For the record, Matthew pushed a fix involving `cairo_fill` and `cairo_paint_with_alpha`: https://github.com/plt/racket/commit/54e42664dfd2aeafbe41c2ff424d86affe171962
Laurent On Mon, Apr 21, 2014 at 4:00 PM, Laurent <laurent.ors...@gmail.com> wrote: > I have a 2htdp/universe program that used to run fast enough a few months > ago, but now it is very slow and not usable. > The slowness seems to be because of the on-screen rendering, and not > because of the generation of the image. > > Here is a stripped-down version that shows this behavior: > https://gist.github.com/Metaxal/11142941 > > In the following log, you see that the `on-key` events are very close one > to the other (in milliseconds after the beginning of the program), but the > corresponding `to-draw` events are separated by more than a second, even > though generating the image (cpu time) takes almost no time: > > on-key a at 6906 > on-key u at 6912 > on-key i at 6912 > on-key e at 6913 > to-draw at 6913 > to-draw: cpu time: 4 real time: 3 gc time: 0 > to-draw at 8598 > to-draw: cpu time: 4 real time: 2 gc time: 0 > to-draw at 11948 > to-draw: cpu time: 4 real time: 2 gc time: 0 > to-draw at 13631 > to-draw: cpu time: 0 real time: 2 gc time: 0 > to-draw at 161839 > to-draw: cpu time: 4 real time: 9 gc time: 0 > > During those long seconds, Xorg is almost at 100% cpu. > > However, using an empty scene instead of an image is fast. > The time also depends on the size of the grid. > > I'm using Ubuntu 12.04 64bits. > I have tried to replicate the behavior on older versions of racket (5.3.1 > and 5.90.0.9) but it's the same. So maybe the problem is not on Racket's > side but something has changed in Ubuntu? > > Does anyone else see the same behavior, either on the same platform or on > a different one? > > Thanks, > Laurent > >
____________________ Racket Users list: http://lists.racket-lang.org/users
