composeLayerRect() pays attention to the offset of the layer, but
composeLayerRegion() ignores it. This patch makes the behavior of
the region case match the behavior of the rect case.
The difference will only be shown by tests which do not position
the saved layer at the origin.
BUG=20288561
R=ccraik@google.com,djsollen@google.com
Change-Id: Ie22737a8614c508cbffe39aa4c55f7d5434277aa
Switch a few places to using android::canvas
instead of SkCanvas as well which eliminated
some JNI
Change-Id: I8f98b56442a06362b82b984cd1bd3a92398d8dbc
Canvas drawText* APIs are inconsistent, and this mismatch snuck through
our tests.
BUG=skia:3636
R=djsollen@google.com
Change-Id: Ibaab512c3f19df8445d582633b2748f1ec4d2902
Also tune the parameters to match a better look designed by UX team.
For ambient part, the acos transformed alpha is not needed any more.
b/19370798
Change-Id: I20d2615d1be502eb900942ed756aecd1660dee48
Add CLOCK_MONOTONIC to signal
when the stat collection started
to allow for tracking stats
over time
Change-Id: Iab7c52f21184eea603ddc8572b9d91a951b45a00
ClipArea's region mode implementation store the region's boundary in
a rectangle, and if the region is that rectangle, set its region to
empty. This is a bug, since as long as the ClipArea is in region
mode it treats its region as determinitive. When the next clipping
operation comes in, it is tested against the empty region instead of
against the valid rect.
Discovered via Skia GM testing.
Change-Id: I3bbc4e81c8d68636061f830156e017e9b5c7e587
More S's for More Speed
Split JankTracker's backing data from the
class to allow for data relocation to/from ashmem regions
Pack the jank tracking data to fit in 256 bytes
Change-Id: Ife86a64b71a328fbd0c8075fe6a0404e081f725b
Also add additional compile time checks to ensure that the Java
and Skia APIs remain in sync.
bug:19890753
Change-Id: I8503cacf2859307e3f480a78603f9f05901b58cc
pthread_cond_signal does not need the mutex
to be held to signal. This results in the thread
waking up trying to grab the lock, failing, the signaling
thread to wake up, release the lock, immediately get preempted
for the signaled thread which can now proceed.
Release the mutex before signaling to avoid the ping-pong
scheduling issue, which shaves another 10us off of this
Change-Id: Ie6bccca031ba6528f357eae8352b74626a6318c7
This prevents an issue where if the signal schedules
hwuiTask it will immediately block and go back to sleep due
to mLock still being held.
This costs 10us in thread scheduling ping-ponging bouncing
between hwuiTask and RenderThread
Change-Id: I47595c1bf5736576483a6aa7aada0b1be1e04268
To obtain the gfxinfo for each process, the static method of RenderProxy is used, which is named outputLogBuffer().
In there,
1. RenderTask is created for getting DisplayList Commands in RenderNode.
2. staticPostAndWait() is called
3. RenderThread's instance is created by 'RenderThread::getInstance()' in staticPostAndWait()
In case of the service, they don't use HW Acceleration, so don't need RenderThread.
But, by the process of No.3, RenderThread is created for all process.
As we know, RenderThread never be destroyed while the process is alive.
This patch checks RenderThread instance before the creation of RenderTask.
And, there is no one, just return to prevent the unnecessay creation of it.
Change-Id: I4fe29d83c9ced3e8b67177c0874c5d8ee62e1870
New ShadowTask with the ShadowDescription key that already
exists in the shadow LruCache will leak as it is not being
added.
Fix adds check for the existing key that is common in the hwui
code but missing for the TessellationCache::precacheShadow
function.
Change-Id: I37fd5ec82f8b8da5d1ec0f2ab9fd04c5f8534367
This removes dependence on SkPath ptrs that HWUI does not control
the lifecycle of. This clears up some errors where the paths are
not generated from Java, but rather the Skia test suites.
Cherry-pick of a change that originally landed in master-skia and is
dependent on a skia merge (ag/655422).
Change-Id: I41b9797a2b0af5d6b4ea51891565469d4f1d832d
