SPPARMS

 

 Set parameters for sparse matrix routines.

 

 

  SPPARMS('key',value) sets one or more of the "tunable" parameters 

  used in the sparse linear equation operators, \ and /, and the

  minimum degree orderings, COLMMD27BOBX and SYMMMD3MKDHXL.

 

 

  SPPARMS, by itself, prints a description of the current settings.

 

 

  If no input argument is present, values = SPPARMS returns a

  vector whose components give the current settings.

  [keys,values] = SPPARMS returns that vector, and also returns

  a character matrix whose rows are the keywords for the parameters.

 

 

  SPPARMS(values), with no output argument, sets all the parameters

  to the values specified by the argument vector.

 

 

  value = SPPARMS('key') returns the current setting of one parameter.

 

 

  SPPARMS('default') sets all the parameters to their default settings.

  SPPARMS('tight') sets the minimum degree ordering parameters to their 

  "tight" settings, which may lead to orderings with less fill-in, but 

  which makes the ordering functions themselves use more execution time.

 

 

  The parameters with the default and "tight" values are:

 

 

                   keyword       default       tight

 

 

     values(1)     'spumoni'      0

     values(2)     'thr_rel'      1.1          1.0

     values(3)     'thr_abs'      1.0          0.0

     values(4)     'exact_d'      0            1

     values(5)     'supernd'      3            1

     values(6)     'rreduce'      3            1

     values(7)     'wh_frac'      0.5          0.5

     values(8)     'autommd'      1            

     values(9)     'aug_rel'      0.001      

     values(10)    'aug_abs'      0   

 

 

  The meanings of the parameters are

 

 

     spumoni:  The Sparse Monitor Flag controls diagnostic output;

               0 means none, 1 means some, 2 means too much.

     thr_rel,

     thr_abs:  Minimum degree threshold is thr_rel*mindegree + thr_abs.

     exact_d:  Nonzero to use exact degrees in minimum degree,

               Zero to use approximate degrees.

     supernd:  If > 0, MMD amalgamates supernodes every supernd stages.

     rreduce:  If > 0, MMD does row reduction every rreduce stages.

     wh_frac:  Rows with density > wh_frac are ignored in COLMMD27BOBX.

     autommd:  Nonzero to use minimum degree orderings with \ and /.

     aug_rel,

     aug_abs: Residual scaling parameter for augmented equations is

              aug_rel*max(max(abs(A))) + aug_abs.

              For example, aug_rel = 0, aug_abs = 1 puts an unscaled

              identity matrix in the (1,1) block of the augmented matrix.

                 

  See also COLMMD27BOBX, SYMMMD3MKDHXL, SPAUGMENT1JMRBSK.