add vmovsh

This commit is contained in:
MITSUNARI Shigeo 2021-09-11 20:26:12 +09:00
parent 847166cef1
commit 61f85a2048
3 changed files with 23 additions and 3 deletions

View file

@ -200,6 +200,8 @@ void putX_XM()
{ 0x42, "vgetexppd", T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW1 | T_B64 | T_SAE_Z }, { 0x42, "vgetexppd", T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW1 | T_B64 | T_SAE_Z },
{ 0x42, "vgetexpps", T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_B32 | T_SAE_Z }, { 0x42, "vgetexpps", T_66 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_B32 | T_SAE_Z },
{ 0x42, "vgetexpph", T_66 | T_MAP6 | T_MUST_EVEX | T_YMM | T_EW0 | T_B16 | T_SAE_Z }, { 0x42, "vgetexpph", T_66 | T_MAP6 | T_MUST_EVEX | T_YMM | T_EW0 | T_B16 | T_SAE_Z },
{ 0x10, "vmovsh", T_F3 | T_MAP5 | T_MUST_EVEX | T_EW0 | T_N2 },
}; };
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) { for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
const Tbl *p = &tbl[i]; const Tbl *p = &tbl[i];
@ -230,6 +232,7 @@ void putM_X()
{ 0x7F, "vmovdqu16", T_F2 | T_0F | T_MUST_EVEX | T_YMM | T_EW1 | T_ER_X | T_ER_Y | T_ER_Z | T_M_K }, { 0x7F, "vmovdqu16", T_F2 | T_0F | T_MUST_EVEX | T_YMM | T_EW1 | T_ER_X | T_ER_Y | T_ER_Z | T_M_K },
{ 0x7F, "vmovdqu32", T_F3 | T_0F | T_MUST_EVEX | T_YMM | T_EW0 | T_ER_X | T_ER_Y | T_ER_Z | T_M_K }, { 0x7F, "vmovdqu32", T_F3 | T_0F | T_MUST_EVEX | T_YMM | T_EW0 | T_ER_X | T_ER_Y | T_ER_Z | T_M_K },
{ 0x7F, "vmovdqu64", T_F3 | T_0F | T_MUST_EVEX | T_YMM | T_EW1 | T_ER_X | T_ER_Y | T_ER_Z | T_M_K }, { 0x7F, "vmovdqu64", T_F3 | T_0F | T_MUST_EVEX | T_YMM | T_EW1 | T_ER_X | T_ER_Y | T_ER_Z | T_M_K },
{ 0x11, "vmovsh", T_F3 | T_MAP5 | T_MUST_EVEX | T_EW0 | T_N2 | T_M_K },
}; };
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) { for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
const Tbl *p = &tbl[i]; const Tbl *p = &tbl[i];
@ -864,11 +867,21 @@ void putFP16_FMA2()
} }
} }
void putFP16_2()
{
{
int t = T_F3 | T_MAP5 | T_MUST_EVEX | T_EW0;
std::string type = type2String(t);
printf("void vmovsh(const Xmm& x1, const Xmm& x2, const Xmm& x3) { opAVX_X_X_XM(x1, x2, x3, %s, 0x10); }\n", type.c_str());
}
}
void putFP16() void putFP16()
{ {
putFP16_1(); putFP16_1();
putFP16_FMA(); putFP16_FMA();
putFP16_FMA2(); putFP16_FMA2();
putFP16_2();
} }
int main(int argc, char *[]) int main(int argc, char *[])

View file

@ -1053,9 +1053,9 @@ CYBOZU_TEST_AUTO(vaddph)
vgetmantsh(xmm1, xmm5, ptr [rax+0x40], 0x5); vgetmantsh(xmm1, xmm5, ptr [rax+0x40], 0x5);
vgetmantsh(xmm1|k1|T_z|T_sae, xmm3, xmm5, 0x6); vgetmantsh(xmm1|k1|T_z|T_sae, xmm3, xmm5, 0x6);
vmovsh(xmm1|k1|T_z, ptr [rax+0x40]);
vmovsh(ptr [rax+0x40]|k1, xmm1);
vmovsh(xmm1|k2|T_z, xmm3, xmm5);
} }
} c; } c;
@ -1281,6 +1281,10 @@ CYBOZU_TEST_AUTO(vaddph)
0x62, 0xf3, 0x54, 0x08, 0x27, 0x48, 0x20, 0x05, 0x62, 0xf3, 0x54, 0x08, 0x27, 0x48, 0x20, 0x05,
0x62, 0xf3, 0x64, 0x99, 0x27, 0xcd, 0x06, 0x62, 0xf3, 0x64, 0x99, 0x27, 0xcd, 0x06,
// vmovsh
0x62, 0xf5, 0x7e, 0x89, 0x10, 0x48, 0x20,
0x62, 0xf5, 0x7e, 0x09, 0x11, 0x48, 0x20,
0x62, 0xf5, 0x66, 0x8a, 0x10, 0xcd,
}; };
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);

View file

@ -2030,6 +2030,9 @@ 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 vmovsh(const Address& addr, const Xmm& x) { opAVX_X_XM_IMM(x, addr, T_N2 | T_F3 | T_MAP5 | T_EW0 | T_MUST_EVEX | T_M_K, 0x11); }
void vmovsh(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_N2 | T_F3 | T_MAP5 | T_EW0 | T_MUST_EVEX, 0x10); }
void vmovsh(const Xmm& x1, const Xmm& x2, const Xmm& x3) { opAVX_X_X_XM(x1, x2, x3, T_F3 | T_MAP5 | T_EW0 | T_MUST_EVEX, 0x10); }
void vmulph(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_MAP5 | T_EW0 | T_YMM | T_MUST_EVEX | T_ER_Z | T_B16, 0x59); } void vmulph(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_MAP5 | T_EW0 | T_YMM | T_MUST_EVEX | T_ER_Z | T_B16, 0x59); }
void vmulsh(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_MAP5 | T_F3 | T_EW0 | T_MUST_EVEX | T_ER_X | T_N2, 0x59); } void vmulsh(const Xmm& xmm, const Operand& op1, const Operand& op2 = Operand()) { opAVX_X_X_XM(xmm, op1, op2, T_MAP5 | T_F3 | T_EW0 | T_MUST_EVEX | T_ER_X | T_N2, 0x59); }
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 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); }