[SciPy-dev] the current state and future directions of the sparse solvers

[Viral Shah]

Hi Robert,

> Hi Nathan,
>
> Nathan Bell wrote:
>>> OK, that sounds very good. I am going to study how to start such a
>>> scikit and we are going to move our additional solvers we have in
>>> sfepy to it (pysparse, soon petsc4py) -- are you ok with that  
>>> Robert?
>>> :) Plus umfpack.  Plus the eigensolvers, like primme, blzpack and
>>> others.
>
> Sure!

I agree its good to have more solvers. Tim Davis' SuiteSparse has a  
whole bunch, and they are the most robust direct sparse solvers out  
there..

>> I just recently saw a talk by some people that work on providing
>> abstract interfaces to the solvers in Trilinos:
>> http://trilinos.sandia.gov/packages/docs/r8.0/packages/stratimikos/doc/html/index.html
>> http://trilinos.sandia.gov/packages/belos/
>>
>> I have no experience with these, but they may provide some guidance.

Trilinos is good idea. By itself, its a huge collection. The relevant  
piece is Amesos for direct solvers. It makes plugging in various  
sparse direct solvers effortless.  Cholmod for sparse cholesky seems  
to not exist yet.

I am wondering how to find out more about the state of iterative  
solvers in scipy. Trilinos has all the methods and preconditioners.  
However, it should be possible to write the iterative solvers and  
preconditioners  completely in .py. For the standard stuff, perhaps  
compiled libraries could be used for speed, but the flexibility is  
nice to have.

The last time I had looked, they didn't have support for complex  
sparse matrices. I don't know if scipy sparse supports complex  
matrices or not yet. But if you want to solve non-symmetric eigenvalue  
problems, you need that..

-viral