[x264-devel] Re: Another CPU related question

[Jelle]

Loren Merritt wrote:
> On Wed, 3 May 2006, Jelle wrote:
> 
>> Loren Merritt wrote:
>>
>>> On Wed, 3 May 2006, Pedro Tumusok wrote:
>>>
>>>> There has been some articles about it being able to decode 48 mpeg2 
>>>> streams at once or something like that, so then to my question, do 
>>>> you belive it would be able to show similar kind of performance in 
>>>> encoding?
>>>
>>>
>>> 48 mpeg2 streams? You don't need a Cell for that, unless they're 
>>> 1080p or something. A dvd takes 2% cpu on my Athlon64 3400, so the 
>>> only thing preventing me from displaying 48 of them is bandwidth to 
>>> the harddrive and video card.
>>
>>
>> That 2% is achieved by offloading a lot of the work to the videocard 
>> (which does the idct and the mc). Unless you have 48 video cards in 
>> your system, it will not be able to decode 48 mpeg streams.
> 
> 
> No, I don't know whether my videocard supports xvmc, but I certainly 
> haven't enabled it in ffmpeg. This is cpu-time as measured by
> `time mplayer -benchmark -vo null -nosound -vc ffmpeg2`.

Ok, I'll run that too then.

Two different machines. I know for sure that neither machines have 
videocards that support Xvmc.

============ Test 1:

Athlon64 3200+ (2GHz/512kb cache)

MPEG2 video 480x576 at 25fps, 2.5Mbit/s (much less than DVD), 4688 seconds 
long.

BENCHMARKs: VC: 612.046s VO:   0.648s A:   0.000s Sys: 127.753s =  740.447s
BENCHMARK%: VC: 82.6589% VO:  0.0875% A:  0.0000% Sys: 17.2535% = 100.0000%

Exiting... (End of file)

real    12m20.753s
user    9m0.006s
sys     0m10.123s

100 * 12*60+20/4688 =~ 16% CPU.

============ Test 2:

P4/3GHz/1M cache

MPEG2 video 720x480 at 29.97fps, 4.5Mbit/s, 1886 seconds long

BENCHMARKs: VC: 224.833s VO:   0.710s A:   0.000s Sys:  24.075s =  249.618s
BENCHMARK%: VC: 90.0707% VO:  0.2846% A:  0.0000% Sys:  9.6446% = 100.0000%

Exiting... (End of file)

real    4m11.951s
user    3m51.214s
sys     0m9.968s

100 * 4*60+12/1886 =~ 13% CPU.

I don't know where you got 2%, but I can't reproduce anything even close 
to it.

2% of a 2.2GHz CPU, or 44MHz, for DVD video? How can I get that?

mpeg2 decoding takes about the same amount of operations per frame as 
JPEG (mjpeg) encoding. If your 2% number is really valid, you should be 
able to transcode mpeg2 to mjpeg 25 times faster than real-time (ramdisk 
to ramdisk if necessary).

I would like that to be possible, but it doens't look like it is.

Jelle.

> 
> --Loren Merritt
> 

-- 
This is the x264-devel mailing-list
To unsubscribe, go to: http://developers.videolan.org/lists.html