PoS_EVM
/
mcl
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

update xbyak v5.941

Browse Source
update-fork
MITSUNARI Shigeo 4 years ago
parent 49e6633eaa
commit 9b86774989
3 changed files with 319 additions and 240 deletions
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Show Stats Download Patch File Download Diff File
  1. 377
      src/xbyak/xbyak.h
  2. 156
      src/xbyak/xbyak_mnemonic.h
  3. 26
      src/xbyak/xbyak_util.h

377
src/xbyak/xbyak.h
Unescape View File

@ -72,13 +72,17 @@
#define XBYAK_STD_UNORDERED_MULTIMAP std::multimap #define XBYAK_STD_UNORDERED_MULTIMAP std::multimap
#endif #endif
#ifdef _WIN32 #ifdef _WIN32
#include <winsock2.h> #ifndef WIN32_LEAN_AND_MEAN
#define WIN32_LEAN_AND_MEAN
#endif
#include <windows.h> #include <windows.h>
#include <malloc.h> #include <malloc.h>
#define XBYAK_TLS __declspec(thread)
#elif defined(__GNUC__) #elif defined(__GNUC__)
#include <unistd.h> #include <unistd.h>
#include <sys/mman.h> #include <sys/mman.h>
#include <stdlib.h> #include <stdlib.h>
#define XBYAK_TLS __thread
#endif #endif
#if defined(__APPLE__) && !defined(XBYAK_DONT_USE_MAP_JIT) #if defined(__APPLE__) && !defined(XBYAK_DONT_USE_MAP_JIT)
#define XBYAK_USE_MAP_JIT #define XBYAK_USE_MAP_JIT
@ -120,7 +124,7 @@ namespace Xbyak {
enum { enum {
DEFAULT_MAX_CODE_SIZE = 4096, DEFAULT_MAX_CODE_SIZE = 4096,
VERSION = 0x5912 /* 0xABCD = A.BC(D) */ VERSION = 0x5941 /* 0xABCD = A.BC(D) */
}; };
#ifndef MIE_INTEGER_TYPE_DEFINED #ifndef MIE_INTEGER_TYPE_DEFINED
@ -194,9 +198,80 @@ enum {
ERR_INVALID_RIP_IN_AUTO_GROW, ERR_INVALID_RIP_IN_AUTO_GROW,
ERR_INVALID_MIB_ADDRESS, ERR_INVALID_MIB_ADDRESS,
ERR_X2APIC_IS_NOT_SUPPORTED, ERR_X2APIC_IS_NOT_SUPPORTED,
ERR_NOT_SUPPORTED,
ERR_INTERNAL // Put it at last. ERR_INTERNAL // Put it at last.
}; };
inline const char *ConvertErrorToString(int err)
{
static const char *errTbl[] = {
"none",
"bad addressing",
"code is too big",
"bad scale",
"esp can't be index",
"bad combination",
"bad size of register",
"imm is too big",
"bad align",
"label is redefined",
"label is too far",
"label is not found",
"code is not copyable",
"bad parameter",
"can't protect",
"can't use 64bit disp(use (void*))",
"offset is too big",
"MEM size is not specified",
"bad mem size",
"bad st combination",
"over local label",
"under local label",
"can't alloc",
"T_SHORT is not supported in AutoGrow",
"bad protect mode",
"bad pNum",
"bad tNum",
"bad vsib addressing",
"can't convert",
"label is not set by L()",
"label is already set by L()",
"bad label string",
"err munmap",
"opmask is already set",
"rounding is already set",
"k0 is invalid",
"evex is invalid",
"sae(suppress all exceptions) is invalid",
"er(embedded rounding) is invalid",
"invalid broadcast",
"invalid opmask with memory",
"invalid zero",
"invalid rip in AutoGrow",
"invalid mib address",
"x2APIC is not supported",
"not supported",
"internal error"
};
assert(ERR_INTERNAL + 1 == sizeof(errTbl) / sizeof(*errTbl));
return err <= ERR_INTERNAL ? errTbl[err] : "unknown err";
}
#ifdef XBYAK_NO_EXCEPTION
namespace local {
static XBYAK_TLS int l_err = 0;
inline void SetError(int err) { if (err) l_err = err; } // keep the first err code
} // local
inline void ClearError() { local::l_err = 0; }
inline int GetError() { return local::l_err; }
#define XBYAK_THROW(err) { local::SetError(err); return; }
#define XBYAK_THROW_RET(err, r) { local::SetError(err); return r; }
#else
class Error : public std::exception { class Error : public std::exception {
int err_; int err_;
public: public:
@ -209,65 +284,24 @@ public:
operator int() const { return err_; } operator int() const { return err_; }
const char *what() const throw() const char *what() const throw()
{ {
static const char *errTbl[] = { return ConvertErrorToString(err_);
"none",
"bad addressing",
"code is too big",
"bad scale",
"esp can't be index",
"bad combination",
"bad size of register",
"imm is too big",
"bad align",
"label is redefined",
"label is too far",
"label is not found",
"code is not copyable",
"bad parameter",
"can't protect",
"can't use 64bit disp(use (void*))",
"offset is too big",
"MEM size is not specified",
"bad mem size",
"bad st combination",
"over local label",
"under local label",
"can't alloc",
"T_SHORT is not supported in AutoGrow",
"bad protect mode",
"bad pNum",
"bad tNum",
"bad vsib addressing",
"can't convert",
"label is not set by L()",
"label is already set by L()",
"bad label string",
"err munmap",
"opmask is already set",
"rounding is already set",
"k0 is invalid",
"evex is invalid",
"sae(suppress all exceptions) is invalid",
"er(embedded rounding) is invalid",
"invalid broadcast",
"invalid opmask with memory",
"invalid zero",
"invalid rip in AutoGrow",
"invalid mib address",
"x2APIC is not supported",
"internal error"
};
assert(err_ <= ERR_INTERNAL);
assert(ERR_INTERNAL + 1 == sizeof(errTbl) / sizeof(*errTbl));
return errTbl[err_];
} }
}; };
// dummy functions
inline void ClearError() { }
inline int GetError() { return 0; }
inline const char *ConvertErrorToString(const Error& err) inline const char *ConvertErrorToString(const Error& err)
{ {
return err.what(); return err.what();
} }
#define XBYAK_THROW(err) { throw Error(err); }
#define XBYAK_THROW_RET(err, r) { throw Error(err); }
#endif
inline void *AlignedMalloc(size_t size, size_t alignment) inline void *AlignedMalloc(size_t size, size_t alignment)
{ {
#ifdef __MINGW32__ #ifdef __MINGW32__
@ -307,7 +341,7 @@ inline bool IsInInt32(uint64 x) { return ~uint64(0x7fffffffu) <= x || x <= 0x7FF
inline uint32 VerifyInInt32(uint64 x) inline uint32 VerifyInInt32(uint64 x)
{ {
#ifdef XBYAK64 #ifdef XBYAK64
if (!IsInInt32(x)) throw Error(ERR_OFFSET_IS_TOO_BIG); if (!IsInInt32(x)) XBYAK_THROW_RET(ERR_OFFSET_IS_TOO_BIG, 0)
#endif #endif
return static_cast<uint32>(x); return static_cast<uint32>(x);
} }
@ -375,7 +409,7 @@ public:
#error "not supported" #error "not supported"
#endif #endif
void *p = mmap(NULL, size, PROT_READ | PROT_WRITE, mode, -1, 0); void *p = mmap(NULL, size, PROT_READ | PROT_WRITE, mode, -1, 0);
if (p == MAP_FAILED) throw Error(ERR_CANT_ALLOC); if (p == MAP_FAILED) XBYAK_THROW_RET(ERR_CANT_ALLOC, 0)
assert(p); assert(p);
sizeList_[(uintptr_t)p] = size; sizeList_[(uintptr_t)p] = size;
return (uint8*)p; return (uint8*)p;
@ -384,8 +418,8 @@ public:
{ {
if (p == 0) return; if (p == 0) return;
SizeList::iterator i = sizeList_.find((uintptr_t)p); SizeList::iterator i = sizeList_.find((uintptr_t)p);
if (i == sizeList_.end()) throw Error(ERR_BAD_PARAMETER); if (i == sizeList_.end()) XBYAK_THROW(ERR_BAD_PARAMETER)
if (munmap((void*)i->first, i->second) < 0) throw Error(ERR_MUNMAP); if (munmap((void*)i->first, i->second) < 0) XBYAK_THROW(ERR_MUNMAP)
sizeList_.erase(i); sizeList_.erase(i);
} }
}; };
@ -397,8 +431,8 @@ class Reg;
class Operand { class Operand {
static const uint8 EXT8BIT = 0x20; static const uint8 EXT8BIT = 0x20;
unsigned int idx_:6; // 0..31 + EXT8BIT = 1 if spl/bpl/sil/dil unsigned int idx_:6; // 0..31 + EXT8BIT = 1 if spl/bpl/sil/dil
unsigned int kind_:9; unsigned int kind_:10;
unsigned int bit_:10; unsigned int bit_:14;
protected: protected:
unsigned int zero_:1; unsigned int zero_:1;
unsigned int mask_:3; unsigned int mask_:3;
@ -415,7 +449,8 @@ public:
YMM = 1 << 5, YMM = 1 << 5,
ZMM = 1 << 6, ZMM = 1 << 6,
OPMASK = 1 << 7, OPMASK = 1 << 7,
BNDREG = 1 << 8 BNDREG = 1 << 8,
TMM = 1 << 9
}; };
enum Code { enum Code {
#ifdef XBYAK64 #ifdef XBYAK64
@ -445,6 +480,7 @@ public:
bool isXMM() const { return is(XMM); } bool isXMM() const { return is(XMM); }
bool isYMM() const { return is(YMM); } bool isYMM() const { return is(YMM); }
bool isZMM() const { return is(ZMM); } bool isZMM() const { return is(ZMM); }
bool isTMM() const { return is(TMM); }
bool isXMEM() const { return is(XMM | MEM); } bool isXMEM() const { return is(XMM | MEM); }
bool isYMEM() const { return is(YMM | MEM); } bool isYMEM() const { return is(YMM | MEM); }
bool isZMEM() const { return is(ZMM | MEM); } bool isZMEM() const { return is(ZMM | MEM); }
@ -463,20 +499,20 @@ public:
int getRounding() const { return rounding_; } int getRounding() const { return rounding_; }
void setKind(Kind kind) void setKind(Kind kind)
{ {
if ((kind & (XMM|YMM|ZMM)) == 0) return; if ((kind & (XMM|YMM|ZMM|TMM)) == 0) return;
kind_ = kind; kind_ = kind;
bit_ = kind == XMM ? 128 : kind == YMM ? 256 : 512; bit_ = kind == XMM ? 128 : kind == YMM ? 256 : kind == ZMM ? 512 : 8192;
} }
// err if MMX/FPU/OPMASK/BNDREG // err if MMX/FPU/OPMASK/BNDREG
void setBit(int bit); void setBit(int bit);
void setOpmaskIdx(int idx, bool /*ignore_idx0*/ = true) void setOpmaskIdx(int idx, bool /*ignore_idx0*/ = true)
{ {
if (mask_) throw Error(ERR_OPMASK_IS_ALREADY_SET); if (mask_) XBYAK_THROW(ERR_OPMASK_IS_ALREADY_SET)
mask_ = idx; mask_ = idx;
} }
void setRounding(int idx) void setRounding(int idx)
{ {
if (rounding_) throw Error(ERR_ROUNDING_IS_ALREADY_SET); if (rounding_) XBYAK_THROW(ERR_ROUNDING_IS_ALREADY_SET)
rounding_ = idx; rounding_ = idx;
} }
void setZero() { zero_ = true; } void setZero() { zero_ = true; }
@ -513,6 +549,11 @@ public:
} else if (isOPMASK()) { } else if (isOPMASK()) {
static const char *tbl[8] = { "k0", "k1", "k2", "k3", "k4", "k5", "k6", "k7" }; static const char *tbl[8] = { "k0", "k1", "k2", "k3", "k4", "k5", "k6", "k7" };
return tbl[idx]; return tbl[idx];
} else if (isTMM()) {
static const char *tbl[8] = {
"tmm0", "tmm1", "tmm2", "tmm3", "tmm4", "tmm5", "tmm6", "tmm7"
};
return tbl[idx];
} else if (isZMM()) { } else if (isZMM()) {
static const char *tbl[32] = { static const char *tbl[32] = {
"zmm0", "zmm1", "zmm2", "zmm3", "zmm4", "zmm5", "zmm6", "zmm7", "zmm8", "zmm9", "zmm10", "zmm11", "zmm12", "zmm13", "zmm14", "zmm15", "zmm0", "zmm1", "zmm2", "zmm3", "zmm4", "zmm5", "zmm6", "zmm7", "zmm8", "zmm9", "zmm10", "zmm11", "zmm12", "zmm13", "zmm14", "zmm15",
@ -541,7 +582,7 @@ public:
static const char *tbl[4] = { "bnd0", "bnd1", "bnd2", "bnd3" }; static const char *tbl[4] = { "bnd0", "bnd1", "bnd2", "bnd3" };
return tbl[idx]; return tbl[idx];
} }
throw Error(ERR_INTERNAL); XBYAK_THROW_RET(ERR_INTERNAL, 0);
} }
bool isEqualIfNotInherited(const Operand& rhs) const { return idx_ == rhs.idx_ && kind_ == rhs.kind_ && bit_ == rhs.bit_ && zero_ == rhs.zero_ && mask_ == rhs.mask_ && rounding_ == rhs.rounding_; } bool isEqualIfNotInherited(const Operand& rhs) const { return idx_ == rhs.idx_ && kind_ == rhs.kind_ && bit_ == rhs.bit_ && zero_ == rhs.zero_ && mask_ == rhs.mask_ && rounding_ == rhs.rounding_; }
bool operator==(const Operand& rhs) const; bool operator==(const Operand& rhs) const;
@ -552,13 +593,13 @@ public:
inline void Operand::setBit(int bit) inline void Operand::setBit(int bit)
{ {
if (bit != 8 && bit != 16 && bit != 32 && bit != 64 && bit != 128 && bit != 256 && bit != 512) goto ERR; if (bit != 8 && bit != 16 && bit != 32 && bit != 64 && bit != 128 && bit != 256 && bit != 512 && bit != 8192) goto ERR;
if (isBit(bit)) return; if (isBit(bit)) return;
if (is(MEM | OPMASK)) { if (is(MEM | OPMASK)) {
bit_ = bit; bit_ = bit;
return; return;
} }
if (is(REG | XMM | YMM | ZMM)) { if (is(REG | XMM | YMM | ZMM | TMM)) {
int idx = getIdx(); int idx = getIdx();
// err if converting ah, bh, ch, dh // err if converting ah, bh, ch, dh
if (isREG(8) && (4 <= idx && idx < 8) && !isExt8bit()) goto ERR; if (isREG(8) && (4 <= idx && idx < 8) && !isExt8bit()) goto ERR;
@ -580,6 +621,7 @@ inline void Operand::setBit(int bit)
case 128: kind = XMM; break; case 128: kind = XMM; break;
case 256: kind = YMM; break; case 256: kind = YMM; break;
case 512: kind = ZMM; break; case 512: kind = ZMM; break;
case 8192: kind = TMM; break;
} }
idx_ = idx; idx_ = idx;
kind_ = kind; kind_ = kind;
@ -590,7 +632,7 @@ inline void Operand::setBit(int bit)
return; return;
} }
ERR: ERR:
throw Error(ERR_CANT_CONVERT); XBYAK_THROW(ERR_CANT_CONVERT)
} }
class Label; class Label;
@ -674,6 +716,12 @@ struct Zmm : public Ymm {
Zmm operator|(const EvexModifierRounding& emr) const { Zmm r(*this); r.setRounding(emr.rounding); return r; } Zmm operator|(const EvexModifierRounding& emr) const { Zmm r(*this); r.setRounding(emr.rounding); return r; }
}; };
#ifdef XBYAK64
struct Tmm : public Reg {
explicit Tmm(int idx = 0, Kind kind = Operand::TMM, int bit = 8192) : Reg(idx, kind, bit) { }
};
#endif
struct Opmask : public Reg { struct Opmask : public Reg {
explicit Opmask(int idx = 0) : Reg(idx, Operand::OPMASK, 64) {} explicit Opmask(int idx = 0) : Reg(idx, Operand::OPMASK, 64) {}
}; };
@ -718,11 +766,11 @@ struct RegRip {
return RegRip(r.disp_ - disp, r.label_, r.isAddr_); return RegRip(r.disp_ - disp, r.label_, r.isAddr_);
} }
friend const RegRip operator+(const RegRip& r, const Label& label) { friend const RegRip operator+(const RegRip& r, const Label& label) {
if (r.label_ || r.isAddr_) throw Error(ERR_BAD_ADDRESSING); if (r.label_ || r.isAddr_) XBYAK_THROW_RET(ERR_BAD_ADDRESSING, RegRip());
return RegRip(r.disp_, &label); return RegRip(r.disp_, &label);
} }
friend const RegRip operator+(const RegRip& r, const void *addr) { friend const RegRip operator+(const RegRip& r, const void *addr) {
if (r.label_ || r.isAddr_) throw Error(ERR_BAD_ADDRESSING); if (r.label_ || r.isAddr_) XBYAK_THROW_RET(ERR_BAD_ADDRESSING, RegRip());
return RegRip(r.disp_ + (sint64)addr, 0, true); return RegRip(r.disp_ + (sint64)addr, 0, true);
} }
}; };
@ -782,9 +830,9 @@ public:
: scale_(scale) : scale_(scale)
, disp_(0) , disp_(0)
{ {
if (!r.isREG(i32e) && !r.is(Reg::XMM|Reg::YMM|Reg::ZMM)) throw Error(ERR_BAD_SIZE_OF_REGISTER); if (!r.isREG(i32e) && !r.is(Reg::XMM|Reg::YMM|Reg::ZMM|Reg::TMM)) XBYAK_THROW(ERR_BAD_SIZE_OF_REGISTER)
if (scale == 0) return; if (scale == 0) return;
if (scale != 1 && scale != 2 && scale != 4 && scale != 8) throw Error(ERR_BAD_SCALE); if (scale != 1 && scale != 2 && scale != 4 && scale != 8) XBYAK_THROW(ERR_BAD_SCALE)
if (r.getBit() >= 128 || scale != 1) { // xmm/ymm is always index if (r.getBit() >= 128 || scale != 1) { // xmm/ymm is always index
index_ = r; index_ = r;
} else { } else {
@ -812,10 +860,10 @@ public:
size_t getDisp() const { return disp_; } size_t getDisp() const { return disp_; }
void verify() const void verify() const
{ {
if (base_.getBit() >= 128) throw Error(ERR_BAD_SIZE_OF_REGISTER); if (base_.getBit() >= 128) XBYAK_THROW(ERR_BAD_SIZE_OF_REGISTER)
if (index_.getBit() && index_.getBit() <= 64) { if (index_.getBit() && index_.getBit() <= 64) {
if (index_.getIdx() == Operand::ESP) throw Error(ERR_ESP_CANT_BE_INDEX); if (index_.getIdx() == Operand::ESP) XBYAK_THROW(ERR_ESP_CANT_BE_INDEX)
if (base_.getBit() && base_.getBit() != index_.getBit()) throw Error(ERR_BAD_SIZE_OF_REGISTER); if (base_.getBit() && base_.getBit() != index_.getBit()) XBYAK_THROW(ERR_BAD_SIZE_OF_REGISTER)
} }
} }
friend RegExp operator+(const RegExp& a, const RegExp& b); friend RegExp operator+(const RegExp& a, const RegExp& b);
@ -838,12 +886,12 @@ private:
inline RegExp operator+(const RegExp& a, const RegExp& b) inline RegExp operator+(const RegExp& a, const RegExp& b)
{ {
if (a.index_.getBit() && b.index_.getBit()) throw Error(ERR_BAD_ADDRESSING); if (a.index_.getBit() && b.index_.getBit()) XBYAK_THROW_RET(ERR_BAD_ADDRESSING, RegExp())
RegExp ret = a; RegExp ret = a;
if (!ret.index_.getBit()) { ret.index_ = b.index_; ret.scale_ = b.scale_; } if (!ret.index_.getBit()) { ret.index_ = b.index_; ret.scale_ = b.scale_; }
if (b.base_.getBit()) { if (b.base_.getBit()) {
if (ret.base_.getBit()) { if (ret.base_.getBit()) {
if (ret.index_.getBit()) throw Error(ERR_BAD_ADDRESSING); if (ret.index_.getBit()) XBYAK_THROW_RET(ERR_BAD_ADDRESSING, RegExp())
// base + base => base + index * 1 // base + base => base + index * 1
ret.index_ = b.base_; ret.index_ = b.base_;
// [reg + esp] => [esp + reg] // [reg + esp] => [esp + reg]
@ -917,7 +965,7 @@ protected:
{ {
const size_t newSize = (std::max<size_t>)(DEFAULT_MAX_CODE_SIZE, maxSize_ * 2); const size_t newSize = (std::max<size_t>)(DEFAULT_MAX_CODE_SIZE, maxSize_ * 2);
uint8 *newTop = alloc_->alloc(newSize); uint8 *newTop = alloc_->alloc(newSize);
if (newTop == 0) throw Error(ERR_CANT_ALLOC); if (newTop == 0) XBYAK_THROW(ERR_CANT_ALLOC)
for (size_t i = 0; i < size_; i++) newTop[i] = top_[i]; for (size_t i = 0; i < size_; i++) newTop[i] = top_[i];
alloc_->free(top_); alloc_->free(top_);
top_ = newTop; top_ = newTop;
@ -949,10 +997,10 @@ public:
, size_(0) , size_(0)
, isCalledCalcJmpAddress_(false) , isCalledCalcJmpAddress_(false)
{ {
if (maxSize_ > 0 && top_ == 0) throw Error(ERR_CANT_ALLOC); if (maxSize_ > 0 && top_ == 0) XBYAK_THROW(ERR_CANT_ALLOC)
if ((type_ == ALLOC_BUF && userPtr != DontSetProtectRWE && useProtect()) && !setProtectMode(PROTECT_RWE, false)) { if ((type_ == ALLOC_BUF && userPtr != DontSetProtectRWE && useProtect()) && !setProtectMode(PROTECT_RWE, false)) {
alloc_->free(top_); alloc_->free(top_);
throw Error(ERR_CANT_PROTECT); XBYAK_THROW(ERR_CANT_PROTECT)
} }
} }
virtual ~CodeArray() virtual ~CodeArray()
@ -966,7 +1014,7 @@ public:
{ {
bool isOK = protect(top_, maxSize_, mode); bool isOK = protect(top_, maxSize_, mode);
if (isOK) return true; if (isOK) return true;
if (throwException) throw Error(ERR_CANT_PROTECT); if (throwException) XBYAK_THROW_RET(ERR_CANT_PROTECT, false)
return false; return false;
} }
bool setProtectModeRE(bool throwException = true) { return setProtectMode(PROTECT_RE, throwException); } bool setProtectModeRE(bool throwException = true) { return setProtectMode(PROTECT_RE, throwException); }
@ -983,7 +1031,7 @@ public:
if (type_ == AUTO_GROW) { if (type_ == AUTO_GROW) {
growMemory(); growMemory();
} else { } else {
throw Error(ERR_CODE_IS_TOO_BIG); XBYAK_THROW(ERR_CODE_IS_TOO_BIG)
} }
} }
top_[size_++] = static_cast<uint8>(code); top_[size_++] = static_cast<uint8>(code);
@ -994,7 +1042,7 @@ public:
} }
void db(uint64 code, size_t codeSize) void db(uint64 code, size_t codeSize)
{ {
if (codeSize > 8) throw Error(ERR_BAD_PARAMETER); if (codeSize > 8) XBYAK_THROW(ERR_BAD_PARAMETER)
for (size_t i = 0; i < codeSize; i++) db(static_cast<uint8>(code >> (i * 8))); for (size_t i = 0; i < codeSize; i++) db(static_cast<uint8>(code >> (i * 8)));
} }
void dw(uint32 code) { db(code, 2); } void dw(uint32 code) { db(code, 2); }
@ -1009,7 +1057,7 @@ public:
size_t getSize() const { return size_; } size_t getSize() const { return size_; }
void setSize(size_t size) void setSize(size_t size)
{ {
if (size > maxSize_) throw Error(ERR_OFFSET_IS_TOO_BIG); if (size > maxSize_) XBYAK_THROW(ERR_OFFSET_IS_TOO_BIG)
size_ = size; size_ = size;
} }
void dump() const void dump() const
@ -1042,7 +1090,7 @@ public:
void rewrite(size_t offset, uint64 disp, size_t size) void rewrite(size_t offset, uint64 disp, size_t size)
{ {
assert(offset < maxSize_); assert(offset < maxSize_);
if (size != 1 && size != 2 && size != 4 && size != 8) throw Error(ERR_BAD_PARAMETER); if (size != 1 && size != 2 && size != 4 && size != 8) XBYAK_THROW(ERR_BAD_PARAMETER)
uint8 *const data = top_ + offset; uint8 *const data = top_ + offset;
for (size_t i = 0; i < size; i++) { for (size_t i = 0; i < size; i++) {
data[i] = static_cast<uint8>(disp >> (i * 8)); data[i] = static_cast<uint8>(disp >> (i * 8));
@ -1271,7 +1319,7 @@ class LabelManager {
// add label // add label
typename DefList::value_type item(labelId, addrOffset); typename DefList::value_type item(labelId, addrOffset);
std::pair<typename DefList::iterator, bool> ret = defList.insert(item); std::pair<typename DefList::iterator, bool> ret = defList.insert(item);
if (!ret.second) throw Error(ERR_LABEL_IS_REDEFINED); if (!ret.second) XBYAK_THROW(ERR_LABEL_IS_REDEFINED)
// search undefined label // search undefined label
for (;;) { for (;;) {
typename UndefList::iterator itr = undefList.find(labelId); typename UndefList::iterator itr = undefList.find(labelId);
@ -1286,9 +1334,9 @@ class LabelManager {
} else { } else {
disp = addrOffset - jmp->endOfJmp + jmp->disp; disp = addrOffset - jmp->endOfJmp + jmp->disp;
#ifdef XBYAK64 #ifdef XBYAK64
if (jmp->jmpSize <= 4 && !inner::IsInInt32(disp)) throw Error(ERR_OFFSET_IS_TOO_BIG); if (jmp->jmpSize <= 4 && !inner::IsInInt32(disp)) XBYAK_THROW(ERR_OFFSET_IS_TOO_BIG)
#endif #endif
if (jmp->jmpSize == 1 && !inner::IsInDisp8((uint32)disp)) throw Error(ERR_LABEL_IS_TOO_FAR); if (jmp->jmpSize == 1 && !inner::IsInDisp8((uint32)disp)) XBYAK_THROW(ERR_LABEL_IS_TOO_FAR)
} }
if (base_->isAutoGrow()) { if (base_->isAutoGrow()) {
base_->save(offset, disp, jmp->jmpSize, jmp->mode); base_->save(offset, disp, jmp->jmpSize, jmp->mode);
@ -1360,6 +1408,7 @@ public:
clabelDefList_.clear(); clabelDefList_.clear();
clabelUndefList_.clear(); clabelUndefList_.clear();
resetLabelPtrList(); resetLabelPtrList();
ClearError();
} }
void enterLocal() void enterLocal()
{ {
@ -1367,14 +1416,14 @@ public:
} }
void leaveLocal() void leaveLocal()
{ {
if (stateList_.size() <= 2) throw Error(ERR_UNDER_LOCAL_LABEL); if (stateList_.size() <= 2) XBYAK_THROW(ERR_UNDER_LOCAL_LABEL)
if (hasUndefinedLabel_inner(stateList_.back().undefList)) throw Error(ERR_LABEL_IS_NOT_FOUND); if (hasUndefinedLabel_inner(stateList_.back().undefList)) XBYAK_THROW(ERR_LABEL_IS_NOT_FOUND)
stateList_.pop_back(); stateList_.pop_back();
} }
void set(CodeArray *base) { base_ = base; } void set(CodeArray *base) { base_ = base; }
void defineSlabel(std::string label) void defineSlabel(std::string label)
{ {
if (label == "@b" || label == "@f") throw Error(ERR_BAD_LABEL_STR); if (label == "@b" || label == "@f") XBYAK_THROW(ERR_BAD_LABEL_STR)
if (label == "@@") { if (label == "@@") {
SlabelDefList& defList = stateList_.front().defList; SlabelDefList& defList = stateList_.front().defList;
SlabelDefList::iterator i = defList.find("@f"); SlabelDefList::iterator i = defList.find("@f");
@ -1401,7 +1450,7 @@ public:
void assign(Label& dst, const Label& src) void assign(Label& dst, const Label& src)
{ {
ClabelDefList::const_iterator i = clabelDefList_.find(src.id); ClabelDefList::const_iterator i = clabelDefList_.find(src.id);
if (i == clabelDefList_.end()) throw Error(ERR_LABEL_ISNOT_SET_BY_L); if (i == clabelDefList_.end()) XBYAK_THROW(ERR_LABEL_ISNOT_SET_BY_L)
define_inner(clabelDefList_, clabelUndefList_, dst.id, i->second.offset); define_inner(clabelDefList_, clabelUndefList_, dst.id, i->second.offset);
dst.mgr = this; dst.mgr = this;
labelPtrList_.insert(&dst); labelPtrList_.insert(&dst);
@ -1413,7 +1462,7 @@ public:
if (defList.find("@f") != defList.end()) { if (defList.find("@f") != defList.end()) {
label = "@f"; label = "@f";
} else if (defList.find("@b") == defList.end()) { } else if (defList.find("@b") == defList.end()) {
throw Error(ERR_LABEL_IS_NOT_FOUND); XBYAK_THROW_RET(ERR_LABEL_IS_NOT_FOUND, false)
} }
} else if (label == "@f") { } else if (label == "@f") {
if (defList.find("@f") != defList.end()) { if (defList.find("@f") != defList.end()) {
@ -1456,7 +1505,7 @@ inline Label::Label(const Label& rhs)
} }
inline Label& Label::operator=(const Label& rhs) inline Label& Label::operator=(const Label& rhs)
{ {
if (id) throw Error(ERR_LABEL_IS_ALREADY_SET_BY_L); if (id) XBYAK_THROW_RET(ERR_LABEL_IS_ALREADY_SET_BY_L, *this)
id = rhs.id; id = rhs.id;
mgr = rhs.mgr; mgr = rhs.mgr;
if (mgr) mgr->incRefCount(id, this); if (mgr) mgr->incRefCount(id, this);
@ -1485,7 +1534,7 @@ private:
CodeGenerator operator=(const CodeGenerator&); // don't call CodeGenerator operator=(const CodeGenerator&); // don't call
#ifdef XBYAK64 #ifdef XBYAK64
enum { i32e = 32 | 64, BIT = 64 }; enum { i32e = 32 | 64, BIT = 64 };
static const size_t dummyAddr = (size_t(0x11223344) << 32) | 55667788; static const uint64 dummyAddr = uint64(0x1122334455667788ull);
typedef Reg64 NativeReg; typedef Reg64 NativeReg;
#else #else
enum { i32e = 32, BIT = 32 }; enum { i32e = 32, BIT = 32 };
@ -1532,7 +1581,7 @@ private:
uint8 rex = 0; uint8 rex = 0;
const Operand *p1 = &op1, *p2 = &op2; const Operand *p1 = &op1, *p2 = &op2;
if (p1->isMEM()) std::swap(p1, p2); if (p1->isMEM()) std::swap(p1, p2);
if (p1->isMEM()) throw Error(ERR_BAD_COMBINATION); if (p1->isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION)
if (p2->isMEM()) { if (p2->isMEM()) {
const Address& addr = p2->getAddress(); const Address& addr = p2->getAddress();
if (BIT == 64 && addr.is32bit()) db(0x67); if (BIT == 64 && addr.is32bit()) db(0x67);
@ -1592,7 +1641,7 @@ private:
bool r = reg.isExtIdx(); bool r = reg.isExtIdx();
bool b = base.isExtIdx(); bool b = base.isExtIdx();
int idx = v ? v->getIdx() : 0; int idx = v ? v->getIdx() : 0;
if ((idx | reg.getIdx() | base.getIdx()) >= 16) throw Error(ERR_BAD_COMBINATION); if ((idx | reg.getIdx() | base.getIdx()) >= 16) XBYAK_THROW(ERR_BAD_COMBINATION)
uint32 pp = (type & T_66) ? 1 : (type & T_F3) ? 2 : (type & T_F2) ? 3 : 0; uint32 pp = (type & T_66) ? 1 : (type & T_F3) ? 2 : (type & T_F2) ? 3 : 0;
uint32 vvvv = (((~idx) & 15) << 3) | (is256 ? 4 : 0) | pp; uint32 vvvv = (((~idx) & 15) << 3) | (is256 ? 4 : 0) | pp;
if (!b && !x && !w && (type & T_0F)) { if (!b && !x && !w && (type & T_0F)) {
@ -1606,23 +1655,23 @@ private:
void verifySAE(const Reg& r, int type) const void verifySAE(const Reg& r, int type) const
{ {
if (((type & T_SAE_X) && r.isXMM()) || ((type & T_SAE_Y) && r.isYMM()) || ((type & T_SAE_Z) && r.isZMM())) return; if (((type & T_SAE_X) && r.isXMM()) || ((type & T_SAE_Y) && r.isYMM()) || ((type & T_SAE_Z) && r.isZMM())) return;
throw Error(ERR_SAE_IS_INVALID); XBYAK_THROW(ERR_SAE_IS_INVALID)
} }
void verifyER(const Reg& r, int type) const void verifyER(const Reg& r, int type) const
{ {
if (((type & T_ER_X) && r.isXMM()) || ((type & T_ER_Y) && r.isYMM()) || ((type & T_ER_Z) && r.isZMM())) return; if (((type & T_ER_X) && r.isXMM()) || ((type & T_ER_Y) && r.isYMM()) || ((type & T_ER_Z) && r.isZMM())) return;
throw Error(ERR_ER_IS_INVALID); XBYAK_THROW(ERR_ER_IS_INVALID)
} }
// (a, b, c) contains non zero two or three values then err // (a, b, c) contains non zero two or three values then err
int verifyDuplicate(int a, int b, int c, int err) int verifyDuplicate(int a, int b, int c, int err)
{ {
int v = a | b | c; int v = a | b | c;
if ((a > 0 && a != v) + (b > 0 && b != v) + (c > 0 && c != v) > 0) return Error(err); if ((a > 0 && a != v) + (b > 0 && b != v) + (c > 0 && c != v) > 0) XBYAK_THROW_RET(err, 0)
return v; return v;
} }
int evex(const Reg& reg, const Reg& base, const Operand *v, int type, int code, bool x = false, bool b = false, int aaa = 0, uint32 VL = 0, bool Hi16Vidx = false) int evex(const Reg& reg, const Reg& base, const Operand *v, int type, int code, bool x = false, bool b = false, int aaa = 0, uint32 VL = 0, bool Hi16Vidx = false)
{ {
if (!(type & (T_EVEX | T_MUST_EVEX))) throw Error(ERR_EVEX_IS_INVALID); if (!(type & (T_EVEX | T_MUST_EVEX))) XBYAK_THROW_RET(ERR_EVEX_IS_INVALID, 0)
int w = (type & T_EW1) ? 1 : 0; int w = (type & T_EW1) ? 1 : 0;
uint32 mm = (type & T_0F) ? 1 : (type & T_0F38) ? 2 : (type & T_0F3A) ? 3 : 0; uint32 mm = (type & T_0F) ? 1 : (type & T_0F38) ? 2 : (type & T_0F3A) ? 3 : 0;
uint32 pp = (type & T_66) ? 1 : (type & T_F3) ? 2 : (type & T_F2) ? 3 : 0; uint32 pp = (type & T_66) ? 1 : (type & T_F3) ? 2 : (type & T_F2) ? 3 : 0;
@ -1680,10 +1729,10 @@ private:
} }
void setSIB(const RegExp& e, int reg, int disp8N = 0) void setSIB(const RegExp& e, int reg, int disp8N = 0)
{ {
size_t disp64 = e.getDisp(); uint64 disp64 = e.getDisp();
#ifdef XBYAK64 #ifdef XBYAK64
size_t high = disp64 >> 32; uint64 high = disp64 >> 32;
if (high != 0 && high != 0xFFFFFFFF) throw Error(ERR_OFFSET_IS_TOO_BIG); if (high != 0 && high != 0xFFFFFFFF) XBYAK_THROW(ERR_OFFSET_IS_TOO_BIG)
#endif #endif
uint32 disp = static_cast<uint32>(disp64); uint32 disp = static_cast<uint32>(disp64);
const Reg& base = e.getBase(); const Reg& base = e.getBase();
@ -1743,23 +1792,23 @@ private:
} }
void opModM(const Address& addr, const Reg& reg, int code0, int code1 = NONE, int code2 = NONE, int immSize = 0) void opModM(const Address& addr, const Reg& reg, int code0, int code1 = NONE, int code2 = NONE, int immSize = 0)
{ {
if (addr.is64bitDisp()) throw Error(ERR_CANT_USE_64BIT_DISP); if (addr.is64bitDisp()) XBYAK_THROW(ERR_CANT_USE_64BIT_DISP)
rex(addr, reg); rex(addr, reg);
db(code0 | (reg.isBit(8) ? 0 : 1)); if (code1 != NONE) db(code1); if (code2 != NONE) db(code2); db(code0 | (reg.isBit(8) ? 0 : 1)); if (code1 != NONE) db(code1); if (code2 != NONE) db(code2);
opAddr(addr, reg.getIdx(), immSize); opAddr(addr, reg.getIdx(), immSize);
} }
void opLoadSeg(const Address& addr, const Reg& reg, int code0, int code1 = NONE) void opLoadSeg(const Address& addr, const Reg& reg, int code0, int code1 = NONE)
{ {
if (addr.is64bitDisp()) throw Error(ERR_CANT_USE_64BIT_DISP); if (addr.is64bitDisp()) XBYAK_THROW(ERR_CANT_USE_64BIT_DISP)
if (reg.isBit(8)) throw Error(ERR_BAD_SIZE_OF_REGISTER); if (reg.isBit(8)) XBYAK_THROW(ERR_BAD_SIZE_OF_REGISTER)
rex(addr, reg); rex(addr, reg);
db(code0); if (code1 != NONE) db(code1); db(code0); if (code1 != NONE) db(code1);
opAddr(addr, reg.getIdx()); opAddr(addr, reg.getIdx());
} }
void opMIB(const Address& addr, const Reg& reg, int code0, int code1) void opMIB(const Address& addr, const Reg& reg, int code0, int code1)
{ {
if (addr.is64bitDisp()) throw Error(ERR_CANT_USE_64BIT_DISP); if (addr.is64bitDisp()) XBYAK_THROW(ERR_CANT_USE_64BIT_DISP)
if (addr.getMode() != Address::M_ModRM) throw Error(ERR_INVALID_MIB_ADDRESS); if (addr.getMode() != Address::M_ModRM) XBYAK_THROW(ERR_INVALID_MIB_ADDRESS)
if (BIT == 64 && addr.is32bit()) db(0x67); if (BIT == 64 && addr.is32bit()) db(0x67);
const RegExp& regExp = addr.getRegExp(false); const RegExp& regExp = addr.getRegExp(false);
uint8 rex = regExp.getRex(); uint8 rex = regExp.getRex();
@ -1775,7 +1824,7 @@ private:
if (type != T_NEAR && inner::IsInDisp8(disp - shortJmpSize)) { if (type != T_NEAR && inner::IsInDisp8(disp - shortJmpSize)) {
db(shortCode); db(disp - shortJmpSize); db(shortCode); db(disp - shortJmpSize);
} else { } else {
if (type == T_SHORT) throw Error(ERR_LABEL_IS_TOO_FAR); if (type == T_SHORT) XBYAK_THROW(ERR_LABEL_IS_TOO_FAR)
if (longPref) db(longPref); if (longPref) db(longPref);
db(longCode); dd(disp - longJmpSize); db(longCode); dd(disp - longJmpSize);
} }
@ -1805,7 +1854,7 @@ private:
void opJmpAbs(const void *addr, LabelType type, uint8 shortCode, uint8 longCode, uint8 longPref = 0) void opJmpAbs(const void *addr, LabelType type, uint8 shortCode, uint8 longCode, uint8 longPref = 0)
{ {
if (isAutoGrow()) { if (isAutoGrow()) {
if (!isNEAR(type)) throw Error(ERR_ONLY_T_NEAR_IS_SUPPORTED_IN_AUTO_GROW); if (!isNEAR(type)) XBYAK_THROW(ERR_ONLY_T_NEAR_IS_SUPPORTED_IN_AUTO_GROW)
if (size_ + 16 >= maxSize_) growMemory(); if (size_ + 16 >= maxSize_) growMemory();
if (longPref) db(longPref); if (longPref) db(longPref);
db(longCode); db(longCode);
@ -1821,7 +1870,7 @@ private:
// disp8N = 0(normal), disp8N = 1(force disp32), disp8N = {2, 4, 8} ; compressed displacement // disp8N = 0(normal), disp8N = 1(force disp32), disp8N = {2, 4, 8} ; compressed displacement
void opAddr(const Address &addr, int reg, int immSize = 0, int disp8N = 0, bool permitVisb = false) void opAddr(const Address &addr, int reg, int immSize = 0, int disp8N = 0, bool permitVisb = false)
{ {
if (!permitVisb && addr.isVsib()) throw Error(ERR_BAD_VSIB_ADDRESSING); if (!permitVisb && addr.isVsib()) XBYAK_THROW(ERR_BAD_VSIB_ADDRESSING)
if (addr.getMode() == Address::M_ModRM) { if (addr.getMode() == Address::M_ModRM) {
setSIB(addr.getRegExp(), reg, disp8N); setSIB(addr.getRegExp(), reg, disp8N);
} else if (addr.getMode() == Address::M_rip || addr.getMode() == Address::M_ripAddr) { } else if (addr.getMode() == Address::M_rip || addr.getMode() == Address::M_ripAddr) {
@ -1831,7 +1880,7 @@ private:
} else { } else {
size_t disp = addr.getDisp(); size_t disp = addr.getDisp();
if (addr.getMode() == Address::M_ripAddr) { if (addr.getMode() == Address::M_ripAddr) {
if (isAutoGrow()) throw Error(ERR_INVALID_RIP_IN_AUTO_GROW); if (isAutoGrow()) XBYAK_THROW(ERR_INVALID_RIP_IN_AUTO_GROW)
disp -= (size_t)getCurr() + 4 + immSize; disp -= (size_t)getCurr() + 4 + immSize;
} }
dd(inner::VerifyInInt32(disp)); dd(inner::VerifyInInt32(disp));
@ -1841,7 +1890,7 @@ private:
/* preCode is for SSSE3/SSE4 */ /* preCode is for SSSE3/SSE4 */
void opGen(const Operand& reg, const Operand& op, int code, int pref, bool isValid(const Operand&, const Operand&), int imm8 = NONE, int preCode = NONE) void opGen(const Operand& reg, const Operand& op, int code, int pref, bool isValid(const Operand&, const Operand&), int imm8 = NONE, int preCode = NONE)
{ {
if (isValid && !isValid(reg, op)) throw Error(ERR_BAD_COMBINATION); if (isValid && !isValid(reg, op)) XBYAK_THROW(ERR_BAD_COMBINATION)
if (pref != NONE) db(pref); if (pref != NONE) db(pref);
if (op.isMEM()) { if (op.isMEM()) {
opModM(op.getAddress(), reg.getReg(), 0x0F, preCode, code, (imm8 != NONE) ? 1 : 0); opModM(op.getAddress(), reg.getReg(), 0x0F, preCode, code, (imm8 != NONE) ? 1 : 0);
@ -1868,7 +1917,7 @@ private:
} else if (op1.isMEM() && op2.isXMM()) { } else if (op1.isMEM() && op2.isXMM()) {
opModM(op1.getAddress(), op2.getReg(), 0x0F, code | 1); opModM(op1.getAddress(), op2.getReg(), 0x0F, code | 1);
} else { } else {
throw Error(ERR_BAD_COMBINATION); XBYAK_THROW(ERR_BAD_COMBINATION)
} }
} }
void opExt(const Operand& op, const Mmx& mmx, int code, int imm, bool hasMMX2 = false) void opExt(const Operand& op, const Mmx& mmx, int code, int imm, bool hasMMX2 = false)
@ -1889,7 +1938,7 @@ private:
} else if (op.isMEM()) { } else if (op.isMEM()) {
opModM(op.getAddress(), Reg(ext, Operand::REG, opBit), code0, code1, code2, immSize); opModM(op.getAddress(), Reg(ext, Operand::REG, opBit), code0, code1, code2, immSize);
} else { } else {
throw Error(ERR_BAD_COMBINATION); XBYAK_THROW(ERR_BAD_COMBINATION)
} }
} }
void opShift(const Operand& op, int imm, int ext) void opShift(const Operand& op, int imm, int ext)
@ -1900,7 +1949,7 @@ private:
} }
void opShift(const Operand& op, const Reg8& _cl, int ext) void opShift(const Operand& op, const Reg8& _cl, int ext)
{ {
if (_cl.getIdx() != Operand::CL) throw Error(ERR_BAD_COMBINATION); if (_cl.getIdx() != Operand::CL) XBYAK_THROW(ERR_BAD_COMBINATION)
opR_ModM(op, 0, ext, 0xD2); opR_ModM(op, 0, ext, 0xD2);
} }
void opModRM(const Operand& op1, const Operand& op2, bool condR, bool condM, int code0, int code1 = NONE, int code2 = NONE, int immSize = 0) void opModRM(const Operand& op1, const Operand& op2, bool condR, bool condM, int code0, int code1 = NONE, int code2 = NONE, int immSize = 0)
@ -1910,12 +1959,12 @@ private:
} else if (condM) { } else if (condM) {
opModM(op2.getAddress(), op1.getReg(), code0, code1, code2, immSize); opModM(op2.getAddress(), op1.getReg(), code0, code1, code2, immSize);
} else { } else {
throw Error(ERR_BAD_COMBINATION); XBYAK_THROW(ERR_BAD_COMBINATION)
} }
} }
void opShxd(const Operand& op, const Reg& reg, uint8 imm, int code, const Reg8 *_cl = 0) void opShxd(const Operand& op, const Reg& reg, uint8 imm, int code, const Reg8 *_cl = 0)
{ {
if (_cl && _cl->getIdx() != Operand::CL) throw Error(ERR_BAD_COMBINATION); if (_cl && _cl->getIdx() != Operand::CL) XBYAK_THROW(ERR_BAD_COMBINATION)
opModRM(reg, op, (op.isREG(16 | i32e) && op.getBit() == reg.getBit()), op.isMEM() && (reg.isREG(16 | i32e)), 0x0F, code | (_cl ? 1 : 0), NONE, _cl ? 0 : 1); opModRM(reg, op, (op.isREG(16 | i32e) && op.getBit() == reg.getBit()), op.isMEM() && (reg.isREG(16 | i32e)), 0x0F, code | (_cl ? 1 : 0), NONE, _cl ? 0 : 1);
if (!_cl) db(imm); if (!_cl) db(imm);
} }
@ -1934,7 +1983,7 @@ private:
verifyMemHasSize(op); verifyMemHasSize(op);
uint32 immBit = inner::IsInDisp8(imm) ? 8 : isInDisp16(imm) ? 16 : 32; uint32 immBit = inner::IsInDisp8(imm) ? 8 : isInDisp16(imm) ? 16 : 32;
if (op.isBit(8)) immBit = 8; if (op.isBit(8)) immBit = 8;
if (op.getBit() < immBit) throw Error(ERR_IMM_IS_TOO_BIG); if (op.getBit() < immBit) XBYAK_THROW(ERR_IMM_IS_TOO_BIG)
if (op.isBit(32|64) && immBit == 16) immBit = 32; /* don't use MEM16 if 32/64bit mode */ if (op.isBit(32|64) && immBit == 16) immBit = 32; /* don't use MEM16 if 32/64bit mode */
if (op.isREG() && op.getIdx() == 0 && (op.getBit() == immBit || (op.isBit(64) && immBit == 32))) { // rax, eax, ax, al if (op.isREG() && op.getIdx() == 0 && (op.getBit() == immBit || (op.isBit(64) && immBit == 32))) { // rax, eax, ax, al
rex(op); rex(op);
@ -1976,21 +2025,21 @@ private:
return; return;
} }
} }
throw Error(ERR_BAD_COMBINATION); XBYAK_THROW(ERR_BAD_COMBINATION)
} }
void verifyMemHasSize(const Operand& op) const void verifyMemHasSize(const Operand& op) const
{ {
if (op.isMEM() && op.getBit() == 0) throw Error(ERR_MEM_SIZE_IS_NOT_SPECIFIED); if (op.isMEM() && op.getBit() == 0) XBYAK_THROW(ERR_MEM_SIZE_IS_NOT_SPECIFIED)
} }
/* /*
mov(r, imm) = db(imm, mov_imm(r, imm)) mov(r, imm) = db(imm, mov_imm(r, imm))
*/ */
int mov_imm(const Reg& reg, size_t imm) int mov_imm(const Reg& reg, uint64 imm)
{ {
int bit = reg.getBit(); int bit = reg.getBit();
const int idx = reg.getIdx(); const int idx = reg.getIdx();
int code = 0xB0 | ((bit == 8 ? 0 : 1) << 3); int code = 0xB0 | ((bit == 8 ? 0 : 1) << 3);
if (bit == 64 && (imm & ~size_t(0xffffffffu)) == 0) { if (bit == 64 && (imm & ~uint64(0xffffffffu)) == 0) {
rex(Reg32(idx)); rex(Reg32(idx));
bit = 32; bit = 32;
} else { } else {
@ -2027,19 +2076,19 @@ private:
} }
void opMovxx(const Reg& reg, const Operand& op, uint8 code) void opMovxx(const Reg& reg, const Operand& op, uint8 code)
{ {
if (op.isBit(32)) throw Error(ERR_BAD_COMBINATION); if (op.isBit(32)) XBYAK_THROW(ERR_BAD_COMBINATION)
int w = op.isBit(16); int w = op.isBit(16);
#ifdef XBYAK64 #ifdef XBYAK64
if (op.isHigh8bit()) throw Error(ERR_BAD_COMBINATION); if (op.isHigh8bit()) XBYAK_THROW(ERR_BAD_COMBINATION)
#endif #endif
bool cond = reg.isREG() && (reg.getBit() > op.getBit()); bool cond = reg.isREG() && (reg.getBit() > op.getBit());
opModRM(reg, op, cond && op.isREG(), cond && op.isMEM(), 0x0F, code | w); opModRM(reg, op, cond && op.isREG(), cond && op.isMEM(), 0x0F, code | w);
} }
void opFpuMem(const Address& addr, uint8 m16, uint8 m32, uint8 m64, uint8 ext, uint8 m64ext) void opFpuMem(const Address& addr, uint8 m16, uint8 m32, uint8 m64, uint8 ext, uint8 m64ext)
{ {
if (addr.is64bitDisp()) throw Error(ERR_CANT_USE_64BIT_DISP); if (addr.is64bitDisp()) XBYAK_THROW(ERR_CANT_USE_64BIT_DISP)
uint8 code = addr.isBit(16) ? m16 : addr.isBit(32) ? m32 : addr.isBit(64) ? m64 : 0; uint8 code = addr.isBit(16) ? m16 : addr.isBit(32) ? m32 : addr.isBit(64) ? m64 : 0;
if (!code) throw Error(ERR_BAD_MEM_SIZE); if (!code) XBYAK_THROW(ERR_BAD_MEM_SIZE)
if (m64ext && addr.isBit(64)) ext = m64ext; if (m64ext && addr.isBit(64)) ext = m64ext;
rex(addr, st0); rex(addr, st0);
@ -2051,7 +2100,7 @@ private:
void opFpuFpu(const Fpu& reg1, const Fpu& reg2, uint32 code1, uint32 code2) void opFpuFpu(const Fpu& reg1, const Fpu& reg2, uint32 code1, uint32 code2)
{ {
uint32 code = reg1.getIdx() == 0 ? code1 : reg2.getIdx() == 0 ? code2 : 0; uint32 code = reg1.getIdx() == 0 ? code1 : reg2.getIdx() == 0 ? code2 : 0;
if (!code) throw Error(ERR_BAD_ST_COMBINATION); if (!code) XBYAK_THROW(ERR_BAD_ST_COMBINATION)
db(uint8(code >> 8)); db(uint8(code >> 8));
db(uint8(code | (reg1.getIdx() | reg2.getIdx()))); db(uint8(code | (reg1.getIdx() | reg2.getIdx())));
} }
@ -2071,10 +2120,10 @@ private:
bool x = index.isExtIdx(); bool x = index.isExtIdx();
if ((type & (T_MUST_EVEX|T_MEM_EVEX)) || r.hasEvex() || (p1 && p1->hasEvex()) || addr.isBroadcast() || addr.getOpmaskIdx()) { if ((type & (T_MUST_EVEX|T_MEM_EVEX)) || r.hasEvex() || (p1 && p1->hasEvex()) || addr.isBroadcast() || addr.getOpmaskIdx()) {
int aaa = addr.getOpmaskIdx(); int aaa = addr.getOpmaskIdx();
if (aaa && !(type & T_M_K)) throw Error(ERR_INVALID_OPMASK_WITH_MEMORY); if (aaa && !(type & T_M_K)) XBYAK_THROW(ERR_INVALID_OPMASK_WITH_MEMORY)
bool b = false; bool b = false;
if (addr.isBroadcast()) { if (addr.isBroadcast()) {
if (!(type & (T_B32 | T_B64))) throw Error(ERR_INVALID_BROADCAST); if (!(type & (T_B32 | T_B64))) XBYAK_THROW(ERR_INVALID_BROADCAST)
b = true; b = true;
} }
int VL = regExp.isVsib() ? index.getBit() : 0; int VL = regExp.isVsib() ? index.getBit() : 0;
@ -2102,7 +2151,7 @@ private:
const Operand *p2 = &op2; const Operand *p2 = &op2;
if (!isR_R_RM) std::swap(p1, p2); if (!isR_R_RM) std::swap(p1, p2);
const unsigned int bit = r.getBit(); const unsigned int bit = r.getBit();
if (p1->getBit() != bit || (p2->isREG() && p2->getBit() != bit)) throw Error(ERR_BAD_COMBINATION); if (p1->getBit() != bit || (p2->isREG() && p2->getBit() != bit)) XBYAK_THROW(ERR_BAD_COMBINATION)
type |= (bit == 64) ? T_W1 : T_W0; type |= (bit == 64) ? T_W1 : T_W0;
opVex(r, p1, *p2, type, code, imm8); opVex(r, p1, *p2, type, code, imm8);
} }
@ -2115,23 +2164,23 @@ private:
op = &op1; op = &op1;
} }
// (x1, x2, op) // (x1, x2, op)
if (!((x1.isXMM() && x2->isXMM()) || ((type & T_YMM) && ((x1.isYMM() && x2->isYMM()) || (x1.isZMM() && x2->isZMM()))))) throw Error(ERR_BAD_COMBINATION); if (!((x1.isXMM() && x2->isXMM()) || ((type & T_YMM) && ((x1.isYMM() && x2->isYMM()) || (x1.isZMM() && x2->isZMM()))))) XBYAK_THROW(ERR_BAD_COMBINATION)
opVex(x1, x2, *op, type, code0, imm8); opVex(x1, x2, *op, type, code0, imm8);
} }
void opAVX_K_X_XM(const Opmask& k, const Xmm& x2, const Operand& op3, int type, int code0, int imm8 = NONE) void opAVX_K_X_XM(const Opmask& k, const Xmm& x2, const Operand& op3, int type, int code0, int imm8 = NONE)
{ {
if (!op3.isMEM() && (x2.getKind() != op3.getKind())) throw Error(ERR_BAD_COMBINATION); if (!op3.isMEM() && (x2.getKind() != op3.getKind())) XBYAK_THROW(ERR_BAD_COMBINATION)
opVex(k, &x2, op3, type, code0, imm8); opVex(k, &x2, op3, type, code0, imm8);
} }
// (x, x/m), (y, x/m256), (z, y/m) // (x, x/m), (y, x/m256), (z, y/m)
void checkCvt1(const Operand& x, const Operand& op) const void checkCvt1(const Operand& x, const Operand& op) const
{ {
if (!op.isMEM() && !(x.is(Operand::XMM | Operand::YMM) && op.isXMM()) && !(x.isZMM() && op.isYMM())) throw Error(ERR_BAD_COMBINATION); if (!op.isMEM() && !(x.is(Operand::XMM | Operand::YMM) && op.isXMM()) && !(x.isZMM() && op.isYMM())) XBYAK_THROW(ERR_BAD_COMBINATION)
} }
// (x, x/m), (x, y/m256), (y, z/m) // (x, x/m), (x, y/m256), (y, z/m)
void checkCvt2(const Xmm& x, const Operand& op) const void checkCvt2(const Xmm& x, const Operand& op) const
{ {
if (!(x.isXMM() && op.is(Operand::XMM | Operand::YMM | Operand::MEM)) && !(x.isYMM() && op.is(Operand::ZMM | Operand::MEM))) throw Error(ERR_BAD_COMBINATION); if (!(x.isXMM() && op.is(Operand::XMM | Operand::YMM | Operand::MEM)) && !(x.isYMM() && op.is(Operand::ZMM | Operand::MEM))) XBYAK_THROW(ERR_BAD_COMBINATION)
} }
void opCvt2(const Xmm& x, const Operand& op, int type, int code) void opCvt2(const Xmm& x, const Operand& op, int type, int code)
{ {
@ -2141,7 +2190,7 @@ private:
} }
void opCvt3(const Xmm& x1, const Xmm& x2, const Operand& op, int type, int type64, int type32, uint8 code) void opCvt3(const Xmm& x1, const Xmm& x2, const Operand& op, int type, int type64, int type32, uint8 code)
{ {
if (!(x1.isXMM() && x2.isXMM() && (op.isREG(i32e) || op.isMEM()))) throw Error(ERR_BAD_SIZE_OF_REGISTER); if (!(x1.isXMM() && x2.isXMM() && (op.isREG(i32e) || op.isMEM()))) XBYAK_THROW(ERR_BAD_SIZE_OF_REGISTER)
Xmm x(op.getIdx()); Xmm x(op.getIdx());
const Operand *p = op.isREG() ? &x : &op; const Operand *p = op.isREG() ? &x : &op;
opVex(x1, &x2, *p, type | (op.isBit(64) ? type64 : type32), code); opVex(x1, &x2, *p, type | (op.isBit(64) ? type64 : type32), code);
@ -2158,16 +2207,16 @@ private:
// QQQ:need to refactor // QQQ:need to refactor
void opSp1(const Reg& reg, const Operand& op, uint8 pref, uint8 code0, uint8 code1) void opSp1(const Reg& reg, const Operand& op, uint8 pref, uint8 code0, uint8 code1)
{ {
if (reg.isBit(8)) throw Error(ERR_BAD_SIZE_OF_REGISTER); if (reg.isBit(8)) XBYAK_THROW(ERR_BAD_SIZE_OF_REGISTER)
bool is16bit = reg.isREG(16) && (op.isREG(16) || op.isMEM()); bool is16bit = reg.isREG(16) && (op.isREG(16) || op.isMEM());
if (!is16bit && !(reg.isREG(i32e) && (op.isREG(reg.getBit()) || op.isMEM()))) throw Error(ERR_BAD_COMBINATION); if (!is16bit && !(reg.isREG(i32e) && (op.isREG(reg.getBit()) || op.isMEM()))) XBYAK_THROW(ERR_BAD_COMBINATION)
if (is16bit) db(0x66); if (is16bit) db(0x66);
db(pref); opModRM(reg.changeBit(i32e == 32 ? 32 : reg.getBit()), op, op.isREG(), true, code0, code1); db(pref); opModRM(reg.changeBit(i32e == 32 ? 32 : reg.getBit()), op, op.isREG(), true, code0, code1);
} }
void opGather(const Xmm& x1, const Address& addr, const Xmm& x2, int type, uint8 code, int mode) void opGather(const Xmm& x1, const Address& addr, const Xmm& x2, int type, uint8 code, int mode)
{ {
const RegExp& regExp = addr.getRegExp(); const RegExp& regExp = addr.getRegExp();
if (!regExp.isVsib(128 | 256)) throw Error(ERR_BAD_VSIB_ADDRESSING); if (!regExp.isVsib(128 | 256)) XBYAK_THROW(ERR_BAD_VSIB_ADDRESSING)
const int y_vx_y = 0; const int y_vx_y = 0;
const int y_vy_y = 1; const int y_vy_y = 1;
// const int x_vy_x = 2; // const int x_vy_x = 2;
@ -2181,7 +2230,7 @@ private:
} else { // x_vy_x } else { // x_vy_x
isOK = !x1.isYMM() && isAddrYMM && !x2.isYMM(); isOK = !x1.isYMM() && isAddrYMM && !x2.isYMM();
} }
if (!isOK) throw Error(ERR_BAD_VSIB_ADDRESSING); if (!isOK) XBYAK_THROW(ERR_BAD_VSIB_ADDRESSING)
} }
opAVX_X_X_XM(isAddrYMM ? Ymm(x1.getIdx()) : x1, isAddrYMM ? Ymm(x2.getIdx()) : x2, addr, type, code); opAVX_X_X_XM(isAddrYMM ? Ymm(x1.getIdx()) : x1, isAddrYMM ? Ymm(x2.getIdx()) : x2, addr, type, code);
} }
@ -2201,11 +2250,11 @@ private:
case xx_xy_yz: if ((x1.isXMM() && x2.isYMM()) || (x1.isYMM() && x2.isZMM())) return; case xx_xy_yz: if ((x1.isXMM() && x2.isYMM()) || (x1.isYMM() && x2.isZMM())) return;
break; break;
} }
throw Error(ERR_BAD_VSIB_ADDRESSING); XBYAK_THROW(ERR_BAD_VSIB_ADDRESSING)
} }
void opGather2(const Xmm& x, const Address& addr, int type, uint8 code, int mode) void opGather2(const Xmm& x, const Address& addr, int type, uint8 code, int mode)
{ {
if (x.hasZero()) throw Error(ERR_INVALID_ZERO); if (x.hasZero()) XBYAK_THROW(ERR_INVALID_ZERO)
checkGather2(x, addr.getRegExp().getIndex(), mode); checkGather2(x, addr.getRegExp().getIndex(), mode);
opVex(x, 0, addr, type, code); opVex(x, 0, addr, type, code);
} }
@ -2216,16 +2265,16 @@ private:
void opVmov(const Operand& op, const Xmm& x, int type, uint8 code, bool mode) void opVmov(const Operand& op, const Xmm& x, int type, uint8 code, bool mode)
{ {
if (mode) { if (mode) {
if (!op.isMEM() && !((op.isXMM() && x.isXMM()) || (op.isXMM() && x.isYMM()) || (op.isYMM() && x.isZMM()))) throw Error(ERR_BAD_COMBINATION); if (!op.isMEM() && !((op.isXMM() && x.isXMM()) || (op.isXMM() && x.isYMM()) || (op.isYMM() && x.isZMM()))) XBYAK_THROW(ERR_BAD_COMBINATION)
} else { } else {
if (!op.isMEM() && !op.isXMM()) throw Error(ERR_BAD_COMBINATION); if (!op.isMEM() && !op.isXMM()) XBYAK_THROW(ERR_BAD_COMBINATION)
} }
opVex(x, 0, op, type, code); opVex(x, 0, op, type, code);
} }
void opGatherFetch(const Address& addr, const Xmm& x, int type, uint8 code, Operand::Kind kind) void opGatherFetch(const Address& addr, const Xmm& x, int type, uint8 code, Operand::Kind kind)
{ {
if (addr.hasZero()) throw Error(ERR_INVALID_ZERO); if (addr.hasZero()) XBYAK_THROW(ERR_INVALID_ZERO)
if (addr.getRegExp().getIndex().getKind() != kind) throw Error(ERR_BAD_VSIB_ADDRESSING); if (addr.getRegExp().getIndex().getKind() != kind) XBYAK_THROW(ERR_BAD_VSIB_ADDRESSING)
opVex(x, 0, addr, type, code); opVex(x, 0, addr, type, code);
} }
void opInOut(const Reg& a, const Reg& d, uint8 code) void opInOut(const Reg& a, const Reg& d, uint8 code)
@ -2237,7 +2286,7 @@ private:
case 32: db(code + 1); return; case 32: db(code + 1); return;
} }
} }
throw Error(ERR_BAD_COMBINATION); XBYAK_THROW(ERR_BAD_COMBINATION)
} }
void opInOut(const Reg& a, uint8 code, uint8 v) void opInOut(const Reg& a, uint8 code, uint8 v)
{ {
@ -2248,8 +2297,17 @@ private:
case 32: db(code + 1); db(v); return; case 32: db(code + 1); db(v); return;
} }
} }
throw Error(ERR_BAD_COMBINATION); XBYAK_THROW(ERR_BAD_COMBINATION)
}
#ifdef XBYAK64
void opAMX(const Tmm& t1, const Address& addr, int type, int code0)
{
// require both base and index
const RegExp exp = addr.getRegExp(false);
if (exp.getBase().getBit() == 0 || exp.getIndex().getBit() == 0) XBYAK_THROW(ERR_NOT_SUPPORTED)
opVex(t1, &tmm0, addr, type, code0);
} }
#endif
public: public:
unsigned int getVersion() const { return VERSION; } unsigned int getVersion() const { return VERSION; }
using CodeArray::db; using CodeArray::db;
@ -2285,6 +2343,7 @@ public:
const Zmm zmm8, zmm9, zmm10, zmm11, zmm12, zmm13, zmm14, zmm15; const Zmm zmm8, zmm9, zmm10, zmm11, zmm12, zmm13, zmm14, zmm15;
const Zmm zmm16, zmm17, zmm18, zmm19, zmm20, zmm21, zmm22, zmm23; const Zmm zmm16, zmm17, zmm18, zmm19, zmm20, zmm21, zmm22, zmm23;
const Zmm zmm24, zmm25, zmm26, zmm27, zmm28, zmm29, zmm30, zmm31; const Zmm zmm24, zmm25, zmm26, zmm27, zmm28, zmm29, zmm30, zmm31;
const Tmm tmm0, tmm1, tmm2, tmm3, tmm4, tmm5, tmm6, tmm7;
const Xmm &xm8, &xm9, &xm10, &xm11, &xm12, &xm13, &xm14, &xm15; // for my convenience const Xmm &xm8, &xm9, &xm10, &xm11, &xm12, &xm13, &xm14, &xm15; // for my convenience
const Xmm &xm16, &xm17, &xm18, &xm19, &xm20, &xm21, &xm22, &xm23; const Xmm &xm16, &xm17, &xm18, &xm19, &xm20, &xm21, &xm22, &xm23;
const Xmm &xm24, &xm25, &xm26, &xm27, &xm28, &xm29, &xm30, &xm31; const Xmm &xm24, &xm25, &xm26, &xm27, &xm28, &xm29, &xm30, &xm31;
@ -2411,7 +2470,7 @@ public:
db(reg1.isREG(8) ? 0xA0 : reg1.isREG() ? 0xA1 : reg2.isREG(8) ? 0xA2 : 0xA3); db(reg1.isREG(8) ? 0xA0 : reg1.isREG() ? 0xA1 : reg2.isREG(8) ? 0xA2 : 0xA3);
db(addr->getDisp(), 8); db(addr->getDisp(), 8);
} else { } else {
throw Error(ERR_BAD_COMBINATION); XBYAK_THROW(ERR_BAD_COMBINATION)
} }
} else } else
#else #else
@ -2425,7 +2484,7 @@ public:
opRM_RM(reg1, reg2, 0x88); opRM_RM(reg1, reg2, 0x88);
} }
} }
void mov(const Operand& op, size_t imm) void mov(const Operand& op, uint64 imm)
{ {
if (op.isREG()) { if (op.isREG()) {
const int size = mov_imm(op.getReg(), imm); const int size = mov_imm(op.getReg(), imm);
@ -2435,15 +2494,15 @@ public:
int immSize = op.getBit() / 8; int immSize = op.getBit() / 8;
if (immSize <= 4) { if (immSize <= 4) {
sint64 s = sint64(imm) >> (immSize * 8); sint64 s = sint64(imm) >> (immSize * 8);
if (s != 0 && s != -1) throw Error(ERR_IMM_IS_TOO_BIG); if (s != 0 && s != -1) XBYAK_THROW(ERR_IMM_IS_TOO_BIG)
} else { } else {
if (!inner::IsInInt32(imm)) throw Error(ERR_IMM_IS_TOO_BIG); if (!inner::IsInInt32(imm)) XBYAK_THROW(ERR_IMM_IS_TOO_BIG)
immSize = 4; immSize = 4;
} }
opModM(op.getAddress(), Reg(0, Operand::REG, op.getBit()), 0xC6, NONE, NONE, immSize); opModM(op.getAddress(), Reg(0, Operand::REG, op.getBit()), 0xC6, NONE, NONE, immSize);
db(static_cast<uint32>(imm), immSize); db(static_cast<uint32>(imm), immSize);
} else { } else {
throw Error(ERR_BAD_COMBINATION); XBYAK_THROW(ERR_BAD_COMBINATION)
} }
} }
void mov(const NativeReg& reg, const char *label) // can't use std::string void mov(const NativeReg& reg, const char *label) // can't use std::string
@ -2466,7 +2525,7 @@ public:
if (p1->isMEM() || (p2->isREG(16 | i32e) && p2->getIdx() == 0)) { if (p1->isMEM() || (p2->isREG(16 | i32e) && p2->getIdx() == 0)) {
p1 = &op2; p2 = &op1; p1 = &op2; p2 = &op1;
} }
if (p1->isMEM()) throw Error(ERR_BAD_COMBINATION); if (p1->isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION)
if (p2->isREG() && (p1->isREG(16 | i32e) && p1->getIdx() == 0) if (p2->isREG() && (p1->isREG(16 | i32e) && p1->getIdx() == 0)
#ifdef XBYAK64 #ifdef XBYAK64
&& (p2->getIdx() != 0 || !p1->isREG(32)) && (p2->getIdx() != 0 || !p1->isREG(32))
@ -2496,7 +2555,7 @@ public:
{ {
switch (seg.getIdx()) { switch (seg.getIdx()) {
case Segment::es: db(0x07); break; case Segment::es: db(0x07); break;
case Segment::cs: throw Error(ERR_BAD_COMBINATION); case Segment::cs: XBYAK_THROW(ERR_BAD_COMBINATION)
case Segment::ss: db(0x17); break; case Segment::ss: db(0x17); break;
case Segment::ds: db(0x1F); break; case Segment::ds: db(0x1F); break;
case Segment::fs: db(0x0F); db(0xA1); break; case Segment::fs: db(0x0F); db(0xA1); break;
@ -2566,6 +2625,7 @@ public:
, zmm8(8), zmm9(9), zmm10(10), zmm11(11), zmm12(12), zmm13(13), zmm14(14), zmm15(15) , zmm8(8), zmm9(9), zmm10(10), zmm11(11), zmm12(12), zmm13(13), zmm14(14), zmm15(15)
, zmm16(16), zmm17(17), zmm18(18), zmm19(19), zmm20(20), zmm21(21), zmm22(22), zmm23(23) , zmm16(16), zmm17(17), zmm18(18), zmm19(19), zmm20(20), zmm21(21), zmm22(22), zmm23(23)
, zmm24(24), zmm25(25), zmm26(26), zmm27(27), zmm28(28), zmm29(29), zmm30(30), zmm31(31) , zmm24(24), zmm25(25), zmm26(26), zmm27(27), zmm28(28), zmm29(29), zmm30(30), zmm31(31)
, tmm0(0), tmm1(1), tmm2(2), tmm3(3), tmm4(4), tmm5(5), tmm6(6), tmm7(7)
// for my convenience // for my convenience
, xm8(xmm8), xm9(xmm9), xm10(xmm10), xm11(xmm11), xm12(xmm12), xm13(xmm13), xm14(xmm14), xm15(xmm15) , xm8(xmm8), xm9(xmm9), xm10(xmm10), xm11(xmm11), xm12(xmm12), xm13(xmm13), xm14(xmm14), xm15(xmm15)
, xm16(xmm16), xm17(xmm17), xm18(xmm18), xm19(xmm19), xm20(xmm20), xm21(xmm21), xm22(xmm22), xm23(xmm23) , xm16(xmm16), xm17(xmm17), xm18(xmm18), xm19(xmm19), xm20(xmm20), xm21(xmm21), xm22(xmm22), xm23(xmm23)
@ -2598,7 +2658,7 @@ public:
*/ */
void ready(ProtectMode mode = PROTECT_RWE) void ready(ProtectMode mode = PROTECT_RWE)
{ {
if (hasUndefinedLabel()) throw Error(ERR_LABEL_IS_NOT_FOUND); if (hasUndefinedLabel()) XBYAK_THROW(ERR_LABEL_IS_NOT_FOUND)
if (isAutoGrow()) { if (isAutoGrow()) {
calcJmpAddress(); calcJmpAddress();
if (useProtect()) setProtectMode(mode); if (useProtect()) setProtectMode(mode);
@ -2663,7 +2723,7 @@ public:
void align(size_t x = 16, bool useMultiByteNop = true) void align(size_t x = 16, bool useMultiByteNop = true)
{ {
if (x == 1) return; if (x == 1) return;
if (x < 1 || (x & (x - 1))) throw Error(ERR_BAD_ALIGN); if (x < 1 || (x & (x - 1))) XBYAK_THROW(ERR_BAD_ALIGN)
if (isAutoGrow() && x > inner::ALIGN_PAGE_SIZE) fprintf(stderr, "warning:autoGrow mode does not support %d align\n", (int)x); if (isAutoGrow() && x > inner::ALIGN_PAGE_SIZE) fprintf(stderr, "warning:autoGrow mode does not support %d align\n", (int)x);
size_t remain = size_t(getCurr()) % x; size_t remain = size_t(getCurr()) % x;
if (remain) { if (remain) {
@ -2702,6 +2762,7 @@ static const Ymm ymm24(24), ymm25(25), ymm26(26), ymm27(27), ymm28(28), ymm29(29
static const Zmm zmm8(8), zmm9(9), zmm10(10), zmm11(11), zmm12(12), zmm13(13), zmm14(14), zmm15(15); static const Zmm zmm8(8), zmm9(9), zmm10(10), zmm11(11), zmm12(12), zmm13(13), zmm14(14), zmm15(15);
static const Zmm zmm16(16), zmm17(17), zmm18(18), zmm19(19), zmm20(20), zmm21(21), zmm22(22), zmm23(23); static const Zmm zmm16(16), zmm17(17), zmm18(18), zmm19(19), zmm20(20), zmm21(21), zmm22(22), zmm23(23);
static const Zmm zmm24(24), zmm25(25), zmm26(26), zmm27(27), zmm28(28), zmm29(29), zmm30(30), zmm31(31); static const Zmm zmm24(24), zmm25(25), zmm26(26), zmm27(27), zmm28(28), zmm29(29), zmm30(30), zmm31(31);
static const Tmm tmm0(0), tmm1(1), tmm2(2), tmm3(3), tmm4(4), tmm5(5), tmm6(6), tmm7(7);
static const RegRip rip; static const RegRip rip;
#endif #endif
#ifndef XBYAK_DISABLE_SEGMENT #ifndef XBYAK_DISABLE_SEGMENT

156
src/xbyak/xbyak_mnemonic.h
Unescape View File

@ -1,4 +1,4 @@
const char *getVersionString() const { return "5.912"; } const char *getVersionString() const { return "5.941"; }
void adc(const Operand& op, uint32 imm) { opRM_I(op, imm, 0x10, 2); } void adc(const Operand& op, uint32 imm) { opRM_I(op, imm, 0x10, 2); }
void adc(const Operand& op1, const Operand& op2) { opRM_RM(op1, op2, 0x10); } void adc(const Operand& op1, const Operand& op2) { opRM_RM(op1, op2, 0x10); }
void adcx(const Reg32e& reg, const Operand& op) { opGen(reg, op, 0xF6, 0x66, isREG32_REG32orMEM, NONE, 0x38); } void adcx(const Reg32e& reg, const Operand& op) { opGen(reg, op, 0xF6, 0x66, isREG32_REG32orMEM, NONE, 0x38); }
@ -266,7 +266,7 @@ void fnsave(const Address& addr) { opModM(addr, Reg32(6), 0xDD, 0x100); }
void fnstcw(const Address& addr) { opModM(addr, Reg32(7), 0xD9, 0x100); } void fnstcw(const Address& addr) { opModM(addr, Reg32(7), 0xD9, 0x100); }
void fnstenv(const Address& addr) { opModM(addr, Reg32(6), 0xD9, 0x100); } void fnstenv(const Address& addr) { opModM(addr, Reg32(6), 0xD9, 0x100); }
void fnstsw(const Address& addr) { opModM(addr, Reg32(7), 0xDD, 0x100); } void fnstsw(const Address& addr) { opModM(addr, Reg32(7), 0xDD, 0x100); }
void fnstsw(const Reg16& r) { if (r.getIdx() != Operand::AX) throw Error(ERR_BAD_PARAMETER); db(0xDF); db(0xE0); } void fnstsw(const Reg16& r) { if (r.getIdx() != Operand::AX) XBYAK_THROW(ERR_BAD_PARAMETER) db(0xDF); db(0xE0); }
void fpatan() { db(0xD9); db(0xF3); } void fpatan() { db(0xD9); db(0xF3); }
void fprem() { db(0xD9); db(0xF8); } void fprem() { db(0xD9); db(0xF8); }
void fprem1() { db(0xD9); db(0xF5); } void fprem1() { db(0xD9); db(0xF5); }
@ -285,7 +285,7 @@ void fstenv(const Address& addr) { db(0x9B); opModM(addr, Reg32(6), 0xD9, 0x100)
void fstp(const Address& addr) { opFpuMem(addr, 0x00, 0xD9, 0xDD, 3, 0); } void fstp(const Address& addr) { opFpuMem(addr, 0x00, 0xD9, 0xDD, 3, 0); }
void fstp(const Fpu& reg) { opFpu(reg, 0xDD, 0xD8); } void fstp(const Fpu& reg) { opFpu(reg, 0xDD, 0xD8); }
void fstsw(const Address& addr) { db(0x9B); opModM(addr, Reg32(7), 0xDD, 0x100); } void fstsw(const Address& addr) { db(0x9B); opModM(addr, Reg32(7), 0xDD, 0x100); }
void fstsw(const Reg16& r) { if (r.getIdx() != Operand::AX) throw Error(ERR_BAD_PARAMETER); db(0x9B); db(0xDF); db(0xE0); } void fstsw(const Reg16& r) { if (r.getIdx() != Operand::AX) XBYAK_THROW(ERR_BAD_PARAMETER) db(0x9B); db(0xDF); db(0xE0); }
void fsub(const Address& addr) { opFpuMem(addr, 0x00, 0xD8, 0xDC, 4, 0); } void fsub(const Address& addr) { opFpuMem(addr, 0x00, 0xD8, 0xDC, 4, 0); }
void fsub(const Fpu& reg1) { opFpuFpu(st0, reg1, 0xD8E0, 0xDCE8); } void fsub(const Fpu& reg1) { opFpuFpu(st0, reg1, 0xD8E0, 0xDCE8); }
void fsub(const Fpu& reg1, const Fpu& reg2) { opFpuFpu(reg1, reg2, 0xD8E0, 0xDCE8); } void fsub(const Fpu& reg1, const Fpu& reg2) { opFpuFpu(reg1, reg2, 0xD8E0, 0xDCE8); }
@ -454,7 +454,7 @@ void jz(std::string label, LabelType type = T_AUTO) { opJmp(label, type, 0x74, 0
void lahf() { db(0x9F); } void lahf() { db(0x9F); }
void lddqu(const Xmm& xmm, const Address& addr) { db(0xF2); opModM(addr, xmm, 0x0F, 0xF0); } void lddqu(const Xmm& xmm, const Address& addr) { db(0xF2); opModM(addr, xmm, 0x0F, 0xF0); }
void ldmxcsr(const Address& addr) { opModM(addr, Reg32(2), 0x0F, 0xAE); } void ldmxcsr(const Address& addr) { opModM(addr, Reg32(2), 0x0F, 0xAE); }
void lea(const Reg& reg, const Address& addr) { if (!reg.isBit(16 | i32e)) throw Error(ERR_BAD_SIZE_OF_REGISTER); opModM(addr, reg, 0x8D); } void lea(const Reg& reg, const Address& addr) { if (!reg.isBit(16 | i32e)) XBYAK_THROW(ERR_BAD_SIZE_OF_REGISTER) opModM(addr, reg, 0x8D); }
void leave() { db(0xC9); } void leave() { db(0xC9); }
void lfence() { db(0x0F); db(0xAE); db(0xE8); } void lfence() { db(0x0F); db(0xAE); db(0xE8); }
void lfs(const Reg& reg, const Address& addr) { opLoadSeg(addr, reg, 0x0F, 0xB4); } void lfs(const Reg& reg, const Address& addr) { opLoadSeg(addr, reg, 0x0F, 0xB4); }
@ -475,7 +475,7 @@ void loopne(std::string label) { opJmp(label, T_SHORT, 0xE0, 0, 0); }
void lss(const Reg& reg, const Address& addr) { opLoadSeg(addr, reg, 0x0F, 0xB2); } void lss(const Reg& reg, const Address& addr) { opLoadSeg(addr, reg, 0x0F, 0xB2); }
void lzcnt(const Reg&reg, const Operand& op) { opSp1(reg, op, 0xF3, 0x0F, 0xBD); } void lzcnt(const Reg&reg, const Operand& op) { opSp1(reg, op, 0xF3, 0x0F, 0xBD); }
void maskmovdqu(const Xmm& reg1, const Xmm& reg2) { db(0x66); opModR(reg1, reg2, 0x0F, 0xF7); } void maskmovdqu(const Xmm& reg1, const Xmm& reg2) { db(0x66); opModR(reg1, reg2, 0x0F, 0xF7); }
void maskmovq(const Mmx& reg1, const Mmx& reg2) { if (!reg1.isMMX() || !reg2.isMMX()) throw Error(ERR_BAD_COMBINATION); opModR(reg1, reg2, 0x0F, 0xF7); } void maskmovq(const Mmx& reg1, const Mmx& reg2) { if (!reg1.isMMX() || !reg2.isMMX()) XBYAK_THROW(ERR_BAD_COMBINATION) opModR(reg1, reg2, 0x0F, 0xF7); }
void maxpd(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x5F, 0x66, isXMM_XMMorMEM); } void maxpd(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x5F, 0x66, isXMM_XMMorMEM); }
void maxps(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x5F, 0x100, isXMM_XMMorMEM); } void maxps(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x5F, 0x100, isXMM_XMMorMEM); }
void maxsd(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x5F, 0xF2, isXMM_XMMorMEM); } void maxsd(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x5F, 0xF2, isXMM_XMMorMEM); }
@ -516,7 +516,7 @@ void movntdqa(const Xmm& xmm, const Address& addr) { db(0x66); opModM(addr, xmm,
void movnti(const Address& addr, const Reg32e& reg) { opModM(addr, reg, 0x0F, 0xC3); } void movnti(const Address& addr, const Reg32e& reg) { opModM(addr, reg, 0x0F, 0xC3); }
void movntpd(const Address& addr, const Xmm& reg) { opModM(addr, Reg16(reg.getIdx()), 0x0F, 0x2B); } void movntpd(const Address& addr, const Xmm& reg) { opModM(addr, Reg16(reg.getIdx()), 0x0F, 0x2B); }
void movntps(const Address& addr, const Xmm& xmm) { opModM(addr, Mmx(xmm.getIdx()), 0x0F, 0x2B); } void movntps(const Address& addr, const Xmm& xmm) { opModM(addr, Mmx(xmm.getIdx()), 0x0F, 0x2B); }
void movntq(const Address& addr, const Mmx& mmx) { if (!mmx.isMMX()) throw Error(ERR_BAD_COMBINATION); opModM(addr, mmx, 0x0F, 0xE7); } void movntq(const Address& addr, const Mmx& mmx) { if (!mmx.isMMX()) XBYAK_THROW(ERR_BAD_COMBINATION) opModM(addr, mmx, 0x0F, 0xE7); }
void movq(const Address& addr, const Mmx& mmx) { if (mmx.isXMM()) db(0x66); opModM(addr, mmx, 0x0F, mmx.isXMM() ? 0xD6 : 0x7F); } void movq(const Address& addr, const Mmx& mmx) { if (mmx.isXMM()) db(0x66); opModM(addr, mmx, 0x0F, mmx.isXMM() ? 0xD6 : 0x7F); }
void movq(const Mmx& mmx, const Operand& op) { if (mmx.isXMM()) db(0xF3); opModRM(mmx, op, (mmx.getKind() == op.getKind()), op.isMEM(), 0x0F, mmx.isXMM() ? 0x7E : 0x6F); } void movq(const Mmx& mmx, const Operand& op) { if (mmx.isXMM()) db(0xF3); opModRM(mmx, op, (mmx.getKind() == op.getKind()), op.isMEM(), 0x0F, mmx.isXMM() ? 0x7E : 0x6F); }
void movq2dq(const Xmm& xmm, const Mmx& mmx) { db(0xF3); opModR(xmm, mmx, 0x0F, 0xD6); } void movq2dq(const Xmm& xmm, const Mmx& mmx) { db(0xF3); opModR(xmm, mmx, 0x0F, 0xD6); }
@ -609,7 +609,7 @@ void phsubsw(const Mmx& mmx, const Operand& op) { opMMX(mmx, op, 0x07, 0x66, NON
void phsubw(const Mmx& mmx, const Operand& op) { opMMX(mmx, op, 0x05, 0x66, NONE, 0x38); } void phsubw(const Mmx& mmx, const Operand& op) { opMMX(mmx, op, 0x05, 0x66, NONE, 0x38); }
void pinsrb(const Xmm& xmm, const Operand& op, uint8 imm) { opGen(xmm, op, 0x20, 0x66, isXMM_REG32orMEM, imm, 0x3A); } void pinsrb(const Xmm& xmm, const Operand& op, uint8 imm) { opGen(xmm, op, 0x20, 0x66, isXMM_REG32orMEM, imm, 0x3A); }
void pinsrd(const Xmm& xmm, const Operand& op, uint8 imm) { opGen(xmm, op, 0x22, 0x66, isXMM_REG32orMEM, imm, 0x3A); } void pinsrd(const Xmm& xmm, const Operand& op, uint8 imm) { opGen(xmm, op, 0x22, 0x66, isXMM_REG32orMEM, imm, 0x3A); }
void pinsrw(const Mmx& mmx, const Operand& op, int imm) { if (!op.isREG(32) && !op.isMEM()) throw Error(ERR_BAD_COMBINATION); opGen(mmx, op, 0xC4, mmx.isXMM() ? 0x66 : NONE, 0, imm); } void pinsrw(const Mmx& mmx, const Operand& op, int imm) { if (!op.isREG(32) && !op.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opGen(mmx, op, 0xC4, mmx.isXMM() ? 0x66 : NONE, 0, imm); }
void pmaddubsw(const Mmx& mmx, const Operand& op) { opMMX(mmx, op, 0x04, 0x66, NONE, 0x38); } void pmaddubsw(const Mmx& mmx, const Operand& op) { opMMX(mmx, op, 0x04, 0x66, NONE, 0x38); }
void pmaddwd(const Mmx& mmx, const Operand& op) { opMMX(mmx, op, 0xF5); } void pmaddwd(const Mmx& mmx, const Operand& op) { opMMX(mmx, op, 0xF5); }
void pmaxsb(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x3C, 0x66, isXMM_XMMorMEM, NONE, 0x38); } void pmaxsb(const Xmm& xmm, const Operand& op) { opGen(xmm, op, 0x3C, 0x66, isXMM_XMMorMEM, NONE, 0x38); }
@ -707,8 +707,8 @@ void rcr(const Operand& op, const Reg8& _cl) { opShift(op, _cl, 3); }
void rcr(const Operand& op, int imm) { opShift(op, imm, 3); } void rcr(const Operand& op, int imm) { opShift(op, imm, 3); }
void rdmsr() { db(0x0F); db(0x32); } void rdmsr() { db(0x0F); db(0x32); }
void rdpmc() { db(0x0F); db(0x33); } void rdpmc() { db(0x0F); db(0x33); }
void rdrand(const Reg& r) { if (r.isBit(8)) throw Error(ERR_BAD_SIZE_OF_REGISTER); opModR(Reg(6, Operand::REG, r.getBit()), r, 0x0F, 0xC7); } void rdrand(const Reg& r) { if (r.isBit(8)) XBYAK_THROW(ERR_BAD_SIZE_OF_REGISTER) opModR(Reg(6, Operand::REG, r.getBit()), r, 0x0F, 0xC7); }
void rdseed(const Reg& r) { if (r.isBit(8)) throw Error(ERR_BAD_SIZE_OF_REGISTER); opModR(Reg(7, Operand::REG, r.getBit()), r, 0x0F, 0xC7); } void rdseed(const Reg& r) { if (r.isBit(8)) XBYAK_THROW(ERR_BAD_SIZE_OF_REGISTER) opModR(Reg(7, Operand::REG, r.getBit()), r, 0x0F, 0xC7); }
void rdtsc() { db(0x0F); db(0x31); } void rdtsc() { db(0x0F); db(0x31); }
void rdtscp() { db(0x0F); db(0x01); db(0xF9); } void rdtscp() { db(0x0F); db(0x01); db(0xF9); }
void rep() { db(0xF3); } void rep() { db(0xF3); }
@ -839,8 +839,8 @@ void vblendvpd(const Xmm& x1, const Xmm& x2, const Operand& op, const Xmm& x4) {
void vblendvps(const Xmm& x1, const Xmm& x2, const Operand& op, const Xmm& x4) { opAVX_X_X_XM(x1, x2, op, T_0F3A | T_66 | T_YMM, 0x4A, x4.getIdx() << 4); } void vblendvps(const Xmm& x1, const Xmm& x2, const Operand& op, const Xmm& x4) { opAVX_X_X_XM(x1, x2, op, T_0F3A | T_66 | T_YMM, 0x4A, x4.getIdx() << 4); }
void vbroadcastf128(const Ymm& y, const Address& addr) { opAVX_X_XM_IMM(y, addr, T_0F38 | T_66 | T_W0 | T_YMM, 0x1A); } void vbroadcastf128(const Ymm& y, const Address& addr) { opAVX_X_XM_IMM(y, addr, T_0F38 | T_66 | T_W0 | T_YMM, 0x1A); }
void vbroadcasti128(const Ymm& y, const Address& addr) { opAVX_X_XM_IMM(y, addr, T_0F38 | T_66 | T_W0 | T_YMM, 0x5A); } void vbroadcasti128(const Ymm& y, const Address& addr) { opAVX_X_XM_IMM(y, addr, T_0F38 | T_66 | T_W0 | T_YMM, 0x5A); }
void vbroadcastsd(const Ymm& y, const Operand& op) { if (!op.isMEM() && !(y.isYMM() && op.isXMM()) && !(y.isZMM() && op.isXMM())) throw Error(ERR_BAD_COMBINATION); opAVX_X_XM_IMM(y, op, T_0F38 | T_66 | T_W0 | T_YMM | T_EVEX | T_EW1 | T_N8, 0x19); } void vbroadcastsd(const Ymm& y, const Operand& op) { if (!op.isMEM() && !(y.isYMM() && op.isXMM()) && !(y.isZMM() && op.isXMM())) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_XM_IMM(y, op, T_0F38 | T_66 | T_W0 | T_YMM | T_EVEX | T_EW1 | T_N8, 0x19); }
void vbroadcastss(const Xmm& x, const Operand& op) { if (!(op.isXMM() || op.isMEM())) throw Error(ERR_BAD_COMBINATION); opAVX_X_XM_IMM(x, op, T_N4 | T_66 | T_0F38 | T_W0 | T_YMM | T_EVEX, 0x18); } void vbroadcastss(const Xmm& x, const Operand& op) { if (!(op.isXMM() || op.isMEM())) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_XM_IMM(x, op, T_N4 | T_66 | T_0F38 | T_W0 | T_YMM | T_EVEX, 0x18); }
void vcmpeq_ospd(const Xmm& x1, const Xmm& x2, const Operand& op) { vcmppd(x1, x2, op, 16); } void vcmpeq_ospd(const Xmm& x1, const Xmm& x2, const Operand& op) { vcmppd(x1, x2, op, 16); }
void vcmpeq_osps(const Xmm& x1, const Xmm& x2, const Operand& op) { vcmpps(x1, x2, op, 16); } void vcmpeq_osps(const Xmm& x1, const Xmm& x2, const Operand& op) { vcmpps(x1, x2, op, 16); }
void vcmpeq_ossd(const Xmm& x1, const Xmm& x2, const Operand& op) { vcmpsd(x1, x2, op, 16); } void vcmpeq_ossd(const Xmm& x1, const Xmm& x2, const Operand& op) { vcmpsd(x1, x2, op, 16); }
@ -999,9 +999,9 @@ void vdivsd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand())
void vdivss(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F3 | T_EW0 | T_EVEX | T_ER_Z | T_N4, 0x5E); } void vdivss(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_0F | T_F3 | T_EW0 | T_EVEX | T_ER_Z | T_N4, 0x5E); }
void vdppd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_W0, 0x41, imm); } void vdppd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_W0, 0x41, imm); }
void vdpps(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_W0 | T_YMM, 0x40, imm); } void vdpps(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_W0 | T_YMM, 0x40, imm); }
void vextractf128(const Operand& op, const Ymm& y, uint8 imm) { if (!(op.isXMEM() && y.isYMM())) throw Error(ERR_BAD_COMBINATION); opVex(y, 0, op, T_0F3A | T_66 | T_W0 | T_YMM, 0x19, imm); } void vextractf128(const Operand& op, const Ymm& y, uint8 imm) { if (!(op.isXMEM() && y.isYMM())) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(y, 0, op, T_0F3A | T_66 | T_W0 | T_YMM, 0x19, imm); }
void vextracti128(const Operand& op, const Ymm& y, uint8 imm) { if (!(op.isXMEM() && y.isYMM())) throw Error(ERR_BAD_COMBINATION); opVex(y, 0, op, T_0F3A | T_66 | T_W0 | T_YMM, 0x39, imm); } void vextracti128(const Operand& op, const Ymm& y, uint8 imm) { if (!(op.isXMEM() && y.isYMM())) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(y, 0, op, T_0F3A | T_66 | T_W0 | T_YMM, 0x39, imm); }
void vextractps(const Operand& op, const Xmm& x, uint8 imm) { if (!((op.isREG(32) || op.isMEM()) && x.isXMM())) throw Error(ERR_BAD_COMBINATION); opVex(x, 0, op, T_0F3A | T_66 | T_W0 | T_EVEX | T_N4, 0x17, imm); } void vextractps(const Operand& op, const Xmm& x, uint8 imm) { if (!((op.isREG(32) || op.isMEM()) && x.isXMM())) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(x, 0, op, T_0F3A | T_66 | T_W0 | T_EVEX | T_N4, 0x17, imm); }
void vfmadd132pd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_W1 | T_EW1 | T_YMM | T_EVEX | T_B64, 0x98); } void vfmadd132pd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_W1 | T_EW1 | T_YMM | T_EVEX | T_B64, 0x98); }
void vfmadd132ps(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_W0 | T_EW0 | T_YMM | T_EVEX | T_B32, 0x98); } void vfmadd132ps(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_W0 | T_EW0 | T_YMM | T_EVEX | T_B32, 0x98); }
void vfmadd132sd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F38 | T_W1 | T_EW1 | T_EVEX | T_ER_X, 0x99); } void vfmadd132sd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F38 | T_W1 | T_EW1 | T_EVEX | T_ER_X, 0x99); }
@ -1073,8 +1073,8 @@ void vhaddpd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand())
void vhaddps(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_F2 | T_0F | T_YMM, 0x7C); } void vhaddps(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_F2 | T_0F | T_YMM, 0x7C); }
void vhsubpd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_66 | T_0F | T_YMM, 0x7D); } void vhsubpd(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_66 | T_0F | T_YMM, 0x7D); }
void vhsubps(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_F2 | T_0F | T_YMM, 0x7D); } void vhsubps(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_F2 | T_0F | T_YMM, 0x7D); }
void vinsertf128(const Ymm& y1, const Ymm& y2, const Operand& op, uint8 imm) { if (!(y1.isYMM() && y2.isYMM() && op.isXMEM())) throw Error(ERR_BAD_COMBINATION); opVex(y1, &y2, op, T_0F3A | T_66 | T_W0 | T_YMM, 0x18, imm); } void vinsertf128(const Ymm& y1, const Ymm& y2, const Operand& op, uint8 imm) { if (!(y1.isYMM() && y2.isYMM() && op.isXMEM())) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(y1, &y2, op, T_0F3A | T_66 | T_W0 | T_YMM, 0x18, imm); }
void vinserti128(const Ymm& y1, const Ymm& y2, const Operand& op, uint8 imm) { if (!(y1.isYMM() && y2.isYMM() && op.isXMEM())) throw Error(ERR_BAD_COMBINATION); opVex(y1, &y2, op, T_0F3A | T_66 | T_W0 | T_YMM, 0x38, imm); } void vinserti128(const Ymm& y1, const Ymm& y2, const Operand& op, uint8 imm) { if (!(y1.isYMM() && y2.isYMM() && op.isXMEM())) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(y1, &y2, op, T_0F3A | T_66 | T_W0 | T_YMM, 0x38, imm); }
void vinsertps(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_N4 | T_66 | T_0F3A | T_W0 | T_EW0 | T_EVEX, 0x21, imm); } void vinsertps(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_N4 | T_66 | T_0F3A | T_W0 | T_EW0 | T_EVEX, 0x21, imm); }
void vlddqu(const Xmm& x, const Address& addr) { opAVX_X_X_XM(x, cvtIdx0(x), addr, T_0F | T_F2 | T_W0 | T_YMM, 0xF0); } void vlddqu(const Xmm& x, const Address& addr) { opAVX_X_X_XM(x, cvtIdx0(x), addr, T_0F | T_F2 | T_W0 | T_YMM, 0xF0); }
void vldmxcsr(const Address& addr) { opAVX_X_X_XM(xm2, xm0, addr, T_0F, 0xAE); } void vldmxcsr(const Address& addr) { opAVX_X_X_XM(xm2, xm0, addr, T_0F, 0xAE); }
@ -1095,25 +1095,25 @@ void vmovapd(const Address& addr, const Xmm& xmm) { opAVX_X_XM_IMM(xmm, addr, T_
void vmovapd(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_66 | T_0F | T_EW1 | T_YMM | T_EVEX, 0x28); } void vmovapd(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_66 | T_0F | T_EW1 | T_YMM | T_EVEX, 0x28); }
void vmovaps(const Address& addr, const Xmm& xmm) { opAVX_X_XM_IMM(xmm, addr, T_0F | T_EW0 | T_YMM | T_EVEX | T_M_K, 0x29); } void vmovaps(const Address& addr, const Xmm& xmm) { opAVX_X_XM_IMM(xmm, addr, T_0F | T_EW0 | T_YMM | T_EVEX | T_M_K, 0x29); }
void vmovaps(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_0F | T_EW0 | T_YMM | T_EVEX, 0x28); } void vmovaps(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_0F | T_EW0 | T_YMM | T_EVEX, 0x28); }
void vmovd(const Operand& op, const Xmm& x) { if (!op.isREG(32) && !op.isMEM()) throw Error(ERR_BAD_COMBINATION); opAVX_X_X_XM(x, xm0, op, T_0F | T_66 | T_W0 | T_EVEX | T_N4, 0x7E); } void vmovd(const Operand& op, const Xmm& x) { if (!op.isREG(32) && !op.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x, xm0, op, T_0F | T_66 | T_W0 | T_EVEX | T_N4, 0x7E); }
void vmovd(const Xmm& x, const Operand& op) { if (!op.isREG(32) && !op.isMEM()) throw Error(ERR_BAD_COMBINATION); opAVX_X_X_XM(x, xm0, op, T_0F | T_66 | T_W0 | T_EVEX | T_N4, 0x6E); } void vmovd(const Xmm& x, const Operand& op) { if (!op.isREG(32) && !op.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x, xm0, op, T_0F | T_66 | T_W0 | T_EVEX | T_N4, 0x6E); }
void vmovddup(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_DUP | T_F2 | T_0F | T_EW1 | T_YMM | T_EVEX | T_ER_X | T_ER_Y | T_ER_Z, 0x12); } void vmovddup(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_DUP | T_F2 | T_0F | T_EW1 | T_YMM | T_EVEX | T_ER_X | T_ER_Y | T_ER_Z, 0x12); }
void vmovdqa(const Address& addr, const Xmm& xmm) { opAVX_X_XM_IMM(xmm, addr, T_66 | T_0F | T_YMM, 0x7F); } void vmovdqa(const Address& addr, const Xmm& xmm) { opAVX_X_XM_IMM(xmm, addr, T_66 | T_0F | T_YMM, 0x7F); }
void vmovdqa(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_66 | T_0F | T_YMM, 0x6F); } void vmovdqa(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_66 | T_0F | T_YMM, 0x6F); }
void vmovdqu(const Address& addr, const Xmm& xmm) { opAVX_X_XM_IMM(xmm, addr, T_F3 | T_0F | T_YMM, 0x7F); } void vmovdqu(const Address& addr, const Xmm& xmm) { opAVX_X_XM_IMM(xmm, addr, T_F3 | T_0F | T_YMM, 0x7F); }
void vmovdqu(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_F3 | T_0F | T_YMM, 0x6F); } void vmovdqu(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_F3 | T_0F | T_YMM, 0x6F); }
void vmovhlps(const Xmm& x1, const Xmm& x2, const Operand& op = Operand()) { if (!op.isNone() && !op.isXMM()) throw Error(ERR_BAD_COMBINATION); opAVX_X_X_XM(x1, x2, op, T_0F | T_EVEX | T_EW0, 0x12); } void vmovhlps(const Xmm& x1, const Xmm& x2, const Operand& op = Operand()) { if (!op.isNone() && !op.isXMM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x1, x2, op, T_0F | T_EVEX | T_EW0, 0x12); }
void vmovhpd(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, T_0F | T_66 | T_EVEX | T_EW1 | T_N8, 0x17); } void vmovhpd(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, T_0F | T_66 | T_EVEX | T_EW1 | T_N8, 0x17); }
void vmovhpd(const Xmm& x, const Operand& op1, const Operand& op2 = Operand()) { if (!op2.isNone() && !op2.isMEM()) throw Error(ERR_BAD_COMBINATION); opAVX_X_X_XM(x, op1, op2, T_0F | T_66 | T_EVEX | T_EW1 | T_N8, 0x16); } void vmovhpd(const Xmm& x, const Operand& op1, const Operand& op2 = Operand()) { if (!op2.isNone() && !op2.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x, op1, op2, T_0F | T_66 | T_EVEX | T_EW1 | T_N8, 0x16); }
void vmovhps(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, T_0F | T_EVEX | T_EW0 | T_N8, 0x17); } void vmovhps(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, T_0F | T_EVEX | T_EW0 | T_N8, 0x17); }
void vmovhps(const Xmm& x, const Operand& op1, const Operand& op2 = Operand()) { if (!op2.isNone() && !op2.isMEM()) throw Error(ERR_BAD_COMBINATION); opAVX_X_X_XM(x, op1, op2, T_0F | T_EVEX | T_EW0 | T_N8, 0x16); } void vmovhps(const Xmm& x, const Operand& op1, const Operand& op2 = Operand()) { if (!op2.isNone() && !op2.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x, op1, op2, T_0F | T_EVEX | T_EW0 | T_N8, 0x16); }
void vmovlhps(const Xmm& x1, const Xmm& x2, const Operand& op = Operand()) { if (!op.isNone() && !op.isXMM()) throw Error(ERR_BAD_COMBINATION); opAVX_X_X_XM(x1, x2, op, T_0F | T_EVEX | T_EW0, 0x16); } void vmovlhps(const Xmm& x1, const Xmm& x2, const Operand& op = Operand()) { if (!op.isNone() && !op.isXMM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x1, x2, op, T_0F | T_EVEX | T_EW0, 0x16); }
void vmovlpd(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, T_0F | T_66 | T_EVEX | T_EW1 | T_N8, 0x13); } void vmovlpd(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, T_0F | T_66 | T_EVEX | T_EW1 | T_N8, 0x13); }
void vmovlpd(const Xmm& x, const Operand& op1, const Operand& op2 = Operand()) { if (!op2.isNone() && !op2.isMEM()) throw Error(ERR_BAD_COMBINATION); opAVX_X_X_XM(x, op1, op2, T_0F | T_66 | T_EVEX | T_EW1 | T_N8, 0x12); } void vmovlpd(const Xmm& x, const Operand& op1, const Operand& op2 = Operand()) { if (!op2.isNone() && !op2.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x, op1, op2, T_0F | T_66 | T_EVEX | T_EW1 | T_N8, 0x12); }
void vmovlps(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, T_0F | T_EVEX | T_EW0 | T_N8, 0x13); } void vmovlps(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, T_0F | T_EVEX | T_EW0 | T_N8, 0x13); }
void vmovlps(const Xmm& x, const Operand& op1, const Operand& op2 = Operand()) { if (!op2.isNone() && !op2.isMEM()) throw Error(ERR_BAD_COMBINATION); opAVX_X_X_XM(x, op1, op2, T_0F | T_EVEX | T_EW0 | T_N8, 0x12); } void vmovlps(const Xmm& x, const Operand& op1, const Operand& op2 = Operand()) { if (!op2.isNone() && !op2.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x, op1, op2, T_0F | T_EVEX | T_EW0 | T_N8, 0x12); }
void vmovmskpd(const Reg& r, const Xmm& x) { if (!r.isBit(i32e)) throw Error(ERR_BAD_COMBINATION); opAVX_X_X_XM(x.isXMM() ? Xmm(r.getIdx()) : Ymm(r.getIdx()), cvtIdx0(x), x, T_0F | T_66 | T_W0 | T_YMM, 0x50); } void vmovmskpd(const Reg& r, const Xmm& x) { if (!r.isBit(i32e)) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x.isXMM() ? Xmm(r.getIdx()) : Ymm(r.getIdx()), cvtIdx0(x), x, T_0F | T_66 | T_W0 | T_YMM, 0x50); }
void vmovmskps(const Reg& r, const Xmm& x) { if (!r.isBit(i32e)) throw Error(ERR_BAD_COMBINATION); opAVX_X_X_XM(x.isXMM() ? Xmm(r.getIdx()) : Ymm(r.getIdx()), cvtIdx0(x), x, T_0F | T_W0 | T_YMM, 0x50); } void vmovmskps(const Reg& r, const Xmm& x) { if (!r.isBit(i32e)) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x.isXMM() ? Xmm(r.getIdx()) : Ymm(r.getIdx()), cvtIdx0(x), x, T_0F | T_W0 | T_YMM, 0x50); }
void vmovntdq(const Address& addr, const Xmm& x) { opVex(x, 0, addr, T_0F | T_66 | T_YMM | T_EVEX | T_EW0, 0xE7); } void vmovntdq(const Address& addr, const Xmm& x) { opVex(x, 0, addr, T_0F | T_66 | T_YMM | T_EVEX | T_EW0, 0xE7); }
void vmovntdqa(const Xmm& x, const Address& addr) { opVex(x, 0, addr, T_0F38 | T_66 | T_YMM | T_EVEX | T_EW0, 0x2A); } void vmovntdqa(const Xmm& x, const Address& addr) { opVex(x, 0, addr, T_0F38 | T_66 | T_YMM | T_EVEX | T_EW0, 0x2A); }
void vmovntpd(const Address& addr, const Xmm& x) { opVex(x, 0, addr, T_0F | T_66 | T_YMM | T_EVEX | T_EW1, 0x2B); } void vmovntpd(const Address& addr, const Xmm& x) { opVex(x, 0, addr, T_0F | T_66 | T_YMM | T_EVEX | T_EW1, 0x2B); }
@ -1123,12 +1123,12 @@ void vmovq(const Xmm& x, const Address& addr) { int type, code; if (x.getIdx() <
void vmovq(const Xmm& x1, const Xmm& x2) { opAVX_X_X_XM(x1, xm0, x2, T_0F | T_F3 | T_EVEX | T_EW1 | T_N8, 0x7E); } void vmovq(const Xmm& x1, const Xmm& x2) { opAVX_X_X_XM(x1, xm0, x2, T_0F | T_F3 | T_EVEX | T_EW1 | T_N8, 0x7E); }
void vmovsd(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, T_N8 | T_F2 | T_0F | T_EW1 | T_EVEX | T_M_K, 0x11); } void vmovsd(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, T_N8 | T_F2 | T_0F | T_EW1 | T_EVEX | T_M_K, 0x11); }
void vmovsd(const Xmm& x, const Address& addr) { opAVX_X_X_XM(x, xm0, addr, T_N8 | T_F2 | T_0F | T_EW1 | T_EVEX, 0x10); } void vmovsd(const Xmm& x, const Address& addr) { opAVX_X_X_XM(x, xm0, addr, T_N8 | T_F2 | T_0F | T_EW1 | T_EVEX, 0x10); }
void vmovsd(const Xmm& x1, const Xmm& x2, const Operand& op = Operand()) { if (!op.isNone() && !op.isXMM()) throw Error(ERR_BAD_COMBINATION); opAVX_X_X_XM(x1, x2, op, T_N8 | T_F2 | T_0F | T_EW1 | T_EVEX, 0x10); } void vmovsd(const Xmm& x1, const Xmm& x2, const Operand& op = Operand()) { if (!op.isNone() && !op.isXMM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x1, x2, op, T_N8 | T_F2 | T_0F | T_EW1 | T_EVEX, 0x10); }
void vmovshdup(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_F3 | T_0F | T_EW0 | T_YMM | T_EVEX, 0x16); } void vmovshdup(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_F3 | T_0F | T_EW0 | T_YMM | T_EVEX, 0x16); }
void vmovsldup(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_F3 | T_0F | T_EW0 | T_YMM | T_EVEX, 0x12); } void vmovsldup(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_F3 | T_0F | T_EW0 | T_YMM | T_EVEX, 0x12); }
void vmovss(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, T_N4 | T_F3 | T_0F | T_EW0 | T_EVEX | T_M_K, 0x11); } void vmovss(const Address& addr, const Xmm& x) { opAVX_X_X_XM(x, xm0, addr, T_N4 | T_F3 | T_0F | T_EW0 | T_EVEX | T_M_K, 0x11); }
void vmovss(const Xmm& x, const Address& addr) { opAVX_X_X_XM(x, xm0, addr, T_N4 | T_F3 | T_0F | T_EW0 | T_EVEX, 0x10); } void vmovss(const Xmm& x, const Address& addr) { opAVX_X_X_XM(x, xm0, addr, T_N4 | T_F3 | T_0F | T_EW0 | T_EVEX, 0x10); }
void vmovss(const Xmm& x1, const Xmm& x2, const Operand& op = Operand()) { if (!op.isNone() && !op.isXMM()) throw Error(ERR_BAD_COMBINATION); opAVX_X_X_XM(x1, x2, op, T_N4 | T_F3 | T_0F | T_EW0 | T_EVEX, 0x10); } void vmovss(const Xmm& x1, const Xmm& x2, const Operand& op = Operand()) { if (!op.isNone() && !op.isXMM()) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_X_XM(x1, x2, op, T_N4 | T_F3 | T_0F | T_EW0 | T_EVEX, 0x10); }
void vmovupd(const Address& addr, const Xmm& xmm) { opAVX_X_XM_IMM(xmm, addr, T_66 | T_0F | T_EW1 | T_YMM | T_EVEX | T_M_K, 0x11); } void vmovupd(const Address& addr, const Xmm& xmm) { opAVX_X_XM_IMM(xmm, addr, T_66 | T_0F | T_EW1 | T_YMM | T_EVEX | T_M_K, 0x11); }
void vmovupd(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_66 | T_0F | T_EW1 | T_YMM | T_EVEX, 0x10); } void vmovupd(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_66 | T_0F | T_EW1 | T_YMM | T_EVEX, 0x10); }
void vmovups(const Address& addr, const Xmm& xmm) { opAVX_X_XM_IMM(xmm, addr, T_0F | T_EW0 | T_YMM | T_EVEX | T_M_K, 0x11); } void vmovups(const Address& addr, const Xmm& xmm) { opAVX_X_XM_IMM(xmm, addr, T_0F | T_EW0 | T_YMM | T_EVEX | T_M_K, 0x11); }
@ -1163,10 +1163,10 @@ void vpavgw(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1,
void vpblendd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_W0 | T_YMM, 0x02, imm); } void vpblendd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_W0 | T_YMM, 0x02, imm); }
void vpblendvb(const Xmm& x1, const Xmm& x2, const Operand& op, const Xmm& x4) { opAVX_X_X_XM(x1, x2, op, T_0F3A | T_66 | T_YMM, 0x4C, x4.getIdx() << 4); } void vpblendvb(const Xmm& x1, const Xmm& x2, const Operand& op, const Xmm& x4) { opAVX_X_X_XM(x1, x2, op, T_0F3A | T_66 | T_YMM, 0x4C, x4.getIdx() << 4); }
void vpblendw(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_W0 | T_YMM, 0x0E, imm); } void vpblendw(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_W0 | T_YMM, 0x0E, imm); }
void vpbroadcastb(const Xmm& x, const Operand& op) { if (!(op.isXMM() || op.isMEM())) throw Error(ERR_BAD_COMBINATION); opAVX_X_XM_IMM(x, op, T_N1 | T_66 | T_0F38 | T_W0 | T_YMM | T_EVEX, 0x78); } void vpbroadcastb(const Xmm& x, const Operand& op) { if (!(op.isXMM() || op.isMEM())) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_XM_IMM(x, op, T_N1 | T_66 | T_0F38 | T_W0 | T_YMM | T_EVEX, 0x78); }
void vpbroadcastd(const Xmm& x, const Operand& op) { if (!(op.isXMM() || op.isMEM())) throw Error(ERR_BAD_COMBINATION); opAVX_X_XM_IMM(x, op, T_N4 | T_66 | T_0F38 | T_W0 | T_YMM | T_EVEX, 0x58); } void vpbroadcastd(const Xmm& x, const Operand& op) { if (!(op.isXMM() || op.isMEM())) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_XM_IMM(x, op, T_N4 | T_66 | T_0F38 | T_W0 | T_YMM | T_EVEX, 0x58); }
void vpbroadcastq(const Xmm& x, const Operand& op) { if (!(op.isXMM() || op.isMEM())) throw Error(ERR_BAD_COMBINATION); opAVX_X_XM_IMM(x, op, T_N8 | T_66 | T_0F38 | T_W0 | T_EW1 | T_YMM | T_EVEX, 0x59); } void vpbroadcastq(const Xmm& x, const Operand& op) { if (!(op.isXMM() || op.isMEM())) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_XM_IMM(x, op, T_N8 | T_66 | T_0F38 | T_W0 | T_EW1 | T_YMM | T_EVEX, 0x59); }
void vpbroadcastw(const Xmm& x, const Operand& op) { if (!(op.isXMM() || op.isMEM())) throw Error(ERR_BAD_COMBINATION); opAVX_X_XM_IMM(x, op, T_N2 | T_66 | T_0F38 | T_W0 | T_YMM | T_EVEX, 0x79); } void vpbroadcastw(const Xmm& x, const Operand& op) { if (!(op.isXMM() || op.isMEM())) XBYAK_THROW(ERR_BAD_COMBINATION) opAVX_X_XM_IMM(x, op, T_N2 | T_66 | T_0F38 | T_W0 | T_YMM | T_EVEX, 0x79); }
void vpclmulqdq(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_W0 | T_YMM | T_EVEX, 0x44, imm); } void vpclmulqdq(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_W0 | T_YMM | T_EVEX, 0x44, imm); }
void vpcmpeqb(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F | T_YMM, 0x74); } void vpcmpeqb(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F | T_YMM, 0x74); }
void vpcmpeqd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F | T_YMM, 0x76); } void vpcmpeqd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F | T_YMM, 0x76); }
@ -1180,8 +1180,8 @@ void vpcmpgtq(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1
void vpcmpgtw(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F | T_YMM, 0x65); } void vpcmpgtw(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F | T_YMM, 0x65); }
void vpcmpistri(const Xmm& xm, const Operand& op, uint8 imm) { opAVX_X_XM_IMM(xm, op, T_66 | T_0F3A, 0x63, imm); } void vpcmpistri(const Xmm& xm, const Operand& op, uint8 imm) { opAVX_X_XM_IMM(xm, op, T_66 | T_0F3A, 0x63, imm); }
void vpcmpistrm(const Xmm& xm, const Operand& op, uint8 imm) { opAVX_X_XM_IMM(xm, op, T_66 | T_0F3A, 0x62, imm); } void vpcmpistrm(const Xmm& xm, const Operand& op, uint8 imm) { opAVX_X_XM_IMM(xm, op, T_66 | T_0F3A, 0x62, imm); }
void vperm2f128(const Ymm& y1, const Ymm& y2, const Operand& op, uint8 imm) { if (!(y1.isYMM() && y2.isYMM() && op.isYMEM())) throw Error(ERR_BAD_COMBINATION); opVex(y1, &y2, op, T_0F3A | T_66 | T_W0 | T_YMM, 0x06, imm); } void vperm2f128(const Ymm& y1, const Ymm& y2, const Operand& op, uint8 imm) { if (!(y1.isYMM() && y2.isYMM() && op.isYMEM())) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(y1, &y2, op, T_0F3A | T_66 | T_W0 | T_YMM, 0x06, imm); }
void vperm2i128(const Ymm& y1, const Ymm& y2, const Operand& op, uint8 imm) { if (!(y1.isYMM() && y2.isYMM() && op.isYMEM())) throw Error(ERR_BAD_COMBINATION); opVex(y1, &y2, op, T_0F3A | T_66 | T_W0 | T_YMM, 0x46, imm); } void vperm2i128(const Ymm& y1, const Ymm& y2, const Operand& op, uint8 imm) { if (!(y1.isYMM() && y2.isYMM() && op.isYMEM())) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(y1, &y2, op, T_0F3A | T_66 | T_W0 | T_YMM, 0x46, imm); }
void vpermd(const Ymm& y1, const Ymm& y2, const Operand& op) { opAVX_X_X_XM(y1, y2, op, T_66 | T_0F38 | T_W0 | T_EW0 | T_YMM | T_EVEX | T_B32, 0x36); } void vpermd(const Ymm& y1, const Ymm& y2, const Operand& op) { opAVX_X_X_XM(y1, y2, op, T_66 | T_0F38 | T_W0 | T_EW0 | T_YMM | T_EVEX | T_B32, 0x36); }
void vpermilpd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_W0 | T_EW1 | T_YMM | T_EVEX | T_B64, 0x0D); } void vpermilpd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_W0 | T_EW1 | T_YMM | T_EVEX | T_B64, 0x0D); }
void vpermilpd(const Xmm& xm, const Operand& op, uint8 imm) { opAVX_X_XM_IMM(xm, op, T_66 | T_0F3A | T_EW1 | T_YMM | T_EVEX | T_B64, 0x05, imm); } void vpermilpd(const Xmm& xm, const Operand& op, uint8 imm) { opAVX_X_XM_IMM(xm, op, T_66 | T_0F3A | T_EW1 | T_YMM | T_EVEX | T_B64, 0x05, imm); }
@ -1192,10 +1192,10 @@ void vpermpd(const Ymm& y1, const Ymm& y2, const Operand& op) { opAVX_X_X_XM(y1,
void vpermps(const Ymm& y1, const Ymm& y2, const Operand& op) { opAVX_X_X_XM(y1, y2, op, T_66 | T_0F38 | T_W0 | T_EW0 | T_YMM | T_EVEX | T_B32, 0x16); } void vpermps(const Ymm& y1, const Ymm& y2, const Operand& op) { opAVX_X_X_XM(y1, y2, op, T_66 | T_0F38 | T_W0 | T_EW0 | T_YMM | T_EVEX | T_B32, 0x16); }
void vpermq(const Ymm& y, const Operand& op, uint8 imm) { opAVX_X_XM_IMM(y, op, T_66 | T_0F3A | T_W1 | T_EW1 | T_YMM | T_EVEX | T_B64, 0x00, imm); } void vpermq(const Ymm& y, const Operand& op, uint8 imm) { opAVX_X_XM_IMM(y, op, T_66 | T_0F3A | T_W1 | T_EW1 | T_YMM | T_EVEX | T_B64, 0x00, imm); }
void vpermq(const Ymm& y1, const Ymm& y2, const Operand& op) { opAVX_X_X_XM(y1, y2, op, T_66 | T_0F38 | T_W0 | T_EW1 | T_YMM | T_EVEX | T_B64, 0x36); } void vpermq(const Ymm& y1, const Ymm& y2, const Operand& op) { opAVX_X_X_XM(y1, y2, op, T_66 | T_0F38 | T_W0 | T_EW1 | T_YMM | T_EVEX | T_B64, 0x36); }
void vpextrb(const Operand& op, const Xmm& x, uint8 imm) { if (!((op.isREG(8|16|i32e) || op.isMEM()) && x.isXMM())) throw Error(ERR_BAD_COMBINATION); opVex(x, 0, op, T_0F3A | T_66 | T_EVEX | T_N1, 0x14, imm); } void vpextrb(const Operand& op, const Xmm& x, uint8 imm) { if (!((op.isREG(8|16|i32e) || op.isMEM()) && x.isXMM())) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(x, 0, op, T_0F3A | T_66 | T_EVEX | T_N1, 0x14, imm); }
void vpextrd(const Operand& op, const Xmm& x, uint8 imm) { if (!((op.isREG(32) || op.isMEM()) && x.isXMM())) throw Error(ERR_BAD_COMBINATION); opVex(x, 0, op, T_0F3A | T_66 | T_W0 | T_EVEX | T_EW0 | T_N4, 0x16, imm); } void vpextrd(const Operand& op, const Xmm& x, uint8 imm) { if (!((op.isREG(32) || op.isMEM()) && x.isXMM())) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(x, 0, op, T_0F3A | T_66 | T_W0 | T_EVEX | T_EW0 | T_N4, 0x16, imm); }
void vpextrq(const Operand& op, const Xmm& x, uint8 imm) { if (!((op.isREG(64) || op.isMEM()) && x.isXMM())) throw Error(ERR_BAD_COMBINATION); opVex(x, 0, op, T_0F3A | T_66 | T_W1 | T_EVEX | T_EW1 | T_N8, 0x16, imm); } void vpextrq(const Operand& op, const Xmm& x, uint8 imm) { if (!((op.isREG(64) || op.isMEM()) && x.isXMM())) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(x, 0, op, T_0F3A | T_66 | T_W1 | T_EVEX | T_EW1 | T_N8, 0x16, imm); }
void vpextrw(const Operand& op, const Xmm& x, uint8 imm) { if (!((op.isREG(16|i32e) || op.isMEM()) && x.isXMM())) throw Error(ERR_BAD_COMBINATION); if (op.isREG() && x.getIdx() < 16) { opAVX_X_X_XM(Xmm(op.getIdx()), xm0, x, T_0F | T_66, 0xC5, imm); } else { opVex(x, 0, op, T_0F3A | T_66 | T_EVEX | T_N2, 0x15, imm); } } void vpextrw(const Operand& op, const Xmm& x, uint8 imm) { if (!((op.isREG(16|i32e) || op.isMEM()) && x.isXMM())) XBYAK_THROW(ERR_BAD_COMBINATION) if (op.isREG() && x.getIdx() < 16) { opAVX_X_X_XM(Xmm(op.getIdx()), xm0, x, T_0F | T_66, 0xC5, imm); } else { opVex(x, 0, op, T_0F3A | T_66 | T_EVEX | T_N2, 0x15, imm); } }
void vpgatherdd(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_YMM | T_VSIB | T_W0, 0x90, 1); } void vpgatherdd(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_YMM | T_VSIB | T_W0, 0x90, 1); }
void vpgatherdq(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_YMM | T_VSIB | T_W1, 0x90, 0); } void vpgatherdq(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_YMM | T_VSIB | T_W1, 0x90, 0); }
void vpgatherqd(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_YMM | T_VSIB | T_W0, 0x91, 2); } void vpgatherqd(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_YMM | T_VSIB | T_W0, 0x91, 2); }
@ -1207,10 +1207,10 @@ void vphminposuw(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_66
void vphsubd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_YMM, 0x06); } void vphsubd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_YMM, 0x06); }
void vphsubsw(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_YMM, 0x07); } void vphsubsw(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_YMM, 0x07); }
void vphsubw(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_YMM, 0x05); } void vphsubw(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_YMM, 0x05); }
void vpinsrb(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { if (!(x1.isXMM() && x2.isXMM() && (op.isREG(32) || op.isMEM()))) throw Error(ERR_BAD_COMBINATION); opVex(x1, &x2, op, T_0F3A | T_66 | T_EVEX | T_N1, 0x20, imm); } void vpinsrb(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { if (!(x1.isXMM() && x2.isXMM() && (op.isREG(32) || op.isMEM()))) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(x1, &x2, op, T_0F3A | T_66 | T_EVEX | T_N1, 0x20, imm); }
void vpinsrd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { if (!(x1.isXMM() && x2.isXMM() && (op.isREG(32) || op.isMEM()))) throw Error(ERR_BAD_COMBINATION); opVex(x1, &x2, op, T_0F3A | T_66 | T_W0 | T_EVEX | T_EW0 | T_N4, 0x22, imm); } void vpinsrd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { if (!(x1.isXMM() && x2.isXMM() && (op.isREG(32) || op.isMEM()))) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(x1, &x2, op, T_0F3A | T_66 | T_W0 | T_EVEX | T_EW0 | T_N4, 0x22, imm); }
void vpinsrq(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { if (!(x1.isXMM() && x2.isXMM() && (op.isREG(64) || op.isMEM()))) throw Error(ERR_BAD_COMBINATION); opVex(x1, &x2, op, T_0F3A | T_66 | T_W1 | T_EVEX | T_EW1 | T_N8, 0x22, imm); } void vpinsrq(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { if (!(x1.isXMM() && x2.isXMM() && (op.isREG(64) || op.isMEM()))) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(x1, &x2, op, T_0F3A | T_66 | T_W1 | T_EVEX | T_EW1 | T_N8, 0x22, imm); }
void vpinsrw(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { if (!(x1.isXMM() && x2.isXMM() && (op.isREG(32) || op.isMEM()))) throw Error(ERR_BAD_COMBINATION); opVex(x1, &x2, op, T_0F | T_66 | T_EVEX | T_N2, 0xC4, imm); } void vpinsrw(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { if (!(x1.isXMM() && x2.isXMM() && (op.isREG(32) || op.isMEM()))) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(x1, &x2, op, T_0F | T_66 | T_EVEX | T_N2, 0xC4, imm); }
void vpmaddubsw(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_YMM | T_EVEX, 0x04); } void vpmaddubsw(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_YMM | T_EVEX, 0x04); }
void vpmaddwd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F | T_YMM | T_EVEX, 0xF5); } void vpmaddwd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F | T_YMM | T_EVEX, 0xF5); }
void vpmaskmovd(const Address& addr, const Xmm& x1, const Xmm& x2) { opAVX_X_X_XM(x2, x1, addr, T_0F38 | T_66 | T_W0 | T_YMM, 0x8E); } void vpmaskmovd(const Address& addr, const Xmm& x1, const Xmm& x2) { opAVX_X_X_XM(x2, x1, addr, T_0F38 | T_66 | T_W0 | T_YMM, 0x8E); }
@ -1229,7 +1229,7 @@ void vpminsw(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1,
void vpminub(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F | T_YMM | T_EVEX, 0xDA); } void vpminub(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F | T_YMM | T_EVEX, 0xDA); }
void vpminud(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_EW0 | T_YMM | T_EVEX | T_B32, 0x3B); } void vpminud(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_EW0 | T_YMM | T_EVEX | T_B32, 0x3B); }
void vpminuw(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_YMM | T_EVEX, 0x3A); } void vpminuw(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_YMM | T_EVEX, 0x3A); }
void vpmovmskb(const Reg32e& r, const Xmm& x) { if (!x.is(Operand::XMM | Operand::YMM)) throw Error(ERR_BAD_COMBINATION); opVex(x.isYMM() ? Ymm(r.getIdx()) : Xmm(r.getIdx()), 0, x, T_0F | T_66 | T_YMM, 0xD7); } void vpmovmskb(const Reg32e& r, const Xmm& x) { if (!x.is(Operand::XMM | Operand::YMM)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(x.isYMM() ? Ymm(r.getIdx()) : Xmm(r.getIdx()), 0, x, T_0F | T_66 | T_YMM, 0xD7); }
void vpmovsxbd(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_N4 | T_N_VL | T_66 | T_0F38 | T_YMM | T_EVEX, 0x21); } void vpmovsxbd(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_N4 | T_N_VL | T_66 | T_0F38 | T_YMM | T_EVEX, 0x21); }
void vpmovsxbq(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_N2 | T_N_VL | T_66 | T_0F38 | T_YMM | T_EVEX, 0x22); } void vpmovsxbq(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_N2 | T_N_VL | T_66 | T_0F38 | T_YMM | T_EVEX, 0x22); }
void vpmovsxbw(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_N8 | T_N_VL | T_66 | T_0F38 | T_YMM | T_EVEX, 0x20); } void vpmovsxbw(const Xmm& xm, const Operand& op) { opAVX_X_XM_IMM(xm, op, T_N8 | T_N_VL | T_66 | T_0F38 | T_YMM | T_EVEX, 0x20); }
@ -1618,15 +1618,27 @@ void cmpxchg16b(const Address& addr) { opModM(addr, Reg64(1), 0x0F, 0xC7); }
void fxrstor64(const Address& addr) { opModM(addr, Reg64(1), 0x0F, 0xAE); } void fxrstor64(const Address& addr) { opModM(addr, Reg64(1), 0x0F, 0xAE); }
void movq(const Reg64& reg, const Mmx& mmx) { if (mmx.isXMM()) db(0x66); opModR(mmx, reg, 0x0F, 0x7E); } void movq(const Reg64& reg, const Mmx& mmx) { if (mmx.isXMM()) db(0x66); opModR(mmx, reg, 0x0F, 0x7E); }
void movq(const Mmx& mmx, const Reg64& reg) { if (mmx.isXMM()) db(0x66); opModR(mmx, reg, 0x0F, 0x6E); } void movq(const Mmx& mmx, const Reg64& reg) { if (mmx.isXMM()) db(0x66); opModR(mmx, reg, 0x0F, 0x6E); }
void movsxd(const Reg64& reg, const Operand& op) { if (!op.isBit(32)) throw Error(ERR_BAD_COMBINATION); opModRM(reg, op, op.isREG(), op.isMEM(), 0x63); } void movsxd(const Reg64& reg, const Operand& op) { if (!op.isBit(32)) XBYAK_THROW(ERR_BAD_COMBINATION) opModRM(reg, op, op.isREG(), op.isMEM(), 0x63); }
void pextrq(const Operand& op, const Xmm& xmm, uint8 imm) { if (!op.isREG(64) && !op.isMEM()) throw Error(ERR_BAD_COMBINATION); opGen(Reg64(xmm.getIdx()), op, 0x16, 0x66, 0, imm, 0x3A); } void pextrq(const Operand& op, const Xmm& xmm, uint8 imm) { if (!op.isREG(64) && !op.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opGen(Reg64(xmm.getIdx()), op, 0x16, 0x66, 0, imm, 0x3A); }
void pinsrq(const Xmm& xmm, const Operand& op, uint8 imm) { if (!op.isREG(64) && !op.isMEM()) throw Error(ERR_BAD_COMBINATION); opGen(Reg64(xmm.getIdx()), op, 0x22, 0x66, 0, imm, 0x3A); } void pinsrq(const Xmm& xmm, const Operand& op, uint8 imm) { if (!op.isREG(64) && !op.isMEM()) XBYAK_THROW(ERR_BAD_COMBINATION) opGen(Reg64(xmm.getIdx()), op, 0x22, 0x66, 0, imm, 0x3A); }
void vcvtss2si(const Reg64& r, const Operand& op) { opAVX_X_X_XM(Xmm(r.getIdx()), xm0, op, T_0F | T_F3 | T_W1 | T_EVEX | T_EW1 | T_ER_X | T_N8, 0x2D); } void vcvtss2si(const Reg64& r, const Operand& op) { opAVX_X_X_XM(Xmm(r.getIdx()), xm0, op, T_0F | T_F3 | T_W1 | T_EVEX | T_EW1 | T_ER_X | T_N8, 0x2D); }
void vcvttss2si(const Reg64& r, const Operand& op) { opAVX_X_X_XM(Xmm(r.getIdx()), xm0, op, T_0F | T_F3 | T_W1 | T_EVEX | T_EW1 | T_SAE_X | T_N8, 0x2C); } void vcvttss2si(const Reg64& r, const Operand& op) { opAVX_X_X_XM(Xmm(r.getIdx()), xm0, op, T_0F | T_F3 | T_W1 | T_EVEX | T_EW1 | T_SAE_X | T_N8, 0x2C); }
void vcvtsd2si(const Reg64& r, const Operand& op) { opAVX_X_X_XM(Xmm(r.getIdx()), xm0, op, T_0F | T_F2 | T_W1 | T_EVEX | T_EW1 | T_N4 | T_ER_X, 0x2D); } void vcvtsd2si(const Reg64& r, const Operand& op) { opAVX_X_X_XM(Xmm(r.getIdx()), xm0, op, T_0F | T_F2 | T_W1 | T_EVEX | T_EW1 | T_N4 | T_ER_X, 0x2D); }
void vcvttsd2si(const Reg64& r, const Operand& op) { opAVX_X_X_XM(Xmm(r.getIdx()), xm0, op, T_0F | T_F2 | T_W1 | T_EVEX | T_EW1 | T_N4 | T_SAE_X, 0x2C); } void vcvttsd2si(const Reg64& r, const Operand& op) { opAVX_X_X_XM(Xmm(r.getIdx()), xm0, op, T_0F | T_F2 | T_W1 | T_EVEX | T_EW1 | T_N4 | T_SAE_X, 0x2C); }
void vmovq(const Xmm& x, const Reg64& r) { opAVX_X_X_XM(x, xm0, Xmm(r.getIdx()), T_66 | T_0F | T_W1 | T_EVEX | T_EW1, 0x6E); } void vmovq(const Xmm& x, const Reg64& r) { opAVX_X_X_XM(x, xm0, Xmm(r.getIdx()), T_66 | T_0F | T_W1 | T_EVEX | T_EW1, 0x6E); }
void vmovq(const Reg64& r, const Xmm& x) { opAVX_X_X_XM(x, xm0, Xmm(r.getIdx()), T_66 | T_0F | T_W1 | T_EVEX | T_EW1, 0x7E); } void vmovq(const Reg64& r, const Xmm& x) { opAVX_X_X_XM(x, xm0, Xmm(r.getIdx()), T_66 | T_0F | T_W1 | T_EVEX | T_EW1, 0x7E); }
void ldtilecfg(const Address& addr) { opVex(tmm0, &tmm0, addr, T_0F38 | T_W0, 0x49); }
void sttilecfg(const Address& addr) { opVex(tmm0, &tmm0, addr, T_66 | T_0F38 | T_W0, 0x49); }
void tileloadd(const Tmm& tm, const Address& addr) { opAMX(tm, addr, T_F2 | T_0F38 | T_W0, 0x4b); }
void tileloaddt1(const Tmm& tm, const Address& addr) { opAMX(tm, addr, T_66 | T_0F38 | T_W0, 0x4b); }
void tilerelease() { db(0xc4); db(0xe2); db(0x78); db(0x49); db(0xc0); }
void tilestored(const Address& addr, const Tmm& tm) { opVex(tm, &tmm0, addr, T_F3 | T_0F38 | T_W0, 0x4b); }
void tilezero(const Tmm& Tmm) { opVex(Tmm, &tmm0, tmm0, T_F2 | T_0F38 | T_W0, 0x49); }
void tdpbssd(const Tmm& x1, const Tmm& x2, const Tmm& x3) { opVex(x1, &x3, x2, T_F2 | T_0F38 | T_W0, 0x5e); }
void tdpbsud(const Tmm& x1, const Tmm& x2, const Tmm& x3) { opVex(x1, &x3, x2, T_F3 | T_0F38 | T_W0, 0x5e); }
void tdpbusd(const Tmm& x1, const Tmm& x2, const Tmm& x3) { opVex(x1, &x3, x2, T_66 | T_0F38 | T_W0, 0x5e); }
void tdpbuud(const Tmm& x1, const Tmm& x2, const Tmm& x3) { opVex(x1, &x3, x2, T_0F38 | T_W0, 0x5e); }
void tdpbf16ps(const Tmm& x1, const Tmm& x2, const Tmm& x3) { opVex(x1, &x3, x2, T_F3 | T_0F38 | T_W0, 0x5c); }
#else #else
void jcxz(std::string label) { db(0x67); opJmp(label, T_SHORT, 0xe3, 0, 0); } void jcxz(std::string label) { db(0x67); opJmp(label, T_SHORT, 0xe3, 0, 0); }
void jcxz(const Label& label) { db(0x67); opJmp(label, T_SHORT, 0xe3, 0, 0); } void jcxz(const Label& label) { db(0x67); opJmp(label, T_SHORT, 0xe3, 0, 0); }
@ -1671,17 +1683,17 @@ void kandnw(const Opmask& r1, const Opmask& r2, const Opmask& r3) { opVex(r1, &r
void kandq(const Opmask& r1, const Opmask& r2, const Opmask& r3) { opVex(r1, &r2, r3, T_L1 | T_0F | T_W1, 0x41); } void kandq(const Opmask& r1, const Opmask& r2, const Opmask& r3) { opVex(r1, &r2, r3, T_L1 | T_0F | T_W1, 0x41); }
void kandw(const Opmask& r1, const Opmask& r2, const Opmask& r3) { opVex(r1, &r2, r3, T_L1 | T_0F | T_W0, 0x41); } void kandw(const Opmask& r1, const Opmask& r2, const Opmask& r3) { opVex(r1, &r2, r3, T_L1 | T_0F | T_W0, 0x41); }
void kmovb(const Address& addr, const Opmask& k) { opVex(k, 0, addr, T_L0 | T_0F | T_66 | T_W0, 0x91); } void kmovb(const Address& addr, const Opmask& k) { opVex(k, 0, addr, T_L0 | T_0F | T_66 | T_W0, 0x91); }
void kmovb(const Opmask& k, const Operand& op) { if (!op.isMEM() && !op.isOPMASK()) throw Error(ERR_BAD_COMBINATION); opVex(k, 0, op, T_L0 | T_0F | T_66 | T_W0, 0x90); } void kmovb(const Opmask& k, const Operand& op) { if (!op.isMEM() && !op.isOPMASK()) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(k, 0, op, T_L0 | T_0F | T_66 | T_W0, 0x90); }
void kmovb(const Opmask& k, const Reg32& r) { opVex(k, 0, r, T_L0 | T_0F | T_66 | T_W0, 0x92); } void kmovb(const Opmask& k, const Reg32& r) { opVex(k, 0, r, T_L0 | T_0F | T_66 | T_W0, 0x92); }
void kmovb(const Reg32& r, const Opmask& k) { opVex(r, 0, k, T_L0 | T_0F | T_66 | T_W0, 0x93); } void kmovb(const Reg32& r, const Opmask& k) { opVex(r, 0, k, T_L0 | T_0F | T_66 | T_W0, 0x93); }
void kmovd(const Address& addr, const Opmask& k) { opVex(k, 0, addr, T_L0 | T_0F | T_66 | T_W1, 0x91); } void kmovd(const Address& addr, const Opmask& k) { opVex(k, 0, addr, T_L0 | T_0F | T_66 | T_W1, 0x91); }
void kmovd(const Opmask& k, const Operand& op) { if (!op.isMEM() && !op.isOPMASK()) throw Error(ERR_BAD_COMBINATION); opVex(k, 0, op, T_L0 | T_0F | T_66 | T_W1, 0x90); } void kmovd(const Opmask& k, const Operand& op) { if (!op.isMEM() && !op.isOPMASK()) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(k, 0, op, T_L0 | T_0F | T_66 | T_W1, 0x90); }
void kmovd(const Opmask& k, const Reg32& r) { opVex(k, 0, r, T_L0 | T_0F | T_F2 | T_W0, 0x92); } void kmovd(const Opmask& k, const Reg32& r) { opVex(k, 0, r, T_L0 | T_0F | T_F2 | T_W0, 0x92); }
void kmovd(const Reg32& r, const Opmask& k) { opVex(r, 0, k, T_L0 | T_0F | T_F2 | T_W0, 0x93); } void kmovd(const Reg32& r, const Opmask& k) { opVex(r, 0, k, T_L0 | T_0F | T_F2 | T_W0, 0x93); }
void kmovq(const Address& addr, const Opmask& k) { opVex(k, 0, addr, T_L0 | T_0F | T_W1, 0x91); } void kmovq(const Address& addr, const Opmask& k) { opVex(k, 0, addr, T_L0 | T_0F | T_W1, 0x91); }
void kmovq(const Opmask& k, const Operand& op) { if (!op.isMEM() && !op.isOPMASK()) throw Error(ERR_BAD_COMBINATION); opVex(k, 0, op, T_L0 | T_0F | T_W1, 0x90); } void kmovq(const Opmask& k, const Operand& op) { if (!op.isMEM() && !op.isOPMASK()) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(k, 0, op, T_L0 | T_0F | T_W1, 0x90); }
void kmovw(const Address& addr, const Opmask& k) { opVex(k, 0, addr, T_L0 | T_0F | T_W0, 0x91); } void kmovw(const Address& addr, const Opmask& k) { opVex(k, 0, addr, T_L0 | T_0F | T_W0, 0x91); }
void kmovw(const Opmask& k, const Operand& op) { if (!op.isMEM() && !op.isOPMASK()) throw Error(ERR_BAD_COMBINATION); opVex(k, 0, op, T_L0 | T_0F | T_W0, 0x90); } void kmovw(const Opmask& k, const Operand& op) { if (!op.isMEM() && !op.isOPMASK()) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(k, 0, op, T_L0 | T_0F | T_W0, 0x90); }
void kmovw(const Opmask& k, const Reg32& r) { opVex(k, 0, r, T_L0 | T_0F | T_W0, 0x92); } void kmovw(const Opmask& k, const Reg32& r) { opVex(k, 0, r, T_L0 | T_0F | T_W0, 0x92); }
void kmovw(const Reg32& r, const Opmask& k) { opVex(r, 0, k, T_L0 | T_0F | T_W0, 0x93); } void kmovw(const Reg32& r, const Opmask& k) { opVex(r, 0, k, T_L0 | T_0F | T_W0, 0x93); }
void knotb(const Opmask& r1, const Opmask& r2) { opVex(r1, 0, r2, T_0F | T_66 | T_W0, 0x44); } void knotb(const Opmask& r1, const Opmask& r2) { opVex(r1, 0, r2, T_0F | T_66 | T_W0, 0x44); }
@ -1777,22 +1789,22 @@ void vexp2pd(const Zmm& z, const Operand& op) { opAVX_X_XM_IMM(z, op, T_66 | T_0
void vexp2ps(const Zmm& z, const Operand& op) { opAVX_X_XM_IMM(z, op, T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_B32 | T_SAE_Z, 0xC8); } void vexp2ps(const Zmm& z, const Operand& op) { opAVX_X_XM_IMM(z, op, T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_B32 | T_SAE_Z, 0xC8); }
void vexpandpd(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX, 0x88); } void vexpandpd(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX, 0x88); }
void vexpandps(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x88); } void vexpandps(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x88); }
void vextractf32x4(const Operand& op, const Ymm& r, uint8 imm) { if (!op.is(Operand::MEM | Operand::XMM)) throw Error(ERR_BAD_COMBINATION); opVex(r, 0, op, T_N16 | T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x19, imm); } void vextractf32x4(const Operand& op, const Ymm& r, uint8 imm) { if (!op.is(Operand::MEM | Operand::XMM)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r, 0, op, T_N16 | T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x19, imm); }
void vextractf32x8(const Operand& op, const Zmm& r, uint8 imm) { if (!op.is(Operand::MEM | Operand::YMM)) throw Error(ERR_BAD_COMBINATION); opVex(r, 0, op, T_N32 | T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x1B, imm); } void vextractf32x8(const Operand& op, const Zmm& r, uint8 imm) { if (!op.is(Operand::MEM | Operand::YMM)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r, 0, op, T_N32 | T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x1B, imm); }
void vextractf64x2(const Operand& op, const Ymm& r, uint8 imm) { if (!op.is(Operand::MEM | Operand::XMM)) throw Error(ERR_BAD_COMBINATION); opVex(r, 0, op, T_N16 | T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x19, imm); } void vextractf64x2(const Operand& op, const Ymm& r, uint8 imm) { if (!op.is(Operand::MEM | Operand::XMM)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r, 0, op, T_N16 | T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x19, imm); }
void vextractf64x4(const Operand& op, const Zmm& r, uint8 imm) { if (!op.is(Operand::MEM | Operand::YMM)) throw Error(ERR_BAD_COMBINATION); opVex(r, 0, op, T_N32 | T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x1B, imm); } void vextractf64x4(const Operand& op, const Zmm& r, uint8 imm) { if (!op.is(Operand::MEM | Operand::YMM)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r, 0, op, T_N32 | T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x1B, imm); }
void vextracti32x4(const Operand& op, const Ymm& r, uint8 imm) { if (!op.is(Operand::MEM | Operand::XMM)) throw Error(ERR_BAD_COMBINATION); opVex(r, 0, op, T_N16 | T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x39, imm); } void vextracti32x4(const Operand& op, const Ymm& r, uint8 imm) { if (!op.is(Operand::MEM | Operand::XMM)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r, 0, op, T_N16 | T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x39, imm); }
void vextracti32x8(const Operand& op, const Zmm& r, uint8 imm) { if (!op.is(Operand::MEM | Operand::YMM)) throw Error(ERR_BAD_COMBINATION); opVex(r, 0, op, T_N32 | T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x3B, imm); } void vextracti32x8(const Operand& op, const Zmm& r, uint8 imm) { if (!op.is(Operand::MEM | Operand::YMM)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r, 0, op, T_N32 | T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x3B, imm); }
void vextracti64x2(const Operand& op, const Ymm& r, uint8 imm) { if (!op.is(Operand::MEM | Operand::XMM)) throw Error(ERR_BAD_COMBINATION); opVex(r, 0, op, T_N16 | T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x39, imm); } void vextracti64x2(const Operand& op, const Ymm& r, uint8 imm) { if (!op.is(Operand::MEM | Operand::XMM)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r, 0, op, T_N16 | T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x39, imm); }
void vextracti64x4(const Operand& op, const Zmm& r, uint8 imm) { if (!op.is(Operand::MEM | Operand::YMM)) throw Error(ERR_BAD_COMBINATION); opVex(r, 0, op, T_N32 | T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x3B, imm); } void vextracti64x4(const Operand& op, const Zmm& r, uint8 imm) { if (!op.is(Operand::MEM | Operand::YMM)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r, 0, op, T_N32 | T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x3B, imm); }
void vfixupimmpd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_EW1 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B64, 0x54, imm); } void vfixupimmpd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_EW1 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B64, 0x54, imm); }
void vfixupimmps(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B32, 0x54, imm); } void vfixupimmps(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F3A | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B32, 0x54, imm); }
void vfixupimmsd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F3A | T_EW1 | T_SAE_Z | T_MUST_EVEX, 0x55, imm); } void vfixupimmsd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F3A | T_EW1 | T_SAE_Z | T_MUST_EVEX, 0x55, imm); }
void vfixupimmss(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_N4 | T_66 | T_0F3A | T_EW0 | T_SAE_Z | T_MUST_EVEX, 0x55, imm); } void vfixupimmss(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_N4 | T_66 | T_0F3A | T_EW0 | T_SAE_Z | T_MUST_EVEX, 0x55, imm); }
void vfpclasspd(const Opmask& k, const Operand& op, uint8 imm) { if (!op.isBit(128|256|512)) throw Error(ERR_BAD_MEM_SIZE); opVex(k.changeBit(op.getBit()), 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_YMM | T_EW1 | T_B64, 0x66, imm); } void vfpclasspd(const Opmask& k, const Operand& op, uint8 imm) { if (!op.isBit(128|256|512)) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(k.changeBit(op.getBit()), 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_YMM | T_EW1 | T_B64, 0x66, imm); }
void vfpclassps(const Opmask& k, const Operand& op, uint8 imm) { if (!op.isBit(128|256|512)) throw Error(ERR_BAD_MEM_SIZE); opVex(k.changeBit(op.getBit()), 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_YMM | T_EW0 | T_B32, 0x66, imm); } void vfpclassps(const Opmask& k, const Operand& op, uint8 imm) { if (!op.isBit(128|256|512)) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(k.changeBit(op.getBit()), 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_YMM | T_EW0 | T_B32, 0x66, imm); }
void vfpclasssd(const Opmask& k, const Operand& op, uint8 imm) { if (!op.isXMEM()) throw Error(ERR_BAD_MEM_SIZE); opVex(k, 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_EW1 | T_N8, 0x67, imm); } void vfpclasssd(const Opmask& k, const Operand& op, uint8 imm) { if (!op.isXMEM()) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(k, 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_EW1 | T_N8, 0x67, imm); }
void vfpclassss(const Opmask& k, const Operand& op, uint8 imm) { if (!op.isXMEM()) throw Error(ERR_BAD_MEM_SIZE); opVex(k, 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_EW0 | T_N4, 0x67, imm); } void vfpclassss(const Opmask& k, const Operand& op, uint8 imm) { if (!op.isXMEM()) XBYAK_THROW(ERR_BAD_MEM_SIZE) opVex(k, 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_EW0 | T_N4, 0x67, imm); }
void vgatherdpd(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_VSIB, 0x92, 1); } void vgatherdpd(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N8 | T_66 | T_0F38 | T_EW1 | T_YMM | T_MUST_EVEX | T_VSIB, 0x92, 1); }
void vgatherdps(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_VSIB, 0x92, 0); } void vgatherdps(const Xmm& x, const Address& addr) { opGather2(x, addr, T_N4 | T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_VSIB, 0x92, 0); }
void vgatherpf0dpd(const Address& addr) { opGatherFetch(addr, zm1, T_N8 | T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_M_K | T_VSIB, 0xC6, Operand::YMM); } void vgatherpf0dpd(const Address& addr) { opGatherFetch(addr, zm1, T_N8 | T_66 | T_0F38 | T_EW1 | T_MUST_EVEX | T_M_K | T_VSIB, 0xC6, Operand::YMM); }
@ -1813,14 +1825,14 @@ void vgetmantpd(const Xmm& x, const Operand& op, uint8 imm) { opAVX_X_XM_IMM(x,
void vgetmantps(const Xmm& x, const Operand& op, uint8 imm) { opAVX_X_XM_IMM(x, op, T_66 | T_0F3A | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B32, 0x26, imm); } void vgetmantps(const Xmm& x, const Operand& op, uint8 imm) { opAVX_X_XM_IMM(x, op, T_66 | T_0F3A | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX | T_B32, 0x26, imm); }
void vgetmantsd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F3A | T_EW1 | T_SAE_X | T_MUST_EVEX, 0x27, imm); } void vgetmantsd(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F3A | T_EW1 | T_SAE_X | T_MUST_EVEX, 0x27, imm); }
void vgetmantss(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_N4 | T_66 | T_0F3A | T_EW0 | T_SAE_X | T_MUST_EVEX, 0x27, imm); } void vgetmantss(const Xmm& x1, const Xmm& x2, const Operand& op, uint8 imm) { opAVX_X_X_XM(x1, x2, op, T_N4 | T_66 | T_0F3A | T_EW0 | T_SAE_X | T_MUST_EVEX, 0x27, imm); }
void vinsertf32x4(const Ymm& r1, const Ymm& r2, const Operand& op, uint8 imm) {if (!(r1.getKind() == r2.getKind() && op.is(Operand::MEM | Operand::XMM))) throw Error(ERR_BAD_COMBINATION); opVex(r1, &r2, op, T_N16 | T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x18, imm); } void vinsertf32x4(const Ymm& r1, const Ymm& r2, const Operand& op, uint8 imm) {if (!(r1.getKind() == r2.getKind() && op.is(Operand::MEM | Operand::XMM))) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r1, &r2, op, T_N16 | T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x18, imm); }
void vinsertf32x8(const Zmm& r1, const Zmm& r2, const Operand& op, uint8 imm) {if (!op.is(Operand::MEM | Operand::YMM)) throw Error(ERR_BAD_COMBINATION); opVex(r1, &r2, op, T_N32 | T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x1A, imm); } void vinsertf32x8(const Zmm& r1, const Zmm& r2, const Operand& op, uint8 imm) {if (!op.is(Operand::MEM | Operand::YMM)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r1, &r2, op, T_N32 | T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x1A, imm); }
void vinsertf64x2(const Ymm& r1, const Ymm& r2, const Operand& op, uint8 imm) {if (!(r1.getKind() == r2.getKind() && op.is(Operand::MEM | Operand::XMM))) throw Error(ERR_BAD_COMBINATION); opVex(r1, &r2, op, T_N16 | T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x18, imm); } void vinsertf64x2(const Ymm& r1, const Ymm& r2, const Operand& op, uint8 imm) {if (!(r1.getKind() == r2.getKind() && op.is(Operand::MEM | Operand::XMM))) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r1, &r2, op, T_N16 | T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x18, imm); }
void vinsertf64x4(const Zmm& r1, const Zmm& r2, const Operand& op, uint8 imm) {if (!op.is(Operand::MEM | Operand::YMM)) throw Error(ERR_BAD_COMBINATION); opVex(r1, &r2, op, T_N32 | T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x1A, imm); } void vinsertf64x4(const Zmm& r1, const Zmm& r2, const Operand& op, uint8 imm) {if (!op.is(Operand::MEM | Operand::YMM)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r1, &r2, op, T_N32 | T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x1A, imm); }
void vinserti32x4(const Ymm& r1, const Ymm& r2, const Operand& op, uint8 imm) {if (!(r1.getKind() == r2.getKind() && op.is(Operand::MEM | Operand::XMM))) throw Error(ERR_BAD_COMBINATION); opVex(r1, &r2, op, T_N16 | T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x38, imm); } void vinserti32x4(const Ymm& r1, const Ymm& r2, const Operand& op, uint8 imm) {if (!(r1.getKind() == r2.getKind() && op.is(Operand::MEM | Operand::XMM))) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r1, &r2, op, T_N16 | T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x38, imm); }
void vinserti32x8(const Zmm& r1, const Zmm& r2, const Operand& op, uint8 imm) {if (!op.is(Operand::MEM | Operand::YMM)) throw Error(ERR_BAD_COMBINATION); opVex(r1, &r2, op, T_N32 | T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x3A, imm); } void vinserti32x8(const Zmm& r1, const Zmm& r2, const Operand& op, uint8 imm) {if (!op.is(Operand::MEM | Operand::YMM)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r1, &r2, op, T_N32 | T_66 | T_0F3A | T_EW0 | T_YMM | T_MUST_EVEX, 0x3A, imm); }
void vinserti64x2(const Ymm& r1, const Ymm& r2, const Operand& op, uint8 imm) {if (!(r1.getKind() == r2.getKind() && op.is(Operand::MEM | Operand::XMM))) throw Error(ERR_BAD_COMBINATION); opVex(r1, &r2, op, T_N16 | T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x38, imm); } void vinserti64x2(const Ymm& r1, const Ymm& r2, const Operand& op, uint8 imm) {if (!(r1.getKind() == r2.getKind() && op.is(Operand::MEM | Operand::XMM))) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r1, &r2, op, T_N16 | T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x38, imm); }
void vinserti64x4(const Zmm& r1, const Zmm& r2, const Operand& op, uint8 imm) {if (!op.is(Operand::MEM | Operand::YMM)) throw Error(ERR_BAD_COMBINATION); opVex(r1, &r2, op, T_N32 | T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x3A, imm); } void vinserti64x4(const Zmm& r1, const Zmm& r2, const Operand& op, uint8 imm) {if (!op.is(Operand::MEM | Operand::YMM)) XBYAK_THROW(ERR_BAD_COMBINATION) opVex(r1, &r2, op, T_N32 | T_66 | T_0F3A | T_EW1 | T_YMM | T_MUST_EVEX, 0x3A, imm); }
void vmovdqa32(const Address& addr, const Xmm& x) { opAVX_X_XM_IMM(x, addr, T_66 | T_0F | T_EW0 | T_YMM | T_ER_X | T_ER_Y | T_ER_Z | T_MUST_EVEX | T_M_K, 0x7F); } void vmovdqa32(const Address& addr, const Xmm& x) { opAVX_X_XM_IMM(x, addr, T_66 | T_0F | T_EW0 | T_YMM | T_ER_X | T_ER_Y | T_ER_Z | T_MUST_EVEX | T_M_K, 0x7F); }
void vmovdqa32(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F | T_EW0 | T_YMM | T_ER_X | T_ER_Y | T_ER_Z | T_MUST_EVEX, 0x6F); } void vmovdqa32(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F | T_EW0 | T_YMM | T_ER_X | T_ER_Y | T_ER_Z | T_MUST_EVEX, 0x6F); }
void vmovdqa64(const Address& addr, const Xmm& x) { opAVX_X_XM_IMM(x, addr, T_66 | T_0F | T_EW1 | T_YMM | T_ER_X | T_ER_Y | T_ER_Z | T_MUST_EVEX | T_M_K, 0x7F); } void vmovdqa64(const Address& addr, const Xmm& x) { opAVX_X_XM_IMM(x, addr, T_66 | T_0F | T_EW1 | T_YMM | T_ER_X | T_ER_Y | T_ER_Z | T_MUST_EVEX | T_M_K, 0x7F); }
@ -1833,8 +1845,8 @@ void vmovdqu64(const Address& addr, const Xmm& x) { opAVX_X_XM_IMM(x, addr, T_F3
void vmovdqu64(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_F3 | T_0F | T_EW1 | T_YMM | T_ER_X | T_ER_Y | T_ER_Z | T_MUST_EVEX, 0x6F); } void vmovdqu64(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_F3 | T_0F | T_EW1 | T_YMM | T_ER_X | T_ER_Y | T_ER_Z | T_MUST_EVEX, 0x6F); }
void vmovdqu8(const Address& addr, const Xmm& x) { opAVX_X_XM_IMM(x, addr, T_F2 | T_0F | T_EW0 | T_YMM | T_ER_X | T_ER_Y | T_ER_Z | T_MUST_EVEX | T_M_K, 0x7F); } void vmovdqu8(const Address& addr, const Xmm& x) { opAVX_X_XM_IMM(x, addr, T_F2 | T_0F | T_EW0 | T_YMM | T_ER_X | T_ER_Y | T_ER_Z | T_MUST_EVEX | T_M_K, 0x7F); }
void vmovdqu8(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_F2 | T_0F | T_EW0 | T_YMM | T_ER_X | T_ER_Y | T_ER_Z | T_MUST_EVEX, 0x6F); } void vmovdqu8(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_F2 | T_0F | T_EW0 | T_YMM | T_ER_X | T_ER_Y | T_ER_Z | T_MUST_EVEX, 0x6F); }
void vp2intersectd(const Opmask& k, const Xmm& x, const Operand& op) { if (k.getOpmaskIdx() != 0) throw Error(ERR_OPMASK_IS_ALREADY_SET); opAVX_K_X_XM(k, x, op, T_F2 | T_0F38 | T_YMM | T_EVEX | T_EW0 | T_B32, 0x68); } void vp2intersectd(const Opmask& k, const Xmm& x, const Operand& op) { if (k.getOpmaskIdx() != 0) XBYAK_THROW(ERR_OPMASK_IS_ALREADY_SET) opAVX_K_X_XM(k, x, op, T_F2 | T_0F38 | T_YMM | T_EVEX | T_EW0 | T_B32, 0x68); }
void vp2intersectq(const Opmask& k, const Xmm& x, const Operand& op) { if (k.getOpmaskIdx() != 0) throw Error(ERR_OPMASK_IS_ALREADY_SET); opAVX_K_X_XM(k, x, op, T_F2 | T_0F38 | T_YMM | T_EVEX | T_EW1 | T_B64, 0x68); } void vp2intersectq(const Opmask& k, const Xmm& x, const Operand& op) { if (k.getOpmaskIdx() != 0) XBYAK_THROW(ERR_OPMASK_IS_ALREADY_SET) opAVX_K_X_XM(k, x, op, T_F2 | T_0F38 | T_YMM | T_EVEX | T_EW1 | T_B64, 0x68); }
void vp4dpwssd(const Zmm& z1, const Zmm& z2, const Address& addr) { opAVX_X_X_XM(z1, z2, addr, T_0F38 | T_F2 | T_EW0 | T_YMM | T_MUST_EVEX | T_N16, 0x52); } void vp4dpwssd(const Zmm& z1, const Zmm& z2, const Address& addr) { opAVX_X_X_XM(z1, z2, addr, T_0F38 | T_F2 | T_EW0 | T_YMM | T_MUST_EVEX | T_N16, 0x52); }
void vp4dpwssds(const Zmm& z1, const Zmm& z2, const Address& addr) { opAVX_X_X_XM(z1, z2, addr, T_0F38 | T_F2 | T_EW0 | T_YMM | T_MUST_EVEX | T_N16, 0x53); } void vp4dpwssds(const Zmm& z1, const Zmm& z2, const Address& addr) { opAVX_X_X_XM(z1, z2, addr, T_0F38 | T_F2 | T_EW0 | T_YMM | T_MUST_EVEX | T_N16, 0x53); }
void vpabsq(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F38 | T_MUST_EVEX | T_EW1 | T_B64 | T_YMM, 0x1F); } void vpabsq(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F38 | T_MUST_EVEX | T_EW1 | T_B64 | T_YMM, 0x1F); }

26
src/xbyak/xbyak_util.h
Unescape View File

@ -220,23 +220,23 @@ public:
int displayModel; // model + extModel int displayModel; // model + extModel
unsigned int getNumCores(IntelCpuTopologyLevel level) { unsigned int getNumCores(IntelCpuTopologyLevel level) {
if (!x2APIC_supported_) throw Error(ERR_X2APIC_IS_NOT_SUPPORTED); if (!x2APIC_supported_) XBYAK_THROW_RET(ERR_X2APIC_IS_NOT_SUPPORTED, 0)
switch (level) { switch (level) {
case SmtLevel: return numCores_[level - 1]; case SmtLevel: return numCores_[level - 1];
case CoreLevel: return numCores_[level - 1] / numCores_[SmtLevel - 1]; case CoreLevel: return numCores_[level - 1] / numCores_[SmtLevel - 1];
default: throw Error(ERR_X2APIC_IS_NOT_SUPPORTED); default: XBYAK_THROW_RET(ERR_X2APIC_IS_NOT_SUPPORTED, 0)
} }
} }
unsigned int getDataCacheLevels() const { return dataCacheLevels_; } unsigned int getDataCacheLevels() const { return dataCacheLevels_; }
unsigned int getCoresSharingDataCache(unsigned int i) const unsigned int getCoresSharingDataCache(unsigned int i) const
{ {
if (i >= dataCacheLevels_) throw Error(ERR_BAD_PARAMETER); if (i >= dataCacheLevels_) XBYAK_THROW_RET(ERR_BAD_PARAMETER, 0)
return coresSharignDataCache_[i]; return coresSharignDataCache_[i];
} }
unsigned int getDataCacheSize(unsigned int i) const unsigned int getDataCacheSize(unsigned int i) const
{ {
if (i >= dataCacheLevels_) throw Error(ERR_BAD_PARAMETER); if (i >= dataCacheLevels_) XBYAK_THROW_RET(ERR_BAD_PARAMETER, 0)
return dataCacheSize_[i]; return dataCacheSize_[i];
} }
@ -353,6 +353,9 @@ public:
static const Type tAVX512_VPOPCNTDQ = uint64(1) << 56; static const Type tAVX512_VPOPCNTDQ = uint64(1) << 56;
static const Type tAVX512_BF16 = uint64(1) << 57; static const Type tAVX512_BF16 = uint64(1) << 57;
static const Type tAVX512_VP2INTERSECT = uint64(1) << 58; static const Type tAVX512_VP2INTERSECT = uint64(1) << 58;
static const Type tAMX_TILE = uint64(1) << 59;
static const Type tAMX_INT8 = uint64(1) << 60;
static const Type tAMX_BF16 = uint64(1) << 61;
Cpu() Cpu()
: type_(NONE) : type_(NONE)
@ -456,6 +459,9 @@ public:
if (EBX & (1U << 14)) type_ |= tMPX; if (EBX & (1U << 14)) type_ |= tMPX;
if (EBX & (1U << 29)) type_ |= tSHA; if (EBX & (1U << 29)) type_ |= tSHA;
if (ECX & (1U << 0)) type_ |= tPREFETCHWT1; if (ECX & (1U << 0)) type_ |= tPREFETCHWT1;
if (EDX & (1U << 24)) type_ |= tAMX_TILE;
if (EDX & (1U << 25)) type_ |= tAMX_INT8;
if (EDX & (1U << 22)) type_ |= tAMX_BF16;
} }
setFamily(); setFamily();
setNumCores(); setNumCores();
@ -558,7 +564,7 @@ public:
{ {
if (n_ == maxTblNum) { if (n_ == maxTblNum) {
fprintf(stderr, "ERR Pack::can't append\n"); fprintf(stderr, "ERR Pack::can't append\n");
throw Error(ERR_BAD_PARAMETER); XBYAK_THROW_RET(ERR_BAD_PARAMETER, *this)
} }
tbl_[n_++] = &t; tbl_[n_++] = &t;
return *this; return *this;
@ -567,7 +573,7 @@ public:
{ {
if (n > maxTblNum) { if (n > maxTblNum) {
fprintf(stderr, "ERR Pack::init bad n=%d\n", (int)n); fprintf(stderr, "ERR Pack::init bad n=%d\n", (int)n);
throw Error(ERR_BAD_PARAMETER); XBYAK_THROW(ERR_BAD_PARAMETER)
} }
n_ = n; n_ = n;
for (size_t i = 0; i < n; i++) { for (size_t i = 0; i < n; i++) {
@ -578,7 +584,7 @@ public:
{ {
if (n >= n_) { if (n >= n_) {
fprintf(stderr, "ERR Pack bad n=%d(%d)\n", (int)n, (int)n_); fprintf(stderr, "ERR Pack bad n=%d(%d)\n", (int)n, (int)n_);
throw Error(ERR_BAD_PARAMETER); XBYAK_THROW_RET(ERR_BAD_PARAMETER, rax)
} }
return *tbl_[n]; return *tbl_[n];
} }
@ -591,7 +597,7 @@ public:
if (num == size_t(-1)) num = n_ - pos; if (num == size_t(-1)) num = n_ - pos;
if (pos + num > n_) { if (pos + num > n_) {
fprintf(stderr, "ERR Pack::sub bad pos=%d, num=%d\n", (int)pos, (int)num); fprintf(stderr, "ERR Pack::sub bad pos=%d, num=%d\n", (int)pos, (int)num);
throw Error(ERR_BAD_PARAMETER); XBYAK_THROW_RET(ERR_BAD_PARAMETER, Pack())
} }
Pack pack; Pack pack;
pack.n_ = num; pack.n_ = num;
@ -666,9 +672,9 @@ public:
, t(t_) , t(t_)
{ {
using namespace Xbyak; using namespace Xbyak;
if (pNum < 0 || pNum > 4) throw Error(ERR_BAD_PNUM); if (pNum < 0 || pNum > 4) XBYAK_THROW(ERR_BAD_PNUM)
const int allRegNum = pNum + tNum_ + (useRcx_ ? 1 : 0) + (useRdx_ ? 1 : 0); const int allRegNum = pNum + tNum_ + (useRcx_ ? 1 : 0) + (useRdx_ ? 1 : 0);
if (tNum_ < 0 || allRegNum > maxRegNum) throw Error(ERR_BAD_TNUM); if (tNum_ < 0 || allRegNum > maxRegNum) XBYAK_THROW(ERR_BAD_TNUM)
const Reg64& _rsp = code->rsp; const Reg64& _rsp = code->rsp;
saveNum_ = (std::max)(0, allRegNum - noSaveNum); saveNum_ = (std::max)(0, allRegNum - noSaveNum);
const int *tbl = getOrderTbl() + noSaveNum; const int *tbl = getOrderTbl() + noSaveNum;

Write Preview
Loading…
Cancel
Save