I have a problem where I have a list of sparse square arrays (csc matrices) A = [A1,A2,...,An] and then a list of dense square arrays B = [B1,B2,...,Bn] I want to calculate the sum C = A1 B1 + A2 B2 + ... + An Bn. The sparse arrays are very sparse, and the sizes of the matrices involved can be up to a million by million. I would like some way to do this in parallel. However, everything I have tried so far has not been able to provide any speed-up. I know that numba is difficult to use, given that we are multiplying sparse with dense matrices. I would like some suggestions to write efficient parallel code for the above. Continue reading...
