While learning to use bempp-cl, I encountered a number of problems. I have posted segmentation faults problem when running EM tutorial example. One response is that using opencl to fix the problem.
For large problem, say 500k unknowns or even more, dense matrix obviously is not a feasible option unless one has a machine with several hundreds GB RAM. I tried using FMM. It is very slow as a previous post already indicated. I am wondering if bempp-cl with FMM can solve > 500k problems.
Thank you,
Jiannan