Ability to get inertia from LDL factorization

[ohinder]

It would be useful if it was possible to get the inertia of the factorization (the number of positive and negative values in D which is equivalent to the number of positive and negative eigenvalues in the original matrix https://en.wikipedia.org/wiki/Sylvester%27s_law_of_inertia). It is commonly used in optimization to determine if a problem is convex.

[lruthotto]

thanks for this suggestion. I'd also like to get more control over the MUMPS package. However, right now I see no time and easy way to fix this myself. Do you have any idea?

Meanwhile, I would also like to point you to https://github.com/dpo/MUMPS.jl/ which provides a much more elegant way to access MUMPS from julia and should probably allow to extract the inertia.

[ohinder]

Thanks, I have posted a issue on https://github.com/dpo/MUMPS.jl/.
I am not sure how do this at the moment - I know it is possible, because some C++ codes do this. If I get time in the next few months I will attempt to solve this problem.