If there is only a CPU without a GPU, you can try using Python nodes for large-scale operations. Using numpy can achieve a computation time of less than 5 seconds. (Note: Under my calculation conditions)
Java snippet can also achieve a computation time of only 5 seconds. I just cleaned up the code.
If using GPU and the environment matching is complex, the general idea is to try TensorFlow’s direct calculation.
Computing Efficiency.knar (60.8 KB)