1Gc(3) OCaml library Gc(3)
2
3
4
6 Gc - Memory management control and statistics; finalised values.
7
9 Module Gc
10
12 Module Gc
13 : sig end
14
15
16 Memory management control and statistics; finalised values.
17
18
19
20
21
22 type stat = {
23 minor_words : float ; (* Number of words allocated in the minor heap
24 since the program was started.
25 *)
26 promoted_words : float ; (* Number of words allocated in the minor
27 heap that survived a minor collection and were moved to the major heap
28 since the program was started.
29 *)
30 major_words : float ; (* Number of words allocated in the major heap,
31 including the promoted words, since the program was started.
32 *)
33 minor_collections : int ; (* Number of minor collections since the
34 program was started.
35 *)
36 major_collections : int ; (* Number of major collection cycles com‐
37 pleted since the program was started.
38 *)
39 heap_words : int ; (* Total size of the major heap, in words.
40 *)
41 heap_chunks : int ; (* Number of contiguous pieces of memory that
42 make up the major heap.
43 *)
44 live_words : int ; (* Number of words of live data in the major heap,
45 including the header words.
46 *)
47 live_blocks : int ; (* Number of live blocks in the major heap.
48 *)
49 free_words : int ; (* Number of words in the free list.
50 *)
51 free_blocks : int ; (* Number of blocks in the free list.
52 *)
53 largest_free : int ; (* Size (in words) of the largest block in the
54 free list.
55 *)
56 fragments : int ; (* Number of wasted words due to fragmentation.
57 These are 1-words free blocks placed between two live blocks. They are
58 not available for allocation.
59 *)
60 compactions : int ; (* Number of heap compactions since the program
61 was started.
62 *)
63 top_heap_words : int ; (* Maximum size reached by the major heap, in
64 words.
65 *)
66 stack_size : int ; (* Current size of the stack, in words.
67
68
69 Since 3.12.0
70 *)
71 }
72
73
74 The memory management counters are returned in a stat record.
75
76 The total amount of memory allocated by the program since it was
77 started is (in words) minor_words + major_words - promoted_words .
78 Multiply by the word size (4 on a 32-bit machine, 8 on a 64-bit
79 machine) to get the number of bytes.
80
81
82 type control = {
83
84 mutable minor_heap_size : int ; (* The size (in words) of the minor
85 heap. Changing this parameter will trigger a minor collection.
86 Default: 256k.
87 *)
88
89 mutable major_heap_increment : int ; (* How much to add to the major
90 heap when increasing it. If this number is less than or equal to 1000,
91 it is a percentage of the current heap size (i.e. setting it to 100
92 will double the heap size at each increase). If it is more than 1000,
93 it is a fixed number of words that will be added to the heap. Default:
94 15.
95 *)
96
97 mutable space_overhead : int ; (* The major GC speed is computed from
98 this parameter. This is the memory that will be "wasted" because the
99 GC does not immediately collect unreachable blocks. It is expressed as
100 a percentage of the memory used for live data. The GC will work more
101 (use more CPU time and collect blocks more eagerly) if space_overhead
102 is smaller. Default: 80.
103 *)
104
105 mutable verbose : int ; (* This value controls the GC messages on
106 standard error output. It is a sum of some of the following flags, to
107 print messages on the corresponding events:
108
109 - 0x001 Start of major GC cycle.
110
111 - 0x002 Minor collection and major GC slice.
112
113 - 0x004 Growing and shrinking of the heap.
114
115 - 0x008 Resizing of stacks and memory manager tables.
116
117 - 0x010 Heap compaction.
118
119 - 0x020 Change of GC parameters.
120
121 - 0x040 Computation of major GC slice size.
122
123 - 0x080 Calling of finalisation functions.
124
125 - 0x100 Bytecode executable and shared library search at start-up.
126
127 - 0x200 Computation of compaction-triggering condition.
128
129 - 0x400 Output GC statistics at program exit. Default: 0.
130
131 *)
132
133 mutable max_overhead : int ; (* Heap compaction is triggered when the
134 estimated amount of "wasted" memory is more than max_overhead percent
135 of the amount of live data. If max_overhead is set to 0, heap com‐
136 paction is triggered at the end of each major GC cycle (this setting is
137 intended for testing purposes only). If max_overhead >= 1000000 , com‐
138 paction is never triggered. If compaction is permanently disabled, it
139 is strongly suggested to set allocation_policy to 2. Default: 500.
140 *)
141
142 mutable stack_limit : int ; (* The maximum size of the stack (in
143 words). This is only relevant to the byte-code runtime, as the native
144 code runtime uses the operating system's stack. Default: 1024k.
145 *)
146
147 mutable allocation_policy : int ; (* The policy used for allocating in
148 the major heap. Possible values are 0, 1 and 2.
149
150
151 -0 is the next-fit policy, which is usually fast but can result in
152 fragmentation, increasing memory consumption.
153
154
155 -1 is the first-fit policy, which avoids fragmentation but has corner
156 cases (in certain realistic workloads) where it is sensibly slower.
157
158
159 -2 is the best-fit policy, which is fast and avoids fragmentation. In
160 our experiments it is faster and uses less memory than both next-fit
161 and first-fit. (since OCaml 4.10)
162
163 The current default is next-fit, as the best-fit policy is new and not
164 yet widely tested. We expect best-fit to become the default in the
165 future.
166
167 On one example that was known to be bad for next-fit and first-fit,
168 next-fit takes 28s using 855Mio of memory, first-fit takes 47s using
169 566Mio of memory, best-fit takes 27s using 545Mio of memory.
170
171 Note: When changing to a low-fragmentation policy, you may need to aug‐
172 ment the space_overhead setting, for example using 100 instead of the
173 default 80 which is tuned for next-fit. Indeed, the difference in frag‐
174 mentation behavior means that different policies will have different
175 proportion of "wasted space" for a given program. Less fragmentation
176 means a smaller heap so, for the same amount of wasted space, a higher
177 proportion of wasted space. This makes the GC work harder, unless you
178 relax it by increasing space_overhead .
179
180 Note: changing the allocation policy at run-time forces a heap com‐
181 paction, which is a lengthy operation unless the heap is small (e.g. at
182 the start of the program).
183
184 Default: 0.
185
186
187 Since 3.11.0
188 *)
189 window_size : int ; (* The size of the window used by the major GC
190 for smoothing out variations in its workload. This is an integer
191 between 1 and 50. Default: 1.
192
193
194 Since 4.03.0
195 *)
196 custom_major_ratio : int ; (* Target ratio of floating garbage to
197 major heap size for out-of-heap memory held by custom values located in
198 the major heap. The GC speed is adjusted to try to use this much memory
199 for dead values that are not yet collected. Expressed as a percentage
200 of major heap size. The default value keeps the out-of-heap floating
201 garbage about the same size as the in-heap overhead. Note: this only
202 applies to values allocated with caml_alloc_custom_mem (e.g. bigar‐
203 rays). Default: 44.
204
205
206 Since 4.08.0
207 *)
208 custom_minor_ratio : int ; (* Bound on floating garbage for
209 out-of-heap memory held by custom values in the minor heap. A minor GC
210 is triggered when this much memory is held by custom values located in
211 the minor heap. Expressed as a percentage of minor heap size. Note:
212 this only applies to values allocated with caml_alloc_custom_mem (e.g.
213 bigarrays). Default: 100.
214
215
216 Since 4.08.0
217 *)
218 custom_minor_max_size : int ; (* Maximum amount of out-of-heap memory
219 for each custom value allocated in the minor heap. When a custom value
220 is allocated on the minor heap and holds more than this many bytes,
221 only this value is counted against custom_minor_ratio and the rest is
222 directly counted against custom_major_ratio . Note: this only applies
223 to values allocated with caml_alloc_custom_mem (e.g. bigarrays).
224 Default: 8192 bytes.
225
226
227 Since 4.08.0
228 *)
229 }
230
231
232 The GC parameters are given as a control record. Note that these
233 parameters can also be initialised by setting the OCAMLRUNPARAM envi‐
234 ronment variable. See the documentation of ocamlrun .
235
236
237
238 val stat : unit -> stat
239
240 Return the current values of the memory management counters in a stat
241 record. This function examines every heap block to get the statistics.
242
243
244
245 val quick_stat : unit -> stat
246
247 Same as stat except that live_words , live_blocks , free_words ,
248 free_blocks , largest_free , and fragments are set to 0. This function
249 is much faster than stat because it does not need to go through the
250 heap.
251
252
253
254 val counters : unit -> float * float * float
255
256 Return (minor_words, promoted_words, major_words) . This function is
257 as fast as quick_stat .
258
259
260
261 val minor_words : unit -> float
262
263 Number of words allocated in the minor heap since the program was
264 started. This number is accurate in byte-code programs, but only an
265 approximation in programs compiled to native code.
266
267 In native code this function does not allocate.
268
269
270 Since 4.04
271
272
273
274 val get : unit -> control
275
276 Return the current values of the GC parameters in a control record.
277
278
279
280 val set : control -> unit
281
282
283 set r changes the GC parameters according to the control record r .
284 The normal usage is: Gc.set { (Gc.get()) with Gc.verbose = 0x00d }
285
286
287
288
289 val minor : unit -> unit
290
291 Trigger a minor collection.
292
293
294
295 val major_slice : int -> int
296
297
298 major_slice n Do a minor collection and a slice of major collection. n
299 is the size of the slice: the GC will do enough work to free (on aver‐
300 age) n words of memory. If n = 0, the GC will try to do enough work to
301 ensure that the next automatic slice has no work to do. This function
302 returns an unspecified integer (currently: 0).
303
304
305
306 val major : unit -> unit
307
308 Do a minor collection and finish the current major collection cycle.
309
310
311
312 val full_major : unit -> unit
313
314 Do a minor collection, finish the current major collection cycle, and
315 perform a complete new cycle. This will collect all currently unreach‐
316 able blocks.
317
318
319
320 val compact : unit -> unit
321
322 Perform a full major collection and compact the heap. Note that heap
323 compaction is a lengthy operation.
324
325
326
327 val print_stat : out_channel -> unit
328
329 Print the current values of the memory management counters (in
330 human-readable form) into the channel argument.
331
332
333
334 val allocated_bytes : unit -> float
335
336 Return the total number of bytes allocated since the program was
337 started. It is returned as a float to avoid overflow problems with int
338 on 32-bit machines.
339
340
341
342 val get_minor_free : unit -> int
343
344 Return the current size of the free space inside the minor heap.
345
346
347 Since 4.03.0
348
349
350
351 val get_bucket : int -> int
352
353
354 get_bucket n returns the current size of the n -th future bucket of the
355 GC smoothing system. The unit is one millionth of a full GC.
356
357
358 Since 4.03.0
359
360
361 Raises Invalid_argument if n is negative, return 0 if n is larger than
362 the smoothing window.
363
364
365
366 val get_credit : unit -> int
367
368
369 get_credit () returns the current size of the "work done in advance"
370 counter of the GC smoothing system. The unit is one millionth of a full
371 GC.
372
373
374 Since 4.03.0
375
376
377
378 val huge_fallback_count : unit -> int
379
380 Return the number of times we tried to map huge pages and had to fall
381 back to small pages. This is always 0 if OCAMLRUNPARAM contains H=1 .
382
383
384 Since 4.03.0
385
386
387
388 val finalise : ('a -> unit) -> 'a -> unit
389
390
391 finalise f v registers f as a finalisation function for v . v must be
392 heap-allocated. f will be called with v as argument at some point
393 between the first time v becomes unreachable (including through weak
394 pointers) and the time v is collected by the GC. Several functions can
395 be registered for the same value, or even several instances of the same
396 function. Each instance will be called once (or never, if the program
397 terminates before v becomes unreachable).
398
399 The GC will call the finalisation functions in the order of dealloca‐
400 tion. When several values become unreachable at the same time (i.e.
401 during the same GC cycle), the finalisation functions will be called in
402 the reverse order of the corresponding calls to finalise . If finalise
403 is called in the same order as the values are allocated, that means
404 each value is finalised before the values it depends upon. Of course,
405 this becomes false if additional dependencies are introduced by assign‐
406 ments.
407
408 In the presence of multiple OCaml threads it should be assumed that any
409 particular finaliser may be executed in any of the threads.
410
411 Anything reachable from the closure of finalisation functions is con‐
412 sidered reachable, so the following code will not work as expected:
413
414 - let v = ... in Gc.finalise (fun _ -> ...v...) v
415
416 Instead you should make sure that v is not in the closure of the final‐
417 isation function by writing:
418
419 - let f = fun x -> ... let v = ... in Gc.finalise f v
420
421 The f function can use all features of OCaml, including assignments
422 that make the value reachable again. It can also loop forever (in this
423 case, the other finalisation functions will not be called during the
424 execution of f, unless it calls finalise_release ). It can call
425 finalise on v or other values to register other functions or even
426 itself. It can raise an exception; in this case the exception will
427 interrupt whatever the program was doing when the function was called.
428
429
430 finalise will raise Invalid_argument if v is not guaranteed to be
431 heap-allocated. Some examples of values that are not heap-allocated
432 are integers, constant constructors, booleans, the empty array, the
433 empty list, the unit value. The exact list of what is heap-allocated
434 or not is implementation-dependent. Some constant values can be
435 heap-allocated but never deallocated during the lifetime of the pro‐
436 gram, for example a list of integer constants; this is also implementa‐
437 tion-dependent. Note that values of types float are sometimes allo‐
438 cated and sometimes not, so finalising them is unsafe, and finalise
439 will also raise Invalid_argument for them. Values of type 'a Lazy.t
440 (for any 'a ) are like float in this respect, except that the compiler
441 sometimes optimizes them in a way that prevents finalise from detecting
442 them. In this case, it will not raise Invalid_argument , but you should
443 still avoid calling finalise on lazy values.
444
445 The results of calling String.make , Bytes.make , Bytes.create ,
446 Array.make , and ref are guaranteed to be heap-allocated and non-con‐
447 stant except when the length argument is 0 .
448
449
450
451 val finalise_last : (unit -> unit) -> 'a -> unit
452
453 same as Gc.finalise except the value is not given as argument. So you
454 can't use the given value for the computation of the finalisation func‐
455 tion. The benefit is that the function is called after the value is
456 unreachable for the last time instead of the first time. So contrary to
457 Gc.finalise the value will never be reachable again or used again. In
458 particular every weak pointer and ephemeron that contained this value
459 as key or data is unset before running the finalisation function. More‐
460 over the finalisation functions attached with Gc.finalise are always
461 called before the finalisation functions attached with Gc.finalise_last
462 .
463
464
465 Since 4.04
466
467
468
469 val finalise_release : unit -> unit
470
471 A finalisation function may call finalise_release to tell the GC that
472 it can launch the next finalisation function without waiting for the
473 current one to return.
474
475
476 type alarm
477
478
479 An alarm is a piece of data that calls a user function at the end of
480 each major GC cycle. The following functions are provided to create
481 and delete alarms.
482
483
484
485 val create_alarm : (unit -> unit) -> alarm
486
487
488 create_alarm f will arrange for f to be called at the end of each major
489 GC cycle, starting with the current cycle or the next one. A value of
490 type alarm is returned that you can use to call delete_alarm .
491
492
493
494 val delete_alarm : alarm -> unit
495
496
497 delete_alarm a will stop the calls to the function associated to a .
498 Calling delete_alarm a again has no effect.
499
500
501
502 val eventlog_pause : unit -> unit
503
504
505 eventlog_pause () will pause the collection of traces in the runtime.
506 Traces are collected if the program is linked to the instrumented run‐
507 time and started with the environment variable OCAML_EVENTLOG_ENABLED.
508 Events are flushed to disk after pausing, and no new events will be
509 recorded until eventlog_resume is called.
510
511
512
513 val eventlog_resume : unit -> unit
514
515
516 eventlog_resume () will resume the collection of traces in the runtime.
517 Traces are collected if the program is linked to the instrumented run‐
518 time and started with the environment variable OCAML_EVENTLOG_ENABLED.
519 This call can be used after calling eventlog_pause , or if the program
520 was started with OCAML_EVENTLOG_ENABLED=p. (which pauses the collection
521 of traces before the first event.)
522
523
524 module Memprof : sig end
525
526
527
528 Memprof is a sampling engine for allocated memory words. Every allo‐
529 cated word has a probability of being sampled equal to a configurable
530 sampling rate. Once a block is sampled, it becomes tracked. A tracked
531 block triggers a user-defined callback as soon as it is allocated, pro‐
532 moted or deallocated.
533
534 Since blocks are composed of several words, a block can potentially be
535 sampled several times. If a block is sampled several times, then each
536 of the callback is called once for each event of this block: the multi‐
537 plicity is given in the n_samples field of the allocation structure.
538
539 This engine makes it possible to implement a low-overhead memory pro‐
540 filer as an OCaml library.
541
542 Note: this API is EXPERIMENTAL. It may change without prior notice.
543
544
545
546
547
548OCamldoc 2020-09-01 Gc(3)