Issues: Different result in step 2 when using MPI version and CUDA version

[Anson-Jiaqi]

Hi Terastitcher developers,
We are currently trying to use CUDA for parallel operations in the second step. I followed your demo “ https://github.com/abria/TeraStitcher/wiki/Demo-and-batch-scripts ” to try it out and found that when I used the MPI version, it came out with the same results as when I used the sequential execution. But when I use CUDA, there are tons of differences. I would like to ask what is causing this?