Expand UMA coverage for compositor-handled events and fling FPS

tools/metrics/histograms/histograms.xml

Index: tools/metrics/histograms/histograms.xml
diff --git a/tools/metrics/histograms/histograms.xml b/tools/metrics/histograms/histograms.xml
index 156913886691d88317b39ac951ec44852e954e9e..2fa015257c46ca17952ed9f794828920459b6b54 100644
--- a/tools/metrics/histograms/histograms.xml
+++ b/tools/metrics/histograms/histograms.xml
@@ -6701,6 +6701,26 @@ Therefore, the affected-histogram name has to have at least one dot in it.
   </summary>
 </histogram>
 
+<histogram name="Event.Frequency.Renderer.FlingAnimate" units="hertz">
+  <owner>rbyers@chromium.org</owner>
+  <summary>
+    Emitted after a renderer process, main-thread fling terminates, for any

[Mark P, 2014/12/01 23:13:01]

nit: omit first comma

ditto other histogram below

[jdduke (slow), 2014/12/02 01:20:52]

Done.

+    reason, reporting the average animation frequency (animations/second) of the
+    fling instance over its lifetime. This is computed as the number of fling
+    animation ticks divided by the fling animation duration.

[Mark P, 2014/12/01 23:13:01]

If I start a fling, keep my finger on the object I'm flinging, out to ten, then
finish my fling, this number will be very low, right?  (We don't need to animate
the object if I'm not moving it.)

If so, I think you should explain mention this, otherwise a low value might be
initially interpreted as bad animation / jank (too few animations per second).

If not, I think you should explain why this pause doesn't affect the animation
total (and hence the result is a good measure of jank).

ditto other histogram below

[jdduke (slow), 2014/12/02 01:20:52]

A fling animation by definition occurs only after you've lifted your finger.  I
guess the terminology may not be self-evident, but scrolling is the motion
induced by physical contact, and the fling is the virtual motion induced by the
velocity at the time of contact release. When the fling animation starts, it
will tick at the rate of the animation subsystem (independent of its velocity).
I tweaked the phrasing to indicating that we're animating a fling curve.

+  </summary>
+</histogram>
+
+<histogram name="Event.Frequency.RendererImpl.FlingAnimate" units="hertz">
+  <owner>rbyers@chromium.org</owner>
+  <summary>
+    Emitted after a renderer process, impl-thread fling terminates, for any
+    reason, reporting the average animation frequency (animations/second) of the
+    fling instance over its lifetime. This is computed as the number of fling
+    animation ticks divided by the fling animation duration.
+  </summary>
+</histogram>
+
 <histogram name="Event.GestureCreated" enum="UIEventType">
   <owner>kuscher@google.com</owner>
   <owner>rbyers@chromium.org</owner>
@@ -7045,6 +7065,18 @@ Therefore, the affected-histogram name has to have at least one dot in it.
   </summary>
 </histogram>
 
+<histogram name="Event.Latency.RendererImpl" units="microseconds">
+  <owner>rbyers@chromium.org</owner>
+  <summary>
+    Time between input event creation and the renderer impl thread receiving and
+    starting to process the event. For touch events on Windows, we measure from
+    when the event reaches Chrome, whereas on other platforms we use the
+    timestamp from the kernel. On Windows, this metric is only reported when
+    |IsHighResNowFastAndReliable| is true, which will introduce some sampling
+    bias.
+  </summary>
+</histogram>
+
 <histogram name="Event.Latency.RendererImpl.GestureScroll" units="microseconds">
   <obsolete>
     Deprecated 12/2013 and replaced by Event.Latency.RendererImpl.GestureScroll2
@@ -60545,6 +60577,7 @@ To add a new entry, add it with any value and run test to compute valid value.
   <suffix name="Undefined" label="For unknown or undefined events."/>
   <affected-histogram name="Event.Latency.Renderer"/>
   <affected-histogram name="Event.Latency.Renderer2"/>
+  <affected-histogram name="Event.Latency.RendererImpl"/>
 </histogram_suffixes>
 
 <histogram_suffixes name="ResourcePrefetchPredictorNetworkTypePrefetch"