Open henricasanova opened 6 years ago
We could maybe pick something in https://github.com/simgrid/SMPI-proxy-apps/ ?
What do you mean by a compelling application ? And what it the pedagogical goal ? If you want a real scientific app to motivate this application I know Frédéric Valentin at LNCC is developping a Galerkin enriched finite element method (MGEM) that exhibits exactly this structure followed by calls to MUMPS (the whole thing being wrapped in an outer loop). If you want to play with various scheduling load balancing strategies in a heterogeneous setup, I don't think you need anything like this. The simpler the better as it will allow students to easily propose and implement new techniques...
The pedagogical goal is exactly what you say at the end: "play with various scheduling load balancing strategies in a heterogeneous setup". And by compelling I mean some "fake" example as a scenario that is somewhat grounded in reality, not at all a real application. So just some idea like "imagine you have a bunch of images to process", but perhaps a little bit more concrete. I don't think it's difficult, but putting it as a github issue may cause us to actually do it (the Martin strategy for better productivity)
We should add a Master-Worker project as I find students oddly confused but it. It's easy, but we would need a reasonable compelling application...