Mismatch in output of nearest neighbor calculation due to FP precision
Here's some commented out code that tests CUDA code vs CPU code for nearest neighbors. Due to the floating point precision, there can be a slight mismatch in the two output: https://gitlab.osti.gov/mtip/spinifel/-/blob/chy/commGroup/spinifel/sequential/nearest_neighbor.py#L115
It would be great if you (Vinay and Christine) could address this and move this to a unit test GitLabGitLab spinifel/sequential/nearest_neighbor.py · chy/commGroup · MTIP / spinifel Applies the M-TIP algorithm to SPI data.