Date: Sat, 18 Dec 1993 10:30:22 +1100
From: Bill Metzenthen <BILLM AT vaxc DOT cc DOT monash DOT edu DOT au>
Subject: Re: suggestion for NPX initialization
To: djgpp AT sun DOT soe DOT clarkson DOT edu


Dieter ("buers AT dg1 DOT chemie DOT uni-konstanz DOT de") writes:

> I probably was wrong. I overlooked, that the `finit' instruction will reset 
> the control word (correct?). The code section for initialization is

yes, finit does a complete initialization of the FPU.

> I was suspicious because I found better results with a math test package
> (paranoia), after setting the control word to 0x137f.

The only difference you should see would be due to the FPU no longer
causing interrupts.

btw, paranoia is quite a good program but it is by no means perfect.
In particular, it does not handle very well the case where the
precision of intermediate results is different from the prevailing
precision _and_ the intermediate results have less than twice the
precision of the prevailing precision:

variables declared       FPU precision     paranoia
as:                      control:          report:
     float                 24 bits          ok
     float                 53 bits          ok
     float                 64 bits          ok
     double                53 bits          ok
     double                64 bits          reports FLAWs

The FLAW reported by paranoia in the last case is not a FLAW in the
FPU, and in my opinion (but I suspect that there are others who would
argue with this) should not be regarded as a FLAW at all.


--Bill