Specifying memory mode (out-of-core or in-core)

- February 6, 2019 at 5:12 am
  
  kaiyeungli
  Subscriber
  
  Hi there,
  
  I am not sure how to specify memory mode (out-of-core or in-core) in ANSYS Workbench.
  
  The documentation says that either the BCSOPTION or DSPOPTION command should be used. Do I just make a one-line APDL command snippet?
  
  For one model, I tried adding the single-line command snippet: dspoption,default,outofcore,300,,,off.
  
  It didn't seem to have any effect however. The model runs on in-core mode by default, and still runs on in-core mode even with the DSPOTION command.
  
  How do I actually specify the memory mode for an analyse?
  
  Kind regards,
  
  Kai-Yeung
- February 6, 2019 at 10:12 pm
  
  kaiyeungli
  Subscriber
  
  Does anyone have any knowledge on this issue?
  
  The situation is that I have a very large model which I plan on running on a HPC cluster. Due to the large amount of physical memory available on the cluster, I want to make sure that my job runs on in-core model, so that it runs faster.
  
  Any help would be much appreciated.
  
  Kai-Yeung
- February 6, 2019 at 10:24 pm
  
  peteroznewman
  Subscriber
  
  You use the command snippet
  
  DSPOPTION,,INCORE
  
  to force the selection of in-core solving if you are using Distributed ANSYS solving.
  
  The solver will automatically select in-core solving if it can. I only use this when the solver decides that it couldn't run in-core and I wanted to force it to, because I knew that it could run in-core. The estimating it does to decide if the model can run in-core is conservative.
  
  If I am wrong and it can't fit in-core, then the solver crashes when it runs out of RAM.
- February 6, 2019 at 10:32 pm
  
  kaiyeungli
  Subscriber
  
  So basically that command only works one way? Changing the memory mode from out-of-core to in-core, but not from in-core to out-of-core? (i.e. If the solver will still run on in-core mode if it can, even with the DSPOPTION command)
- February 6, 2019 at 11:15 pm
  
  peteroznewman
  Subscriber
  
  You said you wanted to make sure your model runs in-core, and that is the command to do it.
  
  A different scenario would be that you know how much RAM is available and there are plenty of cores available and you have two jobs that would each take up more than half the available RAM if they ran in-core. Therefore, you want to limit the RAM each job uses so you can solve two models simultaneously. One thing you could do is to force the jobs to run out-of-core and each would use a small amount of RAM.
  
  DSPOPTION,,OUTOFCORE
  
  If this answers you question, please mark a post with Is Solution to close the discussion, or ask a follow-up question.
- February 7, 2019 at 1:23 am
  
  kaiyeungli
  Subscriber
  
  DSPOPTION,,OUTOFCORE does work. Thanks.
  
  However, when I try to add the 1st (REORD_OPTION) and 3rd argument(MEMORY_SIZE) of the command (i.e. dspoption,default,outofcore,300,,,off), it doesn't work, and the job runs on in-core mode.
  
  Do you have any idea why that is?
- February 7, 2019 at 2:12 am
  
  peteroznewman
  Subscriber
  
  Why would you use anything other than the default reordering scheme? Do you know more than the ANSYS developers?
  
  Why would you need to define the 3rd argument? Once you specify out-of-core, the solver uses a small amount of memory.
  
  You said, " I want to make sure that my job runs on in-core model, so that it runs faster" then you are using out-of-core which is slower than in-core. I'm confused.
- February 7, 2019 at 2:34 am
  
  kaiyeungli
  Subscriber
  
  I was simply wanting to learn how to use DSPOPTION more generally. That was why I was trying to use it on a smaller 2D model, which could easily run in-core on my local desktop, and use the DSOPTION command to force to it go into out-of-core mode to see what happens. I was curious as to why the DSPOPTION command didn't work with work with all of its arguments.
  
  When I run my large models on the cluster, I would obviously be using DSPOPTION,,INCORE
  
  I agree with you however, that practically, I don't have much interest in changing the reordering scheme, or changing from in-core to out-of-core mode. My last post was simply a product of my curiosity...
  
  Thanks for your help Peter.