[Mplayer-dvb] This might be a it off topic but ..

[Arpi]

Hi,

> Already 11% that's pretty much already :) hehehe
> can you send me the files(or patches) you changed to make it go 11% faster
> or tell me where the memcpy's are so i could try eliminating them too ..
> want to try and see if that makes the movie play someone a bit smootly

i don't have patches yet... just measuring cpu usage of things...

now i've finished:

16%  divx decoding

11%  2 x memcpy
 3%  mpeg headers, getpixel
 9%  dct
 1%  quant preparation
17%  quantization (should be rewritten in MMX)
11%  huffman coding
16%  ioctl (sending to DVB card)

it will be hard to pull down drastically.
memcpy - can be eliminated by API changes (direct rendering)
dct - already optimized
quantization - can be mmx optimized, easy work
huffman coding - optimize tables, won't help too much
ioctl - can't change, it depends on card/bus speed (maybe lower if
   we increase quant scale, but it will result worse quality)

ideally we can eliminate about 25-30 %


A'rpi / Astral & ESP-team

--
mailto:arpi at thot.banki.hu
http://esp-team.scene.hu