Bases 33-100 reservations/statuses/primes

[gd_barnes]

Originally Posted by 10metreh

Sorry, but I couldn't get onto the internet in time to see the post. As a result, kar_bon re-computed the base to n=500 and PMed me before I read the suggestion.

That's OK because I'm doing it now. It should be done in about a day. I'll send you the primes that I find for n=501 as well as the k's that were remaining at n=500. You can see if both of them match what Karsten sent you and what you have remaining if you have anything to compare them to.


Gary

[gd_barnes]

Originally Posted by KEP

Ps. Hope that this saved you some admin time

As a matter of fact, it did. Thanks!

[gd_barnes]

KEP,

Question:
Are you certain you are complete on S60 to n=75K with no more primes? The last prime was at n=37209 and there are 28 k's remaining. There were 9 primes for n=26K-38K so this seems suspect. Can you please attach a results file for n=25K-75K? I want to verify that nothing was missed. Thanks.

Request:
You do realize that you have probably 5-10 CPU years worth of work reserved, correct? Due to problems in the past with this, please do not reserve any more bases without finishing up or releasing your current bases. Current reservations:

S58: 119 k's for n=50K-125K (a HUGE effort!)
S60: 28 k's for n=75K-125K
S70: 6 k's for n=25K-100K (I reduced this from n=25K-250K to 25K-100K here)
S383: 50 k's for n=25K-100K (nothing reported since April 21st; a very significant effort for a high base; running on a dual core will take years!)

Is that everything?


Gary

[gd_barnes]

Originally Posted by 10metreh

No. Also, testing just 1 n with so many ks (72360 were remaining at n=500) results in the bizarre problem of srsieve spending its time spewing out thousands of "removed candidate sequence xxxxxxx*51^n-1 from the sieve" errors.

But it would be easier if I could remove the primed ks. Then I would have about 3.3 million less tests to do! I have a feeling inside me that there probably is a way to produce a list of all the ks and then remove the ones with primes below n=500 (I have the primes with n up to 500 separate from the larger ones) to produce a list of the ks that remained at n=500. But I now can't do any more tests due to the complete wiping of the pl_remain.txt file by Mini-Geek's remove_ks.pl script which I am about to detail in "the scripts thread".

After running the starting bases script to n=501 and removing 134 k's that have partial algebraic factors, I show that 72277 k's remain. There were 83 k's with primes for exactly n=501, which means that there are 72360 k's remaining at n=500. That works out to exactly what you have. Very good.

The only thing a little unusual is that I have 83 primes for n=501, which is 56 more than the 27 that you have that are missing. But I'm guessing that you only lost part of the n=501 primes. I'm assuming that you still had the other 56 primes intact. In case there is still a discrepency, here is a complete list of primes for n=501:

134928*51^501-1
261854*51^501-1
471480*51^501-1
708670*51^501-1
903660*51^501-1
947312*51^501-1
1072292*51^501-1
1115560*51^501-1
1270304*51^501-1
1335224*51^501-1
1488400*51^501-1
1592854*51^501-1
1959660*51^501-1
1997074*51^501-1
2035962*51^501-1
2068630*51^501-1
2078470*51^501-1
2088438*51^501-1
2099290*51^501-1
2141764*51^501-1
2325068*51^501-1
2712308*51^501-1
2887488*51^501-1
2936650*51^501-1
3109614*51^501-1
3148068*51^501-1
3169322*51^501-1
3476552*51^501-1
3509538*51^501-1
3741128*51^501-1
3800902*51^501-1
3814150*51^501-1
3868368*51^501-1
3894814*51^501-1
3914220*51^501-1
4040402*51^501-1
4102088*51^501-1
4211884*51^501-1
4238040*51^501-1
4284870*51^501-1
4406448*51^501-1
4465820*51^501-1
4616782*51^501-1
4650270*51^501-1
4924674*51^501-1
5089020*51^501-1
5235990*51^501-1
5255394*51^501-1
5273750*51^501-1
5328948*51^501-1
5385604*51^501-1
5407832*51^501-1
5553344*51^501-1
5689588*51^501-1
5721608*51^501-1
5733248*51^501-1
5769978*51^501-1
5799544*51^501-1
5869258*51^501-1
6116098*51^501-1
6495724*51^501-1
6635422*51^501-1
6641064*51^501-1
6672430*51^501-1
7176404*51^501-1
7287928*51^501-1
7346808*51^501-1
7357620*51^501-1
7477194*51^501-1
7597942*51^501-1
7759560*51^501-1
7882518*51^501-1
7932424*51^501-1
7944652*51^501-1
8024774*51^501-1
8139280*51^501-1
8203178*51^501-1
8448304*51^501-1
8520812*51^501-1
8552194*51^501-1
8594604*51^501-1
8615140*51^501-1
8630254*51^501-1

Also for reference, attached is a list of the 72277 k's that are remaining at n=501. To match with your n=500 remaining list, you'd have to add back the k's for the above primes. I manually removed the k's with partial algebraic factors, which I assume is what you did.

BTW, for future reference, I think you'll find it to be much less hassle to use the starting bases script to test to n=1000 before starting sieving. It may even be less overall CPU time if you set trial factoring to 30% (-f30 switch). It's such a huge hassle to sieve so many k's. I'm guessing that you'd be down to ~50000 k's or less remaining at n=1000. Here, it was only a little over 1 CPU day to script the base to n=501. I'm estimating 4 CPU days to script the base to n=1000 with far less hassle in sieving. Personally, I script everything on all bases to n=2500 but would probably make an exception for conjectures of k>~3M like this. When I script to n=2500, I almost always trial factor to 100%, although 50-70% may be somewhat faster. I haven't tested that possibility.


Gary

[10metreh]

Yes, I only lost 27 of the n=501 primes (the first 27 on your list). I removed the PAFs manually, but it would be useful if the new bases script could identify them.

[gd_barnes]

Originally Posted by 10metreh

Yes, I only lost 27 of the n=501 primes (the first 27 on your list). I removed the PAFs manually, but it would be useful if the new bases script could identify them.

Yes, it would be highly useful. But the # of possible conditions and the limitations of the scripting language make it very difficult. I've visualized 2 different ways that it could be done but both would require inordinately ugly, LONG, and time-consuming code. The # of test conditions required to prove it as 100% correct would also be very large.

Feel free to take a hack at the script if you'd like since you're quite clear on the algebraic factors after proving (to my satisfaction anyway) my conjectures in the generallizing algebraic factors thread.

I do have a hacked-up version of the script that doesn't test squared k's on squared Riesel bases but its use is very limited and makes testing slightly longer on the 95%+ of bases that are not squared Riesel bases. So it's not really useful as a public release. As the scripting language currently exists, it's not easy to even figure out if a value is a perfect square, whether that be the k or the base.

[MyDogBuster]

Riesel 80, all k's, tested n=100K-200K. 1 prime previously reported

Results emailed - Base released

[Rincewind]

So, I've managed to "repair" a computer (its like a zombie with parts from other "death" computers). The performance is not as high as it was before, but I'll take on R36.
I'll use my 'old' sieve file to continue sieving and reserve primetests up to 50.000.
Oh, and I'm making daily backups atm *g*.

[MyDogBuster]

Oh, and I'm making daily backups atm *g*.

Good idea. I'm having a bad weekend also. Lost 1 hard drive (I backed it up just an hour before it went). Luckily I had a spare drive. Also lost 1 case fan and 1 cpu/heatsink fan. Spares for both, but the spare part department is severely depleted.

That's 3. Things happen in threes. Hopefully I'm done.

[Rincewind]

Originally Posted by MyDogBuster

Good idea. I'm having a bad weekend also. Lost 1 hard drive (I backed it up just an hour before it went). Luckily I had a spare drive. Also lost 1 case fan and 1 cpu/heatsink fan. Spares for both, but the spare part department is severely depleted.

That's 3. Things happen in threes. Hopefully I'm done.

WOW, that's hard. But the backup-timing was excellent!
I hope too that you're through this stuff.

[gd_barnes]

Originally Posted by Rincewind

So, I've managed to "repair" a computer (its like a zombie with parts from other "death" computers). The performance is not as high as it was before, but I'll take on R36.
I'll use my 'old' sieve file to continue sieving and reserve primetests up to 50.000.
Oh, and I'm making daily backups atm *g*.

Great to hear that you got it back. I'll Mark you down as reserved on the 74 unreserved k's for R36.

By the way, the sieve file that you provided me had been sieved to P=175G not P=140G as you had said before. I took your file, updated the sieve limit, and removed the primed k so that there are now 74 k's in it. There is a link to it on the Riesel reservations page. You would just need to convert it back to ABCD format for sieving.