[OpenJDK 2D-Dev] [9] Review Request: 8059941 [D3D] The fix for JDK-8029253 should be ported to d3d pipeline

[Phil Race]

Looks good although you probably should edit the OpenGL comment before 
pushing :-

  // transform IntArgbPre intermediate surface to OpenGL surface

-phil.


On 10/15/2014 11:43 AM, Sergey Bylokhov wrote:
> Hello.
> After a few attempts I am able to run 2dbench on my laptop in d3d mode:
>
> Config file:
> http://cr.openjdk.java.net/~serb/8059941/nvidia-d3d-win/drawimage
>
> Results:
> http://cr.openjdk.java.net/~serb/8059941/nvidia-d3d-win/results.txt
>
> Summary:
>   nvidia-d3d-base:
>     Number of tests:  10
>     Overall average:  547.843456221496
>     Best spread:      0.04% variance
>     Worst spread:     18.92% variance
>     (Basis for results comparison)
>
>   nvidia-d3d-fix:
>     Number of tests:  10
>     Overall average:  37294.61933635986
>     Best spread:      0.03% variance
>     Worst spread:     1.83% variance
>     Comparison to basis:
>       Best result:      15509.09% of basis
>       Worst result:     97.27% of basis
>       Number of wins:   9
>       Number of ties:   0
>       Number of losses: 1
>
> On 10.10.2014 14:53, Sergey Bylokhov wrote:
>> Hello.
>> Please review the fix for jdk 9.
>>
>> The new blit was implemented: D3DGeneralTransformedBlit. Performance 
>> difference is 30 seconds vs 25 ms in the test from 
>> https://bugs.openjdk.java.net/browse/JDK-8029253
>> I cannot provide J2DBench results because it simply crash my video 
>> driver(before/after the fix) if d3d pipeline is enabled. I will take 
>> a look at it later.
>>
>> Bug: https://bugs.openjdk.java.net/browse/JDK-8059941
>> Webrev can be found at: 
>> http://cr.openjdk.java.net/~serb/8059941/webrev.00
>>
>
>