num::band_mat¶

Defined in header <linear/banded/banded.hpp>

class band_mat;

LAPACK-style band storage.

Stores \(A_{ij}\) at \(\text{band}(k_l+k_u+i-j,j)\) when \(\max(0,j-k_u)\le i\le \min(n-1,j+k_l)\).

Members¶

member	description
`band_mat(idx n, idx kl, idx ku)`	Construct a zero-filled n-by-n matrix with lower/upper bandwidths kl/ku.
`band_mat(idx n, idx kl, idx ku, real val)`	Construct a banded matrix with every stored entry initialized to val.
`band_mat(const band_mat &)`
`band_mat(band_mat &&) noexcept`
`band_mat &operator=(const band_mat &)`
`band_mat &operator=(band_mat &&) noexcept`
`idx size() const`	Return the square matrix order.
`idx rows() const`
`idx cols() const`
`idx kl() const`	Return the lower bandwidth.
`idx ku() const`	Return the upper bandwidth.
`idx bandwidth() const`	Return the number of mathematical diagonals in the band.
`idx ldab() const`	Return the leading dimension of the LAPACK-compatible storage.
`real &operator()(idx i, idx j)`	Access a mathematical matrix entry inside the stored band.
`real operator()(idx i, idx j) const`
`real &band(idx band_row, idx col)`	Access an entry by physical band-storage coordinates.
`real band(idx band_row, idx col) const`
`real *data()`
`const real *data() const`
`bool in_band(idx i, idx j) const`	Test whether mathematical entry (i,j) is explicitly stored.
`void to_gpu()`
`void to_cpu()`
`real *gpu_data()`
`const real *gpu_data() const`
`bool on_gpu() const`