diff --git a/gen/gen_avx512.cpp b/gen/gen_avx512.cpp index 41b28e9..8283a54 100644 --- a/gen/gen_avx512.cpp +++ b/gen/gen_avx512.cpp @@ -692,29 +692,29 @@ void putMov() int type; int mode; } tbl[] = { - { 0x32, "vpmovqb", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N2 | T_N_VL, false }, - { 0x22, "vpmovsqb", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N2 | T_N_VL, false }, - { 0x12, "vpmovusqb", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N2 | T_N_VL, false }, + { 0x32, "vpmovqb", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N2 | T_N_VL | T_M_K, false }, + { 0x22, "vpmovsqb", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N2 | T_N_VL | T_M_K, false }, + { 0x12, "vpmovusqb", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N2 | T_N_VL | T_M_K, false }, - { 0x34, "vpmovqw", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N4 | T_N_VL, false }, - { 0x24, "vpmovsqw", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N4 | T_N_VL, false }, - { 0x14, "vpmovusqw", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N4 | T_N_VL, false }, + { 0x34, "vpmovqw", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N4 | T_N_VL | T_M_K, false }, + { 0x24, "vpmovsqw", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N4 | T_N_VL | T_M_K, false }, + { 0x14, "vpmovusqw", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N4 | T_N_VL | T_M_K, false }, - { 0x35, "vpmovqd", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N8 | T_N_VL, true }, - { 0x25, "vpmovsqd", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N8 | T_N_VL, true }, - { 0x15, "vpmovusqd", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N8 | T_N_VL, true }, + { 0x35, "vpmovqd", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N8 | T_N_VL | T_M_K, true }, + { 0x25, "vpmovsqd", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N8 | T_N_VL | T_M_K, true }, + { 0x15, "vpmovusqd", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N8 | T_N_VL | T_M_K, true }, - { 0x31, "vpmovdb", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N4 | T_N_VL, false }, - { 0x21, "vpmovsdb", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N4 | T_N_VL, false }, - { 0x11, "vpmovusdb", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N4 | T_N_VL, false }, + { 0x31, "vpmovdb", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N4 | T_N_VL | T_M_K, false }, + { 0x21, "vpmovsdb", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N4 | T_N_VL | T_M_K, false }, + { 0x11, "vpmovusdb", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N4 | T_N_VL | T_M_K, false }, - { 0x33, "vpmovdw", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N8 | T_N_VL, true }, - { 0x23, "vpmovsdw", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N8 | T_N_VL, true }, - { 0x13, "vpmovusdw", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N8 | T_N_VL, true }, + { 0x33, "vpmovdw", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N8 | T_N_VL | T_M_K, true }, + { 0x23, "vpmovsdw", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N8 | T_N_VL | T_M_K, true }, + { 0x13, "vpmovusdw", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N8 | T_N_VL | T_M_K, true }, - { 0x30, "vpmovwb", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N8 | T_N_VL, true }, - { 0x20, "vpmovswb", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N8 | T_N_VL, true }, - { 0x10, "vpmovuswb", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N8 | T_N_VL, true }, + { 0x30, "vpmovwb", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N8 | T_N_VL | T_M_K, true }, + { 0x20, "vpmovswb", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N8 | T_N_VL | T_M_K, true }, + { 0x10, "vpmovuswb", T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0 | T_N8 | T_N_VL | T_M_K, true }, }; for (size_t i = 0; i < NUM_OF_ARRAY(tbl); i++) { const Tbl& p = tbl[i]; diff --git a/test/make_512.cpp b/test/make_512.cpp index 83994ab..39bfa99 100644 --- a/test/make_512.cpp +++ b/test/make_512.cpp @@ -1807,44 +1807,44 @@ public: 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("vpmovqb", XMM_KZ | _MEM | MEM_K, _XMM | _YMM | _ZMM); + put("vpmovsqb", XMM_KZ | _MEM | MEM_K, _XMM | _YMM | _ZMM); + put("vpmovusqb", XMM_KZ | _MEM | MEM_K, _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("vpmovqw", XMM_KZ | _MEM | MEM_K, _XMM | _YMM | _ZMM); + put("vpmovsqw", XMM_KZ | _MEM | MEM_K, _XMM | _YMM | _ZMM); + put("vpmovusqw", XMM_KZ | _MEM | MEM_K, _XMM | _YMM | _ZMM); - put("vpmovqd", XMM_KZ | _MEM, _XMM | _YMM); - put("vpmovqd", YMM_KZ | _MEM, _ZMM); + put("vpmovqd", XMM_KZ | _MEM | MEM_K, _XMM | _YMM); + put("vpmovqd", YMM_KZ | _MEM | MEM_K, _ZMM); - put("vpmovsqd", XMM_KZ | _MEM, _XMM | _YMM); - put("vpmovsqd", YMM_KZ | _MEM, _ZMM); + put("vpmovsqd", XMM_KZ | _MEM | MEM_K, _XMM | _YMM); + put("vpmovsqd", YMM_KZ | _MEM | MEM_K, _ZMM); - put("vpmovusqd", XMM_KZ | _MEM, _XMM | _YMM); - put("vpmovusqd", YMM_KZ | _MEM, _ZMM); + put("vpmovusqd", XMM_KZ | _MEM | MEM_K, _XMM | _YMM); + put("vpmovusqd", YMM_KZ | _MEM | MEM_K, _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("vpmovdb", XMM_KZ | _MEM | MEM_K, _XMM | _YMM | _ZMM); + put("vpmovsdb", XMM_KZ | _MEM | MEM_K, _XMM | _YMM | _ZMM); + put("vpmovusdb", XMM_KZ | _MEM | MEM_K, _XMM | _YMM | _ZMM); - put("vpmovdw", XMM_KZ | _MEM, _XMM | _YMM); - put("vpmovdw", YMM_KZ | _MEM, _ZMM); + put("vpmovdw", XMM_KZ | _MEM | MEM_K, _XMM | _YMM); + put("vpmovdw", YMM_KZ | _MEM | MEM_K, _ZMM); - put("vpmovsdw", XMM_KZ | _MEM, _XMM | _YMM); - put("vpmovsdw", YMM_KZ | _MEM, _ZMM); + put("vpmovsdw", XMM_KZ | _MEM | MEM_K, _XMM | _YMM); + put("vpmovsdw", YMM_KZ | _MEM | MEM_K, _ZMM); - put("vpmovusdw", XMM_KZ | _MEM, _XMM | _YMM); - put("vpmovusdw", YMM_KZ | _MEM, _ZMM); + put("vpmovusdw", XMM_KZ | _MEM | MEM_K, _XMM | _YMM); + put("vpmovusdw", YMM_KZ | _MEM | MEM_K, _ZMM); - put("vpmovwb", XMM_KZ | _MEM, _XMM | _YMM); - put("vpmovwb", YMM_KZ | _MEM, _ZMM); + put("vpmovwb", XMM_KZ | _MEM | MEM_K, _XMM | _YMM); + put("vpmovwb", YMM_KZ | _MEM | MEM_K, _ZMM); - put("vpmovswb", XMM_KZ | _MEM, _XMM | _YMM); - put("vpmovswb", YMM_KZ | _MEM, _ZMM); + put("vpmovswb", XMM_KZ | _MEM | MEM_K, _XMM | _YMM); + put("vpmovswb", YMM_KZ | _MEM | MEM_K, _ZMM); - put("vpmovuswb", XMM_KZ | _MEM, _XMM | _YMM); - put("vpmovuswb", YMM_KZ | _MEM, _ZMM); + put("vpmovuswb", XMM_KZ | _MEM | MEM_K, _XMM | _YMM); + put("vpmovuswb", YMM_KZ | _MEM | MEM_K, _ZMM); } void putRot() { diff --git a/xbyak/xbyak_mnemonic.h b/xbyak/xbyak_mnemonic.h index bf2ab70..52664d0 100644 --- a/xbyak/xbyak_mnemonic.h +++ b/xbyak/xbyak_mnemonic.h @@ -2163,30 +2163,30 @@ void vpminsq(const Xmm& x1, const Xmm& x2, const Operand& op) { opAVX_X_X_XM(x1, void vpminuq(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_MUST_EVEX | T_B64, 0x3B); } void vpmovb2m(const Opmask& k, const Xmm& x) { opVex(k, 0, x, T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0, 0x29); } void vpmovd2m(const Opmask& k, const Xmm& x) { opVex(k, 0, x, T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0, 0x39); } -void vpmovdb(const Operand& op, const Xmm& x) { opVmov(op, x, T_N4 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x31, false); } -void vpmovdw(const Operand& op, const Xmm& x) { opVmov(op, x, T_N8 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x33, true); } +void vpmovdb(const Operand& op, const Xmm& x) { opVmov(op, x, T_N4 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0x31, false); } +void vpmovdw(const Operand& op, const Xmm& x) { opVmov(op, x, T_N8 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0x33, true); } void vpmovm2b(const Xmm& x, const Opmask& k) { opVex(x, 0, k, T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0, 0x28); } void vpmovm2d(const Xmm& x, const Opmask& k) { opVex(x, 0, k, T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW0, 0x38); } void vpmovm2q(const Xmm& x, const Opmask& k) { opVex(x, 0, k, T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW1, 0x38); } void vpmovm2w(const Xmm& x, const Opmask& k) { opVex(x, 0, k, T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW1, 0x28); } void vpmovq2m(const Opmask& k, const Xmm& x) { opVex(k, 0, x, T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW1, 0x39); } -void vpmovqb(const Operand& op, const Xmm& x) { opVmov(op, x, T_N2 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x32, false); } -void vpmovqd(const Operand& op, const Xmm& x) { opVmov(op, x, T_N8 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x35, true); } -void vpmovqw(const Operand& op, const Xmm& x) { opVmov(op, x, T_N4 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x34, false); } -void vpmovsdb(const Operand& op, const Xmm& x) { opVmov(op, x, T_N4 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x21, false); } -void vpmovsdw(const Operand& op, const Xmm& x) { opVmov(op, x, T_N8 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x23, true); } -void vpmovsqb(const Operand& op, const Xmm& x) { opVmov(op, x, T_N2 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x22, false); } -void vpmovsqd(const Operand& op, const Xmm& x) { opVmov(op, x, T_N8 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x25, true); } -void vpmovsqw(const Operand& op, const Xmm& x) { opVmov(op, x, T_N4 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x24, false); } -void vpmovswb(const Operand& op, const Xmm& x) { opVmov(op, x, T_N8 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x20, true); } -void vpmovusdb(const Operand& op, const Xmm& x) { opVmov(op, x, T_N4 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x11, false); } -void vpmovusdw(const Operand& op, const Xmm& x) { opVmov(op, x, T_N8 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x13, true); } -void vpmovusqb(const Operand& op, const Xmm& x) { opVmov(op, x, T_N2 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x12, false); } -void vpmovusqd(const Operand& op, const Xmm& x) { opVmov(op, x, T_N8 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x15, true); } -void vpmovusqw(const Operand& op, const Xmm& x) { opVmov(op, x, T_N4 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x14, false); } -void vpmovuswb(const Operand& op, const Xmm& x) { opVmov(op, x, T_N8 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x10, true); } +void vpmovqb(const Operand& op, const Xmm& x) { opVmov(op, x, T_N2 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0x32, false); } +void vpmovqd(const Operand& op, const Xmm& x) { opVmov(op, x, T_N8 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0x35, true); } +void vpmovqw(const Operand& op, const Xmm& x) { opVmov(op, x, T_N4 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0x34, false); } +void vpmovsdb(const Operand& op, const Xmm& x) { opVmov(op, x, T_N4 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0x21, false); } +void vpmovsdw(const Operand& op, const Xmm& x) { opVmov(op, x, T_N8 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0x23, true); } +void vpmovsqb(const Operand& op, const Xmm& x) { opVmov(op, x, T_N2 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0x22, false); } +void vpmovsqd(const Operand& op, const Xmm& x) { opVmov(op, x, T_N8 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0x25, true); } +void vpmovsqw(const Operand& op, const Xmm& x) { opVmov(op, x, T_N4 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0x24, false); } +void vpmovswb(const Operand& op, const Xmm& x) { opVmov(op, x, T_N8 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0x20, true); } +void vpmovusdb(const Operand& op, const Xmm& x) { opVmov(op, x, T_N4 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0x11, false); } +void vpmovusdw(const Operand& op, const Xmm& x) { opVmov(op, x, T_N8 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0x13, true); } +void vpmovusqb(const Operand& op, const Xmm& x) { opVmov(op, x, T_N2 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0x12, false); } +void vpmovusqd(const Operand& op, const Xmm& x) { opVmov(op, x, T_N8 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0x15, true); } +void vpmovusqw(const Operand& op, const Xmm& x) { opVmov(op, x, T_N4 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0x14, false); } +void vpmovuswb(const Operand& op, const Xmm& x) { opVmov(op, x, T_N8 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0x10, true); } void vpmovw2m(const Opmask& k, const Xmm& x) { opVex(k, 0, x, T_F3 | T_0F38 | T_MUST_EVEX | T_YMM | T_EW1, 0x29); } -void vpmovwb(const Operand& op, const Xmm& x) { opVmov(op, x, T_N8 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX, 0x30, true); } +void vpmovwb(const Operand& op, const Xmm& x) { opVmov(op, x, T_N8 | T_N_VL | T_F3 | T_0F38 | T_EW0 | T_YMM | T_MUST_EVEX | T_M_K, 0x30, true); } void vpmullq(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_MUST_EVEX | T_B64, 0x40); } void vpmultishiftqb(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_MUST_EVEX | T_B64, 0x83); } void vpopcntb(const Xmm& x, const Operand& op) { opAVX_X_XM_IMM(x, op, T_66 | T_0F38 | T_EW0 | T_YMM | T_SAE_Z | T_MUST_EVEX, 0x54); }