Altair RISE

A program to recognize and reward our most engaged community members

Nominate Yourself Now!

About parallelization of Topology Optimization

Dear all,

I have questions on two topics:

1. My top. opt. model has a lot of elements and the in- core requirement for RAM is very high. The optimization model comprises one linear static & one modal analysis loadsteps. Restrictions also contain 5 Eigen frequencies and some nodal displacements, i.e. the constraints come from both load steps. Are Task Based Parallelization and Domain Decomposition Methods useable here? I am thinking of domain decomposition. Do you think the parallelization is advantageous? Another possibility for me is to use a computer cluster with a very high RAM capability. So what is better for topology optimization, TBP, DDM or just very high RAM?

2. I have 125 HWU. Does this set some limits on the amount of parallelizations?

Thanks for the response!

Find more posts tagged with

English

OptiStruct

Accepted answers

All comments

Altair Forum User

Hi Jope,

Please activate PARAM, HASHASSM = Yes : This is applicable to Static, Normal Modes & FRF analysis. It reduces memory requirement.

If you have multiple load cases in any analysis, you can use multiple CPU to solve each load case in Each CPU. There by reducing the run time. ( -cpu, - proc, -nproc, -ncpu, -nt). This also applicable to SMP version.

<?xml version="1.0" encoding="UTF-8"?>

Rahul_P1

A Solver has no limitation in utilizing RAM - the faster the better

For Optistruct one should aim for enough memory to satisfy RAM estimates printed by OptiStruct, preferably for in-core solver mode. If in-core requirements are too high to be practical (depending on expected size of typical run) it is still much better if OptiStruct run will fit in RAM instead of swapping out.

The –ddm run option can be used to activate DDM. Refer to Setting up OptiStruct SPMD and Launching OptiStruct SPMD for information on setting up and launching Domain Decomposition in OptiStruct.

OptiStruct > User's Guide > Running OptiStruct:

Memory Limitations

Altair Forum User

Thanks for your responses.

So, if my RAM is not enough for an in-core solution, is it still preferable to run Optimization 'normally' (with the PARAM, HASHASSM = Yes) rather than using the domain decomposition?

Jope