aarch64: Add support for FEAT_SME2p1 instructions.

Hi,

This patch add support for FEAT_SME2p1 and "movaz" instructions
along with the optional flag +sme2p1.

Following "movaz" instructions are add:
Move and zero two ZA tile slices to vector registers.
Move and zero four ZA tile slices to vector registers.

Regression testing for aarch64-none-elf target and found no regressions.

Ok for binutils-master?

Regards,
Srinath.
This commit is contained in:
Srinath Parvathaneni 2024-01-15 09:34:41 +00:00 committed by Nick Clifton
parent 7e8d2d8757
commit 89e06ec152
13 changed files with 389 additions and 0 deletions

View File

@ -3,6 +3,9 @@
hand-written asm using the new command line option --scfi=experimental on
x86-64. Only System V AMD64 ABI is supported.
* Add support for the AArch64 Scalable Matrix Extension version 2.1 (SME2.1)
instructions.
* Add support for 'armv8.9-a' and 'armv9.4-a' for -march in Arm GAS.
* Initial support for Intel APX: 32 GPRs, NDD, PUSH2/POP2 and PUSHP/POPP.

View File

@ -4492,6 +4492,7 @@ parse_sme_immediate (char **str, int64_t *imm)
[<Wv>, <imm>]
[<Wv>, #<imm>]
[<Ws>, <offsf>:<offsl>]
Return true on success, populating OPND with the parsed index. */
@ -4592,6 +4593,7 @@ parse_sme_za_index (char **str, struct aarch64_indexed_za *opnd)
<Pm>.<T>[<Wv>< #<imm>]
ZA[<Wv>, #<imm>]
<ZAn><HV>.<T>[<Wv>, #<imm>]
<ZAn><HV>.<T>[<Ws>, <offsf>:<offsl>]
FLAGS is as for parse_typed_reg. */
@ -7865,6 +7867,21 @@ parse_operands (char *str, const aarch64_opcode *opcode)
info->qualifier = qualifier;
break;
case AARCH64_OPND_SME_ZA_array_vrsb_1:
case AARCH64_OPND_SME_ZA_array_vrsh_1:
case AARCH64_OPND_SME_ZA_array_vrss_1:
case AARCH64_OPND_SME_ZA_array_vrsd_1:
case AARCH64_OPND_SME_ZA_array_vrsb_2:
case AARCH64_OPND_SME_ZA_array_vrsh_2:
case AARCH64_OPND_SME_ZA_array_vrss_2:
case AARCH64_OPND_SME_ZA_array_vrsd_2:
if (!parse_dual_indexed_reg (&str, REG_TYPE_ZATHV,
&info->indexed_za, &qualifier, 0))
goto failure;
info->qualifier = qualifier;
break;
case AARCH64_OPND_SME_VLxN_10:
case AARCH64_OPND_SME_VLxN_13:
po_strict_enum_or_fail (aarch64_sme_vlxn_array);
@ -10336,6 +10353,7 @@ static const struct aarch64_option_cpu_value_table aarch64_features[] = {
{"d128", AARCH64_FEATURE (D128),
AARCH64_FEATURE (LSE128)},
{"b16b16", AARCH64_FEATURE (B16B16), AARCH64_FEATURE (SVE2)},
{"sme2p1", AARCH64_FEATURE (SME2p1), AARCH64_FEATURE (SME2)},
{NULL, AARCH64_NO_FEATURES, AARCH64_NO_FEATURES},
};

View File

@ -276,6 +276,8 @@ automatically cause those extensions to be disabled.
@tab Enable TRCIT instruction.
@item @code{d128} @tab Armv9.4-A @tab No
@tab Enable the 128-bit Page Descriptor Extension. This implies @code{lse128}.
@item @code{sme2p1} @tab N/A @tab No
@tab Enable the SME2.1 Extension.
@end multitable
@node AArch64 Syntax

View File

@ -0,0 +1,42 @@
#name: Test of SME2.1 movaz instructions.
#as: -march=armv9.4-a+sme2p1
#objdump: -dr
[^:]+: file format .*
[^:]+:
[^:]+:
.*: c006c260 movaz {z0.b-z1.b}, za0v.b \[w14, 6:7\]
.*: c046c260 movaz {z0.h-z1.h}, za0v.h \[w14, 6:7\]
.*: c086c220 movaz {z0.s-z1.s}, za0v.s \[w14, 2:3\]
.*: c0c6c200 movaz {z0.d-z1.d}, za0v.d \[w14, 0:1\]
.*: c00602e0 movaz {z0.b-z1.b}, za0h.b \[w12, 14:15\]
.*: c0462260 movaz {z0.h-z1.h}, za0h.h \[w13, 6:7\]
.*: c0864220 movaz {z0.s-z1.s}, za0h.s \[w14, 2:3\]
.*: c0c66200 movaz {z0.d-z1.d}, za0h.d \[w15, 0:1\]
.*: c006c260 movaz {z0.b-z1.b}, za0v.b \[w14, 6:7\]
.*: c046c2e0 movaz {z0.h-z1.h}, za1v.h \[w14, 6:7\]
.*: c086c2a0 movaz {z0.s-z1.s}, za2v.s \[w14, 2:3\]
.*: c0c6c260 movaz {z0.d-z1.d}, za3v.d \[w14, 0:1\]
.*: c00602e0 movaz {z0.b-z1.b}, za0h.b \[w12, 14:15\]
.*: c04622e0 movaz {z0.h-z1.h}, za1h.h \[w13, 6:7\]
.*: c08642a0 movaz {z0.s-z1.s}, za2h.s \[w14, 2:3\]
.*: c0c66260 movaz {z0.d-z1.d}, za3h.d \[w15, 0:1\]
.*: c006c660 movaz {z0.b-z3.b}, za0v.b \[w14, 12:15\]
.*: c046c620 movaz {z0.h-z3.h}, za0v.h \[w14, 4:7\]
.*: c086c600 movaz {z0.s-z3.s}, za0v.s \[w14, 0:3\]
.*: c0c6c600 movaz {z0.d-z3.d}, za0v.d \[w14, 0:3\]
.*: c0060660 movaz {z0.b-z3.b}, za0h.b \[w12, 12:15\]
.*: c0462620 movaz {z0.h-z3.h}, za0h.h \[w13, 4:7\]
.*: c0864600 movaz {z0.s-z3.s}, za0h.s \[w14, 0:3\]
.*: c0c66600 movaz {z0.d-z3.d}, za0h.d \[w15, 0:3\]
.*: c006c640 movaz {z0.b-z3.b}, za0v.b \[w14, 8:11\]
.*: c046c660 movaz {z0.h-z3.h}, za1v.h \[w14, 4:7\]
.*: c086c640 movaz {z0.s-z3.s}, za2v.s \[w14, 0:3\]
.*: c0c6c660 movaz {z0.d-z3.d}, za3v.d \[w14, 0:3\]
.*: c0060660 movaz {z0.b-z3.b}, za0h.b \[w12, 12:15\]
.*: c0462660 movaz {z0.h-z3.h}, za1h.h \[w13, 4:7\]
.*: c0864640 movaz {z0.s-z3.s}, za2h.s \[w14, 0:3\]
.*: c0c66660 movaz {z0.d-z3.d}, za3h.d \[w15, 0:3\]

View File

@ -0,0 +1,39 @@
movaz {z0.b - z1.b}, ZA0V.B [w14, 6:7]
movaz {z0.h - z1.h}, ZA0V.H [w14, 6:7]
movaz {z0.s - z1.s}, ZA0V.S [w14, 2:3]
movaz {z0.d - z1.d}, ZA0V.D [w14, 0:1]
movaz {z0.b - z1.b}, ZA0H.B [w12, 14:15]
movaz {z0.h - z1.h}, ZA0H.H [w13, 6:7]
movaz {z0.s - z1.s}, ZA0H.S [w14, 2:3]
movaz {z0.d - z1.d}, ZA0H.D [w15, 0:1]
movaz {z0.b - z1.b}, ZA0V.B [w14, 6:7]
movaz {z0.h - z1.h}, ZA1V.H [w14, 6:7]
movaz {z0.s - z1.s}, ZA2V.S [w14, 2:3]
movaz {z0.d - z1.d}, ZA3V.D [w14, 0:1]
movaz {z0.b - z1.b}, ZA0H.B [w12, 14:15]
movaz {z0.h - z1.h}, ZA1H.H [w13, 6:7]
movaz {z0.s - z1.s}, ZA2H.S [w14, 2:3]
movaz {z0.d - z1.d}, ZA3H.D [w15, 0:1]
movaz {z0.b - z3.b}, ZA0V.B [w14, 12:15]
movaz {z0.h - z3.h}, ZA0V.H [w14, 4:7]
movaz {z0.s - z3.s}, ZA0V.S [w14, 0:3]
movaz {z0.d - z3.d}, ZA0V.D [w14, 0:3]
movaz {z0.b - z3.b}, ZA0H.B [w12, 12:15]
movaz {z0.h - z3.h}, ZA0H.H [w13, 4:7]
movaz {z0.s - z3.s}, ZA0H.S [w14, 0:3]
movaz {z0.d - z3.d}, ZA0H.D [w15, 0:3]
movaz {z0.b - z3.b}, ZA0V.B [w14, 8:11]
movaz {z0.h - z3.h}, ZA1V.H [w14, 4:7]
movaz {z0.s - z3.s}, ZA2V.S [w14, 0:3]
movaz {z0.d - z3.d}, ZA3V.D [w14, 0:3]
movaz {z0.b - z3.b}, ZA0H.B [w12, 12:15]
movaz {z0.h - z3.h}, ZA1H.H [w13, 4:7]
movaz {z0.s - z3.s}, ZA2H.S [w14, 0:3]
movaz {z0.d - z3.d}, ZA3H.D [w15, 0:3]

View File

@ -224,6 +224,8 @@ enum aarch64_feature_bit {
AARCH64_FEATURE_SEBEP,
/* SVE2.1 and SME2.1 non-widening BFloat16 instructions. */
AARCH64_FEATURE_B16B16,
/* SME2.1 instructions. */
AARCH64_FEATURE_SME2p1,
AARCH64_NUM_FEATURES
};
@ -705,6 +707,14 @@ enum aarch64_opnd
AARCH64_OPND_SVE_Vd, /* Scalar SIMD&FP register in Vd. */
AARCH64_OPND_SVE_Vm, /* Scalar SIMD&FP register in Vm. */
AARCH64_OPND_SVE_Vn, /* Scalar SIMD&FP register in Vn. */
AARCH64_OPND_SME_ZA_array_vrsb_1, /* Tile to vector, two registers (B). */
AARCH64_OPND_SME_ZA_array_vrsh_1, /* Tile to vector, two registers (H). */
AARCH64_OPND_SME_ZA_array_vrss_1, /* Tile to vector, two registers (S). */
AARCH64_OPND_SME_ZA_array_vrsd_1, /* Tile to vector, two registers (D). */
AARCH64_OPND_SME_ZA_array_vrsb_2, /* Tile to vector, four registers (B). */
AARCH64_OPND_SME_ZA_array_vrsh_2, /* Tile to vector, four registers (H). */
AARCH64_OPND_SME_ZA_array_vrss_2, /* Tile to vector, four registers (S). */
AARCH64_OPND_SME_ZA_array_vrsd_2, /* Tile to vector, four registers (D). */
AARCH64_OPND_SVE_Za_5, /* SVE vector register in Za, bits [9,5]. */
AARCH64_OPND_SVE_Za_16, /* SVE vector register in Za, bits [20,16]. */
AARCH64_OPND_SVE_Zd, /* SVE vector register in Zd. */
@ -962,6 +972,7 @@ enum aarch64_insn_class
sme_start,
sme_stop,
sme2_mov,
sme2_movaz,
sve_cpy,
sve_index,
sve_limm,

View File

@ -1375,6 +1375,76 @@ aarch64_ins_sve_float_zero_one (const aarch64_operand *self,
return true;
}
bool
aarch64_ins_sme_za_vrs1 (const aarch64_operand *self,
const aarch64_opnd_info *info,
aarch64_insn *code,
const aarch64_inst *inst ATTRIBUTE_UNUSED,
aarch64_operand_error *errors ATTRIBUTE_UNUSED)
{
int za_reg = info->indexed_za.regno;
int regno = info->indexed_za.index.regno & 3;
int imm = info->indexed_za.index.imm;
int v = info->indexed_za.v;
int countm1 = info->indexed_za.index.countm1;
insert_field (self->fields[0], code, v, 0);
insert_field (self->fields[1], code, regno, 0);
switch (info->qualifier)
{
case AARCH64_OPND_QLF_S_B:
insert_field (self->fields[2], code, imm / (countm1 + 1), 0);
break;
case AARCH64_OPND_QLF_S_H:
case AARCH64_OPND_QLF_S_S:
insert_field (self->fields[2], code, za_reg, 0);
insert_field (self->fields[3], code, imm / (countm1 + 1), 0);
break;
case AARCH64_OPND_QLF_S_D:
insert_field (self->fields[2], code, za_reg, 0);
break;
default:
return false;
}
return true;
}
bool
aarch64_ins_sme_za_vrs2 (const aarch64_operand *self,
const aarch64_opnd_info *info,
aarch64_insn *code,
const aarch64_inst *inst ATTRIBUTE_UNUSED,
aarch64_operand_error *errors ATTRIBUTE_UNUSED)
{
int za_reg = info->indexed_za.regno;
int regno = info->indexed_za.index.regno & 3;
int imm = info->indexed_za.index.imm;
int v = info->indexed_za.v;
int countm1 = info->indexed_za.index.countm1;
insert_field (self->fields[0], code, v, 0);
insert_field (self->fields[1], code, regno, 0);
switch (info->qualifier)
{
case AARCH64_OPND_QLF_S_B:
insert_field (self->fields[2], code, imm / (countm1 + 1), 0);
break;
case AARCH64_OPND_QLF_S_H:
insert_field (self->fields[2], code, za_reg, 0);
insert_field (self->fields[3], code, imm / (countm1 + 1), 0);
break;
case AARCH64_OPND_QLF_S_S:
case AARCH64_OPND_QLF_S_D:
insert_field (self->fields[2], code, za_reg, 0);
break;
default:
return false;
}
return true;
}
/* Encode in SME instruction such as MOVA ZA tile vector register number,
vector indicator, vector selector and immediate. */
bool
@ -2011,6 +2081,7 @@ aarch64_encode_variant_using_iclass (struct aarch64_inst *inst)
break;
case sme_misc:
case sme2_movaz:
case sve_misc:
/* These instructions have only a single variant. */
break;

View File

@ -100,6 +100,8 @@ AARCH64_DECL_OPD_INSERTER (ins_sve_strided_reglist);
AARCH64_DECL_OPD_INSERTER (ins_sve_scale);
AARCH64_DECL_OPD_INSERTER (ins_sve_shlimm);
AARCH64_DECL_OPD_INSERTER (ins_sve_shrimm);
AARCH64_DECL_OPD_INSERTER (ins_sme_za_vrs1);
AARCH64_DECL_OPD_INSERTER (ins_sme_za_vrs2);
AARCH64_DECL_OPD_INSERTER (ins_sme_za_hv_tiles);
AARCH64_DECL_OPD_INSERTER (ins_sme_za_hv_tiles_range);
AARCH64_DECL_OPD_INSERTER (ins_sme_za_list);

View File

@ -1929,6 +1929,84 @@ aarch64_ext_sme_za_array (const aarch64_operand *self,
return true;
}
/* Decode two ZA tile slice (V, Rv, off3| ZAn ,off2 | ZAn, ol| ZAn) feilds. */
bool
aarch64_ext_sme_za_vrs1 (const aarch64_operand *self,
aarch64_opnd_info *info, aarch64_insn code,
const aarch64_inst *inst,
aarch64_operand_error *errors ATTRIBUTE_UNUSED)
{
int v = extract_field (self->fields[0], code, 0);
int regno = 12 + extract_field (self->fields[1], code, 0);
int imm, za_reg, num_offset = 2;
switch (info->qualifier)
{
case AARCH64_OPND_QLF_S_B:
imm = extract_field (self->fields[2], code, 0);
info->indexed_za.index.imm = imm * num_offset;
break;
case AARCH64_OPND_QLF_S_H:
case AARCH64_OPND_QLF_S_S:
za_reg = extract_field (self->fields[2], code, 0);
imm = extract_field (self->fields[3], code, 0);
info->indexed_za.index.imm = imm * num_offset;
info->indexed_za.regno = za_reg;
break;
case AARCH64_OPND_QLF_S_D:
za_reg = extract_field (self->fields[2], code, 0);
info->indexed_za.regno = za_reg;
break;
default:
return false;
}
info->indexed_za.index.regno = regno;
info->indexed_za.index.countm1 = num_offset - 1;
info->indexed_za.v = v;
info->indexed_za.group_size = get_opcode_dependent_value (inst->opcode);
return true;
}
/* Decode four ZA tile slice (V, Rv, off3| ZAn ,off2 | ZAn, ol| ZAn) feilds. */
bool
aarch64_ext_sme_za_vrs2 (const aarch64_operand *self,
aarch64_opnd_info *info, aarch64_insn code,
const aarch64_inst *inst,
aarch64_operand_error *errors ATTRIBUTE_UNUSED)
{
int v = extract_field (self->fields[0], code, 0);
int regno = 12 + extract_field (self->fields[1], code, 0);
int imm, za_reg, num_offset =4;
switch (info->qualifier)
{
case AARCH64_OPND_QLF_S_B:
imm = extract_field (self->fields[2], code, 0);
info->indexed_za.index.imm = imm * num_offset;
break;
case AARCH64_OPND_QLF_S_H:
za_reg = extract_field (self->fields[2], code, 0);
imm = extract_field (self->fields[3], code, 0);
info->indexed_za.index.imm = imm * num_offset;
info->indexed_za.regno = za_reg;
break;
case AARCH64_OPND_QLF_S_S:
case AARCH64_OPND_QLF_S_D:
za_reg = extract_field (self->fields[2], code, 0);
info->indexed_za.regno = za_reg;
break;
default:
return false;
}
info->indexed_za.index.regno = regno;
info->indexed_za.index.countm1 = num_offset - 1;
info->indexed_za.v = v;
info->indexed_za.group_size = get_opcode_dependent_value (inst->opcode);
return true;
}
bool
aarch64_ext_sme_addr_ri_u4xvl (const aarch64_operand *self,
aarch64_opnd_info *info, aarch64_insn code,
@ -3160,6 +3238,7 @@ aarch64_decode_variant_using_iclass (aarch64_inst *inst)
variant = 3;
break;
case sme2_movaz:
case sme_misc:
case sve_misc:
/* These instructions have only a single variant. */

View File

@ -124,6 +124,8 @@ AARCH64_DECL_OPD_EXTRACTOR (ext_sve_strided_reglist);
AARCH64_DECL_OPD_EXTRACTOR (ext_sve_scale);
AARCH64_DECL_OPD_EXTRACTOR (ext_sve_shlimm);
AARCH64_DECL_OPD_EXTRACTOR (ext_sve_shrimm);
AARCH64_DECL_OPD_EXTRACTOR (ext_sme_za_vrs1);
AARCH64_DECL_OPD_EXTRACTOR (ext_sme_za_vrs2);
AARCH64_DECL_OPD_EXTRACTOR (ext_sme_za_hv_tiles);
AARCH64_DECL_OPD_EXTRACTOR (ext_sme_za_hv_tiles_range);
AARCH64_DECL_OPD_EXTRACTOR (ext_sme_za_list);

View File

@ -400,6 +400,16 @@ const aarch64_field fields[] =
{ 22, 1 }, /* sz: 1-bit element size select. */
{ 22, 2 }, /* type: floating point type field in fp data inst. */
{ 10, 2 }, /* vldst_size: size field in the AdvSIMD load/store inst. */
{ 5, 3 }, /* off3: immediate offset used to calculate slice number in a
ZA tile. */
{ 5, 2 }, /* off2: immediate offset used to calculate slice number in
a ZA tile. */
{ 7, 1 }, /* ZAn_1: name of the 1bit encoded ZA tile. */
{ 5, 1 }, /* ol: immediate offset used to calculate slice number in a ZA
tile. */
{ 6, 2 }, /* ZAn_2: name of the 2bit encoded ZA tile. */
{ 5, 3 }, /* ZAn_3: name of the 3bit encoded ZA tile. */
{ 6, 1 }, /* ZAn: name of the bit encoded ZA tile. */
};
enum aarch64_operand_class
@ -1938,6 +1948,49 @@ operand_general_constraint_met_p (const aarch64_opnd_info *opnds, int idx,
return 0;
break;
case AARCH64_OPND_SME_ZA_array_vrsb_1:
if (!check_za_access (opnd, mismatch_detail, idx, 12, 7, 2,
get_opcode_dependent_value (opcode)))
return 0;
break;
case AARCH64_OPND_SME_ZA_array_vrsh_1:
if (!check_za_access (opnd, mismatch_detail, idx, 12, 3, 2,
get_opcode_dependent_value (opcode)))
return 0;
break;
case AARCH64_OPND_SME_ZA_array_vrss_1:
if (!check_za_access (opnd, mismatch_detail, idx, 12, 1, 2,
get_opcode_dependent_value (opcode)))
return 0;
break;
case AARCH64_OPND_SME_ZA_array_vrsd_1:
if (!check_za_access (opnd, mismatch_detail, idx, 12, 0, 2,
get_opcode_dependent_value (opcode)))
return 0;
break;
case AARCH64_OPND_SME_ZA_array_vrsb_2:
if (!check_za_access (opnd, mismatch_detail, idx, 12, 3, 4,
get_opcode_dependent_value (opcode)))
return 0;
break;
case AARCH64_OPND_SME_ZA_array_vrsh_2:
if (!check_za_access (opnd, mismatch_detail, idx, 12, 1, 4,
get_opcode_dependent_value (opcode)))
return 0;
break;
case AARCH64_OPND_SME_ZA_array_vrss_2:
case AARCH64_OPND_SME_ZA_array_vrsd_2:
if (!check_za_access (opnd, mismatch_detail, idx, 12, 0, 4,
get_opcode_dependent_value (opcode)))
return 0;
break;
case AARCH64_OPND_SME_ZA_HV_idx_srcxN:
case AARCH64_OPND_SME_ZA_HV_idx_destxN:
size = aarch64_get_qualifier_esize (opnd->qualifier);
@ -4103,6 +4156,30 @@ aarch64_print_operand (char *buf, size_t size, bfd_vma pc,
? style_sub_mnem (styler, "vgx4") : "");
break;
case AARCH64_OPND_SME_ZA_array_vrsb_1:
case AARCH64_OPND_SME_ZA_array_vrsh_1:
case AARCH64_OPND_SME_ZA_array_vrss_1:
case AARCH64_OPND_SME_ZA_array_vrsd_1:
case AARCH64_OPND_SME_ZA_array_vrsb_2:
case AARCH64_OPND_SME_ZA_array_vrsh_2:
case AARCH64_OPND_SME_ZA_array_vrss_2:
case AARCH64_OPND_SME_ZA_array_vrsd_2:
snprintf (buf, size, "%s [%s, %s%s%s]",
style_reg (styler, "za%d%c%s%s",
opnd->indexed_za.regno,
opnd->indexed_za.v ? 'v': 'h',
opnd->qualifier == AARCH64_OPND_QLF_NIL ? "" : ".",
(opnd->qualifier == AARCH64_OPND_QLF_NIL
? ""
: aarch64_get_qualifier_name (opnd->qualifier))),
style_reg (styler, "w%d", opnd->indexed_za.index.regno),
style_imm (styler, "%" PRIi64, opnd->indexed_za.index.imm),
opnd->indexed_za.index.countm1 ? ":" : "",
opnd->indexed_za.index.countm1 ? style_imm (styler, "%d",
opnd->indexed_za.index.imm
+ opnd->indexed_za.index.countm1):"");
break;
case AARCH64_OPND_SME_SM_ZA:
snprintf (buf, size, "%s",
style_reg (styler, opnd->reg.regno == 's' ? "sm" : "za"));

View File

@ -210,6 +210,13 @@ enum aarch64_field_kind
FLD_sz,
FLD_type,
FLD_vldst_size,
FLD_off3,
FLD_off2,
FLD_ZAn_1,
FLD_ol,
FLD_ZAn_2,
FLD_ZAn_3,
FLD_ZAn
};
/* Field description. */

View File

@ -1497,6 +1497,10 @@
{ \
QLF2(S_B,S_B), \
}
#define OP_SVE_HH \
{ \
QLF2(S_H,S_H), \
}
#define OP_SVE_BBU \
{ \
QLF3(S_B,S_B,NIL), \
@ -2614,6 +2618,8 @@ static const aarch64_feature_set aarch64_feature_d128_the =
AARCH64_FEATURES (2, D128, THE);
static const aarch64_feature_set aarch64_feature_b16b16 =
AARCH64_FEATURE (B16B16);
static const aarch64_feature_set aarch64_feature_sme2p1 =
AARCH64_FEATURE (SME2p1);
#define CORE &aarch64_feature_v8
#define FP &aarch64_feature_fp
@ -2677,6 +2683,7 @@ static const aarch64_feature_set aarch64_feature_b16b16 =
#define THE &aarch64_feature_the
#define D128_THE &aarch64_feature_d128_the
#define B16B16 &aarch64_feature_b16b16
#define SME2p1 &aarch64_feature_sme2p1
#define CORE_INSN(NAME,OPCODE,MASK,CLASS,OP,OPS,QUALS,FLAGS) \
{ NAME, OPCODE, MASK, CLASS, OP, CORE, OPS, QUALS, FLAGS, 0, 0, NULL }
@ -2743,6 +2750,9 @@ static const aarch64_feature_set aarch64_feature_b16b16 =
#define SVE2_INSN(NAME,OPCODE,MASK,CLASS,OP,OPS,QUALS,FLAGS,TIED) \
{ NAME, OPCODE, MASK, CLASS, OP, SVE2, OPS, QUALS, \
FLAGS | F_STRICT, 0, TIED, NULL }
#define SME2p1_INSN(NAME,OPCODE,MASK,CLASS,OP,OPS,QUALS,FLAGS,TIED) \
{ NAME, OPCODE, MASK, CLASS, OP, SME2p1, OPS, QUALS, \
FLAGS | F_STRICT, 0, TIED, NULL }
#define SVE2_INSNC(NAME,OPCODE,MASK,CLASS,OP,OPS,QUALS,FLAGS,CONSTRAINTS,TIED) \
{ NAME, OPCODE, MASK, CLASS, OP, SVE2, OPS, QUALS, \
FLAGS | F_STRICT, CONSTRAINTS, TIED, NULL }
@ -6289,6 +6299,16 @@ const struct aarch64_opcode aarch64_opcode_table[] =
B16B16_INSN("bfmls", 0x64200c00, 0xffa0fc00, sve_misc, 0, OP3 (SVE_Zd, SVE_Zn, SVE_Zm3_11_INDEX), OP_SVE_VVV_H, 0, 0),
B16B16_INSN("bfmul", 0x64202800, 0xffa0fc00, sve_misc, 0, OP3 (SVE_Zd, SVE_Zn, SVE_Zm3_11_INDEX), OP_SVE_VVV_H, 0, 0),
/* SME2.1 movaz instructions. */
SME2p1_INSN ("movaz", 0xc0060600, 0xffff1f83, sme2_movaz, 0, OP2 (SME_Zdnx4, SME_ZA_array_vrsb_2), OP_SVE_BB, 0, 0),
SME2p1_INSN ("movaz", 0xc0460600, 0xffff1f83, sme2_movaz, 0, OP2 (SME_Zdnx4, SME_ZA_array_vrsh_2), OP_SVE_HH, 0, 0),
SME2p1_INSN ("movaz", 0xc0860600, 0xffff1f83, sme2_movaz, 0, OP2 (SME_Zdnx4, SME_ZA_array_vrss_2), OP_SVE_SS, 0, 0),
SME2p1_INSN ("movaz", 0xc0c60600, 0xffff1f03, sme2_movaz, 0, OP2 (SME_Zdnx4, SME_ZA_array_vrsd_2), OP_SVE_DD, 0, 0),
SME2p1_INSN ("movaz", 0xc0060200, 0xffff1f01, sme2_movaz, 0, OP2 (SME_Zdnx2, SME_ZA_array_vrsb_1), OP_SVE_BB, 0, 0),
SME2p1_INSN ("movaz", 0xc0460200, 0xffff1f01, sme2_movaz, 0, OP2 (SME_Zdnx2, SME_ZA_array_vrsh_1), OP_SVE_HH, 0, 0),
SME2p1_INSN ("movaz", 0xc0860200, 0xffff1f01, sme2_movaz, 0, OP2 (SME_Zdnx2, SME_ZA_array_vrss_1), OP_SVE_SS, 0, 0),
SME2p1_INSN ("movaz", 0xc0c60200, 0xffff1f01, sme2_movaz, 0, OP2 (SME_Zdnx2, SME_ZA_array_vrsd_1), OP_SVE_DD, 0, 0),
{0, 0, 0, 0, 0, 0, {}, {}, 0, 0, 0, NULL},
};
@ -6726,6 +6746,22 @@ const struct aarch64_opcode aarch64_opcode_table[] =
Y(SIMD_REG, regno, "SVE_Vd", 0, F(FLD_SVE_Vd), "a SIMD register") \
Y(SIMD_REG, regno, "SVE_Vm", 0, F(FLD_SVE_Vm), "a SIMD register") \
Y(SIMD_REG, regno, "SVE_Vn", 0, F(FLD_SVE_Vn), "a SIMD register") \
Y(ZA_ACCESS, sme_za_vrs1, "SME_ZA_array_vrsb_1", 0, \
F(FLD_SME_V,FLD_SME_Rv,FLD_off3), "ZA0 tile") \
Y(ZA_ACCESS, sme_za_vrs1, "SME_ZA_array_vrsh_1", 0, \
F(FLD_SME_V,FLD_SME_Rv,FLD_ZAn_1,FLD_off2), "1 bit ZA tile") \
Y(ZA_ACCESS, sme_za_vrs1, "SME_ZA_array_vrss_1", 0, \
F(FLD_SME_V,FLD_SME_Rv,FLD_ZAn_2,FLD_ol), "2 ZA tile") \
Y(ZA_ACCESS, sme_za_vrs1, "SME_ZA_array_vrsd_1", 0, \
F(FLD_SME_V,FLD_SME_Rv,FLD_ZAn_3), "3 ZA tile") \
Y(ZA_ACCESS, sme_za_vrs2, "SME_ZA_array_vrsb_2", 0, \
F(FLD_SME_V,FLD_SME_Rv,FLD_off2), "ZA0 tile") \
Y(ZA_ACCESS, sme_za_vrs2, "SME_ZA_array_vrsh_2", 0, \
F(FLD_SME_V,FLD_SME_Rv,FLD_ZAn,FLD_ol), "1 bit ZA tile") \
Y(ZA_ACCESS, sme_za_vrs2, "SME_ZA_array_vrss_2", 0, \
F(FLD_SME_V,FLD_SME_Rv,FLD_off2), "2 bit ZA tile") \
Y(ZA_ACCESS, sme_za_vrs2, "SME_ZA_array_vrsd_2", 0, \
F(FLD_SME_V,FLD_SME_Rv,FLD_ZAn_3), "3 bit ZA tile") \
Y(SVE_REG, regno, "SVE_Za_5", 0, F(FLD_SVE_Za_5), \
"an SVE vector register") \
Y(SVE_REG, regno, "SVE_Za_16", 0, F(FLD_SVE_Za_16), \