Experimental Physics and
Industrial Control System

[email protected] · Thu, 21 Nov 96 15:47:25 -0500

RE: Hardware configuration/initialization records

I'm not sure that there is a well thought out (or described) goal here.

Suppose in the worst case that there is a system in which there are
configuration options down to the signal level. I for one would not want
to have a hw configuration record for every I/O point -- especially if
it inherits from dbCommon!  That type of config information should be put
into the AI or AO etc records, in a parm field or elsewhere in the 
INP field.

Perhaps allowing the INP field to be parsed according to a description in
a .ascii file for that DTYP is a reasonable idea. This would require the
least amount of memory, and support smart configuration tools.  It would
require a function to pack and unpack the structure by element name.

Proof of principle: cdevData can pack and unpack any data type by element
name, effectively holding an arbitrary "structure".  If CA were extended
to transport such a data object, then the "subfields" would be visible
to any (enhanced) tool.  Further, this extension of CA would improve
the performance of cdev over CA.

I would not want to support adding hardware records unless we change
EPICS to support records which do not inherit from dbCommon.  This does not
imply data only records, because they could still have a "process" method.

Chip

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System