New issue
Advanced search Search tips

Issue 902352 link

Starred by 1 user

Issue metadata

Status: Fixed
Owner:
Closed: Jan 3
Cc:
Components:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 3
Type: Bug



Sign in to add a comment

DOMStorage Onion-Soupify improvements

Project Member Reported by dmu...@chromium.org, Nov 6

Issue description

Bug for bisect jobs to verify improvements
 
Cc: dmu...@chromium.org
📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/124f26e1e40000

[DOMStorage] Enable onion soup codepath by dmurph@chromium.org
https://chromium.googlesource.com/chromium/src/+/80ce30280a48850aad465138b84fe209474733e8
memory:webview:all_processes:reported_by_chrome:malloc:allocated_objects_size: 6.104e+07 → 6.029e+07 (-7.576e+05)

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

Benchmark documentation link:
  None
Verified improvement:
Android Nexus6 WebView Perf/memory.top_10_mobile,
memory:webview:all_processes:reported_by_chrome:malloc:allocated_objects_size
Δavg	-658.1 KiB
%Δavg	-1.1%


Keep in mind that a decrease in malloc memory is totally expected, as we move a bunch of stuff from being malloc allocated to partition_alloc allocated (and indeed, in that benchmark partition_alloc memory went up by about the same amount)
📍 Couldn't reproduce a difference.
https://pinpoint-dot-chromeperf.appspot.com/job/12c4cb69e40000
📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/12453e85e40000

[DOMStorage] Enable onion soup codepath by dmurph@chromium.org
https://chromium.googlesource.com/chromium/src/+/80ce30280a48850aad465138b84fe209474733e8
memory:chrome:all_processes:reported_by_chrome:malloc:allocated_objects_size: 1.777e+08 → 1.749e+08 (-2.825e+06)

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

Benchmark documentation link:
  https://bit.ly/system-health-benchmarks
Cc: rmcilroy@chromium.org
📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/153410e1e40000

[Build] Disable build-time untrusted code mitigations on non-Android platforms and ia32. by rmcilroy@chromium.org
https://chromium.googlesource.com/v8/v8/+/bac7dad666ae203e04b5ff68d838ed2ea5f03e26
React-TodoMVC: 231.1 → 223.9 (-7.208)

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

Benchmark documentation link:
  None
📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/17899049e40000

[Build] Disable build-time untrusted code mitigations on non-Android platforms and ia32. by rmcilroy@chromium.org
https://chromium.googlesource.com/v8/v8/+/bac7dad666ae203e04b5ff68d838ed2ea5f03e26
React-TodoMVC: 5270 → 5180 (-89.48)

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

Benchmark documentation link:
  None
📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/11e4e949e40000

[DOMStorage] Enable onion soup codepath by dmurph@chromium.org
https://chromium.googlesource.com/chromium/src/+/80ce30280a48850aad465138b84fe209474733e8
memory:chrome:all_processes:reported_by_os:system_memory:proportional_resident_size: 9.054e+07 → 8.814e+07 (-2.407e+06)

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

Benchmark documentation link:
  https://bit.ly/system-health-benchmarks
😿 Pinpoint job stopped with an error.
https://pinpoint-dot-chromeperf.appspot.com/job/15c81fcde40000

The swarming task expired. The bots are likely overloaded, dead, or misconfigured.
Interesting data from the memory:chrome:all_processes:reported_by_os:system_memory:proportional_resident_size results:

memory:chrome:all_processes:reported_by_os:private_dirty_size -4.7% 48.6 MiB
memory:chrome:all_processes:reported_by_os:private_footprint_size -2.0% 110.3 MiB
memory:chrome:all_processes:reported_by_os:proportional_resident_size -1.6% 142.6 MiB
memory:chrome:all_processes:reported_by_os:system_memory:native_heap:private_dirty_size -9.1% 21.5 MiB
memory:chrome:all_processes:reported_by_os:system_memory:native_heap:proportional_resident_size -9.1% 21.6 MiB
memory:chrome:all_processes:reported_by_os:system_memory:private_dirty_size -4.7% 48.6 MiB
memory:chrome:all_processes:reported_by_os:system_memory:private_footprint_size -2.0% 110.3 MiB
memory:chrome:all_processes:reported_by_os:system_memory:proportional_resident_size -2.7% 86.3 MiB
memory:chrome:all_processes:reported_by_os:system_memory:stack:private_dirty_size -0.7% 559.2 KiB
memory:chrome:all_processes:reported_by_os:system_memory:stack:proportional_resident_size -0.7% 561.2 KiB

These are all pretty good!

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/10edb035e40000

[DOMStorage] Enable onion soup codepath by dmurph@chromium.org
https://chromium.googlesource.com/chromium/src/+/80ce30280a48850aad465138b84fe209474733e8
tasks_per_frame_total_all: 57.02 → 49.37 (-7.65)

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

Benchmark documentation link:
  https://bit.ly/rendering-benchmarks
Status: Fixed (was: Assigned)
It seems like there wasn't any notable changes from the onionsouping, it was mostly moving around where the memory was living. The speedometer improvement was something else. Oh well!

Sign in to add a comment