Enable asynchronous glReadPixels on Windows.

gpu/command_buffer/client/mapped_memory.h

Index: gpu/command_buffer/client/mapped_memory.h
diff --git a/gpu/command_buffer/client/mapped_memory.h b/gpu/command_buffer/client/mapped_memory.h
index 789e69cf7f89d64fe052d2ca16be2875a93cd626..3689553303c71f5d12a23dddb3be6806581f5c4a 100644
--- a/gpu/command_buffer/client/mapped_memory.h
+++ b/gpu/command_buffer/client/mapped_memory.h
@@ -135,6 +135,7 @@ class GPU_EXPORT MappedMemoryManager {
   }
 
   void set_chunk_size_multiple(unsigned int multiple) {
+    DCHECK_GE(multiple, FencedAllocator::kAllocAlignment);

[piman, 2014/11/15 00:43:43]

Maybe DCHECK(multiple % FencedAllocator::kAllocAlignment == 0)?

[miu, 2014/11/15 01:04:26]

Done.  Right.  Any time the computed chunk_size == alloc_size (and chunk_size is
not divisible by 16), the rounding down issue can also happen.

     chunk_size_multiple_ = multiple;
   }
 
@@ -201,4 +202,3 @@ class GPU_EXPORT MappedMemoryManager {
 }  // namespace gpu
 
 #endif  // GPU_COMMAND_BUFFER_CLIENT_MAPPED_MEMORY_H_
-