Sierp. base 6 team sieve

[mdettweiler]

As requested by Gary, I've done some optimal depth analysis for n=150K-200K.

Test PRP candidate:

51255*6^185002+1 is not prime.  RES64: 6ED5CDA2CB4F4464.  OLD64: 4C8168E861EDCD29  Time: 1466.336 sec.

sr2sieve timings w/all factors removed up to p=10.8T:

Range: 10.8T-10.9T (100G)
Expected # of factors: 35.64
p/sec.: 1409495

Time to do range: 100G / 1409495 = 70947.396
Removal rate: 70947.396 / 35.64 = 1990.7

PRP test at n=185K: 1466.336 sec.

Based on these calculations, the optimal depth for n=150K-200K is approximately 8T. Since we're sieved to almost 11T, breaking 150K-200K immediately is a no-brainer. I will post the updated sieve file, and email 150K-200K to Gary, shortly.

[gd_barnes]

Originally Posted by Anonymous

As requested by Gary, I've done some optimal depth analysis for n=150K-200K.

Test PRP candidate:

51255*6^185002+1 is not prime.  RES64: 6ED5CDA2CB4F4464.  OLD64: 4C8168E861EDCD29  Time: 1466.336 sec.

sr2sieve timings w/all factors removed up to p=10.8T:

Range: 10.8T-10.9T (100G)
Expected # of factors: 35.64
p/sec.: 1409495

Time to do range: 100G / 1409495 = 70947.396
Removal rate: 70947.396 / 35.64 = 1990.7

PRP test at n=185K: 1466.336 sec.

Based on these calculations, the optimal depth for n=150K-200K is approximately 8T. Since we're sieved to almost 11T, breaking 150K-200K immediately is a no-brainer. I will post the updated sieve file, and email 150K-200K to Gary, shortly.

Thanks for taking care of that Anon! Nice work.

We may actually already be at optimal depth for n=200K-250K but wait until after KriZp finishes his next range and then do the same thing using a test at n=235K and of course after removing the 150K-200K piece.

Thanks for the awesome sieving KriZp!

KriZp, after you're done with base 6 here, would you be interested in sieving a large range for either base 4 or base 16? Sierp base 4 will need some sieving for n=1M-3M or 4M (2M-6M or 8M base 2), which will take a long time. Both base 16 conjectures will need sieving from n=200K-750K or 800K (n=800K-3M or 3.2M base 2).

We're only around n=130K on testing for both base 16 conjectures but could bring them up to the current sieved limit at n=200K relatively quickly if a priority was placed on them.

Another possibility would be Riesel base 256. With 47 k's remaining and most k's only tested to n=25K or 50K (n=200K or 400K base 2), it'd be nice to sieve them all up to n=100K or 125K (n=800K or 1M base 2).

Base 4 and base 256 would have more of an immediate impact because we have little to no sieved files for them whereas we do for base 16 for a while yet. Base 256 would be more interesting for sieving and LLRing because it's at a much lower n-range and hence would not take nearly as long to sieve or LLR.


Gary

[KriZp]

I would be interested, but I've turned off a few computers as spring has started, and so my crunching power has diminished. I may turn off all of them in late may/early june, and probably bring them online beginning in august/september.

[mdettweiler]

Gary, would you like me to get a team sieve started for Sierp. base 4 and Riesel base 256? We don't have to turn them into a team drive immediately when we're done with the sieve if you don't think interest is high enough, but I think a team sieve is still the way to go for these large sieving efforts.

[gd_barnes]

Originally Posted by Anonymous

Gary, would you like me to get a team sieve started for Sierp. base 4 and Riesel base 256? We don't have to turn them into a team drive immediately when we're done with the sieve if you don't think interest is high enough, but I think a team sieve is still the way to go for these large sieving efforts.

I wouldn't do Sierp base 4 until we have progress on that mini-drive. I might put a core on it once I have my 6th quad built later this week but I think there will be very limited interest in LLRing it above n=1M (n=2M base 2).

Riesel base 256 would be a good one although t's a little tricky due to the many different testing limits. Why don't you start a thread that asks people if they'd be interested in contributing some sieving to this base? If at least 2 people besides you and me would be interested, then I'd say go for it. Note that I probably wouldn't contribute for a while; at least until I get my Sierp base 3 reservation up to k=30M; ETA is ~3 weeks on that.

Sheesh, I feel like I need 10 quads. lol


Gary

[mdettweiler]

Originally Posted by gd_barnes

I wouldn't do Sierp base 4 until we have progress on that mini-drive. I might put a core on it once I have my 6th quad built later this week but I think there will be very limited interest in LLRing it above n=1M (n=2M base 2).

Riesel base 256 would be a good one although t's a little tricky due to the many different testing limits. Why don't you start a thread that asks people if they'd be interested in contributing some sieving to this base? If at least 2 people besides you and me would be interested, then I'd say go for it. Note that I probably wouldn't contribute for a while; at least until I get my Sierp base 3 reservation up to k=30M; ETA is ~3 weeks on that.

Sheesh, I feel like I need 10 quads. lol


Gary

Okay, Riesel base 256 sounds good. As for the different testing limits, though--I don't think that will be much of a problem, since most of the k's with wacky testing limits are covered by other projects anyway (in fact, quite a lot are covered by NPLB--maybe you should mark those as reserved by NPLB so nobody wastes their time on them when NPLB is progressing so quickly anyway?) So, we can just leave out the few that are at wacky testing limits--since it's only a few k's, it shouldn't be hard for someone to do them on their own if they want to reserve them.

I'm thinking that there should be enough interest for a Riesel base 256 team sieve, considering that the Sierp. base 6 sieve has now progressed to a high enough point that we've got plenty of work ready to last us quite a while, considering the current rate of PRP testing being done. Riesel base 256, by contrast, is in more immediate need of sieving, and thus might attract more contributers, not to mention that it won't be a problem if it steals away resources from the Sierp. base 6 team sieve temporarily (since, as I just mentioned, that sieve is in pretty good shape already).

But, yes, it probably would be a good idea to start a thread requesting opinions on it. I'll get to that as soon as I get the chance.

[gd_barnes]

Originally Posted by Anonymous

Okay, Riesel base 256 sounds good. As for the different testing limits, though--I don't think that will be much of a problem, since most of the k's with wacky testing limits are covered by other projects anyway (in fact, quite a lot are covered by NPLB--maybe you should mark those as reserved by NPLB so nobody wastes their time on them when NPLB is progressing so quickly anyway?) So, we can just leave out the few that are at wacky testing limits--since it's only a few k's, it shouldn't be hard for someone to do them on their own if they want to reserve them.

I'm thinking that there should be enough interest for a Riesel base 256 team sieve, considering that the Sierp. base 6 sieve has now progressed to a high enough point that we've got plenty of work ready to last us quite a while, considering the current rate of PRP testing being done. Riesel base 256, by contrast, is in more immediate need of sieving, and thus might attract more contributers, not to mention that it won't be a problem if it steals away resources from the Sierp. base 6 team sieve temporarily (since, as I just mentioned, that sieve is in pretty good shape already).

But, yes, it probably would be a good idea to start a thread requesting opinions on it. I'll get to that as soon as I get the chance.

No, I don't want to mark any Riesel base 256 as reserved by NPLB or RPS even though some k's reduce to their k-values. That's because base 256 technically only tests every 8th n-value. I'll reserve base 4 vs. base 256 (or base 2 vs. base 16), which would test only every 4th n-value of a base 4 k (base 2 k), but not any higher.

The reasoning is the amount of time. NPLB is at n=~427K on k=400-1001. That converts to only n=~53K base 256. We would like to search k's on base 256 faster than NPLB will progress on base 2 as it relates to base 256. For example, the goal on NPLB is to reach n=600K by year end. That would only be n=75K base 256. There's no reason to wait so long even though it's technically duplicate work, it's a very small percentage of duplicate work in the realm of all k's and n-values.

On the base 256 page, you'll see that I have reserved k's that convert to k's that are being searched by base 4 and base 16 by the individuals or drives that are searching them. I also show the base 2 k's that some of the base 256 k's convert to but don't specifically reserve them for base 256. By showing the converted base 2 k's, it's easy to see what RPS or NPLB k's testing limit to refer to. (Actually I need to update a couple of the ranges.)


Gary

[mdettweiler]

Originally Posted by gd_barnes

No, I don't want to mark any Riesel base 256 as reserved by NPLB or RPS even though some k's reduce to their k-values. That's because base 256 technically only tests every 8th n-value. I'll reserve base 4 vs. base 256 (or base 2 vs. base 16), which would test only every 4th n-value of a base 4 k (base 2 k), but not any higher.

The reasoning is the amount of time. NPLB is at n=~427K on k=400-1001. That converts to only n=~53K base 256. We would like to search k's on base 256 faster than NPLB will progress on base 2 as it relates to base 256. For example, the goal on NPLB is to reach n=600K by year end. That would only be n=75K base 256. There's no reason to wait so long even though it's technically duplicate work, it's a very small percentage of duplicate work in the realm of all k's and n-values.

On the base 256 page, you'll see that I have reserved k's that convert to k's that are being searched by base 4 and base 16 by the individuals or drives that are searching them. I also show the base 2 k's that some of the base 256 k's convert to but don't specifically reserve them for base 256. By showing the converted base 2 k's, it's easy to see what RPS or NPLB k's testing limit to refer to. (Actually I need to update a couple of the ranges.)


Gary

Ah, I see now.

[KriZp]

taking 12T-13T, 11T-12T completes in 7 hours.

[KriZp]

12T-13T complete; reserving 13T-15T.

[KriZp]

13T-15T done, reserving 15T-17T