1 |
NAME |
2 |
Coro - the only real threads in perl |
3 |
|
4 |
SYNOPSIS |
5 |
use Coro; |
6 |
|
7 |
async { |
8 |
# some asynchronous thread of execution |
9 |
print "2\n"; |
10 |
cede; # yield back to main |
11 |
print "4\n"; |
12 |
}; |
13 |
print "1\n"; |
14 |
cede; # yield to coro |
15 |
print "3\n"; |
16 |
cede; # and again |
17 |
|
18 |
# use locking |
19 |
use Coro::Semaphore; |
20 |
my $lock = new Coro::Semaphore; |
21 |
my $locked; |
22 |
|
23 |
$lock->down; |
24 |
$locked = 1; |
25 |
$lock->up; |
26 |
|
27 |
DESCRIPTION |
28 |
For a tutorial-style introduction, please read the Coro::Intro manpage. |
29 |
This manpage mainly contains reference information. |
30 |
|
31 |
This module collection manages continuations in general, most often in |
32 |
the form of cooperative threads (also called coros, or simply "coro" in |
33 |
the documentation). They are similar to kernel threads but don't (in |
34 |
general) run in parallel at the same time even on SMP machines. The |
35 |
specific flavor of thread offered by this module also guarantees you |
36 |
that it will not switch between threads unless necessary, at |
37 |
easily-identified points in your program, so locking and parallel access |
38 |
are rarely an issue, making thread programming much safer and easier |
39 |
than using other thread models. |
40 |
|
41 |
Unlike the so-called "Perl threads" (which are not actually real threads |
42 |
but only the windows process emulation (see section of same name for |
43 |
more details) ported to unix, and as such act as processes), Coro |
44 |
provides a full shared address space, which makes communication between |
45 |
threads very easy. And Coro's threads are fast, too: disabling the |
46 |
Windows process emulation code in your perl and using Coro can easily |
47 |
result in a two to four times speed increase for your programs. A |
48 |
parallel matrix multiplication benchmark runs over 300 times faster on a |
49 |
single core than perl's pseudo-threads on a quad core using all four |
50 |
cores. |
51 |
|
52 |
Coro achieves that by supporting multiple running interpreters that |
53 |
share data, which is especially useful to code pseudo-parallel processes |
54 |
and for event-based programming, such as multiple HTTP-GET requests |
55 |
running concurrently. See Coro::AnyEvent to learn more on how to |
56 |
integrate Coro into an event-based environment. |
57 |
|
58 |
In this module, a thread is defined as "callchain + lexical variables + |
59 |
some package variables + C stack), that is, a thread has its own |
60 |
callchain, its own set of lexicals and its own set of perls most |
61 |
important global variables (see Coro::State for more configuration and |
62 |
background info). |
63 |
|
64 |
See also the "SEE ALSO" section at the end of this document - the Coro |
65 |
module family is quite large. |
66 |
|
67 |
GLOBAL VARIABLES |
68 |
$Coro::main |
69 |
This variable stores the Coro object that represents the main |
70 |
program. While you cna "ready" it and do most other things you can |
71 |
do to coro, it is mainly useful to compare again $Coro::current, to |
72 |
see whether you are running in the main program or not. |
73 |
|
74 |
$Coro::current |
75 |
The Coro object representing the current coro (the last coro that |
76 |
the Coro scheduler switched to). The initial value is $Coro::main |
77 |
(of course). |
78 |
|
79 |
This variable is strictly *read-only*. You can take copies of the |
80 |
value stored in it and use it as any other Coro object, but you must |
81 |
not otherwise modify the variable itself. |
82 |
|
83 |
$Coro::idle |
84 |
This variable is mainly useful to integrate Coro into event loops. |
85 |
It is usually better to rely on Coro::AnyEvent or Coro::EV, as this |
86 |
is pretty low-level functionality. |
87 |
|
88 |
This variable stores a Coro object that is put into the ready queue |
89 |
when there are no other ready threads (without invoking any ready |
90 |
hooks). |
91 |
|
92 |
The default implementation dies with "FATAL: deadlock detected.", |
93 |
followed by a thread listing, because the program has no other way |
94 |
to continue. |
95 |
|
96 |
This hook is overwritten by modules such as "Coro::EV" and |
97 |
"Coro::AnyEvent" to wait on an external event that hopefully wake up |
98 |
a coro so the scheduler can run it. |
99 |
|
100 |
See Coro::EV or Coro::AnyEvent for examples of using this technique. |
101 |
|
102 |
SIMPLE CORO CREATION |
103 |
async { ... } [@args...] |
104 |
Create a new coro and return its Coro object (usually unused). The |
105 |
coro will be put into the ready queue, so it will start running |
106 |
automatically on the next scheduler run. |
107 |
|
108 |
The first argument is a codeblock/closure that should be executed in |
109 |
the coro. When it returns argument returns the coro is automatically |
110 |
terminated. |
111 |
|
112 |
The remaining arguments are passed as arguments to the closure. |
113 |
|
114 |
See the "Coro::State::new" constructor for info about the coro |
115 |
environment in which coro are executed. |
116 |
|
117 |
Calling "exit" in a coro will do the same as calling exit outside |
118 |
the coro. Likewise, when the coro dies, the program will exit, just |
119 |
as it would in the main program. |
120 |
|
121 |
If you do not want that, you can provide a default "die" handler, or |
122 |
simply avoid dieing (by use of "eval"). |
123 |
|
124 |
Example: Create a new coro that just prints its arguments. |
125 |
|
126 |
async { |
127 |
print "@_\n"; |
128 |
} 1,2,3,4; |
129 |
|
130 |
async_pool { ... } [@args...] |
131 |
Similar to "async", but uses a coro pool, so you should not call |
132 |
terminate or join on it (although you are allowed to), and you get a |
133 |
coro that might have executed other code already (which can be good |
134 |
or bad :). |
135 |
|
136 |
On the plus side, this function is about twice as fast as creating |
137 |
(and destroying) a completely new coro, so if you need a lot of |
138 |
generic coros in quick successsion, use "async_pool", not "async". |
139 |
|
140 |
The code block is executed in an "eval" context and a warning will |
141 |
be issued in case of an exception instead of terminating the |
142 |
program, as "async" does. As the coro is being reused, stuff like |
143 |
"on_destroy" will not work in the expected way, unless you call |
144 |
terminate or cancel, which somehow defeats the purpose of pooling |
145 |
(but is fine in the exceptional case). |
146 |
|
147 |
The priority will be reset to 0 after each run, tracing will be |
148 |
disabled, the description will be reset and the default output |
149 |
filehandle gets restored, so you can change all these. Otherwise the |
150 |
coro will be re-used "as-is": most notably if you change other |
151 |
per-coro global stuff such as $/ you *must needs* revert that |
152 |
change, which is most simply done by using local as in: "local $/". |
153 |
|
154 |
The idle pool size is limited to 8 idle coros (this can be adjusted |
155 |
by changing $Coro::POOL_SIZE), but there can be as many non-idle |
156 |
coros as required. |
157 |
|
158 |
If you are concerned about pooled coros growing a lot because a |
159 |
single "async_pool" used a lot of stackspace you can e.g. |
160 |
"async_pool { terminate }" once per second or so to slowly replenish |
161 |
the pool. In addition to that, when the stacks used by a handler |
162 |
grows larger than 32kb (adjustable via $Coro::POOL_RSS) it will also |
163 |
be destroyed. |
164 |
|
165 |
STATIC METHODS |
166 |
Static methods are actually functions that implicitly operate on the |
167 |
current coro. |
168 |
|
169 |
schedule |
170 |
Calls the scheduler. The scheduler will find the next coro that is |
171 |
to be run from the ready queue and switches to it. The next coro to |
172 |
be run is simply the one with the highest priority that is longest |
173 |
in its ready queue. If there is no coro ready, it will call the |
174 |
$Coro::idle hook. |
175 |
|
176 |
Please note that the current coro will *not* be put into the ready |
177 |
queue, so calling this function usually means you will never be |
178 |
called again unless something else (e.g. an event handler) calls |
179 |
"->ready", thus waking you up. |
180 |
|
181 |
This makes "schedule" *the* generic method to use to block the |
182 |
current coro and wait for events: first you remember the current |
183 |
coro in a variable, then arrange for some callback of yours to call |
184 |
"->ready" on that once some event happens, and last you call |
185 |
"schedule" to put yourself to sleep. Note that a lot of things can |
186 |
wake your coro up, so you need to check whether the event indeed |
187 |
happened, e.g. by storing the status in a variable. |
188 |
|
189 |
See HOW TO WAIT FOR A CALLBACK, below, for some ways to wait for |
190 |
callbacks. |
191 |
|
192 |
cede |
193 |
"Cede" to other coros. This function puts the current coro into the |
194 |
ready queue and calls "schedule", which has the effect of giving up |
195 |
the current "timeslice" to other coros of the same or higher |
196 |
priority. Once your coro gets its turn again it will automatically |
197 |
be resumed. |
198 |
|
199 |
This function is often called "yield" in other languages. |
200 |
|
201 |
Coro::cede_notself |
202 |
Works like cede, but is not exported by default and will cede to |
203 |
*any* coro, regardless of priority. This is useful sometimes to |
204 |
ensure progress is made. |
205 |
|
206 |
terminate [arg...] |
207 |
Terminates the current coro with the given status values (see |
208 |
cancel). |
209 |
|
210 |
Coro::on_enter BLOCK, Coro::on_leave BLOCK |
211 |
These function install enter and leave winders in the current scope. |
212 |
The enter block will be executed when on_enter is called and |
213 |
whenever the current coro is re-entered by the scheduler, while the |
214 |
leave block is executed whenever the current coro is blocked by the |
215 |
scheduler, and also when the containing scope is exited (by whatever |
216 |
means, be it exit, die, last etc.). |
217 |
|
218 |
*Neither invoking the scheduler, nor exceptions, are allowed within |
219 |
those BLOCKs*. That means: do not even think about calling "die" |
220 |
without an eval, and do not even think of entering the scheduler in |
221 |
any way. |
222 |
|
223 |
Since both BLOCKs are tied to the current scope, they will |
224 |
automatically be removed when the current scope exits. |
225 |
|
226 |
These functions implement the same concept as "dynamic-wind" in |
227 |
scheme does, and are useful when you want to localise some resource |
228 |
to a specific coro. |
229 |
|
230 |
They slow down thread switching considerably for coros that use them |
231 |
(about 40% for a BLOCK with a single assignment, so thread switching |
232 |
is still reasonably fast if the handlers are fast). |
233 |
|
234 |
These functions are best understood by an example: The following |
235 |
function will change the current timezone to |
236 |
"Antarctica/South_Pole", which requires a call to "tzset", but by |
237 |
using "on_enter" and "on_leave", which remember/change the current |
238 |
timezone and restore the previous value, respectively, the timezone |
239 |
is only changed for the coro that installed those handlers. |
240 |
|
241 |
use POSIX qw(tzset); |
242 |
|
243 |
async { |
244 |
my $old_tz; # store outside TZ value here |
245 |
|
246 |
Coro::on_enter { |
247 |
$old_tz = $ENV{TZ}; # remember the old value |
248 |
|
249 |
$ENV{TZ} = "Antarctica/South_Pole"; |
250 |
tzset; # enable new value |
251 |
}; |
252 |
|
253 |
Coro::on_leave { |
254 |
$ENV{TZ} = $old_tz; |
255 |
tzset; # restore old value |
256 |
}; |
257 |
|
258 |
# at this place, the timezone is Antarctica/South_Pole, |
259 |
# without disturbing the TZ of any other coro. |
260 |
}; |
261 |
|
262 |
This can be used to localise about any resource (locale, uid, |
263 |
current working directory etc.) to a block, despite the existance of |
264 |
other coros. |
265 |
|
266 |
Another interesting example implements time-sliced multitasking |
267 |
using interval timers (this could obviously be optimised, but does |
268 |
the job): |
269 |
|
270 |
# "timeslice" the given block |
271 |
sub timeslice(&) { |
272 |
use Time::HiRes (); |
273 |
|
274 |
Coro::on_enter { |
275 |
# on entering the thread, we set an VTALRM handler to cede |
276 |
$SIG{VTALRM} = sub { cede }; |
277 |
# and then start the interval timer |
278 |
Time::HiRes::setitimer &Time::HiRes::ITIMER_VIRTUAL, 0.01, 0.01; |
279 |
}; |
280 |
Coro::on_leave { |
281 |
# on leaving the thread, we stop the interval timer again |
282 |
Time::HiRes::setitimer &Time::HiRes::ITIMER_VIRTUAL, 0, 0; |
283 |
}; |
284 |
|
285 |
&{+shift}; |
286 |
} |
287 |
|
288 |
# use like this: |
289 |
timeslice { |
290 |
# The following is an endless loop that would normally |
291 |
# monopolise the process. Since it runs in a timesliced |
292 |
# environment, it will regularly cede to other threads. |
293 |
while () { } |
294 |
}; |
295 |
|
296 |
killall |
297 |
Kills/terminates/cancels all coros except the currently running one. |
298 |
|
299 |
Note that while this will try to free some of the main interpreter |
300 |
resources if the calling coro isn't the main coro, but one cannot |
301 |
free all of them, so if a coro that is not the main coro calls this |
302 |
function, there will be some one-time resource leak. |
303 |
|
304 |
CORO OBJECT METHODS |
305 |
These are the methods you can call on coro objects (or to create them). |
306 |
|
307 |
new Coro \&sub [, @args...] |
308 |
Create a new coro and return it. When the sub returns, the coro |
309 |
automatically terminates as if "terminate" with the returned values |
310 |
were called. To make the coro run you must first put it into the |
311 |
ready queue by calling the ready method. |
312 |
|
313 |
See "async" and "Coro::State::new" for additional info about the |
314 |
coro environment. |
315 |
|
316 |
$success = $coro->ready |
317 |
Put the given coro into the end of its ready queue (there is one |
318 |
queue for each priority) and return true. If the coro is already in |
319 |
the ready queue, do nothing and return false. |
320 |
|
321 |
This ensures that the scheduler will resume this coro automatically |
322 |
once all the coro of higher priority and all coro of the same |
323 |
priority that were put into the ready queue earlier have been |
324 |
resumed. |
325 |
|
326 |
$coro->suspend |
327 |
Suspends the specified coro. A suspended coro works just like any |
328 |
other coro, except that the scheduler will not select a suspended |
329 |
coro for execution. |
330 |
|
331 |
Suspending a coro can be useful when you want to keep the coro from |
332 |
running, but you don't want to destroy it, or when you want to |
333 |
temporarily freeze a coro (e.g. for debugging) to resume it later. |
334 |
|
335 |
A scenario for the former would be to suspend all (other) coros |
336 |
after a fork and keep them alive, so their destructors aren't |
337 |
called, but new coros can be created. |
338 |
|
339 |
$coro->resume |
340 |
If the specified coro was suspended, it will be resumed. Note that |
341 |
when the coro was in the ready queue when it was suspended, it might |
342 |
have been unreadied by the scheduler, so an activation might have |
343 |
been lost. |
344 |
|
345 |
To avoid this, it is best to put a suspended coro into the ready |
346 |
queue unconditionally, as every synchronisation mechanism must |
347 |
protect itself against spurious wakeups, and the one in the Coro |
348 |
family certainly do that. |
349 |
|
350 |
$is_ready = $coro->is_ready |
351 |
Returns true iff the Coro object is in the ready queue. Unless the |
352 |
Coro object gets destroyed, it will eventually be scheduled by the |
353 |
scheduler. |
354 |
|
355 |
$is_running = $coro->is_running |
356 |
Returns true iff the Coro object is currently running. Only one Coro |
357 |
object can ever be in the running state (but it currently is |
358 |
possible to have multiple running Coro::States). |
359 |
|
360 |
$is_suspended = $coro->is_suspended |
361 |
Returns true iff this Coro object has been suspended. Suspended |
362 |
Coros will not ever be scheduled. |
363 |
|
364 |
$coro->cancel (arg...) |
365 |
Terminates the given Coro and makes it return the given arguments as |
366 |
status (default: the empty list). Never returns if the Coro is the |
367 |
current Coro. |
368 |
|
369 |
$coro->schedule_to |
370 |
Puts the current coro to sleep (like "Coro::schedule"), but instead |
371 |
of continuing with the next coro from the ready queue, always switch |
372 |
to the given coro object (regardless of priority etc.). The |
373 |
readyness state of that coro isn't changed. |
374 |
|
375 |
This is an advanced method for special cases - I'd love to hear |
376 |
about any uses for this one. |
377 |
|
378 |
$coro->cede_to |
379 |
Like "schedule_to", but puts the current coro into the ready queue. |
380 |
This has the effect of temporarily switching to the given coro, and |
381 |
continuing some time later. |
382 |
|
383 |
This is an advanced method for special cases - I'd love to hear |
384 |
about any uses for this one. |
385 |
|
386 |
$coro->throw ([$scalar]) |
387 |
If $throw is specified and defined, it will be thrown as an |
388 |
exception inside the coro at the next convenient point in time. |
389 |
Otherwise clears the exception object. |
390 |
|
391 |
Coro will check for the exception each time a schedule-like-function |
392 |
returns, i.e. after each "schedule", "cede", |
393 |
"Coro::Semaphore->down", "Coro::Handle->readable" and so on. Most of |
394 |
these functions detect this case and return early in case an |
395 |
exception is pending. |
396 |
|
397 |
The exception object will be thrown "as is" with the specified |
398 |
scalar in $@, i.e. if it is a string, no line number or newline will |
399 |
be appended (unlike with "die"). |
400 |
|
401 |
This can be used as a softer means than "cancel" to ask a coro to |
402 |
end itself, although there is no guarantee that the exception will |
403 |
lead to termination, and if the exception isn't caught it might well |
404 |
end the whole program. |
405 |
|
406 |
You might also think of "throw" as being the moral equivalent of |
407 |
"kill"ing a coro with a signal (in this case, a scalar). |
408 |
|
409 |
$coro->join |
410 |
Wait until the coro terminates and return any values given to the |
411 |
"terminate" or "cancel" functions. "join" can be called concurrently |
412 |
from multiple coro, and all will be resumed and given the status |
413 |
return once the $coro terminates. |
414 |
|
415 |
$coro->on_destroy (\&cb) |
416 |
Registers a callback that is called when this coro thread gets |
417 |
destroyed, but before it is joined. The callback gets passed the |
418 |
terminate arguments, if any, and *must not* die, under any |
419 |
circumstances. |
420 |
|
421 |
There can be any number of "on_destroy" callbacks per coro. |
422 |
|
423 |
$oldprio = $coro->prio ($newprio) |
424 |
Sets (or gets, if the argument is missing) the priority of the coro |
425 |
thread. Higher priority coro get run before lower priority coros. |
426 |
Priorities are small signed integers (currently -4 .. +3), that you |
427 |
can refer to using PRIO_xxx constants (use the import tag :prio to |
428 |
get then): |
429 |
|
430 |
PRIO_MAX > PRIO_HIGH > PRIO_NORMAL > PRIO_LOW > PRIO_IDLE > PRIO_MIN |
431 |
3 > 1 > 0 > -1 > -3 > -4 |
432 |
|
433 |
# set priority to HIGH |
434 |
current->prio (PRIO_HIGH); |
435 |
|
436 |
The idle coro thread ($Coro::idle) always has a lower priority than |
437 |
any existing coro. |
438 |
|
439 |
Changing the priority of the current coro will take effect |
440 |
immediately, but changing the priority of a coro in the ready queue |
441 |
(but not running) will only take effect after the next schedule (of |
442 |
that coro). This is a bug that will be fixed in some future version. |
443 |
|
444 |
$newprio = $coro->nice ($change) |
445 |
Similar to "prio", but subtract the given value from the priority |
446 |
(i.e. higher values mean lower priority, just as in UNIX's nice |
447 |
command). |
448 |
|
449 |
$olddesc = $coro->desc ($newdesc) |
450 |
Sets (or gets in case the argument is missing) the description for |
451 |
this coro thread. This is just a free-form string you can associate |
452 |
with a coro. |
453 |
|
454 |
This method simply sets the "$coro->{desc}" member to the given |
455 |
string. You can modify this member directly if you wish, and in |
456 |
fact, this is often preferred to indicate major processing states |
457 |
that cna then be seen for example in a Coro::Debug session: |
458 |
|
459 |
sub my_long_function { |
460 |
local $Coro::current->{desc} = "now in my_long_function"; |
461 |
... |
462 |
$Coro::current->{desc} = "my_long_function: phase 1"; |
463 |
... |
464 |
$Coro::current->{desc} = "my_long_function: phase 2"; |
465 |
... |
466 |
} |
467 |
|
468 |
GLOBAL FUNCTIONS |
469 |
Coro::nready |
470 |
Returns the number of coro that are currently in the ready state, |
471 |
i.e. that can be switched to by calling "schedule" directory or |
472 |
indirectly. The value 0 means that the only runnable coro is the |
473 |
currently running one, so "cede" would have no effect, and |
474 |
"schedule" would cause a deadlock unless there is an idle handler |
475 |
that wakes up some coro. |
476 |
|
477 |
my $guard = Coro::guard { ... } |
478 |
This function still exists, but is deprecated. Please use the |
479 |
"Guard::guard" function instead. |
480 |
|
481 |
unblock_sub { ... } |
482 |
This utility function takes a BLOCK or code reference and "unblocks" |
483 |
it, returning a new coderef. Unblocking means that calling the new |
484 |
coderef will return immediately without blocking, returning nothing, |
485 |
while the original code ref will be called (with parameters) from |
486 |
within another coro. |
487 |
|
488 |
The reason this function exists is that many event libraries (such |
489 |
as the venerable Event module) are not thread-safe (a weaker form of |
490 |
reentrancy). This means you must not block within event callbacks, |
491 |
otherwise you might suffer from crashes or worse. The only event |
492 |
library currently known that is safe to use without "unblock_sub" is |
493 |
EV (but you might still run into deadlocks if all event loops are |
494 |
blocked). |
495 |
|
496 |
Coro will try to catch you when you block in the event loop |
497 |
("FATAL:$Coro::IDLE blocked itself"), but this is just best effort |
498 |
and only works when you do not run your own event loop. |
499 |
|
500 |
This function allows your callbacks to block by executing them in |
501 |
another coro where it is safe to block. One example where blocking |
502 |
is handy is when you use the Coro::AIO functions to save results to |
503 |
disk, for example. |
504 |
|
505 |
In short: simply use "unblock_sub { ... }" instead of "sub { ... }" |
506 |
when creating event callbacks that want to block. |
507 |
|
508 |
If your handler does not plan to block (e.g. simply sends a message |
509 |
to another coro, or puts some other coro into the ready queue), |
510 |
there is no reason to use "unblock_sub". |
511 |
|
512 |
Note that you also need to use "unblock_sub" for any other callbacks |
513 |
that are indirectly executed by any C-based event loop. For example, |
514 |
when you use a module that uses AnyEvent (and you use |
515 |
Coro::AnyEvent) and it provides callbacks that are the result of |
516 |
some event callback, then you must not block either, or use |
517 |
"unblock_sub". |
518 |
|
519 |
$cb = rouse_cb |
520 |
Create and return a "rouse callback". That's a code reference that, |
521 |
when called, will remember a copy of its arguments and notify the |
522 |
owner coro of the callback. |
523 |
|
524 |
See the next function. |
525 |
|
526 |
@args = rouse_wait [$cb] |
527 |
Wait for the specified rouse callback (or the last one that was |
528 |
created in this coro). |
529 |
|
530 |
As soon as the callback is invoked (or when the callback was invoked |
531 |
before "rouse_wait"), it will return the arguments originally passed |
532 |
to the rouse callback. In scalar context, that means you get the |
533 |
*last* argument, just as if "rouse_wait" had a "return ($a1, $a2, |
534 |
$a3...)" statement at the end. |
535 |
|
536 |
See the section HOW TO WAIT FOR A CALLBACK for an actual usage |
537 |
example. |
538 |
|
539 |
HOW TO WAIT FOR A CALLBACK |
540 |
It is very common for a coro to wait for some callback to be called. |
541 |
This occurs naturally when you use coro in an otherwise event-based |
542 |
program, or when you use event-based libraries. |
543 |
|
544 |
These typically register a callback for some event, and call that |
545 |
callback when the event occured. In a coro, however, you typically want |
546 |
to just wait for the event, simplyifying things. |
547 |
|
548 |
For example "AnyEvent->child" registers a callback to be called when a |
549 |
specific child has exited: |
550 |
|
551 |
my $child_watcher = AnyEvent->child (pid => $pid, cb => sub { ... }); |
552 |
|
553 |
But from within a coro, you often just want to write this: |
554 |
|
555 |
my $status = wait_for_child $pid; |
556 |
|
557 |
Coro offers two functions specifically designed to make this easy, |
558 |
"Coro::rouse_cb" and "Coro::rouse_wait". |
559 |
|
560 |
The first function, "rouse_cb", generates and returns a callback that, |
561 |
when invoked, will save its arguments and notify the coro that created |
562 |
the callback. |
563 |
|
564 |
The second function, "rouse_wait", waits for the callback to be called |
565 |
(by calling "schedule" to go to sleep) and returns the arguments |
566 |
originally passed to the callback. |
567 |
|
568 |
Using these functions, it becomes easy to write the "wait_for_child" |
569 |
function mentioned above: |
570 |
|
571 |
sub wait_for_child($) { |
572 |
my ($pid) = @_; |
573 |
|
574 |
my $watcher = AnyEvent->child (pid => $pid, cb => Coro::rouse_cb); |
575 |
|
576 |
my ($rpid, $rstatus) = Coro::rouse_wait; |
577 |
$rstatus |
578 |
} |
579 |
|
580 |
In the case where "rouse_cb" and "rouse_wait" are not flexible enough, |
581 |
you can roll your own, using "schedule": |
582 |
|
583 |
sub wait_for_child($) { |
584 |
my ($pid) = @_; |
585 |
|
586 |
# store the current coro in $current, |
587 |
# and provide result variables for the closure passed to ->child |
588 |
my $current = $Coro::current; |
589 |
my ($done, $rstatus); |
590 |
|
591 |
# pass a closure to ->child |
592 |
my $watcher = AnyEvent->child (pid => $pid, cb => sub { |
593 |
$rstatus = $_[1]; # remember rstatus |
594 |
$done = 1; # mark $rstatus as valud |
595 |
}); |
596 |
|
597 |
# wait until the closure has been called |
598 |
schedule while !$done; |
599 |
|
600 |
$rstatus |
601 |
} |
602 |
|
603 |
BUGS/LIMITATIONS |
604 |
fork with pthread backend |
605 |
When Coro is compiled using the pthread backend (which isn't |
606 |
recommended but required on many BSDs as their libcs are completely |
607 |
broken), then coro will not survive a fork. There is no known |
608 |
workaround except to fix your libc and use a saner backend. |
609 |
|
610 |
perl process emulation ("threads") |
611 |
This module is not perl-pseudo-thread-safe. You should only ever use |
612 |
this module from the first thread (this requirement might be removed |
613 |
in the future to allow per-thread schedulers, but Coro::State does |
614 |
not yet allow this). I recommend disabling thread support and using |
615 |
processes, as having the windows process emulation enabled under |
616 |
unix roughly halves perl performance, even when not used. |
617 |
|
618 |
coro switching is not signal safe |
619 |
You must not switch to another coro from within a signal handler |
620 |
(only relevant with %SIG - most event libraries provide safe |
621 |
signals), *unless* you are sure you are not interrupting a Coro |
622 |
function. |
623 |
|
624 |
That means you *MUST NOT* call any function that might "block" the |
625 |
current coro - "cede", "schedule" "Coro::Semaphore->down" or |
626 |
anything that calls those. Everything else, including calling |
627 |
"ready", works. |
628 |
|
629 |
WINDOWS PROCESS EMULATION |
630 |
A great many people seem to be confused about ithreads (for example, |
631 |
Chip Salzenberg called me unintelligent, incapable, stupid and gullible, |
632 |
while in the same mail making rather confused statements about perl |
633 |
ithreads (for example, that memory or files would be shared), showing |
634 |
his lack of understanding of this area - if it is hard to understand for |
635 |
Chip, it is probably not obvious to everybody). |
636 |
|
637 |
What follows is an ultra-condensed version of my talk about threads in |
638 |
scripting languages given on the perl workshop 2009: |
639 |
|
640 |
The so-called "ithreads" were originally implemented for two reasons: |
641 |
first, to (badly) emulate unix processes on native win32 perls, and |
642 |
secondly, to replace the older, real thread model ("5.005-threads"). |
643 |
|
644 |
It does that by using threads instead of OS processes. The difference |
645 |
between processes and threads is that threads share memory (and other |
646 |
state, such as files) between threads within a single process, while |
647 |
processes do not share anything (at least not semantically). That means |
648 |
that modifications done by one thread are seen by others, while |
649 |
modifications by one process are not seen by other processes. |
650 |
|
651 |
The "ithreads" work exactly like that: when creating a new ithreads |
652 |
process, all state is copied (memory is copied physically, files and |
653 |
code is copied logically). Afterwards, it isolates all modifications. On |
654 |
UNIX, the same behaviour can be achieved by using operating system |
655 |
processes, except that UNIX typically uses hardware built into the |
656 |
system to do this efficiently, while the windows process emulation |
657 |
emulates this hardware in software (rather efficiently, but of course it |
658 |
is still much slower than dedicated hardware). |
659 |
|
660 |
As mentioned before, loading code, modifying code, modifying data |
661 |
structures and so on is only visible in the ithreads process doing the |
662 |
modification, not in other ithread processes within the same OS process. |
663 |
|
664 |
This is why "ithreads" do not implement threads for perl at all, only |
665 |
processes. What makes it so bad is that on non-windows platforms, you |
666 |
can actually take advantage of custom hardware for this purpose (as |
667 |
evidenced by the forks module, which gives you the (i-) threads API, |
668 |
just much faster). |
669 |
|
670 |
Sharing data is in the i-threads model is done by transfering data |
671 |
structures between threads using copying semantics, which is very slow - |
672 |
shared data simply does not exist. Benchmarks using i-threads which are |
673 |
communication-intensive show extremely bad behaviour with i-threads (in |
674 |
fact, so bad that Coro, which cannot take direct advantage of multiple |
675 |
CPUs, is often orders of magnitude faster because it shares data using |
676 |
real threads, refer to my talk for details). |
677 |
|
678 |
As summary, i-threads *use* threads to implement processes, while the |
679 |
compatible forks module *uses* processes to emulate, uhm, processes. |
680 |
I-threads slow down every perl program when enabled, and outside of |
681 |
windows, serve no (or little) practical purpose, but disadvantages every |
682 |
single-threaded Perl program. |
683 |
|
684 |
This is the reason that I try to avoid the name "ithreads", as it is |
685 |
misleading as it implies that it implements some kind of thread model |
686 |
for perl, and prefer the name "windows process emulation", which |
687 |
describes the actual use and behaviour of it much better. |
688 |
|
689 |
SEE ALSO |
690 |
Event-Loop integration: Coro::AnyEvent, Coro::EV, Coro::Event. |
691 |
|
692 |
Debugging: Coro::Debug. |
693 |
|
694 |
Support/Utility: Coro::Specific, Coro::Util. |
695 |
|
696 |
Locking and IPC: Coro::Signal, Coro::Channel, Coro::Semaphore, |
697 |
Coro::SemaphoreSet, Coro::RWLock. |
698 |
|
699 |
I/O and Timers: Coro::Timer, Coro::Handle, Coro::Socket, Coro::AIO. |
700 |
|
701 |
Compatibility with other modules: Coro::LWP (but see also AnyEvent::HTTP |
702 |
for a better-working alternative), Coro::BDB, Coro::Storable, |
703 |
Coro::Select. |
704 |
|
705 |
XS API: Coro::MakeMaker. |
706 |
|
707 |
Low level Configuration, Thread Environment, Continuations: Coro::State. |
708 |
|
709 |
AUTHOR |
710 |
Marc Lehmann <schmorp@schmorp.de> |
711 |
http://home.schmorp.de/ |
712 |
|