sieve stat question

[Keroberts1]

cool

[MikeH]

Congratulations to hc_grove for taking the top spot in the 14 day average.

And also congratulations to Nuri, who almost unnoticed took the top spot a couple of days ago.

Both very much deserved.

(I've just submitted a few factors, so let's see what my fortunes hold. )

[ceselb]

MikeH, you might want to raise the 'active' window' for double checking.

Your page states 400000 < n < 600000, while my current test for secret shows 739778 (k=19249).


Another thing. Caould you add another page view of the holes?

Based on http://www.aooq73.dsl.pipex.com/gaps...n3_20_p05u.htm

I'd like something like this (One row per not completed range/user pair):

( 90.00G) : 121210-121300 expmod8
( 81.36G+ 183.05G) : 121300-122000 Slatz
( 150.02G+ 81.40G) : 122000-123000 Mystwalker
( 100.00G) : 123400-123500 Moo_the_cow
( 192.41G) : 123600-123800 biwema
( 661.94G) : 123800-125000 cmprince
( 83.48G) : 125000-125100 rosebud
( 176.87G) : 125100-125500 Foobar
( 802.46G) : 126000-128000 Nuri
( 447.76G+ 749.78G) : 126000-128000 Nuri
( 72.41G+ 204.31G+ 377.59G+ 33.23G) : 128000-129000 mklasson
( 106.54G+ 393.46G) : 129000-129500 Troodon
( 380.38G) : 129500-130000 TheCrusher
( 105.82G+ 122.08G+ 120.21G+ 111.46G+ 93.21G) : 130000-131000 hc_grove
( 525.33G+ 334.76G) : 131000-132000 priwo
( 855.03G) : 132000-133000 Mat67
( 500.00G) : 133000-133500 biwema
( 440.03G) : 133500-134000 ceselb
( 173.83G+ 138.09G+ 157.76G+ 151.51G+ 146.22G+ 160.12G+ 147.98G+ 579.37G) : 134000-136000 Mystwalker


ranges added up if it's not too much work. If it's too difficult to do, a simple list would be great. ( I do this manually right now)

121210-121300 expmod8
121300-122000 Slatz
122000-123000 Mystwalker
123400-123500 Moo_the_cow
123600-123800 biwema
123800-125000 cmprince
125000-125100 rosebud
125100-125500 Foobar
126000-128000 Nuri
126000-128000 Nuri
128000-129000 mklasson
129000-129500 Troodon
129500-130000 TheCrusher
130000-131000 hc_grove
131000-132000 priwo
132000-133000 Mat67
133000-133500 biwema
133500-134000 ceselb
134000-136000 Mystwalker

[MikeH]

Your page states 400000 < n < 600000, while my current test for secret shows 739778 (k=19249).

secret performs tests on candidates declared as tested by previous searchers, but for which no residue has been provided. supersecret performs double checking - testing candidates that have only one residue to generate a second (hopefully) matching residue. Louie placed an artificial stop point on supersecret of 400K. Right now supersecret will function as secret. So right now 400K - 600K is correct. If you (sieve and) submit factors for candidates designated for secret, they will score immediately. One such example was submitted yesterday

130.482 4847 743223 225.237 Sun 23-Nov-2003 hc_grove

Another thing. Could you add another page view of the holes?

I'll have a try. Unfortunately I haven't had a lot of time to give to SoB recently (work and family), so can't promise any timescales.

[hc_grove]

Originally posted by MikeH
One such example was submitted yesterday

130.482 4847 743223 225.237 Sun 23-Nov-2003 hc_grove

I did wonder why that factor scored so much, thanks for the explanation.

[ceselb]

Originally posted by MikeH
I'll have a try. [...] can't promise any timescales.

No problems. Do it next year when you've got some time to spare. I've managed so far, so I'll survive a few months more.

[peterzal]

hey i got a quick question about the MikeH's sieving stats. I submitted 11 factors 2 days ago, but only 8 of them showed up in the stats. And whats up with these factors only getting a score of 1.4ish?

Thanks,

Peter Zalewski

[MikeH]

I submitted 11 factors 2 days ago, but only 8 of them showed up in the stats. And whats up with these factors only getting a score of 1.4ish?

Some of the 11 were duplicates, only 8 were unique.

(from your score card )

Sieve 3M<n<20M Range size FactorsU FactorsD FactorsE Score
143000 - 143050 50 8 4 121 17.164

...and the scoring has changed since last you were here. Wecome back

http://www.aooq73.dsl.pipex.com/

[peterzal]

nevermind, i actually took the time to read some of this forum and my questions were answered.

[cmprince]

I thought I was just imagining it, but over the past few days I've noticed the "active window" in the 5M range has moved backward; how does that happen? I thought it tracked with the last assigned proth test.

Last Update: Thu 11-Dec-2003 09:01 (GMT)
Current 'active' windows:
400000 < n < 600000
4879702 < n < 5079702

I'm especially curious since I had a 5.1M factor score as if the window had already passed it.

Thanks,
Chris

[Frodo42]

The active windows is decided by next.txt and it has been stuck at 5359*2^4879702 for some time now.
Why that is I don't know hopefully it only affects the stats, but I dont know therefore i posted this

[MikeH]

I've now changed the sieving stats so that it uses the highest of the 'Current test window'-> 'max n' on this page instead of using next.txt.

Lets hope this stops those crazy jumps

[hc_grove]

Okay, I screwed up a little. Today I've submitted a number of factors without being logged in. Is it possible to have them properly assigned to me? (luckily none of them were on-the-spot).

I've done the submitting from the same IP number as I normally use (80.196.145.71), but a different computer (behind a NAT router) on which I hadn't checked the "remember me" box. I could send somebody the fact.txt's to show which factors it might be (I can't remeber which I have submitted from my regular box, so there will be additional factors in them).

I have a fixed IP, so if any factors have previously been submitted from that by somebody not logged in, that will be me too.

.Henrik

[MikeH]

Okay, I screwed up a little. Today I've submitted a number of factors without being logged in.

If you take a look on your score card I think you'll find they're already sorted.

For any factors where the user wasn't logged in, I check the sieve and P-1 reservations, and match them up that way automatically. So usually you'll only see them as [unknown] if I'm a bit behind on updating my copies of the reservations.

If you can't see your factors, give me a shout and I'll have a deeper look.

[MikeH]

Sieve stats have now been adjusted to reflect the finding of the prime k=5359.

I have gone back to the state of play on Wed 16 Dec 03 03:00, frozen all the k=5359 scores, then re-imported factors submitted between then and now. This means that all factors submitted since Wednesday will now be time stamped as today (sorry if this causes any confusion ).

All factors k=5359 submitted after Wed 16 Dec 03 03:00 will score 0. Any duplicates that have ever been found for k=5359 now score 0. As a result of these changes, most people's scores have decreased a little.

The project page now shows only the remaining ks. It's really good to see the {-3381} for the current 1M slice, and {-67755} for all 20M .

I'm starting an archive of the three main score pages, saving snap shots at significant points (of which this is the first).

There are still a number of k=5359 items floating around on various pages, but if anyone sees anything that now looks wrong, let me know.

[Joe O]

Why does

21.419 21181 716852 21.419 Fri 19-Dec-2003

not appear under

Factors next to enter (double check) 'active window' (@n=600000)

For that matter why don't:

   21.341  28433    625345     21.341  Mon 15-Dec-2003 
   21.322  27653    638457     21.322  Mon 15-Dec-2003 
   13.922  24737    534991     13.922  Mon 15-Sep-2003 
   13.678   4847    629127     13.678  Tue 09-Sep-2003  
   13.371  10223    547901     13.371  Tue 02-Sep-2003                      
    8.265   4847    637767      8.265  Thu 24-Jul-2003

For that matter, there are others that should have appeared there but have now been passed but belonged there before when the window was lower.

Such as:

   13.282  10223    439805     13.282  Sun 31-Aug-2003
   13.256   5359    480526     13.256  Sun 31-Aug-2003

[MikeH]

Here's an indication of how sieving has progressed. I thought it was worth sharing this data.

The x axis are sieve score updates (four per day), so this shows the progress since early August. The y axis is in Gs. The lines relate to the data on the project page.

[Nuri]

Originally posted by Joe O
Why does not appear under
......

I think this is because their p values are lower than 40T. Factors below 40T are scored directly as p/T.

[MikeH]

I think this is because their p values are lower than 40T. Factors below 40T are scored directly as p/T.

Exactly. They will only be listed under the 'next to enter ...' if their score will change as a result.

Hope that helps.

[Joe O]

MikeH,
Could you explain this:

Most recent finds
    p (T)      k         n      Score     Factor found    Score changed  Score was    Score could be  Reqd bias
  152.922  67607  18205451      1.529  Wed 24-Dec-2003  Wed 24-Dec-2003                   158388.595           
  152.905  24737   3819631      1.529  Wed 24-Dec-2003  Wed 24-Dec-2003                     4182.782

Specifically, that there is a date under "Score changed" and this is the original appearance and score.

[Mystwalker]

@Mike:

Could you post a diagram showing the derivation of these graphs? I'd love to see how much performance gain we have. Maybe we can even see some client changes in it?

btw. seems like the lower bounds for DC PRPing is hard-coded to 400,000...

[Joe O]

Last Update: Fri 26-Dec-2003 15:03 (GMT)

Definitely Stuck!

No change as of 2003-12-28-1330 EST

[MikeH]

Specifically, that there is a date under "Score changed" and this is the original appearance and score.

Good question. No idea why, looks like a bug - possibly had something to do with the almost empty results.txt file, need to investigate further.

Definitely Stuck!

Sorry, been visiting family over Christmas. Looks like I had a power failure at 18:00 GMT 26 Dec.

Could you post a diagram showing the derivation of these graphs? I'd love to see how much performance gain we have. Maybe we can even see some client changes in it?

Planning to be out again Monday, but should get change before the year ends

btw. seems like the lower bounds for DC PRPing is hard-coded to 400,000

I've got the code ready, I just wasn't sure when Louie and Dave were going to reactive supersecret, so should be live soon.

[Keroberts1]

Does anyone have a AMD 3200? Just wondering what type of speed they get?

[MikeH]

Specifically, that there is a date under "Score changed" and this is the original appearance and score.

As I suspected it was due to the almost empty results.txt file. As a result of this it was unable to figure out the 90% point properly on the next run, so anything new on that run were scored as negative numbers, so when all was restored to normal the change in score was significant enough for it to log a score changed date. I think that's a long way of saying it's a bug

Could you post a diagram showing the derivation of these graphs? I'd love to see how much performance gain we have. Maybe we can even see some client changes in it?

The attached graph shows the daily change in the various points. I've averaged over 5 days and removed the 100%, 98% 95% and 85% to make things at little smother and more readable.

Don't know why we had a big dip just over two months ago. Any ideas?

I'm now time stamping the data, so any graphs of this type in the future should be more acturate

[ceselb]

Thanks for the nice graphs, very interesting.

I've capped the DC sieve to 75T, could you close out completed ranges above that? Some people have posted about it in the DC thread .

[MikeH]

Hi ceselb,

I've sent my set of n <3M p <75G to Joe O, I think he has more up to date info than me, but really I need everyone that started one of these ranges to either declare as [complete] or [abandoned], that way everything will be neat and tidy.

Another thing. Caould you add another page view of the holes?

Based on http://www.aooq73.dsl.pipex.com/gap..._n3_20_p05u.htm

I'd like something like this (One row per not completed range/user pair):

( 90.00G) : 121210-121300 expmod8
( 81.36G+ 183.05G) : 121300-122000 Slatz
( 150.02G+ 81.40G) : 122000-123000 Mystwalker

How about this (go to the bottom of the page). Is this roughly what you're after?

[ceselb]

Really nice, thank you very much.

[MikeH]

Really nice, thank you very much.

You're welcome.

Now for the n<3M ranges over 75T. With JoeO's help we have narrowed down the following ranges where we don't know whether they have (or will be) completed with the 300K < n < 20M sob.dat, or the new 1M < n < 20M sob.dat.

134000	136000	Mystwalker
137000	139000	Mystwalker
139800	140500	Slatz
140500	143000	cmprince
143050	143250	rosebud
143250	145000	priwo
147300	148000	Mystwalker
148800	149500	expmod8
149500	150000	Mystwalker
151000	152000	Mystwalker

All of the above are still active in the main sieve effort. If you are named above, can you please let me know whether you intend to complete the range with a 300K - 20M sob.dat, or if you have switched to 1M - 20M and you know when you switched can you let me know like this (e.g.)

134000 134500 Mystwalker [complete]
134500 136000 Mystwalker [abandoned]

or if you don't know where you switched (it really isn't a big deal), just like this

134000 136000 Mystwalker [abandoned]

[ceselb]

Mirrored here.

[Mystwalker]

Just declared my DC ranges abandoned in the other thread...

Mike:
The first two links below "The Sieving clients" point to the SoBSieve client, whereas they are titled as prothsieve...

[MikeH]

Mike:
The first two links below "The Sieving clients" point to the SoBSieve client, whereas they are titled as prothsieve...

Thanks Mystwalker ( how long has it been like that?)

[Keroberts1]

As long as i can remember i believe. I thought i made a mention bout it a while ago but never found a reply. However, I can't find my post so maybe it was so late i sumbited the thread wrong. Late night is often when i find time ot visit the forums.

[Slatz]

my range is being completed with the new dat file.

I believe i switched right around 140100

so
139800-140100 complete
140100-140500 abandoned

Slatz

[hc_grove]

I noticed that 6354798418167061747|27653*2^5998713 scored 35 points = as a duplicate factor, but I can't find any other factors for n=5998713 in the result files (checked both the regular and the one with duplicate and excluded factors marked)?

[Mystwalker]

I don't think that's a duplicate.
It's merely still outside of the active window. Normally, this would give you p / 100T points. Put as the 90% sieve point is lower, it will be cut.
But I don't memorize the according equation. It should be somewhere in this thread, though.

[hc_grove]

Originally posted by Mystwalker
I don't think that's a duplicate.
It's merely still outside of the active window. Normally, this would give you p / 100T points. Put as the 90% sieve point is lower, it will be cut.

That's what I expected too, but 90% sieve point / 100T is around 1.8 and not 35 (and it doesn't change as the 90% sieve point grows).

Looking at the scoring rules on Mike's page, I found no other reason why the score would be exactly 35 than it being scored as a duplicate.

[Mystwalker]

Ok, found the scoring conventions. Well, I confused the threads.

It is here

A unique factor will score as follows:

p < 40T, score = p/1T (i.e. as before)
p > 40T, in 'active' window, 0 PRP tests performed, score = (n/1M ^ 2) * 125
p > 40T, in 'active' window, 1 PRP tests performed, score = (n/1M ^ 2) * 125 * 0.6
p > 40T, in 'active' window, 2 PRP tests performed, score = (n/1M ^ 2) * 125 * 0.2
p > 40T, outside 'active' window, score = (as duplicate, see below)

A duplicate factor will score as follows:

score = p/100T, capped at 35.

Thus, you factor is either a duplicate or outside the active window. I guess it's the latter...

[MikeH]

I noticed that 6354798418167061747|27653*2^5998713 scored 35 points

Don't worry it'll get it's moment of glory in about 5-6 weeks time. Might even manage to hold the top spot as the highest scoring factor under 6M (which I'm guessing is what you were after).

....that's provided we don't find a prime for k=27653 before then.

[hc_grove]

Originally posted by MikeH
Don't worry it'll get it's moment of glory in about 5-6 weeks time. Might even manage to hold the top spot as the highest scoring factor under 6M (which I'm guessing is what you were after).

Correct. (And ceselb: I'm going to complete a sensibly looking range to avoid making the coordination page too confusing).

....that's provided we don't find a prime for k=27653 before then.

Yeah, that's the problem with primes, they are bad for sieving scores.