Experimental Physics and
Industrial Control System

"Pearson, Matthew R." <[email protected]> · Thu, 15 Aug 2013 11:29:02 -0400

Hi,

The sseq record supports put_notify:

http://www.aps.anl.gov/bcda/synApps/std/sseqRecord.html

I've used this in the past to replace seq record instances.

Cheers,
Matt

On Aug 15, 2013, at 11:18 AM, Dirk Zimoch <[email protected]> wrote:

> Hello all,
> 
> Is it possible to make a database link wait for completion?
> If not, is this worth to be considered for a future EPICS version?
> 
> Example:
> 
> A seq record S triggers two records, A and B. A has an asynchronous 
> device support (or is the beginning of a processing chain that contains 
> an asynchronous device support).
> 
> record (seq, "S")
> {
>    field (LNK1, "A.PROC PP")
>    field (LNK2, "B.PROC PP")
> }
> 
> record (ai, "A")
> {
>    field (DTYP, "async dev sup")
>    field (INP,  "...")
> }
> 
> record (ao, "B")
> {
>    field (OMSL, "closed_loop")
>    field (DOL,  "A")
>    field (OUT,  "...")
> }
> 
> When I do a caput with callback to S, the call only returns after A and 
> B have both processed, as expected.
> 
> caput -w 10 -c S 1
> 
> However, B starts processing before A has finished. That is because A is 
> asynchronous. Thus A has only just started processing when B is 
> triggered. This is bad if B (or something down its processing chain) 
> depends on A because B gets an old value.
> 
> Now it would be nice to have "put notify" links, e.g:
> 
>    field (LNK1, "A.PROC PN")
> 
> So that LNK2 is only processed when the whole LNK1 chain has finished 
> processing.
> 
> The same syntax and mechanism may be useful for input links too in order 
> to avoid getting old values from asynchronous device supports.
> 
> Any comments?
> 
> Does something like this already exists or is it already planned?
> 
> I think with dbPutNotify the necessary infrastructure is already there.
> 
> Dirk
> 
> 

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System