Add block matrix generators
Broke down the P2P and QBX matrix generators even further to be able to compute random blocks.
A few remaining issues:
-
How to do the split_inameinget_optimized_kernelin when using blocks? -
There seems to be some fundamental difference in passing lp.make_kernela list of domains or just one and I'm not sure I got it right. -
At this point, the SingleSrcTgtListP2PBaseclass is so broken up it probably makes more sense to merge it withP2PComputationBaseand implement the current classes andP2PFromCSRon top of it. Any thoughts on that?
Edited by Alexandru Fikl