AMPL > >Products > >Solvers > >BPMPD Directives for AMPL

BPMPD Directives for AMPL

To set these directives, assign a string specifying their values to the AMPL option bpmpd_options. For example:

ampl: option bpmpd_options 'feas1=1.0e-8 ordering=3 outlev=1';

See the the references given at the BPMPD Home Page for further information on the meaning of these directives.

bargrow     barrier growth bound
barmin      minimum barrier threshold
barset      barrier setup limit
bigbound    limit on finite bounds and ranges
center      centrality force
complimit   improve if compl. < complim
 comppar     improve if compl. < comppar * average compl.
 corstop     correction stop factor
 densgap     density gap param
 denslen     nonzeros in "dense" columns
 dinfs       minimum dual infeasibility reduction
 dslacklim   dual slack variable limit
 dudare      maximum dual steplength
 dumin       minimum initial slack value
 feas1       absolute primal feasibility tolerance
 feas2       absolute dual feasibility tolerance
 inibarr     use initial barrier param.
 lam         minimum value of lambda
 maxccorr    number of maximum corrections
 maxcorr     number of maximum corrections
 maxdense    maximum dense column rate
 maxiter     iteration limit
 maxref      maximum refinements
 maxsnz      maximum nonzeros in one supernode
 memadd      increment (default 0) for scratch array length
 memfac      factor (default 2) for scratch array length
 minccorr    number of minimum corrections
 mincorr     number of minimum corrections
 mindiff     minimum norm diff. to exit scaling
 minmax      1 = minimize, -1 = maximize
 objno       objective number: 0 = none, 1 (default) = first
 objnorm     scale objective maxnorm to this value
 ordering    pivot ordering for solving linear systems:
                 1 ==> minimum degree
                2 ==> minimum fill (default)
                3 ==> minimum fill in augmented system
outlev      1 = print iteration summary, etc.
parfile     name of BPMPD parameter file to read now
pinfs       minimum primal infeasibility reduction
prdare      maximum primal steplength
presolv     presolve methods: default = 511 (all but extended dual test)
prmin       minimum initial variable value
psupnode    primary supernode length
regularize  1 = introduce dummy ranges
rhsnorm     scale RHS maxnorm to this value
safemet     safe method: (1, 2, or 3); default -3
setlam      0 = default; 1, -1 = alternate values
smethod     starting method (1 or 2)
smethod1    scaling method before aggregator
smethod2    scaling method after aggregator
spasses1    maximum passes (< 128) before aggregator
 spasses2    maximum passes (< 128) after aggregator
 ssupnode    secondary supernode length
 stopcor     correction stop parameter
 supdens     super-dense column length
 tabs        absolute pivot tolerance in first factorization
 target      trial steplength improvement
 tfeas1      relative primal feasibility tolerance
 tfeas2      relative dual feasibility tolerance
 tfind       pivot search loop count
 tfixslack   slack reset param.
 tfixvar     variable reset param.
 tlarge      large complementarity bound
 topt1       relative duality gap tolerance
 topt2       average complementarity gap tolerance
 tpiv1       first threshold pivot tolerance
 tpiv2       second threshold pivot tolerance
 tplus       relative addition tolerance
 trabs       absolute pivot tolerance during the algorithm
 tresx       primal residual tolerance
 tresy       dual residual tolerance
 tsdir       search dir. maximum norm tolerance
 tsmall      small complementarity bound
 tzer        relative zero tolerance
 upmax       maximum initial variable value
 wantsol     solution report without -AMPL: sum of
                 1 ==> write .sol file
                2 ==> print primal variable values
                4 ==> print dual variable values
                8 ==> do not print solution message