[MPlayer-dev-eng] Optimizations for Cell Broadband Engine
Reimar.Doeffinger at stud.uni-karlsruhe.de
Mon May 7 17:47:39 CEST 2007
On Mon, May 07, 2007 at 02:30:30PM +0200, Guillaume Poirier wrote:
> There's been several people on this mailing list wondering if there
> were any plans to write optimizations for Cell Broadband Engine, with
> not much luck since now.
I think the ones trying so far have been concentrating on optimizing
H.264 with focus on getting the best possible performance which needs a
really major overhaul of things I fear.
> A few days ago, I've been contacted by a person working at IBM who's
> offering to give one PS3 (maybe more than one, I don't know) to
> someone who would be interested to write such optimizations.
I am interested, though I have no experience with that kind of thing and
I haven't looked into the details of how the cell architecture works (I
know the things you can find in any introduction about it, but that is
not enough by far to know how to optimize things best).
And since my time is limited I don't think I would do any optimizations
that need a lot of code wrangling in the near future (e.g. H.264 above
simple IDCT level).
While I can give other developers ssh access, my internet connection is
rather slow (even more so when I am using it *g*), so I doubt that would
be much fun.
On the plus side though, having access to that kind of hardware would at
least make the PowerPC/bigendian bugfixes a bit simpler - I currently
use tick.rz.uni-augsburg.de for that, but sometimes you need to actually
see things, and X forwarding is really a slow pain and in addition
MPlayer is usually broken when you use a little-endian X server with a
bigendian MPlayer build...
More information about the MPlayer-dev-eng