mirror of
https://github.com/c64scene-ar/llvm-6502.git
synced 2024-08-27 15:29:51 +00:00
call to the MachineCodeEmitter interface and made copying the start line of a function not conditional on whether we're emitting Dwarf debug information. I'll propagate the processDebugLoc() calls to the non-X86 targets in a followup patch. In the long run, it'll probably be better to gather this information through the DwarfWriter, but the DwarfWriter currently depends on the AsmPrinter and TargetAsmInfo, and fixing that would be out of the way for this patch. There's a bug in OProfile 0.9.4 that makes it ignore line numbers for addresses above 4G, and a patch fixing it at http://thread.gmane.org/gmane.linux.oprofile/7634 Sample output: $ sudo opcontrol --reset; sudo opcontrol --start-daemon; sudo opcontrol --start; `pwd`/Debug/bin/lli fib.bc; sudo opcontrol --stop Signalling daemon... done Profiler running. fib(40) == 165580141 Stopping profiling. $ opreport -g -d -l `pwd`/Debug/bin/lli|head -60 Overflow stats not available CPU: Core 2, speed 1998 MHz (estimated) Counted CPU_CLK_UNHALTED events (Clock cycles when not halted) with a unit mask of 0x00 (Unhalted core cycles) count 100000 vma samples % linenr info image name symbol name 00007f67a30370b0 25489 61.2554 fib.c:24 10946.jo fib_left 00007f67a30370b0 1634 6.4106 fib.c:24 00007f67a30370b1 83 0.3256 fib.c:24 00007f67a30370b9 1997 7.8348 fib.c:24 00007f67a30370c6 2080 8.1604 fib.c:27 00007f67a30370c8 988 3.8762 fib.c:27 00007f67a30370cd 1315 5.1591 fib.c:27 00007f67a30370cf 251 0.9847 fib.c:27 00007f67a30370d3 1191 4.6726 fib.c:27 00007f67a30370d6 975 3.8252 fib.c:27 00007f67a30370db 1010 3.9625 fib.c:27 00007f67a30370dd 242 0.9494 fib.c:27 00007f67a30370e1 2782 10.9145 fib.c:28 00007f67a30370e5 3768 14.7828 fib.c:28 00007f67a30370eb 615 2.4128 (no location information) 00007f67a30370f3 6558 25.7287 (no location information) 00007f67a3037100 15603 37.4973 fib.c:29 10946.jo fib_right 00007f67a3037100 1646 10.5493 fib.c:29 00007f67a3037101 45 0.2884 fib.c:29 00007f67a3037109 2372 15.2022 fib.c:29 00007f67a3037116 2234 14.3178 fib.c:32 00007f67a3037118 612 3.9223 fib.c:32 00007f67a303711d 622 3.9864 fib.c:32 00007f67a303711f 385 2.4675 fib.c:32 00007f67a3037123 404 2.5892 fib.c:32 00007f67a3037126 634 4.0633 fib.c:32 00007f67a303712b 870 5.5759 fib.c:32 00007f67a303712d 62 0.3974 fib.c:32 00007f67a3037131 1848 11.8439 fib.c:33 00007f67a3037135 2840 18.2016 fib.c:33 00007f67a303713a 1 0.0064 fib.c:33 00007f67a303713b 1023 6.5564 (no location information) 00007f67a3037143 5 0.0320 (no location information) 000000000080c1e4 15 0.0360 MachineOperand.h:150 lli llvm::MachineOperand::isReg() const 000000000080c1e4 6 40.0000 MachineOperand.h:150 000000000080c1ec 2 13.3333 MachineOperand.h:150 ... git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@76102 91177308-0d34-0410-b5e6-96231b3b80d8
821 lines
29 KiB
C++
821 lines
29 KiB
C++
//===-- X86/X86CodeEmitter.cpp - Convert X86 code to machine code ---------===//
|
|
//
|
|
// 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 pass that transforms the X86 machine instructions into
|
|
// relocatable machine code.
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
#define DEBUG_TYPE "x86-emitter"
|
|
#include "X86InstrInfo.h"
|
|
#include "X86JITInfo.h"
|
|
#include "X86Subtarget.h"
|
|
#include "X86TargetMachine.h"
|
|
#include "X86Relocations.h"
|
|
#include "X86.h"
|
|
#include "llvm/PassManager.h"
|
|
#include "llvm/CodeGen/MachineCodeEmitter.h"
|
|
#include "llvm/CodeGen/JITCodeEmitter.h"
|
|
#include "llvm/CodeGen/ObjectCodeEmitter.h"
|
|
#include "llvm/CodeGen/MachineFunctionPass.h"
|
|
#include "llvm/CodeGen/MachineInstr.h"
|
|
#include "llvm/CodeGen/MachineModuleInfo.h"
|
|
#include "llvm/CodeGen/Passes.h"
|
|
#include "llvm/Function.h"
|
|
#include "llvm/ADT/Statistic.h"
|
|
#include "llvm/Support/Compiler.h"
|
|
#include "llvm/Support/Debug.h"
|
|
#include "llvm/Support/ErrorHandling.h"
|
|
#include "llvm/Target/TargetOptions.h"
|
|
using namespace llvm;
|
|
|
|
STATISTIC(NumEmitted, "Number of machine instructions emitted");
|
|
|
|
namespace {
|
|
template<class CodeEmitter>
|
|
class VISIBILITY_HIDDEN Emitter : public MachineFunctionPass {
|
|
const X86InstrInfo *II;
|
|
const TargetData *TD;
|
|
X86TargetMachine &TM;
|
|
CodeEmitter &MCE;
|
|
intptr_t PICBaseOffset;
|
|
bool Is64BitMode;
|
|
bool IsPIC;
|
|
public:
|
|
static char ID;
|
|
explicit Emitter(X86TargetMachine &tm, CodeEmitter &mce)
|
|
: MachineFunctionPass(&ID), II(0), TD(0), TM(tm),
|
|
MCE(mce), PICBaseOffset(0), Is64BitMode(false),
|
|
IsPIC(TM.getRelocationModel() == Reloc::PIC_) {}
|
|
Emitter(X86TargetMachine &tm, CodeEmitter &mce,
|
|
const X86InstrInfo &ii, const TargetData &td, bool is64)
|
|
: MachineFunctionPass(&ID), II(&ii), TD(&td), TM(tm),
|
|
MCE(mce), PICBaseOffset(0), Is64BitMode(is64),
|
|
IsPIC(TM.getRelocationModel() == Reloc::PIC_) {}
|
|
|
|
bool runOnMachineFunction(MachineFunction &MF);
|
|
|
|
virtual const char *getPassName() const {
|
|
return "X86 Machine Code Emitter";
|
|
}
|
|
|
|
void emitInstruction(const MachineInstr &MI,
|
|
const TargetInstrDesc *Desc);
|
|
|
|
void getAnalysisUsage(AnalysisUsage &AU) const {
|
|
AU.addRequired<MachineModuleInfo>();
|
|
MachineFunctionPass::getAnalysisUsage(AU);
|
|
}
|
|
|
|
private:
|
|
void emitPCRelativeBlockAddress(MachineBasicBlock *MBB);
|
|
void emitGlobalAddress(GlobalValue *GV, unsigned Reloc,
|
|
intptr_t Disp = 0, intptr_t PCAdj = 0,
|
|
bool NeedStub = false, bool Indirect = false);
|
|
void emitExternalSymbolAddress(const char *ES, unsigned Reloc);
|
|
void emitConstPoolAddress(unsigned CPI, unsigned Reloc, intptr_t Disp = 0,
|
|
intptr_t PCAdj = 0);
|
|
void emitJumpTableAddress(unsigned JTI, unsigned Reloc,
|
|
intptr_t PCAdj = 0);
|
|
|
|
void emitDisplacementField(const MachineOperand *RelocOp, int DispVal,
|
|
intptr_t PCAdj = 0);
|
|
|
|
void emitRegModRMByte(unsigned ModRMReg, unsigned RegOpcodeField);
|
|
void emitRegModRMByte(unsigned RegOpcodeField);
|
|
void emitSIBByte(unsigned SS, unsigned Index, unsigned Base);
|
|
void emitConstant(uint64_t Val, unsigned Size);
|
|
|
|
void emitMemModRMByte(const MachineInstr &MI,
|
|
unsigned Op, unsigned RegOpcodeField,
|
|
intptr_t PCAdj = 0);
|
|
|
|
unsigned getX86RegNum(unsigned RegNo) const;
|
|
};
|
|
|
|
template<class CodeEmitter>
|
|
char Emitter<CodeEmitter>::ID = 0;
|
|
}
|
|
|
|
/// createX86CodeEmitterPass - Return a pass that emits the collected X86 code
|
|
/// to the specified templated MachineCodeEmitter object.
|
|
|
|
FunctionPass *llvm::createX86CodeEmitterPass(X86TargetMachine &TM,
|
|
MachineCodeEmitter &MCE) {
|
|
return new Emitter<MachineCodeEmitter>(TM, MCE);
|
|
}
|
|
FunctionPass *llvm::createX86JITCodeEmitterPass(X86TargetMachine &TM,
|
|
JITCodeEmitter &JCE) {
|
|
return new Emitter<JITCodeEmitter>(TM, JCE);
|
|
}
|
|
FunctionPass *llvm::createX86ObjectCodeEmitterPass(X86TargetMachine &TM,
|
|
ObjectCodeEmitter &OCE) {
|
|
return new Emitter<ObjectCodeEmitter>(TM, OCE);
|
|
}
|
|
|
|
template<class CodeEmitter>
|
|
bool Emitter<CodeEmitter>::runOnMachineFunction(MachineFunction &MF) {
|
|
|
|
MCE.setModuleInfo(&getAnalysis<MachineModuleInfo>());
|
|
|
|
II = TM.getInstrInfo();
|
|
TD = TM.getTargetData();
|
|
Is64BitMode = TM.getSubtarget<X86Subtarget>().is64Bit();
|
|
IsPIC = TM.getRelocationModel() == Reloc::PIC_;
|
|
|
|
do {
|
|
DOUT << "JITTing function '" << MF.getFunction()->getName() << "'\n";
|
|
MCE.startFunction(MF);
|
|
for (MachineFunction::iterator MBB = MF.begin(), E = MF.end();
|
|
MBB != E; ++MBB) {
|
|
MCE.StartMachineBasicBlock(MBB);
|
|
for (MachineBasicBlock::const_iterator I = MBB->begin(), E = MBB->end();
|
|
I != E; ++I) {
|
|
const TargetInstrDesc &Desc = I->getDesc();
|
|
emitInstruction(*I, &Desc);
|
|
// MOVPC32r is basically a call plus a pop instruction.
|
|
if (Desc.getOpcode() == X86::MOVPC32r)
|
|
emitInstruction(*I, &II->get(X86::POP32r));
|
|
NumEmitted++; // Keep track of the # of mi's emitted
|
|
}
|
|
}
|
|
} while (MCE.finishFunction(MF));
|
|
|
|
return false;
|
|
}
|
|
|
|
/// emitPCRelativeBlockAddress - This method keeps track of the information
|
|
/// necessary to resolve the address of this block later and emits a dummy
|
|
/// value.
|
|
///
|
|
template<class CodeEmitter>
|
|
void Emitter<CodeEmitter>::emitPCRelativeBlockAddress(MachineBasicBlock *MBB) {
|
|
// Remember where this reference was and where it is to so we can
|
|
// deal with it later.
|
|
MCE.addRelocation(MachineRelocation::getBB(MCE.getCurrentPCOffset(),
|
|
X86::reloc_pcrel_word, MBB));
|
|
MCE.emitWordLE(0);
|
|
}
|
|
|
|
/// emitGlobalAddress - Emit the specified address to the code stream assuming
|
|
/// this is part of a "take the address of a global" instruction.
|
|
///
|
|
template<class CodeEmitter>
|
|
void Emitter<CodeEmitter>::emitGlobalAddress(GlobalValue *GV, unsigned Reloc,
|
|
intptr_t Disp /* = 0 */,
|
|
intptr_t PCAdj /* = 0 */,
|
|
bool NeedStub /* = false */,
|
|
bool Indirect /* = false */) {
|
|
intptr_t RelocCST = 0;
|
|
if (Reloc == X86::reloc_picrel_word)
|
|
RelocCST = PICBaseOffset;
|
|
else if (Reloc == X86::reloc_pcrel_word)
|
|
RelocCST = PCAdj;
|
|
MachineRelocation MR = Indirect
|
|
? MachineRelocation::getIndirectSymbol(MCE.getCurrentPCOffset(), Reloc,
|
|
GV, RelocCST, NeedStub)
|
|
: MachineRelocation::getGV(MCE.getCurrentPCOffset(), Reloc,
|
|
GV, RelocCST, NeedStub);
|
|
MCE.addRelocation(MR);
|
|
// The relocated value will be added to the displacement
|
|
if (Reloc == X86::reloc_absolute_dword)
|
|
MCE.emitDWordLE(Disp);
|
|
else
|
|
MCE.emitWordLE((int32_t)Disp);
|
|
}
|
|
|
|
/// emitExternalSymbolAddress - Arrange for the address of an external symbol to
|
|
/// be emitted to the current location in the function, and allow it to be PC
|
|
/// relative.
|
|
template<class CodeEmitter>
|
|
void Emitter<CodeEmitter>::emitExternalSymbolAddress(const char *ES,
|
|
unsigned Reloc) {
|
|
intptr_t RelocCST = (Reloc == X86::reloc_picrel_word) ? PICBaseOffset : 0;
|
|
MCE.addRelocation(MachineRelocation::getExtSym(MCE.getCurrentPCOffset(),
|
|
Reloc, ES, RelocCST));
|
|
if (Reloc == X86::reloc_absolute_dword)
|
|
MCE.emitDWordLE(0);
|
|
else
|
|
MCE.emitWordLE(0);
|
|
}
|
|
|
|
/// emitConstPoolAddress - Arrange for the address of an constant pool
|
|
/// to be emitted to the current location in the function, and allow it to be PC
|
|
/// relative.
|
|
template<class CodeEmitter>
|
|
void Emitter<CodeEmitter>::emitConstPoolAddress(unsigned CPI, unsigned Reloc,
|
|
intptr_t Disp /* = 0 */,
|
|
intptr_t PCAdj /* = 0 */) {
|
|
intptr_t RelocCST = 0;
|
|
if (Reloc == X86::reloc_picrel_word)
|
|
RelocCST = PICBaseOffset;
|
|
else if (Reloc == X86::reloc_pcrel_word)
|
|
RelocCST = PCAdj;
|
|
MCE.addRelocation(MachineRelocation::getConstPool(MCE.getCurrentPCOffset(),
|
|
Reloc, CPI, RelocCST));
|
|
// The relocated value will be added to the displacement
|
|
if (Reloc == X86::reloc_absolute_dword)
|
|
MCE.emitDWordLE(Disp);
|
|
else
|
|
MCE.emitWordLE((int32_t)Disp);
|
|
}
|
|
|
|
/// emitJumpTableAddress - Arrange for the address of a jump table to
|
|
/// be emitted to the current location in the function, and allow it to be PC
|
|
/// relative.
|
|
template<class CodeEmitter>
|
|
void Emitter<CodeEmitter>::emitJumpTableAddress(unsigned JTI, unsigned Reloc,
|
|
intptr_t PCAdj /* = 0 */) {
|
|
intptr_t RelocCST = 0;
|
|
if (Reloc == X86::reloc_picrel_word)
|
|
RelocCST = PICBaseOffset;
|
|
else if (Reloc == X86::reloc_pcrel_word)
|
|
RelocCST = PCAdj;
|
|
MCE.addRelocation(MachineRelocation::getJumpTable(MCE.getCurrentPCOffset(),
|
|
Reloc, JTI, RelocCST));
|
|
// The relocated value will be added to the displacement
|
|
if (Reloc == X86::reloc_absolute_dword)
|
|
MCE.emitDWordLE(0);
|
|
else
|
|
MCE.emitWordLE(0);
|
|
}
|
|
|
|
template<class CodeEmitter>
|
|
unsigned Emitter<CodeEmitter>::getX86RegNum(unsigned RegNo) const {
|
|
return II->getRegisterInfo().getX86RegNum(RegNo);
|
|
}
|
|
|
|
inline static unsigned char ModRMByte(unsigned Mod, unsigned RegOpcode,
|
|
unsigned RM) {
|
|
assert(Mod < 4 && RegOpcode < 8 && RM < 8 && "ModRM Fields out of range!");
|
|
return RM | (RegOpcode << 3) | (Mod << 6);
|
|
}
|
|
|
|
template<class CodeEmitter>
|
|
void Emitter<CodeEmitter>::emitRegModRMByte(unsigned ModRMReg,
|
|
unsigned RegOpcodeFld){
|
|
MCE.emitByte(ModRMByte(3, RegOpcodeFld, getX86RegNum(ModRMReg)));
|
|
}
|
|
|
|
template<class CodeEmitter>
|
|
void Emitter<CodeEmitter>::emitRegModRMByte(unsigned RegOpcodeFld) {
|
|
MCE.emitByte(ModRMByte(3, RegOpcodeFld, 0));
|
|
}
|
|
|
|
template<class CodeEmitter>
|
|
void Emitter<CodeEmitter>::emitSIBByte(unsigned SS,
|
|
unsigned Index,
|
|
unsigned Base) {
|
|
// SIB byte is in the same format as the ModRMByte...
|
|
MCE.emitByte(ModRMByte(SS, Index, Base));
|
|
}
|
|
|
|
template<class CodeEmitter>
|
|
void Emitter<CodeEmitter>::emitConstant(uint64_t Val, unsigned Size) {
|
|
// Output the constant in little endian byte order...
|
|
for (unsigned i = 0; i != Size; ++i) {
|
|
MCE.emitByte(Val & 255);
|
|
Val >>= 8;
|
|
}
|
|
}
|
|
|
|
/// isDisp8 - Return true if this signed displacement fits in a 8-bit
|
|
/// sign-extended field.
|
|
static bool isDisp8(int Value) {
|
|
return Value == (signed char)Value;
|
|
}
|
|
|
|
static bool gvNeedsNonLazyPtr(const MachineOperand &GVOp,
|
|
const TargetMachine &TM) {
|
|
// For Darwin-64, simulate the linktime GOT by using the same non-lazy-pointer
|
|
// mechanism as 32-bit mode.
|
|
if (TM.getSubtarget<X86Subtarget>().is64Bit() &&
|
|
!TM.getSubtarget<X86Subtarget>().isTargetDarwin())
|
|
return false;
|
|
|
|
// Return true if this is a reference to a stub containing the address of the
|
|
// global, not the global itself.
|
|
return isGlobalStubReference(GVOp.getTargetFlags());
|
|
}
|
|
|
|
template<class CodeEmitter>
|
|
void Emitter<CodeEmitter>::emitDisplacementField(const MachineOperand *RelocOp,
|
|
int DispVal, intptr_t PCAdj) {
|
|
// If this is a simple integer displacement that doesn't require a relocation,
|
|
// emit it now.
|
|
if (!RelocOp) {
|
|
emitConstant(DispVal, 4);
|
|
return;
|
|
}
|
|
|
|
// Otherwise, this is something that requires a relocation. Emit it as such
|
|
// now.
|
|
if (RelocOp->isGlobal()) {
|
|
// In 64-bit static small code model, we could potentially emit absolute.
|
|
// But it's probably not beneficial.
|
|
// 89 05 00 00 00 00 mov %eax,0(%rip) # PC-relative
|
|
// 89 04 25 00 00 00 00 mov %eax,0x0 # Absolute
|
|
unsigned rt = Is64BitMode ? X86::reloc_pcrel_word
|
|
: (IsPIC ? X86::reloc_picrel_word : X86::reloc_absolute_word);
|
|
bool NeedStub = isa<Function>(RelocOp->getGlobal());
|
|
bool Indirect = gvNeedsNonLazyPtr(*RelocOp, TM);
|
|
emitGlobalAddress(RelocOp->getGlobal(), rt, RelocOp->getOffset(),
|
|
PCAdj, NeedStub, Indirect);
|
|
} else if (RelocOp->isCPI()) {
|
|
unsigned rt = Is64BitMode ? X86::reloc_pcrel_word : X86::reloc_picrel_word;
|
|
emitConstPoolAddress(RelocOp->getIndex(), rt,
|
|
RelocOp->getOffset(), PCAdj);
|
|
} else if (RelocOp->isJTI()) {
|
|
unsigned rt = Is64BitMode ? X86::reloc_pcrel_word : X86::reloc_picrel_word;
|
|
emitJumpTableAddress(RelocOp->getIndex(), rt, PCAdj);
|
|
} else {
|
|
llvm_unreachable("Unknown value to relocate!");
|
|
}
|
|
}
|
|
|
|
template<class CodeEmitter>
|
|
void Emitter<CodeEmitter>::emitMemModRMByte(const MachineInstr &MI,
|
|
unsigned Op, unsigned RegOpcodeField,
|
|
intptr_t PCAdj) {
|
|
const MachineOperand &Op3 = MI.getOperand(Op+3);
|
|
int DispVal = 0;
|
|
const MachineOperand *DispForReloc = 0;
|
|
|
|
// Figure out what sort of displacement we have to handle here.
|
|
if (Op3.isGlobal()) {
|
|
DispForReloc = &Op3;
|
|
} else if (Op3.isCPI()) {
|
|
if (Is64BitMode || IsPIC) {
|
|
DispForReloc = &Op3;
|
|
} else {
|
|
DispVal += MCE.getConstantPoolEntryAddress(Op3.getIndex());
|
|
DispVal += Op3.getOffset();
|
|
}
|
|
} else if (Op3.isJTI()) {
|
|
if (Is64BitMode || IsPIC) {
|
|
DispForReloc = &Op3;
|
|
} else {
|
|
DispVal += MCE.getJumpTableEntryAddress(Op3.getIndex());
|
|
}
|
|
} else {
|
|
DispVal = Op3.getImm();
|
|
}
|
|
|
|
const MachineOperand &Base = MI.getOperand(Op);
|
|
const MachineOperand &Scale = MI.getOperand(Op+1);
|
|
const MachineOperand &IndexReg = MI.getOperand(Op+2);
|
|
|
|
unsigned BaseReg = Base.getReg();
|
|
|
|
// Is a SIB byte needed?
|
|
if ((!Is64BitMode || DispForReloc || BaseReg != 0) &&
|
|
IndexReg.getReg() == 0 &&
|
|
(BaseReg == 0 || BaseReg == X86::RIP ||
|
|
getX86RegNum(BaseReg) != N86::ESP)) {
|
|
if (BaseReg == 0 ||
|
|
BaseReg == X86::RIP) { // Just a displacement?
|
|
// Emit special case [disp32] encoding
|
|
MCE.emitByte(ModRMByte(0, RegOpcodeField, 5));
|
|
|
|
emitDisplacementField(DispForReloc, DispVal, PCAdj);
|
|
} else {
|
|
unsigned BaseRegNo = getX86RegNum(BaseReg);
|
|
if (!DispForReloc && DispVal == 0 && BaseRegNo != N86::EBP) {
|
|
// Emit simple indirect register encoding... [EAX] f.e.
|
|
MCE.emitByte(ModRMByte(0, RegOpcodeField, BaseRegNo));
|
|
} else if (!DispForReloc && isDisp8(DispVal)) {
|
|
// Emit the disp8 encoding... [REG+disp8]
|
|
MCE.emitByte(ModRMByte(1, RegOpcodeField, BaseRegNo));
|
|
emitConstant(DispVal, 1);
|
|
} else {
|
|
// Emit the most general non-SIB encoding: [REG+disp32]
|
|
MCE.emitByte(ModRMByte(2, RegOpcodeField, BaseRegNo));
|
|
emitDisplacementField(DispForReloc, DispVal, PCAdj);
|
|
}
|
|
}
|
|
|
|
} else { // We need a SIB byte, so start by outputting the ModR/M byte first
|
|
assert(IndexReg.getReg() != X86::ESP &&
|
|
IndexReg.getReg() != X86::RSP && "Cannot use ESP as index reg!");
|
|
|
|
bool ForceDisp32 = false;
|
|
bool ForceDisp8 = false;
|
|
if (BaseReg == 0) {
|
|
// If there is no base register, we emit the special case SIB byte with
|
|
// MOD=0, BASE=5, to JUST get the index, scale, and displacement.
|
|
MCE.emitByte(ModRMByte(0, RegOpcodeField, 4));
|
|
ForceDisp32 = true;
|
|
} else if (DispForReloc) {
|
|
// Emit the normal disp32 encoding.
|
|
MCE.emitByte(ModRMByte(2, RegOpcodeField, 4));
|
|
ForceDisp32 = true;
|
|
} else if (DispVal == 0 && getX86RegNum(BaseReg) != N86::EBP) {
|
|
// Emit no displacement ModR/M byte
|
|
MCE.emitByte(ModRMByte(0, RegOpcodeField, 4));
|
|
} else if (isDisp8(DispVal)) {
|
|
// Emit the disp8 encoding...
|
|
MCE.emitByte(ModRMByte(1, RegOpcodeField, 4));
|
|
ForceDisp8 = true; // Make sure to force 8 bit disp if Base=EBP
|
|
} else {
|
|
// Emit the normal disp32 encoding...
|
|
MCE.emitByte(ModRMByte(2, RegOpcodeField, 4));
|
|
}
|
|
|
|
// Calculate what the SS field value should be...
|
|
static const unsigned SSTable[] = { ~0, 0, 1, ~0, 2, ~0, ~0, ~0, 3 };
|
|
unsigned SS = SSTable[Scale.getImm()];
|
|
|
|
if (BaseReg == 0) {
|
|
// Handle the SIB byte for the case where there is no base. The
|
|
// displacement has already been output.
|
|
unsigned IndexRegNo;
|
|
if (IndexReg.getReg())
|
|
IndexRegNo = getX86RegNum(IndexReg.getReg());
|
|
else
|
|
IndexRegNo = 4; // For example [ESP+1*<noreg>+4]
|
|
emitSIBByte(SS, IndexRegNo, 5);
|
|
} else {
|
|
unsigned BaseRegNo = getX86RegNum(BaseReg);
|
|
unsigned IndexRegNo;
|
|
if (IndexReg.getReg())
|
|
IndexRegNo = getX86RegNum(IndexReg.getReg());
|
|
else
|
|
IndexRegNo = 4; // For example [ESP+1*<noreg>+4]
|
|
emitSIBByte(SS, IndexRegNo, BaseRegNo);
|
|
}
|
|
|
|
// Do we need to output a displacement?
|
|
if (ForceDisp8) {
|
|
emitConstant(DispVal, 1);
|
|
} else if (DispVal != 0 || ForceDisp32) {
|
|
emitDisplacementField(DispForReloc, DispVal, PCAdj);
|
|
}
|
|
}
|
|
}
|
|
|
|
template<class CodeEmitter>
|
|
void Emitter<CodeEmitter>::emitInstruction(
|
|
const MachineInstr &MI,
|
|
const TargetInstrDesc *Desc) {
|
|
DOUT << MI;
|
|
|
|
MCE.processDebugLoc(MI.getDebugLoc());
|
|
|
|
unsigned Opcode = Desc->Opcode;
|
|
|
|
// Emit the lock opcode prefix as needed.
|
|
if (Desc->TSFlags & X86II::LOCK) MCE.emitByte(0xF0);
|
|
|
|
// Emit segment override opcode prefix as needed.
|
|
switch (Desc->TSFlags & X86II::SegOvrMask) {
|
|
case X86II::FS:
|
|
MCE.emitByte(0x64);
|
|
break;
|
|
case X86II::GS:
|
|
MCE.emitByte(0x65);
|
|
break;
|
|
default: llvm_unreachable("Invalid segment!");
|
|
case 0: break; // No segment override!
|
|
}
|
|
|
|
// Emit the repeat opcode prefix as needed.
|
|
if ((Desc->TSFlags & X86II::Op0Mask) == X86II::REP) MCE.emitByte(0xF3);
|
|
|
|
// Emit the operand size opcode prefix as needed.
|
|
if (Desc->TSFlags & X86II::OpSize) MCE.emitByte(0x66);
|
|
|
|
// Emit the address size opcode prefix as needed.
|
|
if (Desc->TSFlags & X86II::AdSize) MCE.emitByte(0x67);
|
|
|
|
bool Need0FPrefix = false;
|
|
switch (Desc->TSFlags & X86II::Op0Mask) {
|
|
case X86II::TB: // Two-byte opcode prefix
|
|
case X86II::T8: // 0F 38
|
|
case X86II::TA: // 0F 3A
|
|
Need0FPrefix = true;
|
|
break;
|
|
case X86II::REP: break; // already handled.
|
|
case X86II::XS: // F3 0F
|
|
MCE.emitByte(0xF3);
|
|
Need0FPrefix = true;
|
|
break;
|
|
case X86II::XD: // F2 0F
|
|
MCE.emitByte(0xF2);
|
|
Need0FPrefix = true;
|
|
break;
|
|
case X86II::D8: case X86II::D9: case X86II::DA: case X86II::DB:
|
|
case X86II::DC: case X86II::DD: case X86II::DE: case X86II::DF:
|
|
MCE.emitByte(0xD8+
|
|
(((Desc->TSFlags & X86II::Op0Mask)-X86II::D8)
|
|
>> X86II::Op0Shift));
|
|
break; // Two-byte opcode prefix
|
|
default: llvm_unreachable("Invalid prefix!");
|
|
case 0: break; // No prefix!
|
|
}
|
|
|
|
if (Is64BitMode) {
|
|
// REX prefix
|
|
unsigned REX = X86InstrInfo::determineREX(MI);
|
|
if (REX)
|
|
MCE.emitByte(0x40 | REX);
|
|
}
|
|
|
|
// 0x0F escape code must be emitted just before the opcode.
|
|
if (Need0FPrefix)
|
|
MCE.emitByte(0x0F);
|
|
|
|
switch (Desc->TSFlags & X86II::Op0Mask) {
|
|
case X86II::T8: // 0F 38
|
|
MCE.emitByte(0x38);
|
|
break;
|
|
case X86II::TA: // 0F 3A
|
|
MCE.emitByte(0x3A);
|
|
break;
|
|
}
|
|
|
|
// If this is a two-address instruction, skip one of the register operands.
|
|
unsigned NumOps = Desc->getNumOperands();
|
|
unsigned CurOp = 0;
|
|
if (NumOps > 1 && Desc->getOperandConstraint(1, TOI::TIED_TO) != -1)
|
|
++CurOp;
|
|
else if (NumOps > 2 && Desc->getOperandConstraint(NumOps-1, TOI::TIED_TO)== 0)
|
|
// Skip the last source operand that is tied_to the dest reg. e.g. LXADD32
|
|
--NumOps;
|
|
|
|
unsigned char BaseOpcode = II->getBaseOpcodeFor(Desc);
|
|
switch (Desc->TSFlags & X86II::FormMask) {
|
|
default: llvm_unreachable("Unknown FormMask value in X86 MachineCodeEmitter!");
|
|
case X86II::Pseudo:
|
|
// Remember the current PC offset, this is the PIC relocation
|
|
// base address.
|
|
switch (Opcode) {
|
|
default:
|
|
llvm_unreachable("psuedo instructions should be removed before code emission");
|
|
break;
|
|
case TargetInstrInfo::INLINEASM: {
|
|
// We allow inline assembler nodes with empty bodies - they can
|
|
// implicitly define registers, which is ok for JIT.
|
|
if (MI.getOperand(0).getSymbolName()[0]) {
|
|
llvm_report_error("JIT does not support inline asm!");
|
|
}
|
|
break;
|
|
}
|
|
case TargetInstrInfo::DBG_LABEL:
|
|
case TargetInstrInfo::EH_LABEL:
|
|
MCE.emitLabel(MI.getOperand(0).getImm());
|
|
break;
|
|
case TargetInstrInfo::IMPLICIT_DEF:
|
|
case TargetInstrInfo::DECLARE:
|
|
case X86::DWARF_LOC:
|
|
case X86::FP_REG_KILL:
|
|
break;
|
|
case X86::MOVPC32r: {
|
|
// This emits the "call" portion of this pseudo instruction.
|
|
MCE.emitByte(BaseOpcode);
|
|
emitConstant(0, X86InstrInfo::sizeOfImm(Desc));
|
|
// Remember PIC base.
|
|
PICBaseOffset = (intptr_t) MCE.getCurrentPCOffset();
|
|
X86JITInfo *JTI = TM.getJITInfo();
|
|
JTI->setPICBase(MCE.getCurrentPCValue());
|
|
break;
|
|
}
|
|
}
|
|
CurOp = NumOps;
|
|
break;
|
|
case X86II::RawFrm:
|
|
MCE.emitByte(BaseOpcode);
|
|
|
|
if (CurOp != NumOps) {
|
|
const MachineOperand &MO = MI.getOperand(CurOp++);
|
|
|
|
DOUT << "RawFrm CurOp " << CurOp << "\n";
|
|
DOUT << "isMBB " << MO.isMBB() << "\n";
|
|
DOUT << "isGlobal " << MO.isGlobal() << "\n";
|
|
DOUT << "isSymbol " << MO.isSymbol() << "\n";
|
|
DOUT << "isImm " << MO.isImm() << "\n";
|
|
|
|
if (MO.isMBB()) {
|
|
emitPCRelativeBlockAddress(MO.getMBB());
|
|
} else if (MO.isGlobal()) {
|
|
// Assume undefined functions may be outside the Small codespace.
|
|
bool NeedStub =
|
|
(Is64BitMode &&
|
|
(TM.getCodeModel() == CodeModel::Large ||
|
|
TM.getSubtarget<X86Subtarget>().isTargetDarwin())) ||
|
|
Opcode == X86::TAILJMPd;
|
|
emitGlobalAddress(MO.getGlobal(), X86::reloc_pcrel_word,
|
|
MO.getOffset(), 0, NeedStub);
|
|
} else if (MO.isSymbol()) {
|
|
emitExternalSymbolAddress(MO.getSymbolName(), X86::reloc_pcrel_word);
|
|
} else if (MO.isImm()) {
|
|
if (Opcode == X86::CALLpcrel32 || Opcode == X86::CALL64pcrel32) {
|
|
// Fix up immediate operand for pc relative calls.
|
|
intptr_t Imm = (intptr_t)MO.getImm();
|
|
Imm = Imm - MCE.getCurrentPCValue() - 4;
|
|
emitConstant(Imm, X86InstrInfo::sizeOfImm(Desc));
|
|
} else
|
|
emitConstant(MO.getImm(), X86InstrInfo::sizeOfImm(Desc));
|
|
} else {
|
|
llvm_unreachable("Unknown RawFrm operand!");
|
|
}
|
|
}
|
|
break;
|
|
|
|
case X86II::AddRegFrm:
|
|
MCE.emitByte(BaseOpcode + getX86RegNum(MI.getOperand(CurOp++).getReg()));
|
|
|
|
if (CurOp != NumOps) {
|
|
const MachineOperand &MO1 = MI.getOperand(CurOp++);
|
|
unsigned Size = X86InstrInfo::sizeOfImm(Desc);
|
|
if (MO1.isImm())
|
|
emitConstant(MO1.getImm(), Size);
|
|
else {
|
|
unsigned rt = Is64BitMode ? X86::reloc_pcrel_word
|
|
: (IsPIC ? X86::reloc_picrel_word : X86::reloc_absolute_word);
|
|
if (Opcode == X86::MOV64ri64i32)
|
|
rt = X86::reloc_absolute_word; // FIXME: add X86II flag?
|
|
// This should not occur on Darwin for relocatable objects.
|
|
if (Opcode == X86::MOV64ri)
|
|
rt = X86::reloc_absolute_dword; // FIXME: add X86II flag?
|
|
if (MO1.isGlobal()) {
|
|
bool NeedStub = isa<Function>(MO1.getGlobal());
|
|
bool Indirect = gvNeedsNonLazyPtr(MO1, TM);
|
|
emitGlobalAddress(MO1.getGlobal(), rt, MO1.getOffset(), 0,
|
|
NeedStub, Indirect);
|
|
} else if (MO1.isSymbol())
|
|
emitExternalSymbolAddress(MO1.getSymbolName(), rt);
|
|
else if (MO1.isCPI())
|
|
emitConstPoolAddress(MO1.getIndex(), rt);
|
|
else if (MO1.isJTI())
|
|
emitJumpTableAddress(MO1.getIndex(), rt);
|
|
}
|
|
}
|
|
break;
|
|
|
|
case X86II::MRMDestReg: {
|
|
MCE.emitByte(BaseOpcode);
|
|
emitRegModRMByte(MI.getOperand(CurOp).getReg(),
|
|
getX86RegNum(MI.getOperand(CurOp+1).getReg()));
|
|
CurOp += 2;
|
|
if (CurOp != NumOps)
|
|
emitConstant(MI.getOperand(CurOp++).getImm(), X86InstrInfo::sizeOfImm(Desc));
|
|
break;
|
|
}
|
|
case X86II::MRMDestMem: {
|
|
MCE.emitByte(BaseOpcode);
|
|
emitMemModRMByte(MI, CurOp,
|
|
getX86RegNum(MI.getOperand(CurOp + X86AddrNumOperands)
|
|
.getReg()));
|
|
CurOp += X86AddrNumOperands + 1;
|
|
if (CurOp != NumOps)
|
|
emitConstant(MI.getOperand(CurOp++).getImm(), X86InstrInfo::sizeOfImm(Desc));
|
|
break;
|
|
}
|
|
|
|
case X86II::MRMSrcReg:
|
|
MCE.emitByte(BaseOpcode);
|
|
emitRegModRMByte(MI.getOperand(CurOp+1).getReg(),
|
|
getX86RegNum(MI.getOperand(CurOp).getReg()));
|
|
CurOp += 2;
|
|
if (CurOp != NumOps)
|
|
emitConstant(MI.getOperand(CurOp++).getImm(),
|
|
X86InstrInfo::sizeOfImm(Desc));
|
|
break;
|
|
|
|
case X86II::MRMSrcMem: {
|
|
// FIXME: Maybe lea should have its own form?
|
|
int AddrOperands;
|
|
if (Opcode == X86::LEA64r || Opcode == X86::LEA64_32r ||
|
|
Opcode == X86::LEA16r || Opcode == X86::LEA32r)
|
|
AddrOperands = X86AddrNumOperands - 1; // No segment register
|
|
else
|
|
AddrOperands = X86AddrNumOperands;
|
|
|
|
intptr_t PCAdj = (CurOp + AddrOperands + 1 != NumOps) ?
|
|
X86InstrInfo::sizeOfImm(Desc) : 0;
|
|
|
|
MCE.emitByte(BaseOpcode);
|
|
emitMemModRMByte(MI, CurOp+1, getX86RegNum(MI.getOperand(CurOp).getReg()),
|
|
PCAdj);
|
|
CurOp += AddrOperands + 1;
|
|
if (CurOp != NumOps)
|
|
emitConstant(MI.getOperand(CurOp++).getImm(),
|
|
X86InstrInfo::sizeOfImm(Desc));
|
|
break;
|
|
}
|
|
|
|
case X86II::MRM0r: case X86II::MRM1r:
|
|
case X86II::MRM2r: case X86II::MRM3r:
|
|
case X86II::MRM4r: case X86II::MRM5r:
|
|
case X86II::MRM6r: case X86II::MRM7r: {
|
|
MCE.emitByte(BaseOpcode);
|
|
|
|
// Special handling of lfence, mfence, monitor, and mwait.
|
|
if (Desc->getOpcode() == X86::LFENCE ||
|
|
Desc->getOpcode() == X86::MFENCE ||
|
|
Desc->getOpcode() == X86::MONITOR ||
|
|
Desc->getOpcode() == X86::MWAIT) {
|
|
emitRegModRMByte((Desc->TSFlags & X86II::FormMask)-X86II::MRM0r);
|
|
|
|
switch (Desc->getOpcode()) {
|
|
default: break;
|
|
case X86::MONITOR:
|
|
MCE.emitByte(0xC8);
|
|
break;
|
|
case X86::MWAIT:
|
|
MCE.emitByte(0xC9);
|
|
break;
|
|
}
|
|
} else {
|
|
emitRegModRMByte(MI.getOperand(CurOp++).getReg(),
|
|
(Desc->TSFlags & X86II::FormMask)-X86II::MRM0r);
|
|
}
|
|
|
|
if (CurOp != NumOps) {
|
|
const MachineOperand &MO1 = MI.getOperand(CurOp++);
|
|
unsigned Size = X86InstrInfo::sizeOfImm(Desc);
|
|
if (MO1.isImm())
|
|
emitConstant(MO1.getImm(), Size);
|
|
else {
|
|
unsigned rt = Is64BitMode ? X86::reloc_pcrel_word
|
|
: (IsPIC ? X86::reloc_picrel_word : X86::reloc_absolute_word);
|
|
if (Opcode == X86::MOV64ri32)
|
|
rt = X86::reloc_absolute_word; // FIXME: add X86II flag?
|
|
if (MO1.isGlobal()) {
|
|
bool NeedStub = isa<Function>(MO1.getGlobal());
|
|
bool Indirect = gvNeedsNonLazyPtr(MO1, TM);
|
|
emitGlobalAddress(MO1.getGlobal(), rt, MO1.getOffset(), 0,
|
|
NeedStub, Indirect);
|
|
} else if (MO1.isSymbol())
|
|
emitExternalSymbolAddress(MO1.getSymbolName(), rt);
|
|
else if (MO1.isCPI())
|
|
emitConstPoolAddress(MO1.getIndex(), rt);
|
|
else if (MO1.isJTI())
|
|
emitJumpTableAddress(MO1.getIndex(), rt);
|
|
}
|
|
}
|
|
break;
|
|
}
|
|
|
|
case X86II::MRM0m: case X86II::MRM1m:
|
|
case X86II::MRM2m: case X86II::MRM3m:
|
|
case X86II::MRM4m: case X86II::MRM5m:
|
|
case X86II::MRM6m: case X86II::MRM7m: {
|
|
intptr_t PCAdj = (CurOp + X86AddrNumOperands != NumOps) ?
|
|
(MI.getOperand(CurOp+X86AddrNumOperands).isImm() ?
|
|
X86InstrInfo::sizeOfImm(Desc) : 4) : 0;
|
|
|
|
MCE.emitByte(BaseOpcode);
|
|
emitMemModRMByte(MI, CurOp, (Desc->TSFlags & X86II::FormMask)-X86II::MRM0m,
|
|
PCAdj);
|
|
CurOp += X86AddrNumOperands;
|
|
|
|
if (CurOp != NumOps) {
|
|
const MachineOperand &MO = MI.getOperand(CurOp++);
|
|
unsigned Size = X86InstrInfo::sizeOfImm(Desc);
|
|
if (MO.isImm())
|
|
emitConstant(MO.getImm(), Size);
|
|
else {
|
|
unsigned rt = Is64BitMode ? X86::reloc_pcrel_word
|
|
: (IsPIC ? X86::reloc_picrel_word : X86::reloc_absolute_word);
|
|
if (Opcode == X86::MOV64mi32)
|
|
rt = X86::reloc_absolute_word; // FIXME: add X86II flag?
|
|
if (MO.isGlobal()) {
|
|
bool NeedStub = isa<Function>(MO.getGlobal());
|
|
bool Indirect = gvNeedsNonLazyPtr(MO, TM);
|
|
emitGlobalAddress(MO.getGlobal(), rt, MO.getOffset(), 0,
|
|
NeedStub, Indirect);
|
|
} else if (MO.isSymbol())
|
|
emitExternalSymbolAddress(MO.getSymbolName(), rt);
|
|
else if (MO.isCPI())
|
|
emitConstPoolAddress(MO.getIndex(), rt);
|
|
else if (MO.isJTI())
|
|
emitJumpTableAddress(MO.getIndex(), rt);
|
|
}
|
|
}
|
|
break;
|
|
}
|
|
|
|
case X86II::MRMInitReg:
|
|
MCE.emitByte(BaseOpcode);
|
|
// Duplicate register, used by things like MOV8r0 (aka xor reg,reg).
|
|
emitRegModRMByte(MI.getOperand(CurOp).getReg(),
|
|
getX86RegNum(MI.getOperand(CurOp).getReg()));
|
|
++CurOp;
|
|
break;
|
|
}
|
|
|
|
if (!Desc->isVariadic() && CurOp != NumOps) {
|
|
#ifndef NDEBUG
|
|
cerr << "Cannot encode: " << MI << "\n";
|
|
#endif
|
|
llvm_unreachable(0);
|
|
}
|
|
}
|