summaryrefslogtreecommitdiffstats
path: root/lib/ExecutionEngine/JIT/JIT.cpp
blob: 2a7dfa00b62a96d09172b8e9c5f521c6b563f48e (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
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
//===-- JIT.cpp - LLVM Just in Time Compiler ------------------------------===//
//
//                     The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This tool implements a just-in-time compiler for LLVM, allowing direct
// execution of LLVM bitcode in an efficient manner.
//
//===----------------------------------------------------------------------===//

#include "JIT.h"
#include "llvm/ADT/SmallPtrSet.h"
#include "llvm/CodeGen/JITCodeEmitter.h"
#include "llvm/CodeGen/MachineCodeInfo.h"
#include "llvm/Config/config.h"
#include "llvm/Constants.h"
#include "llvm/DataLayout.h"
#include "llvm/DerivedTypes.h"
#include "llvm/ExecutionEngine/GenericValue.h"
#include "llvm/ExecutionEngine/JITEventListener.h"
#include "llvm/ExecutionEngine/JITMemoryManager.h"
#include "llvm/Function.h"
#include "llvm/GlobalVariable.h"
#include "llvm/Instructions.h"
#include "llvm/Support/Dwarf.h"
#include "llvm/Support/DynamicLibrary.h"
#include "llvm/Support/ErrorHandling.h"
#include "llvm/Support/ManagedStatic.h"
#include "llvm/Support/MutexGuard.h"
#include "llvm/Target/TargetJITInfo.h"
#include "llvm/Target/TargetMachine.h"

using namespace llvm;

#ifdef __APPLE__
// Apple gcc defaults to -fuse-cxa-atexit (i.e. calls __cxa_atexit instead
// of atexit). It passes the address of linker generated symbol __dso_handle
// to the function.
// This configuration change happened at version 5330.
# include <AvailabilityMacros.h>
# if defined(MAC_OS_X_VERSION_10_4) && \
     ((MAC_OS_X_VERSION_MIN_REQUIRED > MAC_OS_X_VERSION_10_4) || \
      (MAC_OS_X_VERSION_MIN_REQUIRED == MAC_OS_X_VERSION_10_4 && \
       __APPLE_CC__ >= 5330))
#  ifndef HAVE___DSO_HANDLE
#   define HAVE___DSO_HANDLE 1
#  endif
# endif
#endif

#if HAVE___DSO_HANDLE
extern void *__dso_handle __attribute__ ((__visibility__ ("hidden")));
#endif

namespace {

static struct RegisterJIT {
  RegisterJIT() { JIT::Register(); }
} JITRegistrator;

}

extern "C" void LLVMLinkInJIT() {
}

// Determine whether we can register EH tables.
#if (defined(__GNUC__) && !defined(__ARM_EABI__) && \
     !defined(__USING_SJLJ_EXCEPTIONS__))
#define HAVE_EHTABLE_SUPPORT 1
#else
#define HAVE_EHTABLE_SUPPORT 0
#endif

#if HAVE_EHTABLE_SUPPORT

// libgcc defines the __register_frame function to dynamically register new
// dwarf frames for exception handling. This functionality is not portable
// across compilers and is only provided by GCC. We use the __register_frame
// function here so that code generated by the JIT cooperates with the unwinding
// runtime of libgcc. When JITting with exception handling enable, LLVM
// generates dwarf frames and registers it to libgcc with __register_frame.
//
// The __register_frame function works with Linux.
//
// Unfortunately, this functionality seems to be in libgcc after the unwinding
// library of libgcc for darwin was written. The code for darwin overwrites the
// value updated by __register_frame with a value fetched with "keymgr".
// "keymgr" is an obsolete functionality, which should be rewritten some day.
// In the meantime, since "keymgr" is on all libgccs shipped with apple-gcc, we
// need a workaround in LLVM which uses the "keymgr" to dynamically modify the
// values of an opaque key, used by libgcc to find dwarf tables.

extern "C" void __register_frame(void*);
extern "C" void __deregister_frame(void*);

#if defined(__APPLE__) && MAC_OS_X_VERSION_MAX_ALLOWED <= 1050
# define USE_KEYMGR 1
#else
# define USE_KEYMGR 0
#endif

#if USE_KEYMGR

namespace {

// LibgccObject - This is the structure defined in libgcc. There is no #include
// provided for this structure, so we also define it here. libgcc calls it
// "struct object". The structure is undocumented in libgcc.
struct LibgccObject {
  void *unused1;
  void *unused2;
  void *unused3;

  /// frame - Pointer to the exception table.
  void *frame;

  /// encoding -  The encoding of the object?
  union {
    struct {
      unsigned long sorted : 1;
      unsigned long from_array : 1;
      unsigned long mixed_encoding : 1;
      unsigned long encoding : 8;
      unsigned long count : 21;
    } b;
    size_t i;
  } encoding;

  /// fde_end - libgcc defines this field only if some macro is defined. We
  /// include this field even if it may not there, to make libgcc happy.
  char *fde_end;

  /// next - At least we know it's a chained list!
  struct LibgccObject *next;
};

// "kemgr" stuff. Apparently, all frame tables are stored there.
extern "C" void _keymgr_set_and_unlock_processwide_ptr(int, void *);
extern "C" void *_keymgr_get_and_lock_processwide_ptr(int);
#define KEYMGR_GCC3_DW2_OBJ_LIST        302     /* Dwarf2 object list  */

/// LibgccObjectInfo - libgcc defines this struct as km_object_info. It
/// probably contains all dwarf tables that are loaded.
struct LibgccObjectInfo {

  /// seenObjects - LibgccObjects already parsed by the unwinding runtime.
  ///
  struct LibgccObject* seenObjects;

  /// unseenObjects - LibgccObjects not parsed yet by the unwinding runtime.
  ///
  struct LibgccObject* unseenObjects;

  unsigned unused[2];
};

/// darwin_register_frame - Since __register_frame does not work with darwin's
/// libgcc,we provide our own function, which "tricks" libgcc by modifying the
/// "Dwarf2 object list" key.
void DarwinRegisterFrame(void* FrameBegin) {
  // Get the key.
  LibgccObjectInfo* LOI = (struct LibgccObjectInfo*)
    _keymgr_get_and_lock_processwide_ptr(KEYMGR_GCC3_DW2_OBJ_LIST);
  assert(LOI && "This should be preallocated by the runtime");

  // Allocate a new LibgccObject to represent this frame. Deallocation of this
  // object may be impossible: since darwin code in libgcc was written after
  // the ability to dynamically register frames, things may crash if we
  // deallocate it.
  struct LibgccObject* ob = (struct LibgccObject*)
    malloc(sizeof(struct LibgccObject));

  // Do like libgcc for the values of the field.
  ob->unused1 = (void *)-1;
  ob->unused2 = 0;
  ob->unused3 = 0;
  ob->frame = FrameBegin;
  ob->encoding.i = 0;
  ob->encoding.b.encoding = llvm::dwarf::DW_EH_PE_omit;

  // Put the info on both places, as libgcc uses the first or the second
  // field. Note that we rely on having two pointers here. If fde_end was a
  // char, things would get complicated.
  ob->fde_end = (char*)LOI->unseenObjects;
  ob->next = LOI->unseenObjects;

  // Update the key's unseenObjects list.
  LOI->unseenObjects = ob;

  // Finally update the "key". Apparently, libgcc requires it.
  _keymgr_set_and_unlock_processwide_ptr(KEYMGR_GCC3_DW2_OBJ_LIST,
                                         LOI);

}

}
#endif // __APPLE__
#endif // HAVE_EHTABLE_SUPPORT

/// createJIT - This is the factory method for creating a JIT for the current
/// machine, it does not fall back to the interpreter.  This takes ownership
/// of the module.
ExecutionEngine *JIT::createJIT(Module *M,
                                std::string *ErrorStr,
                                JITMemoryManager *JMM,
                                bool GVsWithCode,
                                TargetMachine *TM) {
  // Try to register the program as a source of symbols to resolve against.
  //
  // FIXME: Don't do this here.
  sys::DynamicLibrary::LoadLibraryPermanently(0, NULL);

  // If the target supports JIT code generation, create the JIT.
  if (TargetJITInfo *TJ = TM->getJITInfo()) {
    return new JIT(M, *TM, *TJ, JMM, GVsWithCode);
  } else {
    if (ErrorStr)
      *ErrorStr = "target does not support JIT code generation";
    return 0;
  }
}

namespace {
/// This class supports the global getPointerToNamedFunction(), which allows
/// bugpoint or gdb users to search for a function by name without any context.
class JitPool {
  SmallPtrSet<JIT*, 1> JITs;  // Optimize for process containing just 1 JIT.
  mutable sys::Mutex Lock;
public:
  void Add(JIT *jit) {
    MutexGuard guard(Lock);
    JITs.insert(jit);
  }
  void Remove(JIT *jit) {
    MutexGuard guard(Lock);
    JITs.erase(jit);
  }
  void *getPointerToNamedFunction(const char *Name) const {
    MutexGuard guard(Lock);
    assert(JITs.size() != 0 && "No Jit registered");
    //search function in every instance of JIT
    for (SmallPtrSet<JIT*, 1>::const_iterator Jit = JITs.begin(),
           end = JITs.end();
         Jit != end; ++Jit) {
      if (Function *F = (*Jit)->FindFunctionNamed(Name))
        return (*Jit)->getPointerToFunction(F);
    }
    // The function is not available : fallback on the first created (will
    // search in symbol of the current program/library)
    return (*JITs.begin())->getPointerToNamedFunction(Name);
  }
};
ManagedStatic<JitPool> AllJits;
}
extern "C" {
  // getPointerToNamedFunction - This function is used as a global wrapper to
  // JIT::getPointerToNamedFunction for the purpose of resolving symbols when
  // bugpoint is debugging the JIT. In that scenario, we are loading an .so and
  // need to resolve function(s) that are being mis-codegenerated, so we need to
  // resolve their addresses at runtime, and this is the way to do it.
  void *getPointerToNamedFunction(const char *Name) {
    return AllJits->getPointerToNamedFunction(Name);
  }
}

JIT::JIT(Module *M, TargetMachine &tm, TargetJITInfo &tji,
         JITMemoryManager *jmm, bool GVsWithCode)
  : ExecutionEngine(M), TM(tm), TJI(tji),
    JMM(jmm ? jmm : JITMemoryManager::CreateDefaultMemManager()),
    AllocateGVsWithCode(GVsWithCode), isAlreadyCodeGenerating(false) {
  setDataLayout(TM.getDataLayout());

  jitstate = new JITState(M);

  // Initialize JCE
  JCE = createEmitter(*this, JMM, TM);

  // Register in global list of all JITs.
  AllJits->Add(this);

  // Add target data
  MutexGuard locked(lock);
  FunctionPassManager &PM = jitstate->getPM(locked);
  PM.add(new DataLayout(*TM.getDataLayout()));

  // Turn the machine code intermediate representation into bytes in memory that
  // may be executed.
  if (TM.addPassesToEmitMachineCode(PM, *JCE)) {
    report_fatal_error("Target does not support machine code emission!");
  }

  // Register routine for informing unwinding runtime about new EH frames
#if HAVE_EHTABLE_SUPPORT
#if USE_KEYMGR
  struct LibgccObjectInfo* LOI = (struct LibgccObjectInfo*)
    _keymgr_get_and_lock_processwide_ptr(KEYMGR_GCC3_DW2_OBJ_LIST);

  // The key is created on demand, and libgcc creates it the first time an
  // exception occurs. Since we need the key to register frames, we create
  // it now.
  if (!LOI)
    LOI = (LibgccObjectInfo*)calloc(sizeof(struct LibgccObjectInfo), 1);
  _keymgr_set_and_unlock_processwide_ptr(KEYMGR_GCC3_DW2_OBJ_LIST, LOI);
  InstallExceptionTableRegister(DarwinRegisterFrame);
  // Not sure about how to deregister on Darwin.
#else
  InstallExceptionTableRegister(__register_frame);
  InstallExceptionTableDeregister(__deregister_frame);
#endif // __APPLE__
#endif // HAVE_EHTABLE_SUPPORT

  // Initialize passes.
  PM.doInitialization();
}

JIT::~JIT() {
  // Unregister all exception tables registered by this JIT.
  DeregisterAllTables();
  // Cleanup.
  AllJits->Remove(this);
  delete jitstate;
  delete JCE;
  // JMM is a ownership of JCE, so we no need delete JMM here.
  delete &TM;
}

/// addModule - Add a new Module to the JIT.  If we previously removed the last
/// Module, we need re-initialize jitstate with a valid Module.
void JIT::addModule(Module *M) {
  MutexGuard locked(lock);

  if (Modules.empty()) {
    assert(!jitstate && "jitstate should be NULL if Modules vector is empty!");

    jitstate = new JITState(M);

    FunctionPassManager &PM = jitstate->getPM(locked);
    PM.add(new DataLayout(*TM.getDataLayout()));

    // Turn the machine code intermediate representation into bytes in memory
    // that may be executed.
    if (TM.addPassesToEmitMachineCode(PM, *JCE)) {
      report_fatal_error("Target does not support machine code emission!");
    }

    // Initialize passes.
    PM.doInitialization();
  }

  ExecutionEngine::addModule(M);
}

/// removeModule - If we are removing the last Module, invalidate the jitstate
/// since the PassManager it contains references a released Module.
bool JIT::removeModule(Module *M) {
  bool result = ExecutionEngine::removeModule(M);

  MutexGuard locked(lock);

  if (jitstate && jitstate->getModule() == M) {
    delete jitstate;
    jitstate = 0;
  }

  if (!jitstate && !Modules.empty()) {
    jitstate = new JITState(Modules[0]);

    FunctionPassManager &PM = jitstate->getPM(locked);
    PM.add(new DataLayout(*TM.getDataLayout()));

    // Turn the machine code intermediate representation into bytes in memory
    // that may be executed.
    if (TM.addPassesToEmitMachineCode(PM, *JCE)) {
      report_fatal_error("Target does not support machine code emission!");
    }

    // Initialize passes.
    PM.doInitialization();
  }
  return result;
}

/// run - Start execution with the specified function and arguments.
///
GenericValue JIT::runFunction(Function *F,
                              const std::vector<GenericValue> &ArgValues) {
  assert(F && "Function *F was null at entry to run()");

  void *FPtr = getPointerToFunction(F);
  assert(FPtr && "Pointer to fn's code was null after getPointerToFunction");
  FunctionType *FTy = F->getFunctionType();
  Type *RetTy = FTy->getReturnType();

  assert((FTy->getNumParams() == ArgValues.size() ||
          (FTy->isVarArg() && FTy->getNumParams() <= ArgValues.size())) &&
         "Wrong number of arguments passed into function!");
  assert(FTy->getNumParams() == ArgValues.size() &&
         "This doesn't support passing arguments through varargs (yet)!");

  // Handle some common cases first.  These cases correspond to common `main'
  // prototypes.
  if (RetTy->isIntegerTy(32) || RetTy->isVoidTy()) {
    switch (ArgValues.size()) {
    case 3:
      if (FTy->getParamType(0)->isIntegerTy(32) &&
          FTy->getParamType(1)->isPointerTy() &&
          FTy->getParamType(2)->isPointerTy()) {
        int (*PF)(int, char **, const char **) =
          (int(*)(int, char **, const char **))(intptr_t)FPtr;

        // Call the function.
        GenericValue rv;
        rv.IntVal = APInt(32, PF(ArgValues[0].IntVal.getZExtValue(),
                                 (char **)GVTOP(ArgValues[1]),
                                 (const char **)GVTOP(ArgValues[2])));
        return rv;
      }
      break;
    case 2:
      if (FTy->getParamType(0)->isIntegerTy(32) &&
          FTy->getParamType(1)->isPointerTy()) {
        int (*PF)(int, char **) = (int(*)(int, char **))(intptr_t)FPtr;

        // Call the function.
        GenericValue rv;
        rv.IntVal = APInt(32, PF(ArgValues[0].IntVal.getZExtValue(),
                                 (char **)GVTOP(ArgValues[1])));
        return rv;
      }
      break;
    case 1:
      if (FTy->getParamType(0)->isIntegerTy(32)) {
        GenericValue rv;
        int (*PF)(int) = (int(*)(int))(intptr_t)FPtr;
        rv.IntVal = APInt(32, PF(ArgValues[0].IntVal.getZExtValue()));
        return rv;
      }
      if (FTy->getParamType(0)->isPointerTy()) {
        GenericValue rv;
        int (*PF)(char *) = (int(*)(char *))(intptr_t)FPtr;
        rv.IntVal = APInt(32, PF((char*)GVTOP(ArgValues[0])));
        return rv;
      }
      break;
    }
  }

  // Handle cases where no arguments are passed first.
  if (ArgValues.empty()) {
    GenericValue rv;
    switch (RetTy->getTypeID()) {
    default: llvm_unreachable("Unknown return type for function call!");
    case Type::IntegerTyID: {
      unsigned BitWidth = cast<IntegerType>(RetTy)->getBitWidth();
      if (BitWidth == 1)
        rv.IntVal = APInt(BitWidth, ((bool(*)())(intptr_t)FPtr)());
      else if (BitWidth <= 8)
        rv.IntVal = APInt(BitWidth, ((char(*)())(intptr_t)FPtr)());
      else if (BitWidth <= 16)
        rv.IntVal = APInt(BitWidth, ((short(*)())(intptr_t)FPtr)());
      else if (BitWidth <= 32)
        rv.IntVal = APInt(BitWidth, ((int(*)())(intptr_t)FPtr)());
      else if (BitWidth <= 64)
        rv.IntVal = APInt(BitWidth, ((int64_t(*)())(intptr_t)FPtr)());
      else
        llvm_unreachable("Integer types > 64 bits not supported");
      return rv;
    }
    case Type::VoidTyID:
      rv.IntVal = APInt(32, ((int(*)())(intptr_t)FPtr)());
      return rv;
    case Type::FloatTyID:
      rv.FloatVal = ((float(*)())(intptr_t)FPtr)();
      return rv;
    case Type::DoubleTyID:
      rv.DoubleVal = ((double(*)())(intptr_t)FPtr)();
      return rv;
    case Type::X86_FP80TyID:
    case Type::FP128TyID:
    case Type::PPC_FP128TyID:
      llvm_unreachable("long double not supported yet");
    case Type::PointerTyID:
      return PTOGV(((void*(*)())(intptr_t)FPtr)());
    }
  }

  // Okay, this is not one of our quick and easy cases.  Because we don't have a
  // full FFI, we have to codegen a nullary stub function that just calls the
  // function we are interested in, passing in constants for all of the
  // arguments.  Make this function and return.

  // First, create the function.
  FunctionType *STy=FunctionType::get(RetTy, false);
  Function *Stub = Function::Create(STy, Function::InternalLinkage, "",
                                    F->getParent());

  // Insert a basic block.
  BasicBlock *StubBB = BasicBlock::Create(F->getContext(), "", Stub);

  // Convert all of the GenericValue arguments over to constants.  Note that we
  // currently don't support varargs.
  SmallVector<Value*, 8> Args;
  for (unsigned i = 0, e = ArgValues.size(); i != e; ++i) {
    Constant *C = 0;
    Type *ArgTy = FTy->getParamType(i);
    const GenericValue &AV = ArgValues[i];
    switch (ArgTy->getTypeID()) {
    default: llvm_unreachable("Unknown argument type for function call!");
    case Type::IntegerTyID:
        C = ConstantInt::get(F->getContext(), AV.IntVal);
        break;
    case Type::FloatTyID:
        C = ConstantFP::get(F->getContext(), APFloat(AV.FloatVal));
        break;
    case Type::DoubleTyID:
        C = ConstantFP::get(F->getContext(), APFloat(AV.DoubleVal));
        break;
    case Type::PPC_FP128TyID:
    case Type::X86_FP80TyID:
    case Type::FP128TyID:
        C = ConstantFP::get(F->getContext(), APFloat(AV.IntVal));
        break;
    case Type::PointerTyID:
      void *ArgPtr = GVTOP(AV);
      if (sizeof(void*) == 4)
        C = ConstantInt::get(Type::getInt32Ty(F->getContext()),
                             (int)(intptr_t)ArgPtr);
      else
        C = ConstantInt::get(Type::getInt64Ty(F->getContext()),
                             (intptr_t)ArgPtr);
      // Cast the integer to pointer
      C = ConstantExpr::getIntToPtr(C, ArgTy);
      break;
    }
    Args.push_back(C);
  }

  CallInst *TheCall = CallInst::Create(F, Args, "", StubBB);
  TheCall->setCallingConv(F->getCallingConv());
  TheCall->setTailCall();
  if (!TheCall->getType()->isVoidTy())
    // Return result of the call.
    ReturnInst::Create(F->getContext(), TheCall, StubBB);
  else
    ReturnInst::Create(F->getContext(), StubBB);           // Just return void.

  // Finally, call our nullary stub function.
  GenericValue Result = runFunction(Stub, std::vector<GenericValue>());
  // Erase it, since no other function can have a reference to it.
  Stub->eraseFromParent();
  // And return the result.
  return Result;
}

void JIT::RegisterJITEventListener(JITEventListener *L) {
  if (L == NULL)
    return;
  MutexGuard locked(lock);
  EventListeners.push_back(L);
}
void JIT::UnregisterJITEventListener(JITEventListener *L) {
  if (L == NULL)
    return;
  MutexGuard locked(lock);
  std::vector<JITEventListener*>::reverse_iterator I=
      std::find(EventListeners.rbegin(), EventListeners.rend(), L);
  if (I != EventListeners.rend()) {
    std::swap(*I, EventListeners.back());
    EventListeners.pop_back();
  }
}
void JIT::NotifyFunctionEmitted(
    const Function &F,
    void *Code, size_t Size,
    const JITEvent_EmittedFunctionDetails &Details) {
  MutexGuard locked(lock);
  for (unsigned I = 0, S = EventListeners.size(); I < S; ++I) {
    EventListeners[I]->NotifyFunctionEmitted(F, Code, Size, Details);
  }
}

void JIT::NotifyFreeingMachineCode(void *OldPtr) {
  MutexGuard locked(lock);
  for (unsigned I = 0, S = EventListeners.size(); I < S; ++I) {
    EventListeners[I]->NotifyFreeingMachineCode(OldPtr);
  }
}

/// runJITOnFunction - Run the FunctionPassManager full of
/// just-in-time compilation passes on F, hopefully filling in
/// GlobalAddress[F] with the address of F's machine code.
///
void JIT::runJITOnFunction(Function *F, MachineCodeInfo *MCI) {
  MutexGuard locked(lock);

  class MCIListener : public JITEventListener {
    MachineCodeInfo *const MCI;
   public:
    MCIListener(MachineCodeInfo *mci) : MCI(mci) {}
    virtual void NotifyFunctionEmitted(const Function &,
                                       void *Code, size_t Size,
                                       const EmittedFunctionDetails &) {
      MCI->setAddress(Code);
      MCI->setSize(Size);
    }
  };
  MCIListener MCIL(MCI);
  if (MCI)
    RegisterJITEventListener(&MCIL);

  runJITOnFunctionUnlocked(F, locked);

  if (MCI)
    UnregisterJITEventListener(&MCIL);
}

void JIT::runJITOnFunctionUnlocked(Function *F, const MutexGuard &locked) {
  assert(!isAlreadyCodeGenerating && "Error: Recursive compilation detected!");

  jitTheFunction(F, locked);

  // If the function referred to another function that had not yet been
  // read from bitcode, and we are jitting non-lazily, emit it now.
  while (!jitstate->getPendingFunctions(locked).empty()) {
    Function *PF = jitstate->getPendingFunctions(locked).back();
    jitstate->getPendingFunctions(locked).pop_back();

    assert(!PF->hasAvailableExternallyLinkage() &&
           "Externally-defined function should not be in pending list.");

    jitTheFunction(PF, locked);

    // Now that the function has been jitted, ask the JITEmitter to rewrite
    // the stub with real address of the function.
    updateFunctionStub(PF);
  }
}

void JIT::jitTheFunction(Function *F, const MutexGuard &locked) {
  isAlreadyCodeGenerating = true;
  jitstate->getPM(locked).run(*F);
  isAlreadyCodeGenerating = false;

  // clear basic block addresses after this function is done
  getBasicBlockAddressMap(locked).clear();
}

/// getPointerToFunction - This method is used to get the address of the
/// specified function, compiling it if necessary.
///
void *JIT::getPointerToFunction(Function *F) {

  if (void *Addr = getPointerToGlobalIfAvailable(F))
    return Addr;   // Check if function already code gen'd

  MutexGuard locked(lock);

  // Now that this thread owns the lock, make sure we read in the function if it
  // exists in this Module.
  std::string ErrorMsg;
  if (F->Materialize(&ErrorMsg)) {
    report_fatal_error("Error reading function '" + F->getName()+
                      "' from bitcode file: " + ErrorMsg);
  }

  // ... and check if another thread has already code gen'd the function.
  if (void *Addr = getPointerToGlobalIfAvailable(F))
    return Addr;

  if (F->isDeclaration() || F->hasAvailableExternallyLinkage()) {
    bool AbortOnFailure = !F->hasExternalWeakLinkage();
    void *Addr = getPointerToNamedFunction(F->getName(), AbortOnFailure);
    addGlobalMapping(F, Addr);
    return Addr;
  }

  runJITOnFunctionUnlocked(F, locked);

  void *Addr = getPointerToGlobalIfAvailable(F);
  assert(Addr && "Code generation didn't add function to GlobalAddress table!");
  return Addr;
}

void JIT::addPointerToBasicBlock(const BasicBlock *BB, void *Addr) {
  MutexGuard locked(lock);

  BasicBlockAddressMapTy::iterator I =
    getBasicBlockAddressMap(locked).find(BB);
  if (I == getBasicBlockAddressMap(locked).end()) {
    getBasicBlockAddressMap(locked)[BB] = Addr;
  } else {
    // ignore repeats: some BBs can be split into few MBBs?
  }
}

void JIT::clearPointerToBasicBlock(const BasicBlock *BB) {
  MutexGuard locked(lock);
  getBasicBlockAddressMap(locked).erase(BB);
}

void *JIT::getPointerToBasicBlock(BasicBlock *BB) {
  // make sure it's function is compiled by JIT
  (void)getPointerToFunction(BB->getParent());

  // resolve basic block address
  MutexGuard locked(lock);

  BasicBlockAddressMapTy::iterator I =
    getBasicBlockAddressMap(locked).find(BB);
  if (I != getBasicBlockAddressMap(locked).end()) {
    return I->second;
  } else {
    llvm_unreachable("JIT does not have BB address for address-of-label, was"
                     " it eliminated by optimizer?");
  }
}

void *JIT::getPointerToNamedFunction(const std::string &Name,
                                     bool AbortOnFailure){
  if (!isSymbolSearchingDisabled()) {
    void *ptr = JMM->getPointerToNamedFunction(Name, false);
    if (ptr)
      return ptr;
  }

  /// If a LazyFunctionCreator is installed, use it to get/create the function.
  if (LazyFunctionCreator)
    if (void *RP = LazyFunctionCreator(Name))
      return RP;

  if (AbortOnFailure) {
    report_fatal_error("Program used external function '"+Name+
                      "' which could not be resolved!");
  }
  return 0;
}


/// getOrEmitGlobalVariable - Return the address of the specified global
/// variable, possibly emitting it to memory if needed.  This is used by the
/// Emitter.
void *JIT::getOrEmitGlobalVariable(const GlobalVariable *GV) {
  MutexGuard locked(lock);

  void *Ptr = getPointerToGlobalIfAvailable(GV);
  if (Ptr) return Ptr;

  // If the global is external, just remember the address.
  if (GV->isDeclaration() || GV->hasAvailableExternallyLinkage()) {
#if HAVE___DSO_HANDLE
    if (GV->getName() == "__dso_handle")
      return (void*)&__dso_handle;
#endif
    Ptr = sys::DynamicLibrary::SearchForAddressOfSymbol(GV->getName());
    if (Ptr == 0) {
      report_fatal_error("Could not resolve external global address: "
                        +GV->getName());
    }
    addGlobalMapping(GV, Ptr);
  } else {
    // If the global hasn't been emitted to memory yet, allocate space and
    // emit it into memory.
    Ptr = getMemoryForGV(GV);
    addGlobalMapping(GV, Ptr);
    EmitGlobalVariable(GV);  // Initialize the variable.
  }
  return Ptr;
}

/// recompileAndRelinkFunction - This method is used to force a function
/// which has already been compiled, to be compiled again, possibly
/// after it has been modified. Then the entry to the old copy is overwritten
/// with a branch to the new copy. If there was no old copy, this acts
/// just like JIT::getPointerToFunction().
///
void *JIT::recompileAndRelinkFunction(Function *F) {
  void *OldAddr = getPointerToGlobalIfAvailable(F);

  // If it's not already compiled there is no reason to patch it up.
  if (OldAddr == 0) { return getPointerToFunction(F); }

  // Delete the old function mapping.
  addGlobalMapping(F, 0);

  // Recodegen the function
  runJITOnFunction(F);

  // Update state, forward the old function to the new function.
  void *Addr = getPointerToGlobalIfAvailable(F);
  assert(Addr && "Code generation didn't add function to GlobalAddress table!");
  TJI.replaceMachineCodeForFunction(OldAddr, Addr);
  return Addr;
}

/// getMemoryForGV - This method abstracts memory allocation of global
/// variable so that the JIT can allocate thread local variables depending
/// on the target.
///
char* JIT::getMemoryForGV(const GlobalVariable* GV) {
  char *Ptr;

  // GlobalVariable's which are not "constant" will cause trouble in a server
  // situation. It's returned in the same block of memory as code which may
  // not be writable.
  if (isGVCompilationDisabled() && !GV->isConstant()) {
    report_fatal_error("Compilation of non-internal GlobalValue is disabled!");
  }

  // Some applications require globals and code to live together, so they may
  // be allocated into the same buffer, but in general globals are allocated
  // through the memory manager which puts them near the code but not in the
  // same buffer.
  Type *GlobalType = GV->getType()->getElementType();
  size_t S = getDataLayout()->getTypeAllocSize(GlobalType);
  size_t A = getDataLayout()->getPreferredAlignment(GV);
  if (GV->isThreadLocal()) {
    MutexGuard locked(lock);
    Ptr = TJI.allocateThreadLocalMemory(S);
  } else if (TJI.allocateSeparateGVMemory()) {
    if (A <= 8) {
      Ptr = (char*)malloc(S);
    } else {
      // Allocate S+A bytes of memory, then use an aligned pointer within that
      // space.
      Ptr = (char*)malloc(S+A);
      unsigned MisAligned = ((intptr_t)Ptr & (A-1));
      Ptr = Ptr + (MisAligned ? (A-MisAligned) : 0);
    }
  } else if (AllocateGVsWithCode) {
    Ptr = (char*)JCE->allocateSpace(S, A);
  } else {
    Ptr = (char*)JCE->allocateGlobal(S, A);
  }
  return Ptr;
}

void JIT::addPendingFunction(Function *F) {
  MutexGuard locked(lock);
  jitstate->getPendingFunctions(locked).push_back(F);
}


JITEventListener::~JITEventListener() {}