> I encountered a situation where I got into an infinite loop in the > routine spx_simplex in the module glpspx2.c. The call to the > function spx_prim_feas located shortly after the feas: label > returned LPX_E_OK and control passed to the label opt:. The call to > the routine spx_prim_opt returned LPX_E_INSTAB which transferred > control back to label feas:. This sequence then repeated without > end.
> I changed the case statement after the spx_prim_opt call such that > the LPX_E_INSTAB return goes to label done: instead of feas:. glp_simplex switches to the phase I if the current basic solution becomes primal infeasible within a tolerance, that sometimes may lead to an infinite loop. This is a lack of implementation. I hope to finish soon a new, more robust and faster simplex code, which will be included in glpk as a basic lp solver. Andrew Makhorin _______________________________________________ Bug-glpk mailing list Bug-glpk@gnu.org http://lists.gnu.org/mailman/listinfo/bug-glpk