Skip to content

Latest commit

 

History

History
76 lines (52 loc) · 3.34 KB

README.md

File metadata and controls

76 lines (52 loc) · 3.34 KB
Raw

Exact version of SCIP now directly on GitHub

The most recent development version of exact SCIP is now directly mirrored on the SCIP github repository

Exact SCIP

This is a development version of the numerically exact variant of MIP solver SCIP, based on SCIP 7.0.

The exact solving mode is based on the framework described in

William J. Cook, Thorsten Koch, Daniel E. Steffy, Kati Wolter: A hybrid branch-and-bound approach for exact rational mixed-integer programming. Math. Program. Comput. 5(3): 305-344 (2013)

That framework has been revised and extended by symbolic presolving, using the parallel presolving library PaPILO, as well as an exact repair heuristic. The results of this revision are described in

Leon Eifler, Ambros Gleixner: A computational status update for exact rational mixed integer programming Math. Program. (2022).

Furthermore, the most recent version features separation and verification of safe Gomory mixed integer cuts.

Certificate printing is supported (although presolving cannot yet be verified), and can be checked using VIPR.

The only input formats that are currently supported for the exact solving mode are ZIMPL and MPS (only in the most recent version).

Versions

  • scipopt-exact-0.1.tgz the version used for the IPCO 2021 paper
  • scipopt-exact-0.2.tgz the version used for the full length 2022 Math. Programming paper
  • scipopt-exact-0.3.tgz the most recent version, featuring separation of safe GMI cuts

Installation instructions

This development version of exact SCIP can only be built using CMake, the Makefile system is not functional.

Dependencies which are optional for the floating point version but mandatory for exact SCIP:

  • GMP for rational arithmetic in ZIMPL, SoPlex, SCIP, and PaPILO
  • Boost multiprecision library (Version should be 1.70 or newer) for rationals in SCIP and PaPILO
  • MPFR for approximating rationals with floating-point numbers in SCIP.

Step by step guide:

1. Extract the tarball

tar -xzvf scipopt-exact-0.1.tgz
cd scipopt-exact-0.1

1. Build Zimpl, SoPlex, PaPILO:

# go into the directory
cd <one of zimpl/soplex/papilo>
mkdir build
cmake ..
make
cd ../..

2. Build SCIP:

cd scip
mkdir build
cmake .. -DSOPLEX_DIR=<PATH_TO_SOPLEX_BUILD_DIRECTORY> -DPAPILO_DIR=<PATH_TO_PAPILO_BUILD_DIRECTORY> -DZIMPL_DIR=<PATH_TO_ZIMPL_BUILD_DIRECTORY>
make

Test if everything works correctly with the MIPEX ctest (vipr needs to be installed for the certificate checking to work)

ctest -R MIPEX

USAGE

Enabling/disabling the exact solving mode is done with parameter exact/exact_enabled, note that this has to be done before reading a problem instance. Further advanced parameters for exact solving can be set in the exact submenu. Alternatively, the provided settings file exactsolve.set can be used.

Certificate printing can be enabled by setting certificate/filename to a non-default value. If cutting planes are enabled (only version 0.3), then the certificate file needs to be completed using the viprcomp script prior to verifiaction.