Make GPU benchmarking numAnimationFrames and totalPaintTime work in software mode.

Make numAnimationFrames and numFramesSentToScreen available in software mode; these two have the same value in that mode.
Make totalPaintTimeInSeconds work in software mode.

This change depends on http://wkbug.com/92184 and http://wkbug.com/92325.


BUG=chromium:137754
TEST=Launch Chrome with --enable-gpu-benchmarking, open a tab to a non-composited page, and type chrome.gpuBenchmarking.renderingStats(). numAnimationFrames, numFramesSentToScreen, and totalPaintTimeInSeconds should be present.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=150259

[dtu, 2012-07-25 00:47:26]

PTAL

[nduca, 2012-07-25 17:14:04]

You should split out the code for the software fix from exposing new counters.
The software fix is a little harder than it might otherwise sound. YOu could
land the new counters right now, whereas the software fix needs some more
refinement.

https://chromiumcodereview.appspot.com/10818038/diff/5001/content/renderer/re...
File content/renderer/render_widget.cc (right):

https://chromiumcodereview.appspot.com/10818038/diff/5001/content/renderer/re...
content/renderer/render_widget.cc:894: total_do_deferred_update_count_++;
name is a bit awkward. Look in the .h for things that count and copy the
convention? Barring that, do_deferred_update_count or num_do_deferred_updates.
Also, you should only increment this when we're !accelerated, since in
accelerated single thread mode, we go through DoDeferredUpdate.

Its important that if you did some frames in software, some frames in hardwaare,
we get a good result. I'm not sure how to make that happen though... maybe get
the webview's data and then add in the render widget data?

https://chromiumcodereview.appspot.com/10818038/diff/5001/content/renderer/re...
content/renderer/render_widget.cc:975: PaintRect(copy_rects[i],
pixel_bounds.origin(), canvas.get());
there is code in here that forces a software paint. That should contribute to
timeSpentPainting

https://chromiumcodereview.appspot.com/10818038/diff/5001/content/renderer/re...
content/renderer/render_widget.cc:1777: if (is_accelerated_compositing_active_)
{
if (is ...) {
  get
  return
}
... unindented software case.

https://chromiumcodereview.appspot.com/10818038/diff/5001/content/renderer/re...
File content/renderer/render_widget.h (left):

https://chromiumcodereview.appspot.com/10818038/diff/5001/content/renderer/re...
content/renderer/render_widget.h:154: 
// Comment?

[dtu, 2012-07-27 01:40:48]

https://chromiumcodereview.appspot.com/10818038/diff/5001/content/renderer/re...
File content/renderer/render_widget.cc (right):

https://chromiumcodereview.appspot.com/10818038/diff/5001/content/renderer/re...
content/renderer/render_widget.cc:894: total_do_deferred_update_count_++;
On 2012/07/25 17:14:04, nduca wrote:
> name is a bit awkward. Look in the .h for things that count and copy the
> convention? Barring that, do_deferred_update_count or num_do_deferred_updates.
> Also, you should only increment this when we're !accelerated, since in
> accelerated single thread mode, we go through DoDeferredUpdate.
> 
> Its important that if you did some frames in software, some frames in
hardwaare,
> we get a good result. I'm not sure how to make that happen though... maybe get
> the webview's data and then add in the render widget data?

Just changed to deferred_update_count_. Hopefully sounds a little less awkward.
Other counters in this class are pending_window_rect_count_ and
num_swapbuffers_complete_pending_.

Done. Even if the current mode is software, it's going to plumb through and get
the frame counts from CCLayerTreeView if it's not null, then add the software
frame count to it in the RenderWidget. Other stats might not be accurate
though... (at this point in time)

https://chromiumcodereview.appspot.com/10818038/diff/5001/content/renderer/re...
content/renderer/render_widget.cc:975: PaintRect(copy_rects[i],
pixel_bounds.origin(), canvas.get());
On 2012/07/25 17:14:04, nduca wrote:
> there is code in here that forces a software paint. That should contribute to
> timeSpentPainting

Done this in WebViewImpl in http://wkbug.com/92325. There is already a histogram
there that records the software paint time, so just piggybacking off of that.

https://chromiumcodereview.appspot.com/10818038/diff/5001/content/renderer/re...
content/renderer/render_widget.cc:1777: if (is_accelerated_compositing_active_)
{
On 2012/07/25 17:14:04, nduca wrote:
> if (is ...) {
>   get
>   return
> }
> ... unindented software case.

Done. With the way the stats are calculated now, the if statement is
unnecessary.

https://chromiumcodereview.appspot.com/10818038/diff/5001/content/renderer/re...
File content/renderer/render_widget.h (left):

https://chromiumcodereview.appspot.com/10818038/diff/5001/content/renderer/re...
content/renderer/render_widget.h:154: 
On 2012/07/25 17:14:04, nduca wrote:
> // Comment?

Done.

[piman, 2012-07-27 03:39:44]

https://chromiumcodereview.appspot.com/10818038/diff/8002/content/renderer/re...
File content/renderer/render_widget.cc (right):

https://chromiumcodereview.appspot.com/10818038/diff/8002/content/renderer/re...
content/renderer/render_widget.cc:895: deferred_update_count_++;
I'm not exactly sure what you're trying to count, but that doesn't quite match
the name of the field :)

[dtu, 2012-07-31 01:35:00]

https://chromiumcodereview.appspot.com/10818038/diff/8002/content/renderer/re...
File content/renderer/render_widget.cc (right):

https://chromiumcodereview.appspot.com/10818038/diff/8002/content/renderer/re...
content/renderer/render_widget.cc:895: deferred_update_count_++;
On 2012/07/27 03:39:44, piman wrote:
> I'm not exactly sure what you're trying to count, but that doesn't quite match
> the name of the field :)

Removed this.

https://chromiumcodereview.appspot.com/10818038/diff/9006/content/renderer/re...
File content/renderer/render_widget.cc (right):

https://chromiumcodereview.appspot.com/10818038/diff/9006/content/renderer/re...
content/renderer/render_widget.cc:697: base::TimeDelta paint_time =
base::TimeTicks::Now() - paint_begin_ticks;
Does base::TimeTicks::Now() have the needed resolution in practice? The comment
on the method says it only has 1ms resolution.

[piman, 2012-07-31 03:32:49]

https://chromiumcodereview.appspot.com/10818038/diff/9006/content/renderer/re...
File content/renderer/render_widget.cc (right):

https://chromiumcodereview.appspot.com/10818038/diff/9006/content/renderer/re...
content/renderer/render_widget.cc:699: software_stats_.totalPaintTimeInSeconds
+= paint_time.InSecondsF();
We should have this in the "optimized_instance" case too above, for
software-mode pepper plugins.

[dtu, 2012-08-03 23:02:47]

https://chromiumcodereview.appspot.com/10818038/diff/9006/content/renderer/re...
File content/renderer/render_widget.cc (right):

https://chromiumcodereview.appspot.com/10818038/diff/9006/content/renderer/re...
content/renderer/render_widget.cc:699: software_stats_.totalPaintTimeInSeconds
+= paint_time.InSecondsF();
On 2012/07/31 03:32:49, piman wrote:
> We should have this in the "optimized_instance" case too above, for
> software-mode pepper plugins.

Done.

[piman, 2012-08-03 23:05:45]

lgtm

[commit-bot: I haz the power, 2012-08-06 21:20:01]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/dtu@chromium.org/10818038/3003

[commit-bot: I haz the power, 2012-08-06 21:20:05]

Failed to apply patch for content/renderer/gpu/gpu_benchmarking_extension.cc:
While running patch -p1 --forward --force;
patching file content/renderer/gpu/gpu_benchmarking_extension.cc
Hunk #1 FAILED at 26.
Hunk #2 succeeded at 159 (offset 46 lines).
1 out of 2 hunks FAILED -- saving rejects to file
content/renderer/gpu/gpu_benchmarking_extension.cc.rej

[commit-bot: I haz the power, 2012-08-06 21:50:23]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/dtu@chromium.org/10818038/3004

[commit-bot: I haz the power, 2012-08-06 22:38:06]

Try job failure for 10818038-3004 (retry) on linux_chromeos for step "compile"
(clobber build).
It's a second try, previously, step "compile" failed.
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=linux_chro...

[commit-bot: I haz the power, 2012-08-07 01:17:55]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/dtu@chromium.org/10818038/3004

[commit-bot: I haz the power, 2012-08-07 03:59:51]

Change committed as 150259