| Re: Performance improvement on IDL 6.2 for Linux/MacOSX? [message #46622 is a reply to message #46574] |
Thu, 01 December 2005 07:57  |
alban
Messages: 5
Registered: November 2005
|
Junior Member |
|
|
I think the key point in IDL performance is the processor. Some
colleagues of mine told me IDL did not work very well -not as well as
expected- on a brand new 6000$ Solaris 64-bit platform two years ago.
In my Mac G4 1.2 GHz it is incredibly slow -just a little bit faster
than an old P3 1GHz- and that has little to do with compiler or
graphics system. It is definitely the processor.
Probably IDL is optimised for x86 architectures alone, which is,
considered the price of one licence, a deeply admirable strategy; the
differences between linux/win performance can easily be explained by
the use of different compilers...
Concerning your question, I do not have many experience working with
Windows and IDL. I had used Matlab on Windows before, and the
multitasking capabilities of the system were far below those of Unix
platforms. Some people have told me this has been improved with XP;
anyway, I performed a test yesterday on a Windows XP computer (P4, 2.4
GHz, 1 GB RAM) using the same code that I usually run on my Linux (same
computer) and MacOS (G4, 1.2GHz, 1.25 GB RAM) computers. I ran the
program -image processor- three times simultaneously taking care of
switching the priorities to low for the three processes. The equivalent
in Mac OS or in Linux is to renice them. Both Linux and Mac OS worked
much better than Windows.
Do not take me wrong, I do think that IDL on Windows works fine -better
as I had expected, it is true-, but the system is not as stable as
others. When performing numerical simulation, or heavy computational
tasks, the main advantage of a system is its stability. I think that
IDL is a language mainly developed for numerical simulations, and so I
am sure there are more suitable machines for that than the Windows
ones. I find it very "funny" that IDL works best on platforms which are
less suitable for its main purpose -although the most spread ones, it
is true-.
Regards.
Marc
> Mark Hadfield wrote:
> alban wrote:
>> Hi, Mark,
>>
>> I do not know, probably just some prejudices. Anyway, the Wintel
>> computer was the same one as the Linux one -differently booted-; I
>> guess I had expected the same performance on the same hardware,
>> independently of the OS.
>
> Oh right, I thought you were expecting poorer performance on Windows.
>
> As one of the gurus pointed out earlier in this thread (too lazy to look
> it up) the OS itself has very little effect on IDL performance. The
> important variables are compiler (Windows IDL has a slight advantage
> here) and the graphics subsystem (the comparison is mixed).
>
>> My "problem" is that I run heavy code on IDL and I need to keep on
>> working after that; that makes the use of IDL on Windows platforms
>> impossible.
>
> Why's that? Memory fragmentation (which I'll accept)? Or inability to
> run IDL at low priority (which I won't )?
>
>
> --
> Mark Hadfield "Kei puwaha te tai nei, Hoea tahi tatou"
> m.hadfield@niwa.co.nz
> National Institute for Water and Atmospheric Research (NIWA)
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
