|
[Sponsors] |
March 5, 2008, 21:26 |
Help understanding Memory Allocation Factor
|
#1 |
Guest
Posts: n/a
|
I am trying to understand what this does. From what I can understand from CFX Help is that it is just an estimation of the memory needed to run the analysis. It then reserves that amount memory to be used by the simulation.
Do I have this right? Also, by increasing it by 200% or 300%, will the simulation run faster or is it just reserving more memory that will not be used? |
|
March 6, 2008, 14:27 |
Re: Help understanding Memory Allocation Factor
|
#2 |
Guest
Posts: n/a
|
The estimation factor just multiplies the estimated memory requirements by the factor you specified. Sometimes you need to increase it, sometimes decrease it, depending on what your memory error is. For instance, if the solver cannot allocate enough memory, a mem factor of 0.9 might help by reducing the memory needed. On the other hand, if the solver did not allocate enough memory, a mem factor greater than 1 is required. If you read the memory error message carefully you'll know what to do.
The solver is pretty smart about how it uses the available memory. To keep the memory overhead down, there are values that it will recalculate when needed. If you increase the available memory, the solver will use it to store these parameters for future use. If you have memory to spare you can increase it to roughly 3x to 5x before the solver runs out of stuff to fill it with. Since pulling some of these parameters from memory can be faster than calculating them, some speedup can occur. -CycLone |
|
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
about numeca harmonic method Memory allocation failed | uqzcpo | Fidelity CFD | 2 | October 7, 2011 23:10 |
CFX CPU time & real time | Nick Strantzias | CFX | 8 | July 23, 2006 18:50 |
how to define memory allocation factor in unix | steven | CFX | 1 | October 25, 2005 11:26 |
Memory allocation problem | shhe | CFX | 4 | January 7, 2004 00:00 |
memory allocation | strudl | CFX | 1 | October 5, 2003 11:26 |