@mikera: is there any way to get parallelization in core.matrix?
Yes but approach depends on use case
use case: you convinced me to finally try out my swarm simulations running on core.matrix
Parallelised implementations are certainly possible
none exist yet though?
you might be pleased that the vectorz-clj implementation beats clatrix (and this is without an underlying KD-tree implementation)
Not to my knowledge. My experience is that you are better off running independent threads rather than a more complex implementation that tries to use multiple threads
interestingly the vectorz-clj implementation without KD-trees is about as fast as the non-matrix-based version with KD-trees
Clatrix is not great for perfornance
so it seems
It could be, but needs a lot more work on the different operations
alright, i think i’ll try for a hybrid core.matrix+KDtree version next then
thank you!