[FFmpeg-devel] Automated builds and some tests

Panagiotis Issaris takis.issaris
Wed Jul 4 18:34:44 CEST 2007

Hash: SHA1

Hi Michael,

Michael Niedermayer wrote:
> Hi
> On Wed, Jul 04, 2007 at 05:15:19PM +0200, Panagiotis Issaris wrote:
> [...]
>>>> * Encoding times (well, actually FPS) for: H.261, H.263, MPEG-4 and Snow
>>> ditto
>> Same here.
> running benchmarks on a system which also runs other stuff like a webserver
> is not good, not in theory and in practice the FPS values look quite noisy

Well, at the moment the FPS values are worthless. I am well aware of
that. I just wanted to share my enthusiasm :) by showing what I was
intending to do.

> no way anyone could notice a 1% change that way ...

No, because currently it is not accurate at all. And because differences
between two revisions are too small.  But, using the database data one
could plot the encoding performance over long periods of time. That
should be nice :)

> what about sigstoping everything before a benchmark is run and running a few
> sync before it too?

I've done some benchmarks for the RTAI project as I mentioned in another
mail. For that project I created a dedicated livecd running nothing but
the benchmarking code. And the actual timing was done from within the
hard realtime threads, but probably I could just nice the process for
ffmpeg benchmarks.

> also run each benchmark several times, discard the first run and print
> average and standard deviation for the scores ...

Yep, I did that same for RTAI [1]. But, haven't gotten to it yet for
this new webpage. Thanks for the suggestions though! :)

With friendly regards,

[1] http://issaris.org/rtai/
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org


More information about the ffmpeg-devel mailing list