Issue metadata
Sign in to add a comment
|
4% regression in speedometer at 543980:544012 |
||||||||||||||||||
Issue descriptionSee the link to graphs below.
,
Mar 19 2018
📍 Pinpoint job started. https://pinpoint-dot-chromeperf.appspot.com/job/14a9d719440000
,
Mar 20 2018
📍 Found a significant difference after 1 commit. https://pinpoint-dot-chromeperf.appspot.com/job/14a9d719440000 Reland "[builtins] Remove off-heap builtins from the snapshot" by jgruber@chromium.org https://chromium.googlesource.com/v8/v8/+/fd70917d520ea7bbcc5e401776c7dac7120c6942 Understanding performance regressions: http://g.co/ChromePerformanceRegressions
,
Mar 20 2018
,
Mar 21 2018
Issue 823930 has been merged into this issue.
,
Mar 21 2018
Issue 823932 has been merged into this issue.
,
Mar 21 2018
Issue 823933 has been merged into this issue.
,
Mar 22 2018
Issue 824669 has been merged into this issue.
,
Mar 22 2018
,
Mar 22 2018
Yep, web-tooling-benchmark is what I've been using to bench locally since it's so convenient. I haven't yet had the time to focus on these regressions. What I know so far is that indirections for constants and external references contribute. It's surprising though that the regressions bisect to this CL, which just switches execution from on-heap code to off-heap, binary-embedded code, without changing the code itself. There's something here that I don't understand yet. I'm currently hacking on landing a few final CLs for the off-heap mechanism, will focus on performance once that is done (early next week). If we don't have a good perf solution in sight by mid next week, we'll flip v8_enable_embedded_builtins back to false for the next release.
,
Mar 27 2018
Duping this into https://crbug.com/819121. See http://goo.gl/vjjyB6 for an analysis of the current state.
,
Mar 27 2018
,
Mar 28 2018
|
|||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||
Comment 1 by 42576172...@developer.gserviceaccount.com
, Mar 19 2018