemZ::R15D).addImm(StartOffset);
if (LowReg == HighReg)
MIB.addReg(0);
MIB.addReg(LowReg, RegState::Kill);
if (LowReg != HighReg)
MIB.addReg(HighReg, RegState::Kill);
// Do a second scan adding regs as being killed by instruction
for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
unsigned Reg = CSI[i].getReg();
// Add the callee-saved register as live-in. It's killed at the spill.
MBB.addLiveIn(Reg);
if (Reg != LowReg && Reg != HighReg)
MIB.addReg(Reg, RegState::ImplicitKill);
}
}
// Save FPRs
for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
unsigned Reg = CSI[i].getReg();
if (SystemZ::FP64RegClass.contains(Reg)) {
MBB.addLiveIn(Reg);
TII.storeRegToStackSlot(MBB, MI, Reg, true, CSI[i].getFrameIdx(),
&SystemZ::FP64RegClass, TRI);
}
}
return true;
}
bool
SystemZFrameLowering::restoreCalleeSavedRegisters(MachineBasicBlock &MBB,
MachineBasicBlock::iterator MI,
const std::vector<CalleeSavedInfo> &CSI,
const TargetRegisterInfo *TRI) const {
if (CSI.empty())
return false;
DebugLoc DL;
if (MI != MBB.end()) DL = MI->getDebugLoc();
MachineFunction &MF = *MBB.getParent();
const TargetInstrInfo &TII = *MF.getTarget().getInstrInfo();
SystemZMachineFunctionInfo *MFI = MF.getInfo<SystemZMachineFunctionInfo>();
// Restore FP registers
for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
unsigned Reg = CSI[i].getReg();
if (SystemZ::FP64RegClass.contains(Reg))
TII.loadRegFromStackSlot(MBB, MI, Reg, CSI[i].getFrameIdx(),
&SystemZ::FP64RegClass, TRI);
}
// Restore GP registers
unsigned LowReg = MFI->getLowReg(), HighReg = MFI->getHighReg();
unsigned StartOffset = RegSpillOffsets[LowReg];
if (StartOffset) {
// Build a load instruction. Use LOAD MULTIPLE instruction if there are many
// registers to load, otherwise - just LOAD.
MachineInstrBuilder MIB =
BuildMI(MBB, MI, DL, TII.get((LowReg == HighReg ?
SystemZ::MOV64rm : SystemZ::MOV64rmm)));
// Add store operands.
MIB.addReg(LowReg, RegState::Define);
if (LowReg != HighReg)
MIB.addReg(HighReg, RegState::Define);
MIB.addReg(hasFP(MF) ? SystemZ::R11D : SystemZ::R15D);
MIB.addImm(StartOffset);
if (LowReg == HighReg)
MIB.addReg(0);
// Do a second scan adding regs as being defined by instruction
for (unsigned i = 0, e = CSI.size(); i != e; ++i) {
unsigned Reg = CSI[i].getReg();
if (Reg != LowReg && Reg != HighReg)
MIB.addReg(Reg, RegState::ImplicitDefine);
}
}
return true;
}
void
SystemZFrameLowering::processFunctionBeforeCalleeSavedScan(MachineFunction &MF,
RegScavenger *RS) const {
// Determine whether R15/R14 will ever be clobbered inside the function. And
// if yes - mark it as 'callee' saved.
MachineFrameInfo *FFI = MF.getFrameInfo();
MachineRegisterInfo &MRI = MF.getRegInfo();
// Check whether high FPRs are ever used, if yes - we need to save R15 as
// well.
static const unsigned HighFPRs[] = {
SystemZ::F8L, SystemZ::F9L, SystemZ::F10L, SystemZ::F11L,
SystemZ::F12L, SystemZ::F13L, SystemZ::F14L, SystemZ::F15L,
SystemZ::F8S, SystemZ::F9S, SystemZ::F10S, SystemZ::F11S,
SystemZ::F12S, SystemZ::F13S, SystemZ::F14S, SystemZ::F15S,
};
bool HighFPRsUsed = false;
for (unsigned i = 0, e = array_lengthof(HighFPRs); i != e; ++i)
HighFPRsUsed |= MRI.isPhysRegUsed(HighFPRs[i]);
if (FFI->hasCalls())
/* FIXME: function is varargs */
/* FIXME: function grabs RA */
/* FIXME: function calls eh_return */
MRI.setPhysRegUsed(SystemZ::R14D);
if (HighFPRsUsed ||
FFI->hasCalls() ||
FFI->getObjectIndexEnd() != 0 || // Contains automatic variables
FFI->hasVarSizedObjects() // Function calls dynamic alloca's
/* FIXME: function is varargs */)
MRI.setPhysRegUsed(SystemZ::R15D);
}
|