[darcs-users] Darcs 2.4 - performance testing (stage 1)

Trent W. Buck twb at cybersource.com.au
Wed Jan 27 00:41:04 UTC 2010

Max Battcher <me at worldmaker.net> writes:

> Also, maybe we should put together a few Wiki pages to store/compare
> benchmark results? (It may be worth tweaking darcs-benchmark's ASCII
> art output to conform to reST tables

It would need to look like this.  Important differences:

  - left-aligned cells (sorry!);
  - mandatory overline and underline of table;
  - whitespace between cells (not "|"); and
  - mandatory name for each column (but can be a comment).

Here's a mockup. It works with python-docutils (the reference
implementation), but appears to be treated by padoc as a simple

    ===============  ======= ===========  ======= ===========
    Test type        darcs-2.3.1          darcs-2.3.98
    ---------------  -------------------  -------------------
    .. comment       elapsed maxresident  elapsed maxresident
    ===============  ======= ===========  ======= ===========
    get (full)       47.6s   10.0M        58.8s   10.0M
    get (lazy, x10)  87.0s   3.0M         74.1s   3.0M
    pull 100         7.9s    24.0M        6.5s    16.0M
    annotate         18.4s   187.0M       36.8s   180.0M
    wh x50           6.2s    0.0M         5.6s    0.0M
    wh mod x50       10.2s   3.0M         7.7s    2.0M
    wh -l x20        34.6s   0.0M         5.7s    0.0M
    check            46.2s   21.0M        44.6s   257.0M
    repair           45.2s   21.0M        40.7s   259.0M
    pull 1000        65.9s   44.0M        61.6s   34.0M
    ===============  ======= ===========  ======= ===========

More information about the darcs-users mailing list