Age | Commit message (Collapse) | Author |
|
In show/hide ImePerfTests we waited for animationStart, however, it also
makes sense to wait for animation end to prevent overlapping and
avoiding reporting incorrect stats.
Fix: 191915803
Test: atest ImePerfTests
Change-Id: I7ef40db249afecc64b6a2418ff1ed7f90ccb4515
|
|
Improve stability of perf tests for ImePerfTests:
1. testShowImeWarm
2. testHideIme
To improve the stability of the test, check IME state and if state is
not as expected, ignore the iteration rather than failing all the
iterations and hence the whole test.
Bug: 180449596
Test: Run the test using steps in Readme.md
Change-Id: Ie42cb49bd8f79af10e7fa48f0a3afb19f253476c
|
|
Add WindowPerfTestBase and WindowPerfRunPreconditionBase
into apct-perftests-utils. So window manager and input
method manager can share the same functions.
Bug: 174292015
Test: WmPerfTests ImePerfTests
Change-Id: Ie2818536d6611d1ba5f4b6cd725cd2d4a95e1cac
|
|
Add Cold IMF startup latency test
Bug: 167947940
Test: atest ImePerfTests#showImeCold
Change-Id: I48be0250c59cb352e3be0558c780294eaae0f7ff
|
|
Measure latency caused by IMF for IME operations like show / hide.
This CL uses apct tests for integration with Crystallball to measure
overall latency and breakdown of critical IMF methods.
In this CL we introduce a BaselineIme with minimal UI to measure
user-preceived delays in IME show/hide.
Refer to design doc in bug.
Bug: 167947940
Test: atest ImePerfTests and also refer to README.md
Change-Id: I8efff52fe25952d452aef7f059400c63d1a9fa4a
|