[XviD-devel] Question about bvop decoding
ed.gomez at free.fr
Sun Jul 18 20:21:58 CEST 2004
I'm still looking at the decoder performance, and i must admit
that profiling is now pretty useless to find candidates for
optimizations. However something did get my attention...
First i must say i'm not a compensation expert, but it seems
that for bvops interpolate and direct blocks, we do spend an
incredible amount of time in decoder_bf_interpolate_mbinter
(something like 14% of cpu time) though it's just a function
that calls more CPU intensive functions like qpel/halfpel
interpolation functions and so on.
So my question is:
Are we doing something extremly silly that no one did notice
yet (like interpolating things more than once without need,
more or less like we were edging frames more than once) ?
Or is this function just as fast as it can be... and to
optimize xvid decoder, i should start doing more hard stuff
like merging some operations together ?
More information about the XviD-devel