Tweak the primary flutter build rule so that fuchsia is more similar to
other platforms in how tests and the shell are built.
Only embedder_unittests and GLFW tests are disabled on Fuchsia now.
TEST: Ran unittests on host/fuchsia; workstation on fuchsia
BUG: fxb/53847, fxb/54056
The embedder.h API layer is an implementation detail of the desktop
embeddings, not part of the public API surface, so should not be part of
the public symbol list for those libraries.
Added a new `FlutterEngineAOTData` argument to `FlutterProjectArgs`. Embedders can instantiate and destroy this object via the new `FlutterEngineCreateAOTData` and `FlutterEngineCollectAOTData` methods provided.
If an embedder provides more than one source of AOT data to `FlutterEngineInitialize` or `FlutterEngineRun` (e.g. snapshots as well as `FlutterEngineAOTData`), the engine will error out.
Resolves: https://github.com/flutter/flutter/issues/50778
This reverts commit ef9e7b1a1365c07ab0df2e2016c4442c3963c2c7 with the following changes to accommodate an embedder for whom the original optimizations caused issues:
* Ensure stable order in the backing stores presented to the embedder. This is a pessimization that will be reverted when the embedder migrates. Tracked in https://github.com/flutter/flutter/issues/51228
* Forego the optimization where the unused layers would be collected before allocation of new layers needs to happen. This is a pessimization that will be reverted when the embedder migrates. Tracked in https://github.com/flutter/flutter/issues/51229
More context in b/146142979.
During the implementation of custom compositor integration, the embedder gets
callbacks on the render thread to prepare render targets (framebuffers,
textures, etc) for the engine to render into, callbacks to present these render
targets along with platform managed contents, and, callbacks to collect render
targets once they can no longer be recycled by the engine in subsequent frames.
During these callbacks, the engine mandates the OpenGL state on the render
thread be preserved. This restriction has been the source of hard to isolate
issues where the embedder trampled on the OpenGL bindings state in the callback
but failed to restore state before control went back to the engine. Due to the
nature of the OpenGL API, such errors are easy to make and overlook. This patch
lifts the restriction from the embedder. Embedders may now freely work with the
OpenGL state in custom compositor callbacks and the engine will make sure to
disregard OpenGL bindings when control flows back to it.
Disregarding current OpenGL state has a certain performance penalty and the
majority of this patch handles refactoring various engine embedder components
such that this happens only once per frame. The most trivial version of this
patch would reset context bindings on every transition of control flow from the
embedder to the engine. However, that naive approach would have necessitated
more than 50 binding resets in existing unit-test cases (depending on the number
of platform view interleaving levels and render target recycling hit rates). In
this implementation, bindings will be reset only once per frame and this does
not depend on the number of platform views in the scene.
The majority of this patch is a refactoring of engine subsystems used in
`ExternalViewEmbedder::SubmitFrame` which is thoroughly documented with each
opportunity for the embedder to invalidate OpenGL state tagged.
The refactoring also enables the implementation of the following optimizations
to engine behavior which should aid in reducing the memory needed for the
creation of render targets. These optimization include:
* The engine will only ask the embedder for render targets in which it expects
to render into. This was a quirk in the way in which root and non-root render
targets were handled. The engine could require the embedder to create a render
target but then realize it didn’t have anything to render into it. In the
presentation callback, it would skip that render target. But the embedder
still had to allocate that extra render target. This will no longer be the
case and should reduce memory use.
* The engine may now skip always realizing (via the embedder render target
creation callback) and presenting the root render target. This was also a side
effect of the same quirk. Previously, the engine would always ask the embedder
to present the root render target even if it was empty. Since this is no
longer the case, few render targets should be allocated which will reduce
memory consumption.
* The engine will now ask the embedder to collect unused render targets before
it asks it to create new ones. The previous behavior was to ask the embedder
for new targets and then collect old ones. This would cause spikes in memory
use when the size of the render targets would change. These memory use spikes
should now be troughs.
* The previous render target cache also considered the platform view ID in cache
viability considerations (instead of just the size of the render target). This
was a bug which has been fixed. This should lead to better cache utilization
in some situations.
These optimizations are now codified in unit-tests and the updated test
expectations are a result of these optimizations now being in place.
* Fixes https://github.com/flutter/flutter/issues/50751
* Fixes https://github.com/flutter/flutter/issues/46911
* Fixes https://github.com/flutter/flutter/issues/43778
* Fixes b/146142979
This was only necessary when the Engine had to build in multiple buildroots
where the sources where checked out at different paths relative to the
buildroot. This is no longer the case and there are already cases GN rules
have been written that mix and match variable usage with the direct
specification of the path to the Flutter sources relative to the sole buildroot.
Tonic used to be used by multiple consumers outside of Flutter Engine. Due to
this, it has an unnecessary abstraction layer as well as utilities duplicated in
FML and other engine subsystems. The sole user of Tonic is now the Flutter
Engine. It is intended that the Flutter Engine team now owns this subsystem,
remove unnecessary utilities and document the headers. This is the first step in
the transition. No history is being imported as the initial history was already
lost in the transition of this component to fuchsia.googlesource. As this
component was unmaintained there, I could see no additional value in importing
the history of the patches there.
No functional change. Just moved the repo from //third_party to
//flutter/third_party and updates GN refs.
This is a duplicate of flutter/engine#13360 with the test switched to use the software backend instead of the GL backend.
After some debugging and testing on another GL embedder I think the issue with the test is some bug having to do with the GL implementation in the test harness specifically.
Fixesflutter/flutter#38903
Previously the cache was disabled on whether or not PlatformViews were
globally enabled. Instead track their existence in the view hierarchy
and only disable RasterCache if a PlatformView is actually present.
This used to only be handled correctly for non-root layer backing stores. This
was mostly a side effect of the fact that we used recording canvases instead of
rendering directly into the backing store. We now use recording canvases
consistently.
Fixes b/143464703
Fixes https://github.com/flutter/flutter/issues/43732
The earlier design speculated that embedders could affect the same
transformations on the layers post engine compositor presentation but before
final composition.
However, the linked issue points out that this design is not suitable for use
with hardware overlay planes. When rendering to the same, to affect the
transformation before composition, embedders would have to render to an
off-screen render target and then apply the transformation before presentation.
This patch negates the need for that off-screen render pass.
To be clear, the previous architecture is still fully viable. Embedders still
have full control over layer transformations before composition. This is an
optimization for the hardware overlay planes use-case.
Fixes b/139758641
This change sets up a "spying canvas" to try and detect empty canvases.
When using platform views with a custom embedder, if a platform view
overlay canvas is known to be empty we skip creating a compositor layer
for that overlay.
change_install_name.py was operating on framework library files
in-place, which breaks GN's timestamp analysis handling since a file
can't be both an input and output of an action. As a result no-op builds
on macOS were not actually no-ops.
This changes the script to operate on an output copy, both fixing the
no-op build issue, and simplifying the GN framework construction scripts
by combining the copy step and the install-name step.
Fixes https://github.com/flutter/flutter/issues/33465
The root canvas is managed by the external view embedder when using a custom
compositor. Due to this, frame submission on the surface will not end up
flushing the same (because the surface doesn’t have it to begin with). Fixed
with tests.
This patch allows embedders to split the Flutter layer tree into multiple
chunks. These chunks are meant to be composed one on top of another. This gives
embedders a chance to interleave their own contents between these chunks.
The Flutter embedder API already provides hooks for the specification of
textures for the Flutter engine to compose within its own hierarchy (for camera
feeds, video, etc..). However, not all embedders can render the contents of such
sources into textures the Flutter engine can accept. Moreover, this composition
model may have overheads that are non-trivial for certain use cases. In such
cases, the embedder may choose to specify multiple render target for Flutter to
render into instead of just one.
The use of this API allows embedders to perform composition very similar to the
iOS embedder. This composition model is used on that platform for the embedding
of UIKit view such and web view and map views within the Flutter hierarchy.
However, do note that iOS also has threading configurations that are currently
not available to custom embedders.
The embedder API updates in this patch are ABI stable and existing embedders
will continue to work are normal. For embedders that want to enable this
composition mode, the API is designed to make it easy to opt into the same in an
incremental manner.
Rendering of contents into the “root” rendering surface remains unchanged.
However, now the application can push “platform views” via a scene builder.
These platform views need to handled by a FlutterCompositor specified in a new
field at the end of the FlutterProjectArgs struct.
When a new platform view in introduced within the layer tree, the compositor
will ask the embedder to create a new render target for that platform view.
Render targets can currently be OpenGL framebuffers, OpenGL textures or software
buffers. The type of the render target returned by the embedder must be
compatible with the root render surface. That is, if the root render surface is
an OpenGL framebuffer, the render target for each platform view must either be a
texture or a framebuffer in the same OpenGL context. New render target types as
well as root renderers for newer APIs like Metal & Vulkan can and will be added
in the future. The addition of these APIs will be done in an ABI & API stable
manner.
As Flutter renders frames, it gives the embedder a callback with information
about the position of the various platform views in the effective hierarchy.
The embedder is then meant to put the contents of the render targets that it
setup and had previously given to the engine onto the screen (of course
interleaving the contents of the platform views).
Unit-tests have been added that test not only the structure and properties of
layer hierarchy given to the compositor, but also the contents of the texels
rendered by a test compositor using both the OpenGL and software rendering
backends.
Fixes b/132812775
Fixesflutter/flutter#35410
embedder.h is a C API, so has no namespace, and only uses 'Flutter' as a
prefix for most symbol names. This creates potential collisions with
other code; for instance, FlutterEngine is the name of a type in
embedder.h, but also an ObjC class in the iOS Flutter API.
This adds a macro that can be set to prefix symbol names, allowing
clients (notably, the macOS embedding) to adjust the names used by the
embedding API internally without breaking ABI or API compatibility for
the standard engine build.
Currently the macro is only applied to FlutterEngine, since that's the
symbol that is currently at issue, but it can be expanded to other
symbols in the future.
This patch reworks image decompression and collection in the following ways
because of misbehavior in the described edge cases.
The current flow for realizing a texture on the GPU from a blob of compressed
bytes is to first pass it to the IO thread for image decompression and then
upload to the GPU. The handle to the texture on the GPU is then passed back to
the UI thread so that it can be included in subsequent layer trees for
rendering. The GPU contexts on the Render & IO threads are in the same
sharegroup so the texture ends up being visible to the Render Thread context
during rendering. This works fine and does not block the UI thread. All
references to the image are owned on UI thread by Dart objects. When the final
reference to the image is dropped, the texture cannot be collected on the UI
thread (because it has not GPU context). Instead, it must be passed to either
the GPU or IO threads. The GPU thread is usually in the middle of a frame
workload so we redirect the same to the IO thread for eventual collection. While
texture collections are usually (comparatively) fast, texture decompression and
upload are slow (order of magnitude of frame intervals).
For application that end up creating (by not necessarily using) numerous large
textures in straight-line execution, it could be the case that texture
collection tasks are pending on the IO task runner after all the image
decompressions (and upload) are done. Put simply, the collection of the first
image could be waiting for the decompression and upload of the last image in the
queue.
This is exacerbated by two other hacks added to workaround unrelated issues.
* First, creating a codec with a single image frame immediately kicks of
decompression and upload of that frame image (even if the frame was never
request from the codec). This hack was added because we wanted to get rid of
the compressed image allocation ASAP. The expectation was codecs would only be
created with the sole purpose of getting the decompressed image bytes.
However, for applications that only create codecs to get image sizes (but
never actually decompress the same), we would end up replacing the compressed
image allocation with a larger allocation (device resident no less) for no
obvious use. This issue is particularly insidious when you consider that the
codec is usually asked for the native image size first before the frame is
requested at a smaller size (usually using a new codec with same data but new
targetsize). This would cause the creation of a whole extra texture (at 1:1)
when the caller was trying to “optimize” for memory use by requesting a
texture of a smaller size.
* Second, all image collections we delayed in by the unref queue by 250ms
because of observations that the calling thread (the UI thread) was being
descheduled unnecessarily when a task with a timeout of zero was posted from
the same (recall that a task has to be posted to the IO thread for the
collection of that texture). 250ms is multiple frame intervals worth of
potentially unnecessary textures.
The net result of these issues is that we may end up creating textures when all
that the application needs is to ask it’s codec for details about the same (but
not necessarily access its bytes). Texture collection could also be delayed
behind other jobs to decompress the textures on the IO thread. Also, all texture
collections are delayed for an arbitrary amount of time.
These issues cause applications to be susceptible to OOM situations. These
situations manifest in various ways. Host memory exhaustion causes the usual OOM
issues. Device memory exhaustion seems to manifest in different ways on iOS and
Android. On Android, allocation of a new texture seems to be causing an
assertion (in the driver). On iOS, the call hangs (presumably waiting for
another thread to release textures which we won’t do because those tasks are
blocked behind the current task completing).
To address peak memory usage, the following changes have been made:
* Image decompression and upload/collection no longer happen on the same thread.
All image decompression will now be handled on a workqueue. The number of
worker threads in this workqueue is equal to the number of processors on the
device. These threads have a lower priority that either the UI or Render
threads. These workers are shared between all Flutter applications in the
process.
* Both the images and their codec now report the correct allocation size to Dart
for GC purposes. The Dart VM uses this to pick objects for collection. Earlier
the image allocation was assumed to 32bpp with no mipmapping overhead
reported. Now, the correct image size is reported and the mipmapping overhead
is accounted for. Image codec sizes were not reported to the VM earlier and
now are. Expect “External” VM allocations to be higher than previously
reported and the numbers in Observatory to line up more closely with actual
memory usage (device and host).
* Decoding images to a specific size used to decode to 1:1 before performing a
resize to the correct dimensions before texture upload. This has now been
reworked so that images are first decompressed to a smaller size supported
natively by the codec before final resizing to the requested target size. The
intermediate copy is now smaller and more promptly collected. Resizing also
happens on the workqueue worker.
* The drain interval of the unref queue is now sub-frame-interval. I am hesitant
to remove the delay entirely because I have not been able to instrument the
performance overhead of the same. That is next on my list. But now, multiple
frame intervals worth of textures no longer stick around.
The following issues have been addressed:
* https://github.com/flutter/flutter/issues/34070 Since this was the first usage
of the concurrent message loops, the number of idle wakes were determined to
be too high and this component has been rewritten to be simpler and not use
the existing task runner and MessageLoopImpl interface.
* Image decoding had no tests. The new `ui_unittests` harness has been added
that sets up a GPU test harness on the host using SwiftShader. Tests have been
added for image decompression, upload and resizing.
* The device memory exhaustion in this benchmark has been addressed. That
benchmark is still not viable for inclusion in any harness however because it
creates 9 million codecs in straight-line execution. Because these codecs are
destroyed in the microtask callbacks, these are referenced till those
callbacks are executed. So now, instead of device memory exhaustion, this will
lead to (slower) exhaustion of host memory. This is expected and working as
intended.
This patch only addresses peak memory use and makes collection of unused images
and textures more prompt. It does NOT address memory use by images referenced
strongly by the application or framework.
Currently, all our host unit-tests that have rendering concerns use the software backend because of OpenGL ES availability and stability issues on the various platforms where we run host tests. Unfortunately, entire subsystems are disabled (and not tested) when rendering with the software backend. This patch pulls in SwiftShader and via pending patches in the buildroot, configures the host unit-tests to optionally use OpenGL ES in a stable manner without relying on the OpenGL drivers being present (and functional).
I have wired up the embedder test fixture in this patch to use the SwiftShader based OpenGL ES driver. I will update the shell and runtime unittests in a subsequent patch as well. The on and offscreen surfaces are configured as 1x1 pbuffer surface because we should be able to write pixel tests using OpenGL directly wihout having to deal with surfaces.
* Allow specifying both Dart and non-Dart fixtures in engine unittests.
This fixes numerous issues in the way in which fixtures were managed
in the engine unit-tests.
* Instead of only being able to specify Dart fixtures, unit-tests may specify
non-Dart fixtures as well. These are simply copied over to the fixtures
directory known to the unit-test at runtime.
* An issue where numerous Dart files could be given to the kernel snapshotter
has been addressed. It was anticipated that such a (legal) invocation to the
kernel snapshotter would produce a snapshot with the contents of all the Dart
files added to the root library. This is incorrect and the behavior in this
case is undefined.
* Dart files referenced by the main Dart file are correctly tracked via a
depfile.
* The snapshotter arguments have been cleaned up to get rid of unused
arguments (`—strong`) and the use of the VM product mode argument has been
corrected to no longer depend on the Flutter product mode.
Currently, all Flutter threads are managed by the engine itself. This works for
all threads except the platform thread. On this thread, the engine cannot see
the underlying event multiplexing mechanism. Using the new task runner
interfaces, the engine can relinquish the task of setting up the event
multiplexing mechanism and instead have the embedder provide one for it during
setup.
This scheme is only wired up for the platform thread. But, the eventual goal
is to expose this message loop interoperability for all threads.
All embedder unit-tests have to setup the Flutter project arguments from scratch
before launching the engine. The boilerplate and having to deal with the low
level C API during each engine launch is a hinderance to writing tests.
This patch introduces an EmbedderTest fixture that sets up all the embedder side snapshots before allowing the unit test to create a FlutterConfigBuilder` that
the test can use to incrementally build and edit the Flutter project
configuration. From the given state state of a configuration, multiple engines
can be launched with their lifecylces managed by appropriate RAII wrappers.
This allows the a fully configured Flutter engine to be launched using 4 lines
of code in a fixture.
```
EmbedderConfigBuilder builder;
builder.SetSoftwareRendererConfig();
builder.SetAssetsPathFromFixture(this);
builder.SetSnapshotsFromFixture(this);
auto engine = builder.LaunchEngine();
```
