Kevin Anderson

Hmm, another idea is to never explicitly invert the matrix at all? For large simulations I see 10-15% faster calls to `engine.run_full_mode` by replacing this: https://github.com/SunPower/pvfactors/blob/e0ea9d5a03d130d25919b39d590645ea2997ad14/pvfactors/engine.py#L223-L226 with `q0 = np.linalg.solve(a_mat,...

Some more progress, this time by combining the two previous ideas: use a sparse linear solver instead of explicitly inverting the matrix. I want to do some more testing to...

After continued efforts to speed up pvfactors (as always, collaborating with @spaneja), I think the timing interpretation I posted above is not wholly correct. tl;dr: I think just using a...

I suppose this is ready for review! Open questions: - The private sparse solve function seems a bit out of place in `engine.py`; should it live somewhere else? - Should...

> If you can permute the matrix so that the non-zero elements are close to the diagonal axis you can further improve the computation simplicity. Although I believe most sparse...

Passing empty timeseries inputs currently works on master but didn't work here; 480b350 fixes that and adds a relevant test. Not sure how valuable it is to support that use...

Ping @campanelli-sunpower -- these changes would be nice to have in the main branch for a project @spaneja and I are doing, any chance this could get merged soonish?

Putting this back to draft -- after further experimentation I'm no longer confident that this PR is the best approach. I will continue the discussion back in #134.

Oh I see, that makes sense. We had the same dilemma with rdtools and solved it using [nbsphinx-link](https://nbsphinx-link.readthedocs.io/en/latest/). It works by putting small shim files in a place that sphinx...

I'm a big fan of automated jobs that deploy docs, upload distributions to PyPI, etc on new releases. It's pretty easy to set up that kind of thing with GitHub...