memory allocation [message #16460] |
Mon, 26 July 1999 00:00  |
Essa Yacoub
Messages: 3 Registered: July 1999
|
Junior Member |
|
|
I continue to get this error message,
(pv-wave)
Unable to allocate memory: to make array.
Not enough space
This is maybe a fragmented memory problem ? Because there should
be enough memory freed. How can I free up memory without fragmenting it
?
Or How do i free more system memory to use during a single session ?
(malloc/ free ? What is call_external?) Does delvar help ?
--
Essa Yacoub
University of Minnesota
Minneapolis, Minnesota
Email: yacoub@cmrr.umn.edu
|
|
|
Re: memory allocation [message #16503 is a reply to message #16460] |
Wed, 28 July 1999 00:00   |
Craig Markwardt
Messages: 1869 Registered: November 1996
|
Senior Member |
|
|
Essa Yacoub <yacoub@cmrr.umn.edu> writes:
>
> Thanks all... The limit/unlimit command seems to be active
> but I cannot change the memoryuse size, and datasize is unlimited.
> the error seems to occur when I reach the memoryuse limit. The
> sysadmin could also not change the size under root. He thought
> he might need to reinstall some kernels.
Hmm... You obviously want to increase the memoryuse limit, since that
controls the amount of physical memory available to you at one time
(as opposed to total swap space for the process).
You may have to increase the "hard" limit. Under the tcsh you do this
with "limit -h memoryuse XXX". Unfortunately, root has to do that, so
I don't know how that resource privelige can be passed to you, Joe
User.
Craig
--
------------------------------------------------------------ --------------
Craig B. Markwardt, Ph.D. EMAIL: craigmnet@cow.physics.wisc.edu
Astrophysics, IDL, Finance, Derivatives | Remove "net" for better response
------------------------------------------------------------ --------------
|
|
|
|
Re: Memory Allocation [message #38057 is a reply to message #16460] |
Sun, 15 February 2004 05:40  |
justspam03
Messages: 36 Registered: October 2003
|
Member |
|
|
Hi Neil,
Which IDL version do you use? It seems that memory handling
has improved quite a bit going from version 5.6 to 6.0.
Under 5.5/5.6 we observerd the following:
version a)
1) allocate large chunk of memory
(1200 objects, ca. 100MByte total)
2) allocate another chunk of memory
and resort data from 1) into this space
3) destroy objects allocated in step 1)
as they're not needed any longer
4) perform calculations on data located
in memory (from step 2)
version b)
exactly as a), but do not destroy objects
in step 3
Surprisingly version a) ran *much* faster (we're
talking about a factor of ~5) than version b)
Any idea why?
Seems this is not the case in 6.0 any longer.
Object destruction is quite a bit faster, too.
Kind regards,
Oliver
|
|
|