<html> <body> At 12:57 PM 2/27/2008, Vladimir Vukicevic wrote: <blockquote type=cite class=cite cite="">One thing to be aware of... while the perf gains are significant, it means that the test suite isn't actually exercising painting an image with alpha transparency with OVER since the images are all opaque.  You should probably replace the images so that the actual OVER compositing path is being tested :)  (Or maybe create new tests, though I don't think there's much value in doing lots and lots of tests with the OVER-to-SOURCE optimization.)</blockquote> The extra tests are already in the perf suite, with: { set_source_image_surface_rgba, "image_rgba" }, I didn't include those tests in my perf sample as they don't get any benefit from the optimization. I've just run them myself to make sure there's no perf regression and the numbers pre/post optimization are + / - 2%. So someone running the full perf suite will see the OVER compositing path be tested with the RGBA source images. Antoine </body> </html>