This repository stores the scripts for generating the results in "Accelerating Optimal Power Flow: Condensed-Space Interior-Point Methods and Automatic Differentiation on GPUs" by Sungho Shin, Francois Pacaud, and Mihai Anitescu.

First, julia v1.9 or higher needs to be installed. We recommend using juliaup.

To benchmark with ma27 linear solver, you can additionally install libHSL. Also, to ensure that AMPL can find the HSL library, make sure that libhsl.so can be found from PATH environment variable.

The script assumes that your computer has an NVIDIA GPU.

Once julia is installed, you can instantiate the project to automatically download the dependencies.

$ cd path/to/opf-on-gpu
$ julia --project -e 'import Pkg; Pkg.develop("path/to/HSL_jll"); Pkg.instantiate()`

Then, the script can run:

$ julia --project -e 'include("example.jl"); include("example.jl")' 

Here, we run it twice just to make sure that compilation time is not affecting the timing results.

Please email @sshin23