Search for multiply-add speedup

[uunet!cwi.nl!dik]

 > To the extent that you can tell anything about these very short time
 > intervals, it looks like the preprocessor doesn't do much good if the data
 > fits in the cache.
 > 
 > Looks like the benefit of multiply-add shows up primarily when the data fits
 > in the cache, even with the benefit of an agressive preprocessor.  
 > Correct conclusion?

I would agree, based on my experiences with i860's in an Alliant FX2800.
The problem appears to be memory bandwidth again.  (For multiply-add
you need three operands for a single result.  The memory bandwith must
cater for those three operands, but in general allows for only one or two.
So out-of-cache multiply-add is not faster than multiply followed by add.)