About myself: I started working with Fortran back in 2013, but I have always worked with “simple” codes so I don’t have an in-depth knowledge of it (where by simple I mean codes with no mean to be extremely efficient). My background varied: B.Sc. in Civil Engineering (IT), M.Sc. in Mathematical Engineering (IT), M2 master in Environmental fluid mechanics (FR), PhD in stochastic sub-grid modeling for ocean models (FR). During my Bachelor thesis I have worked on preconditioners for saddle-point problems and in my Master I have worked more on Krylov based solvers, preconditioners and their parallelization.
While knowing I can be helpful in this project, I have to say that this is the first time joining a project of this level of complexity, but I am eager to learn.
Thank you for your interest @kimala!
I’m not expert in the GSoC selection procedure, I would ask @gnikit@certik to correct this if wrong.
I believe applicants are usually requested to open a PR that is relevant to the project they want to be involved in. You can look at open issues and PRs:
Not an expert in the process either, but I get from here that the full application (with the PR) should be finalized by April 2, from what I understood here?
That leaves time for @kimala to skim through the list of ideas and propose a first PR.
If @kimala and the rest of mentors are ok, I would propose to finalize having some of the different sparse kinds as DTs integrated in stdlib, even if not with all the methods.
That way we can then build on top APIs that have a similar signature for dense matrices following your work @FedericoPerini that take as input a DT for the given sparse type. Any thoughts?