Running eORCA1 at NEMO vn3.6 on a single PE

[frrh]

Context

Here we aim to investigate whether it is possible to run eORCA1 in a 1x1 PE configuration
on the Met Office Cray XC40.

The ability to do that with many component models has been lost as resolutions get higher and higher and machines revert to MPP-style distributed chips and memory. However the Cray XC40 features nodes of 32 or 36 PEs with local memory available to that node.

We have found that it is possible to run eORCA1 (NEMO-CICE) within a single node and conclude that in principle it should be possible to run on a single PE provided that PE has access to the node's full resources.

This is a potentially very useful thing to be able to do for debugging purposes.

Implementation

The configuration we aim to investigate is an eORCA1 NEMO-CICE configuration developed at the Met Office (GO6-GSI8). This configuration will form a component or the basis of various other configurations, including the low res UKESM coupled model and eORCA1 NEMO-CICE-MEDUSA.

We know that it is possible to run this configuration on a Cray XC40 Haswell node (32 PEs)

[frrh]

Details of procedures, investigations and findings are recorded here: wiki:ticket/1721/General

[frrh]

Closing ticket since work not aimed at NEMO trunk inclusion per se.
This is one of those things you don't miss until you can't do it any more and would still be highly desirable
to have documented somewhere for those with an interest in single PE running, debugging techniques etc.
Hence I will transfer relevant details to internal MO systems.