Make GLES2Implementation use a larger chunk size for mapping functions.

gpu/command_buffer/client/mapped_memory.cc

Index: gpu/command_buffer/client/mapped_memory.cc
diff --git a/gpu/command_buffer/client/mapped_memory.cc b/gpu/command_buffer/client/mapped_memory.cc
index 461343e25a0f8a012c64be04ea0f4877fbf8feba..c8d400ec25dc8be835ea8f68c654a923cf6017dd 100644
--- a/gpu/command_buffer/client/mapped_memory.cc
+++ b/gpu/command_buffer/client/mapped_memory.cc
@@ -23,7 +23,8 @@ MemoryChunk::MemoryChunk(
 }
 
 MappedMemoryManager::MappedMemoryManager(CommandBufferHelper* helper)
-    : helper_(helper) {
+    : chunk_size_multiple_(1),
+      helper_(helper) {
 }
 
 MappedMemoryManager::~MappedMemoryManager() {
@@ -52,7 +53,10 @@ void* MappedMemoryManager::Alloc(
 
   // Make a new chunk to satisfy the request.
   CommandBuffer* cmd_buf = helper_->command_buffer();
-  int32 id = cmd_buf->CreateTransferBuffer(size, -1);
+  unsigned int chunk_size =
+      ((size + chunk_size_multiple_ - 1) / chunk_size_multiple_) *
+      chunk_size_multiple_;
+  int32 id = cmd_buf->CreateTransferBuffer(chunk_size, -1);
   if (id == -1) {
     return NULL;
   }