1erl(1) User Commands erl(1)
2
6 erl - The Erlang emulator.
7
9 The erl program starts an Erlang runtime system. The exact details (for
10 example, whether erl is a script or a program and which other programs
11 it calls) are system-dependent.
12 Windows users probably want to use the werl program instead, which runs
14 in its own window with scrollbars and supports command-line editing.
15 The erl program on Windows provides no line editing in its shell, and
16 on Windows 95 there is no way to scroll back to text that has scrolled
17 off the screen. The erl program must be used, however, in pipelines or
18 if you want to redirect standard input or output.
19 Note:
21 As from ERTS 5.9 (Erlang/OTP R15B) the runtime system does by default
22 not bind schedulers to logical processors. For more information, see
23 system flag +sbt.
24
27 erl <arguments>
28 Starts an Erlang runtime system.
30 The arguments can be divided into emulator flags, flags, and
32 plain arguments:
33 * Any argument starting with character + is interpreted as an
35 emulator flag.
36 As indicated by the name, emulator flags control the behav‐
38 ior of the emulator.
39 * Any argument starting with character - (hyphen) is inter‐
41 preted as a flag, which is to be passed to the Erlang part
42 of the runtime system, more specifically to the init system
43 process, see init(3).
44 The init process itself interprets some of these flags, the
46 init flags. It also stores any remaining flags, the user
47 flags. The latter can be retrieved by calling init:get_argu‐
48 ment/1.
49 A small number of "-" flags exist, which now actually are
51 emulator flags, see the description below.
52 * Plain arguments are not interpreted in any way. They are
54 also stored by the init process and can be retrieved by
55 calling init:get_plain_arguments/0. Plain arguments can oc‐
56 cur before the first flag, or after a -- flag. Also, the
57 -extra flag causes everything that follows to become plain
58 arguments.
59 Examples:
61 % erl +W w -sname arnie +R 9 -s my_init -extra +bertie
63 (arnie@host)1> init:get_argument(sname).
64 {ok,[["arnie"]]}
65 (arnie@host)2> init:get_plain_arguments().
66 ["+bertie"]
67 Here +W w and +R 9 are emulator flags. -s my_init is an init
69 flag, interpreted by init. -sname arnie is a user flag, stored
70 by init. It is read by Kernel and causes the Erlang runtime sys‐
71 tem to become distributed. Finally, everything after -extra
72 (that is, +bertie) is considered as plain arguments.
73 % erl -myflag 1
75 1> init:get_argument(myflag).
76 {ok,[["1"]]}
77 2> init:get_plain_arguments().
78 []
79 Here the user flag -myflag 1 is passed to and stored by the init
81 process. It is a user-defined flag, presumably used by some
82 user-defined application.
83
85 In the following list, init flags are marked "(init flag)". Unless oth‐
86 erwise specified, all other flags are user flags, for which the values
87 can be retrieved by calling init:get_argument/1. Notice that the list
88 of user flags is not exhaustive, there can be more application-specific
89 flags that instead are described in the corresponding application docu‐
90 mentation.
91 -- (init flag):
93 Everything following -- up to the next flag (-flag or +flag) is
94 considered plain arguments and can be retrieved using
95 init:get_plain_arguments/0.
96 -Application Par Val:
98 Sets the application configuration parameter Par to the value Val
99 for the application Application; see app(4) and application(3).
100 -args_file FileName:
102 Command-line arguments are read from the file FileName. The argu‐
103 ments read from the file replace flag '-args_file FileName' on the
104 resulting command line.
105 The file FileName is to be a plain text file and can contain com‐
107 ments and command-line arguments. A comment begins with a # charac‐
108 ter and continues until the next end of line character. Backslash
109 (\\) is used as quoting character. All command-line arguments ac‐
110 cepted by erl are allowed, also flag -args_file FileName. Be care‐
111 ful not to cause circular dependencies between files containing
112 flag -args_file, though.
113 The flag -extra is treated in special way. Its scope ends at the
115 end of the file. Arguments following an -extra flag are moved on
116 the command line into the -extra section, that is, the end of the
117 command line following after an -extra flag.
118 -async_shell_start:
120 The initial Erlang shell does not read user input until the system
121 boot procedure has been completed (Erlang/OTP 5.4 and later). This
122 flag disables the start synchronization feature and lets the shell
123 start in parallel with the rest of the system.
124 -boot File:
126 Specifies the name of the boot file, File.boot, which is used to
127 start the system; see init(3). Unless File contains an absolute
128 path, the system searches for File.boot in the current and
129 $ROOT/bin directories.
130 Defaults to $ROOT/bin/start.boot.
132 -boot_var Var Dir:
134 If the boot script contains a path variable Var other than $ROOT,
135 this variable is expanded to Dir. Used when applications are in‐
136 stalled in another directory than $ROOT/lib; see sys‐
137 tools:make_script/1,2 in SASL.
138 -code_path_cache:
140 Enables the code path cache of the code server; see code(3).
141 -compile Mod1 Mod2 ...:
143 Compiles the specified modules and then terminates (with non-zero
144 exit code if the compilation of some file did not succeed). Implies
145 -noinput.
146 Not recommended; use erlc instead.
148 -config Config [Config ...]:
150 Specifies the name of one or more configuration files, Config.con‐
151 fig, which is used to configure applications; see app(4) and appli‐
152 cation(3). See the documentation for the configuration file format
153 for a description of the configuration format and the order in
154 which configuration parameters are read.
155 -configfd FD [FD ...]:
157 Specifies the name of one or more file descriptors (called configu‐
158 ration file descriptors from here on) with configuration data for
159 applications; see app(4) and application(3). See the documentation
160 for the configuration file format for a description of the configu‐
161 ration format and the order in which configuration parameters are
162 read.
163 A configuration file descriptor will be read until its end and will
165 then be closed.
166 The content of a configuration file descriptor is stored so that it
168 can be reused when init:restart/0 or init:restart/1 is called.
169 The parameter -configfd 0 implies -noinput.
171 Note:
173 It is not recommended to use file descriptors 1 (standard output),
174 and 2 (standard error) together with -configfd as these file descrip‐
175 tors are typically used to print information to the console the pro‐
176 gram is running in.
177 Examples (Unix shell):
180 $ erl \ -noshell \ -configfd 3 \ -eval \ 'io:format("~p~n",[application:get_env(kernel, logger_level)]),erlang:halt()' 3< \ <(echo '[{kernel, [{logger_level, warning}]}].')
182 {ok,warning}
183 $ echo '[{kernel, [{logger_level, warning}]}].' > test1.config
186 $ echo '[{kernel, [{logger_level, error}]}].' > test2.config
187 $ erl \ -noshell \ -configfd 3 \ -configfd 4 \ -eval \ 'io:format("~p~n",[application:get_env(kernel, logger_level)]),erlang:halt()' \ 3< test1.config 4< test2.config
188 {ok,error}
189 -connect_all false:
192 This flag is deprecated and has been replaced by the kernel appli‐
193 cation parameter connect_all.
194 -cookie Cookie:
196 Obsolete flag without any effect and common misspelling for -set‐
197 cookie. Use -setcookie instead.
198 -detached:
200 Starts the Erlang runtime system detached from the system console.
201 Useful for running daemons and backgrounds processes. Implies
202 -noinput.
203 -disable-feature feature:
205 Disables the feature feature in the runtime system. The special
206 feature all can be used to disable all non permanent features.
207 -dist_listen true|false:
209 Specifies whether this node should be listening for incoming dis‐
210 tribution connections or not. By default a node will listen for in‐
211 coming connections. Setting this option to false implies -hidden.
212 -emu_args:
214 Useful for debugging. Prints the arguments sent to the emulator.
215 -emu_flavor emu|jit|smp:
217 Start an emulator of a different flavor. Normally only one flavor
218 is available, more can be added by building specific flavors. The
219 currently available flavors are: emu and jit. The smp flavor is an
220 alias for the current default flavor. You can combine this flag
221 with --emu_type. You can get the current flavor at run-time using
222 erlang:system_info(emu_flavor). (The emulator with this flavor must
223 be built. You can build a specific flavor by doing make FLA‐
224 VOR=$FLAVOR in the Erlang/OTP source repository.)
225 -emu_type Type:
227 Start an emulator of a different type. For example, to start the
228 lock-counter emulator, use -emu_type lcnt. You can get the current
229 type at run-time using erlang:system_info(build_type). (The emula‐
230 tor of this type must already be built. Use the configure option
231 --enable-lock-counter to build the lock-counter emulator.)
232 -enable-feature feature:
234 Enables the feature feature in the runtime system. The special fea‐
235 ture all can be used to enable all features.
236 -env Variable Value:
238 Sets the host OS environment variable Variable to the value Value
239 for the Erlang runtime system. Example:
240 % erl -env DISPLAY gin:0
242 In this example, an Erlang runtime system is started with environ‐
244 ment variable DISPLAY set to gin:0.
245 -epmd_module Module (init flag):
247 Configures the module responsible to communicate to epmd. Defaults
248 to erl_epmd.
249 -erl_epmd_port Port (init flag):
251 Configures the port used by erl_epmd to listen for connection and
252 connect to other nodes. See erl_epmd for more details. Defaults to
253 0.
254 -eval Expr (init flag):
256 Makes init evaluate the expression Expr; see init(3).
257 -extra (init flag):
259 Everything following -extra is considered plain arguments and can
260 be retrieved using init:get_plain_arguments/0.
261 -heart:
263 Starts heartbeat monitoring of the Erlang runtime system; see
264 heart(3).
265 -hidden:
267 Starts the Erlang runtime system as a hidden node, if it is run as
268 a distributed node. Hidden nodes always establish hidden connec‐
269 tions to all other nodes except for nodes in the same global group.
270 Hidden connections are not published on any of the connected nodes,
271 that is, none of the connected nodes are part of the result from
272 nodes/0 on the other node. See also hidden global groups;
273 global_group(3).
274 -hosts Hosts:
276 Specifies the IP addresses for the hosts on which Erlang boot
277 servers are running, see erl_boot_server(3). This flag is mandatory
278 if flag -loader inet is present.
279 The IP addresses must be specified in the standard form (four deci‐
281 mal numbers separated by periods, for example, "150.236.20.74".
282 Hosts names are not acceptable, but a broadcast address (preferably
283 limited to the local network) is.
284 -id Id:
286 Specifies the identity of the Erlang runtime system. If it is run
287 as a distributed node, Id must be identical to the name supplied
288 together with flag -sname or -name.
289 -init_debug:
291 Makes init write some debug information while interpreting the boot
292 script.
293 -instr (emulator flag):
295 Selects an instrumented Erlang runtime system (virtual machine) to
296 run, instead of the ordinary one. When running an instrumented run‐
297 time system, some resource usage data can be obtained and analyzed
298 using the instrument module. Functionally, it behaves exactly like
299 an ordinary Erlang runtime system.
300 -loader Loader:
302 Specifies the method used by erl_prim_loader to load Erlang modules
303 into the system; see erl_prim_loader(3). Two Loader methods are
304 supported:
305 * efile, which means use the local file system, this is the de‐
307 fault.
308 * inet, which means use a boot server on another machine. The flags
310 -id, -hosts and -setcookie must also be specified.
311 If Loader is something else, the user-supplied Loader port program
313 is started.
314 -make:
316 Makes the Erlang runtime system invoke make:all() in the current
317 working directory and then terminate; see make(3). Implies -noin‐
318 put.
319 -man Module:
321 Displays the manual page for the Erlang module Module. Only sup‐
322 ported on Unix.
323 -mode interactive | embedded:
325 Modules are auto loaded when they are first referenced if the run‐
326 time system runs in interactive mode, which is the default. In em‐
327 bedded mode modules are not auto loaded. The latter is recommended
328 when the boot script preloads all modules, as conventionally hap‐
329 pens in OTP releases. See code(3).
330 -name Name:
332 Makes the Erlang runtime system into a distributed node. This flag
333 invokes all network servers necessary for a node to become distrib‐
334 uted; see net_kernel(3). It also ensures that epmd runs on the cur‐
335 rent host before Erlang is started (see epmd(1) and the -start_epmd
336 option) and that a magic cookie has been set (see -setcookie).
337 The node name will be Name@Host, where Host is the fully qualified
339 host name of the current host. For short names, use flag -sname in‐
340 stead.
341 If Name is set to undefined the node will be started in a special
343 mode optimized to be the temporary client of another node. The node
344 will then request a dynamic node name from the first node it con‐
345 nects to. Read more in Dynamic Node Name.
346 Warning:
348 Starting a distributed node without also specifying -proto_dist
349 inet_tls will expose the node to attacks that may give the attacker
350 complete access to the node and in extension the cluster. When using
351 un-secure distributed nodes, make sure that the network is configured
352 to keep potential attackers out.
353 -no_epmd:
356 Specifies that the distributed node does not need epmd at all.
357 This option ensures that the Erlang runtime system does not start
359 epmd and does not start the erl_epmd process for distribution ei‐
360 ther.
361 This option only works if Erlang is started as a distributed node
363 with the -proto_dist option using an alternative protocol for Er‐
364 lang distribution which does not rely on epmd for node registration
365 and discovery. For more information, see How to implement an Alter‐
366 native Carrier for the Erlang Distribution.
367 -noinput:
369 Ensures that the Erlang runtime system never tries to read any in‐
370 put. Implies -noshell.
371 -noshell:
373 Starts an Erlang runtime system with no shell. This flag makes it
374 possible to have the Erlang runtime system as a component in a se‐
375 ries of Unix pipes.
376 -nostick:
378 Disables the sticky directory facility of the Erlang code server;
379 see code(3).
380 -oldshell:
382 Invokes the old Erlang shell from Erlang/OTP 3.3. The old shell can
383 still be used.
384 -pa Dir1 Dir2 ...:
386 Adds the specified directories to the beginning of the code path,
387 similar to code:add_pathsa/1. Note that the order of the given di‐
388 rectories will be reversed in the resulting path.
389 As an alternative to -pa, if several directories are to be
391 prepended to the code path and the directories have a common parent
392 directory, that parent directory can be specified in environment
393 variable ERL_LIBS; see code(3).
394 -pz Dir1 Dir2 ...:
396 Adds the specified directories to the end of the code path, similar
397 to code:add_pathsz/1; see code(3).
398 -path Dir1 Dir2 ...:
400 Replaces the path specified in the boot script; see script(4).
401 -proto_dist Proto:
403 Specifies a protocol for Erlang distribution:
406 inet_tcp:
408 TCP over IPv4 (the default)
409 inet_tls:
411 Distribution over TLS/SSL, See the Using SSL for Erlang Distri‐
412 bution User's Guide for details on how to setup a secure distrib‐
413 uted node.
414 inet6_tcp:
416 TCP over IPv6
417 For example, to start up IPv6 distributed nodes:
419 % erl -name test@ipv6node.example.com -proto_dist inet6_tcp
421 -remsh Node:
423 Starts Erlang with a remote shell connected to Node.
424 If no -name or -sname is given the node will be started using
426 -sname undefined. If Node does not contain a hostname, one is auto‐
427 matically taken from -name or -sname
428 Note:
430 Before OTP-23 the user needed to supply a valid -sname or -name for
431 -remsh to work. This is still the case if the target node is not run‐
432 ning OTP-23 or later.
433 Note:
436 The connecting node needs to have a proper shell with terminal emula‐
437 tion. This means that UNIX users must use an Erlang compiled with
438 terminal capabilities and Windows users must use werl(1).
439 -rsh Program:
442 Specifies an alternative to ssh for starting a slave node on a re‐
443 mote host; see slave(3).
444 -run Mod [Func [Arg1, Arg2, ...]] (init flag):
446 Makes init call the specified function. Func defaults to start. If
447 no arguments are provided, the function is assumed to be of arity
448 0. Otherwise it is assumed to be of arity 1, taking the list
449 [Arg1,Arg2,...] as argument. All arguments are passed as strings.
450 See init(3).
451 -s Mod [Func [Arg1, Arg2, ...]] (init flag):
453 Makes init call the specified function. Func defaults to start. If
454 no arguments are provided, the function is assumed to be of arity
455 0. Otherwise it is assumed to be of arity 1, taking the list
456 [Arg1,Arg2,...] as argument. All arguments are passed as atoms. See
457 init(3).
458 -setcookie Cookie:
460 Sets the magic cookie of the node to Cookie; see er‐
461 lang:set_cookie/1. See see section Distributed Erlang in the Erlang
462 Reference Manual for more details.
463 -setcookie Node Cookie:
465 Sets the magic cookie for Node to Cookie; see erlang:set_cookie/2.
466 -shutdown_time Time:
468 Specifies how long time (in milliseconds) the init process is al‐
469 lowed to spend shutting down the system. If Time milliseconds have
470 elapsed, all processes still existing are killed. Defaults to in‐
471 finity.
472 -sname Name:
474 Makes the Erlang runtime system into a distributed node, similar to
475 -name, but the host name portion of the node name Name@Host will be
476 the short name, not fully qualified.
477 This is sometimes the only way to run distributed Erlang if the Do‐
479 main Name System (DNS) is not running. No communication can exist
480 between nodes running with flag -sname and those running with flag
481 -name, as node names must be unique in distributed Erlang systems.
482 If Name is set to undefined the node will be started in a special
484 mode optimized to be the temporary client of another node. The node
485 will then request a dynamic node name from the first node it con‐
486 nects to. Read more in Dynamic Node Name.
487 Warning:
489 Starting a distributed node without also specifying -proto_dist
490 inet_tls will expose the node to attacks that may give the attacker
491 complete access to the node and in extension the cluster. When using
492 un-secure distributed nodes, make sure that the network is configured
493 to keep potential attackers out.
494 -start_epmd true | false:
497 Specifies whether Erlang should start epmd on startup. By default
498 this is true, but if you prefer to start epmd manually, set this to
499 false.
500 This only applies if Erlang is started as a distributed node, i.e.
502 if -name or -sname is specified. Otherwise, epmd is not started
503 even if -start_epmd true is given.
504 Note that a distributed node will fail to start if epmd is not run‐
506 ning.
507 -version (emulator flag):
509 Makes the emulator print its version number. The same as erl +V.
510
512 erl invokes the code for the Erlang emulator (virtual machine), which
513 supports the following flags:
514 +a size:
516 Suggested stack size, in kilowords, for threads in the async thread
517 pool. Valid range is 16-8192 kilowords. The default suggested stack
518 size is 16 kilowords, that is, 64 kilobyte on 32-bit architectures.
519 This small default size has been chosen because the number of async
520 threads can be large. The default size is enough for drivers deliv‐
521 ered with Erlang/OTP, but might not be large enough for other dy‐
522 namically linked-in drivers that use the driver_async() functional‐
523 ity. Notice that the value passed is only a suggestion, and it can
524 even be ignored on some platforms.
525 +A size:
527 Sets the number of threads in async thread pool. Valid range is
528 1-1024. The async thread pool is used by linked-in drivers to han‐
529 dle work that may take a very long time. Since OTP 21 there are
530 very few linked-in drivers in the default Erlang/OTP distribution
531 that uses the async thread pool. Most of them have been migrated to
532 dirty IO schedulers. Defaults to 1.
533 +B [c | d | i]:
535 Option c makes Ctrl-C interrupt the current shell instead of invok‐
536 ing the emulator break handler. Option d (same as specifying +B
537 without an extra option) disables the break handler. Option i makes
538 the emulator ignore any break signal.
539 If option c is used with oldshell on Unix, Ctrl-C will restart the
541 shell process rather than interrupt it.
542 Notice that on Windows, this flag is only applicable for werl, not
544 erl (oldshell). Notice also that Ctrl-Break is used instead of
545 Ctrl-C on Windows.
546 +c true | false:
548 Enables or disables time correction:
549 true:
551 Enables time correction. This is the default if time correction
552 is supported on the specific platform.
553 false:
555 Disables time correction.
556 For backward compatibility, the boolean value can be omitted. This
558 is interpreted as +c false.
559 +C no_time_warp | single_time_warp | multi_time_warp:
561 Sets time warp mode:
562 no_time_warp:
564 No time warp mode (the default)
565 single_time_warp:
567 Single time warp mode
568 multi_time_warp:
570 Multi-time warp mode
571 +d:
573 If the emulator detects an internal error (or runs out of memory),
574 it, by default, generates both a crash dump and a core dump. The
575 core dump is, however, not very useful as the content of process
576 heaps is destroyed by the crash dump generation.
577 Option +d instructs the emulator to produce only a core dump and no
579 crash dump if an internal error is detected.
580 Calling erlang:halt/1 with a string argument still produces a crash
582 dump. On Unix systems, sending an emulator process a SIGUSR1 signal
583 also forces a crash dump.
584 +dcg DecentralizedCounterGroupsLimit:
586 Limits the number of decentralized counter groups used by decen‐
587 tralized counters optimized for update operations in the Erlang
588 runtime system. By default, the limit is 256.
589 When the number of schedulers is less than or equal to the limit,
591 each scheduler has its own group. When the number of schedulers is
592 larger than the groups limit, schedulers share groups. Shared
593 groups degrade the performance for updating counters while many
594 reader groups degrade the performance for reading counters. So, the
595 limit is a tradeoff between performance for update operations and
596 performance for read operations. Each group consumes 64 bytes in
597 each counter.
598 Notice that a runtime system using decentralized counter groups
600 benefits from binding schedulers to logical processors, as the
601 groups are distributed better between schedulers with this option.
602 This option only affects decentralized counters used for the coun‐
604 ters that are keeping track of the memory consumption and the num‐
605 ber of terms in ETS tables of type ordered_set with the write_con‐
606 currency option activated.
607 +e Number:
609 Sets the maximum number of ETS tables. This limit is partially ob‐
610 solete.
611 +ec:
613 Forces option compressed on all ETS tables. Only intended for test
614 and evaluation.
615 +fnl:
617 The virtual machine works with filenames as if they are encoded us‐
618 ing the ISO Latin-1 encoding, disallowing Unicode characters with
619 code points > 255.
620 For more information about Unicode filenames, see section Unicode
622 Filenames in the STDLIB User's Guide. Notice that this value also
623 applies to command-line parameters and environment variables (see
624 section Unicode in Environment and Parameters in the STDLIB User's
625 Guide).
626 +fnu[{w|i|e}]:
628 The virtual machine works with filenames as if they are encoded us‐
629 ing UTF-8 (or some other system-specific Unicode encoding). This is
630 the default on operating systems that enforce Unicode encoding,
631 that is, Windows MacOS X and Android.
632 The +fnu switch can be followed by w, i, or e to control how
634 wrongly encoded filenames are to be reported:
635 * w means that a warning is sent to the error_logger whenever a
637 wrongly encoded filename is "skipped" in directory listings. This
638 is the default.
639 * i means that those wrongly encoded filenames are silently ig‐
641 nored.
642 * e means that the API function returns an error whenever a wrongly
644 encoded filename (or directory name) is encountered.
645 Notice that file:read_link/1 always returns an error if the link
647 points to an invalid filename.
648 For more information about Unicode filenames, see section Unicode
650 Filenames in the STDLIB User's Guide. Notice that this value also
651 applies to command-line parameters and environment variables (see
652 section Unicode in Environment and Parameters in the STDLIB User's
653 Guide).
654 +fna[{w|i|e}]:
656 Selection between +fnl and +fnu is done based on the current locale
657 settings in the OS. This means that if you have set your terminal
658 for UTF-8 encoding, the filesystem is expected to use the same en‐
659 coding for filenames. This is the default on all operating systems,
660 except Android, MacOS X and Windows.
661 The +fna switch can be followed by w, i, or e. This has effect if
663 the locale settings cause the behavior of +fnu to be selected; see
664 the description of +fnu above. If the locale settings cause the be‐
665 havior of +fnl to be selected, then w, i, or e have no effect.
666 For more information about Unicode filenames, see section Unicode
668 Filenames in the STDLIB User's Guide. Notice that this value also
669 applies to command-line parameters and environment variables (see
670 section Unicode in Environment and Parameters in the STDLIB User's
671 Guide).
672 +hms Size:
674 Sets the default heap size of processes to the size Size.
675 +hmbs Size:
677 Sets the default binary virtual heap size of processes to the size
678 Size.
679 +hmax Size:
681 Sets the default maximum heap size of processes to the size Size
682 words. Defaults to 0, which means that no maximum heap size is
683 used. For more information, see process_flag(max_heap_size, Max‐
684 HeapSize).
685 +hmaxel true|false:
687 Sets whether to send an error logger message or not for processes
688 reaching the maximum heap size. Defaults to true. For more informa‐
689 tion, see process_flag(max_heap_size, MaxHeapSize).
690 +hmaxk true|false:
692 Sets whether to kill processes reaching the maximum heap size or
693 not. Default to true. For more information, see
694 process_flag(max_heap_size, MaxHeapSize).
695 +hpds Size:
697 Sets the initial process dictionary size of processes to the size
698 Size.
699 +hmqd off_heap|on_heap:
701 Sets the default value of the message_queue_data process flag. De‐
702 faults to on_heap. If +hmqd is not passed, on_heap will be the de‐
703 fault. For more information, see process_flag(message_queue_data,
704 MQD).
705 +IOp PollSets:
707 Sets the number of IO pollsets to use when polling for I/O. This
708 option is only used on platforms that support concurrent updates of
709 a pollset, otherwise the same number of pollsets are used as IO
710 poll threads. The default is 1.
711 +IOt PollThreads:
713 Sets the number of IO poll threads to use when polling for I/O. The
714 maximum number of poll threads allowed is 1024. The default is 1.
715 A good way to check if more IO poll threads are needed is to use
717 microstate accounting and see what the load of the IO poll thread
718 is. If it is high it could be a good idea to add more threads.
719 +IOPp PollSetsPercentage:
721 Similar to +IOp but uses percentages to set the number of IO
722 pollsets to create, based on the number of poll threads configured.
723 If both +IOPp and +IOp are used, +IOPp is ignored.
724 +IOPt PollThreadsPercentage:
726 Similar to +IOt but uses percentages to set the number of IO poll
727 threads to create, based on the number of schedulers configured. If
728 both +IOPt and +IOt are used, +IOPt is ignored.
729 +IOs true|false:
731 Enable or disable scheduler thread poll optimization. Default is
732 true.
733 If enabled, file descriptors that are frequently read may be moved
735 to a special pollset used by scheduler threads. The objective is to
736 reduce the number of system calls and thereby CPU load, but it can
737 in some cases increase scheduling latency for individual file de‐
738 scriptor input events.
739 +JPperf true|false|dump|map|fp|no_fp:
741 Enables or disables support for the perf profiler when running with
742 the JIT on Linux. Defaults to false.
743 This option can be combined multiple times to enable several op‐
745 tions:
746 dump:
748 Gives perf detailed line information, so that the perf annotate
749 feature works.
750 map:
752 Gives perf a map over all module code, letting it translate ma‐
753 chine code addresses to Erlang source code locations. This also
754 enables frame pointers for Erlang code so that perf can walk the
755 call stacks of Erlang processes, which costs one extra word per
756 stack frame.
757 fp:
759 Enables frame pointers independently of the map option.
760 no_fp:
762 Disables the frame pointers added by the map option.
763 true:
765 Enables map and dump.
766 false:
768 Disables all other options.
769 For more details about how to run perf see the perf support section
771 in the BeamAsm internal documentation.
772 +L:
774 Prevents loading information about source filenames and line num‐
775 bers. This saves some memory, but exceptions do not contain infor‐
776 mation about the filenames and line numbers.
777 +MFlag Value:
779 Memory allocator-specific flags. For more information, see erts_al‐
780 loc(3).
781 +pad true|false:
783 Since: OTP 25.3
784 The boolean value used with the +pad parameter determines the de‐
786 fault value of the async_dist process flag of newly spawned pro‐
787 cesses. By default, if no +pad command line option is passed, the
788 async_dist flag will be set to false.
789 The value used in runtime can be inspected by calling erlang:sys‐
791 tem_info(async_dist).
792 +pc Range:
794 Sets the range of characters that the system considers printable in
795 heuristic detection of strings. This typically affects the shell,
796 debugger, and io:format functions (when ~tp is used in the format
797 string).
798 Two values are supported for Range:
800 latin1:
802 The default. Only characters in the ISO Latin-1 range can be con‐
803 sidered printable. This means that a character with a code point
804 > 255 is never considered printable and that lists containing
805 such characters are displayed as lists of integers rather than
806 text strings by tools.
807 unicode:
809 All printable Unicode characters are considered when determining
810 if a list of integers is to be displayed in string syntax. This
811 can give unexpected results if, for example, your font does not
812 cover all Unicode characters.
813 See also io:printable_range/0 in STDLIB.
815 +P Number:
817 Sets the maximum number of simultaneously existing processes for
818 this system if a Number is passed as value. Valid range for Number
819 is [1024-134217727]
820 NOTE: The actual maximum chosen may be much larger than the Number
822 passed. Currently the runtime system often, but not always, chooses
823 a value that is a power of 2. This might, however, be changed in
824 the future. The actual value chosen can be checked by calling er‐
825 lang:system_info(process_limit).
826 The default value is 262144
828 +Q Number:
830 Sets the maximum number of simultaneously existing ports for this
831 system if a Number is passed as value. Valid range for Number is
832 [1024-134217727]
833 NOTE: The actual maximum chosen may be much larger than the actual
835 Number passed. Currently the runtime system often, but not always,
836 chooses a value that is a power of 2. This might, however, be
837 changed in the future. The actual value chosen can be checked by
838 calling erlang:system_info(port_limit).
839 The default value used is normally 65536. However, if the runtime
841 system is able to determine maximum amount of file descriptors that
842 it is allowed to open and this value is larger than 65536, the cho‐
843 sen value will increased to a value larger or equal to the maximum
844 amount of file descriptors that can be opened.
845 On Windows the default value is set to 8196 because the normal OS
847 limitations are set higher than most machines can handle.
848 +R ReleaseNumber:
850 Sets the compatibility mode.
851 The distribution mechanism is not backward compatible by default.
853 This flag sets the emulator in compatibility mode with an earlier
854 Erlang/OTP release ReleaseNumber. The release number must be in the
855 range <current release>-2..<current release>. This limits the emu‐
856 lator, making it possible for it to communicate with Erlang nodes
857 (as well as C- and Java nodes) running that earlier release.
858 Note:
860 Ensure that all nodes (Erlang-, C-, and Java nodes) of a distributed
861 Erlang system is of the same Erlang/OTP release, or from two differ‐
862 ent Erlang/OTP releases X and Y, where all Y nodes have compatibility
863 mode X.
864 +r:
867 Forces ETS memory block to be moved on realloc.
868 +rg ReaderGroupsLimit:
870 Limits the number of reader groups used by read/write locks opti‐
871 mized for read operations in the Erlang runtime system. By default
872 the reader groups limit is 64.
873 When the number of schedulers is less than or equal to the reader
875 groups limit, each scheduler has its own reader group. When the
876 number of schedulers is larger than the reader groups limit, sched‐
877 ulers share reader groups. Shared reader groups degrade read lock
878 and read unlock performance while many reader groups degrade write
879 lock performance. So, the limit is a tradeoff between performance
880 for read operations and performance for write operations. Each
881 reader group consumes 64 byte in each read/write lock.
882 Notice that a runtime system using shared reader groups benefits
884 from binding schedulers to logical processors, as the reader groups
885 are distributed better between schedulers.
886 +S Schedulers:SchedulerOnline:
888 Sets the number of scheduler threads to create and scheduler
889 threads to set online. The maximum for both values is 1024. If the
890 Erlang runtime system is able to determine the number of logical
891 processors configured and logical processors available, Schedulers
892 defaults to logical processors configured, and SchedulersOnline de‐
893 faults to logical processors available; otherwise the default val‐
894 ues are 1. If the emulator detects that it is subject to a CPU
895 quota, the default value for SchedulersOnline will be limited ac‐
896 cordingly.
897 Schedulers can be omitted if :SchedulerOnline is not and con‐
899 versely. The number of schedulers online can be changed at runtime
900 through erlang:system_flag(schedulers_online, SchedulersOnline).
901 If Schedulers or SchedulersOnline is specified as a negative num‐
903 ber, the value is subtracted from the default number of logical
904 processors configured or logical processors available, respec‐
905 tively.
906 Specifying value 0 for Schedulers or SchedulersOnline resets the
908 number of scheduler threads or scheduler threads online, respec‐
909 tively, to its default value.
910 +SP SchedulersPercentage:SchedulersOnlinePercentage:
912 Similar to +S but uses percentages to set the number of scheduler
913 threads to create, based on logical processors configured, and
914 scheduler threads to set online, based on logical processors avail‐
915 able. Specified values must be > 0. For example, +SP 50:25 sets the
916 number of scheduler threads to 50% of the logical processors con‐
917 figured, and the number of scheduler threads online to 25% of the
918 logical processors available. SchedulersPercentage can be omitted
919 if :SchedulersOnlinePercentage is not and conversely. The number of
920 schedulers online can be changed at runtime through erlang:sys‐
921 tem_flag(schedulers_online, SchedulersOnline).
922 This option interacts with +S settings. For example, on a system
924 with 8 logical cores configured and 8 logical cores available, the
925 combination of the options +S 4:4 +SP 50:25 (in either order) re‐
926 sults in 2 scheduler threads (50% of 4) and 1 scheduler thread on‐
927 line (25% of 4).
928 +SDcpu DirtyCPUSchedulers:DirtyCPUSchedulersOnline:
930 Sets the number of dirty CPU scheduler threads to create and dirty
931 CPU scheduler threads to set online. The maximum for both values is
932 1024, and each value is further limited by the settings for normal
933 schedulers:
934 * The number of dirty CPU scheduler threads created cannot exceed
936 the number of normal scheduler threads created.
937 * The number of dirty CPU scheduler threads online cannot exceed
939 the number of normal scheduler threads online.
940 For details, see the +S and +SP. By default, the number of dirty
942 CPU scheduler threads created equals the number of normal scheduler
943 threads created, and the number of dirty CPU scheduler threads on‐
944 line equals the number of normal scheduler threads online. DirtyC‐
945 PUSchedulers can be omitted if :DirtyCPUSchedulersOnline is not and
946 conversely. The number of dirty CPU schedulers online can be
947 changed at runtime through erlang:system_flag(dirty_cpu_sched‐
948 ulers_online, DirtyCPUSchedulersOnline).
949 The amount of dirty CPU schedulers is limited by the amount of nor‐
951 mal schedulers in order to limit the effect on processes executing
952 on ordinary schedulers. If the amount of dirty CPU schedulers was
953 allowed to be unlimited, dirty CPU bound jobs would potentially
954 starve normal jobs.
955 Typical users of the dirty CPU schedulers are large garbage collec‐
957 tions, json protocol encode/decoders written as nifs and matrix ma‐
958 nipulation libraries.
959 You can use msacc(3) in order to see the current load of the dirty
961 CPU schedulers threads and adjust the number used accordingly.
962 +SDPcpu DirtyCPUSchedulersPercentage:DirtyCPUSchedulersOnlinePercent‐
964 age:
965 Similar to +SDcpu but uses percentages to set the number of dirty
966 CPU scheduler threads to create and the number of dirty CPU sched‐
967 uler threads to set online. Specified values must be > 0. For exam‐
968 ple, +SDPcpu 50:25 sets the number of dirty CPU scheduler threads
969 to 50% of the logical processors configured and the number of dirty
970 CPU scheduler threads online to 25% of the logical processors
971 available. DirtyCPUSchedulersPercentage can be omitted if :DirtyC‐
972 PUSchedulersOnlinePercentage is not and conversely. The number of
973 dirty CPU schedulers online can be changed at runtime through er‐
974 lang:system_flag(dirty_cpu_schedulers_online, DirtyCPUSchedulersOn‐
975 line).
976 This option interacts with +SDcpu settings. For example, on a sys‐
978 tem with 8 logical cores configured and 8 logical cores available,
979 the combination of the options +SDcpu 4:4 +SDPcpu 50:25 (in either
980 order) results in 2 dirty CPU scheduler threads (50% of 4) and 1
981 dirty CPU scheduler thread online (25% of 4).
982 +SDio DirtyIOSchedulers:
984 Sets the number of dirty I/O scheduler threads to create. Valid
985 range is 1-1024. By default, the number of dirty I/O scheduler
986 threads created is 10.
987 The amount of dirty IO schedulers is not limited by the amount of
989 normal schedulers like the amount of dirty CPU schedulers. This
990 since only I/O bound work is expected to execute on dirty I/O
991 schedulers. If the user should schedule CPU bound jobs on dirty I/O
992 schedulers, these jobs might starve ordinary jobs executing on or‐
993 dinary schedulers.
994 Typical users of the dirty IO schedulers are reading and writing to
996 files.
997 You can use msacc(3) in order to see the current load of the dirty
999 IO schedulers threads and adjust the number used accordingly.
1000 +sFlag Value:
1002 Scheduling specific flags.
1003 +sbt BindType:
1005 Sets scheduler bind type.
1006 Schedulers can also be bound using flag +stbt. The only differ‐
1008 ence between these two flags is how the following errors are han‐
1009 dled:
1010 * Binding of schedulers is not supported on the specific plat‐
1012 form.
1013 * No available CPU topology. That is, the runtime system was not
1015 able to detect the CPU topology automatically, and no user-de‐
1016 fined CPU topology was set.
1017 If any of these errors occur when +sbt has been passed, the run‐
1019 time system prints an error message, and refuses to start. If any
1020 of these errors occur when +stbt has been passed, the runtime
1021 system silently ignores the error, and start up using unbound
1022 schedulers.
1023 Valid BindTypes:
1025 u:
1027 unbound - Schedulers are not bound to logical processors, that
1028 is, the operating system decides where the scheduler threads
1029 execute, and when to migrate them. This is the default.
1030 ns:
1032 no_spread - Schedulers with close scheduler identifiers are
1033 bound as close as possible in hardware.
1034 ts:
1036 thread_spread - Thread refers to hardware threads (such as In‐
1037 tel's hyper-threads). Schedulers with low scheduler identi‐
1038 fiers, are bound to the first hardware thread of each core,
1039 then schedulers with higher scheduler identifiers are bound to
1040 the second hardware thread of each core,and so on.
1041 ps:
1043 processor_spread - Schedulers are spread like thread_spread,
1044 but also over physical processor chips.
1045 s:
1047 spread - Schedulers are spread as much as possible.
1048 nnts:
1050 no_node_thread_spread - Like thread_spread, but if multiple
1051 Non-Uniform Memory Access (NUMA) nodes exist, schedulers are
1052 spread over one NUMA node at a time, that is, all logical pro‐
1053 cessors of one NUMA node are bound to schedulers in sequence.
1054 nnps:
1056 no_node_processor_spread - Like processor_spread, but if multi‐
1057 ple NUMA nodes exist, schedulers are spread over one NUMA node
1058 at a time, that is, all logical processors of one NUMA node are
1059 bound to schedulers in sequence.
1060 tnnps:
1062 thread_no_node_processor_spread - A combination of
1063 thread_spread, and no_node_processor_spread. Schedulers are
1064 spread over hardware threads across NUMA nodes, but schedulers
1065 are only spread over processors internally in one NUMA node at
1066 a time.
1067 db:
1069 default_bind - Binds schedulers the default way. Defaults to
1070 thread_no_node_processor_spread (which can change in the fu‐
1071 ture).
1072 Binding of schedulers is only supported on newer Linux, Solaris,
1074 FreeBSD, and Windows systems.
1075 If no CPU topology is available when flag +sbt is processed and
1077 BindType is any other type than u, the runtime system fails to
1078 start. CPU topology can be defined using flag +sct. Notice that
1079 flag +sct can have to be passed before flag +sbt on the command
1080 line (if no CPU topology has been automatically detected).
1081 The runtime system does by default not bind schedulers to logical
1083 processors.
1084 Note:
1086 If the Erlang runtime system is the only operating system process
1087 that binds threads to logical processors, this improves the perfor‐
1088 mance of the runtime system. However, if other operating system
1089 processes (for example another Erlang runtime system) also bind
1090 threads to logical processors, there can be a performance penalty
1091 instead. This performance penalty can sometimes be severe. If so,
1092 you are advised not to bind the schedulers.
1093 How schedulers are bound matters. For example, in situations when
1096 there are fewer running processes than schedulers online, the
1097 runtime system tries to migrate processes to schedulers with low
1098 scheduler identifiers. The more the schedulers are spread over
1099 the hardware, the more resources are available to the runtime
1100 system in such situations.
1101 Note:
1103 If a scheduler fails to bind, this is often silently ignored, as it
1104 is not always possible to verify valid logical processor identi‐
1105 fiers. If an error is reported, it is reported to the error_logger.
1106 If you want to verify that the schedulers have bound as requested,
1107 call erlang:system_info(scheduler_bindings).
1108 +sbwt none|very_short|short|medium|long|very_long:
1111 Sets scheduler busy wait threshold. Defaults to medium. The
1112 threshold determines how long schedulers are to busy wait when
1113 running out of work before going to sleep.
1114 Note:
1116 This flag can be removed or changed at any time without prior no‐
1117 tice.
1118 +sbwtdcpu none|very_short|short|medium|long|very_long:
1121 As +sbwt but affects dirty CPU schedulers. Defaults to short.
1122 Note:
1124 This flag can be removed or changed at any time without prior no‐
1125 tice.
1126 +sbwtdio none|very_short|short|medium|long|very_long:
1129 As +sbwt but affects dirty IO schedulers. Defaults to short.
1130 Note:
1132 This flag can be removed or changed at any time without prior no‐
1133 tice.
1134 +scl true|false:
1137 Enables or disables scheduler compaction of load. By default
1138 scheduler compaction of load is enabled. When enabled, load bal‐
1139 ancing strives for a load distribution, which causes as many
1140 scheduler threads as possible to be fully loaded (that is, not
1141 run out of work). This is accomplished by migrating load (for ex‐
1142 ample, runnable processes) into a smaller set of schedulers when
1143 schedulers frequently run out of work. When disabled, the fre‐
1144 quency with which schedulers run out of work is not taken into
1145 account by the load balancing logic.
1146 +scl false is similar to +sub true, but +sub true also balances
1148 scheduler utilization between schedulers.
1149 +sct CpuTopology:
1151 * <Id> = integer(); when 0 =< <Id> =< 65535
1154 * <IdRange> = <Id>-<Id>
1156 * <IdOrIdRange> = <Id> | <IdRange>
1158 * <IdList> = <IdOrIdRange>,<IdOrIdRange> | <IdOrIdRange>
1160 * <LogicalIds> = L<IdList>
1162 * <ThreadIds> = T<IdList> | t<IdList>
1164 * <CoreIds> = C<IdList> | c<IdList>
1166 * <ProcessorIds> = P<IdList> | p<IdList>
1168 * <NodeIds> = N<IdList> | n<IdList>
1170 * <IdDefs> = <LogicalIds><ThreadIds><CoreIds><Proces‐
1172 sorIds><NodeIds> | <LogicalIds><ThreadIds><Cor‐
1173 eIds><NodeIds><ProcessorIds>
1174 * CpuTopology = <IdDefs>:<IdDefs> | <IdDefs>
1176 Sets a user-defined CPU topology. The user-defined CPU topology
1178 overrides any automatically detected CPU topology. The CPU topol‐
1179 ogy is used when binding schedulers to logical processors.
1180 Uppercase letters signify real identifiers and lowercase letters
1182 signify fake identifiers only used for description of the topol‐
1183 ogy. Identifiers passed as real identifiers can be used by the
1184 runtime system when trying to access specific hardware; if they
1185 are incorrect the behavior is undefined. Faked logical CPU iden‐
1186 tifiers are not accepted, as there is no point in defining the
1187 CPU topology without real logical CPU identifiers. Thread, core,
1188 processor, and node identifiers can be omitted. If omitted, the
1189 thread ID defaults to t0, the core ID defaults to c0, the proces‐
1190 sor ID defaults to p0, and the node ID is left undefined. Either
1191 each logical processor must belong to only one NUMA node, or no
1192 logical processors must belong to any NUMA nodes.
1193 Both increasing and decreasing <IdRange>s are allowed.
1195 NUMA node identifiers are system wide. That is, each NUMA node on
1197 the system must have a unique identifier. Processor identifiers
1198 are also system wide. Core identifiers are processor wide. Thread
1199 identifiers are core wide.
1200 The order of the identifier types implies the hierarchy of the
1202 CPU topology. The valid orders are as follows:
1203 * <LogicalIds><ThreadIds><CoreIds><ProcessorIds><NodeIds>, that
1205 is, thread is part of a core that is part of a processor, which
1206 is part of a NUMA node.
1207 * <LogicalIds><ThreadIds><CoreIds><NodeIds><ProcessorIds>, that
1209 is, thread is part of a core that is part of a NUMA node, which
1210 is part of a processor.
1211 A CPU topology can consist of both processor external, and pro‐
1213 cessor internal NUMA nodes as long as each logical processor be‐
1214 longs to only one NUMA node. If <ProcessorIds> is omitted, its
1215 default position is before <NodeIds>. That is, the default is
1216 processor external NUMA nodes.
1217 If a list of identifiers is used in an <IdDefs>:
1219 * <LogicalIds> must be a list of identifiers.
1221 * At least one other identifier type besides <LogicalIds> must
1223 also have a list of identifiers.
1224 * All lists of identifiers must produce the same number of iden‐
1226 tifiers.
1227 A simple example. A single quad core processor can be described
1229 as follows:
1230 % erl +sct L0-3c0-3
1232 1> erlang:system_info(cpu_topology).
1233 [{processor,[{core,{logical,0}},
1234 {core,{logical,1}},
1235 {core,{logical,2}},
1236 {core,{logical,3}}]}]
1237 A more complicated example with two quad core processors, each
1239 processor in its own NUMA node. The ordering of logical proces‐
1240 sors is a bit weird. This to give a better example of identifier
1241 lists:
1242 % erl +sct L0-1,3-2c0-3p0N0:L7,4,6-5c0-3p1N1
1244 1> erlang:system_info(cpu_topology).
1245 [{node,[{processor,[{core,{logical,0}},
1246 {core,{logical,1}},
1247 {core,{logical,3}},
1248 {core,{logical,2}}]}]},
1249 {node,[{processor,[{core,{logical,7}},
1250 {core,{logical,4}},
1251 {core,{logical,6}},
1252 {core,{logical,5}}]}]}]
1253 As long as real identifiers are correct, it is OK to pass a CPU
1255 topology that is not a correct description of the CPU topology.
1256 When used with care this can be very useful. This to trick the
1257 emulator to bind its schedulers as you want. For example, if you
1258 want to run multiple Erlang runtime systems on the same machine,
1259 you want to reduce the number of schedulers used and manipulate
1260 the CPU topology so that they bind to different logical CPUs. An
1261 example, with two Erlang runtime systems on a quad core machine:
1262 % erl +sct L0-3c0-3 +sbt db +S3:2 -detached -noinput -noshell -sname one
1264 % erl +sct L3-0c0-3 +sbt db +S3:2 -detached -noinput -noshell -sname two
1265 In this example, each runtime system have two schedulers each on‐
1267 line, and all schedulers online will run on different cores. If
1268 we change to one scheduler online on one runtime system, and
1269 three schedulers online on the other, all schedulers online will
1270 still run on different cores.
1271 Notice that a faked CPU topology that does not reflect how the
1273 real CPU topology looks like is likely to decrease the perfor‐
1274 mance of the runtime system.
1275 For more information, see erlang:system_info(cpu_topology).
1277 +ssrct:
1279 Skips reading CPU topology.
1280 Note:
1282 Reading CPU topology slows down startup when starting many parallel
1283 instances of ERTS on systems with large amount of cores, using this
1284 flag might speed up execution in such scenarios.
1285 +sfwi Interval:
1288 Sets scheduler-forced wakeup interval. All run queues are scanned
1289 each Interval milliseconds. While there are sleeping schedulers
1290 in the system, one scheduler is woken for each non-empty run
1291 queue found. Interval default to 0, meaning this feature is dis‐
1292 abled.
1293 Note:
1295 This feature has been introduced as a temporary workaround for
1296 long-executing native code, and native code that does not bump re‐
1297 ductions properly in OTP. When these bugs have been fixed, this
1298 flag will be removed.
1299 +spp Bool:
1302 Sets default scheduler hint for port parallelism. If set to true,
1303 the virtual machine schedules port tasks when it improves paral‐
1304 lelism in the system. If set to false, the virtual machine tries
1305 to perform port tasks immediately, improving latency at the ex‐
1306 pense of parallelism. Default to false. The default used can be
1307 inspected in runtime by calling erlang:system_info(port_parallel‐
1308 ism). The default can be overridden on port creation by passing
1309 option parallelism to erlang:open_port/2.
1310 +sss size:
1312 Suggested stack size, in kilowords, for scheduler threads. Valid
1313 range is 20-8192 kilowords. The default suggested stack size is
1314 128 kilowords.
1315 +sssdcpu size:
1317 Suggested stack size, in kilowords, for dirty CPU scheduler
1318 threads. Valid range is 20-8192 kilowords. The default suggested
1319 stack size is 40 kilowords.
1320 +sssdio size:
1322 Suggested stack size, in kilowords, for dirty IO scheduler
1323 threads. Valid range is 20-8192 kilowords. The default suggested
1324 stack size is 40 kilowords.
1325 +stbt BindType:
1327 Tries to set the scheduler bind type. The same as flag +sbt ex‐
1328 cept how some errors are handled. For more information, see +sbt.
1329 +sub true|false:
1331 Enables or disables scheduler utilization balancing of load. By
1332 default scheduler utilization balancing is disabled and instead
1333 scheduler compaction of load is enabled, which strives for a load
1334 distribution that causes as many scheduler threads as possible to
1335 be fully loaded (that is, not run out of work). When scheduler
1336 utilization balancing is enabled, the system instead tries to
1337 balance scheduler utilization between schedulers. That is, strive
1338 for equal scheduler utilization on all schedulers.
1339 +sub true is only supported on systems where the runtime system
1341 detects and uses a monotonically increasing high-resolution
1342 clock. On other systems, the runtime system fails to start.
1343 +sub true implies +scl false. The difference between +sub true
1345 and +scl false is that +scl false does not try to balance the
1346 scheduler utilization.
1347 +swct very_eager|eager|medium|lazy|very_lazy:
1349 Sets scheduler wake cleanup threshold. Defaults to medium. Con‐
1350 trols how eager schedulers are to be requesting wakeup because of
1351 certain cleanup operations. When a lazy setting is used, more
1352 outstanding cleanup operations can be left undone while a sched‐
1353 uler is idling. When an eager setting is used, schedulers are
1354 more frequently woken, potentially increasing CPU-utilization.
1355 Note:
1357 This flag can be removed or changed at any time without prior no‐
1358 tice.
1359 +sws default|legacy:
1362 Sets scheduler wakeup strategy. Default strategy changed in ERTS
1363 5.10 (Erlang/OTP R16A). This strategy was known as proposal in
1364 Erlang/OTP R15. The legacy strategy was used as default from R13
1365 up to and including R15.
1366 Note:
1368 This flag can be removed or changed at any time without prior no‐
1369 tice.
1370 +swt very_low|low|medium|high|very_high:
1373 Sets scheduler wakeup threshold. Defaults to medium. The thresh‐
1374 old determines when to wake up sleeping schedulers when more work
1375 than can be handled by currently awake schedulers exists. A low
1376 threshold causes earlier wakeups, and a high threshold causes
1377 later wakeups. Early wakeups distribute work over multiple sched‐
1378 ulers faster, but work does more easily bounce between sched‐
1379 ulers.
1380 Note:
1382 This flag can be removed or changed at any time without prior no‐
1383 tice.
1384 +swtdcpu very_low|low|medium|high|very_high:
1387 As +swt but affects dirty CPU schedulers. Defaults to medium.
1388 Note:
1390 This flag can be removed or changed at any time without prior no‐
1391 tice.
1392 +swtdio very_low|low|medium|high|very_high:
1395 As +swt but affects dirty IO schedulers. Defaults to medium.
1396 Note:
1398 This flag can be removed or changed at any time without prior no‐
1399 tice.
1400 +t size:
1403 Sets the maximum number of atoms the virtual machine can handle.
1404 Defaults to 1,048,576.
1405 +T Level:
1407 Enables modified timing and sets the modified timing level. Valid
1408 range is 0-9. The timing of the runtime system is changed. A high
1409 level usually means a greater change than a low level. Changing the
1410 timing can be very useful for finding timing-related bugs.
1411 Modified timing affects the following:
1413 Process spawning:
1415 A process calling spawn, spawn_link, spawn_monitor, or spawn_opt
1416 is scheduled out immediately after completing the call. When
1417 higher modified timing levels are used, the caller also sleeps
1418 for a while after it is scheduled out.
1419 Context reductions:
1421 The number of reductions a process is allowed to use before it is
1422 scheduled out is increased or reduced.
1423 Input reductions:
1425 The number of reductions performed before checking I/O is in‐
1426 creased or reduced.
1427 Note:
1429 Performance suffers when modified timing is enabled. This flag is
1430 only intended for testing and debugging.
1431 return_to and return_from trace messages are lost when tracing on the
1433 spawn BIFs.
1434 This flag can be removed or changed at any time without prior notice.
1436 +v:
1439 Verbose.
1440 +V:
1442 Makes the emulator print its version number.
1443 +W w | i | e:
1445 Sets the mapping of warning messages for error_logger. Messages
1446 sent to the error logger using one of the warning routines can be
1447 mapped to errors (+W e), warnings (+W w), or information reports
1448 (+W i). Defaults to warnings. The current mapping can be retrieved
1449 using error_logger:warning_map/0. For more information, see er‐
1450 ror_logger:warning_map/0 in Kernel.
1451 +zFlag Value:
1453 Miscellaneous flags:
1454 +zdbbl size:
1456 Sets the distribution buffer busy limit (dist_buf_busy_limit) in
1457 kilobytes. Valid range is 1-2097151. Defaults to 1024.
1458 A larger buffer limit allows processes to buffer more outgoing
1460 messages over the distribution. When the buffer limit has been
1461 reached, sending processes will be suspended until the buffer
1462 size has shrunk. The buffer limit is per distribution channel. A
1463 higher limit gives lower latency and higher throughput at the ex‐
1464 pense of higher memory use.
1465 This limit only affects processes that have disabled fully asyn‐
1467 chronous distributed signaling .
1468 +zdntgc time:
1470 Sets the delayed node table garbage collection time (de‐
1471 layed_node_table_gc) in seconds. Valid values are either infinity
1472 or an integer in the range 0-100000000. Defaults to 60.
1473 Node table entries that are not referred linger in the table for
1475 at least the amount of time that this parameter determines. The
1476 lingering prevents repeated deletions and insertions in the ta‐
1477 bles from occurring.
1478 +zosrl limit:
1480 Sets a limit on the amount of outstanding requests made by a sys‐
1481 tem process orchestrating system wide changes. Valid range of
1482 this limit is [1, 134217727]. See erlang:system_flag(outstand‐
1483 ing_system_requests_limit, Limit) for more information.
1484
1486 ERL_CRASH_DUMP:
1487 If the emulator needs to write a crash dump, the value of this
1488 variable is the filename of the crash dump file. If the variable is
1489 not set, the name of the crash dump file is erl_crash.dump in the
1490 current directory.
1491 ERL_CRASH_DUMP_NICE:
1493 Unix systems: If the emulator needs to write a crash dump, it uses
1494 the value of this variable to set the nice value for the process,
1495 thus lowering its priority. Valid range is 1-39 (higher values are
1496 replaced with 39). The highest value, 39, gives the process the
1497 lowest priority.
1498 ERL_CRASH_DUMP_SECONDS:
1500 Unix systems: This variable gives the number of seconds that the
1501 emulator is allowed to spend writing a crash dump. When the given
1502 number of seconds have elapsed, the emulator is terminated.
1503 ERL_CRASH_DUMP_SECONDS=0:
1505 If the variable is set to 0 seconds, the runtime system does not
1506 even attempt to write the crash dump file. It only terminates.
1507 This is the default if option -heart is passed to erl and
1508 ERL_CRASH_DUMP_SECONDS is not set.
1509 ERL_CRASH_DUMP_SECONDS=S:
1511 If the variable is set to a positive value S, wait for S seconds
1512 to complete the crash dump file and then terminates the runtime
1513 system with a SIGALRM signal.
1514 ERL_CRASH_DUMP_SECONDS=-1:
1516 A negative value causes the termination of the runtime system to
1517 wait indefinitely until the crash dump file has been completely
1518 written. This is the default if option -heart is not passed to
1519 erl and ERL_CRASH_DUMP_SECONDS is not set.
1520 See also heart(3).
1522 ERL_CRASH_DUMP_BYTES:
1524 This variable sets the maximum size of a crash dump file in bytes.
1525 The crash dump will be truncated if this limit is exceeded. If the
1526 variable is not set, no size limit is enforced by default. If the
1527 variable is set to 0, the runtime system does not even attempt to
1528 write a crash dump file.
1529 Introduced in ERTS 8.1.2 (Erlang/OTP 19.2).
1531 ERL_AFLAGS:
1533 The content of this variable is added to the beginning of the com‐
1534 mand line for erl.
1535 Flag -extra is treated in a special way. Its scope ends at the end
1537 of the environment variable content. Arguments following an -extra
1538 flag are moved on the command line into section -extra, that is,
1539 the end of the command line following an -extra flag.
1540 ERL_ZFLAGS and ERL_FLAGS:
1542 The content of these variables are added to the end of the command
1543 line for erl.
1544 Flag -extra is treated in a special way. Its scope ends at the end
1546 of the environment variable content. Arguments following an -extra
1547 flag are moved on the command line into section -extra, that is,
1548 the end of the command line following an -extra flag.
1549 ERL_LIBS:
1551 Contains a list of additional library directories that the code
1552 server searches for applications and adds to the code path; see
1553 code(3).
1554 ERL_EPMD_ADDRESS:
1556 Can be set to a comma-separated list of IP addresses, in which case
1557 the epmd daemon listens only on the specified address(es) and on
1558 the loopback address (which is implicitly added to the list if it
1559 has not been specified).
1560 ERL_EPMD_PORT:
1562 Can contain the port number to use when communicating with epmd.
1563 The default port works fine in most cases. A different port can be
1564 specified to allow nodes of independent clusters to co-exist on the
1565 same host. All nodes in a cluster must use the same epmd port num‐
1566 ber.
1567
1569 On Unix systems, the Erlang runtime will interpret two types of sig‐
1570 nals.
1571 SIGUSR1:
1573 A SIGUSR1 signal forces a crash dump.
1574 SIGTERM:
1576 A SIGTERM will produce a stop message to the init process. This is
1577 equivalent to a init:stop/0 call.
1578 Introduced in ERTS 8.3 (Erlang/OTP 19.3)
1580 The signal SIGUSR2 is reserved for internal usage. No other signals are
1582 handled.
1583
1585 The standard Erlang/OTP system can be reconfigured to change the de‐
1586 fault behavior on startup.
1587 The .erlang startup file:
1589 When Erlang/OTP is started, the system searches for a file named
1590 .erlang in the user's home directory and then file‐
1591 name:basedir(user_config, "erlang").
1592 If an .erlang file is found, it is assumed to contain valid Erlang
1594 expressions. These expressions are evaluated as if they were input
1595 to the shell.
1596 A typical .erlang file contains a set of search paths, for example:
1598 io:format("executing user profile in $HOME/.erlang\n",[]).
1600 code:add_path("/home/calvin/test/ebin").
1601 code:add_path("/home/hobbes/bigappl-1.2/ebin").
1602 io:format(".erlang rc finished\n",[]).
1603 user_default and shell_default:
1605 Functions in the shell that are not prefixed by a module name are
1606 assumed to be functional objects (funs), built-in functions (BIFs),
1607 or belong to the module user_default or shell_default.
1608 To include private shell commands, define them in a module user_de‐
1610 fault and add the following argument as the first line in the .er‐
1611 lang file:
1612 code:load_abs("..../user_default").
1614 erl:
1616 If the contents of .erlang are changed and a private version of
1617 user_default is defined, the Erlang/OTP environment can be custom‐
1618 ized. More powerful changes can be made by supplying command-line
1619 arguments in the startup script erl. For more information, see
1620 init(3).
1621
1623 epmd(1), erl_prim_loader(3), erts_alloc(3), init(3), application(3),
1624 auth(3), code(3), erl_boot_server(3), heart(3), net_kernel(3), make(3)
1625Ericsson AB erts 13.2.2.4 erl(1)