[vorbis-dev] vorbis SSE diff

[Felix von Leitner]

I just uploaded a diff I made for libvorbis to

  http://www.fefe.de/diffs/libvorbis-1.0-simd.diff.bz2

It adds a 3dnow! decoding speedup (about 10%), SSE decoding speedup
(about 20%) and SSE encoding speedup (about 25%).

The code selection happens purely at compile time, no CPU flag detection
or so.  I figure someone else will have to do that, I'm short on time.
To activate 3dnow!, add -Dsimd_3dn to CFLAGS.  To activate SSE, compile
with gcc 3.2.2 (others have had alignment problems), compile with
optimizer enabled, and add something like -march=pentium3 -msse to
CFLAGS.

I probably haven't even remotely touched all the parts that would be
beneficial to speed up, but I don't even know how vorbis works, I'm just
some comp.arch guy. ;)  However, I use vorbis to encode the sound track
to recorded TV episodes and encoding time for a 45 min episode is now
down to 5 minutes on my Athlon XP 2000+.  Vorbis already was
impressively fast; now it's on par with lame using --vbr-new.

Hope this helps,

Felix

PS: I'm not on the list, please Cc: me and send questions directly to
me.
--- >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.