#include "base/callback.h"
#include "base/compiler_specific.h"
#include "base/containers/hash_tables.h"
+#include "base/containers/scoped_ptr_hash_map.h"
#include "base/memory/ref_counted.h"
#include "base/memory/weak_ptr.h"
#include "base/observer_list.h"
+#include "gpu/command_buffer/client/gpu_control.h"
#include "gpu/command_buffer/common/command_buffer.h"
#include "gpu/command_buffer/common/command_buffer_shared.h"
-#include "gpu/command_buffer/common/gpu_control.h"
#include "gpu/command_buffer/common/gpu_memory_allocation.h"
#include "ipc/ipc_listener.h"
-#include "media/video/video_decode_accelerator.h"
#include "ui/events/latency_info.h"
struct GPUCommandBufferConsoleMessage;
class SharedMemory;
}
-namespace gfx {
-class GpuMemoryBuffer;
-}
-
namespace gpu {
struct Mailbox;
}
+namespace media {
+class VideoDecodeAccelerator;
+class VideoEncodeAccelerator;
+}
+
namespace content {
class GpuChannelHost;
// returns it as an owned pointer to a media::VideoDecodeAccelerator. Returns
// NULL on failure to create the GpuVideoDecodeAcceleratorHost.
// Note that the GpuVideoDecodeAccelerator may still fail to be created in
- // the GPU process, even if this returns non-NULL. In this case the client is
- // notified of an error later.
- scoped_ptr<media::VideoDecodeAccelerator> CreateVideoDecoder(
- media::VideoCodecProfile profile,
- media::VideoDecodeAccelerator::Client* client);
+ // the GPU process, even if this returns non-NULL. In this case the VDA client
+ // is notified of an error later, after Initialize().
+ scoped_ptr<media::VideoDecodeAccelerator> CreateVideoDecoder();
+
+ // Sends an IPC message to create a GpuVideoEncodeAccelerator. Creates and
+ // returns it as an owned pointer to a media::VideoEncodeAccelerator. Returns
+ // NULL on failure to create the GpuVideoEncodeAcceleratorHost.
+ // Note that the GpuVideoEncodeAccelerator may still fail to be created in
+ // the GPU process, even if this returns non-NULL. In this case the VEA client
+ // is notified of an error later, after Initialize();
+ scoped_ptr<media::VideoEncodeAccelerator> CreateVideoEncoder();
// IPC::Listener implementation:
virtual bool OnMessageReceived(const IPC::Message& message) OVERRIDE;
// CommandBuffer implementation:
virtual bool Initialize() OVERRIDE;
- virtual State GetState() OVERRIDE;
virtual State GetLastState() OVERRIDE;
virtual int32 GetLastToken() OVERRIDE;
virtual void Flush(int32 put_offset) OVERRIDE;
- virtual State FlushSync(int32 put_offset, int32 last_known_get) OVERRIDE;
+ virtual void WaitForTokenInRange(int32 start, int32 end) OVERRIDE;
+ virtual void WaitForGetOffsetInRange(int32 start, int32 end) OVERRIDE;
virtual void SetGetBuffer(int32 shm_id) OVERRIDE;
- virtual void SetGetOffset(int32 get_offset) OVERRIDE;
- virtual gpu::Buffer CreateTransferBuffer(size_t size,
- int32* id) OVERRIDE;
+ virtual scoped_refptr<gpu::Buffer> CreateTransferBuffer(size_t size,
+ int32* id) OVERRIDE;
virtual void DestroyTransferBuffer(int32 id) OVERRIDE;
- virtual gpu::Buffer GetTransferBuffer(int32 id) OVERRIDE;
- virtual void SetToken(int32 token) OVERRIDE;
- virtual void SetParseError(gpu::error::Error error) OVERRIDE;
- virtual void SetContextLostReason(
- gpu::error::ContextLostReason reason) OVERRIDE;
// gpu::GpuControl implementation:
- virtual bool SupportsGpuMemoryBuffer() OVERRIDE;
- virtual gfx::GpuMemoryBuffer* CreateGpuMemoryBuffer(
- size_t width,
- size_t height,
- unsigned internalformat,
- int32* id) OVERRIDE;
+ virtual gpu::Capabilities GetCapabilities() OVERRIDE;
+ virtual gfx::GpuMemoryBuffer* CreateGpuMemoryBuffer(size_t width,
+ size_t height,
+ unsigned internalformat,
+ unsigned usage,
+ int32* id) OVERRIDE;
virtual void DestroyGpuMemoryBuffer(int32 id) OVERRIDE;
- virtual bool GenerateMailboxNames(unsigned num,
- std::vector<gpu::Mailbox>* names) OVERRIDE;
virtual uint32 InsertSyncPoint() OVERRIDE;
+ virtual uint32_t InsertFutureSyncPoint() OVERRIDE;
+ virtual void RetireSyncPoint(uint32_t sync_point) OVERRIDE;
virtual void SignalSyncPoint(uint32 sync_point,
const base::Closure& callback) OVERRIDE;
virtual void SignalQuery(uint32 query,
const base::Closure& callback) OVERRIDE;
- virtual void SendManagedMemoryStats(const gpu::ManagedMemoryStats& stats)
- OVERRIDE;
+ virtual void SetSurfaceVisible(bool visible) OVERRIDE;
+ virtual void Echo(const base::Closure& callback) OVERRIDE;
+ virtual uint32 CreateStreamTexture(uint32 texture_id) OVERRIDE;
int GetRouteID() const;
- bool Echo(const base::Closure& callback);
bool ProduceFrontBuffer(const gpu::Mailbox& mailbox);
void SetChannelErrorCallback(const base::Closure& callback);
void AddDeletionObserver(DeletionObserver* observer);
void RemoveDeletionObserver(DeletionObserver* observer);
- bool DiscardBackbuffer();
bool EnsureBackbuffer();
- // Sends an IPC message with the new state of surface visibility.
- bool SetSurfaceVisible(bool visible);
-
void SetOnConsoleMessageCallback(
const GpuConsoleMessageCallback& callback);
- void SetLatencyInfo(const ui::LatencyInfo& latency_info);
+ void SetLatencyInfo(const std::vector<ui::LatencyInfo>& latency_info);
// TODO(apatrick): this is a temporary optimization while skia is calling
// ContentGLContext::MakeCurrent prior to every GL call. It saves returning 6
GpuChannelHost* channel() const { return channel_; }
+ base::SharedMemoryHandle GetSharedStateHandle() const {
+ return shared_state_shm_->handle();
+ }
+
private:
- typedef std::map<int32, gpu::Buffer> TransferBufferMap;
+ typedef std::map<int32, scoped_refptr<gpu::Buffer> > TransferBufferMap;
typedef base::hash_map<uint32, base::Closure> SignalTaskMap;
- typedef std::map<int32, gfx::GpuMemoryBuffer*> GpuMemoryBufferMap;
+ typedef base::ScopedPtrHashMap<int32, gfx::GpuMemoryBuffer>
+ GpuMemoryBufferMap;
// Send an IPC message over the GPU channel. This is private to fully
// encapsulate the channel; all callers of this function must explicitly
void OnConsoleMessage(const GPUCommandBufferConsoleMessage& message);
void OnSetMemoryAllocation(const gpu::MemoryAllocation& allocation);
void OnSignalSyncPointAck(uint32 id);
- void OnGenerateMailboxNamesReply(const std::vector<std::string>& names);
// Try to read an updated copy of the state from shared memory.
void TryUpdateState();
// The shared memory area used to update state.
- gpu::CommandBufferSharedState* shared_state() const {
- return reinterpret_cast<gpu::CommandBufferSharedState*>(
- shared_state_shm_->memory());
- }
-
- // Local cache of id to transfer buffer mapping.
- TransferBufferMap transfer_buffers_;
+ gpu::CommandBufferSharedState* shared_state() const;
// Unowned list of DeletionObservers.
ObserverList<DeletionObserver> deletion_observers_;
// Local cache of id to gpu memory buffer mapping.
GpuMemoryBufferMap gpu_memory_buffers_;
+ gpu::Capabilities capabilities_;
+
+ std::vector<ui::LatencyInfo> latency_info_;
+
DISALLOW_COPY_AND_ASSIGN(CommandBufferProxyImpl);
};