CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   Memory management incongruence (

Flavio October 31, 2006 14:29

Memory management incongruence
Hi all.

After I just started the solver, the program exits and shows me a message error about memory size. The action required is: "Increase the integer memory stack" or similiar. I searched up in the forum, and I found some issues like that. The standard solution proposed is to open the advanced solver paramter windows and set the memory coefficient higher. Or to choose a cheaper partitioner method. I tried these methods but they don't work. I anticipately thanks you for any suggestion.

Anyway, what i can't really get is why the program returns me such a message if using a bigger .gtm file (and so bigger .def, and higher element number) the solver runs without any problem. The difference between the two files, excepting dimensions, is that the working one has only an interface, while the not-working two interfaces. Notice that they represent the same object.

P.s. If significative for the comprehension, i add that i'm using a pc with 2gb ram, centrino duo processor 1.6ghz.

Thanks a lot.

Flavio October 31, 2006 15:02

Re: Memory management incongruence
I'm replying my own post in order to attach part of the out file. It seems strange to me that the solver stops even if the requested space is less than the free one, as listed:

+--------------------------------------------------------------------+ | | | *** INSUFFICIENT MEMORY ALLOCATED *** | | | | Action required : Increase the integer stack memory size. | | | | If the situation persists please contact the CFX Customer Helpline | | giving the following details:- | | | | Requested space: 16881316 words | | | | Current allocated space: 41159253 words | | Current used space: 23577585 words | | Current free space: 17581668 words | | Number of free areas: 3 | | | +-----------------------------------------------------------------

Furthermore, monitoring RAM usage through Task Manager Windows, it doesn't appear to reache the 2Gb limit at all.

I feel dazed and confused.

opaque October 31, 2006 15:13

Re: Memory management incongruence
Dear Flavio,

The message states that the ANSYS CFX solver did not allocate (estimate) enough memory (nothing to do with your available memory). It also says that internally there are 3 free areas, but none of them matches the requested amount (it is called memory fragmentation)..It is asking you to increase the default memory estimate by increasing the integer stack memory size..

If you are using the ANSYS CFX Solver Manager, you will find that the memory control are under Define Run/Solver tab under Detailed memory overrides..

If you are using the command line, just do cfx5solve -help and search for the flags to modify the integer stack


Flavio November 1, 2006 07:05

Re: Memory management incongruence
I tried to modify those values under "detailed memory overrides": the solver always stops requiring to increase integer memory. This until i set factor multiplier to 5. Then the program returns the same message, but the required memory is real. So i tried to do the same with that parameter but when i reach factor 5 the program returns this message:

*** Run-time memory allocation error *** | | Not enough free memory is currently available on the system. | | Could not allocate requested memory - exiting! | |

I wonder what is the cause that leads to this situation. Thanks for any suggestion you would offer.

opaque November 1, 2006 09:03

Re: Memory management incongruence
Dear Flavio,

You also mentioned that you split the geometry into two domains. If you are using a generalized grid interface (GGI) connection for your domain interface, the memory requirements are a lot higher than if you pick 1:1/Direct.

Are you solving a pseudo-2D problem using large symmetry planes?

It seems that you will end up calling support for this one..


Flavio November 1, 2006 13:18

Re: Memory management incongruence
It seems the problem was a large interface area with a too much refined surface mesh. Thank you very much for helping.

All times are GMT -4. The time now is 15:52.