[mythtv-users] Re: .18 CPU optimizations

[myth joe]

On 5/9/05, Hamish Moffatt <hamish at cloud.net.au> wrote:
> On Mon, May 02, 2005 at 11:07:29PM -0700, myth joe wrote:
> > Nonetheless, the point stands that benchmarks are needed.  What are
> > the bottlenecks in MythTV?  Let's rule out MPEG2 recording and
> > playback since hardware assist exists for those.  That leaves MPEG4
> > encoding and playback and commflagging.
> 
> What hardware decode assistance are you referring to for MPEG-2?
> 
> PVR-350: seems to be particular about the MPEG format, and if you don't
> need analog capture (eg a DVB/HDTV only system) you wouldn't have one.
> 
> XvMC: there seems to be plenty of trouble with this still and it's
> vendor-specific in parts too. Perhaps I'm overstating the situation.

Yes, I was referring to any kind of hardware acceleration.  My
immediate context was an Epia board, which would have VIA's flavor of
XvMC.

However, even assuming no hardware acceleration for MPEG2, I would
assume that MPEG4 results would correlate with MPEG2 results.  It's an
imperfect, yet nonetheless simplifying, assumption.

Anyway, what's a good benchmark?  Ripping a DVD seems like a
reasonable task for benchmarking the combination of raw MPEG2 decoding
and MPEG4 encoding performance, and we could probably agree on a
specific DVD.  How to benchmark playback, though?

JB