In PAMSTAMP, a computation host is more than the name of a computer. It is a saved set of launch settings: where the calculation runs, which solver it uses and how it starts.
Once configured, you can choose it in the start, restart and optimization dialogs. Let us go through those settings so they stop looking like unrelated tabs.
Compare the machine and launcher settings
Create or edit a host through the solver host controls. Give it a label you can recognise, such as a local forming configuration or a particular remote solver.
On Machine, select Local for your computer or Network for another machine. For Network, enter the actual hostname or address and choose its operating system.
On Launch, the solver launcher needs its full path. For a remote host, that path is read from the remote machine’s point of view. A Windows path on your laptop won’t identify a Linux executable on the server.
Choose the solver and parallel mode
On Solver, select the intended solver type, parallelism and precision. The documented options distinguish Mono, SMP and DMP.
DMP exposes additional MPI and node-distribution controls. Use the library configuration intended for that machine and the release, rather than choosing a library because its version number is higher.
The first host in a defined DMP distribution must match the host selected in the Machine panel.
How will it launch?
The documented routes include direct local execution, a custom local script and a solver manager service for local or remote execution.
Solver Manager receives launch requests and starts the calculation. It must be installed and configured on the execution side for that route. The older guide describes configuration files and service behaviour; use the installed manager’s matching documentation before modifying a shared service.
For a Network host, Folders lets you define equivalences between the GUI and computation-side paths. Confirm that both refer to the intended project data and that the job account can access it.
Test with a small model
Save the host, choose it in Start a Computation and run a suitable small example. Read the launch and solver messages. Confirm the output reaches the expected directory before sending a large study through it.
That small test checks the connection between all these settings much more usefully than simply seeing the host label in a list.
Previous: PAM-STAMP Project Files: What Each File Is For
Next: PAM-STAMP Feasibility vs Validation: What Changes?