aboutsummaryrefslogtreecommitdiff
path: root/utils/TableGen/AsmWriterEmitter.cpp
blob: e2cd5dc031013edc3a7de5559e85f6cd00872606 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
//===- AsmWriterEmitter.cpp - Generate an assembly writer -----------------===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This tablegen backend is emits an assembly printer for the current target.
// Note that this is currently fairly skeletal, but will grow over time.
//
//===----------------------------------------------------------------------===//

#include "AsmWriterEmitter.h"
#include "CodeGenTarget.h"
#include "Record.h"
#include "llvm/ADT/StringExtras.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/MathExtras.h"
#include <algorithm>
using namespace llvm;

static bool isIdentChar(char C) {
  return (C >= 'a' && C <= 'z') ||
         (C >= 'A' && C <= 'Z') ||
         (C >= '0' && C <= '9') ||
         C == '_';
}

// This should be an anon namespace, this works around a GCC warning.
namespace llvm {  
  struct AsmWriterOperand {
    enum { isLiteralTextOperand, isMachineInstrOperand } OperandType;

    /// Str - For isLiteralTextOperand, this IS the literal text.  For
    /// isMachineInstrOperand, this is the PrinterMethodName for the operand.
    std::string Str;

    /// MiOpNo - For isMachineInstrOperand, this is the operand number of the
    /// machine instruction.
    unsigned MIOpNo;
    
    /// MiModifier - For isMachineInstrOperand, this is the modifier string for
    /// an operand, specified with syntax like ${opname:modifier}.
    std::string MiModifier;

    // To make VS STL happy
    AsmWriterOperand():OperandType(isLiteralTextOperand) {}

    AsmWriterOperand(const std::string &LitStr)
      : OperandType(isLiteralTextOperand), Str(LitStr) {}

    AsmWriterOperand(const std::string &Printer, unsigned OpNo, 
                     const std::string &Modifier) 
      : OperandType(isMachineInstrOperand), Str(Printer), MIOpNo(OpNo),
      MiModifier(Modifier) {}

    bool operator!=(const AsmWriterOperand &Other) const {
      if (OperandType != Other.OperandType || Str != Other.Str) return true;
      if (OperandType == isMachineInstrOperand)
        return MIOpNo != Other.MIOpNo || MiModifier != Other.MiModifier;
      return false;
    }
    bool operator==(const AsmWriterOperand &Other) const {
      return !operator!=(Other);
    }
    
    /// getCode - Return the code that prints this operand.
    std::string getCode() const;
  };
}

namespace llvm {
  class AsmWriterInst {
  public:
    std::vector<AsmWriterOperand> Operands;
    const CodeGenInstruction *CGI;

    AsmWriterInst(const CodeGenInstruction &CGI, unsigned Variant);

    /// MatchesAllButOneOp - If this instruction is exactly identical to the
    /// specified instruction except for one differing operand, return the
    /// differing operand number.  Otherwise return ~0.
    unsigned MatchesAllButOneOp(const AsmWriterInst &Other) const;

  private:
    void AddLiteralString(const std::string &Str) {
      // If the last operand was already a literal text string, append this to
      // it, otherwise add a new operand.
      if (!Operands.empty() &&
          Operands.back().OperandType == AsmWriterOperand::isLiteralTextOperand)
        Operands.back().Str.append(Str);
      else
        Operands.push_back(AsmWriterOperand(Str));
    }
  };
}


std::string AsmWriterOperand::getCode() const {
  if (OperandType == isLiteralTextOperand)
    return "O << \"" + Str + "\"; ";

  std::string Result = Str + "(MI";
  if (MIOpNo != ~0U)
    Result += ", " + utostr(MIOpNo);
  if (!MiModifier.empty())
    Result += ", \"" + MiModifier + '"';
  return Result + "); ";
}


/// ParseAsmString - Parse the specified Instruction's AsmString into this
/// AsmWriterInst.
///
AsmWriterInst::AsmWriterInst(const CodeGenInstruction &CGI, unsigned Variant) {
  this->CGI = &CGI;
  unsigned CurVariant = ~0U;  // ~0 if we are outside a {.|.|.} region, other #.

  // NOTE: Any extensions to this code need to be mirrored in the 
  // AsmPrinter::printInlineAsm code that executes as compile time (assuming
  // that inline asm strings should also get the new feature)!
  const std::string &AsmString = CGI.AsmString;
  std::string::size_type LastEmitted = 0;
  while (LastEmitted != AsmString.size()) {
    std::string::size_type DollarPos =
      AsmString.find_first_of("${|}\\", LastEmitted);
    if (DollarPos == std::string::npos) DollarPos = AsmString.size();

    // Emit a constant string fragment.
    if (DollarPos != LastEmitted) {
      // TODO: this should eventually handle escaping.
      if (CurVariant == Variant || CurVariant == ~0U)
        AddLiteralString(std::string(AsmString.begin()+LastEmitted,
                                     AsmString.begin()+DollarPos));
      LastEmitted = DollarPos;
    } else if (AsmString[DollarPos] == '\\') {
      if (DollarPos+1 != AsmString.size() &&
          (CurVariant == Variant || CurVariant == ~0U)) {
        if (AsmString[DollarPos+1] == 'n') {
          AddLiteralString("\\n");
        } else if (AsmString[DollarPos+1] == 't') {
          AddLiteralString("\\t");
        } else if (std::string("${|}\\").find(AsmString[DollarPos+1]) 
                   != std::string::npos) {
          AddLiteralString(std::string(1, AsmString[DollarPos+1]));
        } else {
          throw "Non-supported escaped character found in instruction '" +
            CGI.TheDef->getName() + "'!";
        }
        LastEmitted = DollarPos+2;
        continue;
      }
    } else if (AsmString[DollarPos] == '{') {
      if (CurVariant != ~0U)
        throw "Nested variants found for instruction '" +
              CGI.TheDef->getName() + "'!";
      LastEmitted = DollarPos+1;
      CurVariant = 0;   // We are now inside of the variant!
    } else if (AsmString[DollarPos] == '|') {
      if (CurVariant == ~0U)
        throw "'|' character found outside of a variant in instruction '"
          + CGI.TheDef->getName() + "'!";
      ++CurVariant;
      ++LastEmitted;
    } else if (AsmString[DollarPos] == '}') {
      if (CurVariant == ~0U)
        throw "'}' character found outside of a variant in instruction '"
          + CGI.TheDef->getName() + "'!";
      ++LastEmitted;
      CurVariant = ~0U;
    } else if (DollarPos+1 != AsmString.size() &&
               AsmString[DollarPos+1] == '$') {
      if (CurVariant == Variant || CurVariant == ~0U) 
        AddLiteralString("$");  // "$$" -> $
      LastEmitted = DollarPos+2;
    } else {
      // Get the name of the variable.
      std::string::size_type VarEnd = DollarPos+1;

      // handle ${foo}bar as $foo by detecting whether the character following
      // the dollar sign is a curly brace.  If so, advance VarEnd and DollarPos
      // so the variable name does not contain the leading curly brace.
      bool hasCurlyBraces = false;
      if (VarEnd < AsmString.size() && '{' == AsmString[VarEnd]) {
        hasCurlyBraces = true;
        ++DollarPos;
        ++VarEnd;
      }

      while (VarEnd < AsmString.size() && isIdentChar(AsmString[VarEnd]))
        ++VarEnd;
      std::string VarName(AsmString.begin()+DollarPos+1,
                          AsmString.begin()+VarEnd);

      // Modifier - Support ${foo:modifier} syntax, where "modifier" is passed
      // into printOperand.  Also support ${:feature}, which is passed into
      // PrintSpecial.
      std::string Modifier;
      
      // In order to avoid starting the next string at the terminating curly
      // brace, advance the end position past it if we found an opening curly
      // brace.
      if (hasCurlyBraces) {
        if (VarEnd >= AsmString.size())
          throw "Reached end of string before terminating curly brace in '"
                + CGI.TheDef->getName() + "'";
        
        // Look for a modifier string.
        if (AsmString[VarEnd] == ':') {
          ++VarEnd;
          if (VarEnd >= AsmString.size())
            throw "Reached end of string before terminating curly brace in '"
              + CGI.TheDef->getName() + "'";
          
          unsigned ModifierStart = VarEnd;
          while (VarEnd < AsmString.size() && isIdentChar(AsmString[VarEnd]))
            ++VarEnd;
          Modifier = std::string(AsmString.begin()+ModifierStart,
                                 AsmString.begin()+VarEnd);
          if (Modifier.empty())
            throw "Bad operand modifier name in '"+ CGI.TheDef->getName() + "'";
        }
        
        if (AsmString[VarEnd] != '}')
          throw "Variable name beginning with '{' did not end with '}' in '"
                + CGI.TheDef->getName() + "'";
        ++VarEnd;
      }
      if (VarName.empty() && Modifier.empty())
        throw "Stray '$' in '" + CGI.TheDef->getName() +
              "' asm string, maybe you want $$?";

      if (VarName.empty()) {
        // Just a modifier, pass this into PrintSpecial.
        Operands.push_back(AsmWriterOperand("PrintSpecial", ~0U, Modifier));
      } else {
        // Otherwise, normal operand.
        unsigned OpNo = CGI.getOperandNamed(VarName);
        CodeGenInstruction::OperandInfo OpInfo = CGI.OperandList[OpNo];

        if (CurVariant == Variant || CurVariant == ~0U) {
          unsigned MIOp = OpInfo.MIOperandNo;
          Operands.push_back(AsmWriterOperand(OpInfo.PrinterMethodName, MIOp,
                                              Modifier));
        }
      }
      LastEmitted = VarEnd;
    }
  }

  AddLiteralString("\\n");
}

/// MatchesAllButOneOp - If this instruction is exactly identical to the
/// specified instruction except for one differing operand, return the differing
/// operand number.  If more than one operand mismatches, return ~1, otherwise
/// if the instructions are identical return ~0.
unsigned AsmWriterInst::MatchesAllButOneOp(const AsmWriterInst &Other)const{
  if (Operands.size() != Other.Operands.size()) return ~1;

  unsigned MismatchOperand = ~0U;
  for (unsigned i = 0, e = Operands.size(); i != e; ++i) {
    if (Operands[i] != Other.Operands[i]) {
      if (MismatchOperand != ~0U)  // Already have one mismatch?
        return ~1U;
      else
        MismatchOperand = i;
    }
  }
  return MismatchOperand;
}

static void PrintCases(std::vector<std::pair<std::string,
                       AsmWriterOperand> > &OpsToPrint, std::ostream &O) {
  O << "    case " << OpsToPrint.back().first << ": ";
  AsmWriterOperand TheOp = OpsToPrint.back().second;
  OpsToPrint.pop_back();

  // Check to see if any other operands are identical in this list, and if so,
  // emit a case label for them.
  for (unsigned i = OpsToPrint.size(); i != 0; --i)
    if (OpsToPrint[i-1].second == TheOp) {
      O << "\n    case " << OpsToPrint[i-1].first << ": ";
      OpsToPrint.erase(OpsToPrint.begin()+i-1);
    }

  // Finally, emit the code.
  O << TheOp.getCode();
  O << "break;\n";
}


/// EmitInstructions - Emit the last instruction in the vector and any other
/// instructions that are suitably similar to it.
static void EmitInstructions(std::vector<AsmWriterInst> &Insts,
                             std::ostream &O) {
  AsmWriterInst FirstInst = Insts.back();
  Insts.pop_back();

  std::vector<AsmWriterInst> SimilarInsts;
  unsigned DifferingOperand = ~0;
  for (unsigned i = Insts.size(); i != 0; --i) {
    unsigned DiffOp = Insts[i-1].MatchesAllButOneOp(FirstInst);
    if (DiffOp != ~1U) {
      if (DifferingOperand == ~0U)  // First match!
        DifferingOperand = DiffOp;

      // If this differs in the same operand as the rest of the instructions in
      // this class, move it to the SimilarInsts list.
      if (DifferingOperand == DiffOp || DiffOp == ~0U) {
        SimilarInsts.push_back(Insts[i-1]);
        Insts.erase(Insts.begin()+i-1);
      }
    }
  }

  O << "  case " << FirstInst.CGI->Namespace << "::"
    << FirstInst.CGI->TheDef->getName() << ":\n";
  for (unsigned i = 0, e = SimilarInsts.size(); i != e; ++i)
    O << "  case " << SimilarInsts[i].CGI->Namespace << "::"
      << SimilarInsts[i].CGI->TheDef->getName() << ":\n";
  for (unsigned i = 0, e = FirstInst.Operands.size(); i != e; ++i) {
    if (i != DifferingOperand) {
      // If the operand is the same for all instructions, just print it.
      O << "    " << FirstInst.Operands[i].getCode();
    } else {
      // If this is the operand that varies between all of the instructions,
      // emit a switch for just this operand now.
      O << "    switch (MI->getOpcode()) {\n";
      std::vector<std::pair<std::string, AsmWriterOperand> > OpsToPrint;
      OpsToPrint.push_back(std::make_pair(FirstInst.CGI->Namespace + "::" +
                                          FirstInst.CGI->TheDef->getName(),
                                          FirstInst.Operands[i]));

      for (unsigned si = 0, e = SimilarInsts.size(); si != e; ++si) {
        AsmWriterInst &AWI = SimilarInsts[si];
        OpsToPrint.push_back(std::make_pair(AWI.CGI->Namespace+"::"+
                                            AWI.CGI->TheDef->getName(),
                                            AWI.Operands[i]));
      }
      std::reverse(OpsToPrint.begin(), OpsToPrint.end());
      while (!OpsToPrint.empty())
        PrintCases(OpsToPrint, O);
      O << "    }";
    }
    O << "\n";
  }

  O << "    break;\n";
}

void AsmWriterEmitter::
FindUniqueOperandCommands(std::vector<std::string> &UniqueOperandCommands, 
                          std::vector<unsigned> &InstIdxs,
                          std::vector<unsigned> &InstOpsUsed) const {
  InstIdxs.assign(NumberedInstructions.size(), ~0U);
  
  // This vector parallels UniqueOperandCommands, keeping track of which
  // instructions each case are used for.  It is a comma separated string of
  // enums.
  std::vector<std::string> InstrsForCase;
  InstrsForCase.resize(UniqueOperandCommands.size());
  InstOpsUsed.assign(UniqueOperandCommands.size(), 0);
  
  for (unsigned i = 0, e = NumberedInstructions.size(); i != e; ++i) {
    const AsmWriterInst *Inst = getAsmWriterInstByID(i);
    if (Inst == 0) continue;  // PHI, INLINEASM, DBG_LABEL, etc.
    
    std::string Command;
    if (Inst->Operands.empty())
      continue;   // Instruction already done.

    Command = "    " + Inst->Operands[0].getCode() + "\n";

    // If this is the last operand, emit a return.
    if (Inst->Operands.size() == 1)
      Command += "    return true;\n";
    
    // Check to see if we already have 'Command' in UniqueOperandCommands.
    // If not, add it.
    bool FoundIt = false;
    for (unsigned idx = 0, e = UniqueOperandCommands.size(); idx != e; ++idx)
      if (UniqueOperandCommands[idx] == Command) {
        InstIdxs[i] = idx;
        InstrsForCase[idx] += ", ";
        InstrsForCase[idx] += Inst->CGI->TheDef->getName();
        FoundIt = true;
        break;
      }
    if (!FoundIt) {
      InstIdxs[i] = UniqueOperandCommands.size();
      UniqueOperandCommands.push_back(Command);
      InstrsForCase.push_back(Inst->CGI->TheDef->getName());

      // This command matches one operand so far.
      InstOpsUsed.push_back(1);
    }
  }
  
  // For each entry of UniqueOperandCommands, there is a set of instructions
  // that uses it.  If the next command of all instructions in the set are
  // identical, fold it into the command.
  for (unsigned CommandIdx = 0, e = UniqueOperandCommands.size();
       CommandIdx != e; ++CommandIdx) {
    
    for (unsigned Op = 1; ; ++Op) {
      // Scan for the first instruction in the set.
      std::vector<unsigned>::iterator NIT =
        std::find(InstIdxs.begin(), InstIdxs.end(), CommandIdx);
      if (NIT == InstIdxs.end()) break;  // No commonality.

      // If this instruction has no more operands, we isn't anything to merge
      // into this command.
      const AsmWriterInst *FirstInst = 
        getAsmWriterInstByID(NIT-InstIdxs.begin());
      if (!FirstInst || FirstInst->Operands.size() == Op)
        break;

      // Otherwise, scan to see if all of the other instructions in this command
      // set share the operand.
      bool AllSame = true;
      
      for (NIT = std::find(NIT+1, InstIdxs.end(), CommandIdx);
           NIT != InstIdxs.end();
           NIT = std::find(NIT+1, InstIdxs.end(), CommandIdx)) {
        // Okay, found another instruction in this command set.  If the operand
        // matches, we're ok, otherwise bail out.
        const AsmWriterInst *OtherInst = 
          getAsmWriterInstByID(NIT-InstIdxs.begin());
        if (!OtherInst || OtherInst->Operands.size() == Op ||
            OtherInst->Operands[Op] != FirstInst->Operands[Op]) {
          AllSame = false;
          break;
        }
      }
      if (!AllSame) break;
      
      // Okay, everything in this command set has the same next operand.  Add it
      // to UniqueOperandCommands and remember that it was consumed.
      std::string Command = "    " + FirstInst->Operands[Op].getCode() + "\n";
      
      // If this is the last operand, emit a return after the code.
      if (FirstInst->Operands.size() == Op+1)
        Command += "    return true;\n";
      
      UniqueOperandCommands[CommandIdx] += Command;
      InstOpsUsed[CommandIdx]++;
    }
  }
  
  // Prepend some of the instructions each case is used for onto the case val.
  for (unsigned i = 0, e = InstrsForCase.size(); i != e; ++i) {
    std::string Instrs = InstrsForCase[i];
    if (Instrs.size() > 70) {
      Instrs.erase(Instrs.begin()+70, Instrs.end());
      Instrs += "...";
    }
    
    if (!Instrs.empty())
      UniqueOperandCommands[i] = "    // " + Instrs + "\n" + 
        UniqueOperandCommands[i];
  }
}



void AsmWriterEmitter::run(std::ostream &O) {
  EmitSourceFileHeader("Assembly Writer Source Fragment", O);

  CodeGenTarget Target;
  Record *AsmWriter = Target.getAsmWriter();
  std::string ClassName = AsmWriter->getValueAsString("AsmWriterClassName");
  unsigned Variant = AsmWriter->getValueAsInt("Variant");

  O <<
  "/// printInstruction - This method is automatically generated by tablegen\n"
  "/// from the instruction set description.  This method returns true if the\n"
  "/// machine instruction was sufficiently described to print it, otherwise\n"
  "/// it returns false.\n"
    "bool " << Target.getName() << ClassName
            << "::printInstruction(const MachineInstr *MI) {\n";

  std::vector<AsmWriterInst> Instructions;

  for (CodeGenTarget::inst_iterator I = Target.inst_begin(),
         E = Target.inst_end(); I != E; ++I)
    if (!I->second.AsmString.empty())
      Instructions.push_back(AsmWriterInst(I->second, Variant));

  // Get the instruction numbering.
  Target.getInstructionsByEnumValue(NumberedInstructions);
  
  // Compute the CodeGenInstruction -> AsmWriterInst mapping.  Note that not
  // all machine instructions are necessarily being printed, so there may be
  // target instructions not in this map.
  for (unsigned i = 0, e = Instructions.size(); i != e; ++i)
    CGIAWIMap.insert(std::make_pair(Instructions[i].CGI, &Instructions[i]));

  // Build an aggregate string, and build a table of offsets into it.
  std::map<std::string, unsigned> StringOffset;
  std::string AggregateString;
  AggregateString.push_back(0);  // "\0"
  AggregateString.push_back(0);  // "\0"
  
  /// OpcodeInfo - This encodes the index of the string to use for the first
  /// chunk of the output as well as indices used for operand printing.
  std::vector<unsigned> OpcodeInfo;
  
  unsigned MaxStringIdx = 0;
  for (unsigned i = 0, e = NumberedInstructions.size(); i != e; ++i) {
    AsmWriterInst *AWI = CGIAWIMap[NumberedInstructions[i]];
    unsigned Idx;
    if (AWI == 0) {
      // Something not handled by the asmwriter printer.
      Idx = 0;
    } else if (AWI->Operands[0].OperandType != 
                        AsmWriterOperand::isLiteralTextOperand ||
               AWI->Operands[0].Str.empty()) {
      // Something handled by the asmwriter printer, but with no leading string.
      Idx = 1;
    } else {
      unsigned &Entry = StringOffset[AWI->Operands[0].Str];
      if (Entry == 0) {
        // Add the string to the aggregate if this is the first time found.
        MaxStringIdx = Entry = AggregateString.size();
        std::string Str = AWI->Operands[0].Str;
        UnescapeString(Str);
        AggregateString += Str;
        AggregateString += '\0';
      }
      Idx = Entry;

      // Nuke the string from the operand list.  It is now handled!
      AWI->Operands.erase(AWI->Operands.begin());
    }
    OpcodeInfo.push_back(Idx);
  }
  
  // Figure out how many bits we used for the string index.
  unsigned AsmStrBits = Log2_32_Ceil(MaxStringIdx+1);
  
  // To reduce code size, we compactify common instructions into a few bits
  // in the opcode-indexed table.
  unsigned BitsLeft = 32-AsmStrBits;

  std::vector<std::vector<std::string> > TableDrivenOperandPrinters;
  
  bool isFirst = true;
  while (1) {
    std::vector<std::string> UniqueOperandCommands;

    // For the first operand check, add a default value for instructions with
    // just opcode strings to use.
    if (isFirst) {
      UniqueOperandCommands.push_back("    return true;\n");
      isFirst = false;
    }
    
    std::vector<unsigned> InstIdxs;
    std::vector<unsigned> NumInstOpsHandled;
    FindUniqueOperandCommands(UniqueOperandCommands, InstIdxs,
                              NumInstOpsHandled);
    
    // If we ran out of operands to print, we're done.
    if (UniqueOperandCommands.empty()) break;
    
    // Compute the number of bits we need to represent these cases, this is
    // ceil(log2(numentries)).
    unsigned NumBits = Log2_32_Ceil(UniqueOperandCommands.size());
    
    // If we don't have enough bits for this operand, don't include it.
    if (NumBits > BitsLeft) {
      DOUT << "Not enough bits to densely encode " << NumBits
           << " more bits\n";
      break;
    }
    
    // Otherwise, we can include this in the initial lookup table.  Add it in.
    BitsLeft -= NumBits;
    for (unsigned i = 0, e = InstIdxs.size(); i != e; ++i)
      if (InstIdxs[i] != ~0U)
        OpcodeInfo[i] |= InstIdxs[i] << (BitsLeft+AsmStrBits);
    
    // Remove the info about this operand.
    for (unsigned i = 0, e = NumberedInstructions.size(); i != e; ++i) {
      if (AsmWriterInst *Inst = getAsmWriterInstByID(i))
        if (!Inst->Operands.empty()) {
          unsigned NumOps = NumInstOpsHandled[InstIdxs[i]];
          assert(NumOps <= Inst->Operands.size() &&
                 "Can't remove this many ops!");
          Inst->Operands.erase(Inst->Operands.begin(),
                               Inst->Operands.begin()+NumOps);
        }
    }
    
    // Remember the handlers for this set of operands.
    TableDrivenOperandPrinters.push_back(UniqueOperandCommands);
  }
  
  
  
  O<<"  static const unsigned OpInfo[] = {\n";
  for (unsigned i = 0, e = NumberedInstructions.size(); i != e; ++i) {
    O << "    " << OpcodeInfo[i] << "U,\t// "
      << NumberedInstructions[i]->TheDef->getName() << "\n";
  }
  // Add a dummy entry so the array init doesn't end with a comma.
  O << "    0U\n";
  O << "  };\n\n";
  
  // Emit the string itself.
  O << "  const char *AsmStrs = \n    \"";
  unsigned CharsPrinted = 0;
  EscapeString(AggregateString);
  for (unsigned i = 0, e = AggregateString.size(); i != e; ++i) {
    if (CharsPrinted > 70) {
      O << "\"\n    \"";
      CharsPrinted = 0;
    }
    O << AggregateString[i];
    ++CharsPrinted;
    
    // Print escape sequences all together.
    if (AggregateString[i] == '\\') {
      assert(i+1 < AggregateString.size() && "Incomplete escape sequence!");
      if (isdigit(AggregateString[i+1])) {
        assert(isdigit(AggregateString[i+2]) && isdigit(AggregateString[i+3]) &&
               "Expected 3 digit octal escape!");
        O << AggregateString[++i];
        O << AggregateString[++i];
        O << AggregateString[++i];
        CharsPrinted += 3;
      } else {
        O << AggregateString[++i];
        ++CharsPrinted;
      }
    }
  }
  O << "\";\n\n";

  O << "  if (MI->getOpcode() == TargetInstrInfo::INLINEASM) {\n"
    << "    O << \"\\t\";\n"
    << "    printInlineAsm(MI);\n"
    << "    return true;\n"
    << "  } else if (MI->isLabel()) {\n"
    << "    printLabel(MI);\n"
    << "    return true;\n"
    << "  } else if (MI->getOpcode() == TargetInstrInfo::DECLARE) {\n"
    << "    printDeclare(MI);\n"
    << "    return true;\n"
    << "  } else if (MI->getOpcode() == TargetInstrInfo::IMPLICIT_DEF) {\n"
    << "    printImplicitDef(MI);\n"
    << "    return true;\n"
    << "  }\n\n";
  
  O << "  O << \"\\t\";\n\n";

  O << "  // Emit the opcode for the instruction.\n"
    << "  unsigned Bits = OpInfo[MI->getOpcode()];\n"
    << "  if (Bits == 0) return false;\n"
    << "  O << AsmStrs+(Bits & " << (1 << AsmStrBits)-1 << ");\n\n";

  // Output the table driven operand information.
  BitsLeft = 32-AsmStrBits;
  for (unsigned i = 0, e = TableDrivenOperandPrinters.size(); i != e; ++i) {
    std::vector<std::string> &Commands = TableDrivenOperandPrinters[i];

    // Compute the number of bits we need to represent these cases, this is
    // ceil(log2(numentries)).
    unsigned NumBits = Log2_32_Ceil(Commands.size());
    assert(NumBits <= BitsLeft && "consistency error");
    
    // Emit code to extract this field from Bits.
    BitsLeft -= NumBits;
    
    O << "\n  // Fragment " << i << " encoded into " << NumBits
      << " bits for " << Commands.size() << " unique commands.\n";
    
    if (Commands.size() == 2) {
      // Emit two possibilitys with if/else.
      O << "  if ((Bits >> " << (BitsLeft+AsmStrBits) << ") & "
        << ((1 << NumBits)-1) << ") {\n"
        << Commands[1]
        << "  } else {\n"
        << Commands[0]
        << "  }\n\n";
    } else {
      O << "  switch ((Bits >> " << (BitsLeft+AsmStrBits) << ") & "
        << ((1 << NumBits)-1) << ") {\n"
        << "  default:   // unreachable.\n";
      
      // Print out all the cases.
      for (unsigned i = 0, e = Commands.size(); i != e; ++i) {
        O << "  case " << i << ":\n";
        O << Commands[i];
        O << "    break;\n";
      }
      O << "  }\n\n";
    }
  }
  
  // Okay, delete instructions with no operand info left.
  for (unsigned i = 0, e = Instructions.size(); i != e; ++i) {
    // Entire instruction has been emitted?
    AsmWriterInst &Inst = Instructions[i];
    if (Inst.Operands.empty()) {
      Instructions.erase(Instructions.begin()+i);
      --i; --e;
    }
  }

    
  // Because this is a vector, we want to emit from the end.  Reverse all of the
  // elements in the vector.
  std::reverse(Instructions.begin(), Instructions.end());
  
  if (!Instructions.empty()) {
    // Find the opcode # of inline asm.
    O << "  switch (MI->getOpcode()) {\n";
    while (!Instructions.empty())
      EmitInstructions(Instructions, O);

    O << "  }\n";
    O << "  return true;\n";
  }
  
  O << "}\n";
}