500M relations are not enough for L1438 (only 294397430 unique out of 490709259); sieving more to get 600M relations. The duplication rate appears lower for L5895B (328530931 unique out of 497451947)
(5/1: L5895B with 497M raw relations gives 52.16M matrix at density 70; doing more sieving to aim for 550M relations)
It would be really good if we could figure out some way of opening up a 16e queue for user submissions in 2019 - 16e gets much better yields and duplicate rates above GNFS190. GNFS200 is entirely practical with 16e, the linear algebra takes me a month, but sieving GNFS200 on my personal resources takes a season.
Taking L4245A in the meantime; ETA afternoon of 9 Jan
Last fiddled with by fivemack on 2019-01-06 at 14:18
|