1 /*
2  * Copyright (c) 1997, 2017, 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  */
24 
25 // ADLPARSE.CPP - Architecture Description Language Parser
26 // Authors: Chris Vick and Mike Paleczny
27 #include "adlc.hpp"
28 
29 //----------------------------ADLParser----------------------------------------
30 // Create a new ADL parser
ADLParser(FileBuff & buffer,ArchDesc & archDesc)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
37 
38   _preproc_depth = 0;
39   _preproc_not_taken = 0;
40 
41   // Delimit command-line definitions from in-file definitions:
42   _AD._preproc_list.add_signal();
43 }
44 
45 //------------------------------~ADLParser-------------------------------------
46 // Delete an ADL parser.
~ADLParser()47 ADLParser::~ADLParser() {
48   if (!_AD._quiet_mode)
49     fprintf(stderr,"---------------------------- Errors and Warnings ----------------------------\n");
50 #ifndef ASSERT
51   if (!_AD._quiet_mode) {
52     fprintf(stderr, "**************************************************************\n");
53     fprintf(stderr, "***** WARNING: ASSERT is undefined, assertions disabled. *****\n");
54     fprintf(stderr, "**************************************************************\n");
55   }
56 #endif
57   if( _AD._syntax_errs + _AD._semantic_errs + _AD._warnings == 0 ) {
58     if (!_AD._quiet_mode)
59       fprintf(stderr,"No errors or warnings to report from phase-1 parse.\n" );
60   }
61   else {
62     if( _AD._syntax_errs ) {      // Any syntax errors?
63       fprintf(stderr,"%s:  Found %d syntax error", _buf._fp->_name, _AD._syntax_errs);
64       if( _AD._syntax_errs > 1 ) fprintf(stderr,"s.\n\n");
65       else fprintf(stderr,".\n\n");
66     }
67     if( _AD._semantic_errs ) {    // Any semantic errors?
68       fprintf(stderr,"%s:  Found %d semantic error", _buf._fp->_name, _AD._semantic_errs);
69       if( _AD._semantic_errs > 1 ) fprintf(stderr,"s.\n\n");
70       else fprintf(stderr,".\n\n");
71     }
72     if( _AD._warnings ) {         // Any warnings?
73       fprintf(stderr,"%s:  Found %d warning", _buf._fp->_name, _AD._warnings);
74       if( _AD._warnings > 1 ) fprintf(stderr,"s.\n\n");
75       else fprintf(stderr,".\n\n");
76     }
77   }
78   if (!_AD._quiet_mode)
79     fprintf(stderr,"-----------------------------------------------------------------------------\n");
80   _AD._TotalLines += linenum()-1;     // -1 for overshoot in "nextline" routine
81 
82   // Write out information we have stored
83   // // UNIXism == fsync(stderr);
84 }
85 
86 //------------------------------parse------------------------------------------
87 // Each top-level keyword should appear as the first non-whitespace on a line.
88 //
parse()89 void ADLParser::parse() {
90   char *ident;
91 
92   // Iterate over the lines in the file buffer parsing Level 1 objects
93   for( next_line(); _curline != NULL; next_line()) {
94     _ptr = _curline;             // Reset ptr to start of new line
95     skipws();                    // Skip any leading whitespace
96     ident = get_ident();         // Get first token
97     if (ident == NULL) {         // Empty line
98       continue;                  // Get the next line
99     }
100          if (!strcmp(ident, "instruct"))   instr_parse();
101     else if (!strcmp(ident, "operand"))    oper_parse();
102     else if (!strcmp(ident, "opclass"))    opclass_parse();
103     else if (!strcmp(ident, "ins_attrib")) ins_attr_parse();
104     else if (!strcmp(ident, "op_attrib"))  op_attr_parse();
105     else if (!strcmp(ident, "source"))     source_parse();
106     else if (!strcmp(ident, "source_hpp")) source_hpp_parse();
107     else if (!strcmp(ident, "register"))   reg_parse();
108     else if (!strcmp(ident, "frame"))      frame_parse();
109     else if (!strcmp(ident, "encode"))     encode_parse();
110     else if (!strcmp(ident, "pipeline"))   pipe_parse();
111     else if (!strcmp(ident, "definitions")) definitions_parse();
112     else if (!strcmp(ident, "peephole"))   peep_parse();
113     else if (!strcmp(ident, "#line"))      preproc_line();
114     else if (!strcmp(ident, "#define"))    preproc_define();
115     else if (!strcmp(ident, "#undef"))     preproc_undef();
116     else {
117       parse_err(SYNERR, "expected one of - instruct, operand, ins_attrib, op_attrib, source, register, pipeline, encode\n     Found %s",ident);
118     }
119   }
120   // Add reg_class spill_regs after parsing.
121   RegisterForm *regBlock = _AD.get_registers();
122   if (regBlock == NULL) {
123     parse_err(SEMERR, "Did not declare 'register' definitions");
124   }
125   regBlock->addSpillRegClass();
126 
127   // Done with parsing, check consistency.
128 
129   if (_preproc_depth != 0) {
130     parse_err(SYNERR, "End of file inside #ifdef");
131   }
132 
133   // AttributeForms ins_cost and op_cost must be defined for default behaviour
134   if (_globalNames[AttributeForm::_ins_cost] == NULL) {
135     parse_err(SEMERR, "Did not declare 'ins_cost' attribute");
136   }
137   if (_globalNames[AttributeForm::_op_cost] == NULL) {
138     parse_err(SEMERR, "Did not declare 'op_cost' attribute");
139   }
140 }
141 
142 // ******************** Private Level 1 Parse Functions ********************
143 //------------------------------instr_parse------------------------------------
144 // Parse the contents of an instruction definition, build the InstructForm to
145 // represent that instruction, and add it to the InstructForm list.
instr_parse(void)146 void ADLParser::instr_parse(void) {
147   char          *ident;
148   InstructForm  *instr;
149   MatchRule     *rule;
150   int            match_rules_cnt = 0;
151 
152   // First get the name of the instruction
153   if( (ident = get_unique_ident(_globalNames,"instruction")) == NULL )
154     return;
155   instr = new InstructForm(ident); // Create new instruction form
156   instr->_linenum = linenum();
157   _globalNames.Insert(ident, instr); // Add name to the name table
158   // Debugging Stuff
159   if (_AD._adl_debug > 1)
160     fprintf(stderr,"Parsing Instruction Form %s\n", ident);
161 
162   // Then get the operands
163   skipws();
164   if (_curchar != '(') {
165     parse_err(SYNERR, "missing '(' in instruct definition\n");
166   }
167   // Parse the operand list
168   else get_oplist(instr->_parameters, instr->_localNames);
169   skipws();                        // Skip leading whitespace
170   // Check for block delimiter
171   if ( (_curchar != '%')
172        || ( next_char(),  (_curchar != '{')) ) {
173     parse_err(SYNERR, "missing '%%{' in instruction definition\n");
174     return;
175   }
176   next_char();                     // Maintain the invariant
177   do {
178     ident = get_ident();           // Grab next identifier
179     if (ident == NULL) {
180       parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
181       continue;
182     }
183     if      (!strcmp(ident, "predicate")) instr->_predicate = pred_parse();
184     else if      (!strcmp(ident, "match")) {
185       // Allow one instruction have several match rules.
186       rule = instr->_matrule;
187       if (rule == NULL) {
188         // This is first match rule encountered
189         rule = match_parse(instr->_localNames);
190         if (rule) {
191           instr->_matrule = rule;
192           // Special case the treatment of Control instructions.
193           if( instr->is_ideal_control() ) {
194             // Control instructions return a special result, 'Universe'
195             rule->_result = "Universe";
196           }
197           // Check for commutative operations with tree operands.
198           matchrule_clone_and_swap(rule, instr->_ident, match_rules_cnt);
199         }
200       } else {
201         // Find the end of the match rule list
202         while (rule->_next != NULL)
203           rule = rule->_next;
204         // Add the new match rule to the list
205         rule->_next = match_parse(instr->_localNames);
206         if (rule->_next) {
207           rule = rule->_next;
208           if( instr->is_ideal_control() ) {
209             parse_err(SYNERR, "unique match rule expected for %s\n", rule->_name);
210             return;
211           }
212           assert(match_rules_cnt < 100," too many match rule clones");
213           char* buf = (char*) AllocateHeap(strlen(instr->_ident) + 4);
214           sprintf(buf, "%s_%d", instr->_ident, match_rules_cnt++);
215           rule->_result = buf;
216           // Check for commutative operations with tree operands.
217           matchrule_clone_and_swap(rule, instr->_ident, match_rules_cnt);
218         }
219       }
220     }
221     else if (!strcmp(ident, "encode"))  {
222       parse_err(SYNERR, "Instructions specify ins_encode, not encode\n");
223     }
224     else if (!strcmp(ident, "ins_encode"))       ins_encode_parse(*instr);
225     // Parse late expand keyword.
226     else if (!strcmp(ident, "postalloc_expand")) postalloc_expand_parse(*instr);
227     else if (!strcmp(ident, "opcode"))           instr->_opcode    = opcode_parse(instr);
228     else if (!strcmp(ident, "size"))             instr->_size      = size_parse(instr);
229     else if (!strcmp(ident, "effect"))           effect_parse(instr);
230     else if (!strcmp(ident, "expand"))           instr->_exprule   = expand_parse(instr);
231     else if (!strcmp(ident, "rewrite"))          instr->_rewrule   = rewrite_parse();
232     else if (!strcmp(ident, "constraint")) {
233       parse_err(SYNERR, "Instructions do not specify a constraint\n");
234     }
235     else if (!strcmp(ident, "construct")) {
236       parse_err(SYNERR, "Instructions do not specify a construct\n");
237     }
238     else if (!strcmp(ident, "format"))           instr->_format    = format_parse();
239     else if (!strcmp(ident, "interface")) {
240       parse_err(SYNERR, "Instructions do not specify an interface\n");
241     }
242     else if (!strcmp(ident, "ins_pipe"))        ins_pipe_parse(*instr);
243     else {  // Done with staticly defined parts of instruction definition
244       // Check identifier to see if it is the name of an attribute
245       const Form    *form = _globalNames[ident];
246       AttributeForm *attr = form ? form->is_attribute() : NULL;
247       if (attr && (attr->_atype == INS_ATTR)) {
248         // Insert the new attribute into the linked list.
249         Attribute *temp = attr_parse(ident);
250         temp->_next = instr->_attribs;
251         instr->_attribs = temp;
252       } else {
253         parse_err(SYNERR, "expected one of:\n predicate, match, encode, or the name of"
254                   " an instruction attribute at %s\n", ident);
255       }
256     }
257     skipws();
258   } while(_curchar != '%');
259   next_char();
260   if (_curchar != '}') {
261     parse_err(SYNERR, "missing '%%}' in instruction definition\n");
262     return;
263   }
264   // Check for "Set" form of chain rule
265   adjust_set_rule(instr);
266   if (_AD._pipeline) {
267     // No pipe required for late expand.
268     if (instr->expands() || instr->postalloc_expands()) {
269       if (instr->_ins_pipe) {
270         parse_err(WARN, "ins_pipe and expand rule both specified for instruction \"%s\";"
271                   " ins_pipe will be unused\n", instr->_ident);
272       }
273     } else {
274       if (!instr->_ins_pipe) {
275         parse_err(WARN, "No ins_pipe specified for instruction \"%s\"\n", instr->_ident);
276       }
277     }
278   }
279   // Add instruction to tail of instruction list
280   _AD.addForm(instr);
281 
282   // Create instruction form for each additional match rule
283   rule = instr->_matrule;
284   if (rule != NULL) {
285     rule = rule->_next;
286     while (rule != NULL) {
287       ident = (char*)rule->_result;
288       InstructForm *clone = new InstructForm(ident, instr, rule); // Create new instruction form
289       _globalNames.Insert(ident, clone); // Add name to the name table
290       // Debugging Stuff
291       if (_AD._adl_debug > 1)
292         fprintf(stderr,"Parsing Instruction Form %s\n", ident);
293       // Check for "Set" form of chain rule
294       adjust_set_rule(clone);
295       // Add instruction to tail of instruction list
296       _AD.addForm(clone);
297       rule = rule->_next;
298       clone->_matrule->_next = NULL; // One match rule per clone
299     }
300   }
301 }
302 
303 //------------------------------matchrule_clone_and_swap-----------------------
304 // Check for commutative operations with subtree operands,
305 // create clones and swap operands.
matchrule_clone_and_swap(MatchRule * rule,const char * instr_ident,int & match_rules_cnt)306 void ADLParser::matchrule_clone_and_swap(MatchRule* rule, const char* instr_ident, int& match_rules_cnt) {
307   // Check for commutative operations with tree operands.
308   int count = 0;
309   rule->count_commutative_op(count);
310   if (count > 0) {
311     // Clone match rule and swap commutative operation's operands.
312     rule->matchrule_swap_commutative_op(instr_ident, count, match_rules_cnt);
313   }
314 }
315 
316 //------------------------------adjust_set_rule--------------------------------
317 // Check for "Set" form of chain rule
adjust_set_rule(InstructForm * instr)318 void ADLParser::adjust_set_rule(InstructForm *instr) {
319   if (instr->_matrule == NULL || instr->_matrule->_rChild == NULL) return;
320   const char *rch = instr->_matrule->_rChild->_opType;
321   const Form *frm = _globalNames[rch];
322   if( (! strcmp(instr->_matrule->_opType,"Set")) &&
323       frm && frm->is_operand() && (! frm->ideal_only()) ) {
324     // Previous implementation, which missed leaP*, but worked for loadCon*
325     unsigned    position = 0;
326     const char *result   = NULL;
327     const char *name     = NULL;
328     const char *optype   = NULL;
329     MatchNode  *right    = instr->_matrule->_rChild;
330     if (right->base_operand(position, _globalNames, result, name, optype)) {
331       position = 1;
332       const char *result2  = NULL;
333       const char *name2    = NULL;
334       const char *optype2  = NULL;
335       // Can not have additional base operands in right side of match!
336       if ( ! right->base_operand( position, _globalNames, result2, name2, optype2) ) {
337         if (instr->_predicate != NULL)
338           parse_err(SYNERR, "ADLC does not support instruction chain rules with predicates");
339         // Chain from input  _ideal_operand_type_,
340         // Needed for shared roots of match-trees
341         ChainList *lst = (ChainList *)_AD._chainRules[optype];
342         if (lst == NULL) {
343           lst = new ChainList();
344           _AD._chainRules.Insert(optype, lst);
345         }
346         if (!lst->search(instr->_matrule->_lChild->_opType)) {
347           const char *cost = instr->cost();
348           if (cost == NULL) {
349             cost = ((AttributeForm*)_globalNames[AttributeForm::_ins_cost])->_attrdef;
350           }
351           // The ADLC does not support chaining from the ideal operand type
352           // of a predicated user-defined operand
353           if( frm->is_operand() == NULL || frm->is_operand()->_predicate == NULL ) {
354             lst->insert(instr->_matrule->_lChild->_opType,cost,instr->_ident);
355           }
356         }
357         // Chain from input  _user_defined_operand_type_,
358         lst = (ChainList *)_AD._chainRules[result];
359         if (lst == NULL) {
360           lst = new ChainList();
361           _AD._chainRules.Insert(result, lst);
362         }
363         if (!lst->search(instr->_matrule->_lChild->_opType)) {
364           const char *cost = instr->cost();
365           if (cost == NULL) {
366             cost = ((AttributeForm*)_globalNames[AttributeForm::_ins_cost])->_attrdef;
367           }
368           // It is safe to chain from the top-level user-defined operand even
369           // if it has a predicate, since the predicate is checked before
370           // the user-defined type is available.
371           lst->insert(instr->_matrule->_lChild->_opType,cost,instr->_ident);
372         }
373       } else {
374         // May have instruction chain rule if root of right-tree is an ideal
375         OperandForm *rightOp = _globalNames[right->_opType]->is_operand();
376         if( rightOp ) {
377           const Form *rightRoot = _globalNames[rightOp->_matrule->_opType];
378           if( rightRoot && rightRoot->ideal_only() ) {
379             const char *chain_op = NULL;
380             if( rightRoot->is_instruction() )
381               chain_op = rightOp->_ident;
382             if( chain_op ) {
383               // Look-up the operation in chain rule table
384               ChainList *lst = (ChainList *)_AD._chainRules[chain_op];
385               if (lst == NULL) {
386                 lst = new ChainList();
387                 _AD._chainRules.Insert(chain_op, lst);
388               }
389               // if (!lst->search(instr->_matrule->_lChild->_opType)) {
390               const char *cost = instr->cost();
391               if (cost == NULL) {
392                 cost = ((AttributeForm*)_globalNames[AttributeForm::_ins_cost])->_attrdef;
393               }
394               // This chains from a top-level operand whose predicate, if any,
395               // has been checked.
396               lst->insert(instr->_matrule->_lChild->_opType,cost,instr->_ident);
397               // }
398             }
399           }
400         }
401       } // end chain rule from right-tree's ideal root
402     }
403   }
404 }
405 
406 
407 //------------------------------oper_parse-------------------------------------
oper_parse(void)408 void ADLParser::oper_parse(void) {
409   char          *ident;
410   OperandForm   *oper;
411   AttributeForm *attr;
412   MatchRule     *rule;
413 
414   // First get the name of the operand
415   skipws();
416   if( (ident = get_unique_ident(_globalNames,"operand")) == NULL )
417     return;
418   oper = new OperandForm(ident);        // Create new operand form
419   oper->_linenum = linenum();
420   _globalNames.Insert(ident, oper); // Add name to the name table
421 
422   // Debugging Stuff
423   if (_AD._adl_debug > 1) fprintf(stderr,"Parsing Operand Form %s\n", ident);
424 
425   // Get the component operands
426   skipws();
427   if (_curchar != '(') {
428     parse_err(SYNERR, "missing '(' in operand definition\n");
429     return;
430   }
431   else get_oplist(oper->_parameters, oper->_localNames); // Parse the component operand list
432   skipws();
433   // Check for block delimiter
434   if ((_curchar != '%') || (*(_ptr+1) != '{')) { // If not open block
435     parse_err(SYNERR, "missing '%%{' in operand definition\n");
436     return;
437   }
438   next_char(); next_char();        // Skip over "%{" symbol
439   do {
440     ident = get_ident();           // Grab next identifier
441     if (ident == NULL) {
442       parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
443       continue;
444     }
445     if      (!strcmp(ident, "predicate")) oper->_predicate = pred_parse();
446     else if (!strcmp(ident, "match"))     {
447       // Find the end of the match rule list
448       rule = oper->_matrule;
449       if (rule) {
450         while (rule->_next) rule = rule->_next;
451         // Add the new match rule to the list
452         rule->_next = match_parse(oper->_localNames);
453         if (rule->_next) {
454           rule->_next->_result = oper->_ident;
455         }
456       }
457       else {
458         // This is first match rule encountered
459         oper->_matrule = match_parse(oper->_localNames);
460         if (oper->_matrule) {
461           oper->_matrule->_result = oper->_ident;
462         }
463       }
464     }
465     else if (!strcmp(ident, "encode"))    oper->_interface = interface_parse();
466     else if (!strcmp(ident, "ins_encode")) {
467       parse_err(SYNERR, "Operands specify 'encode', not 'ins_encode'\n");
468     }
469     else if (!strcmp(ident, "opcode"))    {
470       parse_err(SYNERR, "Operands do not specify an opcode\n");
471     }
472     else if (!strcmp(ident, "effect"))    {
473       parse_err(SYNERR, "Operands do not specify an effect\n");
474     }
475     else if (!strcmp(ident, "expand"))    {
476       parse_err(SYNERR, "Operands do not specify an expand\n");
477     }
478     else if (!strcmp(ident, "rewrite"))   {
479       parse_err(SYNERR, "Operands do not specify a rewrite\n");
480     }
481     else if (!strcmp(ident, "constraint"))oper->_constraint= constraint_parse();
482     else if (!strcmp(ident, "construct")) oper->_construct = construct_parse();
483     else if (!strcmp(ident, "format"))    oper->_format    = format_parse();
484     else if (!strcmp(ident, "interface")) oper->_interface = interface_parse();
485     // Check identifier to see if it is the name of an attribute
486     else if (((attr = _globalNames[ident]->is_attribute()) != NULL) &&
487              (attr->_atype == OP_ATTR))   oper->_attribs   = attr_parse(ident);
488     else {
489       parse_err(SYNERR, "expected one of - constraint, predicate, match, encode, format, construct, or the name of a defined operand attribute at %s\n", ident);
490     }
491     skipws();
492   } while(_curchar != '%');
493   next_char();
494   if (_curchar != '}') {
495     parse_err(SYNERR, "missing '%%}' in operand definition\n");
496     return;
497   }
498   // Add operand to tail of operand list
499   _AD.addForm(oper);
500 }
501 
502 //------------------------------opclass_parse----------------------------------
503 // Operand Classes are a block with a comma delimited list of operand names
opclass_parse(void)504 void ADLParser::opclass_parse(void) {
505   char          *ident;
506   OpClassForm   *opc;
507   OperandForm   *opForm;
508 
509   // First get the name of the operand class
510   skipws();
511   if( (ident = get_unique_ident(_globalNames,"opclass")) == NULL )
512     return;
513   opc = new OpClassForm(ident);             // Create new operand class form
514   _globalNames.Insert(ident, opc);  // Add name to the name table
515 
516   // Debugging Stuff
517   if (_AD._adl_debug > 1)
518     fprintf(stderr,"Parsing Operand Class Form %s\n", ident);
519 
520   // Get the list of operands
521   skipws();
522   if (_curchar != '(') {
523     parse_err(SYNERR, "missing '(' in operand definition\n");
524     return;
525   }
526   do {
527     next_char();                            // Skip past open paren or comma
528     ident = get_ident();                    // Grab next identifier
529     if (ident == NULL) {
530       parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
531       continue;
532     }
533     // Check identifier to see if it is the name of an operand
534     const Form *form = _globalNames[ident];
535     opForm     = form ? form->is_operand() : NULL;
536     if ( opForm ) {
537       opc->_oplst.addName(ident);           // Add operand to opclass list
538       opForm->_classes.addName(opc->_ident);// Add opclass to operand list
539     }
540     else {
541       parse_err(SYNERR, "expected name of a defined operand at %s\n", ident);
542     }
543     skipws();                               // skip trailing whitespace
544   } while (_curchar == ',');                // Check for the comma
545   // Check for closing ')'
546   if (_curchar != ')') {
547     parse_err(SYNERR, "missing ')' or ',' in opclass definition\n");
548     return;
549   }
550   next_char();                              // Consume the ')'
551   skipws();
552   // Check for closing ';'
553   if (_curchar != ';') {
554     parse_err(SYNERR, "missing ';' in opclass definition\n");
555     return;
556   }
557   next_char();                             // Consume the ';'
558   // Add operand to tail of operand list
559   _AD.addForm(opc);
560 }
561 
562 //------------------------------ins_attr_parse---------------------------------
ins_attr_parse(void)563 void ADLParser::ins_attr_parse(void) {
564   char          *ident;
565   char          *aexpr;
566   AttributeForm *attrib;
567 
568   // get name for the instruction attribute
569   skipws();                      // Skip leading whitespace
570   if( (ident = get_unique_ident(_globalNames,"inst_attrib")) == NULL )
571     return;
572   // Debugging Stuff
573   if (_AD._adl_debug > 1) fprintf(stderr,"Parsing Ins_Attribute Form %s\n", ident);
574 
575   // Get default value of the instruction attribute
576   skipws();                      // Skip whitespace
577   if ((aexpr = get_paren_expr("attribute default expression string")) == NULL) {
578     parse_err(SYNERR, "missing '(' in ins_attrib definition\n");
579     return;
580   }
581   // Debug Stuff
582   if (_AD._adl_debug > 1) fprintf(stderr,"Attribute Expression: %s\n", aexpr);
583 
584   // Check for terminator
585   if (_curchar != ';') {
586     parse_err(SYNERR, "missing ';' in ins_attrib definition\n");
587     return;
588   }
589   next_char();                    // Advance past the ';'
590 
591   // Construct the attribute, record global name, and store in ArchDesc
592   attrib = new AttributeForm(ident, INS_ATTR, aexpr);
593   _globalNames.Insert(ident, attrib);  // Add name to the name table
594   _AD.addForm(attrib);
595 }
596 
597 //------------------------------op_attr_parse----------------------------------
op_attr_parse(void)598 void ADLParser::op_attr_parse(void) {
599   char          *ident;
600   char          *aexpr;
601   AttributeForm *attrib;
602 
603   // get name for the operand attribute
604   skipws();                      // Skip leading whitespace
605   if( (ident = get_unique_ident(_globalNames,"op_attrib")) == NULL )
606     return;
607   // Debugging Stuff
608   if (_AD._adl_debug > 1) fprintf(stderr,"Parsing Op_Attribute Form %s\n", ident);
609 
610   // Get default value of the instruction attribute
611   skipws();                      // Skip whitespace
612   if ((aexpr = get_paren_expr("attribute default expression string")) == NULL) {
613     parse_err(SYNERR, "missing '(' in op_attrib definition\n");
614     return;
615   }
616   // Debug Stuff
617   if (_AD._adl_debug > 1) fprintf(stderr,"Attribute Expression: %s\n", aexpr);
618 
619   // Check for terminator
620   if (_curchar != ';') {
621     parse_err(SYNERR, "missing ';' in op_attrib definition\n");
622     return;
623   }
624   next_char();                    // Advance past the ';'
625 
626   // Construct the attribute, record global name, and store in ArchDesc
627   attrib = new AttributeForm(ident, OP_ATTR, aexpr);
628   _globalNames.Insert(ident, attrib);
629   _AD.addForm(attrib);
630 }
631 
632 //------------------------------definitions_parse-----------------------------------
definitions_parse(void)633 void ADLParser::definitions_parse(void) {
634   skipws();                       // Skip leading whitespace
635   if (_curchar == '%' && *(_ptr+1) == '{') {
636     next_char(); next_char();     // Skip "%{"
637     skipws();
638     while (_curchar != '%' && *(_ptr+1) != '}') {
639       // Process each definition until finding closing string "%}"
640       char *token = get_ident();
641       if (token == NULL) {
642         parse_err(SYNERR, "missing identifier inside definitions block.\n");
643         return;
644       }
645       if (strcmp(token,"int_def")==0)     { int_def_parse(); }
646       // if (strcmp(token,"str_def")==0)   { str_def_parse(); }
647       skipws();
648     }
649   }
650   else {
651     parse_err(SYNERR, "Missing %%{ ... %%} block after definitions keyword.\n");
652     return;
653   }
654 }
655 
656 //------------------------------int_def_parse----------------------------------
657 // Parse Example:
658 // int_def    MEMORY_REF_COST      (         200,  DEFAULT_COST * 2);
659 // <keyword>  <name>               ( <int_value>,   <description>  );
660 //
int_def_parse(void)661 void ADLParser::int_def_parse(void) {
662   char *name        = NULL;         // Name of definition
663   char *value       = NULL;         // its value,
664   int   int_value   = -1;           // positive values only
665   char *description = NULL;         // textual description
666 
667   // Get definition name
668   skipws();                      // Skip whitespace
669   name = get_ident();
670   if (name == NULL) {
671     parse_err(SYNERR, "missing definition name after int_def\n");
672     return;
673   }
674 
675   // Check for value of int_def dname( integer_value [, string_expression ] )
676   skipws();
677   if (_curchar == '(') {
678 
679     // Parse the integer value.
680     next_char();
681     value = get_ident();
682     if (value == NULL) {
683       parse_err(SYNERR, "missing value in int_def\n");
684       return;
685     }
686     if( !is_int_token(value, int_value) ) {
687       parse_err(SYNERR, "value in int_def is not recognized as integer\n");
688       return;
689     }
690     skipws();
691 
692     // Check for description
693     if (_curchar == ',') {
694       next_char();   // skip ','
695 
696       description = get_expr("int_def description", ")");
697       if (description == NULL) {
698         parse_err(SYNERR, "invalid or missing description in int_def\n");
699         return;
700       }
701       trim(description);
702     }
703 
704     if (_curchar != ')') {
705       parse_err(SYNERR, "missing ')' in register definition statement\n");
706       return;
707     }
708     next_char();
709   }
710 
711   // Check for closing ';'
712   skipws();
713   if (_curchar != ';') {
714     parse_err(SYNERR, "missing ';' after int_def\n");
715     return;
716   }
717   next_char();                   // move past ';'
718 
719   // Debug Stuff
720   if (_AD._adl_debug > 1) {
721     fprintf(stderr,"int_def: %s ( %s, %s )\n", name,
722             (value), (description ? description : ""));
723   }
724 
725   // Record new definition.
726   Expr *expr     = new Expr(name, description, int_value, int_value);
727   const Expr *old_expr = _AD.globalDefs().define(name, expr);
728   if (old_expr != NULL) {
729     parse_err(SYNERR, "Duplicate definition\n");
730     return;
731   }
732 
733   return;
734 }
735 
736 
737 //------------------------------source_parse-----------------------------------
source_parse(void)738 void ADLParser::source_parse(void) {
739   SourceForm *source;             // Encode class for instruction/operand
740   char   *rule = NULL;            // String representation of encode rule
741 
742   skipws();                       // Skip leading whitespace
743   if ( (rule = find_cpp_block("source block")) == NULL ) {
744     parse_err(SYNERR, "incorrect or missing block for 'source'.\n");
745     return;
746   }
747   // Debug Stuff
748   if (_AD._adl_debug > 1) fprintf(stderr,"Source Form: %s\n", rule);
749 
750   source = new SourceForm(rule);    // Build new Source object
751   _AD.addForm(source);
752   // skipws();
753 }
754 
755 //------------------------------source_hpp_parse-------------------------------
756 // Parse a source_hpp %{ ... %} block.
757 // The code gets stuck into the ad_<arch>.hpp file.
758 // If the source_hpp block appears before the register block in the AD
759 // file, it goes up at the very top of the ad_<arch>.hpp file, so that
760 // it can be used by register encodings, etc.  Otherwise, it goes towards
761 // the bottom, where it's useful as a global definition to *.cpp files.
source_hpp_parse(void)762 void ADLParser::source_hpp_parse(void) {
763   char   *rule = NULL;            // String representation of encode rule
764 
765   skipws();                       // Skip leading whitespace
766   if ( (rule = find_cpp_block("source_hpp block")) == NULL ) {
767     parse_err(SYNERR, "incorrect or missing block for 'source_hpp'.\n");
768     return;
769   }
770   // Debug Stuff
771   if (_AD._adl_debug > 1) fprintf(stderr,"Header Form: %s\n", rule);
772 
773   if (_AD.get_registers() == NULL) {
774     // Very early in the file, before reg_defs, we collect pre-headers.
775     PreHeaderForm* pre_header = new PreHeaderForm(rule);
776     _AD.addForm(pre_header);
777   } else {
778     // Normally, we collect header info, placed at the bottom of the hpp file.
779     HeaderForm* header = new HeaderForm(rule);
780     _AD.addForm(header);
781   }
782 }
783 
784 //------------------------------reg_parse--------------------------------------
reg_parse(void)785 void ADLParser::reg_parse(void) {
786   RegisterForm *regBlock = _AD.get_registers(); // Information about registers encoding
787   if (regBlock == NULL) {
788     // Create the RegisterForm for the architecture description.
789     regBlock = new RegisterForm();    // Build new Source object
790     _AD.addForm(regBlock);
791   }
792 
793   skipws();                       // Skip leading whitespace
794   if (_curchar == '%' && *(_ptr+1) == '{') {
795     next_char(); next_char();     // Skip "%{"
796     skipws();
797     while (_curchar != '%' && *(_ptr+1) != '}') {
798       char *token = get_ident();
799       if (token == NULL) {
800         parse_err(SYNERR, "missing identifier inside register block.\n");
801         return;
802       }
803       if (strcmp(token,"reg_def")==0)          { reg_def_parse(); }
804       else if (strcmp(token,"reg_class")==0)   { reg_class_parse(); }
805       else if (strcmp(token, "reg_class_dynamic") == 0) { reg_class_dynamic_parse(); }
806       else if (strcmp(token,"alloc_class")==0) { alloc_class_parse(); }
807       else if (strcmp(token,"#define")==0)     { preproc_define(); }
808       else { parse_err(SYNERR, "bad token %s inside register block.\n", token); break; }
809       skipws();
810     }
811   }
812   else {
813     parse_err(SYNERR, "Missing %c{ ... %c} block after register keyword.\n",'%','%');
814     return;
815   }
816 }
817 
818 //------------------------------encode_parse-----------------------------------
encode_parse(void)819 void ADLParser::encode_parse(void) {
820   EncodeForm *encBlock;         // Information about instruction/operand encoding
821 
822   _AD.getForm(&encBlock);
823   if ( encBlock == NULL) {
824     // Create the EncodeForm for the architecture description.
825     encBlock = new EncodeForm();    // Build new Source object
826     _AD.addForm(encBlock);
827   }
828 
829   skipws();                       // Skip leading whitespace
830   if (_curchar == '%' && *(_ptr+1) == '{') {
831     next_char(); next_char();     // Skip "%{"
832     skipws();
833     while (_curchar != '%' && *(_ptr+1) != '}') {
834       char *token = get_ident();
835       if (token == NULL) {
836             parse_err(SYNERR, "missing identifier inside encoding block.\n");
837             return;
838       }
839       if (strcmp(token,"enc_class")==0)   { enc_class_parse(); }
840       skipws();
841     }
842   }
843   else {
844     parse_err(SYNERR, "Missing %c{ ... %c} block after encode keyword.\n",'%','%');
845     return;
846   }
847 }
848 
849 //------------------------------enc_class_parse--------------------------------
enc_class_parse(void)850 void ADLParser::enc_class_parse(void) {
851   char       *ec_name;           // Name of encoding class being defined
852 
853   // Get encoding class name
854   skipws();                      // Skip whitespace
855   ec_name = get_ident();
856   if (ec_name == NULL) {
857     parse_err(SYNERR, "missing encoding class name after encode.\n");
858     return;
859   }
860 
861   EncClass  *encoding = _AD._encode->add_EncClass(ec_name);
862   encoding->_linenum = linenum();
863 
864   skipws();                      // Skip leading whitespace
865   // Check for optional parameter list
866   if (_curchar == '(') {
867     do {
868       char *pType = NULL;        // parameter type
869       char *pName = NULL;        // parameter name
870 
871       next_char();               // skip open paren & comma characters
872       skipws();
873       if (_curchar == ')') break;
874 
875       // Get parameter type
876       pType = get_ident();
877       if (pType == NULL) {
878         parse_err(SYNERR, "parameter type expected at %c\n", _curchar);
879         return;
880       }
881 
882       skipws();
883       // Get parameter name
884       pName = get_ident();
885       if (pName == NULL) {
886         parse_err(SYNERR, "parameter name expected at %c\n", _curchar);
887         return;
888       }
889 
890       // Record parameter type and name
891       encoding->add_parameter( pType, pName );
892 
893       skipws();
894     } while(_curchar == ',');
895 
896     if (_curchar != ')') parse_err(SYNERR, "missing ')'\n");
897     else {
898       next_char();                  // Skip ')'
899     }
900   } // Done with parameter list
901 
902   skipws();
903   // Check for block starting delimiters
904   if ((_curchar != '%') || (*(_ptr+1) != '{')) { // If not open block
905     parse_err(SYNERR, "missing '%c{' in enc_class definition\n", '%');
906     return;
907   }
908   next_char();                      // Skip '%'
909   next_char();                      // Skip '{'
910 
911   enc_class_parse_block(encoding, ec_name);
912 }
913 
914 
enc_class_parse_block(EncClass * encoding,char * ec_name)915 void ADLParser::enc_class_parse_block(EncClass* encoding, char* ec_name) {
916   skipws_no_preproc();              // Skip leading whitespace
917   // Prepend location descriptor, for debugging; cf. ADLParser::find_cpp_block
918   if (_AD._adlocation_debug) {
919     encoding->add_code(get_line_string());
920   }
921 
922   // Collect the parts of the encode description
923   // (1) strings that are passed through to output
924   // (2) replacement/substitution variable, preceeded by a '$'
925   while ( (_curchar != '%') && (*(_ptr+1) != '}') ) {
926 
927     // (1)
928     // Check if there is a string to pass through to output
929     char *start = _ptr;       // Record start of the next string
930     while ((_curchar != '$') && ((_curchar != '%') || (*(_ptr+1) != '}')) ) {
931       // If at the start of a comment, skip past it
932       if( (_curchar == '/') && ((*(_ptr+1) == '/') || (*(_ptr+1) == '*')) ) {
933         skipws_no_preproc();
934       } else {
935         // ELSE advance to the next character, or start of the next line
936         next_char_or_line();
937       }
938     }
939     // If a string was found, terminate it and record in EncClass
940     if ( start != _ptr ) {
941       *_ptr  = '\0';          // Terminate the string
942       encoding->add_code(start);
943     }
944 
945     // (2)
946     // If we are at a replacement variable,
947     // copy it and record in EncClass
948     if (_curchar == '$') {
949       // Found replacement Variable
950       char* rep_var = get_rep_var_ident_dup();
951       // Add flag to _strings list indicating we should check _rep_vars
952       encoding->add_rep_var(rep_var);
953     }
954   } // end while part of format description
955   next_char();                      // Skip '%'
956   next_char();                      // Skip '}'
957 
958   skipws();
959 
960   if (_AD._adlocation_debug) {
961     encoding->add_code(end_line_marker());
962   }
963 
964   // Debug Stuff
965   if (_AD._adl_debug > 1) fprintf(stderr,"EncodingClass Form: %s\n", ec_name);
966 }
967 
968 //------------------------------frame_parse-----------------------------------
frame_parse(void)969 void ADLParser::frame_parse(void) {
970   FrameForm  *frame;              // Information about stack-frame layout
971   char       *desc = NULL;        // String representation of frame
972 
973   skipws();                       // Skip leading whitespace
974 
975   frame = new FrameForm();        // Build new Frame object
976   // Check for open block sequence
977   skipws();                       // Skip leading whitespace
978   if (_curchar == '%' && *(_ptr+1) == '{') {
979     next_char(); next_char();     // Skip "%{"
980     skipws();
981     while (_curchar != '%' && *(_ptr+1) != '}') {
982       char *token = get_ident();
983       if (token == NULL) {
984             parse_err(SYNERR, "missing identifier inside frame block.\n");
985             return;
986       }
987       if (strcmp(token,"stack_direction")==0) {
988         stack_dir_parse(frame);
989       }
990       if (strcmp(token,"sync_stack_slots")==0) {
991         sync_stack_slots_parse(frame);
992       }
993       if (strcmp(token,"frame_pointer")==0) {
994         frame_pointer_parse(frame, false);
995       }
996       if (strcmp(token,"interpreter_frame_pointer")==0) {
997         interpreter_frame_pointer_parse(frame, false);
998       }
999       if (strcmp(token,"inline_cache_reg")==0) {
1000         inline_cache_parse(frame, false);
1001       }
1002       if (strcmp(token,"compiler_method_oop_reg")==0) {
1003         parse_err(WARN, "Using obsolete Token, compiler_method_oop_reg");
1004         skipws();
1005       }
1006       if (strcmp(token,"interpreter_method_oop_reg")==0) {
1007         interpreter_method_oop_parse(frame, false);
1008       }
1009       if (strcmp(token,"cisc_spilling_operand_name")==0) {
1010         cisc_spilling_operand_name_parse(frame, false);
1011       }
1012       if (strcmp(token,"stack_alignment")==0) {
1013         stack_alignment_parse(frame);
1014       }
1015       if (strcmp(token,"return_addr")==0) {
1016         return_addr_parse(frame, false);
1017       }
1018       if (strcmp(token,"in_preserve_stack_slots")==0) {
1019         preserve_stack_parse(frame);
1020       }
1021       if (strcmp(token,"out_preserve_stack_slots")==0) {
1022         parse_err(WARN, "Using obsolete token, out_preserve_stack_slots");
1023         skipws();
1024       }
1025       if (strcmp(token,"varargs_C_out_slots_killed")==0) {
1026         frame->_varargs_C_out_slots_killed = parse_one_arg("varargs C out slots killed");
1027       }
1028       if (strcmp(token,"calling_convention")==0) {
1029         frame->_calling_convention = calling_convention_parse();
1030       }
1031       if (strcmp(token,"return_value")==0) {
1032         frame->_return_value = return_value_parse();
1033       }
1034       if (strcmp(token,"c_frame_pointer")==0) {
1035         frame_pointer_parse(frame, true);
1036       }
1037       if (strcmp(token,"c_return_addr")==0) {
1038         return_addr_parse(frame, true);
1039       }
1040       if (strcmp(token,"c_calling_convention")==0) {
1041         frame->_c_calling_convention = calling_convention_parse();
1042       }
1043       if (strcmp(token,"c_return_value")==0) {
1044         frame->_c_return_value = return_value_parse();
1045       }
1046 
1047       skipws();
1048     }
1049   }
1050   else {
1051     parse_err(SYNERR, "Missing %c{ ... %c} block after encode keyword.\n",'%','%');
1052     return;
1053   }
1054   // All Java versions are required, native versions are optional
1055   if(frame->_frame_pointer == NULL) {
1056     parse_err(SYNERR, "missing frame pointer definition in frame section.\n");
1057     return;
1058   }
1059   // !!!!! !!!!!
1060   // if(frame->_interpreter_frame_ptr_reg == NULL) {
1061   //   parse_err(SYNERR, "missing interpreter frame pointer definition in frame section.\n");
1062   //   return;
1063   // }
1064   if(frame->_alignment == NULL) {
1065     parse_err(SYNERR, "missing alignment definition in frame section.\n");
1066     return;
1067   }
1068   if(frame->_return_addr == NULL) {
1069     parse_err(SYNERR, "missing return address location in frame section.\n");
1070     return;
1071   }
1072   if(frame->_in_preserve_slots == NULL) {
1073     parse_err(SYNERR, "missing stack slot preservation definition in frame section.\n");
1074     return;
1075   }
1076   if(frame->_varargs_C_out_slots_killed == NULL) {
1077     parse_err(SYNERR, "missing varargs C out slots killed definition in frame section.\n");
1078     return;
1079   }
1080   if(frame->_calling_convention == NULL) {
1081     parse_err(SYNERR, "missing calling convention definition in frame section.\n");
1082     return;
1083   }
1084   if(frame->_return_value == NULL) {
1085     parse_err(SYNERR, "missing return value definition in frame section.\n");
1086     return;
1087   }
1088   // Fill natives in identically with the Java versions if not present.
1089   if(frame->_c_frame_pointer == NULL) {
1090     frame->_c_frame_pointer = frame->_frame_pointer;
1091   }
1092   if(frame->_c_return_addr == NULL) {
1093     frame->_c_return_addr = frame->_return_addr;
1094     frame->_c_return_addr_loc = frame->_return_addr_loc;
1095   }
1096   if(frame->_c_calling_convention == NULL) {
1097     frame->_c_calling_convention = frame->_calling_convention;
1098   }
1099   if(frame->_c_return_value == NULL) {
1100     frame->_c_return_value = frame->_return_value;
1101   }
1102 
1103   // Debug Stuff
1104   if (_AD._adl_debug > 1) fprintf(stderr,"Frame Form: %s\n", desc);
1105 
1106   // Create the EncodeForm for the architecture description.
1107   _AD.addForm(frame);
1108   // skipws();
1109 }
1110 
1111 //------------------------------stack_dir_parse--------------------------------
stack_dir_parse(FrameForm * frame)1112 void ADLParser::stack_dir_parse(FrameForm *frame) {
1113   char *direction = parse_one_arg("stack direction entry");
1114   if (strcmp(direction, "TOWARDS_LOW") == 0) {
1115     frame->_direction = false;
1116   }
1117   else if (strcmp(direction, "TOWARDS_HIGH") == 0) {
1118     frame->_direction = true;
1119   }
1120   else {
1121     parse_err(SYNERR, "invalid value inside stack direction entry.\n");
1122     return;
1123   }
1124 }
1125 
1126 //------------------------------sync_stack_slots_parse-------------------------
sync_stack_slots_parse(FrameForm * frame)1127 void ADLParser::sync_stack_slots_parse(FrameForm *frame) {
1128     // Assign value into frame form
1129     frame->_sync_stack_slots = parse_one_arg("sync stack slots entry");
1130 }
1131 
1132 //------------------------------frame_pointer_parse----------------------------
frame_pointer_parse(FrameForm * frame,bool native)1133 void ADLParser::frame_pointer_parse(FrameForm *frame, bool native) {
1134   char *frame_pointer = parse_one_arg("frame pointer entry");
1135   // Assign value into frame form
1136   if (native) { frame->_c_frame_pointer = frame_pointer; }
1137   else        { frame->_frame_pointer   = frame_pointer; }
1138 }
1139 
1140 //------------------------------interpreter_frame_pointer_parse----------------------------
interpreter_frame_pointer_parse(FrameForm * frame,bool native)1141 void ADLParser::interpreter_frame_pointer_parse(FrameForm *frame, bool native) {
1142   frame->_interpreter_frame_pointer_reg = parse_one_arg("interpreter frame pointer entry");
1143 }
1144 
1145 //------------------------------inline_cache_parse-----------------------------
inline_cache_parse(FrameForm * frame,bool native)1146 void ADLParser::inline_cache_parse(FrameForm *frame, bool native) {
1147   frame->_inline_cache_reg = parse_one_arg("inline cache reg entry");
1148 }
1149 
1150 //------------------------------interpreter_method_oop_parse------------------
interpreter_method_oop_parse(FrameForm * frame,bool native)1151 void ADLParser::interpreter_method_oop_parse(FrameForm *frame, bool native) {
1152   frame->_interpreter_method_oop_reg = parse_one_arg("method oop reg entry");
1153 }
1154 
1155 //------------------------------cisc_spilling_operand_parse---------------------
cisc_spilling_operand_name_parse(FrameForm * frame,bool native)1156 void ADLParser::cisc_spilling_operand_name_parse(FrameForm *frame, bool native) {
1157   frame->_cisc_spilling_operand_name = parse_one_arg("cisc spilling operand name");
1158 }
1159 
1160 //------------------------------stack_alignment_parse--------------------------
stack_alignment_parse(FrameForm * frame)1161 void ADLParser::stack_alignment_parse(FrameForm *frame) {
1162   char *alignment = parse_one_arg("stack alignment entry");
1163   // Assign value into frame
1164   frame->_alignment   = alignment;
1165 }
1166 
1167 //------------------------------parse_one_arg-------------------------------
parse_one_arg(const char * description)1168 char *ADLParser::parse_one_arg(const char *description) {
1169   char *token = NULL;
1170   if(_curchar == '(') {
1171     next_char();
1172     skipws();
1173     token = get_expr(description, ")");
1174     if (token == NULL) {
1175       parse_err(SYNERR, "missing value inside %s.\n", description);
1176       return NULL;
1177     }
1178     next_char();           // skip the close paren
1179     if(_curchar != ';') {  // check for semi-colon
1180       parse_err(SYNERR, "missing %c in.\n", ';', description);
1181       return NULL;
1182     }
1183     next_char();           // skip the semi-colon
1184   }
1185   else {
1186     parse_err(SYNERR, "Missing %c in.\n", '(', description);
1187     return NULL;
1188   }
1189 
1190   trim(token);
1191   return token;
1192 }
1193 
1194 //------------------------------return_addr_parse------------------------------
return_addr_parse(FrameForm * frame,bool native)1195 void ADLParser::return_addr_parse(FrameForm *frame, bool native) {
1196   bool in_register  = true;
1197   if(_curchar == '(') {
1198     next_char();
1199     skipws();
1200     char *token = get_ident();
1201     if (token == NULL) {
1202       parse_err(SYNERR, "missing value inside return address entry.\n");
1203       return;
1204     }
1205     // check for valid values for stack/register
1206     if (strcmp(token, "REG") == 0) {
1207       in_register = true;
1208     }
1209     else if (strcmp(token, "STACK") == 0) {
1210       in_register = false;
1211     }
1212     else {
1213       parse_err(SYNERR, "invalid value inside return_address entry.\n");
1214       return;
1215     }
1216     if (native) { frame->_c_return_addr_loc = in_register; }
1217     else        { frame->_return_addr_loc   = in_register; }
1218 
1219     // Parse expression that specifies register or stack position
1220     skipws();
1221     char *token2 = get_expr("return address entry", ")");
1222     if (token2 == NULL) {
1223       parse_err(SYNERR, "missing value inside return address entry.\n");
1224       return;
1225     }
1226     next_char();           // skip the close paren
1227     if (native) { frame->_c_return_addr = token2; }
1228     else        { frame->_return_addr   = token2; }
1229 
1230     if(_curchar != ';') {  // check for semi-colon
1231       parse_err(SYNERR, "missing %c in return address entry.\n", ';');
1232       return;
1233     }
1234     next_char();           // skip the semi-colon
1235   }
1236   else {
1237     parse_err(SYNERR, "Missing %c in return_address entry.\n", '(');
1238   }
1239 }
1240 
1241 //------------------------------preserve_stack_parse---------------------------
preserve_stack_parse(FrameForm * frame)1242 void ADLParser::preserve_stack_parse(FrameForm *frame) {
1243   if(_curchar == '(') {
1244     char *token = get_paren_expr("preserve_stack_slots");
1245     frame->_in_preserve_slots   = token;
1246 
1247     if(_curchar != ';') {  // check for semi-colon
1248       parse_err(SYNERR, "missing %c in preserve stack slot entry.\n", ';');
1249       return;
1250     }
1251     next_char();           // skip the semi-colon
1252   }
1253   else {
1254     parse_err(SYNERR, "Missing %c in preserve stack slot entry.\n", '(');
1255   }
1256 }
1257 
1258 //------------------------------calling_convention_parse-----------------------
calling_convention_parse()1259 char *ADLParser::calling_convention_parse() {
1260   char   *desc = NULL;          // String representation of calling_convention
1261 
1262   skipws();                     // Skip leading whitespace
1263   if ( (desc = find_cpp_block("calling convention block")) == NULL ) {
1264     parse_err(SYNERR, "incorrect or missing block for 'calling_convention'.\n");
1265   }
1266   return desc;
1267 }
1268 
1269 //------------------------------return_value_parse-----------------------------
return_value_parse()1270 char *ADLParser::return_value_parse() {
1271   char   *desc = NULL;          // String representation of calling_convention
1272 
1273   skipws();                     // Skip leading whitespace
1274   if ( (desc = find_cpp_block("return value block")) == NULL ) {
1275     parse_err(SYNERR, "incorrect or missing block for 'return_value'.\n");
1276   }
1277   return desc;
1278 }
1279 
1280 //------------------------------ins_pipe_parse---------------------------------
ins_pipe_parse(InstructForm & instr)1281 void ADLParser::ins_pipe_parse(InstructForm &instr) {
1282   char * ident;
1283 
1284   skipws();
1285   if ( _curchar != '(' ) {       // Check for delimiter
1286     parse_err(SYNERR, "missing \"(\" in ins_pipe definition\n");
1287     return;
1288   }
1289 
1290   next_char();
1291   ident = get_ident();           // Grab next identifier
1292 
1293   if (ident == NULL) {
1294     parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
1295     return;
1296   }
1297 
1298   skipws();
1299   if ( _curchar != ')' ) {       // Check for delimiter
1300     parse_err(SYNERR, "missing \")\" in ins_pipe definition\n");
1301     return;
1302   }
1303 
1304   next_char();                   // skip the close paren
1305   if(_curchar != ';') {          // check for semi-colon
1306     parse_err(SYNERR, "missing %c in return value entry.\n", ';');
1307     return;
1308   }
1309   next_char();                   // skip the semi-colon
1310 
1311   // Check ident for validity
1312   if (_AD._pipeline && !_AD._pipeline->_classlist.search(ident)) {
1313     parse_err(SYNERR, "\"%s\" is not a valid pipeline class\n", ident);
1314     return;
1315   }
1316 
1317   // Add this instruction to the list in the pipeline class
1318   _AD._pipeline->_classdict[ident]->is_pipeclass()->_instructs.addName(instr._ident);
1319 
1320   // Set the name of the pipeline class in the instruction
1321   instr._ins_pipe = ident;
1322   return;
1323 }
1324 
1325 //------------------------------pipe_parse-------------------------------------
pipe_parse(void)1326 void ADLParser::pipe_parse(void) {
1327   PipelineForm *pipeline;         // Encode class for instruction/operand
1328   char * ident;
1329 
1330   pipeline = new PipelineForm();  // Build new Source object
1331   _AD.addForm(pipeline);
1332 
1333   skipws();                       // Skip leading whitespace
1334   // Check for block delimiter
1335   if ( (_curchar != '%')
1336        || ( next_char(),  (_curchar != '{')) ) {
1337     parse_err(SYNERR, "missing '%%{' in pipeline definition\n");
1338     return;
1339   }
1340   next_char();                     // Maintain the invariant
1341   do {
1342     ident = get_ident();           // Grab next identifier
1343     if (ident == NULL) {
1344       parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
1345       continue;
1346     }
1347     if      (!strcmp(ident, "resources" )) resource_parse(*pipeline);
1348     else if (!strcmp(ident, "pipe_desc" )) pipe_desc_parse(*pipeline);
1349     else if (!strcmp(ident, "pipe_class")) pipe_class_parse(*pipeline);
1350     else if (!strcmp(ident, "define")) {
1351       skipws();
1352       if ( (_curchar != '%')
1353            || ( next_char(),  (_curchar != '{')) ) {
1354         parse_err(SYNERR, "expected '%%{'\n");
1355         return;
1356       }
1357       next_char(); skipws();
1358 
1359       char *node_class = get_ident();
1360       if (node_class == NULL) {
1361         parse_err(SYNERR, "expected identifier, found \"%c\"\n", _curchar);
1362         return;
1363       }
1364 
1365       skipws();
1366       if (_curchar != ',' && _curchar != '=') {
1367         parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1368         break;
1369       }
1370       next_char(); skipws();
1371 
1372       char *pipe_class = get_ident();
1373       if (pipe_class == NULL) {
1374         parse_err(SYNERR, "expected identifier, found \"%c\"\n", _curchar);
1375         return;
1376       }
1377       if (_curchar != ';' ) {
1378         parse_err(SYNERR, "expected `;`, found '%c'\n", _curchar);
1379         break;
1380       }
1381       next_char();              // Skip over semi-colon
1382 
1383       skipws();
1384       if ( (_curchar != '%')
1385            || ( next_char(),  (_curchar != '}')) ) {
1386         parse_err(SYNERR, "expected '%%}', found \"%c\"\n", _curchar);
1387       }
1388       next_char();
1389 
1390       // Check ident for validity
1391       if (_AD._pipeline && !_AD._pipeline->_classlist.search(pipe_class)) {
1392         parse_err(SYNERR, "\"%s\" is not a valid pipeline class\n", pipe_class);
1393         return;
1394       }
1395 
1396       // Add this machine node to the list in the pipeline class
1397       _AD._pipeline->_classdict[pipe_class]->is_pipeclass()->_instructs.addName(node_class);
1398 
1399       MachNodeForm *machnode = new MachNodeForm(node_class); // Create new machnode form
1400       machnode->_machnode_pipe = pipe_class;
1401 
1402       _AD.addForm(machnode);
1403     }
1404     else if (!strcmp(ident, "attributes")) {
1405       bool vsi_seen = false;
1406 
1407       skipws();
1408       if ( (_curchar != '%')
1409            || ( next_char(),  (_curchar != '{')) ) {
1410         parse_err(SYNERR, "expected '%%{'\n");
1411         return;
1412       }
1413       next_char(); skipws();
1414 
1415       while (_curchar != '%') {
1416         ident = get_ident();
1417         if (ident == NULL)
1418           break;
1419 
1420         if (!strcmp(ident, "variable_size_instructions")) {
1421           skipws();
1422           if (_curchar == ';') {
1423             next_char(); skipws();
1424           }
1425 
1426           pipeline->_variableSizeInstrs = true;
1427           vsi_seen = true;
1428           continue;
1429         }
1430 
1431         if (!strcmp(ident, "fixed_size_instructions")) {
1432           skipws();
1433           if (_curchar == ';') {
1434             next_char(); skipws();
1435           }
1436 
1437           pipeline->_variableSizeInstrs = false;
1438           vsi_seen = true;
1439           continue;
1440         }
1441 
1442         if (!strcmp(ident, "branch_has_delay_slot")) {
1443           skipws();
1444           if (_curchar == ';') {
1445             next_char(); skipws();
1446           }
1447 
1448           pipeline->_branchHasDelaySlot = true;
1449           continue;
1450         }
1451 
1452         if (!strcmp(ident, "max_instructions_per_bundle")) {
1453           skipws();
1454           if (_curchar != '=') {
1455             parse_err(SYNERR, "expected `=`\n");
1456             break;
1457             }
1458 
1459           next_char(); skipws();
1460           pipeline->_maxInstrsPerBundle = get_int();
1461           skipws();
1462 
1463           if (_curchar == ';') {
1464             next_char(); skipws();
1465           }
1466 
1467           continue;
1468         }
1469 
1470         if (!strcmp(ident, "max_bundles_per_cycle")) {
1471           skipws();
1472           if (_curchar != '=') {
1473             parse_err(SYNERR, "expected `=`\n");
1474             break;
1475             }
1476 
1477           next_char(); skipws();
1478           pipeline->_maxBundlesPerCycle = get_int();
1479           skipws();
1480 
1481           if (_curchar == ';') {
1482             next_char(); skipws();
1483           }
1484 
1485           continue;
1486         }
1487 
1488         if (!strcmp(ident, "instruction_unit_size")) {
1489           skipws();
1490           if (_curchar != '=') {
1491             parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1492             break;
1493             }
1494 
1495           next_char(); skipws();
1496           pipeline->_instrUnitSize = get_int();
1497           skipws();
1498 
1499           if (_curchar == ';') {
1500             next_char(); skipws();
1501           }
1502 
1503           continue;
1504         }
1505 
1506         if (!strcmp(ident, "bundle_unit_size")) {
1507           skipws();
1508           if (_curchar != '=') {
1509             parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1510             break;
1511             }
1512 
1513           next_char(); skipws();
1514           pipeline->_bundleUnitSize = get_int();
1515           skipws();
1516 
1517           if (_curchar == ';') {
1518             next_char(); skipws();
1519           }
1520 
1521           continue;
1522         }
1523 
1524         if (!strcmp(ident, "instruction_fetch_unit_size")) {
1525           skipws();
1526           if (_curchar != '=') {
1527             parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1528             break;
1529             }
1530 
1531           next_char(); skipws();
1532           pipeline->_instrFetchUnitSize = get_int();
1533           skipws();
1534 
1535           if (_curchar == ';') {
1536             next_char(); skipws();
1537           }
1538 
1539           continue;
1540         }
1541 
1542         if (!strcmp(ident, "instruction_fetch_units")) {
1543           skipws();
1544           if (_curchar != '=') {
1545             parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1546             break;
1547             }
1548 
1549           next_char(); skipws();
1550           pipeline->_instrFetchUnits = get_int();
1551           skipws();
1552 
1553           if (_curchar == ';') {
1554             next_char(); skipws();
1555           }
1556 
1557           continue;
1558         }
1559 
1560         if (!strcmp(ident, "nops")) {
1561           skipws();
1562           if (_curchar != '(') {
1563             parse_err(SYNERR, "expected `(`, found '%c'\n", _curchar);
1564             break;
1565             }
1566 
1567           next_char(); skipws();
1568 
1569           while (_curchar != ')') {
1570             ident = get_ident();
1571             if (ident == NULL) {
1572               parse_err(SYNERR, "expected identifier for nop instruction, found '%c'\n", _curchar);
1573               break;
1574             }
1575 
1576             pipeline->_noplist.addName(ident);
1577             pipeline->_nopcnt++;
1578             skipws();
1579 
1580             if (_curchar == ',') {
1581               next_char(); skipws();
1582             }
1583           }
1584 
1585           next_char(); skipws();
1586 
1587           if (_curchar == ';') {
1588             next_char(); skipws();
1589           }
1590 
1591           continue;
1592         }
1593 
1594         parse_err(SYNERR, "unknown specifier \"%s\"\n", ident);
1595       }
1596 
1597       if ( (_curchar != '%')
1598            || ( next_char(),  (_curchar != '}')) ) {
1599         parse_err(SYNERR, "expected '%%}', found \"%c\"\n", _curchar);
1600       }
1601       next_char(); skipws();
1602 
1603       if (pipeline->_maxInstrsPerBundle == 0)
1604         parse_err(SYNERR, "\"max_instructions_per_bundle\" unspecified\n");
1605       if (pipeline->_instrUnitSize == 0 && pipeline->_bundleUnitSize == 0)
1606         parse_err(SYNERR, "\"instruction_unit_size\" and \"bundle_unit_size\" unspecified\n");
1607       if (pipeline->_instrFetchUnitSize == 0)
1608         parse_err(SYNERR, "\"instruction_fetch_unit_size\" unspecified\n");
1609       if (pipeline->_instrFetchUnits == 0)
1610         parse_err(SYNERR, "\"instruction_fetch_units\" unspecified\n");
1611       if (!vsi_seen)
1612         parse_err(SYNERR, "\"variable_size_instruction\" or \"fixed_size_instruction\" unspecified\n");
1613     }
1614     else {  // Done with staticly defined parts of instruction definition
1615       parse_err(SYNERR, "expected one of \"resources\", \"pipe_desc\", \"pipe_class\", found \"%s\"\n", ident);
1616       return;
1617     }
1618     skipws();
1619     if (_curchar == ';')
1620       skipws();
1621   } while(_curchar != '%');
1622 
1623   next_char();
1624   if (_curchar != '}') {
1625     parse_err(SYNERR, "missing \"%%}\" in pipeline definition\n");
1626     return;
1627   }
1628 
1629   next_char();
1630 }
1631 
1632 //------------------------------resource_parse----------------------------
resource_parse(PipelineForm & pipeline)1633 void ADLParser::resource_parse(PipelineForm &pipeline) {
1634   ResourceForm *resource;
1635   char * ident;
1636   char * expr;
1637   unsigned mask;
1638   pipeline._rescount = 0;
1639 
1640   skipws();                       // Skip leading whitespace
1641 
1642   if (_curchar != '(') {
1643     parse_err(SYNERR, "missing \"(\" in resource definition\n");
1644     return;
1645   }
1646 
1647   do {
1648     next_char();                   // Skip "(" or ","
1649     ident = get_ident();           // Grab next identifier
1650 
1651     if (_AD._adl_debug > 1) {
1652       if (ident != NULL) {
1653         fprintf(stderr, "resource_parse: identifier: %s\n", ident);
1654       }
1655     }
1656 
1657     if (ident == NULL) {
1658       parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1659       return;
1660     }
1661     skipws();
1662 
1663     if (_curchar != '=') {
1664       mask = (1 << pipeline._rescount++);
1665     }
1666     else {
1667       next_char(); skipws();
1668       expr = get_ident();          // Grab next identifier
1669       if (expr == NULL) {
1670         parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1671         return;
1672       }
1673       resource = (ResourceForm *) pipeline._resdict[expr];
1674       if (resource == NULL) {
1675         parse_err(SYNERR, "resource \"%s\" is not defined\n", expr);
1676         return;
1677       }
1678       mask = resource->mask();
1679 
1680       skipws();
1681       while (_curchar == '|') {
1682         next_char(); skipws();
1683 
1684         expr = get_ident();          // Grab next identifier
1685         if (expr == NULL) {
1686           parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1687           return;
1688         }
1689 
1690         resource = (ResourceForm *) pipeline._resdict[expr];   // Look up the value
1691         if (resource == NULL) {
1692           parse_err(SYNERR, "resource \"%s\" is not defined\n", expr);
1693           return;
1694         }
1695 
1696         mask |= resource->mask();
1697         skipws();
1698       }
1699     }
1700 
1701     resource = new ResourceForm(mask);
1702 
1703     pipeline._resdict.Insert(ident, resource);
1704     pipeline._reslist.addName(ident);
1705   } while (_curchar == ',');
1706 
1707   if (_curchar != ')') {
1708       parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
1709       return;
1710   }
1711 
1712   next_char();                 // Skip ")"
1713   if (_curchar == ';')
1714     next_char();               // Skip ";"
1715 }
1716 
1717 //------------------------------resource_parse----------------------------
pipe_desc_parse(PipelineForm & pipeline)1718 void ADLParser::pipe_desc_parse(PipelineForm &pipeline) {
1719   char * ident;
1720 
1721   skipws();                       // Skip leading whitespace
1722 
1723   if (_curchar != '(') {
1724     parse_err(SYNERR, "missing \"(\" in pipe_desc definition\n");
1725     return;
1726   }
1727 
1728   do {
1729     next_char();                   // Skip "(" or ","
1730     ident = get_ident();           // Grab next identifier
1731     if (ident == NULL) {
1732       parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1733       return;
1734     }
1735 
1736     // Add the name to the list
1737     pipeline._stages.addName(ident);
1738     pipeline._stagecnt++;
1739 
1740     skipws();
1741   } while (_curchar == ',');
1742 
1743   if (_curchar != ')') {
1744       parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
1745       return;
1746   }
1747 
1748   next_char();                     // Skip ")"
1749   if (_curchar == ';')
1750     next_char();                   // Skip ";"
1751 }
1752 
1753 //------------------------------pipe_class_parse--------------------------
pipe_class_parse(PipelineForm & pipeline)1754 void ADLParser::pipe_class_parse(PipelineForm &pipeline) {
1755   PipeClassForm *pipe_class;
1756   char * ident;
1757   char * stage;
1758   char * read_or_write;
1759   int is_write;
1760   int is_read;
1761   OperandForm  *oper;
1762 
1763   skipws();                       // Skip leading whitespace
1764 
1765   ident = get_ident();            // Grab next identifier
1766 
1767   if (ident == NULL) {
1768     parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1769     return;
1770   }
1771 
1772   // Create a record for the pipe_class
1773   pipe_class = new PipeClassForm(ident, ++pipeline._classcnt);
1774   pipeline._classdict.Insert(ident, pipe_class);
1775   pipeline._classlist.addName(ident);
1776 
1777   // Then get the operands
1778   skipws();
1779   if (_curchar != '(') {
1780     parse_err(SYNERR, "missing \"(\" in pipe_class definition\n");
1781   }
1782   // Parse the operand list
1783   else get_oplist(pipe_class->_parameters, pipe_class->_localNames);
1784   skipws();                        // Skip leading whitespace
1785   // Check for block delimiter
1786   if ( (_curchar != '%')
1787        || ( next_char(),  (_curchar != '{')) ) {
1788     parse_err(SYNERR, "missing \"%%{\" in pipe_class definition\n");
1789     return;
1790   }
1791   next_char();
1792 
1793   do {
1794     ident = get_ident();           // Grab next identifier
1795     if (ident == NULL) {
1796       parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1797       continue;
1798     }
1799     skipws();
1800 
1801     if (!strcmp(ident, "fixed_latency")) {
1802       skipws();
1803       if (_curchar != '(') {
1804         parse_err(SYNERR, "missing \"(\" in latency definition\n");
1805         return;
1806       }
1807       next_char(); skipws();
1808       if( !isdigit(_curchar) ) {
1809         parse_err(SYNERR, "number expected for \"%c\" in latency definition\n", _curchar);
1810         return;
1811       }
1812       int fixed_latency = get_int();
1813       skipws();
1814       if (_curchar != ')') {
1815         parse_err(SYNERR, "missing \")\" in latency definition\n");
1816         return;
1817       }
1818       next_char(); skipws();
1819       if (_curchar != ';') {
1820         parse_err(SYNERR, "missing \";\" in latency definition\n");
1821         return;
1822       }
1823 
1824       pipe_class->setFixedLatency(fixed_latency);
1825       next_char(); skipws();
1826       continue;
1827     }
1828 
1829     if (!strcmp(ident, "zero_instructions") ||
1830         !strcmp(ident, "no_instructions")) {
1831       skipws();
1832       if (_curchar != ';') {
1833         parse_err(SYNERR, "missing \";\" in latency definition\n");
1834         return;
1835       }
1836 
1837       pipe_class->setInstructionCount(0);
1838       next_char(); skipws();
1839       continue;
1840     }
1841 
1842     if (!strcmp(ident, "one_instruction_with_delay_slot") ||
1843         !strcmp(ident, "single_instruction_with_delay_slot")) {
1844       skipws();
1845       if (_curchar != ';') {
1846         parse_err(SYNERR, "missing \";\" in latency definition\n");
1847         return;
1848       }
1849 
1850       pipe_class->setInstructionCount(1);
1851       pipe_class->setBranchDelay(true);
1852       next_char(); skipws();
1853       continue;
1854     }
1855 
1856     if (!strcmp(ident, "one_instruction") ||
1857         !strcmp(ident, "single_instruction")) {
1858       skipws();
1859       if (_curchar != ';') {
1860         parse_err(SYNERR, "missing \";\" in latency definition\n");
1861         return;
1862       }
1863 
1864       pipe_class->setInstructionCount(1);
1865       next_char(); skipws();
1866       continue;
1867     }
1868 
1869     if (!strcmp(ident, "instructions_in_first_bundle") ||
1870         !strcmp(ident, "instruction_count")) {
1871       skipws();
1872 
1873       int number_of_instructions = 1;
1874 
1875       if (_curchar != '(') {
1876         parse_err(SYNERR, "\"(\" expected at \"%c\"\n", _curchar);
1877         continue;
1878       }
1879 
1880       next_char(); skipws();
1881       number_of_instructions = get_int();
1882 
1883       skipws();
1884       if (_curchar != ')') {
1885         parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
1886         continue;
1887       }
1888 
1889       next_char(); skipws();
1890       if (_curchar != ';') {
1891         parse_err(SYNERR, "missing \";\" in latency definition\n");
1892         return;
1893       }
1894 
1895       pipe_class->setInstructionCount(number_of_instructions);
1896       next_char(); skipws();
1897       continue;
1898     }
1899 
1900     if (!strcmp(ident, "multiple_bundles")) {
1901       skipws();
1902       if (_curchar != ';') {
1903         parse_err(SYNERR, "missing \";\" after multiple bundles\n");
1904         return;
1905       }
1906 
1907       pipe_class->setMultipleBundles(true);
1908       next_char(); skipws();
1909       continue;
1910     }
1911 
1912     if (!strcmp(ident, "has_delay_slot")) {
1913       skipws();
1914       if (_curchar != ';') {
1915         parse_err(SYNERR, "missing \";\" after \"has_delay_slot\"\n");
1916         return;
1917       }
1918 
1919       pipe_class->setBranchDelay(true);
1920       next_char(); skipws();
1921       continue;
1922     }
1923 
1924     if (!strcmp(ident, "force_serialization")) {
1925       skipws();
1926       if (_curchar != ';') {
1927         parse_err(SYNERR, "missing \";\" after \"force_serialization\"\n");
1928         return;
1929       }
1930 
1931       pipe_class->setForceSerialization(true);
1932       next_char(); skipws();
1933       continue;
1934     }
1935 
1936     if (!strcmp(ident, "may_have_no_code")) {
1937       skipws();
1938       if (_curchar != ';') {
1939         parse_err(SYNERR, "missing \";\" after \"may_have_no_code\"\n");
1940         return;
1941       }
1942 
1943       pipe_class->setMayHaveNoCode(true);
1944       next_char(); skipws();
1945       continue;
1946     }
1947 
1948     const Form *parm = pipe_class->_localNames[ident];
1949     if (parm != NULL) {
1950       oper = parm->is_operand();
1951       if (oper == NULL && !parm->is_opclass()) {
1952         parse_err(SYNERR, "operand name expected at %s\n", ident);
1953         continue;
1954       }
1955 
1956       if (_curchar != ':') {
1957         parse_err(SYNERR, "\":\" expected at \"%c\"\n", _curchar);
1958         continue;
1959       }
1960       next_char(); skipws();
1961       stage = get_ident();
1962       if (stage == NULL) {
1963         parse_err(SYNERR, "pipeline stage identifier expected at \"%c\"\n", _curchar);
1964         continue;
1965       }
1966 
1967       skipws();
1968       if (_curchar != '(') {
1969         parse_err(SYNERR, "\"(\" expected at \"%c\"\n", _curchar);
1970         continue;
1971       }
1972 
1973       next_char();
1974       read_or_write = get_ident();
1975       if (read_or_write == NULL) {
1976         parse_err(SYNERR, "\"read\" or \"write\" expected at \"%c\"\n", _curchar);
1977         continue;
1978       }
1979 
1980       is_read  = strcmp(read_or_write, "read")   == 0;
1981       is_write = strcmp(read_or_write, "write")  == 0;
1982       if (!is_read && !is_write) {
1983         parse_err(SYNERR, "\"read\" or \"write\" expected at \"%c\"\n", _curchar);
1984         continue;
1985       }
1986 
1987       skipws();
1988       if (_curchar != ')') {
1989         parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
1990         continue;
1991       }
1992 
1993       next_char(); skipws();
1994       int more_instrs = 0;
1995       if (_curchar == '+') {
1996           next_char(); skipws();
1997           if (_curchar < '0' || _curchar > '9') {
1998             parse_err(SYNERR, "<number> expected at \"%c\"\n", _curchar);
1999             continue;
2000           }
2001           while (_curchar >= '0' && _curchar <= '9') {
2002             more_instrs *= 10;
2003             more_instrs += _curchar - '0';
2004             next_char();
2005           }
2006           skipws();
2007       }
2008 
2009       PipeClassOperandForm *pipe_operand = new PipeClassOperandForm(stage, is_write, more_instrs);
2010       pipe_class->_localUsage.Insert(ident, pipe_operand);
2011 
2012       if (_curchar == '%')
2013           continue;
2014 
2015       if (_curchar != ';') {
2016         parse_err(SYNERR, "\";\" expected at \"%c\"\n", _curchar);
2017         continue;
2018       }
2019       next_char(); skipws();
2020       continue;
2021     }
2022 
2023     // Scan for Resource Specifier
2024     const Form *res = pipeline._resdict[ident];
2025     if (res != NULL) {
2026       int cyclecnt = 1;
2027       if (_curchar != ':') {
2028         parse_err(SYNERR, "\":\" expected at \"%c\"\n", _curchar);
2029         continue;
2030       }
2031       next_char(); skipws();
2032       stage = get_ident();
2033       if (stage == NULL) {
2034         parse_err(SYNERR, "pipeline stage identifier expected at \"%c\"\n", _curchar);
2035         continue;
2036       }
2037 
2038       skipws();
2039       if (_curchar == '(') {
2040         next_char();
2041         cyclecnt = get_int();
2042 
2043         skipws();
2044         if (_curchar != ')') {
2045           parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
2046           continue;
2047         }
2048 
2049         next_char(); skipws();
2050       }
2051 
2052       PipeClassResourceForm *resource = new PipeClassResourceForm(ident, stage, cyclecnt);
2053       int stagenum = pipeline._stages.index(stage);
2054       if (pipeline._maxcycleused < (stagenum+cyclecnt))
2055         pipeline._maxcycleused = (stagenum+cyclecnt);
2056       pipe_class->_resUsage.addForm(resource);
2057 
2058       if (_curchar == '%')
2059           continue;
2060 
2061       if (_curchar != ';') {
2062         parse_err(SYNERR, "\";\" expected at \"%c\"\n", _curchar);
2063         continue;
2064       }
2065       next_char(); skipws();
2066       continue;
2067     }
2068 
2069     parse_err(SYNERR, "resource expected at \"%s\"\n", ident);
2070     return;
2071   } while(_curchar != '%');
2072 
2073   next_char();
2074   if (_curchar != '}') {
2075     parse_err(SYNERR, "missing \"%%}\" in pipe_class definition\n");
2076     return;
2077   }
2078 
2079   next_char();
2080 }
2081 
2082 //------------------------------peep_parse-------------------------------------
peep_parse(void)2083 void ADLParser::peep_parse(void) {
2084   Peephole  *peep;                // Pointer to current peephole rule form
2085   char      *desc = NULL;         // String representation of rule
2086 
2087   skipws();                       // Skip leading whitespace
2088 
2089   peep = new Peephole();          // Build new Peephole object
2090   // Check for open block sequence
2091   skipws();                       // Skip leading whitespace
2092   if (_curchar == '%' && *(_ptr+1) == '{') {
2093     next_char(); next_char();     // Skip "%{"
2094     skipws();
2095     while (_curchar != '%' && *(_ptr+1) != '}') {
2096       char *token = get_ident();
2097       if (token == NULL) {
2098         parse_err(SYNERR, "missing identifier inside peephole rule.\n");
2099         return;
2100       }
2101       // check for legal subsections of peephole rule
2102       if (strcmp(token,"peepmatch")==0) {
2103         peep_match_parse(*peep); }
2104       else if (strcmp(token,"peepconstraint")==0) {
2105         peep_constraint_parse(*peep); }
2106       else if (strcmp(token,"peepreplace")==0) {
2107         peep_replace_parse(*peep); }
2108       else {
2109         parse_err(SYNERR, "expected peepmatch, peepconstraint, or peepreplace for identifier %s.\n", token);
2110       }
2111       skipws();
2112     }
2113   }
2114   else {
2115     parse_err(SYNERR, "Missing %%{ ... %%} block after peephole keyword.\n");
2116     return;
2117   }
2118   next_char();                    // Skip past '%'
2119   next_char();                    // Skip past '}'
2120 }
2121 
2122 // ******************** Private Level 2 Parse Functions ********************
2123 //------------------------------constraint_parse------------------------------
constraint_parse(void)2124 Constraint *ADLParser::constraint_parse(void) {
2125   char *func;
2126   char *arg;
2127 
2128   // Check for constraint expression
2129   skipws();
2130   if (_curchar != '(') {
2131     parse_err(SYNERR, "missing constraint expression, (...)\n");
2132     return NULL;
2133   }
2134   next_char();                    // Skip past '('
2135 
2136   // Get constraint function
2137   skipws();
2138   func = get_ident();
2139   if (func == NULL) {
2140     parse_err(SYNERR, "missing function in constraint expression.\n");
2141     return NULL;
2142   }
2143   if (strcmp(func,"ALLOC_IN_RC")==0
2144       || strcmp(func,"IS_R_CLASS")==0) {
2145     // Check for '(' before argument
2146     skipws();
2147     if (_curchar != '(') {
2148       parse_err(SYNERR, "missing '(' for constraint function's argument.\n");
2149       return NULL;
2150     }
2151     next_char();
2152 
2153     // Get it's argument
2154     skipws();
2155     arg = get_ident();
2156     if (arg == NULL) {
2157       parse_err(SYNERR, "missing argument for constraint function %s\n",func);
2158       return NULL;
2159     }
2160     // Check for ')' after argument
2161     skipws();
2162     if (_curchar != ')') {
2163       parse_err(SYNERR, "missing ')' after constraint function argument %s\n",arg);
2164       return NULL;
2165     }
2166     next_char();
2167   } else {
2168     parse_err(SYNERR, "Invalid constraint function %s\n",func);
2169     return NULL;
2170   }
2171 
2172   // Check for closing paren and ';'
2173   skipws();
2174   if (_curchar != ')') {
2175     parse_err(SYNERR, "Missing ')' for constraint function %s\n",func);
2176     return NULL;
2177   }
2178   next_char();
2179   skipws();
2180   if (_curchar != ';') {
2181     parse_err(SYNERR, "Missing ';' after constraint.\n");
2182     return NULL;
2183   }
2184   next_char();
2185 
2186   // Create new "Constraint"
2187   Constraint *constraint = new Constraint(func,arg);
2188   return constraint;
2189 }
2190 
2191 //------------------------------constr_parse-----------------------------------
construct_parse(void)2192 ConstructRule *ADLParser::construct_parse(void) {
2193   return NULL;
2194 }
2195 
2196 
2197 //------------------------------reg_def_parse----------------------------------
reg_def_parse(void)2198 void ADLParser::reg_def_parse(void) {
2199   char *rname;                   // Name of register being defined
2200 
2201   // Get register name
2202   skipws();                      // Skip whitespace
2203   rname = get_ident();
2204   if (rname == NULL) {
2205     parse_err(SYNERR, "missing register name after reg_def\n");
2206     return;
2207   }
2208 
2209   // Check for definition of register calling convention (save on call, ...),
2210   // register save type, and register encoding value.
2211   skipws();
2212   char *callconv  = NULL;
2213   char *c_conv    = NULL;
2214   char *idealtype = NULL;
2215   char *encoding  = NULL;
2216   char *concrete = NULL;
2217   if (_curchar == '(') {
2218     next_char();
2219     callconv = get_ident();
2220     // Parse the internal calling convention, must be NS, SOC, SOE, or AS.
2221     if (callconv == NULL) {
2222       parse_err(SYNERR, "missing register calling convention value\n");
2223       return;
2224     }
2225     if(strcmp(callconv, "SOC") && strcmp(callconv,"SOE") &&
2226        strcmp(callconv, "NS") && strcmp(callconv, "AS")) {
2227       parse_err(SYNERR, "invalid value for register calling convention\n");
2228     }
2229     skipws();
2230     if (_curchar != ',') {
2231       parse_err(SYNERR, "missing comma in register definition statement\n");
2232       return;
2233     }
2234     next_char();
2235 
2236     // Parse the native calling convention, must be NS, SOC, SOE, AS
2237     c_conv = get_ident();
2238     if (c_conv == NULL) {
2239       parse_err(SYNERR, "missing register native calling convention value\n");
2240       return;
2241     }
2242     if(strcmp(c_conv, "SOC") && strcmp(c_conv,"SOE") &&
2243        strcmp(c_conv, "NS") && strcmp(c_conv, "AS")) {
2244       parse_err(SYNERR, "invalid value for register calling convention\n");
2245     }
2246     skipws();
2247     if (_curchar != ',') {
2248       parse_err(SYNERR, "missing comma in register definition statement\n");
2249       return;
2250     }
2251     next_char();
2252     skipws();
2253 
2254     // Parse the ideal save type
2255     idealtype = get_ident();
2256     if (idealtype == NULL) {
2257       parse_err(SYNERR, "missing register save type value\n");
2258       return;
2259     }
2260     skipws();
2261     if (_curchar != ',') {
2262       parse_err(SYNERR, "missing comma in register definition statement\n");
2263       return;
2264     }
2265     next_char();
2266     skipws();
2267 
2268     // Parse the encoding value
2269     encoding = get_expr("encoding", ",");
2270     if (encoding == NULL) {
2271       parse_err(SYNERR, "missing register encoding value\n");
2272       return;
2273     }
2274     trim(encoding);
2275     if (_curchar != ',') {
2276       parse_err(SYNERR, "missing comma in register definition statement\n");
2277       return;
2278     }
2279     next_char();
2280     skipws();
2281     // Parse the concrete name type
2282     // concrete = get_ident();
2283     concrete = get_expr("concrete", ")");
2284     if (concrete == NULL) {
2285       parse_err(SYNERR, "missing vm register name value\n");
2286       return;
2287     }
2288 
2289     if (_curchar != ')') {
2290       parse_err(SYNERR, "missing ')' in register definition statement\n");
2291       return;
2292     }
2293     next_char();
2294   }
2295 
2296   // Check for closing ';'
2297   skipws();
2298   if (_curchar != ';') {
2299     parse_err(SYNERR, "missing ';' after reg_def\n");
2300     return;
2301   }
2302   next_char();                   // move past ';'
2303 
2304   // Debug Stuff
2305   if (_AD._adl_debug > 1) {
2306     fprintf(stderr,"Register Definition: %s ( %s, %s %s )\n", rname,
2307             (callconv ? callconv : ""), (c_conv ? c_conv : ""), concrete);
2308   }
2309 
2310   // Record new register definition.
2311   _AD._register->addRegDef(rname, callconv, c_conv, idealtype, encoding, concrete);
2312   return;
2313 }
2314 
2315 //------------------------------reg_class_parse--------------------------------
reg_class_parse(void)2316 void ADLParser::reg_class_parse(void) {
2317   char *cname;                    // Name of register class being defined
2318 
2319   // Get register class name
2320   skipws();                       // Skip leading whitespace
2321   cname = get_ident();
2322   if (cname == NULL) {
2323     parse_err(SYNERR, "missing register class name after 'reg_class'\n");
2324     return;
2325   }
2326   // Debug Stuff
2327   if (_AD._adl_debug >1) fprintf(stderr,"Register Class: %s\n", cname);
2328 
2329   skipws();
2330   if (_curchar == '(') {
2331     // A register list is defined for the register class.
2332     // Collect registers into a generic RegClass register class.
2333     RegClass* reg_class = _AD._register->addRegClass<RegClass>(cname);
2334 
2335     next_char();                  // Skip '('
2336     skipws();
2337     while (_curchar != ')') {
2338       char *rname = get_ident();
2339       if (rname==NULL) {
2340         parse_err(SYNERR, "missing identifier inside reg_class list.\n");
2341         return;
2342       }
2343       RegDef *regDef = _AD._register->getRegDef(rname);
2344       if (!regDef) {
2345         parse_err(SEMERR, "unknown identifier %s inside reg_class list.\n", rname);
2346       } else {
2347         reg_class->addReg(regDef); // add regDef to regClass
2348       }
2349 
2350       // Check for ',' and position to next token.
2351       skipws();
2352       if (_curchar == ',') {
2353         next_char();              // Skip trailing ','
2354         skipws();
2355       }
2356     }
2357     next_char();                  // Skip closing ')'
2358   } else if (_curchar == '%') {
2359     // A code snippet is defined for the register class.
2360     // Collect the code snippet into a CodeSnippetRegClass register class.
2361     CodeSnippetRegClass* reg_class = _AD._register->addRegClass<CodeSnippetRegClass>(cname);
2362     char *code = find_cpp_block("reg class");
2363     if (code == NULL) {
2364       parse_err(SYNERR, "missing code declaration for reg class.\n");
2365       return;
2366     }
2367     reg_class->set_code_snippet(code);
2368     return;
2369   }
2370 
2371   // Check for terminating ';'
2372   skipws();
2373   if (_curchar != ';') {
2374     parse_err(SYNERR, "missing ';' at end of reg_class definition.\n");
2375     return;
2376   }
2377   next_char();                    // Skip trailing ';'
2378 
2379   // Check RegClass size, must be <= 32 registers in class.
2380 
2381   return;
2382 }
2383 
2384 //------------------------------reg_class_dynamic_parse------------------------
reg_class_dynamic_parse(void)2385 void ADLParser::reg_class_dynamic_parse(void) {
2386   char *cname; // Name of dynamic register class being defined
2387 
2388   // Get register class name
2389   skipws();
2390   cname = get_ident();
2391   if (cname == NULL) {
2392     parse_err(SYNERR, "missing dynamic register class name after 'reg_class_dynamic'\n");
2393     return;
2394   }
2395 
2396   if (_AD._adl_debug > 1) {
2397     fprintf(stdout, "Dynamic Register Class: %s\n", cname);
2398   }
2399 
2400   skipws();
2401   if (_curchar != '(') {
2402     parse_err(SYNERR, "missing '(' at the beginning of reg_class_dynamic definition\n");
2403     return;
2404   }
2405   next_char();
2406   skipws();
2407 
2408   // Collect two register classes and the C++ code representing the condition code used to
2409   // select between the two classes into a ConditionalRegClass register class.
2410   ConditionalRegClass* reg_class = _AD._register->addRegClass<ConditionalRegClass>(cname);
2411   int i;
2412   for (i = 0; i < 2; i++) {
2413     char* name = get_ident();
2414     if (name == NULL) {
2415       parse_err(SYNERR, "missing class identifier inside reg_class_dynamic list.\n");
2416       return;
2417     }
2418     RegClass* rc = _AD._register->getRegClass(name);
2419     if (rc == NULL) {
2420       parse_err(SEMERR, "unknown identifier %s inside reg_class_dynamic list.\n", name);
2421     } else {
2422       reg_class->set_rclass_at_index(i, rc);
2423     }
2424 
2425     skipws();
2426     if (_curchar == ',') {
2427       next_char();
2428       skipws();
2429     } else {
2430       parse_err(SYNERR, "missing separator ',' inside reg_class_dynamic list.\n");
2431     }
2432   }
2433 
2434   // Collect the condition code.
2435   skipws();
2436   if (_curchar == '%') {
2437     char* code = find_cpp_block("reg class dynamic");
2438     if (code == NULL) {
2439        parse_err(SYNERR, "missing code declaration for reg_class_dynamic.\n");
2440        return;
2441     }
2442     reg_class->set_condition_code(code);
2443   } else {
2444     parse_err(SYNERR, "missing %% at the beginning of code block in reg_class_dynamic definition\n");
2445     return;
2446   }
2447 
2448   skipws();
2449   if (_curchar != ')') {
2450     parse_err(SYNERR, "missing ')' at the end of reg_class_dynamic definition\n");
2451     return;
2452   }
2453   next_char();
2454 
2455   skipws();
2456   if (_curchar != ';') {
2457     parse_err(SYNERR, "missing ';' at the end of reg_class_dynamic definition.\n");
2458     return;
2459   }
2460   next_char();                    // Skip trailing ';'
2461 
2462   return;
2463 }
2464 
2465 //------------------------------alloc_class_parse------------------------------
alloc_class_parse(void)2466 void ADLParser::alloc_class_parse(void) {
2467   char *name;                     // Name of allocation class being defined
2468 
2469   // Get allocation class name
2470   skipws();                       // Skip leading whitespace
2471   name = get_ident();
2472   if (name == NULL) {
2473     parse_err(SYNERR, "missing allocation class name after 'reg_class'\n");
2474     return;
2475   }
2476   // Debug Stuff
2477   if (_AD._adl_debug >1) fprintf(stderr,"Allocation Class: %s\n", name);
2478 
2479   AllocClass *alloc_class = _AD._register->addAllocClass(name);
2480 
2481   // Collect registers in class
2482   skipws();
2483   if (_curchar == '(') {
2484     next_char();                  // Skip '('
2485     skipws();
2486     while (_curchar != ')') {
2487       char *rname = get_ident();
2488       if (rname==NULL) {
2489         parse_err(SYNERR, "missing identifier inside reg_class list.\n");
2490         return;
2491       }
2492       // Check if name is a RegDef
2493       RegDef *regDef = _AD._register->getRegDef(rname);
2494       if (regDef) {
2495         alloc_class->addReg(regDef);   // add regDef to allocClass
2496       } else {
2497 
2498         // name must be a RegDef or a RegClass
2499         parse_err(SYNERR, "name %s should be a previously defined reg_def.\n", rname);
2500         return;
2501       }
2502 
2503       // Check for ',' and position to next token.
2504       skipws();
2505       if (_curchar == ',') {
2506         next_char();              // Skip trailing ','
2507         skipws();
2508       }
2509     }
2510     next_char();                  // Skip closing ')'
2511   }
2512 
2513   // Check for terminating ';'
2514   skipws();
2515   if (_curchar != ';') {
2516     parse_err(SYNERR, "missing ';' at end of reg_class definition.\n");
2517     return;
2518   }
2519   next_char();                    // Skip trailing ';'
2520 
2521   return;
2522 }
2523 
2524 //------------------------------peep_match_child_parse-------------------------
peep_match_child_parse(PeepMatch & match,int parent,int & position,int input)2525 InstructForm *ADLParser::peep_match_child_parse(PeepMatch &match, int parent, int &position, int input){
2526   char      *token  = NULL;
2527   int        lparen = 0;          // keep track of parenthesis nesting depth
2528   int        rparen = 0;          // position of instruction at this depth
2529   InstructForm *inst_seen  = NULL;
2530 
2531   // Walk the match tree,
2532   // Record <parent, position, instruction name, input position>
2533   while ( lparen >= rparen ) {
2534     skipws();
2535     // Left paren signals start of an input, collect with recursive call
2536     if (_curchar == '(') {
2537       ++lparen;
2538       next_char();
2539       ( void ) peep_match_child_parse(match, parent, position, rparen);
2540     }
2541     // Right paren signals end of an input, may be more
2542     else if (_curchar == ')') {
2543       ++rparen;
2544       if( rparen == lparen ) { // IF rparen matches an lparen I've seen
2545         next_char();           //    move past ')'
2546       } else {                 // ELSE leave ')' for parent
2547         assert( rparen == lparen + 1, "Should only see one extra ')'");
2548         // if an instruction was not specified for this paren-pair
2549         if( ! inst_seen ) {   // record signal entry
2550           match.add_instruction( parent, position, NameList::_signal, input );
2551           ++position;
2552         }
2553         // ++input;   // TEMPORARY
2554         return inst_seen;
2555       }
2556     }
2557     // if no parens, then check for instruction name
2558     // This instruction is the parent of a sub-tree
2559     else if ((token = get_ident_dup()) != NULL) {
2560       const Form *form = _AD._globalNames[token];
2561       if (form) {
2562         InstructForm *inst = form->is_instruction();
2563         // Record the first instruction at this level
2564         if( inst_seen == NULL ) {
2565           inst_seen = inst;
2566         }
2567         if (inst) {
2568           match.add_instruction( parent, position, token, input );
2569           parent = position;
2570           ++position;
2571         } else {
2572           parse_err(SYNERR, "instruction name expected at identifier %s.\n",
2573                     token);
2574           return inst_seen;
2575         }
2576       }
2577       else {
2578         parse_err(SYNERR, "missing identifier in peepmatch rule.\n");
2579         return NULL;
2580       }
2581     }
2582     else {
2583       parse_err(SYNERR, "missing identifier in peepmatch rule.\n");
2584       return NULL;
2585     }
2586 
2587   } // end while
2588 
2589   assert( false, "ShouldNotReachHere();");
2590   return NULL;
2591 }
2592 
2593 //------------------------------peep_match_parse-------------------------------
2594 // Syntax for a peepmatch rule
2595 //
2596 // peepmatch ( root_instr_name [(instruction subtree)] [,(instruction subtree)]* );
2597 //
peep_match_parse(Peephole & peep)2598 void ADLParser::peep_match_parse(Peephole &peep) {
2599 
2600   skipws();
2601   // Check the structure of the rule
2602   // Check for open paren
2603   if (_curchar != '(') {
2604     parse_err(SYNERR, "missing '(' at start of peepmatch rule.\n");
2605     return;
2606   }
2607   next_char();   // skip '('
2608 
2609   // Construct PeepMatch and parse the peepmatch rule.
2610   PeepMatch *match = new PeepMatch(_ptr);
2611   int  parent   = -1;                   // parent of root
2612   int  position = 0;                    // zero-based positions
2613   int  input    = 0;                    // input position in parent's operands
2614   InstructForm *root= peep_match_child_parse( *match, parent, position, input);
2615   if( root == NULL ) {
2616     parse_err(SYNERR, "missing instruction-name at start of peepmatch.\n");
2617     return;
2618   }
2619 
2620   if( _curchar != ')' ) {
2621     parse_err(SYNERR, "missing ')' at end of peepmatch.\n");
2622     return;
2623   }
2624   next_char();   // skip ')'
2625 
2626   // Check for closing semicolon
2627   skipws();
2628   if( _curchar != ';' ) {
2629     parse_err(SYNERR, "missing ';' at end of peepmatch.\n");
2630     return;
2631   }
2632   next_char();   // skip ';'
2633 
2634   // Store match into peep, and store peep into instruction
2635   peep.add_match(match);
2636   root->append_peephole(&peep);
2637 }
2638 
2639 //------------------------------peep_constraint_parse--------------------------
2640 // Syntax for a peepconstraint rule
2641 // A parenthesized list of relations between operands in peepmatch subtree
2642 //
2643 // peepconstraint %{
2644 // (instruction_number.operand_name
2645 //     relational_op
2646 //  instruction_number.operand_name OR register_name
2647 //  [, ...] );
2648 //
2649 // // instruction numbers are zero-based using topological order in peepmatch
2650 //
peep_constraint_parse(Peephole & peep)2651 void ADLParser::peep_constraint_parse(Peephole &peep) {
2652 
2653   skipws();
2654   // Check the structure of the rule
2655   // Check for open paren
2656   if (_curchar != '(') {
2657     parse_err(SYNERR, "missing '(' at start of peepconstraint rule.\n");
2658     return;
2659   }
2660   else {
2661     next_char();                  // Skip '('
2662   }
2663 
2664   // Check for a constraint
2665   skipws();
2666   while( _curchar != ')' ) {
2667     // Get information on the left instruction and its operand
2668     // left-instructions's number
2669     int left_inst = get_int();
2670     // Left-instruction's operand
2671     skipws();
2672     if( _curchar != '.' ) {
2673       parse_err(SYNERR, "missing '.' in peepconstraint after instruction number.\n");
2674       return;
2675     }
2676     next_char();                  // Skip '.'
2677     char *left_op = get_ident_dup();
2678 
2679     skipws();
2680     // Collect relational operator
2681     char *relation = get_relation_dup();
2682 
2683     skipws();
2684     // Get information on the right instruction and its operand
2685     int right_inst;        // Right-instructions's number
2686     if( isdigit(_curchar) ) {
2687       right_inst = get_int();
2688       // Right-instruction's operand
2689       skipws();
2690       if( _curchar != '.' ) {
2691         parse_err(SYNERR, "missing '.' in peepconstraint after instruction number.\n");
2692         return;
2693       }
2694       next_char();              // Skip '.'
2695     } else {
2696       right_inst = -1;          // Flag as being a register constraint
2697     }
2698 
2699     char *right_op = get_ident_dup();
2700 
2701     // Construct the next PeepConstraint
2702     PeepConstraint *constraint = new PeepConstraint( left_inst, left_op,
2703                                                      relation,
2704                                                      right_inst, right_op );
2705     // And append it to the list for this peephole rule
2706     peep.append_constraint( constraint );
2707 
2708     // Check for another constraint, or end of rule
2709     skipws();
2710     if( _curchar == ',' ) {
2711       next_char();                // Skip ','
2712       skipws();
2713     }
2714     else if( _curchar != ')' ) {
2715       parse_err(SYNERR, "expected ',' or ')' after peephole constraint.\n");
2716       return;
2717     }
2718   } // end while( processing constraints )
2719   next_char();                    // Skip ')'
2720 
2721   // Check for terminating ';'
2722   skipws();
2723   if (_curchar != ';') {
2724     parse_err(SYNERR, "missing ';' at end of peepconstraint.\n");
2725     return;
2726   }
2727   next_char();                    // Skip trailing ';'
2728 }
2729 
2730 
2731 //------------------------------peep_replace_parse-----------------------------
2732 // Syntax for a peepreplace rule
2733 // root instruction name followed by a
2734 // parenthesized list of whitespace separated instruction.operand specifiers
2735 //
2736 // peepreplace ( instr_name  ( [instruction_number.operand_name]* ) );
2737 //
2738 //
peep_replace_parse(Peephole & peep)2739 void ADLParser::peep_replace_parse(Peephole &peep) {
2740   int          lparen = 0;        // keep track of parenthesis nesting depth
2741   int          rparen = 0;        // keep track of parenthesis nesting depth
2742   int          icount = 0;        // count of instructions in rule for naming
2743   char        *str    = NULL;
2744   char        *token  = NULL;
2745 
2746   skipws();
2747   // Check for open paren
2748   if (_curchar != '(') {
2749     parse_err(SYNERR, "missing '(' at start of peepreplace rule.\n");
2750     return;
2751   }
2752   else {
2753     lparen++;
2754     next_char();
2755   }
2756 
2757   // Check for root instruction
2758   char       *inst = get_ident_dup();
2759   const Form *form = _AD._globalNames[inst];
2760   if( form == NULL || form->is_instruction() == NULL ) {
2761     parse_err(SYNERR, "Instruction name expected at start of peepreplace.\n");
2762     return;
2763   }
2764 
2765   // Store string representation of rule into replace
2766   PeepReplace *replace = new PeepReplace(str);
2767   replace->add_instruction( inst );
2768 
2769   skipws();
2770   // Start of root's operand-list
2771   if (_curchar != '(') {
2772     parse_err(SYNERR, "missing '(' at peepreplace root's operand-list.\n");
2773     return;
2774   }
2775   else {
2776     lparen++;
2777     next_char();
2778   }
2779 
2780   skipws();
2781   // Get the list of operands
2782   while( _curchar != ')' ) {
2783     // Get information on an instruction and its operand
2784     // instructions's number
2785     int   inst_num = get_int();
2786     // Left-instruction's operand
2787     skipws();
2788     if( _curchar != '.' ) {
2789       parse_err(SYNERR, "missing '.' in peepreplace after instruction number.\n");
2790       return;
2791     }
2792     next_char();                  // Skip '.'
2793     char *inst_op = get_ident_dup();
2794     if( inst_op == NULL ) {
2795       parse_err(SYNERR, "missing operand identifier in peepreplace.\n");
2796       return;
2797     }
2798 
2799     // Record this operand's position in peepmatch
2800     replace->add_operand( inst_num, inst_op );
2801     skipws();
2802   }
2803 
2804   // Check for the end of operands list
2805   skipws();
2806   assert( _curchar == ')', "While loop should have advanced to ')'.");
2807   next_char();  // Skip ')'
2808 
2809   skipws();
2810   // Check for end of peepreplace
2811   if( _curchar != ')' ) {
2812     parse_err(SYNERR, "missing ')' at end of peepmatch.\n");
2813     parse_err(SYNERR, "Support one replacement instruction.\n");
2814     return;
2815   }
2816   next_char(); // Skip ')'
2817 
2818   // Check for closing semicolon
2819   skipws();
2820   if( _curchar != ';' ) {
2821     parse_err(SYNERR, "missing ';' at end of peepreplace.\n");
2822     return;
2823   }
2824   next_char();   // skip ';'
2825 
2826   // Store replace into peep
2827   peep.add_replace( replace );
2828 }
2829 
2830 //------------------------------pred_parse-------------------------------------
pred_parse(void)2831 Predicate *ADLParser::pred_parse(void) {
2832   Predicate *predicate;           // Predicate class for operand
2833   char      *rule = NULL;         // String representation of predicate
2834 
2835   skipws();                       // Skip leading whitespace
2836   int line = linenum();
2837   if ( (rule = get_paren_expr("pred expression", true)) == NULL ) {
2838     parse_err(SYNERR, "incorrect or missing expression for 'predicate'\n");
2839     return NULL;
2840   }
2841   // Debug Stuff
2842   if (_AD._adl_debug > 1) fprintf(stderr,"Predicate: %s\n", rule);
2843   if (_curchar != ';') {
2844     parse_err(SYNERR, "missing ';' in predicate definition\n");
2845     return NULL;
2846   }
2847   next_char();                     // Point after the terminator
2848 
2849   predicate = new Predicate(rule); // Build new predicate object
2850   skipws();
2851   return predicate;
2852 }
2853 
2854 
2855 //------------------------------ins_encode_parse_block-------------------------
2856 // Parse the block form of ins_encode.  See ins_encode_parse for more details
ins_encode_parse_block(InstructForm & inst)2857 void ADLParser::ins_encode_parse_block(InstructForm& inst) {
2858   // Create a new encoding name based on the name of the instruction
2859   // definition, which should be unique.
2860   const char* prefix = "__ins_encode_";
2861   char* ec_name = (char*) AllocateHeap(strlen(inst._ident) + strlen(prefix) + 1);
2862   sprintf(ec_name, "%s%s", prefix, inst._ident);
2863 
2864   assert(_AD._encode->encClass(ec_name) == NULL, "shouldn't already exist");
2865   EncClass* encoding = _AD._encode->add_EncClass(ec_name);
2866   encoding->_linenum = linenum();
2867 
2868   // synthesize the arguments list for the enc_class from the
2869   // arguments to the instruct definition.
2870   const char* param = NULL;
2871   inst._parameters.reset();
2872   while ((param = inst._parameters.iter()) != NULL) {
2873     OpClassForm* opForm = inst._localNames[param]->is_opclass();
2874     assert(opForm != NULL, "sanity");
2875     encoding->add_parameter(opForm->_ident, param);
2876   }
2877 
2878   if (!inst._is_postalloc_expand) {
2879     // Define a MacroAssembler instance for use by the encoding.  The
2880     // name is chosen to match the __ idiom used for assembly in other
2881     // parts of hotspot and assumes the existence of the standard
2882     // #define __ _masm.
2883     encoding->add_code("    MacroAssembler _masm(&cbuf);\n");
2884   }
2885 
2886   // Parse the following %{ }% block
2887   ins_encode_parse_block_impl(inst, encoding, ec_name);
2888 
2889   // Build an encoding rule which invokes the encoding rule we just
2890   // created, passing all arguments that we received.
2891   InsEncode*   encrule = new InsEncode(); // Encode class for instruction
2892   NameAndList* params  = encrule->add_encode(ec_name);
2893   inst._parameters.reset();
2894   while ((param = inst._parameters.iter()) != NULL) {
2895     params->add_entry(param);
2896   }
2897 
2898   // Check for duplicate ins_encode sections after parsing the block
2899   // so that parsing can continue and find any other errors.
2900   if (inst._insencode != NULL) {
2901     parse_err(SYNERR, "Multiple ins_encode sections defined\n");
2902     return;
2903   }
2904 
2905   // Set encode class of this instruction.
2906   inst._insencode = encrule;
2907 }
2908 
2909 
ins_encode_parse_block_impl(InstructForm & inst,EncClass * encoding,char * ec_name)2910 void ADLParser::ins_encode_parse_block_impl(InstructForm& inst, EncClass* encoding, char* ec_name) {
2911   skipws_no_preproc();              // Skip leading whitespace
2912   // Prepend location descriptor, for debugging; cf. ADLParser::find_cpp_block
2913   if (_AD._adlocation_debug) {
2914     encoding->add_code(get_line_string());
2915   }
2916 
2917   // Collect the parts of the encode description
2918   // (1) strings that are passed through to output
2919   // (2) replacement/substitution variable, preceeded by a '$'
2920   while ((_curchar != '%') && (*(_ptr+1) != '}')) {
2921 
2922     // (1)
2923     // Check if there is a string to pass through to output
2924     char *start = _ptr;       // Record start of the next string
2925     while ((_curchar != '$') && ((_curchar != '%') || (*(_ptr+1) != '}')) ) {
2926       // If at the start of a comment, skip past it
2927       if( (_curchar == '/') && ((*(_ptr+1) == '/') || (*(_ptr+1) == '*')) ) {
2928         skipws_no_preproc();
2929       } else {
2930         // ELSE advance to the next character, or start of the next line
2931         next_char_or_line();
2932       }
2933     }
2934     // If a string was found, terminate it and record in EncClass
2935     if (start != _ptr) {
2936       *_ptr = '\0';          // Terminate the string
2937       encoding->add_code(start);
2938     }
2939 
2940     // (2)
2941     // If we are at a replacement variable,
2942     // copy it and record in EncClass
2943     if (_curchar == '$') {
2944       // Found replacement Variable
2945       char* rep_var = get_rep_var_ident_dup();
2946 
2947       // Add flag to _strings list indicating we should check _rep_vars
2948       encoding->add_rep_var(rep_var);
2949 
2950       skipws();
2951 
2952       // Check if this instruct is a MachConstantNode.
2953       if (strcmp(rep_var, "constanttablebase") == 0) {
2954         // This instruct is a MachConstantNode.
2955         inst.set_needs_constant_base(true);
2956         if (strncmp("MachCall", inst.mach_base_class(_globalNames), strlen("MachCall")) != 0 ) {
2957           inst.set_is_mach_constant(true);
2958         }
2959 
2960         if (_curchar == '(')  {
2961           parse_err(SYNERR, "constanttablebase in instruct %s cannot have an argument "
2962                             "(only constantaddress and constantoffset)", ec_name);
2963           return;
2964         }
2965       }
2966       else if ((strcmp(rep_var, "constantaddress")   == 0) ||
2967                (strcmp(rep_var, "constantoffset")    == 0)) {
2968         // This instruct is a MachConstantNode.
2969         inst.set_is_mach_constant(true);
2970 
2971         // If the constant keyword has an argument, parse it.
2972         if (_curchar == '(')  constant_parse(inst);
2973       }
2974     }
2975   } // end while part of format description
2976   next_char();                      // Skip '%'
2977   next_char();                      // Skip '}'
2978 
2979   skipws();
2980 
2981   if (_AD._adlocation_debug) {
2982     encoding->add_code(end_line_marker());
2983   }
2984 
2985   // Debug Stuff
2986   if (_AD._adl_debug > 1)  fprintf(stderr, "EncodingClass Form: %s\n", ec_name);
2987 }
2988 
2989 
2990 //------------------------------ins_encode_parse-------------------------------
2991 // Encode rules have the form
2992 //   ins_encode( encode_class_name(parameter_list), ... );
2993 //
2994 // The "encode_class_name" must be defined in the encode section
2995 // The parameter list contains $names that are locals.
2996 //
2997 // Alternatively it can be written like this:
2998 //
2999 //   ins_encode %{
3000 //      ... // body
3001 //   %}
3002 //
3003 // which synthesizes a new encoding class taking the same arguments as
3004 // the InstructForm, and automatically prefixes the definition with:
3005 //
3006 //    MacroAssembler masm(&cbuf);\n");
3007 //
3008 //  making it more compact to take advantage of the MacroAssembler and
3009 //  placing the assembly closer to it's use by instructions.
ins_encode_parse(InstructForm & inst)3010 void ADLParser::ins_encode_parse(InstructForm& inst) {
3011 
3012   // Parse encode class name
3013   skipws();                        // Skip whitespace
3014   if (_curchar != '(') {
3015     // Check for ins_encode %{ form
3016     if ((_curchar == '%') && (*(_ptr+1) == '{')) {
3017       next_char();                      // Skip '%'
3018       next_char();                      // Skip '{'
3019 
3020       // Parse the block form of ins_encode
3021       ins_encode_parse_block(inst);
3022       return;
3023     }
3024 
3025     parse_err(SYNERR, "missing '%%{' or '(' in ins_encode definition\n");
3026     return;
3027   }
3028   next_char();                     // move past '('
3029   skipws();
3030 
3031   InsEncode *encrule  = new InsEncode(); // Encode class for instruction
3032   encrule->_linenum = linenum();
3033   char      *ec_name  = NULL;      // String representation of encode rule
3034   // identifier is optional.
3035   while (_curchar != ')') {
3036     ec_name = get_ident();
3037     if (ec_name == NULL) {
3038       parse_err(SYNERR, "Invalid encode class name after 'ins_encode('.\n");
3039       return;
3040     }
3041     // Check that encoding is defined in the encode section
3042     EncClass *encode_class = _AD._encode->encClass(ec_name);
3043     if (encode_class == NULL) {
3044       // Like to defer checking these till later...
3045       // parse_err(WARN, "Using an undefined encode class '%s' in 'ins_encode'.\n", ec_name);
3046     }
3047 
3048     // Get list for encode method's parameters
3049     NameAndList *params = encrule->add_encode(ec_name);
3050 
3051     // Parse the parameters to this encode method.
3052     skipws();
3053     if ( _curchar == '(' ) {
3054       next_char();                 // move past '(' for parameters
3055 
3056       // Parse the encode method's parameters
3057       while (_curchar != ')') {
3058         char *param = get_ident_or_literal_constant("encoding operand");
3059         if ( param != NULL ) {
3060 
3061           // Check if this instruct is a MachConstantNode.
3062           if (strcmp(param, "constanttablebase") == 0) {
3063             // This instruct is a MachConstantNode.
3064             inst.set_needs_constant_base(true);
3065             if (strncmp("MachCall", inst.mach_base_class(_globalNames), strlen("MachCall")) != 0 ) {
3066               inst.set_is_mach_constant(true);
3067             }
3068 
3069             if (_curchar == '(')  {
3070               parse_err(SYNERR, "constanttablebase in instruct %s cannot have an argument "
3071                         "(only constantaddress and constantoffset)", ec_name);
3072               return;
3073             }
3074           } else {
3075             // Found a parameter:
3076             // Check it is a local name, add it to the list, then check for more
3077             // New: allow hex constants as parameters to an encode method.
3078             // New: allow parenthesized expressions as parameters.
3079             // New: allow "primary", "secondary", "tertiary" as parameters.
3080             // New: allow user-defined register name as parameter
3081             if ( (inst._localNames[param] == NULL) &&
3082                  !ADLParser::is_literal_constant(param) &&
3083                  (Opcode::as_opcode_type(param) == Opcode::NOT_AN_OPCODE) &&
3084                  ((_AD._register == NULL ) || (_AD._register->getRegDef(param) == NULL)) ) {
3085               parse_err(SYNERR, "Using non-locally defined parameter %s for encoding %s.\n", param, ec_name);
3086               return;
3087             }
3088           }
3089           params->add_entry(param);
3090 
3091           skipws();
3092           if (_curchar == ',' ) {
3093             // More parameters to come
3094             next_char();           // move past ',' between parameters
3095             skipws();              // Skip to next parameter
3096           }
3097           else if (_curchar == ')') {
3098             // Done with parameter list
3099           }
3100           else {
3101             // Only ',' or ')' are valid after a parameter name
3102             parse_err(SYNERR, "expected ',' or ')' after parameter %s.\n",
3103                       ec_name);
3104             return;
3105           }
3106 
3107         } else {
3108           skipws();
3109           // Did not find a parameter
3110           if (_curchar == ',') {
3111             parse_err(SYNERR, "Expected encode parameter before ',' in encoding %s.\n", ec_name);
3112             return;
3113           }
3114           if (_curchar != ')') {
3115             parse_err(SYNERR, "Expected ')' after encode parameters.\n");
3116             return;
3117           }
3118         }
3119       } // WHILE loop collecting parameters
3120       next_char();                   // move past ')' at end of parameters
3121     } // done with parameter list for encoding
3122 
3123     // Check for ',' or ')' after encoding
3124     skipws();                      // move to character after parameters
3125     if ( _curchar == ',' ) {
3126       // Found a ','
3127       next_char();                 // move past ',' between encode methods
3128       skipws();
3129     }
3130     else if ( _curchar != ')' ) {
3131       // If not a ',' then only a ')' is allowed
3132       parse_err(SYNERR, "Expected ')' after encoding %s.\n", ec_name);
3133       return;
3134     }
3135 
3136     // Check for ',' separating parameters
3137     // if ( _curchar != ',' && _curchar != ')' ) {
3138     //   parse_err(SYNERR, "expected ',' or ')' after encode method inside ins_encode.\n");
3139     //   return NULL;
3140     // }
3141 
3142   } // done parsing ins_encode methods and their parameters
3143   if (_curchar != ')') {
3144     parse_err(SYNERR, "Missing ')' at end of ins_encode description.\n");
3145     return;
3146   }
3147   next_char();                     // move past ')'
3148   skipws();                        // Skip leading whitespace
3149 
3150   if ( _curchar != ';' ) {
3151     parse_err(SYNERR, "Missing ';' at end of ins_encode.\n");
3152     return;
3153   }
3154   next_char();                     // move past ';'
3155   skipws();                        // be friendly to oper_parse()
3156 
3157   // Check for duplicate ins_encode sections after parsing the block
3158   // so that parsing can continue and find any other errors.
3159   if (inst._insencode != NULL) {
3160     parse_err(SYNERR, "Multiple ins_encode sections defined\n");
3161     return;
3162   }
3163 
3164   // Debug Stuff
3165   if (_AD._adl_debug > 1) fprintf(stderr,"Instruction Encode: %s\n", ec_name);
3166 
3167   // Set encode class of this instruction.
3168   inst._insencode = encrule;
3169 }
3170 
3171 //------------------------------postalloc_expand_parse---------------------------
3172 // Encode rules have the form
3173 //   postalloc_expand( encode_class_name(parameter_list) );
3174 //
3175 // The "encode_class_name" must be defined in the encode section.
3176 // The parameter list contains $names that are locals.
3177 //
3178 // This is just a copy of ins_encode_parse without the loop.
postalloc_expand_parse(InstructForm & inst)3179 void ADLParser::postalloc_expand_parse(InstructForm& inst) {
3180   inst._is_postalloc_expand = true;
3181 
3182   // Parse encode class name.
3183   skipws();                        // Skip whitespace.
3184   if (_curchar != '(') {
3185     // Check for postalloc_expand %{ form
3186     if ((_curchar == '%') && (*(_ptr+1) == '{')) {
3187       next_char();                      // Skip '%'
3188       next_char();                      // Skip '{'
3189 
3190       // Parse the block form of postalloc_expand
3191       ins_encode_parse_block(inst);
3192       return;
3193     }
3194 
3195     parse_err(SYNERR, "missing '(' in postalloc_expand definition\n");
3196     return;
3197   }
3198   next_char();                     // Move past '('.
3199   skipws();
3200 
3201   InsEncode *encrule = new InsEncode(); // Encode class for instruction.
3202   encrule->_linenum = linenum();
3203   char      *ec_name = NULL;       // String representation of encode rule.
3204   // identifier is optional.
3205   if (_curchar != ')') {
3206     ec_name = get_ident();
3207     if (ec_name == NULL) {
3208       parse_err(SYNERR, "Invalid postalloc_expand class name after 'postalloc_expand('.\n");
3209       return;
3210     }
3211     // Check that encoding is defined in the encode section.
3212     EncClass *encode_class = _AD._encode->encClass(ec_name);
3213 
3214     // Get list for encode method's parameters
3215     NameAndList *params = encrule->add_encode(ec_name);
3216 
3217     // Parse the parameters to this encode method.
3218     skipws();
3219     if (_curchar == '(') {
3220       next_char();                 // Move past '(' for parameters.
3221 
3222       // Parse the encode method's parameters.
3223       while (_curchar != ')') {
3224         char *param = get_ident_or_literal_constant("encoding operand");
3225         if (param != NULL) {
3226           // Found a parameter:
3227 
3228           // First check for constant table support.
3229 
3230           // Check if this instruct is a MachConstantNode.
3231           if (strcmp(param, "constanttablebase") == 0) {
3232             // This instruct is a MachConstantNode.
3233             inst.set_needs_constant_base(true);
3234             if (strncmp("MachCall", inst.mach_base_class(_globalNames), strlen("MachCall")) != 0 ) {
3235               inst.set_is_mach_constant(true);
3236             }
3237 
3238             if (_curchar == '(') {
3239               parse_err(SYNERR, "constanttablebase in instruct %s cannot have an argument "
3240                         "(only constantaddress and constantoffset)", ec_name);
3241               return;
3242             }
3243           }
3244           else if ((strcmp(param, "constantaddress") == 0) ||
3245                    (strcmp(param, "constantoffset")  == 0))  {
3246             // This instruct is a MachConstantNode.
3247             inst.set_is_mach_constant(true);
3248 
3249             // If the constant keyword has an argument, parse it.
3250             if (_curchar == '(') constant_parse(inst);
3251           }
3252 
3253           // Else check it is a local name, add it to the list, then check for more.
3254           // New: allow hex constants as parameters to an encode method.
3255           // New: allow parenthesized expressions as parameters.
3256           // New: allow "primary", "secondary", "tertiary" as parameters.
3257           // New: allow user-defined register name as parameter.
3258           else if ((inst._localNames[param] == NULL) &&
3259                    !ADLParser::is_literal_constant(param) &&
3260                    (Opcode::as_opcode_type(param) == Opcode::NOT_AN_OPCODE) &&
3261                    ((_AD._register == NULL) || (_AD._register->getRegDef(param) == NULL))) {
3262             parse_err(SYNERR, "Using non-locally defined parameter %s for encoding %s.\n", param, ec_name);
3263             return;
3264           }
3265           params->add_entry(param);
3266 
3267           skipws();
3268           if (_curchar == ',') {
3269             // More parameters to come.
3270             next_char();           // Move past ',' between parameters.
3271             skipws();              // Skip to next parameter.
3272           } else if (_curchar == ')') {
3273             // Done with parameter list
3274           } else {
3275             // Only ',' or ')' are valid after a parameter name.
3276             parse_err(SYNERR, "expected ',' or ')' after parameter %s.\n", ec_name);
3277             return;
3278           }
3279 
3280         } else {
3281           skipws();
3282           // Did not find a parameter.
3283           if (_curchar == ',') {
3284             parse_err(SYNERR, "Expected encode parameter before ',' in postalloc_expand %s.\n", ec_name);
3285             return;
3286           }
3287           if (_curchar != ')') {
3288             parse_err(SYNERR, "Expected ')' after postalloc_expand parameters.\n");
3289             return;
3290           }
3291         }
3292       } // WHILE loop collecting parameters.
3293       next_char();                 // Move past ')' at end of parameters.
3294     } // Done with parameter list for encoding.
3295 
3296     // Check for ',' or ')' after encoding.
3297     skipws();                      // Move to character after parameters.
3298     if (_curchar != ')') {
3299       // Only a ')' is allowed.
3300       parse_err(SYNERR, "Expected ')' after postalloc_expand %s.\n", ec_name);
3301       return;
3302     }
3303   } // Done parsing postalloc_expand method and their parameters.
3304   if (_curchar != ')') {
3305     parse_err(SYNERR, "Missing ')' at end of postalloc_expand description.\n");
3306     return;
3307   }
3308   next_char();                     // Move past ')'.
3309   skipws();                        // Skip leading whitespace.
3310 
3311   if (_curchar != ';') {
3312     parse_err(SYNERR, "Missing ';' at end of postalloc_expand.\n");
3313     return;
3314   }
3315   next_char();                     // Move past ';'.
3316   skipws();                        // Be friendly to oper_parse().
3317 
3318   // Debug Stuff.
3319   if (_AD._adl_debug > 1) fprintf(stderr, "Instruction postalloc_expand: %s\n", ec_name);
3320 
3321   // Set encode class of this instruction.
3322   inst._insencode = encrule;
3323 }
3324 
3325 
3326 //------------------------------constant_parse---------------------------------
3327 // Parse a constant expression.
constant_parse(InstructForm & inst)3328 void ADLParser::constant_parse(InstructForm& inst) {
3329   // Create a new encoding name based on the name of the instruction
3330   // definition, which should be unique.
3331   const char* prefix = "__constant_";
3332   char* ec_name = (char*) AllocateHeap(strlen(inst._ident) + strlen(prefix) + 1);
3333   sprintf(ec_name, "%s%s", prefix, inst._ident);
3334 
3335   assert(_AD._encode->encClass(ec_name) == NULL, "shouldn't already exist");
3336   EncClass* encoding = _AD._encode->add_EncClass(ec_name);
3337   encoding->_linenum = linenum();
3338 
3339   // synthesize the arguments list for the enc_class from the
3340   // arguments to the instruct definition.
3341   const char* param = NULL;
3342   inst._parameters.reset();
3343   while ((param = inst._parameters.iter()) != NULL) {
3344     OpClassForm* opForm = inst._localNames[param]->is_opclass();
3345     assert(opForm != NULL, "sanity");
3346     encoding->add_parameter(opForm->_ident, param);
3347   }
3348 
3349   // Parse the following ( ) expression.
3350   constant_parse_expression(encoding, ec_name);
3351 
3352   // Build an encoding rule which invokes the encoding rule we just
3353   // created, passing all arguments that we received.
3354   InsEncode*   encrule = new InsEncode(); // Encode class for instruction
3355   NameAndList* params  = encrule->add_encode(ec_name);
3356   inst._parameters.reset();
3357   while ((param = inst._parameters.iter()) != NULL) {
3358     params->add_entry(param);
3359   }
3360 
3361   // Set encode class of this instruction.
3362   inst._constant = encrule;
3363 }
3364 
3365 
3366 //------------------------------constant_parse_expression----------------------
constant_parse_expression(EncClass * encoding,char * ec_name)3367 void ADLParser::constant_parse_expression(EncClass* encoding, char* ec_name) {
3368   skipws();
3369 
3370   // Prepend location descriptor, for debugging; cf. ADLParser::find_cpp_block
3371   if (_AD._adlocation_debug) {
3372     encoding->add_code(get_line_string());
3373   }
3374 
3375   // Start code line.
3376   encoding->add_code("    _constant = C->constant_table().add");
3377 
3378   // Parse everything in ( ) expression.
3379   encoding->add_code("(this, ");
3380   next_char();  // Skip '('
3381   int parens_depth = 1;
3382 
3383   // Collect the parts of the constant expression.
3384   // (1) strings that are passed through to output
3385   // (2) replacement/substitution variable, preceeded by a '$'
3386   while (parens_depth > 0) {
3387     if (_curchar == '(') {
3388       parens_depth++;
3389       encoding->add_code("(");
3390       next_char();
3391     }
3392     else if (_curchar == ')') {
3393       parens_depth--;
3394       if (parens_depth > 0)
3395         encoding->add_code(")");
3396       next_char();
3397     }
3398     else {
3399       // (1)
3400       // Check if there is a string to pass through to output
3401       char *start = _ptr;  // Record start of the next string
3402       while ((_curchar != '$') && (_curchar != '(') && (_curchar != ')')) {
3403         next_char();
3404       }
3405       // If a string was found, terminate it and record in EncClass
3406       if (start != _ptr) {
3407         *_ptr = '\0';  // Terminate the string
3408         encoding->add_code(start);
3409       }
3410 
3411       // (2)
3412       // If we are at a replacement variable, copy it and record in EncClass.
3413       if (_curchar == '$') {
3414         // Found replacement Variable
3415         char* rep_var = get_rep_var_ident_dup();
3416         encoding->add_rep_var(rep_var);
3417       }
3418     }
3419   }
3420 
3421   // Finish code line.
3422   encoding->add_code(");");
3423 
3424   if (_AD._adlocation_debug) {
3425     encoding->add_code(end_line_marker());
3426   }
3427 
3428   // Debug Stuff
3429   if (_AD._adl_debug > 1)  fprintf(stderr, "EncodingClass Form: %s\n", ec_name);
3430 }
3431 
3432 
3433 //------------------------------size_parse-----------------------------------
3434 // Parse a 'size(<expr>)' attribute which specifies the size of the
3435 // emitted instructions in bytes. <expr> can be a C++ expression,
3436 // e.g. a constant.
size_parse(InstructForm * instr)3437 char* ADLParser::size_parse(InstructForm *instr) {
3438   char* sizeOfInstr = NULL;
3439 
3440   // Get value of the instruction's size
3441   skipws();
3442 
3443   // Parse size
3444   sizeOfInstr = get_paren_expr("size expression");
3445   if (sizeOfInstr == NULL) {
3446      parse_err(SYNERR, "size of opcode expected at %c\n", _curchar);
3447      return NULL;
3448   }
3449 
3450   skipws();
3451 
3452   // Check for terminator
3453   if (_curchar != ';') {
3454     parse_err(SYNERR, "missing ';' in ins_attrib definition\n");
3455     return NULL;
3456   }
3457   next_char();                     // Advance past the ';'
3458   skipws();                        // necessary for instr_parse()
3459 
3460   // Debug Stuff
3461   if (_AD._adl_debug > 1) {
3462     if (sizeOfInstr != NULL) {
3463       fprintf(stderr,"size of opcode: %s\n", sizeOfInstr);
3464     }
3465   }
3466 
3467   return sizeOfInstr;
3468 }
3469 
3470 
3471 //------------------------------opcode_parse-----------------------------------
opcode_parse(InstructForm * instr)3472 Opcode * ADLParser::opcode_parse(InstructForm *instr) {
3473   char *primary   = NULL;
3474   char *secondary = NULL;
3475   char *tertiary  = NULL;
3476 
3477   char   *val    = NULL;
3478   Opcode *opcode = NULL;
3479 
3480   // Get value of the instruction's opcode
3481   skipws();
3482   if (_curchar != '(') {         // Check for parenthesized operand list
3483     parse_err(SYNERR, "missing '(' in expand instruction declaration\n");
3484     return NULL;
3485   }
3486   next_char();                   // skip open paren
3487   skipws();
3488   if (_curchar != ')') {
3489     // Parse primary, secondary, and tertiary opcodes, if provided.
3490     if ( (primary = get_ident_or_literal_constant("primary opcode")) == NULL ) {
3491           parse_err(SYNERR, "primary hex opcode expected at %c\n", _curchar);
3492         return NULL;
3493     }
3494     skipws();
3495     if (_curchar == ',') {
3496       next_char();
3497       skipws();
3498       // Parse secondary opcode
3499       if ( (secondary = get_ident_or_literal_constant("secondary opcode")) == NULL ) {
3500         parse_err(SYNERR, "secondary hex opcode expected at %c\n", _curchar);
3501         return NULL;
3502       }
3503       skipws();
3504       if (_curchar == ',') {
3505         next_char();
3506         skipws();
3507         // Parse tertiary opcode
3508         if ( (tertiary = get_ident_or_literal_constant("tertiary opcode")) == NULL ) {
3509           parse_err(SYNERR,"tertiary hex opcode expected at %c\n", _curchar);
3510           return NULL;
3511         }
3512         skipws();
3513       }
3514     }
3515     skipws();
3516     if (_curchar != ')') {
3517       parse_err(SYNERR, "Missing ')' in opcode description\n");
3518       return NULL;
3519     }
3520   }
3521   next_char();                     // Skip ')'
3522   skipws();
3523   // Check for terminator
3524   if (_curchar != ';') {
3525     parse_err(SYNERR, "missing ';' in ins_attrib definition\n");
3526     return NULL;
3527   }
3528   next_char();                     // Advance past the ';'
3529   skipws();                        // necessary for instr_parse()
3530 
3531   // Debug Stuff
3532   if (_AD._adl_debug > 1) {
3533     if (primary   != NULL) fprintf(stderr,"primary   opcode: %s\n", primary);
3534     if (secondary != NULL) fprintf(stderr,"secondary opcode: %s\n", secondary);
3535     if (tertiary  != NULL) fprintf(stderr,"tertiary  opcode: %s\n", tertiary);
3536   }
3537 
3538   // Generate new object and return
3539   opcode = new Opcode(primary, secondary, tertiary);
3540   return opcode;
3541 }
3542 
3543 
3544 //------------------------------interface_parse--------------------------------
interface_parse(void)3545 Interface *ADLParser::interface_parse(void) {
3546   char *iface_name  = NULL;      // Name of interface class being used
3547   char *iface_code  = NULL;      // Describe components of this class
3548 
3549   // Get interface class name
3550   skipws();                       // Skip whitespace
3551   if (_curchar != '(') {
3552     parse_err(SYNERR, "Missing '(' at start of interface description.\n");
3553     return NULL;
3554   }
3555   next_char();                    // move past '('
3556   skipws();
3557   iface_name = get_ident();
3558   if (iface_name == NULL) {
3559     parse_err(SYNERR, "missing interface name after 'interface'.\n");
3560     return NULL;
3561   }
3562   skipws();
3563   if (_curchar != ')') {
3564     parse_err(SYNERR, "Missing ')' after name of interface.\n");
3565     return NULL;
3566   }
3567   next_char();                    // move past ')'
3568 
3569   // Get details of the interface,
3570   // for the type of interface indicated by iface_name.
3571   Interface *inter = NULL;
3572   skipws();
3573   if ( _curchar != ';' ) {
3574     if ( strcmp(iface_name,"MEMORY_INTER") == 0 ) {
3575       inter = mem_interface_parse();
3576     }
3577     else if ( strcmp(iface_name,"COND_INTER") == 0 ) {
3578       inter = cond_interface_parse();
3579     }
3580     // The parse routines consume the "%}"
3581 
3582     // Check for probable extra ';' after defining block.
3583     if ( _curchar == ';' ) {
3584       parse_err(SYNERR, "Extra ';' after defining interface block.\n");
3585       next_char();                // Skip ';'
3586       return NULL;
3587     }
3588   } else {
3589     next_char();                  // move past ';'
3590 
3591     // Create appropriate interface object
3592     if ( strcmp(iface_name,"REG_INTER") == 0 ) {
3593       inter = new RegInterface();
3594     }
3595     else if ( strcmp(iface_name,"CONST_INTER") == 0 ) {
3596       inter = new ConstInterface();
3597     }
3598   }
3599   skipws();                       // be friendly to oper_parse()
3600   // Debug Stuff
3601   if (_AD._adl_debug > 1) fprintf(stderr,"Interface Form: %s\n", iface_name);
3602 
3603   // Create appropriate interface object and return.
3604   return inter;
3605 }
3606 
3607 
3608 //------------------------------mem_interface_parse----------------------------
mem_interface_parse(void)3609 Interface *ADLParser::mem_interface_parse(void) {
3610   // Fields for MemInterface
3611   char *base        = NULL;
3612   char *index       = NULL;
3613   char *scale       = NULL;
3614   char *disp        = NULL;
3615 
3616   if (_curchar != '%') {
3617     parse_err(SYNERR, "Missing '%%{' for 'interface' block.\n");
3618     return NULL;
3619   }
3620   next_char();                  // Skip '%'
3621   if (_curchar != '{') {
3622     parse_err(SYNERR, "Missing '%%{' for 'interface' block.\n");
3623     return NULL;
3624   }
3625   next_char();                  // Skip '{'
3626   skipws();
3627   do {
3628     char *field = get_ident();
3629     if (field == NULL) {
3630       parse_err(SYNERR, "Expected keyword, base|index|scale|disp,  or '%%}' ending interface.\n");
3631       return NULL;
3632     }
3633     if ( strcmp(field,"base") == 0 ) {
3634       base  = interface_field_parse();
3635     }
3636     else if ( strcmp(field,"index") == 0 ) {
3637       index = interface_field_parse();
3638     }
3639     else if ( strcmp(field,"scale") == 0 ) {
3640       scale = interface_field_parse();
3641     }
3642     else if ( strcmp(field,"disp") == 0 ) {
3643       disp  = interface_field_parse();
3644     }
3645     else {
3646       parse_err(SYNERR, "Expected keyword, base|index|scale|disp,  or '%%}' ending interface.\n");
3647       return NULL;
3648     }
3649   } while( _curchar != '%' );
3650   next_char();                  // Skip '%'
3651   if ( _curchar != '}' ) {
3652     parse_err(SYNERR, "Missing '%%}' for 'interface' block.\n");
3653     return NULL;
3654   }
3655   next_char();                  // Skip '}'
3656 
3657   // Construct desired object and return
3658   Interface *inter = new MemInterface(base, index, scale, disp);
3659   return inter;
3660 }
3661 
3662 
3663 //------------------------------cond_interface_parse---------------------------
cond_interface_parse(void)3664 Interface *ADLParser::cond_interface_parse(void) {
3665   char *equal;
3666   char *not_equal;
3667   char *less;
3668   char *greater_equal;
3669   char *less_equal;
3670   char *greater;
3671   char *overflow;
3672   char *no_overflow;
3673   const char *equal_format = "eq";
3674   const char *not_equal_format = "ne";
3675   const char *less_format = "lt";
3676   const char *greater_equal_format = "ge";
3677   const char *less_equal_format = "le";
3678   const char *greater_format = "gt";
3679   const char *overflow_format = "o";
3680   const char *no_overflow_format = "no";
3681 
3682   if (_curchar != '%') {
3683     parse_err(SYNERR, "Missing '%%{' for 'cond_interface' block.\n");
3684     return NULL;
3685   }
3686   next_char();                  // Skip '%'
3687   if (_curchar != '{') {
3688     parse_err(SYNERR, "Missing '%%{' for 'cond_interface' block.\n");
3689     return NULL;
3690   }
3691   next_char();                  // Skip '{'
3692   skipws();
3693   do {
3694     char *field = get_ident();
3695     if (field == NULL) {
3696       parse_err(SYNERR, "Expected keyword, base|index|scale|disp,  or '%%}' ending interface.\n");
3697       return NULL;
3698     }
3699     if ( strcmp(field,"equal") == 0 ) {
3700       equal  = interface_field_parse(&equal_format);
3701     }
3702     else if ( strcmp(field,"not_equal") == 0 ) {
3703       not_equal = interface_field_parse(&not_equal_format);
3704     }
3705     else if ( strcmp(field,"less") == 0 ) {
3706       less = interface_field_parse(&less_format);
3707     }
3708     else if ( strcmp(field,"greater_equal") == 0 ) {
3709       greater_equal  = interface_field_parse(&greater_equal_format);
3710     }
3711     else if ( strcmp(field,"less_equal") == 0 ) {
3712       less_equal = interface_field_parse(&less_equal_format);
3713     }
3714     else if ( strcmp(field,"greater") == 0 ) {
3715       greater = interface_field_parse(&greater_format);
3716     }
3717     else if ( strcmp(field,"overflow") == 0 ) {
3718       overflow = interface_field_parse(&overflow_format);
3719     }
3720     else if ( strcmp(field,"no_overflow") == 0 ) {
3721       no_overflow = interface_field_parse(&no_overflow_format);
3722     }
3723     else {
3724       parse_err(SYNERR, "Expected keyword, base|index|scale|disp,  or '%%}' ending interface.\n");
3725       return NULL;
3726     }
3727   } while( _curchar != '%' );
3728   next_char();                  // Skip '%'
3729   if ( _curchar != '}' ) {
3730     parse_err(SYNERR, "Missing '%%}' for 'interface' block.\n");
3731     return NULL;
3732   }
3733   next_char();                  // Skip '}'
3734 
3735   // Construct desired object and return
3736   Interface *inter = new CondInterface(equal,         equal_format,
3737                                        not_equal,     not_equal_format,
3738                                        less,          less_format,
3739                                        greater_equal, greater_equal_format,
3740                                        less_equal,    less_equal_format,
3741                                        greater,       greater_format,
3742                                        overflow,      overflow_format,
3743                                        no_overflow,   no_overflow_format);
3744   return inter;
3745 }
3746 
3747 
3748 //------------------------------interface_field_parse--------------------------
interface_field_parse(const char ** format)3749 char *ADLParser::interface_field_parse(const char ** format) {
3750   char *iface_field = NULL;
3751 
3752   // Get interface field
3753   skipws();                      // Skip whitespace
3754   if (_curchar != '(') {
3755     parse_err(SYNERR, "Missing '(' at start of interface field.\n");
3756     return NULL;
3757   }
3758   next_char();                   // move past '('
3759   skipws();
3760   if ( _curchar != '0' && _curchar != '$' ) {
3761     parse_err(SYNERR, "missing or invalid interface field contents.\n");
3762     return NULL;
3763   }
3764   iface_field = get_rep_var_ident();
3765   if (iface_field == NULL) {
3766     parse_err(SYNERR, "missing or invalid interface field contents.\n");
3767     return NULL;
3768   }
3769   skipws();
3770   if (format != NULL && _curchar == ',') {
3771     next_char();
3772     skipws();
3773     if (_curchar != '"') {
3774       parse_err(SYNERR, "Missing '\"' in field format .\n");
3775       return NULL;
3776     }
3777     next_char();
3778     char *start = _ptr;       // Record start of the next string
3779     while ((_curchar != '"') && (_curchar != '%') && (_curchar != '\n')) {
3780       if (_curchar == '\\')  next_char();  // superquote
3781       if (_curchar == '\n')  parse_err(SYNERR, "newline in string");  // unimplemented!
3782       next_char();
3783     }
3784     if (_curchar != '"') {
3785       parse_err(SYNERR, "Missing '\"' at end of field format .\n");
3786       return NULL;
3787     }
3788     // If a string was found, terminate it and record in FormatRule
3789     if ( start != _ptr ) {
3790       *_ptr  = '\0';          // Terminate the string
3791       *format = start;
3792     }
3793     next_char();
3794     skipws();
3795   }
3796   if (_curchar != ')') {
3797     parse_err(SYNERR, "Missing ')' after interface field.\n");
3798     return NULL;
3799   }
3800   next_char();                   // move past ')'
3801   skipws();
3802   if ( _curchar != ';' ) {
3803     parse_err(SYNERR, "Missing ';' at end of interface field.\n");
3804     return NULL;
3805   }
3806   next_char();                    // move past ';'
3807   skipws();                       // be friendly to interface_parse()
3808 
3809   return iface_field;
3810 }
3811 
3812 
3813 //------------------------------match_parse------------------------------------
match_parse(FormDict & operands)3814 MatchRule *ADLParser::match_parse(FormDict &operands) {
3815   MatchRule *match;               // Match Rule class for instruction/operand
3816   char      *cnstr = NULL;        // Code for constructor
3817   int        depth = 0;           // Counter for matching parentheses
3818   int        numleaves = 0;       // Counter for number of leaves in rule
3819 
3820   // Parse the match rule tree
3821   MatchNode *mnode = matchNode_parse(operands, depth, numleaves, true);
3822 
3823   // Either there is a block with a constructor, or a ';' here
3824   skipws();                       // Skip whitespace
3825   if ( _curchar == ';' ) {        // Semicolon is valid terminator
3826     cnstr = NULL;                 // no constructor for this form
3827     next_char();                  // Move past the ';', replaced with '\0'
3828   }
3829   else if ((cnstr = find_cpp_block("match constructor")) == NULL ) {
3830     parse_err(SYNERR, "invalid construction of match rule\n"
3831               "Missing ';' or invalid '%%{' and '%%}' constructor\n");
3832     return NULL;                  // No MatchRule to return
3833   }
3834   if (_AD._adl_debug > 1)
3835     if (cnstr) fprintf(stderr,"Match Constructor: %s\n", cnstr);
3836   // Build new MatchRule object
3837   match = new MatchRule(_AD, mnode, depth, cnstr, numleaves);
3838   skipws();                       // Skip any trailing whitespace
3839   return match;                   // Return MatchRule object
3840 }
3841 
3842 //------------------------------format_parse-----------------------------------
format_parse(void)3843 FormatRule* ADLParser::format_parse(void) {
3844   char       *desc   = NULL;
3845   FormatRule *format = (new FormatRule(desc));
3846 
3847   // Without expression form, MUST have a code block;
3848   skipws();                       // Skip whitespace
3849   if ( _curchar == ';' ) {        // Semicolon is valid terminator
3850     desc  = NULL;                 // no constructor for this form
3851     next_char();                  // Move past the ';', replaced with '\0'
3852   }
3853   else if ( _curchar == '%' && *(_ptr+1) == '{') {
3854     next_char();                  // Move past the '%'
3855     next_char();                  // Move past the '{'
3856 
3857     skipws();
3858     if (_curchar == '$') {
3859       char* ident = get_rep_var_ident();
3860       if (strcmp(ident, "$$template") == 0) return template_parse();
3861       parse_err(SYNERR, "Unknown \"%s\" directive in format", ident);
3862       return NULL;
3863     }
3864     // Check for the opening '"' inside the format description
3865     if ( _curchar == '"' ) {
3866       next_char();              // Move past the initial '"'
3867       if( _curchar == '"' ) {   // Handle empty format string case
3868         *_ptr = '\0';           // Terminate empty string
3869         format->_strings.addName(_ptr);
3870       }
3871 
3872       // Collect the parts of the format description
3873       // (1) strings that are passed through to tty->print
3874       // (2) replacement/substitution variable, preceeded by a '$'
3875       // (3) multi-token ANSIY C style strings
3876       while ( true ) {
3877         if ( _curchar == '%' || _curchar == '\n' ) {
3878           if ( _curchar != '"' ) {
3879             parse_err(SYNERR, "missing '\"' at end of format block");
3880             return NULL;
3881           }
3882         }
3883 
3884         // (1)
3885         // Check if there is a string to pass through to output
3886         char *start = _ptr;       // Record start of the next string
3887         while ((_curchar != '$') && (_curchar != '"') && (_curchar != '%') && (_curchar != '\n')) {
3888           if (_curchar == '\\') {
3889             next_char();  // superquote
3890             if ((_curchar == '$') || (_curchar == '%'))
3891               // hack to avoid % escapes and warnings about undefined \ escapes
3892               *(_ptr-1) = _curchar;
3893           }
3894           if (_curchar == '\n')  parse_err(SYNERR, "newline in string");  // unimplemented!
3895           next_char();
3896         }
3897         // If a string was found, terminate it and record in FormatRule
3898         if ( start != _ptr ) {
3899           *_ptr  = '\0';          // Terminate the string
3900           format->_strings.addName(start);
3901         }
3902 
3903         // (2)
3904         // If we are at a replacement variable,
3905         // copy it and record in FormatRule
3906         if ( _curchar == '$' ) {
3907           next_char();          // Move past the '$'
3908           char* rep_var = get_ident(); // Nil terminate the variable name
3909           rep_var = strdup(rep_var);// Copy the string
3910           *_ptr   = _curchar;     // and replace Nil with original character
3911           format->_rep_vars.addName(rep_var);
3912           // Add flag to _strings list indicating we should check _rep_vars
3913           format->_strings.addName(NameList::_signal);
3914         }
3915 
3916         // (3)
3917         // Allow very long strings to be broken up,
3918         // using the ANSI C syntax "foo\n" <newline> "bar"
3919         if ( _curchar == '"') {
3920           next_char();           // Move past the '"'
3921           skipws();              // Skip white space before next string token
3922           if ( _curchar != '"') {
3923             break;
3924           } else {
3925             // Found one.  Skip both " and the whitespace in between.
3926             next_char();
3927           }
3928         }
3929       } // end while part of format description
3930 
3931       // Check for closing '"' and '%}' in format description
3932       skipws();                   // Move to closing '%}'
3933       if ( _curchar != '%' ) {
3934         parse_err(SYNERR, "non-blank characters between closing '\"' and '%%' in format");
3935         return NULL;
3936       }
3937     } // Done with format description inside
3938 
3939     skipws();
3940     // Past format description, at '%'
3941     if ( _curchar != '%' || *(_ptr+1) != '}' ) {
3942       parse_err(SYNERR, "missing '%%}' at end of format block");
3943       return NULL;
3944     }
3945     next_char();                  // Move past the '%'
3946     next_char();                  // Move past the '}'
3947   }
3948   else {  // parameter list alone must terminate with a ';'
3949     parse_err(SYNERR, "missing ';' after Format expression");
3950     return NULL;
3951   }
3952   // Debug Stuff
3953   if (_AD._adl_debug > 1) fprintf(stderr,"Format Rule: %s\n", desc);
3954 
3955   skipws();
3956   return format;
3957 }
3958 
3959 
3960 //------------------------------template_parse-----------------------------------
template_parse(void)3961 FormatRule* ADLParser::template_parse(void) {
3962   char       *desc   = NULL;
3963   FormatRule *format = (new FormatRule(desc));
3964 
3965   skipws();
3966   while ( (_curchar != '%') && (*(_ptr+1) != '}') ) {
3967 
3968     // (1)
3969     // Check if there is a string to pass through to output
3970     {
3971       char *start = _ptr;       // Record start of the next string
3972       while ((_curchar != '$') && ((_curchar != '%') || (*(_ptr+1) != '}')) ) {
3973         // If at the start of a comment, skip past it
3974         if( (_curchar == '/') && ((*(_ptr+1) == '/') || (*(_ptr+1) == '*')) ) {
3975           skipws_no_preproc();
3976         } else {
3977           // ELSE advance to the next character, or start of the next line
3978           next_char_or_line();
3979         }
3980       }
3981       // If a string was found, terminate it and record in EncClass
3982       if ( start != _ptr ) {
3983         *_ptr  = '\0';          // Terminate the string
3984         // Add flag to _strings list indicating we should check _rep_vars
3985         format->_strings.addName(NameList::_signal2);
3986         format->_strings.addName(start);
3987       }
3988     }
3989 
3990     // (2)
3991     // If we are at a replacement variable,
3992     // copy it and record in EncClass
3993     if ( _curchar == '$' ) {
3994       // Found replacement Variable
3995       char *rep_var = get_rep_var_ident_dup();
3996       if (strcmp(rep_var, "$emit") == 0) {
3997         // switch to normal format parsing
3998         next_char();
3999         next_char();
4000         skipws();
4001         // Check for the opening '"' inside the format description
4002         if ( _curchar == '"' ) {
4003           next_char();              // Move past the initial '"'
4004           if( _curchar == '"' ) {   // Handle empty format string case
4005             *_ptr = '\0';           // Terminate empty string
4006             format->_strings.addName(_ptr);
4007           }
4008 
4009           // Collect the parts of the format description
4010           // (1) strings that are passed through to tty->print
4011           // (2) replacement/substitution variable, preceeded by a '$'
4012           // (3) multi-token ANSIY C style strings
4013           while ( true ) {
4014             if ( _curchar == '%' || _curchar == '\n' ) {
4015               parse_err(SYNERR, "missing '\"' at end of format block");
4016               return NULL;
4017             }
4018 
4019             // (1)
4020             // Check if there is a string to pass through to output
4021             char *start = _ptr;       // Record start of the next string
4022             while ((_curchar != '$') && (_curchar != '"') && (_curchar != '%') && (_curchar != '\n')) {
4023               if (_curchar == '\\')  next_char();  // superquote
4024               if (_curchar == '\n')  parse_err(SYNERR, "newline in string");  // unimplemented!
4025               next_char();
4026             }
4027             // If a string was found, terminate it and record in FormatRule
4028             if ( start != _ptr ) {
4029               *_ptr  = '\0';          // Terminate the string
4030               format->_strings.addName(start);
4031             }
4032 
4033             // (2)
4034             // If we are at a replacement variable,
4035             // copy it and record in FormatRule
4036             if ( _curchar == '$' ) {
4037               next_char();          // Move past the '$'
4038               char* next_rep_var = get_ident(); // Nil terminate the variable name
4039               next_rep_var = strdup(next_rep_var);// Copy the string
4040               *_ptr   = _curchar;     // and replace Nil with original character
4041               format->_rep_vars.addName(next_rep_var);
4042               // Add flag to _strings list indicating we should check _rep_vars
4043               format->_strings.addName(NameList::_signal);
4044             }
4045 
4046             // (3)
4047             // Allow very long strings to be broken up,
4048             // using the ANSI C syntax "foo\n" <newline> "bar"
4049             if ( _curchar == '"') {
4050               next_char();           // Move past the '"'
4051               skipws();              // Skip white space before next string token
4052               if ( _curchar != '"') {
4053                 break;
4054               } else {
4055                 // Found one.  Skip both " and the whitespace in between.
4056                 next_char();
4057               }
4058             }
4059           } // end while part of format description
4060         }
4061       } else {
4062         // Add flag to _strings list indicating we should check _rep_vars
4063         format->_rep_vars.addName(rep_var);
4064         // Add flag to _strings list indicating we should check _rep_vars
4065         format->_strings.addName(NameList::_signal3);
4066       }
4067     } // end while part of format description
4068   }
4069 
4070   skipws();
4071   // Past format description, at '%'
4072   if ( _curchar != '%' || *(_ptr+1) != '}' ) {
4073     parse_err(SYNERR, "missing '%%}' at end of format block");
4074     return NULL;
4075   }
4076   next_char();                  // Move past the '%'
4077   next_char();                  // Move past the '}'
4078 
4079   // Debug Stuff
4080   if (_AD._adl_debug > 1) fprintf(stderr,"Format Rule: %s\n", desc);
4081 
4082   skipws();
4083   return format;
4084 }
4085 
4086 
4087 //------------------------------effect_parse-----------------------------------
effect_parse(InstructForm * instr)4088 void ADLParser::effect_parse(InstructForm *instr) {
4089   char* desc   = NULL;
4090 
4091   skipws();                      // Skip whitespace
4092   if (_curchar != '(') {
4093     parse_err(SYNERR, "missing '(' in effect definition\n");
4094     return;
4095   }
4096   // Get list of effect-operand pairs and insert into dictionary
4097   else get_effectlist(instr->_effects, instr->_localNames, instr->_has_call);
4098 
4099   // Debug Stuff
4100   if (_AD._adl_debug > 1) fprintf(stderr,"Effect description: %s\n", desc);
4101   if (_curchar != ';') {
4102     parse_err(SYNERR, "missing ';' in Effect definition\n");
4103   }
4104   next_char();                  // Skip ';'
4105 
4106 }
4107 
4108 //------------------------------expand_parse-----------------------------------
expand_parse(InstructForm * instr)4109 ExpandRule* ADLParser::expand_parse(InstructForm *instr) {
4110   char         *ident, *ident2;
4111   NameAndList  *instr_and_operands = NULL;
4112   ExpandRule   *exp = new ExpandRule();
4113 
4114   // Expand is a block containing an ordered list of operands with initializers,
4115   // or instructions, each of which has an ordered list of operands.
4116   // Check for block delimiter
4117   skipws();                        // Skip leading whitespace
4118   if ((_curchar != '%')
4119       || (next_char(), (_curchar != '{')) ) { // If not open block
4120     parse_err(SYNERR, "missing '%%{' in expand definition\n");
4121     return(NULL);
4122   }
4123   next_char();                     // Maintain the invariant
4124   do {
4125     ident = get_ident();           // Grab next identifier
4126     if (ident == NULL) {
4127       parse_err(SYNERR, "identifier expected at %c\n", _curchar);
4128       continue;
4129     }
4130 
4131     // Check whether we should parse an instruction or operand.
4132     const Form *form = _globalNames[ident];
4133     bool parse_oper = false;
4134     bool parse_ins  = false;
4135     if (form == NULL) {
4136       skipws();
4137       // Check whether this looks like an instruction specification.  If so,
4138       // just parse the instruction.  The declaration of the instruction is
4139       // not needed here.
4140       if (_curchar == '(') parse_ins = true;
4141     } else if (form->is_instruction()) {
4142       parse_ins = true;
4143     } else if (form->is_operand()) {
4144       parse_oper = true;
4145     } else {
4146       parse_err(SYNERR, "instruction/operand name expected at %s\n", ident);
4147       continue;
4148     }
4149 
4150     if (parse_oper) {
4151       // This is a new operand
4152       OperandForm *oper = form->is_operand();
4153       if (oper == NULL) {
4154         parse_err(SYNERR, "instruction/operand name expected at %s\n", ident);
4155         continue;
4156       }
4157       // Throw the operand on the _newopers list
4158       skipws();
4159       ident = get_unique_ident(instr->_localNames,"Operand");
4160       if (ident == NULL) {
4161         parse_err(SYNERR, "identifier expected at %c\n", _curchar);
4162         continue;
4163       }
4164       exp->_newopers.addName(ident);
4165       // Add new operand to LocalNames
4166       instr->_localNames.Insert(ident, oper);
4167       // Grab any constructor code and save as a string
4168       char *c = NULL;
4169       skipws();
4170       if (_curchar == '%') { // Need a constructor for the operand
4171         c = find_cpp_block("Operand Constructor");
4172         if (c == NULL) {
4173           parse_err(SYNERR, "Invalid code block for operand constructor\n", _curchar);
4174           continue;
4175         }
4176         // Add constructor to _newopconst Dict
4177         exp->_newopconst.Insert(ident, c);
4178       }
4179       else if (_curchar != ';') { // If no constructor, need a ;
4180         parse_err(SYNERR, "Missing ; in expand rule operand declaration\n");
4181         continue;
4182       }
4183       else next_char(); // Skip the ;
4184       skipws();
4185     }
4186     else {
4187       assert(parse_ins, "sanity");
4188       // Add instruction to list
4189       instr_and_operands = new NameAndList(ident);
4190       // Grab operands, build nameList of them, and then put into dictionary
4191       skipws();
4192       if (_curchar != '(') {         // Check for parenthesized operand list
4193         parse_err(SYNERR, "missing '(' in expand instruction declaration\n");
4194         continue;
4195       }
4196       do {
4197         next_char();                 // skip open paren & comma characters
4198         skipws();
4199         if (_curchar == ')') break;
4200         ident2 = get_ident();
4201         skipws();
4202         if (ident2 == NULL) {
4203           parse_err(SYNERR, "identifier expected at %c\n", _curchar);
4204           continue;
4205         }                            // Check that you have a valid operand
4206         const Form *form2 = instr->_localNames[ident2];
4207         if (!form2) {
4208           parse_err(SYNERR, "operand name expected at %s\n", ident2);
4209           continue;
4210         }
4211         OperandForm *oper = form2->is_operand();
4212         if (oper == NULL && !form2->is_opclass()) {
4213           parse_err(SYNERR, "operand name expected at %s\n", ident2);
4214           continue;
4215         }                            // Add operand to list
4216         instr_and_operands->add_entry(ident2);
4217       } while(_curchar == ',');
4218       if (_curchar != ')') {
4219         parse_err(SYNERR, "missing ')'in expand instruction declaration\n");
4220         continue;
4221       }
4222       next_char();
4223       if (_curchar != ';') {
4224         parse_err(SYNERR, "missing ';'in expand instruction declaration\n");
4225         continue;
4226       }
4227       next_char();
4228 
4229       // Record both instruction name and its operand list
4230       exp->add_instruction(instr_and_operands);
4231 
4232       skipws();
4233     }
4234 
4235   } while(_curchar != '%');
4236   next_char();
4237   if (_curchar != '}') {
4238     parse_err(SYNERR, "missing '%%}' in expand rule definition\n");
4239     return(NULL);
4240   }
4241   next_char();
4242 
4243   // Debug Stuff
4244   if (_AD._adl_debug > 1) fprintf(stderr,"Expand Rule:\n");
4245 
4246   skipws();
4247   return (exp);
4248 }
4249 
4250 //------------------------------rewrite_parse----------------------------------
rewrite_parse(void)4251 RewriteRule* ADLParser::rewrite_parse(void) {
4252   char* params = NULL;
4253   char* desc   = NULL;
4254 
4255 
4256   // This feature targeted for second generation description language.
4257 
4258   skipws();                      // Skip whitespace
4259   // Get parameters for rewrite
4260   if ((params = get_paren_expr("rewrite parameters")) == NULL) {
4261     parse_err(SYNERR, "missing '(' in rewrite rule\n");
4262     return NULL;
4263   }
4264   // Debug Stuff
4265   if (_AD._adl_debug > 1) fprintf(stderr,"Rewrite parameters: %s\n", params);
4266 
4267   // For now, grab entire block;
4268   skipws();
4269   if ( (desc = find_cpp_block("rewrite block")) == NULL ) {
4270     parse_err(SYNERR, "incorrect or missing block for 'rewrite'.\n");
4271     return NULL;
4272   }
4273   // Debug Stuff
4274   if (_AD._adl_debug > 1) fprintf(stderr,"Rewrite Rule: %s\n", desc);
4275 
4276   skipws();
4277   return (new RewriteRule(params,desc));
4278 }
4279 
4280 //------------------------------attr_parse-------------------------------------
attr_parse(char * ident)4281 Attribute *ADLParser::attr_parse(char* ident) {
4282   Attribute *attrib;              // Attribute class
4283   char      *cost = NULL;         // String representation of cost attribute
4284 
4285   skipws();                       // Skip leading whitespace
4286   if ( (cost = get_paren_expr("attribute")) == NULL ) {
4287     parse_err(SYNERR, "incorrect or missing expression for 'attribute'\n");
4288     return NULL;
4289   }
4290   // Debug Stuff
4291   if (_AD._adl_debug > 1) fprintf(stderr,"Attribute: %s\n", cost);
4292   if (_curchar != ';') {
4293     parse_err(SYNERR, "missing ';' in attribute definition\n");
4294     return NULL;
4295   }
4296   next_char();                   // Point after the terminator
4297 
4298   skipws();
4299   attrib = new Attribute(ident,cost,INS_ATTR); // Build new predicate object
4300   return attrib;
4301 }
4302 
4303 
4304 //------------------------------matchNode_parse--------------------------------
matchNode_parse(FormDict & operands,int & depth,int & numleaves,bool atroot)4305 MatchNode *ADLParser::matchNode_parse(FormDict &operands, int &depth, int &numleaves, bool atroot) {
4306   // Count depth of parenthesis nesting for both left and right children
4307   int   lParens = depth;
4308   int   rParens = depth;
4309 
4310   // MatchNode objects for left, right, and root of subtree.
4311   MatchNode *lChild = NULL;
4312   MatchNode *rChild = NULL;
4313   char      *token;               // Identifier which may be opcode or operand
4314 
4315   // Match expression starts with a '('
4316   if (cur_char() != '(')
4317     return NULL;
4318 
4319   next_char();                    // advance past '('
4320 
4321   // Parse the opcode
4322   token = get_ident();            // Get identifier, opcode
4323   if (token == NULL) {
4324     parse_err(SYNERR, "missing opcode in match expression\n");
4325     return NULL;
4326   }
4327 
4328   // Take note if we see one of a few special operations - those that are
4329   // treated differently on different architectures in the sense that on
4330   // one architecture there is a match rule and on another there isn't (so
4331   // a call will eventually be generated).
4332 
4333   for (int i = _last_machine_leaf + 1; i < _last_opcode; i++) {
4334     if (strcmp(token, NodeClassNames[i]) == 0) {
4335       _AD.has_match_rule(i, true);
4336     }
4337   }
4338 
4339   // Lookup the root value in the operands dict to perform substitution
4340   const char  *result    = NULL;  // Result type will be filled in later
4341   const char  *name      = token; // local name associated with this node
4342   const char  *operation = token; // remember valid operation for later
4343   const Form  *form      = operands[token];
4344   OpClassForm *opcForm = form ? form->is_opclass() : NULL;
4345   if (opcForm != NULL) {
4346     // If this token is an entry in the local names table, record its type
4347     if (!opcForm->ideal_only()) {
4348       operation = opcForm->_ident;
4349       result = operation;         // Operands result in their own type
4350     }
4351     // Otherwise it is an ideal type, and so, has no local name
4352     else                        name = NULL;
4353   }
4354 
4355   // Parse the operands
4356   skipws();
4357   if (cur_char() != ')') {
4358 
4359     // Parse the left child
4360     if (strcmp(operation,"Set"))
4361       lChild = matchChild_parse(operands, lParens, numleaves, false);
4362     else
4363       lChild = matchChild_parse(operands, lParens, numleaves, true);
4364 
4365     skipws();
4366     if (cur_char() != ')' ) {
4367       if(strcmp(operation, "Set"))
4368         rChild = matchChild_parse(operands,rParens,numleaves,false);
4369       else
4370         rChild = matchChild_parse(operands,rParens,numleaves,true);
4371     }
4372   }
4373 
4374   // Check for required ')'
4375   skipws();
4376   if (cur_char() != ')') {
4377     parse_err(SYNERR, "missing ')' in match expression\n");
4378     return NULL;
4379   }
4380   next_char();                    // skip the ')'
4381 
4382   MatchNode* mroot = new MatchNode(_AD,result,name,operation,lChild,rChild);
4383 
4384   // If not the root, reduce this subtree to an internal operand
4385   if (!atroot) {
4386     mroot->build_internalop();
4387   }
4388   // depth is greater of left and right paths.
4389   depth = (lParens > rParens) ? lParens : rParens;
4390 
4391   return mroot;
4392 }
4393 
4394 
4395 //------------------------------matchChild_parse-------------------------------
matchChild_parse(FormDict & operands,int & parens,int & numleaves,bool atroot)4396 MatchNode *ADLParser::matchChild_parse(FormDict &operands, int &parens, int &numleaves, bool atroot) {
4397   MatchNode  *child  = NULL;
4398   const char *result = NULL;
4399   const char *token  = NULL;
4400   const char *opType = NULL;
4401 
4402   if (cur_char() == '(') {         // child is an operation
4403     ++parens;
4404     child = matchNode_parse(operands, parens, numleaves, atroot);
4405   }
4406   else {                           // child is an operand
4407     token = get_ident();
4408     const Form  *form    = operands[token];
4409     OpClassForm *opcForm = form ? form->is_opclass() : NULL;
4410     if (opcForm != NULL) {
4411       opType = opcForm->_ident;
4412       result = opcForm->_ident;    // an operand's result matches its type
4413     } else {
4414       parse_err(SYNERR, "undefined operand %s in match rule\n", token);
4415       return NULL;
4416     }
4417 
4418     if (opType == NULL) {
4419       parse_err(SYNERR, "missing type for argument '%s'\n", token);
4420     }
4421 
4422     child = new MatchNode(_AD, result, token, opType);
4423     ++numleaves;
4424   }
4425 
4426   return child;
4427 }
4428 
4429 
4430 
4431 // ******************** Private Utility Functions *************************
4432 
4433 
find_cpp_block(const char * description)4434 char* ADLParser::find_cpp_block(const char* description) {
4435   char *next;                     // Pointer for finding block delimiters
4436   char* cppBlock = NULL;          // Beginning of C++ code block
4437 
4438   if (_curchar == '%') {          // Encoding is a C++ expression
4439     next_char();
4440     if (_curchar != '{') {
4441       parse_err(SYNERR, "missing '{' in %s \n", description);
4442       return NULL;
4443     }
4444     next_char();                  // Skip block delimiter
4445     skipws_no_preproc();          // Skip leading whitespace
4446     cppBlock = _ptr;              // Point to start of expression
4447     int line = linenum();
4448     next = _ptr + 1;
4449     while(((_curchar != '%') || (*next != '}')) && (_curchar != '\0')) {
4450       next_char_or_line();
4451       next = _ptr+1;              // Maintain the next pointer
4452     }                             // Grab string
4453     if (_curchar == '\0') {
4454       parse_err(SYNERR, "invalid termination of %s \n", description);
4455       return NULL;
4456     }
4457     *_ptr = '\0';                 // Terminate string
4458     _ptr += 2;                    // Skip block delimiter
4459     _curchar = *_ptr;             // Maintain invariant
4460 
4461     // Prepend location descriptor, for debugging.
4462     if (_AD._adlocation_debug) {
4463       char* location = get_line_string(line);
4464       char* end_loc  = end_line_marker();
4465       char* result = (char *)AllocateHeap(strlen(location) + strlen(cppBlock) + strlen(end_loc) + 1);
4466       strcpy(result, location);
4467       strcat(result, cppBlock);
4468       strcat(result, end_loc);
4469       cppBlock = result;
4470       free(location);
4471     }
4472   }
4473 
4474   return cppBlock;
4475 }
4476 
4477 // Move to the closing token of the expression we are currently at,
4478 // as defined by stop_chars.  Match parens and quotes.
get_expr(const char * desc,const char * stop_chars)4479 char* ADLParser::get_expr(const char *desc, const char *stop_chars) {
4480   char* expr = NULL;
4481   int   paren = 0;
4482 
4483   expr = _ptr;
4484   while (paren > 0 || !strchr(stop_chars, _curchar)) {
4485     if (_curchar == '(') {        // Down level of nesting
4486       paren++;                    // Bump the parenthesis counter
4487       next_char();                // maintain the invariant
4488     }
4489     else if (_curchar == ')') {   // Up one level of nesting
4490       if (paren == 0) {
4491         // Paren underflow:  We didn't encounter the required stop-char.
4492         parse_err(SYNERR, "too many )'s, did not find %s after %s\n",
4493                   stop_chars, desc);
4494         return NULL;
4495       }
4496       paren--;                    // Drop the parenthesis counter
4497       next_char();                // Maintain the invariant
4498     }
4499     else if (_curchar == '"' || _curchar == '\'') {
4500       int qchar = _curchar;
4501       while (true) {
4502         next_char();
4503         if (_curchar == qchar) { next_char(); break; }
4504         if (_curchar == '\\')  next_char();  // superquote
4505         if (_curchar == '\n' || _curchar == '\0') {
4506           parse_err(SYNERR, "newline in string in %s\n", desc);
4507           return NULL;
4508         }
4509       }
4510     }
4511     else if (_curchar == '%' && (_ptr[1] == '{' || _ptr[1] == '}')) {
4512       // Make sure we do not stray into the next ADLC-level form.
4513       parse_err(SYNERR, "unexpected %%%c in %s\n", _ptr[1], desc);
4514       return NULL;
4515     }
4516     else if (_curchar == '\0') {
4517       parse_err(SYNERR, "unexpected EOF in %s\n", desc);
4518       return NULL;
4519     }
4520     else {
4521       // Always walk over whitespace, comments, preprocessor directives, etc.
4522       char* pre_skip_ptr = _ptr;
4523       skipws();
4524       // If the parser declined to make progress on whitespace,
4525       // skip the next character, which is therefore NOT whitespace.
4526       if (pre_skip_ptr == _ptr) {
4527         next_char();
4528       } else if (pre_skip_ptr+strlen(pre_skip_ptr) != _ptr+strlen(_ptr)) {
4529         parse_err(SYNERR, "unimplemented: preprocessor must not elide subexpression in %s", desc);
4530       }
4531     }
4532   }
4533 
4534   assert(strchr(stop_chars, _curchar), "non-null return must be at stop-char");
4535   *_ptr = '\0';               // Replace ')' or other stop-char with '\0'
4536   return expr;
4537 }
4538 
4539 // Helper function around get_expr
4540 // Sets _curchar to '(' so that get_paren_expr will search for a matching ')'
get_paren_expr(const char * description,bool include_location)4541 char *ADLParser::get_paren_expr(const char *description, bool include_location) {
4542   int line = linenum();
4543   if (_curchar != '(')            // Escape if not valid starting position
4544     return NULL;
4545   next_char();                    // Skip the required initial paren.
4546   char *token2 = get_expr(description, ")");
4547   if (_curchar == ')')
4548     next_char();                  // Skip required final paren.
4549   int junk = 0;
4550   if (include_location && _AD._adlocation_debug && !is_int_token(token2, junk)) {
4551     // Prepend location descriptor, for debugging.
4552     char* location = get_line_string(line);
4553     char* end_loc  = end_line_marker();
4554     char* result = (char *)AllocateHeap(strlen(location) + strlen(token2) + strlen(end_loc) + 1);
4555     strcpy(result, location);
4556     strcat(result, token2);
4557     strcat(result, end_loc);
4558     token2 = result;
4559     free(location);
4560   }
4561   return token2;
4562 }
4563 
4564 //------------------------------get_ident_common-------------------------------
4565 // Looks for an identifier in the buffer, and turns it into a null terminated
4566 // string(still inside the file buffer).  Returns a pointer to the string or
4567 // NULL if some other token is found instead.
get_ident_common(bool do_preproc)4568 char *ADLParser::get_ident_common(bool do_preproc) {
4569   char c;
4570   char *start;                    // Pointer to start of token
4571   char *end;                      // Pointer to end of token
4572 
4573   if( _curline == NULL )          // Return NULL at EOF.
4574     return NULL;
4575 
4576   skipws_common(do_preproc);      // Skip whitespace before identifier
4577   start = end = _ptr;             // Start points at first character
4578   end--;                          // unwind end by one to prepare for loop
4579   do {
4580     end++;                        // Increment end pointer
4581     c = *end;                     // Grab character to test
4582   } while ( ((c >= 'a') && (c <= 'z')) || ((c >= 'A') && (c <= 'Z'))
4583             || ((c >= '0') && (c <= '9'))
4584             || ((c == '_')) || ((c == ':')) || ((c == '#')) );
4585   if (start == end) {             // We popped out on the first try
4586     // It can occur that `start' contains the rest of the input file.
4587     // In this case the output should be truncated.
4588     if (strlen(start) > 24) {
4589       char buf[32];
4590       strncpy(buf, start, 20);
4591       buf[20] = '\0';
4592       strcat(buf, "[...]");
4593       parse_err(SYNERR, "Identifier expected, but found '%s'.", buf);
4594     } else {
4595       parse_err(SYNERR, "Identifier expected, but found '%s'.", start);
4596     }
4597     start = NULL;
4598   }
4599   else {
4600     _curchar = c;                 // Save the first character of next token
4601     *end = '\0';                  // NULL terminate the string in place
4602   }
4603   _ptr = end;                     // Reset _ptr to point to next char after token
4604 
4605   // Make sure we do not try to use #defined identifiers.  If start is
4606   // NULL an error was already reported.
4607   if (do_preproc && start != NULL) {
4608     const char* def = _AD.get_preproc_def(start);
4609     if (def != NULL && strcmp(def, start)) {
4610       const char* def1 = def;
4611       const char* def2 = _AD.get_preproc_def(def1);
4612       // implement up to 2 levels of #define
4613       if (def2 != NULL && strcmp(def2, def1)) {
4614         def = def2;
4615         const char* def3 = _AD.get_preproc_def(def2);
4616         if (def3 != NULL && strcmp(def3, def2) && strcmp(def3, def1)) {
4617           parse_err(SYNERR, "unimplemented: using %s defined as %s => %s => %s",
4618                     start, def1, def2, def3);
4619         }
4620       }
4621       start = strdup(def);
4622     }
4623   }
4624 
4625   return start;                   // Pointer to token in filebuf
4626 }
4627 
4628 //------------------------------get_ident_dup----------------------------------
4629 // Looks for an identifier in the buffer, and returns a duplicate
4630 // or NULL if some other token is found instead.
get_ident_dup(void)4631 char *ADLParser::get_ident_dup(void) {
4632   char *ident = get_ident();
4633 
4634   // Duplicate an identifier before returning and restore string.
4635   if( ident != NULL ) {
4636     ident = strdup(ident);  // Copy the string
4637     *_ptr   = _curchar;         // and replace Nil with original character
4638   }
4639 
4640   return ident;
4641 }
4642 
4643 //----------------------get_ident_or_literal_constant--------------------------
4644 // Looks for an identifier in the buffer, or a parenthesized expression.
get_ident_or_literal_constant(const char * description)4645 char *ADLParser::get_ident_or_literal_constant(const char* description) {
4646   char* param = NULL;
4647   skipws();
4648   if (_curchar == '(') {
4649     // Grab a constant expression.
4650     param = get_paren_expr(description);
4651     if (param[0] != '(') {
4652       char* buf = (char*) AllocateHeap(strlen(param) + 3);
4653       sprintf(buf, "(%s)", param);
4654       param = buf;
4655     }
4656     assert(is_literal_constant(param),
4657            "expr must be recognizable as a constant");
4658   } else {
4659     param = get_ident();
4660   }
4661   return param;
4662 }
4663 
4664 //------------------------------get_rep_var_ident-----------------------------
4665 // Do NOT duplicate,
4666 // Leave nil terminator in buffer
4667 // Preserve initial '$'(s) in string
get_rep_var_ident(void)4668 char *ADLParser::get_rep_var_ident(void) {
4669   // Remember starting point
4670   char *rep_var = _ptr;
4671 
4672   // Check for replacement variable indicator '$' and pass if present
4673   if ( _curchar == '$' ) {
4674     next_char();
4675   }
4676   // Check for a subfield indicator, a second '$', and pass if present
4677   if ( _curchar == '$' ) {
4678     next_char();
4679   }
4680 
4681   // Check for a control indicator, a third '$':
4682   if ( _curchar == '$' ) {
4683     next_char();
4684   }
4685 
4686   // Check for more than three '$'s in sequence, SYNERR
4687   if( _curchar == '$' ) {
4688     parse_err(SYNERR, "Replacement variables and field specifiers can not start with '$$$$'");
4689     next_char();
4690     return NULL;
4691   }
4692 
4693   // Nil terminate the variable name following the '$'
4694   char *rep_var_name = get_ident();
4695   assert( rep_var_name != NULL,
4696           "Missing identifier after replacement variable indicator '$'");
4697 
4698   return rep_var;
4699 }
4700 
4701 
4702 
4703 //------------------------------get_rep_var_ident_dup-------------------------
4704 // Return the next replacement variable identifier, skipping first '$'
4705 // given a pointer into a line of the buffer.
4706 // Null terminates string, still inside the file buffer,
4707 // Returns a pointer to a copy of the string, or NULL on failure
get_rep_var_ident_dup(void)4708 char *ADLParser::get_rep_var_ident_dup(void) {
4709   if( _curchar != '$' ) return NULL;
4710 
4711   next_char();                // Move past the '$'
4712   char *rep_var = _ptr;       // Remember starting point
4713 
4714   // Check for a subfield indicator, a second '$':
4715   if ( _curchar == '$' ) {
4716     next_char();
4717   }
4718 
4719   // Check for a control indicator, a third '$':
4720   if ( _curchar == '$' ) {
4721     next_char();
4722   }
4723 
4724   // Check for more than three '$'s in sequence, SYNERR
4725   if( _curchar == '$' ) {
4726     parse_err(SYNERR, "Replacement variables and field specifiers can not start with '$$$$'");
4727     next_char();
4728     return NULL;
4729   }
4730 
4731   // Nil terminate the variable name following the '$'
4732   char *rep_var_name = get_ident();
4733   assert( rep_var_name != NULL,
4734           "Missing identifier after replacement variable indicator '$'");
4735   rep_var = strdup(rep_var);  // Copy the string
4736   *_ptr   = _curchar;         // and replace Nil with original character
4737 
4738   return rep_var;
4739 }
4740 
4741 
4742 //------------------------------get_unique_ident------------------------------
4743 // Looks for an identifier in the buffer, terminates it with a NULL,
4744 // and checks that it is unique
get_unique_ident(FormDict & dict,const char * nameDescription)4745 char *ADLParser::get_unique_ident(FormDict& dict, const char* nameDescription){
4746   char* ident = get_ident();
4747 
4748   if (ident == NULL) {
4749     parse_err(SYNERR, "missing %s identifier at %c\n", nameDescription, _curchar);
4750   }
4751   else {
4752     if (dict[ident] != NULL) {
4753       parse_err(SYNERR, "duplicate name %s for %s\n", ident, nameDescription);
4754       ident = NULL;
4755     }
4756   }
4757 
4758   return ident;
4759 }
4760 
4761 
4762 //------------------------------get_int----------------------------------------
4763 // Looks for a character string integer in the buffer, and turns it into an int
4764 // invokes a parse_err if the next token is not an integer.
4765 // This routine does not leave the integer null-terminated.
get_int(void)4766 int ADLParser::get_int(void) {
4767   char          c;
4768   char         *start;            // Pointer to start of token
4769   char         *end;              // Pointer to end of token
4770   int           result;           // Storage for integer result
4771 
4772   if( _curline == NULL )          // Return NULL at EOF.
4773     return 0;
4774 
4775   skipws();                       // Skip whitespace before identifier
4776   start = end = _ptr;             // Start points at first character
4777   c = *end;                       // Grab character to test
4778   while ((c >= '0' && c <= '9') || (c == '-' && end == start)) {
4779     end++;                        // Increment end pointer
4780     c = *end;                     // Grab character to test
4781   }
4782   if (start == end) {             // We popped out on the first try
4783     parse_err(SYNERR, "integer expected at %c\n", c);
4784     result = 0;
4785   }
4786   else {
4787     _curchar = c;                 // Save the first character of next token
4788     *end = '\0';                  // NULL terminate the string in place
4789     result = atoi(start);         // Convert the string to an integer
4790     *end = _curchar;              // Restore buffer to original condition
4791   }
4792 
4793   // Reset _ptr to next char after token
4794   _ptr = end;
4795 
4796   return result;                   // integer
4797 }
4798 
4799 
4800 //------------------------------get_relation_dup------------------------------
4801 // Looks for a relational operator in the buffer
4802 // invokes a parse_err if the next token is not a relation
4803 // This routine creates a duplicate of the string in the buffer.
get_relation_dup(void)4804 char *ADLParser::get_relation_dup(void) {
4805   char         *result = NULL;    // relational operator being returned
4806 
4807   if( _curline == NULL )          // Return NULL at EOF.
4808     return  NULL;
4809 
4810   skipws();                       // Skip whitespace before relation
4811   char *start = _ptr;             // Store start of relational operator
4812   char first  = *_ptr;            // the first character
4813   if( (first == '=') || (first == '!') || (first == '<') || (first == '>') ) {
4814     next_char();
4815     char second = *_ptr;          // the second character
4816     if( second == '=' ) {
4817       next_char();
4818       char tmp  = *_ptr;
4819       *_ptr = '\0';               // NULL terminate
4820       result = strdup(start);     // Duplicate the string
4821       *_ptr = tmp;                // restore buffer
4822     } else {
4823       parse_err(SYNERR, "relational operator expected at %s\n", _ptr);
4824     }
4825   } else {
4826     parse_err(SYNERR, "relational operator expected at %s\n", _ptr);
4827   }
4828 
4829   return result;
4830 }
4831 
4832 
4833 
4834 //------------------------------get_oplist-------------------------------------
4835 // Looks for identifier pairs where first must be the name of an operand, and
4836 // second must be a name unique in the scope of this instruction.  Stores the
4837 // names with a pointer to the OpClassForm of their type in a local name table.
get_oplist(NameList & parameters,FormDict & operands)4838 void ADLParser::get_oplist(NameList &parameters, FormDict &operands) {
4839   OpClassForm *opclass = NULL;
4840   char        *ident   = NULL;
4841 
4842   do {
4843     next_char();             // skip open paren & comma characters
4844     skipws();
4845     if (_curchar == ')') break;
4846 
4847     // Get operand type, and check it against global name table
4848     ident = get_ident();
4849     if (ident == NULL) {
4850       parse_err(SYNERR, "optype identifier expected at %c\n", _curchar);
4851       return;
4852     }
4853     else {
4854       const Form  *form = _globalNames[ident];
4855       if( form == NULL ) {
4856         parse_err(SYNERR, "undefined operand type %s\n", ident);
4857         return;
4858       }
4859 
4860       // Check for valid operand type
4861       OpClassForm *opc  = form->is_opclass();
4862       OperandForm *oper = form->is_operand();
4863       if((oper == NULL) && (opc == NULL)) {
4864         parse_err(SYNERR, "identifier %s not operand type\n", ident);
4865         return;
4866       }
4867       opclass = opc;
4868     }
4869     // Debugging Stuff
4870     if (_AD._adl_debug > 1) fprintf(stderr, "\tOperand Type: %s\t", ident);
4871 
4872     // Get name of operand and add it to local name table
4873     if( (ident = get_unique_ident(operands, "operand")) == NULL) {
4874       return;
4875     }
4876     // Parameter names must not be global names.
4877     if( _globalNames[ident] != NULL ) {
4878          parse_err(SYNERR, "Reuse of global name %s as operand.\n",ident);
4879          return;
4880     }
4881     operands.Insert(ident, opclass);
4882     parameters.addName(ident);
4883 
4884     // Debugging Stuff
4885     if (_AD._adl_debug > 1) fprintf(stderr, "\tOperand Name: %s\n", ident);
4886     skipws();
4887   } while(_curchar == ',');
4888 
4889   if (_curchar != ')') parse_err(SYNERR, "missing ')'\n");
4890   else {
4891     next_char();  // set current character position past the close paren
4892   }
4893 }
4894 
4895 
4896 //------------------------------get_effectlist---------------------------------
4897 // Looks for identifier pairs where first must be the name of a pre-defined,
4898 // effect, and the second must be the name of an operand defined in the
4899 // operand list of this instruction.  Stores the names with a pointer to the
4900 // effect form in a local effects table.
get_effectlist(FormDict & effects,FormDict & operands,bool & has_call)4901 void ADLParser::get_effectlist(FormDict &effects, FormDict &operands, bool& has_call) {
4902   OperandForm *opForm;
4903   Effect      *eForm;
4904   char        *ident;
4905 
4906   do {
4907     next_char();             // skip open paren & comma characters
4908     skipws();
4909     if (_curchar == ')') break;
4910 
4911     // Get effect type, and check it against global name table
4912     ident = get_ident();
4913     if (ident == NULL) {
4914       parse_err(SYNERR, "effect type identifier expected at %c\n", _curchar);
4915       return;
4916     }
4917     else {
4918       // Check for valid effect type
4919       const Form *form = _globalNames[ident];
4920       if( form == NULL ) {
4921         parse_err(SYNERR, "undefined effect type %s\n", ident);
4922         return;
4923       }
4924       else {
4925         if( (eForm = form->is_effect()) == NULL) {
4926           parse_err(SYNERR, "identifier %s not effect type\n", ident);
4927           return;
4928         }
4929       }
4930     }
4931       // Debugging Stuff
4932     if (_AD._adl_debug > 1) fprintf(stderr, "\tEffect Type: %s\t", ident);
4933     skipws();
4934     if (eForm->is(Component::CALL)) {
4935       if (_AD._adl_debug > 1) fprintf(stderr, "\n");
4936       has_call = true;
4937     } else {
4938       // Get name of operand and check that it is in the local name table
4939       if( (ident = get_unique_ident(effects, "effect")) == NULL) {
4940         parse_err(SYNERR, "missing operand identifier in effect list\n");
4941         return;
4942       }
4943       const Form *form = operands[ident];
4944       opForm = form ? form->is_operand() : NULL;
4945       if( opForm == NULL ) {
4946         if( form && form->is_opclass() ) {
4947           const char* cname = form->is_opclass()->_ident;
4948           parse_err(SYNERR, "operand classes are illegal in effect lists (found %s %s)\n", cname, ident);
4949         } else {
4950           parse_err(SYNERR, "undefined operand %s in effect list\n", ident);
4951         }
4952         return;
4953       }
4954       // Add the pair to the effects table
4955       effects.Insert(ident, eForm);
4956       // Debugging Stuff
4957       if (_AD._adl_debug > 1) fprintf(stderr, "\tOperand Name: %s\n", ident);
4958     }
4959     skipws();
4960   } while(_curchar == ',');
4961 
4962   if (_curchar != ')') parse_err(SYNERR, "missing ')'\n");
4963   else {
4964     next_char();  // set current character position past the close paren
4965   }
4966 }
4967 
4968 
4969 //-------------------------------preproc_line----------------------------------
4970 // A "#line" keyword has been seen, so parse the rest of the line.
preproc_line(void)4971 void ADLParser::preproc_line(void) {
4972   int line = get_int();
4973   skipws_no_preproc();
4974   const char* file = NULL;
4975   if (_curchar == '"') {
4976     next_char();              // Move past the initial '"'
4977     file = _ptr;
4978     while (true) {
4979       if (_curchar == '\n') {
4980         parse_err(SYNERR, "missing '\"' at end of #line directive");
4981         return;
4982       }
4983       if (_curchar == '"') {
4984         *_ptr  = '\0';          // Terminate the string
4985         next_char();
4986         skipws_no_preproc();
4987         break;
4988       }
4989       next_char();
4990     }
4991   }
4992   ensure_end_of_line();
4993   if (file != NULL)
4994     _AD._ADL_file._name = file;
4995   _buf.set_linenum(line);
4996 }
4997 
4998 //------------------------------preproc_define---------------------------------
4999 // A "#define" keyword has been seen, so parse the rest of the line.
preproc_define(void)5000 void ADLParser::preproc_define(void) {
5001   char* flag = get_ident_no_preproc();
5002   skipws_no_preproc();
5003   // only #define x y is supported for now
5004   char* def = get_ident_no_preproc();
5005   _AD.set_preproc_def(flag, def);
5006   skipws_no_preproc();
5007   if (_curchar != '\n') {
5008     parse_err(SYNERR, "non-identifier in preprocessor definition\n");
5009   }
5010 }
5011 
5012 //------------------------------preproc_undef----------------------------------
5013 // An "#undef" keyword has been seen, so parse the rest of the line.
preproc_undef(void)5014 void ADLParser::preproc_undef(void) {
5015   char* flag = get_ident_no_preproc();
5016   skipws_no_preproc();
5017   ensure_end_of_line();
5018   _AD.set_preproc_def(flag, NULL);
5019 }
5020 
5021 
5022 
5023 //------------------------------parse_err--------------------------------------
5024 // Issue a parser error message, and skip to the end of the current line
parse_err(int flag,const char * fmt,...)5025 void ADLParser::parse_err(int flag, const char *fmt, ...) {
5026   va_list args;
5027 
5028   va_start(args, fmt);
5029   if (flag == 1)
5030     _AD._syntax_errs += _AD.emit_msg(0, flag, linenum(), fmt, args);
5031   else if (flag == 2)
5032     _AD._semantic_errs += _AD.emit_msg(0, flag, linenum(), fmt, args);
5033   else
5034     _AD._warnings += _AD.emit_msg(0, flag, linenum(), fmt, args);
5035 
5036   int error_char = _curchar;
5037   char* error_ptr = _ptr+1;
5038   for(;*_ptr != '\n'; _ptr++) ; // Skip to the end of the current line
5039   _curchar = '\n';
5040   va_end(args);
5041   _AD._no_output = 1;
5042 
5043   if (flag == 1) {
5044     char* error_tail = strchr(error_ptr, '\n');
5045     char tem = *error_ptr;
5046     error_ptr[-1] = '\0';
5047     char* error_head = error_ptr-1;
5048     while (error_head > _curline && *error_head)  --error_head;
5049     if (error_tail)  *error_tail = '\0';
5050     fprintf(stderr, "Error Context:  %s>>>%c<<<%s\n",
5051             error_head, error_char, error_ptr);
5052     if (error_tail)  *error_tail = '\n';
5053     error_ptr[-1] = tem;
5054   }
5055 }
5056 
5057 //---------------------------ensure_start_of_line------------------------------
5058 // A preprocessor directive has been encountered.  Be sure it has fallen at
5059 // the beginning of a line, or else report an error.
ensure_start_of_line(void)5060 void ADLParser::ensure_start_of_line(void) {
5061   if (_curchar == '\n') { next_line(); return; }
5062   assert( _ptr >= _curline && _ptr < _curline+strlen(_curline),
5063           "Must be able to find which line we are in" );
5064 
5065   for (char *s = _curline; s < _ptr; s++) {
5066     if (*s > ' ') {
5067       parse_err(SYNERR, "'%c' must be at beginning of line\n", _curchar);
5068       break;
5069     }
5070   }
5071 }
5072 
5073 //---------------------------ensure_end_of_line--------------------------------
5074 // A preprocessor directive has been parsed.  Be sure there is no trailing
5075 // garbage at the end of this line.  Set the scan point to the beginning of
5076 // the next line.
ensure_end_of_line(void)5077 void ADLParser::ensure_end_of_line(void) {
5078   skipws_no_preproc();
5079   if (_curchar != '\n' && _curchar != '\0') {
5080     parse_err(SYNERR, "garbage char '%c' at end of line\n", _curchar);
5081   } else {
5082     next_char_or_line();
5083   }
5084 }
5085 
5086 //---------------------------handle_preproc------------------------------------
5087 // The '#' character introducing a preprocessor directive has been found.
5088 // Parse the whole directive name (e.g., #define, #endif) and take appropriate
5089 // action.  If we are in an "untaken" span of text, simply keep track of
5090 // #ifdef nesting structure, so we can find out when to start taking text
5091 // again.  (In this state, we "sort of support" C's #if directives, enough
5092 // to disregard their associated #else and #endif lines.)  If we are in a
5093 // "taken" span of text, there are two cases:  "#define" and "#undef"
5094 // directives are preserved and passed up to the caller, which eventually
5095 // passes control to the top-level parser loop, which handles #define and
5096 // #undef directly.  (This prevents these directives from occurring in
5097 // arbitrary positions in the AD file--we require better structure than C.)
5098 // In the other case, and #ifdef, #ifndef, #else, or #endif is silently
5099 // processed as whitespace, with the "taken" state of the text correctly
5100 // updated.  This routine returns "false" exactly in the case of a "taken"
5101 // #define or #undef, which tells the caller that a preprocessor token
5102 // has appeared which must be handled explicitly by the parse loop.
handle_preproc_token()5103 bool ADLParser::handle_preproc_token() {
5104   assert(*_ptr == '#', "must be at start of preproc");
5105   ensure_start_of_line();
5106   next_char();
5107   skipws_no_preproc();
5108   char* start_ident = _ptr;
5109   char* ident = (_curchar == '\n') ? NULL : get_ident_no_preproc();
5110   if (ident == NULL) {
5111     parse_err(SYNERR, "expected preprocessor command, got end of line\n");
5112   } else if (!strcmp(ident, "ifdef") ||
5113              !strcmp(ident, "ifndef")) {
5114     char* flag = get_ident_no_preproc();
5115     ensure_end_of_line();
5116     // Test the identifier only if we are already in taken code:
5117     bool flag_def  = preproc_taken() && (_AD.get_preproc_def(flag) != NULL);
5118     bool now_taken = !strcmp(ident, "ifdef") ? flag_def : !flag_def;
5119     begin_if_def(now_taken);
5120   } else if (!strcmp(ident, "if")) {
5121     if (preproc_taken())
5122       parse_err(SYNERR, "unimplemented: #%s %s", ident, _ptr+1);
5123     next_line();
5124     // Intelligently skip this nested C preprocessor directive:
5125     begin_if_def(true);
5126   } else if (!strcmp(ident, "else")) {
5127     ensure_end_of_line();
5128     invert_if_def();
5129   } else if (!strcmp(ident, "endif")) {
5130     ensure_end_of_line();
5131     end_if_def();
5132   } else if (preproc_taken()) {
5133     // pass this token up to the main parser as "#define" or "#undef"
5134     _ptr = start_ident;
5135     _curchar = *--_ptr;
5136     if( _curchar != '#' ) {
5137       parse_err(SYNERR, "no space allowed after # in #define or #undef");
5138       assert(_curchar == '#', "no space allowed after # in #define or #undef");
5139     }
5140     return false;
5141   }
5142   return true;
5143 }
5144 
5145 //---------------------------skipws_common-------------------------------------
5146 // Skip whitespace, including comments and newlines, while keeping an accurate
5147 // line count.
5148 // Maybe handle certain preprocessor constructs: #ifdef, #ifndef, #else, #endif
skipws_common(bool do_preproc)5149 void ADLParser::skipws_common(bool do_preproc) {
5150   char *start = _ptr;
5151   char *next = _ptr + 1;
5152 
5153   if (*_ptr == '\0') {
5154     // Check for string terminator
5155     if (_curchar > ' ')  return;
5156     if (_curchar == '\n') {
5157       if (!do_preproc)  return;            // let caller handle the newline
5158       next_line();
5159       _ptr = _curline; next = _ptr + 1;
5160     }
5161     else if (_curchar == '#' ||
5162         (_curchar == '/' && (*next == '/' || *next == '*'))) {
5163       parse_err(SYNERR, "unimplemented: comment token in a funny place");
5164     }
5165   }
5166   while(_curline != NULL) {                // Check for end of file
5167     if (*_ptr == '\n') {                   // keep proper track of new lines
5168       if (!do_preproc)  break;             // let caller handle the newline
5169       next_line();
5170       _ptr = _curline; next = _ptr + 1;
5171     }
5172     else if ((*_ptr == '/') && (*next == '/'))      // C++ comment
5173       do { _ptr++; next++; } while(*_ptr != '\n');  // So go to end of line
5174     else if ((*_ptr == '/') && (*next == '*')) {    // C comment
5175       _ptr++; next++;
5176       do {
5177         _ptr++; next++;
5178         if (*_ptr == '\n') {               // keep proper track of new lines
5179           next_line();                     // skip newlines within comments
5180           if (_curline == NULL) {          // check for end of file
5181             parse_err(SYNERR, "end-of-file detected inside comment\n");
5182             break;
5183           }
5184           _ptr = _curline; next = _ptr + 1;
5185         }
5186       } while(!((*_ptr == '*') && (*next == '/'))); // Go to end of comment
5187       _ptr = ++next; next++;               // increment _ptr past comment end
5188     }
5189     else if (do_preproc && *_ptr == '#') {
5190       // Note that this calls skipws_common(false) recursively!
5191       bool preproc_handled = handle_preproc_token();
5192       if (!preproc_handled) {
5193         if (preproc_taken()) {
5194           return;  // short circuit
5195         }
5196         ++_ptr;    // skip the preprocessor character
5197       }
5198       next = _ptr+1;
5199     } else if(*_ptr > ' ' && !(do_preproc && !preproc_taken())) {
5200       break;
5201     }
5202     else if (*_ptr == '"' || *_ptr == '\'') {
5203       assert(do_preproc, "only skip strings if doing preproc");
5204       // skip untaken quoted string
5205       int qchar = *_ptr;
5206       while (true) {
5207         ++_ptr;
5208         if (*_ptr == qchar) { ++_ptr; break; }
5209         if (*_ptr == '\\')  ++_ptr;
5210         if (*_ptr == '\n' || *_ptr == '\0') {
5211           parse_err(SYNERR, "newline in string");
5212           break;
5213         }
5214       }
5215       next = _ptr + 1;
5216     }
5217     else { ++_ptr; ++next; }
5218   }
5219   if( _curline != NULL )            // at end of file _curchar isn't valid
5220     _curchar = *_ptr;               // reset _curchar to maintain invariant
5221 }
5222 
5223 //---------------------------cur_char-----------------------------------------
cur_char()5224 char ADLParser::cur_char() {
5225   return (_curchar);
5226 }
5227 
5228 //---------------------------next_char-----------------------------------------
next_char()5229 void ADLParser::next_char() {
5230   if (_curchar == '\n')  parse_err(WARN, "must call next_line!");
5231   _curchar = *++_ptr;
5232   // if ( _curchar == '\n' ) {
5233   //   next_line();
5234   // }
5235 }
5236 
5237 //---------------------------next_char_or_line---------------------------------
next_char_or_line()5238 void ADLParser::next_char_or_line() {
5239   if ( _curchar != '\n' ) {
5240     _curchar = *++_ptr;
5241   } else {
5242     next_line();
5243     _ptr = _curline;
5244     _curchar = *_ptr;  // maintain invariant
5245   }
5246 }
5247 
5248 //---------------------------next_line-----------------------------------------
next_line()5249 void ADLParser::next_line() {
5250   _curline = _buf.get_line();
5251   _curchar = ' ';
5252 }
5253 
5254 //------------------------get_line_string--------------------------------------
5255 // Prepended location descriptor, for debugging.
5256 // Must return a malloced string (that can be freed if desired).
get_line_string(int linenum)5257 char* ADLParser::get_line_string(int linenum) {
5258   const char* file = _AD._ADL_file._name;
5259   int         line = linenum ? linenum : this->linenum();
5260   char* location = (char *)AllocateHeap(strlen(file) + 100);
5261   sprintf(location, "\n#line %d \"%s\"\n", line, file);
5262   return location;
5263 }
5264 
5265 //-------------------------is_literal_constant---------------------------------
is_literal_constant(const char * param)5266 bool ADLParser::is_literal_constant(const char *param) {
5267   if (param[0] == 0)     return false;  // null string
5268   if (param[0] == '(')   return true;   // parenthesized expression
5269   if (param[0] == '0' && (param[1] == 'x' || param[1] == 'X')) {
5270     // Make sure it's a hex constant.
5271     int i = 2;
5272     do {
5273       if( !ADLParser::is_hex_digit(*(param+i)) )  return false;
5274       ++i;
5275     } while( *(param+i) != 0 );
5276     return true;
5277   }
5278   return false;
5279 }
5280 
5281 //---------------------------is_hex_digit--------------------------------------
is_hex_digit(char digit)5282 bool ADLParser::is_hex_digit(char digit) {
5283   return ((digit >= '0') && (digit <= '9'))
5284        ||((digit >= 'a') && (digit <= 'f'))
5285        ||((digit >= 'A') && (digit <= 'F'));
5286 }
5287 
5288 //---------------------------is_int_token--------------------------------------
is_int_token(const char * token,int & intval)5289 bool ADLParser::is_int_token(const char* token, int& intval) {
5290   const char* cp = token;
5291   while (*cp != '\0' && *cp <= ' ')  cp++;
5292   if (*cp == '-')  cp++;
5293   int ndigit = 0;
5294   while (*cp >= '0' && *cp <= '9')  { cp++; ndigit++; }
5295   while (*cp != '\0' && *cp <= ' ')  cp++;
5296   if (ndigit == 0 || *cp != '\0') {
5297     return false;
5298   }
5299   intval = atoi(token);
5300   return true;
5301 }
5302 
skip_expr_ws(const char * str)5303 static const char* skip_expr_ws(const char* str) {
5304   const char * cp = str;
5305   while (cp[0]) {
5306     if (cp[0] <= ' ') {
5307       ++cp;
5308     } else if (cp[0] == '#') {
5309       ++cp;
5310       while (cp[0] == ' ')  ++cp;
5311       assert(0 == strncmp(cp, "line", 4), "must be a #line directive");
5312       const char* eol = strchr(cp, '\n');
5313       assert(eol != NULL, "must find end of line");
5314       if (eol == NULL)  eol = cp + strlen(cp);
5315       cp = eol;
5316     } else {
5317       break;
5318     }
5319   }
5320   return cp;
5321 }
5322 
5323 //-----------------------equivalent_expressions--------------------------------
equivalent_expressions(const char * str1,const char * str2)5324 bool ADLParser::equivalent_expressions(const char* str1, const char* str2) {
5325   if (str1 == str2)
5326     return true;
5327   else if (str1 == NULL || str2 == NULL)
5328     return false;
5329   const char* cp1 = str1;
5330   const char* cp2 = str2;
5331   char in_quote = '\0';
5332   while (cp1[0] && cp2[0]) {
5333     if (!in_quote) {
5334       // skip spaces and/or cpp directives
5335       const char* cp1a = skip_expr_ws(cp1);
5336       const char* cp2a = skip_expr_ws(cp2);
5337       if (cp1a > cp1 && cp2a > cp2) {
5338         cp1 = cp1a; cp2 = cp2a;
5339         continue;
5340       }
5341       if (cp1a > cp1 || cp2a > cp2)  break; // fail
5342     }
5343     // match one non-space char
5344     if (cp1[0] != cp2[0])  break; // fail
5345     char ch = cp1[0];
5346     cp1++; cp2++;
5347     // watch for quotes
5348     if (in_quote && ch == '\\') {
5349       if (cp1[0] != cp2[0])  break; // fail
5350       if (!cp1[0])  break;
5351       cp1++; cp2++;
5352     }
5353     if (in_quote && ch == in_quote) {
5354       in_quote = '\0';
5355     } else if (!in_quote && (ch == '"' || ch == '\'')) {
5356       in_quote = ch;
5357     }
5358   }
5359   return (!cp1[0] && !cp2[0]);
5360 }
5361 
5362 
5363 //-------------------------------trim------------------------------------------
trim(char * & token)5364 void ADLParser::trim(char* &token) {
5365   while (*token <= ' ')  token++;
5366   char* end = token + strlen(token);
5367   while (end > token && *(end-1) <= ' ')  --end;
5368   *end = '\0';
5369 }
5370