Added rendering benchmark javascript hook

content/public/renderer/rendering_benchmark_results.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 CONTENT_PUBLIC_RENDERER_RENDERING_BENCHMARK_RESULTS_H_
+#define CONTENT_PUBLIC_RENDERER_RENDERING_BENCHMARK_RESULTS_H_
+
+#include <string>
+
+#include "content/common/content_export.h"
+
+namespace content {
+class CONTENT_EXPORT RenderingBenchmarkResults {

[nduca, 2012/07/18 06:06:17]

I'm not sure this abstraction is needed any more. We can have a benchmark have a
public inner struct called Result that has a unit and a value. Run can return
that struct. The name of the benchmark and the result name are redundant. Either
have benchmarks have a name, or have the result have a name, but not both.

[dmurph, 2012/07/18 16:31:41]

I agree, as long as you're fine with only one result per benchmark.  Think about
it - what if you want paint time and canvas count? Wouldn't it be easier to
record both results in the same benchmark?  Although, we would have to change
the naming from 'time_unit/time' to 'value_unit/value', but that would still
work.

[dmurph, 2012/07/18 20:01:57]

Agreed on leaving abstraction for allowing multiple results per test

+ public:
+  virtual ~RenderingBenchmarkResults() { }
+  virtual void addResult(const std::string& benchmark_name,
+                         const std::string& result_name,
+                         const std::string& time_unit,
+                         double time) = 0;
+};
+}
+#endif  // CONTENT_PUBLIC_RENDERER_RENDERING_BENCHMARK_RESULTS_H_