Projects

projects@lrde.epita.fr

1049 discussions

Erreur du système de messagerie - Courrier renvoyé

by Administrateur de courrier

Ce message n'a pas été remis pour la raison suivante : Les adresses de destination suivantes sont inconnues (vérifiez-les puis renvoyez le message) : SMTP <lipe(a)terra.es> Veuillez répondre à<postmaster(a)teleline.es> si vous pensez que ce message n'est pas valide.

20 years, 2 months

[monoburg: 1] Import monoburg.

by Clement Vasseur

Index: ChangeLog from Benoit Perrot <benoit(a)lrde.epita.fr> * monoburg.y: Rename to... * parser.y: This, and let automake automaticaly invoke bison to create `parser.c'. * Makefile.am: Let autoconf add the correct extension to the generated binary. Distribute `sample.brg' as a test. * config/Makefile.am: New. * configure.ac: New. 2004-06-24 David Waite <mass(a)akuma.org> * monoburg.c: change to C90-style comments from C99/C++-style Wed Apr 14 12:40:54 CEST 2004 Paolo Molaro <lupus(a)ximian.com> * monoburg.c, monoburg.h, monoburg.y: changed the grammar so that the same emit code can be easily associated with multiple rules. Coalesce identical emit functions to reduce code size (10 KB - 10 % - with the current unchanged x86 JIT rules). 2002-10-28 Dietmar Maurer <dietmar(a)ximian.com> * monoburg.c (emit_state): use 16bit values for registers, removed reg3 and spilled flag. 2002-10-17 Dietmar Maurer <dietmar(a)ximian.com> * monoburg.y: added missing semicolon 2002-10-11 Dietmar Maurer <dietmar(a)ximian.com> * monoburg.c (emit_tree_match): omit unnecessary compare (emit_label_func): make it possible to print operator names in error messages. 2002-10-09 Dietmar Maurer <dietmar(a)ximian.com> * monoburg.c (check_result): emit a warning instead of an error 2002-10-03 Dietmar Maurer <dietmar(a)ximian.com> * monoburg.c: added new %termprefix mode 2002-09-30 Dietmar Maurer <dietmar(a)ximian.com> * monoburg.c (main): add option to specify default costs, added experimental code to handle several input files. 2002-09-26 Dietmar Maurer <dietmar(a)ximian.com> * monoburg.c (emit_state): include additional fields to handle register allocation in dag_mode 2002-09-25 Dietmar Maurer <dietmar(a)ximian.com> * added -p and -e options. monoburg is now able to work with DAGs. 2002-04-20 Dietmar Maurer <dietmar(a)ximian.com> * monoburg.y (yylex): bug fix in number parsing 2002-04-12 Gonzalo Paniagua Javier <gonzalo(a)ximian.com> * monoburg.c: added option -s to specify the c source file for output. 2002-04-11 Gonzalo Paniagua Javier <gonzalo(a)ximian.com> * monoburg.c: added a default handler for warning messages that just output the messages to stderr instead of stdout. Mon Feb 18 14:28:10 CET 2002 Paolo Molaro <lupus(a)ximian.com> * Makefile.am: fix compatibility problem with automake 1.4. Fri Feb 15 14:20:30 CET 2002 Paolo Molaro <lupus(a)ximian.com> * Makefile.am: avoid automake for build on host stuff. Fri Feb 8 12:31:40 CET 2002 Paolo Molaro <lupus(a)ximian.com> * monoburg.c: make generated arrays const, so that they are shared. Fri Feb 1 15:14:16 CET 2002 Paolo Molaro <lupus(a)ximian.com> * Makefile.am: support cross-compilation. 2001-11-07 Miguel de Icaza <miguel(a)ximian.com> * monoburg.y: Include string.h, stdlib.h to kill warnings. * sample.brg: Include string.h to remove warnings. 2001-09-23 Dietmar Maurer <dietmar(a)ximian.com> * monoburg.c: add a macro MBALLOC_STATE to define the allocation function for MBState. Added an additional user data argument to mono_burg_label - the data can be used in the cost functions. The type can be defined with MBCOST_DATA macro. (emit_cost_func): inline cost functions 2001-09-22 Dietmar Maurer <dietmar(a)ximian.com> * monoburg.y (strndup): removed, use g_strndup instead * monoburg.c (create_term): bug fix: g_strdup strings from the parser 2001-09-21 Miguel de Icaza <miguel(a)ximian.com> * Makefile.am (EXTRA_DIST): Add man page to the distro 2001-09-21 Dietmar Maurer <dietmar(a)ximian.com> * monoburg.y (yylex): bug fix 2001-09-19 Dietmar Maurer <dietmar(a)ximian.com> * monoburg.c (emit_header): bug fix for MBCOND macro Tue Sep 18 13:15:12 CEST 2001 Paolo Molaro <lupus(a)ximian.com> * monoburg.y: fix ANSI C issue. 2001-09-14 Dietmar Maurer <dietmar(a)ximian.com> * monoburg.c (emit_prototypes): add an additional argument to the code emit function - a pointer to the code buffer Tue Sep 11 13:46:35 CEST 2001 Paolo Molaro <lupus(a)ximian.com> * Makefile.am: get it to work on platforms that insist on having a weird extension at the end of an executable name. Mon Sep 10 17:24:45 CEST 2001 Paolo Molaro <lupus(a)ximian.com> * Makefile.am: make it work for make distcheck. 2001-09-09 Nick Drochak <ndrochak(a)gol.com> * Makefile.am: change CLEANFILES line to use just '=' instead of '+=' some versions of automake complain if you try to '+=' before you '=' 2001-09-08 Dietmar Maurer <dietmar(a)ximian.com> * monoburg.c (emit_header): added some convenient macros * monoburg.y (optcfunc): allow arbitrary cost functions 2001-09-06 Dietmar Maurer <dietmar(a)ximian.com> * monoburg.c (emit_header): use macros to access the tree (like in iburg) ++ monoburg.1 (revision 0) .\" .\" monoburg manual page. .\" (C) Ximian, Inc. .\" Author: .\" Dietmar Maurer (dietmar(a)ximian.com) .\" .TH Mono "Mono 1.0" .SH NAME monoburg \- code generator generator .SH SYNOPSIS .PP .B monoburg [\-h] [\-e] [\-p] [\-c VALUE] [\-d HEADER] [\-DVALUE] [FILE...] .SH DESCRIPTION The \fImonoburg\fP program is used to generate tree pattern matchers from BURG specifications. \fImonoburg\fP accepts the following EBNF grammar. .PP .nf .RS .ft CW spec: ccode `%%' { dcl } [`%%' ccode] dcl: `%start' nonterm `%term' { identifier [`=' integer] } `%termprefix' { identifier } nonterm `:' tree [cost] [ `{' ccode `}' ] [costfunc] tree: term `(' tree `,' tree `)' term `(' tree `)' term nonterm cost: `"' string '"' integer costfunc: `cost' `{' ccode `}' .RE .fi .PP Where \fIccode\fP is arbitrary C code. \fIcost\fP and \fIcostfunc\fP are used mutually exclusive, you can't use both in one rule. \fIcost\fP must be a C expression, whereas \fIccode\fP inside \fIcostfunc\fP is the body of a C function. Here are some example rules: .PP .nf .RS .ft CW # define some terminal %term Fetch Four Mul Plus # this rule uses fixed costs addr: Plus (con, Mul (Four, reg)) 2 { printf ("Emit your code here."); } # this one computes costs inside a function reg: Fetch (addr) { printf ("Tree address is %p", tree); } cost { int c; c = 1; /* calculate the costs here */ return c; } .RE .fi .PP A simple pre-processor is included, consisting of: %ifdef, %else and %endif. %ifdef operates on definitions from the command line. .SH OPTIONS The following options are supported: .TP .I "-h" Displays usage instructions. .TP .I "-d HEADER" Writes a separate header file which contains all monoburg definitions. .TP .I "-p" Assume termainals are already defined. Its possible to omit the %term definitions in this mode if you use the %termprefix command. All symbols starting with a prefix specified in %termprefix are considered to be terminals. .TP .I "-e" Extended mode. Enables monoburg to work with DAGs. .TP .I "-c VALUE" Set the default costs to VALUE .TP .I "-Dvar" Defines the variable "var" as true. This is used with %ifdef, %else and %endif in the source files to perform conditional compilation. .PP .SH AUTHOR monoburg was written by Dietmar Maurer. It is based on the papers from Christopher W.\ Fraser, Robert R.\ Henry and Todd A.\ Proebsting: "BURG - Fast Optimal Instruction Selection and Tree Parsing" and "Engineering a Simple, Efficient Code Generator Generator". .SH SEE ALSO .BR monodis(1) .BR pedump(1) ++ Makefile.am (revision 0) ## Makefile.am -- Process this file with automake to produce Makefile.in ## SUBDIRS = config ## Include path INCLUDES = $(GLIB_CFLAGS) -I$(srcdir) ## MonoBURG binary bin_PROGRAMS = monoburg monoburg_SOURCES = \ parser.y \ monoburg.c monoburg.h monoburg_LDADD = \ $(GLIB_LIBS) ## MonoBURG manual man_MANS = monoburg.1 EXTRA_DIST = $(man_MANS) ## Check/Sample check_PROGRAMS = sample TESTS = sample sample.c: monoburg$(EXEEXT) $(srcdir)/sample.brg ./monoburg$(EXEEXT) $(srcdir)/sample.brg > sample.c sample_SOURCES = sample.brg sample_LDADD = sample.o $(GLIB_LIBS) CLEANFILES = sample.c ++ monoburg.c (revision 0) /* * monoburg.c: an iburg like code generator generator * * Author: * Dietmar Maurer (dietmar(a)ximian.com) * * (C) 2001 Ximian, Inc. */ #include <stdio.h> #include <string.h> #include <stdlib.h> #include "monoburg.h" extern void yyparse (void); static GHashTable *term_hash; static GList *term_list; static GHashTable *nonterm_hash; static GList *nonterm_list; static GList *rule_list; static GList *prefix_list; FILE *inputfd; FILE *outputfd; GHashTable *definedvars; static FILE *deffd; static FILE *cfd; static int dag_mode = 0; static int predefined_terms = 0; static int default_cost = 0; static void output (char *fmt, ...) { va_list ap; va_start(ap, fmt); vfprintf (outputfd, fmt, ap); va_end (ap); } Rule* make_rule (char *id, Tree *tree) { Rule *rule = g_new0 (Rule, 1); rule->lhs = nonterm (id); rule->tree = tree; return rule; } void rule_add (Rule *rule, char *code, char *cost, char *cfunc) { if (!cfunc && !cost) cost = g_strdup_printf ("%d", default_cost); rule_list = g_list_append (rule_list, rule); rule->cost = g_strdup (cost); rule->cfunc = g_strdup (cfunc); rule->code = g_strdup (code); if (cfunc) { if (cost) yyerror ("duplicated costs (constant costs and cost function)"); else { if (dag_mode) rule->cost = g_strdup_printf ("mono_burg_cost_%d (p, data)", g_list_length (rule_list)); else rule->cost = g_strdup_printf ("mono_burg_cost_%d (tree, data)", g_list_length (rule_list)); } } rule->lhs->rules = g_list_append (rule->lhs->rules, rule); if (rule->tree->op) rule->tree->op->rules = g_list_append (rule->tree->op->rules, rule); else rule->tree->nonterm->chain = g_list_append (rule->tree->nonterm->chain, rule); } void create_rule (char *id, Tree *tree, char *code, char *cost, char *cfunc) { Rule *rule = make_rule (id, tree); rule_add (rule, code, cost, cfunc); } Tree * create_tree (char *id, Tree *left, Tree *right) { int arity = (left != NULL) + (right != NULL); Term *term = NULL; Tree *tree = g_new0 (Tree, 1); if (term_hash) term = g_hash_table_lookup (term_hash, id); /* try if id has termprefix */ if (!term) { GList *pl; for (pl = prefix_list; pl; pl = pl->next) { char *pfx = (char *)pl->data; if (!strncmp (pfx, id, strlen (pfx))) { term = create_term (id, -1); break; } } } if (term) { if (term->arity == -1) term->arity = arity; if (term->arity != arity) yyerror ("changed arity of terminal %s from %d to %d", id, term->arity, arity); tree->op = term; tree->left = left; tree->right = right; } else { tree->nonterm = nonterm (id); } return tree; } static void check_term_num (char *key, Term *value, int num) { if (num != -1 && value->number == num) yyerror ("duplicate terminal id \"%s\"", key); } void create_term_prefix (char *id) { if (!predefined_terms) yyerror ("%termprefix is only available with -p option"); prefix_list = g_list_prepend (prefix_list, g_strdup (id)); } Term * create_term (char *id, int num) { Term *term; if (!predefined_terms && nonterm_list) yyerror ("terminal definition after nonterminal definition"); if (num < -1) yyerror ("invalid terminal number %d", num); if (!term_hash) term_hash = g_hash_table_new (g_str_hash , g_str_equal); g_hash_table_foreach (term_hash, (GHFunc) check_term_num, (gpointer) num); term = g_new0 (Term, 1); term->name = g_strdup (id); term->number = num; term->arity = -1; term_list = g_list_append (term_list, term); g_hash_table_insert (term_hash, term->name, term); return term; } NonTerm * nonterm (char *id) { NonTerm *nterm; if (!nonterm_hash) nonterm_hash = g_hash_table_new (g_str_hash , g_str_equal); if ((nterm = g_hash_table_lookup (nonterm_hash, id))) return nterm; nterm = g_new0 (NonTerm, 1); nterm->name = g_strdup (id); nonterm_list = g_list_append (nonterm_list, nterm); nterm->number = g_list_length (nonterm_list); g_hash_table_insert (nonterm_hash, nterm->name, nterm); return nterm; } void start_nonterm (char *id) { static gboolean start_def; if (start_def) yyerror ("start symbol redeclared"); start_def = TRUE; nonterm (id); } static void emit_tree_string (Tree *tree) { if (tree->op) { output ("%s", tree->op->name); if (tree->op->arity) { output ("("); emit_tree_string (tree->left); if (tree->right) { output (", "); emit_tree_string (tree->right); } output (")"); } } else output ("%s", tree->nonterm->name); } static void emit_rule_string (Rule *rule, char *fill) { output ("%s/* ", fill); output ("%s: ", rule->lhs->name); emit_tree_string (rule->tree); output (" */\n"); } static int next_term_num () { GList *l = term_list; int i = 1; while (l) { Term *t = (Term *)l->data; if (t->number == i) { l = term_list; i++; } else l = l->next; } return i; } static int term_compare_func (Term *t1, Term *t2) { return t1->number - t2->number; } static void emit_header () { GList *l; output ("#include <glib.h>\n"); output ("\n"); output ("#ifndef MBTREE_TYPE\n#error MBTREE_TYPE undefined\n#endif\n"); output ("#ifndef MBTREE_OP\n#define MBTREE_OP(t) ((t)->op)\n#endif\n"); output ("#ifndef MBTREE_LEFT\n#define MBTREE_LEFT(t) ((t)->left)\n#endif\n"); output ("#ifndef MBTREE_RIGHT\n#define MBTREE_RIGHT(t) ((t)->right)\n#endif\n"); output ("#ifndef MBTREE_STATE\n#define MBTREE_STATE(t) ((t)->state)\n#endif\n"); output ("#ifndef MBCGEN_TYPE\n#define MBCGEN_TYPE int\n#endif\n"); output ("#ifndef MBALLOC_STATE\n#define MBALLOC_STATE g_new (MBState, 1)\n#endif\n"); output ("#ifndef MBCOST_DATA\n#define MBCOST_DATA gpointer\n#endif\n"); output ("\n"); output ("#define MBMAXCOST 32768\n"); output ("\n"); output ("#define MBCOND(x) if (!(x)) return MBMAXCOST;\n"); output ("\n"); for (l = term_list; l; l = l->next) { Term *t = (Term *)l->data; if (t->number == -1) t->number = next_term_num (); } term_list = g_list_sort (term_list, (GCompareFunc)term_compare_func); for (l = term_list; l; l = l->next) { Term *t = (Term *)l->data; if (t->number == -1) t->number = next_term_num (); if (predefined_terms) output ("#define MB_TERM_%s\t %s\n", t->name, t->name); else output ("#define MB_TERM_%s\t %d\n", t->name, t->number); } output ("\n"); } static void emit_nonterm () { GList *l; for (l = nonterm_list; l; l = l->next) { NonTerm *n = (NonTerm *)l->data; output ("#define MB_NTERM_%s\t%d\n", n->name, n->number); } output ("#define MB_MAX_NTERMS\t%d\n", g_list_length (nonterm_list)); output ("\n"); } static void emit_state () { GList *l; int i, j; output ("typedef struct _MBState MBState;\n"); output ("struct _MBState {\n"); output ("\tint\t\t op;\n"); if (dag_mode) { output ("\tMBTREE_TYPE\t *tree;\n"); output ("\tgint32 reg1, reg2;\n"); } output ("\tMBState\t\t*left, *right;\n"); output ("\tguint16\t\tcost[%d];\n", g_list_length (nonterm_list) + 1); for (l = nonterm_list; l; l = l->next) { NonTerm *n = (NonTerm *)l->data; g_assert (g_list_length (n->rules) < 256); i = g_list_length (n->rules); j = 1; while (i >>= 1) j++; output ("\tunsigned int\t rule_%s:%d;\n", n->name, j); } output ("};\n\n"); } static void emit_decoders () { GList *l; GList *rl; for (l = nonterm_list; l; l = l->next) { NonTerm *n = (NonTerm *)l->data; output ("const int mono_burg_decode_%s[] = {\n", n->name); output ("\t0,\n"); for (rl = n->rules; rl; rl = rl->next) { Rule *rule = (Rule *)rl->data; output ("\t%d,\n", g_list_index (rule_list, rule) + 1); } output ("};\n\n"); } } static void emit_tree_match (char *st, Tree *t) { char *tn; int not_first = strcmp (st, "p->"); /* we can omit this check at the top level */ if (not_first) { if (predefined_terms) output ("\t\t\t%sop == %s /* %s */", st, t->op->name, t->op->name); else output ("\t\t\t%sop == %d /* %s */", st, t->op->number, t->op->name); } if (t->left && t->left->op) { tn = g_strconcat (st, "left->", NULL); if (not_first) output (" &&\n"); not_first = 1; emit_tree_match (tn, t->left); g_free (tn); } if (t->right && t->right->op) { tn = g_strconcat (st, "right->", NULL); if (not_first) output (" &&\n"); emit_tree_match (tn, t->right); g_free (tn); } } static void emit_rule_match (Rule *rule) { Tree *t = rule->tree; if ((t->left && t->left->op) || (t->right && t->right->op)) { output ("\t\tif (\n"); emit_tree_match ("p->", t); output ("\n\t\t)\n"); } } static void emit_costs (char *st, Tree *t) { char *tn; if (t->op) { if (t->left) { tn = g_strconcat (st, "left->", NULL); emit_costs (tn, t->left); g_free (tn); } if (t->right) { tn = g_strconcat (st, "right->", NULL); emit_costs (tn, t->right); } } else output ("%scost[MB_NTERM_%s] + ", st, t->nonterm->name); } static void emit_cond_assign (Rule *rule, char *cost, char *fill) { char *rc; if (cost) rc = g_strconcat ("c + ", cost, NULL); else rc = g_strdup ("c"); output ("%sif (%s < p->cost[MB_NTERM_%s]) {\n", fill, rc, rule->lhs->name); output ("%s\tp->cost[MB_NTERM_%s] = %s;\n", fill, rule->lhs->name, rc); output ("%s\tp->rule_%s = %d;\n", fill, rule->lhs->name, g_list_index (rule->lhs->rules, rule) + 1); if (rule->lhs->chain) output ("%s\tclosure_%s (p, %s);\n", fill, rule->lhs->name, rc); output ("%s}\n", fill); g_free (rc); } static void emit_label_func () { GList *l; int i; if (dag_mode) { output ("static void\n"); output ("mono_burg_label_priv (MBTREE_TYPE *tree, MBCOST_DATA *data, MBState *p) {\n"); } else { output ("static MBState *\n"); output ("mono_burg_label_priv (MBTREE_TYPE *tree, MBCOST_DATA *data) {\n"); } output ("\tint arity;\n"); output ("\tint c;\n"); if (!dag_mode) output ("\tMBState *p;\n"); output ("\tMBState *left = NULL, *right = NULL;\n\n"); output ("\tswitch (mono_burg_arity [MBTREE_OP(tree)]) {\n"); output ("\tcase 0:\n"); output ("\t\tbreak;\n"); output ("\tcase 1:\n"); if (dag_mode) { output ("\t\tleft = MBALLOC_STATE;\n"); output ("\t\tmono_burg_label_priv (MBTREE_LEFT(tree), data, left);\n"); } else { output ("\t\tleft = mono_burg_label_priv (MBTREE_LEFT(tree), data);\n"); output ("\t\tright = NULL;\n"); } output ("\t\tbreak;\n"); output ("\tcase 2:\n"); if (dag_mode) { output ("\t\tleft = MBALLOC_STATE;\n"); output ("\t\tmono_burg_label_priv (MBTREE_LEFT(tree), data, left);\n"); output ("\t\tright = MBALLOC_STATE;\n"); output ("\t\tmono_burg_label_priv (MBTREE_RIGHT(tree), data, right);\n"); } else { output ("\t\tleft = mono_burg_label_priv (MBTREE_LEFT(tree), data);\n"); output ("\t\tright = mono_burg_label_priv (MBTREE_RIGHT(tree), data);\n"); } output ("\t}\n\n"); output ("\tarity = (left != NULL) + (right != NULL);\n"); output ("\tg_assert (arity == mono_burg_arity [MBTREE_OP(tree)]);\n\n"); if (!dag_mode) output ("\tp = MBALLOC_STATE;\n"); output ("\tmemset (p, 0, sizeof (MBState));\n"); output ("\tp->op = MBTREE_OP(tree);\n"); output ("\tp->left = left;\n"); output ("\tp->right = right;\n"); if (dag_mode) output ("\tp->tree = tree;\n"); for (l = nonterm_list, i = 0; l; l = l->next) { output ("\tp->cost [%d] = 32767;\n", ++i); } output ("\n"); output ("\tswitch (MBTREE_OP(tree)) {\n"); for (l = term_list; l; l = l->next) { Term *t = (Term *)l->data; GList *rl; if (predefined_terms) output ("\tcase %s: /* %s */\n", t->name, t->name); else output ("\tcase %d: /* %s */\n", t->number, t->name); for (rl = t->rules; rl; rl = rl->next) { Rule *rule = (Rule *)rl->data; Tree *t = rule->tree; emit_rule_string (rule, "\t\t"); emit_rule_match (rule); output ("\t\t{\n"); output ("\t\t\tc = "); emit_costs ("", t); output ("%s;\n", rule->cost); emit_cond_assign (rule, NULL, "\t\t\t"); output ("\t\t}\n"); } output ("\t\tbreak;\n"); } output ("\tdefault:\n"); output ("#ifdef MBGET_OP_NAME\n"); output ("\t\tg_error (\"unknown operator: %%s\", MBGET_OP_NAME(MBTREE_OP(tree)));\n"); output ("#else\n"); output ("\t\tg_error (\"unknown operator: 0x%%04x\", MBTREE_OP(tree));\n"); output ("#endif\n"); output ("\t}\n\n"); if (!dag_mode) { output ("\tMBTREE_STATE(tree) = p;\n"); output ("\treturn p;\n"); } output ("}\n\n"); output ("MBState *\n"); output ("mono_burg_label (MBTREE_TYPE *tree, MBCOST_DATA *data)\n{\n"); if (dag_mode) { output ("\tMBState *p = MBALLOC_STATE;\n"); output ("\tmono_burg_label_priv (tree, data, p);\n"); } else { output ("\tMBState *p = mono_burg_label_priv (tree, data);\n"); } output ("\treturn p->rule_%s ? p : NULL;\n", ((NonTerm *)nonterm_list->data)->name); output ("}\n\n"); } static char * compute_kids (char *ts, Tree *tree, int *n) { char *res; if (tree->nonterm) { return g_strdup_printf ("\t\tkids[%d] = %s;\n", (*n)++, ts); } else if (tree->op && tree->op->arity) { char *res2 = NULL; if (dag_mode) { res = compute_kids (g_strdup_printf ("%s->left", ts), tree->left, n); if (tree->op->arity == 2) res2 = compute_kids (g_strdup_printf ("%s->right", ts), tree->right, n); } else { res = compute_kids (g_strdup_printf ("MBTREE_LEFT(%s)", ts), tree->left, n); if (tree->op->arity == 2) res2 = compute_kids (g_strdup_printf ("MBTREE_RIGHT(%s)", ts), tree->right, n); } return g_strconcat (res, res2, NULL); } return ""; } static void emit_kids () { GList *l, *nl; int i, j, c, n, *si; char **sa; output ("int\n"); output ("mono_burg_rule (MBState *state, int goal)\n{\n"); output ("\tg_return_val_if_fail (state != NULL, 0);\n"); output ("\tg_return_val_if_fail (goal > 0, 0);\n\n"); output ("\tswitch (goal) {\n"); for (nl = nonterm_list; nl; nl = nl->next) { NonTerm *n = (NonTerm *)nl->data; output ("\tcase MB_NTERM_%s:\n", n->name); output ("\t\treturn mono_burg_decode_%s [state->rule_%s];\n", n->name, n->name); } output ("\tdefault: g_assert_not_reached ();\n"); output ("\t}\n"); output ("\treturn 0;\n"); output ("}\n\n"); if (dag_mode) { output ("MBState **\n"); output ("mono_burg_kids (MBState *state, int rulenr, MBState *kids [])\n{\n"); output ("\tg_return_val_if_fail (state != NULL, NULL);\n"); output ("\tg_return_val_if_fail (kids != NULL, NULL);\n\n"); } else { output ("MBTREE_TYPE **\n"); output ("mono_burg_kids (MBTREE_TYPE *tree, int rulenr, MBTREE_TYPE *kids [])\n{\n"); output ("\tg_return_val_if_fail (tree != NULL, NULL);\n"); output ("\tg_return_val_if_fail (kids != NULL, NULL);\n\n"); } output ("\tswitch (rulenr) {\n"); n = g_list_length (rule_list); sa = g_new0 (char *, n); si = g_new0 (int, n); /* compress the case statement */ for (l = rule_list, i = 0, c = 0; l; l = l->next) { Rule *rule = (Rule *)l->data; int kn = 0; char *k; if (dag_mode) k = compute_kids ("state", rule->tree, &kn); else k = compute_kids ("tree", rule->tree, &kn); for (j = 0; j < c; j++) if (!strcmp (sa [j], k)) break; si [i++] = j; if (j == c) sa [c++] = k; } for (i = 0; i < c; i++) { for (l = rule_list, j = 0; l; l = l->next, j++) if (i == si [j]) output ("\tcase %d:\n", j + 1); output ("%s", sa [i]); output ("\t\tbreak;\n"); } output ("\tdefault:\n\t\tg_assert_not_reached ();\n"); output ("\t}\n"); output ("\treturn kids;\n"); output ("}\n\n"); } static void emit_emitter_func () { GList *l; int i, rulen; GHashTable *cache = g_hash_table_new (g_str_hash, g_str_equal); for (l = rule_list, i = 0; l; l = l->next) { Rule *rule = (Rule *)l->data; if (rule->code) { if ((rulen = GPOINTER_TO_INT (g_hash_table_lookup (cache, rule->code)))) { emit_rule_string (rule, ""); output ("#define mono_burg_emit_%d mono_burg_emit_%d\n\n", i, rulen); i++; continue; } output ("static void "); emit_rule_string (rule, ""); if (dag_mode) output ("mono_burg_emit_%d (MBState *state, MBTREE_TYPE *tree, MBCGEN_TYPE *s)\n", i); else output ("mono_burg_emit_%d (MBTREE_TYPE *tree, MBCGEN_TYPE *s)\n", i); output ("{\n"); output ("%s\n", rule->code); output ("}\n\n"); g_hash_table_insert (cache, rule->code, GINT_TO_POINTER (i)); } i++; } g_hash_table_destroy (cache); output ("MBEmitFunc const mono_burg_func [] = {\n"); output ("\tNULL,\n"); for (l = rule_list, i = 0; l; l = l->next) { Rule *rule = (Rule *)l->data; if (rule->code) output ("\tmono_burg_emit_%d,\n", i); else output ("\tNULL,\n"); i++; } output ("};\n\n"); } static void emit_cost_func () { GList *l; int i; for (l = rule_list, i = 0; l; l = l->next) { Rule *rule = (Rule *)l->data; if (rule->cfunc) { output ("inline static guint16\n"); emit_rule_string (rule, ""); if (dag_mode) output ("mono_burg_cost_%d (MBState *state, MBCOST_DATA *data)\n", i + 1); else output ("mono_burg_cost_%d (MBTREE_TYPE *tree, MBCOST_DATA *data)\n", i + 1); output ("{\n"); output ("%s\n", rule->cfunc); output ("}\n\n"); } i++; } } static void emit_closure () { GList *l, *rl; for (l = nonterm_list; l; l = l->next) { NonTerm *n = (NonTerm *)l->data; if (n->chain) output ("static void closure_%s (MBState *p, int c);\n", n->name); } output ("\n"); for (l = nonterm_list; l; l = l->next) { NonTerm *n = (NonTerm *)l->data; if (n->chain) { output ("static void\n"); output ("closure_%s (MBState *p, int c)\n{\n", n->name); for (rl = n->chain; rl; rl = rl->next) { Rule *rule = (Rule *)rl->data; emit_rule_string (rule, "\t"); emit_cond_assign (rule, rule->cost, "\t"); } output ("}\n\n"); } } } static char * compute_nonterms (Tree *tree) { if (!tree) return ""; if (tree->nonterm) { return g_strdup_printf ("MB_NTERM_%s, ", tree->nonterm->name); } else { return g_strconcat (compute_nonterms (tree->left), compute_nonterms (tree->right), NULL); } } static void emit_vardefs () { GList *l; int i, j, c, n, *si; char **sa; if (predefined_terms) { output ("guint8 mono_burg_arity [MBMAX_OPCODES];\n"); output ("void\nmono_burg_init (void)\n{\n"); for (l = term_list, i = 0; l; l = l->next) { Term *t = (Term *)l->data; output ("\tmono_burg_arity [%s] = %d; /* %s */\n", t->name, t->arity, t->name); } output ("}\n\n"); } else { output ("const guint8 mono_burg_arity [] = {\n"); for (l = term_list, i = 0; l; l = l->next) { Term *t = (Term *)l->data; while (i < t->number) { output ("\t0,\n"); i++; } output ("\t%d, /* %s */\n", t->arity, t->name); i++; } output ("};\n\n"); output ("const char *const mono_burg_term_string [] = {\n"); output ("\tNULL,\n"); for (l = term_list, i = 0; l; l = l->next) { Term *t = (Term *)l->data; output ("\t\"%s\",\n", t->name); } output ("};\n\n"); } output ("const char * const mono_burg_rule_string [] = {\n"); output ("\tNULL,\n"); for (l = rule_list, i = 0; l; l = l->next) { Rule *rule = (Rule *)l->data; output ("\t\"%s: ", rule->lhs->name); emit_tree_string (rule->tree); output ("\",\n"); } output ("};\n\n"); n = g_list_length (rule_list); sa = g_new0 (char *, n); si = g_new0 (int, n); /* compress the _nts array */ for (l = rule_list, i = 0, c = 0; l; l = l->next) { Rule *rule = (Rule *)l->data; char *s = compute_nonterms (rule->tree); for (j = 0; j < c; j++) if (!strcmp (sa [j], s)) break; si [i++] = j; if (j == c) { output ("static const guint16 mono_burg_nts_%d [] = { %s0 };\n", c, s); sa [c++] = s; } } output ("\n"); output ("const guint16 *const mono_burg_nts [] = {\n"); output ("\t0,\n"); for (l = rule_list, i = 0; l; l = l->next) { Rule *rule = (Rule *)l->data; output ("\tmono_burg_nts_%d, ", si [i++]); emit_rule_string (rule, ""); } output ("};\n\n"); } static void emit_prototypes () { if (dag_mode) output ("typedef void (*MBEmitFunc) (MBState *state, MBTREE_TYPE *tree, MBCGEN_TYPE *s);\n\n"); else output ("typedef void (*MBEmitFunc) (MBTREE_TYPE *tree, MBCGEN_TYPE *s);\n\n"); output ("extern const char * const mono_burg_term_string [];\n"); output ("extern const char * const mono_burg_rule_string [];\n"); output ("extern const guint16 *const mono_burg_nts [];\n"); output ("extern MBEmitFunc const mono_burg_func [];\n"); output ("MBState *mono_burg_label (MBTREE_TYPE *tree, MBCOST_DATA *data);\n"); output ("int mono_burg_rule (MBState *state, int goal);\n"); if (dag_mode) output ("MBState **mono_burg_kids (MBState *state, int rulenr, MBState *kids []);\n"); else output ("MBTREE_TYPE **mono_burg_kids (MBTREE_TYPE *tree, int rulenr, MBTREE_TYPE *kids []);\n"); output ("extern void mono_burg_init (void);\n"); output ("\n"); } static void check_reach (NonTerm *n); static void mark_reached (Tree *tree) { if (tree->nonterm && !tree->nonterm->reached) check_reach (tree->nonterm); if (tree->left) mark_reached (tree->left); if (tree->right) mark_reached (tree->right); } static void check_reach (NonTerm *n) { GList *l; n->reached = 1; for (l = n->rules; l; l = l->next) { Rule *rule = (Rule *)l->data; mark_reached (rule->tree); } } static void check_result () { GList *l; for (l = term_list; l; l = l->next) { Term *term = (Term *)l->data; if (term->arity == -1) g_warning ("unused terminal \"%s\"",term->name); } check_reach (((NonTerm *)nonterm_list->data)); for (l = nonterm_list; l; l = l->next) { NonTerm *n = (NonTerm *)l->data; if (!n->reached) g_warning ("unreachable nonterm \"%s\"", n->name); } } static void usage () { fprintf (stderr, "Usage is: monoburg -d file.h -s file.c [inputfile] \n"); exit (1); } static void warning_handler (const gchar *log_domain, GLogLevelFlags log_level, const gchar *message, gpointer user_data) { (void) fprintf ((FILE *) user_data, "** WARNING **: %s\n", message); } int main (int argc, char *argv []) { char *cfile = NULL; char *deffile = NULL; GList *infiles = NULL; int i; definedvars = g_hash_table_new (g_str_hash, g_str_equal); g_log_set_handler (NULL, G_LOG_LEVEL_WARNING, warning_handler, stderr); for (i = 1; i < argc; i++){ if (argv [i][0] == '-'){ if (argv [i][1] == 'h') { usage (); } else if (argv [i][1] == 'e') { dag_mode = 1; } else if (argv [i][1] == 'p') { predefined_terms = 1; } else if (argv [i][1] == 'd') { deffile = argv [++i]; } else if (argv [i][1] == 's') { cfile = argv [++i]; } else if (argv [i][1] == 'c') { default_cost = atoi (argv [++i]); } else if (argv [i][1] == 'D') { g_hash_table_insert (definedvars, &argv [i][2], GUINT_TO_POINTER (1)); } else { usage (); } } else { infiles = g_list_append (infiles, argv [i]); } } if (deffile) { if (!(deffd = fopen (deffile, "w"))) { perror ("cant open header output file"); exit (-1); } outputfd = deffd; output ("#ifndef _MONO_BURG_DEFS_\n"); output ("#define _MONO_BURG_DEFS_\n\n"); } else outputfd = stdout; if (infiles) { GList *l = infiles; while (l) { char *infile = (char *)l->data; if (!(inputfd = fopen (infile, "r"))) { perror ("cant open input file"); exit (-1); } yyparse (); reset_parser (); l->data = inputfd; l = l->next; } } else { inputfd = stdin; yyparse (); } check_result (); if (!nonterm_list) g_error ("no start symbol found"); emit_header (); emit_nonterm (); emit_state (); emit_prototypes (); if (deffd) { output ("#endif /* _MONO_BURG_DEFS_ */\n"); fclose (deffd); if (cfile == NULL) outputfd = stdout; else { if (!(cfd = fopen (cfile, "w"))) { perror ("cant open c output file"); (void) remove (deffile); exit (-1); } outputfd = cfd; } output ("#include \"%s\"\n\n", deffile); } emit_vardefs (); emit_cost_func (); emit_emitter_func (); emit_decoders (); emit_closure (); emit_label_func (); emit_kids (); if (infiles) { GList *l = infiles; while (l) { inputfd = l->data; yyparsetail (); fclose (inputfd); l = l->next; } } else { yyparsetail (); } if (cfile) fclose (cfd); return 0; } ++ sample.brg (revision 0) /* * This header (everything before the first "%%") is copied * directly to the output. */ #include <glib.h> #include <stdio.h> #include <string.h> #define MBTREE_TYPE MBTree typedef struct _MBTree MBTree; struct _MBTree { guint16 op; MBTree *left, *right; gpointer state; }; %% these are the monoburg definition # # This is the start of the definitions # comments start with a '#' as first line character # # # we must fisrt define the terminals # with or without numbers # %term Assign Constant Fetch=3 Four=8 Mul=5 Plus=6 AltFetch=7 # # optional start nonterminal # %start reg con: Constant 0 con: Four 0 addr: con 0 addr: Plus(con,reg) { int ern = mono_burg_rule (tree->state, MB_NTERM_addr); printf ("%s\n", mono_burg_rule_string [ern]); } cost { return 1; } addr: Plus(con,Mul(Four,reg)) 2 { int ern = mono_burg_rule (tree->state, MB_NTERM_addr); printf ("%s\n", mono_burg_rule_string [ern]); } reg: AltFetch(addr), reg: Fetch(addr) 1 { int ern = mono_burg_rule (tree->state, MB_NTERM_reg); printf ("%s\n", mono_burg_rule_string [ern]); } reg: Assign(addr,reg) 1 { int ern = mono_burg_rule (tree->state, MB_NTERM_reg); printf ("%s\n", mono_burg_rule_string [ern]); } %% the rest is also copied directly to the output /* everything below the second "%%" is also copied directly * to the output file. */ static MBTree * create_tree (int op, MBTree *left, MBTree *right) { MBTree *t = g_new0 (MBTree, 1); t->op = op; t->left = left; t->right = right; return t; } static void reduce (MBTree *tree, int goal) { MBTree *kids[10]; int ern = mono_burg_rule (tree->state, goal); guint16 *nts = mono_burg_nts [ern]; int i, n; mono_burg_kids (tree, ern, kids); // printf ("TEST %d %d %s %d\n", goal, ern, mono_burg_rule_string [ern], nts [0]); for (i = 0; nts [i]; i++) reduce (kids [i], nts [i]); n = (tree->left != NULL) + (tree->right != NULL); if (n) { /* not a terminal */ // printf ("XXTE %s %d\n", mono_burg_rule_string [ern], n); if (mono_burg_func [ern]) mono_burg_func [ern] (tree, NULL); else g_warning ("no code for rule %s\n", mono_burg_rule_string [ern]); } else { if (mono_burg_func [ern]) g_warning ("unused code in rule %s\n", mono_burg_rule_string [ern]); } } int main () { MBTree *t, *l, *r; MBState *s; l = create_tree (MB_TERM_Constant, NULL, NULL); r = create_tree (MB_TERM_Fetch, l, NULL); l = create_tree (MB_TERM_Constant, NULL, NULL); r = create_tree (MB_TERM_Assign, l, r); l = create_tree (MB_TERM_Four, NULL, NULL); r = create_tree (MB_TERM_Mul, l, r); l = create_tree (MB_TERM_Constant, NULL, NULL); l = create_tree (MB_TERM_Plus, l, r); t = create_tree (MB_TERM_Fetch, l, NULL); s = mono_burg_label (t, NULL); g_assert (s); reduce (t, MB_NTERM_reg); return 0; } ++ README (revision 0) This is an implementation of an IBurg like code generator generator. It is based on the papers from Christopher W. Fraser, Robert R. Henry and Todd A. Proebsting: Reference 1: BURG - Fast Optimal Instruction Selection and Tree Parsing Reference 2: Engineering a Simple, Efficient Code Generator Generator Examples: sample.brg ++ bootstrap (revision 0) #!/bin/sh autoreconf -f -v -i -m Index: configure.ac --- configure.ac (revision 0) +++ configure.ac (revision 0) @@ -0,0 +1,46 @@ +## ---------------------------------------------------------------------------- +## Configure MonoBURG, an IBurg-like code generator generator +## + +## ------------------------------------- +## Project +AC_INIT([MonoBURG], [1.0.1]) + +## ------------------------------------- +## Setup autotools + +# Need autoconf 2.57 at least. +AC_PREREQ([2.57]) + +# Auxiliary files. +AC_CONFIG_AUX_DIR([config]) +AC_CONFIG_FILES([config/Makefile]) + +# Initiate automake. +AM_INIT_AUTOMAKE([foreign 1.7.5 dist-bzip2]) + +## ------------------------------------- +## Development tools + +# Look for a C compiler +AC_PROG_CC + +# Look for a yacc-like program +AC_PROG_YACC + +## ------------------------------------- +## Environment + +# Check for GLIB-2.0 presence +AM_PATH_GLIB_2_0 + +## ------------------------------------- +## Epilogue + +# Ask for Makefile creation +AC_CONFIG_FILES([ + Makefile +]) + +# Instantiate the output files +AC_OUTPUT Index: parser.y --- parser.y (revision 0) +++ parser.y (revision 0) @@ -0,0 +1,363 @@ +%{ +/* + * monoburg.y: yacc input grammer + * + * Author: + * Dietmar Maurer (dietmar(a)ximian.com) + * + * (C) 2001 Ximian, Inc. + */ + +#include <stdio.h> +#include <string.h> +#include <stdlib.h> +#include <unistd.h> +#include <ctype.h> +#include <assert.h> +#include <stdarg.h> + +#include "monoburg.h" + +static int yylineno = 0; +static int yylinepos = 0; + +%} + +%union { + char *text; + int ivalue; + Tree *tree; + Rule *rule; + GList *rule_list; +} + +%token <text> IDENT +%token <text> CODE +%token <text> STRING +%token START +%token COST +%token TERM +%token TERMPREFIX +%token <ivalue> INTEGER + +%type <tree> tree +%type <text> optcost +%type <text> optcfunc +%type <text> optcode +%type <rule> rule +%type <rule_list> rule_list + +%% + +decls : /* empty */ + | START IDENT { start_nonterm ($2); } decls + | TERM tlist decls + | TERMPREFIX plist decls + | rule_list optcost optcode optcfunc { + GList *tmp; + for (tmp = $1; tmp; tmp = tmp->next) { + rule_add (tmp->data, $3, $2, $4); + } + g_list_free ($1); + } decls + ; + +rule : IDENT ':' tree { $$ = make_rule ($1, $3); } + ; + +rule_list : rule { $$ = g_list_append (NULL, $1); } + | rule ',' rule_list { $$ = g_list_prepend ($3, $1); } + ; + +optcode : /* empty */ { $$ = NULL; } + | CODE + ; + +plist : /* empty */ + | plist IDENT { create_term_prefix ($2);} + ; + +tlist : /* empty */ + | tlist IDENT { create_term ($2, -1);} + | tlist IDENT '=' INTEGER { create_term ($2, $4); } + ; + +tree : IDENT { $$ = create_tree ($1, NULL, NULL); } + | IDENT '(' tree ')' { $$ = create_tree ($1, $3, NULL); } + | IDENT '(' tree ',' tree ')' { $$ = create_tree ($1, $3, $5); } + ; + +optcost : /* empty */ {$$ = NULL; } + | STRING + | INTEGER { $$ = g_strdup_printf ("%d", $1); } + ; + +optcfunc : /*empty */ { $$ = NULL; } + | COST CODE { $$ = $2; } + ; +%% + +static char input[2048]; +static char *next = input; + +void +yyerror (char *fmt, ...) +{ + va_list ap; + + va_start(ap, fmt); + + fprintf (stderr, "line %d(%d): ", yylineno, yylinepos); + vfprintf (stderr, fmt, ap); + fprintf(stderr, "\n"); + + va_end (ap); + + exit (-1); +} + +static int state = 0; + +void +reset_parser () +{ + state = 0; +} + +struct pplist { + struct pplist *next; + gboolean ignore; +}; + +static struct pplist *pp = NULL; + +static char* +getvar (const char *input) +{ + char *var = g_strchug (g_strdup (input)); + char *ptr; + + for (ptr = var; *ptr && *ptr != '\n'; ++ptr) { + if (g_ascii_isspace (*ptr)) { + break; + } + } + *ptr = '\0'; + + return var; +} + +static void +push_if (char *input, gboolean flip) +{ + struct pplist *new_pp = g_new (struct pplist, 1); + char *var = getvar (input); + + new_pp->ignore = (g_hash_table_lookup (definedvars, var) == NULL) ^ flip; + new_pp->next = pp; + + new_pp->ignore |= (pp ? pp->ignore : 0); + pp = new_pp; + g_free (var); +} + +static void +flip_if () +{ + if (!pp) + yyerror ("%%else without %%if"); + + pp->ignore = !pp->ignore | (pp->next ? pp->next->ignore : 0); +} + +static void +pop_if () +{ + struct pplist *prev_pp = pp; + + if (!pp) + yyerror ("%%endif without %%if"); + + pp = pp->next; + g_free (prev_pp); +} + +static char +nextchar () +{ + int next_state ; + gboolean ll; + + if (!*next) { + next = input; + *next = 0; + do { + if (!fgets (input, sizeof (input), inputfd)) + return 0; + + ll = (input [0] == '%' && input [1] == '%'); + next_state = state; + + if (state == 1) { + if (!ll && input [0] == '%') { + if (!strncmp (&input [1], "ifdef", 5)) { + push_if (&input [6], FALSE); + ll = TRUE; + continue; + } + else if (!strncmp (&input [1], "ifndef", 6)) { + push_if (&input [7], TRUE); + ll = TRUE; + continue; + } + else if (!strncmp (&input [1], "else", 4)) { + flip_if (); + ll = TRUE; + continue; + } + else if (!strncmp (&input [1], "endif", 5)) { + pop_if (); + ll = TRUE; + continue; + } + } + if (pp && pp->ignore) { + ll = TRUE; + continue; + } + } + + switch (state) { + case 0: + if (ll) { + next_state = 1; + } else + fputs (input, outputfd); + break; + case 1: + if (ll) { + next_state = 2; + *next = 0; + } + break; + default: + return 0; + } + ll = state != 1 || input[0] == '#'; + state = next_state; + yylineno++; + } while (next_state == 2 || ll); + } + + return *next++; +} + +void +yyparsetail (void) +{ + fputs (input, outputfd); + while (fgets (input, sizeof (input), inputfd)) + fputs (input, outputfd); +} + +int +yylex (void) +{ + char c; + + do { + + if (!(c = nextchar ())) + return 0; + + yylinepos = next - input + 1; + + if (isspace (c)) + continue; + + if (c == '%') { + if (!strncmp (next, "start", 5) && isspace (next[5])) { + next += 5; + return START; + } + + if (!strncmp (next, "termprefix", 10) && isspace (next[10])) { + next += 10; + return TERMPREFIX; + } + + if (!strncmp (next, "term", 4) && isspace (next[4])) { + next += 4; + return TERM; + } + return c; + } + + if (isdigit (c)) { + int num = 0; + + do { + num = 10*num + (c - '0'); + } while (isdigit (c = (*next++))); + + yylval.ivalue = num; + return INTEGER; + } + + if (isalpha (c)) { + char *n = next; + int l; + + if (!strncmp (next - 1, "cost", 4) && isspace (next[3])) { + next += 4; + return COST; + } + + while (isalpha (*n) || isdigit (*n) || *n == '_') + n++; + + l = n - next + 1; + yylval.text = g_strndup (next - 1, l); + next = n; + return IDENT; + } + + if (c == '"') { + int i = 0; + static char buf [100000]; + + while ((c = *next++) != '"' && c) + buf [i++] = c; + + buf [i] = '\0'; + yylval.text = g_strdup (buf); + + return STRING; + } + + if (c == '{') { + int i = 0, d = 1; + static char buf [100000]; + + while (d && (c = nextchar ())) { + buf [i++] = c; + assert (i < sizeof (buf)); + + switch (c) { + case '{': d++; break; + case '}': d--; break; + default: + break; + } + } + buf [--i] = '\0'; + yylval.text = g_strdup (buf); + + return CODE; + } + + return c; + + } while (1); +} + Index: config/Makefile.am --- config/Makefile.am (revision 0) +++ config/Makefile.am (revision 0) @@ -0,0 +1,7 @@ +## Makefile.am -- Process this file with automake to produce Makefile.in +## + +MAINTAINERCLEANFILES = \ + depcomp \ + install-sh \ + missing Index: monoburg.h --- monoburg.h (revision 0) +++ monoburg.h (revision 0) @@ -0,0 +1,72 @@ +#ifndef __MONO_MONOBURG_H__ +#define __MONO_MONOBURG_H__ + +#include <glib.h> + +void yyerror (char *fmt, ...); +int yylex (void); + +extern FILE *inputfd; +extern FILE *outputfd; +extern GHashTable *definedvars; + +typedef struct _Rule Rule; + +typedef struct _Term Term; +struct _Term{ + char *name; + int number; + int arity; + GList *rules; /* rules that start with this terminal */ +}; + +typedef struct _NonTerm NonTerm; + +struct _NonTerm { + char *name; + int number; + GList *rules; /* rules with this nonterm on the left side */ + GList *chain; + gboolean reached; +}; + +typedef struct _Tree Tree; + +struct _Tree { + Term *op; + Tree *left; + Tree *right; + NonTerm *nonterm; /* used by chain rules */ +}; + +struct _Rule { + NonTerm *lhs; + Tree *tree; + char *code; + char *cost; + char *cfunc; +}; + + +Tree *create_tree (char *id, Tree *left, Tree *right); + +Term *create_term (char *id, int num); + +void create_term_prefix (char *id); + +NonTerm *nonterm (char *id); + +void start_nonterm (char *id); + +Rule *make_rule (char *id, Tree *tree); + +void rule_add (Rule *rule, char *code, char *cost, char *cfunc); + +void create_rule (char *id, Tree *tree, char *code, char *cost, + char *cfunc); + +void yyparsetail (void); + +void reset_parser (void); + +#endif

20 years, 2 months

[nolimips: 136] Use TC's bison++

by Noe

Index: ChangeLog from Benoît Perrot <benoit(a)lrde.epita.fr> Use a dedicated script to launch bison to control the directory where the generated files are output. * config/move-if-change, config/bison++: Provided by Akim Demaille. * config/Makefile.am: Distribute them. * src/parse/Makefile.am: Use bison++. Index: src/parse/Makefile.am --- src/parse/Makefile.am (revision 135) +++ src/parse/Makefile.am (working copy) @@ -30,18 +30,21 @@ @mv -f $@.tmp $@ asm-parse.yy: parse-asm-parse.stamp -bison.stamp: asm-parse.yy - @rm -rf $@.tmp - @touch $@.tmp - bison -S lalr1.cc -d -ra asm-parse.yy -o asm-parse.cc - @mv -f $@.tmp $@ + +BISONXX = $(top_srcdir)/config/bison++ +EXTRA_DIST = $(srcdir)/bison++.stamp +$(srcdir)/bison++.stamp: $(srcdir)/asm-parse.yy + @rm -rf bison++.stamp.tmp + @touch bison++.stamp.tmp + $(BISONXX) $(srcdir) asm-parse.yy asm-parse.cc -d -ra + @mv -f bison++.stamp.tmp $@ # Run bison if a file that can be created by it is missing: -asm-parse.cc $(FROM_ASM_PARSE_YY): bison.stamp +$(srcdir)/asm-parse.cc $(FROM_ASM_PARSE_YY): $(srcdir)/bison++.stamp @if test ! -f $@; then \ - @rm -f bison.stamp; \ - $(MAKE) $(AM_MAKEFLAGS) bison.stamp; \ + rm -f $(srcdir)/bison++.stamp; \ + $(MAKE) $(AM_MAKEFLAGS) $(srcdir)/bison++.stamp; \ fi MAINTAINERCLEANFILES = asm-scan.ll asm-scan.yy $(FROM_ASM_PARSE_YY) Index: config/Makefile.am --- config/Makefile.am (revision 135) +++ config/Makefile.am (working copy) @@ -1,6 +1,8 @@ ## Makefile.am -- Process this file with automake to produce Makefile.in ## +dist_noinst_SCRIPTS = move-if-change bison++ + EXTRA_DIST = $(SPECIFIC_MACROS) $(STANDARD_MACROS) $(SPECIFIC_TOOLS) MAINTAINERCLEANFILES = \ Index: config/bison++ --- config/bison++ (revision 0) +++ config/bison++ (revision 136) @@ -0,0 +1,49 @@ +#! /bin/sh + +# Any tool failure is a failure of the script. +set -e + +# bison++ SRCDIR INPUT OUTPUT OPTIONS +# ----------------------------------- + +me=$(basename $0) +auxdir=$(cd $(dirname $0) && pwd) + +srcdir=$1 +shift +input=$1 +shift +output=$1 +shift +options="-S lalr1.cc $@" + +# Alexandre Duret-Lutz also notes that in VPATH-builds $(srcdir) can +# be an absolute path depending on how ./configure is called ... +# In that case + +# bison $(srcdir)/parsetiger.yy [...] + +# will hard code the path in the maintainer's tree. Hence, do not use +# paths with Bison, chdir there. + +# A tmp dir. +tmp=$output.dir +rm -rf $tmp +mkdir $tmp + +# Compile in this dir. +# Don't use `ln -s' as its semantics of paths sucks. +cp $srcdir/$input $tmp +cd $tmp +bison $options -S lalr1.cc $input -o $output +for file in * +do + case $file in + $input) ;; + *) $auxdir/move-if-change $file ../$srcdir/$file;; + esac +done + +# Get rid of the tmp dir. +cd .. +rm -rf $tmp Index: config/move-if-change --- config/move-if-change (revision 0) +++ config/move-if-change (revision 136) @@ -0,0 +1,18 @@ +#!/bin/sh +# Like mv $1 $2, but if the files are the same, just delete $1. +# Status is 0 if $2 is changed, 1 otherwise. +if +test -r $2 +then +if +cmp -s $1 $2 +then +echo $2 is unchanged +rm -f $1 +else +mv -f $1 $2 +fi +else +mv -f $1 $2 +fi +

20 years, 2 months

[nolimips: 134,135] Include nodes.mk

by Noe

To avoid svn failure, I suggest you to remove src/inst/Makefile.am before to `svn update'. `bootstrap' must also be launch to create `nodes.mk' for the first time. Index: ChangeLog from Benoît Perrot <benoit(a)lrde.epita.fr> * dev/inst-nodes-mk-gen.py: Split into... * src/inst/Makefile.am: This file. * bootstrap: Ask for nodes.mk generation. 2004-10-03 Benoît Perrot <benoit(a)lrde.epita.fr> Index: src/inst/Makefile.am --- src/inst/Makefile.am (revision 0) +++ src/inst/Makefile.am (revision 134) @@ -0,0 +1,59 @@ +include $(srcdir)/nodes.mk + +INCLUDES = -I $(top_srcdir)/src -I $(top_builddir)/src + +BUILT_SOURCES = \ + $(FROM_INST_NODES_GEN_PY) inst-nodes-gen.stamp \ + program_builder.hh inst-builder-gen.stamp \ + program_solver.hh program_solver.cc inst-solver-gen.stamp + +# This code comes from "Handling Tools that Produce Many Outputs", +# from the Automake documentation. + +nodes.mk: $(top_srcdir)/dev/inst-nodes-mk-gen.py $(top_srcdir)/dev/nolimips.xml + $(top_srcdir)/dev/inst-nodes-mk-gen.py < $(top_srcdir)/dev/nolimips.xml >$@.tmp + mv -f $@.tmp $@ + +inst-nodes-gen.stamp: $(top_srcdir)/dev/inst-nodes-gen.py $(top_srcdir)/dev/nolimips.xml + @rm -rf $@.tmp + @touch $@.tmp + $(top_srcdir)/dev/inst-nodes-gen.py < $(top_srcdir)/dev/nolimips.xml + @mv -f $@.tmp $@ +$(FROM_INST_NODES_GEN_PY): inst-nodes-gen.stamp + +inst-builder-gen.stamp: $(top_srcdir)/dev/inst-builder-gen.py $(top_srcdir)/dev/nolimips.xml + @rm -rf $@.tmp + @touch $@.tmp + $(top_srcdir)/dev/inst-builder-gen.py < $(top_srcdir)/dev/nolimips.xml + @mv -f $@.tmp $@ +program_builder.hh: inst-builder-gen.stamp + +inst-solver-gen.stamp: $(top_srcdir)/dev/inst-solver-gen.py $(top_srcdir)/dev/nolimips.xml + @rm -rf $@.tmp + @touch $@.tmp + $(top_srcdir)/dev/inst-solver-gen.py < $(top_srcdir)/dev/nolimips.xml + @mv -f $@.tmp $@ +program_solver.hh program_solver.cc: inst-solver-gen.stamp + + +MAINTAINERCLEANFILES = \ + $(FROM_INST_NODES_GEN_PY) \ + program_builder.hh \ + program_solver.hh program_solver.cc + +noinst_LIBRARIES = libinst.a +libinst_a_SOURCES = \ + $(FROM_INST_NODES_GEN_PY) inst-nodes-gen.stamp \ + register.hh \ + exp.hh exp.cc \ + exp-visitor.hh \ + inst.hh \ + program.hh \ + label.hh label.cc \ + section.hh \ + text_section.hh text_section.cc \ + data_section.hh data_section.cc \ + program.hh \ + program_builder.hh program_builder.cc inst-builder-gen.stamp \ + program_solver.hh program_solver.cc inst-solver-gen.stamp \ + inst-tasks.hh inst-tasks.cc Index: bootstrap --- bootstrap (revision 133) +++ bootstrap (working copy) @@ -1,4 +1,4 @@ #!/bin/sh -cd src/inst && ../../dev/inst-nodes-mk-gen.py < ../../dev/nolimips.xml && cd - +cd src/inst && ../../dev/inst-nodes-mk-gen.py < ../../dev/nolimips.xml > nodes.mk && cd - autoreconf -f -v -i -m Index: dev/inst-nodes-mk-gen.py --- dev/inst-nodes-mk-gen.py (revision 133) +++ dev/inst-nodes-mk-gen.py (working copy) @@ -29,11 +29,7 @@ parser = nolimips.InstructionSetParser() instructions = parser.parse(sys.stdin) -sys.stdout = file("Makefile.am.tmp", "w") - -print """INCLUDES = -I $(top_srcdir)/src -I $(top_builddir)/src - -FROM_INST_NODES_GEN_PY = \\""" +print "FROM_INST_NODES_GEN_PY = \\" for i in range(0, len(instructions)): if instructions[i].level == "native": line = "\t" @@ -43,62 +39,3 @@ print """ decl.hh \\ all.hh \\ visitor.hh""" - -print """ -BUILT_SOURCES = \\ - $(FROM_INST_NODES_GEN_PY) inst-nodes-gen.stamp \\ - program_builder.hh inst-builder-gen.stamp \\ - program_solver.hh program_solver.cc inst-solver-gen.stamp - -# This code comes from \"Handling Tools that Produce Many Outputs\", -# from the Automake documentation. - -inst-nodes-gen.stamp: $(top_srcdir)/dev/inst-nodes-gen.py $(top_srcdir)/dev/nolimips.xml - @rm -rf $@.tmp - @touch $@.tmp - $(top_srcdir)/dev/inst-nodes-gen.py < $(top_srcdir)/dev/nolimips.xml - @mv -f $@.tmp $@ -$(FROM_INST_NODES_GEN_PY): inst-nodes-gen.stamp - -inst-builder-gen.stamp: $(top_srcdir)/dev/inst-builder-gen.py $(top_srcdir)/dev/nolimips.xml - @rm -rf $@.tmp - @touch $@.tmp - $(top_srcdir)/dev/inst-builder-gen.py < $(top_srcdir)/dev/nolimips.xml - @mv -f $@.tmp $@ -program_builder.hh: inst-builder-gen.stamp - -inst-solver-gen.stamp: $(top_srcdir)/dev/inst-solver-gen.py $(top_srcdir)/dev/nolimips.xml - @rm -rf $@.tmp - @touch $@.tmp - $(top_srcdir)/dev/inst-solver-gen.py < $(top_srcdir)/dev/nolimips.xml - @mv -f $@.tmp $@ -program_solver.hh program_solver.cc: inst-solver-gen.stamp - - -MAINTAINERCLEANFILES = \\ - $(FROM_INST_NODES_GEN_PY) \\ - program_builder.hh \\ - program_solver.hh program_solver.cc - -noinst_LIBRARIES = libinst.a -libinst_a_SOURCES = \\ - $(FROM_INST_NODES_GEN_PY) inst-nodes-gen.stamp \\ - register.hh \\ - exp.hh exp.cc \\ - exp-visitor.hh \\ - inst.hh \\ - program.hh \\ - label.hh label.cc \\ - section.hh \\ - text_section.hh text_section.cc \\ - data_section.hh data_section.cc \\ - program.hh \\ - program_builder.hh program_builder.cc inst-builder-gen.stamp \\ - program_solver.hh program_solver.cc inst-solver-gen.stamp \\ - inst-tasks.hh inst-tasks.cc""" - -## Epilogue -sys.stdout.close() -sys.stdout = sys.__stdout__ -nolimips.lazy_overwrite("Makefile.am", "Makefile.am.tmp") -os.remove("Makefile.am.tmp")

20 years, 2 months

[nolimips: 133] Rename dev/inst-makefile-gen.py

by Noe

Index: ChangeLog from Benoît Perrot <benoit(a)lrde.epita.fr> * dev/inst-makefile-gen.py: Move to... * dev/inst-nodes-mk-gen.py: This file. * Makefile.am, boostrap: Propagate. Index: Makefile.am --- Makefile.am (revision 132) +++ Makefile.am (revision 133) @@ -8,7 +8,7 @@ dev/nolimips.py \ dev/parse-asm-scan-gen.py \ dev/parse-asm-parse-gen.py \ - dev/inst-makefile-gen.py \ + dev/inst-nodes-mk-gen.py \ dev/inst-nodes-gen.py \ dev/inst-builder-gen.py \ dev/inst-solver-gen.py \ Index: bootstrap --- bootstrap (revision 132) +++ bootstrap (revision 133) @@ -1,4 +1,4 @@ #!/bin/sh -cd src/inst && ../../dev/inst-makefile-gen.py < ../../dev/nolimips.xml && cd - +cd src/inst && ../../dev/inst-nodes-mk-gen.py < ../../dev/nolimips.xml && cd - autoreconf -f -v -i -m Index: dev/inst-nodes-mk-gen.py --- dev/inst-nodes-mk-gen.py (revision 0) +++ dev/inst-nodes-mk-gen.py (revision 133) @@ -0,0 +1,104 @@ +#!/usr/bin/env python +## +## This file is part of Nolimips, a MIPS simulator with unlimited registers +## Copyright (C) 2003, 2004 Benoit Perrot <benoit(a)lrde.epita.fr> +## +## Nolimips 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 of the License, or +## (at your option) any later version. +## +## Nolimips 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 +## + +import sys, os +import string, re + +import nolimips +from nolimips import license, depth + +#### Nolimips' inst/Makefile.am generator ------------------ + +parser = nolimips.InstructionSetParser() +instructions = parser.parse(sys.stdin) + +sys.stdout = file("Makefile.am.tmp", "w") + +print """INCLUDES = -I $(top_srcdir)/src -I $(top_builddir)/src + +FROM_INST_NODES_GEN_PY = \\""" +for i in range(0, len(instructions)): + if instructions[i].level == "native": + line = "\t" + for ext in ["hh", "hxx", "cc"]: + line = line + instructions[i].opcode + "." + ext + " " + print line + "\t\t\\" +print """ decl.hh \\ + all.hh \\ + visitor.hh""" + +print """ +BUILT_SOURCES = \\ + $(FROM_INST_NODES_GEN_PY) inst-nodes-gen.stamp \\ + program_builder.hh inst-builder-gen.stamp \\ + program_solver.hh program_solver.cc inst-solver-gen.stamp + +# This code comes from \"Handling Tools that Produce Many Outputs\", +# from the Automake documentation. + +inst-nodes-gen.stamp: $(top_srcdir)/dev/inst-nodes-gen.py $(top_srcdir)/dev/nolimips.xml + @rm -rf $@.tmp + @touch $@.tmp + $(top_srcdir)/dev/inst-nodes-gen.py < $(top_srcdir)/dev/nolimips.xml + @mv -f $@.tmp $@ +$(FROM_INST_NODES_GEN_PY): inst-nodes-gen.stamp + +inst-builder-gen.stamp: $(top_srcdir)/dev/inst-builder-gen.py $(top_srcdir)/dev/nolimips.xml + @rm -rf $@.tmp + @touch $@.tmp + $(top_srcdir)/dev/inst-builder-gen.py < $(top_srcdir)/dev/nolimips.xml + @mv -f $@.tmp $@ +program_builder.hh: inst-builder-gen.stamp + +inst-solver-gen.stamp: $(top_srcdir)/dev/inst-solver-gen.py $(top_srcdir)/dev/nolimips.xml + @rm -rf $@.tmp + @touch $@.tmp + $(top_srcdir)/dev/inst-solver-gen.py < $(top_srcdir)/dev/nolimips.xml + @mv -f $@.tmp $@ +program_solver.hh program_solver.cc: inst-solver-gen.stamp + + +MAINTAINERCLEANFILES = \\ + $(FROM_INST_NODES_GEN_PY) \\ + program_builder.hh \\ + program_solver.hh program_solver.cc + +noinst_LIBRARIES = libinst.a +libinst_a_SOURCES = \\ + $(FROM_INST_NODES_GEN_PY) inst-nodes-gen.stamp \\ + register.hh \\ + exp.hh exp.cc \\ + exp-visitor.hh \\ + inst.hh \\ + program.hh \\ + label.hh label.cc \\ + section.hh \\ + text_section.hh text_section.cc \\ + data_section.hh data_section.cc \\ + program.hh \\ + program_builder.hh program_builder.cc inst-builder-gen.stamp \\ + program_solver.hh program_solver.cc inst-solver-gen.stamp \\ + inst-tasks.hh inst-tasks.cc""" + +## Epilogue +sys.stdout.close() +sys.stdout = sys.__stdout__ +nolimips.lazy_overwrite("Makefile.am", "Makefile.am.tmp") +os.remove("Makefile.am.tmp") Property changes on: dev/inst-nodes-mk-gen.py ___________________________________________________________________ Name: svn:executable + * Index: dev/inst-makefile-gen.py --- dev/inst-makefile-gen.py (revision 132) +++ dev/inst-makefile-gen.py (revision 133) @@ -1,104 +0,0 @@ -#!/usr/bin/env python -## -## This file is part of Nolimips, a MIPS simulator with unlimited registers -## Copyright (C) 2003, 2004 Benoit Perrot <benoit(a)lrde.epita.fr> -## -## Nolimips 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 of the License, or -## (at your option) any later version. -## -## Nolimips 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 -## - -import sys, os -import string, re - -import nolimips -from nolimips import license, depth - -#### Nolimips' inst/Makefile.am generator ------------------ - -parser = nolimips.InstructionSetParser() -instructions = parser.parse(sys.stdin) - -sys.stdout = file("Makefile.am.tmp", "w") - -print """INCLUDES = -I $(top_srcdir)/src -I $(top_builddir)/src - -FROM_INST_NODES_GEN_PY = \\""" -for i in range(0, len(instructions)): - if instructions[i].level == "native": - line = "\t" - for ext in ["hh", "hxx", "cc"]: - line = line + instructions[i].opcode + "." + ext + " " - print line + "\t\t\\" -print """ decl.hh \\ - all.hh \\ - visitor.hh""" - -print """ -BUILT_SOURCES = \\ - $(FROM_INST_NODES_GEN_PY) inst-nodes-gen.stamp \\ - program_builder.hh inst-builder-gen.stamp \\ - program_solver.hh program_solver.cc inst-solver-gen.stamp - -# This code comes from \"Handling Tools that Produce Many Outputs\", -# from the Automake documentation. - -inst-nodes-gen.stamp: $(top_srcdir)/dev/inst-nodes-gen.py $(top_srcdir)/dev/nolimips.xml - @rm -rf $@.tmp - @touch $@.tmp - $(top_srcdir)/dev/inst-nodes-gen.py < $(top_srcdir)/dev/nolimips.xml - @mv -f $@.tmp $@ -$(FROM_INST_NODES_GEN_PY): inst-nodes-gen.stamp - -inst-builder-gen.stamp: $(top_srcdir)/dev/inst-builder-gen.py $(top_srcdir)/dev/nolimips.xml - @rm -rf $@.tmp - @touch $@.tmp - $(top_srcdir)/dev/inst-builder-gen.py < $(top_srcdir)/dev/nolimips.xml - @mv -f $@.tmp $@ -program_builder.hh: inst-builder-gen.stamp - -inst-solver-gen.stamp: $(top_srcdir)/dev/inst-solver-gen.py $(top_srcdir)/dev/nolimips.xml - @rm -rf $@.tmp - @touch $@.tmp - $(top_srcdir)/dev/inst-solver-gen.py < $(top_srcdir)/dev/nolimips.xml - @mv -f $@.tmp $@ -program_solver.hh program_solver.cc: inst-solver-gen.stamp - - -MAINTAINERCLEANFILES = \\ - $(FROM_INST_NODES_GEN_PY) \\ - program_builder.hh \\ - program_solver.hh program_solver.cc - -noinst_LIBRARIES = libinst.a -libinst_a_SOURCES = \\ - $(FROM_INST_NODES_GEN_PY) inst-nodes-gen.stamp \\ - register.hh \\ - exp.hh exp.cc \\ - exp-visitor.hh \\ - inst.hh \\ - program.hh \\ - label.hh label.cc \\ - section.hh \\ - text_section.hh text_section.cc \\ - data_section.hh data_section.cc \\ - program.hh \\ - program_builder.hh program_builder.cc inst-builder-gen.stamp \\ - program_solver.hh program_solver.cc inst-solver-gen.stamp \\ - inst-tasks.hh inst-tasks.cc""" - -## Epilogue -sys.stdout.close() -sys.stdout = sys.__stdout__ -nolimips.lazy_overwrite("Makefile.am", "Makefile.am.tmp") -os.remove("Makefile.am.tmp")

20 years, 2 months

[nolimips: 132] Use time stamp trick for bison generated files.

by Noe

Index: ChangeLog from Benoît Perrot <benoit(a)lrde.epita.fr> * src/parse/Makefile.am: Use time stamp trick for bison generated files. 2004-09-28 Benoît Perrot <benoit(a)lrde.epita.fr> Index: src/parse/Makefile.am --- src/parse/Makefile.am (revision 131) +++ src/parse/Makefile.am (revision 132) @@ -30,8 +30,19 @@ @mv -f $@.tmp $@ asm-parse.yy: parse-asm-parse.stamp -$(FROM_ASM_PARSE_YY): $(srcdir)/asm-parse.yy - bison -S lalr1.cc -d -ra $(srcdir)/asm-parse.yy -o asm-parse.cc +bison.stamp: asm-parse.yy + @rm -rf $@.tmp + @touch $@.tmp + bison -S lalr1.cc -d -ra asm-parse.yy -o asm-parse.cc + @mv -f $@.tmp $@ + +# Run bison if a file that can be created by it is missing: + +asm-parse.cc $(FROM_ASM_PARSE_YY): bison.stamp + @if test ! -f $@; then \ + @rm -f bison.stamp; \ + $(MAKE) $(AM_MAKEFLAGS) bison.stamp; \ + fi MAINTAINERCLEANFILES = asm-scan.ll asm-scan.yy $(FROM_ASM_PARSE_YY)

20 years, 3 months

[nolimips: 131] Do not update timestamps of generators when they fail.

by Noe

I forgot to use the stamp trick on bison generated files, the very next patch will do it. Index: ChangeLog from Benoît Perrot <benoit(a)lrde.epita.fr> Do not update timestamps of generators when they fail. * src/parse/Makefile.am, dev/inst-makefile-gen.py: Suggested by Akim Demaille: Read Automake.texi :) Index: src/parse/Makefile.am --- src/parse/Makefile.am (revision 130) +++ src/parse/Makefile.am (revision 131) @@ -13,17 +13,22 @@ asm-parse.yy parse-asm-parse.stamp \ $(FROM_ASM_PARSE_YY) +# This code comes from "Handling Tools that Produce Many Outputs", +# from the Automake documentation. + parse-asm-scan.stamp: $(top_srcdir)/dev/parse-asm-scan-gen.py $(top_srcdir)/dev/nolimips.xml + @rm -rf $@.tmp + @touch $@.tmp $(top_srcdir)/dev/parse-asm-scan-gen.py < $(top_srcdir)/dev/nolimips.xml - touch $@ + @mv -f $@.tmp $@ asm-scan.ll: parse-asm-scan.stamp - touch $@ parse-asm-parse.stamp: $(top_srcdir)/dev/parse-asm-parse-gen.py $(top_srcdir)/dev/nolimips.xml + @rm -rf $@.tmp + @touch $@.tmp $(top_srcdir)/dev/parse-asm-parse-gen.py < $(top_srcdir)/dev/nolimips.xml - touch $@ + @mv -f $@.tmp $@ asm-parse.yy: parse-asm-parse.stamp - touch $@ $(FROM_ASM_PARSE_YY): $(srcdir)/asm-parse.yy bison -S lalr1.cc -d -ra $(srcdir)/asm-parse.yy -o asm-parse.cc Index: dev/inst-makefile-gen.py --- dev/inst-makefile-gen.py (revision 130) +++ dev/inst-makefile-gen.py (revision 131) @@ -50,20 +50,28 @@ program_builder.hh inst-builder-gen.stamp \\ program_solver.hh program_solver.cc inst-solver-gen.stamp +# This code comes from \"Handling Tools that Produce Many Outputs\", +# from the Automake documentation. inst-nodes-gen.stamp: $(top_srcdir)/dev/inst-nodes-gen.py $(top_srcdir)/dev/nolimips.xml + @rm -rf $@.tmp + @touch $@.tmp $(top_srcdir)/dev/inst-nodes-gen.py < $(top_srcdir)/dev/nolimips.xml - touch $@ + @mv -f $@.tmp $@ $(FROM_INST_NODES_GEN_PY): inst-nodes-gen.stamp inst-builder-gen.stamp: $(top_srcdir)/dev/inst-builder-gen.py $(top_srcdir)/dev/nolimips.xml + @rm -rf $@.tmp + @touch $@.tmp $(top_srcdir)/dev/inst-builder-gen.py < $(top_srcdir)/dev/nolimips.xml - touch $@ + @mv -f $@.tmp $@ program_builder.hh: inst-builder-gen.stamp inst-solver-gen.stamp: $(top_srcdir)/dev/inst-solver-gen.py $(top_srcdir)/dev/nolimips.xml + @rm -rf $@.tmp + @touch $@.tmp $(top_srcdir)/dev/inst-solver-gen.py < $(top_srcdir)/dev/nolimips.xml - touch $@ + @mv -f $@.tmp $@ program_solver.hh program_solver.cc: inst-solver-gen.stamp

20 years, 3 months

Nolimips: src != build

by Akim Demaille

En src != build, nolimips gère incorrectement la reconstruction du parseur :( Ça fait perdre du temps avant de comprendre ce qu'il se passe...

20 years, 3 months

[nolimips: 130] Add data storage directives.

by Noe

Index: ChangeLog from Benoît Perrot <benoit(a)lrde.epita.fr> Add data storage directives. * dev/parse-asm-scan-gen.py, dev/parse-asm-parse-gen.py: Add `.byte', .half', `.ascii' directives. * dev/inst-builder-gen.py: Implement new directives. * src/inst/data_section.hh: Move `.asciiz' implementation to program_builder.hh. 2004-09-26 Benoît Perrot <benoit(a)lrde.epita.fr> Index: src/inst/data_section.hh --- src/inst/data_section.hh (revision 129) +++ src/inst/data_section.hh (revision 130) @@ -81,14 +81,6 @@ add_byte(w / pow); } - void add_asciiz(const std::string* s) - { - for (unsigned i = 0; i < s->size (); ++i) - add_byte((*s)[i]); - add_byte(0); - } - - public: int load_byte(int offset) const { Index: dev/inst-builder-gen.py --- dev/inst-builder-gen.py (revision 129) +++ dev/inst-builder-gen.py (revision 130) @@ -73,18 +73,27 @@ } public: + void add_inst_label(Label *label) + { + program_->text_section ().add_label(label); + } + void add_data_label(Label *label) { program_->data_section ().add_label(label); } - void add_inst_label(Label *label) + + public: + void add_byte(int imm) { - program_->text_section ().add_label(label); + program_->data_section ().add_byte(imm); } - void add_space (int size) + void add_half(int imm) { - program_->data_section ().add_space(size); + // FIXME: check that imm is less than 0x10000! + add_byte((imm >> 8) && 0xff); + add_byte(imm && 0xff); } void add_word (int imm) @@ -98,9 +107,20 @@ add_word (program_->data_section ().get_offset(* label)); } + void add_ascii(const std::string *str) + { + for (unsigned i = 0; i < str->size (); ++i) + program_->data_section ().add_byte((*str)[i]); + } void add_asciiz (const std::string *str) { - program_->data_section ().add_asciiz (str); + add_ascii(str); + add_byte(0); + } + + void add_space (int size) + { + program_->data_section ().add_space(size); } public:""" Index: dev/parse-asm-scan-gen.py --- dev/parse-asm-scan-gen.py (revision 129) +++ dev/parse-asm-scan-gen.py (revision 130) @@ -95,9 +95,13 @@ \".text\" return DIR_TEXT; \".data\" return DIR_DATA; + +\".space\" return DIR_SPACE; +\".byte\" return DIR_BYTE; +\".half\" return DIR_HALF; \".word\" return DIR_WORD; +\".ascii\" return DIR_ASCII; \".asciiz\" return DIR_ASCIIZ; -\".space\" return DIR_SPACE; """ ## Body -------------------------------- Index: dev/parse-asm-parse-gen.py --- dev/parse-asm-parse-gen.py (revision 129) +++ dev/parse-asm-parse-gen.py (revision 130) @@ -92,8 +92,12 @@ %token DIR_TEXT \".text\" %token DIR_DATA \".data\" + %token DIR_SPACE \".space\" +%token DIR_BYTE \".byte\" +%token DIR_HALF \".half\" %token DIR_WORD \".word\" +%token DIR_ASCII \".ascii\" %token DIR_ASCIIZ \".asciiz\" """ @@ -120,14 +124,25 @@ // Label : LABEL_DEF { program_builder.add_data_label($1); } -// Reserve space + +// Uninitialized space | DIR_SPACE INTEGER { program_builder.add_space ($2); } + // Values +| DIR_BYTE INTEGER +{ program_builder.add_byte($2); } + +| DIR_HALF INTEGER +{ program_builder.add_half($2); } + | DIR_WORD INTEGER { program_builder.add_word ($2); } | DIR_WORD LABEL { program_builder.add_word ($2); } + +| DIR_ASCII STRING +{ program_builder.add_ascii($2); } | DIR_ASCIIZ STRING { program_builder.add_asciiz ($2); }

20 years, 3 months

[nolimips: 129] Fix label search for program display

by Noe

Index: ChangeLog from Benoît Perrot <benoit(a)lrde.epita.fr> * src/inst/text_section.cc: Fix label search for program display. 2004-09-26 Benoît Perrot <benoit(a)lrde.epita.fr> Index: src/inst/text_section.cc --- src/inst/text_section.cc (revision 128) +++ src/inst/text_section.cc (revision 129) @@ -44,7 +44,7 @@ for (unsigned i = 0; i < insts_.size(); ++i) { std::map<int, label_list_type>::const_iterator it = - labels_.find(i); + labels_.find(i * 4); if (it != labels_.end()) for (label_list_type::const_iterator l = (*it).second.begin(); l != (*it).second.end(); ++l)

20 years, 3 months

Jump to page:

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

Projects