[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [f-cpu] Re: Floating-Point
On Sun, Aug 19, 2001 at 09:55:25AM +0800, Glenn Alexander wrote:
> I'm new here, but aren't we developing a concept implimentation called FC-0
> and is complicating matters with 128-bit FP really useful at this stage? This
> is (i understand) intended to be a general purpose processor so shouldn't we
> be looking at general purpose cases that apply to normal users? I think if a
> special application needs 128 (or more) FP, they are more likely at this
> stage to be on a purpose-built mainframe. Later that may be different, but
> later FC1 will be done (hey, lets be optimistic) and since the F-architecture
> is already designed to go above 64-bits, implimenting a full 128 (or more)
> F-CPU is just something you do.
There are, in general, two reason why one wants large floating-point
types. One of them is precision (as some of you mentioned, that's a
rather rare case), the other (and more important) one is the increased
range. People often use IEEE double not because they need >6 digit
precision but because single has only 8 bits for the exponent -- and
sometimes 11 bits aren't enough either.
But I agree with you: there's no need to worry about FP with more than
64 bits for the FC0. In fact, the FC0 may not have FP hardware at all
(it's all optional, IIRC).
--
Michael "Tired" Riepe <Michael.Riepe@stud.uni-hannover.de>
"All I wanna do is have a little fun before I die"
*************************************************************
To unsubscribe, send an e-mail to majordomo@seul.org with
unsubscribe f-cpu in the body. http://f-cpu.seul.org/