BenchmarkProfiles.jl documentation

A simple package to plot performance and data profiles.

This package containts Julia translations of original scripts by Elizabeth Dolan, Jorge Moré and Stefan Wild. See http://www.mcs.anl.gov/~wild/dfo/benchmarking.

The original code was not accompanied by an open-source license. Jorge Moré and Stefan Wild have kindly provided their consent in writing to allow distribution of this Julia translation. See here for a full transcription.

It appears that performance profiles date back to at least 1996!

Watch out for the pitfalls of profiles!

How to Install

pkg> add BenchmarkProfiles

No plotting backend is loaded by default so the user can choose among several available plotting backends. Currently, Plots.jl and UnicodePlots.jl are supported. Backends become available when the corresponding package is imported.

Example

julia> using BenchmarkProfiles
julia> T = 10 * rand(25,3);  # 25 problems, 3 solvers
julia> performance_profile(PlotsBackend(), T, ["Solver 1", "Solver 2", "Solver 3"], title="Celebrity Deathmatch")
ERROR: ArgumentError: The backend PlotsBackend() is not loaded. Please load the corresponding AD package.
julia> using Plots
julia> performance_profile(PlotsBackend(), T, ["Solver 1", "Solver 2", "Solver 3"], title="Celebrity Deathmatch")  # Success!

Performance Profile

Supported Backends

Currently supported backends are:

Backends are treated as optional dependencies, but one is required to produce a plot. The user should import one of the supported backends before calling one of the plotting functions.

Adding a New Backend

In order to add a new backend, there are two steps:

  • Edit src/BenchmarkProfiles.jl to define the backend and make it available:
struct SomeNewPlotBackend <: AbstractBackend end
const bp_backends = [:PlotsBackend, :UnicodePlotsBackend, :SomeNewPlotBackend]
  • Edit src/requires.jl to define how to produce the plot from the data:
@require SomeNewPlot = "91a5bcdd-55d7-5caf-9e0b-520d859cae80" begin
  function performance_profile_plot(
    ::SomeNewPlotBackend,
    x_plot,
    y_plot,
    max_ratio,
    xlabel,
    ylabel,
    labels,
    title,
    logscale;
    kwargs...,
  )
    #
    # now produce the plot and return the plot object
    #
  end

  function data_profile_plot(
    ::SomeNewPlotBackend,
    T,
    xs,
    max_data,
    xlabel,
    ylabel,
    labels,
    title;
    kwargs...,
  )
    #
    # do the same here
    #
  end
end

References

  • A. L. Tits and Y. Yang, Globally convergent algorithms for robust pole assignment by state feedback, IEEE Transactions on Automatic Control, 41(10), pages 1432&ndash;1452, 1996. DOI 10.1109/9.539425.
  • E. Dolan and J. Moré, Benchmarking Optimization Software with Performance Profiles, Mathematical Programming 91, pages 201&ndash;213, 2002. DOI 10.1007/s101070100263.
  • J. J. Moré and S. M. Wild, Benchmarking Derivative-Free Optimization Algorithms, SIAM Journal on Optimization, 20(1), pages 172&ndash;191, 2009. DOI 10.1137/080724083.