-
-
Notifications
You must be signed in to change notification settings - Fork 30.7k
/
ceval.c
4915 lines (4459 loc) · 149 KB
/
ceval.c
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
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/* Execute compiled code */
/* XXX TO DO:
XXX speed up searching for keywords by using a dictionary
XXX document it!
*/
/* enable more aggressive intra-module optimizations, where available */
#define PY_LOCAL_AGGRESSIVE
#include "Python.h"
#include "code.h"
#include "frameobject.h"
#include "opcode.h"
#include "structmember.h"
#include <ctype.h>
#ifndef WITH_TSC
#define READ_TIMESTAMP(var)
#else
typedef unsigned long long uint64;
/* PowerPC support.
"__ppc__" appears to be the preprocessor definition to detect on OS X, whereas
"__powerpc__" appears to be the correct one for Linux with GCC
*/
#if defined(__ppc__) || defined (__powerpc__)
#define READ_TIMESTAMP(var) ppc_getcounter(&var)
static void
ppc_getcounter(uint64 *v)
{
unsigned long tbu, tb, tbu2;
loop:
asm volatile ("mftbu %0" : "=r" (tbu) );
asm volatile ("mftb %0" : "=r" (tb) );
asm volatile ("mftbu %0" : "=r" (tbu2));
if (__builtin_expect(tbu != tbu2, 0)) goto loop;
/* The slightly peculiar way of writing the next lines is
compiled better by GCC than any other way I tried. */
((long*)(v))[0] = tbu;
((long*)(v))[1] = tb;
}
#elif defined(__i386__)
/* this is for linux/x86 (and probably any other GCC/x86 combo) */
#define READ_TIMESTAMP(val) \
__asm__ __volatile__("rdtsc" : "=A" (val))
#elif defined(__x86_64__)
/* for gcc/x86_64, the "A" constraint in DI mode means *either* rax *or* rdx;
not edx:eax as it does for i386. Since rdtsc puts its result in edx:eax
even in 64-bit mode, we need to use "a" and "d" for the lower and upper
32-bit pieces of the result. */
#define READ_TIMESTAMP(val) do { \
unsigned int h, l; \
__asm__ __volatile__("rdtsc" : "=a" (l), "=d" (h)); \
(val) = ((uint64)l) | (((uint64)h) << 32); \
} while(0)
#else
#error "Don't know how to implement timestamp counter for this architecture"
#endif
void dump_tsc(int opcode, int ticked, uint64 inst0, uint64 inst1,
uint64 loop0, uint64 loop1, uint64 intr0, uint64 intr1)
{
uint64 intr, inst, loop;
PyThreadState *tstate = PyThreadState_Get();
if (!tstate->interp->tscdump)
return;
intr = intr1 - intr0;
inst = inst1 - inst0 - intr;
loop = loop1 - loop0 - intr;
fprintf(stderr, "opcode=%03d t=%d inst=%06lld loop=%06lld\n",
opcode, ticked, inst, loop);
}
#endif
/* Turn this on if your compiler chokes on the big switch: */
/* #define CASE_TOO_BIG 1 */
#ifdef Py_DEBUG
/* For debugging the interpreter: */
#define LLTRACE 1 /* Low-level trace feature */
#define CHECKEXC 1 /* Double-check exception checking */
#endif
typedef PyObject *(*callproc)(PyObject *, PyObject *, PyObject *);
/* Forward declarations */
#ifdef WITH_TSC
static PyObject * call_function(PyObject ***, int, uint64*, uint64*);
#else
static PyObject * call_function(PyObject ***, int);
#endif
static PyObject * fast_function(PyObject *, PyObject ***, int, int, int);
static PyObject * do_call(PyObject *, PyObject ***, int, int);
static PyObject * ext_do_call(PyObject *, PyObject ***, int, int, int);
static PyObject * update_keyword_args(PyObject *, int, PyObject ***,
PyObject *);
static PyObject * update_star_args(int, int, PyObject *, PyObject ***);
static PyObject * load_args(PyObject ***, int);
#define CALL_FLAG_VAR 1
#define CALL_FLAG_KW 2
#ifdef LLTRACE
static int lltrace;
static int prtrace(PyObject *, char *);
#endif
static int call_trace(Py_tracefunc, PyObject *,
PyThreadState *, PyFrameObject *,
int, PyObject *);
static int call_trace_protected(Py_tracefunc, PyObject *,
PyThreadState *, PyFrameObject *,
int, PyObject *);
static void call_exc_trace(Py_tracefunc, PyObject *,
PyThreadState *, PyFrameObject *);
static int maybe_call_line_trace(Py_tracefunc, PyObject *,
PyThreadState *, PyFrameObject *, int *, int *, int *);
static PyObject * cmp_outcome(int, PyObject *, PyObject *);
static PyObject * import_from(PyObject *, PyObject *);
static int import_all_from(PyObject *, PyObject *);
static void format_exc_check_arg(PyObject *, const char *, PyObject *);
static void format_exc_unbound(PyCodeObject *co, int oparg);
static PyObject * unicode_concatenate(PyObject *, PyObject *,
PyFrameObject *, unsigned char *);
static PyObject * special_lookup(PyObject *, _Py_Identifier *);
#define NAME_ERROR_MSG \
"name '%.200s' is not defined"
#define UNBOUNDLOCAL_ERROR_MSG \
"local variable '%.200s' referenced before assignment"
#define UNBOUNDFREE_ERROR_MSG \
"free variable '%.200s' referenced before assignment" \
" in enclosing scope"
/* Dynamic execution profile */
#ifdef DYNAMIC_EXECUTION_PROFILE
#ifdef DXPAIRS
static long dxpairs[257][256];
#define dxp dxpairs[256]
#else
static long dxp[256];
#endif
#endif
/* Function call profile */
#ifdef CALL_PROFILE
#define PCALL_NUM 11
static int pcall[PCALL_NUM];
#define PCALL_ALL 0
#define PCALL_FUNCTION 1
#define PCALL_FAST_FUNCTION 2
#define PCALL_FASTER_FUNCTION 3
#define PCALL_METHOD 4
#define PCALL_BOUND_METHOD 5
#define PCALL_CFUNCTION 6
#define PCALL_TYPE 7
#define PCALL_GENERATOR 8
#define PCALL_OTHER 9
#define PCALL_POP 10
/* Notes about the statistics
PCALL_FAST stats
FAST_FUNCTION means no argument tuple needs to be created.
FASTER_FUNCTION means that the fast-path frame setup code is used.
If there is a method call where the call can be optimized by changing
the argument tuple and calling the function directly, it gets recorded
twice.
As a result, the relationship among the statistics appears to be
PCALL_ALL == PCALL_FUNCTION + PCALL_METHOD - PCALL_BOUND_METHOD +
PCALL_CFUNCTION + PCALL_TYPE + PCALL_GENERATOR + PCALL_OTHER
PCALL_FUNCTION > PCALL_FAST_FUNCTION > PCALL_FASTER_FUNCTION
PCALL_METHOD > PCALL_BOUND_METHOD
*/
#define PCALL(POS) pcall[POS]++
PyObject *
PyEval_GetCallStats(PyObject *self)
{
return Py_BuildValue("iiiiiiiiiii",
pcall[0], pcall[1], pcall[2], pcall[3],
pcall[4], pcall[5], pcall[6], pcall[7],
pcall[8], pcall[9], pcall[10]);
}
#else
#define PCALL(O)
PyObject *
PyEval_GetCallStats(PyObject *self)
{
Py_INCREF(Py_None);
return Py_None;
}
#endif
#ifdef WITH_THREAD
#define GIL_REQUEST _Py_atomic_load_relaxed(&gil_drop_request)
#else
#define GIL_REQUEST 0
#endif
/* This can set eval_breaker to 0 even though gil_drop_request became
1. We believe this is all right because the eval loop will release
the GIL eventually anyway. */
#define COMPUTE_EVAL_BREAKER() \
_Py_atomic_store_relaxed( \
&eval_breaker, \
GIL_REQUEST | \
_Py_atomic_load_relaxed(&pendingcalls_to_do) | \
pending_async_exc)
#ifdef WITH_THREAD
#define SET_GIL_DROP_REQUEST() \
do { \
_Py_atomic_store_relaxed(&gil_drop_request, 1); \
_Py_atomic_store_relaxed(&eval_breaker, 1); \
} while (0)
#define RESET_GIL_DROP_REQUEST() \
do { \
_Py_atomic_store_relaxed(&gil_drop_request, 0); \
COMPUTE_EVAL_BREAKER(); \
} while (0)
#endif
/* Pending calls are only modified under pending_lock */
#define SIGNAL_PENDING_CALLS() \
do { \
_Py_atomic_store_relaxed(&pendingcalls_to_do, 1); \
_Py_atomic_store_relaxed(&eval_breaker, 1); \
} while (0)
#define UNSIGNAL_PENDING_CALLS() \
do { \
_Py_atomic_store_relaxed(&pendingcalls_to_do, 0); \
COMPUTE_EVAL_BREAKER(); \
} while (0)
#define SIGNAL_ASYNC_EXC() \
do { \
pending_async_exc = 1; \
_Py_atomic_store_relaxed(&eval_breaker, 1); \
} while (0)
#define UNSIGNAL_ASYNC_EXC() \
do { pending_async_exc = 0; COMPUTE_EVAL_BREAKER(); } while (0)
#ifdef WITH_THREAD
#ifdef HAVE_ERRNO_H
#include <errno.h>
#endif
#include "pythread.h"
static PyThread_type_lock pending_lock = 0; /* for pending calls */
static long main_thread = 0;
/* This single variable consolidates all requests to break out of the fast path
in the eval loop. */
static _Py_atomic_int eval_breaker = {0};
/* Request for dropping the GIL */
static _Py_atomic_int gil_drop_request = {0};
/* Request for running pending calls. */
static _Py_atomic_int pendingcalls_to_do = {0};
/* Request for looking at the `async_exc` field of the current thread state.
Guarded by the GIL. */
static int pending_async_exc = 0;
#include "ceval_gil.h"
int
PyEval_ThreadsInitialized(void)
{
return gil_created();
}
void
PyEval_InitThreads(void)
{
if (gil_created())
return;
create_gil();
take_gil(PyThreadState_GET());
main_thread = PyThread_get_thread_ident();
if (!pending_lock)
pending_lock = PyThread_allocate_lock();
}
void
_PyEval_FiniThreads(void)
{
if (!gil_created())
return;
destroy_gil();
assert(!gil_created());
}
void
PyEval_AcquireLock(void)
{
PyThreadState *tstate = PyThreadState_GET();
if (tstate == NULL)
Py_FatalError("PyEval_AcquireLock: current thread state is NULL");
take_gil(tstate);
}
void
PyEval_ReleaseLock(void)
{
/* This function must succeed when the current thread state is NULL.
We therefore avoid PyThreadState_GET() which dumps a fatal error
in debug mode.
*/
drop_gil((PyThreadState*)_Py_atomic_load_relaxed(
&_PyThreadState_Current));
}
void
PyEval_AcquireThread(PyThreadState *tstate)
{
if (tstate == NULL)
Py_FatalError("PyEval_AcquireThread: NULL new thread state");
/* Check someone has called PyEval_InitThreads() to create the lock */
assert(gil_created());
take_gil(tstate);
if (PyThreadState_Swap(tstate) != NULL)
Py_FatalError(
"PyEval_AcquireThread: non-NULL old thread state");
}
void
PyEval_ReleaseThread(PyThreadState *tstate)
{
if (tstate == NULL)
Py_FatalError("PyEval_ReleaseThread: NULL thread state");
if (PyThreadState_Swap(NULL) != tstate)
Py_FatalError("PyEval_ReleaseThread: wrong thread state");
drop_gil(tstate);
}
/* This function is called from PyOS_AfterFork to destroy all threads which are
* not running in the child process, and clear internal locks which might be
* held by those threads. (This could also be done using pthread_atfork
* mechanism, at least for the pthreads implementation.) */
void
PyEval_ReInitThreads(void)
{
_Py_IDENTIFIER(_after_fork);
PyObject *threading, *result;
PyThreadState *current_tstate = PyThreadState_GET();
if (!gil_created())
return;
recreate_gil();
pending_lock = PyThread_allocate_lock();
take_gil(current_tstate);
main_thread = PyThread_get_thread_ident();
/* Update the threading module with the new state.
*/
threading = PyMapping_GetItemString(current_tstate->interp->modules,
"threading");
if (threading == NULL) {
/* threading not imported */
PyErr_Clear();
return;
}
result = _PyObject_CallMethodId(threading, &PyId__after_fork, NULL);
if (result == NULL)
PyErr_WriteUnraisable(threading);
else
Py_DECREF(result);
Py_DECREF(threading);
/* Destroy all threads except the current one */
_PyThreadState_DeleteExcept(current_tstate);
}
#else
static _Py_atomic_int eval_breaker = {0};
static int pending_async_exc = 0;
#endif /* WITH_THREAD */
/* This function is used to signal that async exceptions are waiting to be
raised, therefore it is also useful in non-threaded builds. */
void
_PyEval_SignalAsyncExc(void)
{
SIGNAL_ASYNC_EXC();
}
/* Functions save_thread and restore_thread are always defined so
dynamically loaded modules needn't be compiled separately for use
with and without threads: */
PyThreadState *
PyEval_SaveThread(void)
{
PyThreadState *tstate = PyThreadState_Swap(NULL);
if (tstate == NULL)
Py_FatalError("PyEval_SaveThread: NULL tstate");
#ifdef WITH_THREAD
if (gil_created())
drop_gil(tstate);
#endif
return tstate;
}
void
PyEval_RestoreThread(PyThreadState *tstate)
{
if (tstate == NULL)
Py_FatalError("PyEval_RestoreThread: NULL tstate");
#ifdef WITH_THREAD
if (gil_created()) {
int err = errno;
take_gil(tstate);
/* _Py_Finalizing is protected by the GIL */
if (_Py_Finalizing && tstate != _Py_Finalizing) {
drop_gil(tstate);
PyThread_exit_thread();
assert(0); /* unreachable */
}
errno = err;
}
#endif
PyThreadState_Swap(tstate);
}
/* Mechanism whereby asynchronously executing callbacks (e.g. UNIX
signal handlers or Mac I/O completion routines) can schedule calls
to a function to be called synchronously.
The synchronous function is called with one void* argument.
It should return 0 for success or -1 for failure -- failure should
be accompanied by an exception.
If registry succeeds, the registry function returns 0; if it fails
(e.g. due to too many pending calls) it returns -1 (without setting
an exception condition).
Note that because registry may occur from within signal handlers,
or other asynchronous events, calling malloc() is unsafe!
#ifdef WITH_THREAD
Any thread can schedule pending calls, but only the main thread
will execute them.
There is no facility to schedule calls to a particular thread, but
that should be easy to change, should that ever be required. In
that case, the static variables here should go into the python
threadstate.
#endif
*/
#ifdef WITH_THREAD
/* The WITH_THREAD implementation is thread-safe. It allows
scheduling to be made from any thread, and even from an executing
callback.
*/
#define NPENDINGCALLS 32
static struct {
int (*func)(void *);
void *arg;
} pendingcalls[NPENDINGCALLS];
static int pendingfirst = 0;
static int pendinglast = 0;
int
Py_AddPendingCall(int (*func)(void *), void *arg)
{
int i, j, result=0;
PyThread_type_lock lock = pending_lock;
/* try a few times for the lock. Since this mechanism is used
* for signal handling (on the main thread), there is a (slim)
* chance that a signal is delivered on the same thread while we
* hold the lock during the Py_MakePendingCalls() function.
* This avoids a deadlock in that case.
* Note that signals can be delivered on any thread. In particular,
* on Windows, a SIGINT is delivered on a system-created worker
* thread.
* We also check for lock being NULL, in the unlikely case that
* this function is called before any bytecode evaluation takes place.
*/
if (lock != NULL) {
for (i = 0; i<100; i++) {
if (PyThread_acquire_lock(lock, NOWAIT_LOCK))
break;
}
if (i == 100)
return -1;
}
i = pendinglast;
j = (i + 1) % NPENDINGCALLS;
if (j == pendingfirst) {
result = -1; /* Queue full */
} else {
pendingcalls[i].func = func;
pendingcalls[i].arg = arg;
pendinglast = j;
}
/* signal main loop */
SIGNAL_PENDING_CALLS();
if (lock != NULL)
PyThread_release_lock(lock);
return result;
}
int
Py_MakePendingCalls(void)
{
static int busy = 0;
int i;
int r = 0;
if (!pending_lock) {
/* initial allocation of the lock */
pending_lock = PyThread_allocate_lock();
if (pending_lock == NULL)
return -1;
}
/* only service pending calls on main thread */
if (main_thread && PyThread_get_thread_ident() != main_thread)
return 0;
/* don't perform recursive pending calls */
if (busy)
return 0;
busy = 1;
/* perform a bounded number of calls, in case of recursion */
for (i=0; i<NPENDINGCALLS; i++) {
int j;
int (*func)(void *);
void *arg = NULL;
/* pop one item off the queue while holding the lock */
PyThread_acquire_lock(pending_lock, WAIT_LOCK);
j = pendingfirst;
if (j == pendinglast) {
func = NULL; /* Queue empty */
} else {
func = pendingcalls[j].func;
arg = pendingcalls[j].arg;
pendingfirst = (j + 1) % NPENDINGCALLS;
}
if (pendingfirst != pendinglast)
SIGNAL_PENDING_CALLS();
else
UNSIGNAL_PENDING_CALLS();
PyThread_release_lock(pending_lock);
/* having released the lock, perform the callback */
if (func == NULL)
break;
r = func(arg);
if (r)
break;
}
busy = 0;
return r;
}
#else /* if ! defined WITH_THREAD */
/*
WARNING! ASYNCHRONOUSLY EXECUTING CODE!
This code is used for signal handling in python that isn't built
with WITH_THREAD.
Don't use this implementation when Py_AddPendingCalls() can happen
on a different thread!
There are two possible race conditions:
(1) nested asynchronous calls to Py_AddPendingCall()
(2) AddPendingCall() calls made while pending calls are being processed.
(1) is very unlikely because typically signal delivery
is blocked during signal handling. So it should be impossible.
(2) is a real possibility.
The current code is safe against (2), but not against (1).
The safety against (2) is derived from the fact that only one
thread is present, interrupted by signals, and that the critical
section is protected with the "busy" variable. On Windows, which
delivers SIGINT on a system thread, this does not hold and therefore
Windows really shouldn't use this version.
The two threads could theoretically wiggle around the "busy" variable.
*/
#define NPENDINGCALLS 32
static struct {
int (*func)(void *);
void *arg;
} pendingcalls[NPENDINGCALLS];
static volatile int pendingfirst = 0;
static volatile int pendinglast = 0;
static _Py_atomic_int pendingcalls_to_do = {0};
int
Py_AddPendingCall(int (*func)(void *), void *arg)
{
static volatile int busy = 0;
int i, j;
/* XXX Begin critical section */
if (busy)
return -1;
busy = 1;
i = pendinglast;
j = (i + 1) % NPENDINGCALLS;
if (j == pendingfirst) {
busy = 0;
return -1; /* Queue full */
}
pendingcalls[i].func = func;
pendingcalls[i].arg = arg;
pendinglast = j;
SIGNAL_PENDING_CALLS();
busy = 0;
/* XXX End critical section */
return 0;
}
int
Py_MakePendingCalls(void)
{
static int busy = 0;
if (busy)
return 0;
busy = 1;
UNSIGNAL_PENDING_CALLS();
for (;;) {
int i;
int (*func)(void *);
void *arg;
i = pendingfirst;
if (i == pendinglast)
break; /* Queue empty */
func = pendingcalls[i].func;
arg = pendingcalls[i].arg;
pendingfirst = (i + 1) % NPENDINGCALLS;
if (func(arg) < 0) {
busy = 0;
SIGNAL_PENDING_CALLS(); /* We're not done yet */
return -1;
}
}
busy = 0;
return 0;
}
#endif /* WITH_THREAD */
/* The interpreter's recursion limit */
#ifndef Py_DEFAULT_RECURSION_LIMIT
#define Py_DEFAULT_RECURSION_LIMIT 1000
#endif
static int recursion_limit = Py_DEFAULT_RECURSION_LIMIT;
int _Py_CheckRecursionLimit = Py_DEFAULT_RECURSION_LIMIT;
int
Py_GetRecursionLimit(void)
{
return recursion_limit;
}
void
Py_SetRecursionLimit(int new_limit)
{
recursion_limit = new_limit;
_Py_CheckRecursionLimit = recursion_limit;
}
/* the macro Py_EnterRecursiveCall() only calls _Py_CheckRecursiveCall()
if the recursion_depth reaches _Py_CheckRecursionLimit.
If USE_STACKCHECK, the macro decrements _Py_CheckRecursionLimit
to guarantee that _Py_CheckRecursiveCall() is regularly called.
Without USE_STACKCHECK, there is no need for this. */
int
_Py_CheckRecursiveCall(char *where)
{
PyThreadState *tstate = PyThreadState_GET();
#ifdef USE_STACKCHECK
if (PyOS_CheckStack()) {
--tstate->recursion_depth;
PyErr_SetString(PyExc_MemoryError, "Stack overflow");
return -1;
}
#endif
_Py_CheckRecursionLimit = recursion_limit;
if (tstate->recursion_critical)
/* Somebody asked that we don't check for recursion. */
return 0;
if (tstate->overflowed) {
if (tstate->recursion_depth > recursion_limit + 50) {
/* Overflowing while handling an overflow. Give up. */
Py_FatalError("Cannot recover from stack overflow.");
}
return 0;
}
if (tstate->recursion_depth > recursion_limit) {
--tstate->recursion_depth;
tstate->overflowed = 1;
PyErr_Format(PyExc_RuntimeError,
"maximum recursion depth exceeded%s",
where);
return -1;
}
return 0;
}
/* Status code for main loop (reason for stack unwind) */
enum why_code {
WHY_NOT = 0x0001, /* No error */
WHY_EXCEPTION = 0x0002, /* Exception occurred */
WHY_RETURN = 0x0008, /* 'return' statement */
WHY_BREAK = 0x0010, /* 'break' statement */
WHY_CONTINUE = 0x0020, /* 'continue' statement */
WHY_YIELD = 0x0040, /* 'yield' operator */
WHY_SILENCED = 0x0080 /* Exception silenced by 'with' */
};
static void save_exc_state(PyThreadState *, PyFrameObject *);
static void swap_exc_state(PyThreadState *, PyFrameObject *);
static void restore_and_clear_exc_state(PyThreadState *, PyFrameObject *);
static int do_raise(PyObject *, PyObject *);
static int unpack_iterable(PyObject *, int, int, PyObject **);
/* Records whether tracing is on for any thread. Counts the number of
threads for which tstate->c_tracefunc is non-NULL, so if the value
is 0, we know we don't have to check this thread's c_tracefunc.
This speeds up the if statement in PyEval_EvalFrameEx() after
fast_next_opcode*/
static int _Py_TracingPossible = 0;
PyObject *
PyEval_EvalCode(PyObject *co, PyObject *globals, PyObject *locals)
{
return PyEval_EvalCodeEx(co,
globals, locals,
(PyObject **)NULL, 0,
(PyObject **)NULL, 0,
(PyObject **)NULL, 0,
NULL, NULL);
}
/* Interpreter main loop */
PyObject *
PyEval_EvalFrame(PyFrameObject *f) {
/* This is for backward compatibility with extension modules that
used this API; core interpreter code should call
PyEval_EvalFrameEx() */
return PyEval_EvalFrameEx(f, 0);
}
PyObject *
PyEval_EvalFrameEx(PyFrameObject *f, int throwflag)
{
#ifdef DXPAIRS
int lastopcode = 0;
#endif
PyObject **stack_pointer; /* Next free slot in value stack */
unsigned char *next_instr;
int opcode; /* Current opcode */
int oparg; /* Current opcode argument, if any */
enum why_code why; /* Reason for block stack unwind */
PyObject **fastlocals, **freevars;
PyObject *retval = NULL; /* Return value */
PyThreadState *tstate = PyThreadState_GET();
PyCodeObject *co;
/* when tracing we set things up so that
not (instr_lb <= current_bytecode_offset < instr_ub)
is true when the line being executed has changed. The
initial values are such as to make this false the first
time it is tested. */
int instr_ub = -1, instr_lb = 0, instr_prev = -1;
unsigned char *first_instr;
PyObject *names;
PyObject *consts;
#ifdef LLTRACE
_Py_IDENTIFIER(__ltrace__);
#endif
/* Computed GOTOs, or
the-optimization-commonly-but-improperly-known-as-"threaded code"
using gcc's labels-as-values extension
(http://gcc.gnu.org/onlinedocs/gcc/Labels-as-Values.html).
The traditional bytecode evaluation loop uses a "switch" statement, which
decent compilers will optimize as a single indirect branch instruction
combined with a lookup table of jump addresses. However, since the
indirect jump instruction is shared by all opcodes, the CPU will have a
hard time making the right prediction for where to jump next (actually,
it will be always wrong except in the uncommon case of a sequence of
several identical opcodes).
"Threaded code" in contrast, uses an explicit jump table and an explicit
indirect jump instruction at the end of each opcode. Since the jump
instruction is at a different address for each opcode, the CPU will make a
separate prediction for each of these instructions, which is equivalent to
predicting the second opcode of each opcode pair. These predictions have
a much better chance to turn out valid, especially in small bytecode loops.
A mispredicted branch on a modern CPU flushes the whole pipeline and
can cost several CPU cycles (depending on the pipeline depth),
and potentially many more instructions (depending on the pipeline width).
A correctly predicted branch, however, is nearly free.
At the time of this writing, the "threaded code" version is up to 15-20%
faster than the normal "switch" version, depending on the compiler and the
CPU architecture.
We disable the optimization if DYNAMIC_EXECUTION_PROFILE is defined,
because it would render the measurements invalid.
NOTE: care must be taken that the compiler doesn't try to "optimize" the
indirect jumps by sharing them between all opcodes. Such optimizations
can be disabled on gcc by using the -fno-gcse flag (or possibly
-fno-crossjumping).
*/
#ifdef DYNAMIC_EXECUTION_PROFILE
#undef USE_COMPUTED_GOTOS
#define USE_COMPUTED_GOTOS 0
#endif
#ifdef HAVE_COMPUTED_GOTOS
#ifndef USE_COMPUTED_GOTOS
#define USE_COMPUTED_GOTOS 1
#endif
#else
#if defined(USE_COMPUTED_GOTOS) && USE_COMPUTED_GOTOS
#error "Computed gotos are not supported on this compiler."
#endif
#undef USE_COMPUTED_GOTOS
#define USE_COMPUTED_GOTOS 0
#endif
#if USE_COMPUTED_GOTOS
/* Import the static jump table */
#include "opcode_targets.h"
/* This macro is used when several opcodes defer to the same implementation
(e.g. SETUP_LOOP, SETUP_FINALLY) */
#define TARGET_WITH_IMPL(op, impl) \
TARGET_##op: \
opcode = op; \
if (HAS_ARG(op)) \
oparg = NEXTARG(); \
case op: \
goto impl; \
#define TARGET(op) \
TARGET_##op: \
opcode = op; \
if (HAS_ARG(op)) \
oparg = NEXTARG(); \
case op:
#define DISPATCH() \
{ \
if (!_Py_atomic_load_relaxed(&eval_breaker)) { \
FAST_DISPATCH(); \
} \
continue; \
}
#ifdef LLTRACE
#define FAST_DISPATCH() \
{ \
if (!lltrace && !_Py_TracingPossible) { \
f->f_lasti = INSTR_OFFSET(); \
goto *opcode_targets[*next_instr++]; \
} \
goto fast_next_opcode; \
}
#else
#define FAST_DISPATCH() \
{ \
if (!_Py_TracingPossible) { \
f->f_lasti = INSTR_OFFSET(); \
goto *opcode_targets[*next_instr++]; \
} \
goto fast_next_opcode; \
}
#endif
#else
#define TARGET(op) \
case op:
#define TARGET_WITH_IMPL(op, impl) \
/* silence compiler warnings about `impl` unused */ \
if (0) goto impl; \
case op:
#define DISPATCH() continue
#define FAST_DISPATCH() goto fast_next_opcode
#endif
/* Tuple access macros */
#ifndef Py_DEBUG
#define GETITEM(v, i) PyTuple_GET_ITEM((PyTupleObject *)(v), (i))
#else
#define GETITEM(v, i) PyTuple_GetItem((v), (i))
#endif
#ifdef WITH_TSC
/* Use Pentium timestamp counter to mark certain events:
inst0 -- beginning of switch statement for opcode dispatch
inst1 -- end of switch statement (may be skipped)
loop0 -- the top of the mainloop
loop1 -- place where control returns again to top of mainloop
(may be skipped)
intr1 -- beginning of long interruption
intr2 -- end of long interruption
Many opcodes call out to helper C functions. In some cases, the
time in those functions should be counted towards the time for the
opcode, but not in all cases. For example, a CALL_FUNCTION opcode
calls another Python function; there's no point in charge all the
bytecode executed by the called function to the caller.
It's hard to make a useful judgement statically. In the presence
of operator overloading, it's impossible to tell if a call will
execute new Python code or not.
It's a case-by-case judgement. I'll use intr1 for the following
cases:
IMPORT_STAR
IMPORT_FROM
CALL_FUNCTION (and friends)
*/
uint64 inst0, inst1, loop0, loop1, intr0 = 0, intr1 = 0;
int ticked = 0;
READ_TIMESTAMP(inst0);
READ_TIMESTAMP(inst1);
READ_TIMESTAMP(loop0);
READ_TIMESTAMP(loop1);
/* shut up the compiler */
opcode = 0;
#endif
/* Code access macros */
#define INSTR_OFFSET() ((int)(next_instr - first_instr))
#define NEXTOP() (*next_instr++)
#define NEXTARG() (next_instr += 2, (next_instr[-1]<<8) + next_instr[-2])
#define PEEKARG() ((next_instr[2]<<8) + next_instr[1])
#define JUMPTO(x) (next_instr = first_instr + (x))
#define JUMPBY(x) (next_instr += (x))