On Tue, Feb 28, 2012 at 02:09:23PM +0100, Christophe Gisquet wrote: > 2012/2/28 Mashiat Sarker Shakkhar <[email protected]>: > > https://github.com/shahriman/libav/tree/wmall > > I was in the process of profiling it. A very quick analysis (I'm > sparing you the details) show that > first code blocks of lms_update and lms_predict account for more than > 60% of the decoding time, while the rest is scattered without clear > candidates.
Yes, looks so. also lms_predict is really straightforward - you just obtain a scalar product and shift it, lms_update() is a bit harder but also quite trivial > Most operations seems to involve int16_t, so dspizing should not be > very difficult (famous last words). that's because it's 16-bit only for now :P in reality even reference decoder uses two sets of functions - with 16- and 32-bit precision (which of course can be DSPised). _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
