blob: 287c350d222eacfe5211bda71b87f54524378618 [file] [log] [blame]
/*
* Copyright (c) 2008, Google Inc. All rights reserved.
* Copyright (C) 2009 Dirk Schulze <krit@webkit.org>
* Copyright (C) 2010 Torch Mobile (Beijing) Co. Ltd. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are
* met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above
* copyright notice, this list of conditions and the following disclaimer
* in the documentation and/or other materials provided with the
* distribution.
* * Neither the name of Google Inc. nor the names of its
* contributors may be used to endorse or promote products derived from
* this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "platform/graphics/ImageBuffer.h"
#include <memory>
#include "gpu/command_buffer/client/gles2_interface.h"
#include "gpu/command_buffer/common/mailbox.h"
#include "gpu/command_buffer/common/sync_token.h"
#include "platform/RuntimeEnabledFeatures.h"
#include "platform/geometry/IntRect.h"
#include "platform/graphics/ExpensiveCanvasHeuristicParameters.h"
#include "platform/graphics/GraphicsContext.h"
#include "platform/graphics/ImageBufferClient.h"
#include "platform/graphics/RecordingImageBufferSurface.h"
#include "platform/graphics/StaticBitmapImage.h"
#include "platform/graphics/UnacceleratedImageBufferSurface.h"
#include "platform/graphics/gpu/DrawingBuffer.h"
#include "platform/graphics/gpu/Extensions3DUtil.h"
#include "platform/graphics/paint/PaintImage.h"
#include "platform/graphics/paint/PaintRecord.h"
#include "platform/graphics/skia/SkiaUtils.h"
#include "platform/image-encoders/ImageEncoder.h"
#include "platform/image-encoders/PNGImageEncoder.h"
#include "platform/image-encoders/WEBPImageEncoder.h"
#include "platform/network/mime/MIMETypeRegistry.h"
#include "platform/wtf/CheckedNumeric.h"
#include "platform/wtf/MathExtras.h"
#include "platform/wtf/PtrUtil.h"
#include "platform/wtf/Vector.h"
#include "platform/wtf/text/Base64.h"
#include "platform/wtf/text/WTFString.h"
#include "platform/wtf/typed_arrays/ArrayBufferContents.h"
#include "public/platform/Platform.h"
#include "public/platform/WebGraphicsContext3DProvider.h"
#include "skia/ext/texture_handle.h"
#include "third_party/skia/include/core/SkSwizzle.h"
#include "third_party/skia/include/encode/SkJpegEncoder.h"
#include "third_party/skia/include/gpu/GrContext.h"
#include "third_party/skia/include/gpu/gl/GrGLTypes.h"
namespace blink {
std::unique_ptr<ImageBuffer> ImageBuffer::Create(
std::unique_ptr<ImageBufferSurface> surface) {
if (!surface->IsValid())
return nullptr;
return WTF::WrapUnique(new ImageBuffer(std::move(surface)));
}
std::unique_ptr<ImageBuffer> ImageBuffer::Create(
const IntSize& size,
OpacityMode opacity_mode,
ImageInitializationMode initialization_mode,
const CanvasColorParams& color_params) {
std::unique_ptr<ImageBufferSurface> surface(
WTF::WrapUnique(new UnacceleratedImageBufferSurface(
size, opacity_mode, initialization_mode, color_params)));
if (!surface->IsValid())
return nullptr;
return WTF::WrapUnique(new ImageBuffer(std::move(surface)));
}
ImageBuffer::ImageBuffer(std::unique_ptr<ImageBufferSurface> surface)
: weak_ptr_factory_(this),
snapshot_state_(kInitialSnapshotState),
surface_(std::move(surface)),
client_(0),
gpu_readback_invoked_in_current_frame_(false),
gpu_readback_successive_frames_(0),
gpu_memory_usage_(0) {
surface_->SetImageBuffer(this);
UpdateGPUMemoryUsage();
}
intptr_t ImageBuffer::global_gpu_memory_usage_ = 0;
unsigned ImageBuffer::global_accelerated_image_buffer_count_ = 0;
ImageBuffer::~ImageBuffer() {
if (gpu_memory_usage_) {
DCHECK_GT(global_accelerated_image_buffer_count_, 0u);
global_accelerated_image_buffer_count_--;
}
ImageBuffer::global_gpu_memory_usage_ -= gpu_memory_usage_;
surface_->SetImageBuffer(nullptr);
}
bool ImageBuffer::CanCreateImageBuffer(const IntSize& size) {
if (size.IsEmpty())
return false;
CheckedNumeric<int> area = size.Width();
area *= size.Height();
if (!area.IsValid() || area.ValueOrDie() > kMaxCanvasArea)
return false;
if (size.Width() > kMaxSkiaDim || size.Height() > kMaxSkiaDim)
return false;
return true;
}
PaintCanvas* ImageBuffer::Canvas() const {
return surface_->Canvas();
}
void ImageBuffer::DisableDeferral(DisableDeferralReason reason) const {
return surface_->DisableDeferral(reason);
}
bool ImageBuffer::WritePixels(const SkImageInfo& info,
const void* pixels,
size_t row_bytes,
int x,
int y) {
return surface_->WritePixels(info, pixels, row_bytes, x, y);
}
bool ImageBuffer::IsSurfaceValid() const {
return surface_->IsValid();
}
void ImageBuffer::FinalizeFrame() {
if (IsAccelerated() &&
ExpensiveCanvasHeuristicParameters::kGPUReadbackForcesNoAcceleration &&
!RuntimeEnabledFeatures::canvas2dFixedRenderingModeEnabled()) {
if (gpu_readback_invoked_in_current_frame_) {
gpu_readback_successive_frames_++;
gpu_readback_invoked_in_current_frame_ = false;
} else {
gpu_readback_successive_frames_ = 0;
}
if (gpu_readback_successive_frames_ >=
ExpensiveCanvasHeuristicParameters::kGPUReadbackMinSuccessiveFrames) {
DisableAcceleration();
}
}
surface_->FinalizeFrame();
}
void ImageBuffer::DoPaintInvalidation(const FloatRect& dirty_rect) {
surface_->DoPaintInvalidation(dirty_rect);
}
bool ImageBuffer::RestoreSurface() const {
return surface_->IsValid() || surface_->Restore();
}
void ImageBuffer::NotifySurfaceInvalid() {
if (client_)
client_->NotifySurfaceInvalid();
}
void ImageBuffer::ResetCanvas(PaintCanvas* canvas) const {
if (client_)
client_->RestoreCanvasMatrixClipStack(canvas);
}
sk_sp<SkImage> ImageBuffer::NewSkImageSnapshot(AccelerationHint hint,
SnapshotReason reason) const {
if (snapshot_state_ == kInitialSnapshotState)
snapshot_state_ = kDidAcquireSnapshot;
if (!IsSurfaceValid())
return nullptr;
return surface_->NewImageSnapshot(hint, reason);
}
PassRefPtr<Image> ImageBuffer::NewImageSnapshot(AccelerationHint hint,
SnapshotReason reason) const {
sk_sp<SkImage> snapshot = NewSkImageSnapshot(hint, reason);
if (!snapshot)
return nullptr;
return StaticBitmapImage::Create(std::move(snapshot));
}
void ImageBuffer::DidDraw(const FloatRect& rect) const {
if (snapshot_state_ == kDidAcquireSnapshot)
snapshot_state_ = kDrawnToAfterSnapshot;
surface_->DidDraw(rect);
}
WebLayer* ImageBuffer::PlatformLayer() const {
return surface_->Layer();
}
bool ImageBuffer::CopyToPlatformTexture(SnapshotReason reason,
gpu::gles2::GLES2Interface* gl,
GLenum target,
GLuint texture,
bool premultiply_alpha,
bool flip_y,
const IntPoint& dest_point,
const IntRect& source_sub_rectangle) {
if (!Extensions3DUtil::CanUseCopyTextureCHROMIUM(target))
return false;
if (!IsSurfaceValid())
return false;
sk_sp<const SkImage> texture_image =
surface_->NewImageSnapshot(kPreferAcceleration, reason);
if (!texture_image)
return false;
if (!surface_->IsAccelerated())
return false;
DCHECK(texture_image->isTextureBacked()); // The isAccelerated() check above
// should guarantee this.
// Get the texture ID, flushing pending operations if needed.
const GrGLTextureInfo* texture_info = skia::GrBackendObjectToGrGLTextureInfo(
texture_image->getTextureHandle(true));
if (!texture_info || !texture_info->fID)
return false;
std::unique_ptr<WebGraphicsContext3DProvider> provider =
Platform::Current()->CreateSharedOffscreenGraphicsContext3DProvider();
if (!provider || !provider->GetGrContext())
return false;
gpu::gles2::GLES2Interface* shared_gl = provider->ContextGL();
gpu::Mailbox mailbox;
// Contexts may be in a different share group. We must transfer the texture
// through a mailbox first.
shared_gl->GenMailboxCHROMIUM(mailbox.name);
shared_gl->ProduceTextureDirectCHROMIUM(texture_info->fID,
texture_info->fTarget, mailbox.name);
const GLuint64 shared_fence_sync = shared_gl->InsertFenceSyncCHROMIUM();
shared_gl->Flush();
gpu::SyncToken produce_sync_token;
shared_gl->GenSyncTokenCHROMIUM(shared_fence_sync,
produce_sync_token.GetData());
gl->WaitSyncTokenCHROMIUM(produce_sync_token.GetConstData());
GLuint source_texture =
gl->CreateAndConsumeTextureCHROMIUM(texture_info->fTarget, mailbox.name);
// The canvas is stored in a premultiplied format, so unpremultiply if
// necessary. The canvas is also stored in an inverted position, so the flip
// semantics are reversed.
// It is expected that callers of this method have already allocated
// the platform texture with the appropriate size.
gl->CopySubTextureCHROMIUM(
source_texture, 0, target, texture, 0, dest_point.X(), dest_point.Y(),
source_sub_rectangle.X(), source_sub_rectangle.Y(),
source_sub_rectangle.Width(), source_sub_rectangle.Height(),
flip_y ? GL_FALSE : GL_TRUE, GL_FALSE,
premultiply_alpha ? GL_FALSE : GL_TRUE);
gl->DeleteTextures(1, &source_texture);
const GLuint64 context_fence_sync = gl->InsertFenceSyncCHROMIUM();
gl->Flush();
gpu::SyncToken copy_sync_token;
gl->GenSyncTokenCHROMIUM(context_fence_sync, copy_sync_token.GetData());
shared_gl->WaitSyncTokenCHROMIUM(copy_sync_token.GetConstData());
// This disassociates the texture from the mailbox to avoid leaking the
// mapping between the two.
shared_gl->ProduceTextureDirectCHROMIUM(0, texture_info->fTarget,
mailbox.name);
// Undo grContext texture binding changes introduced in this function.
GrContext* gr_context = provider->GetGrContext();
CHECK(gr_context); // We already check / early-out above if null.
gr_context->resetContext(kTextureBinding_GrGLBackendState);
return true;
}
bool ImageBuffer::CopyRenderingResultsFromDrawingBuffer(
DrawingBuffer* drawing_buffer,
SourceDrawingBuffer source_buffer) {
if (!drawing_buffer || !surface_->IsAccelerated())
return false;
std::unique_ptr<WebGraphicsContext3DProvider> provider =
Platform::Current()->CreateSharedOffscreenGraphicsContext3DProvider();
if (!provider)
return false;
gpu::gles2::GLES2Interface* gl = provider->ContextGL();
GLuint texture_id = surface_->GetBackingTextureHandleForOverwrite();
if (!texture_id)
return false;
gl->Flush();
return drawing_buffer->CopyToPlatformTexture(
gl, GL_TEXTURE_2D, texture_id, true, false, IntPoint(0, 0),
IntRect(IntPoint(0, 0), drawing_buffer->Size()), source_buffer);
}
void ImageBuffer::Draw(GraphicsContext& context,
const FloatRect& dest_rect,
const FloatRect* src_ptr,
SkBlendMode op) {
if (!IsSurfaceValid())
return;
FloatRect src_rect =
src_ptr ? *src_ptr : FloatRect(FloatPoint(), FloatSize(size()));
surface_->Draw(context, dest_rect, src_rect, op);
}
void ImageBuffer::Flush(FlushReason reason) {
if (surface_->Canvas()) {
surface_->Flush(reason);
}
}
void ImageBuffer::FlushGpu(FlushReason reason) {
if (surface_->Canvas()) {
surface_->FlushGpu(reason);
}
}
bool ImageBuffer::GetImageData(Multiply multiplied,
const IntRect& rect,
WTF::ArrayBufferContents& contents) const {
uint8_t bytes_per_pixel = surface_->color_params().BytesPerPixel();
CheckedNumeric<int> data_size = bytes_per_pixel;
data_size *= rect.Width();
data_size *= rect.Height();
if (!data_size.IsValid())
return false;
if (!IsSurfaceValid()) {
size_t alloc_size_in_bytes = rect.Width() * rect.Height() * bytes_per_pixel;
auto data = WTF::ArrayBufferContents::CreateDataHandle(
alloc_size_in_bytes, WTF::ArrayBufferContents::kZeroInitialize);
if (!data)
return false;
WTF::ArrayBufferContents result(std::move(data), alloc_size_in_bytes,
WTF::ArrayBufferContents::kNotShared);
result.Transfer(contents);
return true;
}
DCHECK(Canvas());
sk_sp<SkImage> snapshot = surface_->NewImageSnapshot(
kPreferNoAcceleration, kSnapshotReasonGetImageData);
if (!snapshot)
return false;
const bool may_have_stray_area =
surface_->IsAccelerated() // GPU readback may fail silently
|| rect.X() < 0 || rect.Y() < 0 ||
rect.MaxX() > surface_->size().Width() ||
rect.MaxY() > surface_->size().Height();
size_t alloc_size_in_bytes = rect.Width() * rect.Height() * bytes_per_pixel;
WTF::ArrayBufferContents::InitializationPolicy initialization_policy =
may_have_stray_area ? WTF::ArrayBufferContents::kZeroInitialize
: WTF::ArrayBufferContents::kDontInitialize;
auto data = WTF::ArrayBufferContents::CreateDataHandle(alloc_size_in_bytes,
initialization_policy);
if (!data)
return false;
WTF::ArrayBufferContents result(std::move(data), alloc_size_in_bytes,
WTF::ArrayBufferContents::kNotShared);
SkAlphaType alpha_type = (multiplied == kPremultiplied)
? kPremul_SkAlphaType
: kUnpremul_SkAlphaType;
SkColorType color_type =
(surface_->color_params().GetSkColorType() == kRGBA_F16_SkColorType)
? kRGBA_F16_SkColorType
: kRGBA_8888_SkColorType;
SkImageInfo info = SkImageInfo::Make(
rect.Width(), rect.Height(), color_type, alpha_type,
surface_->color_params().GetSkColorSpaceForSkSurfaces());
snapshot->readPixels(info, result.Data(), bytes_per_pixel * rect.Width(),
rect.X(), rect.Y());
gpu_readback_invoked_in_current_frame_ = true;
result.Transfer(contents);
return true;
}
void ImageBuffer::PutByteArray(Multiply multiplied,
const unsigned char* source,
const IntSize& source_size,
const IntRect& source_rect,
const IntPoint& dest_point) {
if (!IsSurfaceValid())
return;
uint8_t bytes_per_pixel = surface_->color_params().BytesPerPixel();
DCHECK_GT(source_rect.Width(), 0);
DCHECK_GT(source_rect.Height(), 0);
int origin_x = source_rect.X();
int dest_x = dest_point.X() + source_rect.X();
DCHECK_GE(dest_x, 0);
DCHECK_LT(dest_x, surface_->size().Width());
DCHECK_GE(origin_x, 0);
DCHECK_LT(origin_x, source_rect.MaxX());
int origin_y = source_rect.Y();
int dest_y = dest_point.Y() + source_rect.Y();
DCHECK_GE(dest_y, 0);
DCHECK_LT(dest_y, surface_->size().Height());
DCHECK_GE(origin_y, 0);
DCHECK_LT(origin_y, source_rect.MaxY());
const size_t src_bytes_per_row = bytes_per_pixel * source_size.Width();
const void* src_addr =
source + origin_y * src_bytes_per_row + origin_x * bytes_per_pixel;
SkAlphaType alpha_type;
if (kOpaque == surface_->GetOpacityMode()) {
// If the surface is opaque, tell it that we are writing opaque
// pixels. Writing non-opaque pixels to opaque is undefined in
// Skia. There is some discussion about whether it should be
// defined in skbug.com/6157. For now, we can get the desired
// behavior (memcpy) by pretending the write is opaque.
alpha_type = kOpaque_SkAlphaType;
} else {
alpha_type = (multiplied == kPremultiplied) ? kPremul_SkAlphaType
: kUnpremul_SkAlphaType;
}
SkImageInfo info;
if (surface_->color_params().GetSkColorSpaceForSkSurfaces()) {
info = SkImageInfo::Make(
source_rect.Width(), source_rect.Height(),
surface_->color_params().GetSkColorType(), alpha_type,
surface_->color_params().GetSkColorSpaceForSkSurfaces());
} else {
info = SkImageInfo::Make(source_rect.Width(), source_rect.Height(),
kRGBA_8888_SkColorType, alpha_type);
}
surface_->WritePixels(info, src_addr, src_bytes_per_row, dest_x, dest_y);
}
void ImageBuffer::UpdateGPUMemoryUsage() const {
if (this->IsAccelerated()) {
// If image buffer is accelerated, we should keep track of GPU memory usage.
int gpu_buffer_count = 2;
CheckedNumeric<intptr_t> checked_gpu_usage =
surface_->color_params().BytesPerPixel() * gpu_buffer_count;
checked_gpu_usage *= this->size().Width();
checked_gpu_usage *= this->size().Height();
intptr_t gpu_memory_usage =
checked_gpu_usage.ValueOrDefault(std::numeric_limits<intptr_t>::max());
if (!gpu_memory_usage_) // was not accelerated before
global_accelerated_image_buffer_count_++;
global_gpu_memory_usage_ += (gpu_memory_usage - gpu_memory_usage_);
gpu_memory_usage_ = gpu_memory_usage;
} else if (gpu_memory_usage_) {
// In case of switching from accelerated to non-accelerated mode,
// the GPU memory usage needs to be updated too.
DCHECK_GT(global_accelerated_image_buffer_count_, 0u);
global_accelerated_image_buffer_count_--;
global_gpu_memory_usage_ -= gpu_memory_usage_;
gpu_memory_usage_ = 0;
if (client_)
client_->DidDisableAcceleration();
}
}
void ImageBuffer::DisableAcceleration() {
if (!IsAccelerated())
return;
// Create and configure a recording (unaccelerated) surface.
std::unique_ptr<ImageBufferSurface> surface =
WTF::WrapUnique(new RecordingImageBufferSurface(
surface_->size(), RecordingImageBufferSurface::kAllowFallback,
surface_->GetOpacityMode(), surface_->color_params()));
SetSurface(std::move(surface));
}
void ImageBuffer::SetSurface(std::unique_ptr<ImageBufferSurface> surface) {
sk_sp<SkImage> image =
surface_->NewImageSnapshot(kPreferNoAcceleration, kSnapshotReasonPaint);
// image can be null if alloaction failed in which case we should just
// abort the surface switch to reatain the old surface which is still
// functional.
if (!image)
return;
if (surface->IsRecording()) {
// Using a GPU-backed image with RecordingImageBufferSurface
// will fail at playback time.
image = image->makeNonTextureImage();
}
// TODO(vmpstr): Figure out actual values for this.
auto animation_type = PaintImage::AnimationType::UNKNOWN;
auto completion_state = PaintImage::CompletionState::UNKNOWN;
static PaintImage::Id unknown_stable_id = PaintImage::GetNextId();
surface->Canvas()->drawImage(PaintImage(unknown_stable_id, std::move(image),
animation_type, completion_state),
0, 0);
surface->SetImageBuffer(this);
if (client_)
client_->RestoreCanvasMatrixClipStack(surface->Canvas());
surface_ = std::move(surface);
UpdateGPUMemoryUsage();
}
void ImageBuffer::SetNeedsCompositingUpdate() {
if (client_)
client_->SetNeedsCompositingUpdate();
}
bool ImageDataBuffer::EncodeImage(const String& mime_type,
const double& quality,
Vector<unsigned char>* encoded_image) const {
if (mime_type == "image/jpeg") {
SkImageInfo info =
SkImageInfo::Make(Width(), Height(), kRGBA_8888_SkColorType,
kUnpremul_SkAlphaType, nullptr);
size_t rowBytes =
Width() * SkColorTypeBytesPerPixel(kRGBA_8888_SkColorType);
SkPixmap src(info, Pixels(), rowBytes);
SkJpegEncoder::Options options;
options.fQuality = ImageEncoder::ComputeJpegQuality(quality);
options.fAlphaOption = SkJpegEncoder::AlphaOption::kBlendOnBlack;
options.fBlendBehavior = SkTransferFunctionBehavior::kIgnore;
if (options.fQuality == 100) {
options.fDownsample = SkJpegEncoder::Downsample::k444;
}
return ImageEncoder::Encode(encoded_image, src, options);
}
if (mime_type == "image/webp") {
int compression_quality = WEBPImageEncoder::kDefaultCompressionQuality;
if (quality >= 0.0 && quality <= 1.0)
compression_quality = static_cast<int>(quality * 100 + 0.5);
return WEBPImageEncoder::Encode(*this, compression_quality, encoded_image);
}
DCHECK_EQ(mime_type, "image/png");
return PNGImageEncoder::Encode(*this, encoded_image);
}
String ImageDataBuffer::ToDataURL(const String& mime_type,
const double& quality) const {
DCHECK(MIMETypeRegistry::IsSupportedImageMIMETypeForEncoding(mime_type));
Vector<unsigned char> result;
if (!EncodeImage(mime_type, quality, &result))
return "data:,";
return "data:" + mime_type + ";base64," + Base64Encode(result);
}
} // namespace blink