mersenneforum.org Very Prime Riesel and Sierpinski k
 Register FAQ Search Today's Posts Mark Forums Read

 2009-09-29, 08:49 #34 robert44444uk     Jun 2003 Oxford, UK 25·32·7 Posts Robert Your program seems to work quite well, it certainly did not fall down on my machine. I will have to play with the variables to see how to maximise the response, but certainly 11 VPS in a couple of days compares rather favourably to my 2-3 a week, which is the best I could manage recently. None of your Sierpinski VPS coincide with my table which is curious, you might have thought that at least one might have been picked up as the overall approach is not totally different and the range checked looks the same. Whjen you mention in your instructions that an iteration covers 3*10^12 K values, do you mean 3*10^12 payam numbers or 3*10^12 y values? as K is defined as y*E(x).
 2009-09-29, 11:15 #35 robert44444uk     Jun 2003 Oxford, UK 25·32·7 Posts I used the athlon exe file on a beat up old athlon and it is eating numbers!!!! Reserving 82 Sierpinski and Riesel
2009-09-29, 12:12   #36
robert44444uk

Jun 2003
Oxford, UK

25×32×7 Posts

Quote:
 Originally Posted by R. Gerbicz The in.txt file is a little more complicated, there are some obvious parameters. Note that I'm computing also the Nash weight for the sequence, because by this I can predict the number of remaining primes, I'm using a weaker form (not to lost a solution): #(total number of primes found so far)+c0*#(expected number of primes from Nash)+c1.
By this can I infer that a number that would be eliminated by "Smith check" because of insufficient prime count at a given n, will still be calculated at higher n if the Nash weight predicts it should reach a higher "Smith check" hurdle?

2009-09-29, 15:43   #37
R. Gerbicz

"Robert Gerbicz"
Oct 2005
Hungary

2·32·5·17 Posts

Quote:
 Originally Posted by robert44444uk None of your Sierpinski VPS coincide with my table which is curious, you might have thought that at least one might have been picked up as the overall approach is not totally different and the range checked looks the same. Whjen you mention in your instructions that an iteration covers 3*10^12 K values, do you mean 3*10^12 payam numbers or 3*10^12 y values? as K is defined as y*E(x).
In fact I've deleted the known numbers when submitted my post (using this forum and the two excel tables on your site), I've rediscovered on both sides some vps numbers, altogether there were about 5-6 such values, which is not very bad, indicates that the nash check is not very bad.

I'm checking about 3*10^12 y values at once, as you write in your paper only a fraction of them are payam numbers, thanks to a clever sieve I quickly eliminate many of them. In fact as I've written this is not exactly 3*10^12, the correct value is 2869549272527.

ps. Now I see my original post writing 3*10^12 K values, that was a typo.

Last fiddled with by R. Gerbicz on 2009-09-29 at 15:58

2009-09-29, 15:48   #38
R. Gerbicz

"Robert Gerbicz"
Oct 2005
Hungary

27728 Posts

Quote:
 Originally Posted by robert44444uk By this can I infer that a number that would be eliminated by "Smith check" because of insufficient prime count at a given n, will still be calculated at higher n if the Nash weight predicts it should reach a higher "Smith check" hurdle?
No, these checks are independent. So it means that if both of the checks are enabled and only one of them or both of them indicates that we will not reach 100 primes then the code immediately breaks the test on that payam number.

2009-09-29, 16:08   #39
robert44444uk

Jun 2003
Oxford, UK

7E016 Posts

Quote:
 Originally Posted by R. Gerbicz No, these checks are independent. So it means that if both of the checks are enabled and only one of them or both of them indicates that we will not reach 100 primes then the code immediately breaks the test on that payam number.
Intuitively it seems it would be better if one of them indicates that will reach 100/10000 then that prime continues. But intuition does not match up against pure logic, so I must accept your logic.

I also reserve 66 Riesel and 100 Riesel as I experiment.

I am so impressed with the performance of this software! This generates and checks payams faster by 500% than Axn1's software generated payams with no prime checking on the old machine I have. This may be due to the fact that it is also Athlon optimised. I am a happy puppy right now.

The software is so fast that I am generating 74/3000 as my target rather than 100/10000 because I can check those at my leisure.

2009-09-29, 16:10   #40
robert44444uk

Jun 2003
Oxford, UK

25·32·7 Posts

Quote:
 Originally Posted by R. Gerbicz In fact I've deleted the known numbers when submitted my post (using this forum and the two excel tables on your site), I've rediscovered on both sides some vps numbers, altogether there were about 5-6 such values, which is not very bad, indicates that the nash check is not very bad.
Then your rate of discovery is even greater!!!!

 2009-09-29, 19:28 #41 Dougal     Jan 2009 Ireland BA16 Posts my pentium version keeps crashing.and i cant figure out what to put in the in.txt file.
2009-09-29, 19:48   #42
R. Gerbicz

"Robert Gerbicz"
Oct 2005
Hungary

5FA16 Posts

Quote:
 Originally Posted by Dougal my pentium version keeps crashing.and i cant figure out what to put in the in.txt file.
You need also progress.txt file, download that also. (The exe and the two text files should be in the same folder).

If you change nothing then the program should work (it will test E=52), if you want to test other E values then E+1 should be such a prime for that 2 is a primitive root modulo E+1, so for example the good E values up to 140 are: 52,58,60,66,82,100,106,130,138. (For E<52 the program not works.)
If you want to start the computation from the beginning then not change iteration and I, so only modify E in progress.txt. For more information see my post above or ask.

ps.
I can imagine that something is bad with the exe, now I am at an Athlon computer, tomorrow I will see the exe on a Pentium, and if that is still bad then compile the code on that (I've installed gmp on that computer also.).

Last fiddled with by R. Gerbicz on 2009-09-29 at 19:50

 2009-09-30, 01:20 #43 robert44444uk     Jun 2003 Oxford, UK 25·32·7 Posts Athlon and Pentium programs ran all night successfully on three machines and four cores. No problems. I always worry about screensavers, but this did not appear to impact performance.
 2009-10-03, 06:47 #44 robert44444uk     Jun 2003 Oxford, UK 25×32×7 Posts Also reserving Sierpinski 60. Stupendous progress, as of now I generated 26 new values and a further 14 duplicates, in the 4 areas I reserved. No record breakers though yet. Thats 40 in 6 days, compared to the 2-3 a week before. I will post at a convenient time. The software has not fallen down at all. Robert G: Is there a way you could capture a racing table as well, that shows fastest performers to each prime level, each time the software is run? If this would slow things down then please do not change a thing! The softest target is the Riesel record to 100 primes which is only at n=5748. I have found 6 Sierpinskis with better records than that.

 Similar Threads Thread Thread Starter Forum Replies Last Post sweety439 sweety439 11 2020-09-23 01:42 sweety439 Conjectures 'R Us 0 2016-12-07 15:01 robert44444uk Conjectures 'R Us 139 2007-12-17 05:17 rogue Conjectures 'R Us 11 2007-12-17 05:08 michaf Conjectures 'R Us 49 2007-12-17 05:03

All times are UTC. The time now is 09:58.

Mon Jan 17 09:58:32 UTC 2022 up 178 days, 4:27, 0 users, load averages: 0.95, 0.85, 0.90