Bases 6-32 reservations/statuses/primes

[Flatlander]

Yes, that's correct.

[Flatlander]

R base 30 tested to 110k. Sieve to 140k attached, P=3T.

[mdettweiler]

Riesel base 23 is complete to n=570K, no primes. Results are attached for 500K-570K. Releasing this base; 3.5-hour-long tests were becoming a bit too much for me to stomach. (According to my calculations, it would have taken me at least 70 more days to finish this up to my original target of 600K! )

Reserving Riesel base 22 for n=300K+. This base is somewhat heavy-weight (at least compared to other down-to-one-k bases) so hopefully it can be knocked out at a lower n-depth than others.

[gd_barnes]

Originally Posted by mdettweiler

Riesel base 23 is complete to n=570K, no primes. Results are attached for 500K-570K. Releasing this base; 3.5-hour-long tests were becoming a bit too much for me to stomach. (According to my calculations, it would have taken me at least 70 more days to finish this up to my original target of 600K! )

Reserving Riesel base 22 for n=300K+. This base is somewhat heavy-weight (at least compared to other down-to-one-k bases) so hopefully it can be knocked out at a lower n-depth than others.

You didn't, by chance, sieve R23 any higher did you? If so, can you send the file. Otherwise, I'll just post the remaining n=570K-1M as is. There's little doubt that the remaining range would need to be sieved higher than 10T.

R22 is at a lower sieve depth of P=7T than R23 (P=10T). R22 is also higher weight and has a larger k-value, meaning longer tests at some n-levels. The smaller base is almost inconsequential. Therefore R22 will likely need to be sieved more very soon. It may need to be sieved now, even taking into account 32-bit sieving, depending on how high you plan on testing.


Gary

[mdettweiler]

Originally Posted by gd_barnes

You didn't, by chance, sieve R23 any higher did you? If so, can you send the file. Otherwise, I'll just post the remaining n=570K-1M as is. There's little doubt that the remaining range would need to be sieved higher than 10T.

R22 is at a lower sieve depth of P=7T than R23 (P=10T). R22 is also higher weight and has a larger k-value, meaning longer tests at some n-levels. The smaller base is almost inconsequential. Therefore R22 will likely need to be sieved more very soon. It may need to be sieved now, even taking into account 32-bit sieving, depending on how high you plan on testing.


Gary

No, I didn't sieve any more for R23. I figured it might not be sieved quite optimally for the really high stuff I was testing at the end, but that the few candidates which would be removed by further sieving weren't worth the hassle and extra time needed to sieve further. (PRP testing for this was done on 4 cores, as opposed to just one core that I could spare for sieving, hence if I'd sieved further it would have taken somewhat longer in terms of wall-clock time.)

As for R22, I hadn't thought of sieving that one farther...you're right, it probably would be good. I'm not sure how far I'll take this one; that will mostly depend on how long it sticks around without a prime, and how much more the added weight makes the testing drag out. I'll have to think about that a bit.

[gd_barnes]

Originally Posted by mdettweiler

No, I didn't sieve any more for R23. I figured it might not be sieved quite optimally for the really high stuff I was testing at the end, but that the few candidates which would be removed by further sieving weren't worth the hassle and extra time needed to sieve further. (PRP testing for this was done on 4 cores, as opposed to just one core that I could spare for sieving, hence if I'd sieved further it would have taken somewhat longer in terms of wall-clock time.)

As for R22, I hadn't thought of sieving that one farther...you're right, it probably would be good. I'm not sure how far I'll take this one; that will mostly depend on how long it sticks around without a prime, and how much more the added weight makes the testing drag out. I'll have to think about that a bit.

Yeah, when making a reservation that you have to sieve, it's best to figure out how far you'll take it if you don't find a prime. I'll tell you, the chances of finding a prime for the entire n=300K to 1M range are likely less than 50-50 even for a somewhat higher-weight k like this. It's always better to assume you won't find one when sieving.

My guess is that you'll want to do n=300K-500K for this one, although you may want to stop at n=450K, if you're R23 is an indication of how long you like to test high n-range candidates. It appears to be ~20-25% higher weight than R23. So if you sieve n=300K-1M further, you could take your removal rate for an n=440K test and multiply by 3. That's because n=300K-500K is 1/3rd of n=300K-1M. What, in effect, that would do is tell you how fast candidates are being removed from ONLY the range that you plan on testing.

Hey, got an offer for you that will help you out here: Feel free to stop one (or even two) of my cores on the server machine from running NPLB llrnet. Be sure and return the 5 candidates to the server with the LLRnet -c command so they don't sit untested for 2 days. You can use those 1-2 cores for sieving this thing. (Actually you could use 1-2 cores on any one of my Intel's, which are all running NPLB LLRnet.) srxsieve is within the Prime folder on the desktop and then within the various prime1, prime2, prime3, and prime4 subfolders. You may have seen it already.

There's no use to sieve such a high n-depth with 32-bit sieving. Too inefficient.

Note: This won't decrease the amount of time that you sieve (Well, it might decrease it 3-5% since higher P-ranges sieve more quickly than lower ones due to fewer possible prime factors). Twice as fast means almost twice the optimum sieve depth (minus likely 3-5%). What it will do is save you quite a bit more on the testing side of things.

Gary

[gd_barnes]

JapelPrime reported by Email that Sierp base 9 is at n=359K and is continuing.

[mdettweiler]

Originally Posted by gd_barnes

Yeah, when making a reservation that you have to sieve, it's best to figure out how far you'll take it if you don't find a prime. I'll tell you, the chances of finding a prime for the entire n=300K to 1M range are likely less than 50-50 even for a somewhat higher-weight k like this. It's always better to assume you won't find one when sieving.

My guess is that you'll want to do n=300K-500K for this one, although you may want to stop at n=450K, if you're R23 is an indication of how long you like to test high n-range candidates. It appears to be ~20-25% higher weight than R23. So if you sieve n=300K-1M further, you could take your removal rate for an n=440K test and multiply by 3. That's because n=300K-500K is 1/3rd of n=300K-1M. What, in effect, that would do is tell you how fast candidates are being removed from ONLY the range that you plan on testing.

Hey, got an offer for you that will help you out here: Feel free to stop one (or even two) of my cores on the server machine from running NPLB llrnet. Be sure and return the 5 candidates to the server with the LLRnet -c command so they don't sit untested for 2 days. You can use those 1-2 cores for sieving this thing. (Actually you could use 1-2 cores on any one of my Intel's, which are all running NPLB LLRnet.) srxsieve is within the Prime folder on the desktop and then within the various prime1, prime2, prime3, and prime4 subfolders. You may have seen it already.

There's no use to sieve such a high n-depth with 32-bit sieving. Too inefficient.

Note: This won't decrease the amount of time that you sieve (Well, it might decrease it 3-5% since higher P-ranges sieve more quickly than lower ones due to fewer possible prime factors). Twice as fast means almost twice the optimum sieve depth (minus likely 3-5%). What it will do is save you quite a bit more on the testing side of things.

Gary

Thanks for the sieving offer. I don't have time right now to do an optimal depth calculation for this, but I'll try to do something within the next couple of days. You're right, 300K-500K sounds about right...even though 450K would probably be equivalent to where I stopped on base 23 I'd want to make sure I'm prepared to test it all the way to 500K sieving-wise.

[gd_barnes]

Originally Posted by mdettweiler

Thanks for the sieving offer. I don't have time right now to do an optimal depth calculation for this, but I'll try to do something within the next couple of days. You're right, 300K-500K sounds about right...even though 450K would probably be equivalent to where I stopped on base 23 I'd want to make sure I'm prepared to test it all the way to 500K sieving-wise.

Oh, OK; well don't let that stop you from testing, say, n=~300K-325K to start with. Even it ends up being slightly inefficient, it's better than doing nothing.

I could likely do one sometime by Mon. or Tues. Just remind me if you want me to.

[Cruelty]

Base=10 (Sierpinski + Riesel) tested till n=380000.

[gd_barnes]

JapelPrime has reported that he will be releasing Sierp base 9 at n=360K, although isn't quite there yet.

He has sent me his sieve file so I'll reserve it up to n=450K. The file goes to n=500K so if it's not too high weight, I may take it that high.