39.801T 55459 6030298 39.801
Sun 25-Apr-2004
frodo42
Shouldn't this have been found a long time ago?
I ran the siever on this range and found the factor very quickly
39801050200619 | 55459*2^6030298+1
either it was lost or Joe O who was supposed ot have originally sieved this range did not submitted the factor/never sieved the range.
Also there is the much more terrifying possibility that the older version of the siever that was used to sieve this range was missing some factors.
How did the factorer find this factor if it doesn't searc hthe presieved regions? Shouldn't this region of the sieve have been seperated from the region of possible factors searched in by the factorer. Many questions hopefully someone will have an answer Joe O do yo ustill have files that show the factors you found during that time period?
Last edited by Keroberts1; 04-25-2004 at 05:54 PM.
The factoring algorithm finds all smooth factors regardless of size, so if Joe O forgot to submit this factor (or much worse if the siever missed it), the factorer will discover it because it's smooth. The only thing depth parameter is used for in the factorer is determining the optimal bounds.
There was a version quite some time ago that did miss about 1% of factors. That could have been when we were doing the 40T region. Thankfully said version was replaced quite quickly, so if this one does fall into this catagory, then there shouldn't be too many others like it.
I remember that version but i don't htink it was during that region that it was released. Perhaps Mklasson would have details about when it was released or perhaps JoeO could let us know exactly waht version he used on this range. This should probbly be read into a bit more. If anybody still has some of the older versions maybe they should test to see if they all find this particular factor.
Originally posted by Keroberts1
39801050200619 | 55459*2^6030298+1
either it was lost or Joe O <snip> never sieved the range.
Code:Sieve 300K<n<3M Range size FactorsU FactorsD FactorsE Score 39700 - 39900 200 0 3 13 1.195
Joe O
Hi Joe,
I guess he's talking about:
Sieve 3M<n<20M Range size FactorsU FactorsD FactorsE Score
39700 - 39900 200 18 16 129 722.064
where, the expected number of unique factors is ~120 (+-20) vs. the actual figure of 18.
I'm sending you a PM for details.
Cheers,
Nuri
I've made a quick analysis of the results file, and there seems to be there are 46 potential ranges with some holes (it's almost for sure that not all of those ranges contain real holes).
Some of these (like 3650-3700) have been recognized by us previously, and already taken care of (the reason it still shows up is the credit for the additional factors went to the patchers, thus the number of factors for the range on original user's record are still low).
Anyways, after a quick glimpse, it seems like there are two emergin patterns in holes.
1. There are many potential holes (12 of 46) in the 32T-43T area, from various users. I think this might be because of the faulty client version. (~300 potentially missing factors here)
2. For most of the potential holes (8 out of 11) above 78T, there is an enormous number of excluded factors submitted by the users. This reminds of me that, may be in some of those ranges, some of us mistakenly submitted the contents of factexcl.txt and/or factrange.txt instead of fact.txt, or simply forgot to submit fact.txt contents while planning to submit all three. (~230 potentially missing factors here)
Please note that, this is a qucik and dirty analysis. Also, even if all of the 46 suspected ranges (total size of 3.8T) turned out to be actual holes, the estimated number of facotrs we've missed is ~1,300.
I'll take a look at the most promising ones in detail and let you know if I find out anything interesting (PS: might take time).
I know that MikeH has checked a few ranges for holes semi-secretly in the past. You might want to check with him on what he has done to avoid duplication of effort.
Done as part of :
Completed 7/21/03 10:35 am (EDT) with SobSieve 1.28 or 1.30 as one of the first "dual range" sieves.Code:Sieve 3M<n<20M Range size FactorsU FactorsD FactorsE Score 39700 - 39900 200 18 16 129 722.064
Joe O
so if the cient that missed the factors was sobsieve then is it possible that alot of previously sieved ranges may be missing factors? 1300 factors is far to many to risk missing. Joe is your PC overclocked? Perhaps it was just a one tiome glitch that cause this one factor to be missed.
Thanks for the reminder ceselb. I am also aware of that. In 3650-3700 fo example, we patched the holes together with him. When I wrote, "I'll take a look at the most promising ones in detail ", I meant the analysis (i.e. the possible reason of the hole signal - like faulty client or user error or anything else -, if they are already patched by someone etc.). So, I won't be starting to sieve any ranges right away. And when I start, I'll check not to duplicate work. Traces of pathes can be easily tracked through the results.txt file.I know that MikeH has checked a few ranges for holes semi-secretly in the past. You might want to check with him on what he has done to avoid duplication of effort.
I think the 1300 figure covers more than 80% of the potential risk of missing factors. Most of the ranges (95% in terms of number of reservations, and 98% in terms of the total range sieved) seem to appear just fine. I agree 1300 is out of a 3.8T range is like a treasure (though I'm pretty sure significant number of the signals were wrong and we'll end up with much less recovered factors). Still, I'll definitely look into it as I have time.so if the cient that missed the factors was sobsieve then is it possible that alot of previously sieved ranges may be missing factors? 1300 factors is far to many to risk missing.
Joe, I've checked your sobstatus file against the results.txt file. All of the 179 factors were submitted. However, almost all of them turned out to be excluded or duplicate factors. See attached file for details. The factors on the left are from your sobstatus file, and the ones on the right are from the results.txt file. The only difference is 39801050200619 55459 6030298 584 61, which started this tread.Completed 7/21/03 10:35 am (EDT) with SobSieve 1.28 or 1.30 as one of the first "dual range" sieves.
If this is the case for other ranges with many excluded factors, then some of us might have had a faulty client problem for a certain period of time.
well being as I'd love to find a region of the sieve that low that still has factors as soon as a list is made public of which regions has this suspicious look to them I'm be happy to devote a portion of my resources to double checking them. I can't guarentee I'll finish the mhowever if i don't find anything after 100 or so G. At this range I would expect the density ot be around one factor every few G at least. Am i correct?
Yes, that's valid for most of the ranges. Most of the suspected ranges should normally have densities of 0.5 F/G to 1.5 F/G.
However, all of the ranges already have some factors. Total number of normally expected factors less the total of already submitted ones is 1300. That suggests a density of one missing unique factor per 3G on the average. But, of course this should not be taken for granted.
On the double check sieve issue, I do not want any of our resources to be wasted on false alarms. Therefore, please wait for a few days for a more detailed analysis. I hope to find enough time for it during weekend. And of course, in case there really happens to be some ranges that need to be resieved, I'd be more than happy to share the information and the ranges with you all.
No, it was an AMD K6-III/400 that was running at stock. It ran hot from day one, even with an extra case fan. I had to turn it off on hot summer afternoons, or the thermal warning would throttle it down anyway. It has been retired.Originally posted by Keroberts1
Joe is your PC overclocked?
By the way, it passed all stress tests and memtest86+ even till the end.
Joe O
I'll rerun my own ranges when we get to the bottom of this. And I won't stop after a 100G!Originally posted by Keroberts1
a list is made public of which regions has this suspicious look to them I'm be happy to devote a portion of my resources to double checking them.
Joe O
@ Joe: Your ranges that I PMed to you seems to be by far the most promising ones among all. Especially the smallest two of your ranges would yield a good density of missing factors. I guess it would be a wise move to check them. Still, it might be a good idea to hear Mike's opinion on those six ranges as well, before starting a rerun.
On the other potential ranges:
I just found some time today to analyse the ranges a bit further, and noticed that Mike has resieved a particular user's ranges that we have previously suspected, which brought us some of the missing factors.
Taking out that user's ranges, Joe's ranges (which will be sieved by himself), and a few more ranges (due to various reasons, i.e. is not worth resieving due to low potential outcome per G, etc.) we're left with:
- # ranges with possible holes: 26 (dropped from 46)
- total size: 1.3T (dropped from 3.8T)
- maximum* suspected missing factors: 520 (dropped from 1300)
* it's highly likely that the actual figure will be much less.
I'll post more info as I look into other ranges.
Appologies, I haven't been following this thread.
For the record, I a fairly sure that I haven't resieved any of Joe's ranges, so any problems that exist won't have been addressed by me.
You may already know, but one good place to look for a complete record of the factors found is the "all users page". If you take a look at the ranges complete towards the bottom, you see number of factors found by the person that the reserved the range, and in (), the number of factors found by others.
If you're already using this info, sorry for duplication.
Cheers,
Mike.
Hey it's me who found that factor
I've been reading this thread a few times, but it was first now when I started wondering why 39801050200619 | 55459*2^6030298+1 scored so low that it came to my attention that this tread is about that factor (among other things ...)
Just for information it was found using a factor value of 1.5 and 2^47 as what I thought meant lower value ...
I was assuming it was the way you explained here, but was not 100% sure. So, thanks for the clarification.Originally posted by MikeH
If you're already using this info, sorry for duplication.
Anyways, I've eliminated a few more ranges, thus expected number of missing factors and the size of suspicious ranges dropped a bit more.
After that, I started resieving a little portion of each to be sure there exists at least some missing factors.
Here are the results so far:
Original range: 9990 - 9999
User: alexr
Expected: 18 factors
Actual: 12 factors
I checked the whole range, and found four new factors (9998153020439 | 24737*2^11804071+1, 9998204220791 | 24737*2^13269751+1, 9998353852393 | 10223*2^9640217+1, and 9998894415151 | 22699*2^3891430+1).
PS: I dunno why, but they did not show up in Mike's stats (may be they were duplicates??).
Original range: 13200 - 13220
User: ceselb
Expected: 30 factors
Actual: 22 factors
I checked 13210-13220, and found one new factor (13210014234479 | 33661*2^2080032+1).
13200-13210 is available for resieve if anyone is interested. Just drop a note below. EDIT: Taken by Keroberts1
Original range: 22640 - 22700
User: netbrian
Expected: 50 factors
Actual: 31 factors
I checked 22640-22650, and found one new factor (22644560072831 | 10223*2^16939481+1).
22650-22700 is available for resieve if anyone is interested. Just drop a note below.
PS: I dunno why, but it did not show up in Mike's stats (may be it was a duplicate??).
Original range: 26750 - 26800
User: netbrian
Expected: 35 factors
Actual: 15 factors
I checked 26750-26770, and found two new factors (26760255161203 | 10223*2^15506729+1 and 26763155273761 | 10223*2^13811357+1).
26770-26800 is available for resieve if anyone is interested. Just drop a note below.
Original range: 27000 - 27050
User: netbrian
Expected: 35 factors
Actual: 12 factors
I checked 27000-27010, and found one new factor (27001164464707 | 10223*2^3406205+1).
27010-27050 is available for resieve if anyone is interested. Just drop a note below.
I'll post more later.
I also want to note that it's not for sure that you'll find factors. Please use your own discretion.
Last edited by Nuri; 05-01-2004 at 07:07 PM.
Original range: 215000 - 215050
User: BigMajaX
Expected: 5 factors
Actual: 0 unique, 1 duplicate and 60 excluded factors
I did not check this range, and I do not encourage anyone to do so. But, BigMajaX, if you're reading this, could you please check again if you've sent in the contets of the fact.txt file for this range?
13200-13210 I'll check this one and seeif any thing shows
Original range: 215000 - 215050
User: BigMajaX
Expected: 5 factors
Actual: 0 unique, 1 duplicate and 60 excluded factors
yes, i have submitted all the factors found (and a 50G gap with a celeron 400 is quite unforgettable!)
that's why I submitted the excluded too, to "proove" that the range has been sieved... (but I usually do it aniway)
but feel free to resieve the range...
Bye.
That was a real bad luck for you. Thanks for the quick reply.Originally posted by MJX
yes, i have submitted all the factors found (and a 50G gap with a celeron 400 is quite unforgettable!)
27010-27050 i will do the check priwo
13200-13210 nothing there
maybe this one has more promise I'll let ya know how it does
26770-26800 Keroberts1
Last edited by Keroberts1; 05-02-2004 at 02:25 PM.
Originally posted by Nuri
Original range: 13200 - 13220
User: ceselb
Expected: 30 factors
Actual: 22 factors
I checked 13210-13220, and found one new factor (13210014234479 | 33661*2^2080032+1).Well spotted, but wrong user.
I based them on 3M<n<20M ranges reserved here and here.Originally posted by ceselb
Well spotted, but wrong user.
It's no big deal anyway. I think we've done a pretty neat job in sieving. There were a few (around 5) ranges which were exceptionally problematic in the very early phases of the sieve, which were already spotted and taken care of. Almost all of the ranges we're discussing here still have significantly less factors than expected, but as I said, it's no big deal.
Still, for the sake of completion, I'll post more ranges if I encounter one (or more) new factors in 10G-20G checks.
Last edited by Nuri; 05-02-2004 at 06:15 PM.
I'll check 22650-22700
To make sure there is no misunderstanding, these are all regular sieve ranges (i.e. analysis is based on 3m<n<20m range results), not DC sieve ranges.
On a second note: In my opinion, there is no need to go back to the old 3m<n<20m dat file. It would be just fine to use the 11k dat file (1m<n<20m) for resieve.
I found 3 new factors in my range adn i believe all of the ranges that were ot be rechecked have been finished. Are there more suspects?
27010-27050 priwo [complete]
27033331433257 | 10223*2^16790057+1
27040054113179 | 4847*2^6040791+1
27049393456193 | 10223*2^17251337+1
all of them were new !
Yes, there are. But I'm late for work.Originally posted by Keroberts1
I found 3 new factors in my range adn i believe all of the ranges that were ot be rechecked have been finished. Are there more suspects?Will post them when I come back.
Here are some more ranges.
Original range: 16250 - 16270
User: ceselb
Expected: 24 factors
Actual: 17 factors
I checked 16260-16270, and found two new factors (16262672861041 | 4847*2^11161167+1, and 16264171281383 | 4847*2^8343327+1).
16250-16260 is available for resieve if anyone is interested. Just drop a note below.
Original range: 17150 - 17250
User: ceselb
Expected: 110 factors
Actual: 69 factors
I checked 17240-17250, and found one new factor (17243318223409 | 4847*2^2100807+1).
17150-17240 is available for resieve if anyone is interested. Just drop a note below.
Original range: 17250 - 17260
User: priwo
Expected: 11 factors
Actual: 7 factors
I checked 17250-17260, and found two new factors (17252317641127 | 4847*2^16845231+1, and 17256592523101 | 10223*2^7531517+1).
There's nothing left here.
Original range: 21520 - 21605
User: louie
Expected: 75 factors
Actual: 46 factors
I checked 21575-21585, and found two new factors (21576157491797 | 10223*2^10729481+1, and 21578615512453 | 10223*2^15173801+1).
21520-21575 and 21585-21605 are available for resieve if anyone is interested. Just drop a note below.
There are a few more ranges left. And as before, I'll post them if I find something interesting.
22650-22700 completed with 7 new factors.
16250-16260 Death [reserved]
16250959881113 | 10223*2^2105129+1
16253133403439 | 10223*2^7522361+1
both not new. 10h left.
Last edited by Death; 05-06-2004 at 03:36 AM.
wbr, Me. Dead J. Dona \
reserving
17150-17240 priwo
21585-21605 royanee
did you mean low n sieving or 3 < n < 20?
I use this file - SobDat_n1M-20M.zip for checking this range? should I download another file for this?
wbr, Me. Dead J. Dona \
Originally posted by Death
did you mean low n sieving or 3 < n < 20?
I use this file - SobDat_n1M-20M.zip for checking this range? should I download another file for this?
Use the 1m-20m file. It is the correct one to use.
So far, I've resieved 5G, 10G or 20G portions of a total of 22 ranges, a total of 215G.
6 ranges out of these 22 portions did not yield any new factor.
16 ranges out of these 22 portions have yielded at least one new factor.
I have already announced 9 of the 16. An additional 3 of the 16 were Joe's ranges. And the remaining 4 of the 16 are those I have not yet announced (I am waiting for 21520-21575 to be taken by someone before announcing them. If you feel 21520-21575 is too long for you, simply grab a portion of it and please also write the remaining portion as "available for resieve" to avoid confusion).
Please note: The fact that I am finding at least one new factor in all of the ranges I am posting is a strong evidence that there was a problem (either client or user related) which resulted in at least one missing factor. However, this should not be taken as a 100% guarantee that there will be more factors in the remaining portion of the range.
21585-21605 completed with 2 new factors
21520-21550 royanee
This means that 21550-21570 is available for resieve
Posting Permissions
Reply With Quote
