mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Software

Reply
 
Thread Tools
Old 2013-12-20, 21:44   #1
Jean Penné
 
Jean Penné's Avatar
 
May 2004
FRANCE

57510 Posts
Default LLR Version 3.8.11 released

Hi All,

I uploaded today the stable (I hope!) version 3.8.11 of the LLR program.
You can find it now on my personal site :

http://jpenne.free.fr/

The 32bit Windows and Linux compressed binaries are available as usual.
The Linux 64bit binaries are also released here.
I uploaded also the complete source in a compressed file ; it may be used to
build the Mac-Intel executable and also the 64bit Windows binary.
This LLR version is linked with the Version 27.9 of George Woltman's gwnum library.
I released also a cllrd binary, which is linked with the debug version of the
gwnum library.
The main new feature in this version is it can now test the primality of numbers of the form b^n-b^m +/- 1 with n > m ; the header of the input file must then be ABC$a^$b-$a^$c +1 or -1, respectively.
If n > 2*m, only a strong PRP test can be done, but the factored part of the candidate is shown, if the number is PRP. If the factored part is at least 33%, the PFGW program can then be used to complete the proof. If the factored part is lower, it may be necessary to build an helper file, or to use a more general prover...
As usual, I need help to build the 32bit Mac Intel binary, and also the 64bit Mac Intel and Windows ones.
Please, inform me if you encountered any problem while using this new version.

Merry Christmas, Happy new year and Best Regards,
Jean
Jean Penné is offline   Reply With Quote
Old 2013-12-21, 01:51   #2
paulunderwood
 
paulunderwood's Avatar
 
Sep 2002
Database er0rr

3×17×71 Posts
Default

Merry Christmas and a Happy New Year to you, Jean.

I have proved a couple of primes: 10^277200-10^257768-1 and 10^277200-10^99088-1

On an Intel 4770k, hovering around 4.1GHz, the PRP tests run at 1.43msec/bit and the full proving test at 3.4msec/bit. The PRP tests are as quick as OpenPFGW. We were hoping for a bigger speed up, because R*b^n==R*b^m+R (mod b^n-b^m-1) and for smaller "m" I think a quicker special modular reduction is possible, by way of base conversions and an added shift.

One bug is that if "ForcePRP=1" is put llr.ini then I get:
Code:
Starting probable prime test of 10^257768-1

Last fiddled with by paulunderwood on 2013-12-21 at 01:51
paulunderwood is offline   Reply With Quote
Old 2013-12-22, 10:38   #3
Jean Penné
 
Jean Penné's Avatar
 
May 2004
FRANCE

10778 Posts
Default

Quote:
Originally Posted by paulunderwood View Post
Merry Christmas and a Happy New Year to you, Jean.

I have proved a couple of primes: 10^277200-10^257768-1 and 10^277200-10^99088-1

On an Intel 4770k, hovering around 4.1GHz, the PRP tests run at 1.43msec/bit and the full proving test at 3.4msec/bit. The PRP tests are as quick as OpenPFGW. We were hoping for a bigger speed up, because R*b^n==R*b^m+R (mod b^n-b^m-1) and for smaller "m" I think a quicker special modular reduction is possible, by way of base conversions and an added shift.

One bug is that if "ForcePRP=1" is put llr.ini then I get:
Code:
Starting probable prime test of 10^257768-1
Hi Paul,
Yes, it is a bug and I will fix it!
However, would you excuse me, but I don't understand why you need to use the ForcePRP option...
Regards,
Jean
Jean Penné is offline   Reply With Quote
Old 2013-12-22, 14:05   #4
paulunderwood
 
paulunderwood's Avatar
 
Sep 2002
Database er0rr

3·17·71 Posts
Default

Quote:
Originally Posted by Jean Penné View Post
Hi Paul,
Yes, it is a bug and I will fix it!
However, would you excuse me, but I don't understand why you need to use the ForcePRP option...
Regards,
Jean
I was going to try it to see if would be quicker.

Is a special mod reduction George's domain?
paulunderwood is offline   Reply With Quote
Old 2013-12-22, 14:40   #5
Jean Penné
 
Jean Penné's Avatar
 
May 2004
FRANCE

52·23 Posts
Default

Quote:
Originally Posted by paulunderwood View Post
I was going to try it to see if would be quicker.

Is a special mod reduction George's domain?
It is not mine... But I think the best is to ask that to him!
Jean Penné is offline   Reply With Quote
Old 2013-12-22, 20:08   #6
Jean Penné
 
Jean Penné's Avatar
 
May 2004
FRANCE

57510 Posts
Default LLR 3.8.11

Hi,

The bug is now fixed on all released binaries, and on the source.
Regards,
Jean
Jean Penné is offline   Reply With Quote
Old 2013-12-23, 22:21   #7
pinhodecarlos
 
pinhodecarlos's Avatar
 
"Carlos Pinho"
Oct 2011
Milton Keynes, UK

488710 Posts
Default

Seems to be that 3.8.11 is slower than 3.8.9 on LLR for k=5 at n=3.7M.
pinhodecarlos is offline   Reply With Quote
Old 2013-12-24, 17:20   #8
KEP
Quasi Admin Thing
 
KEP's Avatar
 
May 2005

96410 Posts
Default

Quote:
Originally Posted by pinhodecarlos View Post
Seems to be that 3.8.11 is slower than 3.8.9 on LLR for k=5 at n=3.7M.
Have you checked that the FFT length hasn't increased? I have tested hundreds of base 2 tests for a total of 13 different k's and there hasn't been any significant difference in speed, but what you are reporting should of course be investigated, however an explanation could be that you have hit an FFT jump, wich will slow down your testing

Last fiddled with by KEP on 2013-12-24 at 17:21
KEP is online now   Reply With Quote
Old 2013-12-24, 19:50   #9
pinhodecarlos
 
pinhodecarlos's Avatar
 
"Carlos Pinho"
Oct 2011
Milton Keynes, UK

33×181 Posts
Default

Quote:
Originally Posted by KEP View Post
Have you checked that the FFT length hasn't increased? I have tested hundreds of base 2 tests for a total of 13 different k's and there hasn't been any significant difference in speed, but what you are reporting should of course be investigated, however an explanation could be that you have hit an FFT jump, wich will slow down your testing
No FFT Jump. Same numbers, more 100 seconds.

Last fiddled with by pinhodecarlos on 2013-12-24 at 19:59
pinhodecarlos is offline   Reply With Quote
Old 2013-12-24, 23:24   #10
KEP
Quasi Admin Thing
 
KEP's Avatar
 
May 2005

22×241 Posts
Default

Quote:
Originally Posted by pinhodecarlos View Post
No FFT Jump. Same numbers, more 100 seconds.
Is that for a complete test (all the way through from iteration 1 to iteration max) or is it based on timings of each iteration?
KEP is online now   Reply With Quote
Old 2013-12-24, 23:33   #11
pinhodecarlos
 
pinhodecarlos's Avatar
 
"Carlos Pinho"
Oct 2011
Milton Keynes, UK

33×181 Posts
Default

Quote:
Originally Posted by KEP View Post
Is that for a complete test (all the way through from iteration 1 to iteration max) or is it based on timings of each iteration?
Complete test on 4 numbers, went back to previous version.
pinhodecarlos is offline   Reply With Quote
Reply

Thread Tools


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 3.8.15 released Jean Penné Software 28 2015-08-04 04:51
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

All times are UTC. The time now is 15:24.

Mon Apr 19 15:24:28 UTC 2021 up 11 days, 10:05, 0 users, load averages: 1.54, 1.99, 2.14

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.