Wed, 14 Sep 2011 09:22:51 +0200
7077312: Provide a CALL effect for instruct declaration in the ad file
Summary: abstracted way to declare that the MachNode has the effect of a call (kills caller save registers, preserves callee save registers)
Reviewed-by: twisti, never
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 // Add reg_class interpreter_frame_pointer_reg
986 if( _AD._register != NULL ) {
987 RegClass *reg_class = _AD._register->addRegClass("interpreter_frame_pointer_reg");
988 char *interpreter_frame_pointer_reg = frame->_interpreter_frame_pointer_reg;
989 if( interpreter_frame_pointer_reg != NULL ) {
990 RegDef *regDef = _AD._register->getRegDef(interpreter_frame_pointer_reg);
991 reg_class->addReg(regDef); // add regDef to regClass
992 }
993 }
994 }
995 if (strcmp(token,"inline_cache_reg")==0) {
996 inline_cache_parse(frame, false);
997 // Add reg_class inline_cache_reg
998 if( _AD._register != NULL ) {
999 RegClass *reg_class = _AD._register->addRegClass("inline_cache_reg");
1000 char *inline_cache_reg = frame->_inline_cache_reg;
1001 if( inline_cache_reg != NULL ) {
1002 RegDef *regDef = _AD._register->getRegDef(inline_cache_reg);
1003 reg_class->addReg(regDef); // add regDef to regClass
1004 }
1005 }
1006 }
1007 if (strcmp(token,"compiler_method_oop_reg")==0) {
1008 parse_err(WARN, "Using obsolete Token, compiler_method_oop_reg");
1009 skipws();
1010 }
1011 if (strcmp(token,"interpreter_method_oop_reg")==0) {
1012 interpreter_method_oop_parse(frame, false);
1013 // Add reg_class interpreter_method_oop_reg
1014 if( _AD._register != NULL ) {
1015 RegClass *reg_class = _AD._register->addRegClass("interpreter_method_oop_reg");
1016 char *method_oop_reg = frame->_interpreter_method_oop_reg;
1017 if( method_oop_reg != NULL ) {
1018 RegDef *regDef = _AD._register->getRegDef(method_oop_reg);
1019 reg_class->addReg(regDef); // add regDef to regClass
1020 }
1021 }
1022 }
1023 if (strcmp(token,"cisc_spilling_operand_name")==0) {
1024 cisc_spilling_operand_name_parse(frame, false);
1025 }
1026 if (strcmp(token,"stack_alignment")==0) {
1027 stack_alignment_parse(frame);
1028 }
1029 if (strcmp(token,"return_addr")==0) {
1030 return_addr_parse(frame, false);
1031 }
1032 if (strcmp(token,"in_preserve_stack_slots")==0) {
1033 preserve_stack_parse(frame);
1034 }
1035 if (strcmp(token,"out_preserve_stack_slots")==0) {
1036 parse_err(WARN, "Using obsolete token, out_preserve_stack_slots");
1037 skipws();
1038 }
1039 if (strcmp(token,"varargs_C_out_slots_killed")==0) {
1040 frame->_varargs_C_out_slots_killed = parse_one_arg("varargs C out slots killed");
1041 }
1042 if (strcmp(token,"calling_convention")==0) {
1043 frame->_calling_convention = calling_convention_parse();
1044 }
1045 if (strcmp(token,"return_value")==0) {
1046 frame->_return_value = return_value_parse();
1047 }
1048 if (strcmp(token,"c_frame_pointer")==0) {
1049 frame_pointer_parse(frame, true);
1050 }
1051 if (strcmp(token,"c_return_addr")==0) {
1052 return_addr_parse(frame, true);
1053 }
1054 if (strcmp(token,"c_calling_convention")==0) {
1055 frame->_c_calling_convention = calling_convention_parse();
1056 }
1057 if (strcmp(token,"c_return_value")==0) {
1058 frame->_c_return_value = return_value_parse();
1059 }
1061 skipws();
1062 }
1063 }
1064 else {
1065 parse_err(SYNERR, "Missing %c{ ... %c} block after encode keyword.\n",'%','%');
1066 return;
1067 }
1068 // All Java versions are required, native versions are optional
1069 if(frame->_frame_pointer == NULL) {
1070 parse_err(SYNERR, "missing frame pointer definition in frame section.\n");
1071 return;
1072 }
1073 // !!!!! !!!!!
1074 // if(frame->_interpreter_frame_ptr_reg == NULL) {
1075 // parse_err(SYNERR, "missing interpreter frame pointer definition in frame section.\n");
1076 // return;
1077 // }
1078 if(frame->_alignment == NULL) {
1079 parse_err(SYNERR, "missing alignment definition in frame section.\n");
1080 return;
1081 }
1082 if(frame->_return_addr == NULL) {
1083 parse_err(SYNERR, "missing return address location in frame section.\n");
1084 return;
1085 }
1086 if(frame->_in_preserve_slots == NULL) {
1087 parse_err(SYNERR, "missing stack slot preservation definition in frame section.\n");
1088 return;
1089 }
1090 if(frame->_varargs_C_out_slots_killed == NULL) {
1091 parse_err(SYNERR, "missing varargs C out slots killed definition in frame section.\n");
1092 return;
1093 }
1094 if(frame->_calling_convention == NULL) {
1095 parse_err(SYNERR, "missing calling convention definition in frame section.\n");
1096 return;
1097 }
1098 if(frame->_return_value == NULL) {
1099 parse_err(SYNERR, "missing return value definition in frame section.\n");
1100 return;
1101 }
1102 // Fill natives in identically with the Java versions if not present.
1103 if(frame->_c_frame_pointer == NULL) {
1104 frame->_c_frame_pointer = frame->_frame_pointer;
1105 }
1106 if(frame->_c_return_addr == NULL) {
1107 frame->_c_return_addr = frame->_return_addr;
1108 frame->_c_return_addr_loc = frame->_return_addr_loc;
1109 }
1110 if(frame->_c_calling_convention == NULL) {
1111 frame->_c_calling_convention = frame->_calling_convention;
1112 }
1113 if(frame->_c_return_value == NULL) {
1114 frame->_c_return_value = frame->_return_value;
1115 }
1117 // Debug Stuff
1118 if (_AD._adl_debug > 1) fprintf(stderr,"Frame Form: %s\n", desc);
1120 // Create the EncodeForm for the architecture description.
1121 _AD.addForm(frame);
1122 // skipws();
1123 }
1125 //------------------------------stack_dir_parse--------------------------------
1126 void ADLParser::stack_dir_parse(FrameForm *frame) {
1127 char *direction = parse_one_arg("stack direction entry");
1128 if (strcmp(direction, "TOWARDS_LOW") == 0) {
1129 frame->_direction = false;
1130 }
1131 else if (strcmp(direction, "TOWARDS_HIGH") == 0) {
1132 frame->_direction = true;
1133 }
1134 else {
1135 parse_err(SYNERR, "invalid value inside stack direction entry.\n");
1136 return;
1137 }
1138 }
1140 //------------------------------sync_stack_slots_parse-------------------------
1141 void ADLParser::sync_stack_slots_parse(FrameForm *frame) {
1142 // Assign value into frame form
1143 frame->_sync_stack_slots = parse_one_arg("sync stack slots entry");
1144 }
1146 //------------------------------frame_pointer_parse----------------------------
1147 void ADLParser::frame_pointer_parse(FrameForm *frame, bool native) {
1148 char *frame_pointer = parse_one_arg("frame pointer entry");
1149 // Assign value into frame form
1150 if (native) { frame->_c_frame_pointer = frame_pointer; }
1151 else { frame->_frame_pointer = frame_pointer; }
1152 }
1154 //------------------------------interpreter_frame_pointer_parse----------------------------
1155 void ADLParser::interpreter_frame_pointer_parse(FrameForm *frame, bool native) {
1156 frame->_interpreter_frame_pointer_reg = parse_one_arg("interpreter frame pointer entry");
1157 }
1159 //------------------------------inline_cache_parse-----------------------------
1160 void ADLParser::inline_cache_parse(FrameForm *frame, bool native) {
1161 frame->_inline_cache_reg = parse_one_arg("inline cache reg entry");
1162 }
1164 //------------------------------interpreter_method_oop_parse------------------
1165 void ADLParser::interpreter_method_oop_parse(FrameForm *frame, bool native) {
1166 frame->_interpreter_method_oop_reg = parse_one_arg("method oop reg entry");
1167 }
1169 //------------------------------cisc_spilling_operand_parse---------------------
1170 void ADLParser::cisc_spilling_operand_name_parse(FrameForm *frame, bool native) {
1171 frame->_cisc_spilling_operand_name = parse_one_arg("cisc spilling operand name");
1172 }
1174 //------------------------------stack_alignment_parse--------------------------
1175 void ADLParser::stack_alignment_parse(FrameForm *frame) {
1176 char *alignment = parse_one_arg("stack alignment entry");
1177 // Assign value into frame
1178 frame->_alignment = alignment;
1179 }
1181 //------------------------------parse_one_arg-------------------------------
1182 char *ADLParser::parse_one_arg(const char *description) {
1183 char *token = NULL;
1184 if(_curchar == '(') {
1185 next_char();
1186 skipws();
1187 token = get_expr(description, ")");
1188 if (token == NULL) {
1189 parse_err(SYNERR, "missing value inside %s.\n", description);
1190 return NULL;
1191 }
1192 next_char(); // skip the close paren
1193 if(_curchar != ';') { // check for semi-colon
1194 parse_err(SYNERR, "missing %c in.\n", ';', description);
1195 return NULL;
1196 }
1197 next_char(); // skip the semi-colon
1198 }
1199 else {
1200 parse_err(SYNERR, "Missing %c in.\n", '(', description);
1201 return NULL;
1202 }
1204 trim(token);
1205 return token;
1206 }
1208 //------------------------------return_addr_parse------------------------------
1209 void ADLParser::return_addr_parse(FrameForm *frame, bool native) {
1210 bool in_register = true;
1211 if(_curchar == '(') {
1212 next_char();
1213 skipws();
1214 char *token = get_ident();
1215 if (token == NULL) {
1216 parse_err(SYNERR, "missing value inside return address entry.\n");
1217 return;
1218 }
1219 // check for valid values for stack/register
1220 if (strcmp(token, "REG") == 0) {
1221 in_register = true;
1222 }
1223 else if (strcmp(token, "STACK") == 0) {
1224 in_register = false;
1225 }
1226 else {
1227 parse_err(SYNERR, "invalid value inside return_address entry.\n");
1228 return;
1229 }
1230 if (native) { frame->_c_return_addr_loc = in_register; }
1231 else { frame->_return_addr_loc = in_register; }
1233 // Parse expression that specifies register or stack position
1234 skipws();
1235 char *token2 = get_expr("return address entry", ")");
1236 if (token2 == NULL) {
1237 parse_err(SYNERR, "missing value inside return address entry.\n");
1238 return;
1239 }
1240 next_char(); // skip the close paren
1241 if (native) { frame->_c_return_addr = token2; }
1242 else { frame->_return_addr = token2; }
1244 if(_curchar != ';') { // check for semi-colon
1245 parse_err(SYNERR, "missing %c in return address entry.\n", ';');
1246 return;
1247 }
1248 next_char(); // skip the semi-colon
1249 }
1250 else {
1251 parse_err(SYNERR, "Missing %c in return_address entry.\n", '(');
1252 }
1253 }
1255 //------------------------------preserve_stack_parse---------------------------
1256 void ADLParser::preserve_stack_parse(FrameForm *frame) {
1257 if(_curchar == '(') {
1258 char *token = get_paren_expr("preserve_stack_slots");
1259 frame->_in_preserve_slots = token;
1261 if(_curchar != ';') { // check for semi-colon
1262 parse_err(SYNERR, "missing %c in preserve stack slot entry.\n", ';');
1263 return;
1264 }
1265 next_char(); // skip the semi-colon
1266 }
1267 else {
1268 parse_err(SYNERR, "Missing %c in preserve stack slot entry.\n", '(');
1269 }
1270 }
1272 //------------------------------calling_convention_parse-----------------------
1273 char *ADLParser::calling_convention_parse() {
1274 char *desc = NULL; // String representation of calling_convention
1276 skipws(); // Skip leading whitespace
1277 if ( (desc = find_cpp_block("calling convention block")) == NULL ) {
1278 parse_err(SYNERR, "incorrect or missing block for 'calling_convention'.\n");
1279 }
1280 return desc;
1281 }
1283 //------------------------------return_value_parse-----------------------------
1284 char *ADLParser::return_value_parse() {
1285 char *desc = NULL; // String representation of calling_convention
1287 skipws(); // Skip leading whitespace
1288 if ( (desc = find_cpp_block("return value block")) == NULL ) {
1289 parse_err(SYNERR, "incorrect or missing block for 'return_value'.\n");
1290 }
1291 return desc;
1292 }
1294 //------------------------------ins_pipe_parse---------------------------------
1295 void ADLParser::ins_pipe_parse(InstructForm &instr) {
1296 char * ident;
1298 skipws();
1299 if ( _curchar != '(' ) { // Check for delimiter
1300 parse_err(SYNERR, "missing \"(\" in ins_pipe definition\n");
1301 return;
1302 }
1304 next_char();
1305 ident = get_ident(); // Grab next identifier
1307 if (ident == NULL) {
1308 parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
1309 return;
1310 }
1312 skipws();
1313 if ( _curchar != ')' ) { // Check for delimiter
1314 parse_err(SYNERR, "missing \")\" in ins_pipe definition\n");
1315 return;
1316 }
1318 next_char(); // skip the close paren
1319 if(_curchar != ';') { // check for semi-colon
1320 parse_err(SYNERR, "missing %c in return value entry.\n", ';');
1321 return;
1322 }
1323 next_char(); // skip the semi-colon
1325 // Check ident for validity
1326 if (_AD._pipeline && !_AD._pipeline->_classlist.search(ident)) {
1327 parse_err(SYNERR, "\"%s\" is not a valid pipeline class\n", ident);
1328 return;
1329 }
1331 // Add this instruction to the list in the pipeline class
1332 _AD._pipeline->_classdict[ident]->is_pipeclass()->_instructs.addName(instr._ident);
1334 // Set the name of the pipeline class in the instruction
1335 instr._ins_pipe = ident;
1336 return;
1337 }
1339 //------------------------------pipe_parse-------------------------------------
1340 void ADLParser::pipe_parse(void) {
1341 PipelineForm *pipeline; // Encode class for instruction/operand
1342 char * ident;
1344 pipeline = new PipelineForm(); // Build new Source object
1345 _AD.addForm(pipeline);
1347 skipws(); // Skip leading whitespace
1348 // Check for block delimiter
1349 if ( (_curchar != '%')
1350 || ( next_char(), (_curchar != '{')) ) {
1351 parse_err(SYNERR, "missing '%{' in pipeline definition\n");
1352 return;
1353 }
1354 next_char(); // Maintain the invariant
1355 do {
1356 ident = get_ident(); // Grab next identifier
1357 if (ident == NULL) {
1358 parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
1359 continue;
1360 }
1361 if (!strcmp(ident, "resources" )) resource_parse(*pipeline);
1362 else if (!strcmp(ident, "pipe_desc" )) pipe_desc_parse(*pipeline);
1363 else if (!strcmp(ident, "pipe_class")) pipe_class_parse(*pipeline);
1364 else if (!strcmp(ident, "define")) {
1365 skipws();
1366 if ( (_curchar != '%')
1367 || ( next_char(), (_curchar != '{')) ) {
1368 parse_err(SYNERR, "expected '%{'\n");
1369 return;
1370 }
1371 next_char(); skipws();
1373 char *node_class = get_ident();
1374 if (node_class == NULL) {
1375 parse_err(SYNERR, "expected identifier, found \"%c\"\n", _curchar);
1376 return;
1377 }
1379 skipws();
1380 if (_curchar != ',' && _curchar != '=') {
1381 parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1382 break;
1383 }
1384 next_char(); skipws();
1386 char *pipe_class = get_ident();
1387 if (pipe_class == NULL) {
1388 parse_err(SYNERR, "expected identifier, found \"%c\"\n", _curchar);
1389 return;
1390 }
1391 if (_curchar != ';' ) {
1392 parse_err(SYNERR, "expected `;`, found '%c'\n", _curchar);
1393 break;
1394 }
1395 next_char(); // Skip over semi-colon
1397 skipws();
1398 if ( (_curchar != '%')
1399 || ( next_char(), (_curchar != '}')) ) {
1400 parse_err(SYNERR, "expected '%%}', found \"%c\"\n", _curchar);
1401 }
1402 next_char();
1404 // Check ident for validity
1405 if (_AD._pipeline && !_AD._pipeline->_classlist.search(pipe_class)) {
1406 parse_err(SYNERR, "\"%s\" is not a valid pipeline class\n", pipe_class);
1407 return;
1408 }
1410 // Add this machine node to the list in the pipeline class
1411 _AD._pipeline->_classdict[pipe_class]->is_pipeclass()->_instructs.addName(node_class);
1413 MachNodeForm *machnode = new MachNodeForm(node_class); // Create new machnode form
1414 machnode->_machnode_pipe = pipe_class;
1416 _AD.addForm(machnode);
1417 }
1418 else if (!strcmp(ident, "attributes")) {
1419 bool vsi_seen = false, bhds_seen = false;
1421 skipws();
1422 if ( (_curchar != '%')
1423 || ( next_char(), (_curchar != '{')) ) {
1424 parse_err(SYNERR, "expected '%{'\n");
1425 return;
1426 }
1427 next_char(); skipws();
1429 while (_curchar != '%') {
1430 ident = get_ident();
1431 if (ident == NULL)
1432 break;
1434 if (!strcmp(ident, "variable_size_instructions")) {
1435 skipws();
1436 if (_curchar == ';') {
1437 next_char(); skipws();
1438 }
1440 pipeline->_variableSizeInstrs = true;
1441 vsi_seen = true;
1442 continue;
1443 }
1445 if (!strcmp(ident, "fixed_size_instructions")) {
1446 skipws();
1447 if (_curchar == ';') {
1448 next_char(); skipws();
1449 }
1451 pipeline->_variableSizeInstrs = false;
1452 vsi_seen = true;
1453 continue;
1454 }
1456 if (!strcmp(ident, "branch_has_delay_slot")) {
1457 skipws();
1458 if (_curchar == ';') {
1459 next_char(); skipws();
1460 }
1462 pipeline->_branchHasDelaySlot = true;
1463 bhds_seen = true;
1464 continue;
1465 }
1467 if (!strcmp(ident, "max_instructions_per_bundle")) {
1468 skipws();
1469 if (_curchar != '=') {
1470 parse_err(SYNERR, "expected `=`\n");
1471 break;
1472 }
1474 next_char(); skipws();
1475 pipeline->_maxInstrsPerBundle = get_int();
1476 skipws();
1478 if (_curchar == ';') {
1479 next_char(); skipws();
1480 }
1482 continue;
1483 }
1485 if (!strcmp(ident, "max_bundles_per_cycle")) {
1486 skipws();
1487 if (_curchar != '=') {
1488 parse_err(SYNERR, "expected `=`\n");
1489 break;
1490 }
1492 next_char(); skipws();
1493 pipeline->_maxBundlesPerCycle = get_int();
1494 skipws();
1496 if (_curchar == ';') {
1497 next_char(); skipws();
1498 }
1500 continue;
1501 }
1503 if (!strcmp(ident, "instruction_unit_size")) {
1504 skipws();
1505 if (_curchar != '=') {
1506 parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1507 break;
1508 }
1510 next_char(); skipws();
1511 pipeline->_instrUnitSize = get_int();
1512 skipws();
1514 if (_curchar == ';') {
1515 next_char(); skipws();
1516 }
1518 continue;
1519 }
1521 if (!strcmp(ident, "bundle_unit_size")) {
1522 skipws();
1523 if (_curchar != '=') {
1524 parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1525 break;
1526 }
1528 next_char(); skipws();
1529 pipeline->_bundleUnitSize = get_int();
1530 skipws();
1532 if (_curchar == ';') {
1533 next_char(); skipws();
1534 }
1536 continue;
1537 }
1539 if (!strcmp(ident, "instruction_fetch_unit_size")) {
1540 skipws();
1541 if (_curchar != '=') {
1542 parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1543 break;
1544 }
1546 next_char(); skipws();
1547 pipeline->_instrFetchUnitSize = get_int();
1548 skipws();
1550 if (_curchar == ';') {
1551 next_char(); skipws();
1552 }
1554 continue;
1555 }
1557 if (!strcmp(ident, "instruction_fetch_units")) {
1558 skipws();
1559 if (_curchar != '=') {
1560 parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1561 break;
1562 }
1564 next_char(); skipws();
1565 pipeline->_instrFetchUnits = get_int();
1566 skipws();
1568 if (_curchar == ';') {
1569 next_char(); skipws();
1570 }
1572 continue;
1573 }
1575 if (!strcmp(ident, "nops")) {
1576 skipws();
1577 if (_curchar != '(') {
1578 parse_err(SYNERR, "expected `(`, found '%c'\n", _curchar);
1579 break;
1580 }
1582 next_char(); skipws();
1584 while (_curchar != ')') {
1585 ident = get_ident();
1586 if (ident == NULL) {
1587 parse_err(SYNERR, "expected identifier for nop instruction, found '%c'\n", _curchar);
1588 break;
1589 }
1591 pipeline->_noplist.addName(ident);
1592 pipeline->_nopcnt++;
1593 skipws();
1595 if (_curchar == ',') {
1596 next_char(); skipws();
1597 }
1598 }
1600 next_char(); skipws();
1602 if (_curchar == ';') {
1603 next_char(); skipws();
1604 }
1606 continue;
1607 }
1609 parse_err(SYNERR, "unknown specifier \"%s\"\n", ident);
1610 }
1612 if ( (_curchar != '%')
1613 || ( next_char(), (_curchar != '}')) ) {
1614 parse_err(SYNERR, "expected '%}', found \"%c\"\n", _curchar);
1615 }
1616 next_char(); skipws();
1618 if (pipeline->_maxInstrsPerBundle == 0)
1619 parse_err(SYNERR, "\"max_instructions_per_bundle\" unspecified\n");
1620 if (pipeline->_instrUnitSize == 0 && pipeline->_bundleUnitSize == 0)
1621 parse_err(SYNERR, "\"instruction_unit_size\" and \"bundle_unit_size\" unspecified\n");
1622 if (pipeline->_instrFetchUnitSize == 0)
1623 parse_err(SYNERR, "\"instruction_fetch_unit_size\" unspecified\n");
1624 if (pipeline->_instrFetchUnits == 0)
1625 parse_err(SYNERR, "\"instruction_fetch_units\" unspecified\n");
1626 if (!vsi_seen)
1627 parse_err(SYNERR, "\"variable_size_instruction\" or \"fixed_size_instruction\" unspecified\n");
1628 }
1629 else { // Done with staticly defined parts of instruction definition
1630 parse_err(SYNERR, "expected one of \"resources\", \"pipe_desc\", \"pipe_class\", found \"%s\"\n", ident);
1631 return;
1632 }
1633 skipws();
1634 if (_curchar == ';')
1635 skipws();
1636 } while(_curchar != '%');
1638 next_char();
1639 if (_curchar != '}') {
1640 parse_err(SYNERR, "missing \"%}\" in pipeline definition\n");
1641 return;
1642 }
1644 next_char();
1645 }
1647 //------------------------------resource_parse----------------------------
1648 void ADLParser::resource_parse(PipelineForm &pipeline) {
1649 ResourceForm *resource;
1650 char * ident;
1651 char * expr;
1652 unsigned mask;
1653 pipeline._rescount = 0;
1655 skipws(); // Skip leading whitespace
1657 if (_curchar != '(') {
1658 parse_err(SYNERR, "missing \"(\" in resource definition\n");
1659 return;
1660 }
1662 do {
1663 next_char(); // Skip "(" or ","
1664 ident = get_ident(); // Grab next identifier
1666 if (ident == NULL) {
1667 parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1668 return;
1669 }
1670 skipws();
1672 if (_curchar != '=') {
1673 mask = (1 << pipeline._rescount++);
1674 }
1675 else {
1676 next_char(); skipws();
1677 expr = get_ident(); // Grab next identifier
1678 if (expr == NULL) {
1679 parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1680 return;
1681 }
1682 resource = (ResourceForm *) pipeline._resdict[expr];
1683 if (resource == NULL) {
1684 parse_err(SYNERR, "resource \"%s\" is not defined\n", expr);
1685 return;
1686 }
1687 mask = resource->mask();
1689 skipws();
1690 while (_curchar == '|') {
1691 next_char(); skipws();
1693 expr = get_ident(); // Grab next identifier
1694 if (expr == NULL) {
1695 parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1696 return;
1697 }
1699 resource = (ResourceForm *) pipeline._resdict[expr]; // Look up the value
1700 if (resource == NULL) {
1701 parse_err(SYNERR, "resource \"%s\" is not defined\n", expr);
1702 return;
1703 }
1705 mask |= resource->mask();
1706 skipws();
1707 }
1708 }
1710 resource = new ResourceForm(mask);
1712 pipeline._resdict.Insert(ident, resource);
1713 pipeline._reslist.addName(ident);
1714 } while (_curchar == ',');
1716 if (_curchar != ')') {
1717 parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
1718 return;
1719 }
1721 next_char(); // Skip ")"
1722 if (_curchar == ';')
1723 next_char(); // Skip ";"
1724 }
1726 //------------------------------resource_parse----------------------------
1727 void ADLParser::pipe_desc_parse(PipelineForm &pipeline) {
1728 char * ident;
1730 skipws(); // Skip leading whitespace
1732 if (_curchar != '(') {
1733 parse_err(SYNERR, "missing \"(\" in pipe_desc definition\n");
1734 return;
1735 }
1737 do {
1738 next_char(); // Skip "(" or ","
1739 ident = get_ident(); // Grab next identifier
1740 if (ident == NULL) {
1741 parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1742 return;
1743 }
1745 // Add the name to the list
1746 pipeline._stages.addName(ident);
1747 pipeline._stagecnt++;
1749 skipws();
1750 } while (_curchar == ',');
1752 if (_curchar != ')') {
1753 parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
1754 return;
1755 }
1757 next_char(); // Skip ")"
1758 if (_curchar == ';')
1759 next_char(); // Skip ";"
1760 }
1762 //------------------------------pipe_class_parse--------------------------
1763 void ADLParser::pipe_class_parse(PipelineForm &pipeline) {
1764 PipeClassForm *pipe_class;
1765 char * ident;
1766 char * stage;
1767 char * read_or_write;
1768 int is_write;
1769 int is_read;
1770 OperandForm *oper;
1772 skipws(); // Skip leading whitespace
1774 ident = get_ident(); // Grab next identifier
1776 if (ident == NULL) {
1777 parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1778 return;
1779 }
1781 // Create a record for the pipe_class
1782 pipe_class = new PipeClassForm(ident, ++pipeline._classcnt);
1783 pipeline._classdict.Insert(ident, pipe_class);
1784 pipeline._classlist.addName(ident);
1786 // Then get the operands
1787 skipws();
1788 if (_curchar != '(') {
1789 parse_err(SYNERR, "missing \"(\" in pipe_class definition\n");
1790 }
1791 // Parse the operand list
1792 else get_oplist(pipe_class->_parameters, pipe_class->_localNames);
1793 skipws(); // Skip leading whitespace
1794 // Check for block delimiter
1795 if ( (_curchar != '%')
1796 || ( next_char(), (_curchar != '{')) ) {
1797 parse_err(SYNERR, "missing \"%{\" in pipe_class definition\n");
1798 return;
1799 }
1800 next_char();
1802 do {
1803 ident = get_ident(); // Grab next identifier
1804 if (ident == NULL) {
1805 parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1806 continue;
1807 }
1808 skipws();
1810 if (!strcmp(ident, "fixed_latency")) {
1811 skipws();
1812 if (_curchar != '(') {
1813 parse_err(SYNERR, "missing \"(\" in latency definition\n");
1814 return;
1815 }
1816 next_char(); skipws();
1817 if( !isdigit(_curchar) ) {
1818 parse_err(SYNERR, "number expected for \"%c\" in latency definition\n", _curchar);
1819 return;
1820 }
1821 int fixed_latency = get_int();
1822 skipws();
1823 if (_curchar != ')') {
1824 parse_err(SYNERR, "missing \")\" in latency definition\n");
1825 return;
1826 }
1827 next_char(); skipws();
1828 if (_curchar != ';') {
1829 parse_err(SYNERR, "missing \";\" in latency definition\n");
1830 return;
1831 }
1833 pipe_class->setFixedLatency(fixed_latency);
1834 next_char(); skipws();
1835 continue;
1836 }
1838 if (!strcmp(ident, "zero_instructions") ||
1839 !strcmp(ident, "no_instructions")) {
1840 skipws();
1841 if (_curchar != ';') {
1842 parse_err(SYNERR, "missing \";\" in latency definition\n");
1843 return;
1844 }
1846 pipe_class->setInstructionCount(0);
1847 next_char(); skipws();
1848 continue;
1849 }
1851 if (!strcmp(ident, "one_instruction_with_delay_slot") ||
1852 !strcmp(ident, "single_instruction_with_delay_slot")) {
1853 skipws();
1854 if (_curchar != ';') {
1855 parse_err(SYNERR, "missing \";\" in latency definition\n");
1856 return;
1857 }
1859 pipe_class->setInstructionCount(1);
1860 pipe_class->setBranchDelay(true);
1861 next_char(); skipws();
1862 continue;
1863 }
1865 if (!strcmp(ident, "one_instruction") ||
1866 !strcmp(ident, "single_instruction")) {
1867 skipws();
1868 if (_curchar != ';') {
1869 parse_err(SYNERR, "missing \";\" in latency definition\n");
1870 return;
1871 }
1873 pipe_class->setInstructionCount(1);
1874 next_char(); skipws();
1875 continue;
1876 }
1878 if (!strcmp(ident, "instructions_in_first_bundle") ||
1879 !strcmp(ident, "instruction_count")) {
1880 skipws();
1882 int number_of_instructions = 1;
1884 if (_curchar != '(') {
1885 parse_err(SYNERR, "\"(\" expected at \"%c\"\n", _curchar);
1886 continue;
1887 }
1889 next_char(); skipws();
1890 number_of_instructions = get_int();
1892 skipws();
1893 if (_curchar != ')') {
1894 parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
1895 continue;
1896 }
1898 next_char(); skipws();
1899 if (_curchar != ';') {
1900 parse_err(SYNERR, "missing \";\" in latency definition\n");
1901 return;
1902 }
1904 pipe_class->setInstructionCount(number_of_instructions);
1905 next_char(); skipws();
1906 continue;
1907 }
1909 if (!strcmp(ident, "multiple_bundles")) {
1910 skipws();
1911 if (_curchar != ';') {
1912 parse_err(SYNERR, "missing \";\" after multiple bundles\n");
1913 return;
1914 }
1916 pipe_class->setMultipleBundles(true);
1917 next_char(); skipws();
1918 continue;
1919 }
1921 if (!strcmp(ident, "has_delay_slot")) {
1922 skipws();
1923 if (_curchar != ';') {
1924 parse_err(SYNERR, "missing \";\" after \"has_delay_slot\"\n");
1925 return;
1926 }
1928 pipe_class->setBranchDelay(true);
1929 next_char(); skipws();
1930 continue;
1931 }
1933 if (!strcmp(ident, "force_serialization")) {
1934 skipws();
1935 if (_curchar != ';') {
1936 parse_err(SYNERR, "missing \";\" after \"force_serialization\"\n");
1937 return;
1938 }
1940 pipe_class->setForceSerialization(true);
1941 next_char(); skipws();
1942 continue;
1943 }
1945 if (!strcmp(ident, "may_have_no_code")) {
1946 skipws();
1947 if (_curchar != ';') {
1948 parse_err(SYNERR, "missing \";\" after \"may_have_no_code\"\n");
1949 return;
1950 }
1952 pipe_class->setMayHaveNoCode(true);
1953 next_char(); skipws();
1954 continue;
1955 }
1957 const Form *parm = pipe_class->_localNames[ident];
1958 if (parm != NULL) {
1959 oper = parm->is_operand();
1960 if (oper == NULL && !parm->is_opclass()) {
1961 parse_err(SYNERR, "operand name expected at %s\n", ident);
1962 continue;
1963 }
1965 if (_curchar != ':') {
1966 parse_err(SYNERR, "\":\" expected at \"%c\"\n", _curchar);
1967 continue;
1968 }
1969 next_char(); skipws();
1970 stage = get_ident();
1971 if (stage == NULL) {
1972 parse_err(SYNERR, "pipeline stage identifier expected at \"%c\"\n", _curchar);
1973 continue;
1974 }
1976 skipws();
1977 if (_curchar != '(') {
1978 parse_err(SYNERR, "\"(\" expected at \"%c\"\n", _curchar);
1979 continue;
1980 }
1982 next_char();
1983 read_or_write = get_ident();
1984 if (read_or_write == NULL) {
1985 parse_err(SYNERR, "\"read\" or \"write\" expected at \"%c\"\n", _curchar);
1986 continue;
1987 }
1989 is_read = strcmp(read_or_write, "read") == 0;
1990 is_write = strcmp(read_or_write, "write") == 0;
1991 if (!is_read && !is_write) {
1992 parse_err(SYNERR, "\"read\" or \"write\" expected at \"%c\"\n", _curchar);
1993 continue;
1994 }
1996 skipws();
1997 if (_curchar != ')') {
1998 parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
1999 continue;
2000 }
2002 next_char(); skipws();
2003 int more_instrs = 0;
2004 if (_curchar == '+') {
2005 next_char(); skipws();
2006 if (_curchar < '0' || _curchar > '9') {
2007 parse_err(SYNERR, "<number> expected at \"%c\"\n", _curchar);
2008 continue;
2009 }
2010 while (_curchar >= '0' && _curchar <= '9') {
2011 more_instrs *= 10;
2012 more_instrs += _curchar - '0';
2013 next_char();
2014 }
2015 skipws();
2016 }
2018 PipeClassOperandForm *pipe_operand = new PipeClassOperandForm(stage, is_write, more_instrs);
2019 pipe_class->_localUsage.Insert(ident, pipe_operand);
2021 if (_curchar == '%')
2022 continue;
2024 if (_curchar != ';') {
2025 parse_err(SYNERR, "\";\" expected at \"%c\"\n", _curchar);
2026 continue;
2027 }
2028 next_char(); skipws();
2029 continue;
2030 }
2032 // Scan for Resource Specifier
2033 const Form *res = pipeline._resdict[ident];
2034 if (res != NULL) {
2035 int cyclecnt = 1;
2036 if (_curchar != ':') {
2037 parse_err(SYNERR, "\":\" expected at \"%c\"\n", _curchar);
2038 continue;
2039 }
2040 next_char(); skipws();
2041 stage = get_ident();
2042 if (stage == NULL) {
2043 parse_err(SYNERR, "pipeline stage identifier expected at \"%c\"\n", _curchar);
2044 continue;
2045 }
2047 skipws();
2048 if (_curchar == '(') {
2049 next_char();
2050 cyclecnt = get_int();
2052 skipws();
2053 if (_curchar != ')') {
2054 parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
2055 continue;
2056 }
2058 next_char(); skipws();
2059 }
2061 PipeClassResourceForm *resource = new PipeClassResourceForm(ident, stage, cyclecnt);
2062 int stagenum = pipeline._stages.index(stage);
2063 if (pipeline._maxcycleused < (stagenum+cyclecnt))
2064 pipeline._maxcycleused = (stagenum+cyclecnt);
2065 pipe_class->_resUsage.addForm(resource);
2067 if (_curchar == '%')
2068 continue;
2070 if (_curchar != ';') {
2071 parse_err(SYNERR, "\";\" expected at \"%c\"\n", _curchar);
2072 continue;
2073 }
2074 next_char(); skipws();
2075 continue;
2076 }
2078 parse_err(SYNERR, "resource expected at \"%s\"\n", ident);
2079 return;
2080 } while(_curchar != '%');
2082 next_char();
2083 if (_curchar != '}') {
2084 parse_err(SYNERR, "missing \"%}\" in pipe_class definition\n");
2085 return;
2086 }
2088 next_char();
2089 }
2091 //------------------------------peep_parse-------------------------------------
2092 void ADLParser::peep_parse(void) {
2093 Peephole *peep; // Pointer to current peephole rule form
2094 char *desc = NULL; // String representation of rule
2096 skipws(); // Skip leading whitespace
2098 peep = new Peephole(); // Build new Peephole object
2099 // Check for open block sequence
2100 skipws(); // Skip leading whitespace
2101 if (_curchar == '%' && *(_ptr+1) == '{') {
2102 next_char(); next_char(); // Skip "%{"
2103 skipws();
2104 while (_curchar != '%' && *(_ptr+1) != '}') {
2105 char *token = get_ident();
2106 if (token == NULL) {
2107 parse_err(SYNERR, "missing identifier inside peephole rule.\n");
2108 return;
2109 }
2110 // check for legal subsections of peephole rule
2111 if (strcmp(token,"peepmatch")==0) {
2112 peep_match_parse(*peep); }
2113 else if (strcmp(token,"peepconstraint")==0) {
2114 peep_constraint_parse(*peep); }
2115 else if (strcmp(token,"peepreplace")==0) {
2116 peep_replace_parse(*peep); }
2117 else {
2118 parse_err(SYNERR, "expected peepmatch, peepconstraint, or peepreplace for identifier %s.\n", token);
2119 }
2120 skipws();
2121 }
2122 }
2123 else {
2124 parse_err(SYNERR, "Missing %%{ ... %%} block after peephole keyword.\n");
2125 return;
2126 }
2127 next_char(); // Skip past '%'
2128 next_char(); // Skip past '}'
2129 }
2131 // ******************** Private Level 2 Parse Functions ********************
2132 //------------------------------constraint_parse------------------------------
2133 Constraint *ADLParser::constraint_parse(void) {
2134 char *func;
2135 char *arg;
2137 // Check for constraint expression
2138 skipws();
2139 if (_curchar != '(') {
2140 parse_err(SYNERR, "missing constraint expression, (...)\n");
2141 return NULL;
2142 }
2143 next_char(); // Skip past '('
2145 // Get constraint function
2146 skipws();
2147 func = get_ident();
2148 if (func == NULL) {
2149 parse_err(SYNERR, "missing function in constraint expression.\n");
2150 return NULL;
2151 }
2152 if (strcmp(func,"ALLOC_IN_RC")==0
2153 || strcmp(func,"IS_R_CLASS")==0) {
2154 // Check for '(' before argument
2155 skipws();
2156 if (_curchar != '(') {
2157 parse_err(SYNERR, "missing '(' for constraint function's argument.\n");
2158 return NULL;
2159 }
2160 next_char();
2162 // Get it's argument
2163 skipws();
2164 arg = get_ident();
2165 if (arg == NULL) {
2166 parse_err(SYNERR, "missing argument for constraint function %s\n",func);
2167 return NULL;
2168 }
2169 // Check for ')' after argument
2170 skipws();
2171 if (_curchar != ')') {
2172 parse_err(SYNERR, "missing ')' after constraint function argument %s\n",arg);
2173 return NULL;
2174 }
2175 next_char();
2176 } else {
2177 parse_err(SYNERR, "Invalid constraint function %s\n",func);
2178 return NULL;
2179 }
2181 // Check for closing paren and ';'
2182 skipws();
2183 if (_curchar != ')') {
2184 parse_err(SYNERR, "Missing ')' for constraint function %s\n",func);
2185 return NULL;
2186 }
2187 next_char();
2188 skipws();
2189 if (_curchar != ';') {
2190 parse_err(SYNERR, "Missing ';' after constraint.\n");
2191 return NULL;
2192 }
2193 next_char();
2195 // Create new "Constraint"
2196 Constraint *constraint = new Constraint(func,arg);
2197 return constraint;
2198 }
2200 //------------------------------constr_parse-----------------------------------
2201 ConstructRule *ADLParser::construct_parse(void) {
2202 return NULL;
2203 }
2206 //------------------------------reg_def_parse----------------------------------
2207 void ADLParser::reg_def_parse(void) {
2208 char *rname; // Name of register being defined
2210 // Get register name
2211 skipws(); // Skip whitespace
2212 rname = get_ident();
2213 if (rname == NULL) {
2214 parse_err(SYNERR, "missing register name after reg_def\n");
2215 return;
2216 }
2218 // Check for definition of register calling convention (save on call, ...),
2219 // register save type, and register encoding value.
2220 skipws();
2221 char *callconv = NULL;
2222 char *c_conv = NULL;
2223 char *idealtype = NULL;
2224 char *encoding = NULL;
2225 char *concrete = NULL;
2226 if (_curchar == '(') {
2227 next_char();
2228 callconv = get_ident();
2229 // Parse the internal calling convention, must be NS, SOC, SOE, or AS.
2230 if (callconv == NULL) {
2231 parse_err(SYNERR, "missing register calling convention value\n");
2232 return;
2233 }
2234 if(strcmp(callconv, "SOC") && strcmp(callconv,"SOE") &&
2235 strcmp(callconv, "NS") && strcmp(callconv, "AS")) {
2236 parse_err(SYNERR, "invalid value for register calling convention\n");
2237 }
2238 skipws();
2239 if (_curchar != ',') {
2240 parse_err(SYNERR, "missing comma in register definition statement\n");
2241 return;
2242 }
2243 next_char();
2245 // Parse the native calling convention, must be NS, SOC, SOE, AS
2246 c_conv = get_ident();
2247 if (c_conv == NULL) {
2248 parse_err(SYNERR, "missing register native calling convention value\n");
2249 return;
2250 }
2251 if(strcmp(c_conv, "SOC") && strcmp(c_conv,"SOE") &&
2252 strcmp(c_conv, "NS") && strcmp(c_conv, "AS")) {
2253 parse_err(SYNERR, "invalid value for register calling convention\n");
2254 }
2255 skipws();
2256 if (_curchar != ',') {
2257 parse_err(SYNERR, "missing comma in register definition statement\n");
2258 return;
2259 }
2260 next_char();
2261 skipws();
2263 // Parse the ideal save type
2264 idealtype = get_ident();
2265 if (idealtype == NULL) {
2266 parse_err(SYNERR, "missing register save type value\n");
2267 return;
2268 }
2269 skipws();
2270 if (_curchar != ',') {
2271 parse_err(SYNERR, "missing comma in register definition statement\n");
2272 return;
2273 }
2274 next_char();
2275 skipws();
2277 // Parse the encoding value
2278 encoding = get_expr("encoding", ",");
2279 if (encoding == NULL) {
2280 parse_err(SYNERR, "missing register encoding value\n");
2281 return;
2282 }
2283 trim(encoding);
2284 if (_curchar != ',') {
2285 parse_err(SYNERR, "missing comma in register definition statement\n");
2286 return;
2287 }
2288 next_char();
2289 skipws();
2290 // Parse the concrete name type
2291 // concrete = get_ident();
2292 concrete = get_expr("concrete", ")");
2293 if (concrete == NULL) {
2294 parse_err(SYNERR, "missing vm register name value\n");
2295 return;
2296 }
2298 if (_curchar != ')') {
2299 parse_err(SYNERR, "missing ')' in register definition statement\n");
2300 return;
2301 }
2302 next_char();
2303 }
2305 // Check for closing ';'
2306 skipws();
2307 if (_curchar != ';') {
2308 parse_err(SYNERR, "missing ';' after reg_def\n");
2309 return;
2310 }
2311 next_char(); // move past ';'
2313 // Debug Stuff
2314 if (_AD._adl_debug > 1) {
2315 fprintf(stderr,"Register Definition: %s ( %s, %s %s )\n", rname,
2316 (callconv ? callconv : ""), (c_conv ? c_conv : ""), concrete);
2317 }
2319 // Record new register definition.
2320 _AD._register->addRegDef(rname, callconv, c_conv, idealtype, encoding, concrete);
2321 return;
2322 }
2324 //------------------------------reg_class_parse--------------------------------
2325 void ADLParser::reg_class_parse(void) {
2326 char *cname; // Name of register class being defined
2328 // Get register class name
2329 skipws(); // Skip leading whitespace
2330 cname = get_ident();
2331 if (cname == NULL) {
2332 parse_err(SYNERR, "missing register class name after 'reg_class'\n");
2333 return;
2334 }
2335 // Debug Stuff
2336 if (_AD._adl_debug >1) fprintf(stderr,"Register Class: %s\n", cname);
2338 RegClass *reg_class = _AD._register->addRegClass(cname);
2340 // Collect registers in class
2341 skipws();
2342 if (_curchar == '(') {
2343 next_char(); // Skip '('
2344 skipws();
2345 while (_curchar != ')') {
2346 char *rname = get_ident();
2347 if (rname==NULL) {
2348 parse_err(SYNERR, "missing identifier inside reg_class list.\n");
2349 return;
2350 }
2351 RegDef *regDef = _AD._register->getRegDef(rname);
2352 if (!regDef) {
2353 parse_err(SEMERR, "unknown identifier %s inside reg_class list.\n", rname);
2354 } else {
2355 reg_class->addReg(regDef); // add regDef to regClass
2356 }
2358 // Check for ',' and position to next token.
2359 skipws();
2360 if (_curchar == ',') {
2361 next_char(); // Skip trailing ','
2362 skipws();
2363 }
2364 }
2365 next_char(); // Skip closing ')'
2366 }
2368 // Check for terminating ';'
2369 skipws();
2370 if (_curchar != ';') {
2371 parse_err(SYNERR, "missing ';' at end of reg_class definition.\n");
2372 return;
2373 }
2374 next_char(); // Skip trailing ';'
2376 // Check RegClass size, must be <= 32 registers in class.
2378 return;
2379 }
2381 //------------------------------alloc_class_parse------------------------------
2382 void ADLParser::alloc_class_parse(void) {
2383 char *name; // Name of allocation class being defined
2385 // Get allocation class name
2386 skipws(); // Skip leading whitespace
2387 name = get_ident();
2388 if (name == NULL) {
2389 parse_err(SYNERR, "missing allocation class name after 'reg_class'\n");
2390 return;
2391 }
2392 // Debug Stuff
2393 if (_AD._adl_debug >1) fprintf(stderr,"Allocation Class: %s\n", name);
2395 AllocClass *alloc_class = _AD._register->addAllocClass(name);
2397 // Collect registers in class
2398 skipws();
2399 if (_curchar == '(') {
2400 next_char(); // Skip '('
2401 skipws();
2402 while (_curchar != ')') {
2403 char *rname = get_ident();
2404 if (rname==NULL) {
2405 parse_err(SYNERR, "missing identifier inside reg_class list.\n");
2406 return;
2407 }
2408 // Check if name is a RegDef
2409 RegDef *regDef = _AD._register->getRegDef(rname);
2410 if (regDef) {
2411 alloc_class->addReg(regDef); // add regDef to allocClass
2412 } else {
2414 // name must be a RegDef or a RegClass
2415 parse_err(SYNERR, "name %s should be a previously defined reg_def.\n", rname);
2416 return;
2417 }
2419 // Check for ',' and position to next token.
2420 skipws();
2421 if (_curchar == ',') {
2422 next_char(); // Skip trailing ','
2423 skipws();
2424 }
2425 }
2426 next_char(); // Skip closing ')'
2427 }
2429 // Check for terminating ';'
2430 skipws();
2431 if (_curchar != ';') {
2432 parse_err(SYNERR, "missing ';' at end of reg_class definition.\n");
2433 return;
2434 }
2435 next_char(); // Skip trailing ';'
2437 return;
2438 }
2440 //------------------------------peep_match_child_parse-------------------------
2441 InstructForm *ADLParser::peep_match_child_parse(PeepMatch &match, int parent, int &position, int input){
2442 char *token = NULL;
2443 int lparen = 0; // keep track of parenthesis nesting depth
2444 int rparen = 0; // position of instruction at this depth
2445 InstructForm *inst_seen = NULL;
2446 InstructForm *child_seen = NULL;
2448 // Walk the match tree,
2449 // Record <parent, position, instruction name, input position>
2450 while ( lparen >= rparen ) {
2451 skipws();
2452 // Left paren signals start of an input, collect with recursive call
2453 if (_curchar == '(') {
2454 ++lparen;
2455 next_char();
2456 child_seen = peep_match_child_parse(match, parent, position, rparen);
2457 }
2458 // Right paren signals end of an input, may be more
2459 else if (_curchar == ')') {
2460 ++rparen;
2461 if( rparen == lparen ) { // IF rparen matches an lparen I've seen
2462 next_char(); // move past ')'
2463 } else { // ELSE leave ')' for parent
2464 assert( rparen == lparen + 1, "Should only see one extra ')'");
2465 // if an instruction was not specified for this paren-pair
2466 if( ! inst_seen ) { // record signal entry
2467 match.add_instruction( parent, position, NameList::_signal, input );
2468 ++position;
2469 }
2470 // ++input; // TEMPORARY
2471 return inst_seen;
2472 }
2473 }
2474 // if no parens, then check for instruction name
2475 // This instruction is the parent of a sub-tree
2476 else if ((token = get_ident_dup()) != NULL) {
2477 const Form *form = _AD._globalNames[token];
2478 if (form) {
2479 InstructForm *inst = form->is_instruction();
2480 // Record the first instruction at this level
2481 if( inst_seen == NULL ) {
2482 inst_seen = inst;
2483 }
2484 if (inst) {
2485 match.add_instruction( parent, position, token, input );
2486 parent = position;
2487 ++position;
2488 } else {
2489 parse_err(SYNERR, "instruction name expected at identifier %s.\n",
2490 token);
2491 return inst_seen;
2492 }
2493 }
2494 else {
2495 parse_err(SYNERR, "missing identifier in peepmatch rule.\n");
2496 return NULL;
2497 }
2498 }
2499 else {
2500 parse_err(SYNERR, "missing identifier in peepmatch rule.\n");
2501 return NULL;
2502 }
2504 } // end while
2506 assert( false, "ShouldNotReachHere();");
2507 return NULL;
2508 }
2510 //------------------------------peep_match_parse-------------------------------
2511 // Syntax for a peepmatch rule
2512 //
2513 // peepmatch ( root_instr_name [(instruction subtree)] [,(instruction subtree)]* );
2514 //
2515 void ADLParser::peep_match_parse(Peephole &peep) {
2517 skipws();
2518 // Check the structure of the rule
2519 // Check for open paren
2520 if (_curchar != '(') {
2521 parse_err(SYNERR, "missing '(' at start of peepmatch rule.\n");
2522 return;
2523 }
2524 next_char(); // skip '('
2526 // Construct PeepMatch and parse the peepmatch rule.
2527 PeepMatch *match = new PeepMatch(_ptr);
2528 int parent = -1; // parent of root
2529 int position = 0; // zero-based positions
2530 int input = 0; // input position in parent's operands
2531 InstructForm *root= peep_match_child_parse( *match, parent, position, input);
2532 if( root == NULL ) {
2533 parse_err(SYNERR, "missing instruction-name at start of peepmatch.\n");
2534 return;
2535 }
2537 if( _curchar != ')' ) {
2538 parse_err(SYNERR, "missing ')' at end of peepmatch.\n");
2539 return;
2540 }
2541 next_char(); // skip ')'
2543 // Check for closing semicolon
2544 skipws();
2545 if( _curchar != ';' ) {
2546 parse_err(SYNERR, "missing ';' at end of peepmatch.\n");
2547 return;
2548 }
2549 next_char(); // skip ';'
2551 // Store match into peep, and store peep into instruction
2552 peep.add_match(match);
2553 root->append_peephole(&peep);
2554 }
2556 //------------------------------peep_constraint_parse--------------------------
2557 // Syntax for a peepconstraint rule
2558 // A parenthesized list of relations between operands in peepmatch subtree
2559 //
2560 // peepconstraint %{
2561 // (instruction_number.operand_name
2562 // relational_op
2563 // instruction_number.operand_name OR register_name
2564 // [, ...] );
2565 //
2566 // // instruction numbers are zero-based using topological order in peepmatch
2567 //
2568 void ADLParser::peep_constraint_parse(Peephole &peep) {
2570 skipws();
2571 // Check the structure of the rule
2572 // Check for open paren
2573 if (_curchar != '(') {
2574 parse_err(SYNERR, "missing '(' at start of peepconstraint rule.\n");
2575 return;
2576 }
2577 else {
2578 next_char(); // Skip '('
2579 }
2581 // Check for a constraint
2582 skipws();
2583 while( _curchar != ')' ) {
2584 // Get information on the left instruction and its operand
2585 // left-instructions's number
2586 int left_inst = get_int();
2587 // Left-instruction's operand
2588 skipws();
2589 if( _curchar != '.' ) {
2590 parse_err(SYNERR, "missing '.' in peepconstraint after instruction number.\n");
2591 return;
2592 }
2593 next_char(); // Skip '.'
2594 char *left_op = get_ident_dup();
2596 skipws();
2597 // Collect relational operator
2598 char *relation = get_relation_dup();
2600 skipws();
2601 // Get information on the right instruction and its operand
2602 int right_inst; // Right-instructions's number
2603 if( isdigit(_curchar) ) {
2604 right_inst = get_int();
2605 // Right-instruction's operand
2606 skipws();
2607 if( _curchar != '.' ) {
2608 parse_err(SYNERR, "missing '.' in peepconstraint after instruction number.\n");
2609 return;
2610 }
2611 next_char(); // Skip '.'
2612 } else {
2613 right_inst = -1; // Flag as being a register constraint
2614 }
2616 char *right_op = get_ident_dup();
2618 // Construct the next PeepConstraint
2619 PeepConstraint *constraint = new PeepConstraint( left_inst, left_op,
2620 relation,
2621 right_inst, right_op );
2622 // And append it to the list for this peephole rule
2623 peep.append_constraint( constraint );
2625 // Check for another constraint, or end of rule
2626 skipws();
2627 if( _curchar == ',' ) {
2628 next_char(); // Skip ','
2629 skipws();
2630 }
2631 else if( _curchar != ')' ) {
2632 parse_err(SYNERR, "expected ',' or ')' after peephole constraint.\n");
2633 return;
2634 }
2635 } // end while( processing constraints )
2636 next_char(); // Skip ')'
2638 // Check for terminating ';'
2639 skipws();
2640 if (_curchar != ';') {
2641 parse_err(SYNERR, "missing ';' at end of peepconstraint.\n");
2642 return;
2643 }
2644 next_char(); // Skip trailing ';'
2645 }
2648 //------------------------------peep_replace_parse-----------------------------
2649 // Syntax for a peepreplace rule
2650 // root instruction name followed by a
2651 // parenthesized list of whitespace separated instruction.operand specifiers
2652 //
2653 // peepreplace ( instr_name ( [instruction_number.operand_name]* ) );
2654 //
2655 //
2656 void ADLParser::peep_replace_parse(Peephole &peep) {
2657 int lparen = 0; // keep track of parenthesis nesting depth
2658 int rparen = 0; // keep track of parenthesis nesting depth
2659 int icount = 0; // count of instructions in rule for naming
2660 char *str = NULL;
2661 char *token = NULL;
2663 skipws();
2664 // Check for open paren
2665 if (_curchar != '(') {
2666 parse_err(SYNERR, "missing '(' at start of peepreplace rule.\n");
2667 return;
2668 }
2669 else {
2670 lparen++;
2671 next_char();
2672 }
2674 // Check for root instruction
2675 char *inst = get_ident_dup();
2676 const Form *form = _AD._globalNames[inst];
2677 if( form == NULL || form->is_instruction() == NULL ) {
2678 parse_err(SYNERR, "Instruction name expected at start of peepreplace.\n");
2679 return;
2680 }
2682 // Store string representation of rule into replace
2683 PeepReplace *replace = new PeepReplace(str);
2684 replace->add_instruction( inst );
2686 skipws();
2687 // Start of root's operand-list
2688 if (_curchar != '(') {
2689 parse_err(SYNERR, "missing '(' at peepreplace root's operand-list.\n");
2690 return;
2691 }
2692 else {
2693 lparen++;
2694 next_char();
2695 }
2697 skipws();
2698 // Get the list of operands
2699 while( _curchar != ')' ) {
2700 // Get information on an instruction and its operand
2701 // instructions's number
2702 int inst_num = get_int();
2703 // Left-instruction's operand
2704 skipws();
2705 if( _curchar != '.' ) {
2706 parse_err(SYNERR, "missing '.' in peepreplace after instruction number.\n");
2707 return;
2708 }
2709 next_char(); // Skip '.'
2710 char *inst_op = get_ident_dup();
2711 if( inst_op == NULL ) {
2712 parse_err(SYNERR, "missing operand identifier in peepreplace.\n");
2713 return;
2714 }
2716 // Record this operand's position in peepmatch
2717 replace->add_operand( inst_num, inst_op );
2718 skipws();
2719 }
2721 // Check for the end of operands list
2722 skipws();
2723 assert( _curchar == ')', "While loop should have advanced to ')'.");
2724 next_char(); // Skip ')'
2726 skipws();
2727 // Check for end of peepreplace
2728 if( _curchar != ')' ) {
2729 parse_err(SYNERR, "missing ')' at end of peepmatch.\n");
2730 parse_err(SYNERR, "Support one replacement instruction.\n");
2731 return;
2732 }
2733 next_char(); // Skip ')'
2735 // Check for closing semicolon
2736 skipws();
2737 if( _curchar != ';' ) {
2738 parse_err(SYNERR, "missing ';' at end of peepreplace.\n");
2739 return;
2740 }
2741 next_char(); // skip ';'
2743 // Store replace into peep
2744 peep.add_replace( replace );
2745 }
2747 //------------------------------pred_parse-------------------------------------
2748 Predicate *ADLParser::pred_parse(void) {
2749 Predicate *predicate; // Predicate class for operand
2750 char *rule = NULL; // String representation of predicate
2752 skipws(); // Skip leading whitespace
2753 int line = linenum();
2754 if ( (rule = get_paren_expr("pred expression", true)) == NULL ) {
2755 parse_err(SYNERR, "incorrect or missing expression for 'predicate'\n");
2756 return NULL;
2757 }
2758 // Debug Stuff
2759 if (_AD._adl_debug > 1) fprintf(stderr,"Predicate: %s\n", rule);
2760 if (_curchar != ';') {
2761 parse_err(SYNERR, "missing ';' in predicate definition\n");
2762 return NULL;
2763 }
2764 next_char(); // Point after the terminator
2766 predicate = new Predicate(rule); // Build new predicate object
2767 skipws();
2768 return predicate;
2769 }
2772 //------------------------------ins_encode_parse_block-------------------------
2773 // Parse the block form of ins_encode. See ins_encode_parse for more details
2774 void ADLParser::ins_encode_parse_block(InstructForm& inst) {
2775 // Create a new encoding name based on the name of the instruction
2776 // definition, which should be unique.
2777 const char* prefix = "__ins_encode_";
2778 char* ec_name = (char*) malloc(strlen(inst._ident) + strlen(prefix) + 1);
2779 sprintf(ec_name, "%s%s", prefix, inst._ident);
2781 assert(_AD._encode->encClass(ec_name) == NULL, "shouldn't already exist");
2782 EncClass* encoding = _AD._encode->add_EncClass(ec_name);
2783 encoding->_linenum = linenum();
2785 // synthesize the arguments list for the enc_class from the
2786 // arguments to the instruct definition.
2787 const char* param = NULL;
2788 inst._parameters.reset();
2789 while ((param = inst._parameters.iter()) != NULL) {
2790 OperandForm* opForm = (OperandForm*) inst._localNames[param];
2791 encoding->add_parameter(opForm->_ident, param);
2792 }
2794 // Define a MacroAssembler instance for use by the encoding. The
2795 // name is chosen to match the __ idiom used for assembly in other
2796 // parts of hotspot and assumes the existence of the standard
2797 // #define __ _masm.
2798 encoding->add_code(" MacroAssembler _masm(&cbuf);\n");
2800 // Parse the following %{ }% block
2801 ins_encode_parse_block_impl(inst, encoding, ec_name);
2803 // Build an encoding rule which invokes the encoding rule we just
2804 // created, passing all arguments that we received.
2805 InsEncode* encrule = new InsEncode(); // Encode class for instruction
2806 NameAndList* params = encrule->add_encode(ec_name);
2807 inst._parameters.reset();
2808 while ((param = inst._parameters.iter()) != NULL) {
2809 params->add_entry(param);
2810 }
2812 // Check for duplicate ins_encode sections after parsing the block
2813 // so that parsing can continue and find any other errors.
2814 if (inst._insencode != NULL) {
2815 parse_err(SYNERR, "Multiple ins_encode sections defined\n");
2816 return;
2817 }
2819 // Set encode class of this instruction.
2820 inst._insencode = encrule;
2821 }
2824 void ADLParser::ins_encode_parse_block_impl(InstructForm& inst, EncClass* encoding, char* ec_name) {
2825 skipws_no_preproc(); // Skip leading whitespace
2826 // Prepend location descriptor, for debugging; cf. ADLParser::find_cpp_block
2827 if (_AD._adlocation_debug) {
2828 encoding->add_code(get_line_string());
2829 }
2831 // Collect the parts of the encode description
2832 // (1) strings that are passed through to output
2833 // (2) replacement/substitution variable, preceeded by a '$'
2834 while ((_curchar != '%') && (*(_ptr+1) != '}')) {
2836 // (1)
2837 // Check if there is a string to pass through to output
2838 char *start = _ptr; // Record start of the next string
2839 while ((_curchar != '$') && ((_curchar != '%') || (*(_ptr+1) != '}')) ) {
2840 // If at the start of a comment, skip past it
2841 if( (_curchar == '/') && ((*(_ptr+1) == '/') || (*(_ptr+1) == '*')) ) {
2842 skipws_no_preproc();
2843 } else {
2844 // ELSE advance to the next character, or start of the next line
2845 next_char_or_line();
2846 }
2847 }
2848 // If a string was found, terminate it and record in EncClass
2849 if (start != _ptr) {
2850 *_ptr = '\0'; // Terminate the string
2851 encoding->add_code(start);
2852 }
2854 // (2)
2855 // If we are at a replacement variable,
2856 // copy it and record in EncClass
2857 if (_curchar == '$') {
2858 // Found replacement Variable
2859 char* rep_var = get_rep_var_ident_dup();
2861 // Add flag to _strings list indicating we should check _rep_vars
2862 encoding->add_rep_var(rep_var);
2864 skipws();
2866 // Check if this instruct is a MachConstantNode.
2867 if (strcmp(rep_var, "constanttablebase") == 0) {
2868 // This instruct is a MachConstantNode.
2869 inst.set_is_mach_constant(true);
2871 if (_curchar == '(') {
2872 parse_err(SYNERR, "constanttablebase in instruct %s cannot have an argument (only constantaddress and constantoffset)", ec_name);
2873 return;
2874 }
2875 }
2876 else if ((strcmp(rep_var, "constantaddress") == 0) ||
2877 (strcmp(rep_var, "constantoffset") == 0)) {
2878 // This instruct is a MachConstantNode.
2879 inst.set_is_mach_constant(true);
2881 // If the constant keyword has an argument, parse it.
2882 if (_curchar == '(') constant_parse(inst);
2883 }
2884 }
2885 } // end while part of format description
2886 next_char(); // Skip '%'
2887 next_char(); // Skip '}'
2889 skipws();
2891 if (_AD._adlocation_debug) {
2892 encoding->add_code(end_line_marker());
2893 }
2895 // Debug Stuff
2896 if (_AD._adl_debug > 1) fprintf(stderr, "EncodingClass Form: %s\n", ec_name);
2897 }
2900 //------------------------------ins_encode_parse-------------------------------
2901 // Encode rules have the form
2902 // ins_encode( encode_class_name(parameter_list), ... );
2903 //
2904 // The "encode_class_name" must be defined in the encode section
2905 // The parameter list contains $names that are locals.
2906 //
2907 // Alternatively it can be written like this:
2908 //
2909 // ins_encode %{
2910 // ... // body
2911 // %}
2912 //
2913 // which synthesizes a new encoding class taking the same arguments as
2914 // the InstructForm, and automatically prefixes the definition with:
2915 //
2916 // MacroAssembler masm(&cbuf);\n");
2917 //
2918 // making it more compact to take advantage of the MacroAssembler and
2919 // placing the assembly closer to it's use by instructions.
2920 void ADLParser::ins_encode_parse(InstructForm& inst) {
2922 // Parse encode class name
2923 skipws(); // Skip whitespace
2924 if (_curchar != '(') {
2925 // Check for ins_encode %{ form
2926 if ((_curchar == '%') && (*(_ptr+1) == '{')) {
2927 next_char(); // Skip '%'
2928 next_char(); // Skip '{'
2930 // Parse the block form of ins_encode
2931 ins_encode_parse_block(inst);
2932 return;
2933 }
2935 parse_err(SYNERR, "missing '%%{' or '(' in ins_encode definition\n");
2936 return;
2937 }
2938 next_char(); // move past '('
2939 skipws();
2941 InsEncode *encrule = new InsEncode(); // Encode class for instruction
2942 encrule->_linenum = linenum();
2943 char *ec_name = NULL; // String representation of encode rule
2944 // identifier is optional.
2945 while (_curchar != ')') {
2946 ec_name = get_ident();
2947 if (ec_name == NULL) {
2948 parse_err(SYNERR, "Invalid encode class name after 'ins_encode('.\n");
2949 return;
2950 }
2951 // Check that encoding is defined in the encode section
2952 EncClass *encode_class = _AD._encode->encClass(ec_name);
2953 if (encode_class == NULL) {
2954 // Like to defer checking these till later...
2955 // parse_err(WARN, "Using an undefined encode class '%s' in 'ins_encode'.\n", ec_name);
2956 }
2958 // Get list for encode method's parameters
2959 NameAndList *params = encrule->add_encode(ec_name);
2961 // Parse the parameters to this encode method.
2962 skipws();
2963 if ( _curchar == '(' ) {
2964 next_char(); // move past '(' for parameters
2966 // Parse the encode method's parameters
2967 while (_curchar != ')') {
2968 char *param = get_ident_or_literal_constant("encoding operand");
2969 if ( param != NULL ) {
2970 // Found a parameter:
2971 // Check it is a local name, add it to the list, then check for more
2972 // New: allow hex constants as parameters to an encode method.
2973 // New: allow parenthesized expressions as parameters.
2974 // New: allow "primary", "secondary", "tertiary" as parameters.
2975 // New: allow user-defined register name as parameter
2976 if ( (inst._localNames[param] == NULL) &&
2977 !ADLParser::is_literal_constant(param) &&
2978 (Opcode::as_opcode_type(param) == Opcode::NOT_AN_OPCODE) &&
2979 ((_AD._register == NULL ) || (_AD._register->getRegDef(param) == NULL)) ) {
2980 parse_err(SYNERR, "Using non-locally defined parameter %s for encoding %s.\n", param, ec_name);
2981 return;
2982 }
2983 params->add_entry(param);
2985 skipws();
2986 if (_curchar == ',' ) {
2987 // More parameters to come
2988 next_char(); // move past ',' between parameters
2989 skipws(); // Skip to next parameter
2990 }
2991 else if (_curchar == ')') {
2992 // Done with parameter list
2993 }
2994 else {
2995 // Only ',' or ')' are valid after a parameter name
2996 parse_err(SYNERR, "expected ',' or ')' after parameter %s.\n",
2997 ec_name);
2998 return;
2999 }
3001 } else {
3002 skipws();
3003 // Did not find a parameter
3004 if (_curchar == ',') {
3005 parse_err(SYNERR, "Expected encode parameter before ',' in encoding %s.\n", ec_name);
3006 return;
3007 }
3008 if (_curchar != ')') {
3009 parse_err(SYNERR, "Expected ')' after encode parameters.\n");
3010 return;
3011 }
3012 }
3013 } // WHILE loop collecting parameters
3014 next_char(); // move past ')' at end of parameters
3015 } // done with parameter list for encoding
3017 // Check for ',' or ')' after encoding
3018 skipws(); // move to character after parameters
3019 if ( _curchar == ',' ) {
3020 // Found a ','
3021 next_char(); // move past ',' between encode methods
3022 skipws();
3023 }
3024 else if ( _curchar != ')' ) {
3025 // If not a ',' then only a ')' is allowed
3026 parse_err(SYNERR, "Expected ')' after encoding %s.\n", ec_name);
3027 return;
3028 }
3030 // Check for ',' separating parameters
3031 // if ( _curchar != ',' && _curchar != ')' ) {
3032 // parse_err(SYNERR, "expected ',' or ')' after encode method inside ins_encode.\n");
3033 // return NULL;
3034 // }
3036 } // done parsing ins_encode methods and their parameters
3037 if (_curchar != ')') {
3038 parse_err(SYNERR, "Missing ')' at end of ins_encode description.\n");
3039 return;
3040 }
3041 next_char(); // move past ')'
3042 skipws(); // Skip leading whitespace
3044 if ( _curchar != ';' ) {
3045 parse_err(SYNERR, "Missing ';' at end of ins_encode.\n");
3046 return;
3047 }
3048 next_char(); // move past ';'
3049 skipws(); // be friendly to oper_parse()
3051 // Check for duplicate ins_encode sections after parsing the block
3052 // so that parsing can continue and find any other errors.
3053 if (inst._insencode != NULL) {
3054 parse_err(SYNERR, "Multiple ins_encode sections defined\n");
3055 return;
3056 }
3058 // Debug Stuff
3059 if (_AD._adl_debug > 1) fprintf(stderr,"Instruction Encode: %s\n", ec_name);
3061 // Set encode class of this instruction.
3062 inst._insencode = encrule;
3063 }
3066 //------------------------------constant_parse---------------------------------
3067 // Parse a constant expression.
3068 void ADLParser::constant_parse(InstructForm& inst) {
3069 // Create a new encoding name based on the name of the instruction
3070 // definition, which should be unique.
3071 const char* prefix = "__constant_";
3072 char* ec_name = (char*) malloc(strlen(inst._ident) + strlen(prefix) + 1);
3073 sprintf(ec_name, "%s%s", prefix, inst._ident);
3075 assert(_AD._encode->encClass(ec_name) == NULL, "shouldn't already exist");
3076 EncClass* encoding = _AD._encode->add_EncClass(ec_name);
3077 encoding->_linenum = linenum();
3079 // synthesize the arguments list for the enc_class from the
3080 // arguments to the instruct definition.
3081 const char* param = NULL;
3082 inst._parameters.reset();
3083 while ((param = inst._parameters.iter()) != NULL) {
3084 OperandForm* opForm = (OperandForm*) inst._localNames[param];
3085 encoding->add_parameter(opForm->_ident, param);
3086 }
3088 // Parse the following ( ) expression.
3089 constant_parse_expression(encoding, ec_name);
3091 // Build an encoding rule which invokes the encoding rule we just
3092 // created, passing all arguments that we received.
3093 InsEncode* encrule = new InsEncode(); // Encode class for instruction
3094 NameAndList* params = encrule->add_encode(ec_name);
3095 inst._parameters.reset();
3096 while ((param = inst._parameters.iter()) != NULL) {
3097 params->add_entry(param);
3098 }
3100 // Set encode class of this instruction.
3101 inst._constant = encrule;
3102 }
3105 //------------------------------constant_parse_expression----------------------
3106 void ADLParser::constant_parse_expression(EncClass* encoding, char* ec_name) {
3107 skipws();
3109 // Prepend location descriptor, for debugging; cf. ADLParser::find_cpp_block
3110 if (_AD._adlocation_debug) {
3111 encoding->add_code(get_line_string());
3112 }
3114 // Start code line.
3115 encoding->add_code(" _constant = C->constant_table().add");
3117 // Parse everything in ( ) expression.
3118 encoding->add_code("(this, ");
3119 next_char(); // Skip '('
3120 int parens_depth = 1;
3122 // Collect the parts of the constant expression.
3123 // (1) strings that are passed through to output
3124 // (2) replacement/substitution variable, preceeded by a '$'
3125 while (parens_depth > 0) {
3126 if (_curchar == '(') {
3127 parens_depth++;
3128 encoding->add_code("(");
3129 next_char();
3130 }
3131 else if (_curchar == ')') {
3132 parens_depth--;
3133 if (parens_depth > 0)
3134 encoding->add_code(")");
3135 next_char();
3136 }
3137 else {
3138 // (1)
3139 // Check if there is a string to pass through to output
3140 char *start = _ptr; // Record start of the next string
3141 while ((_curchar != '$') && (_curchar != '(') && (_curchar != ')')) {
3142 next_char();
3143 }
3144 // If a string was found, terminate it and record in EncClass
3145 if (start != _ptr) {
3146 *_ptr = '\0'; // Terminate the string
3147 encoding->add_code(start);
3148 }
3150 // (2)
3151 // If we are at a replacement variable, copy it and record in EncClass.
3152 if (_curchar == '$') {
3153 // Found replacement Variable
3154 char* rep_var = get_rep_var_ident_dup();
3155 encoding->add_rep_var(rep_var);
3156 }
3157 }
3158 }
3160 // Finish code line.
3161 encoding->add_code(");");
3163 if (_AD._adlocation_debug) {
3164 encoding->add_code(end_line_marker());
3165 }
3167 // Debug Stuff
3168 if (_AD._adl_debug > 1) fprintf(stderr, "EncodingClass Form: %s\n", ec_name);
3169 }
3172 //------------------------------size_parse-----------------------------------
3173 char* ADLParser::size_parse(InstructForm *instr) {
3174 char* sizeOfInstr = NULL;
3176 // Get value of the instruction's size
3177 skipws();
3179 // Parse size
3180 sizeOfInstr = get_paren_expr("size expression");
3181 if (sizeOfInstr == NULL) {
3182 parse_err(SYNERR, "size of opcode expected at %c\n", _curchar);
3183 return NULL;
3184 }
3186 skipws();
3188 // Check for terminator
3189 if (_curchar != ';') {
3190 parse_err(SYNERR, "missing ';' in ins_attrib definition\n");
3191 return NULL;
3192 }
3193 next_char(); // Advance past the ';'
3194 skipws(); // necessary for instr_parse()
3196 // Debug Stuff
3197 if (_AD._adl_debug > 1) {
3198 if (sizeOfInstr != NULL) {
3199 fprintf(stderr,"size of opcode: %s\n", sizeOfInstr);
3200 }
3201 }
3203 return sizeOfInstr;
3204 }
3207 //------------------------------opcode_parse-----------------------------------
3208 Opcode * ADLParser::opcode_parse(InstructForm *instr) {
3209 char *primary = NULL;
3210 char *secondary = NULL;
3211 char *tertiary = NULL;
3213 char *val = NULL;
3214 Opcode *opcode = NULL;
3216 // Get value of the instruction's opcode
3217 skipws();
3218 if (_curchar != '(') { // Check for parenthesized operand list
3219 parse_err(SYNERR, "missing '(' in expand instruction declaration\n");
3220 return NULL;
3221 }
3222 next_char(); // skip open paren
3223 skipws();
3224 if (_curchar != ')') {
3225 // Parse primary, secondary, and tertiary opcodes, if provided.
3226 if ( ((primary = get_ident_or_literal_constant("primary opcode")) == NULL) ) {
3227 parse_err(SYNERR, "primary hex opcode expected at %c\n", _curchar);
3228 return NULL;
3229 }
3230 skipws();
3231 if (_curchar == ',') {
3232 next_char();
3233 skipws();
3234 // Parse secondary opcode
3235 if ( ((secondary = get_ident_or_literal_constant("secondary opcode")) == NULL) ) {
3236 parse_err(SYNERR, "secondary hex opcode expected at %c\n", _curchar);
3237 return NULL;
3238 }
3239 skipws();
3240 if (_curchar == ',') {
3241 next_char();
3242 skipws();
3243 // Parse tertiary opcode
3244 if ( ((tertiary = get_ident_or_literal_constant("tertiary opcode")) == NULL) ) {
3245 parse_err(SYNERR,"tertiary hex opcode expected at %c\n", _curchar);
3246 return NULL;
3247 }
3248 skipws();
3249 }
3250 }
3251 skipws();
3252 if (_curchar != ')') {
3253 parse_err(SYNERR, "Missing ')' in opcode description\n");
3254 return NULL;
3255 }
3256 }
3257 next_char(); // Skip ')'
3258 skipws();
3259 // Check for terminator
3260 if (_curchar != ';') {
3261 parse_err(SYNERR, "missing ';' in ins_attrib definition\n");
3262 return NULL;
3263 }
3264 next_char(); // Advance past the ';'
3265 skipws(); // necessary for instr_parse()
3267 // Debug Stuff
3268 if (_AD._adl_debug > 1) {
3269 if (primary != NULL) fprintf(stderr,"primary opcode: %s\n", primary);
3270 if (secondary != NULL) fprintf(stderr,"secondary opcode: %s\n", secondary);
3271 if (tertiary != NULL) fprintf(stderr,"tertiary opcode: %s\n", tertiary);
3272 }
3274 // Generate new object and return
3275 opcode = new Opcode(primary, secondary, tertiary);
3276 return opcode;
3277 }
3280 //------------------------------interface_parse--------------------------------
3281 Interface *ADLParser::interface_parse(void) {
3282 char *iface_name = NULL; // Name of interface class being used
3283 char *iface_code = NULL; // Describe components of this class
3285 // Get interface class name
3286 skipws(); // Skip whitespace
3287 if (_curchar != '(') {
3288 parse_err(SYNERR, "Missing '(' at start of interface description.\n");
3289 return NULL;
3290 }
3291 next_char(); // move past '('
3292 skipws();
3293 iface_name = get_ident();
3294 if (iface_name == NULL) {
3295 parse_err(SYNERR, "missing interface name after 'interface'.\n");
3296 return NULL;
3297 }
3298 skipws();
3299 if (_curchar != ')') {
3300 parse_err(SYNERR, "Missing ')' after name of interface.\n");
3301 return NULL;
3302 }
3303 next_char(); // move past ')'
3305 // Get details of the interface,
3306 // for the type of interface indicated by iface_name.
3307 Interface *inter = NULL;
3308 skipws();
3309 if ( _curchar != ';' ) {
3310 if ( strcmp(iface_name,"MEMORY_INTER") == 0 ) {
3311 inter = mem_interface_parse();
3312 }
3313 else if ( strcmp(iface_name,"COND_INTER") == 0 ) {
3314 inter = cond_interface_parse();
3315 }
3316 // The parse routines consume the "%}"
3318 // Check for probable extra ';' after defining block.
3319 if ( _curchar == ';' ) {
3320 parse_err(SYNERR, "Extra ';' after defining interface block.\n");
3321 next_char(); // Skip ';'
3322 return NULL;
3323 }
3324 } else {
3325 next_char(); // move past ';'
3327 // Create appropriate interface object
3328 if ( strcmp(iface_name,"REG_INTER") == 0 ) {
3329 inter = new RegInterface();
3330 }
3331 else if ( strcmp(iface_name,"CONST_INTER") == 0 ) {
3332 inter = new ConstInterface();
3333 }
3334 }
3335 skipws(); // be friendly to oper_parse()
3336 // Debug Stuff
3337 if (_AD._adl_debug > 1) fprintf(stderr,"Interface Form: %s\n", iface_name);
3339 // Create appropriate interface object and return.
3340 return inter;
3341 }
3344 //------------------------------mem_interface_parse----------------------------
3345 Interface *ADLParser::mem_interface_parse(void) {
3346 // Fields for MemInterface
3347 char *base = NULL;
3348 char *index = NULL;
3349 char *scale = NULL;
3350 char *disp = NULL;
3352 if (_curchar != '%') {
3353 parse_err(SYNERR, "Missing '%{' for 'interface' block.\n");
3354 return NULL;
3355 }
3356 next_char(); // Skip '%'
3357 if (_curchar != '{') {
3358 parse_err(SYNERR, "Missing '%{' for 'interface' block.\n");
3359 return NULL;
3360 }
3361 next_char(); // Skip '{'
3362 skipws();
3363 do {
3364 char *field = get_ident();
3365 if (field == NULL) {
3366 parse_err(SYNERR, "Expected keyword, base|index|scale|disp, or '%}' ending interface.\n");
3367 return NULL;
3368 }
3369 if ( strcmp(field,"base") == 0 ) {
3370 base = interface_field_parse();
3371 }
3372 else if ( strcmp(field,"index") == 0 ) {
3373 index = interface_field_parse();
3374 }
3375 else if ( strcmp(field,"scale") == 0 ) {
3376 scale = interface_field_parse();
3377 }
3378 else if ( strcmp(field,"disp") == 0 ) {
3379 disp = interface_field_parse();
3380 }
3381 else {
3382 parse_err(SYNERR, "Expected keyword, base|index|scale|disp, or '%}' ending interface.\n");
3383 return NULL;
3384 }
3385 } while( _curchar != '%' );
3386 next_char(); // Skip '%'
3387 if ( _curchar != '}' ) {
3388 parse_err(SYNERR, "Missing '%}' for 'interface' block.\n");
3389 return NULL;
3390 }
3391 next_char(); // Skip '}'
3393 // Construct desired object and return
3394 Interface *inter = new MemInterface(base, index, scale, disp);
3395 return inter;
3396 }
3399 //------------------------------cond_interface_parse---------------------------
3400 Interface *ADLParser::cond_interface_parse(void) {
3401 char *equal;
3402 char *not_equal;
3403 char *less;
3404 char *greater_equal;
3405 char *less_equal;
3406 char *greater;
3407 const char *equal_format = "eq";
3408 const char *not_equal_format = "ne";
3409 const char *less_format = "lt";
3410 const char *greater_equal_format = "ge";
3411 const char *less_equal_format = "le";
3412 const char *greater_format = "gt";
3414 if (_curchar != '%') {
3415 parse_err(SYNERR, "Missing '%{' for 'cond_interface' block.\n");
3416 return NULL;
3417 }
3418 next_char(); // Skip '%'
3419 if (_curchar != '{') {
3420 parse_err(SYNERR, "Missing '%{' for 'cond_interface' block.\n");
3421 return NULL;
3422 }
3423 next_char(); // Skip '{'
3424 skipws();
3425 do {
3426 char *field = get_ident();
3427 if (field == NULL) {
3428 parse_err(SYNERR, "Expected keyword, base|index|scale|disp, or '%}' ending interface.\n");
3429 return NULL;
3430 }
3431 if ( strcmp(field,"equal") == 0 ) {
3432 equal = interface_field_parse(&equal_format);
3433 }
3434 else if ( strcmp(field,"not_equal") == 0 ) {
3435 not_equal = interface_field_parse(¬_equal_format);
3436 }
3437 else if ( strcmp(field,"less") == 0 ) {
3438 less = interface_field_parse(&less_format);
3439 }
3440 else if ( strcmp(field,"greater_equal") == 0 ) {
3441 greater_equal = interface_field_parse(&greater_equal_format);
3442 }
3443 else if ( strcmp(field,"less_equal") == 0 ) {
3444 less_equal = interface_field_parse(&less_equal_format);
3445 }
3446 else if ( strcmp(field,"greater") == 0 ) {
3447 greater = interface_field_parse(&greater_format);
3448 }
3449 else {
3450 parse_err(SYNERR, "Expected keyword, base|index|scale|disp, or '%}' ending interface.\n");
3451 return NULL;
3452 }
3453 } while( _curchar != '%' );
3454 next_char(); // Skip '%'
3455 if ( _curchar != '}' ) {
3456 parse_err(SYNERR, "Missing '%}' for 'interface' block.\n");
3457 return NULL;
3458 }
3459 next_char(); // Skip '}'
3461 // Construct desired object and return
3462 Interface *inter = new CondInterface(equal, equal_format,
3463 not_equal, not_equal_format,
3464 less, less_format,
3465 greater_equal, greater_equal_format,
3466 less_equal, less_equal_format,
3467 greater, greater_format);
3468 return inter;
3469 }
3472 //------------------------------interface_field_parse--------------------------
3473 char *ADLParser::interface_field_parse(const char ** format) {
3474 char *iface_field = NULL;
3476 // Get interface field
3477 skipws(); // Skip whitespace
3478 if (_curchar != '(') {
3479 parse_err(SYNERR, "Missing '(' at start of interface field.\n");
3480 return NULL;
3481 }
3482 next_char(); // move past '('
3483 skipws();
3484 if ( _curchar != '0' && _curchar != '$' ) {
3485 parse_err(SYNERR, "missing or invalid interface field contents.\n");
3486 return NULL;
3487 }
3488 iface_field = get_rep_var_ident();
3489 if (iface_field == NULL) {
3490 parse_err(SYNERR, "missing or invalid interface field contents.\n");
3491 return NULL;
3492 }
3493 skipws();
3494 if (format != NULL && _curchar == ',') {
3495 next_char();
3496 skipws();
3497 if (_curchar != '"') {
3498 parse_err(SYNERR, "Missing '\"' in field format .\n");
3499 return NULL;
3500 }
3501 next_char();
3502 char *start = _ptr; // Record start of the next string
3503 while ((_curchar != '"') && (_curchar != '%') && (_curchar != '\n')) {
3504 if (_curchar == '\\') next_char(); // superquote
3505 if (_curchar == '\n') parse_err(SYNERR, "newline in string"); // unimplemented!
3506 next_char();
3507 }
3508 if (_curchar != '"') {
3509 parse_err(SYNERR, "Missing '\"' at end of field format .\n");
3510 return NULL;
3511 }
3512 // If a string was found, terminate it and record in FormatRule
3513 if ( start != _ptr ) {
3514 *_ptr = '\0'; // Terminate the string
3515 *format = start;
3516 }
3517 next_char();
3518 skipws();
3519 }
3520 if (_curchar != ')') {
3521 parse_err(SYNERR, "Missing ')' after interface field.\n");
3522 return NULL;
3523 }
3524 next_char(); // move past ')'
3525 skipws();
3526 if ( _curchar != ';' ) {
3527 parse_err(SYNERR, "Missing ';' at end of interface field.\n");
3528 return NULL;
3529 }
3530 next_char(); // move past ';'
3531 skipws(); // be friendly to interface_parse()
3533 return iface_field;
3534 }
3537 //------------------------------match_parse------------------------------------
3538 MatchRule *ADLParser::match_parse(FormDict &operands) {
3539 MatchRule *match; // Match Rule class for instruction/operand
3540 char *cnstr = NULL; // Code for constructor
3541 int depth = 0; // Counter for matching parentheses
3542 int numleaves = 0; // Counter for number of leaves in rule
3544 // Parse the match rule tree
3545 MatchNode *mnode = matchNode_parse(operands, depth, numleaves, true);
3547 // Either there is a block with a constructor, or a ';' here
3548 skipws(); // Skip whitespace
3549 if ( _curchar == ';' ) { // Semicolon is valid terminator
3550 cnstr = NULL; // no constructor for this form
3551 next_char(); // Move past the ';', replaced with '\0'
3552 }
3553 else if ((cnstr = find_cpp_block("match constructor")) == NULL ) {
3554 parse_err(SYNERR, "invalid construction of match rule\n"
3555 "Missing ';' or invalid '%{' and '%}' constructor\n");
3556 return NULL; // No MatchRule to return
3557 }
3558 if (_AD._adl_debug > 1)
3559 if (cnstr) fprintf(stderr,"Match Constructor: %s\n", cnstr);
3560 // Build new MatchRule object
3561 match = new MatchRule(_AD, mnode, depth, cnstr, numleaves);
3562 skipws(); // Skip any trailing whitespace
3563 return match; // Return MatchRule object
3564 }
3566 //------------------------------format_parse-----------------------------------
3567 FormatRule* ADLParser::format_parse(void) {
3568 char *desc = NULL;
3569 FormatRule *format = (new FormatRule(desc));
3571 // Without expression form, MUST have a code block;
3572 skipws(); // Skip whitespace
3573 if ( _curchar == ';' ) { // Semicolon is valid terminator
3574 desc = NULL; // no constructor for this form
3575 next_char(); // Move past the ';', replaced with '\0'
3576 }
3577 else if ( _curchar == '%' && *(_ptr+1) == '{') {
3578 next_char(); // Move past the '%'
3579 next_char(); // Move past the '{'
3581 skipws();
3582 if (_curchar == '$') {
3583 char* ident = get_rep_var_ident();
3584 if (strcmp(ident, "$$template") == 0) return template_parse();
3585 parse_err(SYNERR, "Unknown \"%s\" directive in format", ident);
3586 return NULL;
3587 }
3588 // Check for the opening '"' inside the format description
3589 if ( _curchar == '"' ) {
3590 next_char(); // Move past the initial '"'
3591 if( _curchar == '"' ) { // Handle empty format string case
3592 *_ptr = '\0'; // Terminate empty string
3593 format->_strings.addName(_ptr);
3594 }
3596 // Collect the parts of the format description
3597 // (1) strings that are passed through to tty->print
3598 // (2) replacement/substitution variable, preceeded by a '$'
3599 // (3) multi-token ANSIY C style strings
3600 while ( true ) {
3601 if ( _curchar == '%' || _curchar == '\n' ) {
3602 if ( _curchar != '"' ) {
3603 parse_err(SYNERR, "missing '\"' at end of format block");
3604 return NULL;
3605 }
3606 }
3608 // (1)
3609 // Check if there is a string to pass through to output
3610 char *start = _ptr; // Record start of the next string
3611 while ((_curchar != '$') && (_curchar != '"') && (_curchar != '%') && (_curchar != '\n')) {
3612 if (_curchar == '\\') {
3613 next_char(); // superquote
3614 if ((_curchar == '$') || (_curchar == '%'))
3615 // hack to avoid % escapes and warnings about undefined \ escapes
3616 *(_ptr-1) = _curchar;
3617 }
3618 if (_curchar == '\n') parse_err(SYNERR, "newline in string"); // unimplemented!
3619 next_char();
3620 }
3621 // If a string was found, terminate it and record in FormatRule
3622 if ( start != _ptr ) {
3623 *_ptr = '\0'; // Terminate the string
3624 format->_strings.addName(start);
3625 }
3627 // (2)
3628 // If we are at a replacement variable,
3629 // copy it and record in FormatRule
3630 if ( _curchar == '$' ) {
3631 next_char(); // Move past the '$'
3632 char* rep_var = get_ident(); // Nil terminate the variable name
3633 rep_var = strdup(rep_var);// Copy the string
3634 *_ptr = _curchar; // and replace Nil with original character
3635 format->_rep_vars.addName(rep_var);
3636 // Add flag to _strings list indicating we should check _rep_vars
3637 format->_strings.addName(NameList::_signal);
3638 }
3640 // (3)
3641 // Allow very long strings to be broken up,
3642 // using the ANSI C syntax "foo\n" <newline> "bar"
3643 if ( _curchar == '"') {
3644 next_char(); // Move past the '"'
3645 skipws(); // Skip white space before next string token
3646 if ( _curchar != '"') {
3647 break;
3648 } else {
3649 // Found one. Skip both " and the whitespace in between.
3650 next_char();
3651 }
3652 }
3653 } // end while part of format description
3655 // Check for closing '"' and '%}' in format description
3656 skipws(); // Move to closing '%}'
3657 if ( _curchar != '%' ) {
3658 parse_err(SYNERR, "non-blank characters between closing '\"' and '%' in format");
3659 return NULL;
3660 }
3661 } // Done with format description inside
3663 skipws();
3664 // Past format description, at '%'
3665 if ( _curchar != '%' || *(_ptr+1) != '}' ) {
3666 parse_err(SYNERR, "missing '%}' at end of format block");
3667 return NULL;
3668 }
3669 next_char(); // Move past the '%'
3670 next_char(); // Move past the '}'
3671 }
3672 else { // parameter list alone must terminate with a ';'
3673 parse_err(SYNERR, "missing ';' after Format expression");
3674 return NULL;
3675 }
3676 // Debug Stuff
3677 if (_AD._adl_debug > 1) fprintf(stderr,"Format Rule: %s\n", desc);
3679 skipws();
3680 return format;
3681 }
3684 //------------------------------template_parse-----------------------------------
3685 FormatRule* ADLParser::template_parse(void) {
3686 char *desc = NULL;
3687 FormatRule *format = (new FormatRule(desc));
3689 skipws();
3690 while ( (_curchar != '%') && (*(_ptr+1) != '}') ) {
3692 // (1)
3693 // Check if there is a string to pass through to output
3694 {
3695 char *start = _ptr; // Record start of the next string
3696 while ((_curchar != '$') && ((_curchar != '%') || (*(_ptr+1) != '}')) ) {
3697 // If at the start of a comment, skip past it
3698 if( (_curchar == '/') && ((*(_ptr+1) == '/') || (*(_ptr+1) == '*')) ) {
3699 skipws_no_preproc();
3700 } else {
3701 // ELSE advance to the next character, or start of the next line
3702 next_char_or_line();
3703 }
3704 }
3705 // If a string was found, terminate it and record in EncClass
3706 if ( start != _ptr ) {
3707 *_ptr = '\0'; // Terminate the string
3708 // Add flag to _strings list indicating we should check _rep_vars
3709 format->_strings.addName(NameList::_signal2);
3710 format->_strings.addName(start);
3711 }
3712 }
3714 // (2)
3715 // If we are at a replacement variable,
3716 // copy it and record in EncClass
3717 if ( _curchar == '$' ) {
3718 // Found replacement Variable
3719 char *rep_var = get_rep_var_ident_dup();
3720 if (strcmp(rep_var, "$emit") == 0) {
3721 // switch to normal format parsing
3722 next_char();
3723 next_char();
3724 skipws();
3725 // Check for the opening '"' inside the format description
3726 if ( _curchar == '"' ) {
3727 next_char(); // Move past the initial '"'
3728 if( _curchar == '"' ) { // Handle empty format string case
3729 *_ptr = '\0'; // Terminate empty string
3730 format->_strings.addName(_ptr);
3731 }
3733 // Collect the parts of the format description
3734 // (1) strings that are passed through to tty->print
3735 // (2) replacement/substitution variable, preceeded by a '$'
3736 // (3) multi-token ANSIY C style strings
3737 while ( true ) {
3738 if ( _curchar == '%' || _curchar == '\n' ) {
3739 parse_err(SYNERR, "missing '\"' at end of format block");
3740 return NULL;
3741 }
3743 // (1)
3744 // Check if there is a string to pass through to output
3745 char *start = _ptr; // Record start of the next string
3746 while ((_curchar != '$') && (_curchar != '"') && (_curchar != '%') && (_curchar != '\n')) {
3747 if (_curchar == '\\') next_char(); // superquote
3748 if (_curchar == '\n') parse_err(SYNERR, "newline in string"); // unimplemented!
3749 next_char();
3750 }
3751 // If a string was found, terminate it and record in FormatRule
3752 if ( start != _ptr ) {
3753 *_ptr = '\0'; // Terminate the string
3754 format->_strings.addName(start);
3755 }
3757 // (2)
3758 // If we are at a replacement variable,
3759 // copy it and record in FormatRule
3760 if ( _curchar == '$' ) {
3761 next_char(); // Move past the '$'
3762 char* next_rep_var = get_ident(); // Nil terminate the variable name
3763 next_rep_var = strdup(next_rep_var);// Copy the string
3764 *_ptr = _curchar; // and replace Nil with original character
3765 format->_rep_vars.addName(next_rep_var);
3766 // Add flag to _strings list indicating we should check _rep_vars
3767 format->_strings.addName(NameList::_signal);
3768 }
3770 // (3)
3771 // Allow very long strings to be broken up,
3772 // using the ANSI C syntax "foo\n" <newline> "bar"
3773 if ( _curchar == '"') {
3774 next_char(); // Move past the '"'
3775 skipws(); // Skip white space before next string token
3776 if ( _curchar != '"') {
3777 break;
3778 } else {
3779 // Found one. Skip both " and the whitespace in between.
3780 next_char();
3781 }
3782 }
3783 } // end while part of format description
3784 }
3785 } else {
3786 // Add flag to _strings list indicating we should check _rep_vars
3787 format->_rep_vars.addName(rep_var);
3788 // Add flag to _strings list indicating we should check _rep_vars
3789 format->_strings.addName(NameList::_signal3);
3790 }
3791 } // end while part of format description
3792 }
3794 skipws();
3795 // Past format description, at '%'
3796 if ( _curchar != '%' || *(_ptr+1) != '}' ) {
3797 parse_err(SYNERR, "missing '%}' at end of format block");
3798 return NULL;
3799 }
3800 next_char(); // Move past the '%'
3801 next_char(); // Move past the '}'
3803 // Debug Stuff
3804 if (_AD._adl_debug > 1) fprintf(stderr,"Format Rule: %s\n", desc);
3806 skipws();
3807 return format;
3808 }
3811 //------------------------------effect_parse-----------------------------------
3812 void ADLParser::effect_parse(InstructForm *instr) {
3813 char* desc = NULL;
3815 skipws(); // Skip whitespace
3816 if (_curchar != '(') {
3817 parse_err(SYNERR, "missing '(' in effect definition\n");
3818 return;
3819 }
3820 // Get list of effect-operand pairs and insert into dictionary
3821 else get_effectlist(instr->_effects, instr->_localNames, instr->_has_call);
3823 // Debug Stuff
3824 if (_AD._adl_debug > 1) fprintf(stderr,"Effect description: %s\n", desc);
3825 if (_curchar != ';') {
3826 parse_err(SYNERR, "missing ';' in Effect definition\n");
3827 }
3828 next_char(); // Skip ';'
3830 }
3832 //------------------------------expand_parse-----------------------------------
3833 ExpandRule* ADLParser::expand_parse(InstructForm *instr) {
3834 char *ident, *ident2;
3835 OperandForm *oper;
3836 InstructForm *ins;
3837 NameAndList *instr_and_operands = NULL;
3838 ExpandRule *exp = new ExpandRule();
3840 // Expand is a block containing an ordered list of instructions, each of
3841 // which has an ordered list of operands.
3842 // Check for block delimiter
3843 skipws(); // Skip leading whitespace
3844 if ((_curchar != '%')
3845 || (next_char(), (_curchar != '{')) ) { // If not open block
3846 parse_err(SYNERR, "missing '%{' in expand definition\n");
3847 return(NULL);
3848 }
3849 next_char(); // Maintain the invariant
3850 do {
3851 ident = get_ident(); // Grab next identifier
3852 if (ident == NULL) {
3853 parse_err(SYNERR, "identifier expected at %c\n", _curchar);
3854 continue;
3855 } // Check that you have a valid instruction
3856 const Form *form = _globalNames[ident];
3857 ins = form ? form->is_instruction() : NULL;
3858 if (ins == NULL) {
3859 // This is a new operand
3860 oper = form ? form->is_operand() : NULL;
3861 if (oper == NULL) {
3862 parse_err(SYNERR, "instruction/operand name expected at %s\n", ident);
3863 continue;
3864 }
3865 // Throw the operand on the _newopers list
3866 skipws();
3867 ident = get_unique_ident(instr->_localNames,"Operand");
3868 if (ident == NULL) {
3869 parse_err(SYNERR, "identifier expected at %c\n", _curchar);
3870 continue;
3871 }
3872 exp->_newopers.addName(ident);
3873 // Add new operand to LocalNames
3874 instr->_localNames.Insert(ident, oper);
3875 // Grab any constructor code and save as a string
3876 char *c = NULL;
3877 skipws();
3878 if (_curchar == '%') { // Need a constructor for the operand
3879 c = find_cpp_block("Operand Constructor");
3880 if (c == NULL) {
3881 parse_err(SYNERR, "Invalid code block for operand constructor\n", _curchar);
3882 continue;
3883 }
3884 // Add constructor to _newopconst Dict
3885 exp->_newopconst.Insert(ident, c);
3886 }
3887 else if (_curchar != ';') { // If no constructor, need a ;
3888 parse_err(SYNERR, "Missing ; in expand rule operand declaration\n");
3889 continue;
3890 }
3891 else next_char(); // Skip the ;
3892 skipws();
3893 }
3894 else {
3895 // Add instruction to list
3896 instr_and_operands = new NameAndList(ident);
3897 // Grab operands, build nameList of them, and then put into dictionary
3898 skipws();
3899 if (_curchar != '(') { // Check for parenthesized operand list
3900 parse_err(SYNERR, "missing '(' in expand instruction declaration\n");
3901 continue;
3902 }
3903 do {
3904 next_char(); // skip open paren & comma characters
3905 skipws();
3906 if (_curchar == ')') break;
3907 ident2 = get_ident();
3908 skipws();
3909 if (ident2 == NULL) {
3910 parse_err(SYNERR, "identifier expected at %c\n", _curchar);
3911 continue;
3912 } // Check that you have a valid operand
3913 const Form *form2 = instr->_localNames[ident2];
3914 if (!form2) {
3915 parse_err(SYNERR, "operand name expected at %s\n", ident2);
3916 continue;
3917 }
3918 oper = form2->is_operand();
3919 if (oper == NULL && !form2->is_opclass()) {
3920 parse_err(SYNERR, "operand name expected at %s\n", ident2);
3921 continue;
3922 } // Add operand to list
3923 instr_and_operands->add_entry(ident2);
3924 } while(_curchar == ',');
3925 if (_curchar != ')') {
3926 parse_err(SYNERR, "missing ')'in expand instruction declaration\n");
3927 continue;
3928 }
3929 next_char();
3930 if (_curchar != ';') {
3931 parse_err(SYNERR, "missing ';'in expand instruction declaration\n");
3932 continue;
3933 }
3934 next_char();
3936 // Record both instruction name and its operand list
3937 exp->add_instruction(instr_and_operands);
3939 skipws();
3940 }
3942 } while(_curchar != '%');
3943 next_char();
3944 if (_curchar != '}') {
3945 parse_err(SYNERR, "missing '%}' in expand rule definition\n");
3946 return(NULL);
3947 }
3948 next_char();
3950 // Debug Stuff
3951 if (_AD._adl_debug > 1) fprintf(stderr,"Expand Rule:\n");
3953 skipws();
3954 return (exp);
3955 }
3957 //------------------------------rewrite_parse----------------------------------
3958 RewriteRule* ADLParser::rewrite_parse(void) {
3959 char* params = NULL;
3960 char* desc = NULL;
3963 // This feature targeted for second generation description language.
3965 skipws(); // Skip whitespace
3966 // Get parameters for rewrite
3967 if ((params = get_paren_expr("rewrite parameters")) == NULL) {
3968 parse_err(SYNERR, "missing '(' in rewrite rule\n");
3969 return NULL;
3970 }
3971 // Debug Stuff
3972 if (_AD._adl_debug > 1) fprintf(stderr,"Rewrite parameters: %s\n", params);
3974 // For now, grab entire block;
3975 skipws();
3976 if ( (desc = find_cpp_block("rewrite block")) == NULL ) {
3977 parse_err(SYNERR, "incorrect or missing block for 'rewrite'.\n");
3978 return NULL;
3979 }
3980 // Debug Stuff
3981 if (_AD._adl_debug > 1) fprintf(stderr,"Rewrite Rule: %s\n", desc);
3983 skipws();
3984 return (new RewriteRule(params,desc));
3985 }
3987 //------------------------------attr_parse-------------------------------------
3988 Attribute *ADLParser::attr_parse(char* ident) {
3989 Attribute *attrib; // Attribute class
3990 char *cost = NULL; // String representation of cost attribute
3992 skipws(); // Skip leading whitespace
3993 if ( (cost = get_paren_expr("attribute")) == NULL ) {
3994 parse_err(SYNERR, "incorrect or missing expression for 'attribute'\n");
3995 return NULL;
3996 }
3997 // Debug Stuff
3998 if (_AD._adl_debug > 1) fprintf(stderr,"Attribute: %s\n", cost);
3999 if (_curchar != ';') {
4000 parse_err(SYNERR, "missing ';' in attribute definition\n");
4001 return NULL;
4002 }
4003 next_char(); // Point after the terminator
4005 skipws();
4006 attrib = new Attribute(ident,cost,INS_ATTR); // Build new predicate object
4007 return attrib;
4008 }
4011 //------------------------------matchNode_parse--------------------------------
4012 MatchNode *ADLParser::matchNode_parse(FormDict &operands, int &depth, int &numleaves, bool atroot) {
4013 // Count depth of parenthesis nesting for both left and right children
4014 int lParens = depth;
4015 int rParens = depth;
4017 // MatchNode objects for left, right, and root of subtree.
4018 MatchNode *lChild = NULL;
4019 MatchNode *rChild = NULL;
4020 char *token; // Identifier which may be opcode or operand
4022 // Match expression starts with a '('
4023 if (cur_char() != '(')
4024 return NULL;
4026 next_char(); // advance past '('
4028 // Parse the opcode
4029 token = get_ident(); // Get identifier, opcode
4030 if (token == NULL) {
4031 parse_err(SYNERR, "missing opcode in match expression\n");
4032 return NULL;
4033 }
4035 // Take note if we see one of a few special operations - those that are
4036 // treated differently on different architectures in the sense that on
4037 // one architecture there is a match rule and on another there isn't (so
4038 // a call will eventually be generated).
4040 for (int i = _last_machine_leaf + 1; i < _last_opcode; i++) {
4041 if (strcmp(token, NodeClassNames[i]) == 0) {
4042 _AD.has_match_rule(i, true);
4043 }
4044 }
4046 // Lookup the root value in the operands dict to perform substitution
4047 const char *result = NULL; // Result type will be filled in later
4048 const char *name = token; // local name associated with this node
4049 const char *operation = token; // remember valid operation for later
4050 const Form *form = operands[token];
4051 OpClassForm *opcForm = form ? form->is_opclass() : NULL;
4052 if (opcForm != NULL) {
4053 // If this token is an entry in the local names table, record its type
4054 if (!opcForm->ideal_only()) {
4055 operation = opcForm->_ident;
4056 result = operation; // Operands result in their own type
4057 }
4058 // Otherwise it is an ideal type, and so, has no local name
4059 else name = NULL;
4060 }
4062 // Parse the operands
4063 skipws();
4064 if (cur_char() != ')') {
4066 // Parse the left child
4067 if (strcmp(operation,"Set"))
4068 lChild = matchChild_parse(operands, lParens, numleaves, false);
4069 else
4070 lChild = matchChild_parse(operands, lParens, numleaves, true);
4072 skipws();
4073 if (cur_char() != ')' ) {
4074 if(strcmp(operation, "Set"))
4075 rChild = matchChild_parse(operands,rParens,numleaves,false);
4076 else
4077 rChild = matchChild_parse(operands,rParens,numleaves,true);
4078 }
4079 }
4081 // Check for required ')'
4082 skipws();
4083 if (cur_char() != ')') {
4084 parse_err(SYNERR, "missing ')' in match expression\n");
4085 return NULL;
4086 }
4087 next_char(); // skip the ')'
4089 MatchNode* mroot = new MatchNode(_AD,result,name,operation,lChild,rChild);
4091 // If not the root, reduce this subtree to an internal operand
4092 if (!atroot) {
4093 mroot->build_internalop();
4094 }
4095 // depth is greater of left and right paths.
4096 depth = (lParens > rParens) ? lParens : rParens;
4098 return mroot;
4099 }
4102 //------------------------------matchChild_parse-------------------------------
4103 MatchNode *ADLParser::matchChild_parse(FormDict &operands, int &parens, int &numleaves, bool atroot) {
4104 MatchNode *child = NULL;
4105 const char *result = NULL;
4106 const char *token = NULL;
4107 const char *opType = NULL;
4109 if (cur_char() == '(') { // child is an operation
4110 ++parens;
4111 child = matchNode_parse(operands, parens, numleaves, atroot);
4112 }
4113 else { // child is an operand
4114 token = get_ident();
4115 const Form *form = operands[token];
4116 OpClassForm *opcForm = form ? form->is_opclass() : NULL;
4117 if (opcForm != NULL) {
4118 opType = opcForm->_ident;
4119 result = opcForm->_ident; // an operand's result matches its type
4120 } else {
4121 parse_err(SYNERR, "undefined operand %s in match rule\n", token);
4122 return NULL;
4123 }
4125 if (opType == NULL) {
4126 parse_err(SYNERR, "missing type for argument '%s'\n", token);
4127 }
4129 child = new MatchNode(_AD, result, token, opType);
4130 ++numleaves;
4131 }
4133 return child;
4134 }
4138 // ******************** Private Utility Functions *************************
4141 char* ADLParser::find_cpp_block(const char* description) {
4142 char *next; // Pointer for finding block delimiters
4143 char* cppBlock = NULL; // Beginning of C++ code block
4145 if (_curchar == '%') { // Encoding is a C++ expression
4146 next_char();
4147 if (_curchar != '{') {
4148 parse_err(SYNERR, "missing '{' in %s \n", description);
4149 return NULL;
4150 }
4151 next_char(); // Skip block delimiter
4152 skipws_no_preproc(); // Skip leading whitespace
4153 cppBlock = _ptr; // Point to start of expression
4154 int line = linenum();
4155 next = _ptr + 1;
4156 while(((_curchar != '%') || (*next != '}')) && (_curchar != '\0')) {
4157 next_char_or_line();
4158 next = _ptr+1; // Maintain the next pointer
4159 } // Grab string
4160 if (_curchar == '\0') {
4161 parse_err(SYNERR, "invalid termination of %s \n", description);
4162 return NULL;
4163 }
4164 *_ptr = '\0'; // Terminate string
4165 _ptr += 2; // Skip block delimiter
4166 _curchar = *_ptr; // Maintain invariant
4168 // Prepend location descriptor, for debugging.
4169 if (_AD._adlocation_debug) {
4170 char* location = get_line_string(line);
4171 char* end_loc = end_line_marker();
4172 char* result = (char *)malloc(strlen(location) + strlen(cppBlock) + strlen(end_loc) + 1);
4173 strcpy(result, location);
4174 strcat(result, cppBlock);
4175 strcat(result, end_loc);
4176 cppBlock = result;
4177 free(location);
4178 }
4179 }
4181 return cppBlock;
4182 }
4184 // Move to the closing token of the expression we are currently at,
4185 // as defined by stop_chars. Match parens and quotes.
4186 char* ADLParser::get_expr(const char *desc, const char *stop_chars) {
4187 char* expr = NULL;
4188 int paren = 0;
4190 expr = _ptr;
4191 while (paren > 0 || !strchr(stop_chars, _curchar)) {
4192 if (_curchar == '(') { // Down level of nesting
4193 paren++; // Bump the parenthesis counter
4194 next_char(); // maintain the invariant
4195 }
4196 else if (_curchar == ')') { // Up one level of nesting
4197 if (paren == 0) {
4198 // Paren underflow: We didn't encounter the required stop-char.
4199 parse_err(SYNERR, "too many )'s, did not find %s after %s\n",
4200 stop_chars, desc);
4201 return NULL;
4202 }
4203 paren--; // Drop the parenthesis counter
4204 next_char(); // Maintain the invariant
4205 }
4206 else if (_curchar == '"' || _curchar == '\'') {
4207 int qchar = _curchar;
4208 while (true) {
4209 next_char();
4210 if (_curchar == qchar) { next_char(); break; }
4211 if (_curchar == '\\') next_char(); // superquote
4212 if (_curchar == '\n' || _curchar == '\0') {
4213 parse_err(SYNERR, "newline in string in %s\n", desc);
4214 return NULL;
4215 }
4216 }
4217 }
4218 else if (_curchar == '%' && (_ptr[1] == '{' || _ptr[1] == '}')) {
4219 // Make sure we do not stray into the next ADLC-level form.
4220 parse_err(SYNERR, "unexpected %%%c in %s\n", _ptr[1], desc);
4221 return NULL;
4222 }
4223 else if (_curchar == '\0') {
4224 parse_err(SYNERR, "unexpected EOF in %s\n", desc);
4225 return NULL;
4226 }
4227 else {
4228 // Always walk over whitespace, comments, preprocessor directives, etc.
4229 char* pre_skip_ptr = _ptr;
4230 skipws();
4231 // If the parser declined to make progress on whitespace,
4232 // skip the next character, which is therefore NOT whitespace.
4233 if (pre_skip_ptr == _ptr) {
4234 next_char();
4235 } else if (pre_skip_ptr+strlen(pre_skip_ptr) != _ptr+strlen(_ptr)) {
4236 parse_err(SYNERR, "unimplemented: preprocessor must not elide subexpression in %s", desc);
4237 }
4238 }
4239 }
4241 assert(strchr(stop_chars, _curchar), "non-null return must be at stop-char");
4242 *_ptr = '\0'; // Replace ')' or other stop-char with '\0'
4243 return expr;
4244 }
4246 // Helper function around get_expr
4247 // Sets _curchar to '(' so that get_paren_expr will search for a matching ')'
4248 char *ADLParser::get_paren_expr(const char *description, bool include_location) {
4249 int line = linenum();
4250 if (_curchar != '(') // Escape if not valid starting position
4251 return NULL;
4252 next_char(); // Skip the required initial paren.
4253 char *token2 = get_expr(description, ")");
4254 if (_curchar == ')')
4255 next_char(); // Skip required final paren.
4256 int junk = 0;
4257 if (include_location && _AD._adlocation_debug && !is_int_token(token2, junk)) {
4258 // Prepend location descriptor, for debugging.
4259 char* location = get_line_string(line);
4260 char* end_loc = end_line_marker();
4261 char* result = (char *)malloc(strlen(location) + strlen(token2) + strlen(end_loc) + 1);
4262 strcpy(result, location);
4263 strcat(result, token2);
4264 strcat(result, end_loc);
4265 token2 = result;
4266 free(location);
4267 }
4268 return token2;
4269 }
4271 //------------------------------get_ident_common-------------------------------
4272 // Looks for an identifier in the buffer, and turns it into a null terminated
4273 // string(still inside the file buffer). Returns a pointer to the string or
4274 // NULL if some other token is found instead.
4275 char *ADLParser::get_ident_common(bool do_preproc) {
4276 register char c;
4277 char *start; // Pointer to start of token
4278 char *end; // Pointer to end of token
4280 if( _curline == NULL ) // Return NULL at EOF.
4281 return NULL;
4283 skipws_common(do_preproc); // Skip whitespace before identifier
4284 start = end = _ptr; // Start points at first character
4285 end--; // unwind end by one to prepare for loop
4286 do {
4287 end++; // Increment end pointer
4288 c = *end; // Grab character to test
4289 } while ( ((c >= 'a') && (c <= 'z')) || ((c >= 'A') && (c <= 'Z'))
4290 || ((c >= '0') && (c <= '9'))
4291 || ((c == '_')) || ((c == ':')) || ((c == '#')) );
4292 if (start == end) { // We popped out on the first try
4293 parse_err(SYNERR, "identifier expected at %c\n", c);
4294 start = NULL;
4295 }
4296 else {
4297 _curchar = c; // Save the first character of next token
4298 *end = '\0'; // NULL terminate the string in place
4299 }
4300 _ptr = end; // Reset _ptr to point to next char after token
4302 // Make sure we do not try to use #defined identifiers. If start is
4303 // NULL an error was already reported.
4304 if (do_preproc && start != NULL) {
4305 const char* def = _AD.get_preproc_def(start);
4306 if (def != NULL && strcmp(def, start)) {
4307 const char* def1 = def;
4308 const char* def2 = _AD.get_preproc_def(def1);
4309 // implement up to 2 levels of #define
4310 if (def2 != NULL && strcmp(def2, def1)) {
4311 def = def2;
4312 const char* def3 = _AD.get_preproc_def(def2);
4313 if (def3 != NULL && strcmp(def3, def2) && strcmp(def3, def1)) {
4314 parse_err(SYNERR, "unimplemented: using %s defined as %s => %s => %s",
4315 start, def1, def2, def3);
4316 }
4317 }
4318 start = strdup(def);
4319 }
4320 }
4322 return start; // Pointer to token in filebuf
4323 }
4325 //------------------------------get_ident_dup----------------------------------
4326 // Looks for an identifier in the buffer, and returns a duplicate
4327 // or NULL if some other token is found instead.
4328 char *ADLParser::get_ident_dup(void) {
4329 char *ident = get_ident();
4331 // Duplicate an identifier before returning and restore string.
4332 if( ident != NULL ) {
4333 ident = strdup(ident); // Copy the string
4334 *_ptr = _curchar; // and replace Nil with original character
4335 }
4337 return ident;
4338 }
4340 //----------------------get_ident_or_literal_constant--------------------------
4341 // Looks for an identifier in the buffer, or a parenthesized expression.
4342 char *ADLParser::get_ident_or_literal_constant(const char* description) {
4343 char* param = NULL;
4344 skipws();
4345 if (_curchar == '(') {
4346 // Grab a constant expression.
4347 param = get_paren_expr(description);
4348 if (param[0] != '(') {
4349 char* buf = (char*) malloc(strlen(param) + 3);
4350 sprintf(buf, "(%s)", param);
4351 param = buf;
4352 }
4353 assert(is_literal_constant(param),
4354 "expr must be recognizable as a constant");
4355 } else {
4356 param = get_ident();
4357 }
4358 return param;
4359 }
4361 //------------------------------get_rep_var_ident-----------------------------
4362 // Do NOT duplicate,
4363 // Leave nil terminator in buffer
4364 // Preserve initial '$'(s) in string
4365 char *ADLParser::get_rep_var_ident(void) {
4366 // Remember starting point
4367 char *rep_var = _ptr;
4369 // Check for replacement variable indicator '$' and pass if present
4370 if ( _curchar == '$' ) {
4371 next_char();
4372 }
4373 // Check for a subfield indicator, a second '$', and pass if present
4374 if ( _curchar == '$' ) {
4375 next_char();
4376 }
4378 // Check for a control indicator, a third '$':
4379 if ( _curchar == '$' ) {
4380 next_char();
4381 }
4383 // Check for more than three '$'s in sequence, SYNERR
4384 if( _curchar == '$' ) {
4385 parse_err(SYNERR, "Replacement variables and field specifiers can not start with '$$$$'");
4386 next_char();
4387 return NULL;
4388 }
4390 // Nil terminate the variable name following the '$'
4391 char *rep_var_name = get_ident();
4392 assert( rep_var_name != NULL,
4393 "Missing identifier after replacement variable indicator '$'");
4395 return rep_var;
4396 }
4400 //------------------------------get_rep_var_ident_dup-------------------------
4401 // Return the next replacement variable identifier, skipping first '$'
4402 // given a pointer into a line of the buffer.
4403 // Null terminates string, still inside the file buffer,
4404 // Returns a pointer to a copy of the string, or NULL on failure
4405 char *ADLParser::get_rep_var_ident_dup(void) {
4406 if( _curchar != '$' ) return NULL;
4408 next_char(); // Move past the '$'
4409 char *rep_var = _ptr; // Remember starting point
4411 // Check for a subfield indicator, a second '$':
4412 if ( _curchar == '$' ) {
4413 next_char();
4414 }
4416 // Check for a control indicator, a third '$':
4417 if ( _curchar == '$' ) {
4418 next_char();
4419 }
4421 // Check for more than three '$'s in sequence, SYNERR
4422 if( _curchar == '$' ) {
4423 parse_err(SYNERR, "Replacement variables and field specifiers can not start with '$$$$'");
4424 next_char();
4425 return NULL;
4426 }
4428 // Nil terminate the variable name following the '$'
4429 char *rep_var_name = get_ident();
4430 assert( rep_var_name != NULL,
4431 "Missing identifier after replacement variable indicator '$'");
4432 rep_var = strdup(rep_var); // Copy the string
4433 *_ptr = _curchar; // and replace Nil with original character
4435 return rep_var;
4436 }
4439 //------------------------------get_unique_ident------------------------------
4440 // Looks for an identifier in the buffer, terminates it with a NULL,
4441 // and checks that it is unique
4442 char *ADLParser::get_unique_ident(FormDict& dict, const char* nameDescription){
4443 char* ident = get_ident();
4445 if (ident == NULL) {
4446 parse_err(SYNERR, "missing %s identifier at %c\n", nameDescription, _curchar);
4447 }
4448 else {
4449 if (dict[ident] != NULL) {
4450 parse_err(SYNERR, "duplicate name %s for %s\n", ident, nameDescription);
4451 ident = NULL;
4452 }
4453 }
4455 return ident;
4456 }
4459 //------------------------------get_int----------------------------------------
4460 // Looks for a character string integer in the buffer, and turns it into an int
4461 // invokes a parse_err if the next token is not an integer.
4462 // This routine does not leave the integer null-terminated.
4463 int ADLParser::get_int(void) {
4464 register char c;
4465 char *start; // Pointer to start of token
4466 char *end; // Pointer to end of token
4467 int result; // Storage for integer result
4469 if( _curline == NULL ) // Return NULL at EOF.
4470 return 0;
4472 skipws(); // Skip whitespace before identifier
4473 start = end = _ptr; // Start points at first character
4474 c = *end; // Grab character to test
4475 while ((c >= '0') && (c <= '9')
4476 || ((c == '-') && (end == start))) {
4477 end++; // Increment end pointer
4478 c = *end; // Grab character to test
4479 }
4480 if (start == end) { // We popped out on the first try
4481 parse_err(SYNERR, "integer expected at %c\n", c);
4482 result = 0;
4483 }
4484 else {
4485 _curchar = c; // Save the first character of next token
4486 *end = '\0'; // NULL terminate the string in place
4487 result = atoi(start); // Convert the string to an integer
4488 *end = _curchar; // Restore buffer to original condition
4489 }
4491 // Reset _ptr to next char after token
4492 _ptr = end;
4494 return result; // integer
4495 }
4498 //------------------------------get_relation_dup------------------------------
4499 // Looks for a relational operator in the buffer
4500 // invokes a parse_err if the next token is not a relation
4501 // This routine creates a duplicate of the string in the buffer.
4502 char *ADLParser::get_relation_dup(void) {
4503 char *result = NULL; // relational operator being returned
4505 if( _curline == NULL ) // Return NULL at EOF.
4506 return NULL;
4508 skipws(); // Skip whitespace before relation
4509 char *start = _ptr; // Store start of relational operator
4510 char first = *_ptr; // the first character
4511 if( (first == '=') || (first == '!') || (first == '<') || (first == '>') ) {
4512 next_char();
4513 char second = *_ptr; // the second character
4514 if( (second == '=') ) {
4515 next_char();
4516 char tmp = *_ptr;
4517 *_ptr = '\0'; // NULL terminate
4518 result = strdup(start); // Duplicate the string
4519 *_ptr = tmp; // restore buffer
4520 } else {
4521 parse_err(SYNERR, "relational operator expected at %s\n", _ptr);
4522 }
4523 } else {
4524 parse_err(SYNERR, "relational operator expected at %s\n", _ptr);
4525 }
4527 return result;
4528 }
4532 //------------------------------get_oplist-------------------------------------
4533 // Looks for identifier pairs where first must be the name of an operand, and
4534 // second must be a name unique in the scope of this instruction. Stores the
4535 // names with a pointer to the OpClassForm of their type in a local name table.
4536 void ADLParser::get_oplist(NameList ¶meters, FormDict &operands) {
4537 OpClassForm *opclass = NULL;
4538 char *ident = NULL;
4540 do {
4541 next_char(); // skip open paren & comma characters
4542 skipws();
4543 if (_curchar == ')') break;
4545 // Get operand type, and check it against global name table
4546 ident = get_ident();
4547 if (ident == NULL) {
4548 parse_err(SYNERR, "optype identifier expected at %c\n", _curchar);
4549 return;
4550 }
4551 else {
4552 const Form *form = _globalNames[ident];
4553 if( form == NULL ) {
4554 parse_err(SYNERR, "undefined operand type %s\n", ident);
4555 return;
4556 }
4558 // Check for valid operand type
4559 OpClassForm *opc = form->is_opclass();
4560 OperandForm *oper = form->is_operand();
4561 if((oper == NULL) && (opc == NULL)) {
4562 parse_err(SYNERR, "identifier %s not operand type\n", ident);
4563 return;
4564 }
4565 opclass = opc;
4566 }
4567 // Debugging Stuff
4568 if (_AD._adl_debug > 1) fprintf(stderr, "\tOperand Type: %s\t", ident);
4570 // Get name of operand and add it to local name table
4571 if( (ident = get_unique_ident(operands, "operand")) == NULL) {
4572 return;
4573 }
4574 // Parameter names must not be global names.
4575 if( _globalNames[ident] != NULL ) {
4576 parse_err(SYNERR, "Reuse of global name %s as operand.\n",ident);
4577 return;
4578 }
4579 operands.Insert(ident, opclass);
4580 parameters.addName(ident);
4582 // Debugging Stuff
4583 if (_AD._adl_debug > 1) fprintf(stderr, "\tOperand Name: %s\n", ident);
4584 skipws();
4585 } while(_curchar == ',');
4587 if (_curchar != ')') parse_err(SYNERR, "missing ')'\n");
4588 else {
4589 next_char(); // set current character position past the close paren
4590 }
4591 }
4594 //------------------------------get_effectlist---------------------------------
4595 // Looks for identifier pairs where first must be the name of a pre-defined,
4596 // effect, and the second must be the name of an operand defined in the
4597 // operand list of this instruction. Stores the names with a pointer to the
4598 // effect form in a local effects table.
4599 void ADLParser::get_effectlist(FormDict &effects, FormDict &operands, bool& has_call) {
4600 OperandForm *opForm;
4601 Effect *eForm;
4602 char *ident;
4604 do {
4605 next_char(); // skip open paren & comma characters
4606 skipws();
4607 if (_curchar == ')') break;
4609 // Get effect type, and check it against global name table
4610 ident = get_ident();
4611 if (ident == NULL) {
4612 parse_err(SYNERR, "effect type identifier expected at %c\n", _curchar);
4613 return;
4614 }
4615 else {
4616 // Check for valid effect type
4617 const Form *form = _globalNames[ident];
4618 if( form == NULL ) {
4619 parse_err(SYNERR, "undefined effect type %s\n", ident);
4620 return;
4621 }
4622 else {
4623 if( (eForm = form->is_effect()) == NULL) {
4624 parse_err(SYNERR, "identifier %s not effect type\n", ident);
4625 return;
4626 }
4627 }
4628 }
4629 // Debugging Stuff
4630 if (_AD._adl_debug > 1) fprintf(stderr, "\tEffect Type: %s\t", ident);
4631 skipws();
4632 if (eForm->is(Component::CALL)) {
4633 if (_AD._adl_debug > 1) fprintf(stderr, "\n");
4634 has_call = true;
4635 } else {
4636 // Get name of operand and check that it is in the local name table
4637 if( (ident = get_unique_ident(effects, "effect")) == NULL) {
4638 parse_err(SYNERR, "missing operand identifier in effect list\n");
4639 return;
4640 }
4641 const Form *form = operands[ident];
4642 opForm = form ? form->is_operand() : NULL;
4643 if( opForm == NULL ) {
4644 if( form && form->is_opclass() ) {
4645 const char* cname = form->is_opclass()->_ident;
4646 parse_err(SYNERR, "operand classes are illegal in effect lists (found %s %s)\n", cname, ident);
4647 } else {
4648 parse_err(SYNERR, "undefined operand %s in effect list\n", ident);
4649 }
4650 return;
4651 }
4652 // Add the pair to the effects table
4653 effects.Insert(ident, eForm);
4654 // Debugging Stuff
4655 if (_AD._adl_debug > 1) fprintf(stderr, "\tOperand Name: %s\n", ident);
4656 }
4657 skipws();
4658 } while(_curchar == ',');
4660 if (_curchar != ')') parse_err(SYNERR, "missing ')'\n");
4661 else {
4662 next_char(); // set current character position past the close paren
4663 }
4664 }
4667 //-------------------------------preproc_line----------------------------------
4668 // A "#line" keyword has been seen, so parse the rest of the line.
4669 void ADLParser::preproc_line(void) {
4670 int line = get_int();
4671 skipws_no_preproc();
4672 const char* file = NULL;
4673 if (_curchar == '"') {
4674 next_char(); // Move past the initial '"'
4675 file = _ptr;
4676 while (true) {
4677 if (_curchar == '\n') {
4678 parse_err(SYNERR, "missing '\"' at end of #line directive");
4679 return;
4680 }
4681 if (_curchar == '"') {
4682 *_ptr = '\0'; // Terminate the string
4683 next_char();
4684 skipws_no_preproc();
4685 break;
4686 }
4687 next_char();
4688 }
4689 }
4690 ensure_end_of_line();
4691 if (file != NULL)
4692 _AD._ADL_file._name = file;
4693 _buf.set_linenum(line);
4694 }
4696 //------------------------------preproc_define---------------------------------
4697 // A "#define" keyword has been seen, so parse the rest of the line.
4698 void ADLParser::preproc_define(void) {
4699 char* flag = get_ident_no_preproc();
4700 skipws_no_preproc();
4701 // only #define x y is supported for now
4702 char* def = get_ident_no_preproc();
4703 _AD.set_preproc_def(flag, def);
4704 skipws_no_preproc();
4705 if (_curchar != '\n') {
4706 parse_err(SYNERR, "non-identifier in preprocessor definition\n");
4707 }
4708 }
4710 //------------------------------preproc_undef----------------------------------
4711 // An "#undef" keyword has been seen, so parse the rest of the line.
4712 void ADLParser::preproc_undef(void) {
4713 char* flag = get_ident_no_preproc();
4714 skipws_no_preproc();
4715 ensure_end_of_line();
4716 _AD.set_preproc_def(flag, NULL);
4717 }
4721 //------------------------------parse_err--------------------------------------
4722 // Issue a parser error message, and skip to the end of the current line
4723 void ADLParser::parse_err(int flag, const char *fmt, ...) {
4724 va_list args;
4726 va_start(args, fmt);
4727 if (flag == 1)
4728 _AD._syntax_errs += _AD.emit_msg(0, flag, linenum(), fmt, args);
4729 else if (flag == 2)
4730 _AD._semantic_errs += _AD.emit_msg(0, flag, linenum(), fmt, args);
4731 else
4732 _AD._warnings += _AD.emit_msg(0, flag, linenum(), fmt, args);
4734 int error_char = _curchar;
4735 char* error_ptr = _ptr+1;
4736 for(;*_ptr != '\n'; _ptr++) ; // Skip to the end of the current line
4737 _curchar = '\n';
4738 va_end(args);
4739 _AD._no_output = 1;
4741 if (flag == 1) {
4742 char* error_tail = strchr(error_ptr, '\n');
4743 char tem = *error_ptr;
4744 error_ptr[-1] = '\0';
4745 char* error_head = error_ptr-1;
4746 while (error_head > _curline && *error_head) --error_head;
4747 if (error_tail) *error_tail = '\0';
4748 fprintf(stderr, "Error Context: %s>>>%c<<<%s\n",
4749 error_head, error_char, error_ptr);
4750 if (error_tail) *error_tail = '\n';
4751 error_ptr[-1] = tem;
4752 }
4753 }
4755 //---------------------------ensure_start_of_line------------------------------
4756 // A preprocessor directive has been encountered. Be sure it has fallen at
4757 // the beginning of a line, or else report an error.
4758 void ADLParser::ensure_start_of_line(void) {
4759 if (_curchar == '\n') { next_line(); return; }
4760 assert( _ptr >= _curline && _ptr < _curline+strlen(_curline),
4761 "Must be able to find which line we are in" );
4763 for (char *s = _curline; s < _ptr; s++) {
4764 if (*s > ' ') {
4765 parse_err(SYNERR, "'%c' must be at beginning of line\n", _curchar);
4766 break;
4767 }
4768 }
4769 }
4771 //---------------------------ensure_end_of_line--------------------------------
4772 // A preprocessor directive has been parsed. Be sure there is no trailing
4773 // garbage at the end of this line. Set the scan point to the beginning of
4774 // the next line.
4775 void ADLParser::ensure_end_of_line(void) {
4776 skipws_no_preproc();
4777 if (_curchar != '\n' && _curchar != '\0') {
4778 parse_err(SYNERR, "garbage char '%c' at end of line\n", _curchar);
4779 } else {
4780 next_char_or_line();
4781 }
4782 }
4784 //---------------------------handle_preproc------------------------------------
4785 // The '#' character introducing a preprocessor directive has been found.
4786 // Parse the whole directive name (e.g., #define, #endif) and take appropriate
4787 // action. If we are in an "untaken" span of text, simply keep track of
4788 // #ifdef nesting structure, so we can find out when to start taking text
4789 // again. (In this state, we "sort of support" C's #if directives, enough
4790 // to disregard their associated #else and #endif lines.) If we are in a
4791 // "taken" span of text, there are two cases: "#define" and "#undef"
4792 // directives are preserved and passed up to the caller, which eventually
4793 // passes control to the top-level parser loop, which handles #define and
4794 // #undef directly. (This prevents these directives from occurring in
4795 // arbitrary positions in the AD file--we require better structure than C.)
4796 // In the other case, and #ifdef, #ifndef, #else, or #endif is silently
4797 // processed as whitespace, with the "taken" state of the text correctly
4798 // updated. This routine returns "false" exactly in the case of a "taken"
4799 // #define or #undef, which tells the caller that a preprocessor token
4800 // has appeared which must be handled explicitly by the parse loop.
4801 bool ADLParser::handle_preproc_token() {
4802 assert(*_ptr == '#', "must be at start of preproc");
4803 ensure_start_of_line();
4804 next_char();
4805 skipws_no_preproc();
4806 char* start_ident = _ptr;
4807 char* ident = (_curchar == '\n') ? NULL : get_ident_no_preproc();
4808 if (ident == NULL) {
4809 parse_err(SYNERR, "expected preprocessor command, got end of line\n");
4810 } else if (!strcmp(ident, "ifdef") ||
4811 !strcmp(ident, "ifndef")) {
4812 char* flag = get_ident_no_preproc();
4813 ensure_end_of_line();
4814 // Test the identifier only if we are already in taken code:
4815 bool flag_def = preproc_taken() && (_AD.get_preproc_def(flag) != NULL);
4816 bool now_taken = !strcmp(ident, "ifdef") ? flag_def : !flag_def;
4817 begin_if_def(now_taken);
4818 } else if (!strcmp(ident, "if")) {
4819 if (preproc_taken())
4820 parse_err(SYNERR, "unimplemented: #%s %s", ident, _ptr+1);
4821 next_line();
4822 // Intelligently skip this nested C preprocessor directive:
4823 begin_if_def(true);
4824 } else if (!strcmp(ident, "else")) {
4825 ensure_end_of_line();
4826 invert_if_def();
4827 } else if (!strcmp(ident, "endif")) {
4828 ensure_end_of_line();
4829 end_if_def();
4830 } else if (preproc_taken()) {
4831 // pass this token up to the main parser as "#define" or "#undef"
4832 _ptr = start_ident;
4833 _curchar = *--_ptr;
4834 if( _curchar != '#' ) {
4835 parse_err(SYNERR, "no space allowed after # in #define or #undef");
4836 assert(_curchar == '#', "no space allowed after # in #define or #undef");
4837 }
4838 return false;
4839 }
4840 return true;
4841 }
4843 //---------------------------skipws_common-------------------------------------
4844 // Skip whitespace, including comments and newlines, while keeping an accurate
4845 // line count.
4846 // Maybe handle certain preprocessor constructs: #ifdef, #ifndef, #else, #endif
4847 void ADLParser::skipws_common(bool do_preproc) {
4848 char *start = _ptr;
4849 char *next = _ptr + 1;
4851 if (*_ptr == '\0') {
4852 // Check for string terminator
4853 if (_curchar > ' ') return;
4854 if (_curchar == '\n') {
4855 if (!do_preproc) return; // let caller handle the newline
4856 next_line();
4857 _ptr = _curline; next = _ptr + 1;
4858 }
4859 else if (_curchar == '#' ||
4860 (_curchar == '/' && (*next == '/' || *next == '*'))) {
4861 parse_err(SYNERR, "unimplemented: comment token in a funny place");
4862 }
4863 }
4864 while(_curline != NULL) { // Check for end of file
4865 if (*_ptr == '\n') { // keep proper track of new lines
4866 if (!do_preproc) break; // let caller handle the newline
4867 next_line();
4868 _ptr = _curline; next = _ptr + 1;
4869 }
4870 else if ((*_ptr == '/') && (*next == '/')) // C++ comment
4871 do { _ptr++; next++; } while(*_ptr != '\n'); // So go to end of line
4872 else if ((*_ptr == '/') && (*next == '*')) { // C comment
4873 _ptr++; next++;
4874 do {
4875 _ptr++; next++;
4876 if (*_ptr == '\n') { // keep proper track of new lines
4877 next_line(); // skip newlines within comments
4878 if (_curline == NULL) { // check for end of file
4879 parse_err(SYNERR, "end-of-file detected inside comment\n");
4880 break;
4881 }
4882 _ptr = _curline; next = _ptr + 1;
4883 }
4884 } while(!((*_ptr == '*') && (*next == '/'))); // Go to end of comment
4885 _ptr = ++next; next++; // increment _ptr past comment end
4886 }
4887 else if (do_preproc && *_ptr == '#') {
4888 // Note that this calls skipws_common(false) recursively!
4889 bool preproc_handled = handle_preproc_token();
4890 if (!preproc_handled) {
4891 if (preproc_taken()) {
4892 return; // short circuit
4893 }
4894 ++_ptr; // skip the preprocessor character
4895 }
4896 next = _ptr+1;
4897 } else if(*_ptr > ' ' && !(do_preproc && !preproc_taken())) {
4898 break;
4899 }
4900 else if (*_ptr == '"' || *_ptr == '\'') {
4901 assert(do_preproc, "only skip strings if doing preproc");
4902 // skip untaken quoted string
4903 int qchar = *_ptr;
4904 while (true) {
4905 ++_ptr;
4906 if (*_ptr == qchar) { ++_ptr; break; }
4907 if (*_ptr == '\\') ++_ptr;
4908 if (*_ptr == '\n' || *_ptr == '\0') {
4909 parse_err(SYNERR, "newline in string");
4910 break;
4911 }
4912 }
4913 next = _ptr + 1;
4914 }
4915 else { ++_ptr; ++next; }
4916 }
4917 if( _curline != NULL ) // at end of file _curchar isn't valid
4918 _curchar = *_ptr; // reset _curchar to maintain invariant
4919 }
4921 //---------------------------cur_char-----------------------------------------
4922 char ADLParser::cur_char() {
4923 return (_curchar);
4924 }
4926 //---------------------------next_char-----------------------------------------
4927 void ADLParser::next_char() {
4928 if (_curchar == '\n') parse_err(WARN, "must call next_line!");
4929 _curchar = *++_ptr;
4930 // if ( _curchar == '\n' ) {
4931 // next_line();
4932 // }
4933 }
4935 //---------------------------next_char_or_line---------------------------------
4936 void ADLParser::next_char_or_line() {
4937 if ( _curchar != '\n' ) {
4938 _curchar = *++_ptr;
4939 } else {
4940 next_line();
4941 _ptr = _curline;
4942 _curchar = *_ptr; // maintain invariant
4943 }
4944 }
4946 //---------------------------next_line-----------------------------------------
4947 void ADLParser::next_line() {
4948 _curline = _buf.get_line();
4949 _curchar = ' ';
4950 }
4952 //------------------------get_line_string--------------------------------------
4953 // Prepended location descriptor, for debugging.
4954 // Must return a malloced string (that can be freed if desired).
4955 char* ADLParser::get_line_string(int linenum) {
4956 const char* file = _AD._ADL_file._name;
4957 int line = linenum ? linenum : this->linenum();
4958 char* location = (char *)malloc(strlen(file) + 100);
4959 sprintf(location, "\n#line %d \"%s\"\n", line, file);
4960 return location;
4961 }
4963 //-------------------------is_literal_constant---------------------------------
4964 bool ADLParser::is_literal_constant(const char *param) {
4965 if (param[0] == 0) return false; // null string
4966 if (param[0] == '(') return true; // parenthesized expression
4967 if (param[0] == '0' && (param[1] == 'x' || param[1] == 'X')) {
4968 // Make sure it's a hex constant.
4969 int i = 2;
4970 do {
4971 if( !ADLParser::is_hex_digit(*(param+i)) ) return false;
4972 ++i;
4973 } while( *(param+i) != 0 );
4974 return true;
4975 }
4976 return false;
4977 }
4979 //---------------------------is_hex_digit--------------------------------------
4980 bool ADLParser::is_hex_digit(char digit) {
4981 return ((digit >= '0') && (digit <= '9'))
4982 ||((digit >= 'a') && (digit <= 'f'))
4983 ||((digit >= 'A') && (digit <= 'F'));
4984 }
4986 //---------------------------is_int_token--------------------------------------
4987 bool ADLParser::is_int_token(const char* token, int& intval) {
4988 const char* cp = token;
4989 while (*cp != '\0' && *cp <= ' ') cp++;
4990 if (*cp == '-') cp++;
4991 int ndigit = 0;
4992 while (*cp >= '0' && *cp <= '9') { cp++; ndigit++; }
4993 while (*cp != '\0' && *cp <= ' ') cp++;
4994 if (ndigit == 0 || *cp != '\0') {
4995 return false;
4996 }
4997 intval = atoi(token);
4998 return true;
4999 }
5001 static const char* skip_expr_ws(const char* str) {
5002 const char * cp = str;
5003 while (cp[0]) {
5004 if (cp[0] <= ' ') {
5005 ++cp;
5006 } else if (cp[0] == '#') {
5007 ++cp;
5008 while (cp[0] == ' ') ++cp;
5009 assert(0 == strncmp(cp, "line", 4), "must be a #line directive");
5010 const char* eol = strchr(cp, '\n');
5011 assert(eol != NULL, "must find end of line");
5012 if (eol == NULL) eol = cp + strlen(cp);
5013 cp = eol;
5014 } else {
5015 break;
5016 }
5017 }
5018 return cp;
5019 }
5021 //-----------------------equivalent_expressions--------------------------------
5022 bool ADLParser::equivalent_expressions(const char* str1, const char* str2) {
5023 if (str1 == str2)
5024 return true;
5025 else if (str1 == NULL || str2 == NULL)
5026 return false;
5027 const char* cp1 = str1;
5028 const char* cp2 = str2;
5029 char in_quote = '\0';
5030 while (cp1[0] && cp2[0]) {
5031 if (!in_quote) {
5032 // skip spaces and/or cpp directives
5033 const char* cp1a = skip_expr_ws(cp1);
5034 const char* cp2a = skip_expr_ws(cp2);
5035 if (cp1a > cp1 && cp2a > cp2) {
5036 cp1 = cp1a; cp2 = cp2a;
5037 continue;
5038 }
5039 if (cp1a > cp1 || cp2a > cp2) break; // fail
5040 }
5041 // match one non-space char
5042 if (cp1[0] != cp2[0]) break; // fail
5043 char ch = cp1[0];
5044 cp1++; cp2++;
5045 // watch for quotes
5046 if (in_quote && ch == '\\') {
5047 if (cp1[0] != cp2[0]) break; // fail
5048 if (!cp1[0]) break;
5049 cp1++; cp2++;
5050 }
5051 if (in_quote && ch == in_quote) {
5052 in_quote = '\0';
5053 } else if (!in_quote && (ch == '"' || ch == '\'')) {
5054 in_quote = ch;
5055 }
5056 }
5057 return (!cp1[0] && !cp2[0]);
5058 }
5061 //-------------------------------trim------------------------------------------
5062 void ADLParser::trim(char* &token) {
5063 while (*token <= ' ') token++;
5064 char* end = token + strlen(token);
5065 while (end > token && *(end-1) <= ' ') --end;
5066 *end = '\0';
5067 }