PRPclient 2.4.7 with LLR 3.8.0 now available!

opyrt · 2010-02-09, 09:41

Hi everyone! :)

Jean Penné has now officially released LLR 3.8.0 which gives a nice speed boost when running this project. According to the documentation it's 2-5%, but I've seen a 10% increase when running LLR on all cores on an Intel Quad.

This is a good time to upgrade to PRPclient 2.4.7 aswell as there are known issues with the older versions.

IMPORTANT!!

LLR 3.7.1 and LLR 3.8.0 do not produce compatible checkpoint files. If you are in the middle of a test and upgrade LLR to 3.8.0 it will start over. A way to do this gracefully is to upgrade PRPclient to 2.4.7 and set number of WUs to 0. The client will then finish the current WU and not download a new one. To do this, change these lines (you will obviously have to set them back to 1 once you have upgraded LLR):

server=PSPtest:80:1:www.psp-project.de:7100
server=PSPtestdc:20:1:www.psp-project.de:7101

to this:

server=PSPtest:80:0:www.psp-project.de:7100
server=PSPtestdc:20:0:www.psp-project.de:7101

With the new package, we've set the client to run 80% first pass and 20% second pass as this helps the project to keep up with second pass. Please note that running second pass is optional, and you are free to set these percentages as you please!

You can download PRPclient 2.4.7 with or without the new LLR here: http://tools.1up.no/psp/prpclient/

This package is specially built for this project (genefer is removed) so please do not use this against the PrimeGrid PRPnet servers.

LLR 3.8.0 standalone can be downloaded here (PRPclient uses the cllr version): http://jpenne.free.fr/index2.html

You can upgrade PRPclient and LLR over an existing installation by adding them to the the "programs" folder and running the "install" batch file according to your number of cores (I've only tested that on windows). You might want to update your ini files aswell.

Thanks to:
Rogue for PRPnet 2.4.7
Jean Penné for LLR 3.8.0
Runesk for packaging the programs for this project.

I hope I have covered everything in this post, but please ask if you have any questions. :)

Ltd: If you want to host these packages on the psp-project.de servers, please copy them and alter this post.

Cybertronic · 2010-02-09, 13:35

Attention !

With the new version, I have test my gigantic prime triplet. The 3rd number is not prime, but it is prime.

It is

2072644824759 * 2^33333 -1,+1,+5

shot test-file is

954651922:T:0:2:131
2072644824759 33333

best

Norman

opyrt · 2010-02-09, 14:23

Quote:

Originally Posted by Cybertronic

Attention !

With the new version, I have test my gigantic prime triplet. The 3rd number is not prime, but it is prime.

It is 2072644824759 * 2^33333 -1,+1,+5

I don't understand... Do you mean LLR 3.8.0 says 2072644824759*2^33333+5 isn't prime, but you have verified by other methods that it is?

Cybertronic · 2010-02-09, 14:34

Yes, the triplet was found with LLR 3,7.

Look here:
http://www.sendspace.com/file/cyndmc

opyrt · 2010-02-09, 14:39

I did some testing, and I doubt this is relevant for this project. But I really appreciate you letting us know.

LLR 3.7.1c:

Code:

C:\cllrtestold>cllrold.exe -d -q"2072644824759*2^33333+1"
Starting Proth prime test of 2072644824759*2^33333+1
Using zero-padded FFT length 3584, a = 7
2072644824759*2^33333+1 is prime!  Time : 1.977 sec.

LLR 3.8.0:

Code:

C:\cllrtest>cllr.exe -d -q"2072644824759*2^33333+1"
Starting Proth prime test of 2072644824759*2^33333+1
Using zero-padded FFT length 3584, a = 7
2072644824759*2^33333+1 is prime!  Time : 1.884 sec.

But with the +5 I get this:
LLR 3.7.1c:

Code:

C:\cllrtestold>cllrold.exe -d -q"2072644824759*2^33333+5"
Invalid expression in command line.

LLR 3.8.0:

Code:

C:\cllrtest>cllr.exe -d-q"2072644824759*2^33333+5"
Starting probable prime test of 2072644824759*2^33333+5
Using zero-padded FFT length 3584, a = 3
2072644824759*2^33333+5 is not prime.  RES64: 20A5CD45B0850A08.  OLD64: 61F167D1118F1E15  Time : 1.911 sec.

Mini-Geek · 2010-02-09, 14:44

Quote:

Originally Posted by opyrt

I don't understand... Do you mean LLR 3.8.0 says 2072644824759*2^33333+5 isn't prime, but you have verified by other methods that it is?

LLR 3.7.2: "2072644824759*2^33333+5 is 3-Strong PRP! (3 more test(s))"
LLR 3.8.0: "2072644824759*2^33333+5 is not prime. RES64: 20A5CD45B0850A08. OLD64: 61F167D1118F1E15"
PFGW 3.3.1 (uses GWNUM 25.13): "2072644824759*2^33333+5 is 3-PRP!"
Prime95 25.11: "2072644824759*2^33333+5 is a probable prime!"

I don't know of a good way to independently (i.e. without GWNUM), and in reasonable time, verify which is correct, but it would appear that LLR 3.8.0 has the problem.

opyrt · 2010-02-09, 15:56

FYI: I've notified Jean of this thread. :)

Jean Penné · 2010-02-09, 16:06

Quote:

Originally Posted by Mini-Geek

LLR 3.7.2: "2072644824759*2^33333+5 is 3-Strong PRP! (3 more test(s))"
LLR 3.8.0: "2072644824759*2^33333+5 is not prime. RES64: 20A5CD45B0850A08. OLD64: 61F167D1118F1E15"
PFGW 3.3.1 (uses GWNUM 25.13): "2072644824759*2^33333+5 is 3-PRP!"
Prime95 25.11: "2072644824759*2^33333+5 is a probable prime!"

I don't know of a good way to independently (i.e. without GWNUM), and in reasonable time, verify which is correct, but it would appear that LLR 3.8.0 has the problem.

I also tested with Prime95 25.11 and find this number is PRP, so, I agree with you! Already a problem ? UGH ! I have now to take a look in the new PRP code...
Also, to test without gwnum may be a good idea...
Sorry for this drawback,
Jean

Cybertronic · 2010-02-09, 16:14

Quote:

Originally Posted by Jean Penné

I also tested with Prime95 25.11 and find this number is PRP, so, I agree with you! Already a problem ? UGH ! I have now to take a look in the new PRP code...
Also, to test without gwnum may be a good idea...
Sorry for this drawback,
Jean

Jean, no problem.

For smaller numbers ,like n*2^333 -1,+1,+5 works LLR3.8 correct.
Maybe without my triplet, the bug will be never found by us

Jean Penné · 2010-02-09, 16:46

Quote:

Originally Posted by Cybertronic

Jean, no problem.

For smaller numbers ,like n*2^333 -1,+1,+5 works LLR3.8 correct.
Maybe without my triplet, the bug will be never found by us

Maybe... But I just found your number PRP while using the old PrimeForm program, so gwnum cannot be responsible for this error, and I should prefer to understand its cause!

Thanks!
Jean

rogue · 2010-02-09, 16:53

Quote:

Originally Posted by Jean Penné

Maybe... But I just found your number PRP while using the old PrimeForm program, so gwnum cannot be responsible for this error, and I should prefer to understand its cause!

Thanks!
Jean

The problem could be with gwnum. There are issues with gwnum for some numbers. Look at this thread, http://www.mersenneforum.org/showthr...t=12226&page=2. gwnum does not trigger an error (roundoff, sumout) that PFGW can detect, thus the test result of some primality tests is incorrect. George is aware of the issue and hopes to investigate when he returns from his vacation.

Jean, I suggest you run some of the numbers posted in that thread through LLR 3.8.0 and see if you see some of the same issues.

2010-02-09, 09:41	#1
opyrt Apr 2008 Oslo, Norway 7×31 Posts	PRPclient 2.4.7 with LLR 3.8.0 now available! Hi everyone! :) Jean Penné has now officially released LLR 3.8.0 which gives a nice speed boost when running this project. According to the documentation it's 2-5%, but I've seen a 10% increase when running LLR on all cores on an Intel Quad. This is a good time to upgrade to PRPclient 2.4.7 aswell as there are known issues with the older versions. IMPORTANT!! LLR 3.7.1 and LLR 3.8.0 do not produce compatible checkpoint files. If you are in the middle of a test and upgrade LLR to 3.8.0 it will start over. A way to do this gracefully is to upgrade PRPclient to 2.4.7 and set number of WUs to 0. The client will then finish the current WU and not download a new one. To do this, change these lines (you will obviously have to set them back to 1 once you have upgraded LLR): server=PSPtest:80:1:www.psp-project.de:7100 server=PSPtestdc:20:1:www.psp-project.de:7101 to this: server=PSPtest:80:0:www.psp-project.de:7100 server=PSPtestdc:20:0:www.psp-project.de:7101 With the new package, we've set the client to run 80% first pass and 20% second pass as this helps the project to keep up with second pass. Please note that running second pass is optional, and you are free to set these percentages as you please! You can download PRPclient 2.4.7 with or without the new LLR here: http://tools.1up.no/psp/prpclient/ This package is specially built for this project (genefer is removed) so please do not use this against the PrimeGrid PRPnet servers. LLR 3.8.0 standalone can be downloaded here (PRPclient uses the cllr version): http://jpenne.free.fr/index2.html You can upgrade PRPclient and LLR over an existing installation by adding them to the the "programs" folder and running the "install" batch file according to your number of cores (I've only tested that on windows). You might want to update your ini files aswell. Thanks to: Rogue for PRPnet 2.4.7 Jean Penné for LLR 3.8.0 Runesk for packaging the programs for this project. I hope I have covered everything in this post, but please ask if you have any questions. :) Ltd: If you want to host these packages on the psp-project.de servers, please copy them and alter this post.

2010-02-09, 13:35	#2
Cybertronic Jan 2007 Germany 2³·41 Posts	Attention ! With the new version, I have test my gigantic prime triplet. The 3rd number is not prime, but it is prime. It is 2072644824759 * 2^33333 -1,+1,+5 shot test-file is 954651922:T:0:2:131 2072644824759 33333 best Norman Last fiddled with by Cybertronic on 2010-02-09 at 13:42

2010-02-09, 14:34	#4
Cybertronic Jan 2007 Germany 101001000₂ Posts	Yes, the triplet was found with LLR 3,7. Look here: http://www.sendspace.com/file/cyndmc Last fiddled with by Cybertronic on 2010-02-09 at 14:40

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Tired of all the prpclient console windows?	opyrt	Prime Sierpinski Project	4	2010-05-31 08:42

2010-02-09, 15:56	#7
opyrt Apr 2008 Oslo, Norway 7×31 Posts	FYI: I've notified Jean of this thread. :)