[vorbis-dev] vorbis on playstation 2
Nicolas Pitre
nico at cam.org
Wed Feb 5 16:53:24 PST 2003
On Wed, 5 Feb 2003, Brad BARCLAY wrote:
> The "ogg123" process, as compiled for the PS2 Linux kit, in
> testing took about 90% of the core MIPs processors processing time --
> so the direct recompile is hardly lightweight. Obviously, using the
> vector units will present a major performence enhancement opportunity.
>
> Of that ~90%, 14% was being used by _fpadd_parts. Alot of time
> appears to be wasted while doing floating point arithmetic, probably
> due to the fact that Vorbis uses lots of double-precision floats, which
> are emulated on the PS2.
>
> Of the actual Vorbis sources, 12% of the processing time is
> spent in mdct_backward (mdct.c). the next most used is 9% by
> vorbis_book_decodevv_add (codebook.c), and 4% is used by ov_read
> (vorbisfile.c). Of these, the first two are probably good cantidates
> to porting to vector assembly.
>
> Cache miss is also fairly high -- nearly 50%. Again,
> mdct_backward appears to be the biggest culprit.
It would be interesting if you could do the same analysis with Tremor.
I've made many cache improvements in Tremor's mdct_backward some time ago.
Cache usage clearly improved on an ARM processor compared to the earlier
version of the same function, but I don't know how it compares to the
libvorbis version.
<p>Nicolas
--- >8 ----
List archives: http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-dev-request at xiph.org'
containing only the word 'unsubscribe' in the body. No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.
More information about the Vorbis-dev
mailing list