| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
| |
This change is a temporary workaround for a buffer synchronization issue.
Bug: 6876502
Change-Id: I030b26818ba7e78037a38dcaddd18782cd086b75
|
|
|
|
|
|
|
|
| |
bug:7393228
Single tile per frame is too slow
Change-Id: Ife93976411a54c43313ab2f80706a1ec125e507a
|
|
|
|
|
|
| |
bug:7247750
Change-Id: I8238acc2c20942ab2f42936d16a03226909aebcd
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The function isn't threadsafe, so races would occur and certain textures
wouldn't be removed from ImagesManager's master list. The list would then be
iterated over, all items dereferenced, and bad times would ensue.
The SkRefCnt class uses atomic inc/dec to decide when to destroy an object, but
reading such an integer non-atomically isn't safe. Instead use the real signal
we're looking for - when the ImageTexture is deleted - to know when to remove
the ImageTexture from ImagesManager's list.
Mutual exclusion from editing the list is now maintained by only unref-ing the
ImageTexture within releaseImage, which holds the ImagesManager's m_imagesLock
(which we already do anyway).
bug:6859278
Change-Id: I75ebf79f2617484e7df355d6539226ce64882369
|
|\
| |
| |
| | |
into jb-mr1-dev
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
bug:7002755
This prevents the SurfaceCollectionManager from getting stuck in fast swap mode,
where because swaps are occuring in process mode, and not in draw mode, which
would previously have prevented the flag from being cleared.
Change-Id: Ie888438c73aabeeff0534caad68b58e4acd2c440
|
|/
|
|
|
| |
bug:6968485
Change-Id: I8691d34fc5d4da1d03f31fd45c2d6d6abc39c53b
|
|
|
|
|
|
|
|
| |
bug:6843244
Current implementation only works for HW acceleration.
Change-Id: I009090132a8f3859ab9e57e609c1754f9021d4d7
|
|
|
|
|
|
|
| |
Fast path for same-color rect drawing in InstrumentedPlatformCanvas
Also minimize unnecessary clipping recording-side.
Change-Id: I8e9ebcfce8d137ca0b27d17c42c5045a0d3b3a36
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
bug:6947623
Multiple shaders share the opacity variable, so reset the cache when swapping
programs.
Change-Id: I752575d6089077aa8034c171c5c016e5e4e80606
|
| |
| |
| |
| |
| |
| | |
Everything goes through either PlatformGraphicsContext or recordingCanvas()
Change-Id: I375a4294d2e8d4b467b70c6b8a7f0b96f402f252
|
|\ \
| |/
|/|
| |
| |
| |
| | |
tiles are rendered black"
* commit 'dd5962d2534b8c3cc215492c8dcf463e923f73c2':
Fix browser rendering issue: pure color tiles are rendered black
|
| |\
| | |
| | |
| | |
| | |
| | |
| | | |
rendered black"
* commit '47387cc7716048dcea1cc5f9bf37da65d38e10be':
Fix browser rendering issue: pure color tiles are rendered black
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The “pure color tile” optimization finds the pure-color tiles and
then draws the tiles simply with the color (otherwise the texture).
But the RGBA color pointer type is char, which may overflow when
the value is larger than 127. In such case, the tiles in question
are rendered black, because of the wrong color. This bug is fixed
by defining the pointer as unsigned char.
Change-Id: I5c7214423fa6961e707bde72f0e18c610374745d
Author: Yuyang Du <yuyang.du@intel.com>
Signed-off-by: Yuyang Du <yuyang.du@intel.com>
Signed-off-by: Jack Ren <jack.ren@intel.com>
Signed-off-by: Bruce Beare <bruce.j.beare@intel.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Bug:6875992
Initialize the InstrumentedPlatformCanvas with the correct background color so
that if no painting is done, the background color is returned as the pure color.
Change-Id: Ibe928a4e158f0a4302e0996eca88c3e8b1f1b666
|
|\ \ \
| |/ /
| | |
| | |
| | | |
* commit 'f35e08fefff71bb9b84b679f8d13cf57c6e1d756':
Null check to fix crash
|
| |/
| |
| |
| |
| |
| |
| | |
Need to make sure that img is not zero.
bug: 6864545
Change-Id: I2a2c5440347681b462abaa5ac9101ddb6137c8c4
|
| |
| |
| |
| | |
Change-Id: Ibaa44ce9c24bdf05de0963efe12d9123113b08b6
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Previously toggling hasText for the content of the SurfaceBacking was enough to
cause it to change scale entirely each time the value was changed. Now the
allowZoom flag is sticky to avoid unnecessary full-SurfaceBacking repaints.
Additionally, changes from allowZoom true->false will not trigger delay.
Change-Id: Ie6a13850f86d21f5a12a184c08b940aa2156ef07
|
| |
| |
| |
| |
| |
| |
| |
| | |
Fall back to standard brute-force method.
InstrumentedPlatformCanvas from upstream webkit trunk/Source/WebCore/platform/graphics/blackberry/InstrumentedPlatformCanvas.h as of changeset 113390 ( http://trac.webkit.org/changeset/113390 )
Change-Id: I7b3884b8284d896c0948d23b6ec9fe0d2d98a6d5
|
| |
| |
| |
| | |
Change-Id: I1e8f607a94685f3aa8f63e40fe2a265bca8a084e
|
|\ \ |
|
| | |
| | |
| | |
| | | |
Change-Id: I39a170ce120e4ce5237e1cfc2b0e94cdb42c6124
|
|/ /
| |
| |
| |
| |
| | |
Swap the SkBitmap instead of a copy to save the memory bandwidth.
Change-Id: I0998248a9a87ce4058898e66a4739995da9c003c
|
| |
| |
| |
| |
| |
| |
| | |
Move thread local bitmaps to renderers, so the renderers own them, and can
perform optimizations.
Change-Id: I60d604759f4ce2ec9981990d1613eaf23cd1e625
|
| |
| |
| |
| | |
Change-Id: Ie9d941bd59584b8a77b90e1f41b2c49a0e02e572
|
| |
| |
| |
| |
| |
| |
| | |
Use one thread by default. Further experiments should be done with determining
the correct number of threads, and better scheduling.
Change-Id: I2a87dac7d0b652b5ab627744650d12b0b8842003
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Because no new work is enqueued after a SurfaceCollectionManager::swap(), there
can be significant TexturesGenerator idle period before the next prepare of the
new painting tree. This change removes that delay by preparing the new tree
immediately.
Change-Id: I061983abe7210f75d5e4f44d5dca79a0a47d329f
|
|\ \ |
|
| |/
| |
| |
| | |
Change-Id: I2a53b157d40c630e7ca1361f8e0c8aca1d6e8a4e
|
|/
|
|
|
| |
bug:6653638
Change-Id: I839917feb1803bc1f4ace5aa2367067024a18706
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
An issue occurs on Stingray where calling glTexSubImage2d in order to blit an
update to a texture fails silently if the texture has already been updated this
frame.
With this change, we now draw the tile off-screen to work around the issue just
before the second blit is required. This way, the second blit succeeds.
bug:6630040
Change-Id: I07213954669b0f3d3f5b87a748a591e532b097b4
|
|/
|
|
|
|
|
| |
bug:6579681
Regression from https://android-git.corp.google.com/g/#/c/197452/
Change-Id: If185faec6bb2d014c304afbb8ff752ebfc6e4b51
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
In the slow draw path, we are drawing the small background image one by one.
Now in the fast path, we can use the GL shader to repeat the iamges.
bug:6579681
Change-Id: I8ab413bdd073cb867002b77b36d4a1361290ecb2
|
| |
| |
| |
| |
| |
| |
| |
| | |
Fixes an issue where tiles outside the new bounds weren't invalidated, as the
layer shrank in one dimension.
bug:6609350
Change-Id: I1f6715670654c9e6211dbec71875aa2ea11cd8f8
|
| |
| |
| |
| |
| |
| |
| |
| | |
Base surface can expand dynamically (see BaseLayerAndroid::updatePosition) to
fill the viewport, so disable clipping on it.
bug:6593707
Change-Id: I3afafb28fbe95d3162c76f8e835932577ded65ce
|
|/
|
|
|
|
| |
bug:5683630
Change-Id: I4d617962a8412de884903f150799f733ea996e1c
|
|\ |
|
| |
| |
| |
| |
| |
| | |
bug:5683630
Change-Id: I43f738f2649a79b4ad7865ed27375c07195fa9b9
|
|\ \
| |/
|/| |
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
If the current clip covers the entire view, we don't have to calculate framework
invals from swapping layers - we already draw them immediately.
bug:6530473
Change-Id: I4b639e06781371e37c246f3220b79f28e7d95649
|
|/
|
|
|
|
|
|
| |
Using m_scale would mean that while zooming out, the clipped area would get
larger, but the scale would stay the same.
bug:6561401
Change-Id: I0806ecb8f565c4d227a2be2b7a27a0af6aff7b46
|
|
|
|
|
| |
bug:6535726
Change-Id: Id944448a9dbafa951fd3c7eab0f248380d0a5314
|
|\ |
|
| |
| |
| |
| |
| |
| |
| |
| | |
Previously we used the surface's first layer's area, which is often incorrect in
the case of significant layer merging.
bug:6545187
Change-Id: Ied3e5e3cd894b145ed905f8ba83649b3c4f3b1dd
|
|\ \ |
|
| |/
| |
| |
| |
| |
| | |
bug:6530473 bug:6344165
Change-Id: I40b190036af8c021134d5ac450e3217098ba2296
|