These are the results: with pyproj (without projection cache): 4.0 - 4.5 sec with pyproj (with projection cache): 0.9 - 1.0 sec with utm: 0.4 - 0.5 sec NumPy arrays bring another speed improvement (on ...

Matrix multiplication on GPUs for matrices stored on a CPU. Similar to cublasXt, but ported to both NVIDIA and AMD GPUs.