mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Marin's Mersenne-aries (https://www.mersenneforum.org/forumdisplay.php?f=30)
-   -   Strategic Double Clicking (https://www.mersenneforum.org/showthread.php?t=20372)

Mark Rose 2015-07-30 17:20

[QUOTE=Madpoo;406761]These 2 are still unassigned and need a quint check if anyone is feeling bold:
36742943
46013437[/QUOTE]

I matched you for both.

chalsall 2015-07-30 17:33

[QUOTE=Mark Rose;406906]Yeah, I've found a few factors recently:[/QUOTE]

Just for those interested, the "Strategic DCTF'ers" have completed and returned to Primenet 990 candidates, and found 21 factors.

Madpoo 2015-07-30 18:01

[QUOTE=Mark Rose;406906]Yeah, I've found a few factors recently:

[url=http://www.mersenne.org/report_exponent/?exp_lo=46332389&full=1]46332389[/url]
[url=http://www.mersenne.org/report_exponent/?exp_lo=47549387&full=1]47549387[/url]
[url=http://www.mersenne.org/report_exponent/?exp_lo=50218603&full=1]50218603[/url]
[url=http://www.mersenne.org/report_exponent/?exp_lo=50375903&full=1]50375903[/url]
[url=http://www.mersenne.org/report_exponent/?exp_lo=50556641&full=1]50556641[/url]
[url=http://www.mersenne.org/report_exponent/?exp_lo=52293907&full=1]52293907[/url]
[url=http://www.mersenne.org/report_exponent/?exp_lo=55612153&full=1]55612153[/url]
[url=http://www.mersenne.org/report_exponent/?exp_lo=57909959&full=1]57909959[/url]

Not all of these show up in that post #17 list. I did have some previous assignments, but I don't keep track of my work too closely, and I read there may have been additional lists sent through PM.[/QUOTE]

Nice. Looking at those reminds me again that the "verified (factored)" status is pretty misleading, especially in the cases where 2 mismatched residues appear.

I get that "verified" in this case refers to it's composite-ness, but it's in the LL section and those residues are most definitely NOT verified. :smile:

Oh well... in those cases there's still no way to know if they're bad or good so for my purposes I just ignore it, like the LL never took place, but in cases where there are matching residues and it was factored, it'd be nice to count those as actually "good" results.

Something to tackle later, I guess. Like how I look at mismatches and if one is by an awesome computer, I pretend the other one is bad. I could at least look at these verified but factored and tick the "good" counts up a notch on the computers involved. Might prevent a handful of systems from looking as bad as they do now.

chalsall 2015-07-30 23:56

So, my third batch of 12 from the list of 47 completed today. Eight out of twelve [U][I]matched[/I][/U].

I almost feel disappointed.... :wink:

Madpoo 2015-07-31 01:14

[QUOTE=chalsall;406931]So, my third batch of 12 from the list of 47 completed today. Eight out of twelve [U][I]matched[/I][/U].

I almost feel disappointed.... :wink:[/QUOTE]

Yeah, I know the feeling. I was doing 2nd checks on a few where it was a little iffy if the computer was bad or not (like 1 or 2 bad, no good, but a lot of unknowns). I thought I'd spot check a few of their other ones and see if I could strike gold ... as in find a machine with lots of unknowns that were probably all bad. :smile: Unfortunately most of those were matching... oh well.

I compared it to prospecting... finding a single gold flake or nugget and digging around for a vein. LOL I may have found one but it only has a few more unknowns below 58M, and the rest are all really high up in the 60M-80M range for some reason. I picked a handful of 70M exponents to double-check, so I guess I'll know in a few more days.

Mark Rose 2015-07-31 21:36

Above 53M is done.

One more factor found, for [url=http://www.mersenne.org/report_exponent/?exp_lo=53218939&full=1]M53218939[/url].

Madpoo 2015-07-31 22:45

[QUOTE=Mark Rose;406994]Above 53M is done.

One more factor found, for [url=http://www.mersenne.org/report_exponent/?exp_lo=53218939&full=1]M53218939[/url].[/QUOTE]

Thanks again for doing factoring on those. Since my goal is to work on doing triple-checks of exponents where needed, finding factors on these > 50M exponents really helps quite a bit.

Once I get going on those, I may need some factoring help for those specifically. :smile:

chalsall 2015-07-31 23:06

[QUOTE=Madpoo;407001]Once I get going on those, I may need some factoring help for those specifically. :smile:[/QUOTE]

Feel free to generate and post another list (or several).

The "Strategic DCTF'ers" should have everything currently in the system completed in a week to ten days from now.

Edit: Oh, and let me know if you want them done before, after or concurrent to the current work.

Madpoo 2015-08-01 17:02

Another one that needs a quint check, if anyone was interested:
[URL="http://www.mersenne.org/M40582921"]M40582921[/URL]

It's kind of strange to run across these, or even the ones that now need a quad check. Just strange that 2 or 3 other people checked the same exponent and had a bad run.

I mean... what are the odds? I guess if we assume a residue is bad 3% of the time, the odds of 3 different people getting 3 bad results on the same exponent are .03^3 right? 0.0027 %. And that's assuming my 3% error rate is close.

Well, that's still like 2.7 out of every 100,000 exponents, so yeah, I guess they'll show up. I don't remember if there were any where 4+ (different) people got different residues... I think there might have been. There are a few doozies where the same user kept trying over and over and getting different results, like:
[URL="http://www.mersenne.org/M2397103"]M2397103[/URL]
[URL="http://www.mersenne.org/M2513153"]M2513153[/URL]

Oh chatmate...you tried so hard.

I guess there are a few with 4 bad ones from totally different users before 2 matched. All of the 5+ bad ones were from the same user goofing up repeatedly:
[URL="http://www.mersenne.org/M7021433"]M7021433[/URL]

That's the only one I could find. Pretty slim odds of that happening naturally I guess.

Madpoo 2015-08-01 18:02

[QUOTE=chalsall;407005]Feel free to generate and post another list (or several).[/QUOTE]

Here's a list of candidates for double-checking. Same criteria... 3 times as many bad as good. Some are new based on our improved findings, some are exponents that had been assigned to someone else but expired, and I think some are from previous lists but never got picked up by anyone?

Once we've exhausted these 3x bad/good work, we can do 2x bad/good and there's another 115 of those, give or take. Having twice as many bad as good is still a pretty good indicator that their unknowns are also going to be bad, but I wanted to get the low hanging fruit first. :smile:
[CODE]exponent Bad Good Unk Sus worktodo
37452047 3 0 3 0 DoubleCheck=37452047,71,1
42369917 7 2 5 0 DoubleCheck=42369917,72,1
42469309 4 1 8 1 DoubleCheck=42469309,72,1
42509921 39 9 13 18 DoubleCheck=42509921,72,1
42519247 15 3 5 4 DoubleCheck=42519247,72,1
42522107 10 3 12 1 DoubleCheck=42522107,72,1
42574673 7 2 2 4 DoubleCheck=42574673,72,1
42635947 4 1 2 1 DoubleCheck=42635947,72,1
42636029 4 1 2 1 DoubleCheck=42636029,72,1
42650897 13 3 5 3 DoubleCheck=42650897,72,1
42774437 4 1 8 1 DoubleCheck=42774437,72,1
42802219 14 2 4 8 DoubleCheck=42802219,72,1
42903143 4 1 8 1 DoubleCheck=42903143,72,1
43352831 3 0 3 1 DoubleCheck=43352831,72,1
43370659 4 1 12 0 DoubleCheck=43370659,72,1
52732837 3 0 5 2 DoubleCheck=52732837,73,1
52741279 3 0 5 2 DoubleCheck=52741279,73,1
52806191 17 2 29 14 DoubleCheck=52806191,73,1
52806323 17 2 29 14 DoubleCheck=52806323,73,1
52881599 3 0 5 2 DoubleCheck=52881599,73,1
53205589 17 2 29 14 DoubleCheck=53205589,73,1
53269589 3 0 5 2 DoubleCheck=53269589,73,1
53357987 17 2 29 14 DoubleCheck=53357987,73,1
53389849 17 2 29 14 DoubleCheck=53389849,73,1
54537607 6 0 12 4 DoubleCheck=54537607,73,1
54812297 3 0 4 3 DoubleCheck=54812297,73,1
54880909 17 2 29 14 DoubleCheck=54880909,73,1
54881429 17 2 29 14 DoubleCheck=54881429,73,1
55533703 4 0 6 0 DoubleCheck=55533703,73,1
55612019 6 1 4 3 DoubleCheck=55612019,73,1
55813103 4 0 6 0 DoubleCheck=55813103,73,1
55813129 4 0 6 0 DoubleCheck=55813129,73,1[/CODE]

Mark Rose 2015-08-01 18:24

All those are sufficiently trial factored. Can you post the 2x bad list, too? I think Chris is looking for exponents for GPU72 as the current work there will be done in a week.

chalsall 2015-08-01 18:58

[QUOTE=Mark Rose;407059]I think Chris is looking for exponents for GPU72 as the current work there will be done in a week.[/QUOTE]

That's correct -- anything which might be Strategically double or triple (et al) checked, but not yet appropriately TF'ed.

Mark Rose 2015-08-02 19:52

Above 52M is done. [url=http://www.mersenne.org/report_exponent/default.php?exp_lo=52036211&full=1]M52036211[/url] has a factor.

chalsall 2015-08-04 22:14

[QUOTE=Madpoo;407053]Here's a list of candidates for double-checking. Same criteria... 3 times as many bad as good. Some are new based on our improved findings, some are exponents that had been assigned to someone else but expired, and I think some are from previous lists but never got picked up by anyone?[/QUOTE]

OK, the work I had previously reserved is soon to be finished, so I picked those from this list not yet assigned / completed (Greg had taken a few of the lower ones).

frmky 2015-08-05 00:10

Yes, we need more! :smile:

chalsall 2015-08-07 20:42

[QUOTE=frmky;407252]Yes, we need more! :smile:[/QUOTE]

Just to do a "ping"...

The SDCTF'ers will be mostly done in a few days (there are a few stranglers who will not return their results for a bit longer). The following query gives some idea as to the situation. A\d\d is available for assignment, O\d\d is out for work.

[CODE]+------+------+------+------+------+------+------+-------+----------+----------+----------+
| A70 | O70 | A71 | O71 | A72 | O72 | -- | Avail | Assigned | Factored | Returned |
+------+------+------+------+------+------+------+-------+----------+----------+----------+
| 0 | 241 | 160 | 463 | 0 | 2 | -- | 160 | 706 | 54 | 2633 |
+------+------+------+------+------+------+------+-------+----------+----------+----------+[/CODE]

Aaron, is there any more TF'ing to put into the system for processing? Separately, I don't believe there is any DC'ing work currently posted here for processing not already claimed.

TLDR: Time for some more lists.... :smile:

Madpoo 2015-08-08 00:44

[QUOTE=chalsall;407412]TLDR: Time for some more lists.... :smile:[/QUOTE]

Sorry, I was on vacation for the past week. Had a lot of yummy results from my systems when I got back. Results from doing needed triple-checks, which resulted in lots more systems having an extra "bad" check in their tally.

Here's a new list of the exponents done by systems with 3+ times as many bad as good. 33 on the list. I guess some might be newly expired and available work and some could be from the new data.

[CODE]exponent Bad Good Unk Sus worktodo
36223069 4 1 1 4 DoubleCheck=36223069,71,1
36409657 3 0 1 0 DoubleCheck=36409657,71,1
37560389 7 0 8 3 DoubleCheck=37560389,71,1
37825399 24 1 1 0 DoubleCheck=37825399,71,1
40133851 4 0 1 1 DoubleCheck=40133851,72,1
40819813 5 0 6 3 DoubleCheck=40819813,72,1
42206569 3 0 1 3 DoubleCheck=42206569,71,1
42951593 17 4 4 3 DoubleCheck=42951593,72,1
42960023 15 4 4 4 DoubleCheck=42960023,72,1
42973891 5 1 4 0 DoubleCheck=42973891,72,1
42984763 20 4 5 3 DoubleCheck=42984763,72,1
43008263 16 1 4 1 DoubleCheck=43008263,72,1
43165501 17 4 4 3 DoubleCheck=43165501,72,1
43171423 20 4 5 3 DoubleCheck=43171423,72,1
43477741 11 3 14 7 DoubleCheck=43477741,72,1
43522733 15 4 4 4 DoubleCheck=43522733,72,1
43696487 9 2 1 10 DoubleCheck=43696487,72,1
44259209 3 0 3 1 DoubleCheck=44259209,72,1
45285043 6 1 5 3 DoubleCheck=45285043,70,1
45923573 6 1 5 3 DoubleCheck=45923573,72,1
45940067 5 1 4 0 DoubleCheck=45940067,72,1
46086427 6 1 5 3 DoubleCheck=46086427,72,1
46561309 4 1 3 0 DoubleCheck=46561309,72,1
46583549 4 0 1 5 DoubleCheck=46583549,72,1
47713291 41 9 13 18 DoubleCheck=47713291,72,1
49342991 41 9 13 18 DoubleCheck=49342991,72,1
50060831 41 9 13 18 DoubleCheck=50060831,73,1
50818837 41 9 13 18 DoubleCheck=50818837,73,1
51138553 3 0 2 0 DoubleCheck=51138553,73,1
51138569 3 0 2 0 DoubleCheck=51138569,73,1
51521941 3 0 5 2 DoubleCheck=51521941,73,1
52069961 18 2 28 14 DoubleCheck=52069961,73,1
52078909 18 2 28 14 DoubleCheck=52078909,73,1[/CODE]

Mark Rose 2015-08-08 03:04

[QUOTE=Madpoo;407422]
[CODE]exponent Bad Good Unk Sus worktodo
42206569 3 0 1 3 DoubleCheck=42206569,71,1
45285043 6 1 5 3 DoubleCheck=45285043,70,1[/CODE][/QUOTE]

Only these two need more trial factoring. I'll have that done in a jiffy (approximately 3 hours).

Madpoo 2015-08-08 07:05

[QUOTE=Mark Rose;407430]Only these two need more trial factoring. I'll have that done in a jiffy (approximately 3 hours).[/QUOTE]

That reminds me, do you have a basic chart of how high you prefer to factor exponents up to, for things like this?

For example, there are currently 5,017 unassigned exponents that have had 2 checks without a match.

If I knew how high you like to TF things for different exponent ranges, I could spit out a list of only the exponents needing additional TF work.

Since one of my goals is to work through all of these that need triple checking, it would be great to weed out as many by factoring them first. Not that I alone think I could get through all 5000+ of them, of course. :smile:

Some info along the lines of:
[CODE] < 35M = TF to 72 bits
35M - 40M = TF to 73 bits (or whatever)
[/CODE]
etc.

Mark Rose 2015-08-08 07:12

[QUOTE=Madpoo;407445]That reminds me, do you have a basic chart of how high you prefer to factor exponents up to, for things like this?[/quote]

Yeah, it's pretty easy. Ideal DCTF after an LL test is:

< 40M TF to 71
< 50M TF to 72
< 65M TF to 73
< 80M TF to 74

[quote]
For example, there are currently 5,017 unassigned exponents that have had 2 checks without a match.

If I knew how high you like to TF things for different exponent ranges, I could spit out a list of only the exponents needing additional TF work.[/quote]

I think that's what we're looking for. I just grabbed the last of the TF work for the Strategic Double Checks, so we could use that immediately.

chalsall 2015-08-08 14:51

OK, I've taken 37560389, 37825399, 40133851 and 40819813 so they don't end up being assigned to the "churners".

Actually, the middle two were already assigned, but I'm going to run them any way as they're a 97% chance they won't be completed by the assignee, and a ~50% chance (?) my result won't match the original.

Brian-E 2015-08-08 16:35

[QUOTE=chalsall;407464][...] 97% chance they won't be completed by the assignee [...][/QUOTE]
I love it.
"Ninety-seven percent!! It's the highest mark in the history of Botswana Secretarial College!! ..."
Anyway, it's amazing what you can do these days, isn't it? Who would have thought we'd now be in a position to measure people's commitment to their project-assignments like this and put a nice little number by it?

Sorry, I'm in a silly mood at the moment....:smile:

Madpoo 2015-08-08 17:21

[QUOTE=Mark Rose;407447]Yeah, it's pretty easy. Ideal DCTF after an LL test is:

< 40M TF to 71
< 50M TF to 72
< 65M TF to 73
< 80M TF to 74

I think that's what we're looking for. I just grabbed the last of the TF work for the Strategic Double Checks, so we could use that immediately.[/QUOTE]

Well, let's try this on for size... of the 5000+ there were 103 I saw that could probably use extra TF before anyone does any triple-checks on them. These are all currently unassigned... as assignments expire this list might grow.

There are a couple above 80M, so I don't know what those optimal TF levels would be... I just put them at 75 for a placeholder. Those are actually the only 2 (unassigned) mismatches above 80M. There's an additional one, 100000267, assigned to someone... current TF on that is 72 bits. Feel free to include it as well... assigned a year ago and no progress on it. :smile:

So that "bonus" one would be: Factor=100000267,72,75

[CODE]exponent TF WorkToDo
42159511 71 Factor=42159511,71,72
44336701 70 Factor=44336701,70,72
44337253 70 Factor=44337253,70,72
44339507 70 Factor=44339507,70,72
44651921 71 Factor=44651921,71,72
44704837 70 Factor=44704837,70,72
44705611 70 Factor=44705611,70,72
44706691 70 Factor=44706691,70,72
44710943 70 Factor=44710943,70,72
44712163 70 Factor=44712163,70,72
44713567 70 Factor=44713567,70,72
44717383 70 Factor=44717383,70,72
44720633 70 Factor=44720633,70,72
44724431 70 Factor=44724431,70,72
44726953 70 Factor=44726953,70,72
44743691 70 Factor=44743691,70,72
44751017 70 Factor=44751017,70,72
44757269 70 Factor=44757269,70,72
44763937 70 Factor=44763937,70,72
44768851 70 Factor=44768851,70,72
44778887 70 Factor=44778887,70,72
44791667 70 Factor=44791667,70,72
44791843 70 Factor=44791843,70,72
44795549 70 Factor=44795549,70,72
44796079 70 Factor=44796079,70,72
44797163 70 Factor=44797163,70,72
44798423 70 Factor=44798423,70,72
44798993 70 Factor=44798993,70,72
44800403 70 Factor=44800403,70,72
44800699 70 Factor=44800699,70,72
44802697 70 Factor=44802697,70,72
44809421 70 Factor=44809421,70,72
44814577 70 Factor=44814577,70,72
44816593 70 Factor=44816593,70,72
45250057 70 Factor=45250057,70,72
45252049 70 Factor=45252049,70,72
45289007 70 Factor=45289007,70,72
45302069 70 Factor=45302069,70,72
45370063 70 Factor=45370063,70,72
45385061 70 Factor=45385061,70,72
45400009 70 Factor=45400009,70,72
45429023 70 Factor=45429023,70,72
45561077 70 Factor=45561077,70,72
45596077 70 Factor=45596077,70,72
45633011 70 Factor=45633011,70,72
46078079 70 Factor=46078079,70,72
46082123 70 Factor=46082123,70,72
46123043 70 Factor=46123043,70,72
46163063 70 Factor=46163063,70,72
46179127 70 Factor=46179127,70,72
46210061 70 Factor=46210061,70,72
46218091 70 Factor=46218091,70,72
46273063 70 Factor=46273063,70,72
46304149 70 Factor=46304149,70,72
46400093 70 Factor=46400093,70,72
46428149 70 Factor=46428149,70,72
46441133 70 Factor=46441133,70,72
46457161 70 Factor=46457161,70,72
46471057 70 Factor=46471057,70,72
46503073 70 Factor=46503073,70,72
46504067 70 Factor=46504067,70,72
46535077 70 Factor=46535077,70,72
46562107 70 Factor=46562107,70,72
46568033 70 Factor=46568033,70,72
46618127 70 Factor=46618127,70,72
46622057 70 Factor=46622057,70,72
46654703 70 Factor=46654703,70,72
47027159 70 Factor=47027159,70,72
47079113 70 Factor=47079113,70,72
47125153 70 Factor=47125153,70,72
47133089 70 Factor=47133089,70,72
47148029 70 Factor=47148029,70,72
47485057 70 Factor=47485057,70,72
47488183 70 Factor=47488183,70,72
47505083 70 Factor=47505083,70,72
47540161 70 Factor=47540161,70,72
47573129 70 Factor=47573129,70,72
47590111 70 Factor=47590111,70,72
47606099 70 Factor=47606099,70,72
47711161 70 Factor=47711161,70,72
47732101 70 Factor=47732101,70,72
47754067 70 Factor=47754067,70,72
47842153 70 Factor=47842153,70,72
48014177 70 Factor=48014177,70,72
48019087 70 Factor=48019087,70,72
48021059 70 Factor=48021059,70,72
48027143 70 Factor=48027143,70,72
48042073 70 Factor=48042073,70,72
48044081 70 Factor=48044081,70,72
48062081 70 Factor=48062081,70,72
48151063 70 Factor=48151063,70,72
48154063 70 Factor=48154063,70,72
48178093 70 Factor=48178093,70,72
48318203 70 Factor=48318203,70,72
48367097 70 Factor=48367097,70,72
48517037 70 Factor=48517037,70,72
48531139 70 Factor=48531139,70,72
48572059 70 Factor=48572059,70,72
48657031 70 Factor=48657031,70,72
49319059 70 Factor=49319059,70,72
50317901 71 Factor=50317901,71,73
89985911 71 Factor=89985911,71,75
100139483 73 Factor=100139483,73,75[/CODE]

Mark Rose 2015-08-08 17:59

I'll grab the two oddballs at the end and the one with no progress. I'll leave the rest to be added to the GPU72 hopper.

Though that's less than 1600 GHz-d of work, or only a few hours of GPU72 throughput. Please, sir, I want some more *holds bowl out*

chalsall 2015-08-08 18:29

[QUOTE=Mark Rose;407486]I'll leave the rest to be added to the GPU72 hopper.[/QUOTE]

OK, these are now in the "hopper".

[QUOTE=Mark Rose;407486]Though that's less than 1600 GHz-d of work, or only a few hours of GPU72 throughput. Please, sir, I want some more *holds bowl out*[/QUOTE]

Indeed, it's not much more work. On the other hand, perhaps this is all the TF'ing needed for the SDC'ing.

manfred4 2015-08-08 20:38

If we all just want a bigger List, just to do the TF on you could maybe just run a query to get all exponents from users, who have 1 or more wrong/suspect results just to have a (possibly) bigger buffer of assignments now.

chalsall 2015-08-08 21:36

[QUOTE=manfred4;407496]...who have 1 or more wrong/suspect results just to have a (possibly) bigger buffer of assignments now.[/QUOTE]

Not a bad idea. On the other hand, we might have already (mostly) done this.

Aaron... What says you?

chalsall 2015-08-09 18:10

[QUOTE=chalsall;406771]Gosh... If LaurV doesn't get back in the game soon you're quickly going to pass him in the stats! :wink:[/QUOTE]

And [URL="https://www.gpu72.com/reports/workers/dctf/"]you've done it![/URL] Congrats! :smile:

Mark Rose 2015-08-09 18:24

[QUOTE=chalsall;407541]And [URL="https://www.gpu72.com/reports/workers/dctf/"]you've done it![/URL] Congrats! :smile:[/QUOTE]

I'm finally #1 at something lol

I'm sure LaurV will pass me once he gets all his cards running again.

Next on my list is the guy ahead of me in the Overall Progress charts. :smile:

kladner 2015-08-09 19:42

[QUOTE=chalsall;407541]And [URL="https://www.gpu72.com/reports/workers/dctf/"]you've done it![/URL] Congrats! :smile:[/QUOTE]

It is amazing that Pete is still at number 3.

Madpoo 2015-08-09 20:36

[QUOTE]If we all just want a bigger List, just to do the TF on you could maybe just run a query to get all exponents from users, who have 1 or more wrong/suspect results just to have a (possibly) bigger buffer of assignments now.[/QUOTE]

[QUOTE=chalsall;407502]Not a bad idea. On the other hand, we might have already (mostly) done this.[/QUOTE]

Let me take a look at that. Maybe just to open things up for the TF'ers out there I can really loosen things up and look for cases where a CPU has half as many bad as good, or something like that. I'll play with the numbers a bit until I can work up a good size list of candidates for extra TF work.

I've focused initially on cases where the work a CPU did is > 75% chance of being wrong, but when you get right down to it I'd be happy to find cases where there's a >25% chance that the original was bad. In the most general case there's a ~3% chance of a result being bad, so being able to find cases where the failure rate is higher than average is all good.

chalsall 2015-08-09 20:48

[QUOTE=Madpoo;407555]Let me take a look at that. Maybe just to open things up for the TF'ers out there I can really loosen things up and look for cases where a CPU has half as many bad as good, or something like that. I'll play with the numbers a bit until I can work up a good size list of candidates for extra TF work.[/QUOTE]

That would be good. Like I said before, let's get these off the books so the SDC'ers can drill down as the statistics are refined.

At the end of the day these all have to be TF'ed anyway, so we might as well prioritize those who might get special DC'ing attention in the near(ish) future.

chalsall 2015-08-09 20:53

[QUOTE=kladner;407552]It is amazing that Pete is still at number 3.[/QUOTE]

Indeed. A real shame.... :sad:

Madpoo 2015-08-09 21:16

1 Attachment(s)
[QUOTE=chalsall;407559]That would be good. Like I said before, let's get these off the books so the SDC'ers can drill down as the statistics are refined.

At the end of the day these all have to be TF'ed anyway, so we might as well prioritize those who might get special DC'ing attention in the near(ish) future.[/QUOTE]

Let's try this on for size. I included work where the cpu had half as many bad as good (or at least one bad and zero good). Then filtered down by which ones could use additional TF. I didn't limit it to the < 58M range like I've been doing with strategic DC, so there's a few higher ones... not sure what the ideal TF bit levels are for those so adjust accordingly.

The list is 1702 entries long, most just needing an extra bit, some needing two or more.

chalsall 2015-08-09 21:51

[QUOTE=Madpoo;407561]The list is 1702 entries long, most just needing an extra bit, some needing two or more.[/QUOTE]

Cool. Thanks. Imported and being assigned. This should keep us busy for about a week.

I didn't bring in anything above 58M as I need to check the code to make sure it can handle that; we've never DCTF'ed above 60M before.

Mark Rose 2015-08-09 22:24

It would be nice if I could get GPU72 credit for the higher ones... ;)

I'll start with the couple above 80M manually.

chalsall 2015-08-09 22:34

[QUOTE=Mark Rose;407566]It would be nice if I could get GPU72 credit for the higher ones... ;) I'll start with the couple above 80M manually.[/QUOTE]

Post here what you're working. This will ensure there's no duplication of effort, and when I double check the code I'll not import what you've taken, and will "inform" the system so you get the credit.

Mark Rose 2015-08-09 22:38

[QUOTE=chalsall;407568]Post here what you're working. This will ensure there's no duplication of effort, and when I double check the code I'll not import what you've taken, and will "inform" the system so you get the credit.[/QUOTE]

Oh, I grabbed just these two from the list Aaron posted:

Factor=80060801,72,75
Factor=123456811,73,75

No need to "inform" the system.

Madpoo 2015-08-11 15:23

Running LL tests on factored exponents...
 
So, call me crazy, but...

I looked up all of the exponents that had 2+ LL tests run but no matches, and were then later factored. I figured I could whip out a 3rd verifying test on these and see which of the other runs was a loser, thus being able to count that result as bad.

There's actually quite a few and I don't think I'll be testing all of them, at least not now, but I at least did all of them below 10M (about a hundred) and then any over 10M that had 3+ mismatched results (there were about 20 of those).

There's still another 500 or so higher than 10M with just a pair of mismatched results... since those take longer to run I may just look at that down the road as a possibility.

In many of the cases, I'm sure the CPU that did the bad result doesn't have any work in the unknown/suspect category, so if I do any of the larger ones I'll probably want to weed out only the ones that would help in that regard.

So it was kind of a fun overnight project for a handful of systems... still running a few more through the pipes but it's nearly done.

You should now see some entries like:
[URL="http://www.mersenne.org/M35097967"]M35097967[/URL]

I think in general if an exponent was verified *before* a factor was found, the "bad" results were kept, and it was everything else that just shows up as "verified (factored)"... in the sense that it's verified composite, not that the residue is verified. I just took my results and if it matched one of the others, marked the oddballs as bad. In the database on the backend there's no distinction of an actual verified by residue matching *and* factored result, unfortunately. Would be nice though... "factored and verified", "factored and unverified", "factored and suspect" would all be good distinctions for people like me who care about additional things beyond compositeness or not, like the reliability of systems involved. :smile:

Anyway, not sure if my efforts have generated any additional leads on potentially bad systems, but it was entertaining.

flagrantflowers 2015-08-11 18:25

I sent you a couple of private messages a while ago. I believe it was SCIENCEMAN88 that stated that you can verify is a residue is correct after being factored by doing some simple modular arithmetic. This would allow you to check all exponents that have later been factored very quickly and easily without having to run a very computationally intensive LL test.

science_man_88 2015-08-11 18:51

[QUOTE=flagrantflowers;407681]I sent you a couple of private messages a while ago. I believe it was SCIENCEMAN88 that stated that you can verify is a residue is correct after being factored by doing some simple modular arithmetic. This would allow you to check all exponents that have later been factored very quickly and easily without having to run a very computationally intensive LL test.[/QUOTE]

[URL="http://mersenneforum.org/showpost.php?p=406573&postcount=56"]post 56[/URL] already defeated it in the sense that only the last so many bits of the residue is stored so it's not completely failsafe.

chalsall 2015-08-11 19:33

[QUOTE=Madpoo;407674]In the database on the backend there's no distinction of an actual verified by residue matching *and* factored result, unfortunately. Would be nice though... "factored and verified", "factored and unverified", "factored and suspect" would all be good distinctions for people like me who care about additional things beyond compositeness or not, like the reliability of systems involved. :smile:[/QUOTE]

I know the primary goal of GI[U]MPS[/U] is to find MPs. But...

Your recent work has helped this by focusing on and eliminating possible missed primes by flaky machines.

Would it take much work to add such distinctions to the database (asks I, having no idea as to the Primenet DB schema nor codebase)?

Madpoo 2015-08-11 20:19

[QUOTE=chalsall;407688]Would it take much work to add such distinctions to the database (asks I, having no idea as to the Primenet DB schema nor codebase)?[/QUOTE]

Yeah, it should be possible, maybe without too much work.

Right now, each and every LL test can have one of five different states:
unverified
verified
bad
suspect
factored

Splitting that "factored" into subsets showing the *actual* status of the residue itself would just mean adding a couple additional things.

The trickier parts are updating the code that assigns status to work when a factor for already tested exponents comes in... in general these will almost always be one of two things:
1) factor found for an already verified set of tests ... all verified/bad results would retain that status and just add the "factored" part
2) factor found for unverified results ... it would say something like "factored (unverified)" or "factor (suspect)" if the original test was in fact suspect.

I think right now, if an exponent is verified and has one or more bad results, the bad stuff stays marked as "bad", and it's only the other stuff that shows up as "factored (verified)". There seems to have been a period of time maybe around 2011/2012 when the server wasn't keeping that, and simply marking everything as "factored (verified)"... I went back and found those oddballs and corrected about 150 or so to reflect their actual badness. (I verified each one, FYI... had to exclude those funky "Myrman" residues that were off by 1 that George kindly accepted as valid).

In summary, there are:
~ 37,000 exponents where it was verified before being factored
~ 17,300 exponents where there was a single LL test before being factored
~ 500 exponents that had 2+ mismatching results before being factored
~ 1850 exponents that were verified, but also had some bad results as well (which are now all properly marked as bad)

As you might guess, many of those 37,000 that were already verified have come about by the awesome folks taking smaller exponents to higher TF levels.

chalsall 2015-08-11 20:50

[QUOTE=Madpoo;407694]As you might guess, many of those 37,000 that were already verified have come about by the awesome folks taking smaller exponents to higher TF levels.[/QUOTE]

Sounds like a simple bitmap to me. Should fit in eight bits, but hell, let's use 32 since we're working with modern hardware....

Madpoo 2015-08-11 21:04

[QUOTE=chalsall;407695]Sounds like a simple bitmap to me. Should fit in eight bits, but hell, let's use 32 since we're working with modern hardware....[/QUOTE]

Implementing it on the database is easy enough (it's an INT value) but updating the code for handling incoming results is where it probably needs the most love.

In truth that could all be unchanged and then from time to time on the back end we could have something that auto-marks the *correct* status of factored LL tests. Then it would be some front end changes to properly interpret the new values for display on the various exponent report pages. In fact I think there may be a function involved so it might only be a "one and done" type of update, but I'd leave that for George/James to puzzle out. :smile:

chalsall 2015-08-11 21:18

[QUOTE=Madpoo;407697]Implementing it on the database is easy enough (it's an INT value) but updating the code for handling incoming results is where it probably needs the most love.[/QUOTE]

Been there; done that.

Usually not as easy as it first seems.

Prime95 2015-08-11 22:42

[QUOTE=Madpoo;407694]
Right now, each and every LL test can have one of five different states:
unverified
verified
bad
suspect
factored

Splitting that "factored" into subsets showing the *actual* status of the residue itself would just mean adding a couple additional things.

The trickier parts are updating the code that assigns status to work when a factor for already tested exponents comes in... in general these will almost always be one of two things:
1) factor found for an already verified set of tests ... all verified/bad results would retain that status and just add the "factored" part
2) factor found for unverified results ... it would say something like "factored (unverified)" or "factor (suspect)" if the original test was in fact suspect.[/QUOTE]

Easier would be to create a SQL view that splits the factored state into factored/verified and factored/unverified by doing a subquery looking for a matching residue. Yes, it is slower. But, you are only going to use this view in your bad machine SQL queries and two you don't have to find all the PHP code that looks at the state or maintains the state.

Madpoo 2015-08-11 23:12

[QUOTE=Prime95;407705]Easier would be to create a SQL view that splits the factored state into factored/verified and factored/unverified by doing a subquery looking for a matching residue. Yes, it is slower. But, you are only going to use this view in your bad machine SQL queries and two you don't have to find all the PHP code that looks at the state or maintains the state.[/QUOTE]

Yeah, my current query to do that is slow (but then I'm doing it for the entire set of "result state=factored" in one swell foop)

I thought it might be of some general purpose use for the public, to know if a result was good/bad/unknown even though the # was factored. Eyeballing the details for each exponent would let someone do the same thing, so I guess it's more about programmatic uses...

It's interesting that the vast majority of cases where LL tests exist but it was factored involve actual verified (by double check) residues (as a total of the exponents).

What that really breaks down to is that 37K exponents times at least 2 test / exponent will give us ~75K new "good" result counts, and leave just 17K or so unknowns, plus a smattering of some ~ 4K bad.

Having the known bad ones actually marked as bad is a help... it did actually snag a few extra computers into my net, resulting in some additional exponents < 58M we could DC early. If I'm able to count the good ones as "good" in terms of pushing forward a potential winner in cases where mismatches already occur, it could expose a few more to light.

For now I'll take your advice and do a query of the "good" computers and adjust the "good" counts + 1 when they're part of a matching factored exponent. Whether it makes sense to have that kind of visibility on the site itself, I can only guess that some folks might find it useful? :smile:

chalsall 2015-08-11 23:41

[QUOTE=Madpoo;407709]Yeah, my current query to do that is slow (but then I'm doing it for the entire set of "result state=factored" in one swell foop)[/QUOTE]

So. We have a solution space.

Madpoo 2015-08-12 15:23

Easy pickings...
 
If anyone is looking for what I consider to be low exponent "first time" assignments, I hand selected a few that I'm almost sure were done wrong the first time:
[CODE]exponent Bad Good Unk Sus worktodo
41174801 22 0 8 8 DoubleCheck=41174801,72,1
46583549 4 0 1 5 DoubleCheck=46583549,72,1
44259209 4 0 3 1 DoubleCheck=44259209,72,1
47525339 4 0 3 1 DoubleCheck=47525339,72,1
48014221 3 0 1 3 DoubleCheck=48014221,72,1
51521941 3 0 5 2 DoubleCheck=51521941,73,1[/CODE]

That many bad, zero good, and only a few unknowns (plus some suspect results to boot) tells me that none of those were done right the first time. :smile:

Besides the ones above (which had 3+ bad, zero good, and only single digit uknowns) there's quite a few others I've been looking at. Things like "2 bad, zero good, 1 unknown and varying amounts of suspect results", but I'm less certain about those...

For myself, I'm doing a little fishing expedition... I picked about 5 machines that had 1 bad, zero good, and a LOT of uknowns (20+). I'm wondering what the odds are that their one and only double-checked result so far happened to be bad and the rest are fine... probably not that great.

So I'm running an extra check of one of their other results just to see if I can spot any trends... I may find a honey hole (or two) of some machine that spat out bad results and we just haven't discovered it yet.

Meanwhile, I'll generate another list of "strategic double checks" in a bit... maybe 5 more hours once my fishing expedition gets some results back.

Prime95 2015-08-12 18:27

[QUOTE=Madpoo;407741]If anyone is looking for what I consider to be low exponent "first time" assignments,[/QUOTE]

I took the first 4.

Madpoo 2015-08-12 20:44

Here's another that I'm almost certain was wrong the first time (it had one bad already and another one just came up *probably* bad... this is the only one left from this particular cpu):
[URL="http://www.mersenne.org/M35008399"]M35008399[/URL]
DoubleCheck=35008399,71,1

Madpoo 2015-08-12 20:53

[QUOTE=Madpoo;407777]Here's another that I'm almost certain was wrong the first time (it had one bad already and another one just came up *probably* bad... this is the only one left from this particular cpu)...[/QUOTE]

And a few more that are probably bad based on some new results I just found:
DoubleCheck=36357427,71,1
DoubleCheck=41397511,71,1
DoubleCheck=43072709,71,1

Mark Rose 2015-08-12 21:18

[QUOTE=Madpoo;407778]And a few more that are probably bad based on some new results I just found:
DoubleCheck=41397511,71,1
DoubleCheck=43072709,71,1[/QUOTE]

I'll TF these immediately.

Edit: done. No factors found. New worktodo lines:

DoubleCheck=41397511,72,1
DoubleCheck=43072709,72,1

Madpoo 2015-08-12 22:18

[QUOTE=Mark Rose;407781]I'll TF these immediately.

Edit: done. No factors found. New worktodo lines:

DoubleCheck=41397511,72,1
DoubleCheck=43072709,72,1[/QUOTE]

Thanks...

It occurred to me, what if someone tests one of these and finds a hidden prime? Should I get a co-discoverer credit along with GIMPS? LOL

I'm kidding, but it would still be fun if we found something this way. I still think it's highly possible that there's a hidden prime lurking in there, waiting for a double check.

Madpoo 2015-08-13 03:35

Here's another batch... some have been listed before, I think? And just nobody started an LL test on it or whatever.

Basic set of 3X as many bad as good, or zero good and 3+ bad.

The twist with this list is that it's only including work from cpu's with <= 5 unknowns, so in my mind that says there's a better chance these would be bad too... not a big pool of possibly good out there. :smile:

[CODE]exponent Bad Good Unk Sus worktodo
36249911 3 1 1 1 DoubleCheck=36249911,71,1
36495967 3 1 1 0 DoubleCheck=36495967,71,1
36532159 3 1 3 0 DoubleCheck=36532159,71,1
36615611 3 1 2 0 DoubleCheck=36615611,71,1
36893677 3 1 1 0 DoubleCheck=36893677,71,1
37018711 3 1 2 0 DoubleCheck=37018711,71,1
37369373 11 3 5 9 DoubleCheck=37369373,71,1
37596089 3 1 3 0 DoubleCheck=37596089,71,1
41319587 3 1 2 0 DoubleCheck=41319587,72,1
41772461 3 1 1 0 DoubleCheck=41772461,72,1
42211633 3 1 3 6 DoubleCheck=42211633,72,1
43696487 9 2 1 10 DoubleCheck=43696487,72,1
45040949 3 1 4 3 DoubleCheck=45040949,72,1
45285043 6 1 5 3 DoubleCheck=45285043,72,1
45923573 6 1 5 3 DoubleCheck=45923573,72,1
45940067 5 1 4 0 DoubleCheck=45940067,72,1
46086427 6 1 5 3 DoubleCheck=46086427,72,1
46144303 3 1 4 3 DoubleCheck=46144303,72,1
46561309 4 1 3 0 DoubleCheck=46561309,72,1
48014221 3 0 1 3 DoubleCheck=48014221,72,1
48050389 4 1 3 0 DoubleCheck=48050389,72,1
49280723 3 1 4 3 DoubleCheck=49280723,72,1
49385489 4 1 3 0 DoubleCheck=49385489,72,1
51521941 3 0 5 2 DoubleCheck=51521941,73,1[/CODE]

Madpoo 2015-08-13 03:42

And here's a second list... like the one just before, but from cpu's that had between 6 and 20 unknowns.

[CODE]exponent Bad Good Unk Sus worktodo
35821501 3 1 13 3 DoubleCheck=35821501,71,1
36378401 3 1 13 3 DoubleCheck=36378401,71,1
36591349 3 1 9 0 DoubleCheck=36591349,71,1
36635279 3 1 13 3 DoubleCheck=36635279,71,1
41436847 3 1 6 2 DoubleCheck=41436847,72,1
41482157 3 1 6 2 DoubleCheck=41482157,72,1
42316259 6 2 12 3 DoubleCheck=42316259,72,1
42333931 3 1 7 2 DoubleCheck=42333931,72,1
42832513 6 2 12 3 DoubleCheck=42832513,72,1
42838273 3 1 9 0 DoubleCheck=42838273,72,1
43445537 3 1 6 2 DoubleCheck=43445537,72,1
43695037 3 1 10 9 DoubleCheck=43695037,72,1
43935919 6 2 12 3 DoubleCheck=43935919,72,1
45139733 6 2 12 3 DoubleCheck=45139733,72,1
45676573 3 1 7 2 DoubleCheck=45676573,72,1
46709101 6 2 12 3 DoubleCheck=46709101,72,1
46954631 6 2 12 3 DoubleCheck=46954631,72,1
47106793 43 9 13 17 DoubleCheck=47106793,72,1
47713291 43 9 13 17 DoubleCheck=47713291,72,1
47743589 3 1 10 9 DoubleCheck=47743589,72,1
47780699 3 1 10 9 DoubleCheck=47780699,72,1
47786399 3 1 10 9 DoubleCheck=47786399,72,1
47787829 3 1 10 9 DoubleCheck=47787829,72,1
47843339 6 2 12 3 DoubleCheck=47843339,72,1
48659239 6 2 12 3 DoubleCheck=48659239,72,1
49235143 3 1 7 2 DoubleCheck=49235143,72,1
49342991 43 9 13 17 DoubleCheck=49342991,72,1
49360579 43 9 13 17 DoubleCheck=49360579,72,1
49501717 6 2 12 3 DoubleCheck=49501717,72,1
50060831 43 9 13 17 DoubleCheck=50060831,73,1
50364731 3 1 10 9 DoubleCheck=50364731,73,1
50364827 3 1 10 9 DoubleCheck=50364827,73,1
50364943 3 1 10 9 DoubleCheck=50364943,71,1
50368427 6 2 12 3 DoubleCheck=50368427,73,1
50532067 6 2 12 3 DoubleCheck=50532067,73,1
50725429 3 1 8 0 DoubleCheck=50725429,73,1
50818837 43 9 13 17 DoubleCheck=50818837,73,1
51006671 3 1 7 2 DoubleCheck=51006671,73,1
53740301 3 1 8 0 DoubleCheck=53740301,73,1
54068747 3 1 8 0 DoubleCheck=54068747,73,1
54072163 3 1 8 0 DoubleCheck=54072163,73,1
54072253 3 1 8 0 DoubleCheck=54072253,73,1
54072643 3 1 8 0 DoubleCheck=54072643,73,1
55257889 3 1 12 4 DoubleCheck=55257889,73,1
57414373 3 1 12 4 DoubleCheck=57414373,73,1
57584141 3 1 12 4 DoubleCheck=57584141,73,1[/CODE]

Prime95 2015-08-13 03:54

[QUOTE=Madpoo;407741][code]
41174801 22 0 8 8 DoubleCheck=41174801,72,1
46583549 4 0 1 5 DoubleCheck=46583549,72,1
44259209 4 0 3 1 DoubleCheck=44259209,72,1
47525339 4 0 3 1 DoubleCheck=47525339,72,1
48014221 3 0 1 3 DoubleCheck=48014221,72,1
51521941 3 0 5 2 DoubleCheck=51521941,73,1[/CODE][/QUOTE]

I've now taken all of these though I don't know if they've had enough TF done.

Prime95 2015-08-13 03:56

I've taken those in posts #149 and #150, plus those in #153 below 45 million. Thanks, madpoo.

Mark Rose 2015-08-13 04:42

[QUOTE=Prime95;407797]I've now taken all of these though I don't know if they've had enough TF done.[/QUOTE]

They look appropriately TF'ed.

Madpoo 2015-08-15 22:35

This little gem just became available (old assignment expired). I'm fairly certain it was done wrong the first time.

[CODE]exponent Bad Good Unk Sus worktodo
41019989 22 0 8 8 DoubleCheck=41019989,72,1[/CODE]

endless mike 2015-08-16 01:16

[QUOTE=Madpoo;408045]This little gem just became available (old assignment expired). I'm fairly certain it was done wrong the first time.

[CODE]exponent Bad Good Unk Sus worktodo
41019989 22 0 8 8 DoubleCheck=41019989,72,1[/CODE][/QUOTE]

I'll take it.

S485122 2015-08-16 07:10

[QUOTE=Madpoo;408045][CODE]exponent Bad Good Unk Sus worktodo
41019989 22 0 8 8 DoubleCheck=41019989,72,1[/CODE][/QUOTE]I must have missed something : when I look at LL results or full detailed status for that exponent I see only one unverified LL test. Where are the other 37 LL attempts ? I have the same question about 41174801 and for 47106793 that is supposed to have had more thant 80 results.

Jacob

Brian-E 2015-08-16 09:06

[QUOTE=S485122;408064]I must have missed something [...][/QUOTE]
I thought I'd missed something too, but I think the solution to our confusion lies in the first post of this thread:
[QUOTE=Madpoo;406319][...]I'm doing an analysis of [I]computers[/I] to see which of them have had a sketchy track record, and return unusually high #'s of bad residues.[...][/QUOTE]
So those numerous bad or suspect tests have been done by the same machine which turned in the single test on that one exponent. But they were done on other exponents.

S485122 2015-08-16 10:06

[QUOTE=Brian-E;408069]So those numerous bad or suspect tests have been done by the same machine which turned in the single test on that one exponent. But they were done on other exponents.[/QUOTE]Eureka !

I should have read the beginning of the thread again. Thanks !

Jacob

frmky 2015-08-16 18:43

[QUOTE=Brian-E;408069]But they were done on other exponents.[/QUOTE]
Namely, not including those deemed bad due to mismatch with exceptional machines, 33353981, 33899219, 33995837, 34012283, 34051943, 34109443, 34219027, 34219061, 34239521, 34287101, 34360253, 34594279, 34668677, 34899911, 36648329, 37381459, 37462097, 37751843, and 38126657.

petrw1 2015-08-20 18:51

Looks like I hogged the TF assignments....
 
Anyway my DC-TF will all be completed Friday morning.
Seems I have all the remaining 43M.

airsquirrels 2015-08-21 14:53

If there are any more of these that need TF work done I'm happy to help.

chalsall 2015-08-21 15:46

[QUOTE=airsquirrels;408471]If there are any more of these that need TF work done I'm happy to help.[/QUOTE]

Aaron would have to speak to that, but I'm pretty sure we've cleared everything out.

There's still 75 to finish off (probably today), but so everyone knows... We collectively appropriately TF'ed 6,121 candidates, and factored 100 of them.

Madpoo 2015-08-22 15:08

[QUOTE=S485122;408064]I must have missed something : when I look at LL results or full detailed status for that exponent I see only one unverified LL test. Where are the other 37 LL attempts ? I have the same question about 41174801 and for 47106793 that is supposed to have had more thant 80 results.[/QUOTE]

Yes, as you saw, the number of good/bad/etc are a compilation of the track record of the [I]computer[/I] that did that LL test on each exponent. A way to gauge whether or not we can really trust the result that computer spit out.

Madpoo 2015-08-22 15:14

[QUOTE=frmky;408100]Namely, not including those deemed bad due to mismatch with exceptional machines...[/QUOTE]

Correct. In my analysis of each machine, not only do I tally up the known verified and known bad results for each system, but for the exponents where 2+ mismatching residues exist, I look at the systems that generated each of the residues... if one of those machines has a stellar track record (at least 50 known good results and 1 or fewer bad results) then I'm counting the other result from the other system as bad.

Of course we won't know for sure if it was really bad or not, but generally I've seen that machines that consistently return good results are just that... consistently good.

Machines that return bad results tend to be consistently bad, but it's not always a clear case of "every result you turn in is wrong". In fact, the machines that spew bad residues do get a few correct ones in there. It may be a result of the machine being fixed (bad RAM replaced, wonky overclocking set back to safer values, etc) or maybe it's environmental (better results in the winter when ambient temps are cooler, and bad results in the summer when the fan can't cool your strangely overclocked box). Who knows... I'm guessing. :smile: But the end result is that even the worst systems might occasionally get one right.

Madpoo 2015-08-22 15:19

[QUOTE=chalsall;408477]Aaron would have to speak to that, but I'm pretty sure we've cleared everything out.

There's still 75 to finish off (probably today), but so everyone knows... We collectively appropriately TF'ed 6,121 candidates, and factored 100 of them.[/QUOTE]

I just got back from a nice vacation, so I'll take a look at what we have... I checked in several dozen results from my triple-checks of 34M exponents that needed them. Could be a few new systems that are now considered "bad enough" for inclusion. There are also some assignments that expired so we can try to grab some and see if they need any additional TF work.

Ah... well, here's a list of exponents that [B]need a triple-check[/B] but need some more TF love. Might get lucky and find a factor, eliminating that TC:
[CODE]exponent TF WorkToDo
45819299 70 Factor=45819299,70,72
45821879 70 Factor=45821879,70,72
47035171 70 Factor=47035171,70,72
48140119 70 Factor=48140119,70,72
48214139 70 Factor=48214139,70,72
48225179 70 Factor=48225179,70,72
48340183 70 Factor=48340183,70,72
48434123 70 Factor=48434123,70,72
48611183 70 Factor=48611183,70,72
48648199 70 Factor=48648199,70,72
48685183 70 Factor=48685183,70,72
50293949 71 Factor=50293949,71,73
51111941 71 Factor=51111941,71,73
51360853 71 Factor=51360853,71,73[/CODE]

Madpoo 2015-08-22 15:21

[QUOTE=Madpoo;408524]...here's a list of exponents that [B]need a triple-check[/B] but need some more TF love. Might get lucky and find a factor, eliminating that TC...[/QUOTE]

And here's a list of "suspect" results where some additional TF could be handy:
[CODE]exponent TF WorkToDo
43441361 70 Factor=43441361,70,72
43912217 70 Factor=43912217,70,72
43920089 70 Factor=43920089,70,72
44412197 70 Factor=44412197,70,72
44497403 70 Factor=44497403,70,72
44649079 70 Factor=44649079,70,72
44649329 70 Factor=44649329,70,72
44861221 70 Factor=44861221,70,72
44877403 70 Factor=44877403,70,72
44894191 70 Factor=44894191,70,72
44899973 71 Factor=44899973,71,72
45009593 70 Factor=45009593,70,72
45011831 70 Factor=45011831,70,72
45014471 70 Factor=45014471,70,72
45097189 70 Factor=45097189,70,72
45134189 70 Factor=45134189,70,72
45321187 70 Factor=45321187,70,72
45410191 70 Factor=45410191,70,72
45632999 71 Factor=45632999,71,72
45639193 70 Factor=45639193,70,72
45652237 70 Factor=45652237,70,72
46000949 70 Factor=46000949,70,72
46000987 70 Factor=46000987,70,72
46004983 70 Factor=46004983,70,72
46157291 70 Factor=46157291,70,72
46217657 71 Factor=46217657,71,72
46251281 70 Factor=46251281,70,72
46296221 70 Factor=46296221,70,72
46355293 70 Factor=46355293,70,72
46367329 70 Factor=46367329,70,72
46404997 71 Factor=46404997,71,72
46438913 70 Factor=46438913,70,72
46444301 70 Factor=46444301,70,72
46468049 70 Factor=46468049,70,72
46487227 70 Factor=46487227,70,72
46520293 70 Factor=46520293,70,72
46523201 70 Factor=46523201,70,72
46525229 70 Factor=46525229,70,72
46920089 70 Factor=46920089,70,72
47035169 70 Factor=47035169,70,72
47070227 70 Factor=47070227,70,72
47071253 70 Factor=47071253,70,72
47180167 70 Factor=47180167,70,72
47486983 70 Factor=47486983,70,72
47499113 70 Factor=47499113,70,72
47584337 70 Factor=47584337,70,72
47705291 70 Factor=47705291,70,72
47939503 70 Factor=47939503,70,72
48034853 70 Factor=48034853,70,72
48099251 70 Factor=48099251,70,72
48099301 70 Factor=48099301,70,72
48167233 70 Factor=48167233,70,72
48386189 70 Factor=48386189,70,72
48405233 70 Factor=48405233,70,72
48427231 70 Factor=48427231,70,72
48529279 70 Factor=48529279,70,72
48686327 70 Factor=48686327,70,72
49045063 70 Factor=49045063,70,72
49364213 70 Factor=49364213,70,72
49376213 70 Factor=49376213,70,72
49520011 70 Factor=49520011,70,72
49528057 70 Factor=49528057,70,72
50013547 72 Factor=50013547,72,73
50035823 70 Factor=50035823,70,73
50306941 71 Factor=50306941,71,73
50436173 70 Factor=50436173,70,73
50616283 72 Factor=50616283,72,73
50810173 71 Factor=50810173,71,73
50962609 71 Factor=50962609,71,73
51008873 71 Factor=51008873,71,73
51226867 71 Factor=51226867,71,73
51291859 71 Factor=51291859,71,73
51696443 72 Factor=51696443,72,73
52597933 71 Factor=52597933,71,73
52672157 70 Factor=52672157,70,73
52764557 72 Factor=52764557,72,73
52798831 72 Factor=52798831,72,73
53069911 72 Factor=53069911,72,73
53221277 72 Factor=53221277,72,73
54375773 72 Factor=54375773,72,73
54712289 72 Factor=54712289,72,73
54828667 71 Factor=54828667,71,73
54869987 72 Factor=54869987,72,73
55027309 72 Factor=55027309,72,73
55715117 72 Factor=55715117,72,73
55984871 72 Factor=55984871,72,73
56020703 71 Factor=56020703,71,73
56020733 71 Factor=56020733,71,73
56020819 71 Factor=56020819,71,73
56114419 72 Factor=56114419,72,73
56340967 72 Factor=56340967,72,73
56443847 72 Factor=56443847,72,73
56826013 72 Factor=56826013,72,73
57014579 72 Factor=57014579,72,73
57143551 72 Factor=57143551,72,73
72116249 73 Factor=72116249,73,74
73000217 73 Factor=73000217,73,74
73000237 73 Factor=73000237,73,74
73000241 73 Factor=73000241,73,74
73000267 73 Factor=73000267,73,74
73000273 73 Factor=73000273,73,74
73000289 73 Factor=73000289,73,74
73000471 73 Factor=73000471,73,74
73000549 73 Factor=73000549,73,74
73000561 73 Factor=73000561,73,74
73000583 73 Factor=73000583,73,74
73000607 73 Factor=73000607,73,74
73000619 73 Factor=73000619,73,74
73000639 73 Factor=73000639,73,74
73000687 73 Factor=73000687,73,74
73000883 73 Factor=73000883,73,74
73000913 73 Factor=73000913,73,74
73000927 73 Factor=73000927,73,74
73001119 73 Factor=73001119,73,74
73003673 73 Factor=73003673,73,74
73116257 73 Factor=73116257,73,74
73116607 73 Factor=73116607,73,74
76006829 73 Factor=76006829,73,74[/CODE]

kladner 2015-08-22 16:12

[QUOTE=Madpoo;408524]I just got back from a nice vacation, so I'll take a look at what we have... I checked in several dozen results from my triple-checks of 34M exponents that needed them. Could be a few new systems that are now considered "bad enough" for inclusion. There are also some assignments that expired so we can try to grab some and see if they need any additional TF work.

Ah... well, here's a list of exponents that [B]need a triple-check[/B] but need some more TF love. Might get lucky and find a factor, eliminating that TC:
[CODE]exponent TF WorkToDo
45819299 70 Factor=45819299,70,72
45821879 70 Factor=45821879,70,72
47035171 70 Factor=47035171,70,72
48140119 70 Factor=48140119,70,72
48214139 70 Factor=48214139,70,72
48225179 70 Factor=48225179,70,72
48340183 70 Factor=48340183,70,72
48434123 70 Factor=48434123,70,72
48611183 70 Factor=48611183,70,72
48648199 70 Factor=48648199,70,72
48685183 70 Factor=48685183,70,72
50293949 71 Factor=50293949,71,73
51111941 71 Factor=51111941,71,73
51360853 71 Factor=51360853,71,73[/CODE][/QUOTE]

I will TF this first lot.
EDIT: Running: split between 2 GTX580s. It looks like the 48Ms will take 45-47 minutes each

kladner 2015-08-23 11:19

[QUOTE=kladner;408533]I will TF this first lot.
EDIT: Running: split between 2 GTX580s. It looks like the 48Ms will take 45-47 minutes each[/QUOTE]

Finished overnight. Two factors. :smile:

M48611183 has a factor: 1735077409661214935753 [TF:70:71*:mfaktc 0.21 barrett76_mul32_gs]
M48340183 has a factor: 1314117775062544754639 [TF:70:71*:mfaktc 0.21 barrett76_mul32_gs]

S485122 2015-08-23 18:18

[QUOTE=Madpoo;408523]...
It may be a result of the machine being fixed (bad RAM replaced, wonky overclocking set back to safer values, etc) or maybe it's environmental (better results in the winter when ambient temps are cooler, and bad results in the summer when the fan can't cool your strangely overclocked box). Who knows... I'm guessing.
... :smile: But the end result is that even the worst systems might occasionally get one right.[/QUOTE]Or just one of your children having installed an "official and legit" phone synch application downloaded directly from the phone's own website !

Jacob

airsquirrels 2015-08-24 01:07

[QUOTE=Madpoo;408525]And here's a list of "suspect" results where some additional TF could be handy:
[CODE]exponent TF WorkToDo
43441361 70 Factor=43441361,70,72
43912217 70 Factor=43912217,70,72
43920089 70 Factor=43920089,70,72
44412197 70 Factor=44412197,70,72
44497403 70 Factor=44497403,70,72
44649079 70 Factor=44649079,70,72
44649329 70 Factor=44649329,70,72
44861221 70 Factor=44861221,70,72
44877403 70 Factor=44877403,70,72
44894191 70 Factor=44894191,70,72
44899973 71 Factor=44899973,71,72
45009593 70 Factor=45009593,70,72
45011831 70 Factor=45011831,70,72
45014471 70 Factor=45014471,70,72
45097189 70 Factor=45097189,70,72
45134189 70 Factor=45134189,70,72
45321187 70 Factor=45321187,70,72
45410191 70 Factor=45410191,70,72
45632999 71 Factor=45632999,71,72
45639193 70 Factor=45639193,70,72
45652237 70 Factor=45652237,70,72
46000949 70 Factor=46000949,70,72
46000987 70 Factor=46000987,70,72
46004983 70 Factor=46004983,70,72
46157291 70 Factor=46157291,70,72
46217657 71 Factor=46217657,71,72
46251281 70 Factor=46251281,70,72
46296221 70 Factor=46296221,70,72
46355293 70 Factor=46355293,70,72
46367329 70 Factor=46367329,70,72
46404997 71 Factor=46404997,71,72
46438913 70 Factor=46438913,70,72
46444301 70 Factor=46444301,70,72
46468049 70 Factor=46468049,70,72
46487227 70 Factor=46487227,70,72
46520293 70 Factor=46520293,70,72
46523201 70 Factor=46523201,70,72
46525229 70 Factor=46525229,70,72
46920089 70 Factor=46920089,70,72
47035169 70 Factor=47035169,70,72
47070227 70 Factor=47070227,70,72
47071253 70 Factor=47071253,70,72
47180167 70 Factor=47180167,70,72
47486983 70 Factor=47486983,70,72
47499113 70 Factor=47499113,70,72
47584337 70 Factor=47584337,70,72
47705291 70 Factor=47705291,70,72
47939503 70 Factor=47939503,70,72
48034853 70 Factor=48034853,70,72
48099251 70 Factor=48099251,70,72
48099301 70 Factor=48099301,70,72
48167233 70 Factor=48167233,70,72
48386189 70 Factor=48386189,70,72
48405233 70 Factor=48405233,70,72
48427231 70 Factor=48427231,70,72
48529279 70 Factor=48529279,70,72
48686327 70 Factor=48686327,70,72
49045063 70 Factor=49045063,70,72
49364213 70 Factor=49364213,70,72
49376213 70 Factor=49376213,70,72
49520011 70 Factor=49520011,70,72
49528057 70 Factor=49528057,70,72
50013547 72 Factor=50013547,72,73
50035823 70 Factor=50035823,70,73
50306941 71 Factor=50306941,71,73
50436173 70 Factor=50436173,70,73
50616283 72 Factor=50616283,72,73
50810173 71 Factor=50810173,71,73
50962609 71 Factor=50962609,71,73
51008873 71 Factor=51008873,71,73
51226867 71 Factor=51226867,71,73
51291859 71 Factor=51291859,71,73
51696443 72 Factor=51696443,72,73
52597933 71 Factor=52597933,71,73
52672157 70 Factor=52672157,70,73
52764557 72 Factor=52764557,72,73
52798831 72 Factor=52798831,72,73
53069911 72 Factor=53069911,72,73
53221277 72 Factor=53221277,72,73
54375773 72 Factor=54375773,72,73
54712289 72 Factor=54712289,72,73
54828667 71 Factor=54828667,71,73
54869987 72 Factor=54869987,72,73
55027309 72 Factor=55027309,72,73
55715117 72 Factor=55715117,72,73
55984871 72 Factor=55984871,72,73
56020703 71 Factor=56020703,71,73
56020733 71 Factor=56020733,71,73
56020819 71 Factor=56020819,71,73
56114419 72 Factor=56114419,72,73
56340967 72 Factor=56340967,72,73
56443847 72 Factor=56443847,72,73
56826013 72 Factor=56826013,72,73
57014579 72 Factor=57014579,72,73
57143551 72 Factor=57143551,72,73
72116249 73 Factor=72116249,73,74
73000217 73 Factor=73000217,73,74
73000237 73 Factor=73000237,73,74
73000241 73 Factor=73000241,73,74
73000267 73 Factor=73000267,73,74
73000273 73 Factor=73000273,73,74
73000289 73 Factor=73000289,73,74
73000471 73 Factor=73000471,73,74
73000549 73 Factor=73000549,73,74
73000561 73 Factor=73000561,73,74
73000583 73 Factor=73000583,73,74
73000607 73 Factor=73000607,73,74
73000619 73 Factor=73000619,73,74
73000639 73 Factor=73000639,73,74
73000687 73 Factor=73000687,73,74
73000883 73 Factor=73000883,73,74
73000913 73 Factor=73000913,73,74
73000927 73 Factor=73000927,73,74
73001119 73 Factor=73001119,73,74
73003673 73 Factor=73003673,73,74
73116257 73 Factor=73116257,73,74
73116607 73 Factor=73116607,73,74
76006829 73 Factor=76006829,73,74[/CODE][/QUOTE]

I'll run these

airsquirrels 2015-08-24 07:16

These are done, unfortunately no factors in this batch.

endless mike 2015-08-24 13:28

[QUOTE=Madpoo;408045]This little gem just became available (old assignment expired). I'm fairly certain it was done wrong the first time.

[CODE]exponent Bad Good Unk Sus worktodo
41019989 22 0 8 8 DoubleCheck=41019989,72,1[/CODE][/QUOTE]

[QUOTE=endless mike;408055]I'll take it.[/QUOTE]

Finished overnight; as expected, not a [URL="http://www.mersenne.org/report_exponent/?exp_lo=41019989&full=1"]match[/URL].

Mark Rose 2015-08-24 16:52

[QUOTE=endless mike;408680]Finished overnight; as expected, not a [URL="http://www.mersenne.org/report_exponent/?exp_lo=41019989&full=1"]match[/URL].[/QUOTE]

I'll run a triple check. Should be done next week.

Madpoo 2015-08-25 15:58

different shifts = important
 
And here's why a different shift count matters:
[URL="www.mersenne.org/M37126169"]M37126169[/URL]

Although technically I think this is one of the handful (under 100, I think) of results that got accidentally checked in twice. But if it weren't for looking at shift counts and requiring them to be different, it may well have been accepted as a verified double check... I mean, they came in a year apart. :smile:

Mark Rose 2015-08-25 17:04

endless_mike just beat me to taking that.

endless mike 2015-08-25 17:08

[QUOTE=Madpoo;408759]And here's why a different shift count matters:
[URL="www.mersenne.org/M37126169"]M37126169[/URL]

Although technically I think this is one of the handful (under 100, I think) of results that got accidentally checked in twice. But if it weren't for looking at shift counts and requiring them to be different, it may well have been accepted as a verified double check... I mean, they came in a year apart. :smile:[/QUOTE]

Woo, quad check, I'll take it.

[QUOTE=Mark Rose;408764]endless_mike just beat me to taking that.[/QUOTE]

I type slow

Madpoo 2015-08-25 19:57

my triple checking...
 
FYI, I started out figuring I'd just start doing the necessary triple checks for all the junk that had 2+ mismatching residues. I'm just about done with all of the 34M examples, although some current assignments expire here and there, and our strategic double-checks are finding more mismatches.

That's been helpful in finding some new systems that had bad results where we didn't really have a track record yet, but in many cases the bad system was already a known offender. :smile:

For the 35M+ stuff I thought I'd focus on exponents where both systems are an unknown quantity... Since there's no shortage of exponents with mismatches, I figured this would be the best way to approach it so that I'm actually contributing to the dataset of bad systems.

Any thoughts? Anyone besides myself interested in tackling some of these exponents where TC's are needed? I believe we've already sifted through and found any that need extra factoring (I can double check) so these would be good old LL tests, just to figure out who was right and who was wrong (or both wrong, as I've seen in many cases).

petrw1 2015-08-25 20:25

[QUOTE=Madpoo;408770] Anyone besides myself interested in tackling some of these exponents where TC's are needed?[/QUOTE]

I can take a few....not the fastest of hardware so would prefer in the lower range...
I have 6 cores what were fast 2 years ago that could each complete 1 in about 12 days.

I may have more fire power available later in the year if there is still work to be done on these.

Will you as in the past publish a list now and then or are there just too many?

If no list just send me 18 to test...

Wayne

ric 2015-08-25 21:16

WIP
 
[QUOTE=Madpoo;408770] Anyone besides myself interested in tackling some of these exponents where TC's are needed? .[/QUOTE]

In the last few weeks, I've been TC' and QC'ing a few dozens expos in the 34M6 to 34M8 region, mainly coming from you or Greg, with some more already queued up for completion. Unsurprisingly, all the completed ones confirm 2015's newest results. No big iron here, so my output is on average one check a day, or so. I plan to continue this way for some more time, so if you publish a priority list I can grab a few new ones from there.

Hth,
Ric

Madpoo 2015-08-25 22:22

[QUOTE=petrw1;408773]I can take a few....not the fastest of hardware so would prefer in the lower range...
I have 6 cores what were fast 2 years ago that could each complete 1 in about 12 days.

I may have more fire power available later in the year if there is still work to be done on these.

Will you as in the past publish a list now and then or are there just too many?

If no list just send me 18 to test...[/QUOTE]

I'm trying to nail down how best to query the data now.

I'll start on just the stuff below 36M. There's a total of 470 exponents < 36M that have had 2 checks so far with no matches (and are currently unassigned).

What I want to be able to do is narrow down which among those 470 involve *both* computers with not a lot of historical data, so that by doing the TC now we'll at least build up some stats on good/bad systems for those that are currently unknown.

My first stab at a query for that gave me ~ 340 or so exponents. I snagged about 50 of them (from 35.0M up to 35.3M) but before grabbing more I'm going to QA the list and make sure my query is okay.

If anyone is curious, there are currently 5310 *unassigned* exponents that need a triple check (I've been snagging and clearing the ones that need quad+ checks). But that excludes about 100 more where I'm one of the testers... I didn't want to assign myself any work I've already done. :smile: I have 28 exponents where I'm one of 3 contributors, and then I guess another 75'ish where I'm one of the two mismatched testers. I'm not terribly concerned about those since I've just been assuming mine is the right one, along with other "awesome" systems with low/zero bad results.

Looking at *assigned* work (which could be assigned for extra TF, or for someone already doing a triple check) there's another ~ 2000. 36 of those have 3 mismatched tests... those are the ones that, when they become available, I've been snatching up. It's fun to verify those because I'm marking at least *2* computers as having another bad result with only one test. :smile: Sometimes all 3 previous results are bad, and those are the ones I usually mention here so someone can do a quint check.

Madpoo 2015-08-25 22:47

First, an updated "strategic double check" list
 
Before delving into hunting bad systems by doing triple checks, I thought it was time to put out an updated list of first/only checks that may be bad.

There are some new entries in this list, thanks to a user "eskifullofbeer" that now has 3 bad entries, zero good, and 24 unknown. 2 of those "bad" are actually ones where I did another check and didn't match theirs, so I'm assuming mine is correct. Part of my fishing expedition, looking for systems without much to go on, but seemed like they might be bad.

We also have a new entrant into the "3 times as many bad as good" thanks to user "Arthur Townsend" and some recent tests I completed that bumped his bad count just over the threshold (again, more of my speculative testing of marginal looking boxes).

Those are the ones that have 17 bad, 5 good, 12 unknown.

Anyway... here's the list:
[CODE]exponent Bad Good Unk Sus worktodo
35597701 17 5 12 1 DoubleCheck=35597701,71,1
35751481 17 5 12 1 DoubleCheck=35751481,71,1
36477403 17 5 12 1 DoubleCheck=36477403,71,1
36611753 17 5 12 1 DoubleCheck=36611753,71,1
37373783 17 5 12 1 DoubleCheck=37373783,71,1
37558289 17 5 12 1 DoubleCheck=37558289,71,1
37665493 9 3 5 5 DoubleCheck=37665493,71,1
42069659 3 0 24 1 DoubleCheck=42069659,72,1
42305167 3 0 24 1 DoubleCheck=42305167,72,1
42748019 17 4 5 3 DoubleCheck=42748019,72,1
42983959 3 0 24 1 DoubleCheck=42983959,72,1
43402943 3 1 8 6 DoubleCheck=43402943,72,1
43698703 3 1 2 2 DoubleCheck=43698703,72,1
43717981 3 1 8 6 DoubleCheck=43717981,72,1
43731869 3 0 24 1 DoubleCheck=43731869,72,1
43746137 3 0 24 1 DoubleCheck=43746137,72,1
43768357 3 1 8 6 DoubleCheck=43768357,72,1
44378077 3 0 24 1 DoubleCheck=44378077,72,1
44394173 3 0 24 1 DoubleCheck=44394173,72,1
44781521 3 1 8 6 DoubleCheck=44781521,72,1
44988709 3 0 24 1 DoubleCheck=44988709,72,1
45364853 43 9 13 17 DoubleCheck=45364853,72,1
45388481 43 9 13 17 DoubleCheck=45388481,72,1
45656467 3 1 3 6 DoubleCheck=45656467,72,1
45710999 3 0 24 1 DoubleCheck=45710999,72,1
45857047 3 0 24 1 DoubleCheck=45857047,72,1
46175153 3 0 24 1 DoubleCheck=46175153,72,1
46216699 3 0 24 1 DoubleCheck=46216699,72,1
46378181 3 0 24 1 DoubleCheck=46378181,72,1
46487227 3 1 8 6 DoubleCheck=46487227,72,1
46584451 3 0 24 1 DoubleCheck=46584451,72,1
46605149 3 0 24 1 DoubleCheck=46605149,72,1
46733287 3 0 24 1 DoubleCheck=46733287,72,1
46975139 3 1 8 6 DoubleCheck=46975139,72,1
47780699 4 1 10 8 DoubleCheck=47780699,72,1
47786399 4 1 10 8 DoubleCheck=47786399,72,1
47787829 4 1 10 8 DoubleCheck=47787829,72,1
47962091 3 1 28 3 DoubleCheck=47962091,72,1
48099251 3 0 24 1 DoubleCheck=48099251,72,1
48099301 3 0 24 1 DoubleCheck=48099301,72,1
49342991 43 9 13 17 DoubleCheck=49342991,72,1
49360579 43 9 13 17 DoubleCheck=49360579,72,1
49544023 3 1 28 3 DoubleCheck=49544023,72,1
50060831 43 9 13 17 DoubleCheck=50060831,73,1
50240497 3 1 28 3 DoubleCheck=50240497,73,1
50316473 3 1 28 3 DoubleCheck=50316473,73,1
50331793 3 1 28 3 DoubleCheck=50331793,73,1
50364731 4 1 10 8 DoubleCheck=50364731,73,1
50364827 4 1 10 8 DoubleCheck=50364827,73,1
50364943 4 1 10 8 DoubleCheck=50364943,73,1
50364961 4 1 10 8 DoubleCheck=50364961,73,1
50725429 3 1 8 0 DoubleCheck=50725429,73,1
50818837 43 9 13 17 DoubleCheck=50818837,73,1
51226867 43 9 13 17 DoubleCheck=51226867,73,1
53059973 3 1 28 3 DoubleCheck=53059973,73,1
53375909 3 1 28 3 DoubleCheck=53375909,73,1
53740301 3 1 8 0 DoubleCheck=53740301,73,1
53771789 3 1 28 3 DoubleCheck=53771789,73,1
53948311 3 1 28 3 DoubleCheck=53948311,73,1
54068747 3 1 8 0 DoubleCheck=54068747,73,1
54072163 3 1 8 0 DoubleCheck=54072163,73,1
54072253 3 1 8 0 DoubleCheck=54072253,73,1
54072643 3 1 8 0 DoubleCheck=54072643,73,1
54280231 3 1 28 3 DoubleCheck=54280231,73,1
54513103 3 1 28 3 DoubleCheck=54513103,73,1
54620483 3 1 28 3 DoubleCheck=54620483,73,1
54629573 3 1 28 3 DoubleCheck=54629573,73,1
54629689 3 1 28 3 DoubleCheck=54629689,73,1
54629747 3 1 28 3 DoubleCheck=54629747,73,1
54922663 3 1 28 3 DoubleCheck=54922663,73,1
55075171 3 1 28 3 DoubleCheck=55075171,73,1
55257889 3 1 12 4 DoubleCheck=55257889,73,1
55295731 3 1 28 3 DoubleCheck=55295731,73,1
55421599 3 1 28 3 DoubleCheck=55421599,73,1
55614661 3 1 28 3 DoubleCheck=55614661,73,1
56269861 3 1 28 3 DoubleCheck=56269861,73,1
56959759 3 1 28 3 DoubleCheck=56959759,73,1
57044891 3 1 28 3 DoubleCheck=57044891,73,1
57131161 3 1 28 3 DoubleCheck=57131161,73,1
57414373 3 1 12 4 DoubleCheck=57414373,73,1
57533953 3 1 28 3 DoubleCheck=57533953,73,1
57584141 3 1 12 4 DoubleCheck=57584141,73,1[/CODE]

Madpoo 2015-08-25 23:12

[QUOTE=Madpoo;408785]I'm trying to nail down how best to query the data now.[/QUOTE]

Okay, I think I got it running okay. It'll spit out a list of exponents that need a triple check, and neither of the two machines that already submitted a result are suspected of being bad or good. By focusing on these we can start to get results on both by marking more bad and more good in a single test, which should improve the overall strategy of rooting out bad systems.

To start I just pulled exponents < 35.5M to keep the list smallish... there's still 128 entries though.

As usual, if you start working on one, be sure to let your system check in so it gets assigned to you. And before starting a new one, look online to see if someone else already snagged it.

Or if you're bold and you're going to take a bunch, let people know on here to save some time.

When I manually eyeballed a few of these to see what machines did the tests, many of them have one system that has nothing but good results, but not enough for me to consider them "awesome" (like only 10+ good results, not the 50+ I'm looking for), and the other one has a few bad, but maybe just as many good.

In those cases I'm pretty sure I could predict with 80% accuracy (wild guess) which one is correct, but only a TC will tell for sure.

[CODE]exponent worktodo
35285249 DoubleCheck=35285249,71,1
35302283 DoubleCheck=35302283,71,1
35302507 DoubleCheck=35302507,71,1
35309177 DoubleCheck=35309177,71,1
35309801 DoubleCheck=35309801,71,1
35309947 DoubleCheck=35309947,71,1
35313661 DoubleCheck=35313661,71,1
35316817 DoubleCheck=35316817,71,1
35317423 DoubleCheck=35317423,71,1
35319763 DoubleCheck=35319763,71,1
35321219 DoubleCheck=35321219,71,1
35321899 DoubleCheck=35321899,71,1
35322037 DoubleCheck=35322037,71,1
35322103 DoubleCheck=35322103,71,1
35325373 DoubleCheck=35325373,71,1
35328949 DoubleCheck=35328949,71,1
35329537 DoubleCheck=35329537,71,1
35330017 DoubleCheck=35330017,71,1
35330237 DoubleCheck=35330237,71,1
35330263 DoubleCheck=35330263,71,1
35330749 DoubleCheck=35330749,71,1
35330969 DoubleCheck=35330969,71,1
35331419 DoubleCheck=35331419,71,1
35332223 DoubleCheck=35332223,71,1
35332699 DoubleCheck=35332699,71,1
35340847 DoubleCheck=35340847,71,1
35342929 DoubleCheck=35342929,71,1
35345623 DoubleCheck=35345623,71,1
35346943 DoubleCheck=35346943,71,1
35347567 DoubleCheck=35347567,71,1
35348359 DoubleCheck=35348359,71,1
35349527 DoubleCheck=35349527,71,1
35349997 DoubleCheck=35349997,71,1
35350423 DoubleCheck=35350423,71,1
35354909 DoubleCheck=35354909,71,1
35357389 DoubleCheck=35357389,71,1
35358667 DoubleCheck=35358667,71,1
35361323 DoubleCheck=35361323,71,1
35363239 DoubleCheck=35363239,71,1
35365549 DoubleCheck=35365549,71,1
35366677 DoubleCheck=35366677,71,1
35368717 DoubleCheck=35368717,71,1
35369899 DoubleCheck=35369899,71,1
35369923 DoubleCheck=35369923,71,1
35373341 DoubleCheck=35373341,71,1
35373557 DoubleCheck=35373557,71,1
35373743 DoubleCheck=35373743,71,1
35375731 DoubleCheck=35375731,71,1
35376763 DoubleCheck=35376763,71,1
35377369 DoubleCheck=35377369,71,1
35380067 DoubleCheck=35380067,71,1
35381279 DoubleCheck=35381279,71,1
35381377 DoubleCheck=35381377,71,1
35382337 DoubleCheck=35382337,71,1
35384383 DoubleCheck=35384383,71,1
35384521 DoubleCheck=35384521,71,1
35384801 DoubleCheck=35384801,71,1
35388653 DoubleCheck=35388653,71,1
35390519 DoubleCheck=35390519,71,1
35390633 DoubleCheck=35390633,71,1
35391883 DoubleCheck=35391883,71,1
35392891 DoubleCheck=35392891,71,1
35394083 DoubleCheck=35394083,71,1
35396399 DoubleCheck=35396399,71,1
35398147 DoubleCheck=35398147,71,1
35399597 DoubleCheck=35399597,71,1
35401033 DoubleCheck=35401033,71,1
35401061 DoubleCheck=35401061,71,1
35405561 DoubleCheck=35405561,71,1
35410873 DoubleCheck=35410873,71,1
35412217 DoubleCheck=35412217,71,1
35413999 DoubleCheck=35413999,71,1
35414371 DoubleCheck=35414371,71,1
35415763 DoubleCheck=35415763,71,1
35417969 DoubleCheck=35417969,71,1
35418619 DoubleCheck=35418619,71,1
35423627 DoubleCheck=35423627,71,1
35423747 DoubleCheck=35423747,71,1
35423851 DoubleCheck=35423851,71,1
35424331 DoubleCheck=35424331,71,1
35426077 DoubleCheck=35426077,71,1
35426327 DoubleCheck=35426327,71,1
35428619 DoubleCheck=35428619,71,1
35429333 DoubleCheck=35429333,71,1
35429419 DoubleCheck=35429419,71,1
35432021 DoubleCheck=35432021,71,1
35435489 DoubleCheck=35435489,71,1
35436217 DoubleCheck=35436217,71,1
35438989 DoubleCheck=35438989,71,1
35442499 DoubleCheck=35442499,71,1
35443523 DoubleCheck=35443523,71,1
35444153 DoubleCheck=35444153,71,1
35446403 DoubleCheck=35446403,71,1
35446721 DoubleCheck=35446721,71,1
35448181 DoubleCheck=35448181,71,1
35453389 DoubleCheck=35453389,71,1
35455559 DoubleCheck=35455559,71,1
35456689 DoubleCheck=35456689,71,1
35458231 DoubleCheck=35458231,71,1
35458723 DoubleCheck=35458723,71,1
35463157 DoubleCheck=35463157,71,1
35463739 DoubleCheck=35463739,71,1
35464657 DoubleCheck=35464657,71,1
35466287 DoubleCheck=35466287,71,1
35466397 DoubleCheck=35466397,71,1
35466661 DoubleCheck=35466661,71,1
35467057 DoubleCheck=35467057,71,1
35467109 DoubleCheck=35467109,71,1
35468029 DoubleCheck=35468029,71,1
35469743 DoubleCheck=35469743,71,1
35471729 DoubleCheck=35471729,71,1
35473561 DoubleCheck=35473561,71,1
35474063 DoubleCheck=35474063,71,1
35474849 DoubleCheck=35474849,71,1
35478853 DoubleCheck=35478853,71,1
35479723 DoubleCheck=35479723,71,1
35480339 DoubleCheck=35480339,71,1
35480579 DoubleCheck=35480579,71,1
35480791 DoubleCheck=35480791,71,1
35488613 DoubleCheck=35488613,71,1
35490097 DoubleCheck=35490097,71,1
35490269 DoubleCheck=35490269,71,1
35490529 DoubleCheck=35490529,71,1
35494871 DoubleCheck=35494871,71,1
35496977 DoubleCheck=35496977,71,1
35498587 DoubleCheck=35498587,71,1
35498719 DoubleCheck=35498719,71,1
35499397 DoubleCheck=35499397,71,1[/CODE]

frmky 2015-08-26 01:24

[QUOTE=Madpoo;408790]As usual, if you start working on one, be sure to let your system check in so it gets assigned to you.[/QUOTE]
I do this, but it is a pain for low exponents since the online form won't let me manually reserve them. Currently I stop a running mprime, add them to worktodo.txt, restart it and let it communicate, stop it, remove the updated lines from worktodo.txt, and then restart it. Would it be possible to repurpose the DC LL reservation code from GPU-72 for this project?

Mark Rose 2015-08-26 02:27

You can add the lines to worktodo.add and force manual communication by running ./mprime -c in a new terminal so you don't have to stop mprime.

LaurV 2015-08-26 02:53

[QUOTE=Madpoo;408790]<snip>[/QUOTE]
You can put the N/A key when you create the list, so one just add them to the worktodo, without doing manual insertion (lot of "paste" (CTRL+V)) or reservation.

Madpoo 2015-08-26 03:02

[QUOTE=LaurV;408813]You can put the N/A key when you create the list, so one just add them to the worktodo, without doing manual insertion (lot of "paste" (CTRL+V)) or reservation.[/QUOTE]

I guess I'm not up to date on the syntax there. When I'm just testing stuff manually I don't need the assignment key part at all, it's optional. Others have said that doing that and letting it connect will create the assignment and update your worktodo... So it sounds like you just add the line, force a connection to the server, and that's it?

If there's a better way, let me know. Yeah, it'd be nice if we could do a manual reservation online, but yeah, the smaller exponents can't be done manually since they're in that critical range.

ATH 2015-08-26 03:47

[QUOTE=Mark Rose;408810]You can add the lines to worktodo.add and force manual communication by running ./mprime -c in a new terminal so you don't have to stop mprime.[/QUOTE]

Prime95 checks for "worktodo.add" files like every 15 or 30 min or something like that, so you do not have to do anything other than create that file. I would assume mprime does the same.


[QUOTE=frmky;408800]I do this, but it is a pain for low exponents since the online form won't let me manually reserve them. Currently I stop a running mprime, add them to worktodo.txt, restart it and let it communicate, stop it, remove the updated lines from worktodo.txt, and then restart it. Would it be possible to repurpose the DC LL reservation code from GPU-72 for this project?[/QUOTE]

Why do you need to remove the key after communication?

Mark Rose 2015-08-26 03:56

[QUOTE=ATH;408822]Prime95 checks for "worktodo.add" files like every 15 or 30 min or something like that, so you do not have to do anything other than create that file. I would assume mprime does the same.[/quote]

Forcing communication causes worktodo.add to be merged immediately, if you don't want to wait for whatever reason.

[quote]Why do you need to remove the key after communication?[/QUOTE]

Perhaps he's trying to reserve assignments for processing on a graphics card. Using mprime/Prime95 is the only way to reserve assignments in the critical ranges.

frmky 2015-08-26 06:49

[QUOTE=Mark Rose;408824]Perhaps he's trying to reserve assignments for processing on a graphics card. Using mprime/Prime95 is the only way to reserve assignments in the critical ranges.[/QUOTE]
Exactly. Most of my LL tests are done on 7 GPUs. So using worktodo.add and -c will save stopping it once, but I still need to remove the now-reserved lines from the worktodo.txt so the CPU doesn't do it. I don't need the reservation key for the GPU; I just want to formally reserve it so someone else doesn't grab it while it's running.

Mark Rose 2015-08-26 13:19

[QUOTE=frmky;408837]Exactly. Most of my LL tests are done on 7 GPUs. So using worktodo.add and -c will save stopping it once, but I still need to remove the now-reserved lines from the worktodo.txt so the CPU doesn't do it. I don't need the reservation key for the GPU; I just want to formally reserve it so someone else doesn't grab it while it's running.[/QUOTE]

You may consider using a second copy of mprime in another directory only for assignment reservation. I'd set these values in prime.txt

UsePrimenet=1
V5UserID=frmky
DaysOfWork=0
NoMoreWork=1

and in local.txt

WorkerThreads=1

and run only mprime -c, which will quit immediately after communicating/registering and not process any work, and give you all the assignments in a single [Worker #1] block.


All times are UTC. The time now is 16:46.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.