Forwarding to list ---------- Forwarded message ---------- From: Popolon <popo...@popolon.org> Date: 21 November 2012 01:55 Subject: Re: [Mypaint-discuss] Surface optimizations proposed for merging To: Jon Nordby <jono...@gmail.com>
I made some test with the last git version versus surface optimization version and here are the results in mycase (core i7 4core + hyperthreading ~8core)@3.7Ghz, no opencl support (no intel openCL driver on linux). I tested mosly the deevad brush : deevad/basic_digital_knife_**smudging). this seem to be the slower brush. In all case i growed at max the size of the brush by pressing several time [f] key. All brush are fluid in surface.optimizations version but the deevad/basic_digital_knife_**smudging. In most case this brush is very slow in standard version (perhaps 0.20 to 0.5 s lag). In surface optimization version, it seems to fluidly drawn, when the whole stroke is included in the portion of the canvas seen at screen, but suffer of big slow down every time the brush stroke go out of this area, perhaps simply because the whole surface stroked is bigger ? I made a video record with this brush, but it seems that's the record is not really in real time, probably a little bit accelerated ??? I used gtk-record-mydestkop for this screencast, if someone know a better option, I could make a newone. http://popolon.org/mypaint/**out_seems_accelerated.ogv<http://popolon.org/mypaint/out_seems_accelerated.ogv> I have some slowdown with deevad/watercolor_glazing, deevad/thin_watercolor and deevad/large_watercolor_fringe brushs too. Popolon Le 19/11/2012 13:11, Jon Nordby a écrit : > > On 18 November 2012 03:12, Jon Nordby <jono...@gmail.com > <mailto:jono...@gmail.com>> wrote: > > I have some more ideas for further improve performance, and am working > to document these now. > > Now documented in the surface-optimizations branch, file > brushlib/PERFORMANCE: > http://gitorious.org/mypaint/**mypaint/blobs/surface-** > optimizations/brushlib/**PERFORMANCE<http://gitorious.org/mypaint/mypaint/blobs/surface-optimizations/brushlib/PERFORMANCE> > > Here are the main points: > === TODO: Improve vectorization === > === TODO: More efficient serial code === > === TODO: Try different tile sizes === > === IDEA: Dab masks cache === > === IDEA: Make use of GPU processing: OpenCL and OpenGL === > > -- > Jon Nordby - www.jonnor.com <http://www.jonnor.com> > > > ______________________________**_________________ > Mypaint-discuss mailing list > Mypaint-discuss@gna.org > https://mail.gna.org/listinfo/**mypaint-discuss<https://mail.gna.org/listinfo/mypaint-discuss> > > -- Jon Nordby - www.jonnor.com
_______________________________________________ Mypaint-discuss mailing list Mypaint-discuss@gna.org https://mail.gna.org/listinfo/mypaint-discuss