advantages of fused multiply-add, and a thesis project

Jon L White uunet!lucid.com!jonl%kuwait
Tue Jun 18 12:46:30 PDT 1991


re: [Explanations: single precision rather than double in order that higher-
    precision intermediates can be investigated.  

If I recall correctly, the MC68881 has a Rounding-Precision state, which 
can cause all rounding to be done to 32-bit format, or 64-bit format, 
despite the internal 96-bit registers (80 mantissa, 16 exponent?).  Some 
other hardware also offers this as a "mode" option.  When utilized, these 
modes seem to cause a very noticeable performance degradation; so I would 
suppose they exist for precisely the same purpose you are invoking here?


-- JonL --



More information about the Numeric-interest mailing list