What I did in this case was trace it back with the debugger, to discover
 that I had a divide-by-zero, but I was wondering if there was some way to
 figure out which register caused the error so that it could be fudged (with
 say + infinity)?  The processor in question is a 21161N

 Failing being able to do that, sending an error message with the return
 address to the host processor to be logged sounds like a good idea.

 In this case, the illegal result wasn't in and of itself a problem, but it
 was paired with another value that normally would have prevented the 
illegal
 value from being used, except in the case of an interrupt occuring between
 the updating of the two values!

 Thanks,

 Tom