diff options
Diffstat (limited to 'opcodes/aarch64-tbl.h')
-rw-r--r-- | opcodes/aarch64-tbl.h | 112 |
1 files changed, 112 insertions, 0 deletions
diff --git a/opcodes/aarch64-tbl.h b/opcodes/aarch64-tbl.h index 5876592..da17fa3 100644 --- a/opcodes/aarch64-tbl.h +++ b/opcodes/aarch64-tbl.h @@ -2218,6 +2218,10 @@ { \ QLF3(S_S,S_B,S_B), \ } +#define OP_SVE_VVV_H_B \ +{ \ + QLF3(S_H,S_B,S_B), \ +} #define OP_SVE_VVV_Q_D \ { \ QLF3(S_Q,S_D,S_D), \ @@ -2467,6 +2471,20 @@ QLF3(V_4S, V_16B, S_4B),\ } +/* e.g. FDOT <Vd>.4H, <Vn>.8B, <Vm>.8B. */ +#define QL_V3DOTH \ +{ \ + QLF3(V_4H, V_8B, V_8B), \ + QLF3(V_8H, V_16B, V_16B),\ +} + +/* e.g. FDOT <Vd>.4H, <Vn>.8B, <Vm>.2B[<index>]. */ +#define QL_V2DOTH \ +{ \ + QLF3(V_4H, V_8B, S_2B),\ + QLF3(V_8H, V_16B, S_2B),\ +} + /* e.g. SHA512H <Qd>, <Qn>, <Vm>.2D. */ #define QL_SHA512UPT \ { \ @@ -2533,6 +2551,30 @@ QLF3(V_4S, V_4H, S_H),\ } +/* e.g. FMLALB <Vd>.8H, <Vn>.16B, <Vm>.16B. */ +#define QL_V3FML8H \ +{ \ + QLF3(V_8H, V_16B, V_16B),\ +} + +/* e.g. FMLALB <Vd>.8H, <Vn>.16B, <Vm>.B. */ +#define QL_V2FML8H \ +{ \ + QLF3(V_8H, V_16B, S_B),\ +} + +/* e.g. FMLALLBB <Vd>.4S, <Vn>.16B, <Vm>.16B. */ +#define QL_V3FMLL4S \ +{ \ + QLF3(V_4S, V_16B, V_16B),\ +} + +/* e.g. FMLALLBB <Vd>.4S, <Vn>.16B, <Vm>.B. */ +#define QL_V2FMLL4S \ +{ \ + QLF3(V_4S, V_16B, S_B),\ +} + /* e.g. RMIF <Xn>, #<shift>, #<mask>. */ #define QL_RMIF \ { \ @@ -2768,6 +2810,18 @@ static const aarch64_feature_set aarch64_feature_brbe = AARCH64_FEATURE (BRBE); static const aarch64_feature_set aarch64_feature_sme_lutv2 = AARCH64_FEATURES (3, SME_LUTv2, SME2, SME2p1); +static const aarch64_feature_set aarch64_feature_fp8fma = + AARCH64_FEATURE (FP8FMA); +static const aarch64_feature_set aarch64_feature_fp8dot4 = + AARCH64_FEATURE (FP8DOT4); +static const aarch64_feature_set aarch64_feature_fp8dot2 = + AARCH64_FEATURE (FP8DOT2); +static const aarch64_feature_set aarch64_feature_fp8fma_sve = + AARCH64_FEATURES (2, FP8FMA_SVE, SVE); +static const aarch64_feature_set aarch64_feature_fp8dot4_sve = + AARCH64_FEATURES (2, FP8DOT4_SVE, SVE); +static const aarch64_feature_set aarch64_feature_fp8dot2_sve = + AARCH64_FEATURES (2, FP8DOT2_SVE, SVE); #define CORE &aarch64_feature_v8 #define FP &aarch64_feature_fp @@ -2846,6 +2900,12 @@ static const aarch64_feature_set aarch64_feature_sme_lutv2 = #define LUT_SVE2 &aarch64_feature_lut_sve2 #define BRBE &aarch64_feature_brbe #define LUTv2_SME2 &aarch64_feature_sme_lutv2 +#define FP8FMA &aarch64_feature_fp8fma +#define FP8DOT4 &aarch64_feature_fp8dot4 +#define FP8DOT2 &aarch64_feature_fp8dot2 +#define FP8FMA_SVE &aarch64_feature_fp8fma_sve +#define FP8DOT4_SVE &aarch64_feature_fp8dot4_sve +#define FP8DOT2_SVE &aarch64_feature_fp8dot2_sve #define CORE_INSN(NAME,OPCODE,MASK,CLASS,OP,OPS,QUALS,FLAGS) \ { NAME, OPCODE, MASK, CLASS, OP, CORE, OPS, QUALS, FLAGS, 0, 0, NULL } @@ -3049,6 +3109,18 @@ static const aarch64_feature_set aarch64_feature_sme_lutv2 = #define LUTv2_SME2_INSN(NAME,OPCODE,MASK,CLASS,OPS,QUALS,FLAGS) \ { NAME, OPCODE, MASK, CLASS, 0, LUTv2_SME2, OPS, QUALS, \ FLAGS, 0, 0, NULL } +#define FP8FMA_INSN(NAME,OPCODE,MASK,CLASS,OPS,QUALS,FLAGS) \ + { NAME, OPCODE, MASK, CLASS, 0, FP8FMA, OPS, QUALS, FLAGS, 0, 0, NULL } +#define FP8DOT4_INSN(NAME,OPCODE,MASK,CLASS,OPS,QUALS,FLAGS) \ + { NAME, OPCODE, MASK, CLASS, 0, FP8DOT4, OPS, QUALS, FLAGS, 0, 0, NULL } +#define FP8DOT2_INSN(NAME,OPCODE,MASK,CLASS,OPS,QUALS,FLAGS) \ + { NAME, OPCODE, MASK, CLASS, 0, FP8DOT2, OPS, QUALS, FLAGS, 0, 0, NULL } +#define FP8FMA_SVE_INSNC(NAME,OPCODE,MASK,CLASS,OPS,QUALS,FLAGS,CONSTRAINTS,TIED) \ + { NAME, OPCODE, MASK, CLASS, 0, FP8FMA_SVE, OPS, QUALS, F_STRICT | FLAGS, CONSTRAINTS, TIED, NULL } +#define FP8DOT4_SVE_INSNC(NAME,OPCODE,MASK,CLASS,OPS,QUALS,FLAGS,CONSTRAINTS,TIED) \ + { NAME, OPCODE, MASK, CLASS, 0, FP8DOT4_SVE, OPS, QUALS, F_STRICT | FLAGS, CONSTRAINTS, TIED, NULL } +#define FP8DOT2_SVE_INSNC(NAME,OPCODE,MASK,CLASS,OPS,QUALS,FLAGS,CONSTRAINTS,TIED) \ + { NAME, OPCODE, MASK, CLASS, 0, FP8DOT2_SVE, OPS, QUALS, F_STRICT | FLAGS, CONSTRAINTS, TIED, NULL } #define MOPS_CPY_OP1_OP2_PME_INSN(NAME, OPCODE, MASK, FLAGS, CONSTRAINTS) \ MOPS_INSN (NAME, OPCODE, MASK, 0, \ @@ -6639,6 +6711,41 @@ const struct aarch64_opcode aarch64_opcode_table[] = LUTv2_SME2_INSN ("luti4", 0xc08b0000, 0xffffcc23, sme_size_12_b, OP3 (SME_Zdnx4, SME_ZT0, SME_Znx2_BIT_INDEX), OP_SVE_VUU_B, F_STRICT | 0), LUTv2_SME2_INSN ("luti4", 0xc09b0000, 0xffffcc2c, sme_size_12_b, OP3 (SME_Zdnx4_STRIDED, SME_ZT0, SME_Znx2_BIT_INDEX), OP_SVE_VUU_B, F_STRICT | 0), LUTv2_SME2_INSN ("movt", 0xc04f03e0, 0xffffcfe0, sme_misc, OP2 (SME_ZT0_INDEX2_12, SVE_Zt), {}, 0), + /* FP8 multiplication AdvSIMD instructions. */ + FP8DOT4_INSN("fdot", 0x0e00fc00, 0xbfe0fc00, dotproduct, OP3 (Vd, Vn, Vm), QL_V3DOT, F_SIZEQ), + FP8DOT4_INSN("fdot", 0x0f000000, 0xbfc0f400, dotproduct, OP3 (Vd, Vn, Em), QL_V2DOT, F_SIZEQ), + FP8DOT2_INSN("fdot", 0x0e40fc00, 0xbfe0fc00, dotproduct, OP3 (Vd, Vn, Vm), QL_V3DOTH, F_SIZEQ), + FP8DOT2_INSN("fdot", 0x0f400000, 0xbfc0f400, dotproduct, OP3 (Vd, Vn, Em16), QL_V2DOTH, F_SIZEQ), + FP8FMA_INSN("fmlalb", 0x0ec0fc00, 0xffe0fc00, asimdsame, OP3 (Vd, Vn, Vm), QL_V3FML8H, 0), + FP8FMA_INSN("fmlalt", 0x4ec0fc00, 0xffe0fc00, asimdsame, OP3 (Vd, Vn, Vm), QL_V3FML8H, 0), + FP8FMA_INSN("fmlalb", 0x0fc00000, 0xffc0f400, asimdsame, OP3 (Vd, Vn, Em8), QL_V2FML8H, 0), + FP8FMA_INSN("fmlalt", 0x4fc00000, 0xffc0f400, asimdsame, OP3 (Vd, Vn, Em8), QL_V2FML8H, 0), + FP8FMA_INSN("fmlallbb", 0x0e00c400, 0xffe0fc00, asimdsame, OP3 (Vd, Vn, Vm), QL_V3FMLL4S, 0), + FP8FMA_INSN("fmlallbt", 0x0e40c400, 0xffe0fc00, asimdsame, OP3 (Vd, Vn, Vm), QL_V3FMLL4S, 0), + FP8FMA_INSN("fmlalltb", 0x4e00c400, 0xffe0fc00, asimdsame, OP3 (Vd, Vn, Vm), QL_V3FMLL4S, 0), + FP8FMA_INSN("fmlalltt", 0x4e40c400, 0xffe0fc00, asimdsame, OP3 (Vd, Vn, Vm), QL_V3FMLL4S, 0), + FP8FMA_INSN("fmlallbb", 0x2f008000, 0xffc0f400, asimdsame, OP3 (Vd, Vn, Em8), QL_V2FMLL4S, 0), + FP8FMA_INSN("fmlallbt", 0x2f408000, 0xffc0f400, asimdsame, OP3 (Vd, Vn, Em8), QL_V2FMLL4S, 0), + FP8FMA_INSN("fmlalltb", 0x6f008000, 0xffc0f400, asimdsame, OP3 (Vd, Vn, Em8), QL_V2FMLL4S, 0), + FP8FMA_INSN("fmlalltt", 0x6f408000, 0xffc0f400, asimdsame, OP3 (Vd, Vn, Em8), QL_V2FMLL4S, 0), + + /* FP8 multiplication SVE instructions. */ + FP8DOT4_SVE_INSNC("fdot", 0x64608400, 0xffe0fc00, sve_misc, OP3 (SVE_Zd, SVE_Zn, SVE_Zm_16), OP_SVE_VVV_S_B, 0, C_SCAN_MOVPRFX, 0), + FP8DOT4_SVE_INSNC("fdot", 0x64604400, 0xffe0fc00, sve_misc, OP3 (SVE_Zd, SVE_Zn, SVE_Zm3_INDEX), OP_SVE_VVV_S_B, 0, C_SCAN_MOVPRFX, 0), + FP8DOT2_SVE_INSNC("fdot", 0x64208400, 0xffe0fc00, sve_misc, OP3 (SVE_Zd, SVE_Zn, SVE_Zm_16), OP_SVE_VVV_H_B, 0, C_SCAN_MOVPRFX, 0), + FP8DOT2_SVE_INSNC("fdot", 0x64204400, 0xffe0f400, sve_misc, OP3 (SVE_Zd, SVE_Zn, SVE_Zm3_11_INDEX), OP_SVE_VVV_H_B, 0, C_SCAN_MOVPRFX, 0), + FP8FMA_SVE_INSNC("fmlalb", 0x64a08800, 0xffe0fc00, sve_misc, OP3 (SVE_Zd, SVE_Zn, SVE_Zm_16), OP_SVE_VVV_H_B, 0, C_SCAN_MOVPRFX, 0), + FP8FMA_SVE_INSNC("fmlalb", 0x64205000, 0xffe0f000, sve_misc, OP3 (SVE_Zd, SVE_Zn, SVE_Zm3_10_INDEX), OP_SVE_VVV_H_B, 0, C_SCAN_MOVPRFX, 0), + FP8FMA_SVE_INSNC("fmlallbb", 0x64208800, 0xffe0fc00, sve_misc, OP3 (SVE_Zd, SVE_Zn, SVE_Zm_16), OP_SVE_VVV_S_B, 0, C_SCAN_MOVPRFX, 0), + FP8FMA_SVE_INSNC("fmlallbb", 0x6420c000, 0xffe0f000, sve_misc, OP3 (SVE_Zd, SVE_Zn, SVE_Zm3_10_INDEX), OP_SVE_VVV_S_B, 0, C_SCAN_MOVPRFX, 0), + FP8FMA_SVE_INSNC("fmlallbt", 0x64209800, 0xffe0fc00, sve_misc, OP3 (SVE_Zd, SVE_Zn, SVE_Zm_16), OP_SVE_VVV_S_B, 0, C_SCAN_MOVPRFX, 0), + FP8FMA_SVE_INSNC("fmlallbt", 0x6460c000, 0xffe0f000, sve_misc, OP3 (SVE_Zd, SVE_Zn, SVE_Zm3_10_INDEX), OP_SVE_VVV_S_B, 0, C_SCAN_MOVPRFX, 0), + FP8FMA_SVE_INSNC("fmlalltb", 0x6420a800, 0xffe0fc00, sve_misc, OP3 (SVE_Zd, SVE_Zn, SVE_Zm_16), OP_SVE_VVV_S_B, 0, C_SCAN_MOVPRFX, 0), + FP8FMA_SVE_INSNC("fmlalltb", 0x64a0c000, 0xffe0f000, sve_misc, OP3 (SVE_Zd, SVE_Zn, SVE_Zm3_10_INDEX), OP_SVE_VVV_S_B, 0, C_SCAN_MOVPRFX, 0), + FP8FMA_SVE_INSNC("fmlalltt", 0x6420b800, 0xffe0fc00, sve_misc, OP3 (SVE_Zd, SVE_Zn, SVE_Zm_16), OP_SVE_VVV_S_B, 0, C_SCAN_MOVPRFX, 0), + FP8FMA_SVE_INSNC("fmlalltt", 0x64e0c000, 0xffe0f000, sve_misc, OP3 (SVE_Zd, SVE_Zn, SVE_Zm3_10_INDEX), OP_SVE_VVV_S_B, 0, C_SCAN_MOVPRFX, 0), + FP8FMA_SVE_INSNC("fmlalt", 0x64a09800, 0xffe0fc00, sve_misc, OP3 (SVE_Zd, SVE_Zn, SVE_Zm_16), OP_SVE_VVV_H_B, 0, C_SCAN_MOVPRFX, 0), + FP8FMA_SVE_INSNC("fmlalt", 0x64a05000, 0xffe0f000, sve_misc, OP3 (SVE_Zd, SVE_Zn, SVE_Zm3_10_INDEX), OP_SVE_VVV_H_B, 0, C_SCAN_MOVPRFX, 0), {0, 0, 0, 0, 0, 0, {}, {}, 0, 0, 0, NULL}, }; @@ -6715,6 +6822,8 @@ const struct aarch64_opcode aarch64_opcode_table[] = "a SIMD vector element") \ Y(SIMD_ELEMENT, reglane, "Em16", 0, F(FLD_Rm), \ "a SIMD vector element limited to V0-V15") \ + Y(SIMD_ELEMENT, reglane, "Em8", 0, F(FLD_Rm), \ + "a SIMD vector element limited to V0-V7") \ Y(SIMD_ELEMENT, simple_index, "Em_INDEX1_14", 0, F(FLD_Rm, FLD_imm1_14), \ "a SIMD vector without a type qualifier encoding a bit index") \ Y(SIMD_ELEMENT, simple_index, "Em_INDEX2_13", 0, F(FLD_Rm, FLD_imm2_13), \ @@ -7138,6 +7247,9 @@ const struct aarch64_opcode aarch64_opcode_table[] = Y(SVE_REG, sve_quad_index, "SVE_Zm3_22_INDEX", \ 3 << OPD_F_OD_LSB, F(FLD_SVE_i3h, FLD_SVE_Zm_16), \ "an indexed SVE vector register") \ + Y(SVE_REG, sve_quad_index, "SVE_Zm3_10_INDEX", \ + 3 << OPD_F_OD_LSB, F(FLD_SVE_i3h2, FLD_SVE_i4l2, FLD_SVE_imm3), \ + "an indexed SVE vector register") \ Y(SVE_REG, sve_quad_index, "SVE_Zm4_11_INDEX", \ 4 << OPD_F_OD_LSB, F(FLD_SVE_i2h, FLD_SVE_i3l, FLD_SVE_imm4), \ "an indexed SVE vector register") \ |