Newsgroups: comp.parallel.mpi
From: hym@tcsm.NoSubdomain.NoDomain (H.Y. Mussa)
Subject: looking for STRUCTURED SPARSE MATRIX  PARALLEL EIGENSOLVER
Keywords: parallel eigensolver  for sparse matrix
Organization: Daresbury Lab, Warrington, U.K.
Date: 28 Jun 1996 07:34:07 GMT
Message-ID: <4r01tf$5q5@mserv1.dl.ac.uk>


Hi!
I am looking  for  a parallel Eigensolver to
diagonalise a real sparse symmertic  matrix.
 
The  matrix(diagonally dominant) consists of :
a) a vector  containing the diagonal elements.
b) And an array  containing blocks(blocks do not
have to be of the same size).
 
 
I  need 10% of the eigenvectors and  as many as
possible of the eigenvalues.
 
 
 
Here  is   a diagrammatical  picture of  what the
matrix  looks like.
 
 
 
 
    *
    . *
    ....*
    ......*
    333333333
    3*******3.*
    3*******3...*
    3*******3.....*
    33333333333333333
    3*******3*******3.*
    3*******3*******3...*
    3*******3*******3.....*
    3333333333333333333333333
    ........3*******3*******3.*
    ........3*******3*******3...*
    ........3*******3*******3.....*
    ........3333333333333333333333333
    ................3*******3*******3.*
    .......... .....3*******3*******3...*
    ................3*******3*******3.....*
    ................3333333333333333333333333
    ........................3*******3*******3.*
    ........................3*******3*******3...*
    ...................... .3*******3*******3......*
    ........................33333333333333333........*
 
And  so  on.
 
KEY:
 
 3 and * = non zero elements.
 .       = zero elements.

   
  333333333
  3*******3
  3*******3 - size   would be  1000*1000.
  3*******3
  333333333
 
Number  of blocks  = 50 to 200
 
 
 
 
 
 
 
I would be grateful  for pointers  to  either  programs or  a
suitable  algorithms.
Thank you    for  any  help  you  could offer.
 
Hamse  Mussa (e-mail hym@tcsl.dl.ac.uk)
 

