[FFmpeg-user] Zeranoe Windows builds spiking CPU to 100% randomly

[Andy Sheen]

Jan Ehrhardt wrote on Tue 10 Jul at 7:14 UK time
> Roger Pack in gmane.comp.video.ffmpeg.user (Mon, 9 Jul 2012 12:14:46
> -0600):
>>> (in fact the latest...). I am seeing cpu spiking to 100% randomly on the
>>> first pass of an x264 encode. When this happens the counters freeze and
>>> ...
>>> libs (see below). The ONLY thing that is different is a newer version of
>>> the ffmpeg tree (git-7f83db3 vs git-f514695)
>>
>> Might be interesting to do a git-bisect on it and see which commit is
>> the culprit.  Looks like there were a number of changes...
>> https://github.com/FFmpeg/FFmpeg/compare/f514695...7f83db3
> 
> There is a problem with this approach. You will have to compile FFmpeg
> yourself to know if the spikes still occur. 

Yes - not ideal. I'm just setting up an environment now.... I think I
have a mingw crosscompile setup now (as of 1:30 am this morning at
least). Onto getting the libraries cross compiled.

> I had also seen those spikes
> https://sessiondatabase.net/jwplayer/double5.gif but could not reproduce
> them with my own build of the same core sources (+x264 +libmp3lame
> +libfaac). The only real way to do a bisect is to make an exact copy of
> Zeranoe's build environment. I do not know if he is cross compiling on
> *nix or using MingW onder Windows. But he is using GCC 4.7.1 right now,
> which is not available as Windows binary yet as far as I know.
> 

He's cross compiling on Ubuntu (I think). Which is OK as I have a few
Ubuntu machines here. It's just a pain to get everything setup. Last
time I tried it (a couple of years ago) I had problems with the
threading package not working correctly. We shall see, but if others
have not really noticed the same spikes (I see it on first pass x264
video encodes only) and they have been there since January, someone has
to do the debug....

> Jan
> 

Andy