binutils-gdb/opcodes/cgen-asm.in

456 lines
13 KiB
Plaintext
Raw Normal View History

/* Assembler interface for targets using CGEN. -*- C -*-
CGEN: Cpu tools GENerator
THIS FILE IS MACHINE GENERATED WITH CGEN.
- the resultant file is machine generated, cgen-asm.in isn't
Copyright 1996, 1997, 1998, 1999, 2000, 2001 Free Software Foundation, Inc.
This file is part of the GNU Binutils and GDB, the GNU debugger.
This program 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 2, or (at your option)
any later version.
This program 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; if not, write to the Free Software Foundation, Inc.,
59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. */
/* ??? Eventually more and more of this stuff can go to cpu-independent files.
Keep that in mind. */
#include "sysdep.h"
#include <stdio.h>
#include "ansidecl.h"
#include "bfd.h"
#include "symcat.h"
#include "@prefix@-desc.h"
#include "@prefix@-opc.h"
#include "opintl.h"
#include "xregex.h"
#include "libiberty.h"
2001-10-13 11:38:04 +02:00
#include "safe-ctype.h"
2001-10-13 11:38:04 +02:00
#undef min
#define min(a,b) ((a) < (b) ? (a) : (b))
2001-10-13 11:38:04 +02:00
#undef max
#define max(a,b) ((a) > (b) ? (a) : (b))
static const char * parse_insn_normal
PARAMS ((CGEN_CPU_DESC, const CGEN_INSN *, const char **, CGEN_FIELDS *));
2001-10-13 11:38:04 +02:00
/* -- assembler routines inserted here. */
2001-10-13 11:38:04 +02:00
/* Regex construction routine.
2001-10-13 11:38:04 +02:00
This translates an opcode syntax string into a regex string,
by replacing any non-character syntax element (such as an
opcode) with the pattern '.*'
2001-10-13 11:38:04 +02:00
It then compiles the regex and stores it in the opcode, for
later use by @arch@_cgen_assemble_insn
2001-10-13 11:38:04 +02:00
Returns NULL for success, an error message for failure. */
char *
@arch@_cgen_build_insn_regex (insn)
CGEN_INSN *insn;
{
CGEN_OPCODE *opc = (CGEN_OPCODE *) CGEN_INSN_OPCODE (insn);
const char *mnem = CGEN_INSN_MNEMONIC (insn);
char rxbuf[CGEN_MAX_RX_ELEMENTS];
char *rx = rxbuf;
const CGEN_SYNTAX_CHAR_TYPE *syn;
int reg_err;
syn = CGEN_SYNTAX_STRING (CGEN_OPCODE_SYNTAX (opc));
/* Mnemonics come first in the syntax string. */
if (! CGEN_SYNTAX_MNEMONIC_P (* syn))
return _("missing mnemonic in syntax string");
++syn;
/* Generate a case sensitive regular expression that emulates case
insensitive matching in the "C" locale. We cannot generate a case
insensitive regular expression because in Turkish locales, 'i' and 'I'
are not equal modulo case conversion. */
/* Copy the literal mnemonic out of the insn. */
for (; *mnem; mnem++)
{
char c = *mnem;
if (ISALPHA (c))
{
*rx++ = '[';
*rx++ = TOLOWER (c);
*rx++ = TOUPPER (c);
*rx++ = ']';
}
else
*rx++ = c;
}
/* Copy any remaining literals from the syntax string into the rx. */
for(; * syn != 0 && rx <= rxbuf + (CGEN_MAX_RX_ELEMENTS - 7 - 4); ++syn)
{
if (CGEN_SYNTAX_CHAR_P (* syn))
{
char c = CGEN_SYNTAX_CHAR (* syn);
switch (c)
{
/* Escape any regex metacharacters in the syntax. */
case '.': case '[': case '\\':
case '*': case '^': case '$':
#ifdef CGEN_ESCAPE_EXTENDED_REGEX
case '?': case '{': case '}':
case '(': case ')': case '*':
case '|': case '+': case ']':
#endif
*rx++ = '\\';
*rx++ = c;
break;
default:
if (ISALPHA (c))
{
*rx++ = '[';
*rx++ = TOLOWER (c);
*rx++ = TOUPPER (c);
*rx++ = ']';
}
else
*rx++ = c;
break;
}
}
else
{
/* Replace non-syntax fields with globs. */
*rx++ = '.';
*rx++ = '*';
}
}
/* Trailing whitespace ok. */
* rx++ = '[';
* rx++ = ' ';
* rx++ = '\t';
* rx++ = ']';
* rx++ = '*';
/* But anchor it after that. */
* rx++ = '$';
* rx = '\0';
CGEN_INSN_RX (insn) = xmalloc (sizeof (regex_t));
reg_err = regcomp ((regex_t *) CGEN_INSN_RX (insn), rxbuf, REG_NOSUB);
if (reg_err == 0)
return NULL;
else
{
static char msg[80];
regerror (reg_err, (regex_t *) CGEN_INSN_RX (insn), msg, 80);
regfree ((regex_t *) CGEN_INSN_RX (insn));
free (CGEN_INSN_RX (insn));
(CGEN_INSN_RX (insn)) = NULL;
2001-10-13 11:38:04 +02:00
return msg;
}
}
/* Default insn parser.
The syntax string is scanned and operands are parsed and stored in FIELDS.
Relocs are queued as we go via other callbacks.
??? Note that this is currently an all-or-nothing parser. If we fail to
parse the instruction, we return 0 and the caller will start over from
the beginning. Backtracking will be necessary in parsing subexpressions,
but that can be handled there. Not handling backtracking here may get
expensive in the case of the m68k. Deal with later.
Returns NULL for success, an error message for failure. */
static const char *
parse_insn_normal (cd, insn, strp, fields)
CGEN_CPU_DESC cd;
const CGEN_INSN *insn;
const char **strp;
CGEN_FIELDS *fields;
{
/* ??? Runtime added insns not handled yet. */
const CGEN_SYNTAX *syntax = CGEN_INSN_SYNTAX (insn);
const char *str = *strp;
const char *errmsg;
const char *p;
const CGEN_SYNTAX_CHAR_TYPE * syn;
#ifdef CGEN_MNEMONIC_OPERANDS
/* FIXME: wip */
int past_opcode_p;
#endif
/* For now we assume the mnemonic is first (there are no leading operands).
We can parse it without needing to set up operand parsing.
GAS's input scrubber will ensure mnemonics are lowercase, but we may
not be called from GAS. */
p = CGEN_INSN_MNEMONIC (insn);
2001-10-13 11:38:04 +02:00
while (*p && TOLOWER (*p) == TOLOWER (*str))
++p, ++str;
if (* p)
return _("unrecognized instruction");
#ifndef CGEN_MNEMONIC_OPERANDS
2001-10-13 11:38:04 +02:00
if (* str && ! ISSPACE (* str))
return _("unrecognized instruction");
#endif
CGEN_INIT_PARSE (cd);
cgen_init_parse_operand (cd);
#ifdef CGEN_MNEMONIC_OPERANDS
past_opcode_p = 0;
#endif
/* We don't check for (*str != '\0') here because we want to parse
any trailing fake arguments in the syntax string. */
syn = CGEN_SYNTAX_STRING (syntax);
/* Mnemonics come first for now, ensure valid string. */
if (! CGEN_SYNTAX_MNEMONIC_P (* syn))
abort ();
++syn;
while (* syn != 0)
{
/* Non operand chars must match exactly. */
if (CGEN_SYNTAX_CHAR_P (* syn))
{
/* FIXME: While we allow for non-GAS callers above, we assume the
first char after the mnemonic part is a space. */
/* FIXME: We also take inappropriate advantage of the fact that
GAS's input scrubber will remove extraneous blanks. */
2001-10-13 11:38:04 +02:00
if (TOLOWER (*str) == TOLOWER (CGEN_SYNTAX_CHAR (* syn)))
{
#ifdef CGEN_MNEMONIC_OPERANDS
if (CGEN_SYNTAX_CHAR(* syn) == ' ')
past_opcode_p = 1;
#endif
++ syn;
++ str;
}
else if (*str)
{
/* Syntax char didn't match. Can't be this insn. */
static char msg [80];
/* xgettext:c-format */
sprintf (msg, _("syntax error (expected char `%c', found `%c')"),
CGEN_SYNTAX_CHAR(*syn), *str);
return msg;
}
else
{
/* Ran out of input. */
static char msg [80];
/* xgettext:c-format */
sprintf (msg, _("syntax error (expected char `%c', found end of instruction)"),
CGEN_SYNTAX_CHAR(*syn));
return msg;
}
continue;
}
/* We have an operand of some sort. */
errmsg = cd->parse_operand (cd, CGEN_SYNTAX_FIELD (*syn),
&str, fields);
if (errmsg)
return errmsg;
/* Done with this operand, continue with next one. */
++ syn;
}
/* If we're at the end of the syntax string, we're done. */
if (* syn == 0)
{
/* FIXME: For the moment we assume a valid `str' can only contain
blanks now. IE: We needn't try again with a longer version of
the insn and it is assumed that longer versions of insns appear
before shorter ones (eg: lsr r2,r3,1 vs lsr r2,r3). */
2001-10-13 11:38:04 +02:00
while (ISSPACE (* str))
++ str;
if (* str != '\0')
return _("junk at end of line"); /* FIXME: would like to include `str' */
return NULL;
}
/* We couldn't parse it. */
return _("unrecognized instruction");
}
/* Main entry point.
This routine is called for each instruction to be assembled.
STR points to the insn to be assembled.
We assume all necessary tables have been initialized.
The assembled instruction, less any fixups, is stored in BUF.
Remember that if CGEN_INT_INSN_P then BUF is an int and thus the value
still needs to be converted to target byte order, otherwise BUF is an array
of bytes in target byte order.
The result is a pointer to the insn's entry in the opcode table,
or NULL if an error occured (an error message will have already been
printed).
Note that when processing (non-alias) macro-insns,
this function recurses.
??? It's possible to make this cpu-independent.
One would have to deal with a few minor things.
At this point in time doing so would be more of a curiosity than useful
[for example this file isn't _that_ big], but keeping the possibility in
mind helps keep the design clean. */
const CGEN_INSN *
@arch@_cgen_assemble_insn (cd, str, fields, buf, errmsg)
CGEN_CPU_DESC cd;
const char *str;
CGEN_FIELDS *fields;
CGEN_INSN_BYTES_PTR buf;
char **errmsg;
{
const char *start;
CGEN_INSN_LIST *ilist;
const char *parse_errmsg = NULL;
const char *insert_errmsg = NULL;
int recognized_mnemonic = 0;
/* Skip leading white space. */
2001-10-13 11:38:04 +02:00
while (ISSPACE (* str))
++ str;
/* The instructions are stored in hashed lists.
Get the first in the list. */
ilist = CGEN_ASM_LOOKUP_INSN (cd, str);
/* Keep looking until we find a match. */
start = str;
for ( ; ilist != NULL ; ilist = CGEN_ASM_NEXT_INSN (ilist))
{
const CGEN_INSN *insn = ilist->insn;
recognized_mnemonic = 1;
#ifdef CGEN_VALIDATE_INSN_SUPPORTED
/* Not usually needed as unsupported opcodes
shouldn't be in the hash lists. */
/* Is this insn supported by the selected cpu? */
if (! @arch@_cgen_insn_supported (cd, insn))
continue;
#endif
/* If the RELAXED attribute is set, this is an insn that shouldn't be
chosen immediately. Instead, it is used during assembler/linker
relaxation if possible. */
if (CGEN_INSN_ATTR_VALUE (insn, CGEN_INSN_RELAXED) != 0)
continue;
str = start;
/* Skip this insn if str doesn't look right lexically. */
if (CGEN_INSN_RX (insn) != NULL &&
regexec ((regex_t *) CGEN_INSN_RX (insn), str, 0, NULL, 0) == REG_NOMATCH)
continue;
/* Allow parse/insert handlers to obtain length of insn. */
CGEN_FIELDS_BITSIZE (fields) = CGEN_INSN_BITSIZE (insn);
parse_errmsg = CGEN_PARSE_FN (cd, insn) (cd, insn, & str, fields);
if (parse_errmsg != NULL)
continue;
/* ??? 0 is passed for `pc'. */
insert_errmsg = CGEN_INSERT_FN (cd, insn) (cd, insn, fields, buf,
(bfd_vma) 0);
if (insert_errmsg != NULL)
continue;
/* It is up to the caller to actually output the insn and any
queued relocs. */
return insn;
}
{
static char errbuf[150];
#ifdef CGEN_VERBOSE_ASSEMBLER_ERRORS
const char *tmp_errmsg;
/* If requesting verbose error messages, use insert_errmsg.
Failing that, use parse_errmsg. */
tmp_errmsg = (insert_errmsg ? insert_errmsg :
parse_errmsg ? parse_errmsg :
recognized_mnemonic ?
_("unrecognized form of instruction") :
_("unrecognized instruction"));
if (strlen (start) > 50)
/* xgettext:c-format */
sprintf (errbuf, "%s `%.50s...'", tmp_errmsg, start);
else
/* xgettext:c-format */
sprintf (errbuf, "%s `%.50s'", tmp_errmsg, start);
#else
if (strlen (start) > 50)
/* xgettext:c-format */
sprintf (errbuf, _("bad instruction `%.50s...'"), start);
else
/* xgettext:c-format */
sprintf (errbuf, _("bad instruction `%.50s'"), start);
#endif
*errmsg = errbuf;
return NULL;
}
}
#if 0 /* This calls back to GAS which we can't do without care. */
/* Record each member of OPVALS in the assembler's symbol table.
This lets GAS parse registers for us.
??? Interesting idea but not currently used. */
/* Record each member of OPVALS in the assembler's symbol table.
FIXME: Not currently used. */
void
@arch@_cgen_asm_hash_keywords (cd, opvals)
CGEN_CPU_DESC cd;
CGEN_KEYWORD *opvals;
{
CGEN_KEYWORD_SEARCH search = cgen_keyword_search_init (opvals, NULL);
const CGEN_KEYWORD_ENTRY * ke;
while ((ke = cgen_keyword_search_next (& search)) != NULL)
{
#if 0 /* Unnecessary, should be done in the search routine. */
if (! @arch@_cgen_opval_supported (ke))
continue;
#endif
cgen_asm_record_register (cd, ke->name, ke->value);
}
}
#endif /* 0 */