README.code

Build instructions
==============

dualquery requires OCaml, findlib, and a unix system to build.

It is highly recommended to use a 64 bits machine, as 32bit OCaml will
crash on arrays larger than 2 million entries.

To build dualquery, first configure parmap:

$ cd parmap
$ ./configure
$ cd ..

then, make should do the job:

$ make

you should have a rundq executable.

Running dualquery
==============

dualquery requires an optimizing constraint solver.

In this release we support CPlex 12.5.1.

1.- Edit the file "cplex" and make sure that it points to where your
    cplex executable is.

2.- You can adjust the cplex parameters in the cplex.ml file.
    Typical parameters are:

set workmem 11000
set mip strategy file 2

3.- You should be ready to go.

    The file rundq.ml call the actual experiment. You can use the
    experiments and helpers in icml.ml as a guide.

    You should setup the relevant data and qcache files.

    Call the applications as follows:

    ./rundq | tee dq-run.log

    You should get a rundq-$DATE directory with the results of the
    experiment.

4.- You can generate some plots [Experimental, gnuplot required]

$ ./plot-histo qerror
$ ./plot-histo rerror
$ ./plot-histo qresult