[Telemetry] Output average GPU frequency from power metric

tools/perf/metrics/power.py

Index: tools/perf/metrics/power.py
diff --git a/tools/perf/metrics/power.py b/tools/perf/metrics/power.py
index c115d02e4de4480557823f54a8386791ad6da817..3b8ed2730d6e90980b38db413339a6a478fd455f 100644
--- a/tools/perf/metrics/power.py
+++ b/tools/perf/metrics/power.py
@@ -85,6 +85,10 @@ class PowerMetric(Metric):
     energy_consumption_mwh = self._results['energy_consumption_mwh']
     results.Add('energy_consumption_mwh', 'mWh', energy_consumption_mwh)
 
+    gpu_freq_hz = (
+        self._results['component_utilization']['gpu']['average_frequency_mhz'])
+    results.Add('gpu_average_frequency_hz', 'hz', gpu_freq_hz)

[tonyg, 2014/04/01 16:19:43]

Seems like we're missing the multiplication from mhz to hz.

Perhaps the unittest should check for this stat in the test that is based on a
static powermetrics output file.

Finally, should we check for the existence of 'gpu' and 'average_frequency_mhz'
before trying to access it and just silently not output this stat if
unavailable?

+
     # Add idle wakeup numbers for all processes.
     for process_type in self._results['cpu_stats']:
       trace_name_for_process = 'idle_wakeups_%s' % (process_type.lower())