Copying uncached heap-based images to an accelerated surface is slower with
the new pipeline than in earlier releases, especially when copying
to the screen.
This affects applications which do software-only rendering
(direct pixel manipulation) and then copy images to the screen.
With the attached application:
6u4:
#>java PerfTest
using AWT
using BI->Screen
1367
1622
1641
1636
6u10:
#>java PerfTest
using AWT
using BI->Screen
812
908
997
936
6u4: (BI->BS->screen)
#> java -Dusebs=true PerfTest
using AWT
using BS
1786
2256
2259
2260
6u10: (BI->BS->screen)
#>java -Dusebs=true PerfTest
using AWT
using BS
1204
1334
1382
1381
In both 6u4 and 6u10 the GDI pipeline is the fastest for this
particular application:
6u4:
#> java -Dsun.java2d.noddraw=true PerfTest
using AWT
using BI->Screen
3147
3425
3429
6u10:
#>java -Dsun.java2d.d3d=false PerfTest
using AWT
using BI->Screen
3321
3497
3538