flutter_flutter/shell/common/rasterizer.cc
Chinmay Garde e8f954409d
Allow embedder controlled composition of Flutter layers. (#10195)
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
Fixes flutter/flutter#35410
2019-08-13 14:53:19 -07:00

484 lines
16 KiB
C++

// Copyright 2013 The Flutter Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#include "flutter/shell/common/rasterizer.h"
#include "flutter/shell/common/persistent_cache.h"
#include <utility>
#include "third_party/skia/include/core/SkEncodedImageFormat.h"
#include "third_party/skia/include/core/SkImageEncoder.h"
#include "third_party/skia/include/core/SkPictureRecorder.h"
#include "third_party/skia/include/core/SkSerialProcs.h"
#include "third_party/skia/include/core/SkSurface.h"
#include "third_party/skia/include/core/SkSurfaceCharacterization.h"
#include "third_party/skia/include/utils/SkBase64.h"
namespace flutter {
// The rasterizer will tell Skia to purge cached resources that have not been
// used within this interval.
static constexpr std::chrono::milliseconds kSkiaCleanupExpiration(15000);
// TODO(dnfield): Remove this once internal embedders have caught up.
static Rasterizer::DummyDelegate dummy_delegate_;
Rasterizer::Rasterizer(
TaskRunners task_runners,
std::unique_ptr<flutter::CompositorContext> compositor_context)
: Rasterizer(dummy_delegate_,
std::move(task_runners),
std::move(compositor_context)) {}
Rasterizer::Rasterizer(Delegate& delegate, TaskRunners task_runners)
: Rasterizer(delegate,
std::move(task_runners),
std::make_unique<flutter::CompositorContext>()) {}
Rasterizer::Rasterizer(
Delegate& delegate,
TaskRunners task_runners,
std::unique_ptr<flutter::CompositorContext> compositor_context)
: delegate_(delegate),
task_runners_(std::move(task_runners)),
compositor_context_(std::move(compositor_context)),
user_override_resource_cache_bytes_(false),
weak_factory_(this) {
FML_DCHECK(compositor_context_);
}
Rasterizer::~Rasterizer() = default;
fml::WeakPtr<Rasterizer> Rasterizer::GetWeakPtr() const {
return weak_factory_.GetWeakPtr();
}
void Rasterizer::Setup(std::unique_ptr<Surface> surface) {
surface_ = std::move(surface);
if (max_cache_bytes_.has_value()) {
SetResourceCacheMaxBytes(max_cache_bytes_.value(),
user_override_resource_cache_bytes_);
}
compositor_context_->OnGrContextCreated();
if (surface_->GetExternalViewEmbedder()) {
const auto platform_id =
task_runners_.GetPlatformTaskRunner()->GetTaskQueueId();
const auto gpu_id = task_runners_.GetGPUTaskRunner()->GetTaskQueueId();
gpu_thread_merger_ =
fml::MakeRefCounted<fml::GpuThreadMerger>(platform_id, gpu_id);
}
}
void Rasterizer::Teardown() {
compositor_context_->OnGrContextDestroyed();
surface_.reset();
last_layer_tree_.reset();
}
void Rasterizer::NotifyLowMemoryWarning() const {
if (!surface_) {
FML_DLOG(INFO) << "Rasterizer::PurgeCaches called with no surface.";
return;
}
auto context = surface_->GetContext();
if (!context) {
FML_DLOG(INFO) << "Rasterizer::PurgeCaches called with no GrContext.";
return;
}
context->freeGpuResources();
}
flutter::TextureRegistry* Rasterizer::GetTextureRegistry() {
return &compositor_context_->texture_registry();
}
flutter::LayerTree* Rasterizer::GetLastLayerTree() {
return last_layer_tree_.get();
}
void Rasterizer::DrawLastLayerTree() {
if (!last_layer_tree_ || !surface_) {
return;
}
DrawToSurface(*last_layer_tree_);
}
void Rasterizer::Draw(fml::RefPtr<Pipeline<flutter::LayerTree>> pipeline) {
TRACE_EVENT0("flutter", "GPURasterizer::Draw");
if (gpu_thread_merger_ && !gpu_thread_merger_->IsOnRasterizingThread()) {
// we yield and let this frame be serviced on the right thread.
return;
}
FML_DCHECK(task_runners_.GetGPUTaskRunner()->RunsTasksOnCurrentThread());
RasterStatus raster_status = RasterStatus::kFailed;
Pipeline<flutter::LayerTree>::Consumer consumer =
[&](std::unique_ptr<LayerTree> layer_tree) {
raster_status = DoDraw(std::move(layer_tree));
};
PipelineConsumeResult consume_result = pipeline->Consume(consumer);
// if the raster status is to resubmit the frame, we push the frame to the
// front of the queue and also change the consume status to more available.
if (raster_status == RasterStatus::kResubmit) {
auto front_continuation = pipeline->ProduceToFront();
front_continuation.Complete(std::move(resubmitted_layer_tree_));
consume_result = PipelineConsumeResult::MoreAvailable;
} else if (raster_status == RasterStatus::kEnqueuePipeline) {
consume_result = PipelineConsumeResult::MoreAvailable;
}
// Consume as many pipeline items as possible. But yield the event loop
// between successive tries.
switch (consume_result) {
case PipelineConsumeResult::MoreAvailable: {
task_runners_.GetGPUTaskRunner()->PostTask(
[weak_this = weak_factory_.GetWeakPtr(), pipeline]() {
if (weak_this) {
weak_this->Draw(pipeline);
}
});
break;
}
default:
break;
}
}
RasterStatus Rasterizer::DoDraw(
std::unique_ptr<flutter::LayerTree> layer_tree) {
FML_DCHECK(task_runners_.GetGPUTaskRunner()->RunsTasksOnCurrentThread());
if (!layer_tree || !surface_) {
return RasterStatus::kFailed;
}
FrameTiming timing;
timing.Set(FrameTiming::kBuildStart, layer_tree->build_start());
timing.Set(FrameTiming::kBuildFinish, layer_tree->build_finish());
timing.Set(FrameTiming::kRasterStart, fml::TimePoint::Now());
PersistentCache* persistent_cache = PersistentCache::GetCacheForProcess();
persistent_cache->ResetStoredNewShaders();
RasterStatus raster_status = DrawToSurface(*layer_tree);
if (raster_status == RasterStatus::kSuccess) {
last_layer_tree_ = std::move(layer_tree);
} else if (raster_status == RasterStatus::kResubmit) {
resubmitted_layer_tree_ = std::move(layer_tree);
return raster_status;
}
if (persistent_cache->IsDumpingSkp() &&
persistent_cache->StoredNewShaders()) {
auto screenshot =
ScreenshotLastLayerTree(ScreenshotType::SkiaPicture, false);
persistent_cache->DumpSkp(*screenshot.data);
}
// TODO(liyuqian): in Fuchsia, the rasterization doesn't finish when
// Rasterizer::DoDraw finishes. Future work is needed to adapt the timestamp
// for Fuchsia to capture SceneUpdateContext::ExecutePaintTasks.
timing.Set(FrameTiming::kRasterFinish, fml::TimePoint::Now());
delegate_.OnFrameRasterized(timing);
// Pipeline pressure is applied from a couple of places:
// rasterizer: When there are more items as of the time of Consume.
// animator (via shell): Frame gets produces every vsync.
// Enqueing here is to account for the following scenario:
// T = 1
// - one item (A) in the pipeline
// - rasterizer starts (and merges the threads)
// - pipeline consume result says no items to process
// T = 2
// - animator produces (B) to the pipeline
// - applies pipeline pressure via platform thread.
// T = 3
// - rasterizes finished (and un-merges the threads)
// - |Draw| for B yields as its on the wrong thread.
// This enqueue ensures that we attempt to consume from the right
// thread one more time after un-merge.
if (gpu_thread_merger_) {
if (gpu_thread_merger_->DecrementLease() ==
fml::GpuThreadStatus::kUnmergedNow) {
return RasterStatus::kEnqueuePipeline;
}
}
return raster_status;
}
RasterStatus Rasterizer::DrawToSurface(flutter::LayerTree& layer_tree) {
FML_DCHECK(surface_);
auto frame = surface_->AcquireFrame(layer_tree.frame_size());
if (frame == nullptr) {
return RasterStatus::kFailed;
}
// There is no way for the compositor to know how long the layer tree
// construction took. Fortunately, the layer tree does. Grab that time
// for instrumentation.
compositor_context_->ui_time().SetLapTime(layer_tree.build_time());
auto* external_view_embedder = surface_->GetExternalViewEmbedder();
sk_sp<SkSurface> embedder_root_surface;
if (external_view_embedder != nullptr) {
external_view_embedder->BeginFrame(layer_tree.frame_size(),
surface_->GetContext());
embedder_root_surface = external_view_embedder->GetRootSurface();
}
// If the external view embedder has specified an optional root surface, the
// root surface transformation is set by the embedder instead of
// having to apply it here.
SkMatrix root_surface_transformation =
embedder_root_surface ? SkMatrix{} : surface_->GetRootTransformation();
auto root_surface_canvas = embedder_root_surface
? embedder_root_surface->getCanvas()
: frame->SkiaCanvas();
auto compositor_frame = compositor_context_->AcquireFrame(
surface_->GetContext(), // skia GrContext
root_surface_canvas, // root surface canvas
external_view_embedder, // external view embedder
root_surface_transformation, // root surface transformation
true, // instrumentation enabled
gpu_thread_merger_ // thread merger
);
if (compositor_frame) {
RasterStatus raster_status = compositor_frame->Raster(layer_tree, false);
if (raster_status == RasterStatus::kFailed) {
return raster_status;
}
frame->Submit();
if (external_view_embedder != nullptr) {
external_view_embedder->SubmitFrame(surface_->GetContext());
}
FireNextFrameCallbackIfPresent();
if (surface_->GetContext()) {
surface_->GetContext()->performDeferredCleanup(kSkiaCleanupExpiration);
}
return raster_status;
}
return RasterStatus::kFailed;
}
static sk_sp<SkData> SerializeTypeface(SkTypeface* typeface, void* ctx) {
return typeface->serialize(SkTypeface::SerializeBehavior::kDoIncludeData);
}
static sk_sp<SkData> ScreenshotLayerTreeAsPicture(
flutter::LayerTree* tree,
flutter::CompositorContext& compositor_context) {
FML_DCHECK(tree != nullptr);
SkPictureRecorder recorder;
recorder.beginRecording(
SkRect::MakeWH(tree->frame_size().width(), tree->frame_size().height()));
SkMatrix root_surface_transformation;
root_surface_transformation.reset();
// TODO(amirh): figure out how to take a screenshot with embedded UIView.
// https://github.com/flutter/flutter/issues/23435
auto frame = compositor_context.AcquireFrame(
nullptr, recorder.getRecordingCanvas(), nullptr,
root_surface_transformation, false, nullptr);
frame->Raster(*tree, true);
SkSerialProcs procs = {0};
procs.fTypefaceProc = SerializeTypeface;
return recorder.finishRecordingAsPicture()->serialize(&procs);
}
static sk_sp<SkSurface> CreateSnapshotSurface(GrContext* surface_context,
const SkISize& size) {
const auto image_info = SkImageInfo::MakeN32Premul(
size.width(), size.height(), SkColorSpace::MakeSRGB());
if (surface_context) {
// There is a rendering surface that may contain textures that are going to
// be referenced in the layer tree about to be drawn.
return SkSurface::MakeRenderTarget(surface_context, //
SkBudgeted::kNo, //
image_info //
);
}
// There is no rendering surface, assume no GPU textures are present and
// create a raster surface.
return SkSurface::MakeRaster(image_info);
}
static sk_sp<SkData> ScreenshotLayerTreeAsImage(
flutter::LayerTree* tree,
flutter::CompositorContext& compositor_context,
GrContext* surface_context,
bool compressed) {
// Attempt to create a snapshot surface depending on whether we have access to
// a valid GPU rendering context.
auto snapshot_surface =
CreateSnapshotSurface(surface_context, tree->frame_size());
if (snapshot_surface == nullptr) {
FML_LOG(ERROR) << "Screenshot: unable to create snapshot surface";
return nullptr;
}
// Draw the current layer tree into the snapshot surface.
auto* canvas = snapshot_surface->getCanvas();
// There is no root surface transformation for the screenshot layer. Reset the
// matrix to identity.
SkMatrix root_surface_transformation;
root_surface_transformation.reset();
auto frame = compositor_context.AcquireFrame(surface_context, canvas, nullptr,
root_surface_transformation,
false, nullptr);
canvas->clear(SK_ColorTRANSPARENT);
frame->Raster(*tree, true);
canvas->flush();
// Prepare an image from the surface, this image may potentially be on th GPU.
auto potentially_gpu_snapshot = snapshot_surface->makeImageSnapshot();
if (!potentially_gpu_snapshot) {
FML_LOG(ERROR) << "Screenshot: unable to make image screenshot";
return nullptr;
}
// Copy the GPU image snapshot into CPU memory.
auto cpu_snapshot = potentially_gpu_snapshot->makeRasterImage();
if (!cpu_snapshot) {
FML_LOG(ERROR) << "Screenshot: unable to make raster image";
return nullptr;
}
// If the caller want the pixels to be compressed, there is a Skia utility to
// compress to PNG. Use that.
if (compressed) {
return cpu_snapshot->encodeToData();
}
// Copy it into a bitmap and return the same.
SkPixmap pixmap;
if (!cpu_snapshot->peekPixels(&pixmap)) {
FML_LOG(ERROR) << "Screenshot: unable to obtain bitmap pixels";
return nullptr;
}
return SkData::MakeWithCopy(pixmap.addr32(), pixmap.computeByteSize());
}
Rasterizer::Screenshot Rasterizer::ScreenshotLastLayerTree(
Rasterizer::ScreenshotType type,
bool base64_encode) {
auto* layer_tree = GetLastLayerTree();
if (layer_tree == nullptr) {
FML_LOG(ERROR) << "Last layer tree was null when screenshotting.";
return {};
}
sk_sp<SkData> data = nullptr;
GrContext* surface_context = surface_ ? surface_->GetContext() : nullptr;
switch (type) {
case ScreenshotType::SkiaPicture:
data = ScreenshotLayerTreeAsPicture(layer_tree, *compositor_context_);
break;
case ScreenshotType::UncompressedImage:
data = ScreenshotLayerTreeAsImage(layer_tree, *compositor_context_,
surface_context, false);
break;
case ScreenshotType::CompressedImage:
data = ScreenshotLayerTreeAsImage(layer_tree, *compositor_context_,
surface_context, true);
break;
}
if (data == nullptr) {
FML_LOG(ERROR) << "Screenshot data was null.";
return {};
}
if (base64_encode) {
size_t b64_size = SkBase64::Encode(data->data(), data->size(), nullptr);
auto b64_data = SkData::MakeUninitialized(b64_size);
SkBase64::Encode(data->data(), data->size(), b64_data->writable_data());
return Rasterizer::Screenshot{b64_data, layer_tree->frame_size()};
}
return Rasterizer::Screenshot{data, layer_tree->frame_size()};
}
void Rasterizer::SetNextFrameCallback(fml::closure callback) {
next_frame_callback_ = callback;
}
void Rasterizer::FireNextFrameCallbackIfPresent() {
if (!next_frame_callback_) {
return;
}
// It is safe for the callback to set a new callback.
auto callback = next_frame_callback_;
next_frame_callback_ = nullptr;
callback();
}
void Rasterizer::SetResourceCacheMaxBytes(size_t max_bytes, bool from_user) {
user_override_resource_cache_bytes_ |= from_user;
if (!from_user && user_override_resource_cache_bytes_) {
// We should not update the setting here if a user has explicitly set a
// value for this over the flutter/skia channel.
return;
}
max_cache_bytes_ = max_bytes;
if (!surface_) {
return;
}
GrContext* context = surface_->GetContext();
if (context) {
int max_resources;
context->getResourceCacheLimits(&max_resources, nullptr);
context->setResourceCacheLimits(max_resources, max_bytes);
}
}
std::optional<size_t> Rasterizer::GetResourceCacheMaxBytes() const {
if (!surface_) {
return std::nullopt;
}
GrContext* context = surface_->GetContext();
if (context) {
size_t max_bytes;
context->getResourceCacheLimits(nullptr, &max_bytes);
return max_bytes;
}
return std::nullopt;
}
Rasterizer::Screenshot::Screenshot() {}
Rasterizer::Screenshot::Screenshot(sk_sp<SkData> p_data, SkISize p_size)
: data(std::move(p_data)), frame_size(p_size) {}
Rasterizer::Screenshot::Screenshot(const Screenshot& other) = default;
Rasterizer::Screenshot::~Screenshot() = default;
} // namespace flutter