![]() ![]() HiddenScrollbar, but that turned out to be trivial. Since PyPy does not support Tkinter, I had to stub out This finally prodded me into getting PyPy 1.4 working with > To post to this group, send email to > To unsubscribe from this group, send email to > For more options, visit this group at /group/makerbot?hl=en. > You received this message because you are subscribed to the Google Groups "MakerBot Operators" group. If it could be done, though, I'd love it. > no programmer, though (I'm a hardware guy), so I have no idea how hard > I'm sure this could be done in OpenCL to keep it cross-platform. > I would love to use my 5850 for things other than playing Team > a graphics card like an ATI Radeon HD 5850 with its 1440 stream > realized that the whole process could be accelerated if Skeinforge was > As I was waiting for #11 in the Holiday Prusa Mendel set to slice, I But again, it would probably not be a significant improvement. > meaning to try running it with the latest PyPy, which should be > Cython didn't get any significant improvements though. > that could be speed up by using C extensions. > I spent a bit of time profiling and trying to find parts of Skeinforge Also, Skeinforge has a pretty high development velocity. ![]() However, the big problem will be porting _all_ of > Superskein is taking a stab at this by using Processing (which is Simply porting Skeinforge from Python into C++ or > I'd appreciate any help with this project. The speed improvements are quite remarkable: A test object, needing 9 seconds to carve with Skeinforge, needs less than 40 ms to slice with Pleasant3D. > I wrote a OpenCL version of the carve tool and integrated it into Pleasant3D. > So I decided, instead of trying to port and pimp up the skeinforge sources, it would be much more effective (and fun!) to develop completely new and "native" algorithms, using OpenCL et al. > I was able to write some code optimizations and object caching, but even the advanced version wasn't much faster than 2x. This probably has something to do with the relatively slow handling of object creation in C/C++ (and the heavy use of objects in the original code). The C port was even slower than the python code. Besides the really dull task of porting code to another language, my tests with the ported skeinforge tool "carve" were frustrating. On Dec 11, 7:47 pm, Eberhard Rensch wrote: I'd appreciate any help with this project. I wrote a OpenCL version of the carve tool and integrated it into Pleasant3D. So I decided, instead of trying to port and pimp up the skeinforge sources, it would be much more effective (and fun!) to develop completely new and "native" algorithms, using OpenCL et al. I was able to write some code optimizations and object caching, but even the advanced version wasn't much faster than 2x. > that taking it further by using GPUs would probably feel unnecessary. > even Java would result in such significant performance improvements ![]()
0 Comments
Leave a Reply. |