Uncapacitated Facility Location Problem solver

I. Building

In order to build the code (by invoking 'make all'), you'll need to get the source for BTT from Antonio Frangioni (frangio@di.unipi.it) and replace the appropriate files in the BTT subdirectory.

Why not include BTT?
Because it's freely available only under a weird academic license.
Why not use another non-differentiable optimizer instead of BTT?
Because it's the best one out there and I don't have the expertise or the motivation to write a better one myself (seriously looked into it, though).
What's this other stuff in the subdirectory?
It's a C99 wrapper for BTT written by Paul-Virak Khuong, a friend and colleague who had to kindness to share it with me (thanks Paul!). To put it mildly, BTT is not easily amenable to integration with other programs.

Execution performance has been found to be slightly sensitive to the compiler used, and as of now we recommend Clang over GCC. This recommendation may already be out of date as both compilers are now being constantly improved.

II. Running

Basically, the executable runs the following main loop:

read instance size from stdin
(optional) read resolution parameters from stdin
read instance prices from stdin
solve the UFLP, printing progress information in stderr
print best known solution in stdout
read 'quit' from stdin or go back to 1.

This loop structure exists because the UFLP often occurs as a subproblem within a larger optimization algorithm, in which the prices differ often only slightly from one iteration to the next.

Resolution parameters may also be provided as command-line arguments. A detailed description can be obtained by invoking 'solver --help'.

Example: to solve an instance 'cap_a' stored as './instances/beasley/cap74.bz2' with a time limit of 10 seconds, invoke the following command at a bash prompt: echo bzcat ./instances/beasley/cap74.bz2 quit | ./solver -n cap_74 -t 10

noInstanceName/uflp-solver

Uncapacitated Facility Location Problem solver