| Re: Philosophical Question about NAN [message #63755 is a reply to message #63742] |
Mon, 17 November 2008 08:02   |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
wlandsman writes:
> I agree with the sentiment but also note that always setting /NAN
> incurs a non-trivial performance penalty, e.g.
>
> IDL> a =3D randomn(seed,10000,2000)
> IDL> t =3D systime(1) & b =3D total(a) & print,systime(1)-t
> 0.25451803
> IDL> t =3D systime(1) & b =3D total(a,/nan) & print,systime(1)-t
> 0.35278893
>
> I've thought at times that arrays should carry a hidden bit saying
> whether or not they include NaN values, but this introduces other
> overhead problems.
I guess I would argue that in the overwhelming number of
cases in my experience, the performance penalty is trivial.
I'm calling these routines a couple of times at most. And
I am not arguing for the elimination of the keyword, only
that the default value could be changed. Thus, if I *was*
experiencing a performance penalty, and I was certain I
had good numbers, I could always set the NAN keyword to 0.
Cheers,
David
--
David Fanning, Ph.D.
Fanning Software Consulting, Inc.
Coyote's Guide to IDL Programming: http://www.dfanning.com/
Sepore ma de ni thui. ("Perhaps thou speakest truth.")
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
