You are not logged in.

#1 2010-06-16 08:04:47

kokoko3k
Member
Registered: 2008-11-14
Posts: 2,402

new ffmpeg and h264 decoding performance (!)

It's just an impression or the latest mplayer performs incredibly well?
Some months ago i remember i benchmarked a sample 1280x536 h264 video and cpu usage was very near 100% on a little atom n280@1.67Ghz

Using an mplayer compiled against latest ffpeg-mt, today i tried the same sample video on the same cpu and i got this results:

time mplayer 1280x536.mkv -nosound -vo null -benchmark -endpos 60 -lavdopts skiploopfilter=all:threads=2
[..]
BENCHMARKs: VC:  27.969s VO:   0.009s A:   0.000s Sys:   0.390s =   28.369s
BENCHMARK%: VC: 98.5920% VO:  0.0329% A:  0.0000% Sys:  1.3751% = 100.0000%

Exiting... (End of file)

real    0m28.459s
user    0m49.507s
sys     0m0.140s

28 seconds on a 60 seconds stream means less than 50% of cpu usage!
Note that atom n280 is not a multicore cpu, it just provides hyperthreading, launching mplayer with threads=1 increases the benchmark time by about 2 seconds, which is great too and is the result i get with standard archlinux mplayer package.
On a full frame 1280x720 video cpu usage reach about 67%
I think the new ffmpeg now easilly outperforms even coreAVC, i'm very impressed!

Last edited by kokoko3k (2010-06-16 08:22:08)


Help me to improve ssh-rdp !
Retroarch User? Try my koko-aio shader !

Offline

#2 2010-06-16 08:08:35

PirateJonno
Forum Fellow
From: New Zealand
Registered: 2009-04-13
Posts: 372

Re: new ffmpeg and h264 decoding performance (!)

I use VDPAU, so I can't confirm, but I saw this today: http://www.phoronix.com/scan.php?page=n … &px=ODM0MA


"You can watch for your administrator to install the latest kernel with watch uname -r" - From the watch man page

Offline

#3 2010-06-16 08:12:34

kokoko3k
Member
Registered: 2008-11-14
Posts: 2,402

Re: new ffmpeg and h264 decoding performance (!)

Thanks, I missed that article on phoronix,

"The H.264 and Theora decoders are now significantly faster and the Vorbis decoder has seen important updates."

"significantly" is an euphemism smile

-edit-
with skiploopfilter=none:threads=2

BENCHMARKs: VC:  28.048s VO:   0.010s A:   0.000s Sys:   0.385s =   28.442s
BENCHMARK%: VC: 98.6129% VO:  0.0340% A:  0.0000% Sys:  1.3531% = 100.0000%

Exiting... (End of file)

real    0m28.547s
user    0m49.497s
sys     0m0.173s

skiploopfilter=none:threads=1

BENCHMARKs: VC:  33.144s VO:   0.008s A:   0.000s Sys:   0.300s =   33.452s
BENCHMARK%: VC: 99.0797% VO:  0.0242% A:  0.0000% Sys:  0.8962% = 100.0000%

Exiting... (End of file)

real    0m33.540s
user    0m33.131s
sys     0m0.197s

I'm very curious to compare those results again a coreavc enabled mplayer but i haven't the codec. someone?

Last edited by kokoko3k (2010-06-16 08:23:56)


Help me to improve ssh-rdp !
Retroarch User? Try my koko-aio shader !

Offline

Board footer

Powered by FluxBB