I have asked Ray if he has the time to put together a library of floating point functions so that members who may not have a background in maths can use functions of this type in much the same way as a user of a compiler would.

My own maths is now too far of the pace and Ray has a very extensive maths background so I think we can end up with a very useful library if its design is right.

What Ray has asked me to do is post a request for the type of functions that would be the most use in general purpose terms so that he has some idea of what to work on. My own thinking is that 80 bit precision for FP input/FP output would be useful and that there is also a need for Integer Input/Integer output.

Let me hear your ideas so that I can pass this on to Ray.


Posted on 2002-11-13 02:36:50 by hutch--

Floating point functions will depend on a tradeoff between speed and precission depending on the application.

For example the requirements of speed are far greater than of presicion for a game. At the same time precission far outweights speed for scientific and financial applications.

So basically a floating point library will require both 32 and 80 bit presicion. Can be in the same sourcecode using an equate like

ifdef USE_32FP
; do speed based routines here.
elseif USE_64FP
; do double presicion routines.
elseif USE_80FP
; do extra presicion here.
error You must define wich presicion to use.
Posted on 2002-11-14 09:11:44 by dxantos