llvm.org GIT mirror llvm / 43ad6b3 lib / ExecutionEngine / ExecutionEngine.cpp
43ad6b3

Tree @43ad6b3 (Download .tar.gz)

ExecutionEngine.cpp @43ad6b3raw · history · blame

  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
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
//===-- ExecutionEngine.cpp - Common Implementation shared by EEs ---------===//
//
//                     The LLVM Compiler Infrastructure
//
// This file was developed by the LLVM research group and is distributed under
// the University of Illinois Open Source License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This file defines the common interface used by the various execution engine
// subclasses.
//
//===----------------------------------------------------------------------===//

#define DEBUG_TYPE "jit"
#include "llvm/Constants.h"
#include "llvm/DerivedTypes.h"
#include "llvm/Module.h"
#include "llvm/ModuleProvider.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/Config/alloca.h"
#include "llvm/ExecutionEngine/ExecutionEngine.h"
#include "llvm/ExecutionEngine/GenericValue.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/MutexGuard.h"
#include "llvm/System/DynamicLibrary.h"
#include "llvm/System/Host.h"
#include "llvm/Target/TargetData.h"
#include <math.h>
using namespace llvm;

STATISTIC(NumInitBytes, "Number of bytes of global vars initialized");
STATISTIC(NumGlobals  , "Number of global vars initialized");

ExecutionEngine::EECtorFn ExecutionEngine::JITCtor = 0;
ExecutionEngine::EECtorFn ExecutionEngine::InterpCtor = 0;

ExecutionEngine::ExecutionEngine(ModuleProvider *P) : LazyFunctionCreator(0) {
  LazyCompilationDisabled = false;
  Modules.push_back(P);
  assert(P && "ModuleProvider is null?");
}

ExecutionEngine::~ExecutionEngine() {
  clearAllGlobalMappings();
  for (unsigned i = 0, e = Modules.size(); i != e; ++i)
    delete Modules[i];
}

/// removeModuleProvider - Remove a ModuleProvider from the list of modules.
/// Release module from ModuleProvider.
Module* ExecutionEngine::removeModuleProvider(ModuleProvider *P, 
                                              std::string *ErrInfo) {
  for(SmallVector<ModuleProvider *, 1>::iterator I = Modules.begin(), 
        E = Modules.end(); I != E; ++I) {
    ModuleProvider *MP = *I;
    if (MP == P) {
      Modules.erase(I);
      return MP->releaseModule(ErrInfo);
    }
  }
  return NULL;
}

/// FindFunctionNamed - Search all of the active modules to find the one that
/// defines FnName.  This is very slow operation and shouldn't be used for
/// general code.
Function *ExecutionEngine::FindFunctionNamed(const char *FnName) {
  for (unsigned i = 0, e = Modules.size(); i != e; ++i) {
    if (Function *F = Modules[i]->getModule()->getFunction(FnName))
      return F;
  }
  return 0;
}


/// addGlobalMapping - Tell the execution engine that the specified global is
/// at the specified location.  This is used internally as functions are JIT'd
/// and as global variables are laid out in memory.  It can and should also be
/// used by clients of the EE that want to have an LLVM global overlay
/// existing data in memory.
void ExecutionEngine::addGlobalMapping(const GlobalValue *GV, void *Addr) {
  MutexGuard locked(lock);
  
  void *&CurVal = state.getGlobalAddressMap(locked)[GV];
  assert((CurVal == 0 || Addr == 0) && "GlobalMapping already established!");
  CurVal = Addr;
  
  // If we are using the reverse mapping, add it too
  if (!state.getGlobalAddressReverseMap(locked).empty()) {
    const GlobalValue *&V = state.getGlobalAddressReverseMap(locked)[Addr];
    assert((V == 0 || GV == 0) && "GlobalMapping already established!");
    V = GV;
  }
}

/// clearAllGlobalMappings - Clear all global mappings and start over again
/// use in dynamic compilation scenarios when you want to move globals
void ExecutionEngine::clearAllGlobalMappings() {
  MutexGuard locked(lock);
  
  state.getGlobalAddressMap(locked).clear();
  state.getGlobalAddressReverseMap(locked).clear();
}

/// updateGlobalMapping - Replace an existing mapping for GV with a new
/// address.  This updates both maps as required.  If "Addr" is null, the
/// entry for the global is removed from the mappings.
void ExecutionEngine::updateGlobalMapping(const GlobalValue *GV, void *Addr) {
  MutexGuard locked(lock);
  
  // Deleting from the mapping?
  if (Addr == 0) {
    state.getGlobalAddressMap(locked).erase(GV);
    if (!state.getGlobalAddressReverseMap(locked).empty())
      state.getGlobalAddressReverseMap(locked).erase(Addr);
    return;
  }
  
  void *&CurVal = state.getGlobalAddressMap(locked)[GV];
  if (CurVal && !state.getGlobalAddressReverseMap(locked).empty())
    state.getGlobalAddressReverseMap(locked).erase(CurVal);
  CurVal = Addr;
  
  // If we are using the reverse mapping, add it too
  if (!state.getGlobalAddressReverseMap(locked).empty()) {
    const GlobalValue *&V = state.getGlobalAddressReverseMap(locked)[Addr];
    assert((V == 0 || GV == 0) && "GlobalMapping already established!");
    V = GV;
  }
}

/// getPointerToGlobalIfAvailable - This returns the address of the specified
/// global value if it is has already been codegen'd, otherwise it returns null.
///
void *ExecutionEngine::getPointerToGlobalIfAvailable(const GlobalValue *GV) {
  MutexGuard locked(lock);
  
  std::map<const GlobalValue*, void*>::iterator I =
  state.getGlobalAddressMap(locked).find(GV);
  return I != state.getGlobalAddressMap(locked).end() ? I->second : 0;
}

/// getGlobalValueAtAddress - Return the LLVM global value object that starts
/// at the specified address.
///
const GlobalValue *ExecutionEngine::getGlobalValueAtAddress(void *Addr) {
  MutexGuard locked(lock);

  // If we haven't computed the reverse mapping yet, do so first.
  if (state.getGlobalAddressReverseMap(locked).empty()) {
    for (std::map<const GlobalValue*, void *>::iterator
         I = state.getGlobalAddressMap(locked).begin(),
         E = state.getGlobalAddressMap(locked).end(); I != E; ++I)
      state.getGlobalAddressReverseMap(locked).insert(std::make_pair(I->second,
                                                                     I->first));
  }

  std::map<void *, const GlobalValue*>::iterator I =
    state.getGlobalAddressReverseMap(locked).find(Addr);
  return I != state.getGlobalAddressReverseMap(locked).end() ? I->second : 0;
}

// CreateArgv - Turn a vector of strings into a nice argv style array of
// pointers to null terminated strings.
//
static void *CreateArgv(ExecutionEngine *EE,
                        const std::vector<std::string> &InputArgv) {
  unsigned PtrSize = EE->getTargetData()->getPointerSize();
  char *Result = new char[(InputArgv.size()+1)*PtrSize];

  DOUT << "ARGV = " << (void*)Result << "\n";
  const Type *SBytePtr = PointerType::getUnqual(Type::Int8Ty);

  for (unsigned i = 0; i != InputArgv.size(); ++i) {
    unsigned Size = InputArgv[i].size()+1;
    char *Dest = new char[Size];
    DOUT << "ARGV[" << i << "] = " << (void*)Dest << "\n";

    std::copy(InputArgv[i].begin(), InputArgv[i].end(), Dest);
    Dest[Size-1] = 0;

    // Endian safe: Result[i] = (PointerTy)Dest;
    EE->StoreValueToMemory(PTOGV(Dest), (GenericValue*)(Result+i*PtrSize),
                           SBytePtr);
  }

  // Null terminate it
  EE->StoreValueToMemory(PTOGV(0),
                         (GenericValue*)(Result+InputArgv.size()*PtrSize),
                         SBytePtr);
  return Result;
}


/// runStaticConstructorsDestructors - This method is used to execute all of
/// the static constructors or destructors for a program, depending on the
/// value of isDtors.
void ExecutionEngine::runStaticConstructorsDestructors(bool isDtors) {
  const char *Name = isDtors ? "llvm.global_dtors" : "llvm.global_ctors";
  
  // Execute global ctors/dtors for each module in the program.
  for (unsigned m = 0, e = Modules.size(); m != e; ++m) {
    GlobalVariable *GV = Modules[m]->getModule()->getNamedGlobal(Name);

    // If this global has internal linkage, or if it has a use, then it must be
    // an old-style (llvmgcc3) static ctor with __main linked in and in use.  If
    // this is the case, don't execute any of the global ctors, __main will do
    // it.
    if (!GV || GV->isDeclaration() || GV->hasInternalLinkage()) continue;
  
    // Should be an array of '{ int, void ()* }' structs.  The first value is
    // the init priority, which we ignore.
    ConstantArray *InitList = dyn_cast<ConstantArray>(GV->getInitializer());
    if (!InitList) continue;
    for (unsigned i = 0, e = InitList->getNumOperands(); i != e; ++i)
      if (ConstantStruct *CS = 
          dyn_cast<ConstantStruct>(InitList->getOperand(i))) {
        if (CS->getNumOperands() != 2) break; // Not array of 2-element structs.
      
        Constant *FP = CS->getOperand(1);
        if (FP->isNullValue())
          break;  // Found a null terminator, exit.
      
        if (ConstantExpr *CE = dyn_cast<ConstantExpr>(FP))
          if (CE->isCast())
            FP = CE->getOperand(0);
        if (Function *F = dyn_cast<Function>(FP)) {
          // Execute the ctor/dtor function!
          runFunction(F, std::vector<GenericValue>());
        }
      }
  }
}

/// isTargetNullPtr - Return whether the target pointer stored at Loc is null.
static bool isTargetNullPtr(ExecutionEngine *EE, void *Loc) {
  unsigned PtrSize = EE->getTargetData()->getPointerSize();
  for (unsigned i = 0; i < PtrSize; ++i)
    if (*(i + (uint8_t*)Loc))
      return false;
  return true;
}

/// runFunctionAsMain - This is a helper function which wraps runFunction to
/// handle the common task of starting up main with the specified argc, argv,
/// and envp parameters.
int ExecutionEngine::runFunctionAsMain(Function *Fn,
                                       const std::vector<std::string> &argv,
                                       const char * const * envp) {
  std::vector<GenericValue> GVArgs;
  GenericValue GVArgc;
  GVArgc.IntVal = APInt(32, argv.size());

  // Check main() type
  unsigned NumArgs = Fn->getFunctionType()->getNumParams();
  const FunctionType *FTy = Fn->getFunctionType();
  const Type* PPInt8Ty = 
    PointerType::getUnqual(PointerType::getUnqual(Type::Int8Ty));
  switch (NumArgs) {
  case 3:
   if (FTy->getParamType(2) != PPInt8Ty) {
     cerr << "Invalid type for third argument of main() supplied\n";
     abort();
   }
   // FALLS THROUGH
  case 2:
   if (FTy->getParamType(1) != PPInt8Ty) {
     cerr << "Invalid type for second argument of main() supplied\n";
     abort();
   }
   // FALLS THROUGH
  case 1:
   if (FTy->getParamType(0) != Type::Int32Ty) {
     cerr << "Invalid type for first argument of main() supplied\n";
     abort();
   }
   // FALLS THROUGH
  case 0:
   if (FTy->getReturnType() != Type::Int32Ty &&
       FTy->getReturnType() != Type::VoidTy) {
     cerr << "Invalid return type of main() supplied\n";
     abort();
   }
   break;
  default:
   cerr << "Invalid number of arguments of main() supplied\n";
   abort();
  }
  
  if (NumArgs) {
    GVArgs.push_back(GVArgc); // Arg #0 = argc.
    if (NumArgs > 1) {
      GVArgs.push_back(PTOGV(CreateArgv(this, argv))); // Arg #1 = argv.
      assert(!isTargetNullPtr(this, GVTOP(GVArgs[1])) &&
             "argv[0] was null after CreateArgv");
      if (NumArgs > 2) {
        std::vector<std::string> EnvVars;
        for (unsigned i = 0; envp[i]; ++i)
          EnvVars.push_back(envp[i]);
        GVArgs.push_back(PTOGV(CreateArgv(this, EnvVars))); // Arg #2 = envp.
      }
    }
  }
  return runFunction(Fn, GVArgs).IntVal.getZExtValue();
}

/// If possible, create a JIT, unless the caller specifically requests an
/// Interpreter or there's an error. If even an Interpreter cannot be created,
/// NULL is returned.
///
ExecutionEngine *ExecutionEngine::create(ModuleProvider *MP,
                                         bool ForceInterpreter,
                                         std::string *ErrorStr) {
  ExecutionEngine *EE = 0;

  // Unless the interpreter was explicitly selected, try making a JIT.
  if (!ForceInterpreter && JITCtor)
    EE = JITCtor(MP, ErrorStr);

  // If we can't make a JIT, make an interpreter instead.
  if (EE == 0 && InterpCtor)
    EE = InterpCtor(MP, ErrorStr);

  if (EE) {
    // Make sure we can resolve symbols in the program as well. The zero arg
    // to the function tells DynamicLibrary to load the program, not a library.
    if (sys::DynamicLibrary::LoadLibraryPermanently(0, ErrorStr)) {
      delete EE;
      return 0;
    }
  }

  return EE;
}

ExecutionEngine *ExecutionEngine::create(Module *M) {
  return create(new ExistingModuleProvider(M));
}

/// getPointerToGlobal - This returns the address of the specified global
/// value.  This may involve code generation if it's a function.
///
void *ExecutionEngine::getPointerToGlobal(const GlobalValue *GV) {
  if (Function *F = const_cast<Function*>(dyn_cast<Function>(GV)))
    return getPointerToFunction(F);

  MutexGuard locked(lock);
  void *p = state.getGlobalAddressMap(locked)[GV];
  if (p)
    return p;

  // Global variable might have been added since interpreter started.
  if (GlobalVariable *GVar =
          const_cast<GlobalVariable *>(dyn_cast<GlobalVariable>(GV)))
    EmitGlobalVariable(GVar);
  else
    assert(0 && "Global hasn't had an address allocated yet!");
  return state.getGlobalAddressMap(locked)[GV];
}

/// This function converts a Constant* into a GenericValue. The interesting 
/// part is if C is a ConstantExpr.
/// @brief Get a GenericValue for a Constant*
GenericValue ExecutionEngine::getConstantValue(const Constant *C) {
  // If its undefined, return the garbage.
  if (isa<UndefValue>(C)) 
    return GenericValue();

  // If the value is a ConstantExpr
  if (const ConstantExpr *CE = dyn_cast<ConstantExpr>(C)) {
    Constant *Op0 = CE->getOperand(0);
    switch (CE->getOpcode()) {
    case Instruction::GetElementPtr: {
      // Compute the index 
      GenericValue Result = getConstantValue(Op0);
      SmallVector<Value*, 8> Indices(CE->op_begin()+1, CE->op_end());
      uint64_t Offset =
        TD->getIndexedOffset(Op0->getType(), &Indices[0], Indices.size());

      char* tmp = (char*) Result.PointerVal;
      Result = PTOGV(tmp + Offset);
      return Result;
    }
    case Instruction::Trunc: {
      GenericValue GV = getConstantValue(Op0);
      uint32_t BitWidth = cast<IntegerType>(CE->getType())->getBitWidth();
      GV.IntVal = GV.IntVal.trunc(BitWidth);
      return GV;
    }
    case Instruction::ZExt: {
      GenericValue GV = getConstantValue(Op0);
      uint32_t BitWidth = cast<IntegerType>(CE->getType())->getBitWidth();
      GV.IntVal = GV.IntVal.zext(BitWidth);
      return GV;
    }
    case Instruction::SExt: {
      GenericValue GV = getConstantValue(Op0);
      uint32_t BitWidth = cast<IntegerType>(CE->getType())->getBitWidth();
      GV.IntVal = GV.IntVal.sext(BitWidth);
      return GV;
    }
    case Instruction::FPTrunc: {
      // FIXME long double
      GenericValue GV = getConstantValue(Op0);
      GV.FloatVal = float(GV.DoubleVal);
      return GV;
    }
    case Instruction::FPExt:{
      // FIXME long double
      GenericValue GV = getConstantValue(Op0);
      GV.DoubleVal = double(GV.FloatVal);
      return GV;
    }
    case Instruction::UIToFP: {
      GenericValue GV = getConstantValue(Op0);
      if (CE->getType() == Type::FloatTy)
        GV.FloatVal = float(GV.IntVal.roundToDouble());
      else if (CE->getType() == Type::DoubleTy)
        GV.DoubleVal = GV.IntVal.roundToDouble();
      else if (CE->getType() == Type::X86_FP80Ty) {
        const uint64_t zero[] = {0, 0};
        APFloat apf = APFloat(APInt(80, 2, zero));
        (void)apf.convertFromZeroExtendedInteger(GV.IntVal.getRawData(), 
                               GV.IntVal.getBitWidth(), false,
                               APFloat::rmNearestTiesToEven);
        GV.IntVal = apf.convertToAPInt();
      }
      return GV;
    }
    case Instruction::SIToFP: {
      GenericValue GV = getConstantValue(Op0);
      if (CE->getType() == Type::FloatTy)
        GV.FloatVal = float(GV.IntVal.signedRoundToDouble());
      else if (CE->getType() == Type::DoubleTy)
        GV.DoubleVal = GV.IntVal.signedRoundToDouble();
      else if (CE->getType() == Type::X86_FP80Ty) {
        const uint64_t zero[] = { 0, 0};
        APFloat apf = APFloat(APInt(80, 2, zero));
        (void)apf.convertFromZeroExtendedInteger(GV.IntVal.getRawData(), 
                               GV.IntVal.getBitWidth(), true,
                               APFloat::rmNearestTiesToEven);
        GV.IntVal = apf.convertToAPInt();
      }
      return GV;
    }
    case Instruction::FPToUI: // double->APInt conversion handles sign
    case Instruction::FPToSI: {
      GenericValue GV = getConstantValue(Op0);
      uint32_t BitWidth = cast<IntegerType>(CE->getType())->getBitWidth();
      if (Op0->getType() == Type::FloatTy)
        GV.IntVal = APIntOps::RoundFloatToAPInt(GV.FloatVal, BitWidth);
      else if (Op0->getType() == Type::DoubleTy)
        GV.IntVal = APIntOps::RoundDoubleToAPInt(GV.DoubleVal, BitWidth);
      else if (Op0->getType() == Type::X86_FP80Ty) {
        APFloat apf = APFloat(GV.IntVal);
        uint64_t v;
        (void)apf.convertToInteger(&v, BitWidth,
                                   CE->getOpcode()==Instruction::FPToSI, 
                                   APFloat::rmTowardZero);
        GV.IntVal = v; // endian?
      }
      return GV;
    }
    case Instruction::PtrToInt: {
      GenericValue GV = getConstantValue(Op0);
      uint32_t PtrWidth = TD->getPointerSizeInBits();
      GV.IntVal = APInt(PtrWidth, uintptr_t(GV.PointerVal));
      return GV;
    }
    case Instruction::IntToPtr: {
      GenericValue GV = getConstantValue(Op0);
      uint32_t PtrWidth = TD->getPointerSizeInBits();
      if (PtrWidth != GV.IntVal.getBitWidth())
        GV.IntVal = GV.IntVal.zextOrTrunc(PtrWidth);
      assert(GV.IntVal.getBitWidth() <= 64 && "Bad pointer width");
      GV.PointerVal = PointerTy(uintptr_t(GV.IntVal.getZExtValue()));
      return GV;
    }
    case Instruction::BitCast: {
      GenericValue GV = getConstantValue(Op0);
      const Type* DestTy = CE->getType();
      switch (Op0->getType()->getTypeID()) {
        default: assert(0 && "Invalid bitcast operand");
        case Type::IntegerTyID:
          assert(DestTy->isFloatingPoint() && "invalid bitcast");
          if (DestTy == Type::FloatTy)
            GV.FloatVal = GV.IntVal.bitsToFloat();
          else if (DestTy == Type::DoubleTy)
            GV.DoubleVal = GV.IntVal.bitsToDouble();
          break;
        case Type::FloatTyID: 
          assert(DestTy == Type::Int32Ty && "Invalid bitcast");
          GV.IntVal.floatToBits(GV.FloatVal);
          break;
        case Type::DoubleTyID:
          assert(DestTy == Type::Int64Ty && "Invalid bitcast");
          GV.IntVal.doubleToBits(GV.DoubleVal);
          break;
        case Type::PointerTyID:
          assert(isa<PointerType>(DestTy) && "Invalid bitcast");
          break; // getConstantValue(Op0)  above already converted it
      }
      return GV;
    }
    case Instruction::Add:
    case Instruction::Sub:
    case Instruction::Mul:
    case Instruction::UDiv:
    case Instruction::SDiv:
    case Instruction::URem:
    case Instruction::SRem:
    case Instruction::And:
    case Instruction::Or:
    case Instruction::Xor: {
      GenericValue LHS = getConstantValue(Op0);
      GenericValue RHS = getConstantValue(CE->getOperand(1));
      GenericValue GV;
      switch (CE->getOperand(0)->getType()->getTypeID()) {
      default: assert(0 && "Bad add type!"); abort();
      case Type::IntegerTyID:
        switch (CE->getOpcode()) {
          default: assert(0 && "Invalid integer opcode");
          case Instruction::Add: GV.IntVal = LHS.IntVal + RHS.IntVal; break;
          case Instruction::Sub: GV.IntVal = LHS.IntVal - RHS.IntVal; break;
          case Instruction::Mul: GV.IntVal = LHS.IntVal * RHS.IntVal; break;
          case Instruction::UDiv:GV.IntVal = LHS.IntVal.udiv(RHS.IntVal); break;
          case Instruction::SDiv:GV.IntVal = LHS.IntVal.sdiv(RHS.IntVal); break;
          case Instruction::URem:GV.IntVal = LHS.IntVal.urem(RHS.IntVal); break;
          case Instruction::SRem:GV.IntVal = LHS.IntVal.srem(RHS.IntVal); break;
          case Instruction::And: GV.IntVal = LHS.IntVal & RHS.IntVal; break;
          case Instruction::Or:  GV.IntVal = LHS.IntVal | RHS.IntVal; break;
          case Instruction::Xor: GV.IntVal = LHS.IntVal ^ RHS.IntVal; break;
        }
        break;
      case Type::FloatTyID:
        switch (CE->getOpcode()) {
          default: assert(0 && "Invalid float opcode"); abort();
          case Instruction::Add:  
            GV.FloatVal = LHS.FloatVal + RHS.FloatVal; break;
          case Instruction::Sub:  
            GV.FloatVal = LHS.FloatVal - RHS.FloatVal; break;
          case Instruction::Mul:  
            GV.FloatVal = LHS.FloatVal * RHS.FloatVal; break;
          case Instruction::FDiv: 
            GV.FloatVal = LHS.FloatVal / RHS.FloatVal; break;
          case Instruction::FRem: 
            GV.FloatVal = ::fmodf(LHS.FloatVal,RHS.FloatVal); break;
        }
        break;
      case Type::DoubleTyID:
        switch (CE->getOpcode()) {
          default: assert(0 && "Invalid double opcode"); abort();
          case Instruction::Add:  
            GV.DoubleVal = LHS.DoubleVal + RHS.DoubleVal; break;
          case Instruction::Sub:  
            GV.DoubleVal = LHS.DoubleVal - RHS.DoubleVal; break;
          case Instruction::Mul:  
            GV.DoubleVal = LHS.DoubleVal * RHS.DoubleVal; break;
          case Instruction::FDiv: 
            GV.DoubleVal = LHS.DoubleVal / RHS.DoubleVal; break;
          case Instruction::FRem: 
            GV.DoubleVal = ::fmod(LHS.DoubleVal,RHS.DoubleVal); break;
        }
        break;
      case Type::X86_FP80TyID:
      case Type::PPC_FP128TyID:
      case Type::FP128TyID: {
        APFloat apfLHS = APFloat(LHS.IntVal);
        switch (CE->getOpcode()) {
          default: assert(0 && "Invalid long double opcode"); abort();
          case Instruction::Add:  
            apfLHS.add(APFloat(RHS.IntVal), APFloat::rmNearestTiesToEven);
            GV.IntVal = apfLHS.convertToAPInt();
            break;
          case Instruction::Sub:  
            apfLHS.subtract(APFloat(RHS.IntVal), APFloat::rmNearestTiesToEven);
            GV.IntVal = apfLHS.convertToAPInt();
            break;
          case Instruction::Mul:  
            apfLHS.multiply(APFloat(RHS.IntVal), APFloat::rmNearestTiesToEven);
            GV.IntVal = apfLHS.convertToAPInt();
            break;
          case Instruction::FDiv: 
            apfLHS.divide(APFloat(RHS.IntVal), APFloat::rmNearestTiesToEven);
            GV.IntVal = apfLHS.convertToAPInt();
            break;
          case Instruction::FRem: 
            apfLHS.mod(APFloat(RHS.IntVal), APFloat::rmNearestTiesToEven);
            GV.IntVal = apfLHS.convertToAPInt();
            break;
          }
        }
        break;
      }
      return GV;
    }
    default:
      break;
    }
    cerr << "ConstantExpr not handled: " << *CE << "\n";
    abort();
  }

  GenericValue Result;
  switch (C->getType()->getTypeID()) {
  case Type::FloatTyID: 
    Result.FloatVal = cast<ConstantFP>(C)->getValueAPF().convertToFloat(); 
    break;
  case Type::DoubleTyID:
    Result.DoubleVal = cast<ConstantFP>(C)->getValueAPF().convertToDouble();
    break;
  case Type::X86_FP80TyID:
  case Type::FP128TyID:
  case Type::PPC_FP128TyID:
    Result.IntVal = cast <ConstantFP>(C)->getValueAPF().convertToAPInt();
    break;
  case Type::IntegerTyID:
    Result.IntVal = cast<ConstantInt>(C)->getValue();
    break;
  case Type::PointerTyID:
    if (isa<ConstantPointerNull>(C))
      Result.PointerVal = 0;
    else if (const Function *F = dyn_cast<Function>(C))
      Result = PTOGV(getPointerToFunctionOrStub(const_cast<Function*>(F)));
    else if (const GlobalVariable* GV = dyn_cast<GlobalVariable>(C))
      Result = PTOGV(getOrEmitGlobalVariable(const_cast<GlobalVariable*>(GV)));
    else
      assert(0 && "Unknown constant pointer type!");
    break;
  default:
    cerr << "ERROR: Constant unimplemented for type: " << *C->getType() << "\n";
    abort();
  }
  return Result;
}

/// StoreIntToMemory - Fills the StoreBytes bytes of memory starting from Dst
/// with the integer held in IntVal.
static void StoreIntToMemory(const APInt &IntVal, uint8_t *Dst,
                             unsigned StoreBytes) {
  assert((IntVal.getBitWidth()+7)/8 >= StoreBytes && "Integer too small!");
  uint8_t *Src = (uint8_t *)IntVal.getRawData();

  if (sys::littleEndianHost())
    // Little-endian host - the source is ordered from LSB to MSB.  Order the
    // destination from LSB to MSB: Do a straight copy.
    memcpy(Dst, Src, StoreBytes);
  else {
    // Big-endian host - the source is an array of 64 bit words ordered from
    // LSW to MSW.  Each word is ordered from MSB to LSB.  Order the destination
    // from MSB to LSB: Reverse the word order, but not the bytes in a word.
    while (StoreBytes > sizeof(uint64_t)) {
      StoreBytes -= sizeof(uint64_t);
      // May not be aligned so use memcpy.
      memcpy(Dst + StoreBytes, Src, sizeof(uint64_t));
      Src += sizeof(uint64_t);
    }

    memcpy(Dst, Src + sizeof(uint64_t) - StoreBytes, StoreBytes);
  }
}

/// StoreValueToMemory - Stores the data in Val of type Ty at address Ptr.  Ptr
/// is the address of the memory at which to store Val, cast to GenericValue *.
/// It is not a pointer to a GenericValue containing the address at which to
/// store Val.
void ExecutionEngine::StoreValueToMemory(const GenericValue &Val, GenericValue *Ptr,
                                         const Type *Ty) {
  const unsigned StoreBytes = getTargetData()->getTypeStoreSize(Ty);

  switch (Ty->getTypeID()) {
  case Type::IntegerTyID:
    StoreIntToMemory(Val.IntVal, (uint8_t*)Ptr, StoreBytes);
    break;
  case Type::FloatTyID:
    *((float*)Ptr) = Val.FloatVal;
    break;
  case Type::DoubleTyID:
    *((double*)Ptr) = Val.DoubleVal;
    break;
  case Type::X86_FP80TyID: {
      uint16_t *Dest = (uint16_t*)Ptr;
      const uint16_t *Src = (uint16_t*)Val.IntVal.getRawData();
      // This is endian dependent, but it will only work on x86 anyway.
      Dest[0] = Src[4];
      Dest[1] = Src[0];
      Dest[2] = Src[1];
      Dest[3] = Src[2];
      Dest[4] = Src[3];
      break;
    }
  case Type::PointerTyID:
    // Ensure 64 bit target pointers are fully initialized on 32 bit hosts.
    if (StoreBytes != sizeof(PointerTy))
      memset(Ptr, 0, StoreBytes);

    *((PointerTy*)Ptr) = Val.PointerVal;
    break;
  default:
    cerr << "Cannot store value of type " << *Ty << "!\n";
  }

  if (sys::littleEndianHost() != getTargetData()->isLittleEndian())
    // Host and target are different endian - reverse the stored bytes.
    std::reverse((uint8_t*)Ptr, StoreBytes + (uint8_t*)Ptr);
}

/// LoadIntFromMemory - Loads the integer stored in the LoadBytes bytes starting
/// from Src into IntVal, which is assumed to be wide enough and to hold zero.
static void LoadIntFromMemory(APInt &IntVal, uint8_t *Src, unsigned LoadBytes) {
  assert((IntVal.getBitWidth()+7)/8 >= LoadBytes && "Integer too small!");
  uint8_t *Dst = (uint8_t *)IntVal.getRawData();

  if (sys::littleEndianHost())
    // Little-endian host - the destination must be ordered from LSB to MSB.
    // The source is ordered from LSB to MSB: Do a straight copy.
    memcpy(Dst, Src, LoadBytes);
  else {
    // Big-endian - the destination is an array of 64 bit words ordered from
    // LSW to MSW.  Each word must be ordered from MSB to LSB.  The source is
    // ordered from MSB to LSB: Reverse the word order, but not the bytes in
    // a word.
    while (LoadBytes > sizeof(uint64_t)) {
      LoadBytes -= sizeof(uint64_t);
      // May not be aligned so use memcpy.
      memcpy(Dst, Src + LoadBytes, sizeof(uint64_t));
      Dst += sizeof(uint64_t);
    }

    memcpy(Dst + sizeof(uint64_t) - LoadBytes, Src, LoadBytes);
  }
}

/// FIXME: document
///
void ExecutionEngine::LoadValueFromMemory(GenericValue &Result,
                                                  GenericValue *Ptr,
                                                  const Type *Ty) {
  const unsigned LoadBytes = getTargetData()->getTypeStoreSize(Ty);

  if (sys::littleEndianHost() != getTargetData()->isLittleEndian()) {
    // Host and target are different endian - reverse copy the stored
    // bytes into a buffer, and load from that.
    uint8_t *Src = (uint8_t*)Ptr;
    uint8_t *Buf = (uint8_t*)alloca(LoadBytes);
    std::reverse_copy(Src, Src + LoadBytes, Buf);
    Ptr = (GenericValue*)Buf;
  }

  switch (Ty->getTypeID()) {
  case Type::IntegerTyID:
    // An APInt with all words initially zero.
    Result.IntVal = APInt(cast<IntegerType>(Ty)->getBitWidth(), 0);
    LoadIntFromMemory(Result.IntVal, (uint8_t*)Ptr, LoadBytes);
    break;
  case Type::FloatTyID:
    Result.FloatVal = *((float*)Ptr);
    break;
  case Type::DoubleTyID:
    Result.DoubleVal = *((double*)Ptr);
    break;
  case Type::PointerTyID:
    Result.PointerVal = *((PointerTy*)Ptr);
    break;
  case Type::X86_FP80TyID: {
    // This is endian dependent, but it will only work on x86 anyway.
    // FIXME: Will not trap if loading a signaling NaN.
    uint16_t *p = (uint16_t*)Ptr;
    union {
      uint16_t x[8];
      uint64_t y[2];
    };
    x[0] = p[1];
    x[1] = p[2];
    x[2] = p[3];
    x[3] = p[4];
    x[4] = p[0];
    Result.IntVal = APInt(80, 2, y);
    break;
  }
  default:
    cerr << "Cannot load value of type " << *Ty << "!\n";
    abort();
  }
}

// InitializeMemory - Recursive function to apply a Constant value into the
// specified memory location...
//
void ExecutionEngine::InitializeMemory(const Constant *Init, void *Addr) {
  if (isa<UndefValue>(Init)) {
    return;
  } else if (const ConstantVector *CP = dyn_cast<ConstantVector>(Init)) {
    unsigned ElementSize =
      getTargetData()->getABITypeSize(CP->getType()->getElementType());
    for (unsigned i = 0, e = CP->getNumOperands(); i != e; ++i)
      InitializeMemory(CP->getOperand(i), (char*)Addr+i*ElementSize);
    return;
  } else if (Init->getType()->isFirstClassType()) {
    GenericValue Val = getConstantValue(Init);
    StoreValueToMemory(Val, (GenericValue*)Addr, Init->getType());
    return;
  } else if (isa<ConstantAggregateZero>(Init)) {
    memset(Addr, 0, (size_t)getTargetData()->getABITypeSize(Init->getType()));
    return;
  }

  switch (Init->getType()->getTypeID()) {
  case Type::ArrayTyID: {
    const ConstantArray *CPA = cast<ConstantArray>(Init);
    unsigned ElementSize =
      getTargetData()->getABITypeSize(CPA->getType()->getElementType());
    for (unsigned i = 0, e = CPA->getNumOperands(); i != e; ++i)
      InitializeMemory(CPA->getOperand(i), (char*)Addr+i*ElementSize);
    return;
  }

  case Type::StructTyID: {
    const ConstantStruct *CPS = cast<ConstantStruct>(Init);
    const StructLayout *SL =
      getTargetData()->getStructLayout(cast<StructType>(CPS->getType()));
    for (unsigned i = 0, e = CPS->getNumOperands(); i != e; ++i)
      InitializeMemory(CPS->getOperand(i), (char*)Addr+SL->getElementOffset(i));
    return;
  }

  default:
    cerr << "Bad Type: " << *Init->getType() << "\n";
    assert(0 && "Unknown constant type to initialize memory with!");
  }
}

/// EmitGlobals - Emit all of the global variables to memory, storing their
/// addresses into GlobalAddress.  This must make sure to copy the contents of
/// their initializers into the memory.
///
void ExecutionEngine::emitGlobals() {
  const TargetData *TD = getTargetData();

  // Loop over all of the global variables in the program, allocating the memory
  // to hold them.  If there is more than one module, do a prepass over globals
  // to figure out how the different modules should link together.
  //
  std::map<std::pair<std::string, const Type*>,
           const GlobalValue*> LinkedGlobalsMap;

  if (Modules.size() != 1) {
    for (unsigned m = 0, e = Modules.size(); m != e; ++m) {
      Module &M = *Modules[m]->getModule();
      for (Module::const_global_iterator I = M.global_begin(),
           E = M.global_end(); I != E; ++I) {
        const GlobalValue *GV = I;
        if (GV->hasInternalLinkage() || GV->isDeclaration() ||
            GV->hasAppendingLinkage() || !GV->hasName())
          continue;// Ignore external globals and globals with internal linkage.
          
        const GlobalValue *&GVEntry = 
          LinkedGlobalsMap[std::make_pair(GV->getName(), GV->getType())];

        // If this is the first time we've seen this global, it is the canonical
        // version.
        if (!GVEntry) {
          GVEntry = GV;
          continue;
        }
        
        // If the existing global is strong, never replace it.
        if (GVEntry->hasExternalLinkage() ||
            GVEntry->hasDLLImportLinkage() ||
            GVEntry->hasDLLExportLinkage())
          continue;
        
        // Otherwise, we know it's linkonce/weak, replace it if this is a strong
        // symbol.
        if (GV->hasExternalLinkage() || GVEntry->hasExternalWeakLinkage())
          GVEntry = GV;
      }
    }
  }
  
  std::vector<const GlobalValue*> NonCanonicalGlobals;
  for (unsigned m = 0, e = Modules.size(); m != e; ++m) {
    Module &M = *Modules[m]->getModule();
    for (Module::const_global_iterator I = M.global_begin(), E = M.global_end();
         I != E; ++I) {
      // In the multi-module case, see what this global maps to.
      if (!LinkedGlobalsMap.empty()) {
        if (const GlobalValue *GVEntry = 
              LinkedGlobalsMap[std::make_pair(I->getName(), I->getType())]) {
          // If something else is the canonical global, ignore this one.
          if (GVEntry != &*I) {
            NonCanonicalGlobals.push_back(I);
            continue;
          }
        }
      }
      
      if (!I->isDeclaration()) {
        // Get the type of the global.
        const Type *Ty = I->getType()->getElementType();

        // Allocate some memory for it!
        unsigned Size = TD->getABITypeSize(Ty);
        addGlobalMapping(I, new char[Size]);
      } else {
        // External variable reference. Try to use the dynamic loader to
        // get a pointer to it.
        if (void *SymAddr =
            sys::DynamicLibrary::SearchForAddressOfSymbol(I->getName().c_str()))
          addGlobalMapping(I, SymAddr);
        else {
          cerr << "Could not resolve external global address: "
               << I->getName() << "\n";
          abort();
        }
      }
    }
    
    // If there are multiple modules, map the non-canonical globals to their
    // canonical location.
    if (!NonCanonicalGlobals.empty()) {
      for (unsigned i = 0, e = NonCanonicalGlobals.size(); i != e; ++i) {
        const GlobalValue *GV = NonCanonicalGlobals[i];
        const GlobalValue *CGV =
          LinkedGlobalsMap[std::make_pair(GV->getName(), GV->getType())];
        void *Ptr = getPointerToGlobalIfAvailable(CGV);
        assert(Ptr && "Canonical global wasn't codegen'd!");
        addGlobalMapping(GV, getPointerToGlobalIfAvailable(CGV));
      }
    }
    
    // Now that all of the globals are set up in memory, loop through them all 
    // and initialize their contents.
    for (Module::const_global_iterator I = M.global_begin(), E = M.global_end();
         I != E; ++I) {
      if (!I->isDeclaration()) {
        if (!LinkedGlobalsMap.empty()) {
          if (const GlobalValue *GVEntry = 
                LinkedGlobalsMap[std::make_pair(I->getName(), I->getType())])
            if (GVEntry != &*I)  // Not the canonical variable.
              continue;
        }
        EmitGlobalVariable(I);
      }
    }
  }
}

// EmitGlobalVariable - This method emits the specified global variable to the
// address specified in GlobalAddresses, or allocates new memory if it's not
// already in the map.
void ExecutionEngine::EmitGlobalVariable(const GlobalVariable *GV) {
  void *GA = getPointerToGlobalIfAvailable(GV);
  DOUT << "Global '" << GV->getName() << "' -> " << GA << "\n";

  const Type *ElTy = GV->getType()->getElementType();
  size_t GVSize = (size_t)getTargetData()->getABITypeSize(ElTy);
  if (GA == 0) {
    // If it's not already specified, allocate memory for the global.
    GA = new char[GVSize];
    addGlobalMapping(GV, GA);
  }

  InitializeMemory(GV->getInitializer(), GA);
  NumInitBytes += (unsigned)GVSize;
  ++NumGlobals;
}