Browse Source

milli-opt

master
Marc Alexander Lehmann 14 years ago
parent
commit
c24b26fe30
4 changed files with 207 additions and 39 deletions
  1. +87
    -4
      ev.3
  2. +34
    -31
      ev.c
  3. +1
    -0
      ev.h
  4. +85
    -4
      ev.html

+ 87
- 4
ev.3 View File

@ -129,7 +129,7 @@
.\" ========================================================================
.\"
.IX Title ""<STANDARD INPUT>" 1"
.TH "<STANDARD INPUT>" 1 "2007-11-23" "perl v5.8.8" "User Contributed Perl Documentation"
.TH "<STANDARD INPUT>" 1 "2007-11-24" "perl v5.8.8" "User Contributed Perl Documentation"
.SH "NAME"
libev \- a high performance full\-featured event loop written in C
.SH "SYNOPSIS"
@ -233,6 +233,15 @@ returned by \f(CW\*(C`ev_supported_backends\*(C'\fR, as for example kqueue is br
most BSDs and will not be autodetected unless you explicitly request it
(assuming you know what you are doing). This is the set of backends that
libev will probe for if you specify no backends explicitly.
.IP "unsigned int ev_embeddable_backends ()" 4
.IX Item "unsigned int ev_embeddable_backends ()"
Returns the set of backends that are embeddable in other event loops. This
is the theoretical, all\-platform, value. To find which backends
might be supported on the current system, you would need to look at
\&\f(CW\*(C`ev_embeddable_backends () & ev_supported_backends ()\*(C'\fR, likewise for
recommended ones.
.Sp
See the description of \f(CW\*(C`ev_embed\*(C'\fR watchers for more info.
.IP "ev_set_allocator (void *(*cb)(void *ptr, long size))" 4
.IX Item "ev_set_allocator (void *(*cb)(void *ptr, long size))"
Sets the allocation function to use (the prototype is similar to the
@ -1161,9 +1170,10 @@ Prepare and check watchers are usually (but not always) used in tandem:
prepare watchers get invoked before the process blocks and check watchers
afterwards.
.PP
Their main purpose is to integrate other event mechanisms into libev. This
could be used, for example, to track variable changes, implement your own
watchers, integrate net-snmp or a coroutine library and lots more.
Their main purpose is to integrate other event mechanisms into libev and
their use is somewhat advanced. This could be used, for example, to track
variable changes, implement your own watchers, integrate net-snmp or a
coroutine library and lots more.
.PP
This is done by examining in each prepare call which file descriptors need
to be watched by the other library, registering \f(CW\*(C`ev_io\*(C'\fR watchers for
@ -1193,6 +1203,79 @@ parameters of any kind. There are \f(CW\*(C`ev_prepare_set\*(C'\fR and \f(CW\*(C
macros, but using them is utterly, utterly and completely pointless.
.PP
Example: *TODO*.
.ie n .Sh """ev_embed"" \- when one backend isn't enough"
.el .Sh "\f(CWev_embed\fP \- when one backend isn't enough"
.IX Subsection "ev_embed - when one backend isn't enough"
This is a rather advanced watcher type that lets you embed one event loop
into another.
.PP
There are primarily two reasons you would want that: work around bugs and
prioritise I/O.
.PP
As an example for a bug workaround, the kqueue backend might only support
sockets on some platform, so it is unusable as generic backend, but you
still want to make use of it because you have many sockets and it scales
so nicely. In this case, you would create a kqueue-based loop and embed it
into your default loop (which might use e.g. poll). Overall operation will
be a bit slower because first libev has to poll and then call kevent, but
at least you can use both at what they are best.
.PP
As for prioritising I/O: rarely you have the case where some fds have
to be watched and handled very quickly (with low latency), and even
priorities and idle watchers might have too much overhead. In this case
you would put all the high priority stuff in one loop and all the rest in
a second one, and embed the second one in the first.
.PP
As long as the watcher is started it will automatically handle events. The
callback will be invoked whenever some events have been handled. You can
set the callback to \f(CW0\fR to avoid having to specify one if you are not
interested in that.
.PP
Also, there have not currently been made special provisions for forking:
when you fork, you not only have to call \f(CW\*(C`ev_loop_fork\*(C'\fR on both loops,
but you will also have to stop and restart any \f(CW\*(C`ev_embed\*(C'\fR watchers
yourself.
.PP
Unfortunately, not all backends are embeddable, only the ones returned by
\&\f(CW\*(C`ev_embeddable_backends\*(C'\fR are, which, unfortunately, does not include any
portable one.
.PP
So when you want to use this feature you will always have to be prepared
that you cannot get an embeddable loop. The recommended way to get around
this is to have a separate variables for your embeddable loop, try to
create it, and if that fails, use the normal loop for everything:
.PP
.Vb 3
\& struct ev_loop *loop_hi = ev_default_init (0);
\& struct ev_loop *loop_lo = 0;
\& struct ev_embed embed;
.Ve
.PP
.Vb 5
\& // see if there is a chance of getting one that works
\& // (remember that a flags value of 0 means autodetection)
\& loop_lo = ev_embeddable_backends () & ev_recommended_backends ()
\& ? ev_loop_new (ev_embeddable_backends () & ev_recommended_backends ())
\& : 0;
.Ve
.PP
.Vb 8
\& // if we got one, then embed it, otherwise default to loop_hi
\& if (loop_lo)
\& {
\& ev_embed_init (&embed, 0, loop_lo);
\& ev_embed_start (loop_hi, &embed);
\& }
\& else
\& loop_lo = loop_hi;
.Ve
.IP "ev_embed_init (ev_embed *, callback, struct ev_loop *loop)" 4
.IX Item "ev_embed_init (ev_embed *, callback, struct ev_loop *loop)"
.PD 0
.IP "ev_embed_set (ev_embed *, callback, struct ev_loop *loop)" 4
.IX Item "ev_embed_set (ev_embed *, callback, struct ev_loop *loop)"
.PD
Configures the watcher to embed the given loop, which must be embeddable.
.SH "OTHER FUNCTIONS"
.IX Header "OTHER FUNCTIONS"
There are some other functions of possible interest. Described. Here. Now.


+ 34
- 31
ev.c View File

@ -1250,8 +1250,9 @@ static int loop_done;
void
ev_loop (EV_P_ int flags)
{
double block;
loop_done = flags & (EVLOOP_ONESHOT | EVLOOP_NONBLOCK) ? 1 : 0;
loop_done = flags & (EVLOOP_ONESHOT | EVLOOP_NONBLOCK)
? EVUNLOOP_ONE
: EVUNLOOP_CANCEL;
while (activecnt)
{
@ -1270,43 +1271,45 @@ ev_loop (EV_P_ int flags)
fd_reify (EV_A);
/* calculate blocking time */
{
double block;
/* we only need this for !monotonic clock or timers, but as we basically
always have timers, we just calculate it always */
if (flags & EVLOOP_NONBLOCK || idlecnt)
block = 0.; /* do not block at all */
else
{
/* update time to cancel out callback processing overhead */
#if EV_USE_MONOTONIC
if (expect_true (have_monotonic))
time_update_monotonic (EV_A);
else
if (expect_true (have_monotonic))
time_update_monotonic (EV_A);
else
#endif
{
ev_rt_now = ev_time ();
mn_now = ev_rt_now;
}
{
ev_rt_now = ev_time ();
mn_now = ev_rt_now;
}
if (flags & EVLOOP_NONBLOCK || idlecnt)
block = 0.;
else
{
block = MAX_BLOCKTIME;
block = MAX_BLOCKTIME;
if (timercnt)
{
ev_tstamp to = ((WT)timers [0])->at - mn_now + backend_fudge;
if (block > to) block = to;
}
if (timercnt)
{
ev_tstamp to = ((WT)timers [0])->at - mn_now + backend_fudge;
if (block > to) block = to;
}
#if EV_PERIODICS
if (periodiccnt)
{
ev_tstamp to = ((WT)periodics [0])->at - ev_rt_now + backend_fudge;
if (block > to) block = to;
}
if (periodiccnt)
{
ev_tstamp to = ((WT)periodics [0])->at - ev_rt_now + backend_fudge;
if (block > to) block = to;
}
#endif
if (expect_false (block < 0.)) block = 0.;
}
if (expect_false (block < 0.)) block = 0.;
}
backend_poll (EV_A_ block);
backend_poll (EV_A_ block);
}
/* update ev_rt_now, do magic */
time_update (EV_A);
@ -1331,8 +1334,8 @@ ev_loop (EV_P_ int flags)
break;
}
if (loop_done != 2)
loop_done = 0;
if (loop_done == EVUNLOOP_ONE)
loop_done = EVUNLOOP_CANCEL;
}
void


+ 1
- 0
ev.h View File

@ -329,6 +329,7 @@ unsigned int ev_backend (EV_P);
#define EVLOOP_NONBLOCK 1 /* do not block/wait */
#define EVLOOP_ONESHOT 2 /* block *once* only */
#define EVUNLOOP_CANCEL 0 /* undo unloop */
#define EVUNLOOP_ONE 1 /* unloop once */
#define EVUNLOOP_ALL 2 /* unloop all loops */


+ 85
- 4
ev.html View File

@ -6,7 +6,7 @@
<meta name="description" content="Pod documentation for libev" />
<meta name="inputfile" content="&lt;standard input&gt;" />
<meta name="outputfile" content="&lt;standard output&gt;" />
<meta name="created" content="Fri Nov 23 17:17:04 2007" />
<meta name="created" content="Sat Nov 24 05:58:35 2007" />
<meta name="generator" content="Pod::Xhtml 1.57" />
<link rel="stylesheet" href="http://res.tst.eu/pod.css"/></head>
<body>
@ -34,6 +34,7 @@
<li><a href="#code_ev_child_code_wait_for_pid_stat"><code>ev_child</code> - wait for pid status changes</a></li>
<li><a href="#code_ev_idle_code_when_you_ve_got_no"><code>ev_idle</code> - when you've got nothing better to do</a></li>
<li><a href="#code_ev_prepare_code_and_code_ev_che"><code>ev_prepare</code> and <code>ev_check</code> - customise your event loop</a></li>
<li><a href="#code_ev_embed_code_when_one_backend_"><code>ev_embed</code> - when one backend isn't enough</a></li>
</ul>
</li>
<li><a href="#OTHER_FUNCTIONS">OTHER FUNCTIONS</a></li>
@ -159,6 +160,15 @@ most BSDs and will not be autodetected unless you explicitly request it
(assuming you know what you are doing). This is the set of backends that
libev will probe for if you specify no backends explicitly.</p>
</dd>
<dt>unsigned int ev_embeddable_backends ()</dt>
<dd>
<p>Returns the set of backends that are embeddable in other event loops. This
is the theoretical, all-platform, value. To find which backends
might be supported on the current system, you would need to look at
<code>ev_embeddable_backends () &amp; ev_supported_backends ()</code>, likewise for
recommended ones.</p>
<p>See the description of <code>ev_embed</code> watchers for more info.</p>
</dd>
<dt>ev_set_allocator (void *(*cb)(void *ptr, long size))</dt>
<dd>
<p>Sets the allocation function to use (the prototype is similar to the
@ -933,6 +943,10 @@ of the <code>SIGxxx</code> constants).</p>
</dd>
</dl>
</div>
<h2 id="code_ev_child_code_wait_for_pid_stat"><code>ev_child</code> - wait for pid status changes</h2>
<div id="code_ev_child_code_wait_for_pid_stat-2">
@ -1015,9 +1029,10 @@ callback, free it. Alos, use no error checking, as usual.</p>
<p>Prepare and check watchers are usually (but not always) used in tandem:
prepare watchers get invoked before the process blocks and check watchers
afterwards.</p>
<p>Their main purpose is to integrate other event mechanisms into libev. This
could be used, for example, to track variable changes, implement your own
watchers, integrate net-snmp or a coroutine library and lots more.</p>
<p>Their main purpose is to integrate other event mechanisms into libev and
their use is somewhat advanced. This could be used, for example, to track
variable changes, implement your own watchers, integrate net-snmp or a
coroutine library and lots more.</p>
<p>This is done by examining in each prepare call which file descriptors need
to be watched by the other library, registering <code>ev_io</code> watchers for
them and starting an <code>ev_timer</code> watcher for any timeouts (many libraries
@ -1049,6 +1064,72 @@ macros, but using them is utterly, utterly and completely pointless.</p>
</div>
<h2 id="code_ev_embed_code_when_one_backend_"><code>ev_embed</code> - when one backend isn't enough</h2>
<div id="code_ev_embed_code_when_one_backend_-2">
<p>This is a rather advanced watcher type that lets you embed one event loop
into another.</p>
<p>There are primarily two reasons you would want that: work around bugs and
prioritise I/O.</p>
<p>As an example for a bug workaround, the kqueue backend might only support
sockets on some platform, so it is unusable as generic backend, but you
still want to make use of it because you have many sockets and it scales
so nicely. In this case, you would create a kqueue-based loop and embed it
into your default loop (which might use e.g. poll). Overall operation will
be a bit slower because first libev has to poll and then call kevent, but
at least you can use both at what they are best.</p>
<p>As for prioritising I/O: rarely you have the case where some fds have
to be watched and handled very quickly (with low latency), and even
priorities and idle watchers might have too much overhead. In this case
you would put all the high priority stuff in one loop and all the rest in
a second one, and embed the second one in the first.</p>
<p>As long as the watcher is started it will automatically handle events. The
callback will be invoked whenever some events have been handled. You can
set the callback to <code>0</code> to avoid having to specify one if you are not
interested in that.</p>
<p>Also, there have not currently been made special provisions for forking:
when you fork, you not only have to call <code>ev_loop_fork</code> on both loops,
but you will also have to stop and restart any <code>ev_embed</code> watchers
yourself.</p>
<p>Unfortunately, not all backends are embeddable, only the ones returned by
<code>ev_embeddable_backends</code> are, which, unfortunately, does not include any
portable one.</p>
<p>So when you want to use this feature you will always have to be prepared
that you cannot get an embeddable loop. The recommended way to get around
this is to have a separate variables for your embeddable loop, try to
create it, and if that fails, use the normal loop for everything:</p>
<pre> struct ev_loop *loop_hi = ev_default_init (0);
struct ev_loop *loop_lo = 0;
struct ev_embed embed;
// see if there is a chance of getting one that works
// (remember that a flags value of 0 means autodetection)
loop_lo = ev_embeddable_backends () &amp; ev_recommended_backends ()
? ev_loop_new (ev_embeddable_backends () &amp; ev_recommended_backends ())
: 0;
// if we got one, then embed it, otherwise default to loop_hi
if (loop_lo)
{
ev_embed_init (&amp;embed, 0, loop_lo);
ev_embed_start (loop_hi, &amp;embed);
}
else
loop_lo = loop_hi;
</pre>
<dl>
<dt>ev_embed_init (ev_embed *, callback, struct ev_loop *loop)</dt>
<dt>ev_embed_set (ev_embed *, callback, struct ev_loop *loop)</dt>
<dd>
<p>Configures the watcher to embed the given loop, which must be embeddable.</p>
</dd>
</dl>
</div>
<h1 id="OTHER_FUNCTIONS">OTHER FUNCTIONS</h1><p><a href="#TOP" class="toplink">Top</a></p>
<div id="OTHER_FUNCTIONS_CONTENT">


Loading…
Cancel
Save