Support newMSM
effigies opened this issue · comments
Follow-up to #353. Our initial tests with newMSM show that it does not produce bit-for-bit identical results with the existing MSM (which appears to be 100% reproducible, given the same inputs). We have not quantified the MSM-to-newMSM differences or the run-to-run reliability of newMSM.
newMSM does take advantage of multi-threading, so it could be a significant speed-up over the pretty standard 40 minute registration of MSM.
cc @edickie This was what I wanted to ask you about on the nipreps call. Have you looked into it at all?