How many "secret" tests are left?

[Moo_the_cow]

Does anyone know how many "secret" tests are left, and
whether "secret" checks for all n to 3M or only the n that the
previous prime searchers covered?

[jjjjL]

17099 left

and those are only ones done by previous searchers so some of the k's already have full coverage now. Since there are only a couple people doing it, it's only at 430,000 right now.

-Louie

[Joe O]

Does that count include these 21 tests?


Originally posted by jjjjL
all numbers added. :)

only changes I made before adding the numbers was to remove tests with n < 1000 from the files (SB sometimes crashes with extremely low tests) and I removed these tests for k=4847

2000487
2000607
2000631
2000727
2000943
2001327
2001471
2001543
2001831
2002071
2002143
2002407
2002503
2002767
2003151
2003463
2003727
2004951
2005551
2005647
2005671

because each of these tests would have been assigned to regular users and not "secret" the way the server is setup right now. it is only 21 tests so there's a good chance I could just slip them in and they would finish before most users even noticed they had them but regular users may not be interested in doing double-check work now so i won't make them.

if secret actually burns though all the work it has now, which i think will take at least a few weeks, then i'll manually assign the above tests to myself (or someone who's interested) and do them just to patch any holes in the ranges. also, i highly doubt that those are prime... they were checked the first time by Samidoost and he posted residues for all of them. He'd be the last person i'd expect to miss a prime. ;)

anyway, you can submit factors for the double-check again now. and have fun doing a few dozen proth tests an hour for the next few days if you decide to join in on the "secret". :)

-Louie

[Nuri]

Originally posted by jjjjL
17099 left

and those are only ones done by previous searchers so some of the k's already have full coverage now. Since there are only a couple people doing it, it's only at 430,000 right now.

-Louie
It seems to me that, with the current speed, it will take almost (and perhaps more than) a year to finish those 17099 tests.

So, I decided to switch to secret account for the next couple of weeks to help speed up secret for the next couple of 100,000s.

[Keroberts1]

what aer the chances of gettign a # of secret tests left tally in the stats menu? could be fun to watch the number drop and it might get more people to involve themselves in these tests so we could finish them off quickly

[smh]

also, i highly doubt that those are prime... they were checked the first time by Samidoost and he posted residues for all of them. He'd be the last person i'd expect to miss a prime.

This brings me to the question if results from the client are tested against other programs?

I think that it wouldn't hurt to test a few exponents of each K with another program to see if residues match. Especially exponents neer FFT crossovers.

[jjjjL]

i don't plan on adding a countdown of the "secret" tests since only a few people would understand it.

i also place almost no emphisis on completing them in a timely manner. i don't want people to spend their cpu time on it. i really, really doubt there's a prime in them. it's more an excercise in completeness & a way to help the DC siever's reduce their workload by trimming out the lowest n ranges. don't confuse that with the regular sieve. i don't think that file should ever have it's n-range trimmed. it will speed up as primes are found.

-Louie

[Nuri]

Originally posted by jjjjL
i don't plan on adding a countdown of the "secret" tests since only a few people would understand it. I agree, adding a countdown to the project page might mislead and confuse regular users. Updating the 17099 test figure on this thread from time to time (like once every couple of weeks) would probably be fine though.

i also place almost no emphisis on completing them in a timely manner. i don't want people to spend their cpu time on it. Louie, I just wanted to run secret up to 500K or 600K for a few weeks. I hope you don't mind.

i really, really doubt there's a prime in them. I agree that too. Still, I'd feel more comfortable when they're finished. :D

it's more an excercise in completeness & a way to help the DC siever's reduce their workload by trimming out the lowest n ranges. In fact, since the last algorithm change in the sieve client and Mike's alternative sob.dat file, I guess trimming out the lowest n ranges will not be an issue any more. All we have to do is to patch the holes on lower ranges at DC sieve (especilally those for p<10T). And for ranges p>20T, I really doubt we need to push DC only sieve that further.

don't confuse that with the regular sieve. i don't think that file should ever have it's n-range trimmed. it will speed up as primes are found. As a previous defender of n range trimming, I would have objected such a line a couple of weeks ago. Thanks to Mikael's idea, we now have a client that pushed the optimum level of range trimming to much higher p values so that we will not have to worry about it at least for the next several months.

[Keroberts1]

Hey louie I was wondering if we coudl get an update on how many secret tests are left, just a progress report nothing more. Thanks

[Mystwalker]

9 days have passed, each day the "secret" account completes ~70 tests (http://www.seventeenorbust.com/stats/users/user.mhtml?userID=3) - so there should be around 16,500 tests left...

[Nuri]

The actual figure might be "slightly" less than that. This is mainly because, DC sieving also kills some n values as well, and some (I guess roughly 17%) of the found factors fall within the range of secret's remaining tests.

[jjjjL]

16169

-Louie

[jjjjL]

also, 3 of the k's are totally finished and others are close

here is the breakdown by k

k count
4847 5831
5359 1978
10223 659
19249 1334
21181 1316
22699 689
24737 0
27653 0
28433 3282
33661 699
55459 381
67607 0

-Louie

[Moo_the_cow]

I know that this question is 5 letters off topic ;) , but how many supersecret
tests are left now?

[Nuri]

Louie, could you please give an update on secret user progress?

Thx in advance.

[jjjjL]

14593 secret tests left.

by k it is
4847 5548
5359 1734
10223 469
19249 1227
21181 1144
22699 613
28433 3139
33661 540
55459 179

there are 8 straggler tests left before secret finishes for all n < 500k.

supersecret is still around n=170k and all 7000 residues match.

-Louie

[Mystwalker]

supersecret is still around n=170k

reaching 180k now...
It's not that easy to up that range when I have to work on it alone... :cry:

Louie: could you chop the expire time for a supersecret test to maybe 1 day? That way, there won't be any straggler tests holding the Min n of the current test window low.
Possibly a cron job that releases assignment?
As an alternative, it is possible to assign supersecret's pending block management to me?

btw. I don't have access to these supersecret machines ;) on a regular base and only for short periods of time, so (super)secret tests are the only ones feasible...


all 7000 residues match.

Could still be an implementation error... :D

[jjjjL]

Originally posted by Mystwalker
Louie: could you chop the expire time for a supersecret test to maybe 1 day?

all numbers n < 1M now have 1 day expiration time.

-Louie

[Nuri]

It's not a big issue anyway, but may be we can think about increasing the lower bound of Low n Sieve (and probably only that one, not dual) from 300K to 500K.

Any comments?

[MikeH]

(and probably only that one, not dual) I'd suggest doing it for both.

Another option is to track supersecret, and raise the limit at (say) 100K or 200K thresholds when supersecret is about to hit the current given threshold. I think we have to agree that once candidates have PRPs with matching residues (i.e. double check) there is then zero value in continuing to sieve those ranges.

[MikeH]

Given that the questions here of "how many secret tests are remaining?", "how far through is supersecret?", "how many non-matching residues have you had?" will keep coming back every few weeks, might it be worth adding a couple of extra stats pages to keep everyone who is interested updated daily?

Any stats surrounding the exercise being performed by secret will ultimately have a limited life since it's goal is known, but those of supersecret (and however that evolves) will probably remain of interest for the life of this project.

[Moo_the_cow]

quote:
_______________________________________________
reaching 180k now...
It's not that easy to up that range when I have to work on it alone
_______________________________________________

You're not working on it alone. I switched my PC to supersecret testing
for a while.

[OberonBob]

I have 2 pcs working on secret. I will switch to supersecret when secret is done. we are up to n=500000 in secret. by the time we get to 600000 then we will be finished with 2 more K, as I recall, and the progress should speed up.

[Nuri]

I've also run ~200 secret tests from 434000 to 476000, and switched back to my account a couple of weeks ago. I'm planning another 200 test batch at secret soon.

Anyway, the stats page was showing the figures below for lower n bounds back in february. I guess these are the numbers we originally started.
So, it seems the first k to deplete next is 55459, to be followed with 10223.

k Lower n bound
4847 2006031
5359 875350
10223 610025
19249 1307678
21181 800204
22699 900190
24737 300127
27653 340089
28433 2000353
33661 645696
55459 540046
67607 400091


44131 690012
46157 617063
54767 1023127
65567 981643
69109 1125094

[Nuri]

Originally posted by MikeH
I'd suggest doing it for both.

Another option is to track supersecret, and raise the limit at (say) 100K or 200K thresholds when supersecret is about to hit the current given threshold. I think we have to agree that once candidates have PRPs with matching residues (i.e. double check) there is then zero value in continuing to sieve those ranges.

Mike et al,

Sure, doing both is ok for me too. I just proposed doing only Lower n because of two reasons:

- The marginal speed gain at Lower n would be much higher compared to Dual.
- I thought that some people might oppose a changeover at dual, and did not want to sacrifice a changeover at DC sieve.

But since we do not need factors for n<500K anymore, I guess there won't be many oppositions.

Tracking supersecret is a good idea too. But it really takes not much time prp testing for n<1m. So, what I'd propose is, tracking secret up to n=1m with 100K tresholds, and sticking with that lowerbound until supersecret reaches 1m. We can decide what to do next, looking at the sieve level we're at when supersecret hits 1m.

Any other suggestions?

[Moo_the_cow]

Louie, since we're at n>500 K now for secret, could you please give us an update on "secret" user progress?

[Nuri]

Before Louie updates, as far as I can see, not much effort is being put on secret for the last couple of days. So, it's highly likely that not much has changed (i.e. only 100-200 tests finished since last update).

BTW, Louie, would you consider putting the remaining tests stat to the web site under a normally not seen folder (like the one for the sieve). Something like, http://www.seventeenorbust.com/secret? This way, regular SoB users wouldn't be confused, and we'd see progress as frequent as we want to.

[OberonBob]

yeah, that would be great.

[Keroberts1]

I agree it would be a really good idea

[Moo_the_cow]

I agree too.

[jjjjL]

ok http://www.seventeenorbust.com/secret/

[Frodo42]

Thats neat.
Perhaps I'm just slow but it seems that the (super)secret tests have been removed from some of the other stats so they again are somewhat more reliable, which for my point of view would be the best about it all, to gett the secret-stuff seperated from the "real" stats
The lower n-bound seems to be OK again in the first stats page (even though the explained stats page still seems pretty messed up ...), now I don't know for sure but completed tests today seem pretty high still so I guess (super)secret have not been removed from there, that would be nice (perhaps they will disappear during the next 24 hours?).

[larsivi]

Frodo42 wrote:

now I don't know for sure but completed tests today seem pretty high still so I guess (super)secret have not been removed from there,

I would suppose that the reason for the higher number of completed tests lately would be due to Ars Technicas SoB gauntlet from 9th of June to 9th of July.

larsivi

[Joe O]

Louie, Thank you for this stats page. It's nice to know how things are going. Again, thanks!

[MikeH]

Since there are quite a few people running 'secret' right now. Just a reminder that 'secret' stats can be found at http://www.seventeenorbust.com/secret .

Somehow I'd missed this originally. Only found it searching for something else :rolleyes:

[Keroberts1]

We should be able to inish before the end of the month if we get a few more people to help out. I will be switching a couple CPUs over to it in a few days when i sieving range is finished

[OberonBob]

this has probably been asked before, but when we are done with secret, when all the secret tests are done, is secret going to switch to doing supersecret tests? I mean, will the secret "user" switch to supersecret tests?

I would think that it would be a good idea to do that, the double checks should be done, but the 20 or so people that are doing the secret tests are more than enough, and that number will drop when the 1.1.2 client comes out.

[Keroberts1]

The number of people running the secret account may no actually be enough although the double check is only 1/100th as likely to find a prime as was mentioned before adn may not be so this depth but there was a much better chance of finding a prime in the first place at this level and since these tests ca ne finished alot faster it may be that the optimal level to keep the double check at is far beyond our current level. If anyone has more info on the error % and how it depends on size of N value. Also what is the likelyhood of a number of a certain N value being prime. I believe it is omehow related to the inverse of the natural log but it is differnt because of proth numbers attributes. The final figure i would need is the amount of time required to do a test at different N values. I heard it stated o nthe forum before that doubleing the N value would quadruple the time required is this true or is it more complicated?

[wblipp]

Originally posted by Keroberts1
The number of people running the secret account may no actually be enough although the double check is only 1/100th as likely to find a prime as was mentioned before adn may not be so this depth but there was a much better chance of finding a prime in the first place at this level and since these tests ca ne finished alot faster it may be that the optimal level to keep the double check at is far beyond our current level. If anyone has more info on the error % and how it depends on size of N value. Also what is the likelyhood of a number of a certain N value being prime. I believe it is omehow related to the inverse of the natural log but it is differnt because of proth numbers attributes. The final figure i would need is the amount of time required to do a test at different N values. I heard it stated o nthe forum before that doubleing the N value would quadruple the time required is this true or is it more complicated?
You'll find answers to these questions in the thread on a Resource Allocation Model (http://www.free-dc.org/forum/showthread.php?s=&threadid=2988&highlight=resource+allocation+model). The best estimates of error rates at that time were extremely low - we were experiencing much lower error rates than GIMPS, perhaps because we haven't -attracted as many aggressive overclockers. If error rates have remained low, double checking is a poor use of resources until we get much higher.

William

[Keroberts1]

mostly what i was expecting ot hear but there probably is a certain level that would be optimal and it would probably be easy to have tests at that level handed out with regular tests. It would probably not even get noticed because at this point the tests would finish in about a half hour. Of course we'd need to wait until some more information is gathered about error rates at different levels and such.

[Nuri]

Originally posted by OberonBob
this has probably been asked before, but when we are done with secret, when all the secret tests are done, is secret going to switch to doing supersecret tests? I mean, will the secret "user" switch to supersecret tests?

I would think that it would be a good idea to do that, the double checks should be done, but the 20 or so people that are doing the secret tests are more than enough, and that number will drop when the 1.1.2 client comes out.

Currently, the upper limit of supersecret is blocked at 400k, and anyone using supersecret gets secret tests. If the same logic applies, when there will be no secret tests (provided that some secret tests will be released by that time), both secret and supersecret should start to get supersecret tests.

My guess is Louie is waiting for secret tests to approach completion in order to release some supersecret tests, and with the popuºarity it has now, it will probably finish within the next two weeks.

Since we've switched to the 1m<n<20m dat file at sieve, I guess he will increase the upper limit of supersecret tests to 1 million sometime towards the new year.

[Keroberts1]

If people still can't recieve credit for super secret tests they won't be very popular. With current amount of people running it secret would never reach 1 million, and to add to that these are the numbers that are least likely to yeild any results be cause the error rate is low in them.

[OberonBob]

Well, we have 13 days or so, like you said, and while it's no big deal, I just don't want secret to have any downtime. I hope he increases the supersecret top limit too. or for that matter, why have a limit? I mean other than it shouldn't pass the main effort, but that will never happen anyway.

[OberonBob]

Originally posted by Keroberts1
If people still can't recieve credit for super secret tests they won't be very popular. With current amount of people running it secret would never reach 1 million, and to add to that these are the numbers that are least likely to yeild any results be cause the error rate is low in them.

I like running secret on my slow boxes, it makes the boxes feel like they are do more for the effort by doing 2 tests a day, whereas it would take a month and a half to do a normal test.

[Keroberts1]

wouldn't those resources be more efficient if devoted to numbers that have not already been tested before.

The sieve is good for taht too because they may still find a factor or two a day and down the line those can save alot more effort for the main effort. Same deal with P-1 factoring.

[OberonBob]

I don't know about P-1 factoring, but sieve takes some non-zero amount of effort. the client is automatic. Besides, the double check is important too. gimps has a 3% error rate, we will have some errors too. The double check has to be done, and whats 20 pcs doing double check compared the 3500 doing the main task?

Besides, like I said, once the 1.1.2 client comes out, there will be like 4 computers doing double-check, that is not alot of processing power being allocated to cover that 3%. Gimps has about 6700 computers doing double-check right now with 27,000 doing LL tests, and 4300 doing factoring.

[ceselb]

we probably don't have 3% error rate on those low ranges, hopefully not even on the current ranges. The 3% figure is from much bigger numbers than we have done yet. My guess is 1-1.5% or so for 5M and <0.5% on current double check numbers.

[wblipp]

Originally posted by OberonBob
The double check has to be done, and
Actually, the double check does not need to be done. Seventeen or Bust is not trying to find the smallest prime, we are trying to find any prime. The last time I checked, we hadn't found any errors yet. If the SoB error rate is one in a thousand, then the double check should be approximately 10 times below the first tests (The first time tests about 100 times as long with about one-tenth the probability of finding a prime, so the primes per unit computing is about one thousand times slower). Shade it a bit for the changes in FFT size, and perhaps the ratio is 8 or 9 instead of 10.

[Keroberts1]

but this does support that there is an optimal level to have the double check at. So where would that place it?

[wblipp]

Originally posted by Keroberts1
but this does support that there is an optimal level to have the double check at. So where would that place it?
As the example shows, the optimal point is a bit above the cube root of the error rate.

[Joe O]

Originally posted by wblipp
The last time I checked, we hadn't found any errors yet.
Where would one check to see if there are any errors?

[hc_grove]

Originally posted by OberonBob
I don't know about P-1 factoring, but sieve takes some non-zero amount of effort.


It's the same for p-1.You have to manually reserve a range, tell the client the range and report the factors found.

[ceselb]

Where would one check to see if there are any errors?
That information isn't available, but Louie can check on the server.

[Joe O]

Originally posted by ceselb
That information isn't available

My point exactly!

[bagleyd]

Hmmm 19249 has some interesting secret statistics today....


Ooops, I see the number of tests can be zero and that makes the checked once field "n/a".
:blush:

[Keroberts1]

So the optimal level would be around 500000?

Could the server be programmed to assign secret tests to be done under people names so they could get credit and meanwhile reguardless of how people decide to use their resources they will be allocated properly to find as many primes in the least amount of time. I know this has been discussed before in the resource allocation model thread. Although i still believe that would be the best idea implimenting small trick to boost efficiency can be a big help too.

Also since i believe louie is the only one with error statistics could you give us some idea of how the error rates run through different sized tests. I understand as long as processors aren't overclocked its extremely rare to get a misreported test.

[OberonBob]

yeah, so you mean, like every say 500 tests, a "Supersecret" test goes out in the normal client to normal folks, and is counted normally? how often to be statically detremined by our assumed error rate and the changes on finding a prime during a doubt check?

Yeah, that might be the best way. it would automatic, it could be programed to maximize the odds, and folks get credit. Or, even better, there could be an option in the new client to accept the double check tests, so that only folks that want to will get the tests.

[Keroberts1]

yup :thumb:

It would be nice if we could skip the rest of 400,000 to 1,000,000 unless someone thinks that some of those have been misreported. As far as I've heard we haven't gotten a single non-matching residue yet, this tend to imply a much lower error rate than even 1 in a thousand. Perhaps we should skip ahead to a range thats more likely to have had some probems with errors. It does seem pointless to test numbers that haven't even shown to have an error rate at all yet.

I think the biggest thing that is required if any numbers of people are to ever run the double check is they must recieve credit for their work.


This is a small side question in case anyone knows should we start to expect serious problems whit the reliability of the client once we get toward much larger N values. Being taht we have now found a prime and are getting some boosts in membership we'll probably start running through ranges a bit quicker. Should another prime or two be found the nwe could possible even surpass GIMPS in size of primes we're searching for. I don't know tons about GIMPS but i believe that they have about half the same numbers as us to test before reaching the magic figure 10 milion. However 3 or 4 more primes would eliminate half of them and put us in a position to take lead.

According to Wblipps models we should be able to expect 3 more before reaching 10,000,000 digit primes. And we could always get lucky again. It is december again maybe this'll be a magic month and we'll pull 4 more out in a hurry.