https://community.intel.com/t5/Intel-Embree-Ray-Tracing-Kernels/MLS-Point-Set-Surfaces/m-p/948933#M227 <P>The current version of Embree does not support user defined primitives directly. However, you can of course modify the code to let the ray/primitive intersection do what you need.</P> Mon, 06 May 2013 06:05:00 GMT SvenW_Intel 2013-05-06T06:05:00Z https://community.intel.com/t5/Intel-Embree-Ray-Tracing-Kernels/MLS-Point-Set-Surfaces/m-p/948932#M226 <P>Hi,</P> <P>I'm very interested in implementing basic (projection)&nbsp;<STRONG>moving least squares</STRONG>-based <STRONG>point set surfaces</STRONG> using Embree. However, the library seems mostly optimized for triangle meshes, so I'm curious whether my idea is possible at all with reasonable effort.</P> <P>I already did the same thing using NVIDIA's OptiX library. While the implementation itself was relatively easy thanks to OptiX's support for <STRONG>arbitrary geometry in the BVH</STRONG> and fully <STRONG>programmable ray-surface intersection</STRONG> tests, the resulting performance was rather suboptimal (solving several linear systems per ray stresses the GPU's memory bottleneck).</P> <P>Is Embree suitable for (iteratively refined) polynomial surface definitions / intersections instead of triangle meshes?</P> <P>Thank you very much!</P> <P>Regards, Tim</P> Sat, 04 May 2013 18:00:22 GMT https://community.intel.com/t5/Intel-Embree-Ray-Tracing-Kernels/MLS-Point-Set-Surfaces/m-p/948932#M226 tb3 2013-05-04T18:00:22Z https://community.intel.com/t5/Intel-Embree-Ray-Tracing-Kernels/MLS-Point-Set-Surfaces/m-p/948933#M227 <P>The current version of Embree does not support user defined primitives directly. However, you can of course modify the code to let the ray/primitive intersection do what you need.</P> Mon, 06 May 2013 06:05:00 GMT https://community.intel.com/t5/Intel-Embree-Ray-Tracing-Kernels/MLS-Point-Set-Surfaces/m-p/948933#M227 SvenW_Intel 2013-05-06T06:05:00Z