adlparse.cpp revision 6010:abec000618bf
1194140Simp/*
2213340Sjmallett * Copyright (c) 1997, 2013, Oracle and/or its affiliates. All rights reserved.
3194140Simp * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
4213340Sjmallett *
5213340Sjmallett * This code is free software; you can redistribute it and/or modify it
6213340Sjmallett * under the terms of the GNU General Public License version 2 only, as
7194140Simp * published by the Free Software Foundation.
8194140Simp *
9194140Simp * This code is distributed in the hope that it will be useful, but WITHOUT
10194140Simp * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
11194140Simp * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
12194140Simp * version 2 for more details (a copy is included in the LICENSE file that
13194140Simp * accompanied this code).
14194140Simp *
15213340Sjmallett * You should have received a copy of the GNU General Public License version
16213340Sjmallett * 2 along with this work; if not, write to the Free Software Foundation,
17194140Simp * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
18194140Simp *
19194140Simp * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
20194140Simp * or visit www.oracle.com if you need additional information or have any
21194143Simp * questions.
22194140Simp *
23210311Sjmallett */
24217653Simp
25194140Simp// ADLPARSE.CPP - Architecture Description Language Parser
26194140Simp// Authors: Chris Vick and Mike Paleczny
27194140Simp#include "adlc.hpp"
28199751Simp
29194140Simp//----------------------------ADLParser----------------------------------------
30232913Sjmallett// Create a new ADL parser
31232913SjmallettADLParser::ADLParser(FileBuff& buffer, ArchDesc& archDesc)
32232913Sjmallett  : _buf(buffer), _AD(archDesc),
33202065Simp    _globalNames(archDesc.globalNames()) {
34194140Simp  _AD._syntax_errs = _AD._semantic_errs = 0; // No errors so far this file
35194143Simp  _AD._warnings    = 0;                      // No warnings either
36194140Simp  _curline         = _ptr = NULL;            // No pointers into buffer yet
37194140Simp
38194140Simp  _preproc_depth = 0;
39213340Sjmallett  _preproc_not_taken = 0;
40213340Sjmallett
41242104Sjmallett  // Delimit command-line definitions from in-file definitions:
42244984Sjmallett  _AD._preproc_list.add_signal();
43250429Simp}
44213340Sjmallett
45213340Sjmallett//------------------------------~ADLParser-------------------------------------
46243469Sjmallett// Delete an ADL parser.
47243469SjmallettADLParser::~ADLParser() {
48243469Sjmallett  if (!_AD._quiet_mode)
49243469Sjmallett    fprintf(stderr,"---------------------------- Errors and Warnings ----------------------------\n");
50243469Sjmallett#ifndef ASSERT
51243469Sjmallett  fprintf(stderr, "**************************************************************\n");
52243469Sjmallett  fprintf(stderr, "***** WARNING: ASSERT is undefined, assertions disabled. *****\n");
53213340Sjmallett  fprintf(stderr, "**************************************************************\n");
54213340Sjmallett#endif
55213340Sjmallett  if( _AD._syntax_errs + _AD._semantic_errs + _AD._warnings == 0 ) {
56213340Sjmallett    if (!_AD._quiet_mode)
57213340Sjmallett      fprintf(stderr,"No errors or warnings to report from phase-1 parse.\n" );
58213340Sjmallett  }
59213340Sjmallett  else {
60213340Sjmallett    if( _AD._syntax_errs ) {      // Any syntax errors?
61213340Sjmallett      fprintf(stderr,"%s:  Found %d syntax error", _buf._fp->_name, _AD._syntax_errs);
62213340Sjmallett      if( _AD._syntax_errs > 1 ) fprintf(stderr,"s.\n\n");
63213340Sjmallett      else fprintf(stderr,".\n\n");
64221753Sgonzo    }
65221753Sgonzo    if( _AD._semantic_errs ) {    // Any semantic errors?
66213340Sjmallett      fprintf(stderr,"%s:  Found %d semantic error", _buf._fp->_name, _AD._semantic_errs);
67221753Sgonzo      if( _AD._semantic_errs > 1 ) fprintf(stderr,"s.\n\n");
68213340Sjmallett      else fprintf(stderr,".\n\n");
69213340Sjmallett    }
70213340Sjmallett    if( _AD._warnings ) {         // Any warnings?
71213340Sjmallett      fprintf(stderr,"%s:  Found %d warning", _buf._fp->_name, _AD._warnings);
72213340Sjmallett      if( _AD._warnings > 1 ) fprintf(stderr,"s.\n\n");
73213340Sjmallett      else fprintf(stderr,".\n\n");
74232720Sjmallett    }
75213340Sjmallett  }
76213340Sjmallett  if (!_AD._quiet_mode)
77213340Sjmallett    fprintf(stderr,"-----------------------------------------------------------------------------\n");
78213340Sjmallett  _AD._TotalLines += linenum()-1;     // -1 for overshoot in "nextline" routine
79213340Sjmallett
80213340Sjmallett  // Write out information we have stored
81213340Sjmallett  // // UNIXism == fsync(stderr);
82213340Sjmallett}
83213340Sjmallett
84213340Sjmallett//------------------------------parse------------------------------------------
85213340Sjmallett// Each top-level keyword should appear as the first non-whitespace on a line.
86213340Sjmallett//
87213340Sjmallettvoid ADLParser::parse() {
88213340Sjmallett  char *ident;
89213340Sjmallett
90194140Simp  // Iterate over the lines in the file buffer parsing Level 1 objects
91213340Sjmallett  for( next_line(); _curline != NULL; next_line()) {
92213340Sjmallett    _ptr = _curline;             // Reset ptr to start of new line
93213340Sjmallett    skipws();                    // Skip any leading whitespace
94213340Sjmallett    ident = get_ident();         // Get first token
95213340Sjmallett    if (ident == NULL) {         // Empty line
96213340Sjmallett      continue;                  // Get the next line
97213340Sjmallett    }
98213340Sjmallett         if (!strcmp(ident, "instruct"))   instr_parse();
99213340Sjmallett    else if (!strcmp(ident, "operand"))    oper_parse();
100213340Sjmallett    else if (!strcmp(ident, "opclass"))    opclass_parse();
101194140Simp    else if (!strcmp(ident, "ins_attrib")) ins_attr_parse();
102213340Sjmallett    else if (!strcmp(ident, "op_attrib"))  op_attr_parse();
103213340Sjmallett    else if (!strcmp(ident, "source"))     source_parse();
104194140Simp    else if (!strcmp(ident, "source_hpp")) source_hpp_parse();
105213340Sjmallett    else if (!strcmp(ident, "register"))   reg_parse();
106213340Sjmallett    else if (!strcmp(ident, "frame"))      frame_parse();
107194140Simp    else if (!strcmp(ident, "encode"))     encode_parse();
108220982Smav    else if (!strcmp(ident, "pipeline"))   pipe_parse();
109220982Smav    else if (!strcmp(ident, "definitions")) definitions_parse();
110220982Smav    else if (!strcmp(ident, "peephole"))   peep_parse();
111220982Smav    else if (!strcmp(ident, "#line"))      preproc_line();
112220982Smav    else if (!strcmp(ident, "#define"))    preproc_define();
113194140Simp    else if (!strcmp(ident, "#undef"))     preproc_undef();
114213340Sjmallett    else {
115213340Sjmallett      parse_err(SYNERR, "expected one of - instruct, operand, ins_attrib, op_attrib, source, register, pipeline, encode\n     Found %s",ident);
116213340Sjmallett    }
117194140Simp  }
118213340Sjmallett  // Add reg_class spill_regs after parsing.
119213340Sjmallett  RegisterForm *regBlock = _AD.get_registers();
120213340Sjmallett  if (regBlock == NULL) {
121213340Sjmallett    parse_err(SEMERR, "Did not declare 'register' definitions");
122213340Sjmallett  }
123213340Sjmallett  regBlock->addSpillRegClass();
124213340Sjmallett
125228005Smarius  // Done with parsing, check consistency.
126213340Sjmallett
127213340Sjmallett  if (_preproc_depth != 0) {
128213340Sjmallett    parse_err(SYNERR, "End of file inside #ifdef");
129213340Sjmallett  }
130213340Sjmallett
131265236Sken  // AttributeForms ins_cost and op_cost must be defined for default behaviour
132213340Sjmallett  if (_globalNames[AttributeForm::_ins_cost] == NULL) {
133213340Sjmallett    parse_err(SEMERR, "Did not declare 'ins_cost' attribute");
134210311Sjmallett  }
135213340Sjmallett  if (_globalNames[AttributeForm::_op_cost] == NULL) {
136213340Sjmallett    parse_err(SEMERR, "Did not declare 'op_cost' attribute");
137213340Sjmallett  }
138213340Sjmallett}
139210311Sjmallett
140220982Smav// ******************** Private Level 1 Parse Functions ********************
141220982Smav//------------------------------instr_parse------------------------------------
142213340Sjmallett// Parse the contents of an instruction definition, build the InstructForm to
143213340Sjmallett// represent that instruction, and add it to the InstructForm list.
144213340Sjmallettvoid ADLParser::instr_parse(void) {
145213340Sjmallett  char          *ident;
146220982Smav  InstructForm  *instr;
147235898Smav  MatchRule     *rule;
148194140Simp  int            match_rules_cnt = 0;
149213340Sjmallett
150213340Sjmallett  // First get the name of the instruction
151213340Sjmallett  if( (ident = get_unique_ident(_globalNames,"instruction")) == NULL )
152213340Sjmallett    return;
153213340Sjmallett  instr = new InstructForm(ident); // Create new instruction form
154213340Sjmallett  instr->_linenum = linenum();
155213340Sjmallett  _globalNames.Insert(ident, instr); // Add name to the name table
156213340Sjmallett  // Debugging Stuff
157213340Sjmallett  if (_AD._adl_debug > 1)
158194140Simp    fprintf(stderr,"Parsing Instruction Form %s\n", ident);
159213340Sjmallett
160213340Sjmallett  // Then get the operands
161213340Sjmallett  skipws();
162213340Sjmallett  if (_curchar != '(') {
163213340Sjmallett    parse_err(SYNERR, "missing '(' in instruct definition\n");
164213340Sjmallett  }
165213340Sjmallett  // Parse the operand list
166213340Sjmallett  else get_oplist(instr->_parameters, instr->_localNames);
167213340Sjmallett  skipws();                        // Skip leading whitespace
168194140Simp  // Check for block delimiter
169213340Sjmallett  if ( (_curchar != '%')
170213340Sjmallett       || ( next_char(),  (_curchar != '{')) ) {
171213340Sjmallett    parse_err(SYNERR, "missing '%%{' in instruction definition\n");
172213340Sjmallett    return;
173213340Sjmallett  }
174194140Simp  next_char();                     // Maintain the invariant
175213340Sjmallett  do {
176213340Sjmallett    ident = get_ident();           // Grab next identifier
177213340Sjmallett    if (ident == NULL) {
178213340Sjmallett      parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
179213340Sjmallett      continue;
180213340Sjmallett    }
181213340Sjmallett    if      (!strcmp(ident, "predicate")) instr->_predicate = pred_parse();
182213340Sjmallett    else if      (!strcmp(ident, "match")) {
183213340Sjmallett      // Allow one instruction have several match rules.
184213340Sjmallett      rule = instr->_matrule;
185213340Sjmallett      if (rule == NULL) {
186213340Sjmallett        // This is first match rule encountered
187217214Sjmallett        rule = match_parse(instr->_localNames);
188217214Sjmallett        if (rule) {
189217214Sjmallett          instr->_matrule = rule;
190213762Sjmallett          // Special case the treatment of Control instructions.
191213762Sjmallett          if( instr->is_ideal_control() ) {
192213762Sjmallett            // Control instructions return a special result, 'Universe'
193213762Sjmallett            rule->_result = "Universe";
194213340Sjmallett          }
195213340Sjmallett          // Check for commutative operations with tree operands.
196213340Sjmallett          matchrule_clone_and_swap(rule, instr->_ident, match_rules_cnt);
197213340Sjmallett        }
198280226Sjfv      } else {
199280226Sjfv        // Find the end of the match rule list
200213340Sjmallett        while (rule->_next != NULL)
201213340Sjmallett          rule = rule->_next;
202213340Sjmallett        // Add the new match rule to the list
203213340Sjmallett        rule->_next = match_parse(instr->_localNames);
204213340Sjmallett        if (rule->_next) {
205213340Sjmallett          rule = rule->_next;
206213340Sjmallett          if( instr->is_ideal_control() ) {
207213340Sjmallett            parse_err(SYNERR, "unique match rule expected for %s\n", rule->_name);
208213340Sjmallett            return;
209213340Sjmallett          }
210213340Sjmallett          assert(match_rules_cnt < 100," too many match rule clones");
211213340Sjmallett          char* buf = (char*) malloc(strlen(instr->_ident) + 4);
212213340Sjmallett          sprintf(buf, "%s_%d", instr->_ident, match_rules_cnt++);
213213340Sjmallett          rule->_result = buf;
214213340Sjmallett          // Check for commutative operations with tree operands.
215213340Sjmallett          matchrule_clone_and_swap(rule, instr->_ident, match_rules_cnt);
216213340Sjmallett        }
217213340Sjmallett      }
218213340Sjmallett    }
219213340Sjmallett    else if (!strcmp(ident, "encode"))  {
220213340Sjmallett      parse_err(SYNERR, "Instructions specify ins_encode, not encode\n");
221213340Sjmallett    }
222213340Sjmallett    else if (!strcmp(ident, "ins_encode"))       ins_encode_parse(*instr);
223213340Sjmallett    // Parse late expand keyword.
224213340Sjmallett    else if (!strcmp(ident, "postalloc_expand")) postalloc_expand_parse(*instr);
225213340Sjmallett    else if (!strcmp(ident, "opcode"))           instr->_opcode    = opcode_parse(instr);
226213340Sjmallett    else if (!strcmp(ident, "size"))             instr->_size      = size_parse(instr);
227213340Sjmallett    else if (!strcmp(ident, "effect"))           effect_parse(instr);
228213340Sjmallett    else if (!strcmp(ident, "expand"))           instr->_exprule   = expand_parse(instr);
229213340Sjmallett    else if (!strcmp(ident, "rewrite"))          instr->_rewrule   = rewrite_parse();
230213340Sjmallett    else if (!strcmp(ident, "constraint")) {
231213340Sjmallett      parse_err(SYNERR, "Instructions do not specify a constraint\n");
232213340Sjmallett    }
233213340Sjmallett    else if (!strcmp(ident, "construct")) {
234213340Sjmallett      parse_err(SYNERR, "Instructions do not specify a construct\n");
235213340Sjmallett    }
236213340Sjmallett    else if (!strcmp(ident, "format"))           instr->_format    = format_parse();
237213340Sjmallett    else if (!strcmp(ident, "interface")) {
238213340Sjmallett      parse_err(SYNERR, "Instructions do not specify an interface\n");
239213340Sjmallett    }
240213340Sjmallett    else if (!strcmp(ident, "ins_pipe"))        ins_pipe_parse(*instr);
241213340Sjmallett    else {  // Done with staticly defined parts of instruction definition
242213340Sjmallett      // Check identifier to see if it is the name of an attribute
243213340Sjmallett      const Form    *form = _globalNames[ident];
244213340Sjmallett      AttributeForm *attr = form ? form->is_attribute() : NULL;
245213340Sjmallett      if (attr && (attr->_atype == INS_ATTR)) {
246213340Sjmallett        // Insert the new attribute into the linked list.
247213340Sjmallett        Attribute *temp = attr_parse(ident);
248220185Sadrian        temp->_next = instr->_attribs;
249232409Sjmallett        instr->_attribs = temp;
250213340Sjmallett      } else {
251213340Sjmallett        parse_err(SYNERR, "expected one of:\n predicate, match, encode, or the name of"
252213340Sjmallett                  " an instruction attribute at %s\n", ident);
253213340Sjmallett      }
254213340Sjmallett    }
255213340Sjmallett    skipws();
256213340Sjmallett  } while(_curchar != '%');
257213340Sjmallett  next_char();
258213340Sjmallett  if (_curchar != '}') {
259213340Sjmallett    parse_err(SYNERR, "missing '%%}' in instruction definition\n");
260213340Sjmallett    return;
261213340Sjmallett  }
262213340Sjmallett  // Check for "Set" form of chain rule
263213340Sjmallett  adjust_set_rule(instr);
264213340Sjmallett  if (_AD._pipeline) {
265213340Sjmallett    // No pipe required for late expand.
266213340Sjmallett    if (instr->expands() || instr->postalloc_expands()) {
267213340Sjmallett      if (instr->_ins_pipe) {
268213340Sjmallett        parse_err(WARN, "ins_pipe and expand rule both specified for instruction \"%s\";"
269213340Sjmallett                  " ins_pipe will be unused\n", instr->_ident);
270213340Sjmallett      }
271242953Sjmallett    } else {
272242953Sjmallett      if (!instr->_ins_pipe) {
273242953Sjmallett        parse_err(WARN, "No ins_pipe specified for instruction \"%s\"\n", instr->_ident);
274213340Sjmallett      }
275213340Sjmallett    }
276213340Sjmallett  }
277213340Sjmallett  // Add instruction to tail of instruction list
278213340Sjmallett  _AD.addForm(instr);
279213340Sjmallett
280213340Sjmallett  // Create instruction form for each additional match rule
281213340Sjmallett  rule = instr->_matrule;
282213340Sjmallett  if (rule != NULL) {
283213340Sjmallett    rule = rule->_next;
284213340Sjmallett    while (rule != NULL) {
285213340Sjmallett      ident = (char*)rule->_result;
286213340Sjmallett      InstructForm *clone = new InstructForm(ident, instr, rule); // Create new instruction form
287213340Sjmallett      _globalNames.Insert(ident, clone); // Add name to the name table
288213340Sjmallett      // Debugging Stuff
289213340Sjmallett      if (_AD._adl_debug > 1)
290213340Sjmallett        fprintf(stderr,"Parsing Instruction Form %s\n", ident);
291213340Sjmallett      // Check for "Set" form of chain rule
292213340Sjmallett      adjust_set_rule(clone);
293213340Sjmallett      // Add instruction to tail of instruction list
294213340Sjmallett      _AD.addForm(clone);
295213340Sjmallett      rule = rule->_next;
296213340Sjmallett      clone->_matrule->_next = NULL; // One match rule per clone
297213340Sjmallett    }
298213340Sjmallett  }
299213340Sjmallett}
300213340Sjmallett
301213340Sjmallett//------------------------------matchrule_clone_and_swap-----------------------
302213340Sjmallett// Check for commutative operations with subtree operands,
303213340Sjmallett// create clones and swap operands.
304213340Sjmallettvoid ADLParser::matchrule_clone_and_swap(MatchRule* rule, const char* instr_ident, int& match_rules_cnt) {
305213340Sjmallett  // Check for commutative operations with tree operands.
306213340Sjmallett  int count = 0;
307213340Sjmallett  rule->count_commutative_op(count);
308213340Sjmallett  if (count > 0) {
309225482Sbrueffer    // Clone match rule and swap commutative operation's operands.
310242953Sjmallett    rule->matchrule_swap_commutative_op(instr_ident, count, match_rules_cnt);
311242953Sjmallett  }
312242953Sjmallett}
313242953Sjmallett
314242953Sjmallett//------------------------------adjust_set_rule--------------------------------
315242953Sjmallett// Check for "Set" form of chain rule
316242953Sjmallettvoid ADLParser::adjust_set_rule(InstructForm *instr) {
317242953Sjmallett  if (instr->_matrule == NULL || instr->_matrule->_rChild == NULL) return;
318242953Sjmallett  const char *rch = instr->_matrule->_rChild->_opType;
319242953Sjmallett  const Form *frm = _globalNames[rch];
320242953Sjmallett  if( (! strcmp(instr->_matrule->_opType,"Set")) &&
321      frm && frm->is_operand() && (! frm->ideal_only()) ) {
322    // Previous implementation, which missed leaP*, but worked for loadCon*
323    unsigned    position = 0;
324    const char *result   = NULL;
325    const char *name     = NULL;
326    const char *optype   = NULL;
327    MatchNode  *right    = instr->_matrule->_rChild;
328    if (right->base_operand(position, _globalNames, result, name, optype)) {
329      position = 1;
330      const char *result2  = NULL;
331      const char *name2    = NULL;
332      const char *optype2  = NULL;
333      // Can not have additional base operands in right side of match!
334      if ( ! right->base_operand( position, _globalNames, result2, name2, optype2) ) {
335        if (instr->_predicate != NULL)
336          parse_err(SYNERR, "ADLC does not support instruction chain rules with predicates");
337        // Chain from input  _ideal_operand_type_,
338        // Needed for shared roots of match-trees
339        ChainList *lst = (ChainList *)_AD._chainRules[optype];
340        if (lst == NULL) {
341          lst = new ChainList();
342          _AD._chainRules.Insert(optype, lst);
343        }
344        if (!lst->search(instr->_matrule->_lChild->_opType)) {
345          const char *cost = instr->cost();
346          if (cost == NULL) {
347            cost = ((AttributeForm*)_globalNames[AttributeForm::_ins_cost])->_attrdef;
348          }
349          // The ADLC does not support chaining from the ideal operand type
350          // of a predicated user-defined operand
351          if( frm->is_operand() == NULL || frm->is_operand()->_predicate == NULL ) {
352            lst->insert(instr->_matrule->_lChild->_opType,cost,instr->_ident);
353          }
354        }
355        // Chain from input  _user_defined_operand_type_,
356        lst = (ChainList *)_AD._chainRules[result];
357        if (lst == NULL) {
358          lst = new ChainList();
359          _AD._chainRules.Insert(result, lst);
360        }
361        if (!lst->search(instr->_matrule->_lChild->_opType)) {
362          const char *cost = instr->cost();
363          if (cost == NULL) {
364            cost = ((AttributeForm*)_globalNames[AttributeForm::_ins_cost])->_attrdef;
365          }
366          // It is safe to chain from the top-level user-defined operand even
367          // if it has a predicate, since the predicate is checked before
368          // the user-defined type is available.
369          lst->insert(instr->_matrule->_lChild->_opType,cost,instr->_ident);
370        }
371      } else {
372        // May have instruction chain rule if root of right-tree is an ideal
373        OperandForm *rightOp = _globalNames[right->_opType]->is_operand();
374        if( rightOp ) {
375          const Form *rightRoot = _globalNames[rightOp->_matrule->_opType];
376          if( rightRoot && rightRoot->ideal_only() ) {
377            const char *chain_op = NULL;
378            if( rightRoot->is_instruction() )
379              chain_op = rightOp->_ident;
380            if( chain_op ) {
381              // Look-up the operation in chain rule table
382              ChainList *lst = (ChainList *)_AD._chainRules[chain_op];
383              if (lst == NULL) {
384                lst = new ChainList();
385                _AD._chainRules.Insert(chain_op, lst);
386              }
387              // if (!lst->search(instr->_matrule->_lChild->_opType)) {
388              const char *cost = instr->cost();
389              if (cost == NULL) {
390                cost = ((AttributeForm*)_globalNames[AttributeForm::_ins_cost])->_attrdef;
391              }
392              // This chains from a top-level operand whose predicate, if any,
393              // has been checked.
394              lst->insert(instr->_matrule->_lChild->_opType,cost,instr->_ident);
395              // }
396            }
397          }
398        }
399      } // end chain rule from right-tree's ideal root
400    }
401  }
402}
403
404
405//------------------------------oper_parse-------------------------------------
406void ADLParser::oper_parse(void) {
407  char          *ident;
408  OperandForm   *oper;
409  AttributeForm *attr;
410  MatchRule     *rule;
411
412  // First get the name of the operand
413  skipws();
414  if( (ident = get_unique_ident(_globalNames,"operand")) == NULL )
415    return;
416  oper = new OperandForm(ident);        // Create new operand form
417  oper->_linenum = linenum();
418  _globalNames.Insert(ident, oper); // Add name to the name table
419
420  // Debugging Stuff
421  if (_AD._adl_debug > 1) fprintf(stderr,"Parsing Operand Form %s\n", ident);
422
423  // Get the component operands
424  skipws();
425  if (_curchar != '(') {
426    parse_err(SYNERR, "missing '(' in operand definition\n");
427    return;
428  }
429  else get_oplist(oper->_parameters, oper->_localNames); // Parse the component operand list
430  skipws();
431  // Check for block delimiter
432  if ((_curchar != '%') || (*(_ptr+1) != '{')) { // If not open block
433    parse_err(SYNERR, "missing '%%{' in operand definition\n");
434    return;
435  }
436  next_char(); next_char();        // Skip over "%{" symbol
437  do {
438    ident = get_ident();           // Grab next identifier
439    if (ident == NULL) {
440      parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
441      continue;
442    }
443    if      (!strcmp(ident, "predicate")) oper->_predicate = pred_parse();
444    else if (!strcmp(ident, "match"))     {
445      // Find the end of the match rule list
446      rule = oper->_matrule;
447      if (rule) {
448        while (rule->_next) rule = rule->_next;
449        // Add the new match rule to the list
450        rule->_next = match_parse(oper->_localNames);
451        if (rule->_next) {
452          rule->_next->_result = oper->_ident;
453        }
454      }
455      else {
456        // This is first match rule encountered
457        oper->_matrule = match_parse(oper->_localNames);
458        if (oper->_matrule) {
459          oper->_matrule->_result = oper->_ident;
460        }
461      }
462    }
463    else if (!strcmp(ident, "encode"))    oper->_interface = interface_parse();
464    else if (!strcmp(ident, "ins_encode")) {
465      parse_err(SYNERR, "Operands specify 'encode', not 'ins_encode'\n");
466    }
467    else if (!strcmp(ident, "opcode"))    {
468      parse_err(SYNERR, "Operands do not specify an opcode\n");
469    }
470    else if (!strcmp(ident, "effect"))    {
471      parse_err(SYNERR, "Operands do not specify an effect\n");
472    }
473    else if (!strcmp(ident, "expand"))    {
474      parse_err(SYNERR, "Operands do not specify an expand\n");
475    }
476    else if (!strcmp(ident, "rewrite"))   {
477      parse_err(SYNERR, "Operands do not specify a rewrite\n");
478    }
479    else if (!strcmp(ident, "constraint"))oper->_constraint= constraint_parse();
480    else if (!strcmp(ident, "construct")) oper->_construct = construct_parse();
481    else if (!strcmp(ident, "format"))    oper->_format    = format_parse();
482    else if (!strcmp(ident, "interface")) oper->_interface = interface_parse();
483    // Check identifier to see if it is the name of an attribute
484    else if (((attr = _globalNames[ident]->is_attribute()) != NULL) &&
485             (attr->_atype == OP_ATTR))   oper->_attribs   = attr_parse(ident);
486    else {
487      parse_err(SYNERR, "expected one of - constraint, predicate, match, encode, format, construct, or the name of a defined operand attribute at %s\n", ident);
488    }
489    skipws();
490  } while(_curchar != '%');
491  next_char();
492  if (_curchar != '}') {
493    parse_err(SYNERR, "missing '%%}' in operand definition\n");
494    return;
495  }
496  // Add operand to tail of operand list
497  _AD.addForm(oper);
498}
499
500//------------------------------opclass_parse----------------------------------
501// Operand Classes are a block with a comma delimited list of operand names
502void ADLParser::opclass_parse(void) {
503  char          *ident;
504  OpClassForm   *opc;
505  OperandForm   *opForm;
506
507  // First get the name of the operand class
508  skipws();
509  if( (ident = get_unique_ident(_globalNames,"opclass")) == NULL )
510    return;
511  opc = new OpClassForm(ident);             // Create new operand class form
512  _globalNames.Insert(ident, opc);  // Add name to the name table
513
514  // Debugging Stuff
515  if (_AD._adl_debug > 1)
516    fprintf(stderr,"Parsing Operand Class Form %s\n", ident);
517
518  // Get the list of operands
519  skipws();
520  if (_curchar != '(') {
521    parse_err(SYNERR, "missing '(' in operand definition\n");
522    return;
523  }
524  do {
525    next_char();                            // Skip past open paren or comma
526    ident = get_ident();                    // Grab next identifier
527    if (ident == NULL) {
528      parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
529      continue;
530    }
531    // Check identifier to see if it is the name of an operand
532    const Form *form = _globalNames[ident];
533    opForm     = form ? form->is_operand() : NULL;
534    if ( opForm ) {
535      opc->_oplst.addName(ident);           // Add operand to opclass list
536      opForm->_classes.addName(opc->_ident);// Add opclass to operand list
537    }
538    else {
539      parse_err(SYNERR, "expected name of a defined operand at %s\n", ident);
540    }
541    skipws();                               // skip trailing whitespace
542  } while (_curchar == ',');                // Check for the comma
543  // Check for closing ')'
544  if (_curchar != ')') {
545    parse_err(SYNERR, "missing ')' or ',' in opclass definition\n");
546    return;
547  }
548  next_char();                              // Consume the ')'
549  skipws();
550  // Check for closing ';'
551  if (_curchar != ';') {
552    parse_err(SYNERR, "missing ';' in opclass definition\n");
553    return;
554  }
555  next_char();                             // Consume the ';'
556  // Add operand to tail of operand list
557  _AD.addForm(opc);
558}
559
560//------------------------------ins_attr_parse---------------------------------
561void ADLParser::ins_attr_parse(void) {
562  char          *ident;
563  char          *aexpr;
564  AttributeForm *attrib;
565
566  // get name for the instruction attribute
567  skipws();                      // Skip leading whitespace
568  if( (ident = get_unique_ident(_globalNames,"inst_attrib")) == NULL )
569    return;
570  // Debugging Stuff
571  if (_AD._adl_debug > 1) fprintf(stderr,"Parsing Ins_Attribute Form %s\n", ident);
572
573  // Get default value of the instruction attribute
574  skipws();                      // Skip whitespace
575  if ((aexpr = get_paren_expr("attribute default expression string")) == NULL) {
576    parse_err(SYNERR, "missing '(' in ins_attrib definition\n");
577    return;
578  }
579  // Debug Stuff
580  if (_AD._adl_debug > 1) fprintf(stderr,"Attribute Expression: %s\n", aexpr);
581
582  // Check for terminator
583  if (_curchar != ';') {
584    parse_err(SYNERR, "missing ';' in ins_attrib definition\n");
585    return;
586  }
587  next_char();                    // Advance past the ';'
588
589  // Construct the attribute, record global name, and store in ArchDesc
590  attrib = new AttributeForm(ident, INS_ATTR, aexpr);
591  _globalNames.Insert(ident, attrib);  // Add name to the name table
592  _AD.addForm(attrib);
593}
594
595//------------------------------op_attr_parse----------------------------------
596void ADLParser::op_attr_parse(void) {
597  char          *ident;
598  char          *aexpr;
599  AttributeForm *attrib;
600
601  // get name for the operand attribute
602  skipws();                      // Skip leading whitespace
603  if( (ident = get_unique_ident(_globalNames,"op_attrib")) == NULL )
604    return;
605  // Debugging Stuff
606  if (_AD._adl_debug > 1) fprintf(stderr,"Parsing Op_Attribute Form %s\n", ident);
607
608  // Get default value of the instruction attribute
609  skipws();                      // Skip whitespace
610  if ((aexpr = get_paren_expr("attribute default expression string")) == NULL) {
611    parse_err(SYNERR, "missing '(' in op_attrib definition\n");
612    return;
613  }
614  // Debug Stuff
615  if (_AD._adl_debug > 1) fprintf(stderr,"Attribute Expression: %s\n", aexpr);
616
617  // Check for terminator
618  if (_curchar != ';') {
619    parse_err(SYNERR, "missing ';' in op_attrib definition\n");
620    return;
621  }
622  next_char();                    // Advance past the ';'
623
624  // Construct the attribute, record global name, and store in ArchDesc
625  attrib = new AttributeForm(ident, OP_ATTR, aexpr);
626  _globalNames.Insert(ident, attrib);
627  _AD.addForm(attrib);
628}
629
630//------------------------------definitions_parse-----------------------------------
631void ADLParser::definitions_parse(void) {
632  skipws();                       // Skip leading whitespace
633  if (_curchar == '%' && *(_ptr+1) == '{') {
634    next_char(); next_char();     // Skip "%{"
635    skipws();
636    while (_curchar != '%' && *(_ptr+1) != '}') {
637      // Process each definition until finding closing string "%}"
638      char *token = get_ident();
639      if (token == NULL) {
640        parse_err(SYNERR, "missing identifier inside definitions block.\n");
641        return;
642      }
643      if (strcmp(token,"int_def")==0)     { int_def_parse(); }
644      // if (strcmp(token,"str_def")==0)   { str_def_parse(); }
645      skipws();
646    }
647  }
648  else {
649    parse_err(SYNERR, "Missing %%{ ... %%} block after definitions keyword.\n");
650    return;
651  }
652}
653
654//------------------------------int_def_parse----------------------------------
655// Parse Example:
656// int_def    MEMORY_REF_COST      (         200,  DEFAULT_COST * 2);
657// <keyword>  <name>               ( <int_value>,   <description>  );
658//
659void ADLParser::int_def_parse(void) {
660  char *name        = NULL;         // Name of definition
661  char *value       = NULL;         // its value,
662  int   int_value   = -1;           // positive values only
663  char *description = NULL;         // textual description
664
665  // Get definition name
666  skipws();                      // Skip whitespace
667  name = get_ident();
668  if (name == NULL) {
669    parse_err(SYNERR, "missing definition name after int_def\n");
670    return;
671  }
672
673  // Check for value of int_def dname( integer_value [, string_expression ] )
674  skipws();
675  if (_curchar == '(') {
676
677    // Parse the integer value.
678    next_char();
679    value = get_ident();
680    if (value == NULL) {
681      parse_err(SYNERR, "missing value in int_def\n");
682      return;
683    }
684    if( !is_int_token(value, int_value) ) {
685      parse_err(SYNERR, "value in int_def is not recognized as integer\n");
686      return;
687    }
688    skipws();
689
690    // Check for description
691    if (_curchar == ',') {
692      next_char();   // skip ','
693
694      description = get_expr("int_def description", ")");
695      if (description == NULL) {
696        parse_err(SYNERR, "invalid or missing description in int_def\n");
697        return;
698      }
699      trim(description);
700    }
701
702    if (_curchar != ')') {
703      parse_err(SYNERR, "missing ')' in register definition statement\n");
704      return;
705    }
706    next_char();
707  }
708
709  // Check for closing ';'
710  skipws();
711  if (_curchar != ';') {
712    parse_err(SYNERR, "missing ';' after int_def\n");
713    return;
714  }
715  next_char();                   // move past ';'
716
717  // Debug Stuff
718  if (_AD._adl_debug > 1) {
719    fprintf(stderr,"int_def: %s ( %s, %s )\n", name,
720            (value), (description ? description : ""));
721  }
722
723  // Record new definition.
724  Expr *expr     = new Expr(name, description, int_value, int_value);
725  const Expr *old_expr = _AD.globalDefs().define(name, expr);
726  if (old_expr != NULL) {
727    parse_err(SYNERR, "Duplicate definition\n");
728    return;
729  }
730
731  return;
732}
733
734
735//------------------------------source_parse-----------------------------------
736void ADLParser::source_parse(void) {
737  SourceForm *source;             // Encode class for instruction/operand
738  char   *rule = NULL;            // String representation of encode rule
739
740  skipws();                       // Skip leading whitespace
741  if ( (rule = find_cpp_block("source block")) == NULL ) {
742    parse_err(SYNERR, "incorrect or missing block for 'source'.\n");
743    return;
744  }
745  // Debug Stuff
746  if (_AD._adl_debug > 1) fprintf(stderr,"Source Form: %s\n", rule);
747
748  source = new SourceForm(rule);    // Build new Source object
749  _AD.addForm(source);
750  // skipws();
751}
752
753//------------------------------source_hpp_parse-------------------------------
754// Parse a source_hpp %{ ... %} block.
755// The code gets stuck into the ad_<arch>.hpp file.
756// If the source_hpp block appears before the register block in the AD
757// file, it goes up at the very top of the ad_<arch>.hpp file, so that
758// it can be used by register encodings, etc.  Otherwise, it goes towards
759// the bottom, where it's useful as a global definition to *.cpp files.
760void ADLParser::source_hpp_parse(void) {
761  char   *rule = NULL;            // String representation of encode rule
762
763  skipws();                       // Skip leading whitespace
764  if ( (rule = find_cpp_block("source_hpp block")) == NULL ) {
765    parse_err(SYNERR, "incorrect or missing block for 'source_hpp'.\n");
766    return;
767  }
768  // Debug Stuff
769  if (_AD._adl_debug > 1) fprintf(stderr,"Header Form: %s\n", rule);
770
771  if (_AD.get_registers() == NULL) {
772    // Very early in the file, before reg_defs, we collect pre-headers.
773    PreHeaderForm* pre_header = new PreHeaderForm(rule);
774    _AD.addForm(pre_header);
775  } else {
776    // Normally, we collect header info, placed at the bottom of the hpp file.
777    HeaderForm* header = new HeaderForm(rule);
778    _AD.addForm(header);
779  }
780}
781
782//------------------------------reg_parse--------------------------------------
783void ADLParser::reg_parse(void) {
784  RegisterForm *regBlock = _AD.get_registers(); // Information about registers encoding
785  if (regBlock == NULL) {
786    // Create the RegisterForm for the architecture description.
787    regBlock = new RegisterForm();    // Build new Source object
788    _AD.addForm(regBlock);
789  }
790
791  skipws();                       // Skip leading whitespace
792  if (_curchar == '%' && *(_ptr+1) == '{') {
793    next_char(); next_char();     // Skip "%{"
794    skipws();
795    while (_curchar != '%' && *(_ptr+1) != '}') {
796      char *token = get_ident();
797      if (token == NULL) {
798        parse_err(SYNERR, "missing identifier inside register block.\n");
799        return;
800      }
801      if (strcmp(token,"reg_def")==0)          { reg_def_parse(); }
802      else if (strcmp(token,"reg_class")==0)   { reg_class_parse(); }
803      else if (strcmp(token,"alloc_class")==0) { alloc_class_parse(); }
804      else if (strcmp(token,"#define")==0)     { preproc_define(); }
805      else { parse_err(SYNERR, "bad token %s inside register block.\n", token); break; }
806      skipws();
807    }
808  }
809  else {
810    parse_err(SYNERR, "Missing %c{ ... %c} block after register keyword.\n",'%','%');
811    return;
812  }
813}
814
815//------------------------------encode_parse-----------------------------------
816void ADLParser::encode_parse(void) {
817  EncodeForm *encBlock;         // Information about instruction/operand encoding
818
819  _AD.getForm(&encBlock);
820  if ( encBlock == NULL) {
821    // Create the EncodeForm for the architecture description.
822    encBlock = new EncodeForm();    // Build new Source object
823    _AD.addForm(encBlock);
824  }
825
826  skipws();                       // Skip leading whitespace
827  if (_curchar == '%' && *(_ptr+1) == '{') {
828    next_char(); next_char();     // Skip "%{"
829    skipws();
830    while (_curchar != '%' && *(_ptr+1) != '}') {
831      char *token = get_ident();
832      if (token == NULL) {
833            parse_err(SYNERR, "missing identifier inside encoding block.\n");
834            return;
835      }
836      if (strcmp(token,"enc_class")==0)   { enc_class_parse(); }
837      skipws();
838    }
839  }
840  else {
841    parse_err(SYNERR, "Missing %c{ ... %c} block after encode keyword.\n",'%','%');
842    return;
843  }
844}
845
846//------------------------------enc_class_parse--------------------------------
847void ADLParser::enc_class_parse(void) {
848  char       *ec_name;           // Name of encoding class being defined
849
850  // Get encoding class name
851  skipws();                      // Skip whitespace
852  ec_name = get_ident();
853  if (ec_name == NULL) {
854    parse_err(SYNERR, "missing encoding class name after encode.\n");
855    return;
856  }
857
858  EncClass  *encoding = _AD._encode->add_EncClass(ec_name);
859  encoding->_linenum = linenum();
860
861  skipws();                      // Skip leading whitespace
862  // Check for optional parameter list
863  if (_curchar == '(') {
864    do {
865      char *pType = NULL;        // parameter type
866      char *pName = NULL;        // parameter name
867
868      next_char();               // skip open paren & comma characters
869      skipws();
870      if (_curchar == ')') break;
871
872      // Get parameter type
873      pType = get_ident();
874      if (pType == NULL) {
875        parse_err(SYNERR, "parameter type expected at %c\n", _curchar);
876        return;
877      }
878
879      skipws();
880      // Get parameter name
881      pName = get_ident();
882      if (pName == NULL) {
883        parse_err(SYNERR, "parameter name expected at %c\n", _curchar);
884        return;
885      }
886
887      // Record parameter type and name
888      encoding->add_parameter( pType, pName );
889
890      skipws();
891    } while(_curchar == ',');
892
893    if (_curchar != ')') parse_err(SYNERR, "missing ')'\n");
894    else {
895      next_char();                  // Skip ')'
896    }
897  } // Done with parameter list
898
899  skipws();
900  // Check for block starting delimiters
901  if ((_curchar != '%') || (*(_ptr+1) != '{')) { // If not open block
902    parse_err(SYNERR, "missing '%c{' in enc_class definition\n", '%');
903    return;
904  }
905  next_char();                      // Skip '%'
906  next_char();                      // Skip '{'
907
908  enc_class_parse_block(encoding, ec_name);
909}
910
911
912void ADLParser::enc_class_parse_block(EncClass* encoding, char* ec_name) {
913  skipws_no_preproc();              // Skip leading whitespace
914  // Prepend location descriptor, for debugging; cf. ADLParser::find_cpp_block
915  if (_AD._adlocation_debug) {
916    encoding->add_code(get_line_string());
917  }
918
919  // Collect the parts of the encode description
920  // (1) strings that are passed through to output
921  // (2) replacement/substitution variable, preceeded by a '$'
922  while ( (_curchar != '%') && (*(_ptr+1) != '}') ) {
923
924    // (1)
925    // Check if there is a string to pass through to output
926    char *start = _ptr;       // Record start of the next string
927    while ((_curchar != '$') && ((_curchar != '%') || (*(_ptr+1) != '}')) ) {
928      // If at the start of a comment, skip past it
929      if( (_curchar == '/') && ((*(_ptr+1) == '/') || (*(_ptr+1) == '*')) ) {
930        skipws_no_preproc();
931      } else {
932        // ELSE advance to the next character, or start of the next line
933        next_char_or_line();
934      }
935    }
936    // If a string was found, terminate it and record in EncClass
937    if ( start != _ptr ) {
938      *_ptr  = '\0';          // Terminate the string
939      encoding->add_code(start);
940    }
941
942    // (2)
943    // If we are at a replacement variable,
944    // copy it and record in EncClass
945    if (_curchar == '$') {
946      // Found replacement Variable
947      char* rep_var = get_rep_var_ident_dup();
948      // Add flag to _strings list indicating we should check _rep_vars
949      encoding->add_rep_var(rep_var);
950    }
951  } // end while part of format description
952  next_char();                      // Skip '%'
953  next_char();                      // Skip '}'
954
955  skipws();
956
957  if (_AD._adlocation_debug) {
958    encoding->add_code(end_line_marker());
959  }
960
961  // Debug Stuff
962  if (_AD._adl_debug > 1) fprintf(stderr,"EncodingClass Form: %s\n", ec_name);
963}
964
965//------------------------------frame_parse-----------------------------------
966void ADLParser::frame_parse(void) {
967  FrameForm  *frame;              // Information about stack-frame layout
968  char       *desc = NULL;        // String representation of frame
969
970  skipws();                       // Skip leading whitespace
971
972  frame = new FrameForm();        // Build new Frame object
973  // Check for open block sequence
974  skipws();                       // Skip leading whitespace
975  if (_curchar == '%' && *(_ptr+1) == '{') {
976    next_char(); next_char();     // Skip "%{"
977    skipws();
978    while (_curchar != '%' && *(_ptr+1) != '}') {
979      char *token = get_ident();
980      if (token == NULL) {
981            parse_err(SYNERR, "missing identifier inside frame block.\n");
982            return;
983      }
984      if (strcmp(token,"stack_direction")==0) {
985        stack_dir_parse(frame);
986      }
987      if (strcmp(token,"sync_stack_slots")==0) {
988        sync_stack_slots_parse(frame);
989      }
990      if (strcmp(token,"frame_pointer")==0) {
991        frame_pointer_parse(frame, false);
992      }
993      if (strcmp(token,"interpreter_frame_pointer")==0) {
994        interpreter_frame_pointer_parse(frame, false);
995      }
996      if (strcmp(token,"inline_cache_reg")==0) {
997        inline_cache_parse(frame, false);
998      }
999      if (strcmp(token,"compiler_method_oop_reg")==0) {
1000        parse_err(WARN, "Using obsolete Token, compiler_method_oop_reg");
1001        skipws();
1002      }
1003      if (strcmp(token,"interpreter_method_oop_reg")==0) {
1004        interpreter_method_oop_parse(frame, false);
1005      }
1006      if (strcmp(token,"cisc_spilling_operand_name")==0) {
1007        cisc_spilling_operand_name_parse(frame, false);
1008      }
1009      if (strcmp(token,"stack_alignment")==0) {
1010        stack_alignment_parse(frame);
1011      }
1012      if (strcmp(token,"return_addr")==0) {
1013        return_addr_parse(frame, false);
1014      }
1015      if (strcmp(token,"in_preserve_stack_slots")==0) {
1016        preserve_stack_parse(frame);
1017      }
1018      if (strcmp(token,"out_preserve_stack_slots")==0) {
1019        parse_err(WARN, "Using obsolete token, out_preserve_stack_slots");
1020        skipws();
1021      }
1022      if (strcmp(token,"varargs_C_out_slots_killed")==0) {
1023        frame->_varargs_C_out_slots_killed = parse_one_arg("varargs C out slots killed");
1024      }
1025      if (strcmp(token,"calling_convention")==0) {
1026        frame->_calling_convention = calling_convention_parse();
1027      }
1028      if (strcmp(token,"return_value")==0) {
1029        frame->_return_value = return_value_parse();
1030      }
1031      if (strcmp(token,"c_frame_pointer")==0) {
1032        frame_pointer_parse(frame, true);
1033      }
1034      if (strcmp(token,"c_return_addr")==0) {
1035        return_addr_parse(frame, true);
1036      }
1037      if (strcmp(token,"c_calling_convention")==0) {
1038        frame->_c_calling_convention = calling_convention_parse();
1039      }
1040      if (strcmp(token,"c_return_value")==0) {
1041        frame->_c_return_value = return_value_parse();
1042      }
1043
1044      skipws();
1045    }
1046  }
1047  else {
1048    parse_err(SYNERR, "Missing %c{ ... %c} block after encode keyword.\n",'%','%');
1049    return;
1050  }
1051  // All Java versions are required, native versions are optional
1052  if(frame->_frame_pointer == NULL) {
1053    parse_err(SYNERR, "missing frame pointer definition in frame section.\n");
1054    return;
1055  }
1056  // !!!!! !!!!!
1057  // if(frame->_interpreter_frame_ptr_reg == NULL) {
1058  //   parse_err(SYNERR, "missing interpreter frame pointer definition in frame section.\n");
1059  //   return;
1060  // }
1061  if(frame->_alignment == NULL) {
1062    parse_err(SYNERR, "missing alignment definition in frame section.\n");
1063    return;
1064  }
1065  if(frame->_return_addr == NULL) {
1066    parse_err(SYNERR, "missing return address location in frame section.\n");
1067    return;
1068  }
1069  if(frame->_in_preserve_slots == NULL) {
1070    parse_err(SYNERR, "missing stack slot preservation definition in frame section.\n");
1071    return;
1072  }
1073  if(frame->_varargs_C_out_slots_killed == NULL) {
1074    parse_err(SYNERR, "missing varargs C out slots killed definition in frame section.\n");
1075    return;
1076  }
1077  if(frame->_calling_convention == NULL) {
1078    parse_err(SYNERR, "missing calling convention definition in frame section.\n");
1079    return;
1080  }
1081  if(frame->_return_value == NULL) {
1082    parse_err(SYNERR, "missing return value definition in frame section.\n");
1083    return;
1084  }
1085  // Fill natives in identically with the Java versions if not present.
1086  if(frame->_c_frame_pointer == NULL) {
1087    frame->_c_frame_pointer = frame->_frame_pointer;
1088  }
1089  if(frame->_c_return_addr == NULL) {
1090    frame->_c_return_addr = frame->_return_addr;
1091    frame->_c_return_addr_loc = frame->_return_addr_loc;
1092  }
1093  if(frame->_c_calling_convention == NULL) {
1094    frame->_c_calling_convention = frame->_calling_convention;
1095  }
1096  if(frame->_c_return_value == NULL) {
1097    frame->_c_return_value = frame->_return_value;
1098  }
1099
1100  // Debug Stuff
1101  if (_AD._adl_debug > 1) fprintf(stderr,"Frame Form: %s\n", desc);
1102
1103  // Create the EncodeForm for the architecture description.
1104  _AD.addForm(frame);
1105  // skipws();
1106}
1107
1108//------------------------------stack_dir_parse--------------------------------
1109void ADLParser::stack_dir_parse(FrameForm *frame) {
1110  char *direction = parse_one_arg("stack direction entry");
1111  if (strcmp(direction, "TOWARDS_LOW") == 0) {
1112    frame->_direction = false;
1113  }
1114  else if (strcmp(direction, "TOWARDS_HIGH") == 0) {
1115    frame->_direction = true;
1116  }
1117  else {
1118    parse_err(SYNERR, "invalid value inside stack direction entry.\n");
1119    return;
1120  }
1121}
1122
1123//------------------------------sync_stack_slots_parse-------------------------
1124void ADLParser::sync_stack_slots_parse(FrameForm *frame) {
1125    // Assign value into frame form
1126    frame->_sync_stack_slots = parse_one_arg("sync stack slots entry");
1127}
1128
1129//------------------------------frame_pointer_parse----------------------------
1130void ADLParser::frame_pointer_parse(FrameForm *frame, bool native) {
1131  char *frame_pointer = parse_one_arg("frame pointer entry");
1132  // Assign value into frame form
1133  if (native) { frame->_c_frame_pointer = frame_pointer; }
1134  else        { frame->_frame_pointer   = frame_pointer; }
1135}
1136
1137//------------------------------interpreter_frame_pointer_parse----------------------------
1138void ADLParser::interpreter_frame_pointer_parse(FrameForm *frame, bool native) {
1139  frame->_interpreter_frame_pointer_reg = parse_one_arg("interpreter frame pointer entry");
1140}
1141
1142//------------------------------inline_cache_parse-----------------------------
1143void ADLParser::inline_cache_parse(FrameForm *frame, bool native) {
1144  frame->_inline_cache_reg = parse_one_arg("inline cache reg entry");
1145}
1146
1147//------------------------------interpreter_method_oop_parse------------------
1148void ADLParser::interpreter_method_oop_parse(FrameForm *frame, bool native) {
1149  frame->_interpreter_method_oop_reg = parse_one_arg("method oop reg entry");
1150}
1151
1152//------------------------------cisc_spilling_operand_parse---------------------
1153void ADLParser::cisc_spilling_operand_name_parse(FrameForm *frame, bool native) {
1154  frame->_cisc_spilling_operand_name = parse_one_arg("cisc spilling operand name");
1155}
1156
1157//------------------------------stack_alignment_parse--------------------------
1158void ADLParser::stack_alignment_parse(FrameForm *frame) {
1159  char *alignment = parse_one_arg("stack alignment entry");
1160  // Assign value into frame
1161  frame->_alignment   = alignment;
1162}
1163
1164//------------------------------parse_one_arg-------------------------------
1165char *ADLParser::parse_one_arg(const char *description) {
1166  char *token = NULL;
1167  if(_curchar == '(') {
1168    next_char();
1169    skipws();
1170    token = get_expr(description, ")");
1171    if (token == NULL) {
1172      parse_err(SYNERR, "missing value inside %s.\n", description);
1173      return NULL;
1174    }
1175    next_char();           // skip the close paren
1176    if(_curchar != ';') {  // check for semi-colon
1177      parse_err(SYNERR, "missing %c in.\n", ';', description);
1178      return NULL;
1179    }
1180    next_char();           // skip the semi-colon
1181  }
1182  else {
1183    parse_err(SYNERR, "Missing %c in.\n", '(', description);
1184    return NULL;
1185  }
1186
1187  trim(token);
1188  return token;
1189}
1190
1191//------------------------------return_addr_parse------------------------------
1192void ADLParser::return_addr_parse(FrameForm *frame, bool native) {
1193  bool in_register  = true;
1194  if(_curchar == '(') {
1195    next_char();
1196    skipws();
1197    char *token = get_ident();
1198    if (token == NULL) {
1199      parse_err(SYNERR, "missing value inside return address entry.\n");
1200      return;
1201    }
1202    // check for valid values for stack/register
1203    if (strcmp(token, "REG") == 0) {
1204      in_register = true;
1205    }
1206    else if (strcmp(token, "STACK") == 0) {
1207      in_register = false;
1208    }
1209    else {
1210      parse_err(SYNERR, "invalid value inside return_address entry.\n");
1211      return;
1212    }
1213    if (native) { frame->_c_return_addr_loc = in_register; }
1214    else        { frame->_return_addr_loc   = in_register; }
1215
1216    // Parse expression that specifies register or stack position
1217    skipws();
1218    char *token2 = get_expr("return address entry", ")");
1219    if (token2 == NULL) {
1220      parse_err(SYNERR, "missing value inside return address entry.\n");
1221      return;
1222    }
1223    next_char();           // skip the close paren
1224    if (native) { frame->_c_return_addr = token2; }
1225    else        { frame->_return_addr   = token2; }
1226
1227    if(_curchar != ';') {  // check for semi-colon
1228      parse_err(SYNERR, "missing %c in return address entry.\n", ';');
1229      return;
1230    }
1231    next_char();           // skip the semi-colon
1232  }
1233  else {
1234    parse_err(SYNERR, "Missing %c in return_address entry.\n", '(');
1235  }
1236}
1237
1238//------------------------------preserve_stack_parse---------------------------
1239void ADLParser::preserve_stack_parse(FrameForm *frame) {
1240  if(_curchar == '(') {
1241    char *token = get_paren_expr("preserve_stack_slots");
1242    frame->_in_preserve_slots   = token;
1243
1244    if(_curchar != ';') {  // check for semi-colon
1245      parse_err(SYNERR, "missing %c in preserve stack slot entry.\n", ';');
1246      return;
1247    }
1248    next_char();           // skip the semi-colon
1249  }
1250  else {
1251    parse_err(SYNERR, "Missing %c in preserve stack slot entry.\n", '(');
1252  }
1253}
1254
1255//------------------------------calling_convention_parse-----------------------
1256char *ADLParser::calling_convention_parse() {
1257  char   *desc = NULL;          // String representation of calling_convention
1258
1259  skipws();                     // Skip leading whitespace
1260  if ( (desc = find_cpp_block("calling convention block")) == NULL ) {
1261    parse_err(SYNERR, "incorrect or missing block for 'calling_convention'.\n");
1262  }
1263  return desc;
1264}
1265
1266//------------------------------return_value_parse-----------------------------
1267char *ADLParser::return_value_parse() {
1268  char   *desc = NULL;          // String representation of calling_convention
1269
1270  skipws();                     // Skip leading whitespace
1271  if ( (desc = find_cpp_block("return value block")) == NULL ) {
1272    parse_err(SYNERR, "incorrect or missing block for 'return_value'.\n");
1273  }
1274  return desc;
1275}
1276
1277//------------------------------ins_pipe_parse---------------------------------
1278void ADLParser::ins_pipe_parse(InstructForm &instr) {
1279  char * ident;
1280
1281  skipws();
1282  if ( _curchar != '(' ) {       // Check for delimiter
1283    parse_err(SYNERR, "missing \"(\" in ins_pipe definition\n");
1284    return;
1285  }
1286
1287  next_char();
1288  ident = get_ident();           // Grab next identifier
1289
1290  if (ident == NULL) {
1291    parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
1292    return;
1293  }
1294
1295  skipws();
1296  if ( _curchar != ')' ) {       // Check for delimiter
1297    parse_err(SYNERR, "missing \")\" in ins_pipe definition\n");
1298    return;
1299  }
1300
1301  next_char();                   // skip the close paren
1302  if(_curchar != ';') {          // check for semi-colon
1303    parse_err(SYNERR, "missing %c in return value entry.\n", ';');
1304    return;
1305  }
1306  next_char();                   // skip the semi-colon
1307
1308  // Check ident for validity
1309  if (_AD._pipeline && !_AD._pipeline->_classlist.search(ident)) {
1310    parse_err(SYNERR, "\"%s\" is not a valid pipeline class\n", ident);
1311    return;
1312  }
1313
1314  // Add this instruction to the list in the pipeline class
1315  _AD._pipeline->_classdict[ident]->is_pipeclass()->_instructs.addName(instr._ident);
1316
1317  // Set the name of the pipeline class in the instruction
1318  instr._ins_pipe = ident;
1319  return;
1320}
1321
1322//------------------------------pipe_parse-------------------------------------
1323void ADLParser::pipe_parse(void) {
1324  PipelineForm *pipeline;         // Encode class for instruction/operand
1325  char * ident;
1326
1327  pipeline = new PipelineForm();  // Build new Source object
1328  _AD.addForm(pipeline);
1329
1330  skipws();                       // Skip leading whitespace
1331  // Check for block delimiter
1332  if ( (_curchar != '%')
1333       || ( next_char(),  (_curchar != '{')) ) {
1334    parse_err(SYNERR, "missing '%%{' in pipeline definition\n");
1335    return;
1336  }
1337  next_char();                     // Maintain the invariant
1338  do {
1339    ident = get_ident();           // Grab next identifier
1340    if (ident == NULL) {
1341      parse_err(SYNERR, "keyword identifier expected at %c\n", _curchar);
1342      continue;
1343    }
1344    if      (!strcmp(ident, "resources" )) resource_parse(*pipeline);
1345    else if (!strcmp(ident, "pipe_desc" )) pipe_desc_parse(*pipeline);
1346    else if (!strcmp(ident, "pipe_class")) pipe_class_parse(*pipeline);
1347    else if (!strcmp(ident, "define")) {
1348      skipws();
1349      if ( (_curchar != '%')
1350           || ( next_char(),  (_curchar != '{')) ) {
1351        parse_err(SYNERR, "expected '%%{'\n");
1352        return;
1353      }
1354      next_char(); skipws();
1355
1356      char *node_class = get_ident();
1357      if (node_class == NULL) {
1358        parse_err(SYNERR, "expected identifier, found \"%c\"\n", _curchar);
1359        return;
1360      }
1361
1362      skipws();
1363      if (_curchar != ',' && _curchar != '=') {
1364        parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1365        break;
1366      }
1367      next_char(); skipws();
1368
1369      char *pipe_class = get_ident();
1370      if (pipe_class == NULL) {
1371        parse_err(SYNERR, "expected identifier, found \"%c\"\n", _curchar);
1372        return;
1373      }
1374      if (_curchar != ';' ) {
1375        parse_err(SYNERR, "expected `;`, found '%c'\n", _curchar);
1376        break;
1377      }
1378      next_char();              // Skip over semi-colon
1379
1380      skipws();
1381      if ( (_curchar != '%')
1382           || ( next_char(),  (_curchar != '}')) ) {
1383        parse_err(SYNERR, "expected '%%}', found \"%c\"\n", _curchar);
1384      }
1385      next_char();
1386
1387      // Check ident for validity
1388      if (_AD._pipeline && !_AD._pipeline->_classlist.search(pipe_class)) {
1389        parse_err(SYNERR, "\"%s\" is not a valid pipeline class\n", pipe_class);
1390        return;
1391      }
1392
1393      // Add this machine node to the list in the pipeline class
1394      _AD._pipeline->_classdict[pipe_class]->is_pipeclass()->_instructs.addName(node_class);
1395
1396      MachNodeForm *machnode = new MachNodeForm(node_class); // Create new machnode form
1397      machnode->_machnode_pipe = pipe_class;
1398
1399      _AD.addForm(machnode);
1400    }
1401    else if (!strcmp(ident, "attributes")) {
1402      bool vsi_seen = false;
1403
1404      skipws();
1405      if ( (_curchar != '%')
1406           || ( next_char(),  (_curchar != '{')) ) {
1407        parse_err(SYNERR, "expected '%%{'\n");
1408        return;
1409      }
1410      next_char(); skipws();
1411
1412      while (_curchar != '%') {
1413        ident = get_ident();
1414        if (ident == NULL)
1415          break;
1416
1417        if (!strcmp(ident, "variable_size_instructions")) {
1418          skipws();
1419          if (_curchar == ';') {
1420            next_char(); skipws();
1421          }
1422
1423          pipeline->_variableSizeInstrs = true;
1424          vsi_seen = true;
1425          continue;
1426        }
1427
1428        if (!strcmp(ident, "fixed_size_instructions")) {
1429          skipws();
1430          if (_curchar == ';') {
1431            next_char(); skipws();
1432          }
1433
1434          pipeline->_variableSizeInstrs = false;
1435          vsi_seen = true;
1436          continue;
1437        }
1438
1439        if (!strcmp(ident, "branch_has_delay_slot")) {
1440          skipws();
1441          if (_curchar == ';') {
1442            next_char(); skipws();
1443          }
1444
1445          pipeline->_branchHasDelaySlot = true;
1446          continue;
1447        }
1448
1449        if (!strcmp(ident, "max_instructions_per_bundle")) {
1450          skipws();
1451          if (_curchar != '=') {
1452            parse_err(SYNERR, "expected `=`\n");
1453            break;
1454            }
1455
1456          next_char(); skipws();
1457          pipeline->_maxInstrsPerBundle = get_int();
1458          skipws();
1459
1460          if (_curchar == ';') {
1461            next_char(); skipws();
1462          }
1463
1464          continue;
1465        }
1466
1467        if (!strcmp(ident, "max_bundles_per_cycle")) {
1468          skipws();
1469          if (_curchar != '=') {
1470            parse_err(SYNERR, "expected `=`\n");
1471            break;
1472            }
1473
1474          next_char(); skipws();
1475          pipeline->_maxBundlesPerCycle = get_int();
1476          skipws();
1477
1478          if (_curchar == ';') {
1479            next_char(); skipws();
1480          }
1481
1482          continue;
1483        }
1484
1485        if (!strcmp(ident, "instruction_unit_size")) {
1486          skipws();
1487          if (_curchar != '=') {
1488            parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1489            break;
1490            }
1491
1492          next_char(); skipws();
1493          pipeline->_instrUnitSize = get_int();
1494          skipws();
1495
1496          if (_curchar == ';') {
1497            next_char(); skipws();
1498          }
1499
1500          continue;
1501        }
1502
1503        if (!strcmp(ident, "bundle_unit_size")) {
1504          skipws();
1505          if (_curchar != '=') {
1506            parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1507            break;
1508            }
1509
1510          next_char(); skipws();
1511          pipeline->_bundleUnitSize = get_int();
1512          skipws();
1513
1514          if (_curchar == ';') {
1515            next_char(); skipws();
1516          }
1517
1518          continue;
1519        }
1520
1521        if (!strcmp(ident, "instruction_fetch_unit_size")) {
1522          skipws();
1523          if (_curchar != '=') {
1524            parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1525            break;
1526            }
1527
1528          next_char(); skipws();
1529          pipeline->_instrFetchUnitSize = get_int();
1530          skipws();
1531
1532          if (_curchar == ';') {
1533            next_char(); skipws();
1534          }
1535
1536          continue;
1537        }
1538
1539        if (!strcmp(ident, "instruction_fetch_units")) {
1540          skipws();
1541          if (_curchar != '=') {
1542            parse_err(SYNERR, "expected `=`, found '%c'\n", _curchar);
1543            break;
1544            }
1545
1546          next_char(); skipws();
1547          pipeline->_instrFetchUnits = get_int();
1548          skipws();
1549
1550          if (_curchar == ';') {
1551            next_char(); skipws();
1552          }
1553
1554          continue;
1555        }
1556
1557        if (!strcmp(ident, "nops")) {
1558          skipws();
1559          if (_curchar != '(') {
1560            parse_err(SYNERR, "expected `(`, found '%c'\n", _curchar);
1561            break;
1562            }
1563
1564          next_char(); skipws();
1565
1566          while (_curchar != ')') {
1567            ident = get_ident();
1568            if (ident == NULL) {
1569              parse_err(SYNERR, "expected identifier for nop instruction, found '%c'\n", _curchar);
1570              break;
1571            }
1572
1573            pipeline->_noplist.addName(ident);
1574            pipeline->_nopcnt++;
1575            skipws();
1576
1577            if (_curchar == ',') {
1578              next_char(); skipws();
1579            }
1580          }
1581
1582          next_char(); skipws();
1583
1584          if (_curchar == ';') {
1585            next_char(); skipws();
1586          }
1587
1588          continue;
1589        }
1590
1591        parse_err(SYNERR, "unknown specifier \"%s\"\n", ident);
1592      }
1593
1594      if ( (_curchar != '%')
1595           || ( next_char(),  (_curchar != '}')) ) {
1596        parse_err(SYNERR, "expected '%%}', found \"%c\"\n", _curchar);
1597      }
1598      next_char(); skipws();
1599
1600      if (pipeline->_maxInstrsPerBundle == 0)
1601        parse_err(SYNERR, "\"max_instructions_per_bundle\" unspecified\n");
1602      if (pipeline->_instrUnitSize == 0 && pipeline->_bundleUnitSize == 0)
1603        parse_err(SYNERR, "\"instruction_unit_size\" and \"bundle_unit_size\" unspecified\n");
1604      if (pipeline->_instrFetchUnitSize == 0)
1605        parse_err(SYNERR, "\"instruction_fetch_unit_size\" unspecified\n");
1606      if (pipeline->_instrFetchUnits == 0)
1607        parse_err(SYNERR, "\"instruction_fetch_units\" unspecified\n");
1608      if (!vsi_seen)
1609        parse_err(SYNERR, "\"variable_size_instruction\" or \"fixed_size_instruction\" unspecified\n");
1610    }
1611    else {  // Done with staticly defined parts of instruction definition
1612      parse_err(SYNERR, "expected one of \"resources\", \"pipe_desc\", \"pipe_class\", found \"%s\"\n", ident);
1613      return;
1614    }
1615    skipws();
1616    if (_curchar == ';')
1617      skipws();
1618  } while(_curchar != '%');
1619
1620  next_char();
1621  if (_curchar != '}') {
1622    parse_err(SYNERR, "missing \"%%}\" in pipeline definition\n");
1623    return;
1624  }
1625
1626  next_char();
1627}
1628
1629//------------------------------resource_parse----------------------------
1630void ADLParser::resource_parse(PipelineForm &pipeline) {
1631  ResourceForm *resource;
1632  char * ident;
1633  char * expr;
1634  unsigned mask;
1635  pipeline._rescount = 0;
1636
1637  skipws();                       // Skip leading whitespace
1638
1639  if (_curchar != '(') {
1640    parse_err(SYNERR, "missing \"(\" in resource definition\n");
1641    return;
1642  }
1643
1644  do {
1645    next_char();                   // Skip "(" or ","
1646    ident = get_ident();           // Grab next identifier
1647
1648    if (_AD._adl_debug > 1) {
1649      if (ident != NULL) {
1650        fprintf(stderr, "resource_parse: identifier: %s\n", ident);
1651      }
1652    }
1653
1654    if (ident == NULL) {
1655      parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1656      return;
1657    }
1658    skipws();
1659
1660    if (_curchar != '=') {
1661      mask = (1 << pipeline._rescount++);
1662    }
1663    else {
1664      next_char(); skipws();
1665      expr = get_ident();          // Grab next identifier
1666      if (expr == NULL) {
1667        parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1668        return;
1669      }
1670      resource = (ResourceForm *) pipeline._resdict[expr];
1671      if (resource == NULL) {
1672        parse_err(SYNERR, "resource \"%s\" is not defined\n", expr);
1673        return;
1674      }
1675      mask = resource->mask();
1676
1677      skipws();
1678      while (_curchar == '|') {
1679        next_char(); skipws();
1680
1681        expr = get_ident();          // Grab next identifier
1682        if (expr == NULL) {
1683          parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1684          return;
1685        }
1686
1687        resource = (ResourceForm *) pipeline._resdict[expr];   // Look up the value
1688        if (resource == NULL) {
1689          parse_err(SYNERR, "resource \"%s\" is not defined\n", expr);
1690          return;
1691        }
1692
1693        mask |= resource->mask();
1694        skipws();
1695      }
1696    }
1697
1698    resource = new ResourceForm(mask);
1699
1700    pipeline._resdict.Insert(ident, resource);
1701    pipeline._reslist.addName(ident);
1702  } while (_curchar == ',');
1703
1704  if (_curchar != ')') {
1705      parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
1706      return;
1707  }
1708
1709  next_char();                 // Skip ")"
1710  if (_curchar == ';')
1711    next_char();               // Skip ";"
1712}
1713
1714//------------------------------resource_parse----------------------------
1715void ADLParser::pipe_desc_parse(PipelineForm &pipeline) {
1716  char * ident;
1717
1718  skipws();                       // Skip leading whitespace
1719
1720  if (_curchar != '(') {
1721    parse_err(SYNERR, "missing \"(\" in pipe_desc definition\n");
1722    return;
1723  }
1724
1725  do {
1726    next_char();                   // Skip "(" or ","
1727    ident = get_ident();           // Grab next identifier
1728    if (ident == NULL) {
1729      parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1730      return;
1731    }
1732
1733    // Add the name to the list
1734    pipeline._stages.addName(ident);
1735    pipeline._stagecnt++;
1736
1737    skipws();
1738  } while (_curchar == ',');
1739
1740  if (_curchar != ')') {
1741      parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
1742      return;
1743  }
1744
1745  next_char();                     // Skip ")"
1746  if (_curchar == ';')
1747    next_char();                   // Skip ";"
1748}
1749
1750//------------------------------pipe_class_parse--------------------------
1751void ADLParser::pipe_class_parse(PipelineForm &pipeline) {
1752  PipeClassForm *pipe_class;
1753  char * ident;
1754  char * stage;
1755  char * read_or_write;
1756  int is_write;
1757  int is_read;
1758  OperandForm  *oper;
1759
1760  skipws();                       // Skip leading whitespace
1761
1762  ident = get_ident();            // Grab next identifier
1763
1764  if (ident == NULL) {
1765    parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1766    return;
1767  }
1768
1769  // Create a record for the pipe_class
1770  pipe_class = new PipeClassForm(ident, ++pipeline._classcnt);
1771  pipeline._classdict.Insert(ident, pipe_class);
1772  pipeline._classlist.addName(ident);
1773
1774  // Then get the operands
1775  skipws();
1776  if (_curchar != '(') {
1777    parse_err(SYNERR, "missing \"(\" in pipe_class definition\n");
1778  }
1779  // Parse the operand list
1780  else get_oplist(pipe_class->_parameters, pipe_class->_localNames);
1781  skipws();                        // Skip leading whitespace
1782  // Check for block delimiter
1783  if ( (_curchar != '%')
1784       || ( next_char(),  (_curchar != '{')) ) {
1785    parse_err(SYNERR, "missing \"%%{\" in pipe_class definition\n");
1786    return;
1787  }
1788  next_char();
1789
1790  do {
1791    ident = get_ident();           // Grab next identifier
1792    if (ident == NULL) {
1793      parse_err(SYNERR, "keyword identifier expected at \"%c\"\n", _curchar);
1794      continue;
1795    }
1796    skipws();
1797
1798    if (!strcmp(ident, "fixed_latency")) {
1799      skipws();
1800      if (_curchar != '(') {
1801        parse_err(SYNERR, "missing \"(\" in latency definition\n");
1802        return;
1803      }
1804      next_char(); skipws();
1805      if( !isdigit(_curchar) ) {
1806        parse_err(SYNERR, "number expected for \"%c\" in latency definition\n", _curchar);
1807        return;
1808      }
1809      int fixed_latency = get_int();
1810      skipws();
1811      if (_curchar != ')') {
1812        parse_err(SYNERR, "missing \")\" in latency definition\n");
1813        return;
1814      }
1815      next_char(); skipws();
1816      if (_curchar != ';') {
1817        parse_err(SYNERR, "missing \";\" in latency definition\n");
1818        return;
1819      }
1820
1821      pipe_class->setFixedLatency(fixed_latency);
1822      next_char(); skipws();
1823      continue;
1824    }
1825
1826    if (!strcmp(ident, "zero_instructions") ||
1827        !strcmp(ident, "no_instructions")) {
1828      skipws();
1829      if (_curchar != ';') {
1830        parse_err(SYNERR, "missing \";\" in latency definition\n");
1831        return;
1832      }
1833
1834      pipe_class->setInstructionCount(0);
1835      next_char(); skipws();
1836      continue;
1837    }
1838
1839    if (!strcmp(ident, "one_instruction_with_delay_slot") ||
1840        !strcmp(ident, "single_instruction_with_delay_slot")) {
1841      skipws();
1842      if (_curchar != ';') {
1843        parse_err(SYNERR, "missing \";\" in latency definition\n");
1844        return;
1845      }
1846
1847      pipe_class->setInstructionCount(1);
1848      pipe_class->setBranchDelay(true);
1849      next_char(); skipws();
1850      continue;
1851    }
1852
1853    if (!strcmp(ident, "one_instruction") ||
1854        !strcmp(ident, "single_instruction")) {
1855      skipws();
1856      if (_curchar != ';') {
1857        parse_err(SYNERR, "missing \";\" in latency definition\n");
1858        return;
1859      }
1860
1861      pipe_class->setInstructionCount(1);
1862      next_char(); skipws();
1863      continue;
1864    }
1865
1866    if (!strcmp(ident, "instructions_in_first_bundle") ||
1867        !strcmp(ident, "instruction_count")) {
1868      skipws();
1869
1870      int number_of_instructions = 1;
1871
1872      if (_curchar != '(') {
1873        parse_err(SYNERR, "\"(\" expected at \"%c\"\n", _curchar);
1874        continue;
1875      }
1876
1877      next_char(); skipws();
1878      number_of_instructions = get_int();
1879
1880      skipws();
1881      if (_curchar != ')') {
1882        parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
1883        continue;
1884      }
1885
1886      next_char(); skipws();
1887      if (_curchar != ';') {
1888        parse_err(SYNERR, "missing \";\" in latency definition\n");
1889        return;
1890      }
1891
1892      pipe_class->setInstructionCount(number_of_instructions);
1893      next_char(); skipws();
1894      continue;
1895    }
1896
1897    if (!strcmp(ident, "multiple_bundles")) {
1898      skipws();
1899      if (_curchar != ';') {
1900        parse_err(SYNERR, "missing \";\" after multiple bundles\n");
1901        return;
1902      }
1903
1904      pipe_class->setMultipleBundles(true);
1905      next_char(); skipws();
1906      continue;
1907    }
1908
1909    if (!strcmp(ident, "has_delay_slot")) {
1910      skipws();
1911      if (_curchar != ';') {
1912        parse_err(SYNERR, "missing \";\" after \"has_delay_slot\"\n");
1913        return;
1914      }
1915
1916      pipe_class->setBranchDelay(true);
1917      next_char(); skipws();
1918      continue;
1919    }
1920
1921    if (!strcmp(ident, "force_serialization")) {
1922      skipws();
1923      if (_curchar != ';') {
1924        parse_err(SYNERR, "missing \";\" after \"force_serialization\"\n");
1925        return;
1926      }
1927
1928      pipe_class->setForceSerialization(true);
1929      next_char(); skipws();
1930      continue;
1931    }
1932
1933    if (!strcmp(ident, "may_have_no_code")) {
1934      skipws();
1935      if (_curchar != ';') {
1936        parse_err(SYNERR, "missing \";\" after \"may_have_no_code\"\n");
1937        return;
1938      }
1939
1940      pipe_class->setMayHaveNoCode(true);
1941      next_char(); skipws();
1942      continue;
1943    }
1944
1945    const Form *parm = pipe_class->_localNames[ident];
1946    if (parm != NULL) {
1947      oper = parm->is_operand();
1948      if (oper == NULL && !parm->is_opclass()) {
1949        parse_err(SYNERR, "operand name expected at %s\n", ident);
1950        continue;
1951      }
1952
1953      if (_curchar != ':') {
1954        parse_err(SYNERR, "\":\" expected at \"%c\"\n", _curchar);
1955        continue;
1956      }
1957      next_char(); skipws();
1958      stage = get_ident();
1959      if (stage == NULL) {
1960        parse_err(SYNERR, "pipeline stage identifier expected at \"%c\"\n", _curchar);
1961        continue;
1962      }
1963
1964      skipws();
1965      if (_curchar != '(') {
1966        parse_err(SYNERR, "\"(\" expected at \"%c\"\n", _curchar);
1967        continue;
1968      }
1969
1970      next_char();
1971      read_or_write = get_ident();
1972      if (read_or_write == NULL) {
1973        parse_err(SYNERR, "\"read\" or \"write\" expected at \"%c\"\n", _curchar);
1974        continue;
1975      }
1976
1977      is_read  = strcmp(read_or_write, "read")   == 0;
1978      is_write = strcmp(read_or_write, "write")  == 0;
1979      if (!is_read && !is_write) {
1980        parse_err(SYNERR, "\"read\" or \"write\" expected at \"%c\"\n", _curchar);
1981        continue;
1982      }
1983
1984      skipws();
1985      if (_curchar != ')') {
1986        parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
1987        continue;
1988      }
1989
1990      next_char(); skipws();
1991      int more_instrs = 0;
1992      if (_curchar == '+') {
1993          next_char(); skipws();
1994          if (_curchar < '0' || _curchar > '9') {
1995            parse_err(SYNERR, "<number> expected at \"%c\"\n", _curchar);
1996            continue;
1997          }
1998          while (_curchar >= '0' && _curchar <= '9') {
1999            more_instrs *= 10;
2000            more_instrs += _curchar - '0';
2001            next_char();
2002          }
2003          skipws();
2004      }
2005
2006      PipeClassOperandForm *pipe_operand = new PipeClassOperandForm(stage, is_write, more_instrs);
2007      pipe_class->_localUsage.Insert(ident, pipe_operand);
2008
2009      if (_curchar == '%')
2010          continue;
2011
2012      if (_curchar != ';') {
2013        parse_err(SYNERR, "\";\" expected at \"%c\"\n", _curchar);
2014        continue;
2015      }
2016      next_char(); skipws();
2017      continue;
2018    }
2019
2020    // Scan for Resource Specifier
2021    const Form *res = pipeline._resdict[ident];
2022    if (res != NULL) {
2023      int cyclecnt = 1;
2024      if (_curchar != ':') {
2025        parse_err(SYNERR, "\":\" expected at \"%c\"\n", _curchar);
2026        continue;
2027      }
2028      next_char(); skipws();
2029      stage = get_ident();
2030      if (stage == NULL) {
2031        parse_err(SYNERR, "pipeline stage identifier expected at \"%c\"\n", _curchar);
2032        continue;
2033      }
2034
2035      skipws();
2036      if (_curchar == '(') {
2037        next_char();
2038        cyclecnt = get_int();
2039
2040        skipws();
2041        if (_curchar != ')') {
2042          parse_err(SYNERR, "\")\" expected at \"%c\"\n", _curchar);
2043          continue;
2044        }
2045
2046        next_char(); skipws();
2047      }
2048
2049      PipeClassResourceForm *resource = new PipeClassResourceForm(ident, stage, cyclecnt);
2050      int stagenum = pipeline._stages.index(stage);
2051      if (pipeline._maxcycleused < (stagenum+cyclecnt))
2052        pipeline._maxcycleused = (stagenum+cyclecnt);
2053      pipe_class->_resUsage.addForm(resource);
2054
2055      if (_curchar == '%')
2056          continue;
2057
2058      if (_curchar != ';') {
2059        parse_err(SYNERR, "\";\" expected at \"%c\"\n", _curchar);
2060        continue;
2061      }
2062      next_char(); skipws();
2063      continue;
2064    }
2065
2066    parse_err(SYNERR, "resource expected at \"%s\"\n", ident);
2067    return;
2068  } while(_curchar != '%');
2069
2070  next_char();
2071  if (_curchar != '}') {
2072    parse_err(SYNERR, "missing \"%%}\" in pipe_class definition\n");
2073    return;
2074  }
2075
2076  next_char();
2077}
2078
2079//------------------------------peep_parse-------------------------------------
2080void ADLParser::peep_parse(void) {
2081  Peephole  *peep;                // Pointer to current peephole rule form
2082  char      *desc = NULL;         // String representation of rule
2083
2084  skipws();                       // Skip leading whitespace
2085
2086  peep = new Peephole();          // Build new Peephole object
2087  // Check for open block sequence
2088  skipws();                       // Skip leading whitespace
2089  if (_curchar == '%' && *(_ptr+1) == '{') {
2090    next_char(); next_char();     // Skip "%{"
2091    skipws();
2092    while (_curchar != '%' && *(_ptr+1) != '}') {
2093      char *token = get_ident();
2094      if (token == NULL) {
2095        parse_err(SYNERR, "missing identifier inside peephole rule.\n");
2096        return;
2097      }
2098      // check for legal subsections of peephole rule
2099      if (strcmp(token,"peepmatch")==0) {
2100        peep_match_parse(*peep); }
2101      else if (strcmp(token,"peepconstraint")==0) {
2102        peep_constraint_parse(*peep); }
2103      else if (strcmp(token,"peepreplace")==0) {
2104        peep_replace_parse(*peep); }
2105      else {
2106        parse_err(SYNERR, "expected peepmatch, peepconstraint, or peepreplace for identifier %s.\n", token);
2107      }
2108      skipws();
2109    }
2110  }
2111  else {
2112    parse_err(SYNERR, "Missing %%{ ... %%} block after peephole keyword.\n");
2113    return;
2114  }
2115  next_char();                    // Skip past '%'
2116  next_char();                    // Skip past '}'
2117}
2118
2119// ******************** Private Level 2 Parse Functions ********************
2120//------------------------------constraint_parse------------------------------
2121Constraint *ADLParser::constraint_parse(void) {
2122  char *func;
2123  char *arg;
2124
2125  // Check for constraint expression
2126  skipws();
2127  if (_curchar != '(') {
2128    parse_err(SYNERR, "missing constraint expression, (...)\n");
2129    return NULL;
2130  }
2131  next_char();                    // Skip past '('
2132
2133  // Get constraint function
2134  skipws();
2135  func = get_ident();
2136  if (func == NULL) {
2137    parse_err(SYNERR, "missing function in constraint expression.\n");
2138    return NULL;
2139  }
2140  if (strcmp(func,"ALLOC_IN_RC")==0
2141      || strcmp(func,"IS_R_CLASS")==0) {
2142    // Check for '(' before argument
2143    skipws();
2144    if (_curchar != '(') {
2145      parse_err(SYNERR, "missing '(' for constraint function's argument.\n");
2146      return NULL;
2147    }
2148    next_char();
2149
2150    // Get it's argument
2151    skipws();
2152    arg = get_ident();
2153    if (arg == NULL) {
2154      parse_err(SYNERR, "missing argument for constraint function %s\n",func);
2155      return NULL;
2156    }
2157    // Check for ')' after argument
2158    skipws();
2159    if (_curchar != ')') {
2160      parse_err(SYNERR, "missing ')' after constraint function argument %s\n",arg);
2161      return NULL;
2162    }
2163    next_char();
2164  } else {
2165    parse_err(SYNERR, "Invalid constraint function %s\n",func);
2166    return NULL;
2167  }
2168
2169  // Check for closing paren and ';'
2170  skipws();
2171  if (_curchar != ')') {
2172    parse_err(SYNERR, "Missing ')' for constraint function %s\n",func);
2173    return NULL;
2174  }
2175  next_char();
2176  skipws();
2177  if (_curchar != ';') {
2178    parse_err(SYNERR, "Missing ';' after constraint.\n");
2179    return NULL;
2180  }
2181  next_char();
2182
2183  // Create new "Constraint"
2184  Constraint *constraint = new Constraint(func,arg);
2185  return constraint;
2186}
2187
2188//------------------------------constr_parse-----------------------------------
2189ConstructRule *ADLParser::construct_parse(void) {
2190  return NULL;
2191}
2192
2193
2194//------------------------------reg_def_parse----------------------------------
2195void ADLParser::reg_def_parse(void) {
2196  char *rname;                   // Name of register being defined
2197
2198  // Get register name
2199  skipws();                      // Skip whitespace
2200  rname = get_ident();
2201  if (rname == NULL) {
2202    parse_err(SYNERR, "missing register name after reg_def\n");
2203    return;
2204  }
2205
2206  // Check for definition of register calling convention (save on call, ...),
2207  // register save type, and register encoding value.
2208  skipws();
2209  char *callconv  = NULL;
2210  char *c_conv    = NULL;
2211  char *idealtype = NULL;
2212  char *encoding  = NULL;
2213  char *concrete = NULL;
2214  if (_curchar == '(') {
2215    next_char();
2216    callconv = get_ident();
2217    // Parse the internal calling convention, must be NS, SOC, SOE, or AS.
2218    if (callconv == NULL) {
2219      parse_err(SYNERR, "missing register calling convention value\n");
2220      return;
2221    }
2222    if(strcmp(callconv, "SOC") && strcmp(callconv,"SOE") &&
2223       strcmp(callconv, "NS") && strcmp(callconv, "AS")) {
2224      parse_err(SYNERR, "invalid value for register calling convention\n");
2225    }
2226    skipws();
2227    if (_curchar != ',') {
2228      parse_err(SYNERR, "missing comma in register definition statement\n");
2229      return;
2230    }
2231    next_char();
2232
2233    // Parse the native calling convention, must be NS, SOC, SOE, AS
2234    c_conv = get_ident();
2235    if (c_conv == NULL) {
2236      parse_err(SYNERR, "missing register native calling convention value\n");
2237      return;
2238    }
2239    if(strcmp(c_conv, "SOC") && strcmp(c_conv,"SOE") &&
2240       strcmp(c_conv, "NS") && strcmp(c_conv, "AS")) {
2241      parse_err(SYNERR, "invalid value for register calling convention\n");
2242    }
2243    skipws();
2244    if (_curchar != ',') {
2245      parse_err(SYNERR, "missing comma in register definition statement\n");
2246      return;
2247    }
2248    next_char();
2249    skipws();
2250
2251    // Parse the ideal save type
2252    idealtype = get_ident();
2253    if (idealtype == NULL) {
2254      parse_err(SYNERR, "missing register save type value\n");
2255      return;
2256    }
2257    skipws();
2258    if (_curchar != ',') {
2259      parse_err(SYNERR, "missing comma in register definition statement\n");
2260      return;
2261    }
2262    next_char();
2263    skipws();
2264
2265    // Parse the encoding value
2266    encoding = get_expr("encoding", ",");
2267    if (encoding == NULL) {
2268      parse_err(SYNERR, "missing register encoding value\n");
2269      return;
2270    }
2271    trim(encoding);
2272    if (_curchar != ',') {
2273      parse_err(SYNERR, "missing comma in register definition statement\n");
2274      return;
2275    }
2276    next_char();
2277    skipws();
2278    // Parse the concrete name type
2279    // concrete = get_ident();
2280    concrete = get_expr("concrete", ")");
2281    if (concrete == NULL) {
2282      parse_err(SYNERR, "missing vm register name value\n");
2283      return;
2284    }
2285
2286    if (_curchar != ')') {
2287      parse_err(SYNERR, "missing ')' in register definition statement\n");
2288      return;
2289    }
2290    next_char();
2291  }
2292
2293  // Check for closing ';'
2294  skipws();
2295  if (_curchar != ';') {
2296    parse_err(SYNERR, "missing ';' after reg_def\n");
2297    return;
2298  }
2299  next_char();                   // move past ';'
2300
2301  // Debug Stuff
2302  if (_AD._adl_debug > 1) {
2303    fprintf(stderr,"Register Definition: %s ( %s, %s %s )\n", rname,
2304            (callconv ? callconv : ""), (c_conv ? c_conv : ""), concrete);
2305  }
2306
2307  // Record new register definition.
2308  _AD._register->addRegDef(rname, callconv, c_conv, idealtype, encoding, concrete);
2309  return;
2310}
2311
2312//------------------------------reg_class_parse--------------------------------
2313void ADLParser::reg_class_parse(void) {
2314  char *cname;                    // Name of register class being defined
2315
2316  // Get register class name
2317  skipws();                       // Skip leading whitespace
2318  cname = get_ident();
2319  if (cname == NULL) {
2320    parse_err(SYNERR, "missing register class name after 'reg_class'\n");
2321    return;
2322  }
2323  // Debug Stuff
2324  if (_AD._adl_debug >1) fprintf(stderr,"Register Class: %s\n", cname);
2325
2326  RegClass *reg_class = _AD._register->addRegClass(cname);
2327
2328  // Collect registers in class
2329  skipws();
2330  if (_curchar == '(') {
2331    next_char();                  // Skip '('
2332    skipws();
2333    while (_curchar != ')') {
2334      char *rname = get_ident();
2335      if (rname==NULL) {
2336        parse_err(SYNERR, "missing identifier inside reg_class list.\n");
2337        return;
2338      }
2339      RegDef *regDef = _AD._register->getRegDef(rname);
2340      if (!regDef) {
2341        parse_err(SEMERR, "unknown identifier %s inside reg_class list.\n", rname);
2342      } else {
2343        reg_class->addReg(regDef); // add regDef to regClass
2344      }
2345
2346      // Check for ',' and position to next token.
2347      skipws();
2348      if (_curchar == ',') {
2349        next_char();              // Skip trailing ','
2350        skipws();
2351      }
2352    }
2353    next_char();                  // Skip closing ')'
2354  } else if (_curchar == '%') {
2355    char *code = find_cpp_block("reg class");
2356    if (code == NULL) {
2357      parse_err(SYNERR, "missing code declaration for reg class.\n");
2358      return;
2359    }
2360    reg_class->_user_defined = code;
2361    return;
2362  }
2363
2364  // Check for terminating ';'
2365  skipws();
2366  if (_curchar != ';') {
2367    parse_err(SYNERR, "missing ';' at end of reg_class definition.\n");
2368    return;
2369  }
2370  next_char();                    // Skip trailing ';'
2371
2372  // Check RegClass size, must be <= 32 registers in class.
2373
2374  return;
2375}
2376
2377//------------------------------alloc_class_parse------------------------------
2378void ADLParser::alloc_class_parse(void) {
2379  char *name;                     // Name of allocation class being defined
2380
2381  // Get allocation class name
2382  skipws();                       // Skip leading whitespace
2383  name = get_ident();
2384  if (name == NULL) {
2385    parse_err(SYNERR, "missing allocation class name after 'reg_class'\n");
2386    return;
2387  }
2388  // Debug Stuff
2389  if (_AD._adl_debug >1) fprintf(stderr,"Allocation Class: %s\n", name);
2390
2391  AllocClass *alloc_class = _AD._register->addAllocClass(name);
2392
2393  // Collect registers in class
2394  skipws();
2395  if (_curchar == '(') {
2396    next_char();                  // Skip '('
2397    skipws();
2398    while (_curchar != ')') {
2399      char *rname = get_ident();
2400      if (rname==NULL) {
2401        parse_err(SYNERR, "missing identifier inside reg_class list.\n");
2402        return;
2403      }
2404      // Check if name is a RegDef
2405      RegDef *regDef = _AD._register->getRegDef(rname);
2406      if (regDef) {
2407        alloc_class->addReg(regDef);   // add regDef to allocClass
2408      } else {
2409
2410        // name must be a RegDef or a RegClass
2411        parse_err(SYNERR, "name %s should be a previously defined reg_def.\n", rname);
2412        return;
2413      }
2414
2415      // Check for ',' and position to next token.
2416      skipws();
2417      if (_curchar == ',') {
2418        next_char();              // Skip trailing ','
2419        skipws();
2420      }
2421    }
2422    next_char();                  // Skip closing ')'
2423  }
2424
2425  // Check for terminating ';'
2426  skipws();
2427  if (_curchar != ';') {
2428    parse_err(SYNERR, "missing ';' at end of reg_class definition.\n");
2429    return;
2430  }
2431  next_char();                    // Skip trailing ';'
2432
2433  return;
2434}
2435
2436//------------------------------peep_match_child_parse-------------------------
2437InstructForm *ADLParser::peep_match_child_parse(PeepMatch &match, int parent, int &position, int input){
2438  char      *token  = NULL;
2439  int        lparen = 0;          // keep track of parenthesis nesting depth
2440  int        rparen = 0;          // position of instruction at this depth
2441  InstructForm *inst_seen  = NULL;
2442
2443  // Walk the match tree,
2444  // Record <parent, position, instruction name, input position>
2445  while ( lparen >= rparen ) {
2446    skipws();
2447    // Left paren signals start of an input, collect with recursive call
2448    if (_curchar == '(') {
2449      ++lparen;
2450      next_char();
2451      ( void ) peep_match_child_parse(match, parent, position, rparen);
2452    }
2453    // Right paren signals end of an input, may be more
2454    else if (_curchar == ')') {
2455      ++rparen;
2456      if( rparen == lparen ) { // IF rparen matches an lparen I've seen
2457        next_char();           //    move past ')'
2458      } else {                 // ELSE leave ')' for parent
2459        assert( rparen == lparen + 1, "Should only see one extra ')'");
2460        // if an instruction was not specified for this paren-pair
2461        if( ! inst_seen ) {   // record signal entry
2462          match.add_instruction( parent, position, NameList::_signal, input );
2463          ++position;
2464        }
2465        // ++input;   // TEMPORARY
2466        return inst_seen;
2467      }
2468    }
2469    // if no parens, then check for instruction name
2470    // This instruction is the parent of a sub-tree
2471    else if ((token = get_ident_dup()) != NULL) {
2472      const Form *form = _AD._globalNames[token];
2473      if (form) {
2474        InstructForm *inst = form->is_instruction();
2475        // Record the first instruction at this level
2476        if( inst_seen == NULL ) {
2477          inst_seen = inst;
2478        }
2479        if (inst) {
2480          match.add_instruction( parent, position, token, input );
2481          parent = position;
2482          ++position;
2483        } else {
2484          parse_err(SYNERR, "instruction name expected at identifier %s.\n",
2485                    token);
2486          return inst_seen;
2487        }
2488      }
2489      else {
2490        parse_err(SYNERR, "missing identifier in peepmatch rule.\n");
2491        return NULL;
2492      }
2493    }
2494    else {
2495      parse_err(SYNERR, "missing identifier in peepmatch rule.\n");
2496      return NULL;
2497    }
2498
2499  } // end while
2500
2501  assert( false, "ShouldNotReachHere();");
2502  return NULL;
2503}
2504
2505//------------------------------peep_match_parse-------------------------------
2506// Syntax for a peepmatch rule
2507//
2508// peepmatch ( root_instr_name [(instruction subtree)] [,(instruction subtree)]* );
2509//
2510void ADLParser::peep_match_parse(Peephole &peep) {
2511
2512  skipws();
2513  // Check the structure of the rule
2514  // Check for open paren
2515  if (_curchar != '(') {
2516    parse_err(SYNERR, "missing '(' at start of peepmatch rule.\n");
2517    return;
2518  }
2519  next_char();   // skip '('
2520
2521  // Construct PeepMatch and parse the peepmatch rule.
2522  PeepMatch *match = new PeepMatch(_ptr);
2523  int  parent   = -1;                   // parent of root
2524  int  position = 0;                    // zero-based positions
2525  int  input    = 0;                    // input position in parent's operands
2526  InstructForm *root= peep_match_child_parse( *match, parent, position, input);
2527  if( root == NULL ) {
2528    parse_err(SYNERR, "missing instruction-name at start of peepmatch.\n");
2529    return;
2530  }
2531
2532  if( _curchar != ')' ) {
2533    parse_err(SYNERR, "missing ')' at end of peepmatch.\n");
2534    return;
2535  }
2536  next_char();   // skip ')'
2537
2538  // Check for closing semicolon
2539  skipws();
2540  if( _curchar != ';' ) {
2541    parse_err(SYNERR, "missing ';' at end of peepmatch.\n");
2542    return;
2543  }
2544  next_char();   // skip ';'
2545
2546  // Store match into peep, and store peep into instruction
2547  peep.add_match(match);
2548  root->append_peephole(&peep);
2549}
2550
2551//------------------------------peep_constraint_parse--------------------------
2552// Syntax for a peepconstraint rule
2553// A parenthesized list of relations between operands in peepmatch subtree
2554//
2555// peepconstraint %{
2556// (instruction_number.operand_name
2557//     relational_op
2558//  instruction_number.operand_name OR register_name
2559//  [, ...] );
2560//
2561// // instruction numbers are zero-based using topological order in peepmatch
2562//
2563void ADLParser::peep_constraint_parse(Peephole &peep) {
2564
2565  skipws();
2566  // Check the structure of the rule
2567  // Check for open paren
2568  if (_curchar != '(') {
2569    parse_err(SYNERR, "missing '(' at start of peepconstraint rule.\n");
2570    return;
2571  }
2572  else {
2573    next_char();                  // Skip '('
2574  }
2575
2576  // Check for a constraint
2577  skipws();
2578  while( _curchar != ')' ) {
2579    // Get information on the left instruction and its operand
2580    // left-instructions's number
2581    int left_inst = get_int();
2582    // Left-instruction's operand
2583    skipws();
2584    if( _curchar != '.' ) {
2585      parse_err(SYNERR, "missing '.' in peepconstraint after instruction number.\n");
2586      return;
2587    }
2588    next_char();                  // Skip '.'
2589    char *left_op = get_ident_dup();
2590
2591    skipws();
2592    // Collect relational operator
2593    char *relation = get_relation_dup();
2594
2595    skipws();
2596    // Get information on the right instruction and its operand
2597    int right_inst;        // Right-instructions's number
2598    if( isdigit(_curchar) ) {
2599      right_inst = get_int();
2600      // Right-instruction's operand
2601      skipws();
2602      if( _curchar != '.' ) {
2603        parse_err(SYNERR, "missing '.' in peepconstraint after instruction number.\n");
2604        return;
2605      }
2606      next_char();              // Skip '.'
2607    } else {
2608      right_inst = -1;          // Flag as being a register constraint
2609    }
2610
2611    char *right_op = get_ident_dup();
2612
2613    // Construct the next PeepConstraint
2614    PeepConstraint *constraint = new PeepConstraint( left_inst, left_op,
2615                                                     relation,
2616                                                     right_inst, right_op );
2617    // And append it to the list for this peephole rule
2618    peep.append_constraint( constraint );
2619
2620    // Check for another constraint, or end of rule
2621    skipws();
2622    if( _curchar == ',' ) {
2623      next_char();                // Skip ','
2624      skipws();
2625    }
2626    else if( _curchar != ')' ) {
2627      parse_err(SYNERR, "expected ',' or ')' after peephole constraint.\n");
2628      return;
2629    }
2630  } // end while( processing constraints )
2631  next_char();                    // Skip ')'
2632
2633  // Check for terminating ';'
2634  skipws();
2635  if (_curchar != ';') {
2636    parse_err(SYNERR, "missing ';' at end of peepconstraint.\n");
2637    return;
2638  }
2639  next_char();                    // Skip trailing ';'
2640}
2641
2642
2643//------------------------------peep_replace_parse-----------------------------
2644// Syntax for a peepreplace rule
2645// root instruction name followed by a
2646// parenthesized list of whitespace separated instruction.operand specifiers
2647//
2648// peepreplace ( instr_name  ( [instruction_number.operand_name]* ) );
2649//
2650//
2651void ADLParser::peep_replace_parse(Peephole &peep) {
2652  int          lparen = 0;        // keep track of parenthesis nesting depth
2653  int          rparen = 0;        // keep track of parenthesis nesting depth
2654  int          icount = 0;        // count of instructions in rule for naming
2655  char        *str    = NULL;
2656  char        *token  = NULL;
2657
2658  skipws();
2659  // Check for open paren
2660  if (_curchar != '(') {
2661    parse_err(SYNERR, "missing '(' at start of peepreplace rule.\n");
2662    return;
2663  }
2664  else {
2665    lparen++;
2666    next_char();
2667  }
2668
2669  // Check for root instruction
2670  char       *inst = get_ident_dup();
2671  const Form *form = _AD._globalNames[inst];
2672  if( form == NULL || form->is_instruction() == NULL ) {
2673    parse_err(SYNERR, "Instruction name expected at start of peepreplace.\n");
2674    return;
2675  }
2676
2677  // Store string representation of rule into replace
2678  PeepReplace *replace = new PeepReplace(str);
2679  replace->add_instruction( inst );
2680
2681  skipws();
2682  // Start of root's operand-list
2683  if (_curchar != '(') {
2684    parse_err(SYNERR, "missing '(' at peepreplace root's operand-list.\n");
2685    return;
2686  }
2687  else {
2688    lparen++;
2689    next_char();
2690  }
2691
2692  skipws();
2693  // Get the list of operands
2694  while( _curchar != ')' ) {
2695    // Get information on an instruction and its operand
2696    // instructions's number
2697    int   inst_num = get_int();
2698    // Left-instruction's operand
2699    skipws();
2700    if( _curchar != '.' ) {
2701      parse_err(SYNERR, "missing '.' in peepreplace after instruction number.\n");
2702      return;
2703    }
2704    next_char();                  // Skip '.'
2705    char *inst_op = get_ident_dup();
2706    if( inst_op == NULL ) {
2707      parse_err(SYNERR, "missing operand identifier in peepreplace.\n");
2708      return;
2709    }
2710
2711    // Record this operand's position in peepmatch
2712    replace->add_operand( inst_num, inst_op );
2713    skipws();
2714  }
2715
2716  // Check for the end of operands list
2717  skipws();
2718  assert( _curchar == ')', "While loop should have advanced to ')'.");
2719  next_char();  // Skip ')'
2720
2721  skipws();
2722  // Check for end of peepreplace
2723  if( _curchar != ')' ) {
2724    parse_err(SYNERR, "missing ')' at end of peepmatch.\n");
2725    parse_err(SYNERR, "Support one replacement instruction.\n");
2726    return;
2727  }
2728  next_char(); // Skip ')'
2729
2730  // Check for closing semicolon
2731  skipws();
2732  if( _curchar != ';' ) {
2733    parse_err(SYNERR, "missing ';' at end of peepreplace.\n");
2734    return;
2735  }
2736  next_char();   // skip ';'
2737
2738  // Store replace into peep
2739  peep.add_replace( replace );
2740}
2741
2742//------------------------------pred_parse-------------------------------------
2743Predicate *ADLParser::pred_parse(void) {
2744  Predicate *predicate;           // Predicate class for operand
2745  char      *rule = NULL;         // String representation of predicate
2746
2747  skipws();                       // Skip leading whitespace
2748  int line = linenum();
2749  if ( (rule = get_paren_expr("pred expression", true)) == NULL ) {
2750    parse_err(SYNERR, "incorrect or missing expression for 'predicate'\n");
2751    return NULL;
2752  }
2753  // Debug Stuff
2754  if (_AD._adl_debug > 1) fprintf(stderr,"Predicate: %s\n", rule);
2755  if (_curchar != ';') {
2756    parse_err(SYNERR, "missing ';' in predicate definition\n");
2757    return NULL;
2758  }
2759  next_char();                     // Point after the terminator
2760
2761  predicate = new Predicate(rule); // Build new predicate object
2762  skipws();
2763  return predicate;
2764}
2765
2766
2767//------------------------------ins_encode_parse_block-------------------------
2768// Parse the block form of ins_encode.  See ins_encode_parse for more details
2769void ADLParser::ins_encode_parse_block(InstructForm& inst) {
2770  // Create a new encoding name based on the name of the instruction
2771  // definition, which should be unique.
2772  const char* prefix = "__ins_encode_";
2773  char* ec_name = (char*) malloc(strlen(inst._ident) + strlen(prefix) + 1);
2774  sprintf(ec_name, "%s%s", prefix, inst._ident);
2775
2776  assert(_AD._encode->encClass(ec_name) == NULL, "shouldn't already exist");
2777  EncClass* encoding = _AD._encode->add_EncClass(ec_name);
2778  encoding->_linenum = linenum();
2779
2780  // synthesize the arguments list for the enc_class from the
2781  // arguments to the instruct definition.
2782  const char* param = NULL;
2783  inst._parameters.reset();
2784  while ((param = inst._parameters.iter()) != NULL) {
2785    OperandForm* opForm = (OperandForm*) inst._localNames[param];
2786    encoding->add_parameter(opForm->_ident, param);
2787  }
2788
2789  if (!inst._is_postalloc_expand) {
2790    // Define a MacroAssembler instance for use by the encoding.  The
2791    // name is chosen to match the __ idiom used for assembly in other
2792    // parts of hotspot and assumes the existence of the standard
2793    // #define __ _masm.
2794    encoding->add_code("    MacroAssembler _masm(&cbuf);\n");
2795  }
2796
2797  // Parse the following %{ }% block
2798  ins_encode_parse_block_impl(inst, encoding, ec_name);
2799
2800  // Build an encoding rule which invokes the encoding rule we just
2801  // created, passing all arguments that we received.
2802  InsEncode*   encrule = new InsEncode(); // Encode class for instruction
2803  NameAndList* params  = encrule->add_encode(ec_name);
2804  inst._parameters.reset();
2805  while ((param = inst._parameters.iter()) != NULL) {
2806    params->add_entry(param);
2807  }
2808
2809  // Check for duplicate ins_encode sections after parsing the block
2810  // so that parsing can continue and find any other errors.
2811  if (inst._insencode != NULL) {
2812    parse_err(SYNERR, "Multiple ins_encode sections defined\n");
2813    return;
2814  }
2815
2816  // Set encode class of this instruction.
2817  inst._insencode = encrule;
2818}
2819
2820
2821void ADLParser::ins_encode_parse_block_impl(InstructForm& inst, EncClass* encoding, char* ec_name) {
2822  skipws_no_preproc();              // Skip leading whitespace
2823  // Prepend location descriptor, for debugging; cf. ADLParser::find_cpp_block
2824  if (_AD._adlocation_debug) {
2825    encoding->add_code(get_line_string());
2826  }
2827
2828  // Collect the parts of the encode description
2829  // (1) strings that are passed through to output
2830  // (2) replacement/substitution variable, preceeded by a '$'
2831  while ((_curchar != '%') && (*(_ptr+1) != '}')) {
2832
2833    // (1)
2834    // Check if there is a string to pass through to output
2835    char *start = _ptr;       // Record start of the next string
2836    while ((_curchar != '$') && ((_curchar != '%') || (*(_ptr+1) != '}')) ) {
2837      // If at the start of a comment, skip past it
2838      if( (_curchar == '/') && ((*(_ptr+1) == '/') || (*(_ptr+1) == '*')) ) {
2839        skipws_no_preproc();
2840      } else {
2841        // ELSE advance to the next character, or start of the next line
2842        next_char_or_line();
2843      }
2844    }
2845    // If a string was found, terminate it and record in EncClass
2846    if (start != _ptr) {
2847      *_ptr = '\0';          // Terminate the string
2848      encoding->add_code(start);
2849    }
2850
2851    // (2)
2852    // If we are at a replacement variable,
2853    // copy it and record in EncClass
2854    if (_curchar == '$') {
2855      // Found replacement Variable
2856      char* rep_var = get_rep_var_ident_dup();
2857
2858      // Add flag to _strings list indicating we should check _rep_vars
2859      encoding->add_rep_var(rep_var);
2860
2861      skipws();
2862
2863      // Check if this instruct is a MachConstantNode.
2864      if (strcmp(rep_var, "constanttablebase") == 0) {
2865        // This instruct is a MachConstantNode.
2866        inst.set_needs_constant_base(true);
2867        if (strncmp("MachCall", inst.mach_base_class(_globalNames), strlen("MachCall")) != 0 ) {
2868          inst.set_is_mach_constant(true);
2869        }
2870
2871        if (_curchar == '(')  {
2872          parse_err(SYNERR, "constanttablebase in instruct %s cannot have an argument "
2873                            "(only constantaddress and constantoffset)", ec_name);
2874          return;
2875        }
2876      }
2877      else if ((strcmp(rep_var, "constantaddress")   == 0) ||
2878               (strcmp(rep_var, "constantoffset")    == 0)) {
2879        // This instruct is a MachConstantNode.
2880        inst.set_is_mach_constant(true);
2881
2882        // If the constant keyword has an argument, parse it.
2883        if (_curchar == '(')  constant_parse(inst);
2884      }
2885    }
2886  } // end while part of format description
2887  next_char();                      // Skip '%'
2888  next_char();                      // Skip '}'
2889
2890  skipws();
2891
2892  if (_AD._adlocation_debug) {
2893    encoding->add_code(end_line_marker());
2894  }
2895
2896  // Debug Stuff
2897  if (_AD._adl_debug > 1)  fprintf(stderr, "EncodingClass Form: %s\n", ec_name);
2898}
2899
2900
2901//------------------------------ins_encode_parse-------------------------------
2902// Encode rules have the form
2903//   ins_encode( encode_class_name(parameter_list), ... );
2904//
2905// The "encode_class_name" must be defined in the encode section
2906// The parameter list contains $names that are locals.
2907//
2908// Alternatively it can be written like this:
2909//
2910//   ins_encode %{
2911//      ... // body
2912//   %}
2913//
2914// which synthesizes a new encoding class taking the same arguments as
2915// the InstructForm, and automatically prefixes the definition with:
2916//
2917//    MacroAssembler masm(&cbuf);\n");
2918//
2919//  making it more compact to take advantage of the MacroAssembler and
2920//  placing the assembly closer to it's use by instructions.
2921void ADLParser::ins_encode_parse(InstructForm& inst) {
2922
2923  // Parse encode class name
2924  skipws();                        // Skip whitespace
2925  if (_curchar != '(') {
2926    // Check for ins_encode %{ form
2927    if ((_curchar == '%') && (*(_ptr+1) == '{')) {
2928      next_char();                      // Skip '%'
2929      next_char();                      // Skip '{'
2930
2931      // Parse the block form of ins_encode
2932      ins_encode_parse_block(inst);
2933      return;
2934    }
2935
2936    parse_err(SYNERR, "missing '%%{' or '(' in ins_encode definition\n");
2937    return;
2938  }
2939  next_char();                     // move past '('
2940  skipws();
2941
2942  InsEncode *encrule  = new InsEncode(); // Encode class for instruction
2943  encrule->_linenum = linenum();
2944  char      *ec_name  = NULL;      // String representation of encode rule
2945  // identifier is optional.
2946  while (_curchar != ')') {
2947    ec_name = get_ident();
2948    if (ec_name == NULL) {
2949      parse_err(SYNERR, "Invalid encode class name after 'ins_encode('.\n");
2950      return;
2951    }
2952    // Check that encoding is defined in the encode section
2953    EncClass *encode_class = _AD._encode->encClass(ec_name);
2954    if (encode_class == NULL) {
2955      // Like to defer checking these till later...
2956      // parse_err(WARN, "Using an undefined encode class '%s' in 'ins_encode'.\n", ec_name);
2957    }
2958
2959    // Get list for encode method's parameters
2960    NameAndList *params = encrule->add_encode(ec_name);
2961
2962    // Parse the parameters to this encode method.
2963    skipws();
2964    if ( _curchar == '(' ) {
2965      next_char();                 // move past '(' for parameters
2966
2967      // Parse the encode method's parameters
2968      while (_curchar != ')') {
2969        char *param = get_ident_or_literal_constant("encoding operand");
2970        if ( param != NULL ) {
2971
2972          // Check if this instruct is a MachConstantNode.
2973          if (strcmp(param, "constanttablebase") == 0) {
2974            // This instruct is a MachConstantNode.
2975            inst.set_needs_constant_base(true);
2976            if (strncmp("MachCall", inst.mach_base_class(_globalNames), strlen("MachCall")) != 0 ) {
2977              inst.set_is_mach_constant(true);
2978            }
2979
2980            if (_curchar == '(')  {
2981              parse_err(SYNERR, "constanttablebase in instruct %s cannot have an argument "
2982                        "(only constantaddress and constantoffset)", ec_name);
2983              return;
2984            }
2985          } else {
2986            // Found a parameter:
2987            // Check it is a local name, add it to the list, then check for more
2988            // New: allow hex constants as parameters to an encode method.
2989            // New: allow parenthesized expressions as parameters.
2990            // New: allow "primary", "secondary", "tertiary" as parameters.
2991            // New: allow user-defined register name as parameter
2992            if ( (inst._localNames[param] == NULL) &&
2993                 !ADLParser::is_literal_constant(param) &&
2994                 (Opcode::as_opcode_type(param) == Opcode::NOT_AN_OPCODE) &&
2995                 ((_AD._register == NULL ) || (_AD._register->getRegDef(param) == NULL)) ) {
2996              parse_err(SYNERR, "Using non-locally defined parameter %s for encoding %s.\n", param, ec_name);
2997              return;
2998            }
2999          }
3000          params->add_entry(param);
3001
3002          skipws();
3003          if (_curchar == ',' ) {
3004            // More parameters to come
3005            next_char();           // move past ',' between parameters
3006            skipws();              // Skip to next parameter
3007          }
3008          else if (_curchar == ')') {
3009            // Done with parameter list
3010          }
3011          else {
3012            // Only ',' or ')' are valid after a parameter name
3013            parse_err(SYNERR, "expected ',' or ')' after parameter %s.\n",
3014                      ec_name);
3015            return;
3016          }
3017
3018        } else {
3019          skipws();
3020          // Did not find a parameter
3021          if (_curchar == ',') {
3022            parse_err(SYNERR, "Expected encode parameter before ',' in encoding %s.\n", ec_name);
3023            return;
3024          }
3025          if (_curchar != ')') {
3026            parse_err(SYNERR, "Expected ')' after encode parameters.\n");
3027            return;
3028          }
3029        }
3030      } // WHILE loop collecting parameters
3031      next_char();                   // move past ')' at end of parameters
3032    } // done with parameter list for encoding
3033
3034    // Check for ',' or ')' after encoding
3035    skipws();                      // move to character after parameters
3036    if ( _curchar == ',' ) {
3037      // Found a ','
3038      next_char();                 // move past ',' between encode methods
3039      skipws();
3040    }
3041    else if ( _curchar != ')' ) {
3042      // If not a ',' then only a ')' is allowed
3043      parse_err(SYNERR, "Expected ')' after encoding %s.\n", ec_name);
3044      return;
3045    }
3046
3047    // Check for ',' separating parameters
3048    // if ( _curchar != ',' && _curchar != ')' ) {
3049    //   parse_err(SYNERR, "expected ',' or ')' after encode method inside ins_encode.\n");
3050    //   return NULL;
3051    // }
3052
3053  } // done parsing ins_encode methods and their parameters
3054  if (_curchar != ')') {
3055    parse_err(SYNERR, "Missing ')' at end of ins_encode description.\n");
3056    return;
3057  }
3058  next_char();                     // move past ')'
3059  skipws();                        // Skip leading whitespace
3060
3061  if ( _curchar != ';' ) {
3062    parse_err(SYNERR, "Missing ';' at end of ins_encode.\n");
3063    return;
3064  }
3065  next_char();                     // move past ';'
3066  skipws();                        // be friendly to oper_parse()
3067
3068  // Check for duplicate ins_encode sections after parsing the block
3069  // so that parsing can continue and find any other errors.
3070  if (inst._insencode != NULL) {
3071    parse_err(SYNERR, "Multiple ins_encode sections defined\n");
3072    return;
3073  }
3074
3075  // Debug Stuff
3076  if (_AD._adl_debug > 1) fprintf(stderr,"Instruction Encode: %s\n", ec_name);
3077
3078  // Set encode class of this instruction.
3079  inst._insencode = encrule;
3080}
3081
3082//------------------------------postalloc_expand_parse---------------------------
3083// Encode rules have the form
3084//   postalloc_expand( encode_class_name(parameter_list) );
3085//
3086// The "encode_class_name" must be defined in the encode section.
3087// The parameter list contains $names that are locals.
3088//
3089// This is just a copy of ins_encode_parse without the loop.
3090void ADLParser::postalloc_expand_parse(InstructForm& inst) {
3091  inst._is_postalloc_expand = true;
3092
3093  // Parse encode class name.
3094  skipws();                        // Skip whitespace.
3095  if (_curchar != '(') {
3096    // Check for postalloc_expand %{ form
3097    if ((_curchar == '%') && (*(_ptr+1) == '{')) {
3098      next_char();                      // Skip '%'
3099      next_char();                      // Skip '{'
3100
3101      // Parse the block form of postalloc_expand
3102      ins_encode_parse_block(inst);
3103      return;
3104    }
3105
3106    parse_err(SYNERR, "missing '(' in postalloc_expand definition\n");
3107    return;
3108  }
3109  next_char();                     // Move past '('.
3110  skipws();
3111
3112  InsEncode *encrule = new InsEncode(); // Encode class for instruction.
3113  encrule->_linenum = linenum();
3114  char      *ec_name = NULL;       // String representation of encode rule.
3115  // identifier is optional.
3116  if (_curchar != ')') {
3117    ec_name = get_ident();
3118    if (ec_name == NULL) {
3119      parse_err(SYNERR, "Invalid postalloc_expand class name after 'postalloc_expand('.\n");
3120      return;
3121    }
3122    // Check that encoding is defined in the encode section.
3123    EncClass *encode_class = _AD._encode->encClass(ec_name);
3124
3125    // Get list for encode method's parameters
3126    NameAndList *params = encrule->add_encode(ec_name);
3127
3128    // Parse the parameters to this encode method.
3129    skipws();
3130    if (_curchar == '(') {
3131      next_char();                 // Move past '(' for parameters.
3132
3133      // Parse the encode method's parameters.
3134      while (_curchar != ')') {
3135        char *param = get_ident_or_literal_constant("encoding operand");
3136        if (param != NULL) {
3137          // Found a parameter:
3138
3139          // First check for constant table support.
3140
3141          // Check if this instruct is a MachConstantNode.
3142          if (strcmp(param, "constanttablebase") == 0) {
3143            // This instruct is a MachConstantNode.
3144            inst.set_needs_constant_base(true);
3145            if (strncmp("MachCall", inst.mach_base_class(_globalNames), strlen("MachCall")) != 0 ) {
3146              inst.set_is_mach_constant(true);
3147            }
3148
3149            if (_curchar == '(') {
3150              parse_err(SYNERR, "constanttablebase in instruct %s cannot have an argument "
3151                        "(only constantaddress and constantoffset)", ec_name);
3152              return;
3153            }
3154          }
3155          else if ((strcmp(param, "constantaddress") == 0) ||
3156                   (strcmp(param, "constantoffset")  == 0))  {
3157            // This instruct is a MachConstantNode.
3158            inst.set_is_mach_constant(true);
3159
3160            // If the constant keyword has an argument, parse it.
3161            if (_curchar == '(') constant_parse(inst);
3162          }
3163
3164          // Else check it is a local name, add it to the list, then check for more.
3165          // New: allow hex constants as parameters to an encode method.
3166          // New: allow parenthesized expressions as parameters.
3167          // New: allow "primary", "secondary", "tertiary" as parameters.
3168          // New: allow user-defined register name as parameter.
3169          else if ((inst._localNames[param] == NULL) &&
3170                   !ADLParser::is_literal_constant(param) &&
3171                   (Opcode::as_opcode_type(param) == Opcode::NOT_AN_OPCODE) &&
3172                   ((_AD._register == NULL) || (_AD._register->getRegDef(param) == NULL))) {
3173            parse_err(SYNERR, "Using non-locally defined parameter %s for encoding %s.\n", param, ec_name);
3174            return;
3175          }
3176          params->add_entry(param);
3177
3178          skipws();
3179          if (_curchar == ',') {
3180            // More parameters to come.
3181            next_char();           // Move past ',' between parameters.
3182            skipws();              // Skip to next parameter.
3183          } else if (_curchar == ')') {
3184            // Done with parameter list
3185          } else {
3186            // Only ',' or ')' are valid after a parameter name.
3187            parse_err(SYNERR, "expected ',' or ')' after parameter %s.\n", ec_name);
3188            return;
3189          }
3190
3191        } else {
3192          skipws();
3193          // Did not find a parameter.
3194          if (_curchar == ',') {
3195            parse_err(SYNERR, "Expected encode parameter before ',' in postalloc_expand %s.\n", ec_name);
3196            return;
3197          }
3198          if (_curchar != ')') {
3199            parse_err(SYNERR, "Expected ')' after postalloc_expand parameters.\n");
3200            return;
3201          }
3202        }
3203      } // WHILE loop collecting parameters.
3204      next_char();                 // Move past ')' at end of parameters.
3205    } // Done with parameter list for encoding.
3206
3207    // Check for ',' or ')' after encoding.
3208    skipws();                      // Move to character after parameters.
3209    if (_curchar != ')') {
3210      // Only a ')' is allowed.
3211      parse_err(SYNERR, "Expected ')' after postalloc_expand %s.\n", ec_name);
3212      return;
3213    }
3214  } // Done parsing postalloc_expand method and their parameters.
3215  if (_curchar != ')') {
3216    parse_err(SYNERR, "Missing ')' at end of postalloc_expand description.\n");
3217    return;
3218  }
3219  next_char();                     // Move past ')'.
3220  skipws();                        // Skip leading whitespace.
3221
3222  if (_curchar != ';') {
3223    parse_err(SYNERR, "Missing ';' at end of postalloc_expand.\n");
3224    return;
3225  }
3226  next_char();                     // Move past ';'.
3227  skipws();                        // Be friendly to oper_parse().
3228
3229  // Debug Stuff.
3230  if (_AD._adl_debug > 1) fprintf(stderr, "Instruction postalloc_expand: %s\n", ec_name);
3231
3232  // Set encode class of this instruction.
3233  inst._insencode = encrule;
3234}
3235
3236
3237//------------------------------constant_parse---------------------------------
3238// Parse a constant expression.
3239void ADLParser::constant_parse(InstructForm& inst) {
3240  // Create a new encoding name based on the name of the instruction
3241  // definition, which should be unique.
3242  const char* prefix = "__constant_";
3243  char* ec_name = (char*) malloc(strlen(inst._ident) + strlen(prefix) + 1);
3244  sprintf(ec_name, "%s%s", prefix, inst._ident);
3245
3246  assert(_AD._encode->encClass(ec_name) == NULL, "shouldn't already exist");
3247  EncClass* encoding = _AD._encode->add_EncClass(ec_name);
3248  encoding->_linenum = linenum();
3249
3250  // synthesize the arguments list for the enc_class from the
3251  // arguments to the instruct definition.
3252  const char* param = NULL;
3253  inst._parameters.reset();
3254  while ((param = inst._parameters.iter()) != NULL) {
3255    OperandForm* opForm = (OperandForm*) inst._localNames[param];
3256    encoding->add_parameter(opForm->_ident, param);
3257  }
3258
3259  // Parse the following ( ) expression.
3260  constant_parse_expression(encoding, ec_name);
3261
3262  // Build an encoding rule which invokes the encoding rule we just
3263  // created, passing all arguments that we received.
3264  InsEncode*   encrule = new InsEncode(); // Encode class for instruction
3265  NameAndList* params  = encrule->add_encode(ec_name);
3266  inst._parameters.reset();
3267  while ((param = inst._parameters.iter()) != NULL) {
3268    params->add_entry(param);
3269  }
3270
3271  // Set encode class of this instruction.
3272  inst._constant = encrule;
3273}
3274
3275
3276//------------------------------constant_parse_expression----------------------
3277void ADLParser::constant_parse_expression(EncClass* encoding, char* ec_name) {
3278  skipws();
3279
3280  // Prepend location descriptor, for debugging; cf. ADLParser::find_cpp_block
3281  if (_AD._adlocation_debug) {
3282    encoding->add_code(get_line_string());
3283  }
3284
3285  // Start code line.
3286  encoding->add_code("    _constant = C->constant_table().add");
3287
3288  // Parse everything in ( ) expression.
3289  encoding->add_code("(this, ");
3290  next_char();  // Skip '('
3291  int parens_depth = 1;
3292
3293  // Collect the parts of the constant expression.
3294  // (1) strings that are passed through to output
3295  // (2) replacement/substitution variable, preceeded by a '$'
3296  while (parens_depth > 0) {
3297    if (_curchar == '(') {
3298      parens_depth++;
3299      encoding->add_code("(");
3300      next_char();
3301    }
3302    else if (_curchar == ')') {
3303      parens_depth--;
3304      if (parens_depth > 0)
3305        encoding->add_code(")");
3306      next_char();
3307    }
3308    else {
3309      // (1)
3310      // Check if there is a string to pass through to output
3311      char *start = _ptr;  // Record start of the next string
3312      while ((_curchar != '$') && (_curchar != '(') && (_curchar != ')')) {
3313        next_char();
3314      }
3315      // If a string was found, terminate it and record in EncClass
3316      if (start != _ptr) {
3317        *_ptr = '\0';  // Terminate the string
3318        encoding->add_code(start);
3319      }
3320
3321      // (2)
3322      // If we are at a replacement variable, copy it and record in EncClass.
3323      if (_curchar == '$') {
3324        // Found replacement Variable
3325        char* rep_var = get_rep_var_ident_dup();
3326        encoding->add_rep_var(rep_var);
3327      }
3328    }
3329  }
3330
3331  // Finish code line.
3332  encoding->add_code(");");
3333
3334  if (_AD._adlocation_debug) {
3335    encoding->add_code(end_line_marker());
3336  }
3337
3338  // Debug Stuff
3339  if (_AD._adl_debug > 1)  fprintf(stderr, "EncodingClass Form: %s\n", ec_name);
3340}
3341
3342
3343//------------------------------size_parse-----------------------------------
3344// Parse a 'size(<expr>)' attribute which specifies the size of the
3345// emitted instructions in bytes. <expr> can be a C++ expression,
3346// e.g. a constant.
3347char* ADLParser::size_parse(InstructForm *instr) {
3348  char* sizeOfInstr = NULL;
3349
3350  // Get value of the instruction's size
3351  skipws();
3352
3353  // Parse size
3354  sizeOfInstr = get_paren_expr("size expression");
3355  if (sizeOfInstr == NULL) {
3356     parse_err(SYNERR, "size of opcode expected at %c\n", _curchar);
3357     return NULL;
3358  }
3359
3360  skipws();
3361
3362  // Check for terminator
3363  if (_curchar != ';') {
3364    parse_err(SYNERR, "missing ';' in ins_attrib definition\n");
3365    return NULL;
3366  }
3367  next_char();                     // Advance past the ';'
3368  skipws();                        // necessary for instr_parse()
3369
3370  // Debug Stuff
3371  if (_AD._adl_debug > 1) {
3372    if (sizeOfInstr != NULL) {
3373      fprintf(stderr,"size of opcode: %s\n", sizeOfInstr);
3374    }
3375  }
3376
3377  return sizeOfInstr;
3378}
3379
3380
3381//------------------------------opcode_parse-----------------------------------
3382Opcode * ADLParser::opcode_parse(InstructForm *instr) {
3383  char *primary   = NULL;
3384  char *secondary = NULL;
3385  char *tertiary  = NULL;
3386
3387  char   *val    = NULL;
3388  Opcode *opcode = NULL;
3389
3390  // Get value of the instruction's opcode
3391  skipws();
3392  if (_curchar != '(') {         // Check for parenthesized operand list
3393    parse_err(SYNERR, "missing '(' in expand instruction declaration\n");
3394    return NULL;
3395  }
3396  next_char();                   // skip open paren
3397  skipws();
3398  if (_curchar != ')') {
3399    // Parse primary, secondary, and tertiary opcodes, if provided.
3400    if ( ((primary = get_ident_or_literal_constant("primary opcode")) == NULL) ) {
3401        parse_err(SYNERR, "primary hex opcode expected at %c\n", _curchar);
3402        return NULL;
3403    }
3404    skipws();
3405    if (_curchar == ',') {
3406      next_char();
3407      skipws();
3408      // Parse secondary opcode
3409      if ( ((secondary = get_ident_or_literal_constant("secondary opcode")) == NULL) ) {
3410        parse_err(SYNERR, "secondary hex opcode expected at %c\n", _curchar);
3411        return NULL;
3412      }
3413      skipws();
3414      if (_curchar == ',') {
3415        next_char();
3416        skipws();
3417        // Parse tertiary opcode
3418        if ( ((tertiary = get_ident_or_literal_constant("tertiary opcode")) == NULL) ) {
3419          parse_err(SYNERR,"tertiary hex opcode expected at %c\n", _curchar);
3420          return NULL;
3421        }
3422        skipws();
3423      }
3424    }
3425    skipws();
3426    if (_curchar != ')') {
3427      parse_err(SYNERR, "Missing ')' in opcode description\n");
3428      return NULL;
3429    }
3430  }
3431  next_char();                     // Skip ')'
3432  skipws();
3433  // Check for terminator
3434  if (_curchar != ';') {
3435    parse_err(SYNERR, "missing ';' in ins_attrib definition\n");
3436    return NULL;
3437  }
3438  next_char();                     // Advance past the ';'
3439  skipws();                        // necessary for instr_parse()
3440
3441  // Debug Stuff
3442  if (_AD._adl_debug > 1) {
3443    if (primary   != NULL) fprintf(stderr,"primary   opcode: %s\n", primary);
3444    if (secondary != NULL) fprintf(stderr,"secondary opcode: %s\n", secondary);
3445    if (tertiary  != NULL) fprintf(stderr,"tertiary  opcode: %s\n", tertiary);
3446  }
3447
3448  // Generate new object and return
3449  opcode = new Opcode(primary, secondary, tertiary);
3450  return opcode;
3451}
3452
3453
3454//------------------------------interface_parse--------------------------------
3455Interface *ADLParser::interface_parse(void) {
3456  char *iface_name  = NULL;      // Name of interface class being used
3457  char *iface_code  = NULL;      // Describe components of this class
3458
3459  // Get interface class name
3460  skipws();                       // Skip whitespace
3461  if (_curchar != '(') {
3462    parse_err(SYNERR, "Missing '(' at start of interface description.\n");
3463    return NULL;
3464  }
3465  next_char();                    // move past '('
3466  skipws();
3467  iface_name = get_ident();
3468  if (iface_name == NULL) {
3469    parse_err(SYNERR, "missing interface name after 'interface'.\n");
3470    return NULL;
3471  }
3472  skipws();
3473  if (_curchar != ')') {
3474    parse_err(SYNERR, "Missing ')' after name of interface.\n");
3475    return NULL;
3476  }
3477  next_char();                    // move past ')'
3478
3479  // Get details of the interface,
3480  // for the type of interface indicated by iface_name.
3481  Interface *inter = NULL;
3482  skipws();
3483  if ( _curchar != ';' ) {
3484    if ( strcmp(iface_name,"MEMORY_INTER") == 0 ) {
3485      inter = mem_interface_parse();
3486    }
3487    else if ( strcmp(iface_name,"COND_INTER") == 0 ) {
3488      inter = cond_interface_parse();
3489    }
3490    // The parse routines consume the "%}"
3491
3492    // Check for probable extra ';' after defining block.
3493    if ( _curchar == ';' ) {
3494      parse_err(SYNERR, "Extra ';' after defining interface block.\n");
3495      next_char();                // Skip ';'
3496      return NULL;
3497    }
3498  } else {
3499    next_char();                  // move past ';'
3500
3501    // Create appropriate interface object
3502    if ( strcmp(iface_name,"REG_INTER") == 0 ) {
3503      inter = new RegInterface();
3504    }
3505    else if ( strcmp(iface_name,"CONST_INTER") == 0 ) {
3506      inter = new ConstInterface();
3507    }
3508  }
3509  skipws();                       // be friendly to oper_parse()
3510  // Debug Stuff
3511  if (_AD._adl_debug > 1) fprintf(stderr,"Interface Form: %s\n", iface_name);
3512
3513  // Create appropriate interface object and return.
3514  return inter;
3515}
3516
3517
3518//------------------------------mem_interface_parse----------------------------
3519Interface *ADLParser::mem_interface_parse(void) {
3520  // Fields for MemInterface
3521  char *base        = NULL;
3522  char *index       = NULL;
3523  char *scale       = NULL;
3524  char *disp        = NULL;
3525
3526  if (_curchar != '%') {
3527    parse_err(SYNERR, "Missing '%%{' for 'interface' block.\n");
3528    return NULL;
3529  }
3530  next_char();                  // Skip '%'
3531  if (_curchar != '{') {
3532    parse_err(SYNERR, "Missing '%%{' for 'interface' block.\n");
3533    return NULL;
3534  }
3535  next_char();                  // Skip '{'
3536  skipws();
3537  do {
3538    char *field = get_ident();
3539    if (field == NULL) {
3540      parse_err(SYNERR, "Expected keyword, base|index|scale|disp,  or '%%}' ending interface.\n");
3541      return NULL;
3542    }
3543    if ( strcmp(field,"base") == 0 ) {
3544      base  = interface_field_parse();
3545    }
3546    else if ( strcmp(field,"index") == 0 ) {
3547      index = interface_field_parse();
3548    }
3549    else if ( strcmp(field,"scale") == 0 ) {
3550      scale = interface_field_parse();
3551    }
3552    else if ( strcmp(field,"disp") == 0 ) {
3553      disp  = interface_field_parse();
3554    }
3555    else {
3556      parse_err(SYNERR, "Expected keyword, base|index|scale|disp,  or '%%}' ending interface.\n");
3557      return NULL;
3558    }
3559  } while( _curchar != '%' );
3560  next_char();                  // Skip '%'
3561  if ( _curchar != '}' ) {
3562    parse_err(SYNERR, "Missing '%%}' for 'interface' block.\n");
3563    return NULL;
3564  }
3565  next_char();                  // Skip '}'
3566
3567  // Construct desired object and return
3568  Interface *inter = new MemInterface(base, index, scale, disp);
3569  return inter;
3570}
3571
3572
3573//------------------------------cond_interface_parse---------------------------
3574Interface *ADLParser::cond_interface_parse(void) {
3575  char *equal;
3576  char *not_equal;
3577  char *less;
3578  char *greater_equal;
3579  char *less_equal;
3580  char *greater;
3581  char *overflow;
3582  char *no_overflow;
3583  const char *equal_format = "eq";
3584  const char *not_equal_format = "ne";
3585  const char *less_format = "lt";
3586  const char *greater_equal_format = "ge";
3587  const char *less_equal_format = "le";
3588  const char *greater_format = "gt";
3589  const char *overflow_format = "o";
3590  const char *no_overflow_format = "no";
3591
3592  if (_curchar != '%') {
3593    parse_err(SYNERR, "Missing '%%{' for 'cond_interface' block.\n");
3594    return NULL;
3595  }
3596  next_char();                  // Skip '%'
3597  if (_curchar != '{') {
3598    parse_err(SYNERR, "Missing '%%{' for 'cond_interface' block.\n");
3599    return NULL;
3600  }
3601  next_char();                  // Skip '{'
3602  skipws();
3603  do {
3604    char *field = get_ident();
3605    if (field == NULL) {
3606      parse_err(SYNERR, "Expected keyword, base|index|scale|disp,  or '%%}' ending interface.\n");
3607      return NULL;
3608    }
3609    if ( strcmp(field,"equal") == 0 ) {
3610      equal  = interface_field_parse(&equal_format);
3611    }
3612    else if ( strcmp(field,"not_equal") == 0 ) {
3613      not_equal = interface_field_parse(&not_equal_format);
3614    }
3615    else if ( strcmp(field,"less") == 0 ) {
3616      less = interface_field_parse(&less_format);
3617    }
3618    else if ( strcmp(field,"greater_equal") == 0 ) {
3619      greater_equal  = interface_field_parse(&greater_equal_format);
3620    }
3621    else if ( strcmp(field,"less_equal") == 0 ) {
3622      less_equal = interface_field_parse(&less_equal_format);
3623    }
3624    else if ( strcmp(field,"greater") == 0 ) {
3625      greater = interface_field_parse(&greater_format);
3626    }
3627    else if ( strcmp(field,"overflow") == 0 ) {
3628      overflow = interface_field_parse(&overflow_format);
3629    }
3630    else if ( strcmp(field,"no_overflow") == 0 ) {
3631      no_overflow = interface_field_parse(&no_overflow_format);
3632    }
3633    else {
3634      parse_err(SYNERR, "Expected keyword, base|index|scale|disp,  or '%%}' ending interface.\n");
3635      return NULL;
3636    }
3637  } while( _curchar != '%' );
3638  next_char();                  // Skip '%'
3639  if ( _curchar != '}' ) {
3640    parse_err(SYNERR, "Missing '%%}' for 'interface' block.\n");
3641    return NULL;
3642  }
3643  next_char();                  // Skip '}'
3644
3645  // Construct desired object and return
3646  Interface *inter = new CondInterface(equal,         equal_format,
3647                                       not_equal,     not_equal_format,
3648                                       less,          less_format,
3649                                       greater_equal, greater_equal_format,
3650                                       less_equal,    less_equal_format,
3651                                       greater,       greater_format,
3652                                       overflow,      overflow_format,
3653                                       no_overflow,   no_overflow_format);
3654  return inter;
3655}
3656
3657
3658//------------------------------interface_field_parse--------------------------
3659char *ADLParser::interface_field_parse(const char ** format) {
3660  char *iface_field = NULL;
3661
3662  // Get interface field
3663  skipws();                      // Skip whitespace
3664  if (_curchar != '(') {
3665    parse_err(SYNERR, "Missing '(' at start of interface field.\n");
3666    return NULL;
3667  }
3668  next_char();                   // move past '('
3669  skipws();
3670  if ( _curchar != '0' && _curchar != '$' ) {
3671    parse_err(SYNERR, "missing or invalid interface field contents.\n");
3672    return NULL;
3673  }
3674  iface_field = get_rep_var_ident();
3675  if (iface_field == NULL) {
3676    parse_err(SYNERR, "missing or invalid interface field contents.\n");
3677    return NULL;
3678  }
3679  skipws();
3680  if (format != NULL && _curchar == ',') {
3681    next_char();
3682    skipws();
3683    if (_curchar != '"') {
3684      parse_err(SYNERR, "Missing '\"' in field format .\n");
3685      return NULL;
3686    }
3687    next_char();
3688    char *start = _ptr;       // Record start of the next string
3689    while ((_curchar != '"') && (_curchar != '%') && (_curchar != '\n')) {
3690      if (_curchar == '\\')  next_char();  // superquote
3691      if (_curchar == '\n')  parse_err(SYNERR, "newline in string");  // unimplemented!
3692      next_char();
3693    }
3694    if (_curchar != '"') {
3695      parse_err(SYNERR, "Missing '\"' at end of field format .\n");
3696      return NULL;
3697    }
3698    // If a string was found, terminate it and record in FormatRule
3699    if ( start != _ptr ) {
3700      *_ptr  = '\0';          // Terminate the string
3701      *format = start;
3702    }
3703    next_char();
3704    skipws();
3705  }
3706  if (_curchar != ')') {
3707    parse_err(SYNERR, "Missing ')' after interface field.\n");
3708    return NULL;
3709  }
3710  next_char();                   // move past ')'
3711  skipws();
3712  if ( _curchar != ';' ) {
3713    parse_err(SYNERR, "Missing ';' at end of interface field.\n");
3714    return NULL;
3715  }
3716  next_char();                    // move past ';'
3717  skipws();                       // be friendly to interface_parse()
3718
3719  return iface_field;
3720}
3721
3722
3723//------------------------------match_parse------------------------------------
3724MatchRule *ADLParser::match_parse(FormDict &operands) {
3725  MatchRule *match;               // Match Rule class for instruction/operand
3726  char      *cnstr = NULL;        // Code for constructor
3727  int        depth = 0;           // Counter for matching parentheses
3728  int        numleaves = 0;       // Counter for number of leaves in rule
3729
3730  // Parse the match rule tree
3731  MatchNode *mnode = matchNode_parse(operands, depth, numleaves, true);
3732
3733  // Either there is a block with a constructor, or a ';' here
3734  skipws();                       // Skip whitespace
3735  if ( _curchar == ';' ) {        // Semicolon is valid terminator
3736    cnstr = NULL;                 // no constructor for this form
3737    next_char();                  // Move past the ';', replaced with '\0'
3738  }
3739  else if ((cnstr = find_cpp_block("match constructor")) == NULL ) {
3740    parse_err(SYNERR, "invalid construction of match rule\n"
3741              "Missing ';' or invalid '%%{' and '%%}' constructor\n");
3742    return NULL;                  // No MatchRule to return
3743  }
3744  if (_AD._adl_debug > 1)
3745    if (cnstr) fprintf(stderr,"Match Constructor: %s\n", cnstr);
3746  // Build new MatchRule object
3747  match = new MatchRule(_AD, mnode, depth, cnstr, numleaves);
3748  skipws();                       // Skip any trailing whitespace
3749  return match;                   // Return MatchRule object
3750}
3751
3752//------------------------------format_parse-----------------------------------
3753FormatRule* ADLParser::format_parse(void) {
3754  char       *desc   = NULL;
3755  FormatRule *format = (new FormatRule(desc));
3756
3757  // Without expression form, MUST have a code block;
3758  skipws();                       // Skip whitespace
3759  if ( _curchar == ';' ) {        // Semicolon is valid terminator
3760    desc  = NULL;                 // no constructor for this form
3761    next_char();                  // Move past the ';', replaced with '\0'
3762  }
3763  else if ( _curchar == '%' && *(_ptr+1) == '{') {
3764    next_char();                  // Move past the '%'
3765    next_char();                  // Move past the '{'
3766
3767    skipws();
3768    if (_curchar == '$') {
3769      char* ident = get_rep_var_ident();
3770      if (strcmp(ident, "$$template") == 0) return template_parse();
3771      parse_err(SYNERR, "Unknown \"%s\" directive in format", ident);
3772      return NULL;
3773    }
3774    // Check for the opening '"' inside the format description
3775    if ( _curchar == '"' ) {
3776      next_char();              // Move past the initial '"'
3777      if( _curchar == '"' ) {   // Handle empty format string case
3778        *_ptr = '\0';           // Terminate empty string
3779        format->_strings.addName(_ptr);
3780      }
3781
3782      // Collect the parts of the format description
3783      // (1) strings that are passed through to tty->print
3784      // (2) replacement/substitution variable, preceeded by a '$'
3785      // (3) multi-token ANSIY C style strings
3786      while ( true ) {
3787        if ( _curchar == '%' || _curchar == '\n' ) {
3788          if ( _curchar != '"' ) {
3789            parse_err(SYNERR, "missing '\"' at end of format block");
3790            return NULL;
3791          }
3792        }
3793
3794        // (1)
3795        // Check if there is a string to pass through to output
3796        char *start = _ptr;       // Record start of the next string
3797        while ((_curchar != '$') && (_curchar != '"') && (_curchar != '%') && (_curchar != '\n')) {
3798          if (_curchar == '\\') {
3799            next_char();  // superquote
3800            if ((_curchar == '$') || (_curchar == '%'))
3801              // hack to avoid % escapes and warnings about undefined \ escapes
3802              *(_ptr-1) = _curchar;
3803          }
3804          if (_curchar == '\n')  parse_err(SYNERR, "newline in string");  // unimplemented!
3805          next_char();
3806        }
3807        // If a string was found, terminate it and record in FormatRule
3808        if ( start != _ptr ) {
3809          *_ptr  = '\0';          // Terminate the string
3810          format->_strings.addName(start);
3811        }
3812
3813        // (2)
3814        // If we are at a replacement variable,
3815        // copy it and record in FormatRule
3816        if ( _curchar == '$' ) {
3817          next_char();          // Move past the '$'
3818          char* rep_var = get_ident(); // Nil terminate the variable name
3819          rep_var = strdup(rep_var);// Copy the string
3820          *_ptr   = _curchar;     // and replace Nil with original character
3821          format->_rep_vars.addName(rep_var);
3822          // Add flag to _strings list indicating we should check _rep_vars
3823          format->_strings.addName(NameList::_signal);
3824        }
3825
3826        // (3)
3827        // Allow very long strings to be broken up,
3828        // using the ANSI C syntax "foo\n" <newline> "bar"
3829        if ( _curchar == '"') {
3830          next_char();           // Move past the '"'
3831          skipws();              // Skip white space before next string token
3832          if ( _curchar != '"') {
3833            break;
3834          } else {
3835            // Found one.  Skip both " and the whitespace in between.
3836            next_char();
3837          }
3838        }
3839      } // end while part of format description
3840
3841      // Check for closing '"' and '%}' in format description
3842      skipws();                   // Move to closing '%}'
3843      if ( _curchar != '%' ) {
3844        parse_err(SYNERR, "non-blank characters between closing '\"' and '%%' in format");
3845        return NULL;
3846      }
3847    } // Done with format description inside
3848
3849    skipws();
3850    // Past format description, at '%'
3851    if ( _curchar != '%' || *(_ptr+1) != '}' ) {
3852      parse_err(SYNERR, "missing '%%}' at end of format block");
3853      return NULL;
3854    }
3855    next_char();                  // Move past the '%'
3856    next_char();                  // Move past the '}'
3857  }
3858  else {  // parameter list alone must terminate with a ';'
3859    parse_err(SYNERR, "missing ';' after Format expression");
3860    return NULL;
3861  }
3862  // Debug Stuff
3863  if (_AD._adl_debug > 1) fprintf(stderr,"Format Rule: %s\n", desc);
3864
3865  skipws();
3866  return format;
3867}
3868
3869
3870//------------------------------template_parse-----------------------------------
3871FormatRule* ADLParser::template_parse(void) {
3872  char       *desc   = NULL;
3873  FormatRule *format = (new FormatRule(desc));
3874
3875  skipws();
3876  while ( (_curchar != '%') && (*(_ptr+1) != '}') ) {
3877
3878    // (1)
3879    // Check if there is a string to pass through to output
3880    {
3881      char *start = _ptr;       // Record start of the next string
3882      while ((_curchar != '$') && ((_curchar != '%') || (*(_ptr+1) != '}')) ) {
3883        // If at the start of a comment, skip past it
3884        if( (_curchar == '/') && ((*(_ptr+1) == '/') || (*(_ptr+1) == '*')) ) {
3885          skipws_no_preproc();
3886        } else {
3887          // ELSE advance to the next character, or start of the next line
3888          next_char_or_line();
3889        }
3890      }
3891      // If a string was found, terminate it and record in EncClass
3892      if ( start != _ptr ) {
3893        *_ptr  = '\0';          // Terminate the string
3894        // Add flag to _strings list indicating we should check _rep_vars
3895        format->_strings.addName(NameList::_signal2);
3896        format->_strings.addName(start);
3897      }
3898    }
3899
3900    // (2)
3901    // If we are at a replacement variable,
3902    // copy it and record in EncClass
3903    if ( _curchar == '$' ) {
3904      // Found replacement Variable
3905      char *rep_var = get_rep_var_ident_dup();
3906      if (strcmp(rep_var, "$emit") == 0) {
3907        // switch to normal format parsing
3908        next_char();
3909        next_char();
3910        skipws();
3911        // Check for the opening '"' inside the format description
3912        if ( _curchar == '"' ) {
3913          next_char();              // Move past the initial '"'
3914          if( _curchar == '"' ) {   // Handle empty format string case
3915            *_ptr = '\0';           // Terminate empty string
3916            format->_strings.addName(_ptr);
3917          }
3918
3919          // Collect the parts of the format description
3920          // (1) strings that are passed through to tty->print
3921          // (2) replacement/substitution variable, preceeded by a '$'
3922          // (3) multi-token ANSIY C style strings
3923          while ( true ) {
3924            if ( _curchar == '%' || _curchar == '\n' ) {
3925              parse_err(SYNERR, "missing '\"' at end of format block");
3926              return NULL;
3927            }
3928
3929            // (1)
3930            // Check if there is a string to pass through to output
3931            char *start = _ptr;       // Record start of the next string
3932            while ((_curchar != '$') && (_curchar != '"') && (_curchar != '%') && (_curchar != '\n')) {
3933              if (_curchar == '\\')  next_char();  // superquote
3934              if (_curchar == '\n')  parse_err(SYNERR, "newline in string");  // unimplemented!
3935              next_char();
3936            }
3937            // If a string was found, terminate it and record in FormatRule
3938            if ( start != _ptr ) {
3939              *_ptr  = '\0';          // Terminate the string
3940              format->_strings.addName(start);
3941            }
3942
3943            // (2)
3944            // If we are at a replacement variable,
3945            // copy it and record in FormatRule
3946            if ( _curchar == '$' ) {
3947              next_char();          // Move past the '$'
3948              char* next_rep_var = get_ident(); // Nil terminate the variable name
3949              next_rep_var = strdup(next_rep_var);// Copy the string
3950              *_ptr   = _curchar;     // and replace Nil with original character
3951              format->_rep_vars.addName(next_rep_var);
3952              // Add flag to _strings list indicating we should check _rep_vars
3953              format->_strings.addName(NameList::_signal);
3954            }
3955
3956            // (3)
3957            // Allow very long strings to be broken up,
3958            // using the ANSI C syntax "foo\n" <newline> "bar"
3959            if ( _curchar == '"') {
3960              next_char();           // Move past the '"'
3961              skipws();              // Skip white space before next string token
3962              if ( _curchar != '"') {
3963                break;
3964              } else {
3965                // Found one.  Skip both " and the whitespace in between.
3966                next_char();
3967              }
3968            }
3969          } // end while part of format description
3970        }
3971      } else {
3972        // Add flag to _strings list indicating we should check _rep_vars
3973        format->_rep_vars.addName(rep_var);
3974        // Add flag to _strings list indicating we should check _rep_vars
3975        format->_strings.addName(NameList::_signal3);
3976      }
3977    } // end while part of format description
3978  }
3979
3980  skipws();
3981  // Past format description, at '%'
3982  if ( _curchar != '%' || *(_ptr+1) != '}' ) {
3983    parse_err(SYNERR, "missing '%%}' at end of format block");
3984    return NULL;
3985  }
3986  next_char();                  // Move past the '%'
3987  next_char();                  // Move past the '}'
3988
3989  // Debug Stuff
3990  if (_AD._adl_debug > 1) fprintf(stderr,"Format Rule: %s\n", desc);
3991
3992  skipws();
3993  return format;
3994}
3995
3996
3997//------------------------------effect_parse-----------------------------------
3998void ADLParser::effect_parse(InstructForm *instr) {
3999  char* desc   = NULL;
4000
4001  skipws();                      // Skip whitespace
4002  if (_curchar != '(') {
4003    parse_err(SYNERR, "missing '(' in effect definition\n");
4004    return;
4005  }
4006  // Get list of effect-operand pairs and insert into dictionary
4007  else get_effectlist(instr->_effects, instr->_localNames, instr->_has_call);
4008
4009  // Debug Stuff
4010  if (_AD._adl_debug > 1) fprintf(stderr,"Effect description: %s\n", desc);
4011  if (_curchar != ';') {
4012    parse_err(SYNERR, "missing ';' in Effect definition\n");
4013  }
4014  next_char();                  // Skip ';'
4015
4016}
4017
4018//------------------------------expand_parse-----------------------------------
4019ExpandRule* ADLParser::expand_parse(InstructForm *instr) {
4020  char         *ident, *ident2;
4021  NameAndList  *instr_and_operands = NULL;
4022  ExpandRule   *exp = new ExpandRule();
4023
4024  // Expand is a block containing an ordered list of operands with initializers,
4025  // or instructions, each of which has an ordered list of operands.
4026  // Check for block delimiter
4027  skipws();                        // Skip leading whitespace
4028  if ((_curchar != '%')
4029      || (next_char(), (_curchar != '{')) ) { // If not open block
4030    parse_err(SYNERR, "missing '%%{' in expand definition\n");
4031    return(NULL);
4032  }
4033  next_char();                     // Maintain the invariant
4034  do {
4035    ident = get_ident();           // Grab next identifier
4036    if (ident == NULL) {
4037      parse_err(SYNERR, "identifier expected at %c\n", _curchar);
4038      continue;
4039    }
4040
4041    // Check whether we should parse an instruction or operand.
4042    const Form *form = _globalNames[ident];
4043    bool parse_oper = false;
4044    bool parse_ins  = false;
4045    if (form == NULL) {
4046      skipws();
4047      // Check whether this looks like an instruction specification.  If so,
4048      // just parse the instruction.  The declaration of the instruction is
4049      // not needed here.
4050      if (_curchar == '(') parse_ins = true;
4051    } else if (form->is_instruction()) {
4052      parse_ins = true;
4053    } else if (form->is_operand()) {
4054      parse_oper = true;
4055    } else {
4056      parse_err(SYNERR, "instruction/operand name expected at %s\n", ident);
4057      continue;
4058    }
4059
4060    if (parse_oper) {
4061      // This is a new operand
4062      OperandForm *oper = form->is_operand();
4063      if (oper == NULL) {
4064        parse_err(SYNERR, "instruction/operand name expected at %s\n", ident);
4065        continue;
4066      }
4067      // Throw the operand on the _newopers list
4068      skipws();
4069      ident = get_unique_ident(instr->_localNames,"Operand");
4070      if (ident == NULL) {
4071        parse_err(SYNERR, "identifier expected at %c\n", _curchar);
4072        continue;
4073      }
4074      exp->_newopers.addName(ident);
4075      // Add new operand to LocalNames
4076      instr->_localNames.Insert(ident, oper);
4077      // Grab any constructor code and save as a string
4078      char *c = NULL;
4079      skipws();
4080      if (_curchar == '%') { // Need a constructor for the operand
4081        c = find_cpp_block("Operand Constructor");
4082        if (c == NULL) {
4083          parse_err(SYNERR, "Invalid code block for operand constructor\n", _curchar);
4084          continue;
4085        }
4086        // Add constructor to _newopconst Dict
4087        exp->_newopconst.Insert(ident, c);
4088      }
4089      else if (_curchar != ';') { // If no constructor, need a ;
4090        parse_err(SYNERR, "Missing ; in expand rule operand declaration\n");
4091        continue;
4092      }
4093      else next_char(); // Skip the ;
4094      skipws();
4095    }
4096    else {
4097      assert(parse_ins, "sanity");
4098      // Add instruction to list
4099      instr_and_operands = new NameAndList(ident);
4100      // Grab operands, build nameList of them, and then put into dictionary
4101      skipws();
4102      if (_curchar != '(') {         // Check for parenthesized operand list
4103        parse_err(SYNERR, "missing '(' in expand instruction declaration\n");
4104        continue;
4105      }
4106      do {
4107        next_char();                 // skip open paren & comma characters
4108        skipws();
4109        if (_curchar == ')') break;
4110        ident2 = get_ident();
4111        skipws();
4112        if (ident2 == NULL) {
4113          parse_err(SYNERR, "identifier expected at %c\n", _curchar);
4114          continue;
4115        }                            // Check that you have a valid operand
4116        const Form *form2 = instr->_localNames[ident2];
4117        if (!form2) {
4118          parse_err(SYNERR, "operand name expected at %s\n", ident2);
4119          continue;
4120        }
4121        OperandForm *oper = form2->is_operand();
4122        if (oper == NULL && !form2->is_opclass()) {
4123          parse_err(SYNERR, "operand name expected at %s\n", ident2);
4124          continue;
4125        }                            // Add operand to list
4126        instr_and_operands->add_entry(ident2);
4127      } while(_curchar == ',');
4128      if (_curchar != ')') {
4129        parse_err(SYNERR, "missing ')'in expand instruction declaration\n");
4130        continue;
4131      }
4132      next_char();
4133      if (_curchar != ';') {
4134        parse_err(SYNERR, "missing ';'in expand instruction declaration\n");
4135        continue;
4136      }
4137      next_char();
4138
4139      // Record both instruction name and its operand list
4140      exp->add_instruction(instr_and_operands);
4141
4142      skipws();
4143    }
4144
4145  } while(_curchar != '%');
4146  next_char();
4147  if (_curchar != '}') {
4148    parse_err(SYNERR, "missing '%%}' in expand rule definition\n");
4149    return(NULL);
4150  }
4151  next_char();
4152
4153  // Debug Stuff
4154  if (_AD._adl_debug > 1) fprintf(stderr,"Expand Rule:\n");
4155
4156  skipws();
4157  return (exp);
4158}
4159
4160//------------------------------rewrite_parse----------------------------------
4161RewriteRule* ADLParser::rewrite_parse(void) {
4162  char* params = NULL;
4163  char* desc   = NULL;
4164
4165
4166  // This feature targeted for second generation description language.
4167
4168  skipws();                      // Skip whitespace
4169  // Get parameters for rewrite
4170  if ((params = get_paren_expr("rewrite parameters")) == NULL) {
4171    parse_err(SYNERR, "missing '(' in rewrite rule\n");
4172    return NULL;
4173  }
4174  // Debug Stuff
4175  if (_AD._adl_debug > 1) fprintf(stderr,"Rewrite parameters: %s\n", params);
4176
4177  // For now, grab entire block;
4178  skipws();
4179  if ( (desc = find_cpp_block("rewrite block")) == NULL ) {
4180    parse_err(SYNERR, "incorrect or missing block for 'rewrite'.\n");
4181    return NULL;
4182  }
4183  // Debug Stuff
4184  if (_AD._adl_debug > 1) fprintf(stderr,"Rewrite Rule: %s\n", desc);
4185
4186  skipws();
4187  return (new RewriteRule(params,desc));
4188}
4189
4190//------------------------------attr_parse-------------------------------------
4191Attribute *ADLParser::attr_parse(char* ident) {
4192  Attribute *attrib;              // Attribute class
4193  char      *cost = NULL;         // String representation of cost attribute
4194
4195  skipws();                       // Skip leading whitespace
4196  if ( (cost = get_paren_expr("attribute")) == NULL ) {
4197    parse_err(SYNERR, "incorrect or missing expression for 'attribute'\n");
4198    return NULL;
4199  }
4200  // Debug Stuff
4201  if (_AD._adl_debug > 1) fprintf(stderr,"Attribute: %s\n", cost);
4202  if (_curchar != ';') {
4203    parse_err(SYNERR, "missing ';' in attribute definition\n");
4204    return NULL;
4205  }
4206  next_char();                   // Point after the terminator
4207
4208  skipws();
4209  attrib = new Attribute(ident,cost,INS_ATTR); // Build new predicate object
4210  return attrib;
4211}
4212
4213
4214//------------------------------matchNode_parse--------------------------------
4215MatchNode *ADLParser::matchNode_parse(FormDict &operands, int &depth, int &numleaves, bool atroot) {
4216  // Count depth of parenthesis nesting for both left and right children
4217  int   lParens = depth;
4218  int   rParens = depth;
4219
4220  // MatchNode objects for left, right, and root of subtree.
4221  MatchNode *lChild = NULL;
4222  MatchNode *rChild = NULL;
4223  char      *token;               // Identifier which may be opcode or operand
4224
4225  // Match expression starts with a '('
4226  if (cur_char() != '(')
4227    return NULL;
4228
4229  next_char();                    // advance past '('
4230
4231  // Parse the opcode
4232  token = get_ident();            // Get identifier, opcode
4233  if (token == NULL) {
4234    parse_err(SYNERR, "missing opcode in match expression\n");
4235    return NULL;
4236  }
4237
4238  // Take note if we see one of a few special operations - those that are
4239  // treated differently on different architectures in the sense that on
4240  // one architecture there is a match rule and on another there isn't (so
4241  // a call will eventually be generated).
4242
4243  for (int i = _last_machine_leaf + 1; i < _last_opcode; i++) {
4244    if (strcmp(token, NodeClassNames[i]) == 0) {
4245      _AD.has_match_rule(i, true);
4246    }
4247  }
4248
4249  // Lookup the root value in the operands dict to perform substitution
4250  const char  *result    = NULL;  // Result type will be filled in later
4251  const char  *name      = token; // local name associated with this node
4252  const char  *operation = token; // remember valid operation for later
4253  const Form  *form      = operands[token];
4254  OpClassForm *opcForm = form ? form->is_opclass() : NULL;
4255  if (opcForm != NULL) {
4256    // If this token is an entry in the local names table, record its type
4257    if (!opcForm->ideal_only()) {
4258      operation = opcForm->_ident;
4259      result = operation;         // Operands result in their own type
4260    }
4261    // Otherwise it is an ideal type, and so, has no local name
4262    else                        name = NULL;
4263  }
4264
4265  // Parse the operands
4266  skipws();
4267  if (cur_char() != ')') {
4268
4269    // Parse the left child
4270    if (strcmp(operation,"Set"))
4271      lChild = matchChild_parse(operands, lParens, numleaves, false);
4272    else
4273      lChild = matchChild_parse(operands, lParens, numleaves, true);
4274
4275    skipws();
4276    if (cur_char() != ')' ) {
4277      if(strcmp(operation, "Set"))
4278        rChild = matchChild_parse(operands,rParens,numleaves,false);
4279      else
4280        rChild = matchChild_parse(operands,rParens,numleaves,true);
4281    }
4282  }
4283
4284  // Check for required ')'
4285  skipws();
4286  if (cur_char() != ')') {
4287    parse_err(SYNERR, "missing ')' in match expression\n");
4288    return NULL;
4289  }
4290  next_char();                    // skip the ')'
4291
4292  MatchNode* mroot = new MatchNode(_AD,result,name,operation,lChild,rChild);
4293
4294  // If not the root, reduce this subtree to an internal operand
4295  if (!atroot) {
4296    mroot->build_internalop();
4297  }
4298  // depth is greater of left and right paths.
4299  depth = (lParens > rParens) ? lParens : rParens;
4300
4301  return mroot;
4302}
4303
4304
4305//------------------------------matchChild_parse-------------------------------
4306MatchNode *ADLParser::matchChild_parse(FormDict &operands, int &parens, int &numleaves, bool atroot) {
4307  MatchNode  *child  = NULL;
4308  const char *result = NULL;
4309  const char *token  = NULL;
4310  const char *opType = NULL;
4311
4312  if (cur_char() == '(') {         // child is an operation
4313    ++parens;
4314    child = matchNode_parse(operands, parens, numleaves, atroot);
4315  }
4316  else {                           // child is an operand
4317    token = get_ident();
4318    const Form  *form    = operands[token];
4319    OpClassForm *opcForm = form ? form->is_opclass() : NULL;
4320    if (opcForm != NULL) {
4321      opType = opcForm->_ident;
4322      result = opcForm->_ident;    // an operand's result matches its type
4323    } else {
4324      parse_err(SYNERR, "undefined operand %s in match rule\n", token);
4325      return NULL;
4326    }
4327
4328    if (opType == NULL) {
4329      parse_err(SYNERR, "missing type for argument '%s'\n", token);
4330    }
4331
4332    child = new MatchNode(_AD, result, token, opType);
4333    ++numleaves;
4334  }
4335
4336  return child;
4337}
4338
4339
4340
4341// ******************** Private Utility Functions *************************
4342
4343
4344char* ADLParser::find_cpp_block(const char* description) {
4345  char *next;                     // Pointer for finding block delimiters
4346  char* cppBlock = NULL;          // Beginning of C++ code block
4347
4348  if (_curchar == '%') {          // Encoding is a C++ expression
4349    next_char();
4350    if (_curchar != '{') {
4351      parse_err(SYNERR, "missing '{' in %s \n", description);
4352      return NULL;
4353    }
4354    next_char();                  // Skip block delimiter
4355    skipws_no_preproc();          // Skip leading whitespace
4356    cppBlock = _ptr;              // Point to start of expression
4357    int line = linenum();
4358    next = _ptr + 1;
4359    while(((_curchar != '%') || (*next != '}')) && (_curchar != '\0')) {
4360      next_char_or_line();
4361      next = _ptr+1;              // Maintain the next pointer
4362    }                             // Grab string
4363    if (_curchar == '\0') {
4364      parse_err(SYNERR, "invalid termination of %s \n", description);
4365      return NULL;
4366    }
4367    *_ptr = '\0';                 // Terminate string
4368    _ptr += 2;                    // Skip block delimiter
4369    _curchar = *_ptr;             // Maintain invariant
4370
4371    // Prepend location descriptor, for debugging.
4372    if (_AD._adlocation_debug) {
4373      char* location = get_line_string(line);
4374      char* end_loc  = end_line_marker();
4375      char* result = (char *)malloc(strlen(location) + strlen(cppBlock) + strlen(end_loc) + 1);
4376      strcpy(result, location);
4377      strcat(result, cppBlock);
4378      strcat(result, end_loc);
4379      cppBlock = result;
4380      free(location);
4381    }
4382  }
4383
4384  return cppBlock;
4385}
4386
4387// Move to the closing token of the expression we are currently at,
4388// as defined by stop_chars.  Match parens and quotes.
4389char* ADLParser::get_expr(const char *desc, const char *stop_chars) {
4390  char* expr = NULL;
4391  int   paren = 0;
4392
4393  expr = _ptr;
4394  while (paren > 0 || !strchr(stop_chars, _curchar)) {
4395    if (_curchar == '(') {        // Down level of nesting
4396      paren++;                    // Bump the parenthesis counter
4397      next_char();                // maintain the invariant
4398    }
4399    else if (_curchar == ')') {   // Up one level of nesting
4400      if (paren == 0) {
4401        // Paren underflow:  We didn't encounter the required stop-char.
4402        parse_err(SYNERR, "too many )'s, did not find %s after %s\n",
4403                  stop_chars, desc);
4404        return NULL;
4405      }
4406      paren--;                    // Drop the parenthesis counter
4407      next_char();                // Maintain the invariant
4408    }
4409    else if (_curchar == '"' || _curchar == '\'') {
4410      int qchar = _curchar;
4411      while (true) {
4412        next_char();
4413        if (_curchar == qchar) { next_char(); break; }
4414        if (_curchar == '\\')  next_char();  // superquote
4415        if (_curchar == '\n' || _curchar == '\0') {
4416          parse_err(SYNERR, "newline in string in %s\n", desc);
4417          return NULL;
4418        }
4419      }
4420    }
4421    else if (_curchar == '%' && (_ptr[1] == '{' || _ptr[1] == '}')) {
4422      // Make sure we do not stray into the next ADLC-level form.
4423      parse_err(SYNERR, "unexpected %%%c in %s\n", _ptr[1], desc);
4424      return NULL;
4425    }
4426    else if (_curchar == '\0') {
4427      parse_err(SYNERR, "unexpected EOF in %s\n", desc);
4428      return NULL;
4429    }
4430    else {
4431      // Always walk over whitespace, comments, preprocessor directives, etc.
4432      char* pre_skip_ptr = _ptr;
4433      skipws();
4434      // If the parser declined to make progress on whitespace,
4435      // skip the next character, which is therefore NOT whitespace.
4436      if (pre_skip_ptr == _ptr) {
4437        next_char();
4438      } else if (pre_skip_ptr+strlen(pre_skip_ptr) != _ptr+strlen(_ptr)) {
4439        parse_err(SYNERR, "unimplemented: preprocessor must not elide subexpression in %s", desc);
4440      }
4441    }
4442  }
4443
4444  assert(strchr(stop_chars, _curchar), "non-null return must be at stop-char");
4445  *_ptr = '\0';               // Replace ')' or other stop-char with '\0'
4446  return expr;
4447}
4448
4449// Helper function around get_expr
4450// Sets _curchar to '(' so that get_paren_expr will search for a matching ')'
4451char *ADLParser::get_paren_expr(const char *description, bool include_location) {
4452  int line = linenum();
4453  if (_curchar != '(')            // Escape if not valid starting position
4454    return NULL;
4455  next_char();                    // Skip the required initial paren.
4456  char *token2 = get_expr(description, ")");
4457  if (_curchar == ')')
4458    next_char();                  // Skip required final paren.
4459  int junk = 0;
4460  if (include_location && _AD._adlocation_debug && !is_int_token(token2, junk)) {
4461    // Prepend location descriptor, for debugging.
4462    char* location = get_line_string(line);
4463    char* end_loc  = end_line_marker();
4464    char* result = (char *)malloc(strlen(location) + strlen(token2) + strlen(end_loc) + 1);
4465    strcpy(result, location);
4466    strcat(result, token2);
4467    strcat(result, end_loc);
4468    token2 = result;
4469    free(location);
4470  }
4471  return token2;
4472}
4473
4474//------------------------------get_ident_common-------------------------------
4475// Looks for an identifier in the buffer, and turns it into a null terminated
4476// string(still inside the file buffer).  Returns a pointer to the string or
4477// NULL if some other token is found instead.
4478char *ADLParser::get_ident_common(bool do_preproc) {
4479  register char c;
4480  char *start;                    // Pointer to start of token
4481  char *end;                      // Pointer to end of token
4482
4483  if( _curline == NULL )          // Return NULL at EOF.
4484    return NULL;
4485
4486  skipws_common(do_preproc);      // Skip whitespace before identifier
4487  start = end = _ptr;             // Start points at first character
4488  end--;                          // unwind end by one to prepare for loop
4489  do {
4490    end++;                        // Increment end pointer
4491    c = *end;                     // Grab character to test
4492  } while ( ((c >= 'a') && (c <= 'z')) || ((c >= 'A') && (c <= 'Z'))
4493            || ((c >= '0') && (c <= '9'))
4494            || ((c == '_')) || ((c == ':')) || ((c == '#')) );
4495  if (start == end) {             // We popped out on the first try
4496    // It can occur that `start' contains the rest of the input file.
4497    // In this case the output should be truncated.
4498    if (strlen(start) > 24) {
4499      char buf[32];
4500      strncpy(buf, start, 20);
4501      buf[20] = '\0';
4502      strcat(buf, "[...]");
4503      parse_err(SYNERR, "Identifier expected, but found '%s'.", buf);
4504    } else {
4505      parse_err(SYNERR, "Identifier expected, but found '%s'.", start);
4506    }
4507    start = NULL;
4508  }
4509  else {
4510    _curchar = c;                 // Save the first character of next token
4511    *end = '\0';                  // NULL terminate the string in place
4512  }
4513  _ptr = end;                     // Reset _ptr to point to next char after token
4514
4515  // Make sure we do not try to use #defined identifiers.  If start is
4516  // NULL an error was already reported.
4517  if (do_preproc && start != NULL) {
4518    const char* def = _AD.get_preproc_def(start);
4519    if (def != NULL && strcmp(def, start)) {
4520      const char* def1 = def;
4521      const char* def2 = _AD.get_preproc_def(def1);
4522      // implement up to 2 levels of #define
4523      if (def2 != NULL && strcmp(def2, def1)) {
4524        def = def2;
4525        const char* def3 = _AD.get_preproc_def(def2);
4526        if (def3 != NULL && strcmp(def3, def2) && strcmp(def3, def1)) {
4527          parse_err(SYNERR, "unimplemented: using %s defined as %s => %s => %s",
4528                    start, def1, def2, def3);
4529        }
4530      }
4531      start = strdup(def);
4532    }
4533  }
4534
4535  return start;                   // Pointer to token in filebuf
4536}
4537
4538//------------------------------get_ident_dup----------------------------------
4539// Looks for an identifier in the buffer, and returns a duplicate
4540// or NULL if some other token is found instead.
4541char *ADLParser::get_ident_dup(void) {
4542  char *ident = get_ident();
4543
4544  // Duplicate an identifier before returning and restore string.
4545  if( ident != NULL ) {
4546    ident = strdup(ident);  // Copy the string
4547    *_ptr   = _curchar;         // and replace Nil with original character
4548  }
4549
4550  return ident;
4551}
4552
4553//----------------------get_ident_or_literal_constant--------------------------
4554// Looks for an identifier in the buffer, or a parenthesized expression.
4555char *ADLParser::get_ident_or_literal_constant(const char* description) {
4556  char* param = NULL;
4557  skipws();
4558  if (_curchar == '(') {
4559    // Grab a constant expression.
4560    param = get_paren_expr(description);
4561    if (param[0] != '(') {
4562      char* buf = (char*) malloc(strlen(param) + 3);
4563      sprintf(buf, "(%s)", param);
4564      param = buf;
4565    }
4566    assert(is_literal_constant(param),
4567           "expr must be recognizable as a constant");
4568  } else {
4569    param = get_ident();
4570  }
4571  return param;
4572}
4573
4574//------------------------------get_rep_var_ident-----------------------------
4575// Do NOT duplicate,
4576// Leave nil terminator in buffer
4577// Preserve initial '$'(s) in string
4578char *ADLParser::get_rep_var_ident(void) {
4579  // Remember starting point
4580  char *rep_var = _ptr;
4581
4582  // Check for replacement variable indicator '$' and pass if present
4583  if ( _curchar == '$' ) {
4584    next_char();
4585  }
4586  // Check for a subfield indicator, a second '$', and pass if present
4587  if ( _curchar == '$' ) {
4588    next_char();
4589  }
4590
4591  // Check for a control indicator, a third '$':
4592  if ( _curchar == '$' ) {
4593    next_char();
4594  }
4595
4596  // Check for more than three '$'s in sequence, SYNERR
4597  if( _curchar == '$' ) {
4598    parse_err(SYNERR, "Replacement variables and field specifiers can not start with '$$$$'");
4599    next_char();
4600    return NULL;
4601  }
4602
4603  // Nil terminate the variable name following the '$'
4604  char *rep_var_name = get_ident();
4605  assert( rep_var_name != NULL,
4606          "Missing identifier after replacement variable indicator '$'");
4607
4608  return rep_var;
4609}
4610
4611
4612
4613//------------------------------get_rep_var_ident_dup-------------------------
4614// Return the next replacement variable identifier, skipping first '$'
4615// given a pointer into a line of the buffer.
4616// Null terminates string, still inside the file buffer,
4617// Returns a pointer to a copy of the string, or NULL on failure
4618char *ADLParser::get_rep_var_ident_dup(void) {
4619  if( _curchar != '$' ) return NULL;
4620
4621  next_char();                // Move past the '$'
4622  char *rep_var = _ptr;       // Remember starting point
4623
4624  // Check for a subfield indicator, a second '$':
4625  if ( _curchar == '$' ) {
4626    next_char();
4627  }
4628
4629  // Check for a control indicator, a third '$':
4630  if ( _curchar == '$' ) {
4631    next_char();
4632  }
4633
4634  // Check for more than three '$'s in sequence, SYNERR
4635  if( _curchar == '$' ) {
4636    parse_err(SYNERR, "Replacement variables and field specifiers can not start with '$$$$'");
4637    next_char();
4638    return NULL;
4639  }
4640
4641  // Nil terminate the variable name following the '$'
4642  char *rep_var_name = get_ident();
4643  assert( rep_var_name != NULL,
4644          "Missing identifier after replacement variable indicator '$'");
4645  rep_var = strdup(rep_var);  // Copy the string
4646  *_ptr   = _curchar;         // and replace Nil with original character
4647
4648  return rep_var;
4649}
4650
4651
4652//------------------------------get_unique_ident------------------------------
4653// Looks for an identifier in the buffer, terminates it with a NULL,
4654// and checks that it is unique
4655char *ADLParser::get_unique_ident(FormDict& dict, const char* nameDescription){
4656  char* ident = get_ident();
4657
4658  if (ident == NULL) {
4659    parse_err(SYNERR, "missing %s identifier at %c\n", nameDescription, _curchar);
4660  }
4661  else {
4662    if (dict[ident] != NULL) {
4663      parse_err(SYNERR, "duplicate name %s for %s\n", ident, nameDescription);
4664      ident = NULL;
4665    }
4666  }
4667
4668  return ident;
4669}
4670
4671
4672//------------------------------get_int----------------------------------------
4673// Looks for a character string integer in the buffer, and turns it into an int
4674// invokes a parse_err if the next token is not an integer.
4675// This routine does not leave the integer null-terminated.
4676int ADLParser::get_int(void) {
4677  register char c;
4678  char         *start;            // Pointer to start of token
4679  char         *end;              // Pointer to end of token
4680  int           result;           // Storage for integer result
4681
4682  if( _curline == NULL )          // Return NULL at EOF.
4683    return 0;
4684
4685  skipws();                       // Skip whitespace before identifier
4686  start = end = _ptr;             // Start points at first character
4687  c = *end;                       // Grab character to test
4688  while ((c >= '0') && (c <= '9')
4689         || ((c == '-') && (end == start))) {
4690    end++;                        // Increment end pointer
4691    c = *end;                     // Grab character to test
4692  }
4693  if (start == end) {             // We popped out on the first try
4694    parse_err(SYNERR, "integer expected at %c\n", c);
4695    result = 0;
4696  }
4697  else {
4698    _curchar = c;                 // Save the first character of next token
4699    *end = '\0';                  // NULL terminate the string in place
4700    result = atoi(start);         // Convert the string to an integer
4701    *end = _curchar;              // Restore buffer to original condition
4702  }
4703
4704  // Reset _ptr to next char after token
4705  _ptr = end;
4706
4707  return result;                   // integer
4708}
4709
4710
4711//------------------------------get_relation_dup------------------------------
4712// Looks for a relational operator in the buffer
4713// invokes a parse_err if the next token is not a relation
4714// This routine creates a duplicate of the string in the buffer.
4715char *ADLParser::get_relation_dup(void) {
4716  char         *result = NULL;    // relational operator being returned
4717
4718  if( _curline == NULL )          // Return NULL at EOF.
4719    return  NULL;
4720
4721  skipws();                       // Skip whitespace before relation
4722  char *start = _ptr;             // Store start of relational operator
4723  char first  = *_ptr;            // the first character
4724  if( (first == '=') || (first == '!') || (first == '<') || (first == '>') ) {
4725    next_char();
4726    char second = *_ptr;          // the second character
4727    if( (second == '=') ) {
4728      next_char();
4729      char tmp  = *_ptr;
4730      *_ptr = '\0';               // NULL terminate
4731      result = strdup(start);     // Duplicate the string
4732      *_ptr = tmp;                // restore buffer
4733    } else {
4734      parse_err(SYNERR, "relational operator expected at %s\n", _ptr);
4735    }
4736  } else {
4737    parse_err(SYNERR, "relational operator expected at %s\n", _ptr);
4738  }
4739
4740  return result;
4741}
4742
4743
4744
4745//------------------------------get_oplist-------------------------------------
4746// Looks for identifier pairs where first must be the name of an operand, and
4747// second must be a name unique in the scope of this instruction.  Stores the
4748// names with a pointer to the OpClassForm of their type in a local name table.
4749void ADLParser::get_oplist(NameList &parameters, FormDict &operands) {
4750  OpClassForm *opclass = NULL;
4751  char        *ident   = NULL;
4752
4753  do {
4754    next_char();             // skip open paren & comma characters
4755    skipws();
4756    if (_curchar == ')') break;
4757
4758    // Get operand type, and check it against global name table
4759    ident = get_ident();
4760    if (ident == NULL) {
4761      parse_err(SYNERR, "optype identifier expected at %c\n", _curchar);
4762      return;
4763    }
4764    else {
4765      const Form  *form = _globalNames[ident];
4766      if( form == NULL ) {
4767        parse_err(SYNERR, "undefined operand type %s\n", ident);
4768        return;
4769      }
4770
4771      // Check for valid operand type
4772      OpClassForm *opc  = form->is_opclass();
4773      OperandForm *oper = form->is_operand();
4774      if((oper == NULL) && (opc == NULL)) {
4775        parse_err(SYNERR, "identifier %s not operand type\n", ident);
4776        return;
4777      }
4778      opclass = opc;
4779    }
4780    // Debugging Stuff
4781    if (_AD._adl_debug > 1) fprintf(stderr, "\tOperand Type: %s\t", ident);
4782
4783    // Get name of operand and add it to local name table
4784    if( (ident = get_unique_ident(operands, "operand")) == NULL) {
4785      return;
4786    }
4787    // Parameter names must not be global names.
4788    if( _globalNames[ident] != NULL ) {
4789         parse_err(SYNERR, "Reuse of global name %s as operand.\n",ident);
4790         return;
4791    }
4792    operands.Insert(ident, opclass);
4793    parameters.addName(ident);
4794
4795    // Debugging Stuff
4796    if (_AD._adl_debug > 1) fprintf(stderr, "\tOperand Name: %s\n", ident);
4797    skipws();
4798  } while(_curchar == ',');
4799
4800  if (_curchar != ')') parse_err(SYNERR, "missing ')'\n");
4801  else {
4802    next_char();  // set current character position past the close paren
4803  }
4804}
4805
4806
4807//------------------------------get_effectlist---------------------------------
4808// Looks for identifier pairs where first must be the name of a pre-defined,
4809// effect, and the second must be the name of an operand defined in the
4810// operand list of this instruction.  Stores the names with a pointer to the
4811// effect form in a local effects table.
4812void ADLParser::get_effectlist(FormDict &effects, FormDict &operands, bool& has_call) {
4813  OperandForm *opForm;
4814  Effect      *eForm;
4815  char        *ident;
4816
4817  do {
4818    next_char();             // skip open paren & comma characters
4819    skipws();
4820    if (_curchar == ')') break;
4821
4822    // Get effect type, and check it against global name table
4823    ident = get_ident();
4824    if (ident == NULL) {
4825      parse_err(SYNERR, "effect type identifier expected at %c\n", _curchar);
4826      return;
4827    }
4828    else {
4829      // Check for valid effect type
4830      const Form *form = _globalNames[ident];
4831      if( form == NULL ) {
4832        parse_err(SYNERR, "undefined effect type %s\n", ident);
4833        return;
4834      }
4835      else {
4836        if( (eForm = form->is_effect()) == NULL) {
4837          parse_err(SYNERR, "identifier %s not effect type\n", ident);
4838          return;
4839        }
4840      }
4841    }
4842      // Debugging Stuff
4843    if (_AD._adl_debug > 1) fprintf(stderr, "\tEffect Type: %s\t", ident);
4844    skipws();
4845    if (eForm->is(Component::CALL)) {
4846      if (_AD._adl_debug > 1) fprintf(stderr, "\n");
4847      has_call = true;
4848    } else {
4849      // Get name of operand and check that it is in the local name table
4850      if( (ident = get_unique_ident(effects, "effect")) == NULL) {
4851        parse_err(SYNERR, "missing operand identifier in effect list\n");
4852        return;
4853      }
4854      const Form *form = operands[ident];
4855      opForm = form ? form->is_operand() : NULL;
4856      if( opForm == NULL ) {
4857        if( form && form->is_opclass() ) {
4858          const char* cname = form->is_opclass()->_ident;
4859          parse_err(SYNERR, "operand classes are illegal in effect lists (found %s %s)\n", cname, ident);
4860        } else {
4861          parse_err(SYNERR, "undefined operand %s in effect list\n", ident);
4862        }
4863        return;
4864      }
4865      // Add the pair to the effects table
4866      effects.Insert(ident, eForm);
4867      // Debugging Stuff
4868      if (_AD._adl_debug > 1) fprintf(stderr, "\tOperand Name: %s\n", ident);
4869    }
4870    skipws();
4871  } while(_curchar == ',');
4872
4873  if (_curchar != ')') parse_err(SYNERR, "missing ')'\n");
4874  else {
4875    next_char();  // set current character position past the close paren
4876  }
4877}
4878
4879
4880//-------------------------------preproc_line----------------------------------
4881// A "#line" keyword has been seen, so parse the rest of the line.
4882void ADLParser::preproc_line(void) {
4883  int line = get_int();
4884  skipws_no_preproc();
4885  const char* file = NULL;
4886  if (_curchar == '"') {
4887    next_char();              // Move past the initial '"'
4888    file = _ptr;
4889    while (true) {
4890      if (_curchar == '\n') {
4891        parse_err(SYNERR, "missing '\"' at end of #line directive");
4892        return;
4893      }
4894      if (_curchar == '"') {
4895        *_ptr  = '\0';          // Terminate the string
4896        next_char();
4897        skipws_no_preproc();
4898        break;
4899      }
4900      next_char();
4901    }
4902  }
4903  ensure_end_of_line();
4904  if (file != NULL)
4905    _AD._ADL_file._name = file;
4906  _buf.set_linenum(line);
4907}
4908
4909//------------------------------preproc_define---------------------------------
4910// A "#define" keyword has been seen, so parse the rest of the line.
4911void ADLParser::preproc_define(void) {
4912  char* flag = get_ident_no_preproc();
4913  skipws_no_preproc();
4914  // only #define x y is supported for now
4915  char* def = get_ident_no_preproc();
4916  _AD.set_preproc_def(flag, def);
4917  skipws_no_preproc();
4918  if (_curchar != '\n') {
4919    parse_err(SYNERR, "non-identifier in preprocessor definition\n");
4920  }
4921}
4922
4923//------------------------------preproc_undef----------------------------------
4924// An "#undef" keyword has been seen, so parse the rest of the line.
4925void ADLParser::preproc_undef(void) {
4926  char* flag = get_ident_no_preproc();
4927  skipws_no_preproc();
4928  ensure_end_of_line();
4929  _AD.set_preproc_def(flag, NULL);
4930}
4931
4932
4933
4934//------------------------------parse_err--------------------------------------
4935// Issue a parser error message, and skip to the end of the current line
4936void ADLParser::parse_err(int flag, const char *fmt, ...) {
4937  va_list args;
4938
4939  va_start(args, fmt);
4940  if (flag == 1)
4941    _AD._syntax_errs += _AD.emit_msg(0, flag, linenum(), fmt, args);
4942  else if (flag == 2)
4943    _AD._semantic_errs += _AD.emit_msg(0, flag, linenum(), fmt, args);
4944  else
4945    _AD._warnings += _AD.emit_msg(0, flag, linenum(), fmt, args);
4946
4947  int error_char = _curchar;
4948  char* error_ptr = _ptr+1;
4949  for(;*_ptr != '\n'; _ptr++) ; // Skip to the end of the current line
4950  _curchar = '\n';
4951  va_end(args);
4952  _AD._no_output = 1;
4953
4954  if (flag == 1) {
4955    char* error_tail = strchr(error_ptr, '\n');
4956    char tem = *error_ptr;
4957    error_ptr[-1] = '\0';
4958    char* error_head = error_ptr-1;
4959    while (error_head > _curline && *error_head)  --error_head;
4960    if (error_tail)  *error_tail = '\0';
4961    fprintf(stderr, "Error Context:  %s>>>%c<<<%s\n",
4962            error_head, error_char, error_ptr);
4963    if (error_tail)  *error_tail = '\n';
4964    error_ptr[-1] = tem;
4965  }
4966}
4967
4968//---------------------------ensure_start_of_line------------------------------
4969// A preprocessor directive has been encountered.  Be sure it has fallen at
4970// the beginning of a line, or else report an error.
4971void ADLParser::ensure_start_of_line(void) {
4972  if (_curchar == '\n') { next_line(); return; }
4973  assert( _ptr >= _curline && _ptr < _curline+strlen(_curline),
4974          "Must be able to find which line we are in" );
4975
4976  for (char *s = _curline; s < _ptr; s++) {
4977    if (*s > ' ') {
4978      parse_err(SYNERR, "'%c' must be at beginning of line\n", _curchar);
4979      break;
4980    }
4981  }
4982}
4983
4984//---------------------------ensure_end_of_line--------------------------------
4985// A preprocessor directive has been parsed.  Be sure there is no trailing
4986// garbage at the end of this line.  Set the scan point to the beginning of
4987// the next line.
4988void ADLParser::ensure_end_of_line(void) {
4989  skipws_no_preproc();
4990  if (_curchar != '\n' && _curchar != '\0') {
4991    parse_err(SYNERR, "garbage char '%c' at end of line\n", _curchar);
4992  } else {
4993    next_char_or_line();
4994  }
4995}
4996
4997//---------------------------handle_preproc------------------------------------
4998// The '#' character introducing a preprocessor directive has been found.
4999// Parse the whole directive name (e.g., #define, #endif) and take appropriate
5000// action.  If we are in an "untaken" span of text, simply keep track of
5001// #ifdef nesting structure, so we can find out when to start taking text
5002// again.  (In this state, we "sort of support" C's #if directives, enough
5003// to disregard their associated #else and #endif lines.)  If we are in a
5004// "taken" span of text, there are two cases:  "#define" and "#undef"
5005// directives are preserved and passed up to the caller, which eventually
5006// passes control to the top-level parser loop, which handles #define and
5007// #undef directly.  (This prevents these directives from occurring in
5008// arbitrary positions in the AD file--we require better structure than C.)
5009// In the other case, and #ifdef, #ifndef, #else, or #endif is silently
5010// processed as whitespace, with the "taken" state of the text correctly
5011// updated.  This routine returns "false" exactly in the case of a "taken"
5012// #define or #undef, which tells the caller that a preprocessor token
5013// has appeared which must be handled explicitly by the parse loop.
5014bool ADLParser::handle_preproc_token() {
5015  assert(*_ptr == '#', "must be at start of preproc");
5016  ensure_start_of_line();
5017  next_char();
5018  skipws_no_preproc();
5019  char* start_ident = _ptr;
5020  char* ident = (_curchar == '\n') ? NULL : get_ident_no_preproc();
5021  if (ident == NULL) {
5022    parse_err(SYNERR, "expected preprocessor command, got end of line\n");
5023  } else if (!strcmp(ident, "ifdef") ||
5024             !strcmp(ident, "ifndef")) {
5025    char* flag = get_ident_no_preproc();
5026    ensure_end_of_line();
5027    // Test the identifier only if we are already in taken code:
5028    bool flag_def  = preproc_taken() && (_AD.get_preproc_def(flag) != NULL);
5029    bool now_taken = !strcmp(ident, "ifdef") ? flag_def : !flag_def;
5030    begin_if_def(now_taken);
5031  } else if (!strcmp(ident, "if")) {
5032    if (preproc_taken())
5033      parse_err(SYNERR, "unimplemented: #%s %s", ident, _ptr+1);
5034    next_line();
5035    // Intelligently skip this nested C preprocessor directive:
5036    begin_if_def(true);
5037  } else if (!strcmp(ident, "else")) {
5038    ensure_end_of_line();
5039    invert_if_def();
5040  } else if (!strcmp(ident, "endif")) {
5041    ensure_end_of_line();
5042    end_if_def();
5043  } else if (preproc_taken()) {
5044    // pass this token up to the main parser as "#define" or "#undef"
5045    _ptr = start_ident;
5046    _curchar = *--_ptr;
5047    if( _curchar != '#' ) {
5048      parse_err(SYNERR, "no space allowed after # in #define or #undef");
5049      assert(_curchar == '#', "no space allowed after # in #define or #undef");
5050    }
5051    return false;
5052  }
5053  return true;
5054}
5055
5056//---------------------------skipws_common-------------------------------------
5057// Skip whitespace, including comments and newlines, while keeping an accurate
5058// line count.
5059// Maybe handle certain preprocessor constructs: #ifdef, #ifndef, #else, #endif
5060void ADLParser::skipws_common(bool do_preproc) {
5061  char *start = _ptr;
5062  char *next = _ptr + 1;
5063
5064  if (*_ptr == '\0') {
5065    // Check for string terminator
5066    if (_curchar > ' ')  return;
5067    if (_curchar == '\n') {
5068      if (!do_preproc)  return;            // let caller handle the newline
5069      next_line();
5070      _ptr = _curline; next = _ptr + 1;
5071    }
5072    else if (_curchar == '#' ||
5073        (_curchar == '/' && (*next == '/' || *next == '*'))) {
5074      parse_err(SYNERR, "unimplemented: comment token in a funny place");
5075    }
5076  }
5077  while(_curline != NULL) {                // Check for end of file
5078    if (*_ptr == '\n') {                   // keep proper track of new lines
5079      if (!do_preproc)  break;             // let caller handle the newline
5080      next_line();
5081      _ptr = _curline; next = _ptr + 1;
5082    }
5083    else if ((*_ptr == '/') && (*next == '/'))      // C++ comment
5084      do { _ptr++; next++; } while(*_ptr != '\n');  // So go to end of line
5085    else if ((*_ptr == '/') && (*next == '*')) {    // C comment
5086      _ptr++; next++;
5087      do {
5088        _ptr++; next++;
5089        if (*_ptr == '\n') {               // keep proper track of new lines
5090          next_line();                     // skip newlines within comments
5091          if (_curline == NULL) {          // check for end of file
5092            parse_err(SYNERR, "end-of-file detected inside comment\n");
5093            break;
5094          }
5095          _ptr = _curline; next = _ptr + 1;
5096        }
5097      } while(!((*_ptr == '*') && (*next == '/'))); // Go to end of comment
5098      _ptr = ++next; next++;               // increment _ptr past comment end
5099    }
5100    else if (do_preproc && *_ptr == '#') {
5101      // Note that this calls skipws_common(false) recursively!
5102      bool preproc_handled = handle_preproc_token();
5103      if (!preproc_handled) {
5104        if (preproc_taken()) {
5105          return;  // short circuit
5106        }
5107        ++_ptr;    // skip the preprocessor character
5108      }
5109      next = _ptr+1;
5110    } else if(*_ptr > ' ' && !(do_preproc && !preproc_taken())) {
5111      break;
5112    }
5113    else if (*_ptr == '"' || *_ptr == '\'') {
5114      assert(do_preproc, "only skip strings if doing preproc");
5115      // skip untaken quoted string
5116      int qchar = *_ptr;
5117      while (true) {
5118        ++_ptr;
5119        if (*_ptr == qchar) { ++_ptr; break; }
5120        if (*_ptr == '\\')  ++_ptr;
5121        if (*_ptr == '\n' || *_ptr == '\0') {
5122          parse_err(SYNERR, "newline in string");
5123          break;
5124        }
5125      }
5126      next = _ptr + 1;
5127    }
5128    else { ++_ptr; ++next; }
5129  }
5130  if( _curline != NULL )            // at end of file _curchar isn't valid
5131    _curchar = *_ptr;               // reset _curchar to maintain invariant
5132}
5133
5134//---------------------------cur_char-----------------------------------------
5135char ADLParser::cur_char() {
5136  return (_curchar);
5137}
5138
5139//---------------------------next_char-----------------------------------------
5140void ADLParser::next_char() {
5141  if (_curchar == '\n')  parse_err(WARN, "must call next_line!");
5142  _curchar = *++_ptr;
5143  // if ( _curchar == '\n' ) {
5144  //   next_line();
5145  // }
5146}
5147
5148//---------------------------next_char_or_line---------------------------------
5149void ADLParser::next_char_or_line() {
5150  if ( _curchar != '\n' ) {
5151    _curchar = *++_ptr;
5152  } else {
5153    next_line();
5154    _ptr = _curline;
5155    _curchar = *_ptr;  // maintain invariant
5156  }
5157}
5158
5159//---------------------------next_line-----------------------------------------
5160void ADLParser::next_line() {
5161  _curline = _buf.get_line();
5162  _curchar = ' ';
5163}
5164
5165//------------------------get_line_string--------------------------------------
5166// Prepended location descriptor, for debugging.
5167// Must return a malloced string (that can be freed if desired).
5168char* ADLParser::get_line_string(int linenum) {
5169  const char* file = _AD._ADL_file._name;
5170  int         line = linenum ? linenum : this->linenum();
5171  char* location = (char *)malloc(strlen(file) + 100);
5172  sprintf(location, "\n#line %d \"%s\"\n", line, file);
5173  return location;
5174}
5175
5176//-------------------------is_literal_constant---------------------------------
5177bool ADLParser::is_literal_constant(const char *param) {
5178  if (param[0] == 0)     return false;  // null string
5179  if (param[0] == '(')   return true;   // parenthesized expression
5180  if (param[0] == '0' && (param[1] == 'x' || param[1] == 'X')) {
5181    // Make sure it's a hex constant.
5182    int i = 2;
5183    do {
5184      if( !ADLParser::is_hex_digit(*(param+i)) )  return false;
5185      ++i;
5186    } while( *(param+i) != 0 );
5187    return true;
5188  }
5189  return false;
5190}
5191
5192//---------------------------is_hex_digit--------------------------------------
5193bool ADLParser::is_hex_digit(char digit) {
5194  return ((digit >= '0') && (digit <= '9'))
5195       ||((digit >= 'a') && (digit <= 'f'))
5196       ||((digit >= 'A') && (digit <= 'F'));
5197}
5198
5199//---------------------------is_int_token--------------------------------------
5200bool ADLParser::is_int_token(const char* token, int& intval) {
5201  const char* cp = token;
5202  while (*cp != '\0' && *cp <= ' ')  cp++;
5203  if (*cp == '-')  cp++;
5204  int ndigit = 0;
5205  while (*cp >= '0' && *cp <= '9')  { cp++; ndigit++; }
5206  while (*cp != '\0' && *cp <= ' ')  cp++;
5207  if (ndigit == 0 || *cp != '\0') {
5208    return false;
5209  }
5210  intval = atoi(token);
5211  return true;
5212}
5213
5214static const char* skip_expr_ws(const char* str) {
5215  const char * cp = str;
5216  while (cp[0]) {
5217    if (cp[0] <= ' ') {
5218      ++cp;
5219    } else if (cp[0] == '#') {
5220      ++cp;
5221      while (cp[0] == ' ')  ++cp;
5222      assert(0 == strncmp(cp, "line", 4), "must be a #line directive");
5223      const char* eol = strchr(cp, '\n');
5224      assert(eol != NULL, "must find end of line");
5225      if (eol == NULL)  eol = cp + strlen(cp);
5226      cp = eol;
5227    } else {
5228      break;
5229    }
5230  }
5231  return cp;
5232}
5233
5234//-----------------------equivalent_expressions--------------------------------
5235bool ADLParser::equivalent_expressions(const char* str1, const char* str2) {
5236  if (str1 == str2)
5237    return true;
5238  else if (str1 == NULL || str2 == NULL)
5239    return false;
5240  const char* cp1 = str1;
5241  const char* cp2 = str2;
5242  char in_quote = '\0';
5243  while (cp1[0] && cp2[0]) {
5244    if (!in_quote) {
5245      // skip spaces and/or cpp directives
5246      const char* cp1a = skip_expr_ws(cp1);
5247      const char* cp2a = skip_expr_ws(cp2);
5248      if (cp1a > cp1 && cp2a > cp2) {
5249        cp1 = cp1a; cp2 = cp2a;
5250        continue;
5251      }
5252      if (cp1a > cp1 || cp2a > cp2)  break; // fail
5253    }
5254    // match one non-space char
5255    if (cp1[0] != cp2[0])  break; // fail
5256    char ch = cp1[0];
5257    cp1++; cp2++;
5258    // watch for quotes
5259    if (in_quote && ch == '\\') {
5260      if (cp1[0] != cp2[0])  break; // fail
5261      if (!cp1[0])  break;
5262      cp1++; cp2++;
5263    }
5264    if (in_quote && ch == in_quote) {
5265      in_quote = '\0';
5266    } else if (!in_quote && (ch == '"' || ch == '\'')) {
5267      in_quote = ch;
5268    }
5269  }
5270  return (!cp1[0] && !cp2[0]);
5271}
5272
5273
5274//-------------------------------trim------------------------------------------
5275void ADLParser::trim(char* &token) {
5276  while (*token <= ' ')  token++;
5277  char* end = token + strlen(token);
5278  while (end > token && *(end-1) <= ' ')  --end;
5279  *end = '\0';
5280}
5281