Enable Gemm (local) to be used with MatrixRef

albestro commented 1 year ago

This PR partially address #915 (just local variant)

gemm (local) implementation working with MatrixRef (only, at the moment) with related tests for both full and sub-matrix test-cases
introduces scal tile extension
basic refactoring(=adaptation) of local_matrix and multipliable assert + introduction of same_process_grid
refactor test helper for sub-matrices (subValues and mixValues)
minor: MatrixBase string representation also print information about source_rank_index and offset

TODO

[x] Check preconditions
[x] Extend the test
[x] Move the new function in the right place (see following note)
[x] Evaluate how to deal with multipliable assertion (separate it from algorithm details)
[x] Test also with ROCm (mainly if scal workaround of using gemm with nullptr works there too)

Note: GeneralSub was intended to work on submatrices, while now the sub-matrix concept is hidden by MatrixRef. For this reason, the new gemm helper might just be considered a full general multiplication, and it might be moved in multiplication::general::General (or something like that).

msimberg commented 1 year ago

This looks good to me for the current use cases. I see now what the main difficulty would be with relaxing constraints on the input matrices: ETI. I don't know what the best way around this is. Inheritance works, but if possible I'd rather not go there. We might have to try out and see how bad the compile times would be from exposing the implementations again to allow arbitrary instantiations.