From: Marc Mendez (mmendez++at++silicon-worlds.fr)
Date: 04/03/2003 03:35:42
pfHello,
so maybe noone has faced that problem ?
What does the performer team thinks about that ?
Regards,
Marc.
Marc Mendez wrote:
> pfHello !
>
> Hope everyone is doing fine !
>
> I m working on an abstract, really object-prone, paradigmful
> application based on engines and fluxes and that s all 8)
>
> I ve just run some tests where I write an engine that has an input:
> pfGetFrameTimeFlux() and one output, the time (it just copies the
> value from one flux to another).
> I ve put, for warming up my CPUs, 200 engines chained together. But
> nothing was displayed.
> Then I ve decreased I ve decreased I ve decreased the engine number to
> 10 to see my penguin again ...
>
> Trying this test with 200 engines all connected to the
> pfGetFrameTimeFlux() as an input (in parallel, not in chain) doesn t
> slow the application. Only engine chaining slows everything a lot.
>
> As I d like to build some complex simulations data path where ten or
> twenty engines may be involved, I d like to get more performances here !
> Is this normal ? Is there something to do to get more perfs ? Should I
> code in assembler ?
>
> Thanks for your help/ideas/remarks,
> Marc.
> PS: I almost forgot to tell you that I ve tested with my C__ design
> patterns-full application **AND** pfFluxes and pfEngines directly.
> I may send you the PF sample code based on simple.C
>
-- Marc Mendez Silicon Worlds SA12, rue de Chatillon 75014 Paris Tel : +33 (01) 53 90 11 13 Fax : +33 (01) 53 90 11 12
Without C, we would only have Basi, Pasal and Obol. Anonymous. Computers are useless. They can only give you answers. P. Picasso.
This archive was generated by hypermail 2b29 : Thu Apr 03 2003 - 03:36:16 PST