Should we change to [URL=http://www.moregimps.it/billion/factor4.zip]factor4[/URL]?

It's really faster than factor3_2...

[QUOTE=fetofs]Should we change to [URL=http://www.moregimps.it/billion/factor4.zip]factor4[/URL]?

It's really faster than factor3_2...[/QUOTE]

But does it work properly? I'll change the website once I'm sure that Factor4 actually works well or else I will maybe get a lot of mails telling me that "I can't get the program to work".

I'll watch this closely..

[QUOTE=fetofs]Should we change to [URL=http://www.moregimps.it/billion/factor4.zip]factor4[/URL]?

It's really faster than factor3_2...[/QUOTE]

:wink:

We just need some more QA from other prople (as my factorbase has been correctly replied). The program worked fine in the environments I tested, so if anybody likes to do some tests, he/she is welcome.

Luigi

[QUOTE=ET_]:wink:

We just need some more QA from other prople (as my factorbase has been correctly replied). The program worked fine in the environments I tested, so if anybody likes to do some tests, he/she is welcome.

Luigi[/QUOTE]

The only thing I notice is that the program takes >3 minutes to sieve every 0.1 depth. Are you sure that is optimal?

P.S: And now, if I'm not mistaken, we can do partial ranges (like 2^63 to 2^63.5) :smile:

But the bad fact is that we'll have to stay with factor3_2b while there isn't (and probably there will never be) a factor4_b.

[QUOTE=fetofs]The only thing I notice is that the program takes >3 minutes to sieve every 0.1 depth. Are you sure that is optimal?[/quote]

For what range? Which was the timing for version 3_2 at the same range?

[QUOTE=fetofs]P.S: And now, if I'm not mistaken, we can do partial ranges (like 2^63 to 2^63.5) :smile:[/quote]

It should work even for smaller intervals. :rolleyes:

[QUOTE=fetofs]But the bad fact is that we'll have to stay with factor3_2b while there isn't (and probably there will never be) a factor4_b.[/QUOTE]

There isn't a factor4b version in the to-do list *at the moment*. I'm actually applying to a Riesel-sieve program, but never say never...

Luigi

[QUOTE=ET_]For what range? Which was the timing for version 3_2 at the same range?
[/QUOTE]

I sieved 63.4 to 63.5. Factor3_2 sieved the entire range (63-64) in less than half a minute, although factor4's calculations were about 275% faster in the actual factoring.

63.5 to 63.6 took 200 sec (the sieving, of course)

[QUOTE=fetofs]I sieved 63.4 to 63.5. Factor3_2 sieved the entire range (63-64) in less than half a minute, although factor4's calculations were about 275% faster in the actual factoring.

63.5 to 63.6 took 200 sec (the sieving, of course)[/QUOTE]

Here is another reason why I did not work on a batch version.

I decided to use different structures that take time to initialize. You can use a sieve initialization to check factors from 1 bit to approximately 62.5 bit. There should be an algorithm that minimizes that time based on the bit range you take. From 1 to 50 bit is faster than from 50 to 51 bits.

Anyway, I'll recheck it.

Luigi

Finished this range.
I couldn't quite understand if I should switch to the new factor4.
I am trying it out on 332192897 from 65 to 70. It seems to be working on my P4 with Win XP. What is it doing in the sieve initalizing? It took 124 secs. What is the sieve percentage? It was 95 %.

Thanks Joshua.

Again, please indicate if you want to continue a completed range. For now, I take it you don't...

M332193529 no factor from 63.000 bits to 64.000 bits.

[QUOTE=Joshua2]What is it doing in the sieve initalizing? It took 124 secs. What is the sieve percentage? It was 95 %.[/QUOTE]

During the sieve initializazion we create a bitmap of k's (as in 2kp+1); modulo and prefactoring operations rule out 95% of the candidates. The remaining 5% of k's are submitted to a modpow algorithm that checks if 2[sup]p[/sup] (mod 2kp+1) == 1.

Luigi