Re: where are the source code?

From: Dennis Shea <shea_at_nyahnyahspammersnyahnyah>
Date: Mon, 11 Jun 2007 21:01:36 -0600 (MDT)

Hi Saji

Many of the functions do sight the source. Most frequently in
the Description Section"

Cumulative distribution functions:
The source code is from source routine "CDFBIN" from
DCDFLIB (Double precision Cumulative Distribution Function LIBrary)

EOF routines:
"The eigenvalues and eigenvectors are calculated using LAPACK's "dspevx" routine."

Interpolation:
For a comparison and overview of some of the various interpolation
routines listed below, please see the Ngmath Library Overview [Link].

Random number generators:
The source of this random number generator is from the random section at Netlib. The authors were Brian Brown and James Lovato. The official reference is:

    Authors: P. L'Ecuyer and S. Cote
    Title: Implementing a Random Number Package with Splitting Facilities
    Journal: ACM Transactions on Mathematical Software 17:1, pp 98-111.

Singular Value Decomposition:
Note that the source LAPACK routine (dgesvd) returns V**T, not V.
However, setting optv=1 will result in the return of V.

Spherical Harmonics:
Spherepack

betainc
 The code used is from SLATEC (http://www.netlib.org/slatec/fnlib/).
This returns the same answers as the Numerical Recipes [Cambridge Univ. Press, 1986]
function betai.

------

Generally, when a public domain package is used, NCL does
(generally) site the source. Sometimes the source is inadvertently
not cited.

We will try to be better.

D

On Tue, 12 Jun 2007, Saji Njarackalazhikam Hameed wrote:

> Hi Dennis,
>
> The point raised by the original mailer (i can't read the name) is
> a valid one, i.e, the user would like to have some idea of the
> workings of the subroutine or function s/he is using. It would be
> nice if there was additional documentation outlining what algorithm a
> function uses (or it could point to the relevant literature or
> web-link). As you know some proprietary programs, for instance Matlab,
> provides this kind of information.
>
> Best regards,
>
> saji
>
> * Dennis Shea <shea_at_ucar.edu> [2007-06-11 09:59:09 -0600]:
>
> > ?? ?? wrote:
> > > hi all
> > > Are the source codes of many functions or procedures in NCL available
> > > for all users? I want to better understand the results calculated from
> > > functions or procedures, for example, how dose the procedure
> > > ??uv2sfvpf?? set the sphereic truncation and the mean or basic field for
> > > the final output result "sf" (streamfunction). Maybe only being clear
> > > to the calculation methods can I put more confidence on the accuracy
> > > of those output result. I??m very grateful if anyone could tell me
> > > where could I found the built-in codes?
> > >
> > > Best regards!
> > >
> > > _
> > NCL is not yet "open source". Hence , the raw code is not readily available.
> > NCL's funtions are a combination of C and fortran. Often, there is a C
> > interface
> > to a fortran subroutine.
> >
> > The uv2sfvp is from spherepack. See:
> >
> > http://www.cisl.ucar.edu/css/software/spherepack/
> >
> > Regards
> >
> > _______________________________________________
> > ncl-talk mailing list
> > ncl-talk_at_ucar.edu
> > http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>
>
_______________________________________________
ncl-talk mailing list
ncl-talk_at_ucar.edu
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Mon Jun 11 2007 - 21:01:36 MDT

This archive was generated by hypermail 2.2.0 : Fri Jun 15 2007 - 12:34:56 MDT