Guimark2.Bitmap shows 34% (13 fps, from 38.38 fps to 25.41 fps) regression in graphics-scrum build 2197
comparing to build 2196. Here is the link to Aurora results on fx-win-hi machine:
 http://aurora.russia.sun.com/performance/faces/ChessBoard.xhtml?reportName=FX2-graphics-scrum-trend&parameters=%5Bshownbenchmarks%5D%28benchmarkresults.name+%3D+%27GUIMark2.Bitmap%27%29%5Brelease%5D%28pr.product.productRelease+IS+NOT+NULL%29%5Bbuild%5D%28pr.product.build+%3E+2050%29&splitting=%5BX+axis%5DfxConf%5BComplement%5Dbenchmark%2C+jdkBuild%2C+jdk%2C+benchmarkSuite%5BY+axis%5DfxRelease%2C+fxBuild%5BZ+axis%5Dos%2C+hwclass%2C+jdkRelease%2C+benchmarkName%2C+benchmarkConf&reference=%5BOthers%5DfxRelease%2C+fxBuild%2C+fxConf%2C+benchmark%2C+benchmarkConf%2C+os%2C+hwclass%2C+jdkRelease%2C+jdkBuild%2C+jdk%2C+benchmarkSuite%2C+benchmarkName%5BReference+Set%5D&mixReference=false&flags=&significance=empty&hideDataConfiguration=false&calculateSummary=false&showSummaryExpanded=false&showSummaryContents=true&showComplementAttributes=false&compactTables=true&viewStyle=plot&filter=

The regression is easy reproducible. Running the test on Dell E6410 laptop also
shows the regression from ~18 fps to ~14 fps. JavaPerformanceAnalyzer shows that
the time spent in com.sun.javafx.tk.Toolkit.firePulse() has been increased from
3.4 sec to 9.9 sec. The cause is because of 8 sec spent in javafx.scene.Scene.doCSSPass()
which is because of 7.2 sec spent in com.sun.javafx.css.StyleHelper.transitionToState(javafx.scene.Node).


The "timeline" and "call tree" screenshots for builds 2196 (upper-left) and 2197 (bottom-right) are attached.