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
I have a banded block banded matrix where the blocks themselves can have different bandwidths. Specifically, the bandwidth within the blocks is dependent on which block-diagonal the block is in. e.g., the block at index k ∈ [-l, u] would have a bandwidth of (λ, μ) = (1 - k, 1 + k), so that a block on the subdiagonal k=-1 has a bandwidth of (2, 0), a block on the diagonal k=0 is tridiagonal, and a block on the superdiagonal k=1 has a bandwidth of (0, 2). Can this be represented using this package?
The text was updated successfully, but these errors were encountered:
Do you anticipate doing so would cause any performance regressions? Otherwise, perhaps it would be better to add a new subtype of AbstractBandedBlockBandedMatrix.
No I don’t think there are any performance regressions since it doesn’t really use the fact that all the bandwidths are the same. and if there was anywhere that made a difference we could just overload for the case where the bandwidths are <: Fill
Some implementation details will need to be changed though.
I have a banded block banded matrix where the blocks themselves can have different bandwidths. Specifically, the bandwidth within the blocks is dependent on which block-diagonal the block is in. e.g., the block at index
k ∈ [-l, u]
would have a bandwidth of(λ, μ) = (1 - k, 1 + k)
, so that a block on the subdiagonalk=-1
has a bandwidth of(2, 0)
, a block on the diagonalk=0
is tridiagonal, and a block on the superdiagonalk=1
has a bandwidth of(0, 2)
. Can this be represented using this package?The text was updated successfully, but these errors were encountered: