home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.lang.idl-pvwave
- Path: sparky!uunet!elroy.jpl.nasa.gov!ames!nsisrv!stars.gsfc.nasa.gov!thompson
- From: thompson@stars.gsfc.nasa.gov (William Thompson, code 682.1, x2040)
- Subject: Re: IDL, PV-Wave, Matlab, Khoros
- Message-ID: <21JAN199311590839@stars.gsfc.nasa.gov>
- News-Software: VAX/VMS VNEWS 1.4-b1
- Sender: usenet@nsisrv.gsfc.nasa.gov (Usenet)
- Nntp-Posting-Host: stars.gsfc.nasa.gov
- Organization: NASA/GSFC-Laboratory for Astronomy and Solar Physics
- References: <19JAN199317494582@stars.gsfc.nasa.gov> <1jers7INN9bh@elroy.jpl.nasa.gov> <1993Jan19.175815.8722@iscnvx.lmsc.lockheed.com> <1jjo6jINN1tn@usenet.INS.CWRU.Edu>
- Date: Thu, 21 Jan 1993 15:59:00 GMT
- Lines: 22
-
- In article <1jjo6jINN1tn@usenet.INS.CWRU.Edu>, at913@cleveland.Freenet.Edu (Mirko Vukovic) writes...
-
- (stuff deleted)
-
- >Programs written in IDL can be 30-60 times slower than equivalent programs
- >written in FORTRAN, but if one uses extensive array manipulation facilities
- >execution improves tremendously (factors of 10).
-
- (rest deleted)
-
- Actually our experience has been that, depending on the problem and how the IDL
- code is written, IDL routines are just about as fast as an equivalent FORTRAN
- routine. The secret seems to be the use of loops. If you can write your code
- without any DO loops--and a lot of times you can--then IDL's performance is
- just about the same as a FORTRAN routine. If you need to use loops, then
- performance suffers greatly from the need to reinterpret the commands over and
- over again.
-
- A lot of times I've seen less experienced IDL programs use loops when there
- were ways to avoid them--I've been guilty of it myself.
-
- Bill Thompson
-