Sierp base 3 - mini-drive Ia

[Flatlander]

29548672*3^96142+1
Is prime.

[gd_barnes]

Originally Posted by KEP

I've split the sieve over 3 cores, because it lowered the completion time from ~30-35 days, to ~10 days. I know it is not all to efficient, but the difference is only a few percent, so I decided to do it this way. I may add, that I only sieve the 3 ranges to 150G and not any further. Once the 50M-100M range completes, I'll e-mail you the sieve file, and then continue complete the other 2 ranges, starting with 100M-150M and then six days later complete the 150M-200M range in about 10 days

Regards

KEP

Oh, OK. That sounds great. Thanks for the excellent sieving work!

[mdettweiler]

@Gary: Ah, I see what you mean. Yeah, that sounds good.

[gd_barnes]

Originally Posted by mdettweiler

@Gary: Ah, I see what you mean. Yeah, that sounds good.

I was just thinking of one more thing. To keep the Roman-numeral prefixes in sync, so to speak, I think we'll just call the Sierp drive for k=50M-100M 'Mini-Drive - Sierp base 3 I(b)' after which we'll go II, III, IV, V, etc.

I kind of like the idea of Roman numeral prefixes for drives of the same base.


Gary

[Flatlander]

94-95 and 95-96 complete. No primes.

When the n is below 50,000, say, do you think things would be more manageable if the files were available as ks rather than ns? i.e. 3 or 5 ks per file. It would then just be a case of each tester removing the k from their file when they find a prime.

[gd_barnes]

Originally Posted by Flatlander

94-95 and 95-96 complete. No primes.

When the n is below 50,000, say, do you think things would be more manageable if the files were available as ks rather than ns? i.e. 3 or 5 ks per file. It would then just be a case of each tester removing the k from their file when they find a prime.

Is 93-94 complete?

Funny you mention that. That's how I've been doing the conjecture searches over several cores for quite a while now. For instance, I have Sierp base 31 divided up on 2 cores; one running k=~1 to 3.1M and the other k=~3.1M-6.2M. Much less duplication of testing that way.

I have to think of the best way to design that for a drive. The problem is, one person could end up with 3 k's that never yield a prime over a large n-range. Therefore, it'd have to be something like 3 k's over n=25K-50K or something like that.

The myriad of situations that could arise has prevented me from running a drive like that yet. The main problem being one person ending up with a file that takes far longer to test than another person; and possibly even longer than they care to continue testing. It would be far easier to administer but a little difficult to manage workload properly.


Gary

[Flatlander]

Originally Posted by gd_barnes

Is 93-94 complete?
Gary

Yes, sorry.

[Flatlander]

23243282*3^99009+1
Is prime.

[Flatlander]

96-100 complete.

[gd_barnes]

Originally Posted by Flatlander

96-100 complete.

Thanks, Chris, for your immense contribution to the base 3 efforts!

[Flatlander]

Originally Posted by gd_barnes

...

First, we'll do k=50M-100M for n=25K-100K to get it 'on the same page' as the Riesel side. In the mean time, one person like me (when I free up a resource or 2) or anyone else who wants to volunteer for it, can run k=100M-200M for n=25K-35K or 25K-40K. Later on, we can start a drive for that k-range for n=35K (or 40K) to 100K. It's best that only one person or at most 2 do it. If that works out well, we'll plan on doing it that way for future mini drives.
...
Gary

I'll do it. (Unless this is now not the most efficient way, it light of Karsten's scripts? Perhaps it's now more efficient if I started with n even lower, using the scripts?)

Is it time to rename and unsticky this thread?