Did anyone here ever sucessfully installed AceGen/AceFEM on a cluster?
Does anyone has experience with performing AceFEM simulations on a cluster system?
Is it possible at all?
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
From version 7.402 the AceFEM based finite element simulations can be accelerated on multi-core architectures as well as on cluster type architectures. The AceFEM based finite element simulations on a multi-core or cluster architectures can be accelerated by utilizing several types of parallelization: A) the procedure used to collect the contributions of individual finite elements to the global matrices and vectors is fully parallelized for the multi-core architectures (parallelization of the assembly procedure). This is available on all platforms. B) the solution to the system of linear equations performed by the PARDISO linear direct and iterative solver is parallelized for the multi-core architectures (parallelization of linear solver on multi-core architectures). This is available on all platforms. C) several finite elements simulations can be performed in parallel on multi-core architectures using Mathematica parallel computing capabilities (see LaunchKernels) (parallelization of FE simulations on multi-core architectures). This is available on all platforms. D) the solution to the system of linear equations is parallelized on the cluster type architectures using MKL Cluster Sparse Direct solver or PETSC Krylov Subspace iterative sparse solver (parallelization of the linear solver on cluster architectures). This is available on Linux platforms. E) several finite elements simulations can be performed in parallel on cluster type architectures using standard MPI or Wolfram Lightweight Grid (parallelization of FE simulations on cluster architectures). This is available on Linux platforms.
