Using reinterpret_cast allows conversions between pointer types
and integers which don't have the same size. The corresponding
flags is -Wint-to-pointer-cast.
Change-Id: I8ff0c79d235fa6d07b8d9305edc185d946d5133d
Signed-off-by: Kévin PETIT <kevin.petit@arm.com>
This will allow us to inspect the paint for thing other than
color and xfermode, such as SkColorFilters and SkShaders.
bug: 10650594
Change-Id: I2c3ddd07a3966e1e77af34136307e2b59b2898c1
Somewhat unifies the ortho/perspecive paths - the property matrix
(translate/scale/rotate) is now always a Matrix4.
Change-Id: I36e4fe83d1150ee6e4be5f64f34d0fc8d6525cc6
Fixes the simplifying assumption that shadow casters were always
rectangular.
Java side APIs + plumbing to pass down correct shapes still need to be added.
Change-Id: Ic4fee90af15679951a44bb5cc6ae454b98c4c194
For now, ancestor views signal the acceptance of projections with a
save(0x20)/restore pair.
During the order traversal, each view with the save(0x20) code will
collect descendent views with mProjectToContainedVolume (which still
needs to be renamed) so that they can be drawn out of order later.
- *Temporary* sample code added to HwAccelerationTest.
- Note that a projected displaylist must not be clipped.
Change-Id: I45c493e845961535b958d59c53e8aff3f8891d9f
True 3d transformations are now supported by DisplayLists and the
renderer, initially with the translationZ property on view.
Renderer operations used directly by DisplayList (formerly,
clip/save/restore/saveLayer) are now more simply managed by allocating
them temporarily on the handler's allocator, which exists for a single
frame. This is much simpler than continuing to expand the pool of
pre-allocated DisplayListOps now that more operations are called
directly by DisplayList, especially with z ordered drawing.
Still TODO:
-APIs for camera positioning, shadows
-Make Z apis public, and expose through XML
-Make invalidation / input 3d aware
Change-Id: I95fe6fa03f9b6ddd34a7e0c6ec8dd9fe47c6c6eb
bug:4351353
quickReject and quickRejectNoScissor have been renamed and refactored:
- to make the scissor side effect clear and explicit
- dangerous methods no longer public
- to make the simple quick reject check logic const
- simple quick reject is now conservative
This CL also fixes several issues with line and point quickRejection -
sub-pixel and hairline lines are much less likely to be incorrectly
rejected, especially at small canvas scale.
Additionally, alpha modulation for AA points < 1px in size is now
correct, dumplicating SW behavior (similar to lines and stroked
shapes work).
Change-Id: Ibb0710c721b9fb415d05acf54dd3d2b4d602156a
bug:10863163
This fixes two issues
The check for pure translation was incorrect. It was fixed and renamed
for clarity.
Certain matrix paths weren't setting kTypePositiveScale. For
simplicity (and because positive scale is simple to check) removed
flag in favor of dynamic checking.
Change-Id: Ic5ce235653ef49a68b8b242bd89fc2e95874ecc9
bug:9969358
Instead of storing DeferredDisplayState within an op (thus forcing ops
to be tied to a single state instance), associate each op with a new
state at DeferredDisplayList insertion time.
Now, DisplayLists (and the ops within) can be reused in a single
DeferredDisplayList draw call, as ops will use different state
instances at different points in the frame.
Change-Id: I525ab2abe0c3883679f2fa00b219b293e9ec53d9
Bug #10185769
The assets atlas contains assets that need to be blended and assets
that do not need to be blended. With a single merge id, currently
set to be the pointer to the atlas itself, draw ops merging could
generate batches of commands containing both opaque and translucent
assets. The blend state was chosen from only one of the assets in
the batch, leading either to inefficiencies (blending large opaque
assets) or incorrect behaviors (not blending translucent assets.)
This change introduces two new merge ids in the atlas: an opaque
key and a blend key. These keys are simple booleans set to false
and true respectively (the values do not matter really.) Their
memory addresses are used as the merge ids when createing draw ops
batches, allowing all opaque ops to be batched together and all
translucent ops to be batched together.
Change-Id: I114dba0533c44987e53864b471ccb28c811f2025
By having quickReject round out the window-space geometry bounds, we
prevent the AA perimeter (which falls outside the local bounds passed
in) from drawing outside the clip.
Change-Id: I8ee36be9039a9c47906815ee2f0dbaa5eb910b82
setMatrix() was crashing in native code, only with hw acceleration on.
concat() would throw a NullPointerException. It now ignores null matrices.
Change-Id: Iebd8b410a957d2ba501570c6fbb3f680ff4a1a23
Bug #9589379
If the atlas is terminated/reinitialized and a view does not invalidate
in between it might end up using a stale AssetAtlas::Entry. This change
is similar to how 9patch meshes are cached in DrawPatchOp: we simply
track the generation ID of the cache to make sure we always use the
latest data.
Change-Id: Ib5abb3769d2ce0eabe9adc04e320ca27c422019e
This change also fixes the way batched bitmaps were handled
inside a layer. The layer is now correctly dirtied to minimize
the amount of pixels to blend.
Fix alpha, mode and opaque computations for DrawPatchOp.
Change-Id: I1b6cd581c0f0db66c1002bb4fb1a9811e55bfa78
We were treating immediate mode bounds as pre translate, which is
inconsistent with using them for quickRejection.
This fixes the overdraw counter not drawing correctly (since it uses
immediate mode drawing.
Change-Id: I1c734d367a00942bd7d9b041822c0a9f284e70a8
bug:9464358
Previously, empty and unknown sized ops are assumed to fully cover
their clip. This is now corrected such that empty sized ops are
pre-rejected before defer. Additionally, unknown sized ops disable
overdraw avoidance.
Change-Id: Icf2ce24f98be5ea6299e24ffcf826790373564a1
bug:8951267
If an opaque op, or group of opaque ops covers the invalidate region,
skip draw operations that precede it.
Clipped operations may now be merged, but only if they share a
clipRect - this is a very case for e.g. ListView, where all background
elements may now be a part of the same MergingDrawBatch.
It is this more aggressive merging that groups together clipped
background elements in the ListView case, enabling the overdraw
avoidance skipping the window background.
Change-Id: Ib0961977e272c5ac37f59e4c67d828467422d259
This optimization saves up to 0.3ms per frame on the Play Store's
front page, on a Nexus 4 device.
Change-Id: Iaa4ef33c6e3b37e175efd5b9eea9ef59b43f14f3
bug:8766924
Previously text bounds were calculated to be from 0 to totalAdvance in
the X, and from the font's top to bottom. These are incorrect,
especially in light of the font fallback mechanism.
Now, we calculate the bounds of the text as we layout each glyph.
Since these are much tighter bounds in the common case, this
significantly reduces the amount of clipping required (which in turn
enables more aggressive text merging).
Change-Id: I172e5466bf5975bf837af894a9964c41db538746
When the Android runtime starts, the system preloads a series of assets
in the Zygote process. These assets are shared across all processes.
Unfortunately, each one of these assets is later uploaded in its own
OpenGL texture, once per process. This wastes memory and generates
unnecessary OpenGL state changes.
This CL introduces an asset server that provides an atlas to all processes.
Note: bitmaps used by skia shaders are *not* sampled from the atlas.
It's an uncommon use case and would require extra texture transforms
in the GL shaders.
WHAT IS THE ASSETS ATLAS
The "assets atlas" is a single, shareable graphic buffer that contains
all the system's preloaded bitmap drawables (this includes 9-patches.)
The atlas is made of two distinct objects: the graphic buffer that
contains the actual pixels and the map which indicates where each
preloaded bitmap can be found in the atlas (essentially a pair of
x and y coordinates.)
HOW IS THE ASSETS ATLAS GENERATED
Because we need to support a wide variety of devices and because it
is easy to change the list of preloaded drawables, the atlas is
generated at runtime, during the startup phase of the system process.
There are several steps that lead to the atlas generation:
1. If the device is booting for the first time, or if the device was
updated, we need to find the best atlas configuration. To do so,
the atlas service tries a number of width, height and algorithm
variations that allows us to pack as many assets as possible while
using as little memory as possible. Once a best configuration is found,
it gets written to disk in /data/system/framework_atlas
2. Given a best configuration (algorithm variant, dimensions and
number of bitmaps that can be packed in the atlas), the atlas service
packs all the preloaded bitmaps into a single graphic buffer object.
3. The packing is done using Skia in a temporary native bitmap. The
Skia bitmap is then copied into the graphic buffer using OpenGL ES
to benefit from texture swizzling.
HOW PROCESSES USE THE ATLAS
Whenever a process' hardware renderer initializes its EGL context,
it queries the atlas service for the graphic buffer and the map.
It is important to remember that both the context and the map will
be valid for the lifetime of the hardware renderer (if the system
process goes down, all apps get killed as well.)
Every time the hardware renderer needs to render a bitmap, it first
checks whether the bitmap can be found in the assets atlas. When
the bitmap is part of the atlas, texture coordinates are remapped
appropriately before rendering.
Change-Id: I8eaecf53e7f6a33d90da3d0047c5ceec89ea3af0
The rendering code optimizes by rejecting drawing operations that
lie outside of the bounds of their views. This works in most
situations, but breaks down when containers have called
setClipChildren(false), because we reject drawing that is outside
of that container, but which should be drawn anyway.
Fix is to pass in the value of that flag to the DisplayList drawing
routines which take that flag into account when deciding whether
to quickReject any particular operation.
Issue #8659277 animation clipping
Change-Id: Ief568e4db01b533a97b3c5ea5ad777c03c0eea71
bug:8649215
Previously we prevented ops with non-translate transforms from
merging, but missed the case of the first op in a merging batch
containing a non-translate transform.
This fulfills the assumption of drawText's non-immediate mode that
merged ops will have pure translate transforms.
Change-Id: I6f6db341aff3f7e84e74b4c3ccf970d585a2db1a
Merge simple bitmap draw operations and text operations to avoid
issuing individual gl draws for each operation. Merging other ops to
be done eventually.
The methods are different - the bitmap merging generates a single
mesh for reused, unclipped images (esp. repeated images in a listview)
The text approach queries just defers the normal font rendering until
the last drawText in the sequence that can share the same shader.
Patches are sorted and merged, but don't yet have a multiDraw
implementation. For now, the pretending-to-merge gives better sorting
behavior by keeping similar patches together.
Change-Id: Ic300cdab0a53814cf7b09c58bf54b1bf0f58ccd6
bug:8464795
Changes drawModifiers and alpha to be restored for all operations,
since saveLayer/restore use these values, not just draw operations
Also forces a renderer state restoration before a deferred restore op
is played back, in case it is associated with a saveLayer that
doesn't have the save_clip flag set
Change-Id: I9da5d44fefbfffdee164c98f4f139843dacf85df
bug:8409891
In order to defer fbo-targetting saveLayer operations, it's necessary
to update the snapshot at defer time so that deferred display state
(namely, clip and transform) are fbo relative.
Re-enables deferring, as the issues with saveLayer are fixed.
Change-Id: I74b0779bc732675c747208f0757c3ea85f6dfbed
Removes mMultipliedAlpha, using the snapshot alpha for all
non-overlapping display list alpha control.
Additionally, fixes opacity issues where children of
hasOverlappingRendering=false displaylists (both hw layer sublists and
other sublists with hasOverlappingRendering=false)
Change-Id: I6adc16da855835f9f518f8967628e5d0135c789b
bug:8037003
A recursive drawDisplayList call is now entirely deferred before
playing back to the screen and issuing GL commands. This way, the
entire stream can be inspected, optimized, and batch work (such as
uploading textures) before issuing commands.
Additionally, this fixes an issue where operations draw could move
across restores corresponding to saveLayer(alpha). Those and other
similar cases (such as complex clipping, requiring the stencil) are
now treated as batching barriers, with the operations that change
renderer state in a way that's difficult to defer are just re-issued
at flush time.
Change-Id: Ie7348166662a5ad89fb9b1e87558334fb826b01e
The renderer used to pre-cache glyphs at record time. This then changed
to pre-caching at the beginning of every frame. This unfortunately entails
a lot of duplicate work on every frame, which amounts to 0.5 to 1ms in
some stock applications.
This change is somewhere in the middle: pre-caching happens the first
time a DrawTextOp is deferred or every time the screen-space transform
is different from the last pre-caching operation.
Change-Id: Id6d9e2599d90a5b75010b0f0a28746befbf3c205
