1. 08 Dec, 2020 1 commit
    • Pierre Langlois's avatar
      Add a --baseline option. · 513c8b8b
      Pierre Langlois authored
      This new --baseline flag can be used to indicate which of the data sets
      is to be used as the baseline to compare against, instead of picking the
      first one.
  2. 07 Jan, 2020 1 commit
  3. 25 Sep, 2019 1 commit
  4. 04 Jun, 2019 1 commit
  5. 05 Feb, 2019 1 commit
  6. 04 Feb, 2019 1 commit
  7. 01 Feb, 2019 1 commit
    • Pierre Langlois's avatar
      Introduce Direction flag and optional CSV column. · 12a9541c
      Pierre Langlois authored and Jacob Bramley's avatar Jacob Bramley committed
      Instead of having a global --bigger-is-better/--smaller-is-better setting, we
      can now have a per-row setting with a new optional 'Direction' column in the CSV
      format. It allows plotting different values in the same graph, such as code
      size (bigger is better) and time spent (smaller is better).
      To clean things up, we also introduce a --direction flag that can be set to
      either 'left' or 'right' and indicates which way the biggest values should be
      plotted. As a result, we have the following equivalent flags:
      --bigger-is-better --> --direction=right
      --smaller-is-better --> --direction=left
  8. 17 Jul, 2017 1 commit
  9. 28 Oct, 2016 1 commit
  10. 28 Sep, 2016 1 commit
  11. 29 Jun, 2016 1 commit
    • Jacob Bramley's avatar
      Work around a problem with 'percent'. · fe097f38
      Jacob Bramley authored
      The built-in 'percent' function occasionally produces NaNs when given negative
      inputs. It looks like this has been fixed in more recent versions of R, but the
      workaround is simple.
  12. 28 Jun, 2016 1 commit
    • Jacob Bramley's avatar
      Plot graphs for comparing benchmark results. · 59fb7ac8
      Jacob Bramley authored
      In particular, the graphs are useful for analysing JavaScript benchmark results.
      JavaScript benchmarks often exhibit bimodal (or otherwise noisy) behaviour, and
      take a very long time to run, so simply examining the mean of a set of 10 runs
      tends to be misleading. This tool tries to plot every result in a way such that
      trends and patterns can be spotted.
  13. 25 Jan, 2016 1 commit