RESOLVED DUPLICATE of bug 93185 84131
Dromaeo/cssquery-prototype.html, cssquery-jquery.html, cssquery-dojo.html are very slow
https://bugs.webkit.org/show_bug.cgi?id=84131
Summary Dromaeo/cssquery-prototype.html, cssquery-jquery.html, cssquery-dojo.html are...
Ryosuke Niwa
Reported 2012-04-17 01:24:28 PDT
All of these tests are taking well over 2 minutes each to run on Chromium Mac bots: http://build.webkit.org/builders/Chromium%20Mac%20Release%20%28Perf%29/builds/1616 181.695680 Dromaeo/cssquery-prototype.html 181.681818 Dromaeo/cssquery-jquery.html 172.595614 Dromaeo/dromaeo-object-regexp.html 166.828389 Dromaeo/cssquery-dojo.html 162.882592 Parser/tiny-innerHTML.html 158.915446 Dromaeo/jslib-modify-jquery.html
Attachments
Patch (3.79 KB, patch)
2012-04-17 01:56 PDT, Hajime Morrita
no flags
Hajime Morrita
Comment 1 2012-04-17 01:56:03 PDT
Hajime Morrita
Comment 2 2012-04-17 02:05:35 PDT
The patch illustrates a naive fix. But I'm not sure whether this is the right direction. Slow tests are slow because they contain many small tests, not because they have a few large tests. My feeling is that it's better to just skip less relevant ones. For example, it might be OK just to cover cssquery-jquery.html for "cssquery" family because the world is dominated by jquery not prototype. We also able to skip dromaeo-* since these are almost pure-JS which will be (roughly) covered by SunSpider.
Tony Chang
Comment 3 2012-05-14 10:17:54 PDT
To summarize, those 6 tests take over 17 minutes right now. This change converts the number of runs from 5 to 2, which should bring the running time for these tests to under 7 minutes. It seems worth trying. We can watch the graphs to see how stable the times are. If the times are consistent across runs, then the extra runs probably aren't necessary.
Ryosuke Niwa
Comment 4 2012-05-14 11:19:07 PDT
(In reply to comment #3) > To summarize, those 6 tests take over 17 minutes right now. This change converts the number of runs from 5 to 2, which should bring the running time for these tests to under 7 minutes. We should probably omit reporting stdev, min, max, median in that case because sample size of 5 doesn't give us a meaningful statistics.
Ryosuke Niwa
Comment 5 2012-05-14 11:19:28 PDT
I mean the sample size of "2".
Ryosuke Niwa
Comment 6 2012-09-20 20:08:42 PDT
*** This bug has been marked as a duplicate of bug 93185 ***
Note You need to log in before you can comment on or make changes to this bug.