Sparse Cholesky of Gram Matrix (SuiteSparse?)

[guille]

I’m curious as well. I discarded this “expanded system” solution some time ago since the matrices are quite large, but I can definitely give it a try to see how quickly the problem is solved!

@dpo

Note that the HSL codes are typically only free for academia. MUMPS is open source and can use MPI if your systems are that large. If you have access to it, I find that HSL_MA57 is often the fastest. You can store the factorization and reuse it to solve for multiple right-hand sides.

I actually haven’t tried HSL at all for this (I actually forgot it existed…), so I will do that at some point in the near future. Thanks for the heads up!