Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(354)

Issue 1330723003: Implement a better dashboard for running benchmarks (Closed)

Created:
5 years, 3 months ago by skybrian
Modified:
5 years, 3 months ago
Reviewers:
Søren Gjesse
CC:
reviews_dartlang.org
Base URL:
git@github.com:dart-lang/dart-protoc-plugin.git@master
Target Ref:
refs/heads/master
Visibility:
Public.

Description

Implement a better dashboard for running benchmarks - Provide a 'Run' button instead of running on page reload - Display median samples from saved report as a baseline for comparison. - Provide a menu for choosing the saved report. - Display the median, max, and number of samples gathered. Also: - Simplified the JSON benchmark a bit. - Increase the default number of samples and decrease their duration. - Collect one sample from each benchmark in each loop (round-robin). This was to try to reduce the amount of varation between runs. It seems to have helped somewhat, but benchmark results can still vary by up to about 20% when opening a new tab. BUG= R=sgjesse@google.com Committed: https://github.com/dart-lang/dart-protoc-plugin/commit/3c227864999d1a9700c8623fb7d5f28dad782387

Patch Set 1 #

Total comments: 2
Unified diffs Side-by-side diffs Delta from patch set Stats (+625 lines, -229 lines) Patch
M benchmark/data/index.dart View 1 chunk +3 lines, -1 line 0 comments Download
A benchmark/data/skybrian-macbookpro/0.4.2/json_chrome.pb.json View 1 chunk +62 lines, -0 lines 0 comments Download
A benchmark/data/skybrian-macbookpro/0.4.2/json_vm.pb.json View 1 chunk +63 lines, -0 lines 0 comments Download
M benchmark/lib/benchmark.dart View 3 chunks +3 lines, -3 lines 0 comments Download
A benchmark/lib/dashboard.dart View 1 chunk +118 lines, -0 lines 0 comments Download
A benchmark/lib/dashboard_model.dart View 1 chunk +75 lines, -0 lines 0 comments Download
A benchmark/lib/dashboard_view.dart View 1 chunk +272 lines, -0 lines 2 comments Download
D benchmark/lib/html_view.dart View 1 chunk +0 lines, -197 lines 0 comments Download
M benchmark/lib/suite.dart View 1 chunk +20 lines, -12 lines 0 comments Download
M benchmark/lib/suites/json.dart View 1 chunk +6 lines, -13 lines 0 comments Download
M benchmark/readints.dart View 1 chunk +2 lines, -2 lines 0 comments Download
M benchmark/readints.html View 1 chunk +1 line, -1 line 0 comments Download

Messages

Total messages: 5 (1 generated)
skybrian
5 years, 3 months ago (2015-09-05 06:54:56 UTC) #2
Søren Gjesse
lgtm Maybe look at how the Octane benchmark (https://github.com/chromium/octane) run benchmarks. It seems to have ...
5 years, 3 months ago (2015-09-08 06:43:54 UTC) #3
skybrian
Committed patchset #1 (id:1) manually as 3c227864999d1a9700c8623fb7d5f28dad782387 (presubmit successful).
5 years, 3 months ago (2015-09-08 17:10:37 UTC) #4
skybrian
5 years, 3 months ago (2015-09-08 17:25:11 UTC) #5
Message was sent while issue was closed.
I ran Octane twice in two tabs. The overall average was very close (22008 versus
22370) but the Typescript benchmark's variation was similar to what I'm seeing
(26022 versus 23793). 

So I think it depends on the benchmark, and some of Octane's benchmarks are very
stable so the average is good.

They recommend restarting the browser each time you run the benchmark.

https://chromiumcodereview.appspot.com/1330723003/diff/1/benchmark/lib/dashbo...
File benchmark/lib/dashboard_view.dart (right):

https://chromiumcodereview.appspot.com/1330723003/diff/1/benchmark/lib/dashbo...
benchmark/lib/dashboard_view.dart:5: library protoc.benchmark.html_view;
On 2015/09/08 06:43:54, Søren Gjesse wrote:
> Might have been easier to put more of this into the HTML file and lookup more
> elements by name instead of building the whole DOM in code. I guess you have a
> good reason for doing it this way.

It's because I wanted to keep each view self-contained and all in one place.

The template isn't very readable though. Perhaps setting innerHtml to a
triple-quoted string and querying it for sub-elements would work better.

Powered by Google App Engine
This is Rietveld 408576698