The microblog: 2018.08.31 00:38:10

2018.08.31 00:38:10 (1035295642871054337) from Daniel J. Bernstein, replying to "Marc Stevens (@realhashbreaker)" (1035276289995956226):

Your comparison is bogus. The paper you're claiming to beat sacrificed two orders of magnitude in performance (Figure 1) to allow massive parallelization. Massive parallelization is critical for larger-scale attacks and exactly what _hasn't_ been shown to be possible for sieving.


2018.08.30 15:45:04 (1035161482521178112) from "Marc Stevens (@realhashbreaker)":

We've got new lattice SVP challenge records using sieving (joint work with @martinralbrecht @DucasLeo G.Herold E.Kirshanova @EamonnPostle):

2018.08.30 15:52:57 (1035163466766659584) from "Marc Stevens (@realhashbreaker)", replying to "Marc Stevens (@realhashbreaker)" (1035161482521178112):

We solved SVP dimension 151 in about 460 coredays. Compared to the old record for dim 150 using RSR that took about 180,000 coredays this is in the order of 400 times faster.

2018.08.30 23:21:16 (1035276289995956226) from "Marc Stevens (@realhashbreaker)", replying to "Marc Stevens (@realhashbreaker)" (1035163466766659584):

Exciting, because sieving is asymptotically best, but was not in practice: the best sieving record in HoF was only at dim 116, whereas BKZ+Enum best sol is dim 130 and RSR was the one to beat at dim 150. Now this shows that sieving is best in practice as well, by a large margin.