Experimental Physics and
Industrial Control System

Jeff Hill <[email protected]> · Wed, 23 Feb 2005 10:40:01 -0700

> I was only talking about record field and corresponding 
> CA data types.

A common design goal is to not allow the details of internal data
structures to be exposed by externally visible interfaces. Among
other benefits, this prevents changes in the internal
implementation from impacting clients of the interface (sorry to
restate motherhood and apple pie). However, with a large
distributed system such as EPICS it is quite convenient at times
to peek under the hood at the data structures of the internal
implementation using channel access for debugging purposes.
Therefore, from my perspective, support for unsigned types used
in the internal implementation can also be of value.

Jeff

> -----Original Message-----
> From: Benjamin Franksen [mailto:[email protected]]
> Sent: Wednesday, February 23, 2005 9:10 AM
> To: Jeff Hill
> Cc: 'Marty Kraimer'; 'Andrew Johnson'; 'Bob Dalesio'; 'Ralph
> Lange'; 'Eric Norum'; 'Ken Evans'; 'Ned Arnold'; 'Matej
> Sekoranja'
> Subject: Re: EPICS base V4: iocCore database
> 
> On Wednesday 23 February 2005 16:48, Jeff Hill wrote:
> > > The bottom line is that unsigned types should be used only
> in
> > > the very few cases where they are essential to
> > > faithfully represent raw hardware values.
> >
> > I find that unsigned types are quite useful for situations
> where
> > the operand must never be negative. For example, with array
> > indexes this improves performance and simplifies the code
> when
> > range checking because only one side of the range check need
> be
> > done. Of course, I can't expect that everyone will agree :-)
> 
> I thought it was obvious from the context (but maybe I should
> have stated it
> explicitly) that I was only talking about record field and
> corresponding CA
> data types, not programming in general. Record field values are
> rarely used
> directly as indices into arrays and even if they would be, the
> overhead
> involved in using them would be orders of magnitude larger than
> any overhead
> due to additional range checks.
> 
> Ben

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System