Memory allocation problem
 

I am trying to run Fluent in parallel and test the speedup which I get out of the parallelization. I have been able to estimate the speed up for a grid of about 64^3 cells. now I am expecting the speedup to be better for the finer grid (~122^3 cells) but my problem is that I can not estimate the run time on a serial machine for this grid as the simulation fails after one iteration because of memory problems. I also have to mention that I could not adapt the 64^3 grid to ~122^3 grid on the serial machine due to the same problem with memory, so I refined the grid on two CPU's and then read the refined grid into the serial machine but as I said it crashed right away..so the question is; is there a way of increasing the amount of memory allocated for Fluent simulation? I do not use the Sun utility (SGE) and we do not have batch scheduler..etc I saw thru few websites that some people specify the memory allocation in the batch script or thru the SGE utility.. so any input on this issue will be highly appreciated...

Re: Memory allocation problem
 

please read the Fluent manuals

Re: Memory allocation problem
 

I kinda skimmed through the manual and could not locate this information. so if you know which section talks about this issue pls let me know. thanks for your reply

Re: Memory allocation problem
 

Make sure you have enough memory on the machine. Depending on the platforms you are using, on lnamd64, you may expect around 1GB memory per 1 million cells for single precision.

Re: Memory allocation problem
 

thanks for your response. the machine has about 4 GB of memory. if I use top command during the simulation I find that I am only using 10% of them. If I selected Grid/info/memory usage in Fluent, it shows that I need about 2 GB of memory for the simulation, so I need to devote about 50% of the available memory for my simulation. I am a new linux user and I donnu how to do this. thanks alot

Re: Memory allocation problem
 

What you have sound sufficient for the case. Can you post the error message here? In any case, if you want to study the speedup, you can always run with -t1 for the single-processor performance. What verion of Fluent are you using? Have contacted your Fluent support?