Strategic Double Clicking - Page 70

Mark Rose · 2016-01-26, 17:33

It would be nice if the server only assigned triple checks to only proven reliable machines. Perhaps add a checkbox to prefer triple checks like there is for lowest exponents?

That would automate a lot of this effort, plus keep workers that don't turn in results from holding strategic assignments.

A second tweak would be to double check work done by proven unreliable machines.

A third tweak would be to assign new CPUs double checks from work done by proven reliable machines -- that would quickly identify flaky machines for the second tweak.

I don't know how easy these would be to implement.

Mark Rose · 2016-01-26, 17:43

Quote:

Originally Posted by Madpoo

Here are those available exponents needing quad checks. That smallest one just showed up today after a triple check I ran failed to match either previous run. I'm pretty sure all of these have one test by me:

Code:

DoubleCheck=35540807,71,1

Mine.

chalsall · 2016-01-26, 18:19

Quote:

Originally Posted by Madpoo

I'm pretty sure all of these have one test by me:

Aaron, could I ask you to generate a list of 24 SDCs for me which will help machines be bumped into the good / bad ratio threshold for your heuristics? Preferably different machines for each candidate to be DC'ed/TC'ed.

In my mind it doesn't make sense for my machines to QC a candidate you've already DC'ed/TC'ed, since it will simply confirm what your queries already suspect (no joke intend).

Secondly, I agree with Mark's suggestions above. Perhaps it is time to automate this (if it isn't too much work).

dragonbud20 · 2016-01-27, 00:26

Quote:

Originally Posted by Madpoo

Here are those available exponents needing quad checks.

Code:

DoubleCheck=49654117,72,1
DoubleCheck=50128459,73,1
DoubleCheck=50919889,73,1

I'll take these

Madpoo · 2016-01-28, 16:27

Quote:

Originally Posted by chalsall

Aaron, could I ask you to generate a list of 24 SDCs for me which will help machines be bumped into the good / bad ratio threshold for your heuristics? Preferably different machines for each candidate to be DC'ed/TC'ed....

Here's a list of exponents that have been double-checked with a mismatch. In each case, both machines are within just one bad result away from being thrust unmercifully into the "equal bad and good" category. The result of this triple check will push one or the other.

Code:

exponent	worktodo
36987347	DoubleCheck=36987347,71,1
38555057	DoubleCheck=38555057,71,1
38830921	DoubleCheck=38830921,71,1
38919169	DoubleCheck=38919169,71,1
51344617	DoubleCheck=51344617,73,1
51723311	DoubleCheck=51723311,73,1
51872731	DoubleCheck=51872731,73,1
51913793	DoubleCheck=51913793,73,1
52910057	DoubleCheck=52910057,73,1
53130773	DoubleCheck=53130773,73,1
53860589	DoubleCheck=53860589,73,1
55675049	DoubleCheck=55675049,73,1
56234723	DoubleCheck=56234723,74,1
56539541	DoubleCheck=56539541,73,1
58241527	DoubleCheck=58241527,73,1
58479067	DoubleCheck=58479067,73,1
58805933	DoubleCheck=58805933,73,1
59898451	DoubleCheck=59898451,73,1
60751577	DoubleCheck=60751577,74,1
60790393	DoubleCheck=60790393,74,1
61826683	DoubleCheck=61826683,74,1
62306407	DoubleCheck=62306407,74,1
62621821	DoubleCheck=62621821,74,1
63173237	DoubleCheck=63173237,74,1
63643253	DoubleCheck=63643253,74,1
64466929	DoubleCheck=64466929,74,1
64648789	DoubleCheck=64648789,74,1
64687591	DoubleCheck=64687591,74,1
64792009	DoubleCheck=64792009,74,1
66512779	DoubleCheck=66512779,75,1
67694827	DoubleCheck=67694827,75,1
74049919	DoubleCheck=74049919,75,1

chalsall · 2016-01-28, 16:58

Quote:

Originally Posted by Madpoo

The result of this triple check will push one or the other.

Thanks. Taken.

Madpoo · 2016-01-29, 04:37

Here's a list... smallest available exponent for any machine with more bad than good. The ones that are just slightly over a 1:1 bad/good ratio are interesting, but I like to test another one and see which way it goes before trying to tackle more.

I found one recently that way... started with 1 bad, zero good. It's up to a (presumed) 15 bad, zero good. They were all exponents in the 60-70M range so I just snagged them myself, being larger.

Anyway, maybe we'll get some good ones like that out of this list, and some are already well into the "bad" group, so be sure to snag those while you can.

Code:

Exponent	Bad	Good	Unk	Sus	Solo	Mis	worktodo
37334837	2	1	3	0	3	0	DoubleCheck=37334837,71,1
37865777	3	2	1	0	1	0	DoubleCheck=37865777,71,1
38194643	1	0	3	4	3	4	DoubleCheck=38194643,71,1
38217323	1	0	2	0	2	0	DoubleCheck=38217323,71,1
38347571	1	0	1	1	1	1	DoubleCheck=38347571,71,1
38471309	3	1	1	0	1	0	DoubleCheck=38471309,71,1
38475653	1	0	2	0	1	1	DoubleCheck=38475653,71,1
38592787	1	0	2	0	2	0	DoubleCheck=38592787,71,1
38619751	6	5	2	1	2	1	DoubleCheck=38619751,71,1
38651779	1	0	2	0	2	0	DoubleCheck=38651779,71,1
38652533	5	2	4	0	3	1	DoubleCheck=38652533,71,1
38851957	2	0	1	0	1	0	DoubleCheck=38851957,71,1
38948521	1	0	2	0	2	0	DoubleCheck=38948521,71,1
41268527	1	0	2	0	2	0	DoubleCheck=41268527,72,1
41402099	4	1	2	0	1	1	DoubleCheck=41402099,72,1
41458877	1	0	1	0	1	0	DoubleCheck=41458877,72,1
42036389	6	0	1	1	1	1	DoubleCheck=42036389,72,1
45422269	1	0	3	1	3	1	DoubleCheck=45422269,72,1
46217657	2	0	3	0	2	1	DoubleCheck=46217657,72,1
47673583	1	0	1	0	1	0	DoubleCheck=47673583,72,1
47955311	2	0	1	1	1	1	DoubleCheck=47955311,72,1
48700441	2	0	1	0	1	0	DoubleCheck=48700441,72,1
50145749	2	0	1	0	1	0	DoubleCheck=50145749,73,1
51770489	15	4	2	1	2	1	DoubleCheck=51770489,73,1
54801569	2	0	4	0	4	0	DoubleCheck=54801569,73,1
55380979	4	3	10	0	10	0	DoubleCheck=55380979,74,1
57517913	4	3	1	0	1	0	DoubleCheck=57517913,73,1
57989447	11	7	51	9	51	9	DoubleCheck=57989447,73,1
65095861	1	0	2	2	2	2	DoubleCheck=65095861,74,1
66357407	2	1	2	0	2	0	DoubleCheck=66357407,74,1
67362901	1	0	1	0	1	0	DoubleCheck=67362901,75,1
67665181	1	0	0	3	1	2	DoubleCheck=67665181,75,1
77900497	1	0	1	0	1	0	DoubleCheck=77900497,74,1

kladner · 2016-01-29, 06:41

I'll queue the first one.
DoubleCheck=37334837,71,1

ET_ · 2016-01-29, 10:23

These are mine!

DoubleCheck=37865777,71,1
DoubleCheck=38194643,71,1
DoubleCheck=38217323,71,1

ric · 2016-01-29, 12:23

Quote:

Originally Posted by Madpoo

Here's a list... smallest available exponent for any machine with more bad than good. <snip>

Code:

Exponent	Bad	Good	Unk	Sus	Solo	Mis	worktodo
38347571	1	0	1	1	1	1	DoubleCheck=38347571,71,1
38471309	3	1	1	0	1	0	DoubleCheck=38471309,71,1
38475653	1	0	2	0	1	1	DoubleCheck=38475653,71,1
38592787	1	0	2	0	2	0	DoubleCheck=38592787,71,1
38619751	6	5	2	1	2	1	DoubleCheck=38619751,71,1
38651779	1	0	2	0	2	0	DoubleCheck=38651779,71,1
38652533	5	2	4	0	3	1	DoubleCheck=38652533,71,1
38851957	2	0	1	0	1	0	DoubleCheck=38851957,71,1
38948521	1	0	2	0	2	0	DoubleCheck=38948521,71,1

mine

cuBerBruce · 2016-01-29, 17:24

Quote:

Originally Posted by Madpoo

Here's a list... smallest available exponent for any machine with more bad than good.

Code:

Exponent	Bad	Good	Unk	Sus	Solo	Mis	worktodo
...
41268527	1	0	2	0	2	0	DoubleCheck=41268527,72,1
41402099	4	1	2	0	1	1	DoubleCheck=41402099,72,1
41458877	1	0	1	0	1	0	DoubleCheck=41458877,72,1
42036389	6	0	1	1	1	1	DoubleCheck=42036389,72,1
,,,

These seem to be taken by ANONYMOUS users. Maybe ordinary cat 4 wave assignments?

Quote:

[CODE
45422269 1 0 3 1 3 1 DoubleCheck=45422269,72,1
[/CODE]

I took this one.

2016-01-29, 06:41	#767
kladner "Kieren" Jul 2011 In My Own Galaxy! 27AE₁₆ Posts	I'll queue the first one. DoubleCheck=37334837,71,1 Last fiddled with by kladner on 2016-01-29 at 06:41

2016-01-29, 10:23	#768
ET_ Banned "Luigi" Aug 2002 Team Italia 12D0₁₆ Posts	These are mine! DoubleCheck=37865777,71,1 DoubleCheck=38194643,71,1 DoubleCheck=38217323,71,1 Last fiddled with by ET_ on 2016-01-29 at 10:38

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Double-Double Arithmetic	Mysticial	Software	52	2021-04-23 06:51
Clicking an exponent leads to 404 page	marigonzes	Information & Answers	2	2017-02-14 16:56
x.265 half the size, double the computation; so if you double again? 1/4th?	jasong	jasong	7	2015-08-17 10:56
What about double-checking TF/P-1?	137ben	PrimeNet	6	2012-03-13 04:01
Double the area, Double the volume.	Uncwilly	Puzzles	8	2006-07-03 16:02

2016-01-26, 17:33	#760
Mark Rose "/X\(‘-‘)/X\" Jan 2013 B72₁₆ Posts	It would be nice if the server only assigned triple checks to only proven reliable machines. Perhaps add a checkbox to prefer triple checks like there is for lowest exponents? That would automate a lot of this effort, plus keep workers that don't turn in results from holding strategic assignments. A second tweak would be to double check work done by proven unreliable machines. A third tweak would be to assign new CPUs double checks from work done by proven reliable machines -- that would quickly identify flaky machines for the second tweak. I don't know how easy these would be to implement.