New issue
Advanced search Search tips

Issue 785038 link

Starred by 2 users

Issue metadata

Status: WontFix
Owner:
Closed: Feb 2018
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 3
Type: Bug-Regression
Proj-XR



Sign in to add a comment

Ganesh-induced Performance Regressions

Project Member Reported by bsheedy@chromium.org, Nov 14 2017

Issue description

The introduction of Ganesh introduced a fairly large memory regression, e.g. https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgj5CttQoM. This has been largely alleviated by a subsequent fix that linked Ganesh to the correct GL interface, but it's still ~1MB higher than it was previously.

The regression appears to be solely in the memory:chrome:all_processes:reported_by_chrome:malloc:effective_size_avg measurement.
 
This also appears to affect memory:chrome:all_processes:reported_by_os:gpu_memory:proportional_resident_size_avg and memory:chrome:all_processes:reported_by_os:gpu_memory:gl:proportional_resident_size_avg, see https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgr96LswkM

However, the most recent run seems to have returned to normal for those two measurements, so it may have been inadvertently fixed.

Comment 2 by tiborg@chromium.org, Nov 16 2017

Labels: M-64
Summary: Ganesh-induced Performance Regressions (was: Ganesh-induced Memory Regression)
It seems that the memory regression went away. However, we now see a small regression in text rendering times: https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgr96LswkM

Comment 3 by tiborg@chromium.org, Nov 21 2017

Labels: -M-64 M-65
The text rendering regression seem to have gone away too.
Status: WontFix (was: Assigned)
Closing since regression went away.
Labels: VR-metrics
Labels: Test-Complete
Labels: VR-Perf
Components: Internals>XR

Sign in to add a comment