More style cleanup.
This commit is contained in:
@@ -7,8 +7,8 @@
|
||||
******************************************************************************
|
||||
*/
|
||||
|
||||
#ifndef XENIA_BACKEND_ASSEMBLER_H_
|
||||
#define XENIA_BACKEND_ASSEMBLER_H_
|
||||
#ifndef XENIA_CPU_BACKEND_ASSEMBLER_H_
|
||||
#define XENIA_CPU_BACKEND_ASSEMBLER_H_
|
||||
|
||||
#include <memory>
|
||||
|
||||
@@ -30,7 +30,7 @@ class Backend;
|
||||
|
||||
class Assembler {
|
||||
public:
|
||||
Assembler(Backend* backend);
|
||||
explicit Assembler(Backend* backend);
|
||||
virtual ~Assembler();
|
||||
|
||||
virtual bool Initialize();
|
||||
@@ -49,4 +49,4 @@ class Assembler {
|
||||
} // namespace cpu
|
||||
} // namespace xe
|
||||
|
||||
#endif // XENIA_BACKEND_ASSEMBLER_H_
|
||||
#endif // XENIA_CPU_BACKEND_ASSEMBLER_H_
|
||||
|
||||
@@ -7,8 +7,8 @@
|
||||
******************************************************************************
|
||||
*/
|
||||
|
||||
#ifndef XENIA_BACKEND_BACKEND_H_
|
||||
#define XENIA_BACKEND_BACKEND_H_
|
||||
#ifndef XENIA_CPU_BACKEND_BACKEND_H_
|
||||
#define XENIA_CPU_BACKEND_BACKEND_H_
|
||||
|
||||
#include <memory>
|
||||
|
||||
@@ -31,7 +31,7 @@ class CodeCache;
|
||||
|
||||
class Backend {
|
||||
public:
|
||||
Backend(Processor* processor);
|
||||
explicit Backend(Processor* processor);
|
||||
virtual ~Backend();
|
||||
|
||||
Processor* processor() const { return processor_; }
|
||||
@@ -61,4 +61,4 @@ class Backend {
|
||||
} // namespace cpu
|
||||
} // namespace xe
|
||||
|
||||
#endif // XENIA_BACKEND_BACKEND_H_
|
||||
#endif // XENIA_CPU_BACKEND_BACKEND_H_
|
||||
|
||||
@@ -7,8 +7,8 @@
|
||||
******************************************************************************
|
||||
*/
|
||||
|
||||
#ifndef XENIA_BACKEND_CODE_CACHE_H_
|
||||
#define XENIA_BACKEND_CODE_CACHE_H_
|
||||
#ifndef XENIA_CPU_BACKEND_CODE_CACHE_H_
|
||||
#define XENIA_CPU_BACKEND_CODE_CACHE_H_
|
||||
|
||||
#include <string>
|
||||
|
||||
@@ -39,4 +39,4 @@ class CodeCache {
|
||||
} // namespace cpu
|
||||
} // namespace xe
|
||||
|
||||
#endif // XENIA_BACKEND_CODE_CACHE_H_
|
||||
#endif // XENIA_CPU_BACKEND_CODE_CACHE_H_
|
||||
|
||||
@@ -7,8 +7,8 @@
|
||||
******************************************************************************
|
||||
*/
|
||||
|
||||
#ifndef XENIA_BACKEND_MACHINE_INFO_H_
|
||||
#define XENIA_BACKEND_MACHINE_INFO_H_
|
||||
#ifndef XENIA_CPU_BACKEND_MACHINE_INFO_H_
|
||||
#define XENIA_CPU_BACKEND_MACHINE_INFO_H_
|
||||
|
||||
#include <cstdint>
|
||||
|
||||
@@ -36,4 +36,4 @@ struct MachineInfo {
|
||||
} // namespace cpu
|
||||
} // namespace xe
|
||||
|
||||
#endif // XENIA_BACKEND_MACHINE_INFO_H_
|
||||
#endif // XENIA_CPU_BACKEND_MACHINE_INFO_H_
|
||||
|
||||
@@ -26,9 +26,6 @@ namespace cpu {
|
||||
namespace backend {
|
||||
namespace x64 {
|
||||
|
||||
// TODO(benvanik): remove when enums redefined.
|
||||
using namespace xe::cpu;
|
||||
|
||||
using xe::cpu::hir::HIRBuilder;
|
||||
|
||||
X64Assembler::X64Assembler(X64Backend* backend)
|
||||
@@ -78,7 +75,7 @@ bool X64Assembler::Assemble(GuestFunction* function, HIRBuilder* builder,
|
||||
void* machine_code = nullptr;
|
||||
size_t code_size = 0;
|
||||
if (!emitter_->Emit(function, builder, debug_info_flags, debug_info.get(),
|
||||
machine_code, code_size, function->source_map())) {
|
||||
&machine_code, &code_size, &function->source_map())) {
|
||||
return false;
|
||||
}
|
||||
|
||||
|
||||
@@ -7,8 +7,8 @@
|
||||
******************************************************************************
|
||||
*/
|
||||
|
||||
#ifndef XENIA_BACKEND_X64_X64_ASSEMBLER_H_
|
||||
#define XENIA_BACKEND_X64_X64_ASSEMBLER_H_
|
||||
#ifndef XENIA_CPU_BACKEND_X64_X64_ASSEMBLER_H_
|
||||
#define XENIA_CPU_BACKEND_X64_X64_ASSEMBLER_H_
|
||||
|
||||
#include <memory>
|
||||
#include <vector>
|
||||
@@ -28,7 +28,7 @@ class XbyakAllocator;
|
||||
|
||||
class X64Assembler : public Assembler {
|
||||
public:
|
||||
X64Assembler(X64Backend* backend);
|
||||
explicit X64Assembler(X64Backend* backend);
|
||||
~X64Assembler() override;
|
||||
|
||||
bool Initialize() override;
|
||||
@@ -58,4 +58,4 @@ class X64Assembler : public Assembler {
|
||||
} // namespace cpu
|
||||
} // namespace xe
|
||||
|
||||
#endif // XENIA_BACKEND_X64_X64_ASSEMBLER_H_
|
||||
#endif // XENIA_CPU_BACKEND_X64_X64_ASSEMBLER_H_
|
||||
|
||||
@@ -112,8 +112,6 @@ std::unique_ptr<GuestFunction> X64Backend::CreateGuestFunction(
|
||||
return std::make_unique<X64Function>(module, address);
|
||||
}
|
||||
|
||||
using namespace Xbyak;
|
||||
|
||||
X64ThunkEmitter::X64ThunkEmitter(X64Backend* backend, XbyakAllocator* allocator)
|
||||
: X64Emitter(backend, allocator) {}
|
||||
|
||||
|
||||
@@ -7,8 +7,8 @@
|
||||
******************************************************************************
|
||||
*/
|
||||
|
||||
#ifndef XENIA_BACKEND_X64_X64_BACKEND_H_
|
||||
#define XENIA_BACKEND_X64_X64_BACKEND_H_
|
||||
#ifndef XENIA_CPU_BACKEND_X64_X64_BACKEND_H_
|
||||
#define XENIA_CPU_BACKEND_X64_X64_BACKEND_H_
|
||||
|
||||
#include <gflags/gflags.h>
|
||||
|
||||
@@ -33,9 +33,9 @@ typedef void (*ResolveFunctionThunk)();
|
||||
|
||||
class X64Backend : public Backend {
|
||||
public:
|
||||
const static uint32_t kForceReturnAddress = 0x9FFF0000u;
|
||||
static const uint32_t kForceReturnAddress = 0x9FFF0000u;
|
||||
|
||||
X64Backend(Processor* processor);
|
||||
explicit X64Backend(Processor* processor);
|
||||
~X64Backend() override;
|
||||
|
||||
X64CodeCache* code_cache() const { return code_cache_.get(); }
|
||||
@@ -74,4 +74,4 @@ class X64Backend : public Backend {
|
||||
} // namespace cpu
|
||||
} // namespace xe
|
||||
|
||||
#endif // XENIA_BACKEND_X64_X64_BACKEND_H_
|
||||
#endif // XENIA_CPU_BACKEND_X64_X64_BACKEND_H_
|
||||
|
||||
@@ -7,13 +7,14 @@
|
||||
******************************************************************************
|
||||
*/
|
||||
|
||||
#ifndef XENIA_BACKEND_X64_X64_CODE_CACHE_H_
|
||||
#define XENIA_BACKEND_X64_X64_CODE_CACHE_H_
|
||||
#ifndef XENIA_CPU_BACKEND_X64_X64_CODE_CACHE_H_
|
||||
#define XENIA_CPU_BACKEND_X64_X64_CODE_CACHE_H_
|
||||
|
||||
#include <atomic>
|
||||
#include <memory>
|
||||
#include <mutex>
|
||||
#include <string>
|
||||
#include <utility>
|
||||
#include <vector>
|
||||
|
||||
#include "xenia/base/memory.h"
|
||||
@@ -58,18 +59,18 @@ class X64CodeCache : public CodeCache {
|
||||
protected:
|
||||
// All executable code falls within 0x80000000 to 0x9FFFFFFF, so we can
|
||||
// only map enough for lookups within that range.
|
||||
const static uint64_t kIndirectionTableBase = 0x80000000;
|
||||
const static uint64_t kIndirectionTableSize = 0x1FFFFFFF;
|
||||
static const uint64_t kIndirectionTableBase = 0x80000000;
|
||||
static const uint64_t kIndirectionTableSize = 0x1FFFFFFF;
|
||||
// The code range is 512MB, but we know the total code games will have is
|
||||
// pretty small (dozens of mb at most) and our expansion is reasonablish
|
||||
// so 256MB should be more than enough.
|
||||
const static uint64_t kGeneratedCodeBase = 0xA0000000;
|
||||
const static uint64_t kGeneratedCodeSize = 0x0FFFFFFF;
|
||||
static const uint64_t kGeneratedCodeBase = 0xA0000000;
|
||||
static const uint64_t kGeneratedCodeSize = 0x0FFFFFFF;
|
||||
|
||||
// This is picked to be high enough to cover whatever we can reasonably
|
||||
// expect. If we hit issues with this it probably means some corner case
|
||||
// in analysis triggering.
|
||||
const static size_t kMaximumFunctionCount = 30000;
|
||||
static const size_t kMaximumFunctionCount = 30000;
|
||||
|
||||
struct UnwindReservation {
|
||||
size_t data_size = 0;
|
||||
@@ -119,4 +120,4 @@ class X64CodeCache : public CodeCache {
|
||||
} // namespace cpu
|
||||
} // namespace xe
|
||||
|
||||
#endif // XENIA_BACKEND_X64_X64_CODE_CACHE_H_
|
||||
#endif // XENIA_CPU_BACKEND_X64_X64_CODE_CACHE_H_
|
||||
|
||||
@@ -31,7 +31,7 @@ namespace x64 {
|
||||
|
||||
// Size of unwind info per function.
|
||||
// TODO(benvanik): move this to emitter.
|
||||
const static uint32_t kUnwindInfoSize = 4 + (2 * 1 + 2 + 2);
|
||||
static const uint32_t kUnwindInfoSize = 4 + (2 * 1 + 2 + 2);
|
||||
|
||||
class Win32X64CodeCache : public X64CodeCache {
|
||||
public:
|
||||
|
||||
@@ -42,11 +42,6 @@ namespace cpu {
|
||||
namespace backend {
|
||||
namespace x64 {
|
||||
|
||||
// TODO(benvanik): remove when enums redefined.
|
||||
using namespace xe::cpu::hir;
|
||||
using namespace xe::cpu;
|
||||
|
||||
using namespace Xbyak;
|
||||
using xe::cpu::hir::HIRBuilder;
|
||||
using xe::cpu::hir::Instr;
|
||||
|
||||
@@ -56,7 +51,8 @@ static const size_t kStashOffset = 32;
|
||||
// static const size_t kStashOffsetHigh = 32 + 32;
|
||||
|
||||
const uint32_t X64Emitter::gpr_reg_map_[X64Emitter::GPR_COUNT] = {
|
||||
Operand::RBX, Operand::R12, Operand::R13, Operand::R14, Operand::R15,
|
||||
Xbyak::Operand::RBX, Xbyak::Operand::R12, Xbyak::Operand::R13,
|
||||
Xbyak::Operand::R14, Xbyak::Operand::R15,
|
||||
};
|
||||
|
||||
const uint32_t X64Emitter::xmm_reg_map_[X64Emitter::XMM_COUNT] = {
|
||||
@@ -64,7 +60,7 @@ const uint32_t X64Emitter::xmm_reg_map_[X64Emitter::XMM_COUNT] = {
|
||||
};
|
||||
|
||||
X64Emitter::X64Emitter(X64Backend* backend, XbyakAllocator* allocator)
|
||||
: CodeGenerator(kMaxCodeSize, AutoGrow, allocator),
|
||||
: CodeGenerator(kMaxCodeSize, Xbyak::AutoGrow, allocator),
|
||||
processor_(backend->processor()),
|
||||
backend_(backend),
|
||||
code_cache_(backend->code_cache()),
|
||||
@@ -90,8 +86,8 @@ X64Emitter::~X64Emitter() = default;
|
||||
|
||||
bool X64Emitter::Emit(GuestFunction* function, HIRBuilder* builder,
|
||||
uint32_t debug_info_flags, DebugInfo* debug_info,
|
||||
void*& out_code_address, size_t& out_code_size,
|
||||
std::vector<SourceMapEntry>& out_source_map) {
|
||||
void** out_code_address, size_t* out_code_size,
|
||||
std::vector<SourceMapEntry>* out_source_map) {
|
||||
SCOPE_profile_cpu_f("cpu");
|
||||
|
||||
// Reset.
|
||||
@@ -102,16 +98,16 @@ bool X64Emitter::Emit(GuestFunction* function, HIRBuilder* builder,
|
||||
|
||||
// Fill the generator with code.
|
||||
size_t stack_size = 0;
|
||||
if (!Emit(builder, stack_size)) {
|
||||
if (!Emit(builder, &stack_size)) {
|
||||
return false;
|
||||
}
|
||||
|
||||
// Copy the final code to the cache and relocate it.
|
||||
out_code_size = getSize();
|
||||
out_code_address = Emplace(stack_size, function);
|
||||
*out_code_size = getSize();
|
||||
*out_code_address = Emplace(stack_size, function);
|
||||
|
||||
// Stash source map.
|
||||
source_map_arena_.CloneContents(&out_source_map);
|
||||
source_map_arena_.CloneContents(out_source_map);
|
||||
|
||||
return true;
|
||||
}
|
||||
@@ -129,14 +125,14 @@ void* X64Emitter::Emplace(size_t stack_size, GuestFunction* function) {
|
||||
} else {
|
||||
new_address = code_cache_->PlaceHostCode(0, top_, size_, stack_size);
|
||||
}
|
||||
top_ = (uint8_t*)new_address;
|
||||
top_ = reinterpret_cast<uint8_t*>(new_address);
|
||||
ready();
|
||||
top_ = old_address;
|
||||
reset();
|
||||
return new_address;
|
||||
}
|
||||
|
||||
bool X64Emitter::Emit(HIRBuilder* builder, size_t& out_stack_size) {
|
||||
bool X64Emitter::Emit(HIRBuilder* builder, size_t* out_stack_size) {
|
||||
Xbyak::Label epilog_label;
|
||||
epilog_label_ = &epilog_label;
|
||||
|
||||
@@ -169,7 +165,7 @@ bool X64Emitter::Emit(HIRBuilder* builder, size_t& out_stack_size) {
|
||||
// Adding or changing anything here must be matched!
|
||||
const size_t stack_size = StackLayout::GUEST_STACK_SIZE + stack_offset;
|
||||
assert_true((stack_size + 8) % 16 == 0);
|
||||
out_stack_size = stack_size;
|
||||
*out_stack_size = stack_size;
|
||||
stack_size_ = stack_size;
|
||||
sub(rsp, (uint32_t)stack_size);
|
||||
mov(qword[rsp + StackLayout::GUEST_RCX_HOME], rcx);
|
||||
@@ -190,10 +186,10 @@ bool X64Emitter::Emit(HIRBuilder* builder, size_t& out_stack_size) {
|
||||
static_assert(debug::FunctionTraceData::kFunctionCallerHistoryCount == 4,
|
||||
"bitmask depends on count");
|
||||
mov(rax, qword[low_address(&trace_header->function_call_count)]);
|
||||
and_(rax, B00000011);
|
||||
and_(rax, 0b00000011);
|
||||
|
||||
// Record call history value into slot (guest addr in RDX).
|
||||
mov(dword[RegExp(uint32_t(uint64_t(
|
||||
mov(dword[Xbyak::RegExp(uint32_t(uint64_t(
|
||||
low_address(&trace_header->function_caller_history)))) +
|
||||
rax * 4],
|
||||
edx);
|
||||
@@ -221,7 +217,7 @@ bool X64Emitter::Emit(HIRBuilder* builder, size_t& out_stack_size) {
|
||||
const Instr* instr = block->instr_head;
|
||||
while (instr) {
|
||||
const Instr* new_tail = instr;
|
||||
if (!SelectSequence(*this, instr, &new_tail)) {
|
||||
if (!SelectSequence(this, instr, &new_tail)) {
|
||||
// No sequence found!
|
||||
assert_always();
|
||||
XELOGE("Unable to process HIR opcode %s", instr->opcode->name);
|
||||
@@ -368,7 +364,7 @@ void X64Emitter::Call(const hir::Instr* instr, GuestFunction* function) {
|
||||
}
|
||||
|
||||
// Actually jump/call to rax.
|
||||
if (instr->flags & CALL_TAIL) {
|
||||
if (instr->flags & hir::CALL_TAIL) {
|
||||
// Since we skip the prolog we need to mark the return here.
|
||||
EmitTraceUserCallReturn();
|
||||
|
||||
@@ -385,9 +381,10 @@ void X64Emitter::Call(const hir::Instr* instr, GuestFunction* function) {
|
||||
}
|
||||
}
|
||||
|
||||
void X64Emitter::CallIndirect(const hir::Instr* instr, const Reg64& reg) {
|
||||
void X64Emitter::CallIndirect(const hir::Instr* instr,
|
||||
const Xbyak::Reg64& reg) {
|
||||
// Check if return.
|
||||
if (instr->flags & CALL_POSSIBLE_RETURN) {
|
||||
if (instr->flags & hir::CALL_POSSIBLE_RETURN) {
|
||||
cmp(reg.cvt32(), dword[rsp + StackLayout::GUEST_RET_ADDR]);
|
||||
je(epilog_label(), CodeGenerator::T_NEAR);
|
||||
}
|
||||
@@ -401,7 +398,7 @@ void X64Emitter::CallIndirect(const hir::Instr* instr, const Reg64& reg) {
|
||||
mov(eax, dword[ebx]);
|
||||
|
||||
// Actually jump/call to rax.
|
||||
if (instr->flags & CALL_TAIL) {
|
||||
if (instr->flags & hir::CALL_TAIL) {
|
||||
// Since we skip the prolog we need to mark the return here.
|
||||
EmitTraceUserCallReturn();
|
||||
|
||||
@@ -550,7 +547,7 @@ bool X64Emitter::ConstantFitsIn32Reg(uint64_t v) {
|
||||
return false;
|
||||
}
|
||||
|
||||
void X64Emitter::MovMem64(const RegExp& addr, uint64_t v) {
|
||||
void X64Emitter::MovMem64(const Xbyak::RegExp& addr, uint64_t v) {
|
||||
if ((v & ~0x7FFFFFFF) == 0) {
|
||||
// Fits under 31 bits, so just load using normal mov.
|
||||
mov(qword[addr], v);
|
||||
@@ -651,7 +648,7 @@ uint32_t X64Emitter::PlaceData(Memory* memory) {
|
||||
return ptr;
|
||||
}
|
||||
|
||||
Address X64Emitter::GetXmmConstPtr(XmmConst id) {
|
||||
Xbyak::Address X64Emitter::GetXmmConstPtr(XmmConst id) {
|
||||
// Load through fixed constant table setup by PlaceData.
|
||||
return ptr[rdx + backend_->emitter_data() + sizeof(vec128_t) * id];
|
||||
}
|
||||
@@ -712,7 +709,7 @@ void X64Emitter::LoadConstantXmm(Xbyak::Xmm dest, double v) {
|
||||
}
|
||||
}
|
||||
|
||||
Address X64Emitter::StashXmm(int index, const Xmm& r) {
|
||||
Xbyak::Address X64Emitter::StashXmm(int index, const Xbyak::Xmm& r) {
|
||||
auto addr = ptr[rsp + kStashOffset + (index * 16)];
|
||||
vmovups(addr, r);
|
||||
return addr;
|
||||
|
||||
@@ -7,8 +7,8 @@
|
||||
******************************************************************************
|
||||
*/
|
||||
|
||||
#ifndef XENIA_BACKEND_X64_X64_EMITTER_H_
|
||||
#define XENIA_BACKEND_X64_X64_EMITTER_H_
|
||||
#ifndef XENIA_CPU_BACKEND_X64_X64_EMITTER_H_
|
||||
#define XENIA_CPU_BACKEND_X64_X64_EMITTER_H_
|
||||
|
||||
#include <vector>
|
||||
|
||||
@@ -116,8 +116,8 @@ class X64Emitter : public Xbyak::CodeGenerator {
|
||||
|
||||
bool Emit(GuestFunction* function, hir::HIRBuilder* builder,
|
||||
uint32_t debug_info_flags, DebugInfo* debug_info,
|
||||
void*& out_code_address, size_t& out_code_size,
|
||||
std::vector<SourceMapEntry>& out_source_map);
|
||||
void** out_code_address, size_t* out_code_size,
|
||||
std::vector<SourceMapEntry>* out_source_map);
|
||||
|
||||
static uint32_t PlaceData(Memory* memory);
|
||||
|
||||
@@ -196,7 +196,7 @@ class X64Emitter : public Xbyak::CodeGenerator {
|
||||
|
||||
protected:
|
||||
void* Emplace(size_t stack_size, GuestFunction* function = nullptr);
|
||||
bool Emit(hir::HIRBuilder* builder, size_t& out_stack_size);
|
||||
bool Emit(hir::HIRBuilder* builder, size_t* out_stack_size);
|
||||
void EmitGetCurrentThreadId();
|
||||
void EmitTraceUserCallReturn();
|
||||
|
||||
@@ -228,4 +228,4 @@ class X64Emitter : public Xbyak::CodeGenerator {
|
||||
} // namespace cpu
|
||||
} // namespace xe
|
||||
|
||||
#endif // XENIA_BACKEND_X64_X64_EMITTER_H_
|
||||
#endif // XENIA_CPU_BACKEND_X64_X64_EMITTER_H_
|
||||
|
||||
@@ -7,8 +7,8 @@
|
||||
******************************************************************************
|
||||
*/
|
||||
|
||||
#ifndef XENIA_BACKEND_X64_X64_FUNCTION_H_
|
||||
#define XENIA_BACKEND_X64_X64_FUNCTION_H_
|
||||
#ifndef XENIA_CPU_BACKEND_X64_X64_FUNCTION_H_
|
||||
#define XENIA_CPU_BACKEND_X64_X64_FUNCTION_H_
|
||||
|
||||
#include "xenia/cpu/function.h"
|
||||
#include "xenia/cpu/thread_state.h"
|
||||
@@ -41,4 +41,4 @@ class X64Function : public GuestFunction {
|
||||
} // namespace cpu
|
||||
} // namespace xe
|
||||
|
||||
#endif // XENIA_BACKEND_X64_X64_FUNCTION_H_
|
||||
#endif // XENIA_CPU_BACKEND_X64_X64_FUNCTION_H_
|
||||
|
||||
@@ -1477,16 +1477,16 @@ struct ROUND_F32 : Sequence<ROUND_F32, I<OPCODE_ROUND, F32Op, F32Op>> {
|
||||
static void Emit(X64Emitter& e, const EmitArgType& i) {
|
||||
switch (i.instr->flags) {
|
||||
case ROUND_TO_ZERO:
|
||||
e.vroundss(i.dest, i.src1, B00000011);
|
||||
e.vroundss(i.dest, i.src1, 0b00000011);
|
||||
break;
|
||||
case ROUND_TO_NEAREST:
|
||||
e.vroundss(i.dest, i.src1, B00000000);
|
||||
e.vroundss(i.dest, i.src1, 0b00000000);
|
||||
break;
|
||||
case ROUND_TO_MINUS_INFINITY:
|
||||
e.vroundss(i.dest, i.src1, B00000001);
|
||||
e.vroundss(i.dest, i.src1, 0b00000001);
|
||||
break;
|
||||
case ROUND_TO_POSITIVE_INFINITY:
|
||||
e.vroundss(i.dest, i.src1, B00000010);
|
||||
e.vroundss(i.dest, i.src1, 0b00000010);
|
||||
break;
|
||||
}
|
||||
}
|
||||
@@ -1495,16 +1495,16 @@ struct ROUND_F64 : Sequence<ROUND_F64, I<OPCODE_ROUND, F64Op, F64Op>> {
|
||||
static void Emit(X64Emitter& e, const EmitArgType& i) {
|
||||
switch (i.instr->flags) {
|
||||
case ROUND_TO_ZERO:
|
||||
e.vroundsd(i.dest, i.src1, B00000011);
|
||||
e.vroundsd(i.dest, i.src1, 0b00000011);
|
||||
break;
|
||||
case ROUND_TO_NEAREST:
|
||||
e.vroundsd(i.dest, i.src1, B00000000);
|
||||
e.vroundsd(i.dest, i.src1, 0b00000000);
|
||||
break;
|
||||
case ROUND_TO_MINUS_INFINITY:
|
||||
e.vroundsd(i.dest, i.src1, B00000001);
|
||||
e.vroundsd(i.dest, i.src1, 0b00000001);
|
||||
break;
|
||||
case ROUND_TO_POSITIVE_INFINITY:
|
||||
e.vroundsd(i.dest, i.src1, B00000010);
|
||||
e.vroundsd(i.dest, i.src1, 0b00000010);
|
||||
break;
|
||||
}
|
||||
}
|
||||
@@ -1513,16 +1513,16 @@ struct ROUND_V128 : Sequence<ROUND_V128, I<OPCODE_ROUND, V128Op, V128Op>> {
|
||||
static void Emit(X64Emitter& e, const EmitArgType& i) {
|
||||
switch (i.instr->flags) {
|
||||
case ROUND_TO_ZERO:
|
||||
e.vroundps(i.dest, i.src1, B00000011);
|
||||
e.vroundps(i.dest, i.src1, 0b00000011);
|
||||
break;
|
||||
case ROUND_TO_NEAREST:
|
||||
e.vroundps(i.dest, i.src1, B00000000);
|
||||
e.vroundps(i.dest, i.src1, 0b00000000);
|
||||
break;
|
||||
case ROUND_TO_MINUS_INFINITY:
|
||||
e.vroundps(i.dest, i.src1, B00000001);
|
||||
e.vroundps(i.dest, i.src1, 0b00000001);
|
||||
break;
|
||||
case ROUND_TO_POSITIVE_INFINITY:
|
||||
e.vroundps(i.dest, i.src1, B00000010);
|
||||
e.vroundps(i.dest, i.src1, 0b00000010);
|
||||
break;
|
||||
}
|
||||
}
|
||||
@@ -4791,7 +4791,7 @@ struct DOT_PRODUCT_3_V128
|
||||
[](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {
|
||||
// TODO(benvanik): apparently this is very slow
|
||||
// - find alternative?
|
||||
e.vdpps(dest, src1, src2, B01110001);
|
||||
e.vdpps(dest, src1, src2, 0b01110001);
|
||||
});
|
||||
}
|
||||
};
|
||||
@@ -4809,7 +4809,7 @@ struct DOT_PRODUCT_4_V128
|
||||
[](X64Emitter& e, Xmm dest, Xmm src1, Xmm src2) {
|
||||
// TODO(benvanik): apparently this is very slow
|
||||
// - find alternative?
|
||||
e.vdpps(dest, src1, src2, B11110001);
|
||||
e.vdpps(dest, src1, src2, 0b11110001);
|
||||
});
|
||||
}
|
||||
};
|
||||
@@ -6454,7 +6454,7 @@ struct PACK : Sequence<PACK, I<OPCODE_PACK, V128Op, V128Op, V128Op>> {
|
||||
|
||||
if (e.IsFeatureEnabled(kX64EmitF16C)) {
|
||||
// 0|0|0|0|W|Z|Y|X
|
||||
e.vcvtps2ph(i.dest, i.dest, B00000011);
|
||||
e.vcvtps2ph(i.dest, i.dest, 0b00000011);
|
||||
// Shuffle to X|Y|0|0|0|0|0|0
|
||||
e.vpshufb(i.dest, i.dest, e.GetXmmConstPtr(XMMPackFLOAT16_2));
|
||||
} else {
|
||||
@@ -6481,7 +6481,7 @@ struct PACK : Sequence<PACK, I<OPCODE_PACK, V128Op, V128Op, V128Op>> {
|
||||
|
||||
if (e.IsFeatureEnabled(kX64EmitF16C)) {
|
||||
// 0|0|0|0|W|Z|Y|X
|
||||
e.vcvtps2ph(i.dest, i.src1, B00000011);
|
||||
e.vcvtps2ph(i.dest, i.src1, 0b00000011);
|
||||
// Shuffle to X|Y|Z|W|0|0|0|0
|
||||
e.vpshufb(i.dest, i.dest, e.GetXmmConstPtr(XMMPackFLOAT16_4));
|
||||
} else {
|
||||
@@ -6639,8 +6639,8 @@ struct PACK : Sequence<PACK, I<OPCODE_PACK, V128Op, V128Op, V128Op>> {
|
||||
src2 = e.xmm0;
|
||||
}
|
||||
e.vpackusdw(i.dest, i.src1, src2);
|
||||
e.vpshuflw(i.dest, i.dest, B10110001);
|
||||
e.vpshufhw(i.dest, i.dest, B10110001);
|
||||
e.vpshuflw(i.dest, i.dest, 0b10110001);
|
||||
e.vpshufhw(i.dest, i.dest, 0b10110001);
|
||||
} else {
|
||||
// unsigned -> unsigned
|
||||
assert_always();
|
||||
@@ -7125,11 +7125,11 @@ void RegisterSequences() {
|
||||
Register_OPCODE_ATOMIC_EXCHANGE();
|
||||
}
|
||||
|
||||
bool SelectSequence(X64Emitter& e, const Instr* i, const Instr** new_tail) {
|
||||
bool SelectSequence(X64Emitter* e, const Instr* i, const Instr** new_tail) {
|
||||
const InstrKey key(i);
|
||||
auto it = sequence_table.find(key);
|
||||
if (it != sequence_table.end()) {
|
||||
if (it->second(e, i)) {
|
||||
if (it->second(*e, i)) {
|
||||
*new_tail = i->next;
|
||||
return true;
|
||||
}
|
||||
|
||||
@@ -7,16 +7,10 @@
|
||||
******************************************************************************
|
||||
*/
|
||||
|
||||
#ifndef XENIA_BACKEND_X64_X64_SEQUENCES_H_
|
||||
#define XENIA_BACKEND_X64_X64_SEQUENCES_H_
|
||||
#ifndef XENIA_CPU_BACKEND_X64_X64_SEQUENCES_H_
|
||||
#define XENIA_CPU_BACKEND_X64_X64_SEQUENCES_H_
|
||||
|
||||
namespace xe {
|
||||
namespace cpu {
|
||||
namespace hir {
|
||||
class Instr;
|
||||
} // namespace hir
|
||||
} // namespace cpu
|
||||
} // namespace xe
|
||||
#include "xenia/cpu/hir/instr.h"
|
||||
|
||||
namespace xe {
|
||||
namespace cpu {
|
||||
@@ -26,7 +20,7 @@ namespace x64 {
|
||||
class X64Emitter;
|
||||
|
||||
void RegisterSequences();
|
||||
bool SelectSequence(X64Emitter& e, const hir::Instr* i,
|
||||
bool SelectSequence(X64Emitter* e, const hir::Instr* i,
|
||||
const hir::Instr** new_tail);
|
||||
|
||||
} // namespace x64
|
||||
@@ -34,4 +28,4 @@ bool SelectSequence(X64Emitter& e, const hir::Instr* i,
|
||||
} // namespace cpu
|
||||
} // namespace xe
|
||||
|
||||
#endif // XENIA_BACKEND_X64_X64_SEQUENCES_H_
|
||||
#endif // XENIA_CPU_BACKEND_X64_X64_SEQUENCES_H_
|
||||
|
||||
@@ -7,8 +7,8 @@
|
||||
******************************************************************************
|
||||
*/
|
||||
|
||||
#ifndef XENIA_BACKEND_X64_X64_STACK_LAYOUT_H_
|
||||
#define XENIA_BACKEND_X64_X64_STACK_LAYOUT_H_
|
||||
#ifndef XENIA_CPU_BACKEND_X64_X64_STACK_LAYOUT_H_
|
||||
#define XENIA_CPU_BACKEND_X64_X64_STACK_LAYOUT_H_
|
||||
|
||||
#include "xenia/cpu/backend/x64/x64_backend.h"
|
||||
#include "xenia/cpu/backend/x64/x64_emitter.h"
|
||||
@@ -115,12 +115,12 @@ namespace x64 {
|
||||
|
||||
class StackLayout {
|
||||
public:
|
||||
const static size_t THUNK_STACK_SIZE = 120;
|
||||
static const size_t THUNK_STACK_SIZE = 120;
|
||||
|
||||
const static size_t GUEST_STACK_SIZE = 104;
|
||||
const static size_t GUEST_RCX_HOME = 80;
|
||||
const static size_t GUEST_RET_ADDR = 88;
|
||||
const static size_t GUEST_CALL_RET_ADDR = 96;
|
||||
static const size_t GUEST_STACK_SIZE = 104;
|
||||
static const size_t GUEST_RCX_HOME = 80;
|
||||
static const size_t GUEST_RET_ADDR = 88;
|
||||
static const size_t GUEST_CALL_RET_ADDR = 96;
|
||||
};
|
||||
|
||||
} // namespace x64
|
||||
@@ -128,4 +128,4 @@ class StackLayout {
|
||||
} // namespace cpu
|
||||
} // namespace xe
|
||||
|
||||
#endif // XENIA_BACKEND_X64_X64_STACK_LAYOUT_H_
|
||||
#endif // XENIA_CPU_BACKEND_X64_X64_STACK_LAYOUT_H_
|
||||
|
||||
@@ -58,41 +58,41 @@ uint32_t GetTracingMode() {
|
||||
}
|
||||
|
||||
void TraceString(void* raw_context, const char* str) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
IPRINT("XE[t] :%d: %s\n", thread_state->thread_id(), str);
|
||||
IFLUSH();
|
||||
}
|
||||
|
||||
void TraceContextLoadI8(void* raw_context, uint64_t offset, uint8_t value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("%d (%X) = ctx i8 +%llu\n", (int8_t)value, value, offset);
|
||||
}
|
||||
void TraceContextLoadI16(void* raw_context, uint64_t offset, uint16_t value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("%d (%X) = ctx i16 +%llu\n", (int16_t)value, value, offset);
|
||||
}
|
||||
void TraceContextLoadI32(void* raw_context, uint64_t offset, uint32_t value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("%d (%X) = ctx i32 +%llu\n", (int32_t)value, value, offset);
|
||||
}
|
||||
void TraceContextLoadI64(void* raw_context, uint64_t offset, uint64_t value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("%lld (%llX) = ctx i64 +%llu\n", (int64_t)value, value, offset);
|
||||
}
|
||||
void TraceContextLoadF32(void* raw_context, uint64_t offset, __m128 value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("%e (%X) = ctx f32 +%llu\n", xe::m128_f32<0>(value),
|
||||
xe::m128_i32<0>(value), offset);
|
||||
}
|
||||
void TraceContextLoadF64(void* raw_context, uint64_t offset,
|
||||
const double* value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
auto v = _mm_loadu_pd(value);
|
||||
DPRINT("%le (%llX) = ctx f64 +%llu\n", xe::m128_f64<0>(v), xe::m128_i64<0>(v),
|
||||
offset);
|
||||
}
|
||||
void TraceContextLoadV128(void* raw_context, uint64_t offset, __m128 value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("[%e, %e, %e, %e] [%.8X, %.8X, %.8X, %.8X] = ctx v128 +%llu\n",
|
||||
xe::m128_f32<0>(value), xe::m128_f32<1>(value), xe::m128_f32<2>(value),
|
||||
xe::m128_f32<3>(value), xe::m128_i32<0>(value), xe::m128_i32<1>(value),
|
||||
@@ -100,35 +100,35 @@ void TraceContextLoadV128(void* raw_context, uint64_t offset, __m128 value) {
|
||||
}
|
||||
|
||||
void TraceContextStoreI8(void* raw_context, uint64_t offset, uint8_t value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("ctx i8 +%llu = %d (%X)\n", offset, (int8_t)value, value);
|
||||
}
|
||||
void TraceContextStoreI16(void* raw_context, uint64_t offset, uint16_t value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("ctx i16 +%llu = %d (%X)\n", offset, (int16_t)value, value);
|
||||
}
|
||||
void TraceContextStoreI32(void* raw_context, uint64_t offset, uint32_t value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("ctx i32 +%llu = %d (%X)\n", offset, (int32_t)value, value);
|
||||
}
|
||||
void TraceContextStoreI64(void* raw_context, uint64_t offset, uint64_t value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("ctx i64 +%llu = %lld (%llX)\n", offset, (int64_t)value, value);
|
||||
}
|
||||
void TraceContextStoreF32(void* raw_context, uint64_t offset, __m128 value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("ctx f32 +%llu = %e (%X)\n", offset, xe::m128_f32<0>(value),
|
||||
xe::m128_i32<0>(value));
|
||||
}
|
||||
void TraceContextStoreF64(void* raw_context, uint64_t offset,
|
||||
const double* value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
auto v = _mm_loadu_pd(value);
|
||||
DPRINT("ctx f64 +%llu = %le (%llX)\n", offset, xe::m128_f64<0>(v),
|
||||
xe::m128_i64<0>(v));
|
||||
}
|
||||
void TraceContextStoreV128(void* raw_context, uint64_t offset, __m128 value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("ctx v128 +%llu = [%e, %e, %e, %e] [%.8X, %.8X, %.8X, %.8X]\n", offset,
|
||||
xe::m128_f32<0>(value), xe::m128_f32<1>(value), xe::m128_f32<2>(value),
|
||||
xe::m128_f32<3>(value), xe::m128_i32<0>(value), xe::m128_i32<1>(value),
|
||||
@@ -136,33 +136,33 @@ void TraceContextStoreV128(void* raw_context, uint64_t offset, __m128 value) {
|
||||
}
|
||||
|
||||
void TraceMemoryLoadI8(void* raw_context, uint32_t address, uint8_t value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("%d (%X) = load.i8 %.8X\n", (int8_t)value, value, address);
|
||||
}
|
||||
void TraceMemoryLoadI16(void* raw_context, uint32_t address, uint16_t value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("%d (%X) = load.i16 %.8X\n", (int16_t)value, value, address);
|
||||
}
|
||||
void TraceMemoryLoadI32(void* raw_context, uint32_t address, uint32_t value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("%d (%X) = load.i32 %.8X\n", (int32_t)value, value, address);
|
||||
}
|
||||
void TraceMemoryLoadI64(void* raw_context, uint32_t address, uint64_t value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("%lld (%llX) = load.i64 %.8X\n", (int64_t)value, value, address);
|
||||
}
|
||||
void TraceMemoryLoadF32(void* raw_context, uint32_t address, __m128 value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("%e (%X) = load.f32 %.8X\n", xe::m128_f32<0>(value),
|
||||
xe::m128_i32<0>(value), address);
|
||||
}
|
||||
void TraceMemoryLoadF64(void* raw_context, uint32_t address, __m128 value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("%le (%llX) = load.f64 %.8X\n", xe::m128_f64<0>(value),
|
||||
xe::m128_i64<0>(value), address);
|
||||
}
|
||||
void TraceMemoryLoadV128(void* raw_context, uint32_t address, __m128 value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("[%e, %e, %e, %e] [%.8X, %.8X, %.8X, %.8X] = load.v128 %.8X\n",
|
||||
xe::m128_f32<0>(value), xe::m128_f32<1>(value), xe::m128_f32<2>(value),
|
||||
xe::m128_f32<3>(value), xe::m128_i32<0>(value), xe::m128_i32<1>(value),
|
||||
@@ -170,33 +170,33 @@ void TraceMemoryLoadV128(void* raw_context, uint32_t address, __m128 value) {
|
||||
}
|
||||
|
||||
void TraceMemoryStoreI8(void* raw_context, uint32_t address, uint8_t value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("store.i8 %.8X = %d (%X)\n", address, (int8_t)value, value);
|
||||
}
|
||||
void TraceMemoryStoreI16(void* raw_context, uint32_t address, uint16_t value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("store.i16 %.8X = %d (%X)\n", address, (int16_t)value, value);
|
||||
}
|
||||
void TraceMemoryStoreI32(void* raw_context, uint32_t address, uint32_t value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("store.i32 %.8X = %d (%X)\n", address, (int32_t)value, value);
|
||||
}
|
||||
void TraceMemoryStoreI64(void* raw_context, uint32_t address, uint64_t value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("store.i64 %.8X = %lld (%llX)\n", address, (int64_t)value, value);
|
||||
}
|
||||
void TraceMemoryStoreF32(void* raw_context, uint32_t address, __m128 value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("store.f32 %.8X = %e (%X)\n", address, xe::m128_f32<0>(value),
|
||||
xe::m128_i32<0>(value));
|
||||
}
|
||||
void TraceMemoryStoreF64(void* raw_context, uint32_t address, __m128 value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("store.f64 %.8X = %le (%llX)\n", address, xe::m128_f64<0>(value),
|
||||
xe::m128_i64<0>(value));
|
||||
}
|
||||
void TraceMemoryStoreV128(void* raw_context, uint32_t address, __m128 value) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("store.v128 %.8X = [%e, %e, %e, %e] [%.8X, %.8X, %.8X, %.8X]\n",
|
||||
address, xe::m128_f32<0>(value), xe::m128_f32<1>(value),
|
||||
xe::m128_f32<2>(value), xe::m128_f32<3>(value), xe::m128_i32<0>(value),
|
||||
@@ -206,7 +206,7 @@ void TraceMemoryStoreV128(void* raw_context, uint32_t address, __m128 value) {
|
||||
|
||||
void TraceMemset(void* raw_context, uint32_t address, uint8_t value,
|
||||
uint32_t length) {
|
||||
auto thread_state = *((ThreadState**)raw_context);
|
||||
auto thread_state = *reinterpret_cast<ThreadState**>(raw_context);
|
||||
DPRINT("memset %.8X-%.8X (%d) = %.2X", address, address + length, length,
|
||||
value);
|
||||
}
|
||||
|
||||
@@ -7,8 +7,8 @@
|
||||
******************************************************************************
|
||||
*/
|
||||
|
||||
#ifndef XENIA_BACKEND_X64_X64_TRACERS_H_
|
||||
#define XENIA_BACKEND_X64_X64_TRACERS_H_
|
||||
#ifndef XENIA_CPU_BACKEND_X64_X64_TRACERS_H_
|
||||
#define XENIA_CPU_BACKEND_X64_X64_TRACERS_H_
|
||||
|
||||
#include <xmmintrin.h>
|
||||
#include <cstdint>
|
||||
@@ -72,4 +72,4 @@ void TraceMemset(void* raw_context, uint32_t address, uint8_t value,
|
||||
} // namespace cpu
|
||||
} // namespace xe
|
||||
|
||||
#endif // XENIA_BACKEND_X64_X64_TRACERS_H_
|
||||
#endif // XENIA_CPU_BACKEND_X64_X64_TRACERS_H_
|
||||
|
||||
Reference in New Issue
Block a user