3.1.1. Running Autodyn with Intel MPI

Prerequisite Setup Steps (Distributed Parallel Only)

If you are running an Autodyn job in distributed parallel (master and worker processes distributed across different machines), follow these steps (these can be disregarded for running local parallel jobs):

Intel MPI must be installed on each machine that will be used in the parallel processing environment by following the installation instructions for Intel MPI as provided in Setting up a DMP Analysis. Ensure the same version is installed as you intend to use to run the job.
Once installed, you must ensure that the hydra_service.exe service is installed and running on each machine used. The hydra_service version must match the Intel MPI version you intend to use to run the job. This is done by running the hydra_service.exe in the Intel MPI install using the following commands:
- hydra_service --install
- hydra_service --start
It may be necessary to set the MPI password on each machine used. To do this, call mpiexec with the -register argument the first time you run mpiexec, entering your user password when prompted.

Running an Autodyn Job Using Intel MPI

There are two ways to run parallel Autodyn jobs with Intel MPI: using the AUTODYNWRAPPER.exe or calling mpiexec.exe directly. Using the AUTODYNWRAPPER.exe is the recommended method to use and its usage is documented in Running Autodyn on Windows Platforms.

Note: By default, the AUTODYNWRAPPER.exe uses Intel MPI version 2021.11. To use previous versions of Intel MPI, set the environment variable I_MPI_ROOT to the install of the version that you intend to use before launching the AUTODYNWRAPPER.exe.

Alternatively, an Autodyn job can be run in parallel using the mpiexec.exe executable from the Intel MPI install directly. This may be appropriate if it is necessary to use Intel MPI options that are not currently supported by the AUTODYNWRAPPER.exe.

To run a parallel Autodyn job directly using mpiexec.exe, follow these steps:

In the same folder as the model file you wish to run, create a file called applfile. In this file, enter the following text, making the relevant changes for your parallel model, machine names, and installation path, if the installation is not in the default location:

-localroot -env ANSYS_EXD_MPI_TYPE intelmpi -host [HeadNodeMachineName] -np 1 -wdir (cont'd on next line...)
(...cont'd) "C:\Program Files\ANSYS Inc\v242\aisol\AUTODYN\winx64" autodyn.exe

-host [NodeMachineName] -env ANSYS_EXD_MPI_TYPE intelmpi -np [number of workers] (cont'd on next line...)
(...cont'd) "C:\Program Files\ANSYS Inc\v242\aisol\AUTODYN\winx64\adworker.exe"

Set the environment variable I_MPI_ROOT to the install location of the Intel MPI install you want to run the job with. For example:
```
set I_MPI_ROOT=%AWP_ROOT242%\tp\MPI\Intel\2021.11.0\winx64
```
In the same directory as the model you wish to run, start Autodyn using mpiexec.exe with the following command:
```
"%I_MPI_ROOT%\bin\mpiexec.exe" -configfile applfile
```

When the instructions shown above have been followed, 1 master process (autodyn.exe) will be launched on HeadNodeMachineName and number of workers worker processes will be spawned on NodeMachineName. The working directory is specified by the -wdir option and must be modified if the installation of Autodyn is not in the default location.