We can benefit from parallelizing some of the internal operations to speed up the solve drastrically. Specifically 1. [ ] RAP 2. [ ] SpMV 3. [ ] L1 hybrid smoothers (L1HGS, L1 hybrid block Jacobi, L1HSOR) 4. [ ] HMIS or PMIS for RSAMG 5. [ ] "Parallel Smoothed Aggregation Multigrid : Aggregation Strategies on Massively Parallel Machines" for SAAMG The first two are related to the used matrix format, so one way ahead could be to provide a way to control which matrix format is internally used. Some relevant sources for this are https://www.osti.gov/servlets/purl/1860740 and maybe https://github.com/SINTEF-agentlab/Draugr.jl/ can serve as a source.
We can benefit from parallelizing some of the internal operations to speed up the solve drastrically. Specifically
The first two are related to the used matrix format, so one way ahead could be to provide a way to control which matrix format is internally used.
Some relevant sources for this are https://www.osti.gov/servlets/purl/1860740 and maybe https://github.com/SINTEF-agentlab/Draugr.jl/ can serve as a source.