Experimental Physics and
Industrial Control System

Andrew Johnson <[email protected]> · Thu, 12 Mar 1998 14:54:27 +0000

Benjamin Franksen wrote:
> 
> (64 bits). Is it _guaranteed_ that
> 
>         *((double *)(*private->ppvar) + private->index) = pao->val;
> 
> (such as happens to be in devAoSymb.c) compiles into non-interruptable
> machine code? Is it even guaranteed for the simpler line
> 
>         vxvar = pao->val;
> 
> where vxvar is a globally visible variable of type double?

It should make no difference how the lvalue is derived, nor whether the
variable is being read from or written to.  The answer depends on what
the target CPU is.  If you're running on something without a FP
coprocessor then it's almost certainly not going to happen in a single
indivisible read as you're reading 8 bytes = 2 long words.  With a
coprocessor you have a fighting chance, if the compiler codes the access
as a coprocessor instruction rather than a CPU one.  However I agree
that this shouldn't really be left to chance, and that the intLock is
worth adding for ai/ao, wf and strings.

Of course even an intLock won't solve this problem if the other access
comes from a different CPU in the VME rack...

-- 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