gpu: Thread-safe command buffer state lookup.

gpu/command_buffer/common/command_buffer.h

 // Copyright (c) 2012 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.
 
 #ifndef GPU_COMMAND_BUFFER_COMMON_COMMAND_BUFFER_H_
 #define GPU_COMMAND_BUFFER_COMMON_COMMAND_BUFFER_H_
 
 #include <stddef.h>
 #include <stdint.h>
 
 #include "base/macros.h"
 #include "gpu/command_buffer/common/buffer.h"
 #include "gpu/command_buffer/common/constants.h"
 #include "gpu/gpu_export.h"
 
 namespace gpu {
 
 // Common interface for CommandBuffer implementations.
 class GPU_EXPORT CommandBuffer {
  public:
   struct State {
     State()
         : get_offset(0),
           token(-1),
           error(error::kNoError),
           context_lost_reason(error::kUnknown),
           generation(0) {

[jbauman, 2016/12/09 03:16:03]

initialize release_count

[sunnyps, 2016/12/09 04:09:37]

Done.

     }
 
     // The offset (in entries) from which the reader is reading.
     int32_t get_offset;
 
     // The current token value. This is used by the writer to defer
     // changes to shared memory objects until the reader has reached a certain
     // point in the command buffer. The reader is responsible for updating the
     // token value, for example in response to an asynchronous set token command
     // embedded in the command buffer. The default token value is zero.
     int32_t token;
 
+    uint64_t release_count;
+
     // Error status.
     error::Error error;
 
     // Lost context detail information.
     error::ContextLostReason context_lost_reason;
 
     // Generation index of this state. The generation index is incremented every
     // time a new state is retrieved from the command processor, so that
     // consistency can be kept even if IPC messages are processed out-of-order.
     uint32_t generation;
@@ skipping 17 matching lines @@
   static bool InRange(int32_t start, int32_t end, int32_t value) {
     if (start <= end)
       return start <= value && value <= end;
     else
       return start <= value || value <= end;
   }
 
   // Returns the last state without synchronizing with the service.
   virtual State GetLastState() = 0;
 
-  // Returns the last token without synchronizing with the service. Note that
-  // while you could just call GetLastState().token, GetLastState needs to be
-  // fast as it is called for every command where GetLastToken is only called
-  // by code that needs to know the last token so it can be slower but more up
-  // to date than GetLastState.
-  virtual int32_t GetLastToken() = 0;
-
   // The writer calls this to update its put offset. This ensures the reader
   // sees the latest added commands, and will eventually process them. On the
   // service side, commands are processed up to the given put_offset before
   // subsequent Flushes on the same GpuChannel.
   virtual void Flush(int32_t put_offset) = 0;
 
   // As Flush, ensures that on the service side, commands up to put_offset
   // are processed but before subsequent commands on the same GpuChannel but
   // flushing to the service may be deferred.
   virtual void OrderingBarrier(int32_t put_offset) = 0;
 
   // The writer calls this to wait until the current token is within a
   // specific range, inclusive. Can return early if an error is generated.
-  virtual void WaitForTokenInRange(int32_t start, int32_t end) = 0;
+  virtual State WaitForTokenInRange(int32_t start, int32_t end) = 0;
 
   // The writer calls this to wait until the current get offset is within a
   // specific range, inclusive. Can return early if an error is generated.
-  virtual void WaitForGetOffsetInRange(int32_t start, int32_t end) = 0;
+  virtual State WaitForGetOffsetInRange(int32_t start, int32_t end) = 0;
 
   // Sets the buffer commands are read from.
   // Also resets the get and put offsets to 0.
   virtual void SetGetBuffer(int32_t transfer_buffer_id) = 0;
 
   // Create a transfer buffer of the given size. Returns its ID or -1 on
   // error.
   virtual scoped_refptr<gpu::Buffer> CreateTransferBuffer(size_t size,
                                                           int32_t* id) = 0;
 
   // Destroy a transfer buffer. The ID must be positive.
   virtual void DestroyTransferBuffer(int32_t id) = 0;
 
-// The NaCl Win64 build only really needs the struct definitions above; having
-// GetLastError declared would mean we'd have to also define it, and pull more
-// of gpu in to the NaCl Win64 build.
-#if !defined(NACL_WIN64)
-  // TODO(apatrick): this is a temporary optimization while skia is calling
-  // RendererGLContext::MakeCurrent prior to every GL call. It saves returning 6
-  // ints redundantly when only the error is needed for the CommandBufferProxy
-  // implementation.
-  virtual error::Error GetLastError();
-#endif
-
  private:
   DISALLOW_COPY_AND_ASSIGN(CommandBuffer);
 };
 
 }  // namespace gpu
 
 #endif  // GPU_COMMAND_BUFFER_COMMON_COMMAND_BUFFER_H_