Iter0 Warmstarts

[bknueven]

#331 fixed a bug when warm-starting PH the initial typical trivial bound would be incorrect if the proximal term was active. To fix this, the warmstart wasn't applied until iteration 1 instead, but this meant the iteration 0 solve was wasted effort.

This PR fixes the same bug but preserves the warmstart being applied at iteration 0 by checking if the proximal term was enabled before reporting the iteration 0 "bound". We also clean up some redundant checking for the best_bound_obj_val.

[DLWoodruff]

This was always an issue with reading W, it just seems more stark now that we are doing it pre-iter0: should we check to see if the W is valid? (i.e. E[W] = 0)

[bknueven]

This was always an issue with reading W, it just seems more stark now that we are doing it pre-iter0: should we check to see if the W is valid? (i.e. E[W] = 0)

It looks like we already check by default:

mpi-sppy/mpisppy/utils/wxbarutils.py

Line 93 in a791d8d

def set_W_from_file(fname, PHB, rank, sep_files=False, disable_check=False):