tflovorn / polecalc Goto Github PK

Solve() in selfconsistent.go does nothing to check if the (left, right) returned by eq.Range() actually brackets a root. This needs to be fixed and expanded on; Solve() shouldn't break on multi-root functions.

1. There should be a standard interface for meshes.

2. How should it be organized? Could see speed improvements by passing multiple points at once on the output channel. Would need to profile to see whether this works better than passing a point at a time (as it it now).

Running zerotemp_test with NumProcs = 2 is about 5x slower than NumProcs = 1 (38.44s vs 7.86s) and gives very slightly different results. Why so much slower? What's causing it to give different values?

May just want to make the mesh calculations single-process. All the message-passing there seems kind of bulky and unnecessary.

Throwing everything in one package is kind of awful, but referencing other packages that aren't a part of the local Go installation is kind of painful. What do?

Potential package structure:
--> tools (everything not specific to one region of x/T phase space)
--> zero temp calcs
--> pair temp calcs
etc.