| Re: Numerical Recipes diffs? [message #6355] |
Fri, 21 June 1996 00:00 |
kachun
Messages: 7
Registered: June 1996
|
Junior Member |
|
|
In article <4qchu0$2qc@vixen.cso.uiuc.edu> kspencer@s.psych.uiuc.edu (Kevin Spencer) writes:
> kachun@bogart.Colorado.EDU (Ka Chun Yu) writes:
>
>> I've now tested the IDL3 and IDL4 equivalents of the above routines
>> with the same inputs, and both routines give different results in
>> the two IDL versions.
>
> One possible reason for the different results is that the new routines
> take row-major arrays, not column-major. (I might have just switched
> these two -- anyway, the new routines take the transpose of the old
> format.) All the differences are described in the on-line help.
Yes! That was the problem. One could either enter in the transpose
of the input matrixes, or for some functions, specify the COLUMN key-
word to use column-major input. Thanks to all those who helped.
> Kevin
--
--kachun +** Center for Astrophysics and Space Astronomy **+
+** University of Colorado, Boulder **+
+** Email: kachun@casa.colorado.edu **+
+** http://casa.colorado.edu/~kachun **+
|
|
|
|
| Re: Numerical Recipes diffs? [message #6365 is a reply to message #6355] |
Thu, 20 June 1996 00:00  |
kspencer
Messages: 21
Registered: December 1993
|
Junior Member |
|
|
kachun@bogart.Colorado.EDU (Ka Chun Yu) writes:
> I've now tested the IDL3 and IDL4 equivalents of the above routines
> with the same inputs, and both routines give different results in
> the two IDL versions.
One possible reason for the different results is that the new routines
take row-major arrays, not column-major. (I might have just switched
these two -- anyway, the new routines take the transpose of the old
format.) All the differences are described in the on-line help.
Kevin
-----------------------------------------------------------
Kevin Spencer
Cognitive Psychophysiology Laboratory and Beckman Institute
University of Illinois at Urbana-Champaign
kspence1@uiuc.edu
-----------------------------------------------------------
|
|
|
|
| Re: Numerical Recipes diffs? [message #6366 is a reply to message #6365] |
Thu, 20 June 1996 00:00 |
kachun
Messages: 7
Registered: June 1996
|
Junior Member |
|
|
In article <4qbh8u$r9b@fu-berlin.de> marq@strat01.met.fu-berlin.de writes:
> Ka Chun Yu (kachun@bogart.Colorado.EDU) wrote:
> : Does anyone know what IDL did between IDL3 and IDL4 as far as
> : the numerical recipes subroutines go? I've been testing a fitting
> : program that uses the LUDCMP and LUBKSB routines form _Numerical
> : Recipes_, and I've been using IDL's canned routines. The canned
> : routines got name changes during the switch from IDL3 to IDL4,
> : from "nr_ludcmp" and "nr_lubksb" to "ludc" and "lusol". That's
> : not all though--somehow the code got changed around so that I
> : get different results depending on which version of IDL I use
> : even if the input is exactly the same!
> Check the link 'What was New in IDL version 4.0' / Mathematics Improvements'
> of the IDL Online Help (idlhelp, or ? from the IDL command prompt).
> Probably the most important sentence:
>
> "Note that the old routine names still work, but are undocumented."
Yes I noticed this. Unfortunately this statement does not seem to
hold for LUDCMP and LUBKSB. Running "nr_ludcmp" or "nr_lubksb" gives
me a "Variable is undefined:" error in IDL4.
I've now tested the IDL3 and IDL4 equivalents of the above routines
with the same inputs, and both routines give different results in
the two IDL versions.
> Chris Marquardt (marq@strat01.met.fu-berlin.de)
--
--kachun +** Center for Astrophysics and Space Astronomy **+
+** University of Colorado, Boulder **+
+** Email: kachun@casa.colorado.edu **+
+** http://casa.colorado.edu/~kachun **+
|
|
|
|
| Re: Numerical Recipes diffs? [message #6374 is a reply to message #6365] |
Thu, 20 June 1996 00:00 |
marq
Messages: 12
Registered: February 1996
|
Junior Member |
|
|
Ka Chun Yu (kachun@bogart.Colorado.EDU) wrote:
: Does anyone know what IDL did between IDL3 and IDL4 as far as
: the numerical recipes subroutines go? I've been testing a fitting
: program that uses the LUDCMP and LUBKSB routines form _Numerical
: Recipes_, and I've been using IDL's canned routines. The canned
: routines got name changes during the switch from IDL3 to IDL4,
: from "nr_ludcmp" and "nr_lubksb" to "ludc" and "lusol". That's
: not all though--somehow the code got changed around so that I
: get different results depending on which version of IDL I use
: even if the input is exactly the same!
: I don't think IDL provides the source for these routines in their
: libraries; they leave them as executables. So does anyone know
: what IDL did between versions, or should I just forget about this
: and rewrite everything from scratch (which would be really annoying)?
Check the link 'What was New in IDL version 4.0' / Mathematics Improvements'
of the IDL Online Help (idlhelp, or ? from the IDL command prompt).
Probably the most important sentence:
"Note that the old routine names still work, but are undocumented."
Hope this helps,
Chris Marquardt (marq@strat01.met.fu-berlin.de)
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
