Floating point exception precision

Mitch Bradley wmb at firmworks.com
Sat Mar 17 17:05:28 EDT 2007


The Geode is currently configured for imprecise floating point 
exceptions.  Is that what we want?

(Imprecise exceptions make the FPU run faster, at the expense of 
exceptions potentially happening after the instruction that caused them. 
Precise exceptions limit the FP pipeline to one instruction at a time.)





More information about the Devel mailing list