Fri, 15 Jun 2012 01:25:19 -0700
7119644: Increase superword's vector size up to 256 bits
Summary: Increase vector size up to 256-bits for YMM AVX registers on x86.
Reviewed-by: never, twisti, roland
1 /*
2 * Copyright (c) 1997, 2012, 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 }
118 // Add reg_class spill_regs after parsing.
119 RegisterForm *regBlock = _AD.get_registers();
120 if (regBlock == NULL) {
121 parse_err(SEMERR, "Did not declare 'register' definitions");
122 }
123 regBlock->addSpillRegClass();
125 // Done with parsing, check consistency.
127 if (_preproc_depth != 0) {
128 parse_err(SYNERR, "End of file inside #ifdef");
129 }
131 // AttributeForms ins_cost and op_cost must be defined for default behaviour
132 if (_globalNames[AttributeForm::_ins_cost] == NULL) {
133 parse_err(SEMERR, "Did not declare 'ins_cost' attribute");
134 }
135 if (_globalNames[AttributeForm::_op_cost] == NULL) {
136 parse_err(SEMERR, "Did not declare 'op_cost' attribute");
137 }
138 }
140 // ******************** Private Level 1 Parse Functions ********************
141 //------------------------------instr_parse------------------------------------
142 // Parse the contents of an instruction definition, build the InstructForm to
143 // represent that instruction, and add it to the InstructForm list.
144 void ADLParser::instr_parse(void) {
145 char *ident;
146 InstructForm *instr;
147 MatchRule *rule;
148 int match_rules_cnt = 0;
150 // First get the name of the instruction
151 if( (ident = get_unique_ident(_globalNames,"instruction")) == NULL )
152 return;
153 instr = new InstructForm(ident); // Create new instruction form
154 instr->_linenum = linenum();
155 _globalNames.Insert(ident, instr); // Add name to the name table
156 // Debugging Stuff
157 if (_AD._adl_debug > 1)
158 fprintf(stderr,"Parsing Instruction Form %s\n", ident);
160 // Then get the operands
161 skipws();
162 if (_curchar != '(') {
163 parse_err(SYNERR, "missing '(' in instruct definition\n");
164 }
165 // Parse the operand list
166 else get_oplist(instr->_parameters, instr->_localNames);
167 skipws(); // Skip leading whitespace
168 // Check for block delimiter
169 if ( (_curchar != '%')
170 || ( next_char(), (_curchar != '{')) ) {
171 parse_err(SYNERR, "missing '%{' in instruction definition\n");
172 return;
173 }
174 next_char(); // Maintain the invariant
175 do {
176 ident = get_ident(); // Grab next identifier
177 if (ident == NULL) {
178 parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
179 continue;
180 }
181 if (!strcmp(ident, "predicate")) instr->_predicate = pred_parse();
182 else if (!strcmp(ident, "match")) {
183 // Allow one instruction have several match rules.
184 rule = instr->_matrule;
185 if (rule == NULL) {
186 // This is first match rule encountered
187 rule = match_parse(instr->_localNames);
188 if (rule) {
189 instr->_matrule = rule;
190 // Special case the treatment of Control instructions.
191 if( instr->is_ideal_control() ) {
192 // Control instructions return a special result, 'Universe'
193 rule->_result = "Universe";
194 }
195 // Check for commutative operations with tree operands.
196 matchrule_clone_and_swap(rule, instr->_ident, match_rules_cnt);
197 }
198 } else {
199 // Find the end of the match rule list
200 while (rule->_next != NULL)
201 rule = rule->_next;
202 // Add the new match rule to the list
203 rule->_next = match_parse(instr->_localNames);
204 if (rule->_next) {
205 rule = rule->_next;
206 if( instr->is_ideal_control() ) {
207 parse_err(SYNERR, "unique match rule expected for %s\n", rule->_name);
208 return;
209 }
210 assert(match_rules_cnt < 100," too many match rule clones");
211 char* buf = (char*) malloc(strlen(instr->_ident) + 4);
212 sprintf(buf, "%s_%d", instr->_ident, match_rules_cnt++);
213 rule->_result = buf;
214 // Check for commutative operations with tree operands.
215 matchrule_clone_and_swap(rule, instr->_ident, match_rules_cnt);
216 }
217 }
218 }
219 else if (!strcmp(ident, "encode")) {
220 parse_err(SYNERR, "Instructions specify ins_encode, not encode\n");
221 }
222 else if (!strcmp(ident, "ins_encode")) ins_encode_parse(*instr);
223 else if (!strcmp(ident, "opcode")) instr->_opcode = opcode_parse(instr);
224 else if (!strcmp(ident, "size")) instr->_size = size_parse(instr);
225 else if (!strcmp(ident, "effect")) effect_parse(instr);
226 else if (!strcmp(ident, "expand")) instr->_exprule = expand_parse(instr);
227 else if (!strcmp(ident, "rewrite")) instr->_rewrule = rewrite_parse();
228 else if (!strcmp(ident, "constraint")) {
229 parse_err(SYNERR, "Instructions do not specify a constraint\n");
230 }
231 else if (!strcmp(ident, "construct")) {
232 parse_err(SYNERR, "Instructions do not specify a construct\n");
233 }
234 else if (!strcmp(ident, "format")) instr->_format = format_parse();
235 else if (!strcmp(ident, "interface")) {
236 parse_err(SYNERR, "Instructions do not specify an interface\n");
237 }
238 else if (!strcmp(ident, "ins_pipe")) ins_pipe_parse(*instr);
239 else { // Done with staticly defined parts of instruction definition
240 // Check identifier to see if it is the name of an attribute
241 const Form *form = _globalNames[ident];
242 AttributeForm *attr = form ? form->is_attribute() : NULL;
243 if( attr && (attr->_atype == INS_ATTR) ) {
244 // Insert the new attribute into the linked list.
245 Attribute *temp = attr_parse(ident);
246 temp->_next = instr->_attribs;
247 instr->_attribs = temp;
248 } else {
249 parse_err(SYNERR, "expected one of:\n predicate, match, encode, or the name of an instruction attribute at %s\n", ident);
250 }
251 }
252 skipws();
253 } while(_curchar != '%');
254 next_char();
255 if (_curchar != '}') {
256 parse_err(SYNERR, "missing '%}' in instruction definition\n");
257 return;
258 }
259 // Check for "Set" form of chain rule
260 adjust_set_rule(instr);
261 if (_AD._pipeline ) {
262 if( instr->expands() ) {
263 if( instr->_ins_pipe )
264 parse_err(WARN, "ins_pipe and expand rule both specified for instruction \"%s\"; ins_pipe will be unused\n", instr->_ident);
265 } else {
266 if( !instr->_ins_pipe )
267 parse_err(WARN, "No ins_pipe specified for instruction \"%s\"\n", instr->_ident);
268 }
269 }
270 // Add instruction to tail of instruction list
271 _AD.addForm(instr);
273 // Create instruction form for each additional match rule
274 rule = instr->_matrule;
275 if (rule != NULL) {
276 rule = rule->_next;
277 while (rule != NULL) {
278 ident = (char*)rule->_result;
279 InstructForm *clone = new InstructForm(ident, instr, rule); // Create new instruction form
280 _globalNames.Insert(ident, clone); // Add name to the name table
281 // Debugging Stuff
282 if (_AD._adl_debug > 1)
283 fprintf(stderr,"Parsing Instruction Form %s\n", ident);
284 // Check for "Set" form of chain rule
285 adjust_set_rule(clone);
286 // Add instruction to tail of instruction list
287 _AD.addForm(clone);
288 rule = rule->_next;
289 clone->_matrule->_next = NULL; // One match rule per clone
290 }
291 }
292 }
294 //------------------------------matchrule_clone_and_swap-----------------------
295 // Check for commutative operations with subtree operands,
296 // create clones and swap operands.
297 void ADLParser::matchrule_clone_and_swap(MatchRule* rule, const char* instr_ident, int& match_rules_cnt) {
298 // Check for commutative operations with tree operands.
299 int count = 0;
300 rule->count_commutative_op(count);
301 if (count > 0) {
302 // Clone match rule and swap commutative operation's operands.
303 rule->matchrule_swap_commutative_op(instr_ident, count, match_rules_cnt);
304 }
305 }
307 //------------------------------adjust_set_rule--------------------------------
308 // Check for "Set" form of chain rule
309 void ADLParser::adjust_set_rule(InstructForm *instr) {
310 if (instr->_matrule == NULL || instr->_matrule->_rChild == NULL) return;
311 const char *rch = instr->_matrule->_rChild->_opType;
312 const Form *frm = _globalNames[rch];
313 if( (! strcmp(instr->_matrule->_opType,"Set")) &&
314 frm && frm->is_operand() && (! frm->ideal_only()) ) {
315 // Previous implementation, which missed leaP*, but worked for loadCon*
316 unsigned position = 0;
317 const char *result = NULL;
318 const char *name = NULL;
319 const char *optype = NULL;
320 MatchNode *right = instr->_matrule->_rChild;
321 if (right->base_operand(position, _globalNames, result, name, optype)) {
322 position = 1;
323 const char *result2 = NULL;
324 const char *name2 = NULL;
325 const char *optype2 = NULL;
326 // Can not have additional base operands in right side of match!
327 if ( ! right->base_operand( position, _globalNames, result2, name2, optype2) ) {
328 if (instr->_predicate != NULL)
329 parse_err(SYNERR, "ADLC does not support instruction chain rules with predicates");
330 // Chain from input _ideal_operand_type_,
331 // Needed for shared roots of match-trees
332 ChainList *lst = (ChainList *)_AD._chainRules[optype];
333 if (lst == NULL) {
334 lst = new ChainList();
335 _AD._chainRules.Insert(optype, lst);
336 }
337 if (!lst->search(instr->_matrule->_lChild->_opType)) {
338 const char *cost = instr->cost();
339 if (cost == NULL) {
340 cost = ((AttributeForm*)_globalNames[AttributeForm::_ins_cost])->_attrdef;
341 }
342 // The ADLC does not support chaining from the ideal operand type
343 // of a predicated user-defined operand
344 if( frm->is_operand() == NULL || frm->is_operand()->_predicate == NULL ) {
345 lst->insert(instr->_matrule->_lChild->_opType,cost,instr->_ident);
346 }
347 }
348 // Chain from input _user_defined_operand_type_,
349 lst = (ChainList *)_AD._chainRules[result];
350 if (lst == NULL) {
351 lst = new ChainList();
352 _AD._chainRules.Insert(result, lst);
353 }
354 if (!lst->search(instr->_matrule->_lChild->_opType)) {
355 const char *cost = instr->cost();
356 if (cost == NULL) {
357 cost = ((AttributeForm*)_globalNames[AttributeForm::_ins_cost])->_attrdef;
358 }
359 // It is safe to chain from the top-level user-defined operand even
360 // if it has a predicate, since the predicate is checked before
361 // the user-defined type is available.
362 lst->insert(instr->_matrule->_lChild->_opType,cost,instr->_ident);
363 }
364 } else {
365 // May have instruction chain rule if root of right-tree is an ideal
366 OperandForm *rightOp = _globalNames[right->_opType]->is_operand();
367 if( rightOp ) {
368 const Form *rightRoot = _globalNames[rightOp->_matrule->_opType];
369 if( rightRoot && rightRoot->ideal_only() ) {
370 const char *chain_op = NULL;
371 if( rightRoot->is_instruction() )
372 chain_op = rightOp->_ident;
373 if( chain_op ) {
374 // Look-up the operation in chain rule table
375 ChainList *lst = (ChainList *)_AD._chainRules[chain_op];
376 if (lst == NULL) {
377 lst = new ChainList();
378 _AD._chainRules.Insert(chain_op, lst);
379 }
380 // if (!lst->search(instr->_matrule->_lChild->_opType)) {
381 const char *cost = instr->cost();
382 if (cost == NULL) {
383 cost = ((AttributeForm*)_globalNames[AttributeForm::_ins_cost])->_attrdef;
384 }
385 // This chains from a top-level operand whose predicate, if any,
386 // has been checked.
387 lst->insert(instr->_matrule->_lChild->_opType,cost,instr->_ident);
388 // }
389 }
390 }
391 }
392 } // end chain rule from right-tree's ideal root
393 }
394 }
395 }
398 //------------------------------oper_parse-------------------------------------
399 void ADLParser::oper_parse(void) {
400 char *ident;
401 OperandForm *oper;
402 AttributeForm *attr;
403 MatchRule *rule;
405 // First get the name of the operand
406 skipws();
407 if( (ident = get_unique_ident(_globalNames,"operand")) == NULL )
408 return;
409 oper = new OperandForm(ident); // Create new operand form
410 oper->_linenum = linenum();
411 _globalNames.Insert(ident, oper); // Add name to the name table
413 // Debugging Stuff
414 if (_AD._adl_debug > 1) fprintf(stderr,"Parsing Operand Form %s\n", ident);
416 // Get the component operands
417 skipws();
418 if (_curchar != '(') {
419 parse_err(SYNERR, "missing '(' in operand definition\n");
420 return;
421 }
422 else get_oplist(oper->_parameters, oper->_localNames); // Parse the component operand list
423 skipws();
424 // Check for block delimiter
425 if ((_curchar != '%') || (*(_ptr+1) != '{')) { // If not open block
426 parse_err(SYNERR, "missing '%c{' in operand definition\n","%");
427 return;
428 }
429 next_char(); next_char(); // Skip over "%{" symbol
430 do {
431 ident = get_ident(); // Grab next identifier
432 if (ident == NULL) {
433 parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
434 continue;
435 }
436 if (!strcmp(ident, "predicate")) oper->_predicate = pred_parse();
437 else if (!strcmp(ident, "match")) {
438 // Find the end of the match rule list
439 rule = oper->_matrule;
440 if (rule) {
441 while (rule->_next) rule = rule->_next;
442 // Add the new match rule to the list
443 rule->_next = match_parse(oper->_localNames);
444 if (rule->_next) {
445 rule->_next->_result = oper->_ident;
446 }
447 }
448 else {
449 // This is first match rule encountered
450 oper->_matrule = match_parse(oper->_localNames);
451 if (oper->_matrule) {
452 oper->_matrule->_result = oper->_ident;
453 }
454 }
455 }
456 else if (!strcmp(ident, "encode")) oper->_interface = interface_parse();
457 else if (!strcmp(ident, "ins_encode")) {
458 parse_err(SYNERR, "Operands specify 'encode', not 'ins_encode'\n");
459 }
460 else if (!strcmp(ident, "opcode")) {
461 parse_err(SYNERR, "Operands do not specify an opcode\n");
462 }
463 else if (!strcmp(ident, "effect")) {
464 parse_err(SYNERR, "Operands do not specify an effect\n");
465 }
466 else if (!strcmp(ident, "expand")) {
467 parse_err(SYNERR, "Operands do not specify an expand\n");
468 }
469 else if (!strcmp(ident, "rewrite")) {
470 parse_err(SYNERR, "Operands do not specify a rewrite\n");
471 }
472 else if (!strcmp(ident, "constraint"))oper->_constraint= constraint_parse();
473 else if (!strcmp(ident, "construct")) oper->_construct = construct_parse();
474 else if (!strcmp(ident, "format")) oper->_format = format_parse();
475 else if (!strcmp(ident, "interface")) oper->_interface = interface_parse();
476 // Check identifier to see if it is the name of an attribute
477 else if (((attr = _globalNames[ident]->is_attribute()) != NULL) &&
478 (attr->_atype == OP_ATTR)) oper->_attribs = attr_parse(ident);
479 else {
480 parse_err(SYNERR, "expected one of - constraint, predicate, match, encode, format, construct, or the name of a defined operand attribute at %s\n", ident);
481 }
482 skipws();
483 } while(_curchar != '%');
484 next_char();
485 if (_curchar != '}') {
486 parse_err(SYNERR, "missing '%}' in operand definition\n");
487 return;
488 }
489 // Add operand to tail of operand list
490 _AD.addForm(oper);
491 }
493 //------------------------------opclass_parse----------------------------------
494 // Operand Classes are a block with a comma delimited list of operand names
495 void ADLParser::opclass_parse(void) {
496 char *ident;
497 OpClassForm *opc;
498 OperandForm *opForm;
500 // First get the name of the operand class
501 skipws();
502 if( (ident = get_unique_ident(_globalNames,"opclass")) == NULL )
503 return;
504 opc = new OpClassForm(ident); // Create new operand class form
505 _globalNames.Insert(ident, opc); // Add name to the name table
507 // Debugging Stuff
508 if (_AD._adl_debug > 1)
509 fprintf(stderr,"Parsing Operand Class Form %s\n", ident);
511 // Get the list of operands
512 skipws();
513 if (_curchar != '(') {
514 parse_err(SYNERR, "missing '(' in operand definition\n");
515 return;
516 }
517 do {
518 next_char(); // Skip past open paren or comma
519 ident = get_ident(); // Grab next identifier
520 if (ident == NULL) {
521 parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
522 continue;
523 }
524 // Check identifier to see if it is the name of an operand
525 const Form *form = _globalNames[ident];
526 opForm = form ? form->is_operand() : NULL;
527 if ( opForm ) {
528 opc->_oplst.addName(ident); // Add operand to opclass list
529 opForm->_classes.addName(opc->_ident);// Add opclass to operand list
530 }
531 else {
532 parse_err(SYNERR, "expected name of a defined operand at %s\n", ident);
533 }
534 skipws(); // skip trailing whitespace
535 } while (_curchar == ','); // Check for the comma
536 // Check for closing ')'
537 if (_curchar != ')') {
538 parse_err(SYNERR, "missing ')' or ',' in opclass definition\n");
539 return;
540 }
541 next_char(); // Consume the ')'
542 skipws();
543 // Check for closing ';'
544 if (_curchar != ';') {
545 parse_err(SYNERR, "missing ';' in opclass definition\n");
546 return;
547 }
548 next_char(); // Consume the ';'
549 // Add operand to tail of operand list
550 _AD.addForm(opc);
551 }
553 //------------------------------ins_attr_parse---------------------------------
554 void ADLParser::ins_attr_parse(void) {
555 char *ident;
556 char *aexpr;
557 AttributeForm *attrib;
559 // get name for the instruction attribute
560 skipws(); // Skip leading whitespace
561 if( (ident = get_unique_ident(_globalNames,"inst_attrib")) == NULL )
562 return;
563 // Debugging Stuff
564 if (_AD._adl_debug > 1) fprintf(stderr,"Parsing Ins_Attribute Form %s\n", ident);
566 // Get default value of the instruction attribute
567 skipws(); // Skip whitespace
568 if ((aexpr = get_paren_expr("attribute default expression string")) == NULL) {
569 parse_err(SYNERR, "missing '(' in ins_attrib definition\n");
570 return;
571 }
572 // Debug Stuff
573 if (_AD._adl_debug > 1) fprintf(stderr,"Attribute Expression: %s\n", aexpr);
575 // Check for terminator
576 if (_curchar != ';') {
577 parse_err(SYNERR, "missing ';' in ins_attrib definition\n");
578 return;
579 }
580 next_char(); // Advance past the ';'
582 // Construct the attribute, record global name, and store in ArchDesc
583 attrib = new AttributeForm(ident, INS_ATTR, aexpr);
584 _globalNames.Insert(ident, attrib); // Add name to the name table
585 _AD.addForm(attrib);
586 }
588 //------------------------------op_attr_parse----------------------------------
589 void ADLParser::op_attr_parse(void) {
590 char *ident;
591 char *aexpr;
592 AttributeForm *attrib;
594 // get name for the operand attribute
595 skipws(); // Skip leading whitespace
596 if( (ident = get_unique_ident(_globalNames,"op_attrib")) == NULL )
597 return;
598 // Debugging Stuff
599 if (_AD._adl_debug > 1) fprintf(stderr,"Parsing Op_Attribute Form %s\n", ident);
601 // Get default value of the instruction attribute
602 skipws(); // Skip whitespace
603 if ((aexpr = get_paren_expr("attribute default expression string")) == NULL) {
604 parse_err(SYNERR, "missing '(' in op_attrib definition\n");
605 return;
606 }
607 // Debug Stuff
608 if (_AD._adl_debug > 1) fprintf(stderr,"Attribute Expression: %s\n", aexpr);
610 // Check for terminator
611 if (_curchar != ';') {
612 parse_err(SYNERR, "missing ';' in op_attrib definition\n");
613 return;
614 }
615 next_char(); // Advance past the ';'
617 // Construct the attribute, record global name, and store in ArchDesc
618 attrib = new AttributeForm(ident, OP_ATTR, aexpr);
619 _globalNames.Insert(ident, attrib);
620 _AD.addForm(attrib);
621 }
623 //------------------------------definitions_parse-----------------------------------
624 void ADLParser::definitions_parse(void) {
625 skipws(); // Skip leading whitespace
626 if (_curchar == '%' && *(_ptr+1) == '{') {
627 next_char(); next_char(); // Skip "%{"
628 skipws();
629 while (_curchar != '%' && *(_ptr+1) != '}') {
630 // Process each definition until finding closing string "%}"
631 char *token = get_ident();
632 if (token == NULL) {
633 parse_err(SYNERR, "missing identifier inside definitions block.\n");
634 return;
635 }
636 if (strcmp(token,"int_def")==0) { int_def_parse(); }
637 // if (strcmp(token,"str_def")==0) { str_def_parse(); }
638 skipws();
639 }
640 }
641 else {
642 parse_err(SYNERR, "Missing %%{ ... %%} block after definitions keyword.\n");
643 return;
644 }
645 }
647 //------------------------------int_def_parse----------------------------------
648 // Parse Example:
649 // int_def MEMORY_REF_COST ( 200, DEFAULT_COST * 2);
650 // <keyword> <name> ( <int_value>, <description> );
651 //
652 void ADLParser::int_def_parse(void) {
653 char *name = NULL; // Name of definition
654 char *value = NULL; // its value,
655 int int_value = -1; // positive values only
656 char *description = NULL; // textual description
658 // Get definition name
659 skipws(); // Skip whitespace
660 name = get_ident();
661 if (name == NULL) {
662 parse_err(SYNERR, "missing definition name after int_def\n");
663 return;
664 }
666 // Check for value of int_def dname( integer_value [, string_expression ] )
667 skipws();
668 if (_curchar == '(') {
670 // Parse the integer value.
671 next_char();
672 value = get_ident();
673 if (value == NULL) {
674 parse_err(SYNERR, "missing value in int_def\n");
675 return;
676 }
677 if( !is_int_token(value, int_value) ) {
678 parse_err(SYNERR, "value in int_def is not recognized as integer\n");
679 return;
680 }
681 skipws();
683 // Check for description
684 if (_curchar == ',') {
685 next_char(); // skip ','
687 description = get_expr("int_def description", ")");
688 if (description == NULL) {
689 parse_err(SYNERR, "invalid or missing description in int_def\n");
690 return;
691 }
692 trim(description);
693 }
695 if (_curchar != ')') {
696 parse_err(SYNERR, "missing ')' in register definition statement\n");
697 return;
698 }
699 next_char();
700 }
702 // Check for closing ';'
703 skipws();
704 if (_curchar != ';') {
705 parse_err(SYNERR, "missing ';' after int_def\n");
706 return;
707 }
708 next_char(); // move past ';'
710 // Debug Stuff
711 if (_AD._adl_debug > 1) {
712 fprintf(stderr,"int_def: %s ( %s, %s )\n", name,
713 (value), (description ? description : ""));
714 }
716 // Record new definition.
717 Expr *expr = new Expr(name, description, int_value, int_value);
718 const Expr *old_expr = _AD.globalDefs().define(name, expr);
719 if (old_expr != NULL) {
720 parse_err(SYNERR, "Duplicate definition\n");
721 return;
722 }
724 return;
725 }
728 //------------------------------source_parse-----------------------------------
729 void ADLParser::source_parse(void) {
730 SourceForm *source; // Encode class for instruction/operand
731 char *rule = NULL; // String representation of encode rule
733 skipws(); // Skip leading whitespace
734 if ( (rule = find_cpp_block("source block")) == NULL ) {
735 parse_err(SYNERR, "incorrect or missing block for 'source'.\n");
736 return;
737 }
738 // Debug Stuff
739 if (_AD._adl_debug > 1) fprintf(stderr,"Source Form: %s\n", rule);
741 source = new SourceForm(rule); // Build new Source object
742 _AD.addForm(source);
743 // skipws();
744 }
746 //------------------------------source_hpp_parse-------------------------------
747 // Parse a source_hpp %{ ... %} block.
748 // The code gets stuck into the ad_<arch>.hpp file.
749 // If the source_hpp block appears before the register block in the AD
750 // file, it goes up at the very top of the ad_<arch>.hpp file, so that
751 // it can be used by register encodings, etc. Otherwise, it goes towards
752 // the bottom, where it's useful as a global definition to *.cpp files.
753 void ADLParser::source_hpp_parse(void) {
754 char *rule = NULL; // String representation of encode rule
756 skipws(); // Skip leading whitespace
757 if ( (rule = find_cpp_block("source_hpp block")) == NULL ) {
758 parse_err(SYNERR, "incorrect or missing block for 'source_hpp'.\n");
759 return;
760 }
761 // Debug Stuff
762 if (_AD._adl_debug > 1) fprintf(stderr,"Header Form: %s\n", rule);
764 if (_AD.get_registers() == NULL) {
765 // Very early in the file, before reg_defs, we collect pre-headers.
766 PreHeaderForm* pre_header = new PreHeaderForm(rule);
767 _AD.addForm(pre_header);
768 } else {
769 // Normally, we collect header info, placed at the bottom of the hpp file.
770 HeaderForm* header = new HeaderForm(rule);
771 _AD.addForm(header);
772 }
773 }
775 //------------------------------reg_parse--------------------------------------
776 void ADLParser::reg_parse(void) {
777 RegisterForm *regBlock = _AD.get_registers(); // Information about registers encoding
778 if (regBlock == NULL) {
779 // Create the RegisterForm for the architecture description.
780 regBlock = new RegisterForm(); // Build new Source object
781 _AD.addForm(regBlock);
782 }
784 skipws(); // Skip leading whitespace
785 if (_curchar == '%' && *(_ptr+1) == '{') {
786 next_char(); next_char(); // Skip "%{"
787 skipws();
788 while (_curchar != '%' && *(_ptr+1) != '}') {
789 char *token = get_ident();
790 if (token == NULL) {
791 parse_err(SYNERR, "missing identifier inside register block.\n");
792 return;
793 }
794 if (strcmp(token,"reg_def")==0) { reg_def_parse(); }
795 else if (strcmp(token,"reg_class")==0) { reg_class_parse(); }
796 else if (strcmp(token,"alloc_class")==0) { alloc_class_parse(); }
797 else if (strcmp(token,"#define")==0) { preproc_define(); }
798 else { parse_err(SYNERR, "bad token %s inside register block.\n", token); break; }
799 skipws();
800 }
801 }
802 else {
803 parse_err(SYNERR, "Missing %c{ ... %c} block after register keyword.\n",'%','%');
804 return;
805 }
806 }
808 //------------------------------encode_parse-----------------------------------
809 void ADLParser::encode_parse(void) {
810 EncodeForm *encBlock; // Information about instruction/operand encoding
812 _AD.getForm(&encBlock);
813 if ( encBlock == NULL) {
814 // Create the EncodeForm for the architecture description.
815 encBlock = new EncodeForm(); // Build new Source object
816 _AD.addForm(encBlock);
817 }
819 skipws(); // Skip leading whitespace
820 if (_curchar == '%' && *(_ptr+1) == '{') {
821 next_char(); next_char(); // Skip "%{"
822 skipws();
823 while (_curchar != '%' && *(_ptr+1) != '}') {
824 char *token = get_ident();
825 if (token == NULL) {
826 parse_err(SYNERR, "missing identifier inside encoding block.\n");
827 return;
828 }
829 if (strcmp(token,"enc_class")==0) { enc_class_parse(); }
830 skipws();
831 }
832 }
833 else {
834 parse_err(SYNERR, "Missing %c{ ... %c} block after encode keyword.\n",'%','%');
835 return;
836 }
837 }
839 //------------------------------enc_class_parse--------------------------------
840 void ADLParser::enc_class_parse(void) {
841 char *ec_name; // Name of encoding class being defined
843 // Get encoding class name
844 skipws(); // Skip whitespace
845 ec_name = get_ident();
846 if (ec_name == NULL) {
847 parse_err(SYNERR, "missing encoding class name after encode.\n");
848 return;
849 }
851 EncClass *encoding = _AD._encode->add_EncClass(ec_name);
852 encoding->_linenum = linenum();
854 skipws(); // Skip leading whitespace
855 // Check for optional parameter list
856 if (_curchar == '(') {
857 do {
858 char *pType = NULL; // parameter type
859 char *pName = NULL; // parameter name
861 next_char(); // skip open paren & comma characters
862 skipws();
863 if (_curchar == ')') break;
865 // Get parameter type
866 pType = get_ident();
867 if (pType == NULL) {
868 parse_err(SYNERR, "parameter type expected at %c\n", _curchar);
869 return;
870 }
872 skipws();
873 // Get parameter name
874 pName = get_ident();
875 if (pName == NULL) {
876 parse_err(SYNERR, "parameter name expected at %c\n", _curchar);
877 return;
878 }
880 // Record parameter type and name
881 encoding->add_parameter( pType, pName );
883 skipws();
884 } while(_curchar == ',');
886 if (_curchar != ')') parse_err(SYNERR, "missing ')'\n");
887 else {
888 next_char(); // Skip ')'
889 }
890 } // Done with parameter list
892 skipws();
893 // Check for block starting delimiters
894 if ((_curchar != '%') || (*(_ptr+1) != '{')) { // If not open block
895 parse_err(SYNERR, "missing '%c{' in enc_class definition\n", '%');
896 return;
897 }
898 next_char(); // Skip '%'
899 next_char(); // Skip '{'
901 enc_class_parse_block(encoding, ec_name);
902 }
905 void ADLParser::enc_class_parse_block(EncClass* encoding, char* ec_name) {
906 skipws_no_preproc(); // Skip leading whitespace
907 // Prepend location descriptor, for debugging; cf. ADLParser::find_cpp_block
908 if (_AD._adlocation_debug) {
909 encoding->add_code(get_line_string());
910 }
912 // Collect the parts of the encode description
913 // (1) strings that are passed through to output
914 // (2) replacement/substitution variable, preceeded by a '$'
915 while ( (_curchar != '%') && (*(_ptr+1) != '}') ) {
917 // (1)
918 // Check if there is a string to pass through to output
919 char *start = _ptr; // Record start of the next string
920 while ((_curchar != '$') && ((_curchar != '%') || (*(_ptr+1) != '}')) ) {
921 // If at the start of a comment, skip past it
922 if( (_curchar == '/') && ((*(_ptr+1) == '/') || (*(_ptr+1) == '*')) ) {
923 skipws_no_preproc();
924 } else {
925 // ELSE advance to the next character, or start of the next line
926 next_char_or_line();
927 }
928 }
929 // If a string was found, terminate it and record in EncClass
930 if ( start != _ptr ) {
931 *_ptr = '\0'; // Terminate the string
932 encoding->add_code(start);
933 }
935 // (2)
936 // If we are at a replacement variable,
937 // copy it and record in EncClass
938 if (_curchar == '$') {
939 // Found replacement Variable
940 char* rep_var = get_rep_var_ident_dup();
941 // Add flag to _strings list indicating we should check _rep_vars
942 encoding->add_rep_var(rep_var);
943 }
944 } // end while part of format description
945 next_char(); // Skip '%'
946 next_char(); // Skip '}'
948 skipws();
950 if (_AD._adlocation_debug) {
951 encoding->add_code(end_line_marker());
952 }
954 // Debug Stuff
955 if (_AD._adl_debug > 1) fprintf(stderr,"EncodingClass Form: %s\n", ec_name);
956 }
958 //------------------------------frame_parse-----------------------------------
959 void ADLParser::frame_parse(void) {
960 FrameForm *frame; // Information about stack-frame layout
961 char *desc = NULL; // String representation of frame
963 skipws(); // Skip leading whitespace
965 frame = new FrameForm(); // Build new Frame object
966 // Check for open block sequence
967 skipws(); // Skip leading whitespace
968 if (_curchar == '%' && *(_ptr+1) == '{') {
969 next_char(); next_char(); // Skip "%{"
970 skipws();
971 while (_curchar != '%' && *(_ptr+1) != '}') {
972 char *token = get_ident();
973 if (token == NULL) {
974 parse_err(SYNERR, "missing identifier inside frame block.\n");
975 return;
976 }
977 if (strcmp(token,"stack_direction")==0) {
978 stack_dir_parse(frame);
979 }
980 if (strcmp(token,"sync_stack_slots")==0) {
981 sync_stack_slots_parse(frame);
982 }
983 if (strcmp(token,"frame_pointer")==0) {
984 frame_pointer_parse(frame, false);
985 }
986 if (strcmp(token,"interpreter_frame_pointer")==0) {
987 interpreter_frame_pointer_parse(frame, false);
988 }
989 if (strcmp(token,"inline_cache_reg")==0) {
990 inline_cache_parse(frame, false);
991 }
992 if (strcmp(token,"compiler_method_oop_reg")==0) {
993 parse_err(WARN, "Using obsolete Token, compiler_method_oop_reg");
994 skipws();
995 }
996 if (strcmp(token,"interpreter_method_oop_reg")==0) {
997 interpreter_method_oop_parse(frame, false);
998 }
999 if (strcmp(token,"cisc_spilling_operand_name")==0) {
1000 cisc_spilling_operand_name_parse(frame, false);
1001 }
1002 if (strcmp(token,"stack_alignment")==0) {
1003 stack_alignment_parse(frame);
1004 }
1005 if (strcmp(token,"return_addr")==0) {
1006 return_addr_parse(frame, false);
1007 }
1008 if (strcmp(token,"in_preserve_stack_slots")==0) {
1009 preserve_stack_parse(frame);
1010 }
1011 if (strcmp(token,"out_preserve_stack_slots")==0) {
1012 parse_err(WARN, "Using obsolete token, out_preserve_stack_slots");
1013 skipws();
1014 }
1015 if (strcmp(token,"varargs_C_out_slots_killed")==0) {
1016 frame->_varargs_C_out_slots_killed = parse_one_arg("varargs C out slots killed");
1017 }
1018 if (strcmp(token,"calling_convention")==0) {
1019 frame->_calling_convention = calling_convention_parse();
1020 }
1021 if (strcmp(token,"return_value")==0) {
1022 frame->_return_value = return_value_parse();
1023 }
1024 if (strcmp(token,"c_frame_pointer")==0) {
1025 frame_pointer_parse(frame, true);
1026 }
1027 if (strcmp(token,"c_return_addr")==0) {
1028 return_addr_parse(frame, true);
1029 }
1030 if (strcmp(token,"c_calling_convention")==0) {
1031 frame->_c_calling_convention = calling_convention_parse();
1032 }
1033 if (strcmp(token,"c_return_value")==0) {
1034 frame->_c_return_value = return_value_parse();
1035 }
1037 skipws();
1038 }
1039 }
1040 else {
1041 parse_err(SYNERR, "Missing %c{ ... %c} block after encode keyword.\n",'%','%');
1042 return;
1043 }
1044 // All Java versions are required, native versions are optional
1045 if(frame->_frame_pointer == NULL) {
1046 parse_err(SYNERR, "missing frame pointer definition in frame section.\n");
1047 return;
1048 }
1049 // !!!!! !!!!!
1050 // if(frame->_interpreter_frame_ptr_reg == NULL) {
1051 // parse_err(SYNERR, "missing interpreter frame pointer definition in frame section.\n");
1052 // return;
1053 // }
1054 if(frame->_alignment == NULL) {
1055 parse_err(SYNERR, "missing alignment definition in frame section.\n");
1056 return;
1057 }
1058 if(frame->_return_addr == NULL) {
1059 parse_err(SYNERR, "missing return address location in frame section.\n");
1060 return;
1061 }
1062 if(frame->_in_preserve_slots == NULL) {
1063 parse_err(SYNERR, "missing stack slot preservation definition in frame section.\n");
1064 return;
1065 }
1066 if(frame->_varargs_C_out_slots_killed == NULL) {
1067 parse_err(SYNERR, "missing varargs C out slots killed definition in frame section.\n");
1068 return;
1069 }
1070 if(frame->_calling_convention == NULL) {
1071 parse_err(SYNERR, "missing calling convention definition in frame section.\n");
1072 return;
1073 }
1074 if(frame->_return_value == NULL) {
1075 parse_err(SYNERR, "missing return value definition in frame section.\n");
1076 return;
1077 }
1078 // Fill natives in identically with the Java versions if not present.
1079 if(frame->_c_frame_pointer == NULL) {
1080 frame->_c_frame_pointer = frame->_frame_pointer;
1081 }
1082 if(frame->_c_return_addr == NULL) {
1083 frame->_c_return_addr = frame->_return_addr;
1084 frame->_c_return_addr_loc = frame->_return_addr_loc;
1085 }
1086 if(frame->_c_calling_convention == NULL) {
1087 frame->_c_calling_convention = frame->_calling_convention;
1088 }
1089 if(frame->_c_return_value == NULL) {
1090 frame->_c_return_value = frame->_return_value;
1091 }
1093 // Debug Stuff
1094 if (_AD._adl_debug > 1) fprintf(stderr,"Frame Form: %s\n", desc);
1096 // Create the EncodeForm for the architecture description.
1097 _AD.addForm(frame);
1098 // skipws();
1099 }
1101 //------------------------------stack_dir_parse--------------------------------
1102 void ADLParser::stack_dir_parse(FrameForm *frame) {
1103 char *direction = parse_one_arg("stack direction entry");
1104 if (strcmp(direction, "TOWARDS_LOW") == 0) {
1105 frame->_direction = false;
1106 }
1107 else if (strcmp(direction, "TOWARDS_HIGH") == 0) {
1108 frame->_direction = true;
1109 }
1110 else {
1111 parse_err(SYNERR, "invalid value inside stack direction entry.\n");
1112 return;
1113 }
1114 }
1116 //------------------------------sync_stack_slots_parse-------------------------
1117 void ADLParser::sync_stack_slots_parse(FrameForm *frame) {
1118 // Assign value into frame form
1119 frame->_sync_stack_slots = parse_one_arg("sync stack slots entry");
1120 }
1122 //------------------------------frame_pointer_parse----------------------------
1123 void ADLParser::frame_pointer_parse(FrameForm *frame, bool native) {
1124 char *frame_pointer = parse_one_arg("frame pointer entry");
1125 // Assign value into frame form
1126 if (native) { frame->_c_frame_pointer = frame_pointer; }
1127 else { frame->_frame_pointer = frame_pointer; }
1128 }
1130 //------------------------------interpreter_frame_pointer_parse----------------------------
1131 void ADLParser::interpreter_frame_pointer_parse(FrameForm *frame, bool native) {
1132 frame->_interpreter_frame_pointer_reg = parse_one_arg("interpreter frame pointer entry");
1133 }
1135 //------------------------------inline_cache_parse-----------------------------
1136 void ADLParser::inline_cache_parse(FrameForm *frame, bool native) {
1137 frame->_inline_cache_reg = parse_one_arg("inline cache reg entry");
1138 }
1140 //------------------------------interpreter_method_oop_parse------------------
1141 void ADLParser::interpreter_method_oop_parse(FrameForm *frame, bool native) {
1142 frame->_interpreter_method_oop_reg = parse_one_arg("method oop reg entry");
1143 }
1145 //------------------------------cisc_spilling_operand_parse---------------------
1146 void ADLParser::cisc_spilling_operand_name_parse(FrameForm *frame, bool native) {
1147 frame->_cisc_spilling_operand_name = parse_one_arg("cisc spilling operand name");
1148 }
1150 //------------------------------stack_alignment_parse--------------------------
1151 void ADLParser::stack_alignment_parse(FrameForm *frame) {
1152 char *alignment = parse_one_arg("stack alignment entry");
1153 // Assign value into frame
1154 frame->_alignment = alignment;
1155 }
1157 //------------------------------parse_one_arg-------------------------------
1158 char *ADLParser::parse_one_arg(const char *description) {
1159 char *token = NULL;
1160 if(_curchar == '(') {
1161 next_char();
1162 skipws();
1163 token = get_expr(description, ")");
1164 if (token == NULL) {
1165 parse_err(SYNERR, "missing value inside %s.\n", description);
1166 return NULL;
1167 }
1168 next_char(); // skip the close paren
1169 if(_curchar != ';') { // check for semi-colon
1170 parse_err(SYNERR, "missing %c in.\n", ';', description);
1171 return NULL;
1172 }
1173 next_char(); // skip the semi-colon
1174 }
1175 else {
1176 parse_err(SYNERR, "Missing %c in.\n", '(', description);
1177 return NULL;
1178 }
1180 trim(token);
1181 return token;
1182 }
1184 //------------------------------return_addr_parse------------------------------
1185 void ADLParser::return_addr_parse(FrameForm *frame, bool native) {
1186 bool in_register = true;
1187 if(_curchar == '(') {
1188 next_char();
1189 skipws();
1190 char *token = get_ident();
1191 if (token == NULL) {
1192 parse_err(SYNERR, "missing value inside return address entry.\n");
1193 return;
1194 }
1195 // check for valid values for stack/register
1196 if (strcmp(token, "REG") == 0) {
1197 in_register = true;
1198 }
1199 else if (strcmp(token, "STACK") == 0) {
1200 in_register = false;
1201 }
1202 else {
1203 parse_err(SYNERR, "invalid value inside return_address entry.\n");
1204 return;
1205 }
1206 if (native) { frame->_c_return_addr_loc = in_register; }
1207 else { frame->_return_addr_loc = in_register; }
1209 // Parse expression that specifies register or stack position
1210 skipws();
1211 char *token2 = get_expr("return address entry", ")");
1212 if (token2 == NULL) {
1213 parse_err(SYNERR, "missing value inside return address entry.\n");
1214 return;
1215 }
1216 next_char(); // skip the close paren
1217 if (native) { frame->_c_return_addr = token2; }
1218 else { frame->_return_addr = token2; }
1220 if(_curchar != ';') { // check for semi-colon
1221 parse_err(SYNERR, "missing %c in return address entry.\n", ';');
1222 return;
1223 }
1224 next_char(); // skip the semi-colon
1225 }
1226 else {
1227 parse_err(SYNERR, "Missing %c in return_address entry.\n", '(');
1228 }
1229 }
1231 //------------------------------preserve_stack_parse---------------------------
1232 void ADLParser::preserve_stack_parse(FrameForm *frame) {
1233 if(_curchar == '(') {
1234 char *token = get_paren_expr("preserve_stack_slots");
1235 frame->_in_preserve_slots = token;
1237 if(_curchar != ';') { // check for semi-colon
1238 parse_err(SYNERR, "missing %c in preserve stack slot entry.\n", ';');
1239 return;
1240 }
1241 next_char(); // skip the semi-colon
1242 }
1243 else {
1244 parse_err(SYNERR, "Missing %c in preserve stack slot entry.\n", '(');
1245 }
1246 }
1248 //------------------------------calling_convention_parse-----------------------
1249 char *ADLParser::calling_convention_parse() {
1250 char *desc = NULL; // String representation of calling_convention
1252 skipws(); // Skip leading whitespace
1253 if ( (desc = find_cpp_block("calling convention block")) == NULL ) {
1254 parse_err(SYNERR, "incorrect or missing block for 'calling_convention'.\n");
1255 }
1256 return desc;
1257 }
1259 //------------------------------return_value_parse-----------------------------
1260 char *ADLParser::return_value_parse() {
1261 char *desc = NULL; // String representation of calling_convention
1263 skipws(); // Skip leading whitespace
1264 if ( (desc = find_cpp_block("return value block")) == NULL ) {
1265 parse_err(SYNERR, "incorrect or missing block for 'return_value'.\n");
1266 }
1267 return desc;
1268 }
1270 //------------------------------ins_pipe_parse---------------------------------
1271 void ADLParser::ins_pipe_parse(InstructForm &instr) {
1272 char * ident;
1274 skipws();
1275 if ( _curchar != '(' ) { // Check for delimiter
1276 parse_err(SYNERR, "missing \"(\" in ins_pipe definition\n");
1277 return;
1278 }
1280 next_char();
1281 ident = get_ident(); // Grab next identifier
1283 if (ident == NULL) {
1284 parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
1285 return;
1286 }
1288 skipws();
1289 if ( _curchar != ')' ) { // Check for delimiter
1290 parse_err(SYNERR, "missing \")\" in ins_pipe definition\n");
1291 return;
1292 }
1294 next_char(); // skip the close paren
1295 if(_curchar != ';') { // check for semi-colon
1296 parse_err(SYNERR, "missing %c in return value entry.\n", ';');
1297 return;
1298 }
1299 next_char(); // skip the semi-colon
1301 // Check ident for validity
1302 if (_AD._pipeline && !_AD._pipeline->_classlist.search(ident)) {
1303 parse_err(SYNERR, "\"%s\" is not a valid pipeline class\n", ident);
1304 return;
1305 }
1307 // Add this instruction to the list in the pipeline class
1308 _AD._pipeline->_classdict[ident]->is_pipeclass()->_instructs.addName(instr._ident);
1310 // Set the name of the pipeline class in the instruction
1311 instr._ins_pipe = ident;
1312 return;
1313 }
1315 //------------------------------pipe_parse-------------------------------------
1316 void ADLParser::pipe_parse(void) {
1317 PipelineForm *pipeline; // Encode class for instruction/operand
1318 char * ident;
1320 pipeline = new PipelineForm(); // Build new Source object
1321 _AD.addForm(pipeline);
1323 skipws(); // Skip leading whitespace
1324 // Check for block delimiter
1325 if ( (_curchar != '%')
1326 || ( next_char(), (_curchar != '{')) ) {
1327 parse_err(SYNERR, "missing '%{' in pipeline definition\n");
1328 return;
1329 }
1330 next_char(); // Maintain the invariant
1331 do {
1332 ident = get_ident(); // Grab next identifier
1333 if (ident == NULL) {
1334 parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
1335 continue;
1336 }
1337 if (!strcmp(ident, "resources" )) resource_parse(*pipeline);
1338 else if (!strcmp(ident, "pipe_desc" )) pipe_desc_parse(*pipeline);
1339 else if (!strcmp(ident, "pipe_class")) pipe_class_parse(*pipeline);
1340 else if (!strcmp(ident, "define")) {
1341 skipws();
1342 if ( (_curchar != '%')
1343 || ( next_char(), (_curchar != '{')) ) {
1344 parse_err(SYNERR, "expected '%{'\n");
1345 return;
1346 }
1347 next_char(); skipws();
1349 char *node_class = get_ident();
1350 if (node_class == NULL) {
1351 parse_err(SYNERR, "expected identifier, found \"%c\"\n", _curchar);
1352 return;
1353 }
1355 skipws();
1356 if (_curchar != ',' && _curchar != '=') {
1357 parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1358 break;
1359 }
1360 next_char(); skipws();
1362 char *pipe_class = get_ident();
1363 if (pipe_class == NULL) {
1364 parse_err(SYNERR, "expected identifier, found \"%c\"\n", _curchar);
1365 return;
1366 }
1367 if (_curchar != ';' ) {
1368 parse_err(SYNERR, "expected `;`, found '%c'\n", _curchar);
1369 break;
1370 }
1371 next_char(); // Skip over semi-colon
1373 skipws();
1374 if ( (_curchar != '%')
1375 || ( next_char(), (_curchar != '}')) ) {
1376 parse_err(SYNERR, "expected '%%}', found \"%c\"\n", _curchar);
1377 }
1378 next_char();
1380 // Check ident for validity
1381 if (_AD._pipeline && !_AD._pipeline->_classlist.search(pipe_class)) {
1382 parse_err(SYNERR, "\"%s\" is not a valid pipeline class\n", pipe_class);
1383 return;
1384 }
1386 // Add this machine node to the list in the pipeline class
1387 _AD._pipeline->_classdict[pipe_class]->is_pipeclass()->_instructs.addName(node_class);
1389 MachNodeForm *machnode = new MachNodeForm(node_class); // Create new machnode form
1390 machnode->_machnode_pipe = pipe_class;
1392 _AD.addForm(machnode);
1393 }
1394 else if (!strcmp(ident, "attributes")) {
1395 bool vsi_seen = false, bhds_seen = false;
1397 skipws();
1398 if ( (_curchar != '%')
1399 || ( next_char(), (_curchar != '{')) ) {
1400 parse_err(SYNERR, "expected '%{'\n");
1401 return;
1402 }
1403 next_char(); skipws();
1405 while (_curchar != '%') {
1406 ident = get_ident();
1407 if (ident == NULL)
1408 break;
1410 if (!strcmp(ident, "variable_size_instructions")) {
1411 skipws();
1412 if (_curchar == ';') {
1413 next_char(); skipws();
1414 }
1416 pipeline->_variableSizeInstrs = true;
1417 vsi_seen = true;
1418 continue;
1419 }
1421 if (!strcmp(ident, "fixed_size_instructions")) {
1422 skipws();
1423 if (_curchar == ';') {
1424 next_char(); skipws();
1425 }
1427 pipeline->_variableSizeInstrs = false;
1428 vsi_seen = true;
1429 continue;
1430 }
1432 if (!strcmp(ident, "branch_has_delay_slot")) {
1433 skipws();
1434 if (_curchar == ';') {
1435 next_char(); skipws();
1436 }
1438 pipeline->_branchHasDelaySlot = true;
1439 bhds_seen = true;
1440 continue;
1441 }
1443 if (!strcmp(ident, "max_instructions_per_bundle")) {
1444 skipws();
1445 if (_curchar != '=') {
1446 parse_err(SYNERR, "expected `=`\n");
1447 break;
1448 }
1450 next_char(); skipws();
1451 pipeline->_maxInstrsPerBundle = get_int();
1452 skipws();
1454 if (_curchar == ';') {
1455 next_char(); skipws();
1456 }
1458 continue;
1459 }
1461 if (!strcmp(ident, "max_bundles_per_cycle")) {
1462 skipws();
1463 if (_curchar != '=') {
1464 parse_err(SYNERR, "expected `=`\n");
1465 break;
1466 }
1468 next_char(); skipws();
1469 pipeline->_maxBundlesPerCycle = get_int();
1470 skipws();
1472 if (_curchar == ';') {
1473 next_char(); skipws();
1474 }
1476 continue;
1477 }
1479 if (!strcmp(ident, "instruction_unit_size")) {
1480 skipws();
1481 if (_curchar != '=') {
1482 parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1483 break;
1484 }
1486 next_char(); skipws();
1487 pipeline->_instrUnitSize = get_int();
1488 skipws();
1490 if (_curchar == ';') {
1491 next_char(); skipws();
1492 }
1494 continue;
1495 }
1497 if (!strcmp(ident, "bundle_unit_size")) {
1498 skipws();
1499 if (_curchar != '=') {
1500 parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1501 break;
1502 }
1504 next_char(); skipws();
1505 pipeline->_bundleUnitSize = get_int();
1506 skipws();
1508 if (_curchar == ';') {
1509 next_char(); skipws();
1510 }
1512 continue;
1513 }
1515 if (!strcmp(ident, "instruction_fetch_unit_size")) {
1516 skipws();
1517 if (_curchar != '=') {
1518 parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1519 break;
1520 }
1522 next_char(); skipws();
1523 pipeline->_instrFetchUnitSize = get_int();
1524 skipws();
1526 if (_curchar == ';') {
1527 next_char(); skipws();
1528 }
1530 continue;
1531 }
1533 if (!strcmp(ident, "instruction_fetch_units")) {
1534 skipws();
1535 if (_curchar != '=') {
1536 parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1537 break;
1538 }
1540 next_char(); skipws();
1541 pipeline->_instrFetchUnits = get_int();
1542 skipws();
1544 if (_curchar == ';') {
1545 next_char(); skipws();
1546 }
1548 continue;
1549 }
1551 if (!strcmp(ident, "nops")) {
1552 skipws();
1553 if (_curchar != '(') {
1554 parse_err(SYNERR, "expected `(`, found '%c'\n", _curchar);
1555 break;
1556 }
1558 next_char(); skipws();
1560 while (_curchar != ')') {
1561 ident = get_ident();
1562 if (ident == NULL) {
1563 parse_err(SYNERR, "expected identifier for nop instruction, found '%c'\n", _curchar);
1564 break;
1565 }
1567 pipeline->_noplist.addName(ident);
1568 pipeline->_nopcnt++;
1569 skipws();
1571 if (_curchar == ',') {
1572 next_char(); skipws();
1573 }
1574 }
1576 next_char(); skipws();
1578 if (_curchar == ';') {
1579 next_char(); skipws();
1580 }
1582 continue;
1583 }
1585 parse_err(SYNERR, "unknown specifier \"%s\"\n", ident);
1586 }
1588 if ( (_curchar != '%')
1589 || ( next_char(), (_curchar != '}')) ) {
1590 parse_err(SYNERR, "expected '%}', found \"%c\"\n", _curchar);
1591 }
1592 next_char(); skipws();
1594 if (pipeline->_maxInstrsPerBundle == 0)
1595 parse_err(SYNERR, "\"max_instructions_per_bundle\" unspecified\n");
1596 if (pipeline->_instrUnitSize == 0 && pipeline->_bundleUnitSize == 0)
1597 parse_err(SYNERR, "\"instruction_unit_size\" and \"bundle_unit_size\" unspecified\n");
1598 if (pipeline->_instrFetchUnitSize == 0)
1599 parse_err(SYNERR, "\"instruction_fetch_unit_size\" unspecified\n");
1600 if (pipeline->_instrFetchUnits == 0)
1601 parse_err(SYNERR, "\"instruction_fetch_units\" unspecified\n");
1602 if (!vsi_seen)
1603 parse_err(SYNERR, "\"variable_size_instruction\" or \"fixed_size_instruction\" unspecified\n");
1604 }
1605 else { // Done with staticly defined parts of instruction definition
1606 parse_err(SYNERR, "expected one of \"resources\", \"pipe_desc\", \"pipe_class\", found \"%s\"\n", ident);
1607 return;
1608 }
1609 skipws();
1610 if (_curchar == ';')
1611 skipws();
1612 } while(_curchar != '%');
1614 next_char();
1615 if (_curchar != '}') {
1616 parse_err(SYNERR, "missing \"%}\" in pipeline definition\n");
1617 return;
1618 }
1620 next_char();
1621 }
1623 //------------------------------resource_parse----------------------------
1624 void ADLParser::resource_parse(PipelineForm &pipeline) {
1625 ResourceForm *resource;
1626 char * ident;
1627 char * expr;
1628 unsigned mask;
1629 pipeline._rescount = 0;
1631 skipws(); // Skip leading whitespace
1633 if (_curchar != '(') {
1634 parse_err(SYNERR, "missing \"(\" in resource definition\n");
1635 return;
1636 }
1638 do {
1639 next_char(); // Skip "(" or ","
1640 ident = get_ident(); // Grab next identifier
1642 if (ident == NULL) {
1643 parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1644 return;
1645 }
1646 skipws();
1648 if (_curchar != '=') {
1649 mask = (1 << pipeline._rescount++);
1650 }
1651 else {
1652 next_char(); skipws();
1653 expr = get_ident(); // Grab next identifier
1654 if (expr == NULL) {
1655 parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1656 return;
1657 }
1658 resource = (ResourceForm *) pipeline._resdict[expr];
1659 if (resource == NULL) {
1660 parse_err(SYNERR, "resource \"%s\" is not defined\n", expr);
1661 return;
1662 }
1663 mask = resource->mask();
1665 skipws();
1666 while (_curchar == '|') {
1667 next_char(); skipws();
1669 expr = get_ident(); // Grab next identifier
1670 if (expr == NULL) {
1671 parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1672 return;
1673 }
1675 resource = (ResourceForm *) pipeline._resdict[expr]; // Look up the value
1676 if (resource == NULL) {
1677 parse_err(SYNERR, "resource \"%s\" is not defined\n", expr);
1678 return;
1679 }
1681 mask |= resource->mask();
1682 skipws();
1683 }
1684 }
1686 resource = new ResourceForm(mask);
1688 pipeline._resdict.Insert(ident, resource);
1689 pipeline._reslist.addName(ident);
1690 } while (_curchar == ',');
1692 if (_curchar != ')') {
1693 parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
1694 return;
1695 }
1697 next_char(); // Skip ")"
1698 if (_curchar == ';')
1699 next_char(); // Skip ";"
1700 }
1702 //------------------------------resource_parse----------------------------
1703 void ADLParser::pipe_desc_parse(PipelineForm &pipeline) {
1704 char * ident;
1706 skipws(); // Skip leading whitespace
1708 if (_curchar != '(') {
1709 parse_err(SYNERR, "missing \"(\" in pipe_desc definition\n");
1710 return;
1711 }
1713 do {
1714 next_char(); // Skip "(" or ","
1715 ident = get_ident(); // Grab next identifier
1716 if (ident == NULL) {
1717 parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1718 return;
1719 }
1721 // Add the name to the list
1722 pipeline._stages.addName(ident);
1723 pipeline._stagecnt++;
1725 skipws();
1726 } while (_curchar == ',');
1728 if (_curchar != ')') {
1729 parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
1730 return;
1731 }
1733 next_char(); // Skip ")"
1734 if (_curchar == ';')
1735 next_char(); // Skip ";"
1736 }
1738 //------------------------------pipe_class_parse--------------------------
1739 void ADLParser::pipe_class_parse(PipelineForm &pipeline) {
1740 PipeClassForm *pipe_class;
1741 char * ident;
1742 char * stage;
1743 char * read_or_write;
1744 int is_write;
1745 int is_read;
1746 OperandForm *oper;
1748 skipws(); // Skip leading whitespace
1750 ident = get_ident(); // Grab next identifier
1752 if (ident == NULL) {
1753 parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1754 return;
1755 }
1757 // Create a record for the pipe_class
1758 pipe_class = new PipeClassForm(ident, ++pipeline._classcnt);
1759 pipeline._classdict.Insert(ident, pipe_class);
1760 pipeline._classlist.addName(ident);
1762 // Then get the operands
1763 skipws();
1764 if (_curchar != '(') {
1765 parse_err(SYNERR, "missing \"(\" in pipe_class definition\n");
1766 }
1767 // Parse the operand list
1768 else get_oplist(pipe_class->_parameters, pipe_class->_localNames);
1769 skipws(); // Skip leading whitespace
1770 // Check for block delimiter
1771 if ( (_curchar != '%')
1772 || ( next_char(), (_curchar != '{')) ) {
1773 parse_err(SYNERR, "missing \"%{\" in pipe_class definition\n");
1774 return;
1775 }
1776 next_char();
1778 do {
1779 ident = get_ident(); // Grab next identifier
1780 if (ident == NULL) {
1781 parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1782 continue;
1783 }
1784 skipws();
1786 if (!strcmp(ident, "fixed_latency")) {
1787 skipws();
1788 if (_curchar != '(') {
1789 parse_err(SYNERR, "missing \"(\" in latency definition\n");
1790 return;
1791 }
1792 next_char(); skipws();
1793 if( !isdigit(_curchar) ) {
1794 parse_err(SYNERR, "number expected for \"%c\" in latency definition\n", _curchar);
1795 return;
1796 }
1797 int fixed_latency = get_int();
1798 skipws();
1799 if (_curchar != ')') {
1800 parse_err(SYNERR, "missing \")\" in latency definition\n");
1801 return;
1802 }
1803 next_char(); skipws();
1804 if (_curchar != ';') {
1805 parse_err(SYNERR, "missing \";\" in latency definition\n");
1806 return;
1807 }
1809 pipe_class->setFixedLatency(fixed_latency);
1810 next_char(); skipws();
1811 continue;
1812 }
1814 if (!strcmp(ident, "zero_instructions") ||
1815 !strcmp(ident, "no_instructions")) {
1816 skipws();
1817 if (_curchar != ';') {
1818 parse_err(SYNERR, "missing \";\" in latency definition\n");
1819 return;
1820 }
1822 pipe_class->setInstructionCount(0);
1823 next_char(); skipws();
1824 continue;
1825 }
1827 if (!strcmp(ident, "one_instruction_with_delay_slot") ||
1828 !strcmp(ident, "single_instruction_with_delay_slot")) {
1829 skipws();
1830 if (_curchar != ';') {
1831 parse_err(SYNERR, "missing \";\" in latency definition\n");
1832 return;
1833 }
1835 pipe_class->setInstructionCount(1);
1836 pipe_class->setBranchDelay(true);
1837 next_char(); skipws();
1838 continue;
1839 }
1841 if (!strcmp(ident, "one_instruction") ||
1842 !strcmp(ident, "single_instruction")) {
1843 skipws();
1844 if (_curchar != ';') {
1845 parse_err(SYNERR, "missing \";\" in latency definition\n");
1846 return;
1847 }
1849 pipe_class->setInstructionCount(1);
1850 next_char(); skipws();
1851 continue;
1852 }
1854 if (!strcmp(ident, "instructions_in_first_bundle") ||
1855 !strcmp(ident, "instruction_count")) {
1856 skipws();
1858 int number_of_instructions = 1;
1860 if (_curchar != '(') {
1861 parse_err(SYNERR, "\"(\" expected at \"%c\"\n", _curchar);
1862 continue;
1863 }
1865 next_char(); skipws();
1866 number_of_instructions = get_int();
1868 skipws();
1869 if (_curchar != ')') {
1870 parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
1871 continue;
1872 }
1874 next_char(); skipws();
1875 if (_curchar != ';') {
1876 parse_err(SYNERR, "missing \";\" in latency definition\n");
1877 return;
1878 }
1880 pipe_class->setInstructionCount(number_of_instructions);
1881 next_char(); skipws();
1882 continue;
1883 }
1885 if (!strcmp(ident, "multiple_bundles")) {
1886 skipws();
1887 if (_curchar != ';') {
1888 parse_err(SYNERR, "missing \";\" after multiple bundles\n");
1889 return;
1890 }
1892 pipe_class->setMultipleBundles(true);
1893 next_char(); skipws();
1894 continue;
1895 }
1897 if (!strcmp(ident, "has_delay_slot")) {
1898 skipws();
1899 if (_curchar != ';') {
1900 parse_err(SYNERR, "missing \";\" after \"has_delay_slot\"\n");
1901 return;
1902 }
1904 pipe_class->setBranchDelay(true);
1905 next_char(); skipws();
1906 continue;
1907 }
1909 if (!strcmp(ident, "force_serialization")) {
1910 skipws();
1911 if (_curchar != ';') {
1912 parse_err(SYNERR, "missing \";\" after \"force_serialization\"\n");
1913 return;
1914 }
1916 pipe_class->setForceSerialization(true);
1917 next_char(); skipws();
1918 continue;
1919 }
1921 if (!strcmp(ident, "may_have_no_code")) {
1922 skipws();
1923 if (_curchar != ';') {
1924 parse_err(SYNERR, "missing \";\" after \"may_have_no_code\"\n");
1925 return;
1926 }
1928 pipe_class->setMayHaveNoCode(true);
1929 next_char(); skipws();
1930 continue;
1931 }
1933 const Form *parm = pipe_class->_localNames[ident];
1934 if (parm != NULL) {
1935 oper = parm->is_operand();
1936 if (oper == NULL && !parm->is_opclass()) {
1937 parse_err(SYNERR, "operand name expected at %s\n", ident);
1938 continue;
1939 }
1941 if (_curchar != ':') {
1942 parse_err(SYNERR, "\":\" expected at \"%c\"\n", _curchar);
1943 continue;
1944 }
1945 next_char(); skipws();
1946 stage = get_ident();
1947 if (stage == NULL) {
1948 parse_err(SYNERR, "pipeline stage identifier expected at \"%c\"\n", _curchar);
1949 continue;
1950 }
1952 skipws();
1953 if (_curchar != '(') {
1954 parse_err(SYNERR, "\"(\" expected at \"%c\"\n", _curchar);
1955 continue;
1956 }
1958 next_char();
1959 read_or_write = get_ident();
1960 if (read_or_write == NULL) {
1961 parse_err(SYNERR, "\"read\" or \"write\" expected at \"%c\"\n", _curchar);
1962 continue;
1963 }
1965 is_read = strcmp(read_or_write, "read") == 0;
1966 is_write = strcmp(read_or_write, "write") == 0;
1967 if (!is_read && !is_write) {
1968 parse_err(SYNERR, "\"read\" or \"write\" expected at \"%c\"\n", _curchar);
1969 continue;
1970 }
1972 skipws();
1973 if (_curchar != ')') {
1974 parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
1975 continue;
1976 }
1978 next_char(); skipws();
1979 int more_instrs = 0;
1980 if (_curchar == '+') {
1981 next_char(); skipws();
1982 if (_curchar < '0' || _curchar > '9') {
1983 parse_err(SYNERR, "<number> expected at \"%c\"\n", _curchar);
1984 continue;
1985 }
1986 while (_curchar >= '0' && _curchar <= '9') {
1987 more_instrs *= 10;
1988 more_instrs += _curchar - '0';
1989 next_char();
1990 }
1991 skipws();
1992 }
1994 PipeClassOperandForm *pipe_operand = new PipeClassOperandForm(stage, is_write, more_instrs);
1995 pipe_class->_localUsage.Insert(ident, pipe_operand);
1997 if (_curchar == '%')
1998 continue;
2000 if (_curchar != ';') {
2001 parse_err(SYNERR, "\";\" expected at \"%c\"\n", _curchar);
2002 continue;
2003 }
2004 next_char(); skipws();
2005 continue;
2006 }
2008 // Scan for Resource Specifier
2009 const Form *res = pipeline._resdict[ident];
2010 if (res != NULL) {
2011 int cyclecnt = 1;
2012 if (_curchar != ':') {
2013 parse_err(SYNERR, "\":\" expected at \"%c\"\n", _curchar);
2014 continue;
2015 }
2016 next_char(); skipws();
2017 stage = get_ident();
2018 if (stage == NULL) {
2019 parse_err(SYNERR, "pipeline stage identifier expected at \"%c\"\n", _curchar);
2020 continue;
2021 }
2023 skipws();
2024 if (_curchar == '(') {
2025 next_char();
2026 cyclecnt = get_int();
2028 skipws();
2029 if (_curchar != ')') {
2030 parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
2031 continue;
2032 }
2034 next_char(); skipws();
2035 }
2037 PipeClassResourceForm *resource = new PipeClassResourceForm(ident, stage, cyclecnt);
2038 int stagenum = pipeline._stages.index(stage);
2039 if (pipeline._maxcycleused < (stagenum+cyclecnt))
2040 pipeline._maxcycleused = (stagenum+cyclecnt);
2041 pipe_class->_resUsage.addForm(resource);
2043 if (_curchar == '%')
2044 continue;
2046 if (_curchar != ';') {
2047 parse_err(SYNERR, "\";\" expected at \"%c\"\n", _curchar);
2048 continue;
2049 }
2050 next_char(); skipws();
2051 continue;
2052 }
2054 parse_err(SYNERR, "resource expected at \"%s\"\n", ident);
2055 return;
2056 } while(_curchar != '%');
2058 next_char();
2059 if (_curchar != '}') {
2060 parse_err(SYNERR, "missing \"%}\" in pipe_class definition\n");
2061 return;
2062 }
2064 next_char();
2065 }
2067 //------------------------------peep_parse-------------------------------------
2068 void ADLParser::peep_parse(void) {
2069 Peephole *peep; // Pointer to current peephole rule form
2070 char *desc = NULL; // String representation of rule
2072 skipws(); // Skip leading whitespace
2074 peep = new Peephole(); // Build new Peephole object
2075 // Check for open block sequence
2076 skipws(); // Skip leading whitespace
2077 if (_curchar == '%' && *(_ptr+1) == '{') {
2078 next_char(); next_char(); // Skip "%{"
2079 skipws();
2080 while (_curchar != '%' && *(_ptr+1) != '}') {
2081 char *token = get_ident();
2082 if (token == NULL) {
2083 parse_err(SYNERR, "missing identifier inside peephole rule.\n");
2084 return;
2085 }
2086 // check for legal subsections of peephole rule
2087 if (strcmp(token,"peepmatch")==0) {
2088 peep_match_parse(*peep); }
2089 else if (strcmp(token,"peepconstraint")==0) {
2090 peep_constraint_parse(*peep); }
2091 else if (strcmp(token,"peepreplace")==0) {
2092 peep_replace_parse(*peep); }
2093 else {
2094 parse_err(SYNERR, "expected peepmatch, peepconstraint, or peepreplace for identifier %s.\n", token);
2095 }
2096 skipws();
2097 }
2098 }
2099 else {
2100 parse_err(SYNERR, "Missing %%{ ... %%} block after peephole keyword.\n");
2101 return;
2102 }
2103 next_char(); // Skip past '%'
2104 next_char(); // Skip past '}'
2105 }
2107 // ******************** Private Level 2 Parse Functions ********************
2108 //------------------------------constraint_parse------------------------------
2109 Constraint *ADLParser::constraint_parse(void) {
2110 char *func;
2111 char *arg;
2113 // Check for constraint expression
2114 skipws();
2115 if (_curchar != '(') {
2116 parse_err(SYNERR, "missing constraint expression, (...)\n");
2117 return NULL;
2118 }
2119 next_char(); // Skip past '('
2121 // Get constraint function
2122 skipws();
2123 func = get_ident();
2124 if (func == NULL) {
2125 parse_err(SYNERR, "missing function in constraint expression.\n");
2126 return NULL;
2127 }
2128 if (strcmp(func,"ALLOC_IN_RC")==0
2129 || strcmp(func,"IS_R_CLASS")==0) {
2130 // Check for '(' before argument
2131 skipws();
2132 if (_curchar != '(') {
2133 parse_err(SYNERR, "missing '(' for constraint function's argument.\n");
2134 return NULL;
2135 }
2136 next_char();
2138 // Get it's argument
2139 skipws();
2140 arg = get_ident();
2141 if (arg == NULL) {
2142 parse_err(SYNERR, "missing argument for constraint function %s\n",func);
2143 return NULL;
2144 }
2145 // Check for ')' after argument
2146 skipws();
2147 if (_curchar != ')') {
2148 parse_err(SYNERR, "missing ')' after constraint function argument %s\n",arg);
2149 return NULL;
2150 }
2151 next_char();
2152 } else {
2153 parse_err(SYNERR, "Invalid constraint function %s\n",func);
2154 return NULL;
2155 }
2157 // Check for closing paren and ';'
2158 skipws();
2159 if (_curchar != ')') {
2160 parse_err(SYNERR, "Missing ')' for constraint function %s\n",func);
2161 return NULL;
2162 }
2163 next_char();
2164 skipws();
2165 if (_curchar != ';') {
2166 parse_err(SYNERR, "Missing ';' after constraint.\n");
2167 return NULL;
2168 }
2169 next_char();
2171 // Create new "Constraint"
2172 Constraint *constraint = new Constraint(func,arg);
2173 return constraint;
2174 }
2176 //------------------------------constr_parse-----------------------------------
2177 ConstructRule *ADLParser::construct_parse(void) {
2178 return NULL;
2179 }
2182 //------------------------------reg_def_parse----------------------------------
2183 void ADLParser::reg_def_parse(void) {
2184 char *rname; // Name of register being defined
2186 // Get register name
2187 skipws(); // Skip whitespace
2188 rname = get_ident();
2189 if (rname == NULL) {
2190 parse_err(SYNERR, "missing register name after reg_def\n");
2191 return;
2192 }
2194 // Check for definition of register calling convention (save on call, ...),
2195 // register save type, and register encoding value.
2196 skipws();
2197 char *callconv = NULL;
2198 char *c_conv = NULL;
2199 char *idealtype = NULL;
2200 char *encoding = NULL;
2201 char *concrete = NULL;
2202 if (_curchar == '(') {
2203 next_char();
2204 callconv = get_ident();
2205 // Parse the internal calling convention, must be NS, SOC, SOE, or AS.
2206 if (callconv == NULL) {
2207 parse_err(SYNERR, "missing register calling convention value\n");
2208 return;
2209 }
2210 if(strcmp(callconv, "SOC") && strcmp(callconv,"SOE") &&
2211 strcmp(callconv, "NS") && strcmp(callconv, "AS")) {
2212 parse_err(SYNERR, "invalid value for register calling convention\n");
2213 }
2214 skipws();
2215 if (_curchar != ',') {
2216 parse_err(SYNERR, "missing comma in register definition statement\n");
2217 return;
2218 }
2219 next_char();
2221 // Parse the native calling convention, must be NS, SOC, SOE, AS
2222 c_conv = get_ident();
2223 if (c_conv == NULL) {
2224 parse_err(SYNERR, "missing register native calling convention value\n");
2225 return;
2226 }
2227 if(strcmp(c_conv, "SOC") && strcmp(c_conv,"SOE") &&
2228 strcmp(c_conv, "NS") && strcmp(c_conv, "AS")) {
2229 parse_err(SYNERR, "invalid value for register calling convention\n");
2230 }
2231 skipws();
2232 if (_curchar != ',') {
2233 parse_err(SYNERR, "missing comma in register definition statement\n");
2234 return;
2235 }
2236 next_char();
2237 skipws();
2239 // Parse the ideal save type
2240 idealtype = get_ident();
2241 if (idealtype == NULL) {
2242 parse_err(SYNERR, "missing register save type value\n");
2243 return;
2244 }
2245 skipws();
2246 if (_curchar != ',') {
2247 parse_err(SYNERR, "missing comma in register definition statement\n");
2248 return;
2249 }
2250 next_char();
2251 skipws();
2253 // Parse the encoding value
2254 encoding = get_expr("encoding", ",");
2255 if (encoding == NULL) {
2256 parse_err(SYNERR, "missing register encoding value\n");
2257 return;
2258 }
2259 trim(encoding);
2260 if (_curchar != ',') {
2261 parse_err(SYNERR, "missing comma in register definition statement\n");
2262 return;
2263 }
2264 next_char();
2265 skipws();
2266 // Parse the concrete name type
2267 // concrete = get_ident();
2268 concrete = get_expr("concrete", ")");
2269 if (concrete == NULL) {
2270 parse_err(SYNERR, "missing vm register name value\n");
2271 return;
2272 }
2274 if (_curchar != ')') {
2275 parse_err(SYNERR, "missing ')' in register definition statement\n");
2276 return;
2277 }
2278 next_char();
2279 }
2281 // Check for closing ';'
2282 skipws();
2283 if (_curchar != ';') {
2284 parse_err(SYNERR, "missing ';' after reg_def\n");
2285 return;
2286 }
2287 next_char(); // move past ';'
2289 // Debug Stuff
2290 if (_AD._adl_debug > 1) {
2291 fprintf(stderr,"Register Definition: %s ( %s, %s %s )\n", rname,
2292 (callconv ? callconv : ""), (c_conv ? c_conv : ""), concrete);
2293 }
2295 // Record new register definition.
2296 _AD._register->addRegDef(rname, callconv, c_conv, idealtype, encoding, concrete);
2297 return;
2298 }
2300 //------------------------------reg_class_parse--------------------------------
2301 void ADLParser::reg_class_parse(void) {
2302 char *cname; // Name of register class being defined
2304 // Get register class name
2305 skipws(); // Skip leading whitespace
2306 cname = get_ident();
2307 if (cname == NULL) {
2308 parse_err(SYNERR, "missing register class name after 'reg_class'\n");
2309 return;
2310 }
2311 // Debug Stuff
2312 if (_AD._adl_debug >1) fprintf(stderr,"Register Class: %s\n", cname);
2314 RegClass *reg_class = _AD._register->addRegClass(cname);
2316 // Collect registers in class
2317 skipws();
2318 if (_curchar == '(') {
2319 next_char(); // Skip '('
2320 skipws();
2321 while (_curchar != ')') {
2322 char *rname = get_ident();
2323 if (rname==NULL) {
2324 parse_err(SYNERR, "missing identifier inside reg_class list.\n");
2325 return;
2326 }
2327 RegDef *regDef = _AD._register->getRegDef(rname);
2328 if (!regDef) {
2329 parse_err(SEMERR, "unknown identifier %s inside reg_class list.\n", rname);
2330 } else {
2331 reg_class->addReg(regDef); // add regDef to regClass
2332 }
2334 // Check for ',' and position to next token.
2335 skipws();
2336 if (_curchar == ',') {
2337 next_char(); // Skip trailing ','
2338 skipws();
2339 }
2340 }
2341 next_char(); // Skip closing ')'
2342 } else if (_curchar == '%') {
2343 char *code = find_cpp_block("reg class");
2344 if (code == NULL) {
2345 parse_err(SYNERR, "missing code declaration for reg class.\n");
2346 return;
2347 }
2348 reg_class->_user_defined = code;
2349 return;
2350 }
2352 // Check for terminating ';'
2353 skipws();
2354 if (_curchar != ';') {
2355 parse_err(SYNERR, "missing ';' at end of reg_class definition.\n");
2356 return;
2357 }
2358 next_char(); // Skip trailing ';'
2360 // Check RegClass size, must be <= 32 registers in class.
2362 return;
2363 }
2365 //------------------------------alloc_class_parse------------------------------
2366 void ADLParser::alloc_class_parse(void) {
2367 char *name; // Name of allocation class being defined
2369 // Get allocation class name
2370 skipws(); // Skip leading whitespace
2371 name = get_ident();
2372 if (name == NULL) {
2373 parse_err(SYNERR, "missing allocation class name after 'reg_class'\n");
2374 return;
2375 }
2376 // Debug Stuff
2377 if (_AD._adl_debug >1) fprintf(stderr,"Allocation Class: %s\n", name);
2379 AllocClass *alloc_class = _AD._register->addAllocClass(name);
2381 // Collect registers in class
2382 skipws();
2383 if (_curchar == '(') {
2384 next_char(); // Skip '('
2385 skipws();
2386 while (_curchar != ')') {
2387 char *rname = get_ident();
2388 if (rname==NULL) {
2389 parse_err(SYNERR, "missing identifier inside reg_class list.\n");
2390 return;
2391 }
2392 // Check if name is a RegDef
2393 RegDef *regDef = _AD._register->getRegDef(rname);
2394 if (regDef) {
2395 alloc_class->addReg(regDef); // add regDef to allocClass
2396 } else {
2398 // name must be a RegDef or a RegClass
2399 parse_err(SYNERR, "name %s should be a previously defined reg_def.\n", rname);
2400 return;
2401 }
2403 // Check for ',' and position to next token.
2404 skipws();
2405 if (_curchar == ',') {
2406 next_char(); // Skip trailing ','
2407 skipws();
2408 }
2409 }
2410 next_char(); // Skip closing ')'
2411 }
2413 // Check for terminating ';'
2414 skipws();
2415 if (_curchar != ';') {
2416 parse_err(SYNERR, "missing ';' at end of reg_class definition.\n");
2417 return;
2418 }
2419 next_char(); // Skip trailing ';'
2421 return;
2422 }
2424 //------------------------------peep_match_child_parse-------------------------
2425 InstructForm *ADLParser::peep_match_child_parse(PeepMatch &match, int parent, int &position, int input){
2426 char *token = NULL;
2427 int lparen = 0; // keep track of parenthesis nesting depth
2428 int rparen = 0; // position of instruction at this depth
2429 InstructForm *inst_seen = NULL;
2430 InstructForm *child_seen = NULL;
2432 // Walk the match tree,
2433 // Record <parent, position, instruction name, input position>
2434 while ( lparen >= rparen ) {
2435 skipws();
2436 // Left paren signals start of an input, collect with recursive call
2437 if (_curchar == '(') {
2438 ++lparen;
2439 next_char();
2440 child_seen = peep_match_child_parse(match, parent, position, rparen);
2441 }
2442 // Right paren signals end of an input, may be more
2443 else if (_curchar == ')') {
2444 ++rparen;
2445 if( rparen == lparen ) { // IF rparen matches an lparen I've seen
2446 next_char(); // move past ')'
2447 } else { // ELSE leave ')' for parent
2448 assert( rparen == lparen + 1, "Should only see one extra ')'");
2449 // if an instruction was not specified for this paren-pair
2450 if( ! inst_seen ) { // record signal entry
2451 match.add_instruction( parent, position, NameList::_signal, input );
2452 ++position;
2453 }
2454 // ++input; // TEMPORARY
2455 return inst_seen;
2456 }
2457 }
2458 // if no parens, then check for instruction name
2459 // This instruction is the parent of a sub-tree
2460 else if ((token = get_ident_dup()) != NULL) {
2461 const Form *form = _AD._globalNames[token];
2462 if (form) {
2463 InstructForm *inst = form->is_instruction();
2464 // Record the first instruction at this level
2465 if( inst_seen == NULL ) {
2466 inst_seen = inst;
2467 }
2468 if (inst) {
2469 match.add_instruction( parent, position, token, input );
2470 parent = position;
2471 ++position;
2472 } else {
2473 parse_err(SYNERR, "instruction name expected at identifier %s.\n",
2474 token);
2475 return inst_seen;
2476 }
2477 }
2478 else {
2479 parse_err(SYNERR, "missing identifier in peepmatch rule.\n");
2480 return NULL;
2481 }
2482 }
2483 else {
2484 parse_err(SYNERR, "missing identifier in peepmatch rule.\n");
2485 return NULL;
2486 }
2488 } // end while
2490 assert( false, "ShouldNotReachHere();");
2491 return NULL;
2492 }
2494 //------------------------------peep_match_parse-------------------------------
2495 // Syntax for a peepmatch rule
2496 //
2497 // peepmatch ( root_instr_name [(instruction subtree)] [,(instruction subtree)]* );
2498 //
2499 void ADLParser::peep_match_parse(Peephole &peep) {
2501 skipws();
2502 // Check the structure of the rule
2503 // Check for open paren
2504 if (_curchar != '(') {
2505 parse_err(SYNERR, "missing '(' at start of peepmatch rule.\n");
2506 return;
2507 }
2508 next_char(); // skip '('
2510 // Construct PeepMatch and parse the peepmatch rule.
2511 PeepMatch *match = new PeepMatch(_ptr);
2512 int parent = -1; // parent of root
2513 int position = 0; // zero-based positions
2514 int input = 0; // input position in parent's operands
2515 InstructForm *root= peep_match_child_parse( *match, parent, position, input);
2516 if( root == NULL ) {
2517 parse_err(SYNERR, "missing instruction-name at start of peepmatch.\n");
2518 return;
2519 }
2521 if( _curchar != ')' ) {
2522 parse_err(SYNERR, "missing ')' at end of peepmatch.\n");
2523 return;
2524 }
2525 next_char(); // skip ')'
2527 // Check for closing semicolon
2528 skipws();
2529 if( _curchar != ';' ) {
2530 parse_err(SYNERR, "missing ';' at end of peepmatch.\n");
2531 return;
2532 }
2533 next_char(); // skip ';'
2535 // Store match into peep, and store peep into instruction
2536 peep.add_match(match);
2537 root->append_peephole(&peep);
2538 }
2540 //------------------------------peep_constraint_parse--------------------------
2541 // Syntax for a peepconstraint rule
2542 // A parenthesized list of relations between operands in peepmatch subtree
2543 //
2544 // peepconstraint %{
2545 // (instruction_number.operand_name
2546 // relational_op
2547 // instruction_number.operand_name OR register_name
2548 // [, ...] );
2549 //
2550 // // instruction numbers are zero-based using topological order in peepmatch
2551 //
2552 void ADLParser::peep_constraint_parse(Peephole &peep) {
2554 skipws();
2555 // Check the structure of the rule
2556 // Check for open paren
2557 if (_curchar != '(') {
2558 parse_err(SYNERR, "missing '(' at start of peepconstraint rule.\n");
2559 return;
2560 }
2561 else {
2562 next_char(); // Skip '('
2563 }
2565 // Check for a constraint
2566 skipws();
2567 while( _curchar != ')' ) {
2568 // Get information on the left instruction and its operand
2569 // left-instructions's number
2570 int left_inst = get_int();
2571 // Left-instruction's operand
2572 skipws();
2573 if( _curchar != '.' ) {
2574 parse_err(SYNERR, "missing '.' in peepconstraint after instruction number.\n");
2575 return;
2576 }
2577 next_char(); // Skip '.'
2578 char *left_op = get_ident_dup();
2580 skipws();
2581 // Collect relational operator
2582 char *relation = get_relation_dup();
2584 skipws();
2585 // Get information on the right instruction and its operand
2586 int right_inst; // Right-instructions's number
2587 if( isdigit(_curchar) ) {
2588 right_inst = get_int();
2589 // Right-instruction's operand
2590 skipws();
2591 if( _curchar != '.' ) {
2592 parse_err(SYNERR, "missing '.' in peepconstraint after instruction number.\n");
2593 return;
2594 }
2595 next_char(); // Skip '.'
2596 } else {
2597 right_inst = -1; // Flag as being a register constraint
2598 }
2600 char *right_op = get_ident_dup();
2602 // Construct the next PeepConstraint
2603 PeepConstraint *constraint = new PeepConstraint( left_inst, left_op,
2604 relation,
2605 right_inst, right_op );
2606 // And append it to the list for this peephole rule
2607 peep.append_constraint( constraint );
2609 // Check for another constraint, or end of rule
2610 skipws();
2611 if( _curchar == ',' ) {
2612 next_char(); // Skip ','
2613 skipws();
2614 }
2615 else if( _curchar != ')' ) {
2616 parse_err(SYNERR, "expected ',' or ')' after peephole constraint.\n");
2617 return;
2618 }
2619 } // end while( processing constraints )
2620 next_char(); // Skip ')'
2622 // Check for terminating ';'
2623 skipws();
2624 if (_curchar != ';') {
2625 parse_err(SYNERR, "missing ';' at end of peepconstraint.\n");
2626 return;
2627 }
2628 next_char(); // Skip trailing ';'
2629 }
2632 //------------------------------peep_replace_parse-----------------------------
2633 // Syntax for a peepreplace rule
2634 // root instruction name followed by a
2635 // parenthesized list of whitespace separated instruction.operand specifiers
2636 //
2637 // peepreplace ( instr_name ( [instruction_number.operand_name]* ) );
2638 //
2639 //
2640 void ADLParser::peep_replace_parse(Peephole &peep) {
2641 int lparen = 0; // keep track of parenthesis nesting depth
2642 int rparen = 0; // keep track of parenthesis nesting depth
2643 int icount = 0; // count of instructions in rule for naming
2644 char *str = NULL;
2645 char *token = NULL;
2647 skipws();
2648 // Check for open paren
2649 if (_curchar != '(') {
2650 parse_err(SYNERR, "missing '(' at start of peepreplace rule.\n");
2651 return;
2652 }
2653 else {
2654 lparen++;
2655 next_char();
2656 }
2658 // Check for root instruction
2659 char *inst = get_ident_dup();
2660 const Form *form = _AD._globalNames[inst];
2661 if( form == NULL || form->is_instruction() == NULL ) {
2662 parse_err(SYNERR, "Instruction name expected at start of peepreplace.\n");
2663 return;
2664 }
2666 // Store string representation of rule into replace
2667 PeepReplace *replace = new PeepReplace(str);
2668 replace->add_instruction( inst );
2670 skipws();
2671 // Start of root's operand-list
2672 if (_curchar != '(') {
2673 parse_err(SYNERR, "missing '(' at peepreplace root's operand-list.\n");
2674 return;
2675 }
2676 else {
2677 lparen++;
2678 next_char();
2679 }
2681 skipws();
2682 // Get the list of operands
2683 while( _curchar != ')' ) {
2684 // Get information on an instruction and its operand
2685 // instructions's number
2686 int inst_num = get_int();
2687 // Left-instruction's operand
2688 skipws();
2689 if( _curchar != '.' ) {
2690 parse_err(SYNERR, "missing '.' in peepreplace after instruction number.\n");
2691 return;
2692 }
2693 next_char(); // Skip '.'
2694 char *inst_op = get_ident_dup();
2695 if( inst_op == NULL ) {
2696 parse_err(SYNERR, "missing operand identifier in peepreplace.\n");
2697 return;
2698 }
2700 // Record this operand's position in peepmatch
2701 replace->add_operand( inst_num, inst_op );
2702 skipws();
2703 }
2705 // Check for the end of operands list
2706 skipws();
2707 assert( _curchar == ')', "While loop should have advanced to ')'.");
2708 next_char(); // Skip ')'
2710 skipws();
2711 // Check for end of peepreplace
2712 if( _curchar != ')' ) {
2713 parse_err(SYNERR, "missing ')' at end of peepmatch.\n");
2714 parse_err(SYNERR, "Support one replacement instruction.\n");
2715 return;
2716 }
2717 next_char(); // Skip ')'
2719 // Check for closing semicolon
2720 skipws();
2721 if( _curchar != ';' ) {
2722 parse_err(SYNERR, "missing ';' at end of peepreplace.\n");
2723 return;
2724 }
2725 next_char(); // skip ';'
2727 // Store replace into peep
2728 peep.add_replace( replace );
2729 }
2731 //------------------------------pred_parse-------------------------------------
2732 Predicate *ADLParser::pred_parse(void) {
2733 Predicate *predicate; // Predicate class for operand
2734 char *rule = NULL; // String representation of predicate
2736 skipws(); // Skip leading whitespace
2737 int line = linenum();
2738 if ( (rule = get_paren_expr("pred expression", true)) == NULL ) {
2739 parse_err(SYNERR, "incorrect or missing expression for 'predicate'\n");
2740 return NULL;
2741 }
2742 // Debug Stuff
2743 if (_AD._adl_debug > 1) fprintf(stderr,"Predicate: %s\n", rule);
2744 if (_curchar != ';') {
2745 parse_err(SYNERR, "missing ';' in predicate definition\n");
2746 return NULL;
2747 }
2748 next_char(); // Point after the terminator
2750 predicate = new Predicate(rule); // Build new predicate object
2751 skipws();
2752 return predicate;
2753 }
2756 //------------------------------ins_encode_parse_block-------------------------
2757 // Parse the block form of ins_encode. See ins_encode_parse for more details
2758 void ADLParser::ins_encode_parse_block(InstructForm& inst) {
2759 // Create a new encoding name based on the name of the instruction
2760 // definition, which should be unique.
2761 const char* prefix = "__ins_encode_";
2762 char* ec_name = (char*) malloc(strlen(inst._ident) + strlen(prefix) + 1);
2763 sprintf(ec_name, "%s%s", prefix, inst._ident);
2765 assert(_AD._encode->encClass(ec_name) == NULL, "shouldn't already exist");
2766 EncClass* encoding = _AD._encode->add_EncClass(ec_name);
2767 encoding->_linenum = linenum();
2769 // synthesize the arguments list for the enc_class from the
2770 // arguments to the instruct definition.
2771 const char* param = NULL;
2772 inst._parameters.reset();
2773 while ((param = inst._parameters.iter()) != NULL) {
2774 OperandForm* opForm = (OperandForm*) inst._localNames[param];
2775 encoding->add_parameter(opForm->_ident, param);
2776 }
2778 // Define a MacroAssembler instance for use by the encoding. The
2779 // name is chosen to match the __ idiom used for assembly in other
2780 // parts of hotspot and assumes the existence of the standard
2781 // #define __ _masm.
2782 encoding->add_code(" MacroAssembler _masm(&cbuf);\n");
2784 // Parse the following %{ }% block
2785 ins_encode_parse_block_impl(inst, encoding, ec_name);
2787 // Build an encoding rule which invokes the encoding rule we just
2788 // created, passing all arguments that we received.
2789 InsEncode* encrule = new InsEncode(); // Encode class for instruction
2790 NameAndList* params = encrule->add_encode(ec_name);
2791 inst._parameters.reset();
2792 while ((param = inst._parameters.iter()) != NULL) {
2793 params->add_entry(param);
2794 }
2796 // Check for duplicate ins_encode sections after parsing the block
2797 // so that parsing can continue and find any other errors.
2798 if (inst._insencode != NULL) {
2799 parse_err(SYNERR, "Multiple ins_encode sections defined\n");
2800 return;
2801 }
2803 // Set encode class of this instruction.
2804 inst._insencode = encrule;
2805 }
2808 void ADLParser::ins_encode_parse_block_impl(InstructForm& inst, EncClass* encoding, char* ec_name) {
2809 skipws_no_preproc(); // Skip leading whitespace
2810 // Prepend location descriptor, for debugging; cf. ADLParser::find_cpp_block
2811 if (_AD._adlocation_debug) {
2812 encoding->add_code(get_line_string());
2813 }
2815 // Collect the parts of the encode description
2816 // (1) strings that are passed through to output
2817 // (2) replacement/substitution variable, preceeded by a '$'
2818 while ((_curchar != '%') && (*(_ptr+1) != '}')) {
2820 // (1)
2821 // Check if there is a string to pass through to output
2822 char *start = _ptr; // Record start of the next string
2823 while ((_curchar != '$') && ((_curchar != '%') || (*(_ptr+1) != '}')) ) {
2824 // If at the start of a comment, skip past it
2825 if( (_curchar == '/') && ((*(_ptr+1) == '/') || (*(_ptr+1) == '*')) ) {
2826 skipws_no_preproc();
2827 } else {
2828 // ELSE advance to the next character, or start of the next line
2829 next_char_or_line();
2830 }
2831 }
2832 // If a string was found, terminate it and record in EncClass
2833 if (start != _ptr) {
2834 *_ptr = '\0'; // Terminate the string
2835 encoding->add_code(start);
2836 }
2838 // (2)
2839 // If we are at a replacement variable,
2840 // copy it and record in EncClass
2841 if (_curchar == '$') {
2842 // Found replacement Variable
2843 char* rep_var = get_rep_var_ident_dup();
2845 // Add flag to _strings list indicating we should check _rep_vars
2846 encoding->add_rep_var(rep_var);
2848 skipws();
2850 // Check if this instruct is a MachConstantNode.
2851 if (strcmp(rep_var, "constanttablebase") == 0) {
2852 // This instruct is a MachConstantNode.
2853 inst.set_is_mach_constant(true);
2855 if (_curchar == '(') {
2856 parse_err(SYNERR, "constanttablebase in instruct %s cannot have an argument (only constantaddress and constantoffset)", ec_name);
2857 return;
2858 }
2859 }
2860 else if ((strcmp(rep_var, "constantaddress") == 0) ||
2861 (strcmp(rep_var, "constantoffset") == 0)) {
2862 // This instruct is a MachConstantNode.
2863 inst.set_is_mach_constant(true);
2865 // If the constant keyword has an argument, parse it.
2866 if (_curchar == '(') constant_parse(inst);
2867 }
2868 }
2869 } // end while part of format description
2870 next_char(); // Skip '%'
2871 next_char(); // Skip '}'
2873 skipws();
2875 if (_AD._adlocation_debug) {
2876 encoding->add_code(end_line_marker());
2877 }
2879 // Debug Stuff
2880 if (_AD._adl_debug > 1) fprintf(stderr, "EncodingClass Form: %s\n", ec_name);
2881 }
2884 //------------------------------ins_encode_parse-------------------------------
2885 // Encode rules have the form
2886 // ins_encode( encode_class_name(parameter_list), ... );
2887 //
2888 // The "encode_class_name" must be defined in the encode section
2889 // The parameter list contains $names that are locals.
2890 //
2891 // Alternatively it can be written like this:
2892 //
2893 // ins_encode %{
2894 // ... // body
2895 // %}
2896 //
2897 // which synthesizes a new encoding class taking the same arguments as
2898 // the InstructForm, and automatically prefixes the definition with:
2899 //
2900 // MacroAssembler masm(&cbuf);\n");
2901 //
2902 // making it more compact to take advantage of the MacroAssembler and
2903 // placing the assembly closer to it's use by instructions.
2904 void ADLParser::ins_encode_parse(InstructForm& inst) {
2906 // Parse encode class name
2907 skipws(); // Skip whitespace
2908 if (_curchar != '(') {
2909 // Check for ins_encode %{ form
2910 if ((_curchar == '%') && (*(_ptr+1) == '{')) {
2911 next_char(); // Skip '%'
2912 next_char(); // Skip '{'
2914 // Parse the block form of ins_encode
2915 ins_encode_parse_block(inst);
2916 return;
2917 }
2919 parse_err(SYNERR, "missing '%%{' or '(' in ins_encode definition\n");
2920 return;
2921 }
2922 next_char(); // move past '('
2923 skipws();
2925 InsEncode *encrule = new InsEncode(); // Encode class for instruction
2926 encrule->_linenum = linenum();
2927 char *ec_name = NULL; // String representation of encode rule
2928 // identifier is optional.
2929 while (_curchar != ')') {
2930 ec_name = get_ident();
2931 if (ec_name == NULL) {
2932 parse_err(SYNERR, "Invalid encode class name after 'ins_encode('.\n");
2933 return;
2934 }
2935 // Check that encoding is defined in the encode section
2936 EncClass *encode_class = _AD._encode->encClass(ec_name);
2937 if (encode_class == NULL) {
2938 // Like to defer checking these till later...
2939 // parse_err(WARN, "Using an undefined encode class '%s' in 'ins_encode'.\n", ec_name);
2940 }
2942 // Get list for encode method's parameters
2943 NameAndList *params = encrule->add_encode(ec_name);
2945 // Parse the parameters to this encode method.
2946 skipws();
2947 if ( _curchar == '(' ) {
2948 next_char(); // move past '(' for parameters
2950 // Parse the encode method's parameters
2951 while (_curchar != ')') {
2952 char *param = get_ident_or_literal_constant("encoding operand");
2953 if ( param != NULL ) {
2954 // Found a parameter:
2955 // Check it is a local name, add it to the list, then check for more
2956 // New: allow hex constants as parameters to an encode method.
2957 // New: allow parenthesized expressions as parameters.
2958 // New: allow "primary", "secondary", "tertiary" as parameters.
2959 // New: allow user-defined register name as parameter
2960 if ( (inst._localNames[param] == NULL) &&
2961 !ADLParser::is_literal_constant(param) &&
2962 (Opcode::as_opcode_type(param) == Opcode::NOT_AN_OPCODE) &&
2963 ((_AD._register == NULL ) || (_AD._register->getRegDef(param) == NULL)) ) {
2964 parse_err(SYNERR, "Using non-locally defined parameter %s for encoding %s.\n", param, ec_name);
2965 return;
2966 }
2967 params->add_entry(param);
2969 skipws();
2970 if (_curchar == ',' ) {
2971 // More parameters to come
2972 next_char(); // move past ',' between parameters
2973 skipws(); // Skip to next parameter
2974 }
2975 else if (_curchar == ')') {
2976 // Done with parameter list
2977 }
2978 else {
2979 // Only ',' or ')' are valid after a parameter name
2980 parse_err(SYNERR, "expected ',' or ')' after parameter %s.\n",
2981 ec_name);
2982 return;
2983 }
2985 } else {
2986 skipws();
2987 // Did not find a parameter
2988 if (_curchar == ',') {
2989 parse_err(SYNERR, "Expected encode parameter before ',' in encoding %s.\n", ec_name);
2990 return;
2991 }
2992 if (_curchar != ')') {
2993 parse_err(SYNERR, "Expected ')' after encode parameters.\n");
2994 return;
2995 }
2996 }
2997 } // WHILE loop collecting parameters
2998 next_char(); // move past ')' at end of parameters
2999 } // done with parameter list for encoding
3001 // Check for ',' or ')' after encoding
3002 skipws(); // move to character after parameters
3003 if ( _curchar == ',' ) {
3004 // Found a ','
3005 next_char(); // move past ',' between encode methods
3006 skipws();
3007 }
3008 else if ( _curchar != ')' ) {
3009 // If not a ',' then only a ')' is allowed
3010 parse_err(SYNERR, "Expected ')' after encoding %s.\n", ec_name);
3011 return;
3012 }
3014 // Check for ',' separating parameters
3015 // if ( _curchar != ',' && _curchar != ')' ) {
3016 // parse_err(SYNERR, "expected ',' or ')' after encode method inside ins_encode.\n");
3017 // return NULL;
3018 // }
3020 } // done parsing ins_encode methods and their parameters
3021 if (_curchar != ')') {
3022 parse_err(SYNERR, "Missing ')' at end of ins_encode description.\n");
3023 return;
3024 }
3025 next_char(); // move past ')'
3026 skipws(); // Skip leading whitespace
3028 if ( _curchar != ';' ) {
3029 parse_err(SYNERR, "Missing ';' at end of ins_encode.\n");
3030 return;
3031 }
3032 next_char(); // move past ';'
3033 skipws(); // be friendly to oper_parse()
3035 // Check for duplicate ins_encode sections after parsing the block
3036 // so that parsing can continue and find any other errors.
3037 if (inst._insencode != NULL) {
3038 parse_err(SYNERR, "Multiple ins_encode sections defined\n");
3039 return;
3040 }
3042 // Debug Stuff
3043 if (_AD._adl_debug > 1) fprintf(stderr,"Instruction Encode: %s\n", ec_name);
3045 // Set encode class of this instruction.
3046 inst._insencode = encrule;
3047 }
3050 //------------------------------constant_parse---------------------------------
3051 // Parse a constant expression.
3052 void ADLParser::constant_parse(InstructForm& inst) {
3053 // Create a new encoding name based on the name of the instruction
3054 // definition, which should be unique.
3055 const char* prefix = "__constant_";
3056 char* ec_name = (char*) malloc(strlen(inst._ident) + strlen(prefix) + 1);
3057 sprintf(ec_name, "%s%s", prefix, inst._ident);
3059 assert(_AD._encode->encClass(ec_name) == NULL, "shouldn't already exist");
3060 EncClass* encoding = _AD._encode->add_EncClass(ec_name);
3061 encoding->_linenum = linenum();
3063 // synthesize the arguments list for the enc_class from the
3064 // arguments to the instruct definition.
3065 const char* param = NULL;
3066 inst._parameters.reset();
3067 while ((param = inst._parameters.iter()) != NULL) {
3068 OperandForm* opForm = (OperandForm*) inst._localNames[param];
3069 encoding->add_parameter(opForm->_ident, param);
3070 }
3072 // Parse the following ( ) expression.
3073 constant_parse_expression(encoding, ec_name);
3075 // Build an encoding rule which invokes the encoding rule we just
3076 // created, passing all arguments that we received.
3077 InsEncode* encrule = new InsEncode(); // Encode class for instruction
3078 NameAndList* params = encrule->add_encode(ec_name);
3079 inst._parameters.reset();
3080 while ((param = inst._parameters.iter()) != NULL) {
3081 params->add_entry(param);
3082 }
3084 // Set encode class of this instruction.
3085 inst._constant = encrule;
3086 }
3089 //------------------------------constant_parse_expression----------------------
3090 void ADLParser::constant_parse_expression(EncClass* encoding, char* ec_name) {
3091 skipws();
3093 // Prepend location descriptor, for debugging; cf. ADLParser::find_cpp_block
3094 if (_AD._adlocation_debug) {
3095 encoding->add_code(get_line_string());
3096 }
3098 // Start code line.
3099 encoding->add_code(" _constant = C->constant_table().add");
3101 // Parse everything in ( ) expression.
3102 encoding->add_code("(this, ");
3103 next_char(); // Skip '('
3104 int parens_depth = 1;
3106 // Collect the parts of the constant expression.
3107 // (1) strings that are passed through to output
3108 // (2) replacement/substitution variable, preceeded by a '$'
3109 while (parens_depth > 0) {
3110 if (_curchar == '(') {
3111 parens_depth++;
3112 encoding->add_code("(");
3113 next_char();
3114 }
3115 else if (_curchar == ')') {
3116 parens_depth--;
3117 if (parens_depth > 0)
3118 encoding->add_code(")");
3119 next_char();
3120 }
3121 else {
3122 // (1)
3123 // Check if there is a string to pass through to output
3124 char *start = _ptr; // Record start of the next string
3125 while ((_curchar != '$') && (_curchar != '(') && (_curchar != ')')) {
3126 next_char();
3127 }
3128 // If a string was found, terminate it and record in EncClass
3129 if (start != _ptr) {
3130 *_ptr = '\0'; // Terminate the string
3131 encoding->add_code(start);
3132 }
3134 // (2)
3135 // If we are at a replacement variable, copy it and record in EncClass.
3136 if (_curchar == '$') {
3137 // Found replacement Variable
3138 char* rep_var = get_rep_var_ident_dup();
3139 encoding->add_rep_var(rep_var);
3140 }
3141 }
3142 }
3144 // Finish code line.
3145 encoding->add_code(");");
3147 if (_AD._adlocation_debug) {
3148 encoding->add_code(end_line_marker());
3149 }
3151 // Debug Stuff
3152 if (_AD._adl_debug > 1) fprintf(stderr, "EncodingClass Form: %s\n", ec_name);
3153 }
3156 //------------------------------size_parse-----------------------------------
3157 char* ADLParser::size_parse(InstructForm *instr) {
3158 char* sizeOfInstr = NULL;
3160 // Get value of the instruction's size
3161 skipws();
3163 // Parse size
3164 sizeOfInstr = get_paren_expr("size expression");
3165 if (sizeOfInstr == NULL) {
3166 parse_err(SYNERR, "size of opcode expected at %c\n", _curchar);
3167 return NULL;
3168 }
3170 skipws();
3172 // Check for terminator
3173 if (_curchar != ';') {
3174 parse_err(SYNERR, "missing ';' in ins_attrib definition\n");
3175 return NULL;
3176 }
3177 next_char(); // Advance past the ';'
3178 skipws(); // necessary for instr_parse()
3180 // Debug Stuff
3181 if (_AD._adl_debug > 1) {
3182 if (sizeOfInstr != NULL) {
3183 fprintf(stderr,"size of opcode: %s\n", sizeOfInstr);
3184 }
3185 }
3187 return sizeOfInstr;
3188 }
3191 //------------------------------opcode_parse-----------------------------------
3192 Opcode * ADLParser::opcode_parse(InstructForm *instr) {
3193 char *primary = NULL;
3194 char *secondary = NULL;
3195 char *tertiary = NULL;
3197 char *val = NULL;
3198 Opcode *opcode = NULL;
3200 // Get value of the instruction's opcode
3201 skipws();
3202 if (_curchar != '(') { // Check for parenthesized operand list
3203 parse_err(SYNERR, "missing '(' in expand instruction declaration\n");
3204 return NULL;
3205 }
3206 next_char(); // skip open paren
3207 skipws();
3208 if (_curchar != ')') {
3209 // Parse primary, secondary, and tertiary opcodes, if provided.
3210 if ( ((primary = get_ident_or_literal_constant("primary opcode")) == NULL) ) {
3211 parse_err(SYNERR, "primary hex opcode expected at %c\n", _curchar);
3212 return NULL;
3213 }
3214 skipws();
3215 if (_curchar == ',') {
3216 next_char();
3217 skipws();
3218 // Parse secondary opcode
3219 if ( ((secondary = get_ident_or_literal_constant("secondary opcode")) == NULL) ) {
3220 parse_err(SYNERR, "secondary hex opcode expected at %c\n", _curchar);
3221 return NULL;
3222 }
3223 skipws();
3224 if (_curchar == ',') {
3225 next_char();
3226 skipws();
3227 // Parse tertiary opcode
3228 if ( ((tertiary = get_ident_or_literal_constant("tertiary opcode")) == NULL) ) {
3229 parse_err(SYNERR,"tertiary hex opcode expected at %c\n", _curchar);
3230 return NULL;
3231 }
3232 skipws();
3233 }
3234 }
3235 skipws();
3236 if (_curchar != ')') {
3237 parse_err(SYNERR, "Missing ')' in opcode description\n");
3238 return NULL;
3239 }
3240 }
3241 next_char(); // Skip ')'
3242 skipws();
3243 // Check for terminator
3244 if (_curchar != ';') {
3245 parse_err(SYNERR, "missing ';' in ins_attrib definition\n");
3246 return NULL;
3247 }
3248 next_char(); // Advance past the ';'
3249 skipws(); // necessary for instr_parse()
3251 // Debug Stuff
3252 if (_AD._adl_debug > 1) {
3253 if (primary != NULL) fprintf(stderr,"primary opcode: %s\n", primary);
3254 if (secondary != NULL) fprintf(stderr,"secondary opcode: %s\n", secondary);
3255 if (tertiary != NULL) fprintf(stderr,"tertiary opcode: %s\n", tertiary);
3256 }
3258 // Generate new object and return
3259 opcode = new Opcode(primary, secondary, tertiary);
3260 return opcode;
3261 }
3264 //------------------------------interface_parse--------------------------------
3265 Interface *ADLParser::interface_parse(void) {
3266 char *iface_name = NULL; // Name of interface class being used
3267 char *iface_code = NULL; // Describe components of this class
3269 // Get interface class name
3270 skipws(); // Skip whitespace
3271 if (_curchar != '(') {
3272 parse_err(SYNERR, "Missing '(' at start of interface description.\n");
3273 return NULL;
3274 }
3275 next_char(); // move past '('
3276 skipws();
3277 iface_name = get_ident();
3278 if (iface_name == NULL) {
3279 parse_err(SYNERR, "missing interface name after 'interface'.\n");
3280 return NULL;
3281 }
3282 skipws();
3283 if (_curchar != ')') {
3284 parse_err(SYNERR, "Missing ')' after name of interface.\n");
3285 return NULL;
3286 }
3287 next_char(); // move past ')'
3289 // Get details of the interface,
3290 // for the type of interface indicated by iface_name.
3291 Interface *inter = NULL;
3292 skipws();
3293 if ( _curchar != ';' ) {
3294 if ( strcmp(iface_name,"MEMORY_INTER") == 0 ) {
3295 inter = mem_interface_parse();
3296 }
3297 else if ( strcmp(iface_name,"COND_INTER") == 0 ) {
3298 inter = cond_interface_parse();
3299 }
3300 // The parse routines consume the "%}"
3302 // Check for probable extra ';' after defining block.
3303 if ( _curchar == ';' ) {
3304 parse_err(SYNERR, "Extra ';' after defining interface block.\n");
3305 next_char(); // Skip ';'
3306 return NULL;
3307 }
3308 } else {
3309 next_char(); // move past ';'
3311 // Create appropriate interface object
3312 if ( strcmp(iface_name,"REG_INTER") == 0 ) {
3313 inter = new RegInterface();
3314 }
3315 else if ( strcmp(iface_name,"CONST_INTER") == 0 ) {
3316 inter = new ConstInterface();
3317 }
3318 }
3319 skipws(); // be friendly to oper_parse()
3320 // Debug Stuff
3321 if (_AD._adl_debug > 1) fprintf(stderr,"Interface Form: %s\n", iface_name);
3323 // Create appropriate interface object and return.
3324 return inter;
3325 }
3328 //------------------------------mem_interface_parse----------------------------
3329 Interface *ADLParser::mem_interface_parse(void) {
3330 // Fields for MemInterface
3331 char *base = NULL;
3332 char *index = NULL;
3333 char *scale = NULL;
3334 char *disp = NULL;
3336 if (_curchar != '%') {
3337 parse_err(SYNERR, "Missing '%{' for 'interface' block.\n");
3338 return NULL;
3339 }
3340 next_char(); // Skip '%'
3341 if (_curchar != '{') {
3342 parse_err(SYNERR, "Missing '%{' for 'interface' block.\n");
3343 return NULL;
3344 }
3345 next_char(); // Skip '{'
3346 skipws();
3347 do {
3348 char *field = get_ident();
3349 if (field == NULL) {
3350 parse_err(SYNERR, "Expected keyword, base|index|scale|disp, or '%}' ending interface.\n");
3351 return NULL;
3352 }
3353 if ( strcmp(field,"base") == 0 ) {
3354 base = interface_field_parse();
3355 }
3356 else if ( strcmp(field,"index") == 0 ) {
3357 index = interface_field_parse();
3358 }
3359 else if ( strcmp(field,"scale") == 0 ) {
3360 scale = interface_field_parse();
3361 }
3362 else if ( strcmp(field,"disp") == 0 ) {
3363 disp = interface_field_parse();
3364 }
3365 else {
3366 parse_err(SYNERR, "Expected keyword, base|index|scale|disp, or '%}' ending interface.\n");
3367 return NULL;
3368 }
3369 } while( _curchar != '%' );
3370 next_char(); // Skip '%'
3371 if ( _curchar != '}' ) {
3372 parse_err(SYNERR, "Missing '%}' for 'interface' block.\n");
3373 return NULL;
3374 }
3375 next_char(); // Skip '}'
3377 // Construct desired object and return
3378 Interface *inter = new MemInterface(base, index, scale, disp);
3379 return inter;
3380 }
3383 //------------------------------cond_interface_parse---------------------------
3384 Interface *ADLParser::cond_interface_parse(void) {
3385 char *equal;
3386 char *not_equal;
3387 char *less;
3388 char *greater_equal;
3389 char *less_equal;
3390 char *greater;
3391 const char *equal_format = "eq";
3392 const char *not_equal_format = "ne";
3393 const char *less_format = "lt";
3394 const char *greater_equal_format = "ge";
3395 const char *less_equal_format = "le";
3396 const char *greater_format = "gt";
3398 if (_curchar != '%') {
3399 parse_err(SYNERR, "Missing '%{' for 'cond_interface' block.\n");
3400 return NULL;
3401 }
3402 next_char(); // Skip '%'
3403 if (_curchar != '{') {
3404 parse_err(SYNERR, "Missing '%{' for 'cond_interface' block.\n");
3405 return NULL;
3406 }
3407 next_char(); // Skip '{'
3408 skipws();
3409 do {
3410 char *field = get_ident();
3411 if (field == NULL) {
3412 parse_err(SYNERR, "Expected keyword, base|index|scale|disp, or '%}' ending interface.\n");
3413 return NULL;
3414 }
3415 if ( strcmp(field,"equal") == 0 ) {
3416 equal = interface_field_parse(&equal_format);
3417 }
3418 else if ( strcmp(field,"not_equal") == 0 ) {
3419 not_equal = interface_field_parse(¬_equal_format);
3420 }
3421 else if ( strcmp(field,"less") == 0 ) {
3422 less = interface_field_parse(&less_format);
3423 }
3424 else if ( strcmp(field,"greater_equal") == 0 ) {
3425 greater_equal = interface_field_parse(&greater_equal_format);
3426 }
3427 else if ( strcmp(field,"less_equal") == 0 ) {
3428 less_equal = interface_field_parse(&less_equal_format);
3429 }
3430 else if ( strcmp(field,"greater") == 0 ) {
3431 greater = interface_field_parse(&greater_format);
3432 }
3433 else {
3434 parse_err(SYNERR, "Expected keyword, base|index|scale|disp, or '%}' ending interface.\n");
3435 return NULL;
3436 }
3437 } while( _curchar != '%' );
3438 next_char(); // Skip '%'
3439 if ( _curchar != '}' ) {
3440 parse_err(SYNERR, "Missing '%}' for 'interface' block.\n");
3441 return NULL;
3442 }
3443 next_char(); // Skip '}'
3445 // Construct desired object and return
3446 Interface *inter = new CondInterface(equal, equal_format,
3447 not_equal, not_equal_format,
3448 less, less_format,
3449 greater_equal, greater_equal_format,
3450 less_equal, less_equal_format,
3451 greater, greater_format);
3452 return inter;
3453 }
3456 //------------------------------interface_field_parse--------------------------
3457 char *ADLParser::interface_field_parse(const char ** format) {
3458 char *iface_field = NULL;
3460 // Get interface field
3461 skipws(); // Skip whitespace
3462 if (_curchar != '(') {
3463 parse_err(SYNERR, "Missing '(' at start of interface field.\n");
3464 return NULL;
3465 }
3466 next_char(); // move past '('
3467 skipws();
3468 if ( _curchar != '0' && _curchar != '$' ) {
3469 parse_err(SYNERR, "missing or invalid interface field contents.\n");
3470 return NULL;
3471 }
3472 iface_field = get_rep_var_ident();
3473 if (iface_field == NULL) {
3474 parse_err(SYNERR, "missing or invalid interface field contents.\n");
3475 return NULL;
3476 }
3477 skipws();
3478 if (format != NULL && _curchar == ',') {
3479 next_char();
3480 skipws();
3481 if (_curchar != '"') {
3482 parse_err(SYNERR, "Missing '\"' in field format .\n");
3483 return NULL;
3484 }
3485 next_char();
3486 char *start = _ptr; // Record start of the next string
3487 while ((_curchar != '"') && (_curchar != '%') && (_curchar != '\n')) {
3488 if (_curchar == '\\') next_char(); // superquote
3489 if (_curchar == '\n') parse_err(SYNERR, "newline in string"); // unimplemented!
3490 next_char();
3491 }
3492 if (_curchar != '"') {
3493 parse_err(SYNERR, "Missing '\"' at end of field format .\n");
3494 return NULL;
3495 }
3496 // If a string was found, terminate it and record in FormatRule
3497 if ( start != _ptr ) {
3498 *_ptr = '\0'; // Terminate the string
3499 *format = start;
3500 }
3501 next_char();
3502 skipws();
3503 }
3504 if (_curchar != ')') {
3505 parse_err(SYNERR, "Missing ')' after interface field.\n");
3506 return NULL;
3507 }
3508 next_char(); // move past ')'
3509 skipws();
3510 if ( _curchar != ';' ) {
3511 parse_err(SYNERR, "Missing ';' at end of interface field.\n");
3512 return NULL;
3513 }
3514 next_char(); // move past ';'
3515 skipws(); // be friendly to interface_parse()
3517 return iface_field;
3518 }
3521 //------------------------------match_parse------------------------------------
3522 MatchRule *ADLParser::match_parse(FormDict &operands) {
3523 MatchRule *match; // Match Rule class for instruction/operand
3524 char *cnstr = NULL; // Code for constructor
3525 int depth = 0; // Counter for matching parentheses
3526 int numleaves = 0; // Counter for number of leaves in rule
3528 // Parse the match rule tree
3529 MatchNode *mnode = matchNode_parse(operands, depth, numleaves, true);
3531 // Either there is a block with a constructor, or a ';' here
3532 skipws(); // Skip whitespace
3533 if ( _curchar == ';' ) { // Semicolon is valid terminator
3534 cnstr = NULL; // no constructor for this form
3535 next_char(); // Move past the ';', replaced with '\0'
3536 }
3537 else if ((cnstr = find_cpp_block("match constructor")) == NULL ) {
3538 parse_err(SYNERR, "invalid construction of match rule\n"
3539 "Missing ';' or invalid '%{' and '%}' constructor\n");
3540 return NULL; // No MatchRule to return
3541 }
3542 if (_AD._adl_debug > 1)
3543 if (cnstr) fprintf(stderr,"Match Constructor: %s\n", cnstr);
3544 // Build new MatchRule object
3545 match = new MatchRule(_AD, mnode, depth, cnstr, numleaves);
3546 skipws(); // Skip any trailing whitespace
3547 return match; // Return MatchRule object
3548 }
3550 //------------------------------format_parse-----------------------------------
3551 FormatRule* ADLParser::format_parse(void) {
3552 char *desc = NULL;
3553 FormatRule *format = (new FormatRule(desc));
3555 // Without expression form, MUST have a code block;
3556 skipws(); // Skip whitespace
3557 if ( _curchar == ';' ) { // Semicolon is valid terminator
3558 desc = NULL; // no constructor for this form
3559 next_char(); // Move past the ';', replaced with '\0'
3560 }
3561 else if ( _curchar == '%' && *(_ptr+1) == '{') {
3562 next_char(); // Move past the '%'
3563 next_char(); // Move past the '{'
3565 skipws();
3566 if (_curchar == '$') {
3567 char* ident = get_rep_var_ident();
3568 if (strcmp(ident, "$$template") == 0) return template_parse();
3569 parse_err(SYNERR, "Unknown \"%s\" directive in format", ident);
3570 return NULL;
3571 }
3572 // Check for the opening '"' inside the format description
3573 if ( _curchar == '"' ) {
3574 next_char(); // Move past the initial '"'
3575 if( _curchar == '"' ) { // Handle empty format string case
3576 *_ptr = '\0'; // Terminate empty string
3577 format->_strings.addName(_ptr);
3578 }
3580 // Collect the parts of the format description
3581 // (1) strings that are passed through to tty->print
3582 // (2) replacement/substitution variable, preceeded by a '$'
3583 // (3) multi-token ANSIY C style strings
3584 while ( true ) {
3585 if ( _curchar == '%' || _curchar == '\n' ) {
3586 if ( _curchar != '"' ) {
3587 parse_err(SYNERR, "missing '\"' at end of format block");
3588 return NULL;
3589 }
3590 }
3592 // (1)
3593 // Check if there is a string to pass through to output
3594 char *start = _ptr; // Record start of the next string
3595 while ((_curchar != '$') && (_curchar != '"') && (_curchar != '%') && (_curchar != '\n')) {
3596 if (_curchar == '\\') {
3597 next_char(); // superquote
3598 if ((_curchar == '$') || (_curchar == '%'))
3599 // hack to avoid % escapes and warnings about undefined \ escapes
3600 *(_ptr-1) = _curchar;
3601 }
3602 if (_curchar == '\n') parse_err(SYNERR, "newline in string"); // unimplemented!
3603 next_char();
3604 }
3605 // If a string was found, terminate it and record in FormatRule
3606 if ( start != _ptr ) {
3607 *_ptr = '\0'; // Terminate the string
3608 format->_strings.addName(start);
3609 }
3611 // (2)
3612 // If we are at a replacement variable,
3613 // copy it and record in FormatRule
3614 if ( _curchar == '$' ) {
3615 next_char(); // Move past the '$'
3616 char* rep_var = get_ident(); // Nil terminate the variable name
3617 rep_var = strdup(rep_var);// Copy the string
3618 *_ptr = _curchar; // and replace Nil with original character
3619 format->_rep_vars.addName(rep_var);
3620 // Add flag to _strings list indicating we should check _rep_vars
3621 format->_strings.addName(NameList::_signal);
3622 }
3624 // (3)
3625 // Allow very long strings to be broken up,
3626 // using the ANSI C syntax "foo\n" <newline> "bar"
3627 if ( _curchar == '"') {
3628 next_char(); // Move past the '"'
3629 skipws(); // Skip white space before next string token
3630 if ( _curchar != '"') {
3631 break;
3632 } else {
3633 // Found one. Skip both " and the whitespace in between.
3634 next_char();
3635 }
3636 }
3637 } // end while part of format description
3639 // Check for closing '"' and '%}' in format description
3640 skipws(); // Move to closing '%}'
3641 if ( _curchar != '%' ) {
3642 parse_err(SYNERR, "non-blank characters between closing '\"' and '%' in format");
3643 return NULL;
3644 }
3645 } // Done with format description inside
3647 skipws();
3648 // Past format description, at '%'
3649 if ( _curchar != '%' || *(_ptr+1) != '}' ) {
3650 parse_err(SYNERR, "missing '%}' at end of format block");
3651 return NULL;
3652 }
3653 next_char(); // Move past the '%'
3654 next_char(); // Move past the '}'
3655 }
3656 else { // parameter list alone must terminate with a ';'
3657 parse_err(SYNERR, "missing ';' after Format expression");
3658 return NULL;
3659 }
3660 // Debug Stuff
3661 if (_AD._adl_debug > 1) fprintf(stderr,"Format Rule: %s\n", desc);
3663 skipws();
3664 return format;
3665 }
3668 //------------------------------template_parse-----------------------------------
3669 FormatRule* ADLParser::template_parse(void) {
3670 char *desc = NULL;
3671 FormatRule *format = (new FormatRule(desc));
3673 skipws();
3674 while ( (_curchar != '%') && (*(_ptr+1) != '}') ) {
3676 // (1)
3677 // Check if there is a string to pass through to output
3678 {
3679 char *start = _ptr; // Record start of the next string
3680 while ((_curchar != '$') && ((_curchar != '%') || (*(_ptr+1) != '}')) ) {
3681 // If at the start of a comment, skip past it
3682 if( (_curchar == '/') && ((*(_ptr+1) == '/') || (*(_ptr+1) == '*')) ) {
3683 skipws_no_preproc();
3684 } else {
3685 // ELSE advance to the next character, or start of the next line
3686 next_char_or_line();
3687 }
3688 }
3689 // If a string was found, terminate it and record in EncClass
3690 if ( start != _ptr ) {
3691 *_ptr = '\0'; // Terminate the string
3692 // Add flag to _strings list indicating we should check _rep_vars
3693 format->_strings.addName(NameList::_signal2);
3694 format->_strings.addName(start);
3695 }
3696 }
3698 // (2)
3699 // If we are at a replacement variable,
3700 // copy it and record in EncClass
3701 if ( _curchar == '$' ) {
3702 // Found replacement Variable
3703 char *rep_var = get_rep_var_ident_dup();
3704 if (strcmp(rep_var, "$emit") == 0) {
3705 // switch to normal format parsing
3706 next_char();
3707 next_char();
3708 skipws();
3709 // Check for the opening '"' inside the format description
3710 if ( _curchar == '"' ) {
3711 next_char(); // Move past the initial '"'
3712 if( _curchar == '"' ) { // Handle empty format string case
3713 *_ptr = '\0'; // Terminate empty string
3714 format->_strings.addName(_ptr);
3715 }
3717 // Collect the parts of the format description
3718 // (1) strings that are passed through to tty->print
3719 // (2) replacement/substitution variable, preceeded by a '$'
3720 // (3) multi-token ANSIY C style strings
3721 while ( true ) {
3722 if ( _curchar == '%' || _curchar == '\n' ) {
3723 parse_err(SYNERR, "missing '\"' at end of format block");
3724 return NULL;
3725 }
3727 // (1)
3728 // Check if there is a string to pass through to output
3729 char *start = _ptr; // Record start of the next string
3730 while ((_curchar != '$') && (_curchar != '"') && (_curchar != '%') && (_curchar != '\n')) {
3731 if (_curchar == '\\') next_char(); // superquote
3732 if (_curchar == '\n') parse_err(SYNERR, "newline in string"); // unimplemented!
3733 next_char();
3734 }
3735 // If a string was found, terminate it and record in FormatRule
3736 if ( start != _ptr ) {
3737 *_ptr = '\0'; // Terminate the string
3738 format->_strings.addName(start);
3739 }
3741 // (2)
3742 // If we are at a replacement variable,
3743 // copy it and record in FormatRule
3744 if ( _curchar == '$' ) {
3745 next_char(); // Move past the '$'
3746 char* next_rep_var = get_ident(); // Nil terminate the variable name
3747 next_rep_var = strdup(next_rep_var);// Copy the string
3748 *_ptr = _curchar; // and replace Nil with original character
3749 format->_rep_vars.addName(next_rep_var);
3750 // Add flag to _strings list indicating we should check _rep_vars
3751 format->_strings.addName(NameList::_signal);
3752 }
3754 // (3)
3755 // Allow very long strings to be broken up,
3756 // using the ANSI C syntax "foo\n" <newline> "bar"
3757 if ( _curchar == '"') {
3758 next_char(); // Move past the '"'
3759 skipws(); // Skip white space before next string token
3760 if ( _curchar != '"') {
3761 break;
3762 } else {
3763 // Found one. Skip both " and the whitespace in between.
3764 next_char();
3765 }
3766 }
3767 } // end while part of format description
3768 }
3769 } else {
3770 // Add flag to _strings list indicating we should check _rep_vars
3771 format->_rep_vars.addName(rep_var);
3772 // Add flag to _strings list indicating we should check _rep_vars
3773 format->_strings.addName(NameList::_signal3);
3774 }
3775 } // end while part of format description
3776 }
3778 skipws();
3779 // Past format description, at '%'
3780 if ( _curchar != '%' || *(_ptr+1) != '}' ) {
3781 parse_err(SYNERR, "missing '%}' at end of format block");
3782 return NULL;
3783 }
3784 next_char(); // Move past the '%'
3785 next_char(); // Move past the '}'
3787 // Debug Stuff
3788 if (_AD._adl_debug > 1) fprintf(stderr,"Format Rule: %s\n", desc);
3790 skipws();
3791 return format;
3792 }
3795 //------------------------------effect_parse-----------------------------------
3796 void ADLParser::effect_parse(InstructForm *instr) {
3797 char* desc = NULL;
3799 skipws(); // Skip whitespace
3800 if (_curchar != '(') {
3801 parse_err(SYNERR, "missing '(' in effect definition\n");
3802 return;
3803 }
3804 // Get list of effect-operand pairs and insert into dictionary
3805 else get_effectlist(instr->_effects, instr->_localNames, instr->_has_call);
3807 // Debug Stuff
3808 if (_AD._adl_debug > 1) fprintf(stderr,"Effect description: %s\n", desc);
3809 if (_curchar != ';') {
3810 parse_err(SYNERR, "missing ';' in Effect definition\n");
3811 }
3812 next_char(); // Skip ';'
3814 }
3816 //------------------------------expand_parse-----------------------------------
3817 ExpandRule* ADLParser::expand_parse(InstructForm *instr) {
3818 char *ident, *ident2;
3819 OperandForm *oper;
3820 InstructForm *ins;
3821 NameAndList *instr_and_operands = NULL;
3822 ExpandRule *exp = new ExpandRule();
3824 // Expand is a block containing an ordered list of instructions, each of
3825 // which has an ordered list of operands.
3826 // Check for block delimiter
3827 skipws(); // Skip leading whitespace
3828 if ((_curchar != '%')
3829 || (next_char(), (_curchar != '{')) ) { // If not open block
3830 parse_err(SYNERR, "missing '%{' in expand definition\n");
3831 return(NULL);
3832 }
3833 next_char(); // Maintain the invariant
3834 do {
3835 ident = get_ident(); // Grab next identifier
3836 if (ident == NULL) {
3837 parse_err(SYNERR, "identifier expected at %c\n", _curchar);
3838 continue;
3839 } // Check that you have a valid instruction
3840 const Form *form = _globalNames[ident];
3841 ins = form ? form->is_instruction() : NULL;
3842 if (ins == NULL) {
3843 // This is a new operand
3844 oper = form ? form->is_operand() : NULL;
3845 if (oper == NULL) {
3846 parse_err(SYNERR, "instruction/operand name expected at %s\n", ident);
3847 continue;
3848 }
3849 // Throw the operand on the _newopers list
3850 skipws();
3851 ident = get_unique_ident(instr->_localNames,"Operand");
3852 if (ident == NULL) {
3853 parse_err(SYNERR, "identifier expected at %c\n", _curchar);
3854 continue;
3855 }
3856 exp->_newopers.addName(ident);
3857 // Add new operand to LocalNames
3858 instr->_localNames.Insert(ident, oper);
3859 // Grab any constructor code and save as a string
3860 char *c = NULL;
3861 skipws();
3862 if (_curchar == '%') { // Need a constructor for the operand
3863 c = find_cpp_block("Operand Constructor");
3864 if (c == NULL) {
3865 parse_err(SYNERR, "Invalid code block for operand constructor\n", _curchar);
3866 continue;
3867 }
3868 // Add constructor to _newopconst Dict
3869 exp->_newopconst.Insert(ident, c);
3870 }
3871 else if (_curchar != ';') { // If no constructor, need a ;
3872 parse_err(SYNERR, "Missing ; in expand rule operand declaration\n");
3873 continue;
3874 }
3875 else next_char(); // Skip the ;
3876 skipws();
3877 }
3878 else {
3879 // Add instruction to list
3880 instr_and_operands = new NameAndList(ident);
3881 // Grab operands, build nameList of them, and then put into dictionary
3882 skipws();
3883 if (_curchar != '(') { // Check for parenthesized operand list
3884 parse_err(SYNERR, "missing '(' in expand instruction declaration\n");
3885 continue;
3886 }
3887 do {
3888 next_char(); // skip open paren & comma characters
3889 skipws();
3890 if (_curchar == ')') break;
3891 ident2 = get_ident();
3892 skipws();
3893 if (ident2 == NULL) {
3894 parse_err(SYNERR, "identifier expected at %c\n", _curchar);
3895 continue;
3896 } // Check that you have a valid operand
3897 const Form *form2 = instr->_localNames[ident2];
3898 if (!form2) {
3899 parse_err(SYNERR, "operand name expected at %s\n", ident2);
3900 continue;
3901 }
3902 oper = form2->is_operand();
3903 if (oper == NULL && !form2->is_opclass()) {
3904 parse_err(SYNERR, "operand name expected at %s\n", ident2);
3905 continue;
3906 } // Add operand to list
3907 instr_and_operands->add_entry(ident2);
3908 } while(_curchar == ',');
3909 if (_curchar != ')') {
3910 parse_err(SYNERR, "missing ')'in expand instruction declaration\n");
3911 continue;
3912 }
3913 next_char();
3914 if (_curchar != ';') {
3915 parse_err(SYNERR, "missing ';'in expand instruction declaration\n");
3916 continue;
3917 }
3918 next_char();
3920 // Record both instruction name and its operand list
3921 exp->add_instruction(instr_and_operands);
3923 skipws();
3924 }
3926 } while(_curchar != '%');
3927 next_char();
3928 if (_curchar != '}') {
3929 parse_err(SYNERR, "missing '%}' in expand rule definition\n");
3930 return(NULL);
3931 }
3932 next_char();
3934 // Debug Stuff
3935 if (_AD._adl_debug > 1) fprintf(stderr,"Expand Rule:\n");
3937 skipws();
3938 return (exp);
3939 }
3941 //------------------------------rewrite_parse----------------------------------
3942 RewriteRule* ADLParser::rewrite_parse(void) {
3943 char* params = NULL;
3944 char* desc = NULL;
3947 // This feature targeted for second generation description language.
3949 skipws(); // Skip whitespace
3950 // Get parameters for rewrite
3951 if ((params = get_paren_expr("rewrite parameters")) == NULL) {
3952 parse_err(SYNERR, "missing '(' in rewrite rule\n");
3953 return NULL;
3954 }
3955 // Debug Stuff
3956 if (_AD._adl_debug > 1) fprintf(stderr,"Rewrite parameters: %s\n", params);
3958 // For now, grab entire block;
3959 skipws();
3960 if ( (desc = find_cpp_block("rewrite block")) == NULL ) {
3961 parse_err(SYNERR, "incorrect or missing block for 'rewrite'.\n");
3962 return NULL;
3963 }
3964 // Debug Stuff
3965 if (_AD._adl_debug > 1) fprintf(stderr,"Rewrite Rule: %s\n", desc);
3967 skipws();
3968 return (new RewriteRule(params,desc));
3969 }
3971 //------------------------------attr_parse-------------------------------------
3972 Attribute *ADLParser::attr_parse(char* ident) {
3973 Attribute *attrib; // Attribute class
3974 char *cost = NULL; // String representation of cost attribute
3976 skipws(); // Skip leading whitespace
3977 if ( (cost = get_paren_expr("attribute")) == NULL ) {
3978 parse_err(SYNERR, "incorrect or missing expression for 'attribute'\n");
3979 return NULL;
3980 }
3981 // Debug Stuff
3982 if (_AD._adl_debug > 1) fprintf(stderr,"Attribute: %s\n", cost);
3983 if (_curchar != ';') {
3984 parse_err(SYNERR, "missing ';' in attribute definition\n");
3985 return NULL;
3986 }
3987 next_char(); // Point after the terminator
3989 skipws();
3990 attrib = new Attribute(ident,cost,INS_ATTR); // Build new predicate object
3991 return attrib;
3992 }
3995 //------------------------------matchNode_parse--------------------------------
3996 MatchNode *ADLParser::matchNode_parse(FormDict &operands, int &depth, int &numleaves, bool atroot) {
3997 // Count depth of parenthesis nesting for both left and right children
3998 int lParens = depth;
3999 int rParens = depth;
4001 // MatchNode objects for left, right, and root of subtree.
4002 MatchNode *lChild = NULL;
4003 MatchNode *rChild = NULL;
4004 char *token; // Identifier which may be opcode or operand
4006 // Match expression starts with a '('
4007 if (cur_char() != '(')
4008 return NULL;
4010 next_char(); // advance past '('
4012 // Parse the opcode
4013 token = get_ident(); // Get identifier, opcode
4014 if (token == NULL) {
4015 parse_err(SYNERR, "missing opcode in match expression\n");
4016 return NULL;
4017 }
4019 // Take note if we see one of a few special operations - those that are
4020 // treated differently on different architectures in the sense that on
4021 // one architecture there is a match rule and on another there isn't (so
4022 // a call will eventually be generated).
4024 for (int i = _last_machine_leaf + 1; i < _last_opcode; i++) {
4025 if (strcmp(token, NodeClassNames[i]) == 0) {
4026 _AD.has_match_rule(i, true);
4027 }
4028 }
4030 // Lookup the root value in the operands dict to perform substitution
4031 const char *result = NULL; // Result type will be filled in later
4032 const char *name = token; // local name associated with this node
4033 const char *operation = token; // remember valid operation for later
4034 const Form *form = operands[token];
4035 OpClassForm *opcForm = form ? form->is_opclass() : NULL;
4036 if (opcForm != NULL) {
4037 // If this token is an entry in the local names table, record its type
4038 if (!opcForm->ideal_only()) {
4039 operation = opcForm->_ident;
4040 result = operation; // Operands result in their own type
4041 }
4042 // Otherwise it is an ideal type, and so, has no local name
4043 else name = NULL;
4044 }
4046 // Parse the operands
4047 skipws();
4048 if (cur_char() != ')') {
4050 // Parse the left child
4051 if (strcmp(operation,"Set"))
4052 lChild = matchChild_parse(operands, lParens, numleaves, false);
4053 else
4054 lChild = matchChild_parse(operands, lParens, numleaves, true);
4056 skipws();
4057 if (cur_char() != ')' ) {
4058 if(strcmp(operation, "Set"))
4059 rChild = matchChild_parse(operands,rParens,numleaves,false);
4060 else
4061 rChild = matchChild_parse(operands,rParens,numleaves,true);
4062 }
4063 }
4065 // Check for required ')'
4066 skipws();
4067 if (cur_char() != ')') {
4068 parse_err(SYNERR, "missing ')' in match expression\n");
4069 return NULL;
4070 }
4071 next_char(); // skip the ')'
4073 MatchNode* mroot = new MatchNode(_AD,result,name,operation,lChild,rChild);
4075 // If not the root, reduce this subtree to an internal operand
4076 if (!atroot) {
4077 mroot->build_internalop();
4078 }
4079 // depth is greater of left and right paths.
4080 depth = (lParens > rParens) ? lParens : rParens;
4082 return mroot;
4083 }
4086 //------------------------------matchChild_parse-------------------------------
4087 MatchNode *ADLParser::matchChild_parse(FormDict &operands, int &parens, int &numleaves, bool atroot) {
4088 MatchNode *child = NULL;
4089 const char *result = NULL;
4090 const char *token = NULL;
4091 const char *opType = NULL;
4093 if (cur_char() == '(') { // child is an operation
4094 ++parens;
4095 child = matchNode_parse(operands, parens, numleaves, atroot);
4096 }
4097 else { // child is an operand
4098 token = get_ident();
4099 const Form *form = operands[token];
4100 OpClassForm *opcForm = form ? form->is_opclass() : NULL;
4101 if (opcForm != NULL) {
4102 opType = opcForm->_ident;
4103 result = opcForm->_ident; // an operand's result matches its type
4104 } else {
4105 parse_err(SYNERR, "undefined operand %s in match rule\n", token);
4106 return NULL;
4107 }
4109 if (opType == NULL) {
4110 parse_err(SYNERR, "missing type for argument '%s'\n", token);
4111 }
4113 child = new MatchNode(_AD, result, token, opType);
4114 ++numleaves;
4115 }
4117 return child;
4118 }
4122 // ******************** Private Utility Functions *************************
4125 char* ADLParser::find_cpp_block(const char* description) {
4126 char *next; // Pointer for finding block delimiters
4127 char* cppBlock = NULL; // Beginning of C++ code block
4129 if (_curchar == '%') { // Encoding is a C++ expression
4130 next_char();
4131 if (_curchar != '{') {
4132 parse_err(SYNERR, "missing '{' in %s \n", description);
4133 return NULL;
4134 }
4135 next_char(); // Skip block delimiter
4136 skipws_no_preproc(); // Skip leading whitespace
4137 cppBlock = _ptr; // Point to start of expression
4138 int line = linenum();
4139 next = _ptr + 1;
4140 while(((_curchar != '%') || (*next != '}')) && (_curchar != '\0')) {
4141 next_char_or_line();
4142 next = _ptr+1; // Maintain the next pointer
4143 } // Grab string
4144 if (_curchar == '\0') {
4145 parse_err(SYNERR, "invalid termination of %s \n", description);
4146 return NULL;
4147 }
4148 *_ptr = '\0'; // Terminate string
4149 _ptr += 2; // Skip block delimiter
4150 _curchar = *_ptr; // Maintain invariant
4152 // Prepend location descriptor, for debugging.
4153 if (_AD._adlocation_debug) {
4154 char* location = get_line_string(line);
4155 char* end_loc = end_line_marker();
4156 char* result = (char *)malloc(strlen(location) + strlen(cppBlock) + strlen(end_loc) + 1);
4157 strcpy(result, location);
4158 strcat(result, cppBlock);
4159 strcat(result, end_loc);
4160 cppBlock = result;
4161 free(location);
4162 }
4163 }
4165 return cppBlock;
4166 }
4168 // Move to the closing token of the expression we are currently at,
4169 // as defined by stop_chars. Match parens and quotes.
4170 char* ADLParser::get_expr(const char *desc, const char *stop_chars) {
4171 char* expr = NULL;
4172 int paren = 0;
4174 expr = _ptr;
4175 while (paren > 0 || !strchr(stop_chars, _curchar)) {
4176 if (_curchar == '(') { // Down level of nesting
4177 paren++; // Bump the parenthesis counter
4178 next_char(); // maintain the invariant
4179 }
4180 else if (_curchar == ')') { // Up one level of nesting
4181 if (paren == 0) {
4182 // Paren underflow: We didn't encounter the required stop-char.
4183 parse_err(SYNERR, "too many )'s, did not find %s after %s\n",
4184 stop_chars, desc);
4185 return NULL;
4186 }
4187 paren--; // Drop the parenthesis counter
4188 next_char(); // Maintain the invariant
4189 }
4190 else if (_curchar == '"' || _curchar == '\'') {
4191 int qchar = _curchar;
4192 while (true) {
4193 next_char();
4194 if (_curchar == qchar) { next_char(); break; }
4195 if (_curchar == '\\') next_char(); // superquote
4196 if (_curchar == '\n' || _curchar == '\0') {
4197 parse_err(SYNERR, "newline in string in %s\n", desc);
4198 return NULL;
4199 }
4200 }
4201 }
4202 else if (_curchar == '%' && (_ptr[1] == '{' || _ptr[1] == '}')) {
4203 // Make sure we do not stray into the next ADLC-level form.
4204 parse_err(SYNERR, "unexpected %%%c in %s\n", _ptr[1], desc);
4205 return NULL;
4206 }
4207 else if (_curchar == '\0') {
4208 parse_err(SYNERR, "unexpected EOF in %s\n", desc);
4209 return NULL;
4210 }
4211 else {
4212 // Always walk over whitespace, comments, preprocessor directives, etc.
4213 char* pre_skip_ptr = _ptr;
4214 skipws();
4215 // If the parser declined to make progress on whitespace,
4216 // skip the next character, which is therefore NOT whitespace.
4217 if (pre_skip_ptr == _ptr) {
4218 next_char();
4219 } else if (pre_skip_ptr+strlen(pre_skip_ptr) != _ptr+strlen(_ptr)) {
4220 parse_err(SYNERR, "unimplemented: preprocessor must not elide subexpression in %s", desc);
4221 }
4222 }
4223 }
4225 assert(strchr(stop_chars, _curchar), "non-null return must be at stop-char");
4226 *_ptr = '\0'; // Replace ')' or other stop-char with '\0'
4227 return expr;
4228 }
4230 // Helper function around get_expr
4231 // Sets _curchar to '(' so that get_paren_expr will search for a matching ')'
4232 char *ADLParser::get_paren_expr(const char *description, bool include_location) {
4233 int line = linenum();
4234 if (_curchar != '(') // Escape if not valid starting position
4235 return NULL;
4236 next_char(); // Skip the required initial paren.
4237 char *token2 = get_expr(description, ")");
4238 if (_curchar == ')')
4239 next_char(); // Skip required final paren.
4240 int junk = 0;
4241 if (include_location && _AD._adlocation_debug && !is_int_token(token2, junk)) {
4242 // Prepend location descriptor, for debugging.
4243 char* location = get_line_string(line);
4244 char* end_loc = end_line_marker();
4245 char* result = (char *)malloc(strlen(location) + strlen(token2) + strlen(end_loc) + 1);
4246 strcpy(result, location);
4247 strcat(result, token2);
4248 strcat(result, end_loc);
4249 token2 = result;
4250 free(location);
4251 }
4252 return token2;
4253 }
4255 //------------------------------get_ident_common-------------------------------
4256 // Looks for an identifier in the buffer, and turns it into a null terminated
4257 // string(still inside the file buffer). Returns a pointer to the string or
4258 // NULL if some other token is found instead.
4259 char *ADLParser::get_ident_common(bool do_preproc) {
4260 register char c;
4261 char *start; // Pointer to start of token
4262 char *end; // Pointer to end of token
4264 if( _curline == NULL ) // Return NULL at EOF.
4265 return NULL;
4267 skipws_common(do_preproc); // Skip whitespace before identifier
4268 start = end = _ptr; // Start points at first character
4269 end--; // unwind end by one to prepare for loop
4270 do {
4271 end++; // Increment end pointer
4272 c = *end; // Grab character to test
4273 } while ( ((c >= 'a') && (c <= 'z')) || ((c >= 'A') && (c <= 'Z'))
4274 || ((c >= '0') && (c <= '9'))
4275 || ((c == '_')) || ((c == ':')) || ((c == '#')) );
4276 if (start == end) { // We popped out on the first try
4277 parse_err(SYNERR, "identifier expected at %c\n", c);
4278 start = NULL;
4279 }
4280 else {
4281 _curchar = c; // Save the first character of next token
4282 *end = '\0'; // NULL terminate the string in place
4283 }
4284 _ptr = end; // Reset _ptr to point to next char after token
4286 // Make sure we do not try to use #defined identifiers. If start is
4287 // NULL an error was already reported.
4288 if (do_preproc && start != NULL) {
4289 const char* def = _AD.get_preproc_def(start);
4290 if (def != NULL && strcmp(def, start)) {
4291 const char* def1 = def;
4292 const char* def2 = _AD.get_preproc_def(def1);
4293 // implement up to 2 levels of #define
4294 if (def2 != NULL && strcmp(def2, def1)) {
4295 def = def2;
4296 const char* def3 = _AD.get_preproc_def(def2);
4297 if (def3 != NULL && strcmp(def3, def2) && strcmp(def3, def1)) {
4298 parse_err(SYNERR, "unimplemented: using %s defined as %s => %s => %s",
4299 start, def1, def2, def3);
4300 }
4301 }
4302 start = strdup(def);
4303 }
4304 }
4306 return start; // Pointer to token in filebuf
4307 }
4309 //------------------------------get_ident_dup----------------------------------
4310 // Looks for an identifier in the buffer, and returns a duplicate
4311 // or NULL if some other token is found instead.
4312 char *ADLParser::get_ident_dup(void) {
4313 char *ident = get_ident();
4315 // Duplicate an identifier before returning and restore string.
4316 if( ident != NULL ) {
4317 ident = strdup(ident); // Copy the string
4318 *_ptr = _curchar; // and replace Nil with original character
4319 }
4321 return ident;
4322 }
4324 //----------------------get_ident_or_literal_constant--------------------------
4325 // Looks for an identifier in the buffer, or a parenthesized expression.
4326 char *ADLParser::get_ident_or_literal_constant(const char* description) {
4327 char* param = NULL;
4328 skipws();
4329 if (_curchar == '(') {
4330 // Grab a constant expression.
4331 param = get_paren_expr(description);
4332 if (param[0] != '(') {
4333 char* buf = (char*) malloc(strlen(param) + 3);
4334 sprintf(buf, "(%s)", param);
4335 param = buf;
4336 }
4337 assert(is_literal_constant(param),
4338 "expr must be recognizable as a constant");
4339 } else {
4340 param = get_ident();
4341 }
4342 return param;
4343 }
4345 //------------------------------get_rep_var_ident-----------------------------
4346 // Do NOT duplicate,
4347 // Leave nil terminator in buffer
4348 // Preserve initial '$'(s) in string
4349 char *ADLParser::get_rep_var_ident(void) {
4350 // Remember starting point
4351 char *rep_var = _ptr;
4353 // Check for replacement variable indicator '$' and pass if present
4354 if ( _curchar == '$' ) {
4355 next_char();
4356 }
4357 // Check for a subfield indicator, a second '$', and pass if present
4358 if ( _curchar == '$' ) {
4359 next_char();
4360 }
4362 // Check for a control indicator, a third '$':
4363 if ( _curchar == '$' ) {
4364 next_char();
4365 }
4367 // Check for more than three '$'s in sequence, SYNERR
4368 if( _curchar == '$' ) {
4369 parse_err(SYNERR, "Replacement variables and field specifiers can not start with '$$$$'");
4370 next_char();
4371 return NULL;
4372 }
4374 // Nil terminate the variable name following the '$'
4375 char *rep_var_name = get_ident();
4376 assert( rep_var_name != NULL,
4377 "Missing identifier after replacement variable indicator '$'");
4379 return rep_var;
4380 }
4384 //------------------------------get_rep_var_ident_dup-------------------------
4385 // Return the next replacement variable identifier, skipping first '$'
4386 // given a pointer into a line of the buffer.
4387 // Null terminates string, still inside the file buffer,
4388 // Returns a pointer to a copy of the string, or NULL on failure
4389 char *ADLParser::get_rep_var_ident_dup(void) {
4390 if( _curchar != '$' ) return NULL;
4392 next_char(); // Move past the '$'
4393 char *rep_var = _ptr; // Remember starting point
4395 // Check for a subfield indicator, a second '$':
4396 if ( _curchar == '$' ) {
4397 next_char();
4398 }
4400 // Check for a control indicator, a third '$':
4401 if ( _curchar == '$' ) {
4402 next_char();
4403 }
4405 // Check for more than three '$'s in sequence, SYNERR
4406 if( _curchar == '$' ) {
4407 parse_err(SYNERR, "Replacement variables and field specifiers can not start with '$$$$'");
4408 next_char();
4409 return NULL;
4410 }
4412 // Nil terminate the variable name following the '$'
4413 char *rep_var_name = get_ident();
4414 assert( rep_var_name != NULL,
4415 "Missing identifier after replacement variable indicator '$'");
4416 rep_var = strdup(rep_var); // Copy the string
4417 *_ptr = _curchar; // and replace Nil with original character
4419 return rep_var;
4420 }
4423 //------------------------------get_unique_ident------------------------------
4424 // Looks for an identifier in the buffer, terminates it with a NULL,
4425 // and checks that it is unique
4426 char *ADLParser::get_unique_ident(FormDict& dict, const char* nameDescription){
4427 char* ident = get_ident();
4429 if (ident == NULL) {
4430 parse_err(SYNERR, "missing %s identifier at %c\n", nameDescription, _curchar);
4431 }
4432 else {
4433 if (dict[ident] != NULL) {
4434 parse_err(SYNERR, "duplicate name %s for %s\n", ident, nameDescription);
4435 ident = NULL;
4436 }
4437 }
4439 return ident;
4440 }
4443 //------------------------------get_int----------------------------------------
4444 // Looks for a character string integer in the buffer, and turns it into an int
4445 // invokes a parse_err if the next token is not an integer.
4446 // This routine does not leave the integer null-terminated.
4447 int ADLParser::get_int(void) {
4448 register char c;
4449 char *start; // Pointer to start of token
4450 char *end; // Pointer to end of token
4451 int result; // Storage for integer result
4453 if( _curline == NULL ) // Return NULL at EOF.
4454 return 0;
4456 skipws(); // Skip whitespace before identifier
4457 start = end = _ptr; // Start points at first character
4458 c = *end; // Grab character to test
4459 while ((c >= '0') && (c <= '9')
4460 || ((c == '-') && (end == start))) {
4461 end++; // Increment end pointer
4462 c = *end; // Grab character to test
4463 }
4464 if (start == end) { // We popped out on the first try
4465 parse_err(SYNERR, "integer expected at %c\n", c);
4466 result = 0;
4467 }
4468 else {
4469 _curchar = c; // Save the first character of next token
4470 *end = '\0'; // NULL terminate the string in place
4471 result = atoi(start); // Convert the string to an integer
4472 *end = _curchar; // Restore buffer to original condition
4473 }
4475 // Reset _ptr to next char after token
4476 _ptr = end;
4478 return result; // integer
4479 }
4482 //------------------------------get_relation_dup------------------------------
4483 // Looks for a relational operator in the buffer
4484 // invokes a parse_err if the next token is not a relation
4485 // This routine creates a duplicate of the string in the buffer.
4486 char *ADLParser::get_relation_dup(void) {
4487 char *result = NULL; // relational operator being returned
4489 if( _curline == NULL ) // Return NULL at EOF.
4490 return NULL;
4492 skipws(); // Skip whitespace before relation
4493 char *start = _ptr; // Store start of relational operator
4494 char first = *_ptr; // the first character
4495 if( (first == '=') || (first == '!') || (first == '<') || (first == '>') ) {
4496 next_char();
4497 char second = *_ptr; // the second character
4498 if( (second == '=') ) {
4499 next_char();
4500 char tmp = *_ptr;
4501 *_ptr = '\0'; // NULL terminate
4502 result = strdup(start); // Duplicate the string
4503 *_ptr = tmp; // restore buffer
4504 } else {
4505 parse_err(SYNERR, "relational operator expected at %s\n", _ptr);
4506 }
4507 } else {
4508 parse_err(SYNERR, "relational operator expected at %s\n", _ptr);
4509 }
4511 return result;
4512 }
4516 //------------------------------get_oplist-------------------------------------
4517 // Looks for identifier pairs where first must be the name of an operand, and
4518 // second must be a name unique in the scope of this instruction. Stores the
4519 // names with a pointer to the OpClassForm of their type in a local name table.
4520 void ADLParser::get_oplist(NameList ¶meters, FormDict &operands) {
4521 OpClassForm *opclass = NULL;
4522 char *ident = NULL;
4524 do {
4525 next_char(); // skip open paren & comma characters
4526 skipws();
4527 if (_curchar == ')') break;
4529 // Get operand type, and check it against global name table
4530 ident = get_ident();
4531 if (ident == NULL) {
4532 parse_err(SYNERR, "optype identifier expected at %c\n", _curchar);
4533 return;
4534 }
4535 else {
4536 const Form *form = _globalNames[ident];
4537 if( form == NULL ) {
4538 parse_err(SYNERR, "undefined operand type %s\n", ident);
4539 return;
4540 }
4542 // Check for valid operand type
4543 OpClassForm *opc = form->is_opclass();
4544 OperandForm *oper = form->is_operand();
4545 if((oper == NULL) && (opc == NULL)) {
4546 parse_err(SYNERR, "identifier %s not operand type\n", ident);
4547 return;
4548 }
4549 opclass = opc;
4550 }
4551 // Debugging Stuff
4552 if (_AD._adl_debug > 1) fprintf(stderr, "\tOperand Type: %s\t", ident);
4554 // Get name of operand and add it to local name table
4555 if( (ident = get_unique_ident(operands, "operand")) == NULL) {
4556 return;
4557 }
4558 // Parameter names must not be global names.
4559 if( _globalNames[ident] != NULL ) {
4560 parse_err(SYNERR, "Reuse of global name %s as operand.\n",ident);
4561 return;
4562 }
4563 operands.Insert(ident, opclass);
4564 parameters.addName(ident);
4566 // Debugging Stuff
4567 if (_AD._adl_debug > 1) fprintf(stderr, "\tOperand Name: %s\n", ident);
4568 skipws();
4569 } while(_curchar == ',');
4571 if (_curchar != ')') parse_err(SYNERR, "missing ')'\n");
4572 else {
4573 next_char(); // set current character position past the close paren
4574 }
4575 }
4578 //------------------------------get_effectlist---------------------------------
4579 // Looks for identifier pairs where first must be the name of a pre-defined,
4580 // effect, and the second must be the name of an operand defined in the
4581 // operand list of this instruction. Stores the names with a pointer to the
4582 // effect form in a local effects table.
4583 void ADLParser::get_effectlist(FormDict &effects, FormDict &operands, bool& has_call) {
4584 OperandForm *opForm;
4585 Effect *eForm;
4586 char *ident;
4588 do {
4589 next_char(); // skip open paren & comma characters
4590 skipws();
4591 if (_curchar == ')') break;
4593 // Get effect type, and check it against global name table
4594 ident = get_ident();
4595 if (ident == NULL) {
4596 parse_err(SYNERR, "effect type identifier expected at %c\n", _curchar);
4597 return;
4598 }
4599 else {
4600 // Check for valid effect type
4601 const Form *form = _globalNames[ident];
4602 if( form == NULL ) {
4603 parse_err(SYNERR, "undefined effect type %s\n", ident);
4604 return;
4605 }
4606 else {
4607 if( (eForm = form->is_effect()) == NULL) {
4608 parse_err(SYNERR, "identifier %s not effect type\n", ident);
4609 return;
4610 }
4611 }
4612 }
4613 // Debugging Stuff
4614 if (_AD._adl_debug > 1) fprintf(stderr, "\tEffect Type: %s\t", ident);
4615 skipws();
4616 if (eForm->is(Component::CALL)) {
4617 if (_AD._adl_debug > 1) fprintf(stderr, "\n");
4618 has_call = true;
4619 } else {
4620 // Get name of operand and check that it is in the local name table
4621 if( (ident = get_unique_ident(effects, "effect")) == NULL) {
4622 parse_err(SYNERR, "missing operand identifier in effect list\n");
4623 return;
4624 }
4625 const Form *form = operands[ident];
4626 opForm = form ? form->is_operand() : NULL;
4627 if( opForm == NULL ) {
4628 if( form && form->is_opclass() ) {
4629 const char* cname = form->is_opclass()->_ident;
4630 parse_err(SYNERR, "operand classes are illegal in effect lists (found %s %s)\n", cname, ident);
4631 } else {
4632 parse_err(SYNERR, "undefined operand %s in effect list\n", ident);
4633 }
4634 return;
4635 }
4636 // Add the pair to the effects table
4637 effects.Insert(ident, eForm);
4638 // Debugging Stuff
4639 if (_AD._adl_debug > 1) fprintf(stderr, "\tOperand Name: %s\n", ident);
4640 }
4641 skipws();
4642 } while(_curchar == ',');
4644 if (_curchar != ')') parse_err(SYNERR, "missing ')'\n");
4645 else {
4646 next_char(); // set current character position past the close paren
4647 }
4648 }
4651 //-------------------------------preproc_line----------------------------------
4652 // A "#line" keyword has been seen, so parse the rest of the line.
4653 void ADLParser::preproc_line(void) {
4654 int line = get_int();
4655 skipws_no_preproc();
4656 const char* file = NULL;
4657 if (_curchar == '"') {
4658 next_char(); // Move past the initial '"'
4659 file = _ptr;
4660 while (true) {
4661 if (_curchar == '\n') {
4662 parse_err(SYNERR, "missing '\"' at end of #line directive");
4663 return;
4664 }
4665 if (_curchar == '"') {
4666 *_ptr = '\0'; // Terminate the string
4667 next_char();
4668 skipws_no_preproc();
4669 break;
4670 }
4671 next_char();
4672 }
4673 }
4674 ensure_end_of_line();
4675 if (file != NULL)
4676 _AD._ADL_file._name = file;
4677 _buf.set_linenum(line);
4678 }
4680 //------------------------------preproc_define---------------------------------
4681 // A "#define" keyword has been seen, so parse the rest of the line.
4682 void ADLParser::preproc_define(void) {
4683 char* flag = get_ident_no_preproc();
4684 skipws_no_preproc();
4685 // only #define x y is supported for now
4686 char* def = get_ident_no_preproc();
4687 _AD.set_preproc_def(flag, def);
4688 skipws_no_preproc();
4689 if (_curchar != '\n') {
4690 parse_err(SYNERR, "non-identifier in preprocessor definition\n");
4691 }
4692 }
4694 //------------------------------preproc_undef----------------------------------
4695 // An "#undef" keyword has been seen, so parse the rest of the line.
4696 void ADLParser::preproc_undef(void) {
4697 char* flag = get_ident_no_preproc();
4698 skipws_no_preproc();
4699 ensure_end_of_line();
4700 _AD.set_preproc_def(flag, NULL);
4701 }
4705 //------------------------------parse_err--------------------------------------
4706 // Issue a parser error message, and skip to the end of the current line
4707 void ADLParser::parse_err(int flag, const char *fmt, ...) {
4708 va_list args;
4710 va_start(args, fmt);
4711 if (flag == 1)
4712 _AD._syntax_errs += _AD.emit_msg(0, flag, linenum(), fmt, args);
4713 else if (flag == 2)
4714 _AD._semantic_errs += _AD.emit_msg(0, flag, linenum(), fmt, args);
4715 else
4716 _AD._warnings += _AD.emit_msg(0, flag, linenum(), fmt, args);
4718 int error_char = _curchar;
4719 char* error_ptr = _ptr+1;
4720 for(;*_ptr != '\n'; _ptr++) ; // Skip to the end of the current line
4721 _curchar = '\n';
4722 va_end(args);
4723 _AD._no_output = 1;
4725 if (flag == 1) {
4726 char* error_tail = strchr(error_ptr, '\n');
4727 char tem = *error_ptr;
4728 error_ptr[-1] = '\0';
4729 char* error_head = error_ptr-1;
4730 while (error_head > _curline && *error_head) --error_head;
4731 if (error_tail) *error_tail = '\0';
4732 fprintf(stderr, "Error Context: %s>>>%c<<<%s\n",
4733 error_head, error_char, error_ptr);
4734 if (error_tail) *error_tail = '\n';
4735 error_ptr[-1] = tem;
4736 }
4737 }
4739 //---------------------------ensure_start_of_line------------------------------
4740 // A preprocessor directive has been encountered. Be sure it has fallen at
4741 // the beginning of a line, or else report an error.
4742 void ADLParser::ensure_start_of_line(void) {
4743 if (_curchar == '\n') { next_line(); return; }
4744 assert( _ptr >= _curline && _ptr < _curline+strlen(_curline),
4745 "Must be able to find which line we are in" );
4747 for (char *s = _curline; s < _ptr; s++) {
4748 if (*s > ' ') {
4749 parse_err(SYNERR, "'%c' must be at beginning of line\n", _curchar);
4750 break;
4751 }
4752 }
4753 }
4755 //---------------------------ensure_end_of_line--------------------------------
4756 // A preprocessor directive has been parsed. Be sure there is no trailing
4757 // garbage at the end of this line. Set the scan point to the beginning of
4758 // the next line.
4759 void ADLParser::ensure_end_of_line(void) {
4760 skipws_no_preproc();
4761 if (_curchar != '\n' && _curchar != '\0') {
4762 parse_err(SYNERR, "garbage char '%c' at end of line\n", _curchar);
4763 } else {
4764 next_char_or_line();
4765 }
4766 }
4768 //---------------------------handle_preproc------------------------------------
4769 // The '#' character introducing a preprocessor directive has been found.
4770 // Parse the whole directive name (e.g., #define, #endif) and take appropriate
4771 // action. If we are in an "untaken" span of text, simply keep track of
4772 // #ifdef nesting structure, so we can find out when to start taking text
4773 // again. (In this state, we "sort of support" C's #if directives, enough
4774 // to disregard their associated #else and #endif lines.) If we are in a
4775 // "taken" span of text, there are two cases: "#define" and "#undef"
4776 // directives are preserved and passed up to the caller, which eventually
4777 // passes control to the top-level parser loop, which handles #define and
4778 // #undef directly. (This prevents these directives from occurring in
4779 // arbitrary positions in the AD file--we require better structure than C.)
4780 // In the other case, and #ifdef, #ifndef, #else, or #endif is silently
4781 // processed as whitespace, with the "taken" state of the text correctly
4782 // updated. This routine returns "false" exactly in the case of a "taken"
4783 // #define or #undef, which tells the caller that a preprocessor token
4784 // has appeared which must be handled explicitly by the parse loop.
4785 bool ADLParser::handle_preproc_token() {
4786 assert(*_ptr == '#', "must be at start of preproc");
4787 ensure_start_of_line();
4788 next_char();
4789 skipws_no_preproc();
4790 char* start_ident = _ptr;
4791 char* ident = (_curchar == '\n') ? NULL : get_ident_no_preproc();
4792 if (ident == NULL) {
4793 parse_err(SYNERR, "expected preprocessor command, got end of line\n");
4794 } else if (!strcmp(ident, "ifdef") ||
4795 !strcmp(ident, "ifndef")) {
4796 char* flag = get_ident_no_preproc();
4797 ensure_end_of_line();
4798 // Test the identifier only if we are already in taken code:
4799 bool flag_def = preproc_taken() && (_AD.get_preproc_def(flag) != NULL);
4800 bool now_taken = !strcmp(ident, "ifdef") ? flag_def : !flag_def;
4801 begin_if_def(now_taken);
4802 } else if (!strcmp(ident, "if")) {
4803 if (preproc_taken())
4804 parse_err(SYNERR, "unimplemented: #%s %s", ident, _ptr+1);
4805 next_line();
4806 // Intelligently skip this nested C preprocessor directive:
4807 begin_if_def(true);
4808 } else if (!strcmp(ident, "else")) {
4809 ensure_end_of_line();
4810 invert_if_def();
4811 } else if (!strcmp(ident, "endif")) {
4812 ensure_end_of_line();
4813 end_if_def();
4814 } else if (preproc_taken()) {
4815 // pass this token up to the main parser as "#define" or "#undef"
4816 _ptr = start_ident;
4817 _curchar = *--_ptr;
4818 if( _curchar != '#' ) {
4819 parse_err(SYNERR, "no space allowed after # in #define or #undef");
4820 assert(_curchar == '#', "no space allowed after # in #define or #undef");
4821 }
4822 return false;
4823 }
4824 return true;
4825 }
4827 //---------------------------skipws_common-------------------------------------
4828 // Skip whitespace, including comments and newlines, while keeping an accurate
4829 // line count.
4830 // Maybe handle certain preprocessor constructs: #ifdef, #ifndef, #else, #endif
4831 void ADLParser::skipws_common(bool do_preproc) {
4832 char *start = _ptr;
4833 char *next = _ptr + 1;
4835 if (*_ptr == '\0') {
4836 // Check for string terminator
4837 if (_curchar > ' ') return;
4838 if (_curchar == '\n') {
4839 if (!do_preproc) return; // let caller handle the newline
4840 next_line();
4841 _ptr = _curline; next = _ptr + 1;
4842 }
4843 else if (_curchar == '#' ||
4844 (_curchar == '/' && (*next == '/' || *next == '*'))) {
4845 parse_err(SYNERR, "unimplemented: comment token in a funny place");
4846 }
4847 }
4848 while(_curline != NULL) { // Check for end of file
4849 if (*_ptr == '\n') { // keep proper track of new lines
4850 if (!do_preproc) break; // let caller handle the newline
4851 next_line();
4852 _ptr = _curline; next = _ptr + 1;
4853 }
4854 else if ((*_ptr == '/') && (*next == '/')) // C++ comment
4855 do { _ptr++; next++; } while(*_ptr != '\n'); // So go to end of line
4856 else if ((*_ptr == '/') && (*next == '*')) { // C comment
4857 _ptr++; next++;
4858 do {
4859 _ptr++; next++;
4860 if (*_ptr == '\n') { // keep proper track of new lines
4861 next_line(); // skip newlines within comments
4862 if (_curline == NULL) { // check for end of file
4863 parse_err(SYNERR, "end-of-file detected inside comment\n");
4864 break;
4865 }
4866 _ptr = _curline; next = _ptr + 1;
4867 }
4868 } while(!((*_ptr == '*') && (*next == '/'))); // Go to end of comment
4869 _ptr = ++next; next++; // increment _ptr past comment end
4870 }
4871 else if (do_preproc && *_ptr == '#') {
4872 // Note that this calls skipws_common(false) recursively!
4873 bool preproc_handled = handle_preproc_token();
4874 if (!preproc_handled) {
4875 if (preproc_taken()) {
4876 return; // short circuit
4877 }
4878 ++_ptr; // skip the preprocessor character
4879 }
4880 next = _ptr+1;
4881 } else if(*_ptr > ' ' && !(do_preproc && !preproc_taken())) {
4882 break;
4883 }
4884 else if (*_ptr == '"' || *_ptr == '\'') {
4885 assert(do_preproc, "only skip strings if doing preproc");
4886 // skip untaken quoted string
4887 int qchar = *_ptr;
4888 while (true) {
4889 ++_ptr;
4890 if (*_ptr == qchar) { ++_ptr; break; }
4891 if (*_ptr == '\\') ++_ptr;
4892 if (*_ptr == '\n' || *_ptr == '\0') {
4893 parse_err(SYNERR, "newline in string");
4894 break;
4895 }
4896 }
4897 next = _ptr + 1;
4898 }
4899 else { ++_ptr; ++next; }
4900 }
4901 if( _curline != NULL ) // at end of file _curchar isn't valid
4902 _curchar = *_ptr; // reset _curchar to maintain invariant
4903 }
4905 //---------------------------cur_char-----------------------------------------
4906 char ADLParser::cur_char() {
4907 return (_curchar);
4908 }
4910 //---------------------------next_char-----------------------------------------
4911 void ADLParser::next_char() {
4912 if (_curchar == '\n') parse_err(WARN, "must call next_line!");
4913 _curchar = *++_ptr;
4914 // if ( _curchar == '\n' ) {
4915 // next_line();
4916 // }
4917 }
4919 //---------------------------next_char_or_line---------------------------------
4920 void ADLParser::next_char_or_line() {
4921 if ( _curchar != '\n' ) {
4922 _curchar = *++_ptr;
4923 } else {
4924 next_line();
4925 _ptr = _curline;
4926 _curchar = *_ptr; // maintain invariant
4927 }
4928 }
4930 //---------------------------next_line-----------------------------------------
4931 void ADLParser::next_line() {
4932 _curline = _buf.get_line();
4933 _curchar = ' ';
4934 }
4936 //------------------------get_line_string--------------------------------------
4937 // Prepended location descriptor, for debugging.
4938 // Must return a malloced string (that can be freed if desired).
4939 char* ADLParser::get_line_string(int linenum) {
4940 const char* file = _AD._ADL_file._name;
4941 int line = linenum ? linenum : this->linenum();
4942 char* location = (char *)malloc(strlen(file) + 100);
4943 sprintf(location, "\n#line %d \"%s\"\n", line, file);
4944 return location;
4945 }
4947 //-------------------------is_literal_constant---------------------------------
4948 bool ADLParser::is_literal_constant(const char *param) {
4949 if (param[0] == 0) return false; // null string
4950 if (param[0] == '(') return true; // parenthesized expression
4951 if (param[0] == '0' && (param[1] == 'x' || param[1] == 'X')) {
4952 // Make sure it's a hex constant.
4953 int i = 2;
4954 do {
4955 if( !ADLParser::is_hex_digit(*(param+i)) ) return false;
4956 ++i;
4957 } while( *(param+i) != 0 );
4958 return true;
4959 }
4960 return false;
4961 }
4963 //---------------------------is_hex_digit--------------------------------------
4964 bool ADLParser::is_hex_digit(char digit) {
4965 return ((digit >= '0') && (digit <= '9'))
4966 ||((digit >= 'a') && (digit <= 'f'))
4967 ||((digit >= 'A') && (digit <= 'F'));
4968 }
4970 //---------------------------is_int_token--------------------------------------
4971 bool ADLParser::is_int_token(const char* token, int& intval) {
4972 const char* cp = token;
4973 while (*cp != '\0' && *cp <= ' ') cp++;
4974 if (*cp == '-') cp++;
4975 int ndigit = 0;
4976 while (*cp >= '0' && *cp <= '9') { cp++; ndigit++; }
4977 while (*cp != '\0' && *cp <= ' ') cp++;
4978 if (ndigit == 0 || *cp != '\0') {
4979 return false;
4980 }
4981 intval = atoi(token);
4982 return true;
4983 }
4985 static const char* skip_expr_ws(const char* str) {
4986 const char * cp = str;
4987 while (cp[0]) {
4988 if (cp[0] <= ' ') {
4989 ++cp;
4990 } else if (cp[0] == '#') {
4991 ++cp;
4992 while (cp[0] == ' ') ++cp;
4993 assert(0 == strncmp(cp, "line", 4), "must be a #line directive");
4994 const char* eol = strchr(cp, '\n');
4995 assert(eol != NULL, "must find end of line");
4996 if (eol == NULL) eol = cp + strlen(cp);
4997 cp = eol;
4998 } else {
4999 break;
5000 }
5001 }
5002 return cp;
5003 }
5005 //-----------------------equivalent_expressions--------------------------------
5006 bool ADLParser::equivalent_expressions(const char* str1, const char* str2) {
5007 if (str1 == str2)
5008 return true;
5009 else if (str1 == NULL || str2 == NULL)
5010 return false;
5011 const char* cp1 = str1;
5012 const char* cp2 = str2;
5013 char in_quote = '\0';
5014 while (cp1[0] && cp2[0]) {
5015 if (!in_quote) {
5016 // skip spaces and/or cpp directives
5017 const char* cp1a = skip_expr_ws(cp1);
5018 const char* cp2a = skip_expr_ws(cp2);
5019 if (cp1a > cp1 && cp2a > cp2) {
5020 cp1 = cp1a; cp2 = cp2a;
5021 continue;
5022 }
5023 if (cp1a > cp1 || cp2a > cp2) break; // fail
5024 }
5025 // match one non-space char
5026 if (cp1[0] != cp2[0]) break; // fail
5027 char ch = cp1[0];
5028 cp1++; cp2++;
5029 // watch for quotes
5030 if (in_quote && ch == '\\') {
5031 if (cp1[0] != cp2[0]) break; // fail
5032 if (!cp1[0]) break;
5033 cp1++; cp2++;
5034 }
5035 if (in_quote && ch == in_quote) {
5036 in_quote = '\0';
5037 } else if (!in_quote && (ch == '"' || ch == '\'')) {
5038 in_quote = ch;
5039 }
5040 }
5041 return (!cp1[0] && !cp2[0]);
5042 }
5045 //-------------------------------trim------------------------------------------
5046 void ADLParser::trim(char* &token) {
5047 while (*token <= ' ') token++;
5048 char* end = token + strlen(token);
5049 while (end > token && *(end-1) <= ' ') --end;
5050 *end = '\0';
5051 }