scaler record: int -> float

[Tim Mooney <mooney@aps.anl.gov>]

Would anyone object to the EPICS scaler record changing to use floats
instead of ints for counter and preset values?  The change would allow
the record to support 64-bit counters with minimal modification of the
scaler record, and most clients I'm aware of would not even notice the
difference.

There is the issue of precision: a float has 23 bits of precision, so
the smallest possible difference in two scaler values is roughly one
part in 10^7.  Since scaler time bases I'm aware of are accurate only
to a few parts in 10^5, it seems that we'd lose nothing essential by
going to floats.

Tim Mooney (mooney@aps.anl.gov) (630)252-5417
Beamline Controls & Data Acquisition Group
Advanced Photon Source, Argonne National Lab