[vorbis-dev] ogg vorbis dynamic memory allocation.
segher at chello.nl
Mon Jul 2 00:44:01 PDT 2001
Firelight Multimedia wrote:
> Hi segher,
> Here are some interesting results after some bashing with Michael and
> Jack on irc, and churning some figures out of my memory manager.
> Current is just the amount allocated at the time the song is played..
> maximum at one time is the ceiling amount of any allocs that file did at
> any time. (the more important one)
> codebook.c memusage : current 69184 : maximum at one time 69184
> sharedbook.c memusage : current 510644 : maximum at one time 512248
codebooks use _way_ too much memory, but it's not _that_ easy to get
rid of as Michael suggests...
> vorbisfile.c memusage : current 100 : maximum at one time 100
> block.c memusage : current 102948 : maximum at one time 102948
> framing.c memusage : current 45538 : maximum at one time 74210
Blocking, framing, windowing will be attacked by libogg2, when that
happens... It's 3rd on my current work list, so it should happen
> res0.c memusage : current 31960 : maximum at one time 31960
Residue uses too much.
> floor0.c memusage : current 6176 : maximum at one time 6176
> window.c memusage : current 33792 : maximum at one time 33792
> mdct.c memusage : current 13824 : maximum at one time 13824
> info.c memusage : current 4677 : maximum at one time 4677
> other files memusage : current 6832 : maximum at one time 6832
> michael says "you should be able to kill off 90% of that 500k for decode
Getting it to 100k should be easy, 50k could prove more difficult.
> of files created by the current encoder. Wouldn't be too much work. Ask
> segher, he was going to do it at some point, not sure why he didn't get
> around to it"
The biggie? Really efficient decoding needs a new decoder, not the mixed
encoder/decoder in the reference source. And there are various real-world
reasons why I haven't "just" written a new decoder... Some technical as
well (really should write it tunder for a apecific cpu/os/platform, to get
maximum spped && tightness etc.).
Well anyway, thanks for the report, I'll have a look.
--- >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