Hi Gabor, On Sun, 2010-01-17 at 01:33 +0100, gabor papp wrote: > > In that case, a quick fix might be to save out the primitives from > > fluxus, as they should be quicker to load back in once processed this > > way. > unfortunately that didn't work, because the conversion algorithm still > had to process all indices.
It's a long time since I looked at this code, but could it test if all the indices were the same for a vertex and skip the processing? I'm not sure what the long term answer is though. > i optimized the loading a bit by merging the vector loading and storing > the unified index per face vertex. on a 1 megabyte obj file, the loading > time dropped from 5.6 seconds to 2.1, which is still too slow. Thanks for this - am I missing some faster way, e.g. some way to use separate indices in OpenGL? cheers, dave
