Difference in residue LLR4.0 vs LLR2 1.1.1

diep · 2021-10-16, 19:10

Tested at this internet box Xeon L5420

Different residues:

sllr64 version 4.0 gwnum 30.6 :
9473*2^1024968-1 is not prime. RES64: 20E2F460750CC947. Time : 1280.557 sec.
9473*2^1025002-1 is not prime. RES64: B76C5E01AC969C27. Time : 1233.439 sec.

From primegrid downloadpage llr2 1.1.1 gwnum 30.4 :
9473*2^1024968-1 is not prime. LLR Res64: A64A973DDFB5A0B4 Time : 1081.245 sec.
9473*2^1025002-1 is not prime. LLR Res64: 4AA2BFFF180ABA02 Time : 1081.488 sec.

Is it worrysome or no problem?

ATH · 2021-10-17, 03:19

Prime95 30.7b3
cllr.exe 4.0.0 32bit (Windows)
cllr.exe 3.8.24b3 32bit (Windows)
on an old 5960X all agree with your sllr64 4.0 results.

Code:

Prime95 30.7b3
{"status":"C", "k":9473, "b":2, "n":1024968, "c":-1, "worktype":"PRP-3", "res64":"20E2F460750CC947", "residue-type":1, "res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fft-length":81920, "error-code":"00000000", "security-code":"7D9A1A57", "program":{"name":"Prime95", "version":"30.7", "build":3, "port":4}, "timestamp":"2021-10-17 02:56:29", "errors":{"gerbicz":0}, "proof":{"version":2, "power":6, "hashsize":64, "md5":"be8c612f6bbdc78f13db9aad15be25c2"}, "user":"athath", "computer":"5960X"}
{"status":"C", "k":9473, "b":2, "n":1025002, "c":-1, "worktype":"PRP-3", "res64":"B76C5E01AC969C27", "residue-type":1, "res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fft-length":81920, "error-code":"00000000", "security-code":"7DDE1A9B", "program":{"name":"Prime95", "version":"30.7", "build":3, "port":4}, "timestamp":"2021-10-17 02:58:33", "errors":{"gerbicz":0}, "proof":{"version":2, "power":6, "hashsize":64, "md5":"184b9acc8d9391ba83b8962066a3c5a7"}, "user":"athath", "computer":"5960X"}

cllr.exe 4.0.0 32bit Windows:
cllr.exe -d -q"9473*2^1024968-1"
Starting Fermat PRP test of 9473*2^1024968-1
Using AVX FFT length 80K, Pass1=320, Pass2=256, clm=1, a = 3
9473*2^1024968-1 is not prime.  RES64: 20E2F460750CC947.  Time : 418.243 sec.

cllr.exe -d -q"9473*2^1025002-1"
Starting Fermat PRP test of 9473*2^1025002-1
Using AVX FFT length 80K, Pass1=320, Pass2=256, clm=1, a = 3
9473*2^1025002-1 is not prime.  RES64: B76C5E01AC969C27.  Time : 418.698 sec.

cllr.exe 3.8.24b3 32bit Windows:
cllr.exe -d -q"9473*2^1024968-1"
Starting Fermat PRP test of 9473*2^1024968-1
Using AVX FFT length 80K, Pass1=320, Pass2=256, clm=1, a = 3
9473*2^1024968-1 is not prime.  RES64: 20E2F460750CC947.  Time : 420.067 sec.

cllr.exe -d -q"9473*2^1025002-1"
Starting Fermat PRP test of 9473*2^1025002-1
Using AVX FFT length 80K, Pass1=320, Pass2=256, clm=1, a = 3
9473*2^1025002-1 is not prime.  RES64: B76C5E01AC969C27.  Time : 419.335 sec.

axn · 2021-10-17, 03:54

So the explanation is simple.Two different types of tests, two different residues.

4.0 is doing a PRP test. LLR2 is apparently doing an LLR test (shock! LLR doing an LLR test?!)

ATH · 2021-10-17, 12:54

I just assumed everything was PRP tests, I was not even thinking about LLR tests.

I'm not sure why LLR is not running LLR tests actually, it should when k<2ⁿ ? There is a ForcePRP=1 option, but I did not use it.

axn · 2021-10-17, 13:04

Starting from 3.8.something, PRP + GEC is available, so that is preferred due to the robust error checking.

ATH · 2021-10-17, 13:14

Quote:

Originally Posted by axn

Starting from 3.8.something, PRP + GEC is available, so that is preferred due to the robust error checking.

I tried old version 3.8.0 LLR.exe and cllr.exe both started LLR tests and the residues matched the llr2 1.1.1, so all the results are fine.

So if it runs PRP + GEC, and you find a PRP and wants to make sure it is prime. How do you force an LLR test?

Edit: Seems the forced PRP+GEC started between version 3.8.23 and 3.8.24b3.

henryzz · 2021-10-17, 13:28

Might be wrong but I think it will run an llr test after a prp result

diep · 2021-10-22, 22:46

Using LLR2 now which is 30.4 based.

What would be needed to put in inifile to enforce it use PRP test and use Gerbicz Error Check (GEC)?

2021-10-16, 19:10	#1
diep Sep 2006 The Netherlands 2×17×23 Posts	Difference in residue LLR4.0 vs LLR2 1.1.1 Tested at this internet box Xeon L5420 Different residues: sllr64 version 4.0 gwnum 30.6 : 94732^1024968-1 is not prime. RES64: 20E2F460750CC947. Time : 1280.557 sec. 94732^1025002-1 is not prime. RES64: B76C5E01AC969C27. Time : 1233.439 sec. From primegrid downloadpage llr2 1.1.1 gwnum 30.4 : 94732^1024968-1 is not prime. LLR Res64: A64A973DDFB5A0B4 Time : 1081.245 sec. 94732^1025002-1 is not prime. LLR Res64: 4AA2BFFF180ABA02 Time : 1081.488 sec. Is it worrysome or no problem?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Example of difference between my() and local()?	hansl	PARI/GP	3	2019-05-14 03:19
common difference 0	wildrabbitt	Math	3	2018-12-13 16:41
Spot the Difference	a1call	Miscellaneous Math	34	2017-12-11 00:17
LLR4 and PRP3 bugs	Mystwalker	Prime Sierpinski Project	4	2004-09-17 19:24
60bit, 64 bit - what is the difference??	dave_0273	Data	6	2003-10-04 14:21

2021-10-17, 03:54	#3
axn Jun 2003 12365₈ Posts	So the explanation is simple.Two different types of tests, two different residues. 4.0 is doing a PRP test. LLR2 is apparently doing an LLR test (shock! LLR doing an LLR test?!)

2021-10-17, 12:54	#4
ATH Einyen Dec 2003 Denmark 3,313 Posts	I just assumed everything was PRP tests, I was not even thinking about LLR tests. I'm not sure why LLR is not running LLR tests actually, it should when k<2ⁿ ? There is a ForcePRP=1 option, but I did not use it.

2021-10-17, 13:04	#5
axn Jun 2003 5×29×37 Posts	Starting from 3.8.something, PRP + GEC is available, so that is preferred due to the robust error checking.

2021-10-17, 13:28	#7
henryzz Just call me Henry "David" Sep 2007 Liverpool (GMT/BST) 3²×5×7×19 Posts	Might be wrong but I think it will run an llr test after a prp result

2021-10-22, 22:46	#8
diep Sep 2006 The Netherlands 2×17×23 Posts	Using LLR2 now which is 30.4 based. What would be needed to put in inifile to enforce it use PRP test and use Gerbicz Error Check (GEC)?