GPU Implementation of Second-Order Linear and Nonlinear Programming Solvers

Overview

This repository contains the source code and materials for the paper titled "GPU Implementation of Second-Order Linear and Nonlinear Programming Solvers." The paper presents an overview of GPU-accelerated optimization solvers based on second-order methods, emphasizing pivoting-free interior-point methods for large and sparse linear and nonlinear programs.

Authors

• Alexis Montoison, Argonne National Laboratory
• François Pacaud, Mines Paris-PSL
• Sungho Shin, Massachusetts Institute of Technology
• Mihai Anitescu, Argonne National Laboratory

Benchmark

Dependencies

To reproduce the results in this paper, the following software or license file should be obtained:

• julia: we recommend using juliaup to install the latest stable version of Julia.
• libHSL: a library for sparse linear algebra. After downloading libHSL, please install HSL_jll using the locally downloaded library.

julia -e 'using Pkg; Pkg.develop(path="path/to/libhsl")'

• Gurobi: A free academic license can be obtained from Gurobi's website. After obtaining the license, gurobi.lic should be placed in the home directory.

Running Benchmarks

make -C benchmark