Hi Carl, > Here we see that radial gradients are 17 times slower than linear > gradients on this device. This is a big difference compared to the > results on my x86 laptop where radial gradients are only 2 times > slower than linear gradients. > > So this is definitely a problem spot, and I'm looking forward to > watching how David Turner's gradient improvements help here. > I hope it will improve linear gradients significantly for the reasons explained in my previous post. On the other hand, it shouldn't do squat for the radial ones, mainly because the performance hit corresponds here to the many floating point operations required per pixel, including a square root extraction.
the only way to get decent performance on the ARM here is to replace the computations with fixed point arithmetic; I suspect that using a Newton-Raphson loop to get the results would help tremendously as well. Hope this helps, - David Turner - The FreeType Project (www.freetype.org) _______________________________________________ Performance-list mailing list [email protected] http://mail.gnome.org/mailman/listinfo/performance-list
