Added GetCpuTime to base/ to get total CPU time consumed by process for perf tests.

webrtc/base/cpu_time.h

Index: webrtc/base/cpu_time.h
diff --git a/webrtc/base/cpu_time.h b/webrtc/base/cpu_time.h
new file mode 100644
index 0000000000000000000000000000000000000000..fca337ec0cd8b9cde8126cb999211b8e9f7d8991
--- /dev/null
+++ b/webrtc/base/cpu_time.h
@@ -0,0 +1,28 @@
+/*
+ *  Copyright (c) 2017 The WebRTC project authors. All Rights Reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS.  All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#ifndef WEBRTC_BASE_CPU_TIME_H_
+#define WEBRTC_BASE_CPU_TIME_H_
+
+namespace rtc {
+
+class CpuTime {

[nisse-webrtc, 2017/02/15 12:18:17]

Why a class? Could be just rtc::GetCpuTime.

[ilnik, 2017/02/15 12:53:21]

No particular reason. I've seen it somewhere and thought it is a preferred style
here. At the second look, it is rather an exception than a standard. Will do it
as a function.

+ public:
+  //  Returns total CPU time of a current process.
+  //  Use to calculate difference as some offset may be presented.
+  static double GetCpuTime();

[nisse-webrtc, 2017/02/15 12:18:17]

Any particular reason to use floating point, rather than int64_t with
nanoseconds? If I get the numbers right, 2^63 ns is a few hundred years.

[ilnik, 2017/02/15 12:53:21]

For portability reasons. Initially I thought of using clock()/CLOCKS_PER_SECOND
for POSIX platforms, but it may not align perfectly with nanoseconds. I believe
float will be more logical as a common denominator. But int64_t will do as well.

+
+ private:
+  CpuTime() {}
+};
+
+}  // namespace rtc
+
+#endif  // WEBRTC_BASE_CPU_TIME_H_