For a particular class of non-convex nonlinear programming problems, this FORTRAN implementation of FilterSLP uses sequential linear programming to find a (local) optimal solution.
This version dates from (at the latest) 4 July 2013.
- It won't compile with modern
f90(specificallygcc version 9.4.0 - Ubuntu 9.4.0-1ubuntu1~20.04) due to some FORTRAN usage that's now illegal.
Specifically, two logical statements /= in lines 224 and 229 of rd_prob_da.f can no longer be compiled
This includes
- Edits to allow it to be compiled with modern
f90. Specifically the uses of/=have been replaced by.neqv.
As such, this represents the "best" version of the FORTRAN implementation of FilterSLP that @jajhall can come up with
Note
- It calls EMSOL (see https://github.com/ERGO-Code/EMSOL)
- To compile may require the
-fno-range-checkflag to cope with the mechanism used inEMSV.INCto set the 32nd bit in a 32-bit integer - No version of the FORTRAN implementation of FilterSLP is supported in any way
- Anyone using a FORTRAN implementation of FilterSLP is strongly advised to continue using the version that they have!