Re: float IEEE 754

Well the IEEE-54 standard standard is supposed to be used by everyone implementing floats simply as a standard way for interoperability. Consider using a PC, a game console or a supercomputer. If everyone has their own way of doing things, like it was a few decades ago, then specialized compilers would be needed to calculate binary data for every different platform available Not the way to do business.

Consider the LINPACK benchmark, what good would it be if every machine had a different way of implementing floats.


"Nathan Mates" <nathan@xxxxxxxx> wrote in message news:13tim04smoc1efa@xxxxxxxxxxxxxxxxxxxxx
In article <0kght31p7duierqg5mnlailjrsc2hj0qg5@xxxxxxx>,
Tim Roberts <timr@xxxxxxxxx> wrote:
Many graphics applications use fixed-point numbers in various
formats. It works fine, but those programs have to do their own
arithmetic by hand.

Speaking as someone who programs games for a living, that used to
be more true than it is today. The earliest consoles, and before the
Pentium II era, floating point units were nonexistant (at worst) to
horribly slow (at best). However, starting about 2000, with the era of
the Sega Dreamcast, Sony Playstation 2, Microsoft XBox, and Nintendo's
Gamecube, it's been expected that your platform will have a quality
floating point unit. Once you have such a thing, it's better to use
that floating point unit, because that allows multiple sections of
your CPU to be in use at once.

Even handheld units are starting to get into the floating point
business. The Nintendo DS (top-selling unit) doesn't do floating point
to my recollection. However, the Sony PSP is a cut-down PS2, so it
definitely has good floating point, especially when you use the vector

Nathan Mates
<*> Nathan Mates - personal webpage
# Programmer at Pandemic Studios --
# NOT speaking for Pandemic Studios. "Care not what the neighbors
# think. What are the facts, and to how many decimal places?" -R.A. Heinlein