[LAU] Re: [Jackit-devel] Fwd: Hardware Choices

[Stéphane Letz]

>
> (CC-ed to jack-audio-dev list)
>
>
> "Thomas Vecchione":
>>
>>                   Seablade
>>
>> On 7/20/07, Kjetil S. Matheussen <k.s.matheussen at notam02.no> wrote:
>>>
>>>
>>> "Thomas Vecchione" <seablaede at gmail.com>
>>>>
>>>> I heard Jack (jackdmp?) could only take advantage of two cores  
>>>> right
>>> now?
>>>
>>> No.
>>>
>> Mind being a bit more specific?  As in no it will handle 4,8, or  
>> 16 cores
>> well?
>>
>
> Sorry, that was a short rude answer, especially since I'm not 100%  
> sure of
> the answer either. But I would be surprised if jackdmp was limited  
> to two,
> or any other high-value fixed number of, parallel sound processing
> threads.
>
> And on the other hand, if you are wrong, which I think you are,  
> although
> I'm not 100% sure about that, you could be responsible for
> continuing to spread a(nother) misunderstanding about jackdmp. (The
> previous one was that jackdmp could not work on single processor  
> machines)
>
>


jackdmp can handle *any* number of cores. What jackdmp does in  
changing the jack clients scheduling model from a purely sequential  
one to a "data-flow" like one: as soon a jack clients become  
runnable, they are activated by the OS which is then responsible to  
run them on any available processor. A Jack client become runnable  
when all of it's input clients have been run, and this of course  
depends of the graph topology.

Stephane