Benchmarks - Regression between Merge "Update Wear Compose current release dependencies on Compose libraries to 1.7.0-rc01" into androidx-main and Merge "Revert "Add information about size into WindowInfo"" into androidx-main [360151053]

WAI

Bug

Status Update

No update yet.

Description

pe...@skia-infra-public.iam.gserviceaccount.com

created issue on behalf of an...@google.com #1

Aug 16, 2024 11:55PM

Perf Regression (Low) found, matching 27 tracked metrics from benchmarks.

To triage this regression, see the guide at go/androidx-bench-triage.

Test classes affected:

 androidx.compose.runtime.benchmark.ComposeBenchmark
 androidx.compose.runtime.benchmark.DeepTreeBenchmark
 androidx.compose.runtime.benchmark.SlotTableIntegrationBenchmark
 androidx.wear.compose.material3.benchmark.TextBenchmark

Test methods affected:

 benchmark_01_Compose_OneRect
 benchmark_02_Compose_TenRects
 benchmark_03_Compose_100Rects
 benchmark_10_derivedState_reads_compose
 benchmark_deep_tree_01_depth1_breadth100_wrap2
 benchmark_deep_tree_02_depth7_breadth3_wrap2
 benchmark_deep_tree_03_depth2_breadth10_wrap2
 benchmark_deep_tree_04_depth2_breadth10_wrap6
 create
 layout

Devices affected:

 aosp_mokey
 sargo

API Level:

 API_30_R
 API_31_S
 API_34_U

Comments

pe...@skia-infra-public.iam.gserviceaccount.com <pe...@skia-infra-public.iam.gserviceaccount.com> #2Aug 16, 2024 11:55PM

We could add an API to make the weighted distance customizable, aosp/3271691 has more info.

cc...@google.com <cc...@google.com> #3Aug 22, 2024 04:58PM

Reassigned to cc...@google.com.

Assigning Skia V2 bugs to myself, as they are expected to be dupes of bugs created by the V1 system, which has caught up and is generating bugs again.

For more info, see go/androidx-bench-dashboard-wtf

cc...@google.com <cc...@google.com> #4Sep 5, 2024 10:41PM

Status: Won't Fix (Intended Behavior)

These improvements (in #comment1) are from https://android-review.git.corp.google.com/c/platform/frameworks/support/+/3222662 which reverts a CL that caused the regressions in the first place.

The regressions (in #comment2) appear to be noise - both recovered a few builds later with what appears to be an unrelated CL.

Issue 360151053

Description

Issue summary

Comments

pe...@skia-infra-public.iam.gserviceaccount.com <pe...@skia-infra-public.iam.gserviceaccount.com> #2Aug 16, 2024 11:55PM

cc...@google.com <cc...@google.com> #3Aug 22, 2024 04:58PM

cc...@google.com <cc...@google.com> #4Sep 5, 2024 10:41PM

Add comment

Issue metadata