Report utility process JS memory in task manager.

content/public/browser/process_resource_usage.h

+// Copyright 2015 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_BROWSER_PROCESS_RESOURCE_USAGE_H_
+#define CONTENT_PUBLIC_BROWSER_PROCESS_RESOURCE_USAGE_H_
+
+#include "base/memory/ref_counted.h"
+#include "content/common/content_export.h"
+#include "content/public/browser/browser_thread.h"
+
+namespace content {
+
+// Provides resource usage information about a child process.
+class CONTENT_EXPORT ProcessResourceUsage
+    : public base::RefCountedThreadSafe<ProcessResourceUsage> {
+ public:
+  // Refresh the resource usage information.
+  virtual void Refresh() = 0;

[ncarter (slow), 2015/05/01 21:16:52]

A question for afakhry:

Will you (with the rewrite) be okay with an interface that doesn't provide
notification to the task manager when the V8 stats are updated?

Because that's what this is doing, and it seems that limitation is somewhat
ingrained in the design here, since the model is "ChildProcessHost publishes a
ProcessResourceUsage for all listeners to peek at" rather than the listener
giving a completion callback.

If we want the completion notification it might be good to figure it out now.

[afakhry, 2015/05/04 20:30:03]

I prefer to have the (ChildProcess/Renderer)Tasks listen to V8 and WebCache
stats calculation events rather than having to trigger them on each refresh. If
there's anything that we can keep constant until it actually changes, then that
would be better. There's no point in doing an expensive calculation on each
refresh of the task manager if the values don't change.

[Anand Mistry (off Chromium), 2015/05/05 04:11:21]

It's not here because it's not necessary for this specific change. However,
adding a callback is trivial and necessary for replacing the renderer process V8
usage with this Mojo interface. Take a look at
https://codereview.chromium.org/1081323003/ for details. This still requires
requesting an update instead of passively listening, but it's already the case
that you need to request these stats be updated. This design makes that more
explicit.
The current chrome-IPC approach is to poll for these values, and this change
won't change that. For something like V8 memory usage, it's a value that you
can't observe for changes (and would be expensive to add) so you need to poll
anyway.

+
+  // Get V8 memory usage information.
+  virtual bool ReportsV8MemoryStats() const = 0;
+  virtual size_t GetV8MemoryAllocated() const = 0;
+  virtual size_t GetV8MemoryUsed() const = 0;
+
+ protected:
+  friend class base::RefCountedThreadSafe<ProcessResourceUsage>;
+  virtual ~ProcessResourceUsage() {}
+};
+
+}  // namespace content
+
+#endif  // CONTENT_PUBLIC_BROWSER_PROCESS_RESOURCE_USAGE_H_