SwitchEmu/dynarmic
1
0
Fork 0
mirror of https://git.suyu.dev/suyu/dynarmic.git synced 2026-03-25 08:12:57 +00:00
Code Issues Packages Projects Releases Wiki Activity

Squashed 'externals/xbyak/' changes from d512551e..2794cde7

Browse Source 
2794cde7 add xword, yword, etc. in Xbyak::util
fb9c04e4 fix document for vfpclassps
a51be78b fix test dependency
04fdfb1e update version
e6354f8b add vgf2p8mulb
09a12642 add gf2p8affineqb
d171ba0e add gf2p8affineinvqb
457f4fd0 add vpshufbitqmb
5af0ba39 add vpexpand{b,w}
e450f965 vpopcnt{d,q} supports ptr_b
48499eb1 add vpdpbusd(s), vpdpwssd(s)
9c745109 add vpdpbusd, vpdpbusds
0e1a11b4 add vpopcnt{b,w,d,q}
9acfc132 add vpshrd(v){w,d,q}
ac8de850 add vpshld(v){w,d,q}
f181c259 add vcompressb, vcompressw
5a402477 vpclmulqdq supports AVX-512
9e16b40b vaes* supports AVX-512
7fde08e0 add flags for intel's manual 319433-030.pdf
c5da3778 add test of v4fmaddps, vp4dpwssd, etc.
e4fc9d8a fix mpx encoding
d0b2fb62 add bnd(0xf2) prefix for MPX
f12b5678 use db for array
cd74ab44 remove bat file

git-subtree-dir: externals/xbyak
git-subtree-split: 2794cde79eb71e86490061cac9622ad0067b8d15
This commit is contained in:
MerryMage
2020-04-22 20:45:52 +01:00
parent 4ed09fda06
commit 9fb82036ca
16 changed files with 935 additions and 173 deletions
Download Patch File Download Diff File Expand all files Collapse all files

35
gen/gen_avx512.cpp
View File

@@ -228,6 +228,9 @@ void putXM_X()
{ 0x8B, "vpcompressd", T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N4 },
{ 0x8B, "vpcompressq", T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW1 | T_N8 },
{ 0x63, "vcompressb", T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N1 },
{ 0x63, "vcompressw", T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW1 | T_N2 },
};
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
const Tbl *p = &tbl[i];
@@ -343,6 +346,28 @@ void putX_X_XM_IMM()
{ 0xB4, "vpmadd52luq", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_B64, false },
{ 0xB5, "vpmadd52huq", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_B64, false },
{ 0x70, "vpshldw", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW1 | T_SAE_Z, true },
{ 0x71, "vpshldd", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_B32, true },
{ 0x71, "vpshldq", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW1 | T_SAE_Z | T_B64, true },
{ 0x70, "vpshldvw", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_SAE_Z, false },
{ 0x71, "vpshldvd", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_B32, false },
{ 0x71, "vpshldvq", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_SAE_Z | T_B64, false },
{ 0x72, "vpshrdw", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW1 | T_SAE_Z, true },
{ 0x73, "vpshrdd", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_B32, true },
{ 0x73, "vpshrdq", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW1 | T_SAE_Z | T_B64, true },
{ 0x72, "vpshrdvw", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_SAE_Z, false },
{ 0x73, "vpshrdvd", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_B32, false },
{ 0x73, "vpshrdvq", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_SAE_Z | T_B64, false },
{ 0x50, "vpdpbusd", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_B32, false },
{ 0x51, "vpdpbusds", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_B32, false },
{ 0x52, "vpdpwssd", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_B32, false },
{ 0x53, "vpdpwssds", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_B32, false },
};
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
const Tbl *p = &tbl[i];
@@ -623,6 +648,14 @@ void putX_XM_IMM()
{ 0x56, "vreducepd", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW1 | T_B64 | T_SAE_Z, true },
{ 0x56, "vreduceps", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW0 | T_B32 | T_SAE_Z, true },
{ 0x54, "vpopcntb", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z, false },
{ 0x54, "vpopcntw", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_SAE_Z, false },
{ 0x55, "vpopcntd", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_B32, false },
{ 0x55, "vpopcntq", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_SAE_Z | T_B64, false },
{ 0x62, "vpexpandb", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_SAE_Z | T_N1, false },
{ 0x62, "vpexpandw", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_SAE_Z | T_N2, false },
};
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
const Tbl *p = &tbl[i];
@@ -661,6 +694,8 @@ void putMisc()
puts("void vfpclassps(const Opmask& k, const Operand& op, uint8 imm) { if (!op.isBit(128|256|512)) throw Error(ERR_BAD_MEM_SIZE); Reg x = k; x.setBit(op.getBit()); opVex(x, 0, op, T_66 | T_0F3A | T_MUST_EVEX | T_YMM | T_EW0 | T_B32, 0x66, imm); }");
puts("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); }");
puts("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); }");
puts("void vpshufbitqmb(const Opmask& k, const Xmm& x, const Operand& op) { opVex(k, &x, op, T_66 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x8F); }");
}
void putV4FMA()

19
gen/gen_code.cpp
View File

@@ -60,7 +60,7 @@ void putX_X_XM(bool omitOnly)
{ 0x02, "pblendd", T_0F3A | T_66 | T_W0 | T_YMM, true, true, 2 },
{ 0x0B, "roundsd", T_0F3A | T_66 | T_W0, true, true, 3 },
{ 0x0A, "roundss", T_0F3A | T_66 | T_W0, true, true, 3 },
{ 0x44, "pclmulqdq", T_0F3A | T_66 | T_W0, true, true, 3 },
{ 0x44, "pclmulqdq", T_0F3A | T_66 | T_W0 | T_YMM | T_EVEX, true, true, 3 },
{ 0x0C, "permilps", T_0F38 | T_66 | T_W0 | T_YMM | T_EVEX | T_EW0 | T_B32, false, false, 2 },
{ 0x0D, "permilpd", T_0F38 | T_66 | T_W0 | T_YMM | T_EVEX | T_EW1 | T_B64, false, false, 2 },
@@ -202,6 +202,10 @@ void putX_X_XM(bool omitOnly)
{ 0x14, "unpcklpd", T_0F | T_66 | T_YMM | T_EVEX | T_EW1 | T_B64, false, true, 2 },
{ 0x14, "unpcklps", T_0F | T_YMM | T_EVEX | T_EW0 | T_B32, false, true, 2 },
{ 0xCF, "gf2p8affineinvqb", T_66 | T_0F3A | T_W1 | T_EVEX | T_YMM | T_EW1 | T_SAE_Z | T_B64, true, false, 3 },
{ 0xCE, "gf2p8affineqb", T_66 | T_0F3A | T_W1 | T_EVEX | T_YMM | T_EW1 | T_SAE_Z | T_B64, true, false, 3 },
{ 0xCF, "gf2p8mulb", T_66 | T_0F38 | T_W0 | T_EVEX | T_YMM | T_EW0 | T_SAE_Z, false, false, 3 },
};
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
const Tbl *p = &tbl[i];
@@ -615,6 +619,7 @@ void put()
////////////////////////////////////////////////////////////////
{
const GenericTbl tbl[] = {
{ "bnd", 0xf2 }, /* 0xf2 prefix for MPX */
{ "cbw", 0x66, 0x98 },
{ "cdq", 0x99 },
{ "clc", 0xF8 },
@@ -941,11 +946,11 @@ void put()
puts("void bndcl(const BoundsReg& bnd, const Operand& op) { db(0xF3); opR_ModM(op, i32e, bnd.getIdx(), 0x0F, 0x1A, NONE, !op.isMEM()); }");
puts("void bndcu(const BoundsReg& bnd, const Operand& op) { db(0xF2); opR_ModM(op, i32e, bnd.getIdx(), 0x0F, 0x1A, NONE, !op.isMEM()); }");
puts("void bndcn(const BoundsReg& bnd, const Operand& op) { db(0xF2); opR_ModM(op, i32e, bnd.getIdx(), 0x0F, 0x1B, NONE, !op.isMEM()); }");
puts("void bndldx(const BoundsReg& bnd, const Address& addr) { opModM(addr, bnd, 0x0F, 0x1A); }");
puts("void bndldx(const BoundsReg& bnd, const Address& addr) { opMIB(addr, bnd, 0x0F, 0x1A); }");
puts("void bndmk(const BoundsReg& bnd, const Address& addr) { db(0xF3); opModM(addr, bnd, 0x0F, 0x1B); }");
puts("void bndmov(const BoundsReg& bnd, const Operand& op) { db(0x66); opModRM(bnd, op, op.isBNDREG(), op.isMEM(), 0x0F, 0x1A); }");
puts("void bndmov(const Address& addr, const BoundsReg& bnd) { db(0x66); opModM(addr, bnd, 0x0F, 0x1B); }");
puts("void bndstx(const Address& addr, const BoundsReg& bnd) { opModM(addr, bnd, 0x0F, 0x1B); }");
puts("void bndstx(const Address& addr, const BoundsReg& bnd) { opMIB(addr, bnd, 0x0F, 0x1B); }");
}
// misc
{
@@ -1257,10 +1262,10 @@ void put()
{ 0x7D, "hsubpd", T_0F | T_66 | T_YMM, 3 },
{ 0x7D, "hsubps", T_0F | T_F2 | T_YMM, 3 },
{ 0xDC, "aesenc", T_0F38 | T_66 | T_W0, 3 },
{ 0xDD, "aesenclast", T_0F38 | T_66 | T_W0, 3 },
{ 0xDE, "aesdec", T_0F38 | T_66 | T_W0, 3 },
{ 0xDF, "aesdeclast", T_0F38 | T_66 | T_W0, 3 },
{ 0xDC, "aesenc", T_0F38 | T_66 | T_YMM | T_EVEX, 3 },
{ 0xDD, "aesenclast", T_0F38 | T_66 | T_YMM | T_EVEX, 3 },
{ 0xDE, "aesdec", T_0F38 | T_66 | T_YMM | T_EVEX, 3 },
{ 0xDF, "aesdeclast", T_0F38 | T_66 | T_YMM | T_EVEX, 3 },
};
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
const Tbl *p = &tbl[i];