You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In the MOSEK paper, it states that the matrix M has a fixed nonzero pattern throughout the solution.
Scikit-sparse's interface to Cholmod can create a Factor object which stores the reordering information. . From my tests on some large problems (> 1e6 variables), I am getting a 2-3X speedup by storing this reordering. I would be willing to implement this.
A similar strategy can be used for the scikit-umfpack solver through the low-level interface.
The text was updated successfully, but these errors were encountered:
In the MOSEK paper, it states that the matrix M has a fixed nonzero pattern throughout the solution.
Scikit-sparse's interface to Cholmod can create a Factor object which stores the reordering information. . From my tests on some large problems (> 1e6 variables), I am getting a 2-3X speedup by storing this reordering. I would be willing to implement this.
A similar strategy can be used for the scikit-umfpack solver through the low-level interface.
The text was updated successfully, but these errors were encountered: