/* tc-arm.c -- Assemble for the ARM
- Copyright (C) 1994-2019 Free Software Foundation, Inc.
+ Copyright (C) 1994-2020 Free Software Foundation, Inc.
Contributed by Richard Earnshaw (rwe@pegasus.esprit.ec.org)
Modified by David Taylor (dtaylor@armltd.co.uk)
Cirrus coprocessor mods by Aldy Hernandez (aldyh@redhat.com)
#include "obstack.h"
#include "libiberty.h"
#include "opcode/arm.h"
+#include "cpu-arm.h"
#ifdef OBJ_ELF
#include "elf/arm.h"
should define CPU_DEFAULT here. */
#endif
+/* Perform range checks on positive and negative overflows by checking if the
+ VALUE given fits within the range of an BITS sized immediate. */
+static bfd_boolean out_of_range_p (offsetT value, offsetT bits)
+ {
+ gas_assert (bits < (offsetT)(sizeof (value) * 8));
+ return (value & ~((1 << bits)-1))
+ && ((value & ~((1 << bits)-1)) != ~((1 << bits)-1));
+}
+
#ifndef FPU_DEFAULT
# ifdef TE_LINUX
# define FPU_DEFAULT FPU_ARCH_FPA
static int fix_v4bx = FALSE;
/* Warn on using deprecated features. */
static int warn_on_deprecated = TRUE;
+static int warn_on_restrict_it = FALSE;
/* Understand CodeComposer Studio assembly syntax. */
bfd_boolean codecomposer_syntax = FALSE;
static const arm_feature_set arm_ext_v7 = ARM_FEATURE_CORE_LOW (ARM_EXT_V7);
static const arm_feature_set arm_ext_v7a = ARM_FEATURE_CORE_LOW (ARM_EXT_V7A);
static const arm_feature_set arm_ext_v7r = ARM_FEATURE_CORE_LOW (ARM_EXT_V7R);
+static const arm_feature_set arm_ext_v8r = ARM_FEATURE_CORE_HIGH (ARM_EXT2_V8R);
#ifdef OBJ_ELF
static const arm_feature_set ATTRIBUTE_UNUSED arm_ext_v7m = ARM_FEATURE_CORE_LOW (ARM_EXT_V7M);
#endif
ARM_FEATURE_CORE_HIGH (ARM_EXT2_SB);
static const arm_feature_set arm_ext_predres =
ARM_FEATURE_CORE_HIGH (ARM_EXT2_PREDRES);
+static const arm_feature_set arm_ext_bf16 =
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_BF16);
+static const arm_feature_set arm_ext_i8mm =
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_I8MM);
+static const arm_feature_set arm_ext_crc =
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_CRC);
+static const arm_feature_set arm_ext_cde =
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_CDE);
+static const arm_feature_set arm_ext_cde0 =
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_CDE0);
+static const arm_feature_set arm_ext_cde1 =
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_CDE1);
+static const arm_feature_set arm_ext_cde2 =
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_CDE2);
+static const arm_feature_set arm_ext_cde3 =
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_CDE3);
+static const arm_feature_set arm_ext_cde4 =
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_CDE4);
+static const arm_feature_set arm_ext_cde5 =
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_CDE5);
+static const arm_feature_set arm_ext_cde6 =
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_CDE6);
+static const arm_feature_set arm_ext_cde7 =
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_CDE7);
static const arm_feature_set arm_arch_any = ARM_ANY;
-#ifdef OBJ_ELF
static const arm_feature_set fpu_any = FPU_ANY;
-#endif
static const arm_feature_set arm_arch_full ATTRIBUTE_UNUSED = ARM_FEATURE (-1, -1, -1);
static const arm_feature_set arm_arch_t2 = ARM_ARCH_THUMB2;
static const arm_feature_set arm_arch_none = ARM_ARCH_NONE;
static const arm_feature_set fpu_vfp_v3_or_neon_ext =
ARM_FEATURE_COPROC (FPU_NEON_EXT_V1 | FPU_VFP_EXT_V3);
static const arm_feature_set mve_ext =
- ARM_FEATURE_COPROC (FPU_MVE);
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_MVE);
static const arm_feature_set mve_fp_ext =
- ARM_FEATURE_COPROC (FPU_MVE_FP);
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_MVE_FP);
+/* Note: This has more than one bit set, which means using it with
+ mark_feature_used (which returns if *any* of the bits are set in the current
+ cpu variant) can give surprising results. */
+static const arm_feature_set armv8m_fp =
+ ARM_FEATURE_COPROC (FPU_VFP_V5_SP_D16);
#ifdef OBJ_ELF
static const arm_feature_set fpu_vfp_fp16 =
ARM_FEATURE_COPROC (FPU_VFP_EXT_FP16);
ARM_FEATURE_COPROC (FPU_NEON_EXT_ARMV8);
static const arm_feature_set fpu_crypto_ext_armv8 =
ARM_FEATURE_COPROC (FPU_CRYPTO_EXT_ARMV8);
-static const arm_feature_set crc_ext_armv8 =
- ARM_FEATURE_COPROC (CRC_EXT_ARMV8);
static const arm_feature_set fpu_neon_ext_v8_1 =
ARM_FEATURE_COPROC (FPU_NEON_EXT_RDMA);
static const arm_feature_set fpu_neon_ext_dotprod =
/* Feature bits selected by the last .object_arch directive. */
static arm_feature_set selected_object_arch = ARM_ARCH_NONE;
/* Must be long enough to hold any of the names in arm_cpus. */
+static const struct arm_ext_table * selected_ctx_ext_table = NULL;
static char selected_cpu_name[20];
extern FLONUM_TYPE generic_floating_point_number;
NT_float,
NT_poly,
NT_signed,
+ NT_bfloat,
NT_unsigned
};
unsigned size;
};
-#define NEON_MAX_TYPE_ELS 4
+#define NEON_MAX_TYPE_ELS 5
struct neon_type
{
VPT_INSN, /* The VPT/VPST insn has been parsed. */
MVE_OUTSIDE_PRED_INSN , /* Instruction to indicate a MVE instruction without
a predication code. */
- MVE_UNPREDICABLE_INSN /* MVE instruction that is non-predicable. */
+ MVE_UNPREDICABLE_INSN, /* MVE instruction that is non-predicable. */
};
/* The maximum number of operands we need. */
[REG_TYPE_MMXWCG] = N_("iWMMXt scalar register expected"),
[REG_TYPE_XSCALE] = N_("XScale accumulator register expected"),
[REG_TYPE_MQ] = N_("MVE vector register expected"),
- [REG_TYPE_RNB] = N_("")
+ [REG_TYPE_RNB] = ""
};
/* Some well known registers that we refer to directly elsewhere. */
#define BAD_ADDR_MODE _("instruction does not accept this addressing mode")
#define BAD_BRANCH _("branch must be last instruction in IT block")
#define BAD_BRANCH_OFF _("branch out of range or not a multiple of 2")
+#define BAD_NO_VPT _("instruction not allowed in VPT block")
#define BAD_NOT_IT _("instruction not allowed in IT block")
#define BAD_NOT_VPT _("instruction missing MVE vector predication code")
#define BAD_FPU _("selected FPU does not support instruction")
_("cannot use writeback with PC-relative addressing")
#define BAD_RANGE _("branch out of range")
#define BAD_FP16 _("selected processor does not support fp16 instruction")
+#define BAD_BF16 _("selected processor does not support bf16 instruction")
+#define BAD_CDE _("selected processor does not support cde instruction")
+#define BAD_CDE_COPROC _("coprocessor for insn is not enabled for cde")
#define UNPRED_REG(R) _("using " R " results in unpredictable behaviour")
#define THUMB1_RELOC_ONLY _("relocation valid in thumb1 code only")
#define MVE_NOT_IT _("Warning: instruction is UNPREDICTABLE in an IT " \
prec = 1;
break;
+ /* If this is a bfloat16, then parse it slightly differently, as it
+ does not follow the IEEE specification for floating point numbers
+ exactly. */
+ case 'b':
+ {
+ FLONUM_TYPE generic_float;
+
+ t = atof_ieee_detail (input_line_pointer, 1, 8, words, &generic_float);
+
+ if (t)
+ input_line_pointer = t;
+ else
+ return _("invalid floating point number");
+
+ switch (generic_float.sign)
+ {
+ /* Is +Inf. */
+ case 'P':
+ words[0] = 0x7f80;
+ break;
+
+ /* Is -Inf. */
+ case 'N':
+ words[0] = 0xff80;
+ break;
+
+ /* Is NaN. */
+ /* bfloat16 has two types of NaN - quiet and signalling.
+ Quiet NaN has bit[6] == 1 && faction != 0, whereas
+ signalling NaN's have bit[0] == 0 && fraction != 0.
+ Chosen this specific encoding as it is the same form
+ as used by other IEEE 754 encodings in GAS. */
+ case 0:
+ words[0] = 0x7fff;
+ break;
+
+ default:
+ break;
+ }
+
+ *sizeP = 2;
+
+ md_number_to_chars (litP, (valueT) words[0], sizeof (LITTLENUM_TYPE));
+
+ return NULL;
+ }
case 'f':
case 'F':
case 's':
thissize = 64;
ptr++;
goto done;
+ case 'b':
+ thistype = NT_bfloat;
+ switch (TOLOWER (*(++ptr)))
+ {
+ case 'f':
+ ptr += 1;
+ thissize = strtoul (ptr, &ptr, 10);
+ if (thissize != 16)
+ {
+ as_bad (_("bad size %d in type specifier"), thissize);
+ return FAIL;
+ }
+ goto done;
+ case '0': case '1': case '2': case '3': case '4':
+ case '5': case '6': case '7': case '8': case '9':
+ case ' ': case '.':
+ as_bad (_("unexpected type character `b' -- did you mean `bf'?"));
+ return FAIL;
+ default:
+ break;
+ }
+ break;
default:
as_bad (_("unexpected character `%c' in type specifier"), *ptr);
return FAIL;
const char apsr_str[] = "apsr";
int apsr_str_len = strlen (apsr_str);
- reg = arm_reg_parse (&str, REGLIST_RN);
+ reg = arm_reg_parse (&str, REG_TYPE_RN);
if (etype == REGLIST_CLRM)
{
if (reg == REG_SP || reg == REG_PC)
hash_delete (arm_reg_hsh, name, FALSE);
free ((char *) reg->name);
- if (reg->neon)
- free (reg->neon);
+ free (reg->neon);
free (reg);
/* Also locate the all upper case and all lower case versions.
{
hash_delete (arm_reg_hsh, nbuf, FALSE);
free ((char *) reg->name);
- if (reg->neon)
- free (reg->neon);
+ free (reg->neon);
free (reg);
}
{
hash_delete (arm_reg_hsh, nbuf, FALSE);
free ((char *) reg->name);
- if (reg->neon)
- free (reg->neon);
+ free (reg->neon);
free (reg);
}
}
return;
-error:
+ error:
ignore_rest_of_line ();
}
op = 0xc700 | mask;
add_unwind_opcode (op, 2);
return;
-error:
+ error:
ignore_rest_of_line ();
}
as_warn (_("float16 format cannot be set more than once, ignoring."));
}
-cleanup:
+ cleanup:
*input_line_pointer = saved_char;
ignore_rest_of_line ();
}
{ "extend", float_cons, 'x' },
{ "ldouble", float_cons, 'x' },
{ "packed", float_cons, 'p' },
+ { "bfloat16", float_cons, 'b' },
#ifdef TE_PE
{"secrel32", pe_directive_secrel, 0},
#endif
goto unsupported_psr;
p += 4;
-check_suffix:
+ check_suffix:
if (*p == '_')
{
/* A suffix follows. */
OP_RNDMQ, /* Neon double precision (0..31) or MVE vector register. */
OP_RNDMQR, /* Neon double precision (0..31), MVE vector or ARM register.
*/
+ OP_RNSDMQR, /* Neon single or double precision, MVE vector or ARM register.
+ */
OP_RNQ, /* Neon quad precision register */
OP_RNQMQ, /* Neon quad or MVE vector register. */
OP_RVSD, /* VFP single or double precision register */
OP_RIWG, /* iWMMXt wCG register */
OP_RXA, /* XScale accumulator register */
+ OP_RNSDMQ, /* Neon single, double or MVE vector register */
OP_RNSDQMQ, /* Neon single, double or quad register or MVE vector register
*/
OP_RNSDQMQR, /* Neon single, double or quad register, MVE vector register or
OP_I63s, /* -64 .. 63 */
OP_I64, /* 1 .. 64 */
OP_I64z, /* 0 .. 64 */
+ OP_I127, /* 0 .. 127 */
OP_I255, /* 0 .. 255 */
-
+ OP_I511, /* 0 .. 511 */
+ OP_I4095, /* 0 .. 4095 */
+ OP_I8191, /* 0 .. 8191 */
OP_I4b, /* immediate, prefix optional, 1 .. 4 */
OP_I7b, /* 0 .. 7 */
OP_I15b, /* 0 .. 15 */
OP_oRNSDQ, /* Optional single, double or quad precision vector register */
OP_oRNSDQMQ, /* Optional single, double or quad register or MVE vector
register. */
+ OP_oRNSDMQ, /* Optional single, double register or MVE vector
+ register. */
OP_oSHll, /* LSL immediate */
OP_oSHar, /* ASR immediate */
OP_oSHllar, /* LSL or ASR immediate */
case OP_RVS: po_reg_or_fail (REG_TYPE_VFS); break;
case OP_RVD: po_reg_or_fail (REG_TYPE_VFD); break;
case OP_oRND:
+ case OP_RNSDMQR:
+ po_reg_or_goto (REG_TYPE_VFS, try_rndmqr);
+ break;
+ try_rndmqr:
case OP_RNDMQR:
po_reg_or_goto (REG_TYPE_RN, try_rndmq);
break;
case OP_RVSD_COND:
po_reg_or_goto (REG_TYPE_VFSD, try_cond);
break;
+ case OP_oRNSDMQ:
+ case OP_RNSDMQ:
+ po_reg_or_goto (REG_TYPE_NSD, try_mq2);
+ break;
+ try_mq2:
+ po_reg_or_fail (REG_TYPE_MQ);
+ break;
case OP_oRNSDQ:
case OP_RNSDQ: po_reg_or_fail (REG_TYPE_NSDQ); break;
case OP_RNSDQMQR:
case OP_I63: po_imm_or_fail ( 0, 63, FALSE); break;
case OP_I64: po_imm_or_fail ( 1, 64, FALSE); break;
case OP_I64z: po_imm_or_fail ( 0, 64, FALSE); break;
+ case OP_I127: po_imm_or_fail ( 0, 127, FALSE); break;
case OP_I255: po_imm_or_fail ( 0, 255, FALSE); break;
-
+ case OP_I511: po_imm_or_fail ( 0, 511, FALSE); break;
+ case OP_I4095: po_imm_or_fail ( 0, 4095, FALSE); break;
+ case OP_I8191: po_imm_or_fail ( 0, 8191, FALSE); break;
case OP_I4b: po_imm_or_fail ( 1, 4, TRUE); break;
case OP_oI7b:
case OP_I7b: po_imm_or_fail ( 0, 7, TRUE); break;
#define M_MNEM_vqrshrunt 0xfe801fc0
#define M_MNEM_vqrshrunb 0xfe800fc0
+/* Bfloat16 instruction encoder helpers. */
+#define B_MNEM_vfmat 0xfc300850
+#define B_MNEM_vfmab 0xfc300810
+
/* Neon instruction encoder helpers. */
/* Encodings for the different types for various Neon opcodes. */
X(2, (Q, R), QUAD), \
X(2, (D, I), DOUBLE), \
X(2, (Q, I), QUAD), \
+ X(3, (P, F, I), SINGLE), \
+ X(3, (P, D, I), DOUBLE), \
+ X(3, (P, Q, I), QUAD), \
+ X(4, (P, F, F, I), SINGLE), \
+ X(4, (P, D, D, I), DOUBLE), \
+ X(4, (P, Q, Q, I), QUAD), \
+ X(5, (P, F, F, F, I), SINGLE), \
+ X(5, (P, D, D, D, I), DOUBLE), \
+ X(5, (P, Q, Q, Q, I), QUAD), \
X(3, (D, L, D), DOUBLE), \
X(2, (D, Q), MIXED), \
X(2, (Q, D), MIXED), \
#define S2(A,B) NS_##A##B
#define S3(A,B,C) NS_##A##B##C
#define S4(A,B,C,D) NS_##A##B##C##D
+#define S5(A,B,C,D,E) NS_##A##B##C##D##E
#define X(N, L, C) S##N L
#undef S2
#undef S3
#undef S4
+#undef S5
enum neon_shape_class
{
SE_I,
SE_S,
SE_R,
- SE_L
+ SE_L,
+ SE_P
};
/* Register widths of above. */
0,
32,
32,
+ 0,
0
};
#define S2(A,B) { SE_##A, SE_##B }
#define S3(A,B,C) { SE_##A, SE_##B, SE_##C }
#define S4(A,B,C,D) { SE_##A, SE_##B, SE_##C, SE_##D }
+#define S5(A,B,C,D,E) { SE_##A, SE_##B, SE_##C, SE_##D, SE_##E }
#define X(N, L, C) { N, S##N L }
#undef S2
#undef S3
#undef S4
+#undef S5
/* Bit masks used in type checking given instructions.
'N_EQK' means the type must be the same as (or based on in some way) the key
N_F32 = 0x0080000,
N_F64 = 0x0100000,
N_P64 = 0x0200000,
+ N_BF16 = 0x0400000,
N_KEY = 0x1000000, /* Key element (main type specifier). */
N_EQK = 0x2000000, /* Given operand has the same type & size as the key. */
N_VFP = 0x4000000, /* VFP mode: operand size must match register width. */
matches = 0;
break;
+ case SE_P:
case SE_L:
break;
}
}
break;
+ case NT_bfloat:
+ if (size == 16) return N_BF16;
+ break;
+
default: ;
}
if ((mask & (N_S8 | N_U8 | N_I8 | N_8 | N_P8)) != 0)
*size = 8;
- else if ((mask & (N_S16 | N_U16 | N_I16 | N_16 | N_F16 | N_P16)) != 0)
+ else if ((mask & (N_S16 | N_U16 | N_I16 | N_16 | N_F16 | N_P16 | N_BF16))
+ != 0)
*size = 16;
else if ((mask & (N_S32 | N_U32 | N_I32 | N_32 | N_F32)) != 0)
*size = 32;
*type = NT_poly;
else if ((mask & (N_F_ALL)) != 0)
*type = NT_float;
+ else if ((mask & (N_BF16)) != 0)
+ *type = NT_bfloat;
else
return FAIL;
#define LOW4(R) ((R) & 0xf)
#define HI1(R) (((R) >> 4) & 1)
+#define LOW1(R) ((R) & 0x1)
+#define HI4(R) (((R) >> 1) & 0xf)
static unsigned
mve_get_vcmp_vpt_cond (struct neon_type_el et)
inst.operands[0].present = 1;
}
-static void
-do_vfp_nsyn_push (void)
-{
- nsyn_insert_sp ();
-
- constraint (inst.operands[1].imm < 1 || inst.operands[1].imm > 16,
- _("register list must contain at least 1 and at most 16 "
- "registers"));
-
- if (inst.operands[1].issingle)
- do_vfp_nsyn_opcode ("fstmdbs");
- else
- do_vfp_nsyn_opcode ("fstmdbd");
-}
-
-static void
-do_vfp_nsyn_pop (void)
-{
- nsyn_insert_sp ();
-
- constraint (inst.operands[1].imm < 1 || inst.operands[1].imm > 16,
- _("register list must contain at least 1 and at most 16 "
- "registers"));
-
- if (inst.operands[1].issingle)
- do_vfp_nsyn_opcode ("fldmias");
- else
- do_vfp_nsyn_opcode ("fldmiad");
-}
-
/* Fix up Neon data-processing instructions, ORing in the correct bits for
ARM mode or Thumb mode and moving the encoded bit 24 to bit 28. */
inst.is_neon = 1;
}
+/* Helper function for neon_three_same handling the operands. */
+static void
+neon_three_args (int isquad)
+{
+ inst.instruction |= LOW4 (inst.operands[0].reg) << 12;
+ inst.instruction |= HI1 (inst.operands[0].reg) << 22;
+ inst.instruction |= LOW4 (inst.operands[1].reg) << 16;
+ inst.instruction |= HI1 (inst.operands[1].reg) << 7;
+ inst.instruction |= LOW4 (inst.operands[2].reg);
+ inst.instruction |= HI1 (inst.operands[2].reg) << 5;
+ inst.instruction |= (isquad != 0) << 6;
+ inst.is_neon = 1;
+}
+
/* Encode insns with bit pattern:
|28/24|23|22 |21 20|19 16|15 12|11 8|7|6|5|4|3 0|
static void
neon_three_same (int isquad, int ubit, int size)
{
- inst.instruction |= LOW4 (inst.operands[0].reg) << 12;
- inst.instruction |= HI1 (inst.operands[0].reg) << 22;
- inst.instruction |= LOW4 (inst.operands[1].reg) << 16;
- inst.instruction |= HI1 (inst.operands[1].reg) << 7;
- inst.instruction |= LOW4 (inst.operands[2].reg);
- inst.instruction |= HI1 (inst.operands[2].reg) << 5;
- inst.instruction |= (isquad != 0) << 6;
+ neon_three_args (isquad);
inst.instruction |= (ubit != 0) << 24;
if (size != -1)
inst.instruction |= neon_logbits (size) << 20;
do_mve_vstr_vldr_RQ (int size, int elsize, int load)
{
unsigned os = inst.operands[1].imm >> 5;
+ unsigned type = inst.vectype.el[0].type;
constraint (os != 0 && size == 8,
_("can not shift offsets when accessing less than half-word"));
constraint (os && os != neon_logbits (size),
constraint (inst.operands[0].reg == (inst.operands[1].imm & 0x1f),
_("destination register and offset register may not be"
" the same"));
- constraint (size == elsize && inst.vectype.el[0].type != NT_unsigned,
+ constraint (size == elsize && type == NT_signed, BAD_EL_TYPE);
+ constraint (size != elsize && type != NT_unsigned && type != NT_signed,
BAD_EL_TYPE);
- constraint (inst.vectype.el[0].type != NT_unsigned
- && inst.vectype.el[0].type != NT_signed, BAD_EL_TYPE);
- inst.instruction |= (inst.vectype.el[0].type == NT_unsigned) << 28;
+ inst.instruction |= ((size == elsize) || (type == NT_unsigned)) << 28;
}
else
{
- constraint (inst.vectype.el[0].type != NT_untyped, BAD_EL_TYPE);
+ constraint (type != NT_untyped, BAD_EL_TYPE);
}
inst.instruction |= 1 << 23;
}
}
+static void
+do_bfloat_vfma (void)
+{
+ constraint (!mark_feature_used (&fpu_neon_ext_armv8), _(BAD_FPU));
+ constraint (!mark_feature_used (&arm_ext_bf16), _(BAD_BF16));
+ enum neon_shape rs;
+ int t_bit = 0;
+
+ if (inst.instruction != B_MNEM_vfmab)
+ {
+ t_bit = 1;
+ inst.instruction = B_MNEM_vfmat;
+ }
+
+ if (inst.operands[2].isscalar)
+ {
+ rs = neon_select_shape (NS_QQS, NS_NULL);
+ neon_check_type (3, rs, N_EQK, N_EQK, N_BF16 | N_KEY);
+
+ inst.instruction |= (1 << 25);
+ int index = inst.operands[2].reg & 0xf;
+ constraint (!(index < 4), _("index must be in the range 0 to 3"));
+ inst.operands[2].reg >>= 4;
+ constraint (!(inst.operands[2].reg < 8),
+ _("indexed register must be less than 8"));
+ neon_three_args (t_bit);
+ inst.instruction |= ((index & 1) << 3);
+ inst.instruction |= ((index & 2) << 4);
+ }
+ else
+ {
+ rs = neon_select_shape (NS_QQQ, NS_NULL);
+ neon_check_type (3, rs, N_EQK, N_EQK, N_BF16 | N_KEY);
+ neon_three_args (t_bit);
+ }
+
+}
+
static void
do_neon_fmac (void)
{
if (rs == NS_QQR)
{
+
if (inst.operands[2].reg == REG_SP)
as_tsktsk (MVE_BAD_SP);
else if (inst.operands[2].reg == REG_PC)
neon_dyadic_misc (NT_untyped, N_IF_32, 0);
}
+static void
+do_mve_vfma (void)
+{
+ if (!ARM_CPU_HAS_FEATURE (cpu_variant, arm_ext_bf16) &&
+ inst.cond == COND_ALWAYS)
+ {
+ constraint (!ARM_CPU_HAS_FEATURE (cpu_variant, mve_ext), BAD_FPU);
+ inst.instruction = N_MNEM_vfma;
+ inst.pred_insn_type = INSIDE_VPT_INSN;
+ inst.cond = 0xf;
+ return do_neon_fmac();
+ }
+ else
+ {
+ do_bfloat_vfma();
+ }
+}
+
static void
do_neon_tst (void)
{
enum neon_shape rs = neon_select_shape (NS_HHH, NS_FFF, NS_DDD, NS_DDS,
NS_QQS, NS_QQQ, NS_QQR, NS_NULL);
- if (!ARM_CPU_HAS_FEATURE (cpu_variant, mve_ext)
- && inst.cond == COND_ALWAYS
+ if (inst.cond == COND_ALWAYS
&& ((unsigned)inst.instruction) == M_MNEM_vmullt)
{
+
if (rs == NS_QQQ)
{
-
- struct neon_type_el et = neon_check_type (3, rs, N_EQK , N_EQK,
- N_SUF_32 | N_F64 | N_P8
- | N_P16 | N_I_MVE | N_KEY);
- if (((et.type == NT_poly) && et.size == 8
- && ARM_CPU_IS_ANY (cpu_variant))
- || (et.type == NT_integer) || (et.type == NT_float))
+ if (!ARM_CPU_HAS_FEATURE (cpu_variant, mve_ext))
goto neon_vmul;
}
else
return;
-neon_vmul:
+ neon_vmul:
inst.instruction = N_MNEM_vmul;
inst.cond = 0xb;
if (thumb_mode)
CVT_VAR (f16_u32, N_F16 | N_KEY, N_U32, N_VFP, "fultos", "fuitos", NULL) \
CVT_VAR (u32_f16, N_U32, N_F16 | N_KEY, N_VFP, "ftouls", "ftouis", "ftouizs")\
CVT_VAR (s32_f16, N_S32, N_F16 | N_KEY, N_VFP, "ftosls", "ftosis", "ftosizs")\
+ CVT_VAR (bf16_f32, N_BF16, N_F32, whole_reg, NULL, NULL, NULL) \
/* VFP instructions. */ \
CVT_VAR (f32_f64, N_F32, N_F64, N_VFP, NULL, "fcvtsd", NULL) \
CVT_VAR (f64_f32, N_F64, N_F32, N_VFP, NULL, "fcvtds", NULL) \
}
if (rs == NS_DQ)
- inst.instruction = 0x3b60600;
+ {
+ if (flavour == neon_cvt_flavour_bf16_f32)
+ {
+ if (vfp_or_neon_is_neon (NEON_CHECK_ARCH8) == FAIL)
+ return;
+ constraint (!mark_feature_used (&arm_ext_bf16), _(BAD_BF16));
+ /* VCVT.bf16.f32. */
+ inst.instruction = 0x11b60640;
+ }
+ else
+ /* VCVT.f16.f32. */
+ inst.instruction = 0x3b60600;
+ }
else
+ /* VCVT.f32.f16. */
inst.instruction = 0x3b60700;
inst.instruction |= LOW4 (inst.operands[0].reg) << 12;
inst.error = NULL;
do_neon_cvttb_2 (t, /*to=*/FALSE, /*is_double=*/TRUE);
}
+ else if (neon_check_type (2, rs, N_BF16 | N_VFP, N_F32).type != NT_invtype)
+ {
+ constraint (!mark_feature_used (&arm_ext_bf16), _(BAD_BF16));
+ inst.error = NULL;
+ inst.instruction |= (1 << 8);
+ inst.instruction &= ~(1 << 9);
+ do_neon_cvttb_2 (t, /*to=*/TRUE, /*is_double=*/FALSE);
+ }
else
return;
}
if (ARM_CPU_HAS_FEATURE (cpu_variant, mve_ext))
{
constraint (!inst.operands[1].isreg && !inst.operands[0].isquad, BAD_FPU);
- constraint ((inst.instruction & 0xd00) == 0xd00,
- _("immediate value out of range"));
}
}
static void
do_neon_dyadic_long (void)
{
- enum neon_shape rs = neon_select_shape (NS_QDD, NS_QQQ, NS_QQR, NS_NULL);
+ enum neon_shape rs = neon_select_shape (NS_QDD, NS_HHH, NS_FFF, NS_DDD, NS_NULL);
if (rs == NS_QDD)
{
if (vfp_or_neon_is_neon (NEON_CHECK_ARCH | NEON_CHECK_CC) == FAIL)
| ((elno & 0x1) << 3));
}
-bad_scalar:
+ bad_scalar:
first_error (_("scalar out of range for multiply instruction"));
return 0;
}
0x2. */
int size = -1;
- if (inst.cond != COND_ALWAYS)
- as_warn (_("vfmal/vfmsl with FP16 type cannot be conditional, the "
- "behaviour is UNPREDICTABLE"));
-
- constraint (!ARM_CPU_HAS_FEATURE (cpu_variant, arm_ext_fp16_fml),
- _(BAD_FP16));
-
- constraint (!ARM_CPU_HAS_FEATURE (cpu_variant, fpu_neon_ext_armv8),
- _(BAD_FPU));
-
/* vfmal/vfmsl are in three-same D/Q register format or the third operand can
be a scalar index register. */
if (inst.operands[2].isscalar)
rs = neon_select_shape (NS_DHH, NS_QDD, NS_NULL);
}
- neon_check_type (3, rs, N_EQK, N_EQK, N_KEY | N_F16);
+
+ if (inst.cond != COND_ALWAYS)
+ as_warn (_("vfmal/vfmsl with FP16 type cannot be conditional, the "
+ "behaviour is UNPREDICTABLE"));
+
+ constraint (!ARM_CPU_HAS_FEATURE (cpu_variant, arm_ext_fp16_fml),
+ _(BAD_FP16));
+
+ constraint (!ARM_CPU_HAS_FEATURE (cpu_variant, fpu_neon_ext_armv8),
+ _(BAD_FPU));
/* "opcode" from template has included "ubit", so simply pass 0 here. Also,
the "S" bit in size field has been reused to differentiate vfmal and vfmsl,
inst.instruction &= 0x00ffffff;
inst.instruction |= high8;
-#define LOW1(R) ((R) & 0x1)
-#define HI4(R) (((R) >> 1) & 0xf)
/* Unlike usually NEON three-same, encoding for Vn and Vm will depend on
whether the instruction is in Q form and whether Vm is a scalar indexed
operand. */
static void
do_neon_ldm_stm (void)
{
+ constraint (!ARM_CPU_HAS_FEATURE (cpu_variant, fpu_vfp_ext_v1xd)
+ && !ARM_CPU_HAS_FEATURE (cpu_variant, mve_ext),
+ _(BAD_FPU));
/* P, U and L bits are part of bitmask. */
int is_dbmode = (inst.instruction & (1 << 24)) != 0;
unsigned offsetbits = inst.operands[1].imm * 2;
do_vfp_cond_or_thumb ();
}
+static void
+do_vfp_nsyn_pop (void)
+{
+ nsyn_insert_sp ();
+ if (ARM_CPU_HAS_FEATURE (cpu_variant, mve_ext)) {
+ return do_vfp_nsyn_opcode ("vldm");
+ }
+
+ constraint (!ARM_CPU_HAS_FEATURE (cpu_variant, fpu_vfp_ext_v1xd),
+ _(BAD_FPU));
+
+ constraint (inst.operands[1].imm < 1 || inst.operands[1].imm > 16,
+ _("register list must contain at least 1 and at most 16 "
+ "registers"));
+
+ if (inst.operands[1].issingle)
+ do_vfp_nsyn_opcode ("fldmias");
+ else
+ do_vfp_nsyn_opcode ("fldmiad");
+}
+
+static void
+do_vfp_nsyn_push (void)
+{
+ nsyn_insert_sp ();
+ if (ARM_CPU_HAS_FEATURE (cpu_variant, mve_ext)) {
+ return do_vfp_nsyn_opcode ("vstmdb");
+ }
+
+ constraint (!ARM_CPU_HAS_FEATURE (cpu_variant, fpu_vfp_ext_v1xd),
+ _(BAD_FPU));
+
+ constraint (inst.operands[1].imm < 1 || inst.operands[1].imm > 16,
+ _("register list must contain at least 1 and at most 16 "
+ "registers"));
+
+ if (inst.operands[1].issingle)
+ do_vfp_nsyn_opcode ("fstmdbs");
+ else
+ do_vfp_nsyn_opcode ("fstmdbd");
+}
+
+
static void
do_neon_ldr_str (void)
{
/* VLDR/VSTR. */
else
{
- if (!mark_feature_used (&fpu_vfp_ext_v1xd))
+ if (!mark_feature_used (&fpu_vfp_ext_v1xd)
+ && !ARM_CPU_HAS_FEATURE (cpu_variant, mve_ext))
as_bad (_("Instruction not permitted on this architecture"));
do_neon_ldr_str ();
}
return do_neon_dotproduct (1);
}
-/* Crypto v1 instructions. */
static void
-do_crypto_2op_1 (unsigned elttype, int op)
+do_vusdot (void)
{
+ enum neon_shape rs;
set_pred_insn_type (OUTSIDE_PRED_INSN);
+ if (inst.operands[2].isscalar)
+ {
+ rs = neon_select_shape (NS_DDS, NS_QQS, NS_NULL);
+ neon_check_type (3, rs, N_EQK, N_EQK, N_S8 | N_KEY);
- if (neon_check_type (2, NS_QQ, N_EQK | N_UNT, elttype | N_UNT | N_KEY).type
- == NT_invtype)
- return;
-
- inst.error = NULL;
-
- NEON_ENCODE (INTEGER, inst);
- inst.instruction |= LOW4 (inst.operands[0].reg) << 12;
- inst.instruction |= HI1 (inst.operands[0].reg) << 22;
- inst.instruction |= LOW4 (inst.operands[1].reg);
- inst.instruction |= HI1 (inst.operands[1].reg) << 5;
- if (op != -1)
- inst.instruction |= op << 6;
-
- if (thumb_mode)
- inst.instruction |= 0xfc000000;
+ inst.instruction |= (1 << 25);
+ int index = inst.operands[2].reg & 0xf;
+ constraint ((index != 1 && index != 0), _("index must be 0 or 1"));
+ inst.operands[2].reg >>= 4;
+ constraint (!(inst.operands[2].reg < 16),
+ _("indexed register must be less than 16"));
+ neon_three_args (rs == NS_QQS);
+ inst.instruction |= (index << 5);
+ }
else
- inst.instruction |= 0xf0000000;
+ {
+ inst.instruction |= (1 << 21);
+ rs = neon_select_shape (NS_DDD, NS_QQQ, NS_NULL);
+ neon_check_type (3, rs, N_EQK, N_EQK, N_S8 | N_KEY);
+ neon_three_args (rs == NS_QQQ);
+ }
}
static void
-do_crypto_3op_1 (int u, int op)
+do_vsudot (void)
{
+ enum neon_shape rs;
set_pred_insn_type (OUTSIDE_PRED_INSN);
+ if (inst.operands[2].isscalar)
+ {
+ rs = neon_select_shape (NS_DDS, NS_QQS, NS_NULL);
+ neon_check_type (3, rs, N_EQK, N_EQK, N_U8 | N_KEY);
- if (neon_check_type (3, NS_QQQ, N_EQK | N_UNT, N_EQK | N_UNT,
- N_32 | N_UNT | N_KEY).type == NT_invtype)
- return;
+ inst.instruction |= (1 << 25);
+ int index = inst.operands[2].reg & 0xf;
+ constraint ((index != 1 && index != 0), _("index must be 0 or 1"));
+ inst.operands[2].reg >>= 4;
+ constraint (!(inst.operands[2].reg < 16),
+ _("indexed register must be less than 16"));
+ neon_three_args (rs == NS_QQS);
+ inst.instruction |= (index << 5);
+ }
+}
- inst.error = NULL;
+static void
+do_vsmmla (void)
+{
+ enum neon_shape rs = neon_select_shape (NS_QQQ, NS_NULL);
+ neon_check_type (3, rs, N_EQK, N_EQK, N_S8 | N_KEY);
+
+ set_pred_insn_type (OUTSIDE_PRED_INSN);
+
+ neon_three_args (1);
- NEON_ENCODE (INTEGER, inst);
- neon_three_same (1, u, 8 << op);
}
static void
-do_aese (void)
+do_vummla (void)
{
- do_crypto_2op_1 (N_8, 0);
+ enum neon_shape rs = neon_select_shape (NS_QQQ, NS_NULL);
+ neon_check_type (3, rs, N_EQK, N_EQK, N_U8 | N_KEY);
+
+ set_pred_insn_type (OUTSIDE_PRED_INSN);
+
+ neon_three_args (1);
+
}
static void
-do_aesd (void)
+check_cde_operand (size_t index, int is_dual)
+{
+ unsigned Rx = inst.operands[index].reg;
+ bfd_boolean isvec = inst.operands[index].isvec;
+ if (is_dual == 0 && thumb_mode)
+ constraint (
+ !((Rx <= 14 && Rx != 13) || (Rx == REG_PC && isvec)),
+ _("Register must be r0-r14 except r13, or APSR_nzcv."));
+ else
+ constraint ( !((Rx <= 10 && Rx % 2 == 0 )),
+ _("Register must be an even register between r0-r10."));
+}
+
+static bfd_boolean
+cde_coproc_enabled (unsigned coproc)
+{
+ switch (coproc)
+ {
+ case 0: return mark_feature_used (&arm_ext_cde0);
+ case 1: return mark_feature_used (&arm_ext_cde1);
+ case 2: return mark_feature_used (&arm_ext_cde2);
+ case 3: return mark_feature_used (&arm_ext_cde3);
+ case 4: return mark_feature_used (&arm_ext_cde4);
+ case 5: return mark_feature_used (&arm_ext_cde5);
+ case 6: return mark_feature_used (&arm_ext_cde6);
+ case 7: return mark_feature_used (&arm_ext_cde7);
+ default: return FALSE;
+ }
+}
+
+#define cde_coproc_pos 8
+static void
+cde_handle_coproc (void)
+{
+ unsigned coproc = inst.operands[0].reg;
+ constraint (coproc > 7, _("CDE Coprocessor must be in range 0-7"));
+ constraint (!(cde_coproc_enabled (coproc)), BAD_CDE_COPROC);
+ inst.instruction |= coproc << cde_coproc_pos;
+}
+#undef cde_coproc_pos
+
+static void
+cxn_handle_predication (bfd_boolean is_accum)
+{
+ if (is_accum && conditional_insn ())
+ set_pred_insn_type (INSIDE_IT_INSN);
+ else if (conditional_insn ())
+ /* conditional_insn essentially checks for a suffix, not whether the
+ instruction is inside an IT block or not.
+ The non-accumulator versions should not have suffixes. */
+ inst.error = BAD_SYNTAX;
+ else
+ set_pred_insn_type (OUTSIDE_PRED_INSN);
+}
+
+static void
+do_custom_instruction_1 (int is_dual, bfd_boolean is_accum)
+{
+
+ constraint (!mark_feature_used (&arm_ext_cde), _(BAD_CDE));
+
+ unsigned imm, Rd;
+
+ Rd = inst.operands[1].reg;
+ check_cde_operand (1, is_dual);
+
+ if (is_dual == 1)
+ {
+ constraint (inst.operands[2].reg != Rd + 1,
+ _("cx1d requires consecutive destination registers."));
+ imm = inst.operands[3].imm;
+ }
+ else if (is_dual == 0)
+ imm = inst.operands[2].imm;
+ else
+ abort ();
+
+ inst.instruction |= Rd << 12;
+ inst.instruction |= (imm & 0x1F80) << 9;
+ inst.instruction |= (imm & 0x0040) << 1;
+ inst.instruction |= (imm & 0x003f);
+
+ cde_handle_coproc ();
+ cxn_handle_predication (is_accum);
+}
+
+static void
+do_custom_instruction_2 (int is_dual, bfd_boolean is_accum)
+{
+
+ constraint (!mark_feature_used (&arm_ext_cde), _(BAD_CDE));
+
+ unsigned imm, Rd, Rn;
+
+ Rd = inst.operands[1].reg;
+
+ if (is_dual == 1)
+ {
+ constraint (inst.operands[2].reg != Rd + 1,
+ _("cx2d requires consecutive destination registers."));
+ imm = inst.operands[4].imm;
+ Rn = inst.operands[3].reg;
+ }
+ else if (is_dual == 0)
+ {
+ imm = inst.operands[3].imm;
+ Rn = inst.operands[2].reg;
+ }
+ else
+ abort ();
+
+ check_cde_operand (2 + is_dual, /* is_dual = */0);
+ check_cde_operand (1, is_dual);
+
+ inst.instruction |= Rd << 12;
+ inst.instruction |= Rn << 16;
+
+ inst.instruction |= (imm & 0x0380) << 13;
+ inst.instruction |= (imm & 0x0040) << 1;
+ inst.instruction |= (imm & 0x003f);
+
+ cde_handle_coproc ();
+ cxn_handle_predication (is_accum);
+}
+
+static void
+do_custom_instruction_3 (int is_dual, bfd_boolean is_accum)
+{
+
+ constraint (!mark_feature_used (&arm_ext_cde), _(BAD_CDE));
+
+ unsigned imm, Rd, Rn, Rm;
+
+ Rd = inst.operands[1].reg;
+
+ if (is_dual == 1)
+ {
+ constraint (inst.operands[2].reg != Rd + 1,
+ _("cx3d requires consecutive destination registers."));
+ imm = inst.operands[5].imm;
+ Rn = inst.operands[3].reg;
+ Rm = inst.operands[4].reg;
+ }
+ else if (is_dual == 0)
+ {
+ imm = inst.operands[4].imm;
+ Rn = inst.operands[2].reg;
+ Rm = inst.operands[3].reg;
+ }
+ else
+ abort ();
+
+ check_cde_operand (1, is_dual);
+ check_cde_operand (2 + is_dual, /* is_dual = */0);
+ check_cde_operand (3 + is_dual, /* is_dual = */0);
+
+ inst.instruction |= Rd;
+ inst.instruction |= Rn << 16;
+ inst.instruction |= Rm << 12;
+
+ inst.instruction |= (imm & 0x0038) << 17;
+ inst.instruction |= (imm & 0x0004) << 5;
+ inst.instruction |= (imm & 0x0003) << 4;
+
+ cde_handle_coproc ();
+ cxn_handle_predication (is_accum);
+}
+
+static void
+do_cx1 (void)
+{
+ return do_custom_instruction_1 (0, 0);
+}
+
+static void
+do_cx1a (void)
+{
+ return do_custom_instruction_1 (0, 1);
+}
+
+static void
+do_cx1d (void)
+{
+ return do_custom_instruction_1 (1, 0);
+}
+
+static void
+do_cx1da (void)
+{
+ return do_custom_instruction_1 (1, 1);
+}
+
+static void
+do_cx2 (void)
+{
+ return do_custom_instruction_2 (0, 0);
+}
+
+static void
+do_cx2a (void)
+{
+ return do_custom_instruction_2 (0, 1);
+}
+
+static void
+do_cx2d (void)
+{
+ return do_custom_instruction_2 (1, 0);
+}
+
+static void
+do_cx2da (void)
+{
+ return do_custom_instruction_2 (1, 1);
+}
+
+static void
+do_cx3 (void)
+{
+ return do_custom_instruction_3 (0, 0);
+}
+
+static void
+do_cx3a (void)
+{
+ return do_custom_instruction_3 (0, 1);
+}
+
+static void
+do_cx3d (void)
+{
+ return do_custom_instruction_3 (1, 0);
+}
+
+static void
+do_cx3da (void)
+{
+ return do_custom_instruction_3 (1, 1);
+}
+
+static void
+vcx_assign_vec_d (unsigned regnum)
+{
+ inst.instruction |= HI4 (regnum) << 12;
+ inst.instruction |= LOW1 (regnum) << 22;
+}
+
+static void
+vcx_assign_vec_m (unsigned regnum)
+{
+ inst.instruction |= HI4 (regnum);
+ inst.instruction |= LOW1 (regnum) << 5;
+}
+
+static void
+vcx_assign_vec_n (unsigned regnum)
+{
+ inst.instruction |= HI4 (regnum) << 16;
+ inst.instruction |= LOW1 (regnum) << 7;
+}
+
+enum vcx_reg_type {
+ q_reg,
+ d_reg,
+ s_reg
+};
+
+static enum vcx_reg_type
+vcx_get_reg_type (enum neon_shape ns)
+{
+ gas_assert (ns == NS_PQI
+ || ns == NS_PDI
+ || ns == NS_PFI
+ || ns == NS_PQQI
+ || ns == NS_PDDI
+ || ns == NS_PFFI
+ || ns == NS_PQQQI
+ || ns == NS_PDDDI
+ || ns == NS_PFFFI);
+ if (ns == NS_PQI || ns == NS_PQQI || ns == NS_PQQQI)
+ return q_reg;
+ if (ns == NS_PDI || ns == NS_PDDI || ns == NS_PDDDI)
+ return d_reg;
+ return s_reg;
+}
+
+#define vcx_size_pos 24
+#define vcx_vec_pos 6
+static unsigned
+vcx_handle_shape (enum vcx_reg_type reg_type)
+{
+ unsigned mult = 2;
+ if (reg_type == q_reg)
+ inst.instruction |= 1 << vcx_vec_pos;
+ else if (reg_type == d_reg)
+ inst.instruction |= 1 << vcx_size_pos;
+ else
+ mult = 1;
+ /* NOTE:
+ The documentation says that the Q registers are encoded as 2*N in the D:Vd
+ bits (or equivalent for N and M registers).
+ Similarly the D registers are encoded as N in D:Vd bits.
+ While the S registers are encoded as N in the Vd:D bits.
+
+ Taking into account the maximum values of these registers we can see a
+ nicer pattern for calculation:
+ Q -> 7, D -> 15, S -> 31
+
+ If we say that everything is encoded in the Vd:D bits, then we can say
+ that Q is encoded as 4*N, and D is encoded as 2*N.
+ This way the bits will end up the same, and calculation is simpler.
+ (calculation is now:
+ 1. Multiply by a number determined by the register letter.
+ 2. Encode resulting number in Vd:D bits.)
+
+ This is made a little more complicated by automatic handling of 'Q'
+ registers elsewhere, which means the register number is already 2*N where
+ N is the number the user wrote after the register letter.
+ */
+ return mult;
+}
+#undef vcx_vec_pos
+#undef vcx_size_pos
+
+static void
+vcx_ensure_register_in_range (unsigned R, enum vcx_reg_type reg_type)
+{
+ if (reg_type == q_reg)
+ {
+ gas_assert (R % 2 == 0);
+ constraint (R >= 16, _("'q' register must be in range 0-7"));
+ }
+ else if (reg_type == d_reg)
+ constraint (R >= 16, _("'d' register must be in range 0-15"));
+ else
+ constraint (R >= 32, _("'s' register must be in range 0-31"));
+}
+
+static void (*vcx_assign_vec[3]) (unsigned) = {
+ vcx_assign_vec_d,
+ vcx_assign_vec_m,
+ vcx_assign_vec_n
+};
+
+static void
+vcx_handle_register_arguments (unsigned num_registers,
+ enum vcx_reg_type reg_type)
+{
+ unsigned R, i;
+ unsigned reg_mult = vcx_handle_shape (reg_type);
+ for (i = 0; i < num_registers; i++)
+ {
+ R = inst.operands[i+1].reg;
+ vcx_ensure_register_in_range (R, reg_type);
+ if (num_registers == 3 && i > 0)
+ {
+ if (i == 2)
+ vcx_assign_vec[1] (R * reg_mult);
+ else
+ vcx_assign_vec[2] (R * reg_mult);
+ continue;
+ }
+ vcx_assign_vec[i](R * reg_mult);
+ }
+}
+
+static void
+vcx_handle_insn_block (enum vcx_reg_type reg_type)
+{
+ if (reg_type == q_reg)
+ if (inst.cond > COND_ALWAYS)
+ inst.pred_insn_type = INSIDE_VPT_INSN;
+ else
+ inst.pred_insn_type = MVE_OUTSIDE_PRED_INSN;
+ else if (inst.cond == COND_ALWAYS)
+ inst.pred_insn_type = OUTSIDE_PRED_INSN;
+ else
+ inst.error = BAD_NOT_IT;
+}
+
+static void
+vcx_handle_common_checks (unsigned num_args, enum neon_shape rs)
+{
+ constraint (!mark_feature_used (&arm_ext_cde), _(BAD_CDE));
+ cde_handle_coproc ();
+ enum vcx_reg_type reg_type = vcx_get_reg_type (rs);
+ vcx_handle_register_arguments (num_args, reg_type);
+ vcx_handle_insn_block (reg_type);
+ if (reg_type == q_reg)
+ constraint (!mark_feature_used (&mve_ext),
+ _("vcx instructions with Q registers require MVE"));
+ else
+ constraint (!(ARM_FSET_CPU_SUBSET (armv8m_fp, cpu_variant)
+ && mark_feature_used (&armv8m_fp))
+ && !mark_feature_used (&mve_ext),
+ _("vcx instructions with S or D registers require either MVE"
+ " or Armv8-M floating point etension."));
+}
+
+static void
+do_vcx1 (void)
+{
+ enum neon_shape rs = neon_select_shape (NS_PQI, NS_PDI, NS_PFI, NS_NULL);
+ vcx_handle_common_checks (1, rs);
+
+ unsigned imm = inst.operands[2].imm;
+ inst.instruction |= (imm & 0x03f);
+ inst.instruction |= (imm & 0x040) << 1;
+ inst.instruction |= (imm & 0x780) << 9;
+ if (rs != NS_PQI)
+ constraint (imm >= 2048,
+ _("vcx1 with S or D registers takes immediate within 0-2047"));
+ inst.instruction |= (imm & 0x800) << 13;
+}
+
+static void
+do_vcx2 (void)
+{
+ enum neon_shape rs = neon_select_shape (NS_PQQI, NS_PDDI, NS_PFFI, NS_NULL);
+ vcx_handle_common_checks (2, rs);
+
+ unsigned imm = inst.operands[3].imm;
+ inst.instruction |= (imm & 0x01) << 4;
+ inst.instruction |= (imm & 0x02) << 6;
+ inst.instruction |= (imm & 0x3c) << 14;
+ if (rs != NS_PQQI)
+ constraint (imm >= 64,
+ _("vcx2 with S or D registers takes immediate within 0-63"));
+ inst.instruction |= (imm & 0x40) << 18;
+}
+
+static void
+do_vcx3 (void)
+{
+ enum neon_shape rs = neon_select_shape (NS_PQQQI, NS_PDDDI, NS_PFFFI, NS_NULL);
+ vcx_handle_common_checks (3, rs);
+
+ unsigned imm = inst.operands[4].imm;
+ inst.instruction |= (imm & 0x1) << 4;
+ inst.instruction |= (imm & 0x6) << 19;
+ if (rs != NS_PQQQI)
+ constraint (imm >= 8,
+ _("vcx2 with S or D registers takes immediate within 0-7"));
+ inst.instruction |= (imm & 0x8) << 21;
+}
+
+/* Crypto v1 instructions. */
+static void
+do_crypto_2op_1 (unsigned elttype, int op)
+{
+ set_pred_insn_type (OUTSIDE_PRED_INSN);
+
+ if (neon_check_type (2, NS_QQ, N_EQK | N_UNT, elttype | N_UNT | N_KEY).type
+ == NT_invtype)
+ return;
+
+ inst.error = NULL;
+
+ NEON_ENCODE (INTEGER, inst);
+ inst.instruction |= LOW4 (inst.operands[0].reg) << 12;
+ inst.instruction |= HI1 (inst.operands[0].reg) << 22;
+ inst.instruction |= LOW4 (inst.operands[1].reg);
+ inst.instruction |= HI1 (inst.operands[1].reg) << 5;
+ if (op != -1)
+ inst.instruction |= op << 6;
+
+ if (thumb_mode)
+ inst.instruction |= 0xfc000000;
+ else
+ inst.instruction |= 0xf0000000;
+}
+
+static void
+do_crypto_3op_1 (int u, int op)
+{
+ set_pred_insn_type (OUTSIDE_PRED_INSN);
+
+ if (neon_check_type (3, NS_QQQ, N_EQK | N_UNT, N_EQK | N_UNT,
+ N_32 | N_UNT | N_KEY).type == NT_invtype)
+ return;
+
+ inst.error = NULL;
+
+ NEON_ENCODE (INTEGER, inst);
+ neon_three_same (1, u, 8 << op);
+}
+
+static void
+do_aese (void)
+{
+ do_crypto_2op_1 (N_8, 0);
+}
+
+static void
+do_aesd (void)
{
do_crypto_2op_1 (N_8, 1);
}
do_vfp_cond_or_thumb ();
}
+static void
+do_vdot (void)
+{
+ enum neon_shape rs;
+ constraint (!mark_feature_used (&fpu_neon_ext_armv8), _(BAD_FPU));
+ set_pred_insn_type (OUTSIDE_PRED_INSN);
+ if (inst.operands[2].isscalar)
+ {
+ rs = neon_select_shape (NS_DDS, NS_QQS, NS_NULL);
+ neon_check_type (3, rs, N_EQK, N_EQK, N_BF16 | N_KEY);
+
+ inst.instruction |= (1 << 25);
+ int index = inst.operands[2].reg & 0xf;
+ constraint ((index != 1 && index != 0), _("index must be 0 or 1"));
+ inst.operands[2].reg >>= 4;
+ constraint (!(inst.operands[2].reg < 16),
+ _("indexed register must be less than 16"));
+ neon_three_args (rs == NS_QQS);
+ inst.instruction |= (index << 5);
+ }
+ else
+ {
+ rs = neon_select_shape (NS_DDD, NS_QQQ, NS_NULL);
+ neon_check_type (3, rs, N_EQK, N_EQK, N_BF16 | N_KEY);
+ neon_three_args (rs == NS_QQQ);
+ }
+}
+
+static void
+do_vmmla (void)
+{
+ enum neon_shape rs = neon_select_shape (NS_QQQ, NS_NULL);
+ neon_check_type (3, rs, N_EQK, N_EQK, N_BF16 | N_KEY);
+
+ constraint (!mark_feature_used (&fpu_neon_ext_armv8), _(BAD_FPU));
+ set_pred_insn_type (OUTSIDE_PRED_INSN);
+
+ neon_three_args (1);
+}
+
\f
/* Overall per-instruction processing. */
close_automatic_it_block ();
break;
+ /* Fallthrough. */
case NEUTRAL_IT_INSN:
now_pred.block_length++;
now_pred.insn_cond = TRUE;
handle_pred_state ();
if (now_pred.insn_cond
+ && warn_on_restrict_it
&& !now_pred.warn_deprecated
&& warn_on_deprecated
- && ARM_CPU_HAS_FEATURE (cpu_variant, arm_ext_v8)
+ && (ARM_CPU_HAS_FEATURE (cpu_variant, arm_ext_v8)
+ || ARM_CPU_HAS_FEATURE (cpu_variant, arm_ext_v8r))
&& !ARM_CPU_HAS_FEATURE (cpu_variant, arm_ext_m))
{
if (inst.instruction >= 0x10000)
out of the jump table, and chaos would ensue. */
if (label_is_thumb_function_name
&& (S_GET_NAME (sym)[0] != '.' || S_GET_NAME (sym)[1] != 'L')
- && (bfd_get_section_flags (stdoutput, now_seg) & SEC_CODE) != 0)
+ && (bfd_section_flags (now_seg) & SEC_CODE) != 0)
{
/* When the address of a Thumb function is taken the bottom
bit of that address should be set. This will allow
nUF(sha256su0, _sha2op, 2, (RNQ, RNQ), sha256su0),
#undef ARM_VARIANT
-#define ARM_VARIANT & crc_ext_armv8
+#define ARM_VARIANT & arm_ext_crc
#undef THUMB_VARIANT
-#define THUMB_VARIANT & crc_ext_armv8
+#define THUMB_VARIANT & arm_ext_crc
TUEc("crc32b", 1000040, fac0f080, 3, (RR, oRR, RR), crc32b),
TUEc("crc32h", 1200040, fac0f090, 3, (RR, oRR, RR), crc32h),
TUEc("crc32w", 1400040, fac0f0a0, 3, (RR, oRR, RR), crc32w),
#define THUMB_VARIANT & arm_ext_v6t2
mcCE(vmrs, ef00a10, 2, (APSR_RR, RVC), vmrs),
mcCE(vmsr, ee00a10, 2, (RVC, RR), vmsr),
+ mcCE(fldd, d100b00, 2, (RVD, ADDRGLDC), vfp_dp_ldst),
+ mcCE(fstd, d000b00, 2, (RVD, ADDRGLDC), vfp_dp_ldst),
+ mcCE(flds, d100a00, 2, (RVS, ADDRGLDC), vfp_sp_ldst),
+ mcCE(fsts, d000a00, 2, (RVS, ADDRGLDC), vfp_sp_ldst),
+
+ /* Memory operations. */
+ mcCE(fldmias, c900a00, 2, (RRnpctw, VRSLST), vfp_sp_ldstmia),
+ mcCE(fldmdbs, d300a00, 2, (RRnpctw, VRSLST), vfp_sp_ldstmdb),
+ mcCE(fstmias, c800a00, 2, (RRnpctw, VRSLST), vfp_sp_ldstmia),
+ mcCE(fstmdbs, d200a00, 2, (RRnpctw, VRSLST), vfp_sp_ldstmdb),
#undef THUMB_VARIANT
/* Moves and type conversions. */
cCE("fmxr", ee00a10, 2, (RVC, RR), rn_rd),
/* Memory operations. */
- cCE("flds", d100a00, 2, (RVS, ADDRGLDC), vfp_sp_ldst),
- cCE("fsts", d000a00, 2, (RVS, ADDRGLDC), vfp_sp_ldst),
- cCE("fldmias", c900a00, 2, (RRnpctw, VRSLST), vfp_sp_ldstmia),
cCE("fldmfds", c900a00, 2, (RRnpctw, VRSLST), vfp_sp_ldstmia),
- cCE("fldmdbs", d300a00, 2, (RRnpctw, VRSLST), vfp_sp_ldstmdb),
cCE("fldmeas", d300a00, 2, (RRnpctw, VRSLST), vfp_sp_ldstmdb),
cCE("fldmiax", c900b00, 2, (RRnpctw, VRDLST), vfp_xp_ldstmia),
cCE("fldmfdx", c900b00, 2, (RRnpctw, VRDLST), vfp_xp_ldstmia),
cCE("fldmdbx", d300b00, 2, (RRnpctw, VRDLST), vfp_xp_ldstmdb),
cCE("fldmeax", d300b00, 2, (RRnpctw, VRDLST), vfp_xp_ldstmdb),
- cCE("fstmias", c800a00, 2, (RRnpctw, VRSLST), vfp_sp_ldstmia),
cCE("fstmeas", c800a00, 2, (RRnpctw, VRSLST), vfp_sp_ldstmia),
- cCE("fstmdbs", d200a00, 2, (RRnpctw, VRSLST), vfp_sp_ldstmdb),
cCE("fstmfds", d200a00, 2, (RRnpctw, VRSLST), vfp_sp_ldstmdb),
cCE("fstmiax", c800b00, 2, (RRnpctw, VRDLST), vfp_xp_ldstmia),
cCE("fstmeax", c800b00, 2, (RRnpctw, VRDLST), vfp_xp_ldstmia),
/* Double precision load/store are still present on single precision
implementations. */
- cCE("fldd", d100b00, 2, (RVD, ADDRGLDC), vfp_dp_ldst),
- cCE("fstd", d000b00, 2, (RVD, ADDRGLDC), vfp_dp_ldst),
cCE("fldmiad", c900b00, 2, (RRnpctw, VRDLST), vfp_dp_ldstmia),
cCE("fldmfdd", c900b00, 2, (RRnpctw, VRDLST), vfp_dp_ldstmia),
cCE("fldmdbd", d300b00, 2, (RRnpctw, VRDLST), vfp_dp_ldstmdb),
Individual encoder functions perform additional architecture checks. */
#undef ARM_VARIANT
#define ARM_VARIANT & fpu_vfp_ext_v1xd
+#undef THUMB_VARIANT
+#define THUMB_VARIANT & arm_ext_v6t2
+
+ NCE(vldm, c900b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
+ NCE(vldmia, c900b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
+ NCE(vldmdb, d100b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
+ NCE(vstm, c800b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
+ NCE(vstmia, c800b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
+ NCE(vstmdb, d000b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
+
+ NCE(vpop, 0, 1, (VRSDLST), vfp_nsyn_pop),
+ NCE(vpush, 0, 1, (VRSDLST), vfp_nsyn_push),
+
#undef THUMB_VARIANT
#define THUMB_VARIANT & fpu_vfp_ext_v1xd
nCE(vnmul, _vnmul, 3, (RVSD, RVSD, RVSD), vfp_nsyn_nmul),
nCE(vnmla, _vnmla, 3, (RVSD, RVSD, RVSD), vfp_nsyn_nmul),
nCE(vnmls, _vnmls, 3, (RVSD, RVSD, RVSD), vfp_nsyn_nmul),
- NCE(vpush, 0, 1, (VRSDLST), vfp_nsyn_push),
- NCE(vpop, 0, 1, (VRSDLST), vfp_nsyn_pop),
NCE(vcvtz, 0, 2, (RVSD, RVSD), vfp_nsyn_cvtz),
/* Mnemonics shared by Neon and VFP. */
nCEF(vmls, _vmls, 3, (RNSDQ, oRNSDQ, RNSDQ_RNSC), neon_mac_maybe_scalar),
- NCE(vldm, c900b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
- NCE(vldmia, c900b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
- NCE(vldmdb, d100b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
- NCE(vstm, c800b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
- NCE(vstmia, c800b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
- NCE(vstmdb, d000b00, 2, (RRnpctw, VRSDLST), neon_ldm_stm),
-
mnCEF(vcvt, _vcvt, 3, (RNSDQMQ, RNSDQMQ, oI32z), neon_cvt),
nCEF(vcvtr, _vcvt, 2, (RNSDQ, RNSDQ), neon_cvtr),
MNCEF(vcvtb, eb20a40, 3, (RVSDMQ, RVSDMQ, oI32b), neon_cvtb),
NCE (vins, eb00ac0, 2, (RVS, RVS), neon_movhf),
/* New backported fma/fms instructions optional in v8.2. */
- NCE (vfmal, 810, 3, (RNDQ, RNSD, RNSD_RNSC), neon_vfmal),
- NCE (vfmsl, 810, 3, (RNDQ, RNSD, RNSD_RNSC), neon_vfmsl),
+ NUF (vfmsl, 810, 3, (RNDQ, RNSD, RNSD_RNSC), neon_vfmsl),
+ NUF (vfmal, 810, 3, (RNDQ, RNSD, RNSD_RNSC), neon_vfmal),
#undef THUMB_VARIANT
#define THUMB_VARIANT & fpu_neon_ext_v1
#define ARM_VARIANT & fpu_vfp_ext_fma
#undef THUMB_VARIANT
#define THUMB_VARIANT & fpu_vfp_ext_fma
- /* Mnemonics shared by Neon, VFP and MVE. These are included in the
+ /* Mnemonics shared by Neon, VFP, MVE and BF16. These are included in the
VFP FMA variant; NEON and VFP FMA always includes the NEON
FMA instructions. */
mnCEF(vfma, _vfma, 3, (RNSDQMQ, oRNSDQMQ, RNSDQMQR), neon_fmac),
+ TUF ("vfmat", c300850, fc300850, 3, (RNSDQMQ, oRNSDQMQ, RNSDQ_RNSC_MQ_RR), mve_vfma, mve_vfma),
mnCEF(vfms, _vfms, 3, (RNSDQMQ, oRNSDQMQ, RNSDQMQ), neon_fmac),
/* ffmas/ffmad/ffmss/ffmsd are dummy mnemonics to satisfy gas;
#define ARM_VARIANT & fpu_neon_ext_v1
mnUF(vabd, _vabd, 3, (RNDQMQ, oRNDQMQ, RNDQMQ), neon_dyadic_if_su),
mnUF(vabdl, _vabdl, 3, (RNQMQ, RNDMQ, RNDMQ), neon_dyadic_long),
- mnUF(vaddl, _vaddl, 3, (RNQMQ, RNDMQ, RNDMQR), neon_dyadic_long),
- mnUF(vsubl, _vsubl, 3, (RNQMQ, RNDMQ, RNDMQR), neon_dyadic_long),
+ mnUF(vaddl, _vaddl, 3, (RNSDQMQ, oRNSDMQ, RNSDMQR), neon_dyadic_long),
+ mnUF(vsubl, _vsubl, 3, (RNSDQMQ, oRNSDMQ, RNSDMQR), neon_dyadic_long),
mnUF(vand, _vand, 3, (RNDQMQ, oRNDQMQ, RNDQMQ_Ibig), neon_logic),
mnUF(vbic, _vbic, 3, (RNDQMQ, oRNDQMQ, RNDQMQ_Ibig), neon_logic),
mnUF(vorr, _vorr, 3, (RNDQMQ, oRNDQMQ, RNDQMQ_Ibig), neon_logic),
#define THUMB_VARIANT & arm_ext_v6t2_v8m
MNUF (vcadd, 0, 4, (RNDQMQ, RNDQMQ, RNDQMQ, EXPi), vcadd),
MNUF (vcmla, 0, 4, (RNDQMQ, RNDQMQ, RNDQMQ_RNSC, EXPi), vcmla),
+
+#undef ARM_VARIANT
+#define ARM_VARIANT &arm_ext_bf16
+#undef THUMB_VARIANT
+#define THUMB_VARIANT &arm_ext_bf16
+ TUF ("vdot", c000d00, fc000d00, 3, (RNDQ, RNDQ, RNDQ_RNSC), vdot, vdot),
+ TUF ("vmmla", c000c40, fc000c40, 3, (RNQ, RNQ, RNQ), vmmla, vmmla),
+ TUF ("vfmab", c300810, fc300810, 3, (RNDQ, RNDQ, RNDQ_RNSC), bfloat_vfma, bfloat_vfma),
+
+#undef ARM_VARIANT
+#define ARM_VARIANT &arm_ext_i8mm
+#undef THUMB_VARIANT
+#define THUMB_VARIANT &arm_ext_i8mm
+ TUF ("vsmmla", c200c40, fc200c40, 3, (RNQ, RNQ, RNQ), vsmmla, vsmmla),
+ TUF ("vummla", c200c50, fc200c50, 3, (RNQ, RNQ, RNQ), vummla, vummla),
+ TUF ("vusmmla", ca00c40, fca00c40, 3, (RNQ, RNQ, RNQ), vsmmla, vsmmla),
+ TUF ("vusdot", c800d00, fc800d00, 3, (RNDQ, RNDQ, RNDQ_RNSC), vusdot, vusdot),
+ TUF ("vsudot", c800d10, fc800d10, 3, (RNDQ, RNDQ, RNSC), vsudot, vsudot),
+
+#undef ARM_VARIANT
+#undef THUMB_VARIANT
+#define THUMB_VARIANT &arm_ext_cde
+ ToC ("cx1", ee000000, 3, (RCP, APSR_RR, I8191), cx1),
+ ToC ("cx1a", fe000000, 3, (RCP, APSR_RR, I8191), cx1a),
+ ToC ("cx1d", ee000040, 4, (RCP, RR, APSR_RR, I8191), cx1d),
+ ToC ("cx1da", fe000040, 4, (RCP, RR, APSR_RR, I8191), cx1da),
+
+ ToC ("cx2", ee400000, 4, (RCP, APSR_RR, APSR_RR, I511), cx2),
+ ToC ("cx2a", fe400000, 4, (RCP, APSR_RR, APSR_RR, I511), cx2a),
+ ToC ("cx2d", ee400040, 5, (RCP, RR, APSR_RR, APSR_RR, I511), cx2d),
+ ToC ("cx2da", fe400040, 5, (RCP, RR, APSR_RR, APSR_RR, I511), cx2da),
+
+ ToC ("cx3", ee800000, 5, (RCP, APSR_RR, APSR_RR, APSR_RR, I63), cx3),
+ ToC ("cx3a", fe800000, 5, (RCP, APSR_RR, APSR_RR, APSR_RR, I63), cx3a),
+ ToC ("cx3d", ee800040, 6, (RCP, RR, APSR_RR, APSR_RR, APSR_RR, I63), cx3d),
+ ToC ("cx3da", fe800040, 6, (RCP, RR, APSR_RR, APSR_RR, APSR_RR, I63), cx3da),
+
+ mToC ("vcx1", ec200000, 3, (RCP, RNSDMQ, I4095), vcx1),
+ mToC ("vcx1a", fc200000, 3, (RCP, RNSDMQ, I4095), vcx1),
+
+ mToC ("vcx2", ec300000, 4, (RCP, RNSDMQ, RNSDMQ, I127), vcx2),
+ mToC ("vcx2a", fc300000, 4, (RCP, RNSDMQ, RNSDMQ, I127), vcx2),
+
+ mToC ("vcx3", ec800000, 5, (RCP, RNSDMQ, RNSDMQ, RNSDMQ, I15), vcx3),
+ mToC ("vcx3a", fc800000, 5, (RCP, RNSDMQ, RNSDMQ, RNSDMQ, I15), vcx3),
};
+
#undef ARM_VARIANT
#undef THUMB_VARIANT
#undef TCE
/* PR 21809: Do not set a mapping state for debug sections
- it just confuses other tools. */
- if (bfd_get_section_flags (NULL, now_seg) & SEC_DEBUGGING)
+ if (bfd_section_flags (now_seg) & SEC_DEBUGGING)
return;
frag_thumb_mode = fragP->tc_frag_data.thumb_mode ^ MODE_RECORDED;
const char * text_name;
const char * prefix;
const char * prefix_once;
- const char * group_name;
+ struct elf_section_match match;
char * sec_name;
int type;
int flags;
flags = SHF_ALLOC;
linkonce = 0;
- group_name = 0;
+ memset (&match, 0, sizeof (match));
/* Handle COMDAT group. */
if (prefix != prefix_once && (text_seg->flags & SEC_LINK_ONCE) != 0)
{
- group_name = elf_group_name (text_seg);
- if (group_name == NULL)
+ match.group_name = elf_group_name (text_seg);
+ if (match.group_name == NULL)
{
as_bad (_("Group section `%s' has no group signature"),
segment_name (text_seg));
linkonce = 1;
}
- obj_elf_change_section (sec_name, type, 0, flags, 0, group_name,
+ obj_elf_change_section (sec_name, type, flags, 0, &match,
linkonce, 0);
/* Set the section link for index tables. */
break;
case BFD_RELOC_THUMB_PCREL_BRANCH9: /* Conditional branch. */
- if ((value & ~0xff) && ((value & ~0xff) != ~0xff))
+ if (out_of_range_p (value, 8))
as_bad_where (fixP->fx_file, fixP->fx_line, BAD_RANGE);
if (fixP->fx_done || !seg->use_rela_p)
break;
case BFD_RELOC_THUMB_PCREL_BRANCH12: /* Unconditional branch. */
- if ((value & ~0x7ff) && ((value & ~0x7ff) != ~0x7ff))
+ if (out_of_range_p (value, 11))
as_bad_where (fixP->fx_file, fixP->fx_line, BAD_RANGE);
if (fixP->fx_done || !seg->use_rela_p)
}
break;
+ /* This relocation is misnamed, it should be BRANCH21. */
case BFD_RELOC_THUMB_PCREL_BRANCH20:
if (fixP->fx_addsy
&& (S_GET_SEGMENT (fixP->fx_addsy) == seg)
/* Force a relocation for a branch 20 bits wide. */
fixP->fx_done = 0;
}
- if ((value & ~0x1fffff) && ((value & ~0x0fffff) != ~0x0fffff))
+ if (out_of_range_p (value, 20))
as_bad_where (fixP->fx_file, fixP->fx_line,
_("conditional branch out of range"));
fixP->fx_r_type = BFD_RELOC_THUMB_PCREL_BRANCH23;
#endif
- if ((value & ~0x3fffff) && ((value & ~0x3fffff) != ~0x3fffff))
+ if (out_of_range_p (value, 22))
{
if (!(ARM_CPU_HAS_FEATURE (cpu_variant, arm_ext_v6t2)))
as_bad_where (fixP->fx_file, fixP->fx_line, BAD_RANGE);
- else if ((value & ~0x1ffffff)
- && ((value & ~0x1ffffff) != ~0x1ffffff))
+ else if (out_of_range_p (value, 24))
as_bad_where (fixP->fx_file, fixP->fx_line,
_("Thumb2 branch out of range"));
}
break;
case BFD_RELOC_THUMB_PCREL_BRANCH25:
- if ((value & ~0x0ffffff) && ((value & ~0x0ffffff) != ~0x0ffffff))
+ if (out_of_range_p (value, 24))
as_bad_where (fixP->fx_file, fixP->fx_line, BAD_RANGE);
if (fixP->fx_done || !seg->use_rela_p)
(((unsigned long) fixP->fx_frag->fr_address
+ (unsigned long) fixP->fx_where) & ~3)
+ (unsigned long) value);
+ else if (get_recorded_alignment (seg) < 2)
+ as_warn_where (fixP->fx_file, fixP->fx_line,
+ _("section does not have enough alignment to ensure safe PC-relative loads"));
if (value & ~0x3fc)
as_bad_where (fixP->fx_file, fixP->fx_line,
if (sec != NULL)
{
- bfd_set_section_flags
- (stdoutput, sec, SEC_READONLY | SEC_DEBUGGING /* | SEC_HAS_CONTENTS */);
- bfd_set_section_size (stdoutput, sec, 0);
+ bfd_set_section_flags (sec, SEC_READONLY | SEC_DEBUGGING);
+ bfd_set_section_size (sec, 0);
bfd_set_section_contents (stdoutput, sec, NULL, 0, 0);
}
}
{"mwarn-deprecated", NULL, &warn_on_deprecated, 1, NULL},
{"mno-warn-deprecated", N_("do not warn on use of deprecated feature"),
&warn_on_deprecated, 0, NULL},
+
+ {"mwarn-restrict-it", N_("warn about performance deprecated IT instructions"
+ " in ARMv8-A and ARMv8-R"), &warn_on_restrict_it, 1, NULL},
+ {"mno-warn-restrict-it", NULL, &warn_on_restrict_it, 0, NULL},
+
{"mwarn-syms", N_("warn about symbols that match instruction names [default]"), (int *) (& flag_warn_syms), TRUE, NULL},
{"mno-warn-syms", N_("disable warnings about symobls that match instructions"), (int *) (& flag_warn_syms), FALSE, NULL},
{NULL, NULL, NULL, 0, NULL}
ARM_ARCH_NONE,
FPU_ARCH_NEON_VFP_V4),
ARM_CPU_OPT ("cortex-a32", "Cortex-A32", ARM_ARCH_V8A,
- ARM_FEATURE_COPROC (CRC_EXT_ARMV8),
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_CRC),
FPU_ARCH_CRYPTO_NEON_VFP_ARMV8),
ARM_CPU_OPT ("cortex-a35", "Cortex-A35", ARM_ARCH_V8A,
- ARM_FEATURE_COPROC (CRC_EXT_ARMV8),
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_CRC),
FPU_ARCH_CRYPTO_NEON_VFP_ARMV8),
ARM_CPU_OPT ("cortex-a53", "Cortex-A53", ARM_ARCH_V8A,
- ARM_FEATURE_COPROC (CRC_EXT_ARMV8),
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_CRC),
FPU_ARCH_CRYPTO_NEON_VFP_ARMV8),
ARM_CPU_OPT ("cortex-a55", "Cortex-A55", ARM_ARCH_V8_2A,
ARM_FEATURE_CORE_HIGH (ARM_EXT2_FP16_INST),
FPU_ARCH_CRYPTO_NEON_VFP_ARMV8_DOTPROD),
ARM_CPU_OPT ("cortex-a57", "Cortex-A57", ARM_ARCH_V8A,
- ARM_FEATURE_COPROC (CRC_EXT_ARMV8),
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_CRC),
FPU_ARCH_CRYPTO_NEON_VFP_ARMV8),
ARM_CPU_OPT ("cortex-a72", "Cortex-A72", ARM_ARCH_V8A,
- ARM_FEATURE_COPROC (CRC_EXT_ARMV8),
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_CRC),
FPU_ARCH_CRYPTO_NEON_VFP_ARMV8),
ARM_CPU_OPT ("cortex-a73", "Cortex-A73", ARM_ARCH_V8A,
- ARM_FEATURE_COPROC (CRC_EXT_ARMV8),
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_CRC),
FPU_ARCH_CRYPTO_NEON_VFP_ARMV8),
ARM_CPU_OPT ("cortex-a75", "Cortex-A75", ARM_ARCH_V8_2A,
ARM_FEATURE_CORE_HIGH (ARM_EXT2_FP16_INST),
ARM_FEATURE_CORE_LOW (ARM_EXT_ADIV),
FPU_ARCH_VFP_V3D16),
ARM_CPU_OPT ("cortex-r52", "Cortex-R52", ARM_ARCH_V8R,
- ARM_FEATURE_COPROC (CRC_EXT_ARMV8),
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_CRC),
FPU_ARCH_NEON_VFP_ARMV8),
ARM_CPU_OPT ("cortex-m35p", "Cortex-M35P", ARM_ARCH_V8M_MAIN,
ARM_FEATURE_CORE_LOW (ARM_EXT_V5ExP | ARM_EXT_V6_DSP),
ARM_ARCH_NONE,
FPU_NONE),
ARM_CPU_OPT ("exynos-m1", "Samsung Exynos M1", ARM_ARCH_V8A,
- ARM_FEATURE_COPROC (CRC_EXT_ARMV8),
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_CRC),
FPU_ARCH_CRYPTO_NEON_VFP_ARMV8),
ARM_CPU_OPT ("neoverse-n1", "Neoverse N1", ARM_ARCH_V8_2A,
ARM_FEATURE_CORE_HIGH (ARM_EXT2_FP16_INST),
ARM_ARCH_NONE,
FPU_ARCH_CRYPTO_NEON_VFP_ARMV8),
ARM_CPU_OPT ("xgene2", "APM X-Gene 2", ARM_ARCH_V8A,
- ARM_FEATURE_COPROC (CRC_EXT_ARMV8),
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_CRC),
FPU_ARCH_CRYPTO_NEON_VFP_ARMV8),
{ NULL, 0, ARM_ARCH_NONE, ARM_ARCH_NONE, ARM_ARCH_NONE, NULL }
static const struct arm_ext_table armv8a_ext_table[] =
{
- ARM_ADD ("crc", ARCH_CRC_ARMV8),
+ ARM_ADD ("crc", ARM_FEATURE_CORE_HIGH (ARM_EXT2_CRC)),
ARM_ADD ("simd", FPU_ARCH_NEON_VFP_ARMV8),
ARM_EXT ("crypto", FPU_ARCH_CRYPTO_NEON_VFP_ARMV8,
ARM_FEATURE_COPROC (FPU_CRYPTO_ARMV8)),
ARM_ADD ("simd", FPU_ARCH_NEON_VFP_ARMV8_1),
ARM_ADD ("fp16", FPU_ARCH_NEON_VFP_ARMV8_2_FP16),
ARM_ADD ("fp16fml", FPU_ARCH_NEON_VFP_ARMV8_2_FP16FML),
+ ARM_ADD ("bf16", ARM_FEATURE_CORE_HIGH (ARM_EXT2_BF16)),
+ ARM_ADD ("i8mm", ARM_FEATURE_CORE_HIGH (ARM_EXT2_I8MM)),
ARM_EXT ("crypto", FPU_ARCH_CRYPTO_NEON_VFP_ARMV8_1,
ARM_FEATURE_COPROC (FPU_CRYPTO_ARMV8)),
ARM_ADD ("dotprod", FPU_ARCH_DOTPROD_NEON_VFP_ARMV8),
{
ARM_ADD ("simd", FPU_ARCH_DOTPROD_NEON_VFP_ARMV8),
ARM_ADD ("fp16", FPU_ARCH_NEON_VFP_ARMV8_4_FP16FML),
+ ARM_ADD ("bf16", ARM_FEATURE_CORE_HIGH (ARM_EXT2_BF16)),
+ ARM_ADD ("i8mm", ARM_FEATURE_CORE_HIGH (ARM_EXT2_I8MM)),
ARM_EXT ("crypto", FPU_ARCH_CRYPTO_NEON_VFP_ARMV8_4,
ARM_FEATURE_COPROC (FPU_CRYPTO_ARMV8)),
{
ARM_ADD ("simd", FPU_ARCH_DOTPROD_NEON_VFP_ARMV8),
ARM_ADD ("fp16", FPU_ARCH_NEON_VFP_ARMV8_4_FP16FML),
+ ARM_ADD ("bf16", ARM_FEATURE_CORE_HIGH (ARM_EXT2_BF16)),
+ ARM_ADD ("i8mm", ARM_FEATURE_CORE_HIGH (ARM_EXT2_I8MM)),
ARM_EXT ("crypto", FPU_ARCH_CRYPTO_NEON_VFP_ARMV8_4,
ARM_FEATURE_COPROC (FPU_CRYPTO_ARMV8)),
{ NULL, 0, ARM_ARCH_NONE, ARM_ARCH_NONE }
};
+static const struct arm_ext_table armv86a_ext_table[] =
+{
+ ARM_ADD ("i8mm", ARM_FEATURE_CORE_HIGH (ARM_EXT2_I8MM)),
+ { NULL, 0, ARM_ARCH_NONE, ARM_ARCH_NONE }
+};
+
+#define CDE_EXTENSIONS \
+ ARM_ADD ("cdecp0", ARM_FEATURE_CORE_HIGH (ARM_EXT2_CDE | ARM_EXT2_CDE0)), \
+ ARM_ADD ("cdecp1", ARM_FEATURE_CORE_HIGH (ARM_EXT2_CDE | ARM_EXT2_CDE1)), \
+ ARM_ADD ("cdecp2", ARM_FEATURE_CORE_HIGH (ARM_EXT2_CDE | ARM_EXT2_CDE2)), \
+ ARM_ADD ("cdecp3", ARM_FEATURE_CORE_HIGH (ARM_EXT2_CDE | ARM_EXT2_CDE3)), \
+ ARM_ADD ("cdecp4", ARM_FEATURE_CORE_HIGH (ARM_EXT2_CDE | ARM_EXT2_CDE4)), \
+ ARM_ADD ("cdecp5", ARM_FEATURE_CORE_HIGH (ARM_EXT2_CDE | ARM_EXT2_CDE5)), \
+ ARM_ADD ("cdecp6", ARM_FEATURE_CORE_HIGH (ARM_EXT2_CDE | ARM_EXT2_CDE6)), \
+ ARM_ADD ("cdecp7", ARM_FEATURE_CORE_HIGH (ARM_EXT2_CDE | ARM_EXT2_CDE7))
+
static const struct arm_ext_table armv8m_main_ext_table[] =
{
- ARM_EXT ("dsp", ARM_FEATURE_CORE_LOW (ARM_EXT_V5ExP | ARM_EXT_V6_DSP),
- ARM_FEATURE_CORE_LOW (ARM_EXT_V5ExP | ARM_EXT_V6_DSP)),
+ ARM_EXT ("dsp", ARM_FEATURE_CORE_LOW (ARM_AEXT_V8M_MAIN_DSP),
+ ARM_FEATURE_CORE_LOW (ARM_AEXT_V8M_MAIN_DSP)),
ARM_EXT ("fp", FPU_ARCH_VFP_V5_SP_D16, ALL_FP),
ARM_ADD ("fp.dp", FPU_ARCH_VFP_V5D16),
+ CDE_EXTENSIONS,
{ NULL, 0, ARM_ARCH_NONE, ARM_ARCH_NONE }
};
+
static const struct arm_ext_table armv8_1m_main_ext_table[] =
{
- ARM_EXT ("dsp", ARM_FEATURE_CORE_LOW (ARM_EXT_V5ExP | ARM_EXT_V6_DSP),
- ARM_FEATURE_CORE_LOW (ARM_EXT_V5ExP | ARM_EXT_V6_DSP)),
+ ARM_EXT ("dsp", ARM_FEATURE_CORE_LOW (ARM_AEXT_V8M_MAIN_DSP),
+ ARM_FEATURE_CORE_LOW (ARM_AEXT_V8M_MAIN_DSP)),
ARM_EXT ("fp",
ARM_FEATURE (0, ARM_EXT2_FP16_INST,
FPU_VFP_V5_SP_D16 | FPU_VFP_EXT_FP16 | FPU_VFP_EXT_FMA),
ARM_ADD ("fp.dp",
ARM_FEATURE (0, ARM_EXT2_FP16_INST,
FPU_VFP_V5D16 | FPU_VFP_EXT_FP16 | FPU_VFP_EXT_FMA)),
- ARM_EXT ("mve", ARM_FEATURE_COPROC (FPU_MVE),
- ARM_FEATURE_COPROC (FPU_MVE | FPU_MVE_FP)),
+ ARM_EXT ("mve", ARM_FEATURE (ARM_AEXT_V8M_MAIN_DSP, ARM_EXT2_MVE, 0),
+ ARM_FEATURE_CORE_HIGH (ARM_EXT2_MVE | ARM_EXT2_MVE_FP)),
ARM_ADD ("mve.fp",
- ARM_FEATURE (0, ARM_EXT2_FP16_INST,
- FPU_MVE | FPU_MVE_FP | FPU_VFP_V5_SP_D16 |
- FPU_VFP_EXT_FP16 | FPU_VFP_EXT_FMA)),
+ ARM_FEATURE (ARM_AEXT_V8M_MAIN_DSP,
+ ARM_EXT2_FP16_INST | ARM_EXT2_MVE | ARM_EXT2_MVE_FP,
+ FPU_VFP_V5_SP_D16 | FPU_VFP_EXT_FP16 | FPU_VFP_EXT_FMA)),
+ CDE_EXTENSIONS,
{ NULL, 0, ARM_ARCH_NONE, ARM_ARCH_NONE }
};
+#undef CDE_EXTENSIONS
+
static const struct arm_ext_table armv8r_ext_table[] =
{
- ARM_ADD ("crc", ARCH_CRC_ARMV8),
+ ARM_ADD ("crc", ARM_FEATURE_CORE_HIGH (ARM_EXT2_CRC)),
ARM_ADD ("simd", FPU_ARCH_NEON_VFP_ARMV8),
ARM_EXT ("crypto", FPU_ARCH_CRYPTO_NEON_VFP_ARMV8,
ARM_FEATURE_COPROC (FPU_CRYPTO_ARMV8)),
ARM_ARCH_OPT2 ("armv8-r", ARM_ARCH_V8R, FPU_ARCH_VFP, armv8r),
ARM_ARCH_OPT2 ("armv8.4-a", ARM_ARCH_V8_4A, FPU_ARCH_VFP, armv84a),
ARM_ARCH_OPT2 ("armv8.5-a", ARM_ARCH_V8_5A, FPU_ARCH_VFP, armv85a),
+ ARM_ARCH_OPT2 ("armv8.6-a", ARM_ARCH_V8_6A, FPU_ARCH_VFP, armv86a),
ARM_ARCH_OPT ("xscale", ARM_ARCH_XSCALE, FPU_ARCH_VFP),
ARM_ARCH_OPT ("iwmmxt", ARM_ARCH_IWMMXT, FPU_ARCH_VFP),
ARM_ARCH_OPT ("iwmmxt2", ARM_ARCH_IWMMXT2, FPU_ARCH_VFP),
use the context sensitive approach using arm_ext_table's. */
static const struct arm_option_extension_value_table arm_extensions[] =
{
- ARM_EXT_OPT ("crc", ARCH_CRC_ARMV8, ARM_FEATURE_COPROC (CRC_EXT_ARMV8),
+ ARM_EXT_OPT ("crc", ARM_FEATURE_CORE_HIGH(ARM_EXT2_CRC),
+ ARM_FEATURE_CORE_HIGH(ARM_EXT2_CRC),
ARM_FEATURE_CORE_LOW (ARM_EXT_V8)),
ARM_EXT_OPT ("crypto", FPU_ARCH_CRYPTO_NEON_VFP_ARMV8,
ARM_FEATURE_COPROC (FPU_CRYPTO_ARMV8),
march_ext_opt = XNEW (arm_feature_set);
*march_ext_opt = arm_arch_none;
march_fpu_opt = &opt->default_fpu;
+ selected_ctx_ext_table = opt->ext_table;
strcpy (selected_cpu_name, opt->name);
if (ext != NULL)
{TAG_CPU_ARCH_V8, ARM_ARCH_V8_4A},
{TAG_CPU_ARCH_V8, ARM_ARCH_V8_5A},
{TAG_CPU_ARCH_V8_1M_MAIN, ARM_ARCH_V8_1M_MAIN},
- {-1, ARM_ARCH_NONE}
+ {TAG_CPU_ARCH_V8, ARM_ARCH_V8_6A},
+ {-1, ARM_ARCH_NONE}
};
/* Set an attribute if it has not already been set by the user. */
if (p_ver_ret == NULL)
return -1;
-found:
+ found:
/* Tag_CPU_arch_profile. */
- if (ARM_CPU_HAS_FEATURE (p_ver_ret->flags, arm_ext_v7a)
- || ARM_CPU_HAS_FEATURE (p_ver_ret->flags, arm_ext_v8)
- || (ARM_CPU_HAS_FEATURE (p_ver_ret->flags, arm_ext_atomics)
- && !ARM_CPU_HAS_FEATURE (p_ver_ret->flags, arm_ext_v8m_m_only)))
+ if (!ARM_CPU_HAS_FEATURE (p_ver_ret->flags, arm_ext_v8r)
+ && (ARM_CPU_HAS_FEATURE (p_ver_ret->flags, arm_ext_v7a)
+ || ARM_CPU_HAS_FEATURE (p_ver_ret->flags, arm_ext_v8)
+ || (ARM_CPU_HAS_FEATURE (p_ver_ret->flags, arm_ext_atomics)
+ && !ARM_CPU_HAS_FEATURE (p_ver_ret->flags, arm_ext_v8m_m_only))))
*profile = 'A';
- else if (ARM_CPU_HAS_FEATURE (p_ver_ret->flags, arm_ext_v7r))
+ else if (ARM_CPU_HAS_FEATURE (p_ver_ret->flags, arm_ext_v7r)
+ || ARM_CPU_HAS_FEATURE (p_ver_ret->flags, arm_ext_v8r))
*profile = 'R';
else if (ARM_CPU_HAS_FEATURE (p_ver_ret->flags, arm_ext_m))
*profile = 'M';
if (streq (opt->name, name))
{
selected_arch = opt->value;
+ selected_ctx_ext_table = opt->ext_table;
selected_ext = arm_arch_none;
selected_cpu = selected_arch;
strcpy (selected_cpu_name, opt->name);
name += 2;
}
+ /* Check the context specific extension table */
+ if (selected_ctx_ext_table)
+ {
+ const struct arm_ext_table * ext_opt;
+ for (ext_opt = selected_ctx_ext_table; ext_opt->name != NULL; ext_opt++)
+ {
+ if (streq (ext_opt->name, name))
+ {
+ if (adding_value)
+ {
+ if (ARM_FEATURE_ZERO (ext_opt->merge))
+ /* TODO: Option not supported. When we remove the
+ legacy table this case should error out. */
+ continue;
+ ARM_MERGE_FEATURE_SETS (selected_ext, selected_ext,
+ ext_opt->merge);
+ }
+ else
+ ARM_CLEAR_FEATURE (selected_ext, selected_ext, ext_opt->clear);
+
+ ARM_MERGE_FEATURE_SETS (selected_cpu, selected_arch, selected_ext);
+ ARM_MERGE_FEATURE_SETS (cpu_variant, selected_cpu, selected_fpu);
+ *input_line_pointer = saved_char;
+ demand_empty_rest_of_line ();
+ return;
+ }
+ }
+ }
+
for (opt = arm_extensions; opt->name != NULL; opt++)
if (streq (opt->name, name))
{
if (streq (opt->name, name))
{
selected_fpu = opt->value;
+ ARM_CLEAR_FEATURE (selected_cpu, selected_cpu, fpu_any);
#ifndef CPU_DEFAULT
if (no_cpu_selected ())
ARM_MERGE_FEATURE_SETS (cpu_variant, arm_arch_any, selected_fpu);
projects / deliverable / binutils-gdb.git / blobdiff