New issue
Advanced search Search tips
Note: Color blocks (like █ or █) mean that a user may not be available. Tooltip shows the reason.

Issue 756084 link

Starred by 1 user
Status: Assigned
Owner:
█ fran...@chromium.org
Last visit > 30 days ago
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 2
Type: Bug-Regression



Sign in to add a comment

9.7% regression in v8/Massive-asm_wasm/Poppler on ChromeOS at 47369:47369

Project Member Reported by bbudge@chromium.org, Aug 16 2017 
Performance dashboard identified a 9.7% regression in v8/Massive-asm_wasm/Poppler on ChromeOS at revision range 47369:47369. Graph: https://chromeperf.appspot.com/report?masters=internal.client.v8&bots=ChromeOS&tests=v8%2FMassive-asm_wasm%2FPoppler&checked=Poppler%2CPoppler_ref%2Cref&rev=47369

Poppler is a PDF renderer - a very large codebase that uses lots of JS / Asm/Wasm features. The regression seems pretty correlated to this change.

Comment 1 by fran...@chromium.org, Aug 16 2017 

I know the change regresses certain cases. Maybe it's a micro benchmark hitting exactly that spot. I reverted and will have a look at the benchmark tomorrow.

Comment 2 by bbudge@chromium.org, Aug 17 2017 

Hi Franziska,

It looks like the revert or your CL causes a bunch more 2-10% regressions on other benchmarks. In other words, please re-land your change at your convenience, as it seems to help much more than it hurts.

Comment 8 by fran...@chromium.org, Aug 19 2017 

I created that microbenchmark specifically for working on Set. It's expected that it goes down a lot with the first changes. We can safely ignore that microbenchmark.
►

Sign in to add a comment