Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Thu, 14 Jan 2016 12:29:14 -0600

On 01/14/2016 11:59 AM, Lingyun Yang wrote:
> This would be the type of solution that I am looking for. But I am not
> quite familiar with the device support yet.

To be strictly correct, this is not device support.

> What I am doing now is define my data analysis function, register and
> put it in dbd file.
> Is this dbPutLink called in my data analysis function ? I guess not.
> Or I duplicate the device support work done by src/rec/aSubRecord.c or
> even the dbd file, but change its process to update dbPutLink according
> to my mask ?

You put the necessary dbPutLink() calls in your aSubRecord subroutine,
the one whose name appears in the SNAM field of the record, so yes
that's in your data analysis function. You should not change the record
type code at all; by returning a non-zero value from your analysis
function you will be telling the record not to do the puts, because your
code has already done them.

Now the code in the record itself has been optimized to put all the
output links at once in a single loop:

>         for (i = 0; i < NUM_ARGS; i++)
>             dbPutLink(&(&prec->outa)[i], (&prec->ftva)[i],
>                 (&prec->vala)[i], (&prec->neva)[i]);

In your case though you may want to do each put separately when you have
finished calculating each output field's value. The result might look
something like this:

    /* Calculate vala ... */

    dbPutLink(&prec->outa, prec->ftva, prec->vala, prec->neva);

    /* Calculate valb and valc ... */

    if (doAction) {
        dbPutLink(&prec->outb, prec->ftvb, prec->valb, prec->nevb);
        dbPutLink(&prec->outc, prec->ftvc, prec->valc, prec->nevd);
    }

    /* Calculate vald ... */

    dbPutLink(&prec->outd, prec->ftvd, prec->vald, prec->nevc);

    return -1; /* Don't put output links */

- Andrew

-- 
There are only two hard problems in distributed systems:
  2. Exactly-once delivery
  1. Guaranteed order of messages
  2. Exactly-once delivery
 -- Mathias Verraes

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System