Limits for Prime95/mprime

[Peter Nelson]

Can anyone tell me what the limits are for the Prime95 (windows) and mprime(linux) software (probably the same).

I would like to know what is the largest exponent whose Mersenne number can be TRIAL FACTORED using the software.

I would like to know what is the largest exponent whose Mersenne number can be LUCAS-LEHMER tested for primality using the software.

I could not find this documented, but am already aware of several alternative software for these purposes. I just wanted to know what limitations of Prime95 are. For example, I believe, (but don't know for sure) that Prime95 cannot be used for TF on billion digits Mersennes.

For bonus marks, you may also know the largest exponent whose Mersenne number can be attempted to find factor using P-1 method in the program also (assuming that memory limitations on pc are not an issue, or maybe reach 4GB max).

Thankyou.

[Xyzzy]

Using "AdvancedFactor" you can trial factor a lot higher than the default 79.3M exponents... It may not be 100% reliable though... Look at the >79.3M forum for more info...

[Peter Nelson]

Thanks for trying to help with your suggestion, but I already read the LMH > .... and Operation Billion Digits threads. For example, I already have Luigi's Factor3_2 program.

I found this post:

At the current time, we use Prime95 by manually putting the line AdvancedFactor=start_exponent,end_exponent,b1,b2
into worktodo.ini where b1 and b2 are the starting bit and the ending bit inclusive. However, Prime95 misses some factors (by some indications lots) for exponents above 79300000. Also, advanced factor is a slow, out of date feature. At some point, someone may make a program with comparable speed to Prime95 to do this job.

SO, MY QUESTION REMAINS

What is the largest exponent Prime95 can Trial Factor?
What is the largest exponent Prime95 can LL test?

I am interested in capabilities in current range, 10M digit, 100M digit, billion digit etc.

I will further qualify these with "RELIABLY" ie at what size exponent does Prime95 begin to miss factors?

False positives are not such a big problem as factors are easily verified.

I assume latest versions eg version 24.6 or similar.

[Prime95]

LL test (and P-1):

79.3 million for x87
77.9 million for SSE2

Factoring:

I'm not sure. I just eyeballed the code and see no reason why p up to 2^31 wouldn't work. If someone can send me an example of a missed factor, I can debug the cause and come up with a limit.

[Peter Nelson]

Thanks George those answers are exactly what I was after.

If I notice discrepancies of skipped factors versus output of (say) factor3_2 then I will post details as requested.

[ET_]

Originally Posted by Peter Nelson

Thanks George those answers are exactly what I was after.

If I notice discrepancies of skipped factors versus output of (say) factor3_2 then I will post details as requested.

Found those threads on the forum:

http://www.mersenneforum.org/showthread.php?t=1921
http://www.mersenneforum.org/showthread.php?t=1925
http://www.mersenneforum.org/showthread.php?t=1946

HTH

Luigi

[Prime95]

ET_, I tested these:

AdvancedFactor=195004883,195004883,45,50
AdvancedFactor=195000258,195000272,0,50

and found the appropriate factors.

When reporting bugs, let me know the machine type. There is different code for 486, Pentium, and PentiumPro and later machines.

[ET_]

Originally Posted by Prime95

ET_, I tested these:

AdvancedFactor=195004883,195004883,45,50
AdvancedFactor=195000258,195000272,0,50

and found the appropriate factors.

When reporting bugs, let me know the machine type. There is different code for 486, Pentium, and PentiumPro and later machines.

At that time I was using an Athlon, and version prime95 22.12 or 23.5.

Hbock worked on the104_105 range with Prime95 and sent me the results.txt file, but I did not have the time to recheck it.

Luigi