| call external speed [message #12686] |
Wed, 02 September 1998 00:00  |
curtis volin
Messages: 1
Registered: September 1998
|
Junior Member |
|
|
I use the IDL SVD command very frequently, so I was wondering if it would
be any faster to create a DLL and call it with call external. I suppose
that the fundamental question is, are IDL routines optimized for speed?
Thanks,
Curtis Volin
cev@u.arizona.edu
Research Associate
Optical Sciences Center
University of Arizona
|
|
|
|
| Re: call external speed [message #12756 is a reply to message #12686] |
Fri, 04 September 1998 00:00  |
steinhh
Messages: 260
Registered: June 1994
|
Senior Member |
|
|
In article <6soch3$5da$1@nnrp1.dejanews.com>
rmlongfield@my-dejanews.com writes:
> I have a question though: Which chapters in the Advanced Development guide
> are related to LINKIMAGE? There seems to be a mixture of topics. Do you have
> a short example? I can't find any in the IDL documentation. (I work with
> UNIX on an SGI workstation.)
Now I don't have the printed documentation, but the online help
does have some examples (IDLv5.1). Under the "Contents" in the
normal online help, double-click on the *book* "IDL External
Development Guide", double-click on "Open IDL Ext....". Then
press the Contents button, and double-click again on "External
Development Guide" - a list of chapters should appear.
"Adding system routines" is crucial to LINKIMAGE. Most of the
other chapters are relevant, too, depending on what you want
to do in your routines.
The Caution under LINKIMAGE is appropriate, but only to a
certain extent. You do *not* have to have "intimate knowledge
of the internals of IDL" in the sense that you have to qualify
for work in RSI's development group - not by a long shot.
The very least you have to know about is how variables are
*stored*, *passed*, and returned. That's basically it (for
writing something simple that works - for the right reasons).
Using Callable IDL has nothing as such to do with LINKIMAGE
versus CALL_EXTERNAL, all permutations are possible.
However, you *have* to use Callable IDL to install new system
procedures from inside CALL_EXTERNAL routines (i.e., *not*
using LINKIMAGE, but the IDL_AddSystemRoutine function).
On the other hand, if you're using LINKIMAGE, Callable
IDL provides some *very* useful routines/macros to do checks
on your input data etc. For example, IDL_ENSURE_ARRAY will
issue standard "system" error messages and associated
behaviour (stopping etc), in exactly the same way that e.g.
MEDIAN does:
IDL> print,median(0)
% MEDIAN: Expression must be an array in this context: <INT ( 0)>.
% Execution halted at: $MAIN$
Callable IDL also provides routines for keyword processing,
type conversion etc. *Very* handy.
One thing that I cannot find in the online help is a good
listing of "these routines and macros are available when
linking your shareable objects with the idl library
(i.e., using Callable IDL)". There is a list for RPC
function calls, but not Callable IDL. Thus, you have to
guess, and read through many chapters and a lot of text
to find what you need (or just drop the whole thing
because you don't know what you're missing!).
Regards,
Stein Vidar
|
|
|
|
|
|
|
|
|
|
| Re: call external speed [message #12909 is a reply to message #12686] |
Mon, 14 September 1998 00:00 |
Karl Krieger
Messages: 8
Registered: April 1995
|
Junior Member |
|
|
On Tue, 8 Sep 1998, David Foster wrote:
> ... Just for kicks, I decided to compare the speed of my
> routine and IDL's INTERPOLATE() routine. The results were *very*
> similar. I would say that it is fairly safe to say that IDL's
> routines are well optimized.
It really depends on the application. I wrote a LINKIMAGE wrapper for the
FFTW package ( http://theory.lcs.mit.edu/~fftw ) and compared the speed to
IDL's native FFT routine. The speed gain for single precision
real->complex 2d transforms is about 2.5 on a SUN Ultra/170 and about 2.3
on a Pentium/133 under WinNT, so it's really worth the effort if you want
to do FFT of large data sets.
Karl
--
Max-Planck-Institute for Plasma Physics
Boltzmannstr.2, 85740 Garching, Germany Email: krieger@ipp.mpg.de
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
