Hi Sergey,
could you please clarify why you have eliminated OGLC_VENDOR_SUN?
Thanks,
Andrew
On 6/11/2014 5:01 PM, Sergey Bylokhov wrote:
Hello.
Please review the fix for jdk 9, which is also targeted for jdk 8u20.
Description:
We have two codepaths to draw a raster to the opengl:
- via glDrawPixels
- via intermediate texture.
We empirically checked, what way is better, based on benchmarks.
Became obvious that on intel devices the glDrawPixels is slower than
intermediate texture, and this is a bottleneck when the scrolling is
used.
j2bench is not properly cover retina case, but according to some of my
tests results is 2-3 times better.
There is no regressions on windows:
Summary:
ogl-base:
Number of tests: 60
Overall average: 1452524.357368045
Best spread: 0.17% variance
Worst spread: 248.44% variance
(Basis for results comparison)
ogl-fix:
Number of tests: 60
Overall average: 1469528.5484618822
Best spread: 0.12% variance
Worst spread: 246.59% variance
Comparison to basis:
Best result: 161.48% of basis
Worst result: 93.54% of basis
Number of wins: 38
Number of ties: 17
Number of losses: 5
http://cr.openjdk.java.net/~serb/8029253/perf/windows.txt
Bug: https://bugs.openjdk.java.net/browse/JDK-8029253
Webrev can be found at:
http://cr.openjdk.java.net/~serb/8029253/webrev.00
