RE: [AD] div-by-zero... math3d.c

[Vincent Penquerc'h <Vincent.Penquerch@xxxxxxxxxx>]

Title: RE: [AD] div-by-zero... math3d.c

> They're not only Intel FPU numbers - they're part of the IEEE-754
> standard for floating-point numbers. Granted, that standard
> was mostly
> writen by Intel, but just about every other general-purpose CPU
> implementing an FPU supports them.

Ah, that's interesting. I thought other FPUs didn't bother with
denormals and yielded zero or infinity, depending on which was
appropriate. Thanks for the info.

> Technically, they shouldn't be any slower than normalized FP numbers.
> It's just that you don't want to double the number of transistors
> dedicated to the FPU to speed up some computations on some
> numbers that
> don't appear very often :)

Maybe, I have no idea what this entails in hardware :)

--
Vincent Penquerc'h