To improve the return path latency, we want to keep the CPU at at least WFI state (Idle_1). The PMQos cpu_dma_latency knob prevents the CPU from falling below WFI state. This makes the return path CPU wakeup latency very good. Check with wvw@, the power impact shouldn't be too significant.
The average energy cost per inference dropped from 3.85 to 3.47 mJ. The reason why the power number is lower WITH disable-idle is that, due to better latency, we get to run more inferences given the same amount of time. This makes the average power consumption lower.
Measurement:
MLPerf IC model Latency (ms) Power (mW) Energy/inference (mJ) MLPerf scores
Default 1.35 2837 3.85 560
Disable CPU Idle 0.98 3539 3.47 826
https://docs.google.com/presentation/d/1zx7sLkhOClmuRTCrq8-l3N1mZrrv7f-CtcdMuzV0eaI/edit?pli=1#slide=id.g12dd9e50b4b_0_0
Bug: 232183574
Test: MLPerf on Android T. Performance improved. Verified on Perfetto.
Change-Id: If067e0851bea0475043ef2127a25ed3a5fdab093
Skip I signal by setting kI to 0 and use UclampMinInit as the display
update boost value.
Bug: 232992641
Test: Manual Test
Change-Id: Iee49cb054817d49d18498ceced61446916669887
Disable the powerhint for F2FS gc_urgent mode, since we are going to
replace this with the smart idle maintenance service.
Test: turn on/off the screen & check /sys/fs/f2fs/dm-*/gc_urgent
Bug: 214430685
Signed-off-by: Daeho Jeong <daehojeong@google.com>
Change-Id: I6acf8e9e8a8d7e33d89acb1be44ea340e3f02241
- redefine mute duration of primary output from 160ms to 40ms to avoid
cutting of preview notification.
Bug: 233176336
Test: preview notification/voice end call tone
Signed-off-by: Jasmine Cha <chajasmine@google.com>
Change-Id: Ib696638eca99b1f58300533cca586cab209b188d
we need disable uclamp_threshold and reduce_prefer_idle
under launch scenarios.
Bug: 232593932
Test: build pass
Signed-off-by: chungkai <chungkai@google.com>
Change-Id: I83573ed4640cd841d99cff0b52df0335332623df
No need to prevent uclamp.min update rate by Granularity.
There is no perf impact.
Bug: 232336917
Test: /data/nativetest64/libperfmgr_test/libperfmgr_test
Change-Id: Ic38f133f7e6b1abe3d743fcfae71ce5a1b4780fb
Task packing has slightly impact for camera performance, so disable
it when camera is running.
Bug: 230554223
Test: build pass
Change-Id: Id20fb06b30fa8448c49c8163574e336dfa89ffe4
Jankiness is experience during BATOILO triggering. Lowering GPU DVFS
state and Raising TPU DVFS state to re-tune throttling mix.
Bug: 231575776
Test: Boot and confirm settings
Signed-off-by: George Lee <geolee@google.com>
Change-Id: Ife93560c8864f4340e94d9afc2e8d2b47b97fc67
Task packing may hurt app launch time, so disable it when app launch.
Bug: 201259891
Test: hint works
Change-Id: I7c8f7d1625f84a0968f74835fb34f4fc5d4123aa
set 160 back to 161 since RT task will select
little cores
Test: build pass
Bug: 230247757
Signed-off-by: chungkai <chungkai@google.com>
Change-Id: I71b74b647f1b9904f4a6688a07c5e7b3a2841bd9
To avoid task being placed into little cluster,
and keep same freq for middle cluster, increase
the uclamp min to 129,
Bug: 224649226
Test: freq is same as before
Signed-off-by: chungkai <chungkai@google.com>
Change-Id: Iabbb06b2a6c066ebe7bd8bdb3aba94ccf25b0d0e
add mic gain on path "speaker-mic" from 0 to 22
Bug: 228224014
Test: audio record
Signed-off-by: Carter Hsu <carterhsu@google.com>
Change-Id: Idafad36fdfb5c5f54e2ad3c1718b814b956ac17c
update new boost uclamp value since we update
cpu capacity
Test: PTS UIBench has no obvious difference
Bug: 224649226
Signed-off-by: chungkai <chungkai@google.com>
Change-Id: I63f72a2f7aa48c3cd915089cfba39d666adcb903
[O6/R4]
Phasing out config_roundedCornerMultipleRadius config from device
overlay package.
Bug: 224688507
Test: Build ROM and pass presubmit tes
Change-Id: Ie0b0c7b9e3591f11e4a481440943cf0a8b328340
