Zhi Chen

switched to evolving Ye using dyedt instead of updating Xn separately. Also added `nse_T_abar_from_e` to get consistent abar from nse_state. I was able to get reasonable results for detonation: ...

GPROF from detonation: ``` | time | seconds | seconds | calls | name | |-------+---------+---------+------------+----------------------------------| | 56.74 | 400.61 | 400.61 | 92644127 | void nse_hybrid_solver | | 24.48...

updated using `state.T` as initial guess decreases runtime from ~1200s to ~550s for detonation

seems like I can just use a central difference instead of the 5-stencil for dabardT and still get second order convergence, might just stick with that ``` | density |...

Current state of this pr: Its takes ~400s to run detonation with NSE and ~250s without NSE. (just run with USE_NSE_NET=TRUE and USE_SIMPLIFIED_SDC=TRUE with inputs-det-x.nse_net) Profile for detonation, without NSE:...

previously only ran up to 128 res because its slow, but I decided to run 256 res to check the convergence, and for whatever reason, its not good... ``` density...

I think this is fine too. I just copied what's there for P_radial and added a corresponding version for P_theta.

I think also update mom_flux_has_p and related stuff

also need to modify cfl_violation/timestep constraint since in spherical2d dy is just dtheta but we want physical length , so rdtheta, when checking and determining timestep

I think its a reasonable thing to do if we were to use these loosen-constraint parameters. But right now I tend to not use them because I think it would...