#1627: Merge rewrite branch of McLachlan ------------------------------------+--------------------------------------- Reporter: eschnett | Owner: Type: enhancement | Status: new Priority: major | Milestone: Component: EinsteinToolkit thorn | Version: development version Resolution: | Keywords: ------------------------------------+---------------------------------------
Comment (by eschnett):
Merged.
Answers to Ian's questions:
- Performance benchmark results were posted to the mailing list. The upshot is that performance remains largely the same, except for McLachlan's built-in dissipation that is now much faster, probably slightly faster than thorn Dissipation. - Turning off dissipation: Yes, this is still a Kranc-time choice; this is necessary for performance. This the new performance is as good as or better than thorn Dissipation, I think this is now a non-issue. - Advection terms: The current code is not slowed down by the advection terms. I expect that other restructuring led to the same speed improvement that we gained from splitting out the advection terms previously.
- {{{ML_}}} prefix: I've kept this for backward compatibility. - FromBSSNCalc calculates everything that can be calculated from the BSSN state vector. This is not just the RHS, it also includes constraints, ADM variables, etc. Maybe "EverythingFromBSSNCalc" would be a good name. This change would be invisible to the user. - Renaming "phiW" to "cf": If so, then this needs to be done NOW. Please open a discussion on the mailing list. - Changing parameter names: I'd like to not do this for backward compatibility. Many parameter names don't change, and changing all of them is more disruptive than changing just a few gauge-related ones. - The helper thorn could be avoided. This seemed an orthogonal change, so I've omitted it for the time begin. We can look at it in a week or two, in case there's an unexpected problem with the rewrite.