Closed maltezfaria closed 3 months ago
The decision was to:
SparseArrays
, at least for now (no extension)HMatrix
does not check that the sparse entries are inside a dense block, as that cannot be guaranteed even in the BIE example discussed above. It does check, however, that after the multiplication the memory of the resulting matrix is not much greater than before.
In some boundary integral equation applications, it makes sense to add a sparse correction matrix to an
HMatrix
approximation. In those cases, as it turns out, the sparse matrix only has non-zero entries precisely where theHMatrix
has dense blocks, so adding the two should be rather simple and cost-effective (i.e. no increase in the rank).SparseArrays
is loaded, theaxpy!
oraxpby!
function is defined to add a sparse matrix in-place to anHMatrix
HMatrix
is low-rank, decide what to do