[x264-devel] h.264 encoding live stream from a web cam

[Jason Garrett-Glaser]

On Sun, Aug 23, 2009 at 4:12 PM, i-love-spam<i-love-spam at yandex.ru> wrote:
> 24.08.09, 02:45, "Jason Garrett-Glaser" <darkshikari at gmail.com>:
> ..
>> You should use the x264 library directly, not ffmpeg, if possible.
>> x264's CPU usage depends on the encoding settings you use; see the
>> presets in x264.c for examples.  --preset ultrafast, for example, is
>> over 100 times faster than --preset placebo.
>> Additionally, you may want to make sure that you didn't compile
>> without the assembly code enabled.
> ..
>> Tell your debugger to ignore thread terminations, or pipe the output
>> to "grep -v "Win32 Thread".
>> > and one small question: what's that project direct264 and what's the difference? It seems that x264 doesn't compile out of the box on MSVC
>> MSVC is not a supported compiler; use GCC or ICC.  Direct264 is an
>> unsupported fork of x264 for use as a DirectShow filter.
>
> Thanks for prompt reply!
>
> I prefer to use ffmpeg, since I use h263 from ffmepg, and many other utilities.
> Are there any drawbacks to using ffmpeg than x264 directly?

ffmpeg's option system with regards to x264 is braindead; the defaults
that it uses are completely broken.  You basically have to re-set all
the x264 defaults and then go back through and apply your settings.
Of course, if you can do all this and be sure you don't screw up,
you're fine.

The presets in ffpresets/libx264* in the ffmpeg source tree show
command-line examples of how to set sane x264 settings.  The "default"
preset has what used to be the x264 defaults, though it is now
slightly inaccurate since the x264 defaults changed to High Profile.

> By the way, are the any info on running x264 on wince devices? Is there arm optimized code, or x264 is x86 only?

You're asking at a good time--in a day or two now we will be
committing the results of one of the two successful Google Summer of
Code projects this for year--ARM support and assembly optimizations
for NEON SIMD.  This should vastly improve performance on Cortex A8
and Cortex A9 ARM chips, allowing for high-performance mobile encoding
at ultra low power.

Dark Shikari