LLR / Gwnums version

[Cruelty]

Are there any plans to implement Gwnums 24.13 in LLR? Latest LLR v.3.6 is based on 24.11

Are there any plans to implement Gwnums 24.13 in LLR? Latest LLR v.3.6 is based on 24.11

I dont know, maybe George or Jean can answer this?
I'm not sure it is nessesary, without knowing the exact difference.

[Cruelty]

Originally Posted by TTn

I'm not sure it is nessesary, without knowing the exact difference.

Looking at the speed increase from prime95 24.11 to 24.13 I guess it would be a good idea to update LLR too.

Looking at the speed increase from prime95 24.11 to 24.13 I guess it would be a good idea to update LLR too.

I particularly meant, the exact difference in code.
Usually speed increases, translate to errors.
Since George and Jean know exactly what they are doing, maybe that is why it hasn't been done yet. Or otherwise pesonal time is a factor.

It is a valid question though.
TTn

[Kosmaj]

Jean mentioned that he is testing the new gwnum version, have a look here post #6 (and subsequent posts about the performance on HT cpu's). The new PRP based on 24.13 has been just released but note that brand new versions are risky without proper testing. LLR-3.6 is very well tested and unless any new versions are more than 10% faster, IMO it's not worth the trouble to migrate.

[Cruelty]

Am I the only one to notice new 3.6.2 version of LLR? I will give it a try later today... but perhaps somebody is already using it?

[Jean Penné]

Originally Posted by Cruelty

Am I the only one to notice new 3.6.2 version of LLR? I will give it a try later today... but perhaps somebody is already using it?

Congratulations for your watch !

LLR Version 3.6.2 is identical to LLR Version 3.6 about the features, but with two advantages :

1) It is linked with very last gwnum version : V24.14
2) I tested it on all the verified Riesel an Proth primes from the Chris Caldwell database, and found neither false negative result nor error message of any kind.

So, I think all LLR users have better to use this version now!

Happy hunting and Best Regards,

Jean

[Pconfig]

If i use llrnet, can i just replace llr.exe?

[Jean Penné]

Originally Posted by Pconfig

If i use llrnet, can i just replace llr.exe?

No, the llrnet executable is llrnet.exe, not llr.exe, and its last version is based
on llr version 3.5, I will ask to Vincent for upgrading it, but actually, I think he is too much busy...

[Citrix]

Jean,

Is multiplying number mod a^a-1 as fast as multiplying numbers mod k*2^n+1. Which one is faster? Secondly can LLR support a^a-1/a-1 and a^a+1/a+1?

Citrix

[Jean Penné]

Originally Posted by Citrix

Jean,

Is multiplying number mod a^a-1 as fast as multiplying numbers mod k*2^n+1. Which one is faster? Secondly can LLR support a^a-1/a-1 and a^a+1/a+1?

Citrix

Presently, the LLR program is a special one, it is devised to prove the primality of numbers of the k*2^n-1 and k*2^n+1 forms. On numbers of any other form, it can only do PRP tests, and is then equivalent to George Woltman's PRP program.

Its speed performances are mainly due to the using of the gwnum library code, so, they are the best for calculus modulo k*2^n+b or k*2^n-b, with k and b not larger than 2^20, either for primality proving or PRP tests.

Jean