eden-emu/xbyak
17
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

mov vcmp{pd,ps,sd,ss} to xbyak_avx512.h

Browse source
This commit is contained in:
MITSUNARI Shigeo 2016-06-28 15:56:13 +09:00
parent 8f88dfa439
commit 5cac6dcf4f
5 changed files with 40 additions and 25 deletions
Download patch file Download diff file Expand all files Collapse all files

23
gen/gen_avx512.cpp
View file

@ -92,9 +92,32 @@ void putOpmask()
puts("void kmovd(const Reg32& r, const Opmask& k) { opVex(r, 0, k, T_L0 | T_0F | T_F2 | T_W0, 0x93); }"); puts("void kmovd(const Reg32& r, const Opmask& k) { opVex(r, 0, k, T_L0 | T_0F | T_F2 | T_W0, 0x93); }");
} }
// vcmppd(k, x, op)
void putVcmp()
{
const struct Tbl {
uint8 code;
const char *name;
int type;
bool hasIMM;
} tbl[] = {
{ 0xC2, "cmppd", T_0F | T_EVEX | T_MUST_EVEX | T_EW1 | T_SAE | T_YMM | T_66, true },
{ 0xC2, "cmpps", T_0F | T_EVEX | T_MUST_EVEX | T_EW0 | T_SAE | T_YMM, true },
{ 0xC2, "cmpsd", T_0F | T_EVEX | T_MUST_EVEX | T_EW1 | T_SAE | T_F2, true },
{ 0xC2, "cmpss", T_0F | T_EVEX | T_MUST_EVEX | T_EW0 | T_SAE | T_F3, true },
};
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
const Tbl *p = &tbl[i];
std::string type = type2String(p->type);
printf("void v%s(const Opmask& k, const Xmm& x, const Operand& op%s) { opAVX_K_X_XM(k, x, op, %s, 0x%02X%s); }\n"
, p->name, p->hasIMM ? ", uint8 imm" : "", type.c_str(), p->code, p->hasIMM ? ", imm" : "");
}
}
int main() int main()
{ {
puts("#ifdef XBYAK_AVX512"); puts("#ifdef XBYAK_AVX512");
putOpmask(); putOpmask();
putVcmp();
puts("#endif"); puts("#endif");
} }

21
gen/gen_code.cpp
View file

@ -1664,27 +1664,6 @@ void put()
printf("void %s(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_W%d, 0x%x, %d); }\n", p.name, p.w, p.code, p.mode); printf("void %s(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_W%d, 0x%x, %d); }\n", p.name, p.w, p.code, p.mode);
} }
} }
// AVX-512
// vcmppd(k, x, op)
{
const struct Tbl {
uint8 code;
const char *name;
int type;
bool hasIMM;
} tbl[] = {
{ 0xC2, "cmppd", T_0F | T_EVEX | T_MUST_EVEX | T_EW1 | T_SAE | T_YMM | T_66, true },
{ 0xC2, "cmpps", T_0F | T_EVEX | T_MUST_EVEX | T_EW0 | T_SAE | T_YMM, true },
{ 0xC2, "cmpsd", T_0F | T_EVEX | T_MUST_EVEX | T_EW1 | T_SAE | T_F2, true },
{ 0xC2, "cmpss", T_0F | T_EVEX | T_MUST_EVEX | T_EW0 | T_SAE | T_F3, true },
};
for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) {
const Tbl *p = &tbl[i];
std::string type = type2String(p->type);
printf("void v%s(const Opmask& k, const Xmm& x, const Operand& op%s) { opAVX_K_X_XM(k, x, op, %s, 0x%02X%s); }\n"
, p->name, p->hasIMM ? ", uint8 imm" : "", type.c_str(), p->code, p->hasIMM ? ", imm" : "");
}
}
} }
int main() int main()

13
readme.txt
View file

@ -125,6 +125,19 @@ vaddps(xmm2, xmm3, ptr [rax]); // メモリアクセスはptrで
vfmadd231pd(xmm1, xmm2, xmm3); // xmm1 <- (xmm2 * xmm3) + xmm1 vfmadd231pd(xmm1, xmm2, xmm3); // xmm1 <- (xmm2 * xmm3) + xmm1
・AVX-512
vaddpd zmm2, zmm5, zmm30 --> vaddpd(zmm2, zmm5, zmm30);
vaddpd zmm2{k5}, zmm4, zmm2 --> vaddpd(zmm2 | k5, zmm4, zmm2);
vaddpd zmm2{k5}{z}, zmm4, zmm2 --> vaddpd(zmm2 | k5 | T_z, zmm4, zmm2);
vaddpd zmm2{k5}{z}, zmm4, zmm2,{rd-sae} --> vaddpd(zmm2 | k5 | T_z, zmm4, zmm2 | T_rd_sae);
vcmppd k4{k3}, zmm1, zmm2, {sae}, 5 --> vcmppd(k4 | k3, zmm1, zmm2 | T_sae, 5);
注意
* k1, ..., k7 は新しいopmaskレジスタです。
* z, sae, rn-sae, rd-sae, ru-sae, rz-saeの代わりにT_z, T_sae, T_rn_sae, T_rd_sae, T_ru_sae, T_rz_saeを使ってください。
* `k4 | k3`と`k3 | k4`は意味が異なります。
・ラベル ・ラベル
L(文字列); L(文字列);

4
xbyak/xbyak_avx512.h
View file

@ -64,4 +64,8 @@ void kmovb(const Opmask& k, const Reg32& r) { opVex(k, 0, r, T_L0 | T_0F | T_66
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 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 vcmppd(const Opmask& k, const Xmm& x, const Operand& op, uint8 imm) { opAVX_K_X_XM(k, x, op, T_0F | T_66 | T_EW1 | T_YMM | T_EVEX | T_SAE | T_MUST_EVEX, 0xC2, imm); }
void vcmpps(const Opmask& k, const Xmm& x, const Operand& op, uint8 imm) { opAVX_K_X_XM(k, x, op, T_0F | T_EW0 | T_YMM | T_EVEX | T_SAE | T_MUST_EVEX, 0xC2, imm); }
void vcmpsd(const Opmask& k, const Xmm& x, const Operand& op, uint8 imm) { opAVX_K_X_XM(k, x, op, T_0F | T_F2 | T_EW1 | T_EVEX | T_SAE | T_MUST_EVEX, 0xC2, imm); }
void vcmpss(const Opmask& k, const Xmm& x, const Operand& op, uint8 imm) { opAVX_K_X_XM(k, x, op, T_0F | T_F3 | T_EW0 | T_EVEX | T_SAE | T_MUST_EVEX, 0xC2, imm); }
#endif #endif

4
xbyak/xbyak_mnemonic.h
View file

@ -1519,7 +1519,3 @@ void vpgatherdd(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1
void vpgatherqd(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_W0, 0x91, 2); } void vpgatherqd(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_W0, 0x91, 2); }
void vpgatherdq(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_W1, 0x90, 0); } void vpgatherdq(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_W1, 0x90, 0); }
void vpgatherqq(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_W1, 0x91, 1); } void vpgatherqq(const Xmm& x1, const Address& addr, const Xmm& x2) { opGather(x1, addr, x2, T_0F38 | T_66 | T_W1, 0x91, 1); }
void vcmppd(const Opmask& k, const Xmm& x, const Operand& op, uint8 imm) { opAVX_K_X_XM(k, x, op, T_0F | T_66 | T_EW1 | T_YMM | T_EVEX | T_SAE | T_MUST_EVEX, 0xC2, imm); }
void vcmpps(const Opmask& k, const Xmm& x, const Operand& op, uint8 imm) { opAVX_K_X_XM(k, x, op, T_0F | T_EW0 | T_YMM | T_EVEX | T_SAE | T_MUST_EVEX, 0xC2, imm); }
void vcmpsd(const Opmask& k, const Xmm& x, const Operand& op, uint8 imm) { opAVX_K_X_XM(k, x, op, T_0F | T_F2 | T_EW1 | T_EVEX | T_SAE | T_MUST_EVEX, 0xC2, imm); }
void vcmpss(const Opmask& k, const Xmm& x, const Operand& op, uint8 imm) { opAVX_K_X_XM(k, x, op, T_0F | T_F3 | T_EW0 | T_EVEX | T_SAE | T_MUST_EVEX, 0xC2, imm); }