Closed PetrKryslUCSD closed 2 years ago
Issue described here: https://discourse.julialang.org/t/is-suitesparse-cholmod-ldlt-working-correctly/78951
For a sparse matrix (CSC) ldlt is extremely slow (~170 slower than cholesky). I can't believe that the problem is fallback to dense linear algebra: the matrix is too large for this to go through. But go through it does: correct solution is returned.
I believe the ldlt is a low-performance "teaching" piece of code. Only cholesky from Cholmod actually performs well.
Issue described here: https://discourse.julialang.org/t/is-suitesparse-cholmod-ldlt-working-correctly/78951
For a sparse matrix (CSC) ldlt is extremely slow (~170 slower than cholesky). I can't believe that the problem is fallback to dense linear algebra: the matrix is too large for this to go through. But go through it does: correct solution is returned.