Re: Are comedi commands broken for ni_pcimio?

Hey Frank,

Nevermind my apparent troubles with comedi commands on the ni_pcimio 
driver.  I realized that the problem was due to bugs in my code.  The 
ni_pcimio driver is fine.

My bug was related to the fact that the driver sometimes gives you more 
than one scan (but less than two) when it calls back your COMEDI_CB_EOS 
callback.  The bug I had was I was reading all the data available in the 
async buffer -- which caused scans to become desynchronized.

Note to self:  Always read data off the async buffer in multiples of the 
scansize otherwise your scans can get totally desyncrhonized!!

-Calin


On Thu, 29 Dec 2005, Calin A. Culianu wrote:

>
>
> On Wed, 28 Dec 2005, Frank Mori Hess wrote:
>
>>
>>  You can try the cvs, I don't think I've broken e-series support.  And if I
>>  have, it might as well be you who tells me, right? :)  It's possible the
>>  cvs will work better, as some of the code for querying the progress of the
>>  dma transfers has been refined.  I haven't looked into your specific
>>  problem yet though (maybe this weekend).
>> 
>
> Well CVS E-series support isn't any more 'broken' than in the stable release, 
> so that's good.  :)
>
> However my problem persists:  namely in comedi asynch acquisition with more 
> than 1 channel, the channels get garbled/mangled in the channel array: so for 
> instance channel 0 isn't always in the sample place, etc.
>
> Try it with 16 channels at once.  It doesn't always happen with low channel 
> counts.
>
> Anyway, if you can look at this at any time, I would be very grateful...
>
> Thanks!
>
> -Calin
>
> _______________________________________________
> comedi mailing list
> comedi_at_comedi.org
> https://cvs.comedi.org/cgi-bin/mailman/listinfo/comedi
>
>

Received on 2005-12-30Z18:23:05