

SHERMAN3       matrix filename -- HB boeing matrix 
               [matrix located in ../matrices] 

ras            class of parallel preconditioners used  
               -- Schur complement 
               -- Restrictive Additive Schwarz 
               -- Additive Schwarz [Block Jacobi]
            
arms           local preconditioner one of 
               ilu0, 
               iluk, 
               ilut, or 
               arms
               
0.01           eps -- tolerance for inner iteration if applicable
               Inner iteration is used only for last Schur complement
               system.. This is indicated by the parameter kim 
               (krylov subspace size for inner iteration) -- see below.

1.0e-6         eps1 : tolerance for outer iteration [stopping criterion]

5              nlev: Maximum number of levels for ARMS, ILUK
             
0              indicator for type of permutaions used in 
               arms: 
                   0 == symmetric permutation used
                   1 == nonsymmetric permutations [ddPQ] 

400000         bsize: block size for block independent sets
                
1.0            tolind: tolerance used for  independent set or 
               nonsymmetric permutations to enforce diagonal
               dominance on B-block. The larger tolind the 
               more diagonally dominant B -- but the smaller B.
               good values are anywhere between 0.1 and 1..

10             im  dimension of krylov subspace for outer iteration
               
400            maxits Maximum number of outer fgmres steps 

5              kim = dimension of krylov subspace for Schur complement
               system. If kim = 0, no inner iterations are performed --
               only one preconditioning operation is done. 

5              itsgmr = max. number of iterations for  Schur complement
               system. 

0              Type of permutations used for interlevel B blocks in arms:
                 -- 0 use symmetric permutations [indset] 
                 -- 1 use nonsymmetric permutations [ddPQ] 

0              permutations of columns for interlevel blocks (1:yes, 0:no)

0              row scaling for interlevel blocks (1:yes, 0:no)

0              column scaling for interlevel blocks (1:yes, 0:no)
               option unused!

1             nonsymmetric permutation for last Schur complement 
              whether or not to apply nonsymmetric reordering (ddPQ) for
              last-level Schur complement. 

1             column permutations for ILU in last Schur complement 
              1: use ILUTP 
              0: use ILUT. 
              [recommended : always use 1] 
    
1             row scaling for last Schur complement (1:yes, 0:no)

1             column scaling for last Schur complement (1:yes, 0:no)

10            lfil0(ilut, iluk and arms for lfil[0-3])

10            lfil4(schur)


10            lfil5(ILUT L, ILUT U)


0.001         droptol0(droptol[0=3], L, U, L^{-1}F, EU^{-1}

0.001          droptol4(for schur complements at each level)


0.001         droptol5(for ILUT in last level Schur Complement)
