OptiProfiler: a platform for benchmarking optimization solvers

🌐 Website: www.optprof.com

Install OptiProfiler from PyPI:

pip install optiprofiler

You can also install OptiProfiler from conda-forge:

conda install conda-forge::optiprofiler

Python package removal follows distribution ownership. Running python -m pip uninstall optiprofiler removes the core package and its bundled S2MPJ contents, but it does not remove independently installed problem-library distributions, their external runtimes, or their caches. Those plugins require OptiProfiler and become usable again after the core is reinstalled. Remove one adapter separately with, for example, python -m pip uninstall optiprofiler-pycutest. This does not remove the independently managed PyCUTEst or CUTEst runtime and problem definitions. The complete workflow for installing, discovering, updating, and removing providers is documented under doc/source/user/problem_libraries/.

Download the MATLAB-only ZIP from the OptiProfiler release page and extract it to a writable directory. In MATLAB, navigate to the extracted directory, where setup.m is located, and run:

setup

If the optional SOLAR MATLAB adapter is not installed locally, setup asks whether to download it. For non-interactive scripts, pass the choice explicitly:

setup(struct('install_solar', true))
setup(struct('install_solar', false))

When accepted or requested, setup clones optiprofiler/solar_matlab under prefdir/optiprofiler/problem_libraries by default, adds the adapter root to the MATLAB path, and registers its canonical functions. Pass problem_library_root to choose another writable parent directory. The adapter is not a submodule of the OptiProfiler repository. Passing install_solar=false skips the prompt and optional download when the adapter is not installed locally, which is useful in CI and batch scripts. The public problem-library name remains solar; use plibs={'solar'} in MATLAB benchmark options. The library-specific solar_select and solar_load functions remain callable after setup. Generic code may instead call library = resolveProblemLibrary('solar') and then use library.select and library.load.

To detach an optional or custom library without deleting its source, runtime, cache, or compiled artifacts, call unregisterProblemLibrary('solar'). Running setup uninstall removes the OptiProfiler core and setup-managed search paths, including registered adapter roots, but preserves external-library registrations and data. The complete workflow, including storage locations, updates, and the difference between detaching and deleting files, is documented under doc/source/user/problem_libraries/.