cc: Improve texture per second calculation and overhead

This keeps a sorted history of the texture upload rate and increases
the history size to improve consistency/accuracy and overhead.

Prior to this patch, calling ThrottledTextureUploader::estimatedTexturesPerSecond
ran a sort operation, which will be slow for larger history sizes.

Instead of keeping a FIFO history, it keeps sorted history and discards
the min and max values.
    
BUG=155172
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=163828

[reveman, 2012-10-11 17:59:46]

I'm generally OK with this but a bit worried about having up to a 1000 queries
around. Especially as it also affects single threaded mode. Maybe we should make
it not affect single threaded mode first?

https://codereview.chromium.org/11028129/diff/1/cc/ThrottledTextureUploader.cpp
File cc/ThrottledTextureUploader.cpp (right):

https://codereview.chromium.org/11028129/diff/1/cc/ThrottledTextureUploader.c...
cc/ThrottledTextureUploader.cpp:128: std::sort(sortedHistory.begin(),
sortedHistory.end());
with the history now relatively large, does it makes sense to do something a bit
smarter than sorting every time estimatedTexturesPerSecond() is called? seems
like a balanced binary tree is just what we need here?

[brianderson, 2012-10-12 00:31:48]

Addressed comments.  I decided not to keep a parallel FIFO history structure
because discarding the min/max values instead works really well.

> I'm generally OK with this but a bit worried about having up to a 1000 queries
around.

This should not affect the number of pending queries.

[reveman, 2012-10-12 01:41:00]

I think the increased history size and removal of max/min values are great!
Could we keep the details like using 3/4 median separately?

https://codereview.chromium.org/11028129/diff/4001/cc/ThrottledTextureUploade...
File cc/ThrottledTextureUploader.cpp (right):

https://codereview.chromium.org/11028129/diff/4001/cc/ThrottledTextureUploade...
cc/ThrottledTextureUploader.cpp:126: // Use the 3/4 median as our optimistic
estimate.
I'm not convinced that 3/4 median magic number we should use here. Can we tune
these details separately? Preferably after we land your 'fill upload queue
patch'.

https://codereview.chromium.org/11028129/diff/4001/cc/ThrottledTextureUploade...
cc/ThrottledTextureUploader.cpp:128: std::advance(median,
-(int)m_texturesPerSecondHistory.size() / 4);
nit: use static_cast<> here

https://codereview.chromium.org/11028129/diff/4001/cc/ThrottledTextureUploade...
cc/ThrottledTextureUploader.cpp:180:
m_texturesPerSecondHistory.insert(texturesPerSecond);
maybe do the remove before adding the new value?

https://codereview.chromium.org/11028129/diff/4001/cc/ThrottledTextureUploade...
cc/ThrottledTextureUploader.cpp:182: if (m_historyRemovalCount & 0x3)
how about we save the 3/4 and 1/4 details for later?

lets get rid of m_historyRemovalCount and just do multiple removals here when
size() >= uploadHistorySizeMax.

[reveman, 2012-10-23 21:06:58]

lgtm. would be great if you removed the global variable while you're at it.

https://codereview.chromium.org/11028129/diff/10001/cc/texture_uploader.cc
File cc/texture_uploader.cc (right):

https://codereview.chromium.org/11028129/diff/10001/cc/texture_uploader.cc#ne...
cc/texture_uploader.cc:29: static double estimatedTexturesPerSecondGlobal = 48.0
* 60.0;
Please make this a constant instead. The TextureUploader is only recreated when
we lose the context so I don't think it makes sense to cache this between
instances.

[brianderson, 2012-10-23 21:14:20]

Adding enne for owners review.

[enne (OOO), 2012-10-23 21:18:24]

lgtm

[commit-bot: I haz the power, 2012-10-23 21:21:30]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/brianderson@chromium.org/11028129/14001

[commit-bot: I haz the power, 2012-10-23 21:37:28]

Sorry for I got bad news for ya.
Compile failed with a clobber build on android_dbg.
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=android_db...
Your code is likely broken or HEAD is junk. Please ensure your
code is not broken then alert the build sheriffs.
Look at the try server FAQ for more details.

[commit-bot: I haz the power, 2012-10-23 21:37:34]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/brianderson@chromium.org/11028129/19001

[commit-bot: I haz the power, 2012-10-24 15:12:49]

Change committed as 163828