LLR Version 3.8.5 is available!

[Jean Penné]

Hi All,

The new version 3.8.5 of LLR is now available on my personal site :

http://jpenne.free.fr/index2.html

This version is almost identical to V 3.8.4, but uses Version 26.5 of
George Woltman's gwnum library.

Also, two bugs are fixed, and one option is added ; please see Readme.txt file
for details.

Best Regards,
Jean

P.S. : I still need help to build the Mac Intel binary...

[Christenson]

Silly noob question: LLR = ?? (And I'm not running any macs, so no help with the binary)

[mdettweiler]

Originally Posted by Christenson

Silly noob question: LLR = ?? (And I'm not running any macs, so no help with the binary)

LLR is a program, based on Prime95, which tests numbers of the form k*b^n+/-c for primality. The Lucas-Lehmer (LL) test that Prime95 uses on Mersenne numbers (2^n-1) can be extended with some slight modifications to the more general case of k*2^n-1--known as the Lucas Lehmer Riesel (LLR) test, since the extension was discovered by a certain Hans Riesel (the numbers k*2^n-1 are also known as Riesel numbers after him). The LLR program was originally designed to perform just LLR tests, but has since been extended to also handle Proth tests (for Proth numbers, k*2^n+1) and N-1/N+1 tests (for non-base-2 k*b^n+/-1 numbers). For |c|!=1, LLR can perform a PRP (probable prime) test. (Prior to recent versions, it used a PRP test on non-base-2 k*b^n+/-1 numbers as well; it produced probable primes which needed to be proven separately using a program such as PFGW to perform an N-1 or N+1 test. Since version 3.8.0, LLR can now do these full-proof tests directly.)

LLR isn't used by GIMPS, since Mersenne numbers (a subset of the Riesel numbers, with k=1) can be tested faster with the LL test than with LLR. Nonetheless, LLR is used quite heavily by a number of other project searching for different kinds of primes. A number of these are hosted in the "Prime Search Projects" section of the mersenneforum, though perhaps the biggest user of LLR is PrimeGrid, a project that has modified it to be run through the BOINC framework and which has picked up quite a large group of participants. (I'm not sure but they may even be bigger than GIMPS at this point.)

[otutusaus]

Thanks for the new option, now the tests on relatively small n's don't stress the HDD anymore, plus some time saved.

[nuggetprime]

btw,rounding error problem testing 245*830^492-1 seems to be fixed in this version.
Nugget

[otutusaus]

I am currently running LLR on machines to which I don't have access on an everyday basis. In such cases, when the job is done, since I don't have the chance to add more input files, the computer sits idle until I again have access to it. I know I could add the next job at the end of the input file, and later on edit lresults.txt to get the data. But altogether it gets messy and there is always the slight chance of data loss when having to modify files.

I would suggest to allow LLR to take more than one input file to overcome this issue. My thinking would be to have more than one input file in the folder (input1.txt, input2.txt, ... as an example) and include these extra lines in the llr.ini:

PgenInputFile=input1.txt
PgenInputFile2=input2.txt
(...)
PgenInUse=[1/2/3/...]

LLR would take input files sequentially according to PgenInputFile[x] and PgenInUse. This will require creating a new results file, which will have to take a different name (maybe lresults.txt, lresult2.txt, ...?).

Finally, in my opinion this feature should be disabled by default for compatibility purposes with other projects (PRPNet, LLRNet) and would be enabled with an extra option MultiPgen=1.

Thanks, Jean!

Oscar

[mdettweiler]

Hmm...interesting idea. I wonder, though, if it could be accomplished about as easily with an external script (batch or shell, for Windows or Linux respectively). For instance, you could use a batch file like this:

cllr -d input1.txt
ren lresults.txt lresults1.txt
cllr -d input2.txt
ren lresults.txt lresults2.txt
cllr -d input3.txt
ren lresults.txt lresults3.txt
(etc.)

Of course, this would requite a bit of manual fiddling if you had to stop and restart the program; but with a bit more advanced batch scripting, it could be designed to (for instance) skip straight to input3.txt if lresults1.txt and lresults2.txt exist.

Note that this uses the command-line Windows version of LLR, cllr.exe, instead of the GUI LLR.exe. (I use it all the time myself so I can string up multiple jobs like this and avoid idle "down time".) If you're familiar with the Linux version of LLR, the command-line Windows version should be exactly the same.

[otutusaus]

The external script option looks a viable solution, but I am attached to the GUI version.
I basically don't have any machine running under Linux so I currently only use the GUI app.
In addition, as you say, when using the script rebooting the computer would interrupt the batch process (which happens too often when running on a shared computer).

[Jean Penné]

Hi All,

Thanks to Iain Bethune, the new LLR version binary for Mac Intel is now available from my personal pages :

http://jpenne.free.fr/index2.html

Regards,
Jean

[ATH]

Is the Lucas+Frobenius PRP test a "normal" Lucas Pseudoprime test with some other parameters?

I'm wondering if it would be "easy" to add the choice of running the Lucas PRP test with Selfridge parameters from the BPSW test, with D being the first number from the list 5,-7,9,-11,13, ..., for which the jacobi symbol (D/n)=-1 :
http://mathworld.wolfram.com/Baillie...alityTest.html

[Jean Penné]

Originally Posted by ATH

Is the Lucas+Frobenius PRP test a "normal" Lucas Pseudoprime test with some other parameters?

I'm wondering if it would be "easy" to add the choice of running the Lucas PRP test with Selfridge parameters from the BPSW test, with D being the first number from the list 5,-7,9,-11,13, ..., for which the jacobi symbol (D/n)=-1 :
http://mathworld.wolfram.com/Baillie...alityTest.html

Presently, this test uses the smallest positive D = P^2-4*Q such as kronecker (D, N) = -1, so, P is generally not 1

I think it would not be difficult to implement another version of the same test using the Selfridge parameters. It may be a BPSW option... I shall add this in my work to do list for the next version of LLR.
Regards,
Jean