Re: MCC PCI-DAS1602/16

On Sat, 01.11.03, John Conner <conner_at_empiredi.com> wrote:

> You would think that a 512 sample fifo would be large enough
> to handle 64000 samples per second. It takes 8 ms to fill the
> fifo. Even with the puny 450MHz CPU we really shouldn't be
> having problems. I'll try Monday or sooner to run the same 
> test from the same disk in a 2GHz machine.

While the 450MHz CPU should be fast enough for a throughput of
64kSamples/sec it is still very possible that your latency is higher than 4
(or even 8) ms. A vanilla linux kernel is not a realtime kernel.

However, you can try a few things:

Set semi-realtime priority (SCHED_FIFO) using sched_setscheduler()
Turn off paging to disk (swap).
Applying low-latency kernel patch

You might want to take a look at
http://www.zip.com.au/~akpm/linux/schedlat.html

hope this helps,

Nils

Received on 2003-11-03Z09:58:12