Mon, 25 Jun 2012 21:33:35 -0400
7178670: runtime/7158800/BadUtf8.java fails in SymbolTable::rehash_table
Summary: Cannot delete _buckets and HashtableEntries in shared space (CDS)
Reviewed-by: acorn, kvn, dlong, dcubed, kamg
1 /*
2 * Copyright (c) 1997, 2011, Oracle and/or its affiliates. All rights reserved.
3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
4 *
5 * This code is free software; you can redistribute it and/or modify it
6 * under the terms of the GNU General Public License version 2 only, as
7 * published by the Free Software Foundation.
8 *
9 * This code is distributed in the hope that it will be useful, but WITHOUT
10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
12 * version 2 for more details (a copy is included in the LICENSE file that
13 * accompanied this code).
14 *
15 * You should have received a copy of the GNU General Public License version
16 * 2 along with this work; if not, write to the Free Software Foundation,
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
18 *
19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
20 * or visit www.oracle.com if you need additional information or have any
21 * questions.
22 *
23 */
25 // ADLPARSE.CPP - Architecture Description Language Parser
26 // Authors: Chris Vick and Mike Paleczny
27 #include "adlc.hpp"
29 //----------------------------ADLParser----------------------------------------
30 // Create a new ADL parser
31 ADLParser::ADLParser(FileBuff& buffer, ArchDesc& archDesc)
32 : _buf(buffer), _AD(archDesc),
33 _globalNames(archDesc.globalNames()) {
34 _AD._syntax_errs = _AD._semantic_errs = 0; // No errors so far this file
35 _AD._warnings = 0; // No warnings either
36 _curline = _ptr = NULL; // No pointers into buffer yet
38 _preproc_depth = 0;
39 _preproc_not_taken = 0;
41 // Delimit command-line definitions from in-file definitions:
42 _AD._preproc_list.add_signal();
43 }
45 //------------------------------~ADLParser-------------------------------------
46 // Delete an ADL parser.
47 ADLParser::~ADLParser() {
48 if (!_AD._quiet_mode)
49 fprintf(stderr,"---------------------------- Errors and Warnings ----------------------------\n");
50 #ifndef ASSERT
51 fprintf(stderr, "**************************************************************\n");
52 fprintf(stderr, "***** WARNING: ASSERT is undefined, assertions disabled. *****\n");
53 fprintf(stderr, "**************************************************************\n");
54 #endif
55 if( _AD._syntax_errs + _AD._semantic_errs + _AD._warnings == 0 ) {
56 if (!_AD._quiet_mode)
57 fprintf(stderr,"No errors or warnings to report from phase-1 parse.\n" );
58 }
59 else {
60 if( _AD._syntax_errs ) { // Any syntax errors?
61 fprintf(stderr,"%s: Found %d syntax error", _buf._fp->_name, _AD._syntax_errs);
62 if( _AD._syntax_errs > 1 ) fprintf(stderr,"s.\n\n");
63 else fprintf(stderr,".\n\n");
64 }
65 if( _AD._semantic_errs ) { // Any semantic errors?
66 fprintf(stderr,"%s: Found %d semantic error", _buf._fp->_name, _AD._semantic_errs);
67 if( _AD._semantic_errs > 1 ) fprintf(stderr,"s.\n\n");
68 else fprintf(stderr,".\n\n");
69 }
70 if( _AD._warnings ) { // Any warnings?
71 fprintf(stderr,"%s: Found %d warning", _buf._fp->_name, _AD._warnings);
72 if( _AD._warnings > 1 ) fprintf(stderr,"s.\n\n");
73 else fprintf(stderr,".\n\n");
74 }
75 }
76 if (!_AD._quiet_mode)
77 fprintf(stderr,"-----------------------------------------------------------------------------\n");
78 _AD._TotalLines += linenum()-1; // -1 for overshoot in "nextline" routine
80 // Write out information we have stored
81 // // UNIXism == fsync(stderr);
82 }
84 //------------------------------parse------------------------------------------
85 // Each top-level keyword should appear as the first non-whitespace on a line.
86 //
87 void ADLParser::parse() {
88 char *ident;
90 // Iterate over the lines in the file buffer parsing Level 1 objects
91 for( next_line(); _curline != NULL; next_line()) {
92 _ptr = _curline; // Reset ptr to start of new line
93 skipws(); // Skip any leading whitespace
94 ident = get_ident(); // Get first token
95 if (ident == NULL) { // Empty line
96 continue; // Get the next line
97 }
98 if (!strcmp(ident, "instruct")) instr_parse();
99 else if (!strcmp(ident, "operand")) oper_parse();
100 else if (!strcmp(ident, "opclass")) opclass_parse();
101 else if (!strcmp(ident, "ins_attrib")) ins_attr_parse();
102 else if (!strcmp(ident, "op_attrib")) op_attr_parse();
103 else if (!strcmp(ident, "source")) source_parse();
104 else if (!strcmp(ident, "source_hpp")) source_hpp_parse();
105 else if (!strcmp(ident, "register")) reg_parse();
106 else if (!strcmp(ident, "frame")) frame_parse();
107 else if (!strcmp(ident, "encode")) encode_parse();
108 else if (!strcmp(ident, "pipeline")) pipe_parse();
109 else if (!strcmp(ident, "definitions")) definitions_parse();
110 else if (!strcmp(ident, "peephole")) peep_parse();
111 else if (!strcmp(ident, "#line")) preproc_line();
112 else if (!strcmp(ident, "#define")) preproc_define();
113 else if (!strcmp(ident, "#undef")) preproc_undef();
114 else {
115 parse_err(SYNERR, "expected one of - instruct, operand, ins_attrib, op_attrib, source, register, pipeline, encode\n Found %s",ident);
116 }
117 }
119 // Done with parsing, check consistency.
121 if (_preproc_depth != 0) {
122 parse_err(SYNERR, "End of file inside #ifdef");
123 }
125 // AttributeForms ins_cost and op_cost must be defined for default behaviour
126 if (_globalNames[AttributeForm::_ins_cost] == NULL) {
127 parse_err(SEMERR, "Did not declare 'ins_cost' attribute");
128 }
129 if (_globalNames[AttributeForm::_op_cost] == NULL) {
130 parse_err(SEMERR, "Did not declare 'op_cost' attribute");
131 }
132 }
134 // ******************** Private Level 1 Parse Functions ********************
135 //------------------------------instr_parse------------------------------------
136 // Parse the contents of an instruction definition, build the InstructForm to
137 // represent that instruction, and add it to the InstructForm list.
138 void ADLParser::instr_parse(void) {
139 char *ident;
140 InstructForm *instr;
141 MatchRule *rule;
142 int match_rules_cnt = 0;
144 // First get the name of the instruction
145 if( (ident = get_unique_ident(_globalNames,"instruction")) == NULL )
146 return;
147 instr = new InstructForm(ident); // Create new instruction form
148 instr->_linenum = linenum();
149 _globalNames.Insert(ident, instr); // Add name to the name table
150 // Debugging Stuff
151 if (_AD._adl_debug > 1)
152 fprintf(stderr,"Parsing Instruction Form %s\n", ident);
154 // Then get the operands
155 skipws();
156 if (_curchar != '(') {
157 parse_err(SYNERR, "missing '(' in instruct definition\n");
158 }
159 // Parse the operand list
160 else get_oplist(instr->_parameters, instr->_localNames);
161 skipws(); // Skip leading whitespace
162 // Check for block delimiter
163 if ( (_curchar != '%')
164 || ( next_char(), (_curchar != '{')) ) {
165 parse_err(SYNERR, "missing '%{' in instruction definition\n");
166 return;
167 }
168 next_char(); // Maintain the invariant
169 do {
170 ident = get_ident(); // Grab next identifier
171 if (ident == NULL) {
172 parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
173 continue;
174 }
175 if (!strcmp(ident, "predicate")) instr->_predicate = pred_parse();
176 else if (!strcmp(ident, "match")) {
177 // Allow one instruction have several match rules.
178 rule = instr->_matrule;
179 if (rule == NULL) {
180 // This is first match rule encountered
181 rule = match_parse(instr->_localNames);
182 if (rule) {
183 instr->_matrule = rule;
184 // Special case the treatment of Control instructions.
185 if( instr->is_ideal_control() ) {
186 // Control instructions return a special result, 'Universe'
187 rule->_result = "Universe";
188 }
189 // Check for commutative operations with tree operands.
190 matchrule_clone_and_swap(rule, instr->_ident, match_rules_cnt);
191 }
192 } else {
193 // Find the end of the match rule list
194 while (rule->_next != NULL)
195 rule = rule->_next;
196 // Add the new match rule to the list
197 rule->_next = match_parse(instr->_localNames);
198 if (rule->_next) {
199 rule = rule->_next;
200 if( instr->is_ideal_control() ) {
201 parse_err(SYNERR, "unique match rule expected for %s\n", rule->_name);
202 return;
203 }
204 assert(match_rules_cnt < 100," too many match rule clones");
205 char* buf = (char*) malloc(strlen(instr->_ident) + 4);
206 sprintf(buf, "%s_%d", instr->_ident, match_rules_cnt++);
207 rule->_result = buf;
208 // Check for commutative operations with tree operands.
209 matchrule_clone_and_swap(rule, instr->_ident, match_rules_cnt);
210 }
211 }
212 }
213 else if (!strcmp(ident, "encode")) {
214 parse_err(SYNERR, "Instructions specify ins_encode, not encode\n");
215 }
216 else if (!strcmp(ident, "ins_encode")) ins_encode_parse(*instr);
217 else if (!strcmp(ident, "opcode")) instr->_opcode = opcode_parse(instr);
218 else if (!strcmp(ident, "size")) instr->_size = size_parse(instr);
219 else if (!strcmp(ident, "effect")) effect_parse(instr);
220 else if (!strcmp(ident, "expand")) instr->_exprule = expand_parse(instr);
221 else if (!strcmp(ident, "rewrite")) instr->_rewrule = rewrite_parse();
222 else if (!strcmp(ident, "constraint")) {
223 parse_err(SYNERR, "Instructions do not specify a constraint\n");
224 }
225 else if (!strcmp(ident, "construct")) {
226 parse_err(SYNERR, "Instructions do not specify a construct\n");
227 }
228 else if (!strcmp(ident, "format")) instr->_format = format_parse();
229 else if (!strcmp(ident, "interface")) {
230 parse_err(SYNERR, "Instructions do not specify an interface\n");
231 }
232 else if (!strcmp(ident, "ins_pipe")) ins_pipe_parse(*instr);
233 else { // Done with staticly defined parts of instruction definition
234 // Check identifier to see if it is the name of an attribute
235 const Form *form = _globalNames[ident];
236 AttributeForm *attr = form ? form->is_attribute() : NULL;
237 if( attr && (attr->_atype == INS_ATTR) ) {
238 // Insert the new attribute into the linked list.
239 Attribute *temp = attr_parse(ident);
240 temp->_next = instr->_attribs;
241 instr->_attribs = temp;
242 } else {
243 parse_err(SYNERR, "expected one of:\n predicate, match, encode, or the name of an instruction attribute at %s\n", ident);
244 }
245 }
246 skipws();
247 } while(_curchar != '%');
248 next_char();
249 if (_curchar != '}') {
250 parse_err(SYNERR, "missing '%}' in instruction definition\n");
251 return;
252 }
253 // Check for "Set" form of chain rule
254 adjust_set_rule(instr);
255 if (_AD._pipeline ) {
256 if( instr->expands() ) {
257 if( instr->_ins_pipe )
258 parse_err(WARN, "ins_pipe and expand rule both specified for instruction \"%s\"; ins_pipe will be unused\n", instr->_ident);
259 } else {
260 if( !instr->_ins_pipe )
261 parse_err(WARN, "No ins_pipe specified for instruction \"%s\"\n", instr->_ident);
262 }
263 }
264 // Add instruction to tail of instruction list
265 _AD.addForm(instr);
267 // Create instruction form for each additional match rule
268 rule = instr->_matrule;
269 if (rule != NULL) {
270 rule = rule->_next;
271 while (rule != NULL) {
272 ident = (char*)rule->_result;
273 InstructForm *clone = new InstructForm(ident, instr, rule); // Create new instruction form
274 _globalNames.Insert(ident, clone); // Add name to the name table
275 // Debugging Stuff
276 if (_AD._adl_debug > 1)
277 fprintf(stderr,"Parsing Instruction Form %s\n", ident);
278 // Check for "Set" form of chain rule
279 adjust_set_rule(clone);
280 // Add instruction to tail of instruction list
281 _AD.addForm(clone);
282 rule = rule->_next;
283 clone->_matrule->_next = NULL; // One match rule per clone
284 }
285 }
286 }
288 //------------------------------matchrule_clone_and_swap-----------------------
289 // Check for commutative operations with subtree operands,
290 // create clones and swap operands.
291 void ADLParser::matchrule_clone_and_swap(MatchRule* rule, const char* instr_ident, int& match_rules_cnt) {
292 // Check for commutative operations with tree operands.
293 int count = 0;
294 rule->count_commutative_op(count);
295 if (count > 0) {
296 // Clone match rule and swap commutative operation's operands.
297 rule->matchrule_swap_commutative_op(instr_ident, count, match_rules_cnt);
298 }
299 }
301 //------------------------------adjust_set_rule--------------------------------
302 // Check for "Set" form of chain rule
303 void ADLParser::adjust_set_rule(InstructForm *instr) {
304 if (instr->_matrule == NULL || instr->_matrule->_rChild == NULL) return;
305 const char *rch = instr->_matrule->_rChild->_opType;
306 const Form *frm = _globalNames[rch];
307 if( (! strcmp(instr->_matrule->_opType,"Set")) &&
308 frm && frm->is_operand() && (! frm->ideal_only()) ) {
309 // Previous implementation, which missed leaP*, but worked for loadCon*
310 unsigned position = 0;
311 const char *result = NULL;
312 const char *name = NULL;
313 const char *optype = NULL;
314 MatchNode *right = instr->_matrule->_rChild;
315 if (right->base_operand(position, _globalNames, result, name, optype)) {
316 position = 1;
317 const char *result2 = NULL;
318 const char *name2 = NULL;
319 const char *optype2 = NULL;
320 // Can not have additional base operands in right side of match!
321 if ( ! right->base_operand( position, _globalNames, result2, name2, optype2) ) {
322 if (instr->_predicate != NULL)
323 parse_err(SYNERR, "ADLC does not support instruction chain rules with predicates");
324 // Chain from input _ideal_operand_type_,
325 // Needed for shared roots of match-trees
326 ChainList *lst = (ChainList *)_AD._chainRules[optype];
327 if (lst == NULL) {
328 lst = new ChainList();
329 _AD._chainRules.Insert(optype, lst);
330 }
331 if (!lst->search(instr->_matrule->_lChild->_opType)) {
332 const char *cost = instr->cost();
333 if (cost == NULL) {
334 cost = ((AttributeForm*)_globalNames[AttributeForm::_ins_cost])->_attrdef;
335 }
336 // The ADLC does not support chaining from the ideal operand type
337 // of a predicated user-defined operand
338 if( frm->is_operand() == NULL || frm->is_operand()->_predicate == NULL ) {
339 lst->insert(instr->_matrule->_lChild->_opType,cost,instr->_ident);
340 }
341 }
342 // Chain from input _user_defined_operand_type_,
343 lst = (ChainList *)_AD._chainRules[result];
344 if (lst == NULL) {
345 lst = new ChainList();
346 _AD._chainRules.Insert(result, lst);
347 }
348 if (!lst->search(instr->_matrule->_lChild->_opType)) {
349 const char *cost = instr->cost();
350 if (cost == NULL) {
351 cost = ((AttributeForm*)_globalNames[AttributeForm::_ins_cost])->_attrdef;
352 }
353 // It is safe to chain from the top-level user-defined operand even
354 // if it has a predicate, since the predicate is checked before
355 // the user-defined type is available.
356 lst->insert(instr->_matrule->_lChild->_opType,cost,instr->_ident);
357 }
358 } else {
359 // May have instruction chain rule if root of right-tree is an ideal
360 OperandForm *rightOp = _globalNames[right->_opType]->is_operand();
361 if( rightOp ) {
362 const Form *rightRoot = _globalNames[rightOp->_matrule->_opType];
363 if( rightRoot && rightRoot->ideal_only() ) {
364 const char *chain_op = NULL;
365 if( rightRoot->is_instruction() )
366 chain_op = rightOp->_ident;
367 if( chain_op ) {
368 // Look-up the operation in chain rule table
369 ChainList *lst = (ChainList *)_AD._chainRules[chain_op];
370 if (lst == NULL) {
371 lst = new ChainList();
372 _AD._chainRules.Insert(chain_op, lst);
373 }
374 // if (!lst->search(instr->_matrule->_lChild->_opType)) {
375 const char *cost = instr->cost();
376 if (cost == NULL) {
377 cost = ((AttributeForm*)_globalNames[AttributeForm::_ins_cost])->_attrdef;
378 }
379 // This chains from a top-level operand whose predicate, if any,
380 // has been checked.
381 lst->insert(instr->_matrule->_lChild->_opType,cost,instr->_ident);
382 // }
383 }
384 }
385 }
386 } // end chain rule from right-tree's ideal root
387 }
388 }
389 }
392 //------------------------------oper_parse-------------------------------------
393 void ADLParser::oper_parse(void) {
394 char *ident;
395 OperandForm *oper;
396 AttributeForm *attr;
397 MatchRule *rule;
399 // First get the name of the operand
400 skipws();
401 if( (ident = get_unique_ident(_globalNames,"operand")) == NULL )
402 return;
403 oper = new OperandForm(ident); // Create new operand form
404 oper->_linenum = linenum();
405 _globalNames.Insert(ident, oper); // Add name to the name table
407 // Debugging Stuff
408 if (_AD._adl_debug > 1) fprintf(stderr,"Parsing Operand Form %s\n", ident);
410 // Get the component operands
411 skipws();
412 if (_curchar != '(') {
413 parse_err(SYNERR, "missing '(' in operand definition\n");
414 return;
415 }
416 else get_oplist(oper->_parameters, oper->_localNames); // Parse the component operand list
417 skipws();
418 // Check for block delimiter
419 if ((_curchar != '%') || (*(_ptr+1) != '{')) { // If not open block
420 parse_err(SYNERR, "missing '%c{' in operand definition\n","%");
421 return;
422 }
423 next_char(); next_char(); // Skip over "%{" symbol
424 do {
425 ident = get_ident(); // Grab next identifier
426 if (ident == NULL) {
427 parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
428 continue;
429 }
430 if (!strcmp(ident, "predicate")) oper->_predicate = pred_parse();
431 else if (!strcmp(ident, "match")) {
432 // Find the end of the match rule list
433 rule = oper->_matrule;
434 if (rule) {
435 while (rule->_next) rule = rule->_next;
436 // Add the new match rule to the list
437 rule->_next = match_parse(oper->_localNames);
438 if (rule->_next) {
439 rule->_next->_result = oper->_ident;
440 }
441 }
442 else {
443 // This is first match rule encountered
444 oper->_matrule = match_parse(oper->_localNames);
445 if (oper->_matrule) {
446 oper->_matrule->_result = oper->_ident;
447 }
448 }
449 }
450 else if (!strcmp(ident, "encode")) oper->_interface = interface_parse();
451 else if (!strcmp(ident, "ins_encode")) {
452 parse_err(SYNERR, "Operands specify 'encode', not 'ins_encode'\n");
453 }
454 else if (!strcmp(ident, "opcode")) {
455 parse_err(SYNERR, "Operands do not specify an opcode\n");
456 }
457 else if (!strcmp(ident, "effect")) {
458 parse_err(SYNERR, "Operands do not specify an effect\n");
459 }
460 else if (!strcmp(ident, "expand")) {
461 parse_err(SYNERR, "Operands do not specify an expand\n");
462 }
463 else if (!strcmp(ident, "rewrite")) {
464 parse_err(SYNERR, "Operands do not specify a rewrite\n");
465 }
466 else if (!strcmp(ident, "constraint"))oper->_constraint= constraint_parse();
467 else if (!strcmp(ident, "construct")) oper->_construct = construct_parse();
468 else if (!strcmp(ident, "format")) oper->_format = format_parse();
469 else if (!strcmp(ident, "interface")) oper->_interface = interface_parse();
470 // Check identifier to see if it is the name of an attribute
471 else if (((attr = _globalNames[ident]->is_attribute()) != NULL) &&
472 (attr->_atype == OP_ATTR)) oper->_attribs = attr_parse(ident);
473 else {
474 parse_err(SYNERR, "expected one of - constraint, predicate, match, encode, format, construct, or the name of a defined operand attribute at %s\n", ident);
475 }
476 skipws();
477 } while(_curchar != '%');
478 next_char();
479 if (_curchar != '}') {
480 parse_err(SYNERR, "missing '%}' in operand definition\n");
481 return;
482 }
483 // Add operand to tail of operand list
484 _AD.addForm(oper);
485 }
487 //------------------------------opclass_parse----------------------------------
488 // Operand Classes are a block with a comma delimited list of operand names
489 void ADLParser::opclass_parse(void) {
490 char *ident;
491 OpClassForm *opc;
492 OperandForm *opForm;
494 // First get the name of the operand class
495 skipws();
496 if( (ident = get_unique_ident(_globalNames,"opclass")) == NULL )
497 return;
498 opc = new OpClassForm(ident); // Create new operand class form
499 _globalNames.Insert(ident, opc); // Add name to the name table
501 // Debugging Stuff
502 if (_AD._adl_debug > 1)
503 fprintf(stderr,"Parsing Operand Class Form %s\n", ident);
505 // Get the list of operands
506 skipws();
507 if (_curchar != '(') {
508 parse_err(SYNERR, "missing '(' in operand definition\n");
509 return;
510 }
511 do {
512 next_char(); // Skip past open paren or comma
513 ident = get_ident(); // Grab next identifier
514 if (ident == NULL) {
515 parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
516 continue;
517 }
518 // Check identifier to see if it is the name of an operand
519 const Form *form = _globalNames[ident];
520 opForm = form ? form->is_operand() : NULL;
521 if ( opForm ) {
522 opc->_oplst.addName(ident); // Add operand to opclass list
523 opForm->_classes.addName(opc->_ident);// Add opclass to operand list
524 }
525 else {
526 parse_err(SYNERR, "expected name of a defined operand at %s\n", ident);
527 }
528 skipws(); // skip trailing whitespace
529 } while (_curchar == ','); // Check for the comma
530 // Check for closing ')'
531 if (_curchar != ')') {
532 parse_err(SYNERR, "missing ')' or ',' in opclass definition\n");
533 return;
534 }
535 next_char(); // Consume the ')'
536 skipws();
537 // Check for closing ';'
538 if (_curchar != ';') {
539 parse_err(SYNERR, "missing ';' in opclass definition\n");
540 return;
541 }
542 next_char(); // Consume the ';'
543 // Add operand to tail of operand list
544 _AD.addForm(opc);
545 }
547 //------------------------------ins_attr_parse---------------------------------
548 void ADLParser::ins_attr_parse(void) {
549 char *ident;
550 char *aexpr;
551 AttributeForm *attrib;
553 // get name for the instruction attribute
554 skipws(); // Skip leading whitespace
555 if( (ident = get_unique_ident(_globalNames,"inst_attrib")) == NULL )
556 return;
557 // Debugging Stuff
558 if (_AD._adl_debug > 1) fprintf(stderr,"Parsing Ins_Attribute Form %s\n", ident);
560 // Get default value of the instruction attribute
561 skipws(); // Skip whitespace
562 if ((aexpr = get_paren_expr("attribute default expression string")) == NULL) {
563 parse_err(SYNERR, "missing '(' in ins_attrib definition\n");
564 return;
565 }
566 // Debug Stuff
567 if (_AD._adl_debug > 1) fprintf(stderr,"Attribute Expression: %s\n", aexpr);
569 // Check for terminator
570 if (_curchar != ';') {
571 parse_err(SYNERR, "missing ';' in ins_attrib definition\n");
572 return;
573 }
574 next_char(); // Advance past the ';'
576 // Construct the attribute, record global name, and store in ArchDesc
577 attrib = new AttributeForm(ident, INS_ATTR, aexpr);
578 _globalNames.Insert(ident, attrib); // Add name to the name table
579 _AD.addForm(attrib);
580 }
582 //------------------------------op_attr_parse----------------------------------
583 void ADLParser::op_attr_parse(void) {
584 char *ident;
585 char *aexpr;
586 AttributeForm *attrib;
588 // get name for the operand attribute
589 skipws(); // Skip leading whitespace
590 if( (ident = get_unique_ident(_globalNames,"op_attrib")) == NULL )
591 return;
592 // Debugging Stuff
593 if (_AD._adl_debug > 1) fprintf(stderr,"Parsing Op_Attribute Form %s\n", ident);
595 // Get default value of the instruction attribute
596 skipws(); // Skip whitespace
597 if ((aexpr = get_paren_expr("attribute default expression string")) == NULL) {
598 parse_err(SYNERR, "missing '(' in op_attrib definition\n");
599 return;
600 }
601 // Debug Stuff
602 if (_AD._adl_debug > 1) fprintf(stderr,"Attribute Expression: %s\n", aexpr);
604 // Check for terminator
605 if (_curchar != ';') {
606 parse_err(SYNERR, "missing ';' in op_attrib definition\n");
607 return;
608 }
609 next_char(); // Advance past the ';'
611 // Construct the attribute, record global name, and store in ArchDesc
612 attrib = new AttributeForm(ident, OP_ATTR, aexpr);
613 _globalNames.Insert(ident, attrib);
614 _AD.addForm(attrib);
615 }
617 //------------------------------definitions_parse-----------------------------------
618 void ADLParser::definitions_parse(void) {
619 skipws(); // Skip leading whitespace
620 if (_curchar == '%' && *(_ptr+1) == '{') {
621 next_char(); next_char(); // Skip "%{"
622 skipws();
623 while (_curchar != '%' && *(_ptr+1) != '}') {
624 // Process each definition until finding closing string "%}"
625 char *token = get_ident();
626 if (token == NULL) {
627 parse_err(SYNERR, "missing identifier inside definitions block.\n");
628 return;
629 }
630 if (strcmp(token,"int_def")==0) { int_def_parse(); }
631 // if (strcmp(token,"str_def")==0) { str_def_parse(); }
632 skipws();
633 }
634 }
635 else {
636 parse_err(SYNERR, "Missing %%{ ... %%} block after definitions keyword.\n");
637 return;
638 }
639 }
641 //------------------------------int_def_parse----------------------------------
642 // Parse Example:
643 // int_def MEMORY_REF_COST ( 200, DEFAULT_COST * 2);
644 // <keyword> <name> ( <int_value>, <description> );
645 //
646 void ADLParser::int_def_parse(void) {
647 char *name = NULL; // Name of definition
648 char *value = NULL; // its value,
649 int int_value = -1; // positive values only
650 char *description = NULL; // textual description
652 // Get definition name
653 skipws(); // Skip whitespace
654 name = get_ident();
655 if (name == NULL) {
656 parse_err(SYNERR, "missing definition name after int_def\n");
657 return;
658 }
660 // Check for value of int_def dname( integer_value [, string_expression ] )
661 skipws();
662 if (_curchar == '(') {
664 // Parse the integer value.
665 next_char();
666 value = get_ident();
667 if (value == NULL) {
668 parse_err(SYNERR, "missing value in int_def\n");
669 return;
670 }
671 if( !is_int_token(value, int_value) ) {
672 parse_err(SYNERR, "value in int_def is not recognized as integer\n");
673 return;
674 }
675 skipws();
677 // Check for description
678 if (_curchar == ',') {
679 next_char(); // skip ','
681 description = get_expr("int_def description", ")");
682 if (description == NULL) {
683 parse_err(SYNERR, "invalid or missing description in int_def\n");
684 return;
685 }
686 trim(description);
687 }
689 if (_curchar != ')') {
690 parse_err(SYNERR, "missing ')' in register definition statement\n");
691 return;
692 }
693 next_char();
694 }
696 // Check for closing ';'
697 skipws();
698 if (_curchar != ';') {
699 parse_err(SYNERR, "missing ';' after int_def\n");
700 return;
701 }
702 next_char(); // move past ';'
704 // Debug Stuff
705 if (_AD._adl_debug > 1) {
706 fprintf(stderr,"int_def: %s ( %s, %s )\n", name,
707 (value), (description ? description : ""));
708 }
710 // Record new definition.
711 Expr *expr = new Expr(name, description, int_value, int_value);
712 const Expr *old_expr = _AD.globalDefs().define(name, expr);
713 if (old_expr != NULL) {
714 parse_err(SYNERR, "Duplicate definition\n");
715 return;
716 }
718 return;
719 }
722 //------------------------------source_parse-----------------------------------
723 void ADLParser::source_parse(void) {
724 SourceForm *source; // Encode class for instruction/operand
725 char *rule = NULL; // String representation of encode rule
727 skipws(); // Skip leading whitespace
728 if ( (rule = find_cpp_block("source block")) == NULL ) {
729 parse_err(SYNERR, "incorrect or missing block for 'source'.\n");
730 return;
731 }
732 // Debug Stuff
733 if (_AD._adl_debug > 1) fprintf(stderr,"Source Form: %s\n", rule);
735 source = new SourceForm(rule); // Build new Source object
736 _AD.addForm(source);
737 // skipws();
738 }
740 //------------------------------source_hpp_parse-------------------------------
741 // Parse a source_hpp %{ ... %} block.
742 // The code gets stuck into the ad_<arch>.hpp file.
743 // If the source_hpp block appears before the register block in the AD
744 // file, it goes up at the very top of the ad_<arch>.hpp file, so that
745 // it can be used by register encodings, etc. Otherwise, it goes towards
746 // the bottom, where it's useful as a global definition to *.cpp files.
747 void ADLParser::source_hpp_parse(void) {
748 char *rule = NULL; // String representation of encode rule
750 skipws(); // Skip leading whitespace
751 if ( (rule = find_cpp_block("source_hpp block")) == NULL ) {
752 parse_err(SYNERR, "incorrect or missing block for 'source_hpp'.\n");
753 return;
754 }
755 // Debug Stuff
756 if (_AD._adl_debug > 1) fprintf(stderr,"Header Form: %s\n", rule);
758 if (_AD.get_registers() == NULL) {
759 // Very early in the file, before reg_defs, we collect pre-headers.
760 PreHeaderForm* pre_header = new PreHeaderForm(rule);
761 _AD.addForm(pre_header);
762 } else {
763 // Normally, we collect header info, placed at the bottom of the hpp file.
764 HeaderForm* header = new HeaderForm(rule);
765 _AD.addForm(header);
766 }
767 }
769 //------------------------------reg_parse--------------------------------------
770 void ADLParser::reg_parse(void) {
772 // Create the RegisterForm for the architecture description.
773 RegisterForm *regBlock = new RegisterForm(); // Build new Source object
774 regBlock->_linenum = linenum();
775 _AD.addForm(regBlock);
777 skipws(); // Skip leading whitespace
778 if (_curchar == '%' && *(_ptr+1) == '{') {
779 next_char(); next_char(); // Skip "%{"
780 skipws();
781 while (_curchar != '%' && *(_ptr+1) != '}') {
782 char *token = get_ident();
783 if (token == NULL) {
784 parse_err(SYNERR, "missing identifier inside register block.\n");
785 return;
786 }
787 if (strcmp(token,"reg_def")==0) { reg_def_parse(); }
788 else if (strcmp(token,"reg_class")==0) { reg_class_parse(); }
789 else if (strcmp(token,"alloc_class")==0) { alloc_class_parse(); }
790 else if (strcmp(token,"#define")==0) { preproc_define(); }
791 else { parse_err(SYNERR, "bad token %s inside register block.\n", token); break; }
792 skipws();
793 }
794 }
795 else {
796 parse_err(SYNERR, "Missing %c{ ... %c} block after register keyword.\n",'%','%');
797 return;
798 }
800 // Add reg_class spill_regs
801 regBlock->addSpillRegClass();
802 }
804 //------------------------------encode_parse-----------------------------------
805 void ADLParser::encode_parse(void) {
806 EncodeForm *encBlock; // Information about instruction/operand encoding
807 char *desc = NULL; // String representation of encode rule
809 _AD.getForm(&encBlock);
810 if ( encBlock == NULL) {
811 // Create the EncodeForm for the architecture description.
812 encBlock = new EncodeForm(); // Build new Source object
813 _AD.addForm(encBlock);
814 }
816 skipws(); // Skip leading whitespace
817 if (_curchar == '%' && *(_ptr+1) == '{') {
818 next_char(); next_char(); // Skip "%{"
819 skipws();
820 while (_curchar != '%' && *(_ptr+1) != '}') {
821 char *token = get_ident();
822 if (token == NULL) {
823 parse_err(SYNERR, "missing identifier inside encoding block.\n");
824 return;
825 }
826 if (strcmp(token,"enc_class")==0) { enc_class_parse(); }
827 skipws();
828 }
829 }
830 else {
831 parse_err(SYNERR, "Missing %c{ ... %c} block after encode keyword.\n",'%','%');
832 return;
833 }
834 }
836 //------------------------------enc_class_parse--------------------------------
837 void ADLParser::enc_class_parse(void) {
838 char *ec_name; // Name of encoding class being defined
840 // Get encoding class name
841 skipws(); // Skip whitespace
842 ec_name = get_ident();
843 if (ec_name == NULL) {
844 parse_err(SYNERR, "missing encoding class name after encode.\n");
845 return;
846 }
848 EncClass *encoding = _AD._encode->add_EncClass(ec_name);
849 encoding->_linenum = linenum();
851 skipws(); // Skip leading whitespace
852 // Check for optional parameter list
853 if (_curchar == '(') {
854 do {
855 char *pType = NULL; // parameter type
856 char *pName = NULL; // parameter name
858 next_char(); // skip open paren & comma characters
859 skipws();
860 if (_curchar == ')') break;
862 // Get parameter type
863 pType = get_ident();
864 if (pType == NULL) {
865 parse_err(SYNERR, "parameter type expected at %c\n", _curchar);
866 return;
867 }
869 skipws();
870 // Get parameter name
871 pName = get_ident();
872 if (pName == NULL) {
873 parse_err(SYNERR, "parameter name expected at %c\n", _curchar);
874 return;
875 }
877 // Record parameter type and name
878 encoding->add_parameter( pType, pName );
880 skipws();
881 } while(_curchar == ',');
883 if (_curchar != ')') parse_err(SYNERR, "missing ')'\n");
884 else {
885 next_char(); // Skip ')'
886 }
887 } // Done with parameter list
889 skipws();
890 // Check for block starting delimiters
891 if ((_curchar != '%') || (*(_ptr+1) != '{')) { // If not open block
892 parse_err(SYNERR, "missing '%c{' in enc_class definition\n", '%');
893 return;
894 }
895 next_char(); // Skip '%'
896 next_char(); // Skip '{'
898 enc_class_parse_block(encoding, ec_name);
899 }
902 void ADLParser::enc_class_parse_block(EncClass* encoding, char* ec_name) {
903 skipws_no_preproc(); // Skip leading whitespace
904 // Prepend location descriptor, for debugging; cf. ADLParser::find_cpp_block
905 if (_AD._adlocation_debug) {
906 encoding->add_code(get_line_string());
907 }
909 // Collect the parts of the encode description
910 // (1) strings that are passed through to output
911 // (2) replacement/substitution variable, preceeded by a '$'
912 while ( (_curchar != '%') && (*(_ptr+1) != '}') ) {
914 // (1)
915 // Check if there is a string to pass through to output
916 char *start = _ptr; // Record start of the next string
917 while ((_curchar != '$') && ((_curchar != '%') || (*(_ptr+1) != '}')) ) {
918 // If at the start of a comment, skip past it
919 if( (_curchar == '/') && ((*(_ptr+1) == '/') || (*(_ptr+1) == '*')) ) {
920 skipws_no_preproc();
921 } else {
922 // ELSE advance to the next character, or start of the next line
923 next_char_or_line();
924 }
925 }
926 // If a string was found, terminate it and record in EncClass
927 if ( start != _ptr ) {
928 *_ptr = '\0'; // Terminate the string
929 encoding->add_code(start);
930 }
932 // (2)
933 // If we are at a replacement variable,
934 // copy it and record in EncClass
935 if (_curchar == '$') {
936 // Found replacement Variable
937 char* rep_var = get_rep_var_ident_dup();
938 // Add flag to _strings list indicating we should check _rep_vars
939 encoding->add_rep_var(rep_var);
940 }
941 } // end while part of format description
942 next_char(); // Skip '%'
943 next_char(); // Skip '}'
945 skipws();
947 if (_AD._adlocation_debug) {
948 encoding->add_code(end_line_marker());
949 }
951 // Debug Stuff
952 if (_AD._adl_debug > 1) fprintf(stderr,"EncodingClass Form: %s\n", ec_name);
953 }
955 //------------------------------frame_parse-----------------------------------
956 void ADLParser::frame_parse(void) {
957 FrameForm *frame; // Information about stack-frame layout
958 char *desc = NULL; // String representation of frame
960 skipws(); // Skip leading whitespace
962 frame = new FrameForm(); // Build new Frame object
963 // Check for open block sequence
964 skipws(); // Skip leading whitespace
965 if (_curchar == '%' && *(_ptr+1) == '{') {
966 next_char(); next_char(); // Skip "%{"
967 skipws();
968 while (_curchar != '%' && *(_ptr+1) != '}') {
969 char *token = get_ident();
970 if (token == NULL) {
971 parse_err(SYNERR, "missing identifier inside frame block.\n");
972 return;
973 }
974 if (strcmp(token,"stack_direction")==0) {
975 stack_dir_parse(frame);
976 }
977 if (strcmp(token,"sync_stack_slots")==0) {
978 sync_stack_slots_parse(frame);
979 }
980 if (strcmp(token,"frame_pointer")==0) {
981 frame_pointer_parse(frame, false);
982 }
983 if (strcmp(token,"interpreter_frame_pointer")==0) {
984 interpreter_frame_pointer_parse(frame, false);
985 }
986 if (strcmp(token,"inline_cache_reg")==0) {
987 inline_cache_parse(frame, false);
988 }
989 if (strcmp(token,"compiler_method_oop_reg")==0) {
990 parse_err(WARN, "Using obsolete Token, compiler_method_oop_reg");
991 skipws();
992 }
993 if (strcmp(token,"interpreter_method_oop_reg")==0) {
994 interpreter_method_oop_parse(frame, false);
995 }
996 if (strcmp(token,"cisc_spilling_operand_name")==0) {
997 cisc_spilling_operand_name_parse(frame, false);
998 }
999 if (strcmp(token,"stack_alignment")==0) {
1000 stack_alignment_parse(frame);
1001 }
1002 if (strcmp(token,"return_addr")==0) {
1003 return_addr_parse(frame, false);
1004 }
1005 if (strcmp(token,"in_preserve_stack_slots")==0) {
1006 preserve_stack_parse(frame);
1007 }
1008 if (strcmp(token,"out_preserve_stack_slots")==0) {
1009 parse_err(WARN, "Using obsolete token, out_preserve_stack_slots");
1010 skipws();
1011 }
1012 if (strcmp(token,"varargs_C_out_slots_killed")==0) {
1013 frame->_varargs_C_out_slots_killed = parse_one_arg("varargs C out slots killed");
1014 }
1015 if (strcmp(token,"calling_convention")==0) {
1016 frame->_calling_convention = calling_convention_parse();
1017 }
1018 if (strcmp(token,"return_value")==0) {
1019 frame->_return_value = return_value_parse();
1020 }
1021 if (strcmp(token,"c_frame_pointer")==0) {
1022 frame_pointer_parse(frame, true);
1023 }
1024 if (strcmp(token,"c_return_addr")==0) {
1025 return_addr_parse(frame, true);
1026 }
1027 if (strcmp(token,"c_calling_convention")==0) {
1028 frame->_c_calling_convention = calling_convention_parse();
1029 }
1030 if (strcmp(token,"c_return_value")==0) {
1031 frame->_c_return_value = return_value_parse();
1032 }
1034 skipws();
1035 }
1036 }
1037 else {
1038 parse_err(SYNERR, "Missing %c{ ... %c} block after encode keyword.\n",'%','%');
1039 return;
1040 }
1041 // All Java versions are required, native versions are optional
1042 if(frame->_frame_pointer == NULL) {
1043 parse_err(SYNERR, "missing frame pointer definition in frame section.\n");
1044 return;
1045 }
1046 // !!!!! !!!!!
1047 // if(frame->_interpreter_frame_ptr_reg == NULL) {
1048 // parse_err(SYNERR, "missing interpreter frame pointer definition in frame section.\n");
1049 // return;
1050 // }
1051 if(frame->_alignment == NULL) {
1052 parse_err(SYNERR, "missing alignment definition in frame section.\n");
1053 return;
1054 }
1055 if(frame->_return_addr == NULL) {
1056 parse_err(SYNERR, "missing return address location in frame section.\n");
1057 return;
1058 }
1059 if(frame->_in_preserve_slots == NULL) {
1060 parse_err(SYNERR, "missing stack slot preservation definition in frame section.\n");
1061 return;
1062 }
1063 if(frame->_varargs_C_out_slots_killed == NULL) {
1064 parse_err(SYNERR, "missing varargs C out slots killed definition in frame section.\n");
1065 return;
1066 }
1067 if(frame->_calling_convention == NULL) {
1068 parse_err(SYNERR, "missing calling convention definition in frame section.\n");
1069 return;
1070 }
1071 if(frame->_return_value == NULL) {
1072 parse_err(SYNERR, "missing return value definition in frame section.\n");
1073 return;
1074 }
1075 // Fill natives in identically with the Java versions if not present.
1076 if(frame->_c_frame_pointer == NULL) {
1077 frame->_c_frame_pointer = frame->_frame_pointer;
1078 }
1079 if(frame->_c_return_addr == NULL) {
1080 frame->_c_return_addr = frame->_return_addr;
1081 frame->_c_return_addr_loc = frame->_return_addr_loc;
1082 }
1083 if(frame->_c_calling_convention == NULL) {
1084 frame->_c_calling_convention = frame->_calling_convention;
1085 }
1086 if(frame->_c_return_value == NULL) {
1087 frame->_c_return_value = frame->_return_value;
1088 }
1090 // Debug Stuff
1091 if (_AD._adl_debug > 1) fprintf(stderr,"Frame Form: %s\n", desc);
1093 // Create the EncodeForm for the architecture description.
1094 _AD.addForm(frame);
1095 // skipws();
1096 }
1098 //------------------------------stack_dir_parse--------------------------------
1099 void ADLParser::stack_dir_parse(FrameForm *frame) {
1100 char *direction = parse_one_arg("stack direction entry");
1101 if (strcmp(direction, "TOWARDS_LOW") == 0) {
1102 frame->_direction = false;
1103 }
1104 else if (strcmp(direction, "TOWARDS_HIGH") == 0) {
1105 frame->_direction = true;
1106 }
1107 else {
1108 parse_err(SYNERR, "invalid value inside stack direction entry.\n");
1109 return;
1110 }
1111 }
1113 //------------------------------sync_stack_slots_parse-------------------------
1114 void ADLParser::sync_stack_slots_parse(FrameForm *frame) {
1115 // Assign value into frame form
1116 frame->_sync_stack_slots = parse_one_arg("sync stack slots entry");
1117 }
1119 //------------------------------frame_pointer_parse----------------------------
1120 void ADLParser::frame_pointer_parse(FrameForm *frame, bool native) {
1121 char *frame_pointer = parse_one_arg("frame pointer entry");
1122 // Assign value into frame form
1123 if (native) { frame->_c_frame_pointer = frame_pointer; }
1124 else { frame->_frame_pointer = frame_pointer; }
1125 }
1127 //------------------------------interpreter_frame_pointer_parse----------------------------
1128 void ADLParser::interpreter_frame_pointer_parse(FrameForm *frame, bool native) {
1129 frame->_interpreter_frame_pointer_reg = parse_one_arg("interpreter frame pointer entry");
1130 }
1132 //------------------------------inline_cache_parse-----------------------------
1133 void ADLParser::inline_cache_parse(FrameForm *frame, bool native) {
1134 frame->_inline_cache_reg = parse_one_arg("inline cache reg entry");
1135 }
1137 //------------------------------interpreter_method_oop_parse------------------
1138 void ADLParser::interpreter_method_oop_parse(FrameForm *frame, bool native) {
1139 frame->_interpreter_method_oop_reg = parse_one_arg("method oop reg entry");
1140 }
1142 //------------------------------cisc_spilling_operand_parse---------------------
1143 void ADLParser::cisc_spilling_operand_name_parse(FrameForm *frame, bool native) {
1144 frame->_cisc_spilling_operand_name = parse_one_arg("cisc spilling operand name");
1145 }
1147 //------------------------------stack_alignment_parse--------------------------
1148 void ADLParser::stack_alignment_parse(FrameForm *frame) {
1149 char *alignment = parse_one_arg("stack alignment entry");
1150 // Assign value into frame
1151 frame->_alignment = alignment;
1152 }
1154 //------------------------------parse_one_arg-------------------------------
1155 char *ADLParser::parse_one_arg(const char *description) {
1156 char *token = NULL;
1157 if(_curchar == '(') {
1158 next_char();
1159 skipws();
1160 token = get_expr(description, ")");
1161 if (token == NULL) {
1162 parse_err(SYNERR, "missing value inside %s.\n", description);
1163 return NULL;
1164 }
1165 next_char(); // skip the close paren
1166 if(_curchar != ';') { // check for semi-colon
1167 parse_err(SYNERR, "missing %c in.\n", ';', description);
1168 return NULL;
1169 }
1170 next_char(); // skip the semi-colon
1171 }
1172 else {
1173 parse_err(SYNERR, "Missing %c in.\n", '(', description);
1174 return NULL;
1175 }
1177 trim(token);
1178 return token;
1179 }
1181 //------------------------------return_addr_parse------------------------------
1182 void ADLParser::return_addr_parse(FrameForm *frame, bool native) {
1183 bool in_register = true;
1184 if(_curchar == '(') {
1185 next_char();
1186 skipws();
1187 char *token = get_ident();
1188 if (token == NULL) {
1189 parse_err(SYNERR, "missing value inside return address entry.\n");
1190 return;
1191 }
1192 // check for valid values for stack/register
1193 if (strcmp(token, "REG") == 0) {
1194 in_register = true;
1195 }
1196 else if (strcmp(token, "STACK") == 0) {
1197 in_register = false;
1198 }
1199 else {
1200 parse_err(SYNERR, "invalid value inside return_address entry.\n");
1201 return;
1202 }
1203 if (native) { frame->_c_return_addr_loc = in_register; }
1204 else { frame->_return_addr_loc = in_register; }
1206 // Parse expression that specifies register or stack position
1207 skipws();
1208 char *token2 = get_expr("return address entry", ")");
1209 if (token2 == NULL) {
1210 parse_err(SYNERR, "missing value inside return address entry.\n");
1211 return;
1212 }
1213 next_char(); // skip the close paren
1214 if (native) { frame->_c_return_addr = token2; }
1215 else { frame->_return_addr = token2; }
1217 if(_curchar != ';') { // check for semi-colon
1218 parse_err(SYNERR, "missing %c in return address entry.\n", ';');
1219 return;
1220 }
1221 next_char(); // skip the semi-colon
1222 }
1223 else {
1224 parse_err(SYNERR, "Missing %c in return_address entry.\n", '(');
1225 }
1226 }
1228 //------------------------------preserve_stack_parse---------------------------
1229 void ADLParser::preserve_stack_parse(FrameForm *frame) {
1230 if(_curchar == '(') {
1231 char *token = get_paren_expr("preserve_stack_slots");
1232 frame->_in_preserve_slots = token;
1234 if(_curchar != ';') { // check for semi-colon
1235 parse_err(SYNERR, "missing %c in preserve stack slot entry.\n", ';');
1236 return;
1237 }
1238 next_char(); // skip the semi-colon
1239 }
1240 else {
1241 parse_err(SYNERR, "Missing %c in preserve stack slot entry.\n", '(');
1242 }
1243 }
1245 //------------------------------calling_convention_parse-----------------------
1246 char *ADLParser::calling_convention_parse() {
1247 char *desc = NULL; // String representation of calling_convention
1249 skipws(); // Skip leading whitespace
1250 if ( (desc = find_cpp_block("calling convention block")) == NULL ) {
1251 parse_err(SYNERR, "incorrect or missing block for 'calling_convention'.\n");
1252 }
1253 return desc;
1254 }
1256 //------------------------------return_value_parse-----------------------------
1257 char *ADLParser::return_value_parse() {
1258 char *desc = NULL; // String representation of calling_convention
1260 skipws(); // Skip leading whitespace
1261 if ( (desc = find_cpp_block("return value block")) == NULL ) {
1262 parse_err(SYNERR, "incorrect or missing block for 'return_value'.\n");
1263 }
1264 return desc;
1265 }
1267 //------------------------------ins_pipe_parse---------------------------------
1268 void ADLParser::ins_pipe_parse(InstructForm &instr) {
1269 char * ident;
1271 skipws();
1272 if ( _curchar != '(' ) { // Check for delimiter
1273 parse_err(SYNERR, "missing \"(\" in ins_pipe definition\n");
1274 return;
1275 }
1277 next_char();
1278 ident = get_ident(); // Grab next identifier
1280 if (ident == NULL) {
1281 parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
1282 return;
1283 }
1285 skipws();
1286 if ( _curchar != ')' ) { // Check for delimiter
1287 parse_err(SYNERR, "missing \")\" in ins_pipe definition\n");
1288 return;
1289 }
1291 next_char(); // skip the close paren
1292 if(_curchar != ';') { // check for semi-colon
1293 parse_err(SYNERR, "missing %c in return value entry.\n", ';');
1294 return;
1295 }
1296 next_char(); // skip the semi-colon
1298 // Check ident for validity
1299 if (_AD._pipeline && !_AD._pipeline->_classlist.search(ident)) {
1300 parse_err(SYNERR, "\"%s\" is not a valid pipeline class\n", ident);
1301 return;
1302 }
1304 // Add this instruction to the list in the pipeline class
1305 _AD._pipeline->_classdict[ident]->is_pipeclass()->_instructs.addName(instr._ident);
1307 // Set the name of the pipeline class in the instruction
1308 instr._ins_pipe = ident;
1309 return;
1310 }
1312 //------------------------------pipe_parse-------------------------------------
1313 void ADLParser::pipe_parse(void) {
1314 PipelineForm *pipeline; // Encode class for instruction/operand
1315 char * ident;
1317 pipeline = new PipelineForm(); // Build new Source object
1318 _AD.addForm(pipeline);
1320 skipws(); // Skip leading whitespace
1321 // Check for block delimiter
1322 if ( (_curchar != '%')
1323 || ( next_char(), (_curchar != '{')) ) {
1324 parse_err(SYNERR, "missing '%{' in pipeline definition\n");
1325 return;
1326 }
1327 next_char(); // Maintain the invariant
1328 do {
1329 ident = get_ident(); // Grab next identifier
1330 if (ident == NULL) {
1331 parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
1332 continue;
1333 }
1334 if (!strcmp(ident, "resources" )) resource_parse(*pipeline);
1335 else if (!strcmp(ident, "pipe_desc" )) pipe_desc_parse(*pipeline);
1336 else if (!strcmp(ident, "pipe_class")) pipe_class_parse(*pipeline);
1337 else if (!strcmp(ident, "define")) {
1338 skipws();
1339 if ( (_curchar != '%')
1340 || ( next_char(), (_curchar != '{')) ) {
1341 parse_err(SYNERR, "expected '%{'\n");
1342 return;
1343 }
1344 next_char(); skipws();
1346 char *node_class = get_ident();
1347 if (node_class == NULL) {
1348 parse_err(SYNERR, "expected identifier, found \"%c\"\n", _curchar);
1349 return;
1350 }
1352 skipws();
1353 if (_curchar != ',' && _curchar != '=') {
1354 parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1355 break;
1356 }
1357 next_char(); skipws();
1359 char *pipe_class = get_ident();
1360 if (pipe_class == NULL) {
1361 parse_err(SYNERR, "expected identifier, found \"%c\"\n", _curchar);
1362 return;
1363 }
1364 if (_curchar != ';' ) {
1365 parse_err(SYNERR, "expected `;`, found '%c'\n", _curchar);
1366 break;
1367 }
1368 next_char(); // Skip over semi-colon
1370 skipws();
1371 if ( (_curchar != '%')
1372 || ( next_char(), (_curchar != '}')) ) {
1373 parse_err(SYNERR, "expected '%%}', found \"%c\"\n", _curchar);
1374 }
1375 next_char();
1377 // Check ident for validity
1378 if (_AD._pipeline && !_AD._pipeline->_classlist.search(pipe_class)) {
1379 parse_err(SYNERR, "\"%s\" is not a valid pipeline class\n", pipe_class);
1380 return;
1381 }
1383 // Add this machine node to the list in the pipeline class
1384 _AD._pipeline->_classdict[pipe_class]->is_pipeclass()->_instructs.addName(node_class);
1386 MachNodeForm *machnode = new MachNodeForm(node_class); // Create new machnode form
1387 machnode->_machnode_pipe = pipe_class;
1389 _AD.addForm(machnode);
1390 }
1391 else if (!strcmp(ident, "attributes")) {
1392 bool vsi_seen = false, bhds_seen = false;
1394 skipws();
1395 if ( (_curchar != '%')
1396 || ( next_char(), (_curchar != '{')) ) {
1397 parse_err(SYNERR, "expected '%{'\n");
1398 return;
1399 }
1400 next_char(); skipws();
1402 while (_curchar != '%') {
1403 ident = get_ident();
1404 if (ident == NULL)
1405 break;
1407 if (!strcmp(ident, "variable_size_instructions")) {
1408 skipws();
1409 if (_curchar == ';') {
1410 next_char(); skipws();
1411 }
1413 pipeline->_variableSizeInstrs = true;
1414 vsi_seen = true;
1415 continue;
1416 }
1418 if (!strcmp(ident, "fixed_size_instructions")) {
1419 skipws();
1420 if (_curchar == ';') {
1421 next_char(); skipws();
1422 }
1424 pipeline->_variableSizeInstrs = false;
1425 vsi_seen = true;
1426 continue;
1427 }
1429 if (!strcmp(ident, "branch_has_delay_slot")) {
1430 skipws();
1431 if (_curchar == ';') {
1432 next_char(); skipws();
1433 }
1435 pipeline->_branchHasDelaySlot = true;
1436 bhds_seen = true;
1437 continue;
1438 }
1440 if (!strcmp(ident, "max_instructions_per_bundle")) {
1441 skipws();
1442 if (_curchar != '=') {
1443 parse_err(SYNERR, "expected `=`\n");
1444 break;
1445 }
1447 next_char(); skipws();
1448 pipeline->_maxInstrsPerBundle = get_int();
1449 skipws();
1451 if (_curchar == ';') {
1452 next_char(); skipws();
1453 }
1455 continue;
1456 }
1458 if (!strcmp(ident, "max_bundles_per_cycle")) {
1459 skipws();
1460 if (_curchar != '=') {
1461 parse_err(SYNERR, "expected `=`\n");
1462 break;
1463 }
1465 next_char(); skipws();
1466 pipeline->_maxBundlesPerCycle = get_int();
1467 skipws();
1469 if (_curchar == ';') {
1470 next_char(); skipws();
1471 }
1473 continue;
1474 }
1476 if (!strcmp(ident, "instruction_unit_size")) {
1477 skipws();
1478 if (_curchar != '=') {
1479 parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1480 break;
1481 }
1483 next_char(); skipws();
1484 pipeline->_instrUnitSize = get_int();
1485 skipws();
1487 if (_curchar == ';') {
1488 next_char(); skipws();
1489 }
1491 continue;
1492 }
1494 if (!strcmp(ident, "bundle_unit_size")) {
1495 skipws();
1496 if (_curchar != '=') {
1497 parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1498 break;
1499 }
1501 next_char(); skipws();
1502 pipeline->_bundleUnitSize = get_int();
1503 skipws();
1505 if (_curchar == ';') {
1506 next_char(); skipws();
1507 }
1509 continue;
1510 }
1512 if (!strcmp(ident, "instruction_fetch_unit_size")) {
1513 skipws();
1514 if (_curchar != '=') {
1515 parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1516 break;
1517 }
1519 next_char(); skipws();
1520 pipeline->_instrFetchUnitSize = get_int();
1521 skipws();
1523 if (_curchar == ';') {
1524 next_char(); skipws();
1525 }
1527 continue;
1528 }
1530 if (!strcmp(ident, "instruction_fetch_units")) {
1531 skipws();
1532 if (_curchar != '=') {
1533 parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1534 break;
1535 }
1537 next_char(); skipws();
1538 pipeline->_instrFetchUnits = get_int();
1539 skipws();
1541 if (_curchar == ';') {
1542 next_char(); skipws();
1543 }
1545 continue;
1546 }
1548 if (!strcmp(ident, "nops")) {
1549 skipws();
1550 if (_curchar != '(') {
1551 parse_err(SYNERR, "expected `(`, found '%c'\n", _curchar);
1552 break;
1553 }
1555 next_char(); skipws();
1557 while (_curchar != ')') {
1558 ident = get_ident();
1559 if (ident == NULL) {
1560 parse_err(SYNERR, "expected identifier for nop instruction, found '%c'\n", _curchar);
1561 break;
1562 }
1564 pipeline->_noplist.addName(ident);
1565 pipeline->_nopcnt++;
1566 skipws();
1568 if (_curchar == ',') {
1569 next_char(); skipws();
1570 }
1571 }
1573 next_char(); skipws();
1575 if (_curchar == ';') {
1576 next_char(); skipws();
1577 }
1579 continue;
1580 }
1582 parse_err(SYNERR, "unknown specifier \"%s\"\n", ident);
1583 }
1585 if ( (_curchar != '%')
1586 || ( next_char(), (_curchar != '}')) ) {
1587 parse_err(SYNERR, "expected '%}', found \"%c\"\n", _curchar);
1588 }
1589 next_char(); skipws();
1591 if (pipeline->_maxInstrsPerBundle == 0)
1592 parse_err(SYNERR, "\"max_instructions_per_bundle\" unspecified\n");
1593 if (pipeline->_instrUnitSize == 0 && pipeline->_bundleUnitSize == 0)
1594 parse_err(SYNERR, "\"instruction_unit_size\" and \"bundle_unit_size\" unspecified\n");
1595 if (pipeline->_instrFetchUnitSize == 0)
1596 parse_err(SYNERR, "\"instruction_fetch_unit_size\" unspecified\n");
1597 if (pipeline->_instrFetchUnits == 0)
1598 parse_err(SYNERR, "\"instruction_fetch_units\" unspecified\n");
1599 if (!vsi_seen)
1600 parse_err(SYNERR, "\"variable_size_instruction\" or \"fixed_size_instruction\" unspecified\n");
1601 }
1602 else { // Done with staticly defined parts of instruction definition
1603 parse_err(SYNERR, "expected one of \"resources\", \"pipe_desc\", \"pipe_class\", found \"%s\"\n", ident);
1604 return;
1605 }
1606 skipws();
1607 if (_curchar == ';')
1608 skipws();
1609 } while(_curchar != '%');
1611 next_char();
1612 if (_curchar != '}') {
1613 parse_err(SYNERR, "missing \"%}\" in pipeline definition\n");
1614 return;
1615 }
1617 next_char();
1618 }
1620 //------------------------------resource_parse----------------------------
1621 void ADLParser::resource_parse(PipelineForm &pipeline) {
1622 ResourceForm *resource;
1623 char * ident;
1624 char * expr;
1625 unsigned mask;
1626 pipeline._rescount = 0;
1628 skipws(); // Skip leading whitespace
1630 if (_curchar != '(') {
1631 parse_err(SYNERR, "missing \"(\" in resource definition\n");
1632 return;
1633 }
1635 do {
1636 next_char(); // Skip "(" or ","
1637 ident = get_ident(); // Grab next identifier
1639 if (ident == NULL) {
1640 parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1641 return;
1642 }
1643 skipws();
1645 if (_curchar != '=') {
1646 mask = (1 << pipeline._rescount++);
1647 }
1648 else {
1649 next_char(); skipws();
1650 expr = get_ident(); // Grab next identifier
1651 if (expr == NULL) {
1652 parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1653 return;
1654 }
1655 resource = (ResourceForm *) pipeline._resdict[expr];
1656 if (resource == NULL) {
1657 parse_err(SYNERR, "resource \"%s\" is not defined\n", expr);
1658 return;
1659 }
1660 mask = resource->mask();
1662 skipws();
1663 while (_curchar == '|') {
1664 next_char(); skipws();
1666 expr = get_ident(); // Grab next identifier
1667 if (expr == NULL) {
1668 parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1669 return;
1670 }
1672 resource = (ResourceForm *) pipeline._resdict[expr]; // Look up the value
1673 if (resource == NULL) {
1674 parse_err(SYNERR, "resource \"%s\" is not defined\n", expr);
1675 return;
1676 }
1678 mask |= resource->mask();
1679 skipws();
1680 }
1681 }
1683 resource = new ResourceForm(mask);
1685 pipeline._resdict.Insert(ident, resource);
1686 pipeline._reslist.addName(ident);
1687 } while (_curchar == ',');
1689 if (_curchar != ')') {
1690 parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
1691 return;
1692 }
1694 next_char(); // Skip ")"
1695 if (_curchar == ';')
1696 next_char(); // Skip ";"
1697 }
1699 //------------------------------resource_parse----------------------------
1700 void ADLParser::pipe_desc_parse(PipelineForm &pipeline) {
1701 char * ident;
1703 skipws(); // Skip leading whitespace
1705 if (_curchar != '(') {
1706 parse_err(SYNERR, "missing \"(\" in pipe_desc definition\n");
1707 return;
1708 }
1710 do {
1711 next_char(); // Skip "(" or ","
1712 ident = get_ident(); // Grab next identifier
1713 if (ident == NULL) {
1714 parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1715 return;
1716 }
1718 // Add the name to the list
1719 pipeline._stages.addName(ident);
1720 pipeline._stagecnt++;
1722 skipws();
1723 } while (_curchar == ',');
1725 if (_curchar != ')') {
1726 parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
1727 return;
1728 }
1730 next_char(); // Skip ")"
1731 if (_curchar == ';')
1732 next_char(); // Skip ";"
1733 }
1735 //------------------------------pipe_class_parse--------------------------
1736 void ADLParser::pipe_class_parse(PipelineForm &pipeline) {
1737 PipeClassForm *pipe_class;
1738 char * ident;
1739 char * stage;
1740 char * read_or_write;
1741 int is_write;
1742 int is_read;
1743 OperandForm *oper;
1745 skipws(); // Skip leading whitespace
1747 ident = get_ident(); // Grab next identifier
1749 if (ident == NULL) {
1750 parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1751 return;
1752 }
1754 // Create a record for the pipe_class
1755 pipe_class = new PipeClassForm(ident, ++pipeline._classcnt);
1756 pipeline._classdict.Insert(ident, pipe_class);
1757 pipeline._classlist.addName(ident);
1759 // Then get the operands
1760 skipws();
1761 if (_curchar != '(') {
1762 parse_err(SYNERR, "missing \"(\" in pipe_class definition\n");
1763 }
1764 // Parse the operand list
1765 else get_oplist(pipe_class->_parameters, pipe_class->_localNames);
1766 skipws(); // Skip leading whitespace
1767 // Check for block delimiter
1768 if ( (_curchar != '%')
1769 || ( next_char(), (_curchar != '{')) ) {
1770 parse_err(SYNERR, "missing \"%{\" in pipe_class definition\n");
1771 return;
1772 }
1773 next_char();
1775 do {
1776 ident = get_ident(); // Grab next identifier
1777 if (ident == NULL) {
1778 parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1779 continue;
1780 }
1781 skipws();
1783 if (!strcmp(ident, "fixed_latency")) {
1784 skipws();
1785 if (_curchar != '(') {
1786 parse_err(SYNERR, "missing \"(\" in latency definition\n");
1787 return;
1788 }
1789 next_char(); skipws();
1790 if( !isdigit(_curchar) ) {
1791 parse_err(SYNERR, "number expected for \"%c\" in latency definition\n", _curchar);
1792 return;
1793 }
1794 int fixed_latency = get_int();
1795 skipws();
1796 if (_curchar != ')') {
1797 parse_err(SYNERR, "missing \")\" in latency definition\n");
1798 return;
1799 }
1800 next_char(); skipws();
1801 if (_curchar != ';') {
1802 parse_err(SYNERR, "missing \";\" in latency definition\n");
1803 return;
1804 }
1806 pipe_class->setFixedLatency(fixed_latency);
1807 next_char(); skipws();
1808 continue;
1809 }
1811 if (!strcmp(ident, "zero_instructions") ||
1812 !strcmp(ident, "no_instructions")) {
1813 skipws();
1814 if (_curchar != ';') {
1815 parse_err(SYNERR, "missing \";\" in latency definition\n");
1816 return;
1817 }
1819 pipe_class->setInstructionCount(0);
1820 next_char(); skipws();
1821 continue;
1822 }
1824 if (!strcmp(ident, "one_instruction_with_delay_slot") ||
1825 !strcmp(ident, "single_instruction_with_delay_slot")) {
1826 skipws();
1827 if (_curchar != ';') {
1828 parse_err(SYNERR, "missing \";\" in latency definition\n");
1829 return;
1830 }
1832 pipe_class->setInstructionCount(1);
1833 pipe_class->setBranchDelay(true);
1834 next_char(); skipws();
1835 continue;
1836 }
1838 if (!strcmp(ident, "one_instruction") ||
1839 !strcmp(ident, "single_instruction")) {
1840 skipws();
1841 if (_curchar != ';') {
1842 parse_err(SYNERR, "missing \";\" in latency definition\n");
1843 return;
1844 }
1846 pipe_class->setInstructionCount(1);
1847 next_char(); skipws();
1848 continue;
1849 }
1851 if (!strcmp(ident, "instructions_in_first_bundle") ||
1852 !strcmp(ident, "instruction_count")) {
1853 skipws();
1855 int number_of_instructions = 1;
1857 if (_curchar != '(') {
1858 parse_err(SYNERR, "\"(\" expected at \"%c\"\n", _curchar);
1859 continue;
1860 }
1862 next_char(); skipws();
1863 number_of_instructions = get_int();
1865 skipws();
1866 if (_curchar != ')') {
1867 parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
1868 continue;
1869 }
1871 next_char(); skipws();
1872 if (_curchar != ';') {
1873 parse_err(SYNERR, "missing \";\" in latency definition\n");
1874 return;
1875 }
1877 pipe_class->setInstructionCount(number_of_instructions);
1878 next_char(); skipws();
1879 continue;
1880 }
1882 if (!strcmp(ident, "multiple_bundles")) {
1883 skipws();
1884 if (_curchar != ';') {
1885 parse_err(SYNERR, "missing \";\" after multiple bundles\n");
1886 return;
1887 }
1889 pipe_class->setMultipleBundles(true);
1890 next_char(); skipws();
1891 continue;
1892 }
1894 if (!strcmp(ident, "has_delay_slot")) {
1895 skipws();
1896 if (_curchar != ';') {
1897 parse_err(SYNERR, "missing \";\" after \"has_delay_slot\"\n");
1898 return;
1899 }
1901 pipe_class->setBranchDelay(true);
1902 next_char(); skipws();
1903 continue;
1904 }
1906 if (!strcmp(ident, "force_serialization")) {
1907 skipws();
1908 if (_curchar != ';') {
1909 parse_err(SYNERR, "missing \";\" after \"force_serialization\"\n");
1910 return;
1911 }
1913 pipe_class->setForceSerialization(true);
1914 next_char(); skipws();
1915 continue;
1916 }
1918 if (!strcmp(ident, "may_have_no_code")) {
1919 skipws();
1920 if (_curchar != ';') {
1921 parse_err(SYNERR, "missing \";\" after \"may_have_no_code\"\n");
1922 return;
1923 }
1925 pipe_class->setMayHaveNoCode(true);
1926 next_char(); skipws();
1927 continue;
1928 }
1930 const Form *parm = pipe_class->_localNames[ident];
1931 if (parm != NULL) {
1932 oper = parm->is_operand();
1933 if (oper == NULL && !parm->is_opclass()) {
1934 parse_err(SYNERR, "operand name expected at %s\n", ident);
1935 continue;
1936 }
1938 if (_curchar != ':') {
1939 parse_err(SYNERR, "\":\" expected at \"%c\"\n", _curchar);
1940 continue;
1941 }
1942 next_char(); skipws();
1943 stage = get_ident();
1944 if (stage == NULL) {
1945 parse_err(SYNERR, "pipeline stage identifier expected at \"%c\"\n", _curchar);
1946 continue;
1947 }
1949 skipws();
1950 if (_curchar != '(') {
1951 parse_err(SYNERR, "\"(\" expected at \"%c\"\n", _curchar);
1952 continue;
1953 }
1955 next_char();
1956 read_or_write = get_ident();
1957 if (read_or_write == NULL) {
1958 parse_err(SYNERR, "\"read\" or \"write\" expected at \"%c\"\n", _curchar);
1959 continue;
1960 }
1962 is_read = strcmp(read_or_write, "read") == 0;
1963 is_write = strcmp(read_or_write, "write") == 0;
1964 if (!is_read && !is_write) {
1965 parse_err(SYNERR, "\"read\" or \"write\" expected at \"%c\"\n", _curchar);
1966 continue;
1967 }
1969 skipws();
1970 if (_curchar != ')') {
1971 parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
1972 continue;
1973 }
1975 next_char(); skipws();
1976 int more_instrs = 0;
1977 if (_curchar == '+') {
1978 next_char(); skipws();
1979 if (_curchar < '0' || _curchar > '9') {
1980 parse_err(SYNERR, "<number> expected at \"%c\"\n", _curchar);
1981 continue;
1982 }
1983 while (_curchar >= '0' && _curchar <= '9') {
1984 more_instrs *= 10;
1985 more_instrs += _curchar - '0';
1986 next_char();
1987 }
1988 skipws();
1989 }
1991 PipeClassOperandForm *pipe_operand = new PipeClassOperandForm(stage, is_write, more_instrs);
1992 pipe_class->_localUsage.Insert(ident, pipe_operand);
1994 if (_curchar == '%')
1995 continue;
1997 if (_curchar != ';') {
1998 parse_err(SYNERR, "\";\" expected at \"%c\"\n", _curchar);
1999 continue;
2000 }
2001 next_char(); skipws();
2002 continue;
2003 }
2005 // Scan for Resource Specifier
2006 const Form *res = pipeline._resdict[ident];
2007 if (res != NULL) {
2008 int cyclecnt = 1;
2009 if (_curchar != ':') {
2010 parse_err(SYNERR, "\":\" expected at \"%c\"\n", _curchar);
2011 continue;
2012 }
2013 next_char(); skipws();
2014 stage = get_ident();
2015 if (stage == NULL) {
2016 parse_err(SYNERR, "pipeline stage identifier expected at \"%c\"\n", _curchar);
2017 continue;
2018 }
2020 skipws();
2021 if (_curchar == '(') {
2022 next_char();
2023 cyclecnt = get_int();
2025 skipws();
2026 if (_curchar != ')') {
2027 parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
2028 continue;
2029 }
2031 next_char(); skipws();
2032 }
2034 PipeClassResourceForm *resource = new PipeClassResourceForm(ident, stage, cyclecnt);
2035 int stagenum = pipeline._stages.index(stage);
2036 if (pipeline._maxcycleused < (stagenum+cyclecnt))
2037 pipeline._maxcycleused = (stagenum+cyclecnt);
2038 pipe_class->_resUsage.addForm(resource);
2040 if (_curchar == '%')
2041 continue;
2043 if (_curchar != ';') {
2044 parse_err(SYNERR, "\";\" expected at \"%c\"\n", _curchar);
2045 continue;
2046 }
2047 next_char(); skipws();
2048 continue;
2049 }
2051 parse_err(SYNERR, "resource expected at \"%s\"\n", ident);
2052 return;
2053 } while(_curchar != '%');
2055 next_char();
2056 if (_curchar != '}') {
2057 parse_err(SYNERR, "missing \"%}\" in pipe_class definition\n");
2058 return;
2059 }
2061 next_char();
2062 }
2064 //------------------------------peep_parse-------------------------------------
2065 void ADLParser::peep_parse(void) {
2066 Peephole *peep; // Pointer to current peephole rule form
2067 char *desc = NULL; // String representation of rule
2069 skipws(); // Skip leading whitespace
2071 peep = new Peephole(); // Build new Peephole object
2072 // Check for open block sequence
2073 skipws(); // Skip leading whitespace
2074 if (_curchar == '%' && *(_ptr+1) == '{') {
2075 next_char(); next_char(); // Skip "%{"
2076 skipws();
2077 while (_curchar != '%' && *(_ptr+1) != '}') {
2078 char *token = get_ident();
2079 if (token == NULL) {
2080 parse_err(SYNERR, "missing identifier inside peephole rule.\n");
2081 return;
2082 }
2083 // check for legal subsections of peephole rule
2084 if (strcmp(token,"peepmatch")==0) {
2085 peep_match_parse(*peep); }
2086 else if (strcmp(token,"peepconstraint")==0) {
2087 peep_constraint_parse(*peep); }
2088 else if (strcmp(token,"peepreplace")==0) {
2089 peep_replace_parse(*peep); }
2090 else {
2091 parse_err(SYNERR, "expected peepmatch, peepconstraint, or peepreplace for identifier %s.\n", token);
2092 }
2093 skipws();
2094 }
2095 }
2096 else {
2097 parse_err(SYNERR, "Missing %%{ ... %%} block after peephole keyword.\n");
2098 return;
2099 }
2100 next_char(); // Skip past '%'
2101 next_char(); // Skip past '}'
2102 }
2104 // ******************** Private Level 2 Parse Functions ********************
2105 //------------------------------constraint_parse------------------------------
2106 Constraint *ADLParser::constraint_parse(void) {
2107 char *func;
2108 char *arg;
2110 // Check for constraint expression
2111 skipws();
2112 if (_curchar != '(') {
2113 parse_err(SYNERR, "missing constraint expression, (...)\n");
2114 return NULL;
2115 }
2116 next_char(); // Skip past '('
2118 // Get constraint function
2119 skipws();
2120 func = get_ident();
2121 if (func == NULL) {
2122 parse_err(SYNERR, "missing function in constraint expression.\n");
2123 return NULL;
2124 }
2125 if (strcmp(func,"ALLOC_IN_RC")==0
2126 || strcmp(func,"IS_R_CLASS")==0) {
2127 // Check for '(' before argument
2128 skipws();
2129 if (_curchar != '(') {
2130 parse_err(SYNERR, "missing '(' for constraint function's argument.\n");
2131 return NULL;
2132 }
2133 next_char();
2135 // Get it's argument
2136 skipws();
2137 arg = get_ident();
2138 if (arg == NULL) {
2139 parse_err(SYNERR, "missing argument for constraint function %s\n",func);
2140 return NULL;
2141 }
2142 // Check for ')' after argument
2143 skipws();
2144 if (_curchar != ')') {
2145 parse_err(SYNERR, "missing ')' after constraint function argument %s\n",arg);
2146 return NULL;
2147 }
2148 next_char();
2149 } else {
2150 parse_err(SYNERR, "Invalid constraint function %s\n",func);
2151 return NULL;
2152 }
2154 // Check for closing paren and ';'
2155 skipws();
2156 if (_curchar != ')') {
2157 parse_err(SYNERR, "Missing ')' for constraint function %s\n",func);
2158 return NULL;
2159 }
2160 next_char();
2161 skipws();
2162 if (_curchar != ';') {
2163 parse_err(SYNERR, "Missing ';' after constraint.\n");
2164 return NULL;
2165 }
2166 next_char();
2168 // Create new "Constraint"
2169 Constraint *constraint = new Constraint(func,arg);
2170 return constraint;
2171 }
2173 //------------------------------constr_parse-----------------------------------
2174 ConstructRule *ADLParser::construct_parse(void) {
2175 return NULL;
2176 }
2179 //------------------------------reg_def_parse----------------------------------
2180 void ADLParser::reg_def_parse(void) {
2181 char *rname; // Name of register being defined
2183 // Get register name
2184 skipws(); // Skip whitespace
2185 rname = get_ident();
2186 if (rname == NULL) {
2187 parse_err(SYNERR, "missing register name after reg_def\n");
2188 return;
2189 }
2191 // Check for definition of register calling convention (save on call, ...),
2192 // register save type, and register encoding value.
2193 skipws();
2194 char *callconv = NULL;
2195 char *c_conv = NULL;
2196 char *idealtype = NULL;
2197 char *encoding = NULL;
2198 char *concrete = NULL;
2199 if (_curchar == '(') {
2200 next_char();
2201 callconv = get_ident();
2202 // Parse the internal calling convention, must be NS, SOC, SOE, or AS.
2203 if (callconv == NULL) {
2204 parse_err(SYNERR, "missing register calling convention value\n");
2205 return;
2206 }
2207 if(strcmp(callconv, "SOC") && strcmp(callconv,"SOE") &&
2208 strcmp(callconv, "NS") && strcmp(callconv, "AS")) {
2209 parse_err(SYNERR, "invalid value for register calling convention\n");
2210 }
2211 skipws();
2212 if (_curchar != ',') {
2213 parse_err(SYNERR, "missing comma in register definition statement\n");
2214 return;
2215 }
2216 next_char();
2218 // Parse the native calling convention, must be NS, SOC, SOE, AS
2219 c_conv = get_ident();
2220 if (c_conv == NULL) {
2221 parse_err(SYNERR, "missing register native calling convention value\n");
2222 return;
2223 }
2224 if(strcmp(c_conv, "SOC") && strcmp(c_conv,"SOE") &&
2225 strcmp(c_conv, "NS") && strcmp(c_conv, "AS")) {
2226 parse_err(SYNERR, "invalid value for register calling convention\n");
2227 }
2228 skipws();
2229 if (_curchar != ',') {
2230 parse_err(SYNERR, "missing comma in register definition statement\n");
2231 return;
2232 }
2233 next_char();
2234 skipws();
2236 // Parse the ideal save type
2237 idealtype = get_ident();
2238 if (idealtype == NULL) {
2239 parse_err(SYNERR, "missing register save type value\n");
2240 return;
2241 }
2242 skipws();
2243 if (_curchar != ',') {
2244 parse_err(SYNERR, "missing comma in register definition statement\n");
2245 return;
2246 }
2247 next_char();
2248 skipws();
2250 // Parse the encoding value
2251 encoding = get_expr("encoding", ",");
2252 if (encoding == NULL) {
2253 parse_err(SYNERR, "missing register encoding value\n");
2254 return;
2255 }
2256 trim(encoding);
2257 if (_curchar != ',') {
2258 parse_err(SYNERR, "missing comma in register definition statement\n");
2259 return;
2260 }
2261 next_char();
2262 skipws();
2263 // Parse the concrete name type
2264 // concrete = get_ident();
2265 concrete = get_expr("concrete", ")");
2266 if (concrete == NULL) {
2267 parse_err(SYNERR, "missing vm register name value\n");
2268 return;
2269 }
2271 if (_curchar != ')') {
2272 parse_err(SYNERR, "missing ')' in register definition statement\n");
2273 return;
2274 }
2275 next_char();
2276 }
2278 // Check for closing ';'
2279 skipws();
2280 if (_curchar != ';') {
2281 parse_err(SYNERR, "missing ';' after reg_def\n");
2282 return;
2283 }
2284 next_char(); // move past ';'
2286 // Debug Stuff
2287 if (_AD._adl_debug > 1) {
2288 fprintf(stderr,"Register Definition: %s ( %s, %s %s )\n", rname,
2289 (callconv ? callconv : ""), (c_conv ? c_conv : ""), concrete);
2290 }
2292 // Record new register definition.
2293 _AD._register->addRegDef(rname, callconv, c_conv, idealtype, encoding, concrete);
2294 return;
2295 }
2297 //------------------------------reg_class_parse--------------------------------
2298 void ADLParser::reg_class_parse(void) {
2299 char *cname; // Name of register class being defined
2301 // Get register class name
2302 skipws(); // Skip leading whitespace
2303 cname = get_ident();
2304 if (cname == NULL) {
2305 parse_err(SYNERR, "missing register class name after 'reg_class'\n");
2306 return;
2307 }
2308 // Debug Stuff
2309 if (_AD._adl_debug >1) fprintf(stderr,"Register Class: %s\n", cname);
2311 RegClass *reg_class = _AD._register->addRegClass(cname);
2313 // Collect registers in class
2314 skipws();
2315 if (_curchar == '(') {
2316 next_char(); // Skip '('
2317 skipws();
2318 while (_curchar != ')') {
2319 char *rname = get_ident();
2320 if (rname==NULL) {
2321 parse_err(SYNERR, "missing identifier inside reg_class list.\n");
2322 return;
2323 }
2324 RegDef *regDef = _AD._register->getRegDef(rname);
2325 if (!regDef) {
2326 parse_err(SEMERR, "unknown identifier %s inside reg_class list.\n", rname);
2327 } else {
2328 reg_class->addReg(regDef); // add regDef to regClass
2329 }
2331 // Check for ',' and position to next token.
2332 skipws();
2333 if (_curchar == ',') {
2334 next_char(); // Skip trailing ','
2335 skipws();
2336 }
2337 }
2338 next_char(); // Skip closing ')'
2339 } else if (_curchar == '%') {
2340 char *code = find_cpp_block("reg class");
2341 if (code == NULL) {
2342 parse_err(SYNERR, "missing code declaration for reg class.\n");
2343 return;
2344 }
2345 reg_class->_user_defined = code;
2346 return;
2347 }
2349 // Check for terminating ';'
2350 skipws();
2351 if (_curchar != ';') {
2352 parse_err(SYNERR, "missing ';' at end of reg_class definition.\n");
2353 return;
2354 }
2355 next_char(); // Skip trailing ';'
2357 // Check RegClass size, must be <= 32 registers in class.
2359 return;
2360 }
2362 //------------------------------alloc_class_parse------------------------------
2363 void ADLParser::alloc_class_parse(void) {
2364 char *name; // Name of allocation class being defined
2366 // Get allocation class name
2367 skipws(); // Skip leading whitespace
2368 name = get_ident();
2369 if (name == NULL) {
2370 parse_err(SYNERR, "missing allocation class name after 'reg_class'\n");
2371 return;
2372 }
2373 // Debug Stuff
2374 if (_AD._adl_debug >1) fprintf(stderr,"Allocation Class: %s\n", name);
2376 AllocClass *alloc_class = _AD._register->addAllocClass(name);
2378 // Collect registers in class
2379 skipws();
2380 if (_curchar == '(') {
2381 next_char(); // Skip '('
2382 skipws();
2383 while (_curchar != ')') {
2384 char *rname = get_ident();
2385 if (rname==NULL) {
2386 parse_err(SYNERR, "missing identifier inside reg_class list.\n");
2387 return;
2388 }
2389 // Check if name is a RegDef
2390 RegDef *regDef = _AD._register->getRegDef(rname);
2391 if (regDef) {
2392 alloc_class->addReg(regDef); // add regDef to allocClass
2393 } else {
2395 // name must be a RegDef or a RegClass
2396 parse_err(SYNERR, "name %s should be a previously defined reg_def.\n", rname);
2397 return;
2398 }
2400 // Check for ',' and position to next token.
2401 skipws();
2402 if (_curchar == ',') {
2403 next_char(); // Skip trailing ','
2404 skipws();
2405 }
2406 }
2407 next_char(); // Skip closing ')'
2408 }
2410 // Check for terminating ';'
2411 skipws();
2412 if (_curchar != ';') {
2413 parse_err(SYNERR, "missing ';' at end of reg_class definition.\n");
2414 return;
2415 }
2416 next_char(); // Skip trailing ';'
2418 return;
2419 }
2421 //------------------------------peep_match_child_parse-------------------------
2422 InstructForm *ADLParser::peep_match_child_parse(PeepMatch &match, int parent, int &position, int input){
2423 char *token = NULL;
2424 int lparen = 0; // keep track of parenthesis nesting depth
2425 int rparen = 0; // position of instruction at this depth
2426 InstructForm *inst_seen = NULL;
2427 InstructForm *child_seen = NULL;
2429 // Walk the match tree,
2430 // Record <parent, position, instruction name, input position>
2431 while ( lparen >= rparen ) {
2432 skipws();
2433 // Left paren signals start of an input, collect with recursive call
2434 if (_curchar == '(') {
2435 ++lparen;
2436 next_char();
2437 child_seen = peep_match_child_parse(match, parent, position, rparen);
2438 }
2439 // Right paren signals end of an input, may be more
2440 else if (_curchar == ')') {
2441 ++rparen;
2442 if( rparen == lparen ) { // IF rparen matches an lparen I've seen
2443 next_char(); // move past ')'
2444 } else { // ELSE leave ')' for parent
2445 assert( rparen == lparen + 1, "Should only see one extra ')'");
2446 // if an instruction was not specified for this paren-pair
2447 if( ! inst_seen ) { // record signal entry
2448 match.add_instruction( parent, position, NameList::_signal, input );
2449 ++position;
2450 }
2451 // ++input; // TEMPORARY
2452 return inst_seen;
2453 }
2454 }
2455 // if no parens, then check for instruction name
2456 // This instruction is the parent of a sub-tree
2457 else if ((token = get_ident_dup()) != NULL) {
2458 const Form *form = _AD._globalNames[token];
2459 if (form) {
2460 InstructForm *inst = form->is_instruction();
2461 // Record the first instruction at this level
2462 if( inst_seen == NULL ) {
2463 inst_seen = inst;
2464 }
2465 if (inst) {
2466 match.add_instruction( parent, position, token, input );
2467 parent = position;
2468 ++position;
2469 } else {
2470 parse_err(SYNERR, "instruction name expected at identifier %s.\n",
2471 token);
2472 return inst_seen;
2473 }
2474 }
2475 else {
2476 parse_err(SYNERR, "missing identifier in peepmatch rule.\n");
2477 return NULL;
2478 }
2479 }
2480 else {
2481 parse_err(SYNERR, "missing identifier in peepmatch rule.\n");
2482 return NULL;
2483 }
2485 } // end while
2487 assert( false, "ShouldNotReachHere();");
2488 return NULL;
2489 }
2491 //------------------------------peep_match_parse-------------------------------
2492 // Syntax for a peepmatch rule
2493 //
2494 // peepmatch ( root_instr_name [(instruction subtree)] [,(instruction subtree)]* );
2495 //
2496 void ADLParser::peep_match_parse(Peephole &peep) {
2498 skipws();
2499 // Check the structure of the rule
2500 // Check for open paren
2501 if (_curchar != '(') {
2502 parse_err(SYNERR, "missing '(' at start of peepmatch rule.\n");
2503 return;
2504 }
2505 next_char(); // skip '('
2507 // Construct PeepMatch and parse the peepmatch rule.
2508 PeepMatch *match = new PeepMatch(_ptr);
2509 int parent = -1; // parent of root
2510 int position = 0; // zero-based positions
2511 int input = 0; // input position in parent's operands
2512 InstructForm *root= peep_match_child_parse( *match, parent, position, input);
2513 if( root == NULL ) {
2514 parse_err(SYNERR, "missing instruction-name at start of peepmatch.\n");
2515 return;
2516 }
2518 if( _curchar != ')' ) {
2519 parse_err(SYNERR, "missing ')' at end of peepmatch.\n");
2520 return;
2521 }
2522 next_char(); // skip ')'
2524 // Check for closing semicolon
2525 skipws();
2526 if( _curchar != ';' ) {
2527 parse_err(SYNERR, "missing ';' at end of peepmatch.\n");
2528 return;
2529 }
2530 next_char(); // skip ';'
2532 // Store match into peep, and store peep into instruction
2533 peep.add_match(match);
2534 root->append_peephole(&peep);
2535 }
2537 //------------------------------peep_constraint_parse--------------------------
2538 // Syntax for a peepconstraint rule
2539 // A parenthesized list of relations between operands in peepmatch subtree
2540 //
2541 // peepconstraint %{
2542 // (instruction_number.operand_name
2543 // relational_op
2544 // instruction_number.operand_name OR register_name
2545 // [, ...] );
2546 //
2547 // // instruction numbers are zero-based using topological order in peepmatch
2548 //
2549 void ADLParser::peep_constraint_parse(Peephole &peep) {
2551 skipws();
2552 // Check the structure of the rule
2553 // Check for open paren
2554 if (_curchar != '(') {
2555 parse_err(SYNERR, "missing '(' at start of peepconstraint rule.\n");
2556 return;
2557 }
2558 else {
2559 next_char(); // Skip '('
2560 }
2562 // Check for a constraint
2563 skipws();
2564 while( _curchar != ')' ) {
2565 // Get information on the left instruction and its operand
2566 // left-instructions's number
2567 int left_inst = get_int();
2568 // Left-instruction's operand
2569 skipws();
2570 if( _curchar != '.' ) {
2571 parse_err(SYNERR, "missing '.' in peepconstraint after instruction number.\n");
2572 return;
2573 }
2574 next_char(); // Skip '.'
2575 char *left_op = get_ident_dup();
2577 skipws();
2578 // Collect relational operator
2579 char *relation = get_relation_dup();
2581 skipws();
2582 // Get information on the right instruction and its operand
2583 int right_inst; // Right-instructions's number
2584 if( isdigit(_curchar) ) {
2585 right_inst = get_int();
2586 // Right-instruction's operand
2587 skipws();
2588 if( _curchar != '.' ) {
2589 parse_err(SYNERR, "missing '.' in peepconstraint after instruction number.\n");
2590 return;
2591 }
2592 next_char(); // Skip '.'
2593 } else {
2594 right_inst = -1; // Flag as being a register constraint
2595 }
2597 char *right_op = get_ident_dup();
2599 // Construct the next PeepConstraint
2600 PeepConstraint *constraint = new PeepConstraint( left_inst, left_op,
2601 relation,
2602 right_inst, right_op );
2603 // And append it to the list for this peephole rule
2604 peep.append_constraint( constraint );
2606 // Check for another constraint, or end of rule
2607 skipws();
2608 if( _curchar == ',' ) {
2609 next_char(); // Skip ','
2610 skipws();
2611 }
2612 else if( _curchar != ')' ) {
2613 parse_err(SYNERR, "expected ',' or ')' after peephole constraint.\n");
2614 return;
2615 }
2616 } // end while( processing constraints )
2617 next_char(); // Skip ')'
2619 // Check for terminating ';'
2620 skipws();
2621 if (_curchar != ';') {
2622 parse_err(SYNERR, "missing ';' at end of peepconstraint.\n");
2623 return;
2624 }
2625 next_char(); // Skip trailing ';'
2626 }
2629 //------------------------------peep_replace_parse-----------------------------
2630 // Syntax for a peepreplace rule
2631 // root instruction name followed by a
2632 // parenthesized list of whitespace separated instruction.operand specifiers
2633 //
2634 // peepreplace ( instr_name ( [instruction_number.operand_name]* ) );
2635 //
2636 //
2637 void ADLParser::peep_replace_parse(Peephole &peep) {
2638 int lparen = 0; // keep track of parenthesis nesting depth
2639 int rparen = 0; // keep track of parenthesis nesting depth
2640 int icount = 0; // count of instructions in rule for naming
2641 char *str = NULL;
2642 char *token = NULL;
2644 skipws();
2645 // Check for open paren
2646 if (_curchar != '(') {
2647 parse_err(SYNERR, "missing '(' at start of peepreplace rule.\n");
2648 return;
2649 }
2650 else {
2651 lparen++;
2652 next_char();
2653 }
2655 // Check for root instruction
2656 char *inst = get_ident_dup();
2657 const Form *form = _AD._globalNames[inst];
2658 if( form == NULL || form->is_instruction() == NULL ) {
2659 parse_err(SYNERR, "Instruction name expected at start of peepreplace.\n");
2660 return;
2661 }
2663 // Store string representation of rule into replace
2664 PeepReplace *replace = new PeepReplace(str);
2665 replace->add_instruction( inst );
2667 skipws();
2668 // Start of root's operand-list
2669 if (_curchar != '(') {
2670 parse_err(SYNERR, "missing '(' at peepreplace root's operand-list.\n");
2671 return;
2672 }
2673 else {
2674 lparen++;
2675 next_char();
2676 }
2678 skipws();
2679 // Get the list of operands
2680 while( _curchar != ')' ) {
2681 // Get information on an instruction and its operand
2682 // instructions's number
2683 int inst_num = get_int();
2684 // Left-instruction's operand
2685 skipws();
2686 if( _curchar != '.' ) {
2687 parse_err(SYNERR, "missing '.' in peepreplace after instruction number.\n");
2688 return;
2689 }
2690 next_char(); // Skip '.'
2691 char *inst_op = get_ident_dup();
2692 if( inst_op == NULL ) {
2693 parse_err(SYNERR, "missing operand identifier in peepreplace.\n");
2694 return;
2695 }
2697 // Record this operand's position in peepmatch
2698 replace->add_operand( inst_num, inst_op );
2699 skipws();
2700 }
2702 // Check for the end of operands list
2703 skipws();
2704 assert( _curchar == ')', "While loop should have advanced to ')'.");
2705 next_char(); // Skip ')'
2707 skipws();
2708 // Check for end of peepreplace
2709 if( _curchar != ')' ) {
2710 parse_err(SYNERR, "missing ')' at end of peepmatch.\n");
2711 parse_err(SYNERR, "Support one replacement instruction.\n");
2712 return;
2713 }
2714 next_char(); // Skip ')'
2716 // Check for closing semicolon
2717 skipws();
2718 if( _curchar != ';' ) {
2719 parse_err(SYNERR, "missing ';' at end of peepreplace.\n");
2720 return;
2721 }
2722 next_char(); // skip ';'
2724 // Store replace into peep
2725 peep.add_replace( replace );
2726 }
2728 //------------------------------pred_parse-------------------------------------
2729 Predicate *ADLParser::pred_parse(void) {
2730 Predicate *predicate; // Predicate class for operand
2731 char *rule = NULL; // String representation of predicate
2733 skipws(); // Skip leading whitespace
2734 int line = linenum();
2735 if ( (rule = get_paren_expr("pred expression", true)) == NULL ) {
2736 parse_err(SYNERR, "incorrect or missing expression for 'predicate'\n");
2737 return NULL;
2738 }
2739 // Debug Stuff
2740 if (_AD._adl_debug > 1) fprintf(stderr,"Predicate: %s\n", rule);
2741 if (_curchar != ';') {
2742 parse_err(SYNERR, "missing ';' in predicate definition\n");
2743 return NULL;
2744 }
2745 next_char(); // Point after the terminator
2747 predicate = new Predicate(rule); // Build new predicate object
2748 skipws();
2749 return predicate;
2750 }
2753 //------------------------------ins_encode_parse_block-------------------------
2754 // Parse the block form of ins_encode. See ins_encode_parse for more details
2755 void ADLParser::ins_encode_parse_block(InstructForm& inst) {
2756 // Create a new encoding name based on the name of the instruction
2757 // definition, which should be unique.
2758 const char* prefix = "__ins_encode_";
2759 char* ec_name = (char*) malloc(strlen(inst._ident) + strlen(prefix) + 1);
2760 sprintf(ec_name, "%s%s", prefix, inst._ident);
2762 assert(_AD._encode->encClass(ec_name) == NULL, "shouldn't already exist");
2763 EncClass* encoding = _AD._encode->add_EncClass(ec_name);
2764 encoding->_linenum = linenum();
2766 // synthesize the arguments list for the enc_class from the
2767 // arguments to the instruct definition.
2768 const char* param = NULL;
2769 inst._parameters.reset();
2770 while ((param = inst._parameters.iter()) != NULL) {
2771 OperandForm* opForm = (OperandForm*) inst._localNames[param];
2772 encoding->add_parameter(opForm->_ident, param);
2773 }
2775 // Define a MacroAssembler instance for use by the encoding. The
2776 // name is chosen to match the __ idiom used for assembly in other
2777 // parts of hotspot and assumes the existence of the standard
2778 // #define __ _masm.
2779 encoding->add_code(" MacroAssembler _masm(&cbuf);\n");
2781 // Parse the following %{ }% block
2782 ins_encode_parse_block_impl(inst, encoding, ec_name);
2784 // Build an encoding rule which invokes the encoding rule we just
2785 // created, passing all arguments that we received.
2786 InsEncode* encrule = new InsEncode(); // Encode class for instruction
2787 NameAndList* params = encrule->add_encode(ec_name);
2788 inst._parameters.reset();
2789 while ((param = inst._parameters.iter()) != NULL) {
2790 params->add_entry(param);
2791 }
2793 // Check for duplicate ins_encode sections after parsing the block
2794 // so that parsing can continue and find any other errors.
2795 if (inst._insencode != NULL) {
2796 parse_err(SYNERR, "Multiple ins_encode sections defined\n");
2797 return;
2798 }
2800 // Set encode class of this instruction.
2801 inst._insencode = encrule;
2802 }
2805 void ADLParser::ins_encode_parse_block_impl(InstructForm& inst, EncClass* encoding, char* ec_name) {
2806 skipws_no_preproc(); // Skip leading whitespace
2807 // Prepend location descriptor, for debugging; cf. ADLParser::find_cpp_block
2808 if (_AD._adlocation_debug) {
2809 encoding->add_code(get_line_string());
2810 }
2812 // Collect the parts of the encode description
2813 // (1) strings that are passed through to output
2814 // (2) replacement/substitution variable, preceeded by a '$'
2815 while ((_curchar != '%') && (*(_ptr+1) != '}')) {
2817 // (1)
2818 // Check if there is a string to pass through to output
2819 char *start = _ptr; // Record start of the next string
2820 while ((_curchar != '$') && ((_curchar != '%') || (*(_ptr+1) != '}')) ) {
2821 // If at the start of a comment, skip past it
2822 if( (_curchar == '/') && ((*(_ptr+1) == '/') || (*(_ptr+1) == '*')) ) {
2823 skipws_no_preproc();
2824 } else {
2825 // ELSE advance to the next character, or start of the next line
2826 next_char_or_line();
2827 }
2828 }
2829 // If a string was found, terminate it and record in EncClass
2830 if (start != _ptr) {
2831 *_ptr = '\0'; // Terminate the string
2832 encoding->add_code(start);
2833 }
2835 // (2)
2836 // If we are at a replacement variable,
2837 // copy it and record in EncClass
2838 if (_curchar == '$') {
2839 // Found replacement Variable
2840 char* rep_var = get_rep_var_ident_dup();
2842 // Add flag to _strings list indicating we should check _rep_vars
2843 encoding->add_rep_var(rep_var);
2845 skipws();
2847 // Check if this instruct is a MachConstantNode.
2848 if (strcmp(rep_var, "constanttablebase") == 0) {
2849 // This instruct is a MachConstantNode.
2850 inst.set_is_mach_constant(true);
2852 if (_curchar == '(') {
2853 parse_err(SYNERR, "constanttablebase in instruct %s cannot have an argument (only constantaddress and constantoffset)", ec_name);
2854 return;
2855 }
2856 }
2857 else if ((strcmp(rep_var, "constantaddress") == 0) ||
2858 (strcmp(rep_var, "constantoffset") == 0)) {
2859 // This instruct is a MachConstantNode.
2860 inst.set_is_mach_constant(true);
2862 // If the constant keyword has an argument, parse it.
2863 if (_curchar == '(') constant_parse(inst);
2864 }
2865 }
2866 } // end while part of format description
2867 next_char(); // Skip '%'
2868 next_char(); // Skip '}'
2870 skipws();
2872 if (_AD._adlocation_debug) {
2873 encoding->add_code(end_line_marker());
2874 }
2876 // Debug Stuff
2877 if (_AD._adl_debug > 1) fprintf(stderr, "EncodingClass Form: %s\n", ec_name);
2878 }
2881 //------------------------------ins_encode_parse-------------------------------
2882 // Encode rules have the form
2883 // ins_encode( encode_class_name(parameter_list), ... );
2884 //
2885 // The "encode_class_name" must be defined in the encode section
2886 // The parameter list contains $names that are locals.
2887 //
2888 // Alternatively it can be written like this:
2889 //
2890 // ins_encode %{
2891 // ... // body
2892 // %}
2893 //
2894 // which synthesizes a new encoding class taking the same arguments as
2895 // the InstructForm, and automatically prefixes the definition with:
2896 //
2897 // MacroAssembler masm(&cbuf);\n");
2898 //
2899 // making it more compact to take advantage of the MacroAssembler and
2900 // placing the assembly closer to it's use by instructions.
2901 void ADLParser::ins_encode_parse(InstructForm& inst) {
2903 // Parse encode class name
2904 skipws(); // Skip whitespace
2905 if (_curchar != '(') {
2906 // Check for ins_encode %{ form
2907 if ((_curchar == '%') && (*(_ptr+1) == '{')) {
2908 next_char(); // Skip '%'
2909 next_char(); // Skip '{'
2911 // Parse the block form of ins_encode
2912 ins_encode_parse_block(inst);
2913 return;
2914 }
2916 parse_err(SYNERR, "missing '%%{' or '(' in ins_encode definition\n");
2917 return;
2918 }
2919 next_char(); // move past '('
2920 skipws();
2922 InsEncode *encrule = new InsEncode(); // Encode class for instruction
2923 encrule->_linenum = linenum();
2924 char *ec_name = NULL; // String representation of encode rule
2925 // identifier is optional.
2926 while (_curchar != ')') {
2927 ec_name = get_ident();
2928 if (ec_name == NULL) {
2929 parse_err(SYNERR, "Invalid encode class name after 'ins_encode('.\n");
2930 return;
2931 }
2932 // Check that encoding is defined in the encode section
2933 EncClass *encode_class = _AD._encode->encClass(ec_name);
2934 if (encode_class == NULL) {
2935 // Like to defer checking these till later...
2936 // parse_err(WARN, "Using an undefined encode class '%s' in 'ins_encode'.\n", ec_name);
2937 }
2939 // Get list for encode method's parameters
2940 NameAndList *params = encrule->add_encode(ec_name);
2942 // Parse the parameters to this encode method.
2943 skipws();
2944 if ( _curchar == '(' ) {
2945 next_char(); // move past '(' for parameters
2947 // Parse the encode method's parameters
2948 while (_curchar != ')') {
2949 char *param = get_ident_or_literal_constant("encoding operand");
2950 if ( param != NULL ) {
2951 // Found a parameter:
2952 // Check it is a local name, add it to the list, then check for more
2953 // New: allow hex constants as parameters to an encode method.
2954 // New: allow parenthesized expressions as parameters.
2955 // New: allow "primary", "secondary", "tertiary" as parameters.
2956 // New: allow user-defined register name as parameter
2957 if ( (inst._localNames[param] == NULL) &&
2958 !ADLParser::is_literal_constant(param) &&
2959 (Opcode::as_opcode_type(param) == Opcode::NOT_AN_OPCODE) &&
2960 ((_AD._register == NULL ) || (_AD._register->getRegDef(param) == NULL)) ) {
2961 parse_err(SYNERR, "Using non-locally defined parameter %s for encoding %s.\n", param, ec_name);
2962 return;
2963 }
2964 params->add_entry(param);
2966 skipws();
2967 if (_curchar == ',' ) {
2968 // More parameters to come
2969 next_char(); // move past ',' between parameters
2970 skipws(); // Skip to next parameter
2971 }
2972 else if (_curchar == ')') {
2973 // Done with parameter list
2974 }
2975 else {
2976 // Only ',' or ')' are valid after a parameter name
2977 parse_err(SYNERR, "expected ',' or ')' after parameter %s.\n",
2978 ec_name);
2979 return;
2980 }
2982 } else {
2983 skipws();
2984 // Did not find a parameter
2985 if (_curchar == ',') {
2986 parse_err(SYNERR, "Expected encode parameter before ',' in encoding %s.\n", ec_name);
2987 return;
2988 }
2989 if (_curchar != ')') {
2990 parse_err(SYNERR, "Expected ')' after encode parameters.\n");
2991 return;
2992 }
2993 }
2994 } // WHILE loop collecting parameters
2995 next_char(); // move past ')' at end of parameters
2996 } // done with parameter list for encoding
2998 // Check for ',' or ')' after encoding
2999 skipws(); // move to character after parameters
3000 if ( _curchar == ',' ) {
3001 // Found a ','
3002 next_char(); // move past ',' between encode methods
3003 skipws();
3004 }
3005 else if ( _curchar != ')' ) {
3006 // If not a ',' then only a ')' is allowed
3007 parse_err(SYNERR, "Expected ')' after encoding %s.\n", ec_name);
3008 return;
3009 }
3011 // Check for ',' separating parameters
3012 // if ( _curchar != ',' && _curchar != ')' ) {
3013 // parse_err(SYNERR, "expected ',' or ')' after encode method inside ins_encode.\n");
3014 // return NULL;
3015 // }
3017 } // done parsing ins_encode methods and their parameters
3018 if (_curchar != ')') {
3019 parse_err(SYNERR, "Missing ')' at end of ins_encode description.\n");
3020 return;
3021 }
3022 next_char(); // move past ')'
3023 skipws(); // Skip leading whitespace
3025 if ( _curchar != ';' ) {
3026 parse_err(SYNERR, "Missing ';' at end of ins_encode.\n");
3027 return;
3028 }
3029 next_char(); // move past ';'
3030 skipws(); // be friendly to oper_parse()
3032 // Check for duplicate ins_encode sections after parsing the block
3033 // so that parsing can continue and find any other errors.
3034 if (inst._insencode != NULL) {
3035 parse_err(SYNERR, "Multiple ins_encode sections defined\n");
3036 return;
3037 }
3039 // Debug Stuff
3040 if (_AD._adl_debug > 1) fprintf(stderr,"Instruction Encode: %s\n", ec_name);
3042 // Set encode class of this instruction.
3043 inst._insencode = encrule;
3044 }
3047 //------------------------------constant_parse---------------------------------
3048 // Parse a constant expression.
3049 void ADLParser::constant_parse(InstructForm& inst) {
3050 // Create a new encoding name based on the name of the instruction
3051 // definition, which should be unique.
3052 const char* prefix = "__constant_";
3053 char* ec_name = (char*) malloc(strlen(inst._ident) + strlen(prefix) + 1);
3054 sprintf(ec_name, "%s%s", prefix, inst._ident);
3056 assert(_AD._encode->encClass(ec_name) == NULL, "shouldn't already exist");
3057 EncClass* encoding = _AD._encode->add_EncClass(ec_name);
3058 encoding->_linenum = linenum();
3060 // synthesize the arguments list for the enc_class from the
3061 // arguments to the instruct definition.
3062 const char* param = NULL;
3063 inst._parameters.reset();
3064 while ((param = inst._parameters.iter()) != NULL) {
3065 OperandForm* opForm = (OperandForm*) inst._localNames[param];
3066 encoding->add_parameter(opForm->_ident, param);
3067 }
3069 // Parse the following ( ) expression.
3070 constant_parse_expression(encoding, ec_name);
3072 // Build an encoding rule which invokes the encoding rule we just
3073 // created, passing all arguments that we received.
3074 InsEncode* encrule = new InsEncode(); // Encode class for instruction
3075 NameAndList* params = encrule->add_encode(ec_name);
3076 inst._parameters.reset();
3077 while ((param = inst._parameters.iter()) != NULL) {
3078 params->add_entry(param);
3079 }
3081 // Set encode class of this instruction.
3082 inst._constant = encrule;
3083 }
3086 //------------------------------constant_parse_expression----------------------
3087 void ADLParser::constant_parse_expression(EncClass* encoding, char* ec_name) {
3088 skipws();
3090 // Prepend location descriptor, for debugging; cf. ADLParser::find_cpp_block
3091 if (_AD._adlocation_debug) {
3092 encoding->add_code(get_line_string());
3093 }
3095 // Start code line.
3096 encoding->add_code(" _constant = C->constant_table().add");
3098 // Parse everything in ( ) expression.
3099 encoding->add_code("(this, ");
3100 next_char(); // Skip '('
3101 int parens_depth = 1;
3103 // Collect the parts of the constant expression.
3104 // (1) strings that are passed through to output
3105 // (2) replacement/substitution variable, preceeded by a '$'
3106 while (parens_depth > 0) {
3107 if (_curchar == '(') {
3108 parens_depth++;
3109 encoding->add_code("(");
3110 next_char();
3111 }
3112 else if (_curchar == ')') {
3113 parens_depth--;
3114 if (parens_depth > 0)
3115 encoding->add_code(")");
3116 next_char();
3117 }
3118 else {
3119 // (1)
3120 // Check if there is a string to pass through to output
3121 char *start = _ptr; // Record start of the next string
3122 while ((_curchar != '$') && (_curchar != '(') && (_curchar != ')')) {
3123 next_char();
3124 }
3125 // If a string was found, terminate it and record in EncClass
3126 if (start != _ptr) {
3127 *_ptr = '\0'; // Terminate the string
3128 encoding->add_code(start);
3129 }
3131 // (2)
3132 // If we are at a replacement variable, copy it and record in EncClass.
3133 if (_curchar == '$') {
3134 // Found replacement Variable
3135 char* rep_var = get_rep_var_ident_dup();
3136 encoding->add_rep_var(rep_var);
3137 }
3138 }
3139 }
3141 // Finish code line.
3142 encoding->add_code(");");
3144 if (_AD._adlocation_debug) {
3145 encoding->add_code(end_line_marker());
3146 }
3148 // Debug Stuff
3149 if (_AD._adl_debug > 1) fprintf(stderr, "EncodingClass Form: %s\n", ec_name);
3150 }
3153 //------------------------------size_parse-----------------------------------
3154 char* ADLParser::size_parse(InstructForm *instr) {
3155 char* sizeOfInstr = NULL;
3157 // Get value of the instruction's size
3158 skipws();
3160 // Parse size
3161 sizeOfInstr = get_paren_expr("size expression");
3162 if (sizeOfInstr == NULL) {
3163 parse_err(SYNERR, "size of opcode expected at %c\n", _curchar);
3164 return NULL;
3165 }
3167 skipws();
3169 // Check for terminator
3170 if (_curchar != ';') {
3171 parse_err(SYNERR, "missing ';' in ins_attrib definition\n");
3172 return NULL;
3173 }
3174 next_char(); // Advance past the ';'
3175 skipws(); // necessary for instr_parse()
3177 // Debug Stuff
3178 if (_AD._adl_debug > 1) {
3179 if (sizeOfInstr != NULL) {
3180 fprintf(stderr,"size of opcode: %s\n", sizeOfInstr);
3181 }
3182 }
3184 return sizeOfInstr;
3185 }
3188 //------------------------------opcode_parse-----------------------------------
3189 Opcode * ADLParser::opcode_parse(InstructForm *instr) {
3190 char *primary = NULL;
3191 char *secondary = NULL;
3192 char *tertiary = NULL;
3194 char *val = NULL;
3195 Opcode *opcode = NULL;
3197 // Get value of the instruction's opcode
3198 skipws();
3199 if (_curchar != '(') { // Check for parenthesized operand list
3200 parse_err(SYNERR, "missing '(' in expand instruction declaration\n");
3201 return NULL;
3202 }
3203 next_char(); // skip open paren
3204 skipws();
3205 if (_curchar != ')') {
3206 // Parse primary, secondary, and tertiary opcodes, if provided.
3207 if ( ((primary = get_ident_or_literal_constant("primary opcode")) == NULL) ) {
3208 parse_err(SYNERR, "primary hex opcode expected at %c\n", _curchar);
3209 return NULL;
3210 }
3211 skipws();
3212 if (_curchar == ',') {
3213 next_char();
3214 skipws();
3215 // Parse secondary opcode
3216 if ( ((secondary = get_ident_or_literal_constant("secondary opcode")) == NULL) ) {
3217 parse_err(SYNERR, "secondary hex opcode expected at %c\n", _curchar);
3218 return NULL;
3219 }
3220 skipws();
3221 if (_curchar == ',') {
3222 next_char();
3223 skipws();
3224 // Parse tertiary opcode
3225 if ( ((tertiary = get_ident_or_literal_constant("tertiary opcode")) == NULL) ) {
3226 parse_err(SYNERR,"tertiary hex opcode expected at %c\n", _curchar);
3227 return NULL;
3228 }
3229 skipws();
3230 }
3231 }
3232 skipws();
3233 if (_curchar != ')') {
3234 parse_err(SYNERR, "Missing ')' in opcode description\n");
3235 return NULL;
3236 }
3237 }
3238 next_char(); // Skip ')'
3239 skipws();
3240 // Check for terminator
3241 if (_curchar != ';') {
3242 parse_err(SYNERR, "missing ';' in ins_attrib definition\n");
3243 return NULL;
3244 }
3245 next_char(); // Advance past the ';'
3246 skipws(); // necessary for instr_parse()
3248 // Debug Stuff
3249 if (_AD._adl_debug > 1) {
3250 if (primary != NULL) fprintf(stderr,"primary opcode: %s\n", primary);
3251 if (secondary != NULL) fprintf(stderr,"secondary opcode: %s\n", secondary);
3252 if (tertiary != NULL) fprintf(stderr,"tertiary opcode: %s\n", tertiary);
3253 }
3255 // Generate new object and return
3256 opcode = new Opcode(primary, secondary, tertiary);
3257 return opcode;
3258 }
3261 //------------------------------interface_parse--------------------------------
3262 Interface *ADLParser::interface_parse(void) {
3263 char *iface_name = NULL; // Name of interface class being used
3264 char *iface_code = NULL; // Describe components of this class
3266 // Get interface class name
3267 skipws(); // Skip whitespace
3268 if (_curchar != '(') {
3269 parse_err(SYNERR, "Missing '(' at start of interface description.\n");
3270 return NULL;
3271 }
3272 next_char(); // move past '('
3273 skipws();
3274 iface_name = get_ident();
3275 if (iface_name == NULL) {
3276 parse_err(SYNERR, "missing interface name after 'interface'.\n");
3277 return NULL;
3278 }
3279 skipws();
3280 if (_curchar != ')') {
3281 parse_err(SYNERR, "Missing ')' after name of interface.\n");
3282 return NULL;
3283 }
3284 next_char(); // move past ')'
3286 // Get details of the interface,
3287 // for the type of interface indicated by iface_name.
3288 Interface *inter = NULL;
3289 skipws();
3290 if ( _curchar != ';' ) {
3291 if ( strcmp(iface_name,"MEMORY_INTER") == 0 ) {
3292 inter = mem_interface_parse();
3293 }
3294 else if ( strcmp(iface_name,"COND_INTER") == 0 ) {
3295 inter = cond_interface_parse();
3296 }
3297 // The parse routines consume the "%}"
3299 // Check for probable extra ';' after defining block.
3300 if ( _curchar == ';' ) {
3301 parse_err(SYNERR, "Extra ';' after defining interface block.\n");
3302 next_char(); // Skip ';'
3303 return NULL;
3304 }
3305 } else {
3306 next_char(); // move past ';'
3308 // Create appropriate interface object
3309 if ( strcmp(iface_name,"REG_INTER") == 0 ) {
3310 inter = new RegInterface();
3311 }
3312 else if ( strcmp(iface_name,"CONST_INTER") == 0 ) {
3313 inter = new ConstInterface();
3314 }
3315 }
3316 skipws(); // be friendly to oper_parse()
3317 // Debug Stuff
3318 if (_AD._adl_debug > 1) fprintf(stderr,"Interface Form: %s\n", iface_name);
3320 // Create appropriate interface object and return.
3321 return inter;
3322 }
3325 //------------------------------mem_interface_parse----------------------------
3326 Interface *ADLParser::mem_interface_parse(void) {
3327 // Fields for MemInterface
3328 char *base = NULL;
3329 char *index = NULL;
3330 char *scale = NULL;
3331 char *disp = NULL;
3333 if (_curchar != '%') {
3334 parse_err(SYNERR, "Missing '%{' for 'interface' block.\n");
3335 return NULL;
3336 }
3337 next_char(); // Skip '%'
3338 if (_curchar != '{') {
3339 parse_err(SYNERR, "Missing '%{' for 'interface' block.\n");
3340 return NULL;
3341 }
3342 next_char(); // Skip '{'
3343 skipws();
3344 do {
3345 char *field = get_ident();
3346 if (field == NULL) {
3347 parse_err(SYNERR, "Expected keyword, base|index|scale|disp, or '%}' ending interface.\n");
3348 return NULL;
3349 }
3350 if ( strcmp(field,"base") == 0 ) {
3351 base = interface_field_parse();
3352 }
3353 else if ( strcmp(field,"index") == 0 ) {
3354 index = interface_field_parse();
3355 }
3356 else if ( strcmp(field,"scale") == 0 ) {
3357 scale = interface_field_parse();
3358 }
3359 else if ( strcmp(field,"disp") == 0 ) {
3360 disp = interface_field_parse();
3361 }
3362 else {
3363 parse_err(SYNERR, "Expected keyword, base|index|scale|disp, or '%}' ending interface.\n");
3364 return NULL;
3365 }
3366 } while( _curchar != '%' );
3367 next_char(); // Skip '%'
3368 if ( _curchar != '}' ) {
3369 parse_err(SYNERR, "Missing '%}' for 'interface' block.\n");
3370 return NULL;
3371 }
3372 next_char(); // Skip '}'
3374 // Construct desired object and return
3375 Interface *inter = new MemInterface(base, index, scale, disp);
3376 return inter;
3377 }
3380 //------------------------------cond_interface_parse---------------------------
3381 Interface *ADLParser::cond_interface_parse(void) {
3382 char *equal;
3383 char *not_equal;
3384 char *less;
3385 char *greater_equal;
3386 char *less_equal;
3387 char *greater;
3388 const char *equal_format = "eq";
3389 const char *not_equal_format = "ne";
3390 const char *less_format = "lt";
3391 const char *greater_equal_format = "ge";
3392 const char *less_equal_format = "le";
3393 const char *greater_format = "gt";
3395 if (_curchar != '%') {
3396 parse_err(SYNERR, "Missing '%{' for 'cond_interface' block.\n");
3397 return NULL;
3398 }
3399 next_char(); // Skip '%'
3400 if (_curchar != '{') {
3401 parse_err(SYNERR, "Missing '%{' for 'cond_interface' block.\n");
3402 return NULL;
3403 }
3404 next_char(); // Skip '{'
3405 skipws();
3406 do {
3407 char *field = get_ident();
3408 if (field == NULL) {
3409 parse_err(SYNERR, "Expected keyword, base|index|scale|disp, or '%}' ending interface.\n");
3410 return NULL;
3411 }
3412 if ( strcmp(field,"equal") == 0 ) {
3413 equal = interface_field_parse(&equal_format);
3414 }
3415 else if ( strcmp(field,"not_equal") == 0 ) {
3416 not_equal = interface_field_parse(¬_equal_format);
3417 }
3418 else if ( strcmp(field,"less") == 0 ) {
3419 less = interface_field_parse(&less_format);
3420 }
3421 else if ( strcmp(field,"greater_equal") == 0 ) {
3422 greater_equal = interface_field_parse(&greater_equal_format);
3423 }
3424 else if ( strcmp(field,"less_equal") == 0 ) {
3425 less_equal = interface_field_parse(&less_equal_format);
3426 }
3427 else if ( strcmp(field,"greater") == 0 ) {
3428 greater = interface_field_parse(&greater_format);
3429 }
3430 else {
3431 parse_err(SYNERR, "Expected keyword, base|index|scale|disp, or '%}' ending interface.\n");
3432 return NULL;
3433 }
3434 } while( _curchar != '%' );
3435 next_char(); // Skip '%'
3436 if ( _curchar != '}' ) {
3437 parse_err(SYNERR, "Missing '%}' for 'interface' block.\n");
3438 return NULL;
3439 }
3440 next_char(); // Skip '}'
3442 // Construct desired object and return
3443 Interface *inter = new CondInterface(equal, equal_format,
3444 not_equal, not_equal_format,
3445 less, less_format,
3446 greater_equal, greater_equal_format,
3447 less_equal, less_equal_format,
3448 greater, greater_format);
3449 return inter;
3450 }
3453 //------------------------------interface_field_parse--------------------------
3454 char *ADLParser::interface_field_parse(const char ** format) {
3455 char *iface_field = NULL;
3457 // Get interface field
3458 skipws(); // Skip whitespace
3459 if (_curchar != '(') {
3460 parse_err(SYNERR, "Missing '(' at start of interface field.\n");
3461 return NULL;
3462 }
3463 next_char(); // move past '('
3464 skipws();
3465 if ( _curchar != '0' && _curchar != '$' ) {
3466 parse_err(SYNERR, "missing or invalid interface field contents.\n");
3467 return NULL;
3468 }
3469 iface_field = get_rep_var_ident();
3470 if (iface_field == NULL) {
3471 parse_err(SYNERR, "missing or invalid interface field contents.\n");
3472 return NULL;
3473 }
3474 skipws();
3475 if (format != NULL && _curchar == ',') {
3476 next_char();
3477 skipws();
3478 if (_curchar != '"') {
3479 parse_err(SYNERR, "Missing '\"' in field format .\n");
3480 return NULL;
3481 }
3482 next_char();
3483 char *start = _ptr; // Record start of the next string
3484 while ((_curchar != '"') && (_curchar != '%') && (_curchar != '\n')) {
3485 if (_curchar == '\\') next_char(); // superquote
3486 if (_curchar == '\n') parse_err(SYNERR, "newline in string"); // unimplemented!
3487 next_char();
3488 }
3489 if (_curchar != '"') {
3490 parse_err(SYNERR, "Missing '\"' at end of field format .\n");
3491 return NULL;
3492 }
3493 // If a string was found, terminate it and record in FormatRule
3494 if ( start != _ptr ) {
3495 *_ptr = '\0'; // Terminate the string
3496 *format = start;
3497 }
3498 next_char();
3499 skipws();
3500 }
3501 if (_curchar != ')') {
3502 parse_err(SYNERR, "Missing ')' after interface field.\n");
3503 return NULL;
3504 }
3505 next_char(); // move past ')'
3506 skipws();
3507 if ( _curchar != ';' ) {
3508 parse_err(SYNERR, "Missing ';' at end of interface field.\n");
3509 return NULL;
3510 }
3511 next_char(); // move past ';'
3512 skipws(); // be friendly to interface_parse()
3514 return iface_field;
3515 }
3518 //------------------------------match_parse------------------------------------
3519 MatchRule *ADLParser::match_parse(FormDict &operands) {
3520 MatchRule *match; // Match Rule class for instruction/operand
3521 char *cnstr = NULL; // Code for constructor
3522 int depth = 0; // Counter for matching parentheses
3523 int numleaves = 0; // Counter for number of leaves in rule
3525 // Parse the match rule tree
3526 MatchNode *mnode = matchNode_parse(operands, depth, numleaves, true);
3528 // Either there is a block with a constructor, or a ';' here
3529 skipws(); // Skip whitespace
3530 if ( _curchar == ';' ) { // Semicolon is valid terminator
3531 cnstr = NULL; // no constructor for this form
3532 next_char(); // Move past the ';', replaced with '\0'
3533 }
3534 else if ((cnstr = find_cpp_block("match constructor")) == NULL ) {
3535 parse_err(SYNERR, "invalid construction of match rule\n"
3536 "Missing ';' or invalid '%{' and '%}' constructor\n");
3537 return NULL; // No MatchRule to return
3538 }
3539 if (_AD._adl_debug > 1)
3540 if (cnstr) fprintf(stderr,"Match Constructor: %s\n", cnstr);
3541 // Build new MatchRule object
3542 match = new MatchRule(_AD, mnode, depth, cnstr, numleaves);
3543 skipws(); // Skip any trailing whitespace
3544 return match; // Return MatchRule object
3545 }
3547 //------------------------------format_parse-----------------------------------
3548 FormatRule* ADLParser::format_parse(void) {
3549 char *desc = NULL;
3550 FormatRule *format = (new FormatRule(desc));
3552 // Without expression form, MUST have a code block;
3553 skipws(); // Skip whitespace
3554 if ( _curchar == ';' ) { // Semicolon is valid terminator
3555 desc = NULL; // no constructor for this form
3556 next_char(); // Move past the ';', replaced with '\0'
3557 }
3558 else if ( _curchar == '%' && *(_ptr+1) == '{') {
3559 next_char(); // Move past the '%'
3560 next_char(); // Move past the '{'
3562 skipws();
3563 if (_curchar == '$') {
3564 char* ident = get_rep_var_ident();
3565 if (strcmp(ident, "$$template") == 0) return template_parse();
3566 parse_err(SYNERR, "Unknown \"%s\" directive in format", ident);
3567 return NULL;
3568 }
3569 // Check for the opening '"' inside the format description
3570 if ( _curchar == '"' ) {
3571 next_char(); // Move past the initial '"'
3572 if( _curchar == '"' ) { // Handle empty format string case
3573 *_ptr = '\0'; // Terminate empty string
3574 format->_strings.addName(_ptr);
3575 }
3577 // Collect the parts of the format description
3578 // (1) strings that are passed through to tty->print
3579 // (2) replacement/substitution variable, preceeded by a '$'
3580 // (3) multi-token ANSIY C style strings
3581 while ( true ) {
3582 if ( _curchar == '%' || _curchar == '\n' ) {
3583 if ( _curchar != '"' ) {
3584 parse_err(SYNERR, "missing '\"' at end of format block");
3585 return NULL;
3586 }
3587 }
3589 // (1)
3590 // Check if there is a string to pass through to output
3591 char *start = _ptr; // Record start of the next string
3592 while ((_curchar != '$') && (_curchar != '"') && (_curchar != '%') && (_curchar != '\n')) {
3593 if (_curchar == '\\') {
3594 next_char(); // superquote
3595 if ((_curchar == '$') || (_curchar == '%'))
3596 // hack to avoid % escapes and warnings about undefined \ escapes
3597 *(_ptr-1) = _curchar;
3598 }
3599 if (_curchar == '\n') parse_err(SYNERR, "newline in string"); // unimplemented!
3600 next_char();
3601 }
3602 // If a string was found, terminate it and record in FormatRule
3603 if ( start != _ptr ) {
3604 *_ptr = '\0'; // Terminate the string
3605 format->_strings.addName(start);
3606 }
3608 // (2)
3609 // If we are at a replacement variable,
3610 // copy it and record in FormatRule
3611 if ( _curchar == '$' ) {
3612 next_char(); // Move past the '$'
3613 char* rep_var = get_ident(); // Nil terminate the variable name
3614 rep_var = strdup(rep_var);// Copy the string
3615 *_ptr = _curchar; // and replace Nil with original character
3616 format->_rep_vars.addName(rep_var);
3617 // Add flag to _strings list indicating we should check _rep_vars
3618 format->_strings.addName(NameList::_signal);
3619 }
3621 // (3)
3622 // Allow very long strings to be broken up,
3623 // using the ANSI C syntax "foo\n" <newline> "bar"
3624 if ( _curchar == '"') {
3625 next_char(); // Move past the '"'
3626 skipws(); // Skip white space before next string token
3627 if ( _curchar != '"') {
3628 break;
3629 } else {
3630 // Found one. Skip both " and the whitespace in between.
3631 next_char();
3632 }
3633 }
3634 } // end while part of format description
3636 // Check for closing '"' and '%}' in format description
3637 skipws(); // Move to closing '%}'
3638 if ( _curchar != '%' ) {
3639 parse_err(SYNERR, "non-blank characters between closing '\"' and '%' in format");
3640 return NULL;
3641 }
3642 } // Done with format description inside
3644 skipws();
3645 // Past format description, at '%'
3646 if ( _curchar != '%' || *(_ptr+1) != '}' ) {
3647 parse_err(SYNERR, "missing '%}' at end of format block");
3648 return NULL;
3649 }
3650 next_char(); // Move past the '%'
3651 next_char(); // Move past the '}'
3652 }
3653 else { // parameter list alone must terminate with a ';'
3654 parse_err(SYNERR, "missing ';' after Format expression");
3655 return NULL;
3656 }
3657 // Debug Stuff
3658 if (_AD._adl_debug > 1) fprintf(stderr,"Format Rule: %s\n", desc);
3660 skipws();
3661 return format;
3662 }
3665 //------------------------------template_parse-----------------------------------
3666 FormatRule* ADLParser::template_parse(void) {
3667 char *desc = NULL;
3668 FormatRule *format = (new FormatRule(desc));
3670 skipws();
3671 while ( (_curchar != '%') && (*(_ptr+1) != '}') ) {
3673 // (1)
3674 // Check if there is a string to pass through to output
3675 {
3676 char *start = _ptr; // Record start of the next string
3677 while ((_curchar != '$') && ((_curchar != '%') || (*(_ptr+1) != '}')) ) {
3678 // If at the start of a comment, skip past it
3679 if( (_curchar == '/') && ((*(_ptr+1) == '/') || (*(_ptr+1) == '*')) ) {
3680 skipws_no_preproc();
3681 } else {
3682 // ELSE advance to the next character, or start of the next line
3683 next_char_or_line();
3684 }
3685 }
3686 // If a string was found, terminate it and record in EncClass
3687 if ( start != _ptr ) {
3688 *_ptr = '\0'; // Terminate the string
3689 // Add flag to _strings list indicating we should check _rep_vars
3690 format->_strings.addName(NameList::_signal2);
3691 format->_strings.addName(start);
3692 }
3693 }
3695 // (2)
3696 // If we are at a replacement variable,
3697 // copy it and record in EncClass
3698 if ( _curchar == '$' ) {
3699 // Found replacement Variable
3700 char *rep_var = get_rep_var_ident_dup();
3701 if (strcmp(rep_var, "$emit") == 0) {
3702 // switch to normal format parsing
3703 next_char();
3704 next_char();
3705 skipws();
3706 // Check for the opening '"' inside the format description
3707 if ( _curchar == '"' ) {
3708 next_char(); // Move past the initial '"'
3709 if( _curchar == '"' ) { // Handle empty format string case
3710 *_ptr = '\0'; // Terminate empty string
3711 format->_strings.addName(_ptr);
3712 }
3714 // Collect the parts of the format description
3715 // (1) strings that are passed through to tty->print
3716 // (2) replacement/substitution variable, preceeded by a '$'
3717 // (3) multi-token ANSIY C style strings
3718 while ( true ) {
3719 if ( _curchar == '%' || _curchar == '\n' ) {
3720 parse_err(SYNERR, "missing '\"' at end of format block");
3721 return NULL;
3722 }
3724 // (1)
3725 // Check if there is a string to pass through to output
3726 char *start = _ptr; // Record start of the next string
3727 while ((_curchar != '$') && (_curchar != '"') && (_curchar != '%') && (_curchar != '\n')) {
3728 if (_curchar == '\\') next_char(); // superquote
3729 if (_curchar == '\n') parse_err(SYNERR, "newline in string"); // unimplemented!
3730 next_char();
3731 }
3732 // If a string was found, terminate it and record in FormatRule
3733 if ( start != _ptr ) {
3734 *_ptr = '\0'; // Terminate the string
3735 format->_strings.addName(start);
3736 }
3738 // (2)
3739 // If we are at a replacement variable,
3740 // copy it and record in FormatRule
3741 if ( _curchar == '$' ) {
3742 next_char(); // Move past the '$'
3743 char* next_rep_var = get_ident(); // Nil terminate the variable name
3744 next_rep_var = strdup(next_rep_var);// Copy the string
3745 *_ptr = _curchar; // and replace Nil with original character
3746 format->_rep_vars.addName(next_rep_var);
3747 // Add flag to _strings list indicating we should check _rep_vars
3748 format->_strings.addName(NameList::_signal);
3749 }
3751 // (3)
3752 // Allow very long strings to be broken up,
3753 // using the ANSI C syntax "foo\n" <newline> "bar"
3754 if ( _curchar == '"') {
3755 next_char(); // Move past the '"'
3756 skipws(); // Skip white space before next string token
3757 if ( _curchar != '"') {
3758 break;
3759 } else {
3760 // Found one. Skip both " and the whitespace in between.
3761 next_char();
3762 }
3763 }
3764 } // end while part of format description
3765 }
3766 } else {
3767 // Add flag to _strings list indicating we should check _rep_vars
3768 format->_rep_vars.addName(rep_var);
3769 // Add flag to _strings list indicating we should check _rep_vars
3770 format->_strings.addName(NameList::_signal3);
3771 }
3772 } // end while part of format description
3773 }
3775 skipws();
3776 // Past format description, at '%'
3777 if ( _curchar != '%' || *(_ptr+1) != '}' ) {
3778 parse_err(SYNERR, "missing '%}' at end of format block");
3779 return NULL;
3780 }
3781 next_char(); // Move past the '%'
3782 next_char(); // Move past the '}'
3784 // Debug Stuff
3785 if (_AD._adl_debug > 1) fprintf(stderr,"Format Rule: %s\n", desc);
3787 skipws();
3788 return format;
3789 }
3792 //------------------------------effect_parse-----------------------------------
3793 void ADLParser::effect_parse(InstructForm *instr) {
3794 char* desc = NULL;
3796 skipws(); // Skip whitespace
3797 if (_curchar != '(') {
3798 parse_err(SYNERR, "missing '(' in effect definition\n");
3799 return;
3800 }
3801 // Get list of effect-operand pairs and insert into dictionary
3802 else get_effectlist(instr->_effects, instr->_localNames, instr->_has_call);
3804 // Debug Stuff
3805 if (_AD._adl_debug > 1) fprintf(stderr,"Effect description: %s\n", desc);
3806 if (_curchar != ';') {
3807 parse_err(SYNERR, "missing ';' in Effect definition\n");
3808 }
3809 next_char(); // Skip ';'
3811 }
3813 //------------------------------expand_parse-----------------------------------
3814 ExpandRule* ADLParser::expand_parse(InstructForm *instr) {
3815 char *ident, *ident2;
3816 OperandForm *oper;
3817 InstructForm *ins;
3818 NameAndList *instr_and_operands = NULL;
3819 ExpandRule *exp = new ExpandRule();
3821 // Expand is a block containing an ordered list of instructions, each of
3822 // which has an ordered list of operands.
3823 // Check for block delimiter
3824 skipws(); // Skip leading whitespace
3825 if ((_curchar != '%')
3826 || (next_char(), (_curchar != '{')) ) { // If not open block
3827 parse_err(SYNERR, "missing '%{' in expand definition\n");
3828 return(NULL);
3829 }
3830 next_char(); // Maintain the invariant
3831 do {
3832 ident = get_ident(); // Grab next identifier
3833 if (ident == NULL) {
3834 parse_err(SYNERR, "identifier expected at %c\n", _curchar);
3835 continue;
3836 } // Check that you have a valid instruction
3837 const Form *form = _globalNames[ident];
3838 ins = form ? form->is_instruction() : NULL;
3839 if (ins == NULL) {
3840 // This is a new operand
3841 oper = form ? form->is_operand() : NULL;
3842 if (oper == NULL) {
3843 parse_err(SYNERR, "instruction/operand name expected at %s\n", ident);
3844 continue;
3845 }
3846 // Throw the operand on the _newopers list
3847 skipws();
3848 ident = get_unique_ident(instr->_localNames,"Operand");
3849 if (ident == NULL) {
3850 parse_err(SYNERR, "identifier expected at %c\n", _curchar);
3851 continue;
3852 }
3853 exp->_newopers.addName(ident);
3854 // Add new operand to LocalNames
3855 instr->_localNames.Insert(ident, oper);
3856 // Grab any constructor code and save as a string
3857 char *c = NULL;
3858 skipws();
3859 if (_curchar == '%') { // Need a constructor for the operand
3860 c = find_cpp_block("Operand Constructor");
3861 if (c == NULL) {
3862 parse_err(SYNERR, "Invalid code block for operand constructor\n", _curchar);
3863 continue;
3864 }
3865 // Add constructor to _newopconst Dict
3866 exp->_newopconst.Insert(ident, c);
3867 }
3868 else if (_curchar != ';') { // If no constructor, need a ;
3869 parse_err(SYNERR, "Missing ; in expand rule operand declaration\n");
3870 continue;
3871 }
3872 else next_char(); // Skip the ;
3873 skipws();
3874 }
3875 else {
3876 // Add instruction to list
3877 instr_and_operands = new NameAndList(ident);
3878 // Grab operands, build nameList of them, and then put into dictionary
3879 skipws();
3880 if (_curchar != '(') { // Check for parenthesized operand list
3881 parse_err(SYNERR, "missing '(' in expand instruction declaration\n");
3882 continue;
3883 }
3884 do {
3885 next_char(); // skip open paren & comma characters
3886 skipws();
3887 if (_curchar == ')') break;
3888 ident2 = get_ident();
3889 skipws();
3890 if (ident2 == NULL) {
3891 parse_err(SYNERR, "identifier expected at %c\n", _curchar);
3892 continue;
3893 } // Check that you have a valid operand
3894 const Form *form2 = instr->_localNames[ident2];
3895 if (!form2) {
3896 parse_err(SYNERR, "operand name expected at %s\n", ident2);
3897 continue;
3898 }
3899 oper = form2->is_operand();
3900 if (oper == NULL && !form2->is_opclass()) {
3901 parse_err(SYNERR, "operand name expected at %s\n", ident2);
3902 continue;
3903 } // Add operand to list
3904 instr_and_operands->add_entry(ident2);
3905 } while(_curchar == ',');
3906 if (_curchar != ')') {
3907 parse_err(SYNERR, "missing ')'in expand instruction declaration\n");
3908 continue;
3909 }
3910 next_char();
3911 if (_curchar != ';') {
3912 parse_err(SYNERR, "missing ';'in expand instruction declaration\n");
3913 continue;
3914 }
3915 next_char();
3917 // Record both instruction name and its operand list
3918 exp->add_instruction(instr_and_operands);
3920 skipws();
3921 }
3923 } while(_curchar != '%');
3924 next_char();
3925 if (_curchar != '}') {
3926 parse_err(SYNERR, "missing '%}' in expand rule definition\n");
3927 return(NULL);
3928 }
3929 next_char();
3931 // Debug Stuff
3932 if (_AD._adl_debug > 1) fprintf(stderr,"Expand Rule:\n");
3934 skipws();
3935 return (exp);
3936 }
3938 //------------------------------rewrite_parse----------------------------------
3939 RewriteRule* ADLParser::rewrite_parse(void) {
3940 char* params = NULL;
3941 char* desc = NULL;
3944 // This feature targeted for second generation description language.
3946 skipws(); // Skip whitespace
3947 // Get parameters for rewrite
3948 if ((params = get_paren_expr("rewrite parameters")) == NULL) {
3949 parse_err(SYNERR, "missing '(' in rewrite rule\n");
3950 return NULL;
3951 }
3952 // Debug Stuff
3953 if (_AD._adl_debug > 1) fprintf(stderr,"Rewrite parameters: %s\n", params);
3955 // For now, grab entire block;
3956 skipws();
3957 if ( (desc = find_cpp_block("rewrite block")) == NULL ) {
3958 parse_err(SYNERR, "incorrect or missing block for 'rewrite'.\n");
3959 return NULL;
3960 }
3961 // Debug Stuff
3962 if (_AD._adl_debug > 1) fprintf(stderr,"Rewrite Rule: %s\n", desc);
3964 skipws();
3965 return (new RewriteRule(params,desc));
3966 }
3968 //------------------------------attr_parse-------------------------------------
3969 Attribute *ADLParser::attr_parse(char* ident) {
3970 Attribute *attrib; // Attribute class
3971 char *cost = NULL; // String representation of cost attribute
3973 skipws(); // Skip leading whitespace
3974 if ( (cost = get_paren_expr("attribute")) == NULL ) {
3975 parse_err(SYNERR, "incorrect or missing expression for 'attribute'\n");
3976 return NULL;
3977 }
3978 // Debug Stuff
3979 if (_AD._adl_debug > 1) fprintf(stderr,"Attribute: %s\n", cost);
3980 if (_curchar != ';') {
3981 parse_err(SYNERR, "missing ';' in attribute definition\n");
3982 return NULL;
3983 }
3984 next_char(); // Point after the terminator
3986 skipws();
3987 attrib = new Attribute(ident,cost,INS_ATTR); // Build new predicate object
3988 return attrib;
3989 }
3992 //------------------------------matchNode_parse--------------------------------
3993 MatchNode *ADLParser::matchNode_parse(FormDict &operands, int &depth, int &numleaves, bool atroot) {
3994 // Count depth of parenthesis nesting for both left and right children
3995 int lParens = depth;
3996 int rParens = depth;
3998 // MatchNode objects for left, right, and root of subtree.
3999 MatchNode *lChild = NULL;
4000 MatchNode *rChild = NULL;
4001 char *token; // Identifier which may be opcode or operand
4003 // Match expression starts with a '('
4004 if (cur_char() != '(')
4005 return NULL;
4007 next_char(); // advance past '('
4009 // Parse the opcode
4010 token = get_ident(); // Get identifier, opcode
4011 if (token == NULL) {
4012 parse_err(SYNERR, "missing opcode in match expression\n");
4013 return NULL;
4014 }
4016 // Take note if we see one of a few special operations - those that are
4017 // treated differently on different architectures in the sense that on
4018 // one architecture there is a match rule and on another there isn't (so
4019 // a call will eventually be generated).
4021 for (int i = _last_machine_leaf + 1; i < _last_opcode; i++) {
4022 if (strcmp(token, NodeClassNames[i]) == 0) {
4023 _AD.has_match_rule(i, true);
4024 }
4025 }
4027 // Lookup the root value in the operands dict to perform substitution
4028 const char *result = NULL; // Result type will be filled in later
4029 const char *name = token; // local name associated with this node
4030 const char *operation = token; // remember valid operation for later
4031 const Form *form = operands[token];
4032 OpClassForm *opcForm = form ? form->is_opclass() : NULL;
4033 if (opcForm != NULL) {
4034 // If this token is an entry in the local names table, record its type
4035 if (!opcForm->ideal_only()) {
4036 operation = opcForm->_ident;
4037 result = operation; // Operands result in their own type
4038 }
4039 // Otherwise it is an ideal type, and so, has no local name
4040 else name = NULL;
4041 }
4043 // Parse the operands
4044 skipws();
4045 if (cur_char() != ')') {
4047 // Parse the left child
4048 if (strcmp(operation,"Set"))
4049 lChild = matchChild_parse(operands, lParens, numleaves, false);
4050 else
4051 lChild = matchChild_parse(operands, lParens, numleaves, true);
4053 skipws();
4054 if (cur_char() != ')' ) {
4055 if(strcmp(operation, "Set"))
4056 rChild = matchChild_parse(operands,rParens,numleaves,false);
4057 else
4058 rChild = matchChild_parse(operands,rParens,numleaves,true);
4059 }
4060 }
4062 // Check for required ')'
4063 skipws();
4064 if (cur_char() != ')') {
4065 parse_err(SYNERR, "missing ')' in match expression\n");
4066 return NULL;
4067 }
4068 next_char(); // skip the ')'
4070 MatchNode* mroot = new MatchNode(_AD,result,name,operation,lChild,rChild);
4072 // If not the root, reduce this subtree to an internal operand
4073 if (!atroot) {
4074 mroot->build_internalop();
4075 }
4076 // depth is greater of left and right paths.
4077 depth = (lParens > rParens) ? lParens : rParens;
4079 return mroot;
4080 }
4083 //------------------------------matchChild_parse-------------------------------
4084 MatchNode *ADLParser::matchChild_parse(FormDict &operands, int &parens, int &numleaves, bool atroot) {
4085 MatchNode *child = NULL;
4086 const char *result = NULL;
4087 const char *token = NULL;
4088 const char *opType = NULL;
4090 if (cur_char() == '(') { // child is an operation
4091 ++parens;
4092 child = matchNode_parse(operands, parens, numleaves, atroot);
4093 }
4094 else { // child is an operand
4095 token = get_ident();
4096 const Form *form = operands[token];
4097 OpClassForm *opcForm = form ? form->is_opclass() : NULL;
4098 if (opcForm != NULL) {
4099 opType = opcForm->_ident;
4100 result = opcForm->_ident; // an operand's result matches its type
4101 } else {
4102 parse_err(SYNERR, "undefined operand %s in match rule\n", token);
4103 return NULL;
4104 }
4106 if (opType == NULL) {
4107 parse_err(SYNERR, "missing type for argument '%s'\n", token);
4108 }
4110 child = new MatchNode(_AD, result, token, opType);
4111 ++numleaves;
4112 }
4114 return child;
4115 }
4119 // ******************** Private Utility Functions *************************
4122 char* ADLParser::find_cpp_block(const char* description) {
4123 char *next; // Pointer for finding block delimiters
4124 char* cppBlock = NULL; // Beginning of C++ code block
4126 if (_curchar == '%') { // Encoding is a C++ expression
4127 next_char();
4128 if (_curchar != '{') {
4129 parse_err(SYNERR, "missing '{' in %s \n", description);
4130 return NULL;
4131 }
4132 next_char(); // Skip block delimiter
4133 skipws_no_preproc(); // Skip leading whitespace
4134 cppBlock = _ptr; // Point to start of expression
4135 int line = linenum();
4136 next = _ptr + 1;
4137 while(((_curchar != '%') || (*next != '}')) && (_curchar != '\0')) {
4138 next_char_or_line();
4139 next = _ptr+1; // Maintain the next pointer
4140 } // Grab string
4141 if (_curchar == '\0') {
4142 parse_err(SYNERR, "invalid termination of %s \n", description);
4143 return NULL;
4144 }
4145 *_ptr = '\0'; // Terminate string
4146 _ptr += 2; // Skip block delimiter
4147 _curchar = *_ptr; // Maintain invariant
4149 // Prepend location descriptor, for debugging.
4150 if (_AD._adlocation_debug) {
4151 char* location = get_line_string(line);
4152 char* end_loc = end_line_marker();
4153 char* result = (char *)malloc(strlen(location) + strlen(cppBlock) + strlen(end_loc) + 1);
4154 strcpy(result, location);
4155 strcat(result, cppBlock);
4156 strcat(result, end_loc);
4157 cppBlock = result;
4158 free(location);
4159 }
4160 }
4162 return cppBlock;
4163 }
4165 // Move to the closing token of the expression we are currently at,
4166 // as defined by stop_chars. Match parens and quotes.
4167 char* ADLParser::get_expr(const char *desc, const char *stop_chars) {
4168 char* expr = NULL;
4169 int paren = 0;
4171 expr = _ptr;
4172 while (paren > 0 || !strchr(stop_chars, _curchar)) {
4173 if (_curchar == '(') { // Down level of nesting
4174 paren++; // Bump the parenthesis counter
4175 next_char(); // maintain the invariant
4176 }
4177 else if (_curchar == ')') { // Up one level of nesting
4178 if (paren == 0) {
4179 // Paren underflow: We didn't encounter the required stop-char.
4180 parse_err(SYNERR, "too many )'s, did not find %s after %s\n",
4181 stop_chars, desc);
4182 return NULL;
4183 }
4184 paren--; // Drop the parenthesis counter
4185 next_char(); // Maintain the invariant
4186 }
4187 else if (_curchar == '"' || _curchar == '\'') {
4188 int qchar = _curchar;
4189 while (true) {
4190 next_char();
4191 if (_curchar == qchar) { next_char(); break; }
4192 if (_curchar == '\\') next_char(); // superquote
4193 if (_curchar == '\n' || _curchar == '\0') {
4194 parse_err(SYNERR, "newline in string in %s\n", desc);
4195 return NULL;
4196 }
4197 }
4198 }
4199 else if (_curchar == '%' && (_ptr[1] == '{' || _ptr[1] == '}')) {
4200 // Make sure we do not stray into the next ADLC-level form.
4201 parse_err(SYNERR, "unexpected %%%c in %s\n", _ptr[1], desc);
4202 return NULL;
4203 }
4204 else if (_curchar == '\0') {
4205 parse_err(SYNERR, "unexpected EOF in %s\n", desc);
4206 return NULL;
4207 }
4208 else {
4209 // Always walk over whitespace, comments, preprocessor directives, etc.
4210 char* pre_skip_ptr = _ptr;
4211 skipws();
4212 // If the parser declined to make progress on whitespace,
4213 // skip the next character, which is therefore NOT whitespace.
4214 if (pre_skip_ptr == _ptr) {
4215 next_char();
4216 } else if (pre_skip_ptr+strlen(pre_skip_ptr) != _ptr+strlen(_ptr)) {
4217 parse_err(SYNERR, "unimplemented: preprocessor must not elide subexpression in %s", desc);
4218 }
4219 }
4220 }
4222 assert(strchr(stop_chars, _curchar), "non-null return must be at stop-char");
4223 *_ptr = '\0'; // Replace ')' or other stop-char with '\0'
4224 return expr;
4225 }
4227 // Helper function around get_expr
4228 // Sets _curchar to '(' so that get_paren_expr will search for a matching ')'
4229 char *ADLParser::get_paren_expr(const char *description, bool include_location) {
4230 int line = linenum();
4231 if (_curchar != '(') // Escape if not valid starting position
4232 return NULL;
4233 next_char(); // Skip the required initial paren.
4234 char *token2 = get_expr(description, ")");
4235 if (_curchar == ')')
4236 next_char(); // Skip required final paren.
4237 int junk = 0;
4238 if (include_location && _AD._adlocation_debug && !is_int_token(token2, junk)) {
4239 // Prepend location descriptor, for debugging.
4240 char* location = get_line_string(line);
4241 char* end_loc = end_line_marker();
4242 char* result = (char *)malloc(strlen(location) + strlen(token2) + strlen(end_loc) + 1);
4243 strcpy(result, location);
4244 strcat(result, token2);
4245 strcat(result, end_loc);
4246 token2 = result;
4247 free(location);
4248 }
4249 return token2;
4250 }
4252 //------------------------------get_ident_common-------------------------------
4253 // Looks for an identifier in the buffer, and turns it into a null terminated
4254 // string(still inside the file buffer). Returns a pointer to the string or
4255 // NULL if some other token is found instead.
4256 char *ADLParser::get_ident_common(bool do_preproc) {
4257 register char c;
4258 char *start; // Pointer to start of token
4259 char *end; // Pointer to end of token
4261 if( _curline == NULL ) // Return NULL at EOF.
4262 return NULL;
4264 skipws_common(do_preproc); // Skip whitespace before identifier
4265 start = end = _ptr; // Start points at first character
4266 end--; // unwind end by one to prepare for loop
4267 do {
4268 end++; // Increment end pointer
4269 c = *end; // Grab character to test
4270 } while ( ((c >= 'a') && (c <= 'z')) || ((c >= 'A') && (c <= 'Z'))
4271 || ((c >= '0') && (c <= '9'))
4272 || ((c == '_')) || ((c == ':')) || ((c == '#')) );
4273 if (start == end) { // We popped out on the first try
4274 parse_err(SYNERR, "identifier expected at %c\n", c);
4275 start = NULL;
4276 }
4277 else {
4278 _curchar = c; // Save the first character of next token
4279 *end = '\0'; // NULL terminate the string in place
4280 }
4281 _ptr = end; // Reset _ptr to point to next char after token
4283 // Make sure we do not try to use #defined identifiers. If start is
4284 // NULL an error was already reported.
4285 if (do_preproc && start != NULL) {
4286 const char* def = _AD.get_preproc_def(start);
4287 if (def != NULL && strcmp(def, start)) {
4288 const char* def1 = def;
4289 const char* def2 = _AD.get_preproc_def(def1);
4290 // implement up to 2 levels of #define
4291 if (def2 != NULL && strcmp(def2, def1)) {
4292 def = def2;
4293 const char* def3 = _AD.get_preproc_def(def2);
4294 if (def3 != NULL && strcmp(def3, def2) && strcmp(def3, def1)) {
4295 parse_err(SYNERR, "unimplemented: using %s defined as %s => %s => %s",
4296 start, def1, def2, def3);
4297 }
4298 }
4299 start = strdup(def);
4300 }
4301 }
4303 return start; // Pointer to token in filebuf
4304 }
4306 //------------------------------get_ident_dup----------------------------------
4307 // Looks for an identifier in the buffer, and returns a duplicate
4308 // or NULL if some other token is found instead.
4309 char *ADLParser::get_ident_dup(void) {
4310 char *ident = get_ident();
4312 // Duplicate an identifier before returning and restore string.
4313 if( ident != NULL ) {
4314 ident = strdup(ident); // Copy the string
4315 *_ptr = _curchar; // and replace Nil with original character
4316 }
4318 return ident;
4319 }
4321 //----------------------get_ident_or_literal_constant--------------------------
4322 // Looks for an identifier in the buffer, or a parenthesized expression.
4323 char *ADLParser::get_ident_or_literal_constant(const char* description) {
4324 char* param = NULL;
4325 skipws();
4326 if (_curchar == '(') {
4327 // Grab a constant expression.
4328 param = get_paren_expr(description);
4329 if (param[0] != '(') {
4330 char* buf = (char*) malloc(strlen(param) + 3);
4331 sprintf(buf, "(%s)", param);
4332 param = buf;
4333 }
4334 assert(is_literal_constant(param),
4335 "expr must be recognizable as a constant");
4336 } else {
4337 param = get_ident();
4338 }
4339 return param;
4340 }
4342 //------------------------------get_rep_var_ident-----------------------------
4343 // Do NOT duplicate,
4344 // Leave nil terminator in buffer
4345 // Preserve initial '$'(s) in string
4346 char *ADLParser::get_rep_var_ident(void) {
4347 // Remember starting point
4348 char *rep_var = _ptr;
4350 // Check for replacement variable indicator '$' and pass if present
4351 if ( _curchar == '$' ) {
4352 next_char();
4353 }
4354 // Check for a subfield indicator, a second '$', and pass if present
4355 if ( _curchar == '$' ) {
4356 next_char();
4357 }
4359 // Check for a control indicator, a third '$':
4360 if ( _curchar == '$' ) {
4361 next_char();
4362 }
4364 // Check for more than three '$'s in sequence, SYNERR
4365 if( _curchar == '$' ) {
4366 parse_err(SYNERR, "Replacement variables and field specifiers can not start with '$$$$'");
4367 next_char();
4368 return NULL;
4369 }
4371 // Nil terminate the variable name following the '$'
4372 char *rep_var_name = get_ident();
4373 assert( rep_var_name != NULL,
4374 "Missing identifier after replacement variable indicator '$'");
4376 return rep_var;
4377 }
4381 //------------------------------get_rep_var_ident_dup-------------------------
4382 // Return the next replacement variable identifier, skipping first '$'
4383 // given a pointer into a line of the buffer.
4384 // Null terminates string, still inside the file buffer,
4385 // Returns a pointer to a copy of the string, or NULL on failure
4386 char *ADLParser::get_rep_var_ident_dup(void) {
4387 if( _curchar != '$' ) return NULL;
4389 next_char(); // Move past the '$'
4390 char *rep_var = _ptr; // Remember starting point
4392 // Check for a subfield indicator, a second '$':
4393 if ( _curchar == '$' ) {
4394 next_char();
4395 }
4397 // Check for a control indicator, a third '$':
4398 if ( _curchar == '$' ) {
4399 next_char();
4400 }
4402 // Check for more than three '$'s in sequence, SYNERR
4403 if( _curchar == '$' ) {
4404 parse_err(SYNERR, "Replacement variables and field specifiers can not start with '$$$$'");
4405 next_char();
4406 return NULL;
4407 }
4409 // Nil terminate the variable name following the '$'
4410 char *rep_var_name = get_ident();
4411 assert( rep_var_name != NULL,
4412 "Missing identifier after replacement variable indicator '$'");
4413 rep_var = strdup(rep_var); // Copy the string
4414 *_ptr = _curchar; // and replace Nil with original character
4416 return rep_var;
4417 }
4420 //------------------------------get_unique_ident------------------------------
4421 // Looks for an identifier in the buffer, terminates it with a NULL,
4422 // and checks that it is unique
4423 char *ADLParser::get_unique_ident(FormDict& dict, const char* nameDescription){
4424 char* ident = get_ident();
4426 if (ident == NULL) {
4427 parse_err(SYNERR, "missing %s identifier at %c\n", nameDescription, _curchar);
4428 }
4429 else {
4430 if (dict[ident] != NULL) {
4431 parse_err(SYNERR, "duplicate name %s for %s\n", ident, nameDescription);
4432 ident = NULL;
4433 }
4434 }
4436 return ident;
4437 }
4440 //------------------------------get_int----------------------------------------
4441 // Looks for a character string integer in the buffer, and turns it into an int
4442 // invokes a parse_err if the next token is not an integer.
4443 // This routine does not leave the integer null-terminated.
4444 int ADLParser::get_int(void) {
4445 register char c;
4446 char *start; // Pointer to start of token
4447 char *end; // Pointer to end of token
4448 int result; // Storage for integer result
4450 if( _curline == NULL ) // Return NULL at EOF.
4451 return 0;
4453 skipws(); // Skip whitespace before identifier
4454 start = end = _ptr; // Start points at first character
4455 c = *end; // Grab character to test
4456 while ((c >= '0') && (c <= '9')
4457 || ((c == '-') && (end == start))) {
4458 end++; // Increment end pointer
4459 c = *end; // Grab character to test
4460 }
4461 if (start == end) { // We popped out on the first try
4462 parse_err(SYNERR, "integer expected at %c\n", c);
4463 result = 0;
4464 }
4465 else {
4466 _curchar = c; // Save the first character of next token
4467 *end = '\0'; // NULL terminate the string in place
4468 result = atoi(start); // Convert the string to an integer
4469 *end = _curchar; // Restore buffer to original condition
4470 }
4472 // Reset _ptr to next char after token
4473 _ptr = end;
4475 return result; // integer
4476 }
4479 //------------------------------get_relation_dup------------------------------
4480 // Looks for a relational operator in the buffer
4481 // invokes a parse_err if the next token is not a relation
4482 // This routine creates a duplicate of the string in the buffer.
4483 char *ADLParser::get_relation_dup(void) {
4484 char *result = NULL; // relational operator being returned
4486 if( _curline == NULL ) // Return NULL at EOF.
4487 return NULL;
4489 skipws(); // Skip whitespace before relation
4490 char *start = _ptr; // Store start of relational operator
4491 char first = *_ptr; // the first character
4492 if( (first == '=') || (first == '!') || (first == '<') || (first == '>') ) {
4493 next_char();
4494 char second = *_ptr; // the second character
4495 if( (second == '=') ) {
4496 next_char();
4497 char tmp = *_ptr;
4498 *_ptr = '\0'; // NULL terminate
4499 result = strdup(start); // Duplicate the string
4500 *_ptr = tmp; // restore buffer
4501 } else {
4502 parse_err(SYNERR, "relational operator expected at %s\n", _ptr);
4503 }
4504 } else {
4505 parse_err(SYNERR, "relational operator expected at %s\n", _ptr);
4506 }
4508 return result;
4509 }
4513 //------------------------------get_oplist-------------------------------------
4514 // Looks for identifier pairs where first must be the name of an operand, and
4515 // second must be a name unique in the scope of this instruction. Stores the
4516 // names with a pointer to the OpClassForm of their type in a local name table.
4517 void ADLParser::get_oplist(NameList ¶meters, FormDict &operands) {
4518 OpClassForm *opclass = NULL;
4519 char *ident = NULL;
4521 do {
4522 next_char(); // skip open paren & comma characters
4523 skipws();
4524 if (_curchar == ')') break;
4526 // Get operand type, and check it against global name table
4527 ident = get_ident();
4528 if (ident == NULL) {
4529 parse_err(SYNERR, "optype identifier expected at %c\n", _curchar);
4530 return;
4531 }
4532 else {
4533 const Form *form = _globalNames[ident];
4534 if( form == NULL ) {
4535 parse_err(SYNERR, "undefined operand type %s\n", ident);
4536 return;
4537 }
4539 // Check for valid operand type
4540 OpClassForm *opc = form->is_opclass();
4541 OperandForm *oper = form->is_operand();
4542 if((oper == NULL) && (opc == NULL)) {
4543 parse_err(SYNERR, "identifier %s not operand type\n", ident);
4544 return;
4545 }
4546 opclass = opc;
4547 }
4548 // Debugging Stuff
4549 if (_AD._adl_debug > 1) fprintf(stderr, "\tOperand Type: %s\t", ident);
4551 // Get name of operand and add it to local name table
4552 if( (ident = get_unique_ident(operands, "operand")) == NULL) {
4553 return;
4554 }
4555 // Parameter names must not be global names.
4556 if( _globalNames[ident] != NULL ) {
4557 parse_err(SYNERR, "Reuse of global name %s as operand.\n",ident);
4558 return;
4559 }
4560 operands.Insert(ident, opclass);
4561 parameters.addName(ident);
4563 // Debugging Stuff
4564 if (_AD._adl_debug > 1) fprintf(stderr, "\tOperand Name: %s\n", ident);
4565 skipws();
4566 } while(_curchar == ',');
4568 if (_curchar != ')') parse_err(SYNERR, "missing ')'\n");
4569 else {
4570 next_char(); // set current character position past the close paren
4571 }
4572 }
4575 //------------------------------get_effectlist---------------------------------
4576 // Looks for identifier pairs where first must be the name of a pre-defined,
4577 // effect, and the second must be the name of an operand defined in the
4578 // operand list of this instruction. Stores the names with a pointer to the
4579 // effect form in a local effects table.
4580 void ADLParser::get_effectlist(FormDict &effects, FormDict &operands, bool& has_call) {
4581 OperandForm *opForm;
4582 Effect *eForm;
4583 char *ident;
4585 do {
4586 next_char(); // skip open paren & comma characters
4587 skipws();
4588 if (_curchar == ')') break;
4590 // Get effect type, and check it against global name table
4591 ident = get_ident();
4592 if (ident == NULL) {
4593 parse_err(SYNERR, "effect type identifier expected at %c\n", _curchar);
4594 return;
4595 }
4596 else {
4597 // Check for valid effect type
4598 const Form *form = _globalNames[ident];
4599 if( form == NULL ) {
4600 parse_err(SYNERR, "undefined effect type %s\n", ident);
4601 return;
4602 }
4603 else {
4604 if( (eForm = form->is_effect()) == NULL) {
4605 parse_err(SYNERR, "identifier %s not effect type\n", ident);
4606 return;
4607 }
4608 }
4609 }
4610 // Debugging Stuff
4611 if (_AD._adl_debug > 1) fprintf(stderr, "\tEffect Type: %s\t", ident);
4612 skipws();
4613 if (eForm->is(Component::CALL)) {
4614 if (_AD._adl_debug > 1) fprintf(stderr, "\n");
4615 has_call = true;
4616 } else {
4617 // Get name of operand and check that it is in the local name table
4618 if( (ident = get_unique_ident(effects, "effect")) == NULL) {
4619 parse_err(SYNERR, "missing operand identifier in effect list\n");
4620 return;
4621 }
4622 const Form *form = operands[ident];
4623 opForm = form ? form->is_operand() : NULL;
4624 if( opForm == NULL ) {
4625 if( form && form->is_opclass() ) {
4626 const char* cname = form->is_opclass()->_ident;
4627 parse_err(SYNERR, "operand classes are illegal in effect lists (found %s %s)\n", cname, ident);
4628 } else {
4629 parse_err(SYNERR, "undefined operand %s in effect list\n", ident);
4630 }
4631 return;
4632 }
4633 // Add the pair to the effects table
4634 effects.Insert(ident, eForm);
4635 // Debugging Stuff
4636 if (_AD._adl_debug > 1) fprintf(stderr, "\tOperand Name: %s\n", ident);
4637 }
4638 skipws();
4639 } while(_curchar == ',');
4641 if (_curchar != ')') parse_err(SYNERR, "missing ')'\n");
4642 else {
4643 next_char(); // set current character position past the close paren
4644 }
4645 }
4648 //-------------------------------preproc_line----------------------------------
4649 // A "#line" keyword has been seen, so parse the rest of the line.
4650 void ADLParser::preproc_line(void) {
4651 int line = get_int();
4652 skipws_no_preproc();
4653 const char* file = NULL;
4654 if (_curchar == '"') {
4655 next_char(); // Move past the initial '"'
4656 file = _ptr;
4657 while (true) {
4658 if (_curchar == '\n') {
4659 parse_err(SYNERR, "missing '\"' at end of #line directive");
4660 return;
4661 }
4662 if (_curchar == '"') {
4663 *_ptr = '\0'; // Terminate the string
4664 next_char();
4665 skipws_no_preproc();
4666 break;
4667 }
4668 next_char();
4669 }
4670 }
4671 ensure_end_of_line();
4672 if (file != NULL)
4673 _AD._ADL_file._name = file;
4674 _buf.set_linenum(line);
4675 }
4677 //------------------------------preproc_define---------------------------------
4678 // A "#define" keyword has been seen, so parse the rest of the line.
4679 void ADLParser::preproc_define(void) {
4680 char* flag = get_ident_no_preproc();
4681 skipws_no_preproc();
4682 // only #define x y is supported for now
4683 char* def = get_ident_no_preproc();
4684 _AD.set_preproc_def(flag, def);
4685 skipws_no_preproc();
4686 if (_curchar != '\n') {
4687 parse_err(SYNERR, "non-identifier in preprocessor definition\n");
4688 }
4689 }
4691 //------------------------------preproc_undef----------------------------------
4692 // An "#undef" keyword has been seen, so parse the rest of the line.
4693 void ADLParser::preproc_undef(void) {
4694 char* flag = get_ident_no_preproc();
4695 skipws_no_preproc();
4696 ensure_end_of_line();
4697 _AD.set_preproc_def(flag, NULL);
4698 }
4702 //------------------------------parse_err--------------------------------------
4703 // Issue a parser error message, and skip to the end of the current line
4704 void ADLParser::parse_err(int flag, const char *fmt, ...) {
4705 va_list args;
4707 va_start(args, fmt);
4708 if (flag == 1)
4709 _AD._syntax_errs += _AD.emit_msg(0, flag, linenum(), fmt, args);
4710 else if (flag == 2)
4711 _AD._semantic_errs += _AD.emit_msg(0, flag, linenum(), fmt, args);
4712 else
4713 _AD._warnings += _AD.emit_msg(0, flag, linenum(), fmt, args);
4715 int error_char = _curchar;
4716 char* error_ptr = _ptr+1;
4717 for(;*_ptr != '\n'; _ptr++) ; // Skip to the end of the current line
4718 _curchar = '\n';
4719 va_end(args);
4720 _AD._no_output = 1;
4722 if (flag == 1) {
4723 char* error_tail = strchr(error_ptr, '\n');
4724 char tem = *error_ptr;
4725 error_ptr[-1] = '\0';
4726 char* error_head = error_ptr-1;
4727 while (error_head > _curline && *error_head) --error_head;
4728 if (error_tail) *error_tail = '\0';
4729 fprintf(stderr, "Error Context: %s>>>%c<<<%s\n",
4730 error_head, error_char, error_ptr);
4731 if (error_tail) *error_tail = '\n';
4732 error_ptr[-1] = tem;
4733 }
4734 }
4736 //---------------------------ensure_start_of_line------------------------------
4737 // A preprocessor directive has been encountered. Be sure it has fallen at
4738 // the beginning of a line, or else report an error.
4739 void ADLParser::ensure_start_of_line(void) {
4740 if (_curchar == '\n') { next_line(); return; }
4741 assert( _ptr >= _curline && _ptr < _curline+strlen(_curline),
4742 "Must be able to find which line we are in" );
4744 for (char *s = _curline; s < _ptr; s++) {
4745 if (*s > ' ') {
4746 parse_err(SYNERR, "'%c' must be at beginning of line\n", _curchar);
4747 break;
4748 }
4749 }
4750 }
4752 //---------------------------ensure_end_of_line--------------------------------
4753 // A preprocessor directive has been parsed. Be sure there is no trailing
4754 // garbage at the end of this line. Set the scan point to the beginning of
4755 // the next line.
4756 void ADLParser::ensure_end_of_line(void) {
4757 skipws_no_preproc();
4758 if (_curchar != '\n' && _curchar != '\0') {
4759 parse_err(SYNERR, "garbage char '%c' at end of line\n", _curchar);
4760 } else {
4761 next_char_or_line();
4762 }
4763 }
4765 //---------------------------handle_preproc------------------------------------
4766 // The '#' character introducing a preprocessor directive has been found.
4767 // Parse the whole directive name (e.g., #define, #endif) and take appropriate
4768 // action. If we are in an "untaken" span of text, simply keep track of
4769 // #ifdef nesting structure, so we can find out when to start taking text
4770 // again. (In this state, we "sort of support" C's #if directives, enough
4771 // to disregard their associated #else and #endif lines.) If we are in a
4772 // "taken" span of text, there are two cases: "#define" and "#undef"
4773 // directives are preserved and passed up to the caller, which eventually
4774 // passes control to the top-level parser loop, which handles #define and
4775 // #undef directly. (This prevents these directives from occurring in
4776 // arbitrary positions in the AD file--we require better structure than C.)
4777 // In the other case, and #ifdef, #ifndef, #else, or #endif is silently
4778 // processed as whitespace, with the "taken" state of the text correctly
4779 // updated. This routine returns "false" exactly in the case of a "taken"
4780 // #define or #undef, which tells the caller that a preprocessor token
4781 // has appeared which must be handled explicitly by the parse loop.
4782 bool ADLParser::handle_preproc_token() {
4783 assert(*_ptr == '#', "must be at start of preproc");
4784 ensure_start_of_line();
4785 next_char();
4786 skipws_no_preproc();
4787 char* start_ident = _ptr;
4788 char* ident = (_curchar == '\n') ? NULL : get_ident_no_preproc();
4789 if (ident == NULL) {
4790 parse_err(SYNERR, "expected preprocessor command, got end of line\n");
4791 } else if (!strcmp(ident, "ifdef") ||
4792 !strcmp(ident, "ifndef")) {
4793 char* flag = get_ident_no_preproc();
4794 ensure_end_of_line();
4795 // Test the identifier only if we are already in taken code:
4796 bool flag_def = preproc_taken() && (_AD.get_preproc_def(flag) != NULL);
4797 bool now_taken = !strcmp(ident, "ifdef") ? flag_def : !flag_def;
4798 begin_if_def(now_taken);
4799 } else if (!strcmp(ident, "if")) {
4800 if (preproc_taken())
4801 parse_err(SYNERR, "unimplemented: #%s %s", ident, _ptr+1);
4802 next_line();
4803 // Intelligently skip this nested C preprocessor directive:
4804 begin_if_def(true);
4805 } else if (!strcmp(ident, "else")) {
4806 ensure_end_of_line();
4807 invert_if_def();
4808 } else if (!strcmp(ident, "endif")) {
4809 ensure_end_of_line();
4810 end_if_def();
4811 } else if (preproc_taken()) {
4812 // pass this token up to the main parser as "#define" or "#undef"
4813 _ptr = start_ident;
4814 _curchar = *--_ptr;
4815 if( _curchar != '#' ) {
4816 parse_err(SYNERR, "no space allowed after # in #define or #undef");
4817 assert(_curchar == '#', "no space allowed after # in #define or #undef");
4818 }
4819 return false;
4820 }
4821 return true;
4822 }
4824 //---------------------------skipws_common-------------------------------------
4825 // Skip whitespace, including comments and newlines, while keeping an accurate
4826 // line count.
4827 // Maybe handle certain preprocessor constructs: #ifdef, #ifndef, #else, #endif
4828 void ADLParser::skipws_common(bool do_preproc) {
4829 char *start = _ptr;
4830 char *next = _ptr + 1;
4832 if (*_ptr == '\0') {
4833 // Check for string terminator
4834 if (_curchar > ' ') return;
4835 if (_curchar == '\n') {
4836 if (!do_preproc) return; // let caller handle the newline
4837 next_line();
4838 _ptr = _curline; next = _ptr + 1;
4839 }
4840 else if (_curchar == '#' ||
4841 (_curchar == '/' && (*next == '/' || *next == '*'))) {
4842 parse_err(SYNERR, "unimplemented: comment token in a funny place");
4843 }
4844 }
4845 while(_curline != NULL) { // Check for end of file
4846 if (*_ptr == '\n') { // keep proper track of new lines
4847 if (!do_preproc) break; // let caller handle the newline
4848 next_line();
4849 _ptr = _curline; next = _ptr + 1;
4850 }
4851 else if ((*_ptr == '/') && (*next == '/')) // C++ comment
4852 do { _ptr++; next++; } while(*_ptr != '\n'); // So go to end of line
4853 else if ((*_ptr == '/') && (*next == '*')) { // C comment
4854 _ptr++; next++;
4855 do {
4856 _ptr++; next++;
4857 if (*_ptr == '\n') { // keep proper track of new lines
4858 next_line(); // skip newlines within comments
4859 if (_curline == NULL) { // check for end of file
4860 parse_err(SYNERR, "end-of-file detected inside comment\n");
4861 break;
4862 }
4863 _ptr = _curline; next = _ptr + 1;
4864 }
4865 } while(!((*_ptr == '*') && (*next == '/'))); // Go to end of comment
4866 _ptr = ++next; next++; // increment _ptr past comment end
4867 }
4868 else if (do_preproc && *_ptr == '#') {
4869 // Note that this calls skipws_common(false) recursively!
4870 bool preproc_handled = handle_preproc_token();
4871 if (!preproc_handled) {
4872 if (preproc_taken()) {
4873 return; // short circuit
4874 }
4875 ++_ptr; // skip the preprocessor character
4876 }
4877 next = _ptr+1;
4878 } else if(*_ptr > ' ' && !(do_preproc && !preproc_taken())) {
4879 break;
4880 }
4881 else if (*_ptr == '"' || *_ptr == '\'') {
4882 assert(do_preproc, "only skip strings if doing preproc");
4883 // skip untaken quoted string
4884 int qchar = *_ptr;
4885 while (true) {
4886 ++_ptr;
4887 if (*_ptr == qchar) { ++_ptr; break; }
4888 if (*_ptr == '\\') ++_ptr;
4889 if (*_ptr == '\n' || *_ptr == '\0') {
4890 parse_err(SYNERR, "newline in string");
4891 break;
4892 }
4893 }
4894 next = _ptr + 1;
4895 }
4896 else { ++_ptr; ++next; }
4897 }
4898 if( _curline != NULL ) // at end of file _curchar isn't valid
4899 _curchar = *_ptr; // reset _curchar to maintain invariant
4900 }
4902 //---------------------------cur_char-----------------------------------------
4903 char ADLParser::cur_char() {
4904 return (_curchar);
4905 }
4907 //---------------------------next_char-----------------------------------------
4908 void ADLParser::next_char() {
4909 if (_curchar == '\n') parse_err(WARN, "must call next_line!");
4910 _curchar = *++_ptr;
4911 // if ( _curchar == '\n' ) {
4912 // next_line();
4913 // }
4914 }
4916 //---------------------------next_char_or_line---------------------------------
4917 void ADLParser::next_char_or_line() {
4918 if ( _curchar != '\n' ) {
4919 _curchar = *++_ptr;
4920 } else {
4921 next_line();
4922 _ptr = _curline;
4923 _curchar = *_ptr; // maintain invariant
4924 }
4925 }
4927 //---------------------------next_line-----------------------------------------
4928 void ADLParser::next_line() {
4929 _curline = _buf.get_line();
4930 _curchar = ' ';
4931 }
4933 //------------------------get_line_string--------------------------------------
4934 // Prepended location descriptor, for debugging.
4935 // Must return a malloced string (that can be freed if desired).
4936 char* ADLParser::get_line_string(int linenum) {
4937 const char* file = _AD._ADL_file._name;
4938 int line = linenum ? linenum : this->linenum();
4939 char* location = (char *)malloc(strlen(file) + 100);
4940 sprintf(location, "\n#line %d \"%s\"\n", line, file);
4941 return location;
4942 }
4944 //-------------------------is_literal_constant---------------------------------
4945 bool ADLParser::is_literal_constant(const char *param) {
4946 if (param[0] == 0) return false; // null string
4947 if (param[0] == '(') return true; // parenthesized expression
4948 if (param[0] == '0' && (param[1] == 'x' || param[1] == 'X')) {
4949 // Make sure it's a hex constant.
4950 int i = 2;
4951 do {
4952 if( !ADLParser::is_hex_digit(*(param+i)) ) return false;
4953 ++i;
4954 } while( *(param+i) != 0 );
4955 return true;
4956 }
4957 return false;
4958 }
4960 //---------------------------is_hex_digit--------------------------------------
4961 bool ADLParser::is_hex_digit(char digit) {
4962 return ((digit >= '0') && (digit <= '9'))
4963 ||((digit >= 'a') && (digit <= 'f'))
4964 ||((digit >= 'A') && (digit <= 'F'));
4965 }
4967 //---------------------------is_int_token--------------------------------------
4968 bool ADLParser::is_int_token(const char* token, int& intval) {
4969 const char* cp = token;
4970 while (*cp != '\0' && *cp <= ' ') cp++;
4971 if (*cp == '-') cp++;
4972 int ndigit = 0;
4973 while (*cp >= '0' && *cp <= '9') { cp++; ndigit++; }
4974 while (*cp != '\0' && *cp <= ' ') cp++;
4975 if (ndigit == 0 || *cp != '\0') {
4976 return false;
4977 }
4978 intval = atoi(token);
4979 return true;
4980 }
4982 static const char* skip_expr_ws(const char* str) {
4983 const char * cp = str;
4984 while (cp[0]) {
4985 if (cp[0] <= ' ') {
4986 ++cp;
4987 } else if (cp[0] == '#') {
4988 ++cp;
4989 while (cp[0] == ' ') ++cp;
4990 assert(0 == strncmp(cp, "line", 4), "must be a #line directive");
4991 const char* eol = strchr(cp, '\n');
4992 assert(eol != NULL, "must find end of line");
4993 if (eol == NULL) eol = cp + strlen(cp);
4994 cp = eol;
4995 } else {
4996 break;
4997 }
4998 }
4999 return cp;
5000 }
5002 //-----------------------equivalent_expressions--------------------------------
5003 bool ADLParser::equivalent_expressions(const char* str1, const char* str2) {
5004 if (str1 == str2)
5005 return true;
5006 else if (str1 == NULL || str2 == NULL)
5007 return false;
5008 const char* cp1 = str1;
5009 const char* cp2 = str2;
5010 char in_quote = '\0';
5011 while (cp1[0] && cp2[0]) {
5012 if (!in_quote) {
5013 // skip spaces and/or cpp directives
5014 const char* cp1a = skip_expr_ws(cp1);
5015 const char* cp2a = skip_expr_ws(cp2);
5016 if (cp1a > cp1 && cp2a > cp2) {
5017 cp1 = cp1a; cp2 = cp2a;
5018 continue;
5019 }
5020 if (cp1a > cp1 || cp2a > cp2) break; // fail
5021 }
5022 // match one non-space char
5023 if (cp1[0] != cp2[0]) break; // fail
5024 char ch = cp1[0];
5025 cp1++; cp2++;
5026 // watch for quotes
5027 if (in_quote && ch == '\\') {
5028 if (cp1[0] != cp2[0]) break; // fail
5029 if (!cp1[0]) break;
5030 cp1++; cp2++;
5031 }
5032 if (in_quote && ch == in_quote) {
5033 in_quote = '\0';
5034 } else if (!in_quote && (ch == '"' || ch == '\'')) {
5035 in_quote = ch;
5036 }
5037 }
5038 return (!cp1[0] && !cp2[0]);
5039 }
5042 //-------------------------------trim------------------------------------------
5043 void ADLParser::trim(char* &token) {
5044 while (*token <= ' ') token++;
5045 char* end = token + strlen(token);
5046 while (end > token && *(end-1) <= ' ') --end;
5047 *end = '\0';
5048 }