Running Parsemark

この記事はまだ日本語に翻訳されていません。MDN の翻訳はボランティアによって行われています。是非 MDN に登録し、私たちの力になって下さい。

Parsemark is a command-line testing utility for measuring parse/emit performance of the SpiderMonkey front-end. The intent is to run a dozen or so large-payload JS sites globbed up from across the web. Yes, it includes GMail.

Here's how you get your mitts on it. The test files are not checked into the source tree because they're non-Mozilla JS code.

  1. Download the suite of parse tests from Bug 548621 and decompress them.
  2. Within js/src/tests, run python parsemark.py -h to see help documentation.

Generally you'll want to capture JSON results for your baseline and compare them to the results for the "current" version of your shell. A typical run can be done like so:

cd js/src/tests

python parsemark.py /path/to/baseline/js /path/to/parse-tests-dir/ -q > /tmp/baseline.json

python parsemark.py /path/to/current/js /path/to/parse-tests-dir/ -q > /tmp/current.json

python compare_bench.py /tmp/current.json /tmp/baseline.json

NOTE:  Unfortunately the comparisons done are very noisy and not reliable!

Document Tags and Contributors

Contributors to this page: cdleary, nnethercote
最終更新者: nnethercote,