PowerPC floating point exceptions

[Peter Dufault]

I noticed something funny working with the MVME5500 and RTEMS 4.7 last 
week.

I'm doing a control application.  The analog input is double-buffered, 
and detects an overrun if I don't get the data off of it and re-enable 
the first buffer before the second buffer fills.

I noticed that when I used garbage memory for doing the floating point 
control calculations (not on purpose), that I would get input overruns 
but no errors.  No tasks would die, no complaints, just the input 
overruns.  The mechanical engineer correctly started interpreting the 
input overruns as floating point issues, and he'd call me over to fix 
it.

Does this make any sense?  How can I set it up so that anything funny 
about the floating point causes an exception and kills my task?  I 
looked at the PowerPC supplement and didn't see anything.

Thanks,

Peter Dufault
HD Associates, Inc.