## 09/25/1999 ##########################################################
#                                                                      #
#  The directory blzpack/drv.                                          #
#                                                                      #
########################################################################

   The current directory is organized as follows:

   subdirectory data:
   -----------------

   A.dat      : the upper triangle, in coordinate format (row,column,
                value), of a symmetric test matrix of dimension 10.
   A.eig      : eigenvectors (x) and corresponding eigenvalues (eig) 
                computed with Matlab for the matrix given in the 
                file `A.dat'.
   A.eig.dat  : eigenvectors (x) and corresponding eigenvalues (eig) 
                computed with Matlab for the matrix given in the 
                file `A.dat'.
   AB1.eig    : eigenvectors (x) and corresponding eigenvalues (eig) 
                computed with Matlab for (A)*(x)-eig*(B)*(x)=(0),
                where (A) corresponds to `A.dat' and (B) to
               `B1.dat'.
   AB2.eig.1  : eigenvectors (x) and corresponding eigenvalues (eig) 
                computed with Matlab for (A)*(x)-eig*(B)*(x)=(0),
                where (A) corresponds to `A.dat' and (B) to
                `B2.dat'.
   AB2.eig.2  : eigenvectors (x) and corresponding eigenvalues (eig) 
                computed with Matlab for (A)*(x)-eig*(B)*(x)=(0) in
                buckling mode, where (A) corresponds to `A.dat' and
                (B) to `B2.dat'.
   Ainv.dat   : the upper triangle, in coordinate format (row,column,
                value), of the inverse of the matrix given in the 
                file `A.dat'.
   Ainv.eig   : eigenvectors (x) and corresponding eigenvalues (eig) 
                computed with Matlab for the matrix given in the file
                `Ainv.dat' (the eigenvalues are just the reciprocals
                of those given in `A.eig', while the eigenvectors
                are the same).
   B1.dat     : diagonal matrix of dimension 10 to test `drvgp2.f'.
   B2.dat     : the upper triangle, in coordinate format (row,column,
                value), of a matrix of dimension 10 with the same
                pattern of `A.dat' to test `drvgp3.f'.
   C1.dat     : a 10-by-12 matrix stored in coordinate format (row,
                column,value). It can be used to test `drvsp3.f'.
   C1.svd     : singular value decomposition of the matrix given in
                the file `C1.dat' (that is what `drvsp3.f' does).
   C2.dat     : a 12-by-10 matrix stored in coordinate format (row,
                column,value). It can be used to test `drvsp4.f'.
   C2.svd     : singular value decomposition of the matrix given in
                the file `C2.dat' (that is what `drvsp4.f' does).
   Makefile   : creates executables and runs tests
   drvgp1.dat : data for drvgp1.f  
   drvgp2.dat : data for drvgp2.f  
   drvgp3.dat : data for drvgp3.f  
   drvgp4.dat : data for drvgp4.f  
   drvsp1.dat : data for drvsp1.f, drvmpi.f and drvpvm.f
   drvsp2.dat : data for drvsp2.f  
   drvsp2.eig : data for drvsp2.f  
   drvsp3.dat : data for drvsp3.f  
   drvsp4.dat : data for drvsp4.f  

   subdirectories double/single:
   ----------------------------

   drvgp1.f : driver model 1 for (A)*(x)-eig*(B)*(x)=(0). It uses the
              routine MA47, which is part of the Harwell Subroutine
              Library, to factor matrices and solve systems of linear
              equations. (B) is set to the identity matrix.
   drvgp2.f : driver model 2 for (A)*(x)-eig*(B)*(x)=(0). It uses the
              routine MA47, which is part of the Harwell Subroutine
              Library, to factor matrices and solve systems of linear
              equations. (B) is assumed to be a diagonal matrix.
   drvgp3.f : driver model 3 for (A)*(x)-eig*(B)*(x)=(0). It uses the
              routine MA47, which is part of the Harwell Subroutine
              Library, to factor matrices and solve systems of linear
              equations. (B) is assumed to have the pattern of (A).
   drvgp4.f : driver model 4 for (A)*(x)-eig*(B)*(x)=(0) in buckling
              mode. It uses the routine MA47, which is part of the
              Harwell Subroutine Library, to factor matrices and solve
              systems of linear equations. (B) is assumed to have the
              pattern of (A).
   drvmpi.f : a parallel version of drvsp1.f using MPI.
   drvpvm.f : a parallel version of drvsp1.f using PVM.
   drvsp1.f : driver model 1 for (A)*(x)-eig*(x)=(0). It uses a BLAS 
              kernel to perform matrix-vector products.
   drvsp2.f : driver model 2 for (A)*(x)-eig*(x)=(0). It uses a sparse
              matrix-vector multiplication.
   drvsp3.f : driver model 3 for (A)*(x)-eig*(x)=(0) where (A) is equal
              to (C)^T*(C) (that is the SVD of (C)). It uses a sparse
              matrix-vector multiplication.
   drvsp4.f : driver model 4 for (A)*(x)-eig*(x)=(0) where (A) is an 
              augmented matrix [(0) (C); (C)^T (0)] (that is the SVD
              of (C)). It uses a sparse matrix-vector multiplication.
   prteig.f : subroutine to store EIG and X (which are outputs of 
              BLZDRS/D) in two separate files, `EIG' and `X'.

end of README ##########################################################
