New cef branch 2062

Having problems with building or using CEF's C/C++ APIs? This forum is here to help. Please do not post bug reports or feature requests here.

New cef branch 2062

Postby amitkanfer » Thu Aug 28, 2014 11:03 am

Hello guys,
i'm comparing the CPU load between branch 1916 and branch 2062, on a linux machine, and it seems that 1916 has much better results.
any idea why ?

where can i find updates release notes? on the CEF website the last release notes is for branch 1750...

thanks alot!
Amit
amitkanfer
Techie
 
Posts: 31
Joined: Mon Jul 21, 2014 4:10 am

Re: New cef branch 2062

Postby amitkanfer » Thu Aug 28, 2014 3:57 pm

OK,
it seems like i found the reason in this post:
https://code.google.com/p/chromiumembedded/issues/detail?id=1257

unless anyone has some other idea... :)

thanks
amitkanfer
Techie
 
Posts: 31
Joined: Mon Jul 21, 2014 4:10 am

Re: New cef branch 2062

Postby magreenblatt » Thu Aug 28, 2014 4:01 pm

If you're using off-screen rendering that's likely the important change. All changes are listed here: https://code.google.com/p/chromiumembedded/source/list
magreenblatt
Site Admin
 
Posts: 12382
Joined: Fri May 29, 2009 6:57 pm

Re: New cef branch 2062

Postby DOCaCola » Fri Aug 29, 2014 7:27 am

yes. the latest off screen implementation seems to be a lot slower than the one in older builds (tested using win32).
currently you are better off using the older branch in that regard.
DOCaCola
Newbie
 
Posts: 3
Joined: Sat May 24, 2014 4:01 am

Re: New cef branch 2062

Postby magreenblatt » Fri Aug 29, 2014 9:23 am

DOCaCola wrote:yes. the latest off screen implementation seems to be a lot slower than the one in older builds (tested using win32).
currently you are better off using the older branch in that regard.

I don't necessarily agree with this. What hardware (CPU/GPU/ram) are you using, and how are you testing the performance? Are you using a Release build of CEF?
magreenblatt
Site Admin
 
Posts: 12382
Joined: Fri May 29, 2009 6:57 pm

Re: New cef branch 2062

Postby amitkanfer » Fri Aug 29, 2014 9:29 am

I've tested it on monstress server with 32 cores and 256 GB of RAM but no GPU.

I will test it with GPU next week and update.
amitkanfer
Techie
 
Posts: 31
Joined: Mon Jul 21, 2014 4:10 am

Re: New cef branch 2062

Postby magreenblatt » Fri Aug 29, 2014 10:05 am

amitkanfer wrote:I've tested it on monstress server with 32 cores and 256 GB of RAM but no GPU.

I will test it with GPU next week and update.

OK, the lack of GPU will definitely hurt performance compared to the old implementation.
magreenblatt
Site Admin
 
Posts: 12382
Joined: Fri May 29, 2009 6:57 pm

Re: New cef branch 2062

Postby hlevring » Tue Sep 02, 2014 3:02 am

@amitkanfer

Did you get to continue with your tests ? When you said that your "monstress server" does not have a GPU, I presume you mean it does not have a discrete GPU but I guess it must have some kind of iGPU (onboard), can you be specific on that ..Intel HD2000 , HD3000 ?
hlevring
Newbie
 
Posts: 3
Joined: Fri May 31, 2013 10:02 am

Re: New cef branch 2062

Postby amitkanfer » Tue Sep 02, 2014 3:09 am

Hi,
i used a spec called 'c3.8xlarge' on AWS.
you can see the specifics here:
http://aws.amazon.com/ec2/instance-types/
it doesn't have any GPU.

Anyway - i did try to use 'g2.2xlarge', which does have GPU, but i haven't succeeded in getting any improvement in the CPU load yet...

will update once i have some results.
amitkanfer
Techie
 
Posts: 31
Joined: Mon Jul 21, 2014 4:10 am

Re: New cef branch 2062

Postby amitkanfer » Tue Sep 02, 2014 8:23 am

OK,
I've tested branch 2062 against amazon's g2.2xlarge instance.
Took me half a day to install the NVIDIA driver on it,
but once installed properly, and after installing Xorg server - i saw CEF creating child processes with 'type=gpu-process' and 'gpu-vendor-id' etc`.

i also verified that the GPU is utilized by watching 'nvidia-smi' modifications - and indeed it is utilized.

The results are disturbing.. Running 8 instances of CEF with GPU, browsing to a site like instagram.com (that has endless "OnPaint" events) - gave an average CPU which was HIGHER than running the exact same test, on the same server, with the same CEF code, but with different X server (Xvfb) and by that -- with no GPU.

bottom line - Somehow the GPU is only making things slower...
anyway - for now i'm sticking to branch 1916, which doesn't use the GPU, but the fact that it has the 'dirty rectangles' - makes a BIG difference (CPU-wise).

Perhaps i'm missing something... i don't know.
amitkanfer
Techie
 
Posts: 31
Joined: Mon Jul 21, 2014 4:10 am

Next

Return to Support Forum

Who is online

Users browsing this forum: No registered users and 30 guests