Re: [fluid-dev] Multi-kernel system test

[Bernd Casper]

Hi David,

this is quite interesting.

 

In a real-time scenario using two kernels *slows down* latency and overall performance here, when using many programs simultaneously, while seemingly increases polyphony performance at te same moment (personal impression of my ears - stereo horizon is enlarged). I use buffers/size/polyphony 6/384/1024, and 8 instances of FS.

 

Is this what's intended?

 

To me it's very important not to use a QuadCore high-end computer system - on such a system the kernel settings would not affect latency, I assume. I use a Dual-Core mid-size system ("average").

 

Regards

Bernd.

----- Folgende Nachricht wurde empfangen -----

Absender: David Henningsson

Empfänger: fluid-dev

Zeit: 2010-09-30, 08:12:22

Betreff: Re: [fluid-dev] Multi-kernel system test

On 2010-09-27 18:28, Bernd Casper wrote:
> Hi,
> I'm testing the activation of more than one processor kernels, at a
> Dual-Core system.
> Since I experience nothing when changing the number of kernels, could
> you please point me to what kind of improvement I should expect, by
> doing this.
> Many thanks
> Bernd.
>

I hope the gain will be less CPU usage if 1) you have a lot of voices
and 2) you have big latency/buffer settings. How much "a lot" and "big"
actually means would depend on your hardware. I've tested it with
fast-render mostly, where I had one typical midi file taking 15 s to
render with one core and 10 s to render with two cores. That was with -z
8192 (if I ran with -z 64, there was no improvement at all).

// David