8042dc93e8
0140eeff Merge branch 'dev' 1efe14b2 change the original behavior of SetError 83c89c7a rename and fix indent 8be7ca93 Merge branch 'sbogusev-master' into dev 070b4c09 make l_err() inline with block scope static TLS l_error 9a4e6579 v5.97 d0ced1bc XBYAK_ONLY_CLASS_CPU is for only util::Cpu bb967ae7 replace uint32 with uint32_t etc. c306b8e5 update to v5.95 605e4224 use noexcept if C++11 or later 7a17c2c8 remove warning 5dfa4462 use constexpr if c++14 or later 18c9caaa Merge branch 'densamoilov-fix-mov-interface' into dev 3966ba9d fix mov interface be492be1 change the behavior of push((byte|word), imm) to cast imm to int8_t/int16_t d9696b54 Merge pull request #102 from igorsafo/master ea73267f Cpu: make getNumCores constant ff0b10e9 Merge pull request #101 from densamoilov/use-thread_local-when-supported 0c4eafc3 use thread_local for XBYAK_TLS when supported c1aea35e CodeGenerator::reset() calls ClearError() b4df97b1 Merge branch 'cursey-no-winsock2-header' 6a47bb0e v5.94 9a1749e6 define WIN32_LEAN_AND_MEAN for including winsock2.h after xbyak.h 42dddb74 Remove #include <winsock2.h> 615b85fa update doc 9cd796a9 rename XBYAK_NOEXCEPTION to XBYAK_NO_EXCEPTION 7cdf227f use static to avoid multiple instance 38a28dec test_nm.bat supports noexcept 0fdffc6b XBYAK_NOEXCEPTION for -fno-exceptions eda6e2a3 v5.92 5c26c8bb mov(rax, imm64) on 32-bit env with XBYAK64 6208e3ae throw exception if not supported amx sibmem 2 c6737d14 mov amx insts from avx512 34ea5c16 throw exception if not supported amx sibmem 6f93fe35 fix test of sizeof(Operand) 5b89c3b2 remove T_TMM 5ce32858 gen_amx.cpp is merged into gen_avx512.cpp fe4f965f remove my alias for tmm registers 92f904d8 bit_ contains 8192 98b51da9 extend mnemonics with Intel(R) AMX ISA 8d1b4c9e add generation of Intel(R) AMX ISA mnemonics 8ded45d1 add support of Intel(R) AMX ISA b23c4b02 v5.912 ffe32a60 Merge branch 'rsdubtso-master' e7b7fd2f use MAP_JIT on macOS regardless of Xcode version 82b70e66 v5.911 ; XBYAK_USE_MMAP_ALLOCATOR is defined 2f6d9e34 fix test for mac a7d10a1e add link to GitHub Sponsor 96076265 accept k0 mask register (it means no mask) 7e3167e4 kmov{b,w,d,q} throws for unsupported reg f487d7b7 Merge pull request #91 from marcelotrevisani/patch-1 dc9e6a79 Possibility to specify a different PREFIX 5fc69fc8 remove warning of test e69e0b42 fix typo of type of Zmi 34f797e8 perf does not recognize too short function name 6cc0f4df Consider max defined as a macro on Windows 5722393d fix for zeroed-out 0xb leaf 6a4459a8 Merge branch 'tyfkda-feature/fix-segfault-in-calc' 47922ed9 Fix segmentation fault in calc sample 8f696e93 add test_avx512 to bat 00114d79 add .travis.yml a29fa27b refactor test 508b543c fix error of vfpclasspd 0d54f1b1 fix for windows 4da8fd4e add setDefaultJmpNEAR da7f7317 revert to the behavior before v5.84 if -fno-operator-names is defined 7dac9f61 update to v5.85 fe639332 enable MAP_JIT only if mojave or later 4443d791 specify MAP_JIT mmap flag on macOS 20ee4c2d update doc ca0e8395 [changed] XBYAK_NO_OP_NAMES is defined f32836da remove exit(1) a1e9adf2 v5.82 08b8b1ba Support AMD Zen New Instructions. 2501ba9a remove *.user and *.vcproj 5c2ea988 Merge branch 'jrmwng-feature/upgrade-to-vs2017/jrmwng' 35847f7a Merge branch 'feature/upgrade-to-vs2017/jrmwng' of https://github.com/jrmwng/xbyak into jrmwng-feature/upgrade-to-vs2017/jrmwng ef267775 address "warning LNK4075: ignoring '/EDITANDCONTINUE' due to '/SAFESEH' specification" 4a6c59bb address a conflict of sharing intermediate directory by different projects 9577cbf3 inherit "some output locations" from parent or project defaults 6c5f7186 upgrade projects from VS2018 to VS2017 4ca0434b v5.81 72b4e95d add lds/lss/les/lfs/lgs cc8f037c fix ; move ERR_INTERNAL to the end 9e9ec1c3 add repe, repne, repne, prez eea0edc3 add some fpu mnemonics 06235fa6 add loop/loope/loopne 7fc0c2bb add enter/leave 9fa2ef3c add in_, out_ df208648 add lods{b,w,d,q}, outs{b,w,d} 4672d2cb add int3, int_, into 431977cb add pushfq, popfq 81c4749f syscall, sysenter, sysexit, sysret 1f1b53c4 add clflushopt, fldenv, fnstw b765db33 Profiler uses append mode 44dc3546 add Profiler class 42949334 update version to v5.802 91cb919b Merge branch 'vpirogov-master' a6452f82 fixed avx512_bf16 detection f41da5aa tweak ; vcvtneps2bf16 calls opCvt2 b12460ba [sample] fix typo of quantize.cpp b22f5881 add set_opt.bat for test on Windows f402faad add vp2intersectd/vp2intersectq 4cfd5208 add avx512_bf16 4033564c fix vcmppd/vcmpps for ptr_b git-subtree-dir: externals/xbyak git-subtree-split: 0140eeff1fffcf5069dea3abb57095695320971c
2200 lines
62 KiB
C++
2200 lines
62 KiB
C++
#include <stdio.h>
|
|
#include "xbyak/xbyak.h"
|
|
#include <stdlib.h>
|
|
#include <string.h>
|
|
#include "cybozu/inttype.hpp"
|
|
#define NUM_OF_ARRAY(x) (sizeof(x) / sizeof(x[0]))
|
|
|
|
using namespace Xbyak;
|
|
|
|
const int bitEnd = 64;
|
|
|
|
const uint64_t YMM_SAE = 1ULL << 0;
|
|
const uint64_t _XMM = 1ULL << 1;
|
|
const uint64_t _MEM = 1ULL << 2;
|
|
const uint64_t _REG32 = 1ULL << 3;
|
|
const uint64_t EAX = 1ULL << 4;
|
|
const uint64_t IMM32 = 1ULL << 5;
|
|
const uint64_t IMM8 = 1ULL << 6;
|
|
const uint64_t _REG8 = 1ULL << 7;
|
|
const uint64_t _REG16 = 1ULL << 8;
|
|
const uint64_t XMM_K = 1ULL << 9;
|
|
const uint64_t YMM_K = 1ULL << 10;
|
|
const uint64_t ZMM_K = 1ULL << 11;
|
|
const uint64_t AX = 1ULL << 12;
|
|
const uint64_t AL = 1ULL << 13;
|
|
const uint64_t IMM_1 = 1ULL << 14;
|
|
const uint64_t MEM8 = 1ULL << 15;
|
|
const uint64_t MEM16 = 1ULL << 16;
|
|
const uint64_t MEM32 = 1ULL << 17;
|
|
const uint64_t VM32Z = 1ULL << 19;
|
|
const uint64_t K_K = 1ULL << 20;
|
|
const uint64_t MEM_ONLY_DISP = 1ULL << 21;
|
|
const uint64_t VM32X_K = 1ULL << 23;
|
|
const uint64_t _YMM = 1ULL << 24;
|
|
const uint64_t VM32X_32 = 1ULL << 39;
|
|
const uint64_t VM32X_64 = 1ULL << 40;
|
|
const uint64_t VM32Y_32 = 1ULL << 41;
|
|
const uint64_t VM32Y_64 = 1ULL << 42;
|
|
const uint64_t VM32Z_K = 1ULL << 32;
|
|
#ifdef XBYAK64
|
|
const uint64_t _MEMe = 1ULL << 25;
|
|
const uint64_t REG32_2 = 1ULL << 26; // r8d, ...
|
|
const uint64_t REG16_2 = 1ULL << 27; // r8w, ...
|
|
const uint64_t REG8_2 = 1ULL << 28; // r8b, ...
|
|
const uint64_t REG8_3 = 1ULL << 29; // spl, ...
|
|
const uint64_t _REG64 = 1ULL << 30; // rax, ...
|
|
const uint64_t _REG64_2 = 1ULL << 31; // r8, ...
|
|
const uint64_t _XMM2 = 1ULL << 33;
|
|
const uint64_t _YMM2 = 1ULL << 34;
|
|
const uint64_t VM32X = VM32X_32 | VM32X_64;
|
|
const uint64_t VM32Y = VM32Y_32 | VM32Y_64;
|
|
#else
|
|
const uint64_t _MEMe = 0;
|
|
const uint64_t REG32_2 = 0;
|
|
const uint64_t REG16_2 = 0;
|
|
const uint64_t REG8_2 = 0;
|
|
const uint64_t REG8_3 = 0;
|
|
const uint64_t _REG64 = 0;
|
|
const uint64_t _REG64_2 = 0;
|
|
const uint64_t _XMM2 = 0;
|
|
const uint64_t _YMM2 = 0;
|
|
const uint64_t VM32X = VM32X_32;
|
|
const uint64_t VM32Y = VM32Y_32;
|
|
#endif
|
|
const uint64_t REG64 = _REG64 | _REG64_2;
|
|
const uint64_t REG32 = _REG32 | REG32_2 | EAX;
|
|
const uint64_t REG16 = _REG16 | REG16_2 | AX;
|
|
const uint64_t REG32e = REG32 | REG64;
|
|
const uint64_t REG8 = _REG8 | REG8_2|AL;
|
|
const uint64_t MEM = _MEM | _MEMe;
|
|
const uint64_t MEM64 = 1ULL << 35;
|
|
const uint64_t YMM_ER = 1ULL << 36;
|
|
const uint64_t VM32Y_K = 1ULL << 37;
|
|
const uint64_t IMM_2 = 1ULL << 38;
|
|
const uint64_t IMM = IMM_1 | IMM_2;
|
|
const uint64_t YMM = _YMM | _YMM2;
|
|
const uint64_t K = 1ULL << 43;
|
|
const uint64_t _ZMM = 1ULL << 44;
|
|
const uint64_t _ZMM2 = 1ULL << 45;
|
|
#ifdef XBYAK64
|
|
const uint64_t ZMM = _ZMM | _ZMM2;
|
|
const uint64_t _YMM3 = 1ULL << 46;
|
|
#else
|
|
const uint64_t ZMM = _ZMM;
|
|
const uint64_t _YMM3 = 0;
|
|
#endif
|
|
const uint64_t K2 = 1ULL << 47;
|
|
const uint64_t ZMM_SAE = 1ULL << 48;
|
|
const uint64_t ZMM_ER = 1ULL << 49;
|
|
#ifdef XBYAK64
|
|
const uint64_t _XMM3 = 1ULL << 50;
|
|
#else
|
|
const uint64_t _XMM3 = 0;
|
|
#endif
|
|
const uint64_t XMM = _XMM | _XMM2 | _XMM3;
|
|
const uint64_t XMM_SAE = 1ULL << 51;
|
|
#ifdef XBYAK64
|
|
const uint64_t XMM_KZ = 1ULL << 52;
|
|
const uint64_t YMM_KZ = 1ULL << 53;
|
|
const uint64_t ZMM_KZ = 1ULL << 54;
|
|
#else
|
|
const uint64_t XMM_KZ = 0;
|
|
const uint64_t YMM_KZ = 0;
|
|
const uint64_t ZMM_KZ = 0;
|
|
#endif
|
|
const uint64_t MEM_K = 1ULL << 55;
|
|
const uint64_t M_1to2 = 1ULL << 56;
|
|
const uint64_t M_1to4 = 1ULL << 57;
|
|
const uint64_t M_1to8 = 1ULL << 58;
|
|
const uint64_t M_1to16 = 1ULL << 59;
|
|
const uint64_t XMM_ER = 1ULL << 60;
|
|
const uint64_t M_xword = 1ULL << 61;
|
|
const uint64_t M_yword = 1ULL << 62;
|
|
const uint64_t MY_1to4 = 1ULL << 18;
|
|
|
|
const uint64_t NOPARA = 1ULL << (bitEnd - 1);
|
|
|
|
class Test {
|
|
Test(const Test&);
|
|
void operator=(const Test&);
|
|
const bool isXbyak_;
|
|
int funcNum_;
|
|
// check all op1, op2, op3
|
|
void put(const std::string& nm, uint64_t op1 = NOPARA, uint64_t op2 = NOPARA, uint64_t op3 = NOPARA, uint64_t op4 = NOPARA) const
|
|
{
|
|
for (int i = 0; i < bitEnd; i++) {
|
|
if ((op1 & (1ULL << i)) == 0) continue;
|
|
for (int j = 0; j < bitEnd; j++) {
|
|
if ((op2 & (1ULL << j)) == 0) continue;
|
|
for (int k = 0; k < bitEnd; k++) {
|
|
if ((op3 & (1ULL << k)) == 0) continue;
|
|
for (int s = 0; s < bitEnd; s++) {
|
|
if ((op4 & (1ULL << s)) == 0) continue;
|
|
printf("%s ", nm.c_str());
|
|
if (isXbyak_) printf("(");
|
|
if (!(op1 & NOPARA)) printf("%s", get(1ULL << i));
|
|
if (!(op2 & NOPARA)) printf(", %s", get(1ULL << j));
|
|
if (!(op3 & NOPARA)) printf(", %s", get(1ULL << k));
|
|
if (!(op4 & NOPARA)) printf(", %s", get(1ULL << s));
|
|
if (isXbyak_) printf("); dump();");
|
|
printf("\n");
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
void put(const char *nm, uint64_t op, const char *xbyak, const char *nasm) const
|
|
{
|
|
for (int i = 0; i < bitEnd; i++) {
|
|
if ((op & (1ULL << i)) == 0) continue;
|
|
printf("%s ", nm);
|
|
if (isXbyak_) printf("(");
|
|
if (!(op & NOPARA)) printf("%s", get(1ULL << i));
|
|
printf(", %s", isXbyak_ ? xbyak : nasm);
|
|
if (isXbyak_) printf("); dump();");
|
|
printf("\n");
|
|
}
|
|
}
|
|
void put(const char *nm, const char *xbyak, const char *nasm = 0, uint64_t op = NOPARA) const
|
|
{
|
|
if (nasm == 0) nasm = xbyak;
|
|
for (int i = 0; i < bitEnd; i++) {
|
|
if ((op & (1ULL << i)) == 0) continue;
|
|
printf("%s ", nm);
|
|
if (isXbyak_) printf("(");
|
|
printf("%s ", isXbyak_ ? xbyak : nasm);
|
|
if (!(op & NOPARA)) printf(", %s", get(1ULL << i));
|
|
if (isXbyak_) printf("); dump();");
|
|
printf("\n");
|
|
}
|
|
}
|
|
const char *get(uint64_t type) const
|
|
{
|
|
int idx = (rand() / 31) & 7;
|
|
switch (type) {
|
|
case _XMM:
|
|
{
|
|
static const char tbl[][6] = {
|
|
"xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6", "xmm7",
|
|
};
|
|
return tbl[idx];
|
|
}
|
|
case _YMM:
|
|
{
|
|
static const char tbl[][6] = {
|
|
"ymm0", "ymm1", "ymm2", "ymm3", "ymm4", "ymm5", "ymm6", "ymm7"
|
|
};
|
|
return tbl[idx];
|
|
}
|
|
case _ZMM:
|
|
{
|
|
static const char tbl[][6] = {
|
|
"zmm0", "zmm1", "zmm2", "zmm3", "zmm4", "zmm5", "zmm6", "zmm7"
|
|
};
|
|
return tbl[idx];
|
|
}
|
|
#ifdef XBYAK64
|
|
case _XMM2:
|
|
{
|
|
static const char tbl[][6] = {
|
|
"xmm8", "xmm9", "xmm10", "xmm11", "xmm12", "xmm13", "xmm14", "xmm15"
|
|
};
|
|
return tbl[idx];
|
|
}
|
|
case _XMM3:
|
|
{
|
|
static const char tbl[][6] = {
|
|
"xmm16", "xmm17", "xmm18", "xmm19", "xmm20", "xmm21", "xmm22", "xmm23"
|
|
};
|
|
return tbl[idx];
|
|
}
|
|
case _YMM2:
|
|
{
|
|
static const char tbl[][6] = {
|
|
"ymm8", "ymm9", "ymm10", "ymm11", "ymm12", "ymm13", "ymm14", "ymm15",
|
|
};
|
|
return tbl[idx];
|
|
}
|
|
case _YMM3:
|
|
{
|
|
static const char tbl[][6] = {
|
|
"ymm16", "ymm17", "ymm18", "ymm19", "ymm20", "ymm21", "ymm22", "ymm23",
|
|
};
|
|
return tbl[idx];
|
|
}
|
|
case _ZMM2:
|
|
{
|
|
static const char tbl[][6] = {
|
|
"zmm8", "zmm9", "zmm10", "zmm11", "zmm28", "zmm29", "zmm30", "zmm31",
|
|
};
|
|
return tbl[idx];
|
|
}
|
|
#endif
|
|
case _MEM:
|
|
return isXbyak_ ? "ptr[eax+ecx+64]" : "[eax+ecx+64]"; // QQQ
|
|
// return isXbyak_ ? "ptr[eax+ecx+6]" : "[eax+ecx+6]";
|
|
case _MEMe:
|
|
{
|
|
static int ccc = 1;
|
|
#ifdef USE_YASM
|
|
ccc++;
|
|
#endif
|
|
if (ccc & 1) {
|
|
return isXbyak_ ? "ptr[rdx+r15+0x12]" : "[rdx+r15+0x12]";
|
|
} else {
|
|
return isXbyak_ ? "ptr[rip - 0x13456+1-3]" : "[rip - 0x13456+1-3]";
|
|
}
|
|
}
|
|
case MEM8:
|
|
return "byte [eax+edx]";
|
|
case MEM16:
|
|
return "word [esi]";
|
|
case MEM32:
|
|
return "dword [eax+64]";
|
|
case MEM64:
|
|
return "qword [rax+64]";
|
|
case MEM_ONLY_DISP:
|
|
return isXbyak_ ? "ptr[(void*)0x123]" : "[0x123]";
|
|
case _REG16: // not ax
|
|
{
|
|
static const char Reg16Tbl[][4] = {
|
|
"ax", "cx", "dx", "bx", "sp", "bp", "si", "di"
|
|
};
|
|
return Reg16Tbl[(idx % 7) + 1];
|
|
}
|
|
case _REG8: // not al
|
|
{
|
|
static const char Reg8Tbl[][4] = {
|
|
#ifdef XBYAK64 // QQQ
|
|
"al", "cl", "dl", "bl", "al", "cl", "dl", "bl"
|
|
#else
|
|
"al", "cl", "dl", "bl", "ah", "ch", "dh", "bh"
|
|
#endif
|
|
};
|
|
return Reg8Tbl[(idx % 7) + 1];
|
|
}
|
|
case _REG32: // not eax
|
|
{
|
|
static const char Reg32Tbl[][4] = {
|
|
"eax", "ecx", "edx", "ebx", "esp", "ebp", "esi", "edi"
|
|
};
|
|
return Reg32Tbl[(idx % 7) + 1];
|
|
}
|
|
#ifdef XBYAK64
|
|
case _REG64: // not rax
|
|
{
|
|
static const char Reg64Tbl[][4] = {
|
|
"rax", "rcx", "rdx", "rbx", "rsp", "rbp", "rsi", "rdi"
|
|
};
|
|
return Reg64Tbl[(idx % 7) + 1];
|
|
}
|
|
case _REG64_2:
|
|
{
|
|
static const char Reg64_2Tbl[][4] = {
|
|
"r8", "r9", "r10", "r11", "r12", "r13", "r14", "r15"
|
|
};
|
|
return Reg64_2Tbl[idx];
|
|
}
|
|
case REG32_2:
|
|
{
|
|
static const char Reg32eTbl[][5] = {
|
|
"r8d", "r9d", "r10d", "r11d", "r12d", "r13d", "r14d", "r15d"
|
|
};
|
|
return Reg32eTbl[idx];
|
|
}
|
|
case REG16_2:
|
|
{
|
|
static const char Reg16eTbl[][5] = {
|
|
"r8w", "r9w", "r10w", "r11w", "r12w", "r13w", "r14w", "r15w"
|
|
};
|
|
return Reg16eTbl[idx];
|
|
}
|
|
case REG8_2:
|
|
{
|
|
static const char Reg8_2Tbl[][5] = {
|
|
"r8b", "r9b", "r10b", "r11b", "r12b", "r13b", "r14b", "r15b"
|
|
};
|
|
return Reg8_2Tbl[idx];
|
|
}
|
|
case REG8_3:
|
|
{
|
|
static const char Reg8_3Tbl[][5] = {
|
|
"spl", "bpl", "sil", "dil", "spl", "bpl", "sil", "dil"
|
|
};
|
|
return Reg8_3Tbl[idx];
|
|
}
|
|
#endif
|
|
case EAX:
|
|
return "eax";
|
|
case AX:
|
|
return "ax";
|
|
case AL:
|
|
return "al";
|
|
case K_K:
|
|
return isXbyak_ ? "k5 | k3" : "k5{k3}";
|
|
case IMM32:
|
|
return isXbyak_ ? "12345678" : "dword 12345678";
|
|
case IMM8:
|
|
return isXbyak_ ? "4" : "byte 4";
|
|
case IMM_1:
|
|
return "4";
|
|
case IMM_2:
|
|
return isXbyak_ ? "0xda" : "0xda";
|
|
case VM32X_32:
|
|
return isXbyak_ ? "ptr [ebp+64+xmm1*8]" : "[ebp+64+xmm1*8]";
|
|
case VM32X_64:
|
|
return isXbyak_ ? "ptr [rax+64+xmm13*2]" : "[rax+64+xmm13*2]";
|
|
case VM32Y_32:
|
|
return isXbyak_ ? "ptr [ymm4]" : "[ymm4]";
|
|
case VM32Y_64:
|
|
return isXbyak_ ? "ptr [64+ymm13*2+r13]" : "[64+ymm13*2+r13]";
|
|
case VM32X_K:
|
|
return isXbyak_ ? "ptr [64+xmm13*2+r13] | k6" : "[64+xmm13*2+r13]{k6}";
|
|
case VM32Y_K:
|
|
return isXbyak_ ? "ptr [64+ymm13*2+r13] | k6" : "[64+ymm13*2+r13]{k6}";
|
|
case VM32Z_K:
|
|
if (idx & 1) return isXbyak_ ? "ptr [64+zmm10*8+r9] | k6" : "[64+zmm10*8+r9]{k6}";
|
|
return isXbyak_ ? "ptr [64+zmm30*2+r13] | k6" : "[64+zmm30*2+r13]{k6}";
|
|
case VM32Z:
|
|
return isXbyak_ ? "ptr [64+zmm13*2+rcx]" : "[64+zmm13*2+rcx]";
|
|
case M_1to2: return isXbyak_ ? "ptr_b [eax+32]" : "[eax+32]{1to2}";
|
|
case M_1to4: return isXbyak_ ? "ptr_b [eax+32]" : "[eax+32]{1to4}";
|
|
case M_1to8: return isXbyak_ ? "ptr_b [eax+32]" : "[eax+32]{1to8}";
|
|
case M_1to16: return isXbyak_ ? "ptr_b [eax+32]" : "[eax+32]{1to16}";
|
|
|
|
case M_xword: return isXbyak_ ? "ptr [eax+32]" : "oword [eax+32]";
|
|
case M_yword: return isXbyak_ ? "yword [eax+32]" : "yword [eax+32]";
|
|
case MY_1to4: return isXbyak_ ? "yword_b [eax+32]" : "[eax+32]{1to4}";
|
|
case K:
|
|
{
|
|
static const char kTbl[][5] = {
|
|
"k1", "k2", "k3", "k4", "k5", "k6", "k7",
|
|
};
|
|
return kTbl[idx % 7];
|
|
}
|
|
case K2:
|
|
return isXbyak_ ? "k3 | k5" : "k3{k5}";
|
|
#ifdef XBYAK64
|
|
case XMM_SAE:
|
|
return isXbyak_ ? "xmm25 | T_sae" : "xmm25, {sae}";
|
|
case YMM_SAE:
|
|
return isXbyak_ ? "ymm25 | T_sae" : "ymm25, {sae}";
|
|
case ZMM_SAE:
|
|
return isXbyak_ ? "zmm25 | T_sae" : "zmm25, {sae}";
|
|
case XMM_ER:
|
|
return isXbyak_ ? "xmm4 | T_rd_sae" : "xmm4, {rd-sae}";
|
|
case YMM_ER:
|
|
return isXbyak_ ? "ymm20 | T_rd_sae" : "ymm20, {rd-sae}";
|
|
case ZMM_ER:
|
|
return isXbyak_ ? "zmm20 | T_rd_sae" : "zmm20, {rd-sae}";
|
|
case XMM_KZ:
|
|
return isXbyak_ ? "xmm5 | k5" : "xmm5{k5}";
|
|
case YMM_KZ:
|
|
return isXbyak_ ? "ymm2 |k3|T_z" : "ymm2{k3}{z}";
|
|
case ZMM_KZ:
|
|
return isXbyak_ ? "zmm7|k1" : "zmm7{k1}";
|
|
case MEM_K:
|
|
return isXbyak_ ? "ptr [rax] | k1" : "[rax]{k1}";
|
|
#else
|
|
case XMM_SAE:
|
|
return isXbyak_ ? "xmm5 | T_sae" : "xmm5, {sae}";
|
|
case YMM_SAE:
|
|
return isXbyak_ ? "ymm5 | T_sae" : "ymm5, {sae}";
|
|
case ZMM_SAE:
|
|
return isXbyak_ ? "zmm5 | T_sae" : "zmm5, {sae}";
|
|
case XMM_ER:
|
|
return isXbyak_ ? "xmm30 | T_rd_sae" : "xmm30, {rd-sae}";
|
|
case YMM_ER:
|
|
return isXbyak_ ? "ymm2 | T_rd_sae" : "ymm2, {rd-sae}";
|
|
case ZMM_ER:
|
|
return isXbyak_ ? "zmm2 | T_rd_sae" : "zmm2, {rd-sae}";
|
|
case MEM_K:
|
|
return isXbyak_ ? "ptr [eax] | k1" : "[eax]{k1}";
|
|
#endif
|
|
case XMM_K:
|
|
return isXbyak_ ? "xmm5 | k7" : "xmm5{k7}";
|
|
case YMM_K:
|
|
return isXbyak_ ? "ymm5 | k4" : "ymm5{k4}";
|
|
case ZMM_K:
|
|
return isXbyak_ ? "zmm5 | k3" : "zmm5{k3}";
|
|
}
|
|
return 0;
|
|
}
|
|
public:
|
|
Test(bool isXbyak)
|
|
: isXbyak_(isXbyak)
|
|
, funcNum_(1)
|
|
{
|
|
if (!isXbyak_) return;
|
|
printf("%s",
|
|
" void gen0()\n"
|
|
" {\n");
|
|
}
|
|
/*
|
|
gcc and vc give up to compile this source,
|
|
so I split functions.
|
|
*/
|
|
void separateFunc()
|
|
{
|
|
if (!isXbyak_) return;
|
|
printf(
|
|
" }\n"
|
|
" void gen%d()\n"
|
|
" {\n", funcNum_++);
|
|
}
|
|
~Test()
|
|
{
|
|
if (!isXbyak_) return;
|
|
printf("%s",
|
|
" }\n"
|
|
" void gen()\n"
|
|
" {\n");
|
|
for (int i = 0; i < funcNum_; i++) {
|
|
printf(
|
|
" gen%d();\n", i);
|
|
}
|
|
printf(
|
|
" }\n");
|
|
}
|
|
void put()
|
|
{
|
|
putAVX512();
|
|
}
|
|
void putOpmask()
|
|
{
|
|
{
|
|
const char *tbl[] = {
|
|
"kadd",
|
|
"kand",
|
|
"kandn",
|
|
"kor",
|
|
"kxnor",
|
|
"kxor",
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
std::string name = tbl[i];
|
|
put(name + "b", K, K, K);
|
|
put(name + "w", K, K, K);
|
|
put(name + "q", K, K, K);
|
|
put(name + "d", K, K, K);
|
|
}
|
|
put("kunpckbw", K, K, K);
|
|
put("kunpckwd", K, K, K);
|
|
put("kunpckdq", K, K, K);
|
|
}
|
|
{
|
|
const char *tbl[] = {
|
|
"knot",
|
|
"kortest",
|
|
"ktest",
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
std::string name = tbl[i];
|
|
put(name + "b", K, K);
|
|
put(name + "w", K, K);
|
|
put(name + "q", K, K);
|
|
put(name + "d", K, K);
|
|
}
|
|
}
|
|
{
|
|
const char *tbl[] = {
|
|
"kshiftl",
|
|
"kshiftr",
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
std::string name = tbl[i];
|
|
put(name + "b", K, K, IMM8);
|
|
put(name + "w", K, K, IMM8);
|
|
put(name + "q", K, K, IMM8);
|
|
put(name + "d", K, K, IMM8);
|
|
}
|
|
}
|
|
put("kmovw", K, K | MEM | REG32);
|
|
put("kmovq", K, K | MEM);
|
|
put("kmovb", K, K | MEM | REG32);
|
|
put("kmovd", K, K | MEM | REG32);
|
|
|
|
put("kmovw", MEM | REG32, K);
|
|
put("kmovq", MEM, K);
|
|
put("kmovb", MEM | REG32, K);
|
|
put("kmovd", MEM | REG32, K);
|
|
#ifdef XBYAK64
|
|
put("kmovq", K, REG64);
|
|
put("kmovq", REG64, K);
|
|
#endif
|
|
}
|
|
void put_vaddpd(const char *r1, const char *r2, const char *r3, int kIdx = 0, bool z = false, int sae = 0)
|
|
{
|
|
std::string modifier;
|
|
char pk[16] = "";
|
|
const char *pz = "";
|
|
const char *saeTblXbyak[] = { "", "|T_rn_sae", "|T_rd_sae", "|T_ru_sae", "|T_rz_sae" };
|
|
const char *saeTblNASM[] = { "", ",{rn-sae}", ",{rd-sae}", ",{ru-sae}", ",{rz-sae}" };
|
|
if (isXbyak_) {
|
|
if (kIdx) CYBOZU_SNPRINTF(pk, sizeof(pk), "|k%d", kIdx);
|
|
if (z) pz = "|T_z";
|
|
printf("vaddpd(%s%s%s, %s, %s%s); dump();\n", r1, pk, pz, r2, r3, saeTblXbyak[sae]);
|
|
} else {
|
|
if (kIdx) CYBOZU_SNPRINTF(pk, sizeof(pk), "{k%d}", kIdx);
|
|
if (z && kIdx) pz = "{z}";
|
|
printf("vaddpd %s%s%s, %s, %s%s\n", r1, pk, pz, r2, r3, saeTblNASM[sae]);
|
|
}
|
|
}
|
|
void putCombi()
|
|
{
|
|
const char *xTbl[] = {
|
|
"xmm2",
|
|
#ifdef XBYAK64
|
|
"xmm8", "xmm31"
|
|
#else
|
|
"xmm5", "xmm6"
|
|
#endif
|
|
};
|
|
const char *yTbl[] = {
|
|
"ymm0",
|
|
#ifdef XBYAK64
|
|
"ymm15", "ymm31"
|
|
#else
|
|
"ymm4", "ymm2"
|
|
#endif
|
|
};
|
|
const char *zTbl[] = {
|
|
"zmm1",
|
|
#ifdef XBYAK64
|
|
"zmm9", "zmm30"
|
|
#else
|
|
"zmm3", "zmm7"
|
|
#endif
|
|
};
|
|
const size_t N = NUM_OF_ARRAY(zTbl);
|
|
for (size_t i = 0; i < N; i++) {
|
|
for (size_t j = 0; j < N; j++) {
|
|
separateFunc();
|
|
for (size_t k = 0; k < N; k++) {
|
|
#ifdef XBYAK64
|
|
for (int kIdx = 0; kIdx < 8; kIdx++) {
|
|
put_vaddpd(xTbl[i], xTbl[j], xTbl[k], kIdx);
|
|
put_vaddpd(yTbl[i], yTbl[j], yTbl[k], kIdx);
|
|
for (int z = 0; z < 2; z++) {
|
|
for (int sae = 0; sae < 5; sae++) {
|
|
put_vaddpd(zTbl[i], zTbl[j], zTbl[k], kIdx, z == 1, sae);
|
|
}
|
|
}
|
|
}
|
|
#else
|
|
put_vaddpd(xTbl[i], xTbl[j], xTbl[k]);
|
|
put_vaddpd(yTbl[i], yTbl[j], yTbl[k]);
|
|
for (int sae = 0; sae < 5; sae++) {
|
|
put_vaddpd(zTbl[i], zTbl[j], zTbl[k], sae);
|
|
}
|
|
#endif
|
|
}
|
|
}
|
|
}
|
|
put("vaddpd", XMM, XMM, _MEM);
|
|
put("vaddpd", YMM, YMM, _MEM);
|
|
put("vaddpd", ZMM, ZMM, _MEM);
|
|
}
|
|
void putCmpK()
|
|
{
|
|
{
|
|
const struct Tbl {
|
|
const char *name;
|
|
bool supportYMM;
|
|
} tbl[] = {
|
|
{ "vcmppd", true },
|
|
{ "vcmpps", true },
|
|
{ "vcmpsd", false },
|
|
{ "vcmpss", false },
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const Tbl *p = &tbl[i];
|
|
put(p->name, K, XMM, _XMM | MEM, IMM8);
|
|
if (!p->supportYMM) continue;
|
|
put(p->name, K, _YMM, _YMM | MEM, IMM8);
|
|
put(p->name, K, _ZMM, _ZMM | MEM, IMM8);
|
|
}
|
|
put("vcmppd", K, XMM, M_1to2, IMM8);
|
|
put("vcmppd", K, YMM, M_1to4, IMM8);
|
|
put("vcmppd", K, ZMM, M_1to8, IMM8);
|
|
|
|
put("vcmpps", K, XMM, M_1to4, IMM8);
|
|
put("vcmpps", K, YMM, M_1to8, IMM8);
|
|
put("vcmpps", K, ZMM, M_1to16, IMM8);
|
|
}
|
|
put("vcmppd", K2, ZMM, ZMM_SAE, IMM);
|
|
#ifdef XBYAK64
|
|
{
|
|
const struct Tbl {
|
|
const char *name;
|
|
} tbl[] = {
|
|
{ "vcomisd" },
|
|
{ "vcomiss" },
|
|
{ "vucomisd" },
|
|
{ "vucomiss" },
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const Tbl *p = &tbl[i];
|
|
put(p->name, XMM, XMM_SAE | XMM | MEM);
|
|
}
|
|
}
|
|
put("vcomiss", XMM, _XMM3 | MEM);
|
|
put("vcomiss", XMM, XMM_SAE);
|
|
#endif
|
|
}
|
|
void putBroadcastSub(int idx, int disp)
|
|
{
|
|
#ifdef XBYAK64
|
|
const char *a = "rax";
|
|
#else
|
|
const char *a = "eax";
|
|
#endif
|
|
if (isXbyak_) {
|
|
printf("vaddpd(zmm%d, zmm1, ptr_b[%s+%d]);dump();\n", idx, a, disp);
|
|
printf("vaddpd(ymm%d, ymm1, ptr_b[%s+%d]);dump();\n", idx, a, disp);
|
|
printf("vaddpd(xmm%d, xmm1, ptr_b[%s+%d]);dump();\n", idx, a, disp);
|
|
} else {
|
|
printf("vaddpd zmm%d, zmm1, [%s+%d]{1to8}\n", idx, a, disp);
|
|
printf("vaddpd ymm%d, ymm1, [%s+%d]{1to4}\n", idx, a, disp);
|
|
printf("vaddpd xmm%d, xmm1, [%s+%d]{1to2}\n", idx, a, disp);
|
|
}
|
|
}
|
|
void putBroadcast()
|
|
{
|
|
for (int i = 0; i < 9; i++) {
|
|
putBroadcastSub(0, i);
|
|
#ifdef XBYAK64
|
|
putBroadcastSub(10, i);
|
|
putBroadcastSub(20, i);
|
|
#endif
|
|
}
|
|
put("vpbroadcastb", XMM_KZ | ZMM_KZ, REG8 | _MEM);
|
|
put("vpbroadcastw", XMM_KZ | ZMM_KZ, REG16 | _MEM);
|
|
put("vpbroadcastd", XMM_KZ | ZMM_KZ, REG32 | _MEM);
|
|
#ifdef XBYAK64
|
|
put("vpbroadcastq", XMM_KZ | ZMM_KZ, REG64 | _MEM);
|
|
#endif
|
|
{
|
|
const char *tbl[] = {
|
|
"vpbroadcastb",
|
|
"vpbroadcastw",
|
|
"vpbroadcastd",
|
|
"vpbroadcastq",
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
put(tbl[i], XMM_KZ | ZMM_KZ, XMM | _MEM);
|
|
}
|
|
}
|
|
put("vbroadcasti32x2", XMM_KZ | YMM_KZ | ZMM_KZ, XMM | _MEM);
|
|
put("vbroadcasti32x4", YMM_KZ | ZMM_KZ, _MEM);
|
|
put("vbroadcasti64x2", YMM_KZ | ZMM_KZ, _MEM);
|
|
put("vbroadcasti32x8", ZMM_KZ, _MEM);
|
|
put("vbroadcasti64x4", ZMM_KZ, _MEM);
|
|
}
|
|
void putMisc1()
|
|
{
|
|
put("vmaskmovps", _XMM, _XMM, MEM);
|
|
put("vmaskmovps", YMM, YMM, MEM);
|
|
|
|
put("vmaskmovpd", YMM, YMM, MEM);
|
|
put("vmaskmovpd", _XMM, _XMM, MEM);
|
|
|
|
put("vmaskmovps", MEM, _XMM, _XMM);
|
|
put("vmaskmovpd", MEM, _XMM, _XMM);
|
|
|
|
put("vbroadcastf128", YMM, MEM);
|
|
put("vbroadcasti128", YMM, MEM);
|
|
put("vbroadcastsd", YMM|_YMM3, XMM|MEM);
|
|
put("vbroadcastsd", ZMM, XMM|MEM);
|
|
{
|
|
const char *tbl[] = {
|
|
"vbroadcastss",
|
|
"vpbroadcastb",
|
|
"vpbroadcastw",
|
|
"vpbroadcastd",
|
|
"vpbroadcastq",
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
put(tbl[i], XMM | YMM | ZMM, XMM|MEM);
|
|
}
|
|
}
|
|
|
|
put("vinsertf128", YMM, YMM, _XMM | _XMM2 | MEM, IMM8);
|
|
put("vinserti128", YMM, YMM, _XMM | _XMM2 | MEM, IMM8);
|
|
put("vperm2f128", YMM, YMM, YMM | MEM, IMM8);
|
|
put("vperm2i128", YMM, YMM, YMM | MEM, IMM8);
|
|
|
|
{
|
|
const char *tbl[] = {
|
|
"vpmaskmovd", "vpmaskmovq"
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const char *name = tbl[i];
|
|
put(name, _XMM, _XMM, MEM);
|
|
put(name, YMM, YMM, MEM);
|
|
put(name, MEM, _XMM, _XMM);
|
|
put(name, MEM, YMM, YMM);
|
|
}
|
|
}
|
|
{
|
|
const char *tbl[] = {
|
|
"vpermd", "vpermps",
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const char *name = tbl[i];
|
|
put(name, YMM, YMM, YMM | MEM);
|
|
}
|
|
}
|
|
{
|
|
const char *tbl[] = {
|
|
"vpermq", "vpermpd",
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const char *name = tbl[i];
|
|
put(name, YMM, YMM | MEM, IMM8);
|
|
}
|
|
}
|
|
put("vpextrw", REG32e | MEM, XMM, IMM); // nasm is ok, yasm generate redundant code
|
|
}
|
|
void putAVX512_M_X()
|
|
{
|
|
const char *tbl[] = {
|
|
"vmovapd",
|
|
"vmovaps",
|
|
"vmovupd",
|
|
"vmovups",
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const char *name = tbl[i];
|
|
put(name, MEM, ZMM);
|
|
put(name, ZMM, MEM);
|
|
#ifdef XBYAK64
|
|
put(name, MEM, XMM);
|
|
put(name, XMM, MEM);
|
|
#endif
|
|
}
|
|
}
|
|
void put_vmov()
|
|
{
|
|
#ifdef XBYAK64
|
|
put("vmovd", XMM, MEM|REG32);
|
|
put("vmovd", MEM|REG32, XMM);
|
|
put("vmovq", XMM, MEM|REG64|XMM);
|
|
put("vmovq", MEM|REG64|XMM, XMM);
|
|
put("vmovhlps", XMM, _XMM3, _XMM3);
|
|
put("vmovlhps", XMM, _XMM3, _XMM3);
|
|
put("vmovntdqa", XMM|_YMM3|ZMM, MEM);
|
|
put("vmovntdq", MEM, XMM | _YMM3 | ZMM);
|
|
put("vmovntpd", MEM, XMM | _YMM3 | ZMM);
|
|
put("vmovntps", MEM, XMM | _YMM3 | ZMM);
|
|
|
|
put("vmovsd", XMM_KZ, XMM, _XMM3);
|
|
put("vmovsd", XMM_KZ, MEM);
|
|
put("vmovsd", MEM_K, XMM);
|
|
put("vmovss", XMM_KZ, XMM, _XMM3);
|
|
put("vmovss", XMM_KZ, MEM);
|
|
put("vmovss", MEM_K, XMM);
|
|
|
|
put("vmovshdup", _ZMM, _ZMM);
|
|
put("vmovsldup", _ZMM, _ZMM);
|
|
|
|
|
|
{
|
|
const char *tbl[] = {
|
|
"valignd",
|
|
"valignq",
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const char *name = tbl[i];
|
|
put(name, XMM_KZ, XMM, _XMM | MEM, IMM);
|
|
put(name, _YMM3, _YMM3, _YMM3 | _MEM, IMM);
|
|
put(name, _ZMM, _ZMM, _ZMM | _MEM, IMM);
|
|
}
|
|
}
|
|
{
|
|
const char tbl[][16] = {
|
|
"vmovhpd",
|
|
"vmovhps",
|
|
"vmovlpd",
|
|
"vmovlps",
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
put(tbl[i], XMM, _XMM3, MEM);
|
|
put(tbl[i], MEM, _XMM3);
|
|
}
|
|
}
|
|
#endif
|
|
}
|
|
void put512_X_XM()
|
|
{
|
|
const struct Tbl {
|
|
const char *name;
|
|
bool M_X;
|
|
} tbl[] = {
|
|
{ "vmovddup", false },
|
|
{ "vmovdqa32", true },
|
|
{ "vmovdqa64", true },
|
|
{ "vmovdqu8", true },
|
|
{ "vmovdqu16", true },
|
|
{ "vmovdqu32", true },
|
|
{ "vmovdqu64", true },
|
|
{ "vpabsb", false },
|
|
{ "vpabsw", false },
|
|
{ "vpabsd", false },
|
|
{ "vpabsq", false },
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const Tbl& p = tbl[i];
|
|
put(p.name, XMM|XMM_KZ, _XMM|MEM);
|
|
put(p.name, _YMM|YMM_KZ, _YMM|MEM);
|
|
put(p.name, _ZMM|ZMM_KZ, _ZMM|MEM);
|
|
if (!p.M_X) continue;
|
|
put(p.name, MEM|MEM_K, XMM);
|
|
put(p.name, MEM|MEM_K, _YMM);
|
|
put(p.name, MEM|MEM_K, _ZMM);
|
|
}
|
|
put("vsqrtpd", XMM_KZ, M_1to2 | _MEM);
|
|
put("vsqrtpd", YMM_KZ, M_1to4 | _MEM);
|
|
put("vsqrtpd", ZMM_KZ, M_1to8 | _MEM);
|
|
put("vsqrtpd", ZMM_KZ, ZMM_ER);
|
|
|
|
put("vsqrtps", XMM_KZ, M_1to4 | _MEM);
|
|
put("vsqrtps", YMM_KZ, M_1to8 | _MEM);
|
|
put("vsqrtps", ZMM_KZ, M_1to16 | _MEM);
|
|
put("vsqrtps", ZMM_KZ, ZMM_ER);
|
|
|
|
put("vpabsd", ZMM_KZ, M_1to16 | _MEM);
|
|
put("vpabsq", ZMM_KZ, M_1to8 | _MEM);
|
|
|
|
put("vbroadcastf32x2", YMM_KZ | ZMM_KZ, XMM | _MEM);
|
|
put("vbroadcastf32x4", YMM_KZ | ZMM_KZ, _MEM);
|
|
|
|
put("vbroadcastf64x2", YMM_KZ | ZMM_KZ, _MEM);
|
|
put("vbroadcastf64x4", ZMM_KZ, _MEM);
|
|
put("vbroadcastf32x8", ZMM_KZ, _MEM);
|
|
}
|
|
void put512_X_X_XM()
|
|
{
|
|
const struct Tbl {
|
|
const char *name;
|
|
uint64_t mem;
|
|
} tbl[] = {
|
|
{ "vsqrtsd", MEM },
|
|
{ "vsqrtss", MEM },
|
|
{ "vunpckhpd", M_1to2 },
|
|
{ "vunpckhps", M_1to4 },
|
|
{ "vunpcklpd", M_1to2 },
|
|
{ "vunpcklps", M_1to4 },
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const Tbl& p = tbl[i];
|
|
put(p.name, XMM_KZ, XMM, _XMM|p.mem);
|
|
}
|
|
}
|
|
void put512_X3()
|
|
{
|
|
#ifdef XBYAK64
|
|
const struct Tbl {
|
|
const char *name;
|
|
uint64_t x1;
|
|
uint64_t x2;
|
|
uint64_t xm;
|
|
} tbl[] = {
|
|
{ "vpacksswb", XMM_KZ, XMM, _XMM | _MEM },
|
|
{ "vpacksswb", YMM_KZ, _YMM, _YMM | _MEM },
|
|
{ "vpacksswb", ZMM_KZ, _ZMM, _ZMM | _MEM },
|
|
|
|
{ "vpackssdw", XMM_KZ, XMM, _XMM | M_1to4 | _MEM },
|
|
{ "vpackssdw", YMM_KZ, _YMM, _YMM | M_1to8 | _MEM },
|
|
{ "vpackssdw", ZMM_KZ, _ZMM, _ZMM | M_1to16 | _MEM },
|
|
|
|
{ "vpackusdw", XMM_KZ, XMM, _XMM | M_1to4 | _MEM },
|
|
{ "vpackusdw", YMM_KZ, _YMM, _YMM | M_1to8 | _MEM },
|
|
{ "vpackusdw", ZMM_KZ, _ZMM, _ZMM | M_1to16 | _MEM },
|
|
|
|
{ "vpackuswb", XMM_KZ, XMM, _XMM | _MEM },
|
|
{ "vpackuswb", YMM_KZ, _YMM, _YMM | _MEM },
|
|
{ "vpackuswb", ZMM_KZ, _ZMM, _ZMM | _MEM },
|
|
|
|
{ "vpaddb", XMM_KZ, XMM, _XMM | _MEM },
|
|
{ "vpaddw", XMM_KZ, _XMM, _XMM | _MEM },
|
|
{ "vpaddd", XMM_KZ, _XMM, _XMM | M_1to4 | _MEM },
|
|
{ "vpaddq", ZMM_KZ, _ZMM, M_1to8 | _MEM },
|
|
|
|
{ "vpaddsb", XMM_KZ, XMM, _XMM | _MEM },
|
|
{ "vpaddsb", ZMM_KZ, _ZMM, _ZMM | _MEM },
|
|
|
|
{ "vpaddsw", XMM_KZ, XMM, _XMM | _MEM },
|
|
{ "vpaddsw", ZMM_KZ, _ZMM, _ZMM | _MEM },
|
|
|
|
{ "vpaddusb", XMM_KZ, XMM, _XMM | MEM },
|
|
{ "vpaddusb", ZMM_KZ, _ZMM, _ZMM | MEM },
|
|
|
|
{ "vpaddusw", XMM_KZ, XMM, _XMM | MEM },
|
|
{ "vpaddusw", ZMM_KZ, _ZMM, _ZMM | MEM },
|
|
|
|
{ "vpsubb", XMM_KZ, XMM, _XMM | _MEM },
|
|
{ "vpsubw", XMM_KZ, XMM, _XMM | _MEM },
|
|
{ "vpsubd", XMM_KZ, XMM, _XMM | M_1to4 | _MEM },
|
|
{ "vpsubq", ZMM_KZ, _ZMM, M_1to8 | _MEM },
|
|
|
|
{ "vpsubsb", XMM_KZ, XMM, _XMM | _MEM },
|
|
{ "vpsubsb", ZMM_KZ, _ZMM, _ZMM | _MEM },
|
|
|
|
{ "vpsubsw", XMM_KZ, XMM, _XMM | _MEM },
|
|
{ "vpsubsw", ZMM_KZ, _ZMM, _ZMM | _MEM },
|
|
|
|
{ "vpsubusb", XMM_KZ, XMM, _XMM | MEM },
|
|
{ "vpsubusb", ZMM_KZ, _ZMM, _ZMM | MEM },
|
|
|
|
{ "vpsubusw", XMM_KZ, XMM, _XMM | MEM },
|
|
{ "vpsubusw", ZMM_KZ, _ZMM, _ZMM | MEM },
|
|
|
|
{ "vpandd", ZMM_KZ, _ZMM, _ZMM | M_1to16 | _MEM },
|
|
{ "vpandq", ZMM_KZ, _ZMM, _ZMM | M_1to8 | _MEM },
|
|
|
|
{ "vpandnd", ZMM_KZ, _ZMM, _ZMM | M_1to16 | _MEM },
|
|
{ "vpandnq", ZMM_KZ, _ZMM, _ZMM | M_1to8 | _MEM },
|
|
|
|
{ "vpavgb", ZMM_KZ, _ZMM, _ZMM },
|
|
{ "vpavgw", ZMM_KZ, _ZMM, _ZMM },
|
|
|
|
{ "vpcmpeqb", K2, _ZMM, _ZMM | _MEM },
|
|
{ "vpcmpeqw", K2, _ZMM, _ZMM | _MEM },
|
|
{ "vpcmpeqd", K2, _ZMM, _ZMM | M_1to16 | _MEM },
|
|
{ "vpcmpeqq", K2, _ZMM, _ZMM | M_1to8 | _MEM },
|
|
|
|
{ "vpcmpgtb", K2, _ZMM, _ZMM | _MEM },
|
|
{ "vpcmpgtw", K2, _ZMM, _ZMM | _MEM },
|
|
{ "vpcmpgtd", K2, _ZMM, _ZMM | M_1to16 | _MEM },
|
|
{ "vpcmpgtq", K2, _ZMM, _ZMM | M_1to8 | _MEM },
|
|
|
|
{ "vpmaddubsw", ZMM_KZ, _ZMM, _ZMM | _MEM },
|
|
{ "vpmaddwd", ZMM_KZ, _ZMM, _ZMM | _MEM },
|
|
|
|
{ "vpmaxsb", ZMM_KZ, _ZMM, _ZMM | _MEM },
|
|
{ "vpmaxsw", ZMM_KZ, _ZMM, _ZMM | _MEM },
|
|
{ "vpmaxsd", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to16 },
|
|
{ "vpmaxsq", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to8 },
|
|
|
|
{ "vpmaxub", ZMM_KZ, _ZMM, _ZMM | _MEM },
|
|
{ "vpmaxuw", ZMM_KZ, _ZMM, _ZMM | _MEM },
|
|
{ "vpmaxud", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to16 },
|
|
{ "vpmaxuq", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to8 },
|
|
|
|
{ "vpminsb", ZMM_KZ, _ZMM, _ZMM | _MEM },
|
|
{ "vpminsw", ZMM_KZ, _ZMM, _ZMM | _MEM },
|
|
{ "vpminsd", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to16 },
|
|
{ "vpminsq", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to8 },
|
|
|
|
{ "vpminub", ZMM_KZ, _ZMM, _ZMM | _MEM },
|
|
{ "vpminuw", ZMM_KZ, _ZMM, _ZMM | _MEM },
|
|
{ "vpminud", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to16 },
|
|
{ "vpminuq", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to8 },
|
|
|
|
{ "vpslldq", XMM, _XMM3 | _MEM, IMM8 },
|
|
{ "vpslldq", _YMM3, _YMM3 | _MEM, IMM8 },
|
|
{ "vpslldq", _ZMM, _ZMM | _MEM, IMM8 },
|
|
|
|
{ "vpsrldq", XMM, _XMM3 | _MEM, IMM8 },
|
|
{ "vpsrldq", _YMM3, _YMM3 | _MEM, IMM8 },
|
|
{ "vpsrldq", _ZMM, _ZMM | _MEM, IMM8 },
|
|
|
|
{ "vpsraw", XMM_KZ, XMM | _MEM, IMM8 },
|
|
{ "vpsraw", ZMM_KZ, _ZMM | _MEM, IMM8 },
|
|
|
|
{ "vpsrad", XMM_KZ, XMM | M_1to4 | _MEM, IMM8 },
|
|
{ "vpsrad", ZMM_KZ, _ZMM | M_1to16 | _MEM, IMM8 },
|
|
|
|
{ "vpsraq", XMM, XMM, IMM8 },
|
|
{ "vpsraq", XMM_KZ, XMM | M_1to2 | _MEM, IMM8 },
|
|
{ "vpsraq", ZMM_KZ, _ZMM | M_1to8 | _MEM, IMM8 },
|
|
|
|
{ "vpsllw", XMM, _XMM3 | _MEM, IMM8 },
|
|
{ "vpslld", XMM, _XMM3 | _MEM | M_1to4, IMM8 },
|
|
{ "vpsllq", XMM, _XMM3 | _MEM | M_1to2, IMM8 },
|
|
|
|
{ "vpsrlw", XMM_KZ, XMM | _MEM, IMM8 },
|
|
{ "vpsrlw", ZMM_KZ, _ZMM | _MEM, IMM8 },
|
|
|
|
{ "vpsrld", XMM_KZ, XMM | M_1to4 | _MEM, IMM8 },
|
|
{ "vpsrld", ZMM_KZ, _ZMM | M_1to16 | _MEM, IMM8 },
|
|
|
|
{ "vpsrlq", XMM, _XMM3 | _MEM | M_1to2, IMM8 },
|
|
{ "vpsrlq", _ZMM, _ZMM | _MEM | M_1to8, IMM8 },
|
|
|
|
{ "vpsravw", XMM_KZ | XMM, _XMM, _XMM | _MEM },
|
|
{ "vpsravw", _ZMM, _ZMM, _MEM },
|
|
|
|
{ "vpsravd", XMM_KZ | XMM, _XMM, _XMM | _MEM },
|
|
{ "vpsravd", _ZMM, _ZMM, M_1to16 | _MEM },
|
|
|
|
{ "vpsravq", XMM_KZ | XMM, _XMM, _XMM | _MEM },
|
|
{ "vpsravq", _ZMM, _ZMM, M_1to8 | _MEM },
|
|
|
|
{ "vpsllvw", XMM_KZ | XMM, _XMM, _XMM | _MEM },
|
|
{ "vpsllvw", _ZMM, _ZMM, _MEM },
|
|
|
|
{ "vpsllvd", XMM_KZ | XMM, _XMM, _XMM | _MEM },
|
|
{ "vpsllvd", _ZMM, _ZMM, M_1to16 | _MEM },
|
|
|
|
{ "vpsllvq", XMM_KZ | XMM, _XMM, _XMM | _MEM },
|
|
{ "vpsllvq", _ZMM, _ZMM, M_1to8 | _MEM },
|
|
|
|
{ "vpsrlvw", XMM_KZ | XMM, _XMM, _XMM | _MEM },
|
|
{ "vpsrlvw", _ZMM, _ZMM, _MEM },
|
|
|
|
{ "vpsrlvd", XMM_KZ | XMM, _XMM, _XMM | _MEM },
|
|
{ "vpsrlvd", _ZMM, _ZMM, M_1to16 | _MEM },
|
|
|
|
{ "vpsrlvq", XMM_KZ | XMM, _XMM, _XMM | _MEM },
|
|
{ "vpsrlvq", _ZMM, _ZMM, M_1to8 | _MEM },
|
|
|
|
{ "vpshufb", XMM | XMM_KZ, _XMM, _XMM | _MEM },
|
|
{ "vpshufb", ZMM_KZ, _ZMM, _MEM },
|
|
|
|
{ "vpshufhw", XMM | XMM_KZ, _XMM | _MEM, IMM8 },
|
|
{ "vpshufhw", ZMM_KZ, _MEM, IMM8 },
|
|
|
|
{ "vpshuflw", XMM | XMM_KZ, _XMM | _MEM, IMM8 },
|
|
{ "vpshuflw", ZMM_KZ, _MEM, IMM8 },
|
|
|
|
{ "vpshufd", XMM | XMM_KZ, _XMM | M_1to4 | _MEM, IMM8 },
|
|
{ "vpshufd", _ZMM | ZMM_KZ, _ZMM | M_1to16 | _MEM, IMM8 },
|
|
|
|
{ "vpord", XMM | XMM_KZ, _XMM, _XMM | M_1to4 | _MEM },
|
|
{ "vpord", _ZMM | ZMM_KZ, _ZMM, M_1to16 | _MEM },
|
|
|
|
{ "vporq", XMM | XMM_KZ, _XMM, _XMM | M_1to2 | _MEM },
|
|
{ "vporq", _ZMM | ZMM_KZ, _ZMM, M_1to8 | _MEM },
|
|
|
|
{ "vpxord", XMM | XMM_KZ, _XMM, _XMM | M_1to4 | _MEM },
|
|
{ "vpxord", _ZMM | ZMM_KZ, _ZMM, M_1to16 | _MEM },
|
|
|
|
{ "vpxorq", XMM | XMM_KZ, _XMM, _XMM | M_1to2 | _MEM },
|
|
{ "vpxorq", _ZMM | ZMM_KZ, _ZMM, M_1to8 | _MEM },
|
|
|
|
{ "vpsadbw", XMM, _XMM, _XMM | _MEM },
|
|
{ "vpsadbw", _ZMM, _ZMM, _MEM },
|
|
|
|
{ "vpmuldq", XMM, _XMM, _XMM | M_1to2 | _MEM },
|
|
{ "vpmuldq", ZMM_KZ, _ZMM, M_1to8 | _MEM },
|
|
|
|
{ "vpmulhrsw", XMM, _XMM, _XMM | _MEM },
|
|
{ "vpmulhrsw", ZMM_KZ, _ZMM, _MEM },
|
|
|
|
{ "vpmulhuw", XMM, _XMM, _XMM | _MEM },
|
|
{ "vpmulhuw", ZMM_KZ, _ZMM, _MEM },
|
|
|
|
{ "vpmulhw", XMM, _XMM, _XMM | _MEM },
|
|
{ "vpmulhw", ZMM_KZ, _ZMM, _MEM },
|
|
|
|
{ "vpmullw", XMM, _XMM, _XMM | _MEM },
|
|
{ "vpmullw", ZMM_KZ, _ZMM, _MEM },
|
|
|
|
{ "vpmulld", XMM, _XMM, M_1to4 | _MEM },
|
|
{ "vpmulld", ZMM_KZ, _ZMM, M_1to16 | _MEM },
|
|
|
|
{ "vpmullq", XMM, _XMM, M_1to2 | _MEM },
|
|
{ "vpmullq", ZMM_KZ, _ZMM, M_1to8 | _MEM },
|
|
|
|
{ "vpmuludq", XMM, _XMM, M_1to2 | _MEM },
|
|
{ "vpmuludq", ZMM_KZ, _ZMM, M_1to8 | _MEM },
|
|
|
|
{ "vpunpckhbw", XMM, _XMM, _XMM | _MEM },
|
|
{ "vpunpckhbw", _ZMM, _ZMM, _MEM },
|
|
|
|
{ "vpunpckhwd", XMM, _XMM, _XMM | _MEM },
|
|
{ "vpunpckhwd", _ZMM, _ZMM, _MEM },
|
|
|
|
{ "vpunpckhdq", XMM, _XMM, M_1to4 | _MEM },
|
|
{ "vpunpckhdq", _ZMM, _ZMM, M_1to16 | _MEM },
|
|
|
|
{ "vpunpckhqdq", XMM, _XMM, M_1to2 | _MEM },
|
|
{ "vpunpckhqdq", _ZMM, _ZMM, M_1to8 | _MEM },
|
|
|
|
{ "vpunpcklbw", XMM, _XMM, _XMM | _MEM },
|
|
{ "vpunpcklbw", _ZMM, _ZMM, _MEM },
|
|
|
|
{ "vpunpcklwd", XMM, _XMM, _XMM | _MEM },
|
|
{ "vpunpcklwd", _ZMM, _ZMM, _MEM },
|
|
|
|
{ "vpunpckldq", XMM, _XMM, M_1to4 | _MEM },
|
|
{ "vpunpckldq", _ZMM, _ZMM, M_1to16 | _MEM },
|
|
|
|
{ "vpunpcklqdq", XMM, _XMM, M_1to2 | _MEM },
|
|
{ "vpunpcklqdq", _ZMM, _ZMM, M_1to8 | _MEM },
|
|
|
|
{ "vextractf32x4", _XMM | XMM_KZ | _MEM, _YMM | _ZMM, IMM8 },
|
|
{ "vextractf64x2", _XMM | XMM_KZ | _MEM, _YMM | _ZMM, IMM8 },
|
|
{ "vextractf32x8", _YMM | YMM_KZ | _MEM, _ZMM, IMM8 },
|
|
{ "vextractf64x4", _YMM | YMM_KZ | _MEM, _ZMM, IMM8 },
|
|
|
|
{ "vextracti32x4", _XMM | XMM_KZ | _MEM, _YMM | _ZMM, IMM8 },
|
|
{ "vextracti64x2", _XMM | XMM_KZ | _MEM, _YMM | _ZMM, IMM8 },
|
|
{ "vextracti32x8", _YMM | YMM_KZ | _MEM, _ZMM, IMM8 },
|
|
{ "vextracti64x4", _YMM | YMM_KZ | _MEM, _ZMM, IMM8 },
|
|
|
|
{ "vextractps", REG32 | _MEM, XMM, IMM8 },
|
|
|
|
{ "vpermb", XMM_KZ, _XMM, _XMM | _MEM },
|
|
{ "vpermb", ZMM_KZ, _ZMM, _ZMM | _MEM },
|
|
|
|
{ "vpermw", XMM_KZ, _XMM, _XMM | _MEM },
|
|
{ "vpermw", ZMM_KZ, _ZMM, _ZMM | _MEM },
|
|
|
|
{ "vpermd", YMM_KZ, _YMM, _YMM | M_1to8 | _MEM },
|
|
{ "vpermd", ZMM_KZ, _ZMM, _ZMM | M_1to16 | _MEM },
|
|
|
|
{ "vpermilpd", XMM_KZ, _XMM, _XMM | M_1to2 | _MEM },
|
|
{ "vpermilpd", ZMM_KZ, _ZMM, M_1to8 | _MEM },
|
|
{ "vpermilpd", XMM_KZ, M_1to2 | _MEM, IMM8 },
|
|
{ "vpermilpd", ZMM_KZ, M_1to8 | _MEM, IMM8 },
|
|
|
|
{ "vpermilps", XMM_KZ, _XMM, _XMM | _MEM | M_1to4 },
|
|
{ "vpermilps", ZMM_KZ, _ZMM, _MEM | M_1to16 },
|
|
{ "vpermilps", XMM_KZ, _MEM | M_1to4 | _MEM, IMM8 },
|
|
{ "vpermilps", ZMM_KZ, _MEM | M_1to16 | _MEM, IMM8 },
|
|
|
|
{ "vpermpd", YMM_KZ, _YMM | M_1to4 | _MEM, IMM8 },
|
|
{ "vpermpd", ZMM_KZ, _ZMM | M_1to8 | _MEM, IMM8 },
|
|
{ "vpermpd", YMM_KZ, _YMM, M_1to4 | _MEM },
|
|
{ "vpermpd", ZMM_KZ, _ZMM, M_1to8 | _MEM },
|
|
|
|
{ "vpermps", YMM_KZ, _YMM, M_1to8 | _MEM },
|
|
{ "vpermps", ZMM_KZ, _ZMM, M_1to16 | _MEM },
|
|
|
|
{ "vpermq", YMM_KZ, _YMM | M_1to4 | _MEM, IMM8 },
|
|
{ "vpermq", ZMM_KZ, _ZMM | M_1to8 | _MEM, IMM8 },
|
|
{ "vpermq", YMM_KZ, _YMM, M_1to4 | _MEM },
|
|
{ "vpermq", ZMM_KZ, _ZMM, M_1to8 | _MEM },
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const Tbl& p = tbl[i];
|
|
put(p.name, p.x1, p.x2, p.xm);
|
|
}
|
|
#endif
|
|
}
|
|
void put512_X3_I()
|
|
{
|
|
const struct Tbl {
|
|
const char *name;
|
|
uint64_t x1;
|
|
uint64_t x2;
|
|
uint64_t xm;
|
|
} tbl[] = {
|
|
#ifdef XBYAK64
|
|
{ "vinsertps", XMM, _XMM, _XMM3 | _MEM },
|
|
|
|
{ "vshufpd", XMM_KZ, _XMM, M_1to2 | _MEM },
|
|
{ "vshufpd", ZMM_KZ, _ZMM, M_1to8 | _MEM },
|
|
|
|
{ "vshufps", XMM_KZ, _XMM, M_1to4 | _MEM },
|
|
{ "vshufps", ZMM_KZ, _ZMM, M_1to16 | _MEM },
|
|
|
|
{ "vinsertf32x4", _YMM | YMM_KZ, _YMM, _XMM | _MEM },
|
|
{ "vinsertf32x4", _ZMM | ZMM_KZ, _ZMM, _XMM | _MEM },
|
|
|
|
{ "vinsertf64x2", _YMM | YMM_KZ, _YMM, _XMM | _MEM },
|
|
{ "vinsertf64x2", _ZMM | ZMM_KZ, _ZMM, _XMM | _MEM },
|
|
|
|
{ "vinsertf32x8", _ZMM | ZMM_KZ, _ZMM, _YMM | _MEM },
|
|
{ "vinsertf64x4", _ZMM | ZMM_KZ, _ZMM, _YMM | _MEM },
|
|
|
|
{ "vinserti32x4", _YMM | YMM_KZ, _YMM, _XMM | _MEM },
|
|
{ "vinserti32x4", _ZMM | ZMM_KZ, _ZMM, _XMM | _MEM },
|
|
|
|
{ "vinserti64x2", _YMM | YMM_KZ, _YMM, _XMM | _MEM },
|
|
{ "vinserti64x2", _ZMM | ZMM_KZ, _ZMM, _XMM | _MEM },
|
|
|
|
{ "vinserti32x8", _ZMM | ZMM_KZ, _ZMM, _YMM | _MEM },
|
|
{ "vinserti64x4", _ZMM | ZMM_KZ, _ZMM, _YMM | _MEM },
|
|
#endif
|
|
{ "vpalignr", ZMM_KZ, _ZMM, _ZMM | _MEM },
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const Tbl& p = tbl[i];
|
|
put(p.name, p.x1, p.x2, p.xm, IMM8);
|
|
}
|
|
#ifdef XBYAK64
|
|
put("vpextrb", _REG64 | _MEM, XMM, IMM8);
|
|
put("vpextrw", _REG64 | _MEM, XMM, IMM8);
|
|
put("vpextrd", _REG32 | _MEM, XMM, IMM8);
|
|
put("vpextrq", _REG64 | _MEM, XMM, IMM8);
|
|
put("vpinsrb", XMM, _XMM3, _REG32 | _MEM, IMM8);
|
|
put("vpinsrw", XMM, _XMM3, _REG32 | _MEM, IMM8);
|
|
put("vpinsrd", XMM, _XMM3, _REG32 | _MEM, IMM8);
|
|
put("vpinsrq", XMM, _XMM3, _REG64 | _MEM, IMM8);
|
|
#endif
|
|
}
|
|
void put512_FMA()
|
|
{
|
|
const struct Tbl {
|
|
const char *name;
|
|
bool supportYMM;
|
|
} tbl[] = {
|
|
{ "vfmadd", true },
|
|
{ "vfmadd", false },
|
|
{ "vfmaddsub", true },
|
|
{ "vfmsubadd", true },
|
|
{ "vfmsub", true },
|
|
{ "vfmsub", false },
|
|
{ "vfnmadd", true },
|
|
{ "vfnmadd", false },
|
|
{ "vfnmsub", true },
|
|
{ "vfnmsub", false },
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const Tbl& p = tbl[i];
|
|
const struct Ord {
|
|
const char *name;
|
|
} ord[] = {
|
|
{ "132" },
|
|
{ "213" },
|
|
{ "231" },
|
|
};
|
|
for (size_t j = 0; j < NUM_OF_ARRAY(ord); j++) {
|
|
const char sufTbl[][2][8] = {
|
|
{ "pd", "ps" },
|
|
{ "sd", "ss" },
|
|
};
|
|
for (size_t k = 0; k < 2; k++) {
|
|
const std::string suf = sufTbl[p.supportYMM ? 0 : 1][k];
|
|
uint64_t mem = 0;
|
|
if (suf == "pd") {
|
|
mem = M_1to2;
|
|
} else if (suf == "ps") {
|
|
mem = M_1to4;
|
|
} else {
|
|
mem = XMM_ER;
|
|
}
|
|
std::string name = std::string(p.name) + ord[j].name + suf;
|
|
const char *q = name.c_str();
|
|
put(q, XMM_KZ, _XMM, mem | _MEM);
|
|
if (!p.supportYMM) continue;
|
|
if (suf == "pd") {
|
|
mem = M_1to8;
|
|
} else if (suf == "ps") {
|
|
mem = M_1to16;
|
|
} else {
|
|
mem = XMM_ER;
|
|
}
|
|
put(q, _ZMM, _ZMM, mem | _MEM);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
void put512_Y_XM()
|
|
{
|
|
const struct Tbl {
|
|
const char *name;
|
|
bool all_xmm; // 2nd param
|
|
} tbl[] = {
|
|
{ "vpmovsxbw", false },
|
|
{ "vpmovsxbd", true },
|
|
{ "vpmovsxbq", true },
|
|
{ "vpmovsxwd", false },
|
|
{ "vpmovsxwq", true },
|
|
{ "vpmovsxdq", false },
|
|
|
|
{ "vpmovzxbw", false },
|
|
{ "vpmovzxbd", true },
|
|
{ "vpmovzxbq", true },
|
|
{ "vpmovzxwd", false },
|
|
{ "vpmovzxwq", true },
|
|
{ "vpmovzxdq", false },
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const Tbl& p = tbl[i];
|
|
const char *name = p.name;
|
|
put(name, XMM_KZ | YMM, _XMM | _MEM);
|
|
if (p.all_xmm) {
|
|
put(name, ZMM, _XMM | _MEM);
|
|
} else {
|
|
put(name, ZMM, YMM | _MEM);
|
|
}
|
|
}
|
|
}
|
|
void put512_AVX1()
|
|
{
|
|
#ifdef XBYAK64
|
|
const struct Tbl {
|
|
std::string name;
|
|
bool only_pd_ps;
|
|
} tbl[] = {
|
|
{ "vadd", false },
|
|
{ "vsub", false },
|
|
{ "vmul", false },
|
|
{ "vdiv", false },
|
|
{ "vmax", false },
|
|
{ "vmin", false },
|
|
{ "vand", true },
|
|
{ "vandn", true },
|
|
{ "vor", true },
|
|
{ "vxor", true },
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const struct Suf {
|
|
const char *suf;
|
|
bool supportYMM;
|
|
} sufTbl[] = {
|
|
{ "pd", true },
|
|
{ "ps", true },
|
|
{ "sd", false },
|
|
{ "ss", false },
|
|
};
|
|
for (size_t j = 0; j < NUM_OF_ARRAY(sufTbl); j++) {
|
|
if (tbl[i].only_pd_ps && j == 2) break;
|
|
std::string suf = sufTbl[j].suf;
|
|
std::string name = tbl[i].name + suf;
|
|
const char *p = name.c_str();
|
|
uint64_t mem = 0;
|
|
if (suf == "pd") {
|
|
mem = M_1to2;
|
|
} else if (suf == "ps") {
|
|
mem = M_1to4;
|
|
}
|
|
put(p, XMM | XMM_KZ, _XMM, mem | _MEM);
|
|
if (!sufTbl[j].supportYMM) continue;
|
|
mem = 0;
|
|
if (suf == "pd") {
|
|
mem = M_1to8;
|
|
} else if (suf == "ps") {
|
|
mem = M_1to16;
|
|
}
|
|
put(p, _ZMM, _ZMM, mem | _MEM);
|
|
}
|
|
}
|
|
#endif
|
|
}
|
|
void putAVX1()
|
|
{
|
|
const struct Tbl {
|
|
const char *name;
|
|
bool only_pd_ps;
|
|
} tbl[] = {
|
|
{ "add", false },
|
|
{ "sub", false },
|
|
{ "mul", false },
|
|
{ "div", false },
|
|
{ "max", false },
|
|
{ "min", false },
|
|
{ "and", true },
|
|
{ "andn", true },
|
|
{ "or", true },
|
|
{ "xor", true },
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const struct Suf {
|
|
const char *suf;
|
|
bool supportYMM;
|
|
} suf[] = {
|
|
{ "pd", true },
|
|
{ "ps", true },
|
|
{ "sd", false },
|
|
{ "ss", false },
|
|
};
|
|
for (size_t j = 0; j < NUM_OF_ARRAY(suf); j++) {
|
|
if (tbl[i].only_pd_ps && j == 2) break;
|
|
std::string name = std::string("v") + tbl[i].name + suf[j].suf;
|
|
const char *p = name.c_str();
|
|
put(p, XMM, XMM | MEM);
|
|
put(p, XMM, XMM, XMM | MEM);
|
|
if (!suf[j].supportYMM) continue;
|
|
put(p, YMM, YMM | MEM);
|
|
put(p, YMM, YMM, YMM | MEM);
|
|
put(p, ZMM, ZMM, ZMM | MEM);
|
|
}
|
|
}
|
|
}
|
|
void put512_cvt()
|
|
{
|
|
#ifdef XBYAK64
|
|
put("vcvtdq2pd", XMM_KZ, _XMM | _MEM | M_1to2);
|
|
put("vcvtdq2pd", YMM_KZ, _XMM | _MEM | M_1to4);
|
|
put("vcvtdq2pd", ZMM_KZ, _YMM | _MEM | M_1to8);
|
|
|
|
put("vcvtdq2ps", XMM_KZ, _XMM | _MEM | M_1to4);
|
|
put("vcvtdq2ps", YMM_KZ, _YMM | _MEM | M_1to8);
|
|
put("vcvtdq2ps", ZMM_KZ, _ZMM | _MEM | M_1to16 | ZMM_ER);
|
|
|
|
put("vcvtpd2dq", XMM_KZ, _XMM | M_xword | M_1to2);
|
|
put("vcvtpd2dq", XMM_KZ, _YMM | M_yword | MY_1to4);
|
|
put("vcvtpd2dq", YMM_KZ, ZMM | _MEM | M_1to8 | ZMM_ER);
|
|
|
|
put("vcvtpd2ps", XMM_KZ, _XMM | M_xword | M_1to2);
|
|
put("vcvtpd2ps", XMM_KZ, _YMM | M_yword | MY_1to4);
|
|
put("vcvtpd2ps", YMM_KZ, ZMM | _MEM | M_1to8 | ZMM_ER);
|
|
|
|
put("vcvtpd2qq", XMM_KZ, _XMM | _MEM | M_1to2);
|
|
put("vcvtpd2qq", YMM_KZ, _YMM | _MEM | M_1to4);
|
|
put("vcvtpd2qq", ZMM_KZ, _ZMM | _MEM | M_1to8 | ZMM_ER);
|
|
|
|
put("vcvtpd2udq", XMM_KZ, _XMM | M_xword | M_1to2);
|
|
put("vcvtpd2udq", XMM_KZ, _YMM | M_yword | MY_1to4);
|
|
put("vcvtpd2udq", YMM_KZ, ZMM | _MEM | M_1to8 | ZMM_ER);
|
|
|
|
put("vcvtpd2uqq", XMM_KZ, _XMM | _MEM | M_1to2);
|
|
put("vcvtpd2uqq", YMM_KZ, _YMM | _MEM | M_1to4);
|
|
put("vcvtpd2uqq", ZMM_KZ, _ZMM | _MEM | M_1to8 | ZMM_ER);
|
|
|
|
put("vcvtph2ps", XMM_KZ, _XMM | _MEM);
|
|
put("vcvtph2ps", YMM_KZ, _XMM | _MEM);
|
|
put("vcvtph2ps", ZMM_KZ, _YMM | _MEM | YMM_SAE);
|
|
|
|
put("vcvtps2ph", XMM_KZ | _MEM, _XMM, IMM8);
|
|
put("vcvtps2ph", XMM_KZ | _MEM, _YMM, IMM8);
|
|
put("vcvtps2ph", YMM_KZ | _MEM, _ZMM, IMM8);
|
|
put("vcvtps2ph", YMM_KZ, ZMM_SAE, IMM8);
|
|
|
|
put("vcvtps2dq", XMM_KZ, _XMM | _MEM | M_1to4);
|
|
put("vcvtps2dq", YMM_KZ, _YMM | _MEM | M_1to8);
|
|
put("vcvtps2dq", ZMM_KZ, _ZMM | _MEM | M_1to16 | ZMM_ER);
|
|
|
|
put("vcvtps2udq", XMM_KZ, _XMM | M_1to4);
|
|
put("vcvtps2udq", YMM_KZ, _YMM | M_1to8);
|
|
put("vcvtps2udq", ZMM_KZ, _ZMM | _MEM | M_1to16 | ZMM_ER);
|
|
|
|
put("vcvtps2qq", XMM_KZ, _XMM | _MEM | M_1to2);
|
|
put("vcvtps2qq", YMM_KZ, _XMM | _MEM | M_1to4);
|
|
put("vcvtps2qq", ZMM_KZ, _YMM | _MEM | M_1to8 | YMM_ER);
|
|
|
|
put("vcvtps2uqq", XMM_KZ, _XMM | _MEM | M_1to2);
|
|
put("vcvtps2uqq", YMM_KZ, _XMM | _MEM | M_1to4);
|
|
put("vcvtps2uqq", ZMM_KZ, _YMM | _MEM | M_1to8 | YMM_ER);
|
|
|
|
put("vcvtps2pd", XMM_KZ, _XMM | _MEM | M_1to2);
|
|
put("vcvtps2pd", YMM_KZ, _XMM | _MEM | M_1to4);
|
|
put("vcvtps2pd", ZMM_KZ, _YMM | _MEM | M_1to8 | YMM_SAE);
|
|
|
|
put("vcvtqq2pd", XMM_KZ, _XMM | _MEM | M_1to2);
|
|
put("vcvtqq2pd", YMM_KZ, _YMM | _MEM | M_1to4);
|
|
put("vcvtqq2pd", ZMM_KZ, _ZMM | _MEM | M_1to8 | ZMM_ER);
|
|
|
|
put("vcvtqq2ps", XMM_KZ, _XMM | M_xword | M_1to2);
|
|
put("vcvtqq2ps", XMM_KZ, _YMM | M_yword | MY_1to4);
|
|
put("vcvtqq2ps", YMM_KZ, ZMM | _MEM | M_1to8 | ZMM_ER);
|
|
|
|
put("vcvtsd2si", REG32 | REG64, XMM | _MEM | XMM_ER);
|
|
|
|
put("vcvtsd2usi", REG32 | REG64, XMM | _MEM | XMM_ER);
|
|
|
|
put("vcvtsd2ss", XMM_KZ, XMM, _XMM3 | _MEM | XMM_ER);
|
|
|
|
put("vcvtsi2sd", XMM, _XMM3, REG32 | REG64 | MEM32 | MEM64);
|
|
put("vcvtsi2sd", XMM, XMM_ER, REG64);
|
|
|
|
put("vcvtsi2ss", XMM, _XMM3, REG32 | REG64 | MEM32 | MEM64);
|
|
put("vcvtsi2ss", XMM, XMM_ER, REG32 | REG64);
|
|
|
|
put("vcvtss2sd", XMM_KZ, XMM, _XMM3 | _MEM | XMM_SAE);
|
|
|
|
put("vcvtss2si", REG32 | REG64, XMM | _MEM | XMM_ER);
|
|
|
|
put("vcvtss2usi", REG32 | REG64, XMM | _MEM | XMM_ER);
|
|
|
|
put("vcvtpd2dq", XMM_KZ, _XMM | M_xword | M_1to2);
|
|
put("vcvtpd2dq", XMM_KZ, _YMM | M_yword | MY_1to4);
|
|
put("vcvtpd2dq", YMM_KZ, _ZMM | _MEM | M_1to8 | ZMM_ER);
|
|
|
|
put("vcvttpd2qq", XMM_KZ, _XMM | _MEM | M_1to2);
|
|
put("vcvttpd2qq", YMM_KZ, _YMM | _MEM | M_1to4);
|
|
put("vcvttpd2qq", ZMM_KZ, _ZMM | _MEM | M_1to8 | ZMM_SAE);
|
|
|
|
put("vcvttpd2udq", XMM_KZ, _XMM | M_xword | M_1to2);
|
|
put("vcvttpd2udq", XMM_KZ, _YMM | M_yword | MY_1to4);
|
|
put("vcvttpd2udq", YMM_KZ, ZMM | _MEM | M_1to8 | ZMM_SAE);
|
|
|
|
put("vcvttpd2uqq", XMM_KZ, _XMM | _MEM | M_1to2);
|
|
put("vcvttpd2uqq", YMM_KZ, _YMM | _MEM | M_1to4);
|
|
put("vcvttpd2uqq", ZMM_KZ, _ZMM | _MEM | M_1to8 | ZMM_SAE);
|
|
|
|
put("vcvttps2dq", XMM_KZ, _XMM | _MEM | M_1to4);
|
|
put("vcvttps2dq", YMM_KZ, _YMM | _MEM | M_1to8);
|
|
put("vcvttps2dq", ZMM_KZ, _ZMM | _MEM | M_1to16 | ZMM_SAE);
|
|
|
|
put("vcvttps2udq", XMM_KZ, _XMM | M_1to4);
|
|
put("vcvttps2udq", YMM_KZ, _YMM | M_1to8);
|
|
put("vcvttps2udq", ZMM_KZ, _ZMM | _MEM | M_1to16 | ZMM_SAE);
|
|
|
|
put("vcvttps2qq", XMM_KZ, _XMM | _MEM | M_1to2);
|
|
put("vcvttps2qq", YMM_KZ, _XMM | _MEM | M_1to4);
|
|
put("vcvttps2qq", ZMM_KZ, _YMM | _MEM | M_1to8 | YMM_SAE);
|
|
|
|
put("vcvttps2uqq", XMM_KZ, _XMM | _MEM | M_1to2);
|
|
put("vcvttps2uqq", YMM_KZ, _XMM | _MEM | M_1to4);
|
|
put("vcvttps2uqq", ZMM_KZ, _YMM | _MEM | M_1to8 | YMM_SAE);
|
|
|
|
put("vcvttsd2si", REG32 | REG64, XMM | _MEM | XMM_SAE);
|
|
|
|
put("vcvttsd2usi", REG32 | REG64, XMM | _MEM | XMM_SAE);
|
|
|
|
put("vcvttss2si", REG32 | REG64, XMM | _MEM | XMM_SAE);
|
|
|
|
put("vcvttss2usi", REG32 | REG64, XMM | _MEM | XMM_SAE);
|
|
|
|
put("vcvtudq2pd", XMM_KZ, _XMM | _MEM | M_1to2);
|
|
put("vcvtudq2pd", YMM_KZ, _XMM | _MEM | M_1to4);
|
|
put("vcvtudq2pd", ZMM_KZ, _YMM | _MEM | M_1to8);
|
|
|
|
put("vcvtudq2ps", XMM_KZ, _XMM | _MEM | M_1to4);
|
|
put("vcvtudq2ps", YMM_KZ, _YMM | _MEM | M_1to8);
|
|
put("vcvtudq2ps", ZMM_KZ, _ZMM | _MEM | M_1to16 | ZMM_ER);
|
|
|
|
put("vcvtuqq2pd", XMM_KZ, _XMM | _MEM | M_1to2);
|
|
put("vcvtuqq2pd", YMM_KZ, _YMM | _MEM | M_1to4);
|
|
put("vcvtuqq2pd", ZMM_KZ, _ZMM | _MEM | M_1to8 | ZMM_ER);
|
|
|
|
put("vcvtuqq2ps", XMM_KZ, _XMM | M_xword | M_1to2);
|
|
put("vcvtuqq2ps", XMM_KZ, _YMM | M_yword | MY_1to4);
|
|
put("vcvtuqq2ps", YMM_KZ, ZMM | _MEM | M_1to8 | ZMM_ER);
|
|
|
|
put("vcvtusi2sd", XMM, _XMM3, REG32 | REG64 | MEM32 | MEM64);
|
|
put("vcvtusi2sd", XMM, XMM_ER, REG64);
|
|
|
|
put("vcvtusi2ss", XMM, _XMM3, REG32 | REG64 | MEM32 | MEM64);
|
|
put("vcvtusi2ss", XMM, XMM_ER, REG32 | REG64);
|
|
#endif
|
|
}
|
|
enum {
|
|
xx_yy_zz,
|
|
xx_yx_zy,
|
|
xx_xy_yz
|
|
};
|
|
void putGather()
|
|
{
|
|
#ifdef XBYAK64
|
|
const struct Tbl {
|
|
const char *name;
|
|
int mode;
|
|
} tbl[] = {
|
|
{ "vpgatherdd", xx_yy_zz },
|
|
{ "vpgatherdq", xx_yx_zy },
|
|
{ "vpgatherqd", xx_xy_yz },
|
|
{ "vpgatherqq", xx_yy_zz },
|
|
{ "vgatherdps", xx_yy_zz },
|
|
{ "vgatherdpd", xx_yx_zy },
|
|
{ "vgatherqps", xx_xy_yz },
|
|
{ "vgatherqpd", xx_yy_zz },
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const Tbl& p = tbl[i];
|
|
switch (p.mode) {
|
|
case xx_yy_zz:
|
|
put(p.name, XMM_K, VM32X);
|
|
put(p.name, YMM_K, VM32Y);
|
|
put(p.name, ZMM_K, VM32Z);
|
|
break;
|
|
case xx_yx_zy:
|
|
put(p.name, XMM_K, VM32X);
|
|
put(p.name, YMM_K, VM32X);
|
|
put(p.name, ZMM_K, VM32Y);
|
|
break;
|
|
case xx_xy_yz:
|
|
put(p.name, XMM_K, VM32X);
|
|
put(p.name, XMM_K, VM32Y);
|
|
put(p.name, YMM_K, VM32Z);
|
|
break;
|
|
}
|
|
}
|
|
#endif
|
|
}
|
|
void putScatter()
|
|
{
|
|
#ifdef XBYAK64
|
|
const struct Tbl {
|
|
const char *name;
|
|
int mode;
|
|
} tbl[] = {
|
|
{ "vpscatterdd", xx_yy_zz },
|
|
{ "vpscatterdq", xx_xy_yz },
|
|
{ "vpscatterqd", xx_yx_zy },
|
|
{ "vpscatterqq", xx_yy_zz },
|
|
|
|
{ "vscatterdps", xx_yy_zz },
|
|
{ "vscatterdpd", xx_xy_yz },
|
|
{ "vscatterqps", xx_yx_zy },
|
|
{ "vscatterqpd", xx_yy_zz },
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const Tbl& p = tbl[i];
|
|
switch (p.mode) {
|
|
case xx_yy_zz:
|
|
put(p.name, VM32X_K, _XMM);
|
|
put(p.name, VM32Y_K, _YMM);
|
|
put(p.name, VM32Z_K, _ZMM);
|
|
break;
|
|
case xx_yx_zy:
|
|
put(p.name, VM32X_K, _XMM);
|
|
put(p.name, VM32Y_K, _XMM);
|
|
put(p.name, VM32Z_K, _YMM);
|
|
break;
|
|
case xx_xy_yz:
|
|
put(p.name, VM32X_K, _XMM);
|
|
put(p.name, VM32X_K, _YMM);
|
|
put(p.name, VM32Y_K, _ZMM);
|
|
break;
|
|
}
|
|
}
|
|
#endif
|
|
}
|
|
void putBlend()
|
|
{
|
|
put("vblendmpd", XMM_KZ, _XMM, _XMM | _MEM | M_1to2);
|
|
put("vblendmpd", YMM_KZ, _YMM, _YMM | _MEM | M_1to4);
|
|
put("vblendmpd", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to8);
|
|
|
|
put("vblendmps", XMM_KZ, _XMM, _XMM | _MEM | M_1to4);
|
|
put("vblendmps", YMM_KZ, _YMM, _YMM | _MEM | M_1to8);
|
|
put("vblendmps", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to16);
|
|
|
|
put("vpblendmb", XMM_KZ, _XMM, _XMM | _MEM);
|
|
put("vpblendmb", YMM_KZ, _YMM, _YMM | _MEM);
|
|
put("vpblendmb", ZMM_KZ, _ZMM, _ZMM | _MEM);
|
|
|
|
put("vpblendmb", XMM_KZ, _XMM, _XMM | _MEM);
|
|
put("vpblendmb", YMM_KZ, _YMM, _YMM | _MEM);
|
|
put("vpblendmb", ZMM_KZ, _ZMM, _ZMM | _MEM);
|
|
|
|
put("vpblendmd", XMM_KZ, _XMM, _XMM | _MEM | M_1to4);
|
|
put("vpblendmd", YMM_KZ, _YMM, _YMM | _MEM | M_1to8);
|
|
put("vpblendmd", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to16);
|
|
|
|
put("vpblendmq", XMM_KZ, _XMM, _XMM | _MEM | M_1to2);
|
|
put("vpblendmq", YMM_KZ, _YMM, _YMM | _MEM | M_1to4);
|
|
put("vpblendmq", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to8);
|
|
}
|
|
void putVpcmp()
|
|
{
|
|
const uint64_t b0Tbl[] = { 0, 0, 0 };
|
|
const uint64_t b4Tbl[] = { M_1to4, M_1to8, M_1to16 };
|
|
const uint64_t b2Tbl[] = { M_1to2, M_1to4, M_1to8 };
|
|
const struct Tbl {
|
|
const char *name;
|
|
uint64_t b;
|
|
} tbl[] = {
|
|
{ "vpcmpb", 0 },
|
|
{ "vpcmpub", 0 },
|
|
{ "vpcmpw", 0 },
|
|
{ "vpcmpuw", 0 },
|
|
{ "vpcmpd", M_1to4 },
|
|
{ "vpcmpud", M_1to4 },
|
|
{ "vpcmpq", M_1to2 },
|
|
{ "vpcmpuq", M_1to2 },
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const Tbl& p = tbl[i];
|
|
const uint64_t *bTbl = p.b == 0 ? b0Tbl : p.b == M_1to4 ? b4Tbl : b2Tbl;
|
|
put(p.name, K_K, _XMM, _XMM | _MEM | bTbl[0], IMM8);
|
|
put(p.name, K_K, _YMM, _YMM | _MEM | bTbl[1], IMM8);
|
|
put(p.name, K_K, _ZMM, _ZMM | _MEM | bTbl[2], IMM8);
|
|
}
|
|
}
|
|
void putVtest()
|
|
{
|
|
const uint64_t b0Tbl[] = { 0, 0, 0 };
|
|
const uint64_t b4Tbl[] = { M_1to4, M_1to8, M_1to16 };
|
|
const uint64_t b2Tbl[] = { M_1to2, M_1to4, M_1to8 };
|
|
const struct Tbl {
|
|
const char *name;
|
|
uint64_t b;
|
|
} tbl[] = {
|
|
{ "vptestmb", 0 },
|
|
{ "vptestmw", 0 },
|
|
{ "vptestmd", M_1to4 },
|
|
{ "vptestmq", M_1to2 },
|
|
|
|
{ "vptestnmb", 0 },
|
|
{ "vptestnmw", 0 },
|
|
{ "vptestnmd", M_1to4 },
|
|
{ "vptestnmq", M_1to2 },
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const Tbl& p = tbl[i];
|
|
const uint64_t *bTbl = p.b == 0 ? b0Tbl : p.b == M_1to4 ? b4Tbl : b2Tbl;
|
|
put(p.name, K_K, _XMM, _XMM | _MEM | bTbl[0]);
|
|
put(p.name, K_K, _YMM, _YMM | _MEM | bTbl[1]);
|
|
put(p.name, K_K, _ZMM, _ZMM | _MEM | bTbl[2]);
|
|
}
|
|
}
|
|
void putCompExp()
|
|
{
|
|
{
|
|
const char *tbl[] = {
|
|
"vcompresspd",
|
|
"vcompressps",
|
|
"vpcompressd",
|
|
"vpcompressq",
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const char *name = tbl[i];
|
|
put(name, XMM_KZ | _MEM, _XMM);
|
|
put(name, YMM_KZ | _MEM, _YMM);
|
|
put(name, ZMM_KZ | _MEM, _ZMM);
|
|
}
|
|
}
|
|
{
|
|
const char *tbl[] = {
|
|
"vexpandpd",
|
|
"vexpandps",
|
|
"vpexpandd",
|
|
"vpexpandq",
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const char *name = tbl[i];
|
|
put(name, XMM_KZ, _XMM | _MEM);
|
|
put(name, YMM_KZ, _YMM | _MEM);
|
|
put(name, ZMM_KZ, _ZMM | _MEM);
|
|
}
|
|
}
|
|
}
|
|
void putPerm()
|
|
{
|
|
const uint64_t b0Tbl[] = { 0, 0, 0 };
|
|
const uint64_t b4Tbl[] = { M_1to4, M_1to8, M_1to16 };
|
|
const uint64_t b2Tbl[] = { M_1to2, M_1to4, M_1to8 };
|
|
const struct Tbl {
|
|
const char *name;
|
|
uint64_t b;
|
|
} tbl[] = {
|
|
{ "vpermt2b", 0 },
|
|
{ "vpermt2w", 0 },
|
|
{ "vpermt2d", M_1to4 },
|
|
{ "vpermt2q", M_1to2 },
|
|
{ "vpermt2ps", M_1to4 },
|
|
{ "vpermt2pd", M_1to2 },
|
|
|
|
{ "vpermi2b", 0 },
|
|
{ "vpermi2w", 0 },
|
|
{ "vpermi2d", M_1to4 },
|
|
{ "vpermi2q", M_1to2 },
|
|
{ "vpermi2ps", M_1to4 },
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
const Tbl& p = tbl[i];
|
|
const uint64_t *bTbl = p.b == 0 ? b0Tbl : p.b == M_1to4 ? b4Tbl : b2Tbl;
|
|
put(p.name, XMM_KZ, _XMM, _XMM | _MEM | bTbl[0]);
|
|
put(p.name, YMM_KZ, _YMM, _YMM | _MEM | bTbl[1]);
|
|
put(p.name, ZMM_KZ, _ZMM, _ZMM | _MEM | bTbl[2]);
|
|
}
|
|
}
|
|
void putShuff()
|
|
{
|
|
put("vshuff32x4", YMM_KZ, _YMM, _YMM | _MEM | M_1to8, IMM8);
|
|
put("vshuff32x4", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to16, IMM8);
|
|
|
|
put("vshuff64x2", YMM_KZ, _YMM, _YMM | _MEM | M_1to4, IMM8);
|
|
put("vshuff64x2", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to8, IMM8);
|
|
|
|
put("vshufi32x4", YMM_KZ, _YMM, _YMM | _MEM | M_1to8, IMM8);
|
|
put("vshufi32x4", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to16, IMM8);
|
|
|
|
put("vshufi64x2", YMM_KZ, _YMM, _YMM | _MEM | M_1to4, IMM8);
|
|
put("vshufi64x2", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to8, IMM8);
|
|
}
|
|
void putMov()
|
|
{
|
|
put("vpmovm2b", _XMM | _YMM | _ZMM, K);
|
|
put("vpmovm2w", _XMM | _YMM | _ZMM, K);
|
|
put("vpmovm2d", _XMM | _YMM | _ZMM, K);
|
|
put("vpmovm2q", _XMM | _YMM | _ZMM, K);
|
|
|
|
put("vpmovb2m", K, _XMM | _YMM | _ZMM);
|
|
put("vpmovw2m", K, _XMM | _YMM | _ZMM);
|
|
put("vpmovd2m", K, _XMM | _YMM | _ZMM);
|
|
put("vpmovq2m", K, _XMM | _YMM | _ZMM);
|
|
|
|
put("vpmovqb", XMM_KZ | _MEM, _XMM | _YMM | _ZMM);
|
|
put("vpmovsqb", XMM_KZ | _MEM, _XMM | _YMM | _ZMM);
|
|
put("vpmovusqb", XMM_KZ | _MEM, _XMM | _YMM | _ZMM);
|
|
|
|
put("vpmovqw", XMM_KZ | _MEM, _XMM | _YMM | _ZMM);
|
|
put("vpmovsqw", XMM_KZ | _MEM, _XMM | _YMM | _ZMM);
|
|
put("vpmovusqw", XMM_KZ | _MEM, _XMM | _YMM | _ZMM);
|
|
|
|
put("vpmovqd", XMM_KZ | _MEM, _XMM | _YMM);
|
|
put("vpmovqd", YMM_KZ | _MEM, _ZMM);
|
|
|
|
put("vpmovsqd", XMM_KZ | _MEM, _XMM | _YMM);
|
|
put("vpmovsqd", YMM_KZ | _MEM, _ZMM);
|
|
|
|
put("vpmovusqd", XMM_KZ | _MEM, _XMM | _YMM);
|
|
put("vpmovusqd", YMM_KZ | _MEM, _ZMM);
|
|
|
|
put("vpmovdb", XMM_KZ | _MEM, _XMM | _YMM | _ZMM);
|
|
put("vpmovsdb", XMM_KZ | _MEM, _XMM | _YMM | _ZMM);
|
|
put("vpmovusdb", XMM_KZ | _MEM, _XMM | _YMM | _ZMM);
|
|
|
|
put("vpmovdw", XMM_KZ | _MEM, _XMM | _YMM);
|
|
put("vpmovdw", YMM_KZ | _MEM, _ZMM);
|
|
|
|
put("vpmovsdw", XMM_KZ | _MEM, _XMM | _YMM);
|
|
put("vpmovsdw", YMM_KZ | _MEM, _ZMM);
|
|
|
|
put("vpmovusdw", XMM_KZ | _MEM, _XMM | _YMM);
|
|
put("vpmovusdw", YMM_KZ | _MEM, _ZMM);
|
|
|
|
put("vpmovwb", XMM_KZ | _MEM, _XMM | _YMM);
|
|
put("vpmovwb", YMM_KZ | _MEM, _ZMM);
|
|
|
|
put("vpmovswb", XMM_KZ | _MEM, _XMM | _YMM);
|
|
put("vpmovswb", YMM_KZ | _MEM, _ZMM);
|
|
|
|
put("vpmovuswb", XMM_KZ | _MEM, _XMM | _YMM);
|
|
put("vpmovuswb", YMM_KZ | _MEM, _ZMM);
|
|
}
|
|
void putRot()
|
|
{
|
|
put("vprolvd", XMM_KZ, _XMM, _XMM | _MEM | M_1to4);
|
|
put("vprolvd", YMM_KZ, _YMM, _YMM | _MEM | M_1to8);
|
|
put("vprolvd", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to16);
|
|
|
|
put("vprolvq", XMM_KZ, _XMM, _XMM | _MEM | M_1to2);
|
|
put("vprolvq", YMM_KZ, _YMM, _YMM | _MEM | M_1to4);
|
|
put("vprolvq", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to8);
|
|
|
|
put("vprorvd", XMM_KZ, _XMM, _XMM | _MEM | M_1to4);
|
|
put("vprorvd", YMM_KZ, _YMM, _YMM | _MEM | M_1to8);
|
|
put("vprorvd", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to16);
|
|
|
|
put("vprorvq", XMM_KZ, _XMM, _XMM | _MEM | M_1to2);
|
|
put("vprorvq", YMM_KZ, _YMM, _YMM | _MEM | M_1to4);
|
|
put("vprorvq", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to8);
|
|
|
|
put("vprold", XMM_KZ, _XMM | _MEM | M_1to4, IMM8);
|
|
put("vprold", YMM_KZ, _YMM | _MEM | M_1to8, IMM8);
|
|
put("vprold", ZMM_KZ, _ZMM | _MEM | M_1to16, IMM8);
|
|
|
|
put("vprolq", XMM_KZ, _XMM | _MEM | M_1to2, IMM8);
|
|
put("vprolq", YMM_KZ, _YMM | _MEM | M_1to4, IMM8);
|
|
put("vprolq", ZMM_KZ, _ZMM | _MEM | M_1to8, IMM8);
|
|
|
|
put("vprord", XMM_KZ, _XMM | _MEM | M_1to4, IMM8);
|
|
put("vprord", YMM_KZ, _YMM | _MEM | M_1to8, IMM8);
|
|
put("vprord", ZMM_KZ, _ZMM | _MEM | M_1to16, IMM8);
|
|
|
|
put("vprorq", XMM_KZ, _XMM | _MEM | M_1to2, IMM8);
|
|
put("vprorq", YMM_KZ, _YMM | _MEM | M_1to4, IMM8);
|
|
put("vprorq", ZMM_KZ, _ZMM | _MEM | M_1to8, IMM8);
|
|
}
|
|
void putMisc2()
|
|
{
|
|
#ifdef XBYAK64
|
|
put("vpternlogd", XMM_KZ, _XMM, _XMM | _MEM | M_1to4, IMM8);
|
|
put("vpternlogd", YMM_KZ, _YMM, _YMM | _MEM | M_1to8, IMM8);
|
|
put("vpternlogd", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to16, IMM8);
|
|
|
|
put("vpternlogq", XMM_KZ, _XMM, _XMM | _MEM | M_1to2, IMM8);
|
|
put("vpternlogq", YMM_KZ, _YMM, _YMM | _MEM | M_1to4, IMM8);
|
|
put("vpternlogq", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to8, IMM8);
|
|
|
|
put("vgetexppd", XMM_KZ, _XMM | MEM | M_1to2);
|
|
put("vgetexppd", YMM_KZ, _YMM | MEM | M_1to4);
|
|
put("vgetexppd", ZMM_KZ, _ZMM | MEM | M_1to8 | ZMM_SAE);
|
|
|
|
put("vgetexpps", XMM_KZ, _XMM | MEM | M_1to4);
|
|
put("vgetexpps", YMM_KZ, _YMM | MEM | M_1to8);
|
|
put("vgetexpps", ZMM_KZ, _ZMM | MEM | M_1to16 | ZMM_SAE);
|
|
|
|
put("vgetexpsd", XMM_KZ, _XMM, _XMM | _MEM | XMM_SAE);
|
|
put("vgetexpss", XMM_KZ, _XMM, _XMM | _MEM | XMM_SAE);
|
|
|
|
put("vgetmantpd", XMM_KZ, _XMM | _MEM | M_1to2, IMM8);
|
|
put("vgetmantpd", YMM_KZ, _YMM | _MEM | M_1to4, IMM8);
|
|
put("vgetmantpd", ZMM_KZ, _ZMM | _MEM | M_1to8, IMM8);
|
|
|
|
put("vgetmantps", XMM_KZ, _XMM | _MEM | M_1to4, IMM8);
|
|
put("vgetmantps", YMM_KZ, _YMM | _MEM | M_1to8, IMM8);
|
|
put("vgetmantps", ZMM_KZ, _ZMM | _MEM | M_1to16, IMM8);
|
|
|
|
put("vgetmantsd", XMM_KZ, _XMM, _XMM | _MEM | XMM_SAE, IMM8);
|
|
put("vgetmantss", XMM_KZ, _XMM, _XMM | _MEM | XMM_SAE, IMM8);
|
|
|
|
put("vfixupimmpd", XMM_KZ, _XMM, _XMM | _MEM | M_1to2, IMM8);
|
|
put("vfixupimmpd", YMM_KZ, _YMM, _YMM | _MEM | M_1to4, IMM8);
|
|
put("vfixupimmpd", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to8, IMM8);
|
|
|
|
put("vfixupimmps", XMM_KZ, _XMM, _XMM | _MEM | M_1to4, IMM8);
|
|
put("vfixupimmps", YMM_KZ, _YMM, _YMM | _MEM | M_1to8, IMM8);
|
|
put("vfixupimmps", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to16, IMM8);
|
|
|
|
put("vfixupimmsd", XMM_KZ, _XMM, _XMM | _MEM, IMM8);
|
|
put("vfixupimmss", XMM_KZ, _XMM, _XMM | _MEM, IMM8);
|
|
|
|
put("vrcp14pd", XMM_KZ, _XMM | _MEM | M_1to2);
|
|
put("vrcp14pd", YMM_KZ, _YMM | _MEM | M_1to4);
|
|
put("vrcp14pd", ZMM_KZ, _ZMM | _MEM | M_1to8);
|
|
|
|
put("vrcp14ps", XMM_KZ, _XMM | _MEM | M_1to4);
|
|
put("vrcp14ps", YMM_KZ, _YMM | _MEM | M_1to8);
|
|
put("vrcp14ps", ZMM_KZ, _ZMM | _MEM | M_1to16);
|
|
|
|
put("vrcp14sd", XMM_KZ, _XMM, _XMM | _MEM);
|
|
|
|
put("vrcp14ss", XMM_KZ, _XMM, _XMM | _MEM);
|
|
|
|
put("vrsqrt14pd", XMM_KZ, _XMM | _MEM | M_1to2);
|
|
put("vrsqrt14pd", YMM_KZ, _YMM | _MEM | M_1to4);
|
|
put("vrsqrt14pd", ZMM_KZ, _ZMM | _MEM | M_1to8);
|
|
|
|
put("vrsqrt14ps", XMM_KZ, _XMM | _MEM | M_1to4);
|
|
put("vrsqrt14ps", YMM_KZ, _YMM | _MEM | M_1to8);
|
|
put("vrsqrt14ps", ZMM_KZ, _ZMM | _MEM | M_1to16);
|
|
|
|
put("vrsqrt14sd", XMM_KZ, _XMM, _XMM | _MEM);
|
|
|
|
put("vrsqrt14ss", XMM_KZ, _XMM, _XMM | _MEM);
|
|
|
|
put("vrndscalepd", XMM_KZ, _XMM | _MEM | M_1to2, IMM8);
|
|
put("vrndscalepd", YMM_KZ, _YMM | _MEM | M_1to4, IMM8);
|
|
put("vrndscalepd", ZMM_KZ, _ZMM | _MEM | M_1to8, IMM8);
|
|
|
|
put("vrndscaleps", XMM_KZ, _XMM | _MEM | M_1to4, IMM8);
|
|
put("vrndscaleps", YMM_KZ, _YMM | _MEM | M_1to8, IMM8);
|
|
put("vrndscaleps", ZMM_KZ, _ZMM | _MEM | M_1to16, IMM8);
|
|
|
|
put("vrndscalesd", XMM_KZ, _XMM, _XMM | _MEM, IMM8);
|
|
|
|
put("vrndscaless", XMM_KZ, _XMM, _XMM | _MEM, IMM8);
|
|
|
|
put("vscalefpd", XMM_KZ, _XMM, _XMM | _MEM | M_1to2);
|
|
put("vscalefpd", YMM_KZ, _YMM, _YMM | _MEM | M_1to4);
|
|
put("vscalefpd", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to8 | ZMM_ER);
|
|
|
|
put("vscalefps", XMM_KZ, _XMM, _XMM | _MEM | M_1to4);
|
|
put("vscalefps", YMM_KZ, _YMM, _YMM | _MEM | M_1to8);
|
|
put("vscalefps", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to16 | ZMM_ER);
|
|
|
|
put("vscalefsd", XMM_KZ, _XMM, _XMM | _MEM | XMM_ER);
|
|
put("vscalefss", XMM_KZ, _XMM, _XMM | _MEM | XMM_ER);
|
|
|
|
put("vdbpsadbw", XMM_KZ, _XMM, _XMM | _MEM, IMM8);
|
|
put("vdbpsadbw", YMM_KZ, _YMM, _YMM | _MEM, IMM8);
|
|
put("vdbpsadbw", ZMM_KZ, _ZMM, _ZMM | _MEM, IMM8);
|
|
|
|
put("vpmultishiftqb", XMM_KZ, _XMM, _XMM | _MEM | M_1to2);
|
|
put("vpmultishiftqb", YMM_KZ, _YMM, _YMM | _MEM | M_1to4);
|
|
put("vpmultishiftqb", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to8);
|
|
|
|
put("vpconflictd", XMM_KZ, _XMM | _MEM | M_1to4);
|
|
put("vpconflictd", YMM_KZ, _YMM | _MEM | M_1to8);
|
|
put("vpconflictd", ZMM_KZ, _ZMM | _MEM | M_1to16);
|
|
|
|
put("vpconflictq", XMM_KZ, _XMM | _MEM | M_1to2);
|
|
put("vpconflictq", YMM_KZ, _YMM | _MEM | M_1to4);
|
|
put("vpconflictq", ZMM_KZ, _ZMM | _MEM | M_1to8);
|
|
|
|
put("vplzcntd", XMM_KZ, _XMM | _MEM | M_1to4);
|
|
put("vplzcntd", YMM_KZ, _YMM | _MEM | M_1to8);
|
|
put("vplzcntd", ZMM_KZ, _ZMM | _MEM | M_1to16);
|
|
|
|
put("vplzcntq", XMM_KZ, _XMM | _MEM | M_1to2);
|
|
put("vplzcntq", YMM_KZ, _YMM | _MEM | M_1to4);
|
|
put("vplzcntq", ZMM_KZ, _ZMM | _MEM | M_1to8);
|
|
|
|
put("vpbroadcastmb2q", _XMM | _YMM | _ZMM, K);
|
|
put("vpbroadcastmw2d", _XMM | _YMM | _ZMM, K);
|
|
|
|
put("vexp2pd", ZMM_KZ, _ZMM | _MEM | M_1to8 | ZMM_SAE);
|
|
put("vexp2ps", ZMM_KZ, _ZMM | _MEM | M_1to16 | ZMM_SAE);
|
|
|
|
put("vrcp28pd", ZMM_KZ, _ZMM | _MEM | M_1to8 | ZMM_SAE);
|
|
put("vrcp28ps", ZMM_KZ, _ZMM | _MEM | M_1to16 | ZMM_SAE);
|
|
put("vrcp28sd", XMM_KZ, _XMM, _XMM | _MEM | XMM_SAE);
|
|
put("vrcp28ss", XMM_KZ, _XMM, _XMM | _MEM | XMM_SAE);
|
|
|
|
put("vrsqrt28pd", ZMM_KZ, _ZMM | _MEM | M_1to8 | ZMM_SAE);
|
|
put("vrsqrt28ps", ZMM_KZ, _ZMM | _MEM | M_1to16 | ZMM_SAE);
|
|
put("vrsqrt28sd", XMM_KZ, _XMM, _XMM | _MEM | XMM_SAE);
|
|
put("vrsqrt28ss", XMM_KZ, _XMM, _XMM | _MEM | XMM_SAE);
|
|
|
|
put("vgatherpf0dps", VM32Z_K);
|
|
put("vgatherpf0qps", VM32Z_K);
|
|
put("vgatherpf0dpd", VM32Y_K);
|
|
put("vgatherpf0qpd", VM32Z_K);
|
|
|
|
put("vgatherpf1dps", VM32Z_K);
|
|
put("vgatherpf1qps", VM32Z_K);
|
|
put("vgatherpf1dpd", VM32Y_K);
|
|
put("vgatherpf1qpd", VM32Z_K);
|
|
|
|
put("vscatterpf0dps", VM32Z_K);
|
|
put("vscatterpf0qps", VM32Z_K);
|
|
put("vscatterpf0dpd", VM32Y_K);
|
|
put("vscatterpf0qpd", VM32Z_K);
|
|
|
|
put("vscatterpf1dps", VM32Z_K);
|
|
put("vscatterpf1qps", VM32Z_K);
|
|
put("vscatterpf1dpd", VM32Y_K);
|
|
put("vscatterpf1qpd", VM32Z_K);
|
|
|
|
put("vrangepd", XMM_KZ, _XMM, _XMM | _MEM | M_1to2, IMM8);
|
|
put("vrangepd", YMM_KZ, _YMM, _YMM | _MEM | M_1to4, IMM8);
|
|
put("vrangepd", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to8 | ZMM_SAE, IMM8);
|
|
|
|
put("vrangeps", XMM_KZ, _XMM, _XMM | _MEM | M_1to4, IMM8);
|
|
put("vrangeps", YMM_KZ, _YMM, _YMM | _MEM | M_1to8, IMM8);
|
|
put("vrangeps", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to16 | ZMM_SAE, IMM8);
|
|
|
|
put("vrangesd", XMM_KZ, _XMM, _XMM | _MEM | XMM_SAE, IMM8);
|
|
put("vrangess", XMM_KZ, _XMM, _XMM | _MEM | XMM_SAE, IMM8);
|
|
|
|
put("vreducepd", XMM_KZ, _XMM | _MEM | M_1to2, IMM8);
|
|
put("vreducepd", YMM_KZ, _YMM | _MEM | M_1to4, IMM8);
|
|
put("vreducepd", ZMM_KZ, _ZMM | _MEM | M_1to8 | ZMM_SAE, IMM8);
|
|
|
|
put("vreduceps", XMM_KZ, _XMM | _MEM | M_1to4, IMM8);
|
|
put("vreduceps", YMM_KZ, _YMM | _MEM | M_1to8, IMM8);
|
|
put("vreduceps", ZMM_KZ, _ZMM | _MEM | M_1to16 | ZMM_SAE, IMM8);
|
|
|
|
put("vreducesd", XMM_KZ, _XMM, _XMM | _MEM | XMM_SAE, IMM8);
|
|
put("vreducess", XMM_KZ, _XMM, _XMM | _MEM | XMM_SAE, IMM8);
|
|
|
|
put("vpmadd52luq", XMM_KZ, _XMM, _XMM | _MEM | M_1to2);
|
|
put("vpmadd52luq", YMM_KZ, _YMM, _YMM | _MEM | M_1to4);
|
|
put("vpmadd52luq", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to8);
|
|
|
|
put("vpmadd52huq", XMM_KZ, _XMM, _XMM | _MEM | M_1to2);
|
|
put("vpmadd52huq", YMM_KZ, _YMM, _YMM | _MEM | M_1to4);
|
|
put("vpmadd52huq", ZMM_KZ, _ZMM, _ZMM | _MEM | M_1to8);
|
|
#endif
|
|
}
|
|
void classSubMem(const char *nm, char x, bool broadcast, int size)
|
|
{
|
|
printf("%s ", nm);
|
|
if (isXbyak_) {
|
|
printf("(k5|k3, %cword%s [rax+64], 5);dump();\n", x, broadcast ? "_b" : "");
|
|
} else {
|
|
if (broadcast) {
|
|
int d = x == 'x' ? 128 / size : x == 'y' ? 256 / size : 512 / size;
|
|
printf("k5{k3}, [rax+64]{1to%d}, 5\n", d);
|
|
} else {
|
|
if (x == 'x') x = 'o'; // nasm
|
|
printf("k5{k3}, %cword [rax+64], 5\n", x);
|
|
}
|
|
}
|
|
}
|
|
void putClassSub(const char *name, int size)
|
|
{
|
|
put(name, K_K, _XMM | _YMM | _ZMM, IMM8);
|
|
for (int i = 0; i < 2; i++) {
|
|
classSubMem(name, 'x', i == 0, size);
|
|
classSubMem(name, 'y', i == 0, size);
|
|
classSubMem(name, 'z', i == 0, size);
|
|
}
|
|
}
|
|
void putClass()
|
|
{
|
|
#ifdef XBYAK64
|
|
putClassSub("vfpclasspd", 64);
|
|
putClassSub("vfpclassps", 32);
|
|
put("vfpclasssd", K_K, _XMM | _MEM, IMM8);
|
|
put("vfpclassss", K_K, _XMM | _MEM, IMM8);
|
|
#endif
|
|
}
|
|
void putMin()
|
|
{
|
|
#ifdef XBYAK64
|
|
put("vextractf32x4", XMM_KZ, _YMM, IMM8);
|
|
#endif
|
|
}
|
|
void putDisp8N()
|
|
{
|
|
{
|
|
const int tbl[] = {
|
|
-129, -128, -127, 0, 1, 64, 65, 127, 128
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
char xs[128], ns[128];
|
|
int v = tbl[i];
|
|
CYBOZU_SNPRINTF(xs, sizeof(xs), "xmm0, ptr[eax%+d]", v);
|
|
CYBOZU_SNPRINTF(ns, sizeof(ns), "xmm0, [eax%+d]", v);
|
|
put("vpbroadcastb", xs, ns);
|
|
}
|
|
}
|
|
{
|
|
const int tbl[] = {
|
|
-1024, -512 -256, -128, -64, -32, -16, -8, -4, -2, -1,
|
|
0, 1, 2, 4, 8, 16, 32, 64, 128, 256, 512
|
|
};
|
|
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
|
|
char xs[128], ns[128];
|
|
int v = tbl[i];
|
|
CYBOZU_SNPRINTF(xs, sizeof(xs), "zmm0, zmm1, ptr_b[eax%+d]", v);
|
|
CYBOZU_SNPRINTF(ns, sizeof(ns), "zmm0, zmm1, [eax%+d]{1to16}", v);
|
|
put("vaddps", xs, ns);
|
|
}
|
|
}
|
|
#ifdef XBYAK64
|
|
put("vfmadd231ps", "zmm8, zmm31, ptr_b[r14+rbp-0x1e4]", "zmm8, zmm31, [r14+rbp-0x1e4]{1to16}");
|
|
#endif
|
|
}
|
|
void putAVX512()
|
|
{
|
|
#ifdef MIN_TEST
|
|
putMin();
|
|
#else
|
|
putOpmask();
|
|
separateFunc();
|
|
putCombi();
|
|
separateFunc();
|
|
putCmpK();
|
|
separateFunc();
|
|
putBroadcast();
|
|
separateFunc();
|
|
putAVX512_M_X();
|
|
separateFunc();
|
|
put_vmov();
|
|
separateFunc();
|
|
put512_X_XM();
|
|
separateFunc();
|
|
put512_X_X_XM();
|
|
separateFunc();
|
|
put512_X3();
|
|
separateFunc();
|
|
put512_X3_I();
|
|
separateFunc();
|
|
put512_FMA();
|
|
separateFunc();
|
|
put512_Y_XM();
|
|
separateFunc();
|
|
put512_AVX1();
|
|
separateFunc();
|
|
put512_cvt();
|
|
separateFunc();
|
|
putMisc1();
|
|
separateFunc();
|
|
putGather();
|
|
separateFunc();
|
|
putBlend();
|
|
separateFunc();
|
|
putVpcmp();
|
|
separateFunc();
|
|
putVtest();
|
|
separateFunc();
|
|
putCompExp();
|
|
separateFunc();
|
|
putPerm();
|
|
separateFunc();
|
|
putShuff();
|
|
separateFunc();
|
|
putMisc2();
|
|
separateFunc();
|
|
putMov();
|
|
separateFunc();
|
|
putRot();
|
|
separateFunc();
|
|
putScatter();
|
|
separateFunc();
|
|
putClass();
|
|
putDisp8N();
|
|
#endif
|
|
}
|
|
};
|
|
|
|
int main(int argc, char *[])
|
|
{
|
|
Test test(argc > 1);
|
|
test.put();
|
|
}
|