/* aarch64-opc.h -- Header file for aarch64-opc.c and aarch64-opc-2.c.
Copyright (C) 2012-2020 Free Software Foundation, Inc.
Contributed by ARM Ltd.
This file is part of the GNU opcodes library.
This library is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 3, or (at your option)
any later version.
It is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public
License for more details.
You should have received a copy of the GNU General Public License
along with this program; see the file COPYING3. If not,
see . */
#ifndef OPCODES_AARCH64_OPC_H
#define OPCODES_AARCH64_OPC_H
#include
#include "opcode/aarch64.h"
/* Instruction fields.
Keep synced with fields. */
enum aarch64_field_kind
{
FLD_NIL,
FLD_cond2,
FLD_nzcv,
FLD_defgh,
FLD_abc,
FLD_imm19,
FLD_immhi,
FLD_immlo,
FLD_size,
FLD_vldst_size,
FLD_op,
FLD_Q,
FLD_Rt,
FLD_Rd,
FLD_Rn,
FLD_Rt2,
FLD_Ra,
FLD_op2,
FLD_CRm,
FLD_CRn,
FLD_op1,
FLD_op0,
FLD_imm3,
FLD_cond,
FLD_opcode,
FLD_cmode,
FLD_asisdlso_opcode,
FLD_len,
FLD_Rm,
FLD_Rs,
FLD_option,
FLD_S,
FLD_hw,
FLD_opc,
FLD_opc1,
FLD_shift,
FLD_type,
FLD_ldst_size,
FLD_imm6,
FLD_imm6_2,
FLD_imm4,
FLD_imm4_2,
FLD_imm4_3,
FLD_imm5,
FLD_imm7,
FLD_imm8,
FLD_imm9,
FLD_imm12,
FLD_imm14,
FLD_imm16,
FLD_imm16_2,
FLD_imm26,
FLD_imms,
FLD_immr,
FLD_immb,
FLD_immh,
FLD_S_imm10,
FLD_N,
FLD_index,
FLD_index2,
FLD_sf,
FLD_lse_sz,
FLD_H,
FLD_L,
FLD_M,
FLD_b5,
FLD_b40,
FLD_scale,
FLD_SVE_M_4,
FLD_SVE_M_14,
FLD_SVE_M_16,
FLD_SVE_N,
FLD_SVE_Pd,
FLD_SVE_Pg3,
FLD_SVE_Pg4_5,
FLD_SVE_Pg4_10,
FLD_SVE_Pg4_16,
FLD_SVE_Pm,
FLD_SVE_Pn,
FLD_SVE_Pt,
FLD_SVE_Rm,
FLD_SVE_Rn,
FLD_SVE_Vd,
FLD_SVE_Vm,
FLD_SVE_Vn,
FLD_SVE_Za_5,
FLD_SVE_Za_16,
FLD_SVE_Zd,
FLD_SVE_Zm_5,
FLD_SVE_Zm_16,
FLD_SVE_Zn,
FLD_SVE_Zt,
FLD_SVE_i1,
FLD_SVE_i3h,
FLD_SVE_i3l,
FLD_SVE_i3h2,
FLD_SVE_i2h,
FLD_SVE_imm3,
FLD_SVE_imm4,
FLD_SVE_imm5,
FLD_SVE_imm5b,
FLD_SVE_imm6,
FLD_SVE_imm7,
FLD_SVE_imm8,
FLD_SVE_imm9,
FLD_SVE_immr,
FLD_SVE_imms,
FLD_SVE_msz,
FLD_SVE_pattern,
FLD_SVE_prfop,
FLD_SVE_rot1,
FLD_SVE_rot2,
FLD_SVE_rot3,
FLD_SVE_sz,
FLD_SVE_size,
FLD_SVE_sz2,
FLD_SVE_tsz,
FLD_SVE_tszh,
FLD_SVE_tszl_8,
FLD_SVE_tszl_19,
FLD_SVE_xs_14,
FLD_SVE_xs_22,
FLD_rotate1,
FLD_rotate2,
FLD_rotate3,
FLD_SM3_imm2,
FLD_sz
};
/* Field description. */
struct aarch64_field
{
int lsb;
int width;
};
typedef struct aarch64_field aarch64_field;
extern const aarch64_field fields[];
/* Operand description. */
struct aarch64_operand
{
enum aarch64_operand_class op_class;
/* Name of the operand code; used mainly for the purpose of internal
debugging. */
const char *name;
unsigned int flags;
/* The associated instruction bit-fields; no operand has more than 4
bit-fields */
enum aarch64_field_kind fields[4];
/* Brief description */
const char *desc;
};
typedef struct aarch64_operand aarch64_operand;
extern const aarch64_operand aarch64_operands[];
enum err_type
verify_constraints (const struct aarch64_inst *, const aarch64_insn, bfd_vma,
bfd_boolean, aarch64_operand_error *, aarch64_instr_sequence*);
/* Operand flags. */
#define OPD_F_HAS_INSERTER 0x00000001
#define OPD_F_HAS_EXTRACTOR 0x00000002
#define OPD_F_SEXT 0x00000004 /* Require sign-extension. */
#define OPD_F_SHIFT_BY_2 0x00000008 /* Need to left shift the field
value by 2 to get the value
of an immediate operand. */
#define OPD_F_MAYBE_SP 0x00000010 /* May potentially be SP. */
#define OPD_F_OD_MASK 0x000000e0 /* Operand-dependent data. */
#define OPD_F_OD_LSB 5
#define OPD_F_NO_ZR 0x00000100 /* ZR index not allowed. */
#define OPD_F_SHIFT_BY_4 0x00000200 /* Need to left shift the field
value by 4 to get the value
of an immediate operand. */
/* Register flags. */
#undef F_DEPRECATED
#define F_DEPRECATED (1 << 0) /* Deprecated system register. */
#undef F_ARCHEXT
#define F_ARCHEXT (1 << 1) /* Architecture dependent system register. */
#undef F_HASXT
#define F_HASXT (1 << 2) /* System instruction register
operand. */
#undef F_REG_READ
#define F_REG_READ (1 << 3) /* Register can only be used to read values
out of. */
#undef F_REG_WRITE
#define F_REG_WRITE (1 << 4) /* Register can only be written to but not
read from. */
/* HINT operand flags. */
#define HINT_OPD_F_NOPRINT (1 << 0) /* Should not be printed. */
/* Encode 7-bit HINT #imm in the lower 8 bits. Use higher bits for flags. */
#define HINT_ENCODE(flag, val) ((flag << 8) | val)
#define HINT_FLAG(val) (val >> 8)
#define HINT_VAL(val) (val & 0xff)
static inline bfd_boolean
operand_has_inserter (const aarch64_operand *operand)
{
return (operand->flags & OPD_F_HAS_INSERTER) ? TRUE : FALSE;
}
static inline bfd_boolean
operand_has_extractor (const aarch64_operand *operand)
{
return (operand->flags & OPD_F_HAS_EXTRACTOR) ? TRUE : FALSE;
}
static inline bfd_boolean
operand_need_sign_extension (const aarch64_operand *operand)
{
return (operand->flags & OPD_F_SEXT) ? TRUE : FALSE;
}
static inline bfd_boolean
operand_need_shift_by_two (const aarch64_operand *operand)
{
return (operand->flags & OPD_F_SHIFT_BY_2) ? TRUE : FALSE;
}
static inline bfd_boolean
operand_need_shift_by_four (const aarch64_operand *operand)
{
return (operand->flags & OPD_F_SHIFT_BY_4) ? TRUE : FALSE;
}
static inline bfd_boolean
operand_maybe_stack_pointer (const aarch64_operand *operand)
{
return (operand->flags & OPD_F_MAYBE_SP) ? TRUE : FALSE;
}
/* Return the value of the operand-specific data field (OPD_F_OD_MASK). */
static inline unsigned int
get_operand_specific_data (const aarch64_operand *operand)
{
return (operand->flags & OPD_F_OD_MASK) >> OPD_F_OD_LSB;
}
/* Return the width of field number N of operand *OPERAND. */
static inline unsigned
get_operand_field_width (const aarch64_operand *operand, unsigned n)
{
assert (operand->fields[n] != FLD_NIL);
return fields[operand->fields[n]].width;
}
/* Return the total width of the operand *OPERAND. */
static inline unsigned
get_operand_fields_width (const aarch64_operand *operand)
{
int i = 0;
unsigned width = 0;
while (operand->fields[i] != FLD_NIL)
width += fields[operand->fields[i++]].width;
assert (width > 0 && width < 32);
return width;
}
static inline const aarch64_operand *
get_operand_from_code (enum aarch64_opnd code)
{
return aarch64_operands + code;
}
/* Operand qualifier and operand constraint checking. */
int aarch64_match_operands_constraint (aarch64_inst *,
aarch64_operand_error *);
/* Operand qualifier related functions. */
const char* aarch64_get_qualifier_name (aarch64_opnd_qualifier_t);
unsigned char aarch64_get_qualifier_nelem (aarch64_opnd_qualifier_t);
aarch64_insn aarch64_get_qualifier_standard_value (aarch64_opnd_qualifier_t);
int aarch64_find_best_match (const aarch64_inst *,
const aarch64_opnd_qualifier_seq_t *,
int, aarch64_opnd_qualifier_t *);
static inline void
reset_operand_qualifier (aarch64_inst *inst, int idx)
{
assert (idx >=0 && idx < aarch64_num_of_operands (inst->opcode));
inst->operands[idx].qualifier = AARCH64_OPND_QLF_NIL;
}
/* Inline functions operating on instruction bit-field(s). */
/* Generate a mask that has WIDTH number of consecutive 1s. */
static inline aarch64_insn
gen_mask (int width)
{
return ((aarch64_insn) 1 << width) - 1;
}
/* LSB_REL is the relative location of the lsb in the sub field, starting from 0. */
static inline int
gen_sub_field (enum aarch64_field_kind kind, int lsb_rel, int width, aarch64_field *ret)
{
const aarch64_field *field = &fields[kind];
if (lsb_rel < 0 || width <= 0 || lsb_rel + width > field->width)
return 0;
ret->lsb = field->lsb + lsb_rel;
ret->width = width;
return 1;
}
/* Insert VALUE into FIELD of CODE. MASK can be zero or the base mask
of the opcode. */
static inline void
insert_field_2 (const aarch64_field *field, aarch64_insn *code,
aarch64_insn value, aarch64_insn mask)
{
assert (field->width < 32 && field->width >= 1 && field->lsb >= 0
&& field->lsb + field->width <= 32);
value &= gen_mask (field->width);
value <<= field->lsb;
/* In some opcodes, field can be part of the base opcode, e.g. the size
field in FADD. The following helps avoid corrupt the base opcode. */
value &= ~mask;
*code |= value;
}
/* Extract FIELD of CODE and return the value. MASK can be zero or the base
mask of the opcode. */
static inline aarch64_insn
extract_field_2 (const aarch64_field *field, aarch64_insn code,
aarch64_insn mask)
{
aarch64_insn value;
/* Clear any bit that is a part of the base opcode. */
code &= ~mask;
value = (code >> field->lsb) & gen_mask (field->width);
return value;
}
/* Insert VALUE into field KIND of CODE. MASK can be zero or the base mask
of the opcode. */
static inline void
insert_field (enum aarch64_field_kind kind, aarch64_insn *code,
aarch64_insn value, aarch64_insn mask)
{
insert_field_2 (&fields[kind], code, value, mask);
}
/* Extract field KIND of CODE and return the value. MASK can be zero or the
base mask of the opcode. */
static inline aarch64_insn
extract_field (enum aarch64_field_kind kind, aarch64_insn code,
aarch64_insn mask)
{
return extract_field_2 (&fields[kind], code, mask);
}
extern aarch64_insn
extract_fields (aarch64_insn code, aarch64_insn mask, ...);
/* Inline functions selecting operand to do the encoding/decoding for a
certain instruction bit-field. */
/* Select the operand to do the encoding/decoding of the 'sf' field.
The heuristic-based rule is that the result operand is respected more. */
static inline int
select_operand_for_sf_field_coding (const aarch64_opcode *opcode)
{
int idx = -1;
if (aarch64_get_operand_class (opcode->operands[0])
== AARCH64_OPND_CLASS_INT_REG)
/* normal case. */
idx = 0;
else if (aarch64_get_operand_class (opcode->operands[1])
== AARCH64_OPND_CLASS_INT_REG)
/* e.g. float2fix. */
idx = 1;
else
{ assert (0); abort (); }
return idx;
}
/* Select the operand to do the encoding/decoding of the 'type' field in
the floating-point instructions.
The heuristic-based rule is that the source operand is respected more. */
static inline int
select_operand_for_fptype_field_coding (const aarch64_opcode *opcode)
{
int idx;
if (aarch64_get_operand_class (opcode->operands[1])
== AARCH64_OPND_CLASS_FP_REG)
/* normal case. */
idx = 1;
else if (aarch64_get_operand_class (opcode->operands[0])
== AARCH64_OPND_CLASS_FP_REG)
/* e.g. float2fix. */
idx = 0;
else
{ assert (0); abort (); }
return idx;
}
/* Select the operand to do the encoding/decoding of the 'size' field in
the AdvSIMD scalar instructions.
The heuristic-based rule is that the destination operand is respected
more. */
static inline int
select_operand_for_scalar_size_field_coding (const aarch64_opcode *opcode)
{
int src_size = 0, dst_size = 0;
if (aarch64_get_operand_class (opcode->operands[0])
== AARCH64_OPND_CLASS_SISD_REG)
dst_size = aarch64_get_qualifier_esize (opcode->qualifiers_list[0][0]);
if (aarch64_get_operand_class (opcode->operands[1])
== AARCH64_OPND_CLASS_SISD_REG)
src_size = aarch64_get_qualifier_esize (opcode->qualifiers_list[0][1]);
if (src_size == dst_size && src_size == 0)
{ assert (0); abort (); }
/* When the result is not a sisd register or it is a long operantion. */
if (dst_size == 0 || dst_size == src_size << 1)
return 1;
else
return 0;
}
/* Select the operand to do the encoding/decoding of the 'size:Q' fields in
the AdvSIMD instructions. */
int aarch64_select_operand_for_sizeq_field_coding (const aarch64_opcode *);
/* Miscellaneous. */
aarch64_insn aarch64_get_operand_modifier_value (enum aarch64_modifier_kind);
enum aarch64_modifier_kind
aarch64_get_operand_modifier_from_value (aarch64_insn, bfd_boolean);
bfd_boolean aarch64_wide_constant_p (uint64_t, int, unsigned int *);
bfd_boolean aarch64_logical_immediate_p (uint64_t, int, aarch64_insn *);
int aarch64_shrink_expanded_imm8 (uint64_t);
/* Copy the content of INST->OPERANDS[SRC] to INST->OPERANDS[DST]. */
static inline void
copy_operand_info (aarch64_inst *inst, int dst, int src)
{
assert (dst >= 0 && src >= 0 && dst < AARCH64_MAX_OPND_NUM
&& src < AARCH64_MAX_OPND_NUM);
memcpy (&inst->operands[dst], &inst->operands[src],
sizeof (aarch64_opnd_info));
inst->operands[dst].idx = dst;
}
/* A primitive log caculator. */
static inline unsigned int
get_logsz (unsigned int size)
{
const unsigned char ls[16] =
{0, 1, -1, 2, -1, -1, -1, 3, -1, -1, -1, -1, -1, -1, -1, 4};
if (size > 16)
{
assert (0);
return -1;
}
assert (ls[size - 1] != (unsigned char)-1);
return ls[size - 1];
}
#endif /* OPCODES_AARCH64_OPC_H */