Two months ago, I asked about GPU programming for PerfectTIN. I now have the routine, which I'd like implemented in GPU, working. It takes a block of dots, all in one triangle, computes the difference in elevation between the dots and the triangle, computes how much moving the corners of the triangle up or down would affect the previous figure, does some matrix arithmetic, and produces a small square matrix and a small column vector as results. The code is in adjelev.cpp in https://github.com/phma/perfecttin .
I am about to change the sequence of block sizes. Among other things, it will make most of the block sizes multiples of 256, which should avoid waste when using the GPU to compute. Pierre -- gau do li'i co'e kei do _______________________________________________ Discuss mailing list Discuss@lists.osgeo.org https://lists.osgeo.org/mailman/listinfo/discuss
