Use inner HWND for accelerated rendering on windows

chrome/gpu/gpu_command_buffer_stub.cc

 // Copyright (c) 2010 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #if defined(ENABLE_GPU)
 
 #include "base/process_util.h"
 #include "base/shared_memory.h"
 #include "build/build_config.h"
 #include "chrome/common/child_thread.h"
 #include "chrome/common/gpu_messages.h"
 #include "chrome/gpu/gpu_channel.h"
 #include "chrome/gpu/gpu_command_buffer_stub.h"
 
 using gpu::Buffer;
 
+#if defined(OS_WIN)
+#define kCompositorWindowOwner L"CompositorWindowOwner"
+#endif
+
 GpuCommandBufferStub::GpuCommandBufferStub(
     GpuChannel* channel,
     gfx::PluginWindowHandle handle,
     GpuCommandBufferStub* parent,
     const gfx::Size& size,
     const std::string& allowed_extensions,
     const std::vector<int32>& attribs,
     uint32 parent_texture_id,
     int32 route_id,
     int32 renderer_id,
     int32 render_view_id)
     : channel_(channel),
       handle_(handle),
       parent_(
           parent ? parent->AsWeakPtr() : base::WeakPtr<GpuCommandBufferStub>()),
       initial_size_(size),
       allowed_extensions_(allowed_extensions),
       requested_attribs_(attribs),
       parent_texture_id_(parent_texture_id),
       route_id_(route_id),
       renderer_id_(renderer_id),
       render_view_id_(render_view_id) {
 }
 
+#if defined(OS_WIN)
+static LRESULT CALLBACK CompositorWindowProc(
+    HWND hwnd,
+    UINT message,
+    WPARAM wparam,
+    LPARAM lparam) {
+  switch (message) {
+    case WM_ERASEBKGND:
+      return 0;
+    case WM_DESTROY:
+      RemoveProp(hwnd, kCompositorWindowOwner);
+      return 0;
+    case WM_PAINT: {
+      PAINTSTRUCT paint;
+      HDC dc = BeginPaint(hwnd, &paint);
+      if (dc) {
+        HANDLE h = GetProp(hwnd, kCompositorWindowOwner);
+        if (h) {
+          GpuCommandBufferStub* stub =
+              reinterpret_cast<GpuCommandBufferStub*>(h);
+          stub->OnCompositorWindowPainted();
+        }
+        EndPaint(hwnd, &paint);
+      }
+      break;
+    }
+    default:
+      return DefWindowProc(hwnd, message, wparam, lparam);
+  }
+  return 0;
+}
+
+bool GpuCommandBufferStub::CreateCompositorWindow() {
+  DCHECK(handle_ != gfx::kNullPluginWindow);
+
+  // Ask the browser to create the the host window.
+  ChildThread* gpu_thread = ChildThread::current();
+  gfx::PluginWindowHandle host_window_id = gfx::kNullPluginWindow;
+    gpu_thread->Send(new GpuHostMsg_CreateCompositorHostWindow(
+        renderer_id_,
+        render_view_id_,
+        &host_window_id));
+  if (host_window_id == gfx::kNullPluginWindow)
+    return false;
+  HWND host_window = static_cast<HWND>(host_window_id);
+
+  // Create the compositor window itself.
+  DCHECK(host_window);
+  static ATOM window_class = 0;
+  if (!window_class) {
+    WNDCLASSEX wcex;
+    wcex.cbSize         = sizeof(wcex);
+    wcex.style          = 0;
+    wcex.lpfnWndProc    = CompositorWindowProc;
+    wcex.cbClsExtra     = 0;
+    wcex.cbWndExtra     = 0;
+    wcex.hInstance      = GetModuleHandle(NULL);
+    wcex.hIcon          = 0;
+    wcex.hCursor        = 0;
+    wcex.hbrBackground  = NULL;
+    wcex.lpszMenuName   = 0;
+    wcex.lpszClassName  = L"CompositorWindowClass";
+    wcex.hIconSm        = 0;
+    window_class = RegisterClassEx(&wcex);
+    DCHECK(window_class);
+  }
+
+  HWND compositor_window = CreateWindowEx(
+      WS_EX_LEFT | WS_EX_LTRREADING | WS_EX_RIGHTSCROLLBAR,
+      MAKEINTATOM(window_class),
+      0,
+      WS_CHILD | WS_CLIPCHILDREN | WS_CLIPSIBLINGS | WS_DISABLED,
+      0, 0,
+      0, 0,
+      host_window,
+      0,
+      GetModuleHandle(NULL),
+      0);
+  if (!compositor_window) {
+    compositor_window_ = gfx::kNullPluginWindow;
+    return false;
+  }
+  SetProp(compositor_window, kCompositorWindowOwner,
+      reinterpret_cast<HANDLE>(this));
+
+  RECT parent_rect;
+  GetClientRect(host_window, &parent_rect);
+
+  UINT flags = SWP_NOSENDCHANGING | SWP_NOCOPYBITS | SWP_NOZORDER |
+      SWP_NOACTIVATE | SWP_DEFERERASE | SWP_SHOWWINDOW;
+  SetWindowPos(compositor_window,
+      NULL,
+      0, 0,
+      parent_rect.right - parent_rect.left,
+      parent_rect.bottom - parent_rect.top,
+      flags);
+  compositor_window_ = static_cast<gfx::PluginWindowHandle>(compositor_window);
+  return true;
+}
+
+void GpuCommandBufferStub::OnCompositorWindowPainted() {
+  ChildThread* gpu_thread = ChildThread::current();
+  gpu_thread->Send(new GpuHostMsg_ScheduleComposite(
+      renderer_id_, render_view_id_));
+}
+#endif  // defined(OS_WIN)
+
+
 GpuCommandBufferStub::~GpuCommandBufferStub() {
   if (processor_.get()) {
     processor_->Destroy();
   }
+#if defined(OS_WIN)
+  if (compositor_window_) {
+    DestroyWindow(static_cast<HWND>(compositor_window_));
+    compositor_window_ = NULL;
+  }
+#endif
 }
 
 void GpuCommandBufferStub::OnMessageReceived(const IPC::Message& message) {
   IPC_BEGIN_MESSAGE_MAP(GpuCommandBufferStub, message)
     IPC_MESSAGE_HANDLER(GpuCommandBufferMsg_Initialize, OnInitialize);
     IPC_MESSAGE_HANDLER(GpuCommandBufferMsg_GetState, OnGetState);
     IPC_MESSAGE_HANDLER(GpuCommandBufferMsg_AsyncGetState, OnAsyncGetState);
     IPC_MESSAGE_HANDLER(GpuCommandBufferMsg_Flush, OnFlush);
     IPC_MESSAGE_HANDLER(GpuCommandBufferMsg_AsyncFlush, OnAsyncFlush);
     IPC_MESSAGE_HANDLER(GpuCommandBufferMsg_CreateTransferBuffer,
@@ skipping 17 matching lines @@
 
 void GpuCommandBufferStub::OnInitialize(
     int32 size,
     base::SharedMemoryHandle* ring_buffer) {
   DCHECK(!command_buffer_.get());
 
   *ring_buffer = base::SharedMemory::NULLHandle();
 
   command_buffer_.reset(new gpu::CommandBufferService);
 
+  // Create the child window, if needed
+#if defined(OS_WIN)
+  gfx::PluginWindowHandle output_window_handle;
+  if (handle_) {
+    if (!CreateCompositorWindow()) {
+      return;
+    }
+    output_window_handle = compositor_window_;
+  } else {
+    output_window_handle = handle_;
+  }
+#else
+  gfx::PluginWindowHandle output_window_handle = handle_;
+#endif
+
   // Initialize the CommandBufferService and GPUProcessor.
   if (command_buffer_->Initialize(size)) {
     Buffer buffer = command_buffer_->GetRingBuffer();
     if (buffer.shared_memory) {
       gpu::GPUProcessor* parent_processor =
           parent_ ? parent_->processor_.get() : NULL;
       processor_.reset(new gpu::GPUProcessor(command_buffer_.get(), NULL));
       if (processor_->Initialize(
-          handle_,
+          output_window_handle,
           initial_size_,
           allowed_extensions_.c_str(),
           requested_attribs_,
           parent_processor,
           parent_texture_id_)) {
         command_buffer_->SetPutOffsetChangeCallback(
             NewCallback(processor_.get(),
                         &gpu::GPUProcessor::ProcessCommands));
         processor_->SetSwapBuffersCallback(
             NewCallback(this, &GpuCommandBufferStub::OnSwapBuffers));
 
         // Assume service is responsible for duplicating the handle from the
         // calling process.
         buffer.shared_memory->ShareToProcess(channel_->renderer_handle(),
                                              ring_buffer);
 #if defined(OS_MACOSX)
         if (handle_) {
           // This context conceptually puts its output directly on the
           // screen, rendered by the accelerated plugin layer in
           // RenderWidgetHostViewMac. Set up a pathway to notify the
           // browser process when its contents change.
           processor_->SetSwapBuffersCallback(
               NewCallback(this,
                           &GpuCommandBufferStub::SwapBuffersCallback));
         }
-#elif defined(OS_LINUX)
+#elif defined(OS_LINUX) || defined(OS_WIN)
         if (handle_) {
-          // Set up a pathway to allow the Gpu process to ask the browser
-          // for a window resize.
+          // Set up a pathway for resizing the output window at the right time
+          // relative to other GL commands.
           processor_->SetResizeCallback(
               NewCallback(this,
                           &GpuCommandBufferStub::ResizeCallback));
         }
 #endif
       } else {
         processor_.reset();
         command_buffer_.reset();
       }
     }
@@ skipping 90 matching lines @@
 }
 
 void GpuCommandBufferStub::AcceleratedSurfaceBuffersSwapped(
     uint64 swap_buffers_count) {
   processor_->set_acknowledged_swap_buffers_count(swap_buffers_count);
   // Wake up the GpuProcessor to start doing work again.
   processor_->ScheduleProcessCommands();
 }
 #endif  // defined(OS_MACOSX)
 
+void GpuCommandBufferStub::ResizeCallback(gfx::Size size) {
+  if (handle_ == gfx::kNullPluginWindow)
+    return;
+
 #if defined(OS_LINUX)
-void GpuCommandBufferStub::ResizeCallback(gfx::Size size) {
   ChildThread* gpu_thread = ChildThread::current();
   bool result = false;
   gpu_thread->Send(
       new GpuHostMsg_ResizeXID(handle_, size, &result));
+#elif defined(OS_WIN)
+  HWND hwnd = static_cast<HWND>(compositor_window_);
+  UINT swp_flags = SWP_NOSENDCHANGING | SWP_NOOWNERZORDER | SWP_NOCOPYBITS |
+    SWP_NOMOVE | SWP_NOZORDER | SWP_NOACTIVATE | SWP_DEFERERASE;
+  SetWindowPos(hwnd, NULL, 0, 0, size.width(), size.height(), swp_flags);
+#endif
 }
-#endif  // defined(OS_LINUX)
 
 #endif  // ENABLE_GPU