Should Sparse matrices have the BlackBox Trait ?

Skip to first unread message

Clement Pernet

unread,
Oct 6, 2015, 4:20:24 AM10/6/15
to linbox...@googlegroups.com
Hi,

There seems to be an issue with the way we treat sparse matrices:
In linbox/matrix/sparse-matrix.h line 94, I see that the MatrixContainerTrait for any sparse matrix
is set to "Blackbox" (the former value "Container" is commented out).

This implies that converting such a matrix to a DenseMatrix, will result in calling n matrix-vector
products (apply) to the canonical vectors, which also seems to be broken btw.
This is what happens when calling (solve(A,b,Method::Elimination)

Reverting the MatrixContainerTrait to Container, solves it (as it offers to the consrtuctor of a
dense matrix, the getEntry methods), but this change makes the test-rank-md test-rank-u32 fail on
the sparse matrix ELL_R.

I suggest to use the Container trait and report a bug on the test-rank suite.

Please comment if you disagree on this change.

Clément
Reply all
Reply to author
Forward
0 new messages