Optimization Solver Benchmark

📋 Project Overview

A reproducible benchmarking framework for optimization solvers across LP, QP, SOCP, and SDP problems using standardized libraries (DIMACS, SDPLIB). The system conducts systematic performance evaluation establishing "out of the box" performance baselines rather than optimized configurations, and publishes transparent results for research community use. The benchmarking code is available at https://github.com/napinoco/optimization-solver-benchmark

👤 Author: Naoki Ito

Solvers Tested

Problems Tested

120

Libraries

Problem Types

🔧 Solvers Tested

Total Solvers: 11

Solver Names:

cvxpy_clarabel
cvxpy_cvxopt
cvxpy_ecos
cvxpy_highs
cvxpy_osqp
cvxpy_scip
cvxpy_scs
cvxpy_sdpa
matlab_sdpt3
matlab_sedumi
scipy_linprog

📚 Problems Tested by Library and Type

Library	Problem Type	Count
DIMACS	SDP	17
DIMACS	SOCP	17
SDPLIB	SDP	86

📊 Performance Analysis

🚧 To Be Determined

Performance ranking methodology is currently under development.
Evaluation criteria for determining "best performers" are being established
to ensure fair and meaningful comparisons across different solver types.

Please refer to the Results Matrix for detailed performance data in the meantime.

🔧 Environment Information

Git Commit Hashes: 0d9be48d, 2ef15f0b, 443b599d, 86bfae9b, 96ca0437, d31cb753, f95b7a82

⚠️ Multiple environments detected: Results from 7 different Git commits

Platform: Darwin (8CPU, 24GB)

Python Version: CPython 3.12.2

Operating System: Darwin 23.5.0

CPU: arm (8 cores)

Memory: 24.0 GB

MATLAB: Available (version detection via solver results)