[CSC 435] Preconditioning

[Andrew J. Pounds]

Steve -- (and all).   I looked back at the reference you sent me and 
they are doing "standard" GMRES condintioning.  I am fine with you using 
this method.  The only issue is that we will need to all agree on it and 
also modify our function arguments to include the array b in Ax=b.

The SGEBAL method will improve the accuracy of the eigenvalues of your 
sparse system.  The Krylov method I sent out earlier does something 
similar for large sparse matrices.  Now, through some more operations, 
you can use those results to also improve your linear system solution.  
If, however, you would prefer a more direct -- a.k.a - GMRES method that 
skips the ancillary linear algebra operations, then I wholeheartedly 
approve.


On 02/13/14 16:30, Steve Hussung wrote:
> Hello Dr. Pounds,
>
> While figuring out preconditioning, I ran across several articles 
> similar to 
> http://www.cs.ucdavis.edu/~bai/ECS231/returnsfinal/Fuentes.pdf 
> <http://www.cs.ucdavis.edu/%7Ebai/ECS231/returnsfinal/Fuentes.pdf>
>
> They all describe preconditioning as something which is done to both 
> the matrix A and the vector b for a linear system of equations Ax=b. 
> Our method is required to take in only the matrix A. I have looked in 
> our primary parallel programming textbook, and I didn't see anything 
> under preconditioning, conditioning, or matrix that looked relevant.
>
> Any ideas? I'm not sure where else to look.
>
> Thanks
> Steve Hussung


-- 
Andrew J. Pounds, Ph.D.  (pounds_aj at mercer.edu)
Professor of Chemistry and Computer Science
Mercer University,  Macon, GA 31207   (478) 301-5627
http://faculty.mercer.edu/pounds_aj