Hi folks! I want to discuss AWT 2D Performance Issue, especially performance of the BufferedImage rendering.
In order to recieve best performance we have to use hardware acceleration. Thus is appearing good thought using DirectDraw or Direct3D Surface as cache of BufferedImage Raster for rendering operations. In other words all drawing operations for BufferedImage Graphics we are doing on the BufferedImage cache and then draw BufferedImage cache on the screen with hardware acceleration. But sometimes we need Java access to the Raster, for example BufferedImage has getRGB and setRGB methods. Appear synchronization issue. We can use flags that inform us about cache or Raster was chaneged, after that we have to update cache or Raster. But customer can to recieve full access to the DataBuffer data (DataBuffer classes have getData methods which return link for arrays which store data) and we don't know when customer release data. Thus we have to update cache before each rendering operation what will slow down performance. I have done package private class which can safely give and release link for DataBuffer data and we can use this class for our private API, but end customer know nothing about this possibility :(. Have anybody ideas about how solve this problem. Wellcome to the discussion. Best regards, Igor
