On 1/10/10 9:36 PM, Frank A. Stevenson wrote: > On Sun, 2010-01-10 at 17:15 +0100, M vd S wrote: > > >> Here you go: >> >> 16 back: 57% illegal states >> 32 back: 68% illegal states >> 48 back: 74% illegal states >> 64 back: 78% illegal states >> 80 back: 82% illegal states >> 96 back: 84% illegal states >> >> These numbers are based on one chain. (de001bc0006f0000 [0] -> >> 36dc483fb4fe0000) >> >> I verified superficially with chains >> de001bc034563200 [0] -> d59ebd1df40d8000 >> f1234678e5f7d600 [0] -> 61df56b387178000 >> >> and both give the same pattern. Clocking back 150 times (the maximum in >> practice) gives 89% illegal states. >> >> > > I think it is should be compared with what you get from "purely random" > states, it could be that our round function has a propensity for > generating invalid states, and that would be a major flaw in our > approach. (which advance value are you using?) > >
This is based on random states (or as random as rand() with the proper shifts to get some 64 random bits will get) I tested every listed number of backclockings 10 times on 100000 random states and calculated average and stddev. The count is the number of illegal states. Nback 16 32 48 64 80 96 150 sample0 58001 68167 74857 78955 82275 84354 89039 sample1 57809 68260 74744 78995 82037 84279 89039 sample2 58029 68212 74667 78767 82173 84525 89307 sample3 57799 68346 74557 78992 81982 84189 89033 sample4 58113 68234 74818 78803 82046 84605 88926 sample5 57995 68290 74516 78997 81868 84544 88991 sample6 58096 68228 74966 78988 82251 84416 88953 sample7 57975 68150 74848 79162 81934 84474 88884 sample8 57874 68208 74884 79011 82355 84399 88998 sample9 57827 68327 74533 79020 81951 84469 89152 avg 57951.8 68242.2 74739 78969 82087.2 84425.4 89032.2 stddev 116.8720098 64.18168655 161.6230182 111.8133365 165.5064953 126.36209 121.2626168 pct 58.0% 68.2% 74.7% 79.0% 82.1% 84.4% 89.0% So that's almost identical. M. _______________________________________________ A51 mailing list [email protected] http://lists.lists.reflextor.com/cgi-bin/mailman/listinfo/a51
