mersenneforum.org > Data Thinking out loud about getting under 20M unfactored exponents
 User Name Remember Me? Password
 Register FAQ Search Today's Posts Mark Forums Read

2021-07-22, 21:25   #584
SethTro

"Seth"
Apr 2019

36210 Posts

Quote:
 Originally Posted by petrw1 Yes, it depends a lot on how deep the P1 is; ...
I was previously working on a manyfactors project so TF would find factors at ~1/bitlevel they would just already be known. I was still thinking in that context and forgot all these candidates have the prior that P-1 didn't find any factors.

I checked the report again and many of these have reasonable B1/B2 (e.g. 180000/3375000).

using 1/90: 6% of the time I should have found zero or one factor.
using 1/100: 8% of the time I should have found zero or one factor.

Bad luck but not even a natural 1 on a d20.

 2021-07-24, 03:56 #585 petrw1 1976 Toyota Corona years forever!     "Wayne" Nov 2006 Saskatchewan, Canada 7·677 Posts 2021-07-24 Update - 4 years since I gathered the first stats It is 4 years since I gathered my first stats for this project. The numbers below (60ish%) suggest we are quite possibly more than half done. That is considering there are a lot of tough ranges left BUT hardware is getting faster and interest is growing. CHARGE!!!! 26 more ranges cleared: 3.4, 3.7, 3.9, 4.9, 7.6, 9.0, 9.6, 10.0, 10.1, 15.7, 18.3, 19.0, 20.0, 21.5, 23.0, 23.5, 26.7, 27.9, 28.8, 29.2, 29.5, 32.4, 35.1, 39.6, 42.6, 48.4 TOTALS to date: 293 total ranges cleared or 58.95% (204 ranges remaining) 2,899 more factored (33,630)....60.89% total factored. A new set of stats: ... I hope I can make these clear: The first column of numbers is as of the start of this project (2017-07-24) The second column of numbers is as of today. The first two rows will be addressed via this project. The last two rows will be addressed by Prime95 with no help required here. --- REPEAT: This project will NOT need to help with these 2 rows Code: Category Start Today Lowest ToDo 1.8 4.1 (The lowest range still over 1999) Highest ToDo 86.3 49.6 (The highest remaining for this project ... all higher will be done via standard Prime95 protocols) Low UnDone 86.4 105.2 (The lowest still over 1999 that Prime95 will eventually clear on its own) High UnDone 999.8 999.8 (The highest remaining for Prime95) My current activity/status: There is only 1 ranges remaining in 4xM. 49.6. It has been very aggressively P1'd. Completion will required 2 more bits of TF; to 76.... or P+1. Currently, Anton Repko has been dabbling here. There are only 8 rows remaining in 3xM. I am currently focusing all my CPUs doing P1 there. I am also receiving help from others with P1 and some TF. I expect to complete P1 near year end 2021. Then these 8 ranges will require TF to 75. At that point there may be a couple ranges that required a tad more P1 or TF. I am using my GPUs for TF in the closest 2xM ranges. Several others are doing TF of P1 there. There are some concerted efforts in the 1xM and 0xM ranges too. Thanks all I see light at the end of the tunnel. Where can you best help? If you have a GPU: TF 49.6 to 76 bits To avoid toe stepping the two following will best work if you use GPU72 as some are already working here: TF the remaining 3xM to 75 bits. TF all of 2xM to 73 bits. If you have a good CPU with ample RAM a LOT more P1 is required in all the remaining ranges. Best to note your intentions in this thread to again avoid toe stepping. Thanks again
2021-07-24, 06:07   #586
LaurV
Romulan Interpreter

Jun 2011
Thailand

975310 Posts

Quote:
 Originally Posted by petrw1 It is 4 years since I gathered my first stats for this project.
Wow, so much time wasted! You should have mobilized your resources at front PRP and find two new mersenne primes in 4 years...

Last fiddled with by LaurV on 2021-07-24 at 06:08

2021-07-24, 14:06   #587
petrw1
1976 Toyota Corona years forever!

"Wayne"
Nov 2006
Saskatchewan, Canada

7·677 Posts

Quote:
 Originally Posted by LaurV Wow, so much time wasted! You should have mobilized your resources at front PRP and find two new mersenne primes in 4 years...
Seems good in theory until I calculate that Ben Delo does as much PRP in 3 days that my farm did in those 4 years and he hasn't quite found 2 MPs yet...

 2021-07-24, 15:21 #588 Viliam Furik   "Viliam Furík" Jul 2018 Martin, Slovakia 24·41 Posts I will TF the 49.6 range to 75 bits, then I will continue based on how many factors are found - to help with the LaurV's sub-200 project... Later I would like to do 11M to 70 bits, then 12M, 13M, and 14M. When I get to it, I will ask which ones are available.
 2021-07-24, 16:04 #589 LaurV Romulan Interpreter     Jun 2011 Thailand 3·3,251 Posts I finished 11.9M few days ago, and found 81 factors from the 95 wanted. Therefore there are still 2014 candidates there, and 15 more factors to be found. I will come back to it later. I am still doing some P+1 there, which is terrible inefficient - it didn't find yet a factor, albeit 28 cores invested for about 3 days and another 10 cores for a week after - but I hope for a big P+1 hit, so I still let the 10 cores run there for a while. During the work there, I repeatedly adjusted B1 and B2 (mostly increasing them) till the point where the number of factors per a fixed number of candidates was "convenient", but I still could not catch up, so the missing 15 factors remained hidden. Interesting, in the upper 11.9M I found more factors (when B1 and B2 were smaller) and then in the lower 11.9M (after increasing and increasing the bounds) I found less factors. Yet, 81 factors was not a piece of cake task. I will come back to this range later with some more P-1 or some TF, whatever will fit appropriate at the time (hopefully, I will learn some more by doing the rest of the 11M ranges). Now, 11.9M was mostly factored to 70 bits, with few exceptions which were at 69, and negligible amount of expos which are higher, 71+ bitlevel. As the P-1 progressed into the 11.8M range, with the B1 and B2 fixed by the former 11.9 range, I was surprised to discover that I am finding too many factors, The reason is the fact that 11.8M (and lower ranges too) are factored only to 69 bits (and with few exception higher, 70+). In this case, the difference in probability for P-1 is about 1% (including the TF and low-P-1 done), so I expect to find about 20 factors more than needed for the 2093 exponents that I started with. But this wisdom only came to me later, after seeing the results. This was not considered when I calculated the limits (read "choose", I didn't do much of calculation, as I said, I was just increasing the limits because I was founding too less factors in the lower side of the 11.9M). So, in summary, I went to 40% of the 11.8M and found already 50% of the 94 needed factors (i.e. 47 factors). Meantime I progressed a bit further, close to 46% of the range, and having 56 factors found (38 to go). I will, anyhow, continue downward towards 11800000, find the (about) 110 factors (94 needed totally) and then the plan is to skip 11.7M because there are a lot more factors needed there, go directly to .6, .4, .1, .3, .2, .0, in order, and do the .7 at the end (in order of needed factors, so I get more experience and "feeling" with those boundaries, and increase them a little for each range if needed). Edit: currently I went through 47% of the 11.8M range and found 58 factors, two factors popped up as I was writing this post, so only 36 more needed while 52% of the range is "to go". I don't thing this range will need any more TF or other work after, even if I get into the "unlucky" side of the force . (hopefully didn't jinx it, hehe) Last fiddled with by LaurV on 2021-07-24 at 16:30
 2021-07-24, 18:33 #590 lycorn     "GIMFS" Sep 2002 Oeiras, Portugal 27238 Posts ... or, most likely, 4 MP in 2 years.
 2021-07-24, 19:00 #591 De Wandelaar     "Yves" Jul 2017 Belgium 22·17 Posts 30.5 M : 2 factors found during the first trials and thereafter more than 330 exponents processed without one single factor . Probability of 330 unsuccessfull trials seems to be about about 1.2 % : (74/75) 330. I've been shutting down and rebooting the computer, testing successfully (factor found) an exponent with a known factor. Should I nevertheless suspect a problem with the GPU/system/... before shutting down/rebooting and redo the TF work since the last found factor ? Yves
2021-07-24, 21:14   #592
Viliam Furik

"Viliam Furík"
Jul 2018
Martin, Slovakia

24·41 Posts

Quote:
 Originally Posted by De Wandelaar Should I nevertheless suspect a problem with the GPU/system/... before shutting down/rebooting and redo the TF work since the last found factor ?
Even an infinitesimally small probability doesn't mean something can't happen. It just gives a perspective of how often it happens.

I've had many such long runs without a factor, most of them when doing TF on <10G exponents. Giving your numbers into a perspective:

The event of having 330 (assuming consecutive) no factor events occurs with probability of about the mentioned 1.19%. That means that every 1/0,0119 = 84th of such runs will be factor-free. For that to happen once, one only needs to test about 27686 exponents. (I know that's not exactly how it works, but I think it's good enough to prove a point.)

So I think you shouldn't waste time rechecking 330 exponents for what is maybe one factor with not exactly high probability. You are better off using that time to check the next 330 exponents, which were not checked before. That's a lot more efficient with regard to probability and worth of the outcome.

2021-07-25, 02:55   #593
LaurV
Romulan Interpreter

Jun 2011
Thailand

100110000110012 Posts

Quote:
 Originally Posted by Viliam Furik So I think you shouldn't waste time rechecking 330 exponents for what is maybe one factor with not exactly high probability.
+1. I doesn't worth rechecking, even if you missed one or two factors (which has low probability anyhow). The total effort (for two runs) would be too large to justify the gain. We all had long empty strikes.

On a better note, we passed 50% of 11.8M (51% done) and fond 61 factors from the 94 needed. There is 49% of the range to go and 33 factors more to find, so this range will be for sure an over-achiever. We will keep the bound for now (lazy to change), and the order was changed to be .8, .4, .1, .6, .3, .2, .0, .7 for the future, in all worktodo files.

Last fiddled with by LaurV on 2021-07-25 at 03:02

 2021-07-25, 10:44 #594 De Wandelaar     "Yves" Jul 2017 Belgium 22×17 Posts Thanks Villiam & LaurV, I'll do as you propose. Last fiddled with by De Wandelaar on 2021-07-25 at 10:47

 Thread Tools

 Similar Threads Thread Thread Starter Forum Replies Last Post jschwar313 GPU to 72 3 2016-01-31 00:50 Batalov Factoring 6 2011-12-27 22:40 jasong jasong 1 2008-11-11 09:43 devarajkandadai Math 4 2007-07-25 03:01 WraithX GMP-ECM 1 2006-03-19 22:16

All times are UTC. The time now is 02:59.

Sat Sep 25 02:59:00 UTC 2021 up 63 days, 21:27, 0 users, load averages: 2.28, 2.60, 3.11

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.