LLR Version 3.8.16 released - Page 2

pepi37 · 2015-08-06, 20:19

./llr64 -d -q"51956602*63^19370-1"
Base factorized as : 3^2*7
Base prime factor(s) taken : 3, 7
Starting Lucas sequence for 51956602*63^19370-1...
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 10
51956602*63^19370-1 may be prime, trying to compute gcd's
U((N+1)/7) is coprime to N!
51956602*63^19370-1 may be prime, but N divides U((N+1)/3), P = 10
Restarting Lucas sequence with P = 14
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 14
51956602*63^19370-1 may be prime, trying to compute gcd's
51956602*63^19370-1 may be prime, but N divides U((N+1)/3), P = 14
Restarting Lucas sequence with P = 15
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 15
51956602*63^19370-1 may be prime, trying to compute gcd's
51956602*63^19370-1 may be prime, but N divides U((N+1)/3), P = 15
Restarting Lucas sequence with P = 16
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 16
51956602*63^19370-1 may be prime, trying to compute gcd's
51956602*63^19370-1 may be prime, but N divides U((N+1)/3), P = 16
Restarting Lucas sequence with P = 17
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 17
51956602*63^19370-1 may be prime, trying to compute gcd's
51956602*63^19370-1 may be prime, but N divides U((N+1)/3), P = 17
Restarting Lucas sequence with P = 18
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 18
51956602*63^19370-1 may be prime, trying to compute gcd's
51956602*63^19370-1 may be prime, but N divides U((N+1)/3), P = 18
Restarting Lucas sequence with P = 19
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 19
51956602*63^19370-1 may be prime, trying to compute gcd's
51956602*63^19370-1 may be prime, but N divides U((N+1)/3), P = 19
Giving up after 11 restarts...
Time : 141.817 sec.
This is with updated LLR 3.8.16
Why llr cannot confirm this is prime or not?

Also I do test on AVX and FMA3 test of this prime http://primes.utm.edu/primes/page.ph...18946#comments
Sergei told there is mixed bag of results...

wombatman · 2015-08-06, 20:39

LLR v 3.8.13 does the same.

Code:

Base factorized as : 3^2*7
Base prime factor(s) taken : 3, 7
Starting N+1 prime test of 51956602*63^19370-1
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, a = 3
51956602*63^19370-1 may be prime. Starting Lucas sequence...
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 3
51956602*63^19370-1 may be prime, trying to compute gcd's
U((N+1)/7) is coprime to N!
Restarting Lucas sequence with P = 4
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 4
51956602*63^19370-1 may be prime, trying to compute gcd's
Restarting Lucas sequence with P = 7
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 7
51956602*63^19370-1 may be prime, trying to compute gcd's
Restarting Lucas sequence with P = 9
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 9
51956602*63^19370-1 may be prime, trying to compute gcd's
Restarting Lucas sequence with P = 10
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 10
51956602*63^19370-1 may be prime, trying to compute gcd's
Restarting Lucas sequence with P = 14
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 14
51956602*63^19370-1 may be prime, trying to compute gcd's
Restarting Lucas sequence with P = 15
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 15
51956602*63^19370-1 may be prime, trying to compute gcd's
Restarting Lucas sequence with P = 16
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 16
51956602*63^19370-1 may be prime, trying to compute gcd's
Restarting Lucas sequence with P = 17
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 17
51956602*63^19370-1 may be prime, trying to compute gcd's
Restarting Lucas sequence with P = 18
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 18
51956602*63^19370-1 may be prime, trying to compute gcd's
Restarting Lucas sequence with P = 19
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 19
51956602*63^19370-1 may be prime, trying to compute gcd's
Giving up after 11 restarts...
  Time : 393.591 sec.

rebirther · 2015-08-06, 23:33

Quote:

Originally Posted by pepi37

./llr64 -d -q"51956602*63^19370-1"
Base factorized as : 3^2*7
Base prime factor(s) taken : 3, 7
Starting Lucas sequence for 51956602*63^19370-1...
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 10
51956602*63^19370-1 may be prime, trying to compute gcd's
U((N+1)/7) is coprime to N!
51956602*63^19370-1 may be prime, but N divides U((N+1)/3), P = 10
Restarting Lucas sequence with P = 14
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 14
51956602*63^19370-1 may be prime, trying to compute gcd's
51956602*63^19370-1 may be prime, but N divides U((N+1)/3), P = 14
Restarting Lucas sequence with P = 15
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 15
51956602*63^19370-1 may be prime, trying to compute gcd's
51956602*63^19370-1 may be prime, but N divides U((N+1)/3), P = 15
Restarting Lucas sequence with P = 16
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 16
51956602*63^19370-1 may be prime, trying to compute gcd's
51956602*63^19370-1 may be prime, but N divides U((N+1)/3), P = 16
Restarting Lucas sequence with P = 17
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 17
51956602*63^19370-1 may be prime, trying to compute gcd's
51956602*63^19370-1 may be prime, but N divides U((N+1)/3), P = 17
Restarting Lucas sequence with P = 18
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 18
51956602*63^19370-1 may be prime, trying to compute gcd's
51956602*63^19370-1 may be prime, but N divides U((N+1)/3), P = 18
Restarting Lucas sequence with P = 19
Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 19
51956602*63^19370-1 may be prime, trying to compute gcd's
51956602*63^19370-1 may be prime, but N divides U((N+1)/3), P = 19
Giving up after 11 restarts...
Time : 141.817 sec.
This is with updated LLR 3.8.16
Why llr cannot confirm this is prime or not?

Also I do test on AVX and FMA3 test of this prime http://primes.utm.edu/primes/page.ph...18946#comments
Sergei told there is mixed bag of results...

Jean gave me this option last time for this situation:

-oMaxrestarts=20

Increase it to 40, this could be enough.

IBethune · 2015-08-07, 08:04

Quote:

Originally Posted by pepi37

./llr64 -d -q"51956602*63^19370-1"
...
Giving up after 11 restarts...
Time : 141.817 sec.
This is with updated LLR 3.8.16
Why llr cannot confirm this is prime or not?

Also I do test on AVX and FMA3 test of this prime http://primes.utm.edu/primes/page.ph...18946#comments
Sergei told there is mixed bag of results...

Quote:

Originally Posted by rebirther

Jean gave me this option last time for this situation:

-oMaxrestarts=20

Increase it to 40, this could be enough.

Re: 51956602*63^19370-1 - this number is prime, and can be proven with both LLR 3.8.15 and 3.8.16 - this is not a bug, but rather for an N+1 test as well as showing a^(n+1) == 1 (mod n), there is also a divisibility test (see eq 4 in http://www.curtisbright.com/bln/2013...imality-tests/). For this candidate, lots of values of a (called P by LLr) fail this test, so it must make repeated attempts to find an a that satisfies both equations.

Re: 1024*3^1877301+1, which was discussed in http://primes.utm.edu/primes/page.ph...18946#comments - this is prime, and LLR 3.8.16 confirms it is prime. The uncertainty was due to a bug in gwnum which was fixed in the previous release.

- Iain

Batalov · 2015-08-07, 16:57

It reminds me to ping Chris who was waiting for the software to mature to replace his UTM server side LLR and PFGW and then trigger the standard re-processing and have the 'External' tag removed. I'll email him that LLR 3.8.16 is what he wants to use from now on. As well as the latest PFGW.

LaurV · 2015-08-08, 03:53

Quote:

Originally Posted by Batalov

LLR 3.8.16 is what he wants to use from now on. As well as the latest PFGW.

Is this valid for us (small CRUS hunters) too? i.e should I get a new cllr/pfgw too? (the old one seems to be ok and I usually don't crunch over n=200-300k or so).

(edit: I didn't follow all technical discussion but I have learned to pay attention to your posts

)

rebirther · 2015-08-13, 15:39

I have found some errors in results:

Number sent to gwsetup is too large for the FFTs to handle.

Happened on a GenuineIntel Intel(R) Celeron(R) D CPU 3.46GHz [Family 15 Model 6 Stepping 5]

After 5 sec it has finished the WU.

rogue · 2015-08-13, 16:12

Quote:

Originally Posted by rebirther

I have found some errors in results:

Number sent to gwsetup is too large for the FFTs to handle.

Happened on a GenuineIntel Intel(R) Celeron(R) D CPU 3.46GHz [Family 15 Model 6 Stepping 5]

After 5 sec it has finished the WU.

What was the number that this failed on?

rebirther · 2015-08-13, 16:26

Quote:

Originally Posted by rogue

What was the number that this failed on?

Different, some S6 and 1 R430:

34910*430^153946-1
112783*6^1607505+1

Jean Penné · 2015-08-13, 20:05

Hi,

Thanks to Rebirther, the two Win64 binaries are now also updated accordingly to the minor change I made on the 06/08/15.

Regards,
Jean

pepi37 · 2015-08-13, 20:35

Quote:

Originally Posted by Jean Penné

Hi,

Thanks to Rebirther, the two Win64 binaries are now also updated accordingly to the minor change I made on the 06/08/15.

Regards,
Jean

Linux binaries on your page is newest?

2015-08-06, 20:19	#12
pepi37 Dec 2011 After milion nines:) 10110000000₂ Posts	./llr64 -d -q"5195660263^19370-1" Base factorized as : 3^27 Base prime factor(s) taken : 3, 7 Starting Lucas sequence for 5195660263^19370-1... Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 10 5195660263^19370-1 may be prime, trying to compute gcd's U((N+1)/7) is coprime to N! 5195660263^19370-1 may be prime, but N divides U((N+1)/3), P = 10 Restarting Lucas sequence with P = 14 Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 14 5195660263^19370-1 may be prime, trying to compute gcd's 5195660263^19370-1 may be prime, but N divides U((N+1)/3), P = 14 Restarting Lucas sequence with P = 15 Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 15 5195660263^19370-1 may be prime, trying to compute gcd's 5195660263^19370-1 may be prime, but N divides U((N+1)/3), P = 15 Restarting Lucas sequence with P = 16 Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 16 5195660263^19370-1 may be prime, trying to compute gcd's 5195660263^19370-1 may be prime, but N divides U((N+1)/3), P = 16 Restarting Lucas sequence with P = 17 Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 17 5195660263^19370-1 may be prime, trying to compute gcd's 5195660263^19370-1 may be prime, but N divides U((N+1)/3), P = 17 Restarting Lucas sequence with P = 18 Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 18 5195660263^19370-1 may be prime, trying to compute gcd's 5195660263^19370-1 may be prime, but N divides U((N+1)/3), P = 18 Restarting Lucas sequence with P = 19 Using zero-padded AVX FFT length 15K, Pass1=320, Pass2=48, P = 19 5195660263^19370-1 may be prime, trying to compute gcd's 5195660263^19370-1 may be prime, but N divides U((N+1)/3), P = 19 Giving up after 11 restarts... Time : 141.817 sec. This is with updated LLR 3.8.16 Why llr cannot confirm this is prime or not? Also I do test on AVX and FMA3 test of this prime http://primes.utm.edu/primes/page.ph...18946#comments Sergei told there is mixed bag of results... Last fiddled with by pepi37 on 2015-08-06 at 20:25*

2015-08-13, 20:05	#21
Jean Penné May 2004 FRANCE 571 Posts	Updated Win64 binaries now uploaded Hi, Thanks to Rebirther, the two Win64 binaries are now also updated accordingly to the minor change I made on the 06/08/15. Regards, Jean

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
LLR Version 3.8.20 released	Jean Penné	Software	30	2018-08-13 20:00
LLR Version 3.8.19 released	Jean Penné	Software	11	2017-02-23 08:52
LLR Version 3.8.11 released	Jean Penné	Software	37	2014-01-29 16:32
LLR Version 3.8.9 released	Jean Penné	Software	37	2013-10-31 08:45
llr 3.8.2 released as dev-version	opyrt	Prime Sierpinski Project	11	2010-11-18 18:24

2015-08-07, 16:57	#16
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 3²×17×61 Posts	It reminds me to ping Chris who was waiting for the software to mature to replace his UTM server side LLR and PFGW and then trigger the standard re-processing and have the 'External' tag removed. I'll email him that LLR 3.8.16 is what he wants to use from now on. As well as the latest PFGW.

2015-08-13, 15:39	#18
rebirther Sep 2011 Germany 5²×109 Posts	I have found some errors in results: Number sent to gwsetup is too large for the FFTs to handle. Happened on a GenuineIntel Intel(R) Celeron(R) D CPU 3.46GHz [Family 15 Model 6 Stepping 5] After 5 sec it has finished the WU.