java2d performance java7 / java8

[Florian Bruckner (3kraft)]

Hi Sergey,

I have built a minimalistic test case that exposes the problem. It creates a JFrame with an editor 
pane in it to show some HTML content (a javadoc page). The test is to just scroll from the bottom to 
the top and back to the bottom by mouse wheel as quickly as possible, I did 5 iterations in each 
test run.

Wall clock timing was approx. 15 seconds with Java 6 and approx 45 seconds with java 8. As the 
problem worsens with increased screen resolutions I decided to use the full resolution of my new UHD 
display (which is 3840x2160), the machine is a

MacBook Pro
15-inch, Early 2011
Processor  2 GHz Intel Core i7
Memory  16 GB 1333 MHz DDR3
Graphics  AMD Radeon HD 6490M 256 MB
Software  OS X 10.9.4 (13E28)

These are the java versions:

$ java -version
java version "1.8.0_11"
Java(TM) SE Runtime Environment (build 1.8.0_11-b12)
Java HotSpot(TM) 64-Bit Server VM (build 25.11-b03, mixed mode)

$ /System/Library/Frameworks/JavaVM.framework/Home/bin/java -version
java version "1.6.0_65"
Java(TM) SE Runtime Environment (build 1.6.0_65-b14-462-11M4609)
Java HotSpot(TM) 64-Bit Server VM (build 20.65-b04-462, mixed mode)

With java 6, the application felt agile and responding, whereas with java 8 it was barely usable.

Attachments were rejected from the list, therefore I have uploaded them to our web storage (this 
link expires Aug 31st):

https://owncloud.3kraft.com/public.php?service=files&t=ae0f609d6bd42b27a5ea38261507a38c

Please let me know if those results are usable for you.

regards,

Florian

On 05.08.14 15:37, Sergey Bylokhov wrote:
> Hi, Florian.
> On 04.08.2014 19:40, Florian Bruckner (3kraft) wrote:
>> Are you guys aware of a performance regression from java6 on osx? Is this something you think is
>> worth investigating?
>>
>> If yes - I can offer systematic testing on a range of different MBP models, starting with
>> pre-unibody MBP up until quite recent retina MBPs, all attached to the same screen to get
>> comparable results. Building and modifying the rendering pipeline is unfortunately beyond my
>> capabilities (at least without your support).
> Yes, we already have some reports on this problem, but not all of them can be reproduced by us
> locally.
> It would be good to find the bottleneck using a native profilers(ex TimeProfiler which is part of
> XCode), on different systems with different video cards/screen size.
> In your case would be good to know the differences in profiler results between before/after
> SQLDeveloper is started.
>>
>> thanks and regards,
>>
>> Florian
>
>