-Switch to LZ4 for better overall performance
-Set max_comp_streams to num of cores since upstream also moves this to
percpu.
-Set page-cluster to 0 as the incremental cost of reading 1 page at a
time is negligible for zram
Bug: 38249616
Test: boot and run zram-perf showing better performance
Change-Id: I0b92b246d773db85aa03d033b2cecee54347cbd1
Reserving more than 1.5gb sharply decreases free inodes, resulting
in gapps install/restore issues.
Change-Id: I16c0c649b7e06aea6203e34a81d0d0350b80588c
GPU takes long time(>8ms) when app was launching and closing.
This may because GPU is handling some heavy task like color conversion.
By setting NegativeOffset as 5ms to allow SF have more time to handle
when doing animation transaction and shift back 5ms with
early_app_phase_offset_ns such that app still has the same amount of
time to handle each frame.
Test: 1. Open Chrome from homescreen
2. Swipe vertically to go back to home screen
3. Measure the gpu completion time and "FrameMissed" count from systrace result
Bug: 147339695
Change-Id: I157cb11e8ed8c1ca296fc65612752fbb3d985dcd
Since we frequently hit GL composition during transitions, it
makes sense to enable early phase offsets on crosshatch.
Test: Open apps, observe no jank
Bug: 79173695
Change-Id: Ie44b116b1a4f2c8fde1ba4decbbc6e3316686a5d
super.img is designed to be used for factory use, which is not
meaningful for AOSP targets. And it would cause build failure when
building w/o vendor.img.
Bug: 135752763
Test: `m dist PRODUCT-aosp_coral-userdebug` no longer generates
super.img.
Test: Build the CL in a branch that doesn't have vendor.img.
Change-Id: If020ef21ca8881c1765b38ec8fa78ead1ded97ee
* restores original logic and fixes this error:
08-27 02:38:48.695 W/libperfmgr(814): Failed to write to node: /sys/class/devfreq/soc:qcom,cpu-llcc-ddr-bw/min_freq with value: 8137, fd: 11
We're not using perfd so this is useless.
Signed-off-by: Subhajeet Muhuri <subhajeet.muhuri@gmail.com>
Change-Id: I3b436f3f747c6955335513a133da81d27a6fbf60
Allows us to check the idle state of the display in order to use perf
hints to improve system performance.
Bug: 124470767
Test: Booted and cated the affected files
Change-Id: Ib22b5209c57725c50f3545c578af7b72e7f08322
It is no longer a surprise for anyone to see Xiaomi shipping the same
device with different displays and different displays usually mean
different touchscreen drivers.
The current double tap handling in the power HAL only takes into
account one specific node, meant to be set in the device tree.
However, if the touchscreen was registered as a different input device
it would cause the feature to be broken. Address the issue by iterating
through the input devices in order to find one supported touchscreen
input to be used.
Co-authored-by: LuK1337 <priv.luk@gmail.com>
Change-Id: I2f7fb4a8b0c9cd1a16d6c2b93602d285a191f170