forked from OSchip/llvm-project
[llvm-exegesis][NFC] Some code style cleanup
Apply review comments of https://reviews.llvm.org/D54185 to other target as well, specifically: 1. make anonymous namespaces as small as possible, avoid using static inside anonymous namespaces 2. Add missing header to some files 3. GetLoadImmediateOpcodem-> getLoadImmediateOpcode 4. Fix typo Differential Revision: https://reviews.llvm.org/D54343 llvm-svn: 347309
This commit is contained in:
parent
c9cc6cca42
commit
56c74cff70
|
@ -20,11 +20,9 @@
|
|||
namespace llvm {
|
||||
namespace exegesis {
|
||||
|
||||
namespace {
|
||||
|
||||
// Returns an error if we cannot handle the memory references in this
|
||||
// instruction.
|
||||
Error isInvalidMemoryInstr(const Instruction &Instr) {
|
||||
static Error isInvalidMemoryInstr(const Instruction &Instr) {
|
||||
switch (Instr.Description->TSFlags & X86II::FormMask) {
|
||||
default:
|
||||
llvm_unreachable("Unknown FormMask value");
|
||||
|
@ -169,78 +167,90 @@ static llvm::Error IsInvalidOpcode(const Instruction &Instr) {
|
|||
return llvm::Error::success();
|
||||
}
|
||||
|
||||
static unsigned GetX86FPFlags(const Instruction &Instr) {
|
||||
static unsigned getX86FPFlags(const Instruction &Instr) {
|
||||
return Instr.Description->TSFlags & llvm::X86II::FPTypeMask;
|
||||
}
|
||||
|
||||
namespace {
|
||||
class X86LatencySnippetGenerator : public LatencySnippetGenerator {
|
||||
public:
|
||||
using LatencySnippetGenerator::LatencySnippetGenerator;
|
||||
|
||||
llvm::Expected<std::vector<CodeTemplate>>
|
||||
generateCodeTemplates(const Instruction &Instr) const override {
|
||||
if (auto E = IsInvalidOpcode(Instr))
|
||||
return std::move(E);
|
||||
|
||||
switch (GetX86FPFlags(Instr)) {
|
||||
case llvm::X86II::NotFP:
|
||||
return LatencySnippetGenerator::generateCodeTemplates(Instr);
|
||||
case llvm::X86II::ZeroArgFP:
|
||||
case llvm::X86II::OneArgFP:
|
||||
case llvm::X86II::SpecialFP:
|
||||
case llvm::X86II::CompareFP:
|
||||
case llvm::X86II::CondMovFP:
|
||||
return llvm::make_error<BenchmarkFailure>("Unsupported x87 Instruction");
|
||||
case llvm::X86II::OneArgFPRW:
|
||||
case llvm::X86II::TwoArgFP:
|
||||
// These are instructions like
|
||||
// - `ST(0) = fsqrt(ST(0))` (OneArgFPRW)
|
||||
// - `ST(0) = ST(0) + ST(i)` (TwoArgFP)
|
||||
// They are intrinsically serial and do not modify the state of the stack.
|
||||
return generateSelfAliasingCodeTemplates(Instr);
|
||||
default:
|
||||
llvm_unreachable("Unknown FP Type!");
|
||||
}
|
||||
}
|
||||
generateCodeTemplates(const Instruction &Instr) const override;
|
||||
};
|
||||
} // namespace
|
||||
|
||||
llvm::Expected<std::vector<CodeTemplate>>
|
||||
X86LatencySnippetGenerator::generateCodeTemplates(
|
||||
const Instruction &Instr) const {
|
||||
if (auto E = IsInvalidOpcode(Instr))
|
||||
return std::move(E);
|
||||
|
||||
switch (getX86FPFlags(Instr)) {
|
||||
case llvm::X86II::NotFP:
|
||||
return LatencySnippetGenerator::generateCodeTemplates(Instr);
|
||||
case llvm::X86II::ZeroArgFP:
|
||||
case llvm::X86II::OneArgFP:
|
||||
case llvm::X86II::SpecialFP:
|
||||
case llvm::X86II::CompareFP:
|
||||
case llvm::X86II::CondMovFP:
|
||||
return llvm::make_error<BenchmarkFailure>("Unsupported x87 Instruction");
|
||||
case llvm::X86II::OneArgFPRW:
|
||||
case llvm::X86II::TwoArgFP:
|
||||
// These are instructions like
|
||||
// - `ST(0) = fsqrt(ST(0))` (OneArgFPRW)
|
||||
// - `ST(0) = ST(0) + ST(i)` (TwoArgFP)
|
||||
// They are intrinsically serial and do not modify the state of the stack.
|
||||
return generateSelfAliasingCodeTemplates(Instr);
|
||||
default:
|
||||
llvm_unreachable("Unknown FP Type!");
|
||||
}
|
||||
}
|
||||
|
||||
namespace {
|
||||
class X86UopsSnippetGenerator : public UopsSnippetGenerator {
|
||||
public:
|
||||
using UopsSnippetGenerator::UopsSnippetGenerator;
|
||||
|
||||
llvm::Expected<std::vector<CodeTemplate>>
|
||||
generateCodeTemplates(const Instruction &Instr) const override {
|
||||
if (auto E = IsInvalidOpcode(Instr))
|
||||
return std::move(E);
|
||||
|
||||
switch (GetX86FPFlags(Instr)) {
|
||||
case llvm::X86II::NotFP:
|
||||
return UopsSnippetGenerator::generateCodeTemplates(Instr);
|
||||
case llvm::X86II::ZeroArgFP:
|
||||
case llvm::X86II::OneArgFP:
|
||||
case llvm::X86II::SpecialFP:
|
||||
return llvm::make_error<BenchmarkFailure>("Unsupported x87 Instruction");
|
||||
case llvm::X86II::OneArgFPRW:
|
||||
case llvm::X86II::TwoArgFP:
|
||||
// These are instructions like
|
||||
// - `ST(0) = fsqrt(ST(0))` (OneArgFPRW)
|
||||
// - `ST(0) = ST(0) + ST(i)` (TwoArgFP)
|
||||
// They are intrinsically serial and do not modify the state of the stack.
|
||||
// We generate the same code for latency and uops.
|
||||
return generateSelfAliasingCodeTemplates(Instr);
|
||||
case llvm::X86II::CompareFP:
|
||||
case llvm::X86II::CondMovFP:
|
||||
// We can compute uops for any FP instruction that does not grow or shrink
|
||||
// the stack (either do not touch the stack or push as much as they pop).
|
||||
return generateUnconstrainedCodeTemplates(
|
||||
Instr, "instruction does not grow/shrink the FP stack");
|
||||
default:
|
||||
llvm_unreachable("Unknown FP Type!");
|
||||
}
|
||||
}
|
||||
generateCodeTemplates(const Instruction &Instr) const override;
|
||||
};
|
||||
} // namespace
|
||||
|
||||
static unsigned GetLoadImmediateOpcode(unsigned RegBitWidth) {
|
||||
llvm::Expected<std::vector<CodeTemplate>>
|
||||
X86UopsSnippetGenerator::generateCodeTemplates(
|
||||
const Instruction &Instr) const {
|
||||
if (auto E = IsInvalidOpcode(Instr))
|
||||
return std::move(E);
|
||||
|
||||
switch (getX86FPFlags(Instr)) {
|
||||
case llvm::X86II::NotFP:
|
||||
return UopsSnippetGenerator::generateCodeTemplates(Instr);
|
||||
case llvm::X86II::ZeroArgFP:
|
||||
case llvm::X86II::OneArgFP:
|
||||
case llvm::X86II::SpecialFP:
|
||||
return llvm::make_error<BenchmarkFailure>("Unsupported x87 Instruction");
|
||||
case llvm::X86II::OneArgFPRW:
|
||||
case llvm::X86II::TwoArgFP:
|
||||
// These are instructions like
|
||||
// - `ST(0) = fsqrt(ST(0))` (OneArgFPRW)
|
||||
// - `ST(0) = ST(0) + ST(i)` (TwoArgFP)
|
||||
// They are intrinsically serial and do not modify the state of the stack.
|
||||
// We generate the same code for latency and uops.
|
||||
return generateSelfAliasingCodeTemplates(Instr);
|
||||
case llvm::X86II::CompareFP:
|
||||
case llvm::X86II::CondMovFP:
|
||||
// We can compute uops for any FP instruction that does not grow or shrink
|
||||
// the stack (either do not touch the stack or push as much as they pop).
|
||||
return generateUnconstrainedCodeTemplates(
|
||||
Instr, "instruction does not grow/shrink the FP stack");
|
||||
default:
|
||||
llvm_unreachable("Unknown FP Type!");
|
||||
}
|
||||
}
|
||||
|
||||
static unsigned getLoadImmediateOpcode(unsigned RegBitWidth) {
|
||||
switch (RegBitWidth) {
|
||||
case 8:
|
||||
return llvm::X86::MOV8ri;
|
||||
|
@ -259,7 +269,7 @@ static llvm::MCInst loadImmediate(unsigned Reg, unsigned RegBitWidth,
|
|||
const llvm::APInt &Value) {
|
||||
if (Value.getBitWidth() > RegBitWidth)
|
||||
llvm_unreachable("Value must fit in the Register");
|
||||
return llvm::MCInstBuilder(GetLoadImmediateOpcode(RegBitWidth))
|
||||
return llvm::MCInstBuilder(getLoadImmediateOpcode(RegBitWidth))
|
||||
.addReg(Reg)
|
||||
.addImm(Value.getZExtValue());
|
||||
}
|
||||
|
@ -308,181 +318,123 @@ static llvm::MCInst releaseStackSpace(unsigned Bytes) {
|
|||
|
||||
// Reserves some space on the stack, fills it with the content of the provided
|
||||
// constant and provide methods to load the stack value into a register.
|
||||
namespace {
|
||||
struct ConstantInliner {
|
||||
explicit ConstantInliner(const llvm::APInt &Constant) : Constant_(Constant) {}
|
||||
|
||||
std::vector<llvm::MCInst> loadAndFinalize(unsigned Reg, unsigned RegBitWidth,
|
||||
unsigned Opcode) {
|
||||
assert((RegBitWidth & 7) == 0 &&
|
||||
"RegBitWidth must be a multiple of 8 bits");
|
||||
initStack(RegBitWidth / 8);
|
||||
add(loadToReg(Reg, Opcode));
|
||||
add(releaseStackSpace(RegBitWidth / 8));
|
||||
return std::move(Instructions);
|
||||
}
|
||||
unsigned Opcode);
|
||||
|
||||
std::vector<llvm::MCInst> loadX87STAndFinalize(unsigned Reg) {
|
||||
initStack(kF80Bytes);
|
||||
add(llvm::MCInstBuilder(llvm::X86::LD_F80m)
|
||||
// Address = ESP
|
||||
.addReg(llvm::X86::RSP) // BaseReg
|
||||
.addImm(1) // ScaleAmt
|
||||
.addReg(0) // IndexReg
|
||||
.addImm(0) // Disp
|
||||
.addReg(0)); // Segment
|
||||
if (Reg != llvm::X86::ST0)
|
||||
add(llvm::MCInstBuilder(llvm::X86::ST_Frr).addReg(Reg));
|
||||
add(releaseStackSpace(kF80Bytes));
|
||||
return std::move(Instructions);
|
||||
}
|
||||
std::vector<llvm::MCInst> loadX87STAndFinalize(unsigned Reg);
|
||||
|
||||
std::vector<llvm::MCInst> loadX87FPAndFinalize(unsigned Reg) {
|
||||
initStack(kF80Bytes);
|
||||
add(llvm::MCInstBuilder(llvm::X86::LD_Fp80m)
|
||||
.addReg(Reg)
|
||||
// Address = ESP
|
||||
.addReg(llvm::X86::RSP) // BaseReg
|
||||
.addImm(1) // ScaleAmt
|
||||
.addReg(0) // IndexReg
|
||||
.addImm(0) // Disp
|
||||
.addReg(0)); // Segment
|
||||
add(releaseStackSpace(kF80Bytes));
|
||||
return std::move(Instructions);
|
||||
}
|
||||
std::vector<llvm::MCInst> loadX87FPAndFinalize(unsigned Reg);
|
||||
|
||||
std::vector<llvm::MCInst> popFlagAndFinalize() {
|
||||
initStack(8);
|
||||
add(llvm::MCInstBuilder(llvm::X86::POPF64));
|
||||
return std::move(Instructions);
|
||||
}
|
||||
std::vector<llvm::MCInst> popFlagAndFinalize();
|
||||
|
||||
private:
|
||||
static constexpr const unsigned kF80Bytes = 10; // 80 bits.
|
||||
|
||||
ConstantInliner &add(const llvm::MCInst &Inst) {
|
||||
Instructions.push_back(Inst);
|
||||
return *this;
|
||||
}
|
||||
|
||||
void initStack(unsigned Bytes) {
|
||||
assert(Constant_.getBitWidth() <= Bytes * 8 &&
|
||||
"Value does not have the correct size");
|
||||
const llvm::APInt WideConstant = Constant_.getBitWidth() < Bytes * 8
|
||||
? Constant_.sext(Bytes * 8)
|
||||
: Constant_;
|
||||
add(allocateStackSpace(Bytes));
|
||||
size_t ByteOffset = 0;
|
||||
for (; Bytes - ByteOffset >= 4; ByteOffset += 4)
|
||||
add(fillStackSpace(
|
||||
llvm::X86::MOV32mi, ByteOffset,
|
||||
WideConstant.extractBits(32, ByteOffset * 8).getZExtValue()));
|
||||
if (Bytes - ByteOffset >= 2) {
|
||||
add(fillStackSpace(
|
||||
llvm::X86::MOV16mi, ByteOffset,
|
||||
WideConstant.extractBits(16, ByteOffset * 8).getZExtValue()));
|
||||
ByteOffset += 2;
|
||||
}
|
||||
if (Bytes - ByteOffset >= 1)
|
||||
add(fillStackSpace(
|
||||
llvm::X86::MOV8mi, ByteOffset,
|
||||
WideConstant.extractBits(8, ByteOffset * 8).getZExtValue()));
|
||||
}
|
||||
void initStack(unsigned Bytes);
|
||||
|
||||
static constexpr const unsigned kF80Bytes = 10; // 80 bits.
|
||||
|
||||
llvm::APInt Constant_;
|
||||
std::vector<llvm::MCInst> Instructions;
|
||||
};
|
||||
} // namespace
|
||||
|
||||
std::vector<llvm::MCInst> ConstantInliner::loadAndFinalize(unsigned Reg,
|
||||
unsigned RegBitWidth,
|
||||
unsigned Opcode) {
|
||||
assert((RegBitWidth & 7) == 0 && "RegBitWidth must be a multiple of 8 bits");
|
||||
initStack(RegBitWidth / 8);
|
||||
add(loadToReg(Reg, Opcode));
|
||||
add(releaseStackSpace(RegBitWidth / 8));
|
||||
return std::move(Instructions);
|
||||
}
|
||||
|
||||
std::vector<llvm::MCInst> ConstantInliner::loadX87STAndFinalize(unsigned Reg) {
|
||||
initStack(kF80Bytes);
|
||||
add(llvm::MCInstBuilder(llvm::X86::LD_F80m)
|
||||
// Address = ESP
|
||||
.addReg(llvm::X86::RSP) // BaseReg
|
||||
.addImm(1) // ScaleAmt
|
||||
.addReg(0) // IndexReg
|
||||
.addImm(0) // Disp
|
||||
.addReg(0)); // Segment
|
||||
if (Reg != llvm::X86::ST0)
|
||||
add(llvm::MCInstBuilder(llvm::X86::ST_Frr).addReg(Reg));
|
||||
add(releaseStackSpace(kF80Bytes));
|
||||
return std::move(Instructions);
|
||||
}
|
||||
|
||||
std::vector<llvm::MCInst> ConstantInliner::loadX87FPAndFinalize(unsigned Reg) {
|
||||
initStack(kF80Bytes);
|
||||
add(llvm::MCInstBuilder(llvm::X86::LD_Fp80m)
|
||||
.addReg(Reg)
|
||||
// Address = ESP
|
||||
.addReg(llvm::X86::RSP) // BaseReg
|
||||
.addImm(1) // ScaleAmt
|
||||
.addReg(0) // IndexReg
|
||||
.addImm(0) // Disp
|
||||
.addReg(0)); // Segment
|
||||
add(releaseStackSpace(kF80Bytes));
|
||||
return std::move(Instructions);
|
||||
}
|
||||
|
||||
std::vector<llvm::MCInst> ConstantInliner::popFlagAndFinalize() {
|
||||
initStack(8);
|
||||
add(llvm::MCInstBuilder(llvm::X86::POPF64));
|
||||
return std::move(Instructions);
|
||||
}
|
||||
|
||||
void ConstantInliner::initStack(unsigned Bytes) {
|
||||
assert(Constant_.getBitWidth() <= Bytes * 8 &&
|
||||
"Value does not have the correct size");
|
||||
const llvm::APInt WideConstant = Constant_.getBitWidth() < Bytes * 8
|
||||
? Constant_.sext(Bytes * 8)
|
||||
: Constant_;
|
||||
add(allocateStackSpace(Bytes));
|
||||
size_t ByteOffset = 0;
|
||||
for (; Bytes - ByteOffset >= 4; ByteOffset += 4)
|
||||
add(fillStackSpace(
|
||||
llvm::X86::MOV32mi, ByteOffset,
|
||||
WideConstant.extractBits(32, ByteOffset * 8).getZExtValue()));
|
||||
if (Bytes - ByteOffset >= 2) {
|
||||
add(fillStackSpace(
|
||||
llvm::X86::MOV16mi, ByteOffset,
|
||||
WideConstant.extractBits(16, ByteOffset * 8).getZExtValue()));
|
||||
ByteOffset += 2;
|
||||
}
|
||||
if (Bytes - ByteOffset >= 1)
|
||||
add(fillStackSpace(
|
||||
llvm::X86::MOV8mi, ByteOffset,
|
||||
WideConstant.extractBits(8, ByteOffset * 8).getZExtValue()));
|
||||
}
|
||||
|
||||
#include "X86GenExegesis.inc"
|
||||
|
||||
namespace {
|
||||
class ExegesisX86Target : public ExegesisTarget {
|
||||
public:
|
||||
ExegesisX86Target() : ExegesisTarget(X86CpuPfmCounters) {}
|
||||
|
||||
private:
|
||||
void addTargetSpecificPasses(llvm::PassManagerBase &PM) const override {
|
||||
// Lowers FP pseudo-instructions, e.g. ABS_Fp32 -> ABS_F.
|
||||
PM.add(llvm::createX86FloatingPointStackifierPass());
|
||||
}
|
||||
void addTargetSpecificPasses(llvm::PassManagerBase &PM) const override;
|
||||
|
||||
unsigned getScratchMemoryRegister(const llvm::Triple &TT) const override {
|
||||
if (!TT.isArch64Bit()) {
|
||||
// FIXME: This would require popping from the stack, so we would have to
|
||||
// add some additional setup code.
|
||||
return 0;
|
||||
}
|
||||
return TT.isOSWindows() ? llvm::X86::RCX : llvm::X86::RDI;
|
||||
}
|
||||
unsigned getScratchMemoryRegister(const llvm::Triple &TT) const override;
|
||||
|
||||
unsigned getMaxMemoryAccessSize() const override { return 64; }
|
||||
|
||||
void fillMemoryOperands(InstructionTemplate &IT, unsigned Reg,
|
||||
unsigned Offset) const override {
|
||||
assert(!isInvalidMemoryInstr(IT.Instr) &&
|
||||
"fillMemoryOperands requires a valid memory instruction");
|
||||
int MemOpIdx = X86II::getMemoryOperandNo(IT.Instr.Description->TSFlags);
|
||||
assert(MemOpIdx >= 0 && "invalid memory operand index");
|
||||
// getMemoryOperandNo() ignores tied operands, so we have to add them back.
|
||||
for (unsigned I = 0; I <= static_cast<unsigned>(MemOpIdx); ++I) {
|
||||
const auto &Op = IT.Instr.Operands[I];
|
||||
if (Op.isTied() && Op.getTiedToIndex() < I) {
|
||||
++MemOpIdx;
|
||||
}
|
||||
}
|
||||
// Now fill in the memory operands.
|
||||
const auto SetOp = [&IT](int OpIdx, const MCOperand &OpVal) {
|
||||
const auto Op = IT.Instr.Operands[OpIdx];
|
||||
assert(Op.isMemory() && Op.isExplicit() && "invalid memory pattern");
|
||||
IT.getValueFor(Op) = OpVal;
|
||||
};
|
||||
SetOp(MemOpIdx + 0, MCOperand::createReg(Reg)); // BaseReg
|
||||
SetOp(MemOpIdx + 1, MCOperand::createImm(1)); // ScaleAmt
|
||||
SetOp(MemOpIdx + 2, MCOperand::createReg(0)); // IndexReg
|
||||
SetOp(MemOpIdx + 3, MCOperand::createImm(Offset)); // Disp
|
||||
SetOp(MemOpIdx + 4, MCOperand::createReg(0)); // Segment
|
||||
}
|
||||
unsigned Offset) const override;
|
||||
|
||||
std::vector<llvm::MCInst> setRegTo(const llvm::MCSubtargetInfo &STI,
|
||||
unsigned Reg,
|
||||
const llvm::APInt &Value) const override {
|
||||
if (llvm::X86::GR8RegClass.contains(Reg))
|
||||
return {loadImmediate(Reg, 8, Value)};
|
||||
if (llvm::X86::GR16RegClass.contains(Reg))
|
||||
return {loadImmediate(Reg, 16, Value)};
|
||||
if (llvm::X86::GR32RegClass.contains(Reg))
|
||||
return {loadImmediate(Reg, 32, Value)};
|
||||
if (llvm::X86::GR64RegClass.contains(Reg))
|
||||
return {loadImmediate(Reg, 64, Value)};
|
||||
ConstantInliner CI(Value);
|
||||
if (llvm::X86::VR64RegClass.contains(Reg))
|
||||
return CI.loadAndFinalize(Reg, 64, llvm::X86::MMX_MOVQ64rm);
|
||||
if (llvm::X86::VR128XRegClass.contains(Reg)) {
|
||||
if (STI.getFeatureBits()[llvm::X86::FeatureAVX512])
|
||||
return CI.loadAndFinalize(Reg, 128, llvm::X86::VMOVDQU32Z128rm);
|
||||
if (STI.getFeatureBits()[llvm::X86::FeatureAVX])
|
||||
return CI.loadAndFinalize(Reg, 128, llvm::X86::VMOVDQUrm);
|
||||
return CI.loadAndFinalize(Reg, 128, llvm::X86::MOVDQUrm);
|
||||
}
|
||||
if (llvm::X86::VR256XRegClass.contains(Reg)) {
|
||||
if (STI.getFeatureBits()[llvm::X86::FeatureAVX512])
|
||||
return CI.loadAndFinalize(Reg, 256, llvm::X86::VMOVDQU32Z256rm);
|
||||
if (STI.getFeatureBits()[llvm::X86::FeatureAVX])
|
||||
return CI.loadAndFinalize(Reg, 256, llvm::X86::VMOVDQUYrm);
|
||||
}
|
||||
if (llvm::X86::VR512RegClass.contains(Reg))
|
||||
if (STI.getFeatureBits()[llvm::X86::FeatureAVX512])
|
||||
return CI.loadAndFinalize(Reg, 512, llvm::X86::VMOVDQU32Zrm);
|
||||
if (llvm::X86::RSTRegClass.contains(Reg)) {
|
||||
return CI.loadX87STAndFinalize(Reg);
|
||||
}
|
||||
if (llvm::X86::RFP32RegClass.contains(Reg) ||
|
||||
llvm::X86::RFP64RegClass.contains(Reg) ||
|
||||
llvm::X86::RFP80RegClass.contains(Reg)) {
|
||||
return CI.loadX87FPAndFinalize(Reg);
|
||||
}
|
||||
if (Reg == llvm::X86::EFLAGS)
|
||||
return CI.popFlagAndFinalize();
|
||||
return {}; // Not yet implemented.
|
||||
}
|
||||
const llvm::APInt &Value) const override;
|
||||
|
||||
std::unique_ptr<SnippetGenerator>
|
||||
createLatencySnippetGenerator(const LLVMState &State) const override {
|
||||
|
@ -498,9 +450,94 @@ private:
|
|||
return Arch == llvm::Triple::x86_64 || Arch == llvm::Triple::x86;
|
||||
}
|
||||
};
|
||||
|
||||
} // namespace
|
||||
|
||||
void ExegesisX86Target::addTargetSpecificPasses(
|
||||
llvm::PassManagerBase &PM) const {
|
||||
// Lowers FP pseudo-instructions, e.g. ABS_Fp32 -> ABS_F.
|
||||
PM.add(llvm::createX86FloatingPointStackifierPass());
|
||||
}
|
||||
|
||||
unsigned
|
||||
ExegesisX86Target::getScratchMemoryRegister(const llvm::Triple &TT) const {
|
||||
if (!TT.isArch64Bit()) {
|
||||
// FIXME: This would require popping from the stack, so we would have to
|
||||
// add some additional setup code.
|
||||
return 0;
|
||||
}
|
||||
return TT.isOSWindows() ? llvm::X86::RCX : llvm::X86::RDI;
|
||||
}
|
||||
|
||||
void ExegesisX86Target::fillMemoryOperands(InstructionTemplate &IT,
|
||||
unsigned Reg,
|
||||
unsigned Offset) const {
|
||||
assert(!isInvalidMemoryInstr(IT.Instr) &&
|
||||
"fillMemoryOperands requires a valid memory instruction");
|
||||
int MemOpIdx = X86II::getMemoryOperandNo(IT.Instr.Description->TSFlags);
|
||||
assert(MemOpIdx >= 0 && "invalid memory operand index");
|
||||
// getMemoryOperandNo() ignores tied operands, so we have to add them back.
|
||||
for (unsigned I = 0; I <= static_cast<unsigned>(MemOpIdx); ++I) {
|
||||
const auto &Op = IT.Instr.Operands[I];
|
||||
if (Op.isTied() && Op.getTiedToIndex() < I) {
|
||||
++MemOpIdx;
|
||||
}
|
||||
}
|
||||
// Now fill in the memory operands.
|
||||
const auto SetOp = [&IT](int OpIdx, const MCOperand &OpVal) {
|
||||
const auto Op = IT.Instr.Operands[OpIdx];
|
||||
assert(Op.isMemory() && Op.isExplicit() && "invalid memory pattern");
|
||||
IT.getValueFor(Op) = OpVal;
|
||||
};
|
||||
SetOp(MemOpIdx + 0, MCOperand::createReg(Reg)); // BaseReg
|
||||
SetOp(MemOpIdx + 1, MCOperand::createImm(1)); // ScaleAmt
|
||||
SetOp(MemOpIdx + 2, MCOperand::createReg(0)); // IndexReg
|
||||
SetOp(MemOpIdx + 3, MCOperand::createImm(Offset)); // Disp
|
||||
SetOp(MemOpIdx + 4, MCOperand::createReg(0)); // Segment
|
||||
}
|
||||
|
||||
std::vector<llvm::MCInst>
|
||||
ExegesisX86Target::setRegTo(const llvm::MCSubtargetInfo &STI, unsigned Reg,
|
||||
const llvm::APInt &Value) const {
|
||||
if (llvm::X86::GR8RegClass.contains(Reg))
|
||||
return {loadImmediate(Reg, 8, Value)};
|
||||
if (llvm::X86::GR16RegClass.contains(Reg))
|
||||
return {loadImmediate(Reg, 16, Value)};
|
||||
if (llvm::X86::GR32RegClass.contains(Reg))
|
||||
return {loadImmediate(Reg, 32, Value)};
|
||||
if (llvm::X86::GR64RegClass.contains(Reg))
|
||||
return {loadImmediate(Reg, 64, Value)};
|
||||
ConstantInliner CI(Value);
|
||||
if (llvm::X86::VR64RegClass.contains(Reg))
|
||||
return CI.loadAndFinalize(Reg, 64, llvm::X86::MMX_MOVQ64rm);
|
||||
if (llvm::X86::VR128XRegClass.contains(Reg)) {
|
||||
if (STI.getFeatureBits()[llvm::X86::FeatureAVX512])
|
||||
return CI.loadAndFinalize(Reg, 128, llvm::X86::VMOVDQU32Z128rm);
|
||||
if (STI.getFeatureBits()[llvm::X86::FeatureAVX])
|
||||
return CI.loadAndFinalize(Reg, 128, llvm::X86::VMOVDQUrm);
|
||||
return CI.loadAndFinalize(Reg, 128, llvm::X86::MOVDQUrm);
|
||||
}
|
||||
if (llvm::X86::VR256XRegClass.contains(Reg)) {
|
||||
if (STI.getFeatureBits()[llvm::X86::FeatureAVX512])
|
||||
return CI.loadAndFinalize(Reg, 256, llvm::X86::VMOVDQU32Z256rm);
|
||||
if (STI.getFeatureBits()[llvm::X86::FeatureAVX])
|
||||
return CI.loadAndFinalize(Reg, 256, llvm::X86::VMOVDQUYrm);
|
||||
}
|
||||
if (llvm::X86::VR512RegClass.contains(Reg))
|
||||
if (STI.getFeatureBits()[llvm::X86::FeatureAVX512])
|
||||
return CI.loadAndFinalize(Reg, 512, llvm::X86::VMOVDQU32Zrm);
|
||||
if (llvm::X86::RSTRegClass.contains(Reg)) {
|
||||
return CI.loadX87STAndFinalize(Reg);
|
||||
}
|
||||
if (llvm::X86::RFP32RegClass.contains(Reg) ||
|
||||
llvm::X86::RFP64RegClass.contains(Reg) ||
|
||||
llvm::X86::RFP80RegClass.contains(Reg)) {
|
||||
return CI.loadX87FPAndFinalize(Reg);
|
||||
}
|
||||
if (Reg == llvm::X86::EFLAGS)
|
||||
return CI.popFlagAndFinalize();
|
||||
return {}; // Not yet implemented.
|
||||
}
|
||||
|
||||
static ExegesisTarget *getTheExegesisX86Target() {
|
||||
static ExegesisX86Target Target;
|
||||
return &Target;
|
||||
|
|
|
@ -1,3 +1,12 @@
|
|||
//===-- TargetTest.cpp ------------------------------------------*- C++ -*-===//
|
||||
//
|
||||
// The LLVM Compiler Infrastructure
|
||||
//
|
||||
// This file is distributed under the University of Illinois Open Source
|
||||
// License. See LICENSE.TXT for details.
|
||||
//
|
||||
//===----------------------------------------------------------------------===//
|
||||
|
||||
#include "Target.h"
|
||||
|
||||
#include <cassert>
|
||||
|
|
|
@ -1,3 +1,12 @@
|
|||
//===-- AnalysisTest.cpp ---------------------------------------*- C++ -*-===//
|
||||
//
|
||||
// The LLVM Compiler Infrastructure
|
||||
//
|
||||
// This file is distributed under the University of Illinois Open Source
|
||||
// License. See LICENSE.TXT for details.
|
||||
//
|
||||
//===----------------------------------------------------------------------===//
|
||||
|
||||
#include "Analysis.h"
|
||||
|
||||
#include <cassert>
|
||||
|
@ -28,7 +37,7 @@ protected:
|
|||
}
|
||||
STI.reset(TheTarget->createMCSubtargetInfo(TT, "haswell", ""));
|
||||
|
||||
// Compute the ProxResIdx of ports unes in tests.
|
||||
// Compute the ProxResIdx of ports uses in tests.
|
||||
const auto &SM = STI->getSchedModel();
|
||||
for (unsigned I = 0, E = SM.getNumProcResourceKinds(); I < E; ++I) {
|
||||
const std::string Name = SM.getProcResource(I)->Name;
|
||||
|
|
|
@ -1,3 +1,13 @@
|
|||
//===-- RegisterAliasingTest.cpp --------------------------------*- C++ -*-===//
|
||||
//
|
||||
// The LLVM Compiler Infrastructure
|
||||
//
|
||||
// This file is distributed under the University of Illinois Open Source
|
||||
// License. See LICENSE.TXT for details.
|
||||
//
|
||||
//===----------------------------------------------------------------------===//
|
||||
|
||||
|
||||
#include "RegisterAliasing.h"
|
||||
|
||||
#include <cassert>
|
||||
|
|
|
@ -1,3 +1,12 @@
|
|||
//===-- TargetTest.cpp -----------------------------------------*- C++ -*-===//
|
||||
//
|
||||
// The LLVM Compiler Infrastructure
|
||||
//
|
||||
// This file is distributed under the University of Illinois Open Source
|
||||
// License. See LICENSE.TXT for details.
|
||||
//
|
||||
//===----------------------------------------------------------------------===//
|
||||
|
||||
#include "Target.h"
|
||||
|
||||
#include <cassert>
|
||||
|
|
Loading…
Reference in New Issue