Search for prime Gaussian-Mersenne norms (and G-M-cofactors) - Page 14

Citrix · 2014-09-05, 00:42

Serge,

I understand your point, but one of the reasons I like prime testing/distributed computing is to be as efficient I can. I like and value efficiency.

Why don't you go ahead with the range. I have stumbled across something very interesting that I am currently working on and won't have the CPU resources to work on GM search for the next few months to a whole year. I will join back once I have some free resources. I can pick up a smaller range in the 5M+ region, when I have some free computing resources.

Batalov · 2014-09-05, 01:00

Ok. I will start by GPU sieving de novo (I want to give gmqfaktc a good test).
We can compare the survivor candidates, later. We can compare some residues later, too. I will run the AVX binary on the survivors.

After that I will finish the 4.7-5M GQ-only PRP tests as well. Those are enormous. (Caldwell's double check with PFGW took 6.7 days, as far as I remember. but of course that was before they got a new, shiny, fast server for UTM.)

For this particular project, I was on a hiatus since April, too.

Batalov · 2014-09-06, 16:34

2^3704053 + 2^1852027 + 1 is prime

There is ~5 AVX CPU-years of work in this interval (or ~8 non-AVX years), and this was waiting in the middle of it.

Citrix · 2014-09-06, 22:05

Congratulations.

Batalov · 2014-09-06, 22:38

Thanks! ;-)

Granted, the expected number of GM+GQs was << 1 in this interval.

You mentioned that you had ~8 cores? With them, you could have reached this number, from end of Feb to mid- or end of summer. But aw well.

Citrix · 2014-09-07, 02:22

I did not do the work sequentially. I had done part of that range. Just a matter of luck.

Cruelty · 2014-09-07, 08:44

Congratulations

Batalov · 2014-09-08, 01:16

The 3.5-3.85M range is complete. The report is attached. 1 GM was found.

The J.Penne file <<gmfcandidates.txt>> had 1537 candidates removed out of 23172 prime numbers in this interval. For QC, these were double-checked and the tiny factors are included in an addendum(<<RES_GMQ_35_to32bit.txt>>). Each of the 23172 prime numbers in this interval is accounted for, on both GM and GQ sides. Pre-factoring was done using gmqfaktc-0.20 to 60 bits. All factors are validated with Pari/GP.

Light doublecheck for randomly chosen candidates (n=40) produced full agreement in the 80 RES64s with the initial run (of note: LLR implements random bit-shifts for tests of GM candidates, so these were independent just like within GIMPS). More double-checks are, of course, welcome.

Batalov · 2015-01-01, 00:29

I'll have a look at adapting the Cyclo program for Gaussian- (and Eisenstein-)Mersennes.
Then we may have a great boost to find the next sequence members on GPUs.
And we already have a GPU sieve.

Cruelty · 2015-01-02, 10:52

Quote:

Originally Posted by Batalov

I'll have a look at adapting the Cyclo program for Gaussian- (and Eisenstein-)Mersennes.
Then we may have a great boost to find the next sequence members on GPUs.
And we already have a GPU sieve.

By GPU you mean nVidia or is there a software for AMD too?

Batalov · 2015-01-02, 16:17

Well, I have modified mfaktc, but I have never looked at mfakto.
I don't know how different they are.

2014-09-06, 16:34	#146
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 2²·23·103 Posts	2^3704053 + 2^1852027 + 1 is prime There is ~5 AVX CPU-years of work in this interval (or ~8 non-AVX years), and this was waiting in the middle of it. Last fiddled with by Batalov on 2014-09-06 at 17:01 Reason: formatting

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
New PC dedicated to Mersenne Prime Search	Taiy	Hardware	12	2018-01-02 15:54
Gaussian integers- use of norms	devarajkandadai	Number Theory Discussion Group	11	2017-10-28 20:58
Low clock speeds on Mersenne Prime search	Ammonia	Hardware	2	2016-01-21 17:46
Testing Mersenne cofactors for primality?	CRGreathouse	Computer Science & Computational Number Theory	18	2013-06-08 19:12
Can I specify the range to search the Mersenne Prime?	Unregistered	Information & Answers	22	2012-03-20 11:38

2014-09-05, 00:42	#144
Citrix Jun 2003 2×7×113 Posts	Serge, I understand your point, but one of the reasons I like prime testing/distributed computing is to be as efficient I can. I like and value efficiency. Why don't you go ahead with the range. I have stumbled across something very interesting that I am currently working on and won't have the CPU resources to work on GM search for the next few months to a whole year. I will join back once I have some free resources. I can pick up a smaller range in the 5M+ region, when I have some free computing resources.

2014-09-05, 01:00	#145
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 2²·23·103 Posts	Ok. I will start by GPU sieving de novo (I want to give gmqfaktc a good test). We can compare the survivor candidates, later. We can compare some residues later, too. I will run the AVX binary on the survivors. After that I will finish the 4.7-5M GQ-only PRP tests as well. Those are enormous. (Caldwell's double check with PFGW took 6.7 days, as far as I remember. but of course that was before they got a new, shiny, fast server for UTM.) For this particular project, I was on a hiatus since April, too.

2014-09-06, 22:05	#147
Citrix Jun 2003 2×7×113 Posts	Congratulations.

2014-09-06, 22:38	#148
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 22404₈ Posts	Thanks! ;-) Granted, the expected number of GM+GQs was << 1 in this interval. You mentioned that you had ~8 cores? With them, you could have reached this number, from end of Feb to mid- or end of summer. But aw well.

2014-09-07, 02:22	#149
Citrix Jun 2003 2×7×113 Posts	I did not do the work sequentially. I had done part of that range. Just a matter of luck.

2014-09-07, 08:44	#150
Cruelty May 2005 2³·7·29 Posts	Congratulations

2015-01-01, 00:29	#152
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 2²·23·103 Posts	I'll have a look at adapting the Cyclo program for Gaussian- (and Eisenstein-)Mersennes. Then we may have a great boost to find the next sequence members on GPUs. And we already have a GPU sieve.

2015-01-02, 16:17	#154
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 2²×23×103 Posts	Well, I have modified mfaktc, but I have never looked at mfakto. I don't know how different they are.