Experimental Physics and
Industrial Control System

Benjamin Franksen <[email protected]> · Sat, 25 Feb 2006 20:00:44 +0100

On Friday 24 February 2006 23:22, Steve Lewis wrote:
> >Of course, Erlang's runtime system is written in (fine-tuned,
> >hand-optimized) C...
>
> I have always considered EPICS record support, channel access, and
> the related code to /be/ a "fine-tuned, hand-optimized run-time
> system."  That's where the hard work has gone.  

Yes, of course. That's why I made this last remark ;-)

I am very fond of the high code equality in the EPICS base.

> Other than the 
> process routines, there is not much other code needed.

Not much, but still.

Often you have to write additional code, for instance in SNL. And you 
can't really argue that SNL is a high-level language. It is more or 
less a downgraded C plus a simplified CA plus state machine semantics. 
No abstraction facilities to speak of. It is barely adequate for simple 
jobs and gets extremely annoting if you want to do complex things.

Also, record support is a bit more than just a linear process() routine 
(init_record, special, cvt_dbaddr, special code for asynch processing, 
especially in conjunction with bumpless reboot).

> We already have what Erlang--or any other language+runtime with the
> same qualities--would bring to the table.

We have much of it, agreed. But there are also things missing. See the 
wish-list for EPICS V4. Off the top of my head:

- online add/delete of records
- unlimited data sizes (strings, arrays)
- atomic read/write of structured data
- asynchronous link processing
- replication

And from my personal wish list

- hot-swap of code
- exception handling / supervisor processes
- automatic memory management
- decent high-level processing language

It's not that there isn't anything to do to make EPICS even better than 
it is...

Ben

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System