Age | Commit message (Collapse) | Author |
|
Change-Id: I437662886cb9bb0e4e4f5aab557cb4949728f2fb
|
|
Add testNotificationListScroll() for GfxMonitor on scrolling
Bug : 71900748
Test: atest SystemUiJankTests#testNotificationListScroll
Change-Id: Id0633850ad9a5a2b21d0d3fb0e1219a8b466e76f
|
|
Change-Id: I57b9d6ba466d078574a7a843765540f3f255d767
|
|
am: d9251c36f6
Change-Id: I44db13add0bfda33fb90a73b4ed3960da82df49d
|
|
am: 32e63c965a
Change-Id: I0f0a2dc4c34f090a93ca53cadb62ec44a24f4305
|
|
am: c6bbd72750
Change-Id: I09d865a96b986a112bf30a384decc8f12ca62c2f
|
|
Exempt-From-Owner-Approval: meta data
BUG:69316045
Change-Id: I5ad41b5e6915357c67075a0bdf899fb6e1ef91d3
|
|
Change-Id: I403bdc2d8f059ffd47c5bd796bf8ef2bdbc41689
|
|
|
|
Change-Id: I48288672d18effc7f5d10260dbb9fe4bb28be891
|
|
The Workload class can now automatically install missing dependencies
stored in a workload's "packages" variable. Modify the SystemUi
workload to allow installing UbSystemUiJankTests automatically.
Also eliminate the "package" variable since the device will always
have the SystemUi package installed.
Test: Ran ./experiments/runsysui.py on a device without the tests installed
Change-Id: I415f0dd9309d5572ccacfdcceb91eff10cdb478c
Signed-off-by: Connor O'Brien <connoro@google.com>
|
|
Adds a method that plots a histogram of irq events. The irq can be
designated by name or by number. Also adds the histogram plot to the
Youtube_example notebook.
Fixes: 67362250
Test: run the notebook at examples/android/workloads/Android_YouTube.ipynb
Change-Id: Ie1653a44046d952cac7b5ea07bbe28944af78f50
|
|
* changes:
power_profile: update the cpu freq script to support experiment
experiments/power: add active and cluster costs
experiments/power: Refactor run_cpu_freq
experiments/power: Continue previous experiment
experiments/power: Taskset dhrystone
experiments/power: Sleep until dhrystone dies
experiments/power: Correctly update cgroups in cpu freq
experiments/power: Hold wakelock during cpu freq
experiments/power: Disable thermal throttling
|
|
* changes:
tools: Add a tool to compare Janbench results
experiments: Add a jankbench experiment
|
|
Change-Id: If8c477dd6a8056f1f6a63157aea730616f2f73c7
|
|
Change-Id: I359c9ef0780e5036a0c07df00585d0401176025f
|
|
am: a02cc5035c
Change-Id: Ic9f71e64e95e5d07870cf393c5c02bc312472110
|
|
am: ac819aec74 -s ours
Change-Id: I4fe11f706ac257fb36dd6c4fb53a0eb8be364386
|
|
Exempt-From-Owner-Approval: Changes already landed internally
Change-Id: I4eb1ab4d690329edd4b24ffcfa84ca861fe47adc
|
|
The cpu frequency power experiment has been updated to
calculate the base and cluster costs. Update the analysis and
power profile generator to support it.
Test: ./generate_power_profile.py
Change-Id: I282f268e8b0aa3c288f13727ed6c7cf6b0362b28
|
|
Add measurements that can be used to calculate active costs
and accurate cluster costs. These measurements can be
used by EAS and the power profile to model cpu power.
Test: ./run_cpu_frequency.py
Change-Id: I225b388fa4bab9dfc28715f9dada08df936bcb87
|
|
Put the current power tests for a single cluster
into a seperate function because next we will be adding
tests accross multiple clusters. Rename the outfiles
because it will help disambiguate which cpus at what
frequency belong to which cluster.
Test: ./run_cpu_frequency.py
Change-Id: I8e8f705579edf2a3242efa538c6bcdeb78a04421
|
|
Allow a previous run_cpu_frequency experiment to continue. This
experiment takes a significant amount of time to run. This will
allow an run to be stopped in the middle and continued sometime
later.
Test: ./run_cpu_frequency.py
Change-Id: I7651d354f2b16d7f5730b222e1fb803115a06601
|
|
Force dhrystone to run on each core. Previously, dhrystone would
sometimes run multiple threads on the same core which cause
problems with the power measurements.
Test: ./run_cpu_frequency.py
Change-Id: Ie112dddd743d494a01e045de1b35beaf1a18c357
|
|
Killall does not work on background processes.
Instead, just sleep until the benchmark has finished.
A long sleep will also allow the hardware to cool down
and stop power measurements to be affected by heating.
Test: ./run_cpu_frequency.py
Change-Id: I78528aea52adb6c0e63755b20a0b8103406afb3f
|
|
Cpu freq was not updating the offline cpu cgroup.
Test: ./run_cpu_frequency.py
Change-Id: Ieeef71160ea80b85012e183801bb12d7b5f19e8f
|
|
Hold a wakelock to prevent suspending at any point.
Test: run ./run_cpu_frequency.py
Change-Id: I8e3040177ffb675771ec7ef53ade1d0061cad370
|
|
Disable thermal engine and perfd that can cause thermal throttling
and interfere with the power measurements.
Test: run ./run_cpu_frequency.py
Change-Id: Ifc811f29cb9ac1a3b9213ff9869ce16b664917b3
|
|
Change-Id: I8c038220b691b0266acd0e17c31d86332761f0e9
|
|
am: d988c21a3f
Change-Id: I21b0f2bde76ea2d3c2b6f2df73b7949ed6e052ee
|
|
Change-Id: I26ce33f0db82b217da46290e7dd7784a66a08eb3
|
|
Change-Id: I6d63c3a5f5411e42b17b82a685d6a49f31b74a60
|
|
Change-Id: I5c2e3ac54d2e1d27ad849cb9ae98e1366cf54252
|
|
|
|
|
|
Change-Id: If0903f9aba40dc45d71dd05b521d32bbef5af761
|
|
Its often difficult to find regressions in jankbench results visually.
This tool allows to compare multiple test results between 2 different
kernels or test runs. In the output of this tool, each statistic gets a
_diff column which shows the delta between baseline and the test.
Negative values in the diff column mean improvements (lower frame
durations).
Example run:
./compare_jankbench.py --baseline='./results/Jankbench_baseline'
--compare-with='./results/Jankbench_kernel_change'
The output will be something like (only showing 25% and 50%):
25% compare 25%_diff 50%_compare 50%_diff
test_name
image_list_view 2.11249 0.0178108 5.7952 0.0242445
list_view 2.02227 -3.65839 5.74957 -0.095421
shadow_grid 6.00877 -0.000898 6.23746 -0.0057695
high_hitrate_text 5.81625 0.0264913 6.03504 0.0017795
Change-Id: I2ba9e43309ab93b11059563e8db014ecc28c3fc7
Signed-off-by: Joel Fernandes <joelaf@google.com>
|
|
Also prints Frame duration statistics to logs.
Change-Id: I78785a1e0c31b59d73d9e2efa10e8fdde0aa0cd7
Signed-off-by: Joel Fernandes <joelaf@google.com>
|
|
Change-Id: Ia398d4820ada92208bb9c74fdcff89bdef1f2a7d
Signed-off-by: Joel Fernandes <joelaf@google.com>
|
|
Also fix spelling errors and add a new 'overdraw' test. Also make it
possible to return a list of available tests.
Change-Id: I2360f82cdc2e9c3c0be4336a1f4a45d71ed61500
Signed-off-by: Joel Fernandes <joelaf@google.com>
|
|
Upstream changed 'comm' to 'TaskName', which broke parsing.
CPU residency analysis was not in the integration testing loop, so this
includes some changes so that running experiments/run_uibench_cgroup.py
is sufficient smoke testing for this analysis feature.
If running experiments/run_uibench_cgroup.py, the charts no longer pop
up onscreen. This helps with testing, as well as is more consistent with
other analysis scripts work. (e.g. FrequencyAnalysis). pylab grabs the
output image in the notebook, so the behavior in the .ipynb files is the
same.
Fixes: 68655983
Test: Run experiments/run_uibench_cgroup.py, and verify the charts
produced in the results directory.
Test: Run notebooks/residency/task_residencies_uibench.ipynb and
make sure that the charts appear correctly.
Change-Id: I5c8cc54e9d8fde352c779a9a47c9762bd0dd9905
|
|
This fixes the problem that was causing the new unit test from upstream
to fail.
Fixes: 68331831
Bug: 68655983
Test: Run 'lisa-test tests/lisa' and see all 64 unit tests pass.
Change-Id: I5059a2cb31556a09cd623acda3f6df2c2925d6cf
|
|
Change-Id: I7ee30f4fec89cef111436d954f6d0af084a1fd80
|
|
Test: grep for files with no license
Change-Id: I9f31aeb9273f82c77762a1aee6bc2390598a5cc3
Signed-off-by: Connor O'Brien <connoro@google.com>
|
|
|
|
Change-Id: I17c106c389a7659e34e81ca8b1d1ae25035b4cb5
|
|
sync_with_upstream
Change-Id: Ifa298ecf70b3da280b5404c9105273144ed06aa4
|
|
58e57434aa am: b307ff397b
am: c1267cd7df
Change-Id: I030bd52b3fe0a0150ba98dcdd718524d6a796382
|
|
am: b307ff397b
Change-Id: If0c167ea3a19c3936457a34d413e4e00dd03419a
|
|
am: 58e57434aa
Change-Id: I6d287a076467a537ce69fbe764d6c9d0fbf660b5
|