JuMP benchmarks

This repository contains benchmarks for JuMP-dev.

Visualizations are available at jump.dev/benchmarks.

Adding new benchmarks

This repository is open for contributions! There are two types of benchmarks we track:

• /src/latency contains end-to-end scripts. We run these from the command line, so they include compilation latency. Add new benchmarks here that matter to end-users.
• /src/micro contains a collection of micro-benchmarks, organized by package. Each benchmark is a function beginning with benchmark_, and tests a small, self-contained feature of the package.

Running the benchmark script

Run the benchmarks using scripts/benchmarks.jl:

julia --project=. scripts/benchmarks.jl --tune
julia --project=. scripts/benchmarks.jl --run
julia --project=. scripts/benchmarks.jl --publish

Dashboard improvements

The visualizations at jump.dev/benchmarks are preliminary. If you're interested in front-end web development and have suggestions for how we could improve things, please get in touch.

Hardware

These benchmarks run on a 2022 Apple Mac Mini with 16GB of memory:

Julia Version 1.7.2
Commit bf53498635 (2022-02-06 15:21 UTC)
Platform Info:
  OS: macOS (x86_64-apple-darwin19.5.0)
  CPU: Apple M1
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-12.0.1 (ORCJIT, westmere)

Performance on other machines may differ. However, we're mostly interested in long-term trends.