eden-emu/xbyak
19
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

add vrsqrtph

Browse source
This commit is contained in:
MITSUNARI Shigeo 2021-09-07 09:49:53 +09:00
parent b1ff7891d4
commit 2ec6a7ab4c
3 changed files with 18 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

2
gen/gen_avx512.cpp
View file

@ -659,6 +659,8 @@ void putX_XM_IMM()
{ 0x4E, "vrsqrt14pd", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_B64, false }, { 0x4E, "vrsqrt14pd", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW1 | T_B64, false },
{ 0x4E, "vrsqrt14ps", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_B32, false }, { 0x4E, "vrsqrt14ps", T_66 | T_0F38 | T_YMM | T_MUST_EVEX | T_EW0 | T_B32, false },
{ 0x4E, "vrsqrtph", T_66 | T_MAP6 | T_YMM | T_MUST_EVEX | T_EW0 | T_B16, false },
{ 0x09, "vrndscalepd", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW1 | T_B64, true }, { 0x09, "vrndscalepd", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW1 | T_B64, true },
{ 0x08, "vrndscaleps", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW0 | T_B32, true }, { 0x08, "vrndscaleps", T_66 | T_0F3A | T_YMM | T_MUST_EVEX | T_EW0 | T_B32, true },

15
test/misc.cpp
View file

@ -976,6 +976,13 @@ CYBOZU_TEST_AUTO(vaddph)
vrcpph(zmm1, ptr_b [rax+0x40]); vrcpph(zmm1, ptr_b [rax+0x40]);
vrcpsh(xmm1, xmm3, ptr [rax+0x40]); vrcpsh(xmm1, xmm3, ptr [rax+0x40]);
vrsqrtph(xmm1, ptr [rax+0x40]);
vrsqrtph(xmm1, ptr_b [rax+0x40]);
vrsqrtph(ymm2, ptr [rax+0x40]);
vrsqrtph(ymm2, ptr_b [rax+0x40]);
vrsqrtph(zmm2, ptr [rax+0x40]);
vrsqrtph(zmm2, ptr_b [rax+0x40]);
} }
} c; } c;
const uint8_t tbl[] = { const uint8_t tbl[] = {
@ -1106,6 +1113,14 @@ CYBOZU_TEST_AUTO(vaddph)
// vrcpsh // vrcpsh
0x62, 0xf6, 0x65, 0x08, 0x4d, 0x48, 0x20, 0x62, 0xf6, 0x65, 0x08, 0x4d, 0x48, 0x20,
// vrsqrtph
0x62, 0xf6, 0x7d, 0x08, 0x4e, 0x48, 0x04,
0x62, 0xf6, 0x7d, 0x18, 0x4e, 0x48, 0x20,
0x62, 0xf6, 0x7d, 0x28, 0x4e, 0x50, 0x02,
0x62, 0xf6, 0x7d, 0x38, 0x4e, 0x50, 0x20,
0x62, 0xf6, 0x7d, 0x48, 0x4e, 0x50, 0x01,
0x62, 0xf6, 0x7d, 0x58, 0x4e, 0x50, 0x20,
}; };
const size_t n = sizeof(tbl) / sizeof(tbl[0]); const size_t n = sizeof(tbl) / sizeof(tbl[0]);
CYBOZU_TEST_EQUAL(c.getSize(), n); CYBOZU_TEST_EQUAL(c.getSize(), n);

1
xbyak/xbyak_mnemonic.h
View file

@ -2201,6 +2201,7 @@ void vrsqrt28pd(const Zmm& z, const Operand& op) { opAVX_X_XM_IMM(z, op, T_66 |
void vrsqrt28ps(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, 0xCC); } void vrsqrt28ps(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, 0xCC); }
void vrsqrt28sd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F38 | T_EW1 | T_SAE_X | T_MUST_EVEX, 0xCD); } void vrsqrt28sd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F38 | T_EW1 | T_SAE_X | T_MUST_EVEX, 0xCD); }
void vrsqrt28ss(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N4 | T_66 | T_0F38 | T_EW0 | T_SAE_X | T_MUST_EVEX, 0xCD); } void vrsqrt28ss(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N4 | T_66 | T_0F38 | T_EW0 | T_SAE_X | T_MUST_EVEX, 0xCD); }
void vrsqrtph(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_MAP6 | T_EW0 | T_YMM | T_MUST_EVEX | T_B16, 0x4E); }
void vscalefpd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_EW1 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B64, 0x2C); } void vscalefpd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_66 | T_0F38 | T_EW1 | T_YMM | T_ER_Z | T_MUST_EVEX | T_B64, 0x2C); }
void vscalefps(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_ER_Z | T_MUST_EVEX | T_B32, 0x2C); } void vscalefps(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_ER_Z | T_MUST_EVEX | T_B32, 0x2C); }
void vscalefsd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F38 | T_EW1 | T_ER_X | T_MUST_EVEX, 0x2D); } void vscalefsd(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, x2, op, T_N8 | T_66 | T_0F38 | T_EW1 | T_ER_X | T_MUST_EVEX, 0x2D); }