Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Fri, 13 Mar 1998 10:17:49 +0000

Jeff Hill wrote:
> 
> I still worry that an exception might occur while intLock() is applied,
> but perhaps this risk is worth taking in order to allow unconstrained synchronized
> access to globally known variables.

It may be possible to find a solution which doesn't use an intLock, at
the expense of a tiny bit of performance.  If the device support were to
read the value twice and compare the two, it could be fairly confident
that nothing else had changed the variable in the meantime (the chances
of both reads having been split and both giving the same wrong value are
likely to be pretty low, and it can only happen when the write rate is
somewhat higher than the time between reads).  The problem arises with
what to do if the two copies are different - read again and compare with
the later of the first two, but if these are different *again* then how
long do we carry on trying?  Perhaps raise an alarm if the second
comparison fails.

-- Andrew
   ___
 .'   `.	Andrew Johnson, Head of Electronics
/ Royal )	Royal Greenwich Observatory
\ Greenwich	Madingley Road, Cambridge, CB3 0EZ
| Observatory	Tel: +44 1223 374823   Fax: 374700
+----------	WWW: http://www.ast.cam.ac.uk/~anj

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System