Experimental Physics and
Industrial Control System

"Kalantari Babak" <[email protected]> · Thu, 16 Jun 2011 22:50:52 +0200

I also think having an standard API in devLib is indeed necessary to allow easy usage of external codes/drivers which depend on DMA. And yes we have our own wheel but I should confess that I did not look much around at the time when I started design and implementation. The not-fully-complete documentation of my code can be found here: http://controls.web.psi.ch/cgi-bin/twiki/view/Main/VxWorksDmaDriver#API_and_Usage <http://controls.web.psi.ch/cgi-bin/twiki/view/Main/VxWorksDmaDriver#API_and_Usage> . It basically supports Universe-2 and Tempe for vxWorks.

I can offer some help maybe not in every aspect of this issue but at least and especially I would like to participate in details of the design and behaviour of API and/or code as we have quite some running applications already depending on (our) DMA driver.    

Regards,

Babak

________________________________

From: [email protected] on behalf of Andrew Johnson
Sent: Thu 6/16/2011 7:20 PM
To: EPICS tech-talk
Subject: Re: FW: EPICS driver for SIS3302

There is obviously a need for a VME DMA API in devLib.  APS has been using one
I designed here for many years, and there are tech-talk messages archiving my
implementations of this for vxWorks for all 3 of the VME interface chips we
use here (VMEChip2, Universe-2, Tempe).  Recently Mark Rivers started using it
for some of his code and now Diamond have looked at it.  I believe that Till
has a slightly different API for RTEMS with implementations for the Universe-2
and Tempe chips, and PSI also wrote their own but I don't know what chips or
OSs that runs on.

I've already suggested that Michael Davidsaver add a pluggable DMA API to
devLib so we can stop reinventing this particular wheel, but I know he's busy
with work on NSLS-II so he might appreciate some help.  If anyone discusses
this topic in the future, /please/ use tech-talk (or core-talk) so others can
also join in.

- Andrew

On 2011-06-16 Ernest Williams wrote:
> From: Williams Jr., Ernest L.
> Sent: Wednesday, June 15, 2011 10:39 PM
> To: Korhonen Timo; Keahey, Pelham; Hu, Yong
> Cc: Szalata, Zenon M.; Straumann, Till; Kim, Kukhee; [email protected];
>  Allison, Stephanie; [email protected]; Williams Jr., Ernest L. Subject:
>  RE: FW: EPICS driver for SIS3302
>
> Hi,
>
> This sounds like a great codeathon topic.
>
> Till are you interested in this as well?
>
> Can we get this kind of thing in devLib2 --> libCom ?
> Michael, what do you think?
>
>
>
> Cheers,
> Ernest
> ________________________________________
> From: Korhonen Timo [[email protected]]
> Sent: Wednesday, June 15, 2011 9:47 PM
> To: Keahey, Pelham; Williams Jr., Ernest L.; Hu, Yong
> Cc: Szalata, Zenon M.
> Subject: RE: FW: EPICS driver for SIS3302
>
> Hi all,
>
> sorry for the late reply.
> The complicated part of our driver is that we make heavy use of DMA, and
> we have some libraries that are on top of the vxWorks BSPs. So it
>  unfortunately will need some work to get it to run on RTEMS; libCom did
>  not have this support (as far as I remember). It should, though - but this
>  is another issue.
>
> In any case we would be happy to share the driver. To understand it
>  somebody needs to look at the vxWorks dependent part, to be able to
>  replicate the API functions in RTEMS.
>
> best regards,
>
> Timo
--
Optimization is the process of taking something that works and
replacing it with something that almost works, but costs less.
-- Roger Needham

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System