For what I most often do (quadrature and solving integral equations)
the reasons for not using the highest possible precision are not memory
limitations but, most importantly lack of patience (quad precision takes
longer than double), and less importantly the desire to try my programs
with several different compilers: many selected_real_kind(33) intrinsics
are missing from g95, which is otherwise a good compiler. I don't use
REAL64 and friends because that's not available in either g95 or our
rather old (2014) version of Sun Fortran.
On Mon, 3 Apr 2017, Phillip Helbig wrote:
> Date: Mon, 3 Apr 2017 18:24:51 +0200
> From: Phillip Helbig <[log in to unmask]>
> Reply-To: Fortran 90 List <[log in to unmask]>
> To: [log in to unmask]
> Subject: Re: REAL64 what is it good for!
>
> On a related note (limited connectivity while in hospital, thus not a usual
> elegant post), is there anything these dazs lige G-Float and D-Float on
> the VAX (same number of bits, different precision)?
>
> these days like
>
> Yes, memory is cheap these days, so there is the temptation of overkill,
> but I am sure that there are still many memory-limited applications, so
> if one needs just more precision but not much range, or just more range
> but not much precision, then the same number of bits but split up
> differently could be attractive.
>
> Possibly related question: is everything IEEE these days?
>
-- John Harper, School of Mathematics and Statistics
Victoria University, PO Box 600, Wellington 6140, New Zealand
e-mail [log in to unmask] phone (+64)(4)463 5276 fax (+64)(4)463 5045
|