[SciPy-dev] Tri-diagonal LAPACK Routines - Shall I interface them?

[Nils Wagner]

On Sun, 6 Dec 2009 18:31:54 -0800
  David Goldsmith <d.l.goldsmith@gmail.com> wrote:
> On Sun, Dec 6, 2009 at 6:13 PM, Nathan Bell 
><wnbell@gmail.com> wrote:
> 
>> On Wed, Dec 2, 2009 at 3:43 AM, Benny Malengier
>> <benny.malengier@gmail.com> wrote:
>> >
>> > Interesting, this is exactly the function I needed for 
>>my problem, but
>> > I was looking in scipy.sparse.linalg, so did not 
>>notice banded matrix
>> > solver was present in scipy.linalg.
>> >
>> > In my logic, the "matrix diagonal orded form" of
>> >
>> http://docs.scipy.org/doc/scipy/reference/generated/scipy.linalg.solve_banded.html#scipy.linalg.solve_banded
>> > would be a type of sparse matrix one can manipulate. 
>>This would allow
>> > things like changing matrix diagonal orded form sparse 
>>matrix to a csr
>> > matrix, adding some extra elements off the diagonals, 
>>and then calling
>> > a more generic solver.
>> >
>>
>> Can't you do that already with scipy.sparse.dia_matrix? 
>> If I'm not
>> mistaken, dia_matrix is (slightly) more general than the 
>>banded format
>> but similarly efficient.
>>
>> In an ideal world scipy.sparse.spsolve() would detect 
>>the case that A
>> was a dia_matrix (with small bandwidth) and invoke the 
>>LAPACK method
>> in scipy.linalg instead of using the general sparse LU 
>>solver.
>>
> 
> Perhaps you might file this as an "enhancement" "issue"?
> 
What do you make of it ?

http://projects.scipy.org/scipy/ticket/456

Nils