By ensuring randstrobes are sorted by all their fields.
I thought that the comparison function I added in #386 (commit 2e4ff9500e68d6e465735dd276d362cf71851dcd) was good enough, but now that I ran on some other datasets that I hadn’t used for testing previously, it became apparent that that’s not the case because I was getting non-reproducible results.
This slows down index creation again a little bit (now takes ~61s instead of ~60s for CHM13), but it’s still a lot faster than before parallel sorting.
By ensuring randstrobes are sorted by all their fields.
I thought that the comparison function I added in #386 (commit 2e4ff9500e68d6e465735dd276d362cf71851dcd) was good enough, but now that I ran on some other datasets that I hadn’t used for testing previously, it became apparent that that’s not the case because I was getting non-reproducible results.
This slows down index creation again a little bit (now takes ~61s instead of ~60s for CHM13), but it’s still a lot faster than before parallel sorting.