Experimental Physics and
Industrial Control System

[email protected] (Ralph Lange) · Tue, 18 Aug 1998 11:15:56 +0200 (METDST)

> I am not sure what is being asked. Is this it?
> 
> The change will be made to mbbiRecord, mbbiDirectRecord, mbboRecord, and
> mbboDirectRecord.
> 
> For each, SPC_NOMOD will no longer be specified for SHFT.

This will enable Raw Soft Supported records pointing to arbitrary bit
subsets of other records.

> For all existing mbbi, mbbo, mbbiDirect, mbboDirect device support in
> base statements like the following
> 
>  pmbbo->shft = pvmeio->signal;
> 
> will be changed to
> 
>   if(pmbbo->shft==0) pmbbo->shft = pvmeio->signal;
> 
> 1) Do we really want to do the later?

I suggest not doing this change:

If SHFT depends on the INP/OUT address field, Device Support should set
it. Not setting it when SHIFT <> 0 will allow rubbish in the database and
lead to errors that are hard to track down. Setting SHFT for Raw Soft
Support will be allowed anyway, since Raw Soft Support won't override it.

I guess overwriting SHFT unconditionally will minimize the change in the
system's behaviour.

> 2) What about all the support not in base?

Doesn't have to be changed, then. (I like that aspect...)

> 3) Where will this be documented?

Release Notes plus Record Reference Manual. Since this only affects Raw
Soft Support, I would regard it a minor change.

> This also does not seem to address Bob Dalesio's concern about what
> happens when links are modified dynamically.

Bob wants to modify hardware links dynamically. This is an entirely new
feature, which is strongly device (and Device Support) dependent and will
most likely only be available with certain Device Supports for a long, long
time.

The proper way to do this would be having the INP/OUT/NOBT fields SPC_MOD
(for hardware related records) and the Device Support recalculate SHFT and
MASK whenever these fields are changed. This would imply major changes to
Record and Device Support. Bob's approach of recalculating SHFT and MASK on
each process() may be sub-optimal in terms of performance, but ... well,
sounds reasonable until this feature is to be implemented generally.

> For bi, bo records signal is used to select bit in register. For
> mbbi,mbbi it was used to select low order bit in register. Do we really
> want to change this convention?

Not really!

Ralph

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System