[MEncoder-users] Performance on multi core machines - a specific enquiry re avisynth mt input

[Michael Rozdoba]

James Hastings-Trew wrote:
> Michael Rozdoba wrote:

>> Can anyone comment on why avisynth is being restricted to one core 
>> /only/ when it's being used as input to mencoder?
>>
>> I could really use a solution as I've just discovered my alternative 
>> solution - vdub & ffdshow's huffy, is introducing a yellow colour cast 
>> (probably as it requires an additional converttorgb step at the end of 
>> the avisynth script, as vdub will only accept rgb).

Aside: fixed this - it wasn't the conversion at the end of the script to 
rgb (or yuy2 which had the same effect), but the fact that within vdub I 
had huffyuv configured to compress in yv12 space; that conversion was 
producing the cast. Changing the script to output yuy2 (it had been 
working in yv12 which vdub wouldn't accept) & configuring huffyuv to 
compress in yuy2 space removed the colour cast. Not too surprising, as 
this eliminated the one down conversion that was occurring.

>> If no one here knows, any suggestions where I could ask? The issue does 
>> seem to be specific to mencoder.
>>
>>   
> I think you are missing the point... you don't need to do the 
> intermediate file, avisynth script, or anything like that. Just do it 
> all in MEncoder. All that other software and extra steps are 
> unnecessary, and only making the process slower in the long run than 
> worrying about processing the image on each pass.

Quite possibly but there are several issues involved here. I'll list the 
most significant to me first:

It will take me a long time to get to know MEncoder allowing me to get 
what I want out of it; this job requires a solution now.

I don't like loose ends. Even if I intended switching right now & knew 
how to get better results with MEncoder, I'd still want to solve my 
current problem.

Regardless of how many threads I configure MEncoder to use, it still 
won't run above 25% cpu, strongly suggesting its only using one core 
(despite having an affinity with all four), so that problem would need 
solving in any case.

-- 
Michael Rozdoba