improving UI of FPS counter

improving UI of FPS counter

The FPS counter shows the frame rate based on a time-stamp history and is located in the compositor's HUDLayer.

This change includes:
- moving to top right corner
- plotting the graph's X-axis based on time between frames
- measuring min/max value instead of standard deviation
- showing histogram of time spent in each value
- scaling to the max value in case of very high fps (disabled vsync)

comparison images:
https://docs.google.com/folder/d/0B8Y78t3tjy1XTC1OYzUySFgwODQ/edit

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=167765

[egraether, 2012-11-12 23:44:05]

https://codereview.chromium.org/11369200/diff/1/cc/frame_rate_counter.cc
File cc/frame_rate_counter.cc (left):

https://codereview.chromium.org/11369200/diff/1/cc/frame_rate_counter.cc#oldc...
cc/frame_rate_counter.cc:110: double x = 1.0 / delta.InSecondsF();
This average calculation is wrong, because x is already a speed value and
therefore an average. Averaging averages leads to a wrong result. The error was
in most cases really small though.

https://codereview.chromium.org/11369200/diff/1/cc/frame_rate_counter.cc#oldc...
cc/frame_rate_counter.cc:114: fpsVarianceNumerator += deltaFromAverage * (x -
averageFPS);
The standard deviation calculation was thereon wrong as well. It is no longer
used, so I removed it completely.

https://codereview.chromium.org/11369200/diff/1/cc/frame_rate_counter.cc
File cc/frame_rate_counter.cc (right):

https://codereview.chromium.org/11369200/diff/1/cc/frame_rate_counter.cc#newc...
cc/frame_rate_counter.cc:15: const double FrameRateCounter::kFrameTooSlow = 1.0;
Reduced the kFrameTooSlow value so that only time deltas of above one second get
sorted out. This change happened due to feedback complaining that really slow
framerates, e.g. during lags, are not showing up.

https://codereview.chromium.org/11369200/diff/1/cc/frame_rate_counter.cc#newc...
cc/frame_rate_counter.cc:84: double FrameRateCounter::getAverageFPS() const
Method returns now the average only. Standard deviation was removed due to wrong
calculation.

https://codereview.chromium.org/11369200/diff/1/cc/frame_rate_counter.cc#newc...
cc/frame_rate_counter.cc:102: while (frameIndex(frameNumber) !=
frameIndex(m_currentFrameNumber) && frameNumber >= 0 && frameDeltas < 1.0) {
I added a check for a maximum time of 1 second, so that the average is not too
widespread.

https://codereview.chromium.org/11369200/diff/1/cc/frame_rate_counter.cc#newc...
cc/frame_rate_counter.cc:115: averageFPS = frameCount / frameDeltas;
The average needs to be calculated by total count divided by total time.

https://codereview.chromium.org/11369200/diff/1/cc/heads_up_display_layer.cc
File cc/heads_up_display_layer.cc (right):

https://codereview.chromium.org/11369200/diff/1/cc/heads_up_display_layer.cc#...
cc/heads_up_display_layer.cc:44:
matrix.translate(layerTreeHost()->deviceViewportSize().width() - 256, 0);
Moves the HUDLayer to the right of the screen if only the FPS counter is shown.

https://codereview.chromium.org/11369200/diff/1/cc/heads_up_display_layer_imp...
File cc/heads_up_display_layer_impl.cc (right):

https://codereview.chromium.org/11369200/diff/1/cc/heads_up_display_layer_imp...
cc/heads_up_display_layer_impl.cc:7: #include <float.h>
Necessary to set member m_minFPS to DBL_MAX. Worked without this as well though,
but I thought I rather add it.

https://codereview.chromium.org/11369200/diff/1/cc/heads_up_display_layer_imp...
cc/heads_up_display_layer_impl.cc:216: std::string minMaxText =
base::StringPrintf("%.0f-%.0f", std::min( m_minFPS, m_maxFPS), m_maxFPS);
Shows the min/max fps instead of standard deviation.

https://codereview.chromium.org/11369200/diff/1/cc/heads_up_display_layer_imp...
cc/heads_up_display_layer_impl.cc:230: const double hiFPS = std::max(m_maxFPS +
10.0, 80.0);
Setting the upper fps bound of graph and histogram to m_maxFPS if it exceeds the
default value of 80fps. This is especially helpful when GPU vsync is disabled.

https://codereview.chromium.org/11369200/diff/1/cc/heads_up_display_layer_imp...
cc/heads_up_display_layer_impl.cc:287: x += delta.InSecondsF() * timeScale;
Move x based on time spent between frames.

https://codereview.chromium.org/11369200/diff/1/cc/heads_up_display_layer_imp...
cc/heads_up_display_layer_impl.cc:300: double barWidth = histogram[i] /
maxBucketValue * histogramBounds.width();
Scale the width of the bar to the maximum in the histogram, so no bar ever goes
out the graph.

[nduca, 2012-11-13 00:51:46]

nits

https://codereview.chromium.org/11369200/diff/1/cc/frame_rate_counter.cc
File cc/frame_rate_counter.cc (right):

https://codereview.chromium.org/11369200/diff/1/cc/frame_rate_counter.cc#newc...
cc/frame_rate_counter.cc:15: const double FrameRateCounter::kFrameTooSlow = 1.0;
How about .25 second? I'm thinking about pauses as you scroll down...

https://codereview.chromium.org/11369200/diff/1/cc/frame_rate_counter.cc#newc...
cc/frame_rate_counter.cc:115: averageFPS = frameCount / frameDeltas;
frameDeltas is an oddly named var. Can you make better?

https://codereview.chromium.org/11369200/diff/1/cc/heads_up_display_layer_imp...
File cc/heads_up_display_layer_impl.cc (right):

https://codereview.chromium.org/11369200/diff/1/cc/heads_up_display_layer_imp...
cc/heads_up_display_layer_impl.cc:7: #include <float.h>
I think there is a numeric limts thing for this....  git grep limits  | grep
numeric to see what the right thing is.

[nduca, 2012-11-14 01:09:35]

lgtm, nice

[commit-bot: I haz the power, 2012-11-14 19:10:44]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/egraether@chromium.org/11369200/12001

[commit-bot: I haz the power, 2012-11-14 23:34:56]

Change committed as 167765