Closed kostrzewa closed 2 years ago
@urbach @martin-ueding @pittlerf @marcus-petschlies anybody willing to work on this?
@urbach @martin-ueding @pittlerf @marcus-petschlies anybody willing to work on this?
Important topic, however, given my recent failure: it does not seem realistic for me timewise to work on this in a reasonable timeframe.
To help with strong scaling, pipelining should be exposed as a setting for solvers. The details of how to do this in a fine-grained enough fashion to be useful are a bit complicated I think. See https://github.com/lattice/quda/issues/692 for some hints.