Patch in datashift2 addresses the case where no spikes are detected in the last few batches of data. When this is the case, the # of motion estimate batches will be less than the data batches. Since the shifted data is written in-place with no overlaps, I chose to omit those batches altogether. Not sure if you think it's cleaner to expand the dshift matrix with identity shifts? @oliche
Patch in
datashift2
addresses the case where no spikes are detected in the last few batches of data. When this is the case, the # of motion estimate batches will be less than the data batches. Since the shifted data is written in-place with no overlaps, I chose to omit those batches altogether. Not sure if you think it's cleaner to expand thedshift
matrix with identity shifts? @oliche