Experimental Physics and
Industrial Control System

[email protected] (Ralph Lange) · Tue, 12 Nov 1996 17:27:15 +0100 (MET)

Hi there.

I mailed this comment to Marty and he encouraged me to post it to tech-talk
to have the discussion continue there, so - here it is:

First of all - You're right in that there's no basic disagreement from our
side:

Full agreement on the statement that to get all the functionality these
hardware records will have to be almost full size records - it's the prize
we will have to pay. (Maybe improves a lot of device supporter's under-
standing of record support ...?)

Full agreement on the idea to introduce support levels to change things
smoothly and only as much as needed.

Just to prove we're not fully asleep (yet) some additional thought:

----

I find underlying field buses quite a useful thing to configure using
hardware records (we almost CAN see it coming). Especially with on-line add
and delete adding new fieldbus nodes without rebooting the IOC would be
possible. But, in this case, there must be a way to have some kind of
hierarchical organized hardware records. E.g.:

One IOC has several VME cards (with card specific configuration like
interrupt level, card address) serving two field bus segments each (with
segment specific configuration info like segment speed).  On each field bus
segment there's an arbitrary number of devices (specific configuration info
like node number, device name, type). Each device may have an arbitrary
number of accessible variables (with a certain data type, access type, id,
name each).

This example isn't over-generalized, it reflects exactly the structure of
our CAN subsystem. Right now, most of this information is squeezed into a
most cryptical INST_IO link. We would prefer having a symbolic variable
name in the link and resolving that into a hardware record with all the
configuration info in it.

So we get one hardware record type for the device support itself, one with
an instance for every VME card, one with an instance for every field bus
segment, one with an instance for every node and one with an instance for
every variable. All these should be level 2 (or level 1 at least).

Is this generally too much? How could such hierarchical records be linked
together in a way that device supports can (simply?) traverse them and
initialize all the needed things in the driver?

----

Is it really possible to allow CA links in hardware records? Doesn't that
cause problems within the IOC boot sequence, i.e. is it allowed to make CA
calls during iocInit?

----

What do you need processing hardware records for? (Just give an idea, I
just can't think of a useful application...)

----

So much for today...

Ralph
-- 
      __  Ralph Lange                         Email:       [email protected] 
     /\ \                                     WWW: http://www.bessy.de/~lange
    /  \ \  BESSY II                          
   / /\ \ \  Berliner Elektronenspeicherring- Snail:                 BESSY II
  / / /\ \ \  Gesellschaft fuer Synchrotron-               Rudower Chaussee 5
 / / /__\_\ \  strahlung m.b.H.                       D-12489 Berlin, Germany
/ / /________\                                Phone:         +49 30 6392-4862
\/___________/ Control System Group           Fax:                ...   -4859

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System