libPhenom is an eventing framework for building high performance and high scalability systems in C
libPhenom is known to compile and pass its test suite on:
- Linux systems with
epoll - OS X
libPhenom has been known to compile and pass its test suite on these systems, but they have not been tried in a little while, so may require a little bit of TLC:
- BSDish systems that have the
kqueue(2)facility, including FreeBSD 9.1 and OpenBSD 5.2 - Illumos and Solaris style systems that have
port_create(3C).
libPhenom depends on:
- c-ares for DNS resolution.
It expects to find it via
pkg-config; you need to provide this in order for phenom to build successfully. - Concurrency Kit for its excellent concurrency primitives and key data structures. We include CK with phenom.
- autoconf and automake are required to build libPhenom
libPhenom works best if built with GCC version 4.3 or later, but should be able to build with any C99 compiler.
- Memory management with counters - record how much of which kinds of memory your application is using.
- Jobs - decompose your application into portions of work and let the phenom scheduler manage getting them done
- streaming I/O with buffers
- Handy data structures (hash tables, lists, queues)
- Variant data type to enable serialization and deserialization of JSON
- A printf implementation with registerable object formatting
- Balance ease of use with performance
- Aim to be neutral wrt. your choice of threaded or event-based dispatch and work well with both.
- Where possible, avoid contention points in our implementation so as to avoid limiting scalability with the number of cores in the system.
If you're reading these on http://facebook.github.io/libphenom, simply start
typing and the search box will suggest topics. You may select topics from the
Topics menu or browse the header files via the Headers menu.
You can obtain the sources from https://github.com/facebook/libphenom:
$ git clone https://github.com/facebook/libphenom.git$ ./autogen.sh
$ ./configure
$ make
$ make checkYou'll want to set up the main loop using something like this:
#include "phenom/defs.h"
#include "phenom/configuration.h"
#include "phenom/job.h"
#include "phenom/log.h"
#include "phenom/sysutil.h"
int main(int argc, char **argv)
{
// Must be called prior to calling any other phenom functions
ph_library_init();
// Optional config file for tuning internals
ph_config_load_config_file("/path/to/my/config.json");
// Enable the non-blocking IO manager
ph_nbio_init(0);
// Do stuff here to register client/server stuff.
// This enables a very simple request/response console
// that allows you to run diagnostic commands:
// `echo memory | nc -UC /tmp/phenom-debug-console`
// The code behind this is in
// https://github.com/facebook/libphenom/blob/master/corelib/debug_console.c
ph_debug_console_start("/tmp/phenom-debug-console");
// Run
ph_sched_run();
return 0;
}And compile it using something like this:
$ gcc main.c `pkg-config libphenom --cflags --libs`(this will link against libphenom and libcares).
Want more inspiration? Take a look at the code in the test suite.
We're still hacking and evolving this library, so there may be some rough edges. We're very open to feedback; check out the Contributing section below.
If you're thinking of hacking on libPhenom we'd love to hear from you! Feel free to use the Github issue tracker and pull requests to discuss and submit code changes.
We (Facebook) have to ask for a "Contributor License Agreement" from someone who sends in a patch or code that we want to include in the codebase. This is a legal requirement; a similar situation applies to Apache and other ASF projects.
If we ask you to fill out a CLA we'll direct you to our online CLA page where you can complete it easily. We use the same form as the Apache CLA so that friction is minimal.
libPhenom is made available under the terms of the Apache License 2.0. See the LICENSE file that accompanies this distribution for the full text of the license.