Hello,

first of all, thank you for Eigen, I really like it and I'm going to include it into my programs (scientific stuff).

I have some questions on sparse matrices:

(i) generally, do you plan to work further on the sparse components (make it stable). if yes, what is planned? can it be expected that the sparse matrix component behaves similar in perfomance than the dense component (which as I've learned outperforms them all ) ?


(ii) I have a rather large sparse matrix that is filled in random order, so I use DynamicSparseMatrix. Is there a way to copy it to a 'normal' SparseMatrix without using the constructor? Is the storage needed twice as big, or does it copy row-by-row / column-by-column.

(iii) The sparse matrix is symmetric. Can I exploit that somehow in hte SparseMat-DenseVec multiplication? I tried SymmetricView, but got sometimes compiler errors (I specify on this when needed). Is the symmetric multiplication expected to be faster for sparse matrices?

Best regards,
David

(i) yes, betting a stable Sparse module my first priority regarding Eigen.
(ii) operator = works too but note that DynamicSparseMatrix will be deprecated
(iii) there is n perf. advantage in using a selfadjpintView but this allow you to store only one half.

Thank you for the answer.

when copying from DynamicSparseMatrix to SparseMatrix, does one need twice the storage?
Can you estimate when the stable version will be ready?
what will replace DynamicSparseMatrix?