mirrors/ladybird
0
0
Fork 0
mirror of https://github.com/LadybirdBrowser/ladybird.git synced 2025-08-02 14:19:48 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 9

LibWeb: Move painting surface allocation into rendering thread

Browse source 
Skia has a check in debug mode to verify that surface is only used
within one thread. Before this change we were violating this by
allocating surfaces on the main thread while using and destructing them
on the rendering thread.
This commit is contained in:
Aliaksandr Kalenik 2025-04-03 18:03:17 +02:00 • committed by Alexander Kalenik
commit 12a2aebeb6
Notes: github-actions[bot] 2025-04-03 20:02:46 +00:00
6 changed files with 73 additions and 52 deletions
Download patch file Download diff file Expand all files Collapse all files

53
Libraries/LibWeb/HTML/RenderingThread.cpp
View file

@ -6,6 +6,8 @@
#include <LibCore/EventLoop.h>
#include <LibWeb/HTML/RenderingThread.h>
#include <LibWeb/HTML/TraversableNavigable.h>
#include <LibWeb/Painting/BackingStore.h>
namespace Web::HTML {
@ -21,8 +23,9 @@ RenderingThread::~RenderingThread()
(void)m_thread->join();
}
void RenderingThread::start()
void RenderingThread::start(DisplayListPlayerType display_list_player_type)
{
m_display_list_player_type = display_list_player_type;
VERIFY(m_skia_player);
m_thread = Threading::Thread::construct([this] {
rendering_thread_loop();
@ -36,6 +39,10 @@ void RenderingThread::rendering_thread_loop()
while (true) {
auto task = [this]() -> Optional<Task> {
Threading::MutexLocker const locker { m_rendering_task_mutex };
if (m_needs_to_clear_bitmap_to_surface_cache) {
m_bitmap_to_surface.clear();
m_needs_to_clear_bitmap_to_surface_cache = false;
}
while (m_rendering_tasks.is_empty() && !m_exit) {
m_rendering_task_ready_wake_condition.wait();
}
@ -49,18 +56,56 @@ void RenderingThread::rendering_thread_loop()
break;
}
m_skia_player->execute(*task->display_list, task->painting_surface);
auto painting_surface = painting_surface_for_backing_store(task->backing_store);
m_skia_player->execute(*task->display_list, painting_surface);
m_main_thread_event_loop.deferred_invoke([callback = move(task->callback)] {
callback();
});
}
}
void RenderingThread::enqueue_rendering_task(NonnullRefPtr<Painting::DisplayList> display_list, NonnullRefPtr<Gfx::PaintingSurface> painting_surface, Function<void()>&& callback)
void RenderingThread::enqueue_rendering_task(NonnullRefPtr<Painting::DisplayList> display_list, NonnullRefPtr<Painting::BackingStore> backing_store, Function<void()>&& callback)
{
Threading::MutexLocker const locker { m_rendering_task_mutex };
m_rendering_tasks.enqueue(Task { move(display_list), move(painting_surface), move(callback) });
m_rendering_tasks.enqueue(Task { move(display_list), move(backing_store), move(callback) });
m_rendering_task_ready_wake_condition.signal();
}
NonnullRefPtr<Gfx::PaintingSurface> RenderingThread::painting_surface_for_backing_store(Painting::BackingStore& backing_store)
{
auto& bitmap = backing_store.bitmap();
auto cached_surface = m_bitmap_to_surface.find(&bitmap);
if (cached_surface != m_bitmap_to_surface.end())
return cached_surface->value;
RefPtr<Gfx::PaintingSurface> new_surface;
if (m_display_list_player_type == DisplayListPlayerType::SkiaGPUIfAvailable && m_skia_backend_context) {
#ifdef USE_VULKAN
// Vulkan: Try to create an accelerated surface.
new_surface = Gfx::PaintingSurface::create_with_size(m_skia_backend_context, backing_store.size(), Gfx::BitmapFormat::BGRA8888, Gfx::AlphaType::Premultiplied);
new_surface->on_flush = [backing_store = static_cast<NonnullRefPtr<Painting::BackingStore>>(backing_store)](auto& surface) { surface.read_into_bitmap(backing_store->bitmap()); };
#endif
#ifdef AK_OS_MACOS
// macOS: Wrap an IOSurface if available.
if (is<Painting::IOSurfaceBackingStore>(backing_store)) {
auto& iosurface_backing_store = static_cast<Painting::IOSurfaceBackingStore&>(backing_store);
new_surface = Gfx::PaintingSurface::wrap_iosurface(iosurface_backing_store.iosurface_handle(), *m_skia_backend_context);
}
#endif
}
// CPU and fallback: wrap the backing store bitmap directly.
if (!new_surface)
new_surface = Gfx::PaintingSurface::wrap_bitmap(bitmap);
m_bitmap_to_surface.set(&bitmap, *new_surface);
return *new_surface;
}
void RenderingThread::clear_bitmap_to_surface_cache()
{
Threading::MutexLocker const locker { m_rendering_task_mutex };
m_needs_to_clear_bitmap_to_surface_cache = true;
}
}