[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Higher precision arithmetic ?
|
From: |
Rhys Ulerich |
|
Subject: |
Re: Higher precision arithmetic ? |
|
Date: |
Thu, 8 Nov 2012 09:06:06 -0600 |
>> Is there some way to get more resolution ? I.e. being able to compute the
>> inverse of matrices whose the ratio between the greatest and the smallest
>> eigenvalues module span 12 decades ?
> Perhaps we should add iterative refinement to the solver - it's cheap and it
> looks like it would be easy.
>
> ...
>
> I haven't played with lapack's refinement stuff on a really hard
> problem so I'm curious to see how it does.
Playing with LAPACK's ?GESVX routines, e.g.
http://www.netlib.no/netlib/lapack/double/dgesvx.f would provide both
row and column rescaling in addition to performing iterative
refinement. It sounds the the former may also be important for this
problem.
- Rhys