Tutorial on selecting simulation parameters #48
Conversation
|
I think this is a nice summary of the important issues. A lot of the discussion is from the perspective of running atomistic biomolecular simulations, which is certainly a major use case. However, some more comments on what to consider for other kinds of simulations might be good, especially since this is where the defaults are less likely to be appropriate, and users will actually need to think more carefully about choosing parameters. Of course, the best answer for how to pick a certain parameter might be "it depends on the system details, and without more information, you need to test to find an appropriate value," but we can emphasize that. Along these lines, monitoring energy drift over time is something that users might want to do to pick an appropriate timestep for a new kind of system they want to simulate, see whether or not they can use single precision and looser tolerances, etc. Since this is a tutorial, it could be good to have a code example showing how to do this (or perhaps that would be better suited for a cookbook entry). |
|
I was debating about whether to give an example of measuring energy drift. Doing it correctly is kind of complicated, but I suppose that's a reason to include it. If we don't show how to do it correctly, people will write their own code that does it incorrectly. |
At one point a long time ago I became a member of this set of people, and I have also seen others trying to figure out how to do this properly, which is why I think it would be good to have. |
|
Here's an updated version adding the sections you suggested. |
|
This is now finished. It should be ready to merge if it looks ok. |
2 participants
This is a rough draft of a tutorial on how to select values for simulation parameters. For the final version I plan to add figures, documentation links, etc. But first I'd like to get feedback on whether this is a good approach to the subject.