> Date: Thu, 11 Mar 2004 07:35:58 -0800
> From: Richard Maine <[log in to unmask]>
> I wrote:
>
> >> I find selected_real_kind(12,30) to be a pretty
> >> verbose and nonintuitive spelling of 64-bit real.
>
> On Mar 10, 2004, at 6:34 PM, Robin wrote:
>
> > For 64-bit real, "double precision" and
> > "real (kind=kind(1.0d0))"
>
> No thanks. Been there. Done that. Regretted it. Wasted way too
> much of my time already converting codes that made this unjustified
> assumption.
>
> And for that matter, real(kind=kind(1.d0)) isn't exactly my idea of
> an intuitive way to spell 64-bits. Take any 100 programmers that
> don't already know f90 and ask them what that means. Would you like
> to take bets on the fraction that are even close? Go ahead and
> stack the deck my making them all competent programers, and they can
> even be f77 experts, as long as they don't know f90.
>
> > and "selected_real_kind(12)" might do it better.
>
> If you think it obvious that this means 64-bits, be my guest.
If you know of a better way, then use it.
It is not "verbose" if you do it properly (one line).
Nor is it appropriate to ask for "64 bits", as that's not
the precision you get when you want a 64-bit word.
Theres no guarantee that a machine will have 64-bit
f.p. numbers. Could be 60. Could be 80. Could be
anything.
> This differs only in the omission of the ",30" from what I was
> complaining about.
Well, that's an improvement, albeit small, n'est-ce pas?
> Sorry, but when I see 12, it doesn't
> immediately make me think of 64. Not to mention the fact that
> it takes half of a line to type (I always thought "double precision"
> was bad enough that way).
Most folks think in terns of the number of decimal digits required.
They don't think of 64 (bears no relation to the actual precision).
> --
> Richard Maine
|