Some more comments: On Wed, Dec 05, 2012 at 05:26:56PM +0000, Micael wrote: > I have made an attempt at optimizing the algorithm as well as the > performance with big radius dabs.
I have re-run the benchmark, it's good now for big dabs. I haven't benchmarked small dabs (I think there are some charcoal brushes that make really a lot of them, if you want to check). > The overall quality of the antialiasing effect should be better too due a > previous mistake I had made which is now corrected. I can't think of a > way to get rid of the division though. I believe it's not that much of a > problem though, since we're only going this way for small radius dabs. Yes I agree. Another thing: it may be worth double-checking that you never divide by zero, or something close to it. If I understood your code right, this was only a possible problem with very large dabs, so I think it's fine. > I don't know how to run the benchmark to see the results myself. I have > tried to "python test_performance.py" but every "print" statement generated > an error, and after fixing that it complained about the gtk imports. Sounds like you have run python3 instead of python2. You also need matplotlib installed (for python2 obviously). > I wasn't able to spot the step at the threshold value of radius=3.0 with > both the mouse and a wacom intuos 3. I tested with different brushes, I have also tested, and even tried to craft a brush that triggers such a problem, but I haven't found a way to trigger a glitch, neither. I didn't really understand the final delta calculations in your code, but the calculations above look correct to me. Overall I think the behaviour is good (with drawbacks mentioned in my other mail), certainly better than the gappy lines. -- Martin Renold _______________________________________________ Mypaint-discuss mailing list [email protected] https://mail.gna.org/listinfo/mypaint-discuss
