GSPH update #273
GSPH update #273
Conversation
removing GHLLC (depricate), adding a ArtificialViscosity implementation for direct comparison in the GSPH paper. Might need to rename to make it less confusing with the AV class heirarchy in Spheral
update policy is now used to handle the Riemann gradients that are stored between steps. for problems with const bc and initial pressure gradients (hydrostatic atmospheres etc...) the pressure gradient will need to be given initial conditions similar to eps or rho or any other field
…ty gradient on start up
this commit removes the isFirstCycle switch which I was using to hackishly initialize the spatial derivatives. When switching over to using the initializeOnStartUp method the switch never flipped messing up the gradients.
only commenting out the Hstretch stuff b/c I'll want to create a H evolution phys package after mike's update which recreates what Hopkins did in his 2015 paper
There was a problem hiding this comment.
Looks pretty good. Two questions occur to me:
- You have added a reimplementaion of the limited second-order artificial viscosity. Isn't there some way to share the code between our two implementations a bit more?
- The Lloyds algorithm added to the Yee vortext test -- should this be folded into our existing centroidal relaxation method?
Yeah I had the same thought when I put it together. In the end it seemed cleaner to me to do a reimplementation rather than modify GSPH to take AV or Riemann solver classes. The two classes differ in who owns the velocity gradient and the things they return to the hydro in the eval derivs. I was also leaning towards keeping the modification modular. Mind if we keep the somewhat redundant version in the GSPH module for now? I want to do a comparison study.
oo yeah i think they are same thing. I'll take it out |
There was a problem hiding this comment.
This is fine as is (keeping the AV distinct between GSPH and the other hydros). I'd like to revisit this to combine and remove as much redundancy as possible, but that can happen in a later change.
This should be OK to merge once the CI succeeds (currently failed for some reason).
Summary
ToDo :
RELEASE_NOTES.mdwith notable changes.