PrimeGrid
Please visit donation page to help the project cover running costs for this month

Toggle Menu

Join PrimeGrid

Returning Participants

Community

Leader Boards

Results

Other

drummers-lowrise

Advanced search

Message boards : Sierpinski/Riesel Base 5 Problem : Strange Times

Author Message
NeoMetal*
Volunteer tester
Avatar
Send message
Joined: 25 Mar 11
Posts: 418
ID: 92179
Credit: 1,747,428,303
RAC: 0
Eliminated 1 conjecture "k"Found 1 prime in the 2018 Tour de Primes321 LLR Turquoise: Earned 5,000,000 credits (7,698,294)Cullen LLR Turquoise: Earned 5,000,000 credits (6,655,866)ESP LLR Turquoise: Earned 5,000,000 credits (8,198,062)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,817,032)PPS LLR Jade: Earned 10,000,000 credits (10,016,096)PSP LLR Jade: Earned 10,000,000 credits (10,891,779)SoB LLR Jade: Earned 10,000,000 credits (11,111,741)SR5 LLR Turquoise: Earned 5,000,000 credits (7,278,494)SGS LLR Turquoise: Earned 5,000,000 credits (7,457,856)TRP LLR Turquoise: Earned 5,000,000 credits (7,714,186)Woodall LLR Turquoise: Earned 5,000,000 credits (5,726,778)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,094,130)Generalized Cullen/Woodall Sieve Jade: Earned 10,000,000 credits (11,628,594)PPS Sieve Double Amethyst: Earned 1,000,000,000 credits (1,040,865,445)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Sapphire: Earned 20,000,000 credits (20,235,150)TRP Sieve (suspended) Sapphire: Earned 20,000,000 credits (20,376,844)AP 26/27 Jade: Earned 10,000,000 credits (11,615,539)GFN Double Silver: Earned 200,000,000 credits (427,593,017)PSA Double Bronze: Earned 100,000,000 credits (125,462,363)
Message 67072 - Posted: 19 Jun 2013 | 10:58:14 UTC

These WUs completed and validated despite strange times:

http://www.primegrid.com/workunit.php?wuid=339340463
http://www.primegrid.com/workunit.php?wuid=339363656
http://www.primegrid.com/workunit.php?wuid=339391030

There's more but these are just 3 examples.

Also why does credit vary only slightly when WUs vary in length by up to 2x from the smallest to largest?
____________
Largest Primes to Date:
As Double Checker: SR5 109208*5^1816285+1 Dgts-1,269,534
As Initial Finder: SR5 243944*5^1258576-1 Dgts-879,713


Profile ConanProject donor
Avatar
Send message
Joined: 24 Mar 09
Posts: 218
ID: 37336
Credit: 52,930,317
RAC: 2
321 LLR Gold: Earned 500,000 credits (500,863)Cullen LLR Gold: Earned 500,000 credits (522,970)ESP LLR Gold: Earned 500,000 credits (516,922)Generalized Cullen/Woodall LLR Gold: Earned 500,000 credits (543,312)PPS LLR Gold: Earned 500,000 credits (567,107)PSP LLR Amethyst: Earned 1,000,000 credits (1,019,589)SoB LLR Amethyst: Earned 1,000,000 credits (1,055,482)SR5 LLR Gold: Earned 500,000 credits (522,539)SGS LLR Gold: Earned 500,000 credits (532,524)TRP LLR Gold: Earned 500,000 credits (564,910)Woodall LLR Amethyst: Earned 1,000,000 credits (1,022,336)321 Sieve Silver: Earned 100,000 credits (431,508)Cullen/Woodall Sieve (suspended) Silver: Earned 100,000 credits (204,463)Generalized Cullen/Woodall Sieve Amethyst: Earned 1,000,000 credits (1,015,683)PPS Sieve Sapphire: Earned 20,000,000 credits (37,926,771)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,226,498)TRP Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,515,247)AP 26/27 Gold: Earned 500,000 credits (509,418)GFN Ruby: Earned 2,000,000 credits (2,506,619)PSA Silver: Earned 100,000 credits (225,555)
Message 67073 - Posted: 19 Jun 2013 | 11:38:04 UTC
Last modified: 19 Jun 2013 | 11:46:01 UTC

I have seen a few of these also on one of my Windows systems, not seen it on Linux.

339384219
339370895
339399932

How can the run time be mere seconds but the CPU time is thousands of seconds?

Conan
____________


Profile Michael GoetzProject donor
Volunteer moderator
Project administrator
Project scientist
Avatar
Send message
Joined: 21 Jan 10
Posts: 12675
ID: 53948
Credit: 184,626,545
RAC: 147,034
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,063,182)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Ruby: Earned 2,000,000 credits (4,146,403)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Ruby: Earned 2,000,000 credits (2,773,744)PSP LLR Ruby: Earned 2,000,000 credits (2,632,269)SoB LLR Sapphire: Earned 20,000,000 credits (34,158,496)SR5 LLR Turquoise: Earned 5,000,000 credits (8,293,415)SGS LLR Ruby: Earned 2,000,000 credits (2,012,222)TRP LLR Ruby: Earned 2,000,000 credits (2,737,347)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Turquoise: Earned 5,000,000 credits (5,046,112)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (20,110,788)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Turquoise: Earned 5,000,000 credits (7,090,096)GFN Emerald: Earned 50,000,000 credits (64,594,991)PSA Jade: Earned 10,000,000 credits (10,305,150)
Message 67074 - Posted: 19 Jun 2013 | 12:34:39 UTC - in response to Message 67072.
Last modified: 19 Jun 2013 | 12:36:35 UTC

These WUs completed and validated despite strange times:

http://www.primegrid.com/workunit.php?wuid=339340463
http://www.primegrid.com/workunit.php?wuid=339363656
http://www.primegrid.com/workunit.php?wuid=339391030

There's more but these are just 3 examples.


Because another app (Genefer) tracks elapsed time on its own, and it ALWAYS says it uses more time than BOINC reports, I've known for quite a while that the BOINC client is under-reporting the elapsed time. I've also noticed, over the last couple of days, numerous examples where the elapsed time was *slightly* less than the CPU time, which should be impossible unless you're running a multi-threaded app. None of our boinc apps use multiple CPU cores.

What's new is that I'm now seeing ridiculously low elapsed times from BOINC, in some cases under a minute.

That probably isn't going to be a problem, but we'll review the validation process -- especially the sanity checks -- to make sure this won't break something somewhere. So far I've not seen any examples of where this has caused a problem.

We'll also take a look and make sure this isn't something we've managed to break somehow.

Also why does credit vary only slightly when WUs vary in length by up to 2x from the smallest to largest?


This is something we'll be looking into.
____________
Please do not PM me with support questions. Ask on the forums instead. Thank you!

My lucky number is 75898524288+1

Profile Michael GoetzProject donor
Volunteer moderator
Project administrator
Project scientist
Avatar
Send message
Joined: 21 Jan 10
Posts: 12675
ID: 53948
Credit: 184,626,545
RAC: 147,034
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,063,182)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Ruby: Earned 2,000,000 credits (4,146,403)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Ruby: Earned 2,000,000 credits (2,773,744)PSP LLR Ruby: Earned 2,000,000 credits (2,632,269)SoB LLR Sapphire: Earned 20,000,000 credits (34,158,496)SR5 LLR Turquoise: Earned 5,000,000 credits (8,293,415)SGS LLR Ruby: Earned 2,000,000 credits (2,012,222)TRP LLR Ruby: Earned 2,000,000 credits (2,737,347)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Turquoise: Earned 5,000,000 credits (5,046,112)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (20,110,788)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Turquoise: Earned 5,000,000 credits (7,090,096)GFN Emerald: Earned 50,000,000 credits (64,594,991)PSA Jade: Earned 10,000,000 credits (10,305,150)
Message 67076 - Posted: 19 Jun 2013 | 13:07:41 UTC

The very short elapsed times have only been happening since yesterday.

Slightly short times have been happening for a very long time. At least months, if not years. I'm not concerned with those.

But we will look into the very short times.
____________
Please do not PM me with support questions. Ask on the forums instead. Thank you!

My lucky number is 75898524288+1

Profile ConanProject donor
Avatar
Send message
Joined: 24 Mar 09
Posts: 218
ID: 37336
Credit: 52,930,317
RAC: 2
321 LLR Gold: Earned 500,000 credits (500,863)Cullen LLR Gold: Earned 500,000 credits (522,970)ESP LLR Gold: Earned 500,000 credits (516,922)Generalized Cullen/Woodall LLR Gold: Earned 500,000 credits (543,312)PPS LLR Gold: Earned 500,000 credits (567,107)PSP LLR Amethyst: Earned 1,000,000 credits (1,019,589)SoB LLR Amethyst: Earned 1,000,000 credits (1,055,482)SR5 LLR Gold: Earned 500,000 credits (522,539)SGS LLR Gold: Earned 500,000 credits (532,524)TRP LLR Gold: Earned 500,000 credits (564,910)Woodall LLR Amethyst: Earned 1,000,000 credits (1,022,336)321 Sieve Silver: Earned 100,000 credits (431,508)Cullen/Woodall Sieve (suspended) Silver: Earned 100,000 credits (204,463)Generalized Cullen/Woodall Sieve Amethyst: Earned 1,000,000 credits (1,015,683)PPS Sieve Sapphire: Earned 20,000,000 credits (37,926,771)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,226,498)TRP Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,515,247)AP 26/27 Gold: Earned 500,000 credits (509,418)GFN Ruby: Earned 2,000,000 credits (2,506,619)PSA Silver: Earned 100,000 credits (225,555)
Message 67091 - Posted: 19 Jun 2013 | 23:54:35 UTC

Thanks for that Michael.

Conan
____________


TroubledBunnyProject donor
Volunteer tester
Avatar
Send message
Joined: 4 Jun 07
Posts: 605
ID: 8931
Credit: 385,015,100
RAC: 0
Discovered 1 mega prime321 LLR Gold: Earned 500,000 credits (622,449)Cullen LLR Amethyst: Earned 1,000,000 credits (1,019,085)ESP LLR Amethyst: Earned 1,000,000 credits (1,006,207)Generalized Cullen/Woodall LLR Gold: Earned 500,000 credits (769,754)PPS LLR Ruby: Earned 2,000,000 credits (3,839,000)PSP LLR Amethyst: Earned 1,000,000 credits (1,147,411)SoB LLR Gold: Earned 500,000 credits (521,611)SR5 LLR Amethyst: Earned 1,000,000 credits (1,036,817)SGS LLR Amethyst: Earned 1,000,000 credits (1,069,052)TRP LLR Amethyst: Earned 1,000,000 credits (1,004,422)Woodall LLR Gold: Earned 500,000 credits (531,350)321 Sieve Silver: Earned 100,000 credits (223,275)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (2,655,968)Generalized Cullen/Woodall Sieve Amethyst: Earned 1,000,000 credits (1,006,509)PPS Sieve Double Silver: Earned 200,000,000 credits (343,952,410)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,014,690)TRP Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,141,121)AP 26/27 Amethyst: Earned 1,000,000 credits (1,422,266)GFN Ruby: Earned 2,000,000 credits (4,753,065)PSA Jade: Earned 10,000,000 credits (16,276,495)
Message 67329 - Posted: 25 Jun 2013 | 22:15:40 UTC
Last modified: 25 Jun 2013 | 22:23:13 UTC

Not strange times but horrid...

460515435 340441469 360567 24 Jun 2013 | 18:26:17 UTC 24 Jun 2013 | 19:47:10 UTC Completed and validated 3,251.36 3,200.17 264.08 Sierpinski/Riesel Base 5 Problem (LLR) v6.15 460509170 340437906 360567 24 Jun 2013 | 17:35:14 UTC 24 Jun 2013 | 19:11:15 UTC Completed and validated 4,224.83 4,159.89 263.87 Sierpinski/Riesel Base 5 Problem (LLR) v6.15

Workunit 340437906 had FFT length 280K and WU 340441469 had FFT lengtgh 240K
I do understand the problem and reason but 30% more computing time for the same credit :(
I know there's nothing to be done about this but it still makes me a little (very little) grumpy.

As an aside: I was crunching this project on PRPNet until it migrated, I like using PRPNet occasionally. I turn on the verbose option and enjoy seeing a little more detail. One other good thing is that you never know what credit you are being 'paid' per unit under PRPNet so I never had a reason to moan!

Profile Michael GoetzProject donor
Volunteer moderator
Project administrator
Project scientist
Avatar
Send message
Joined: 21 Jan 10
Posts: 12675
ID: 53948
Credit: 184,626,545
RAC: 147,034
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,063,182)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Ruby: Earned 2,000,000 credits (4,146,403)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Ruby: Earned 2,000,000 credits (2,773,744)PSP LLR Ruby: Earned 2,000,000 credits (2,632,269)SoB LLR Sapphire: Earned 20,000,000 credits (34,158,496)SR5 LLR Turquoise: Earned 5,000,000 credits (8,293,415)SGS LLR Ruby: Earned 2,000,000 credits (2,012,222)TRP LLR Ruby: Earned 2,000,000 credits (2,737,347)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Turquoise: Earned 5,000,000 credits (5,046,112)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (20,110,788)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Turquoise: Earned 5,000,000 credits (7,090,096)GFN Emerald: Earned 50,000,000 credits (64,594,991)PSA Jade: Earned 10,000,000 credits (10,305,150)
Message 67334 - Posted: 25 Jun 2013 | 22:52:51 UTC - in response to Message 67329.

One other good thing is that you never know what credit you are being 'paid' per unit under PRPNet so I never had a reason to moan!


It's the exact same formula on PRPNet as BOINC, so you were seeing the same discrepancies on PRPNet. It just didn't bother you because you never saw the individual tasks, only the aggregate.

I do understand the problem and reason but 30% more computing time for the same credit :(


I suppose we COULD go back to the old BOINC credit system, based on CPU time and CPU benchmarks. You know, the system where credit could vary by 1000%, and AVX CPUs get half the credit for the same task as older non-AVX computers?

If we're only off by 30%, that's an improvement of biblical proportions.
____________
Please do not PM me with support questions. Ask on the forums instead. Thank you!

My lucky number is 75898524288+1

Profile sergeProject donor
Avatar
Send message
Joined: 21 Jun 12
Posts: 110
ID: 144858
Credit: 199,518,009
RAC: 104,661
Eliminated 2 conjecture "k"s321 LLR Turquoise: Earned 5,000,000 credits (5,342,838)Cullen LLR Silver: Earned 100,000 credits (158,983)ESP LLR Gold: Earned 500,000 credits (505,527)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,268,642)PPS LLR Sapphire: Earned 20,000,000 credits (23,395,331)PSP LLR Silver: Earned 100,000 credits (120,541)SoB LLR Amethyst: Earned 1,000,000 credits (1,280,028)SR5 LLR Ruby: Earned 2,000,000 credits (2,649,390)SGS LLR Gold: Earned 500,000 credits (518,356)TRP LLR Silver: Earned 100,000 credits (141,131)Woodall LLR Silver: Earned 100,000 credits (116,405)Cullen/Woodall Sieve (suspended) Silver: Earned 100,000 credits (235,088)PPS Sieve Amethyst: Earned 1,000,000 credits (1,031,526)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Gold: Earned 500,000 credits (500,445)TRP Sieve (suspended) Gold: Earned 500,000 credits (507,938)GFN Double Bronze: Earned 100,000,000 credits (160,454,410)PSA Silver: Earned 100,000 credits (291,533)
Message 67339 - Posted: 25 Jun 2013 | 23:51:56 UTC

What is needed is a sufficiently representative sampling of future tests (which is an easily solvable problem because you have them all known in advance) by running "llr -d in.npg", promptly killing it after registering the time per iteration, and then from an array of datapoints, building a simple model of the expected runtime. This is likely to be a blocky lookup table of a certain weight function of (range of k, range of n) multiplied by n. I had already run such experiment, but didn't build the model (I don't have any use for it, but you do - you could re-run this; I also ran on a very heterogeneous cluster, so times had been confounded with CPU type).

The simple observations that I can easily share are these:
1. for any n, there's at least two or three ranges of k with distinctly different runtimes: roughly, four-digit k's are in the class of their own (with the run time under 3000s on a 3770K, let's call it T); 1e4<k<1.5e5 would have runtime ~1.15T, and k>1.5e5 would have runtime ~(1.3-1.5)T.
2. the FFT-size crossover boundaries over n ranges will be different for different k's - need to sample from the population and tabulate; this is easier than "make up" a formula from one's imagination.

If you don't put k dependency in your credit formula, you may see high k's WUs aborted systematically by impatient users: there's no reason for them to run a 1.5 time longer job for the same credit as a short one with low k.

My $0.02.


____________
My lucky number is Phi(4, 2^2396029-1)/2.

Profile Michael GoetzProject donor
Volunteer moderator
Project administrator
Project scientist
Avatar
Send message
Joined: 21 Jan 10
Posts: 12675
ID: 53948
Credit: 184,626,545
RAC: 147,034
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,063,182)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Ruby: Earned 2,000,000 credits (4,146,403)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Ruby: Earned 2,000,000 credits (2,773,744)PSP LLR Ruby: Earned 2,000,000 credits (2,632,269)SoB LLR Sapphire: Earned 20,000,000 credits (34,158,496)SR5 LLR Turquoise: Earned 5,000,000 credits (8,293,415)SGS LLR Ruby: Earned 2,000,000 credits (2,012,222)TRP LLR Ruby: Earned 2,000,000 credits (2,737,347)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Turquoise: Earned 5,000,000 credits (5,046,112)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (20,110,788)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Turquoise: Earned 5,000,000 credits (7,090,096)GFN Emerald: Earned 50,000,000 credits (64,594,991)PSA Jade: Earned 10,000,000 credits (10,305,150)
Message 67341 - Posted: 26 Jun 2013 | 2:00:14 UTC - in response to Message 67339.

If you don't put k dependency in your credit formula, you may see high k's WUs aborted systematically by impatient users: there's no reason for them to run a 1.5 time longer job for the same credit as a short one with low k.


We're not going to have 100 different credit calculations for 1 project, and we're not going to run 100s of tests to find out where all the FFT shift points are. The shift points vary by processor type, because different processors will use different FFTs. Even if we wanted to do this -- which we definitely do not -- it wouldn't work anyway because the behavior and timing will vary from system to system.

So, plain and simple, what you're suggesting isn't going to happen.

If we find that people are trying to game the system, there's no shortage of punitive strategies that can be employed to discourage such behavior.

____________
Please do not PM me with support questions. Ask on the forums instead. Thank you!

My lucky number is 75898524288+1

Profile x3mEn
Avatar
Send message
Joined: 21 Jul 10
Posts: 344
ID: 64131
Credit: 55,738,523
RAC: 25,545
321 LLR Amethyst: Earned 1,000,000 credits (1,137,410)Cullen LLR Amethyst: Earned 1,000,000 credits (1,005,544)ESP LLR Amethyst: Earned 1,000,000 credits (1,529,595)Generalized Cullen/Woodall LLR Amethyst: Earned 1,000,000 credits (1,020,167)PPS LLR Ruby: Earned 2,000,000 credits (2,023,233)PSP LLR Ruby: Earned 2,000,000 credits (2,065,730)SoB LLR Ruby: Earned 2,000,000 credits (2,036,592)SR5 LLR Ruby: Earned 2,000,000 credits (2,008,007)SGS LLR Ruby: Earned 2,000,000 credits (2,001,244)TRP LLR Ruby: Earned 2,000,000 credits (2,002,759)Woodall LLR Amethyst: Earned 1,000,000 credits (1,017,997)321 Sieve Ruby: Earned 2,000,000 credits (2,005,104)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,000,304)Generalized Cullen/Woodall Sieve Ruby: Earned 2,000,000 credits (2,087,920)PPS Sieve Jade: Earned 10,000,000 credits (11,092,841)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,313,121)TRP Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,284,804)AP 26/27 Amethyst: Earned 1,000,000 credits (1,750,619)GFN Turquoise: Earned 5,000,000 credits (5,599,058)PSA Turquoise: Earned 5,000,000 credits (8,756,474)
Message 67342 - Posted: 26 Jun 2013 | 3:46:40 UTC - in response to Message 67339.
Last modified: 26 Jun 2013 | 3:48:13 UTC


If you don't put k dependency in your credit formula, you may see high k's WUs aborted systematically by impatient users: there's no reason for them to run a 1.5 time longer job for the same credit as a short one with low k.

I see you are already using this strategy successfully in you action... :)

And what punitive measures we will take for them?
Public reprimand?.. Flogging?.. Guillotine?..
)))
____________

Profile chip
Avatar
Send message
Joined: 12 Apr 11
Posts: 129
ID: 94709
Credit: 164,479,462
RAC: 65,102
321 LLR Amethyst: Earned 1,000,000 credits (1,431,220)ESP LLR Ruby: Earned 2,000,000 credits (2,006,207)PPS LLR Ruby: Earned 2,000,000 credits (3,000,162)PSP LLR Amethyst: Earned 1,000,000 credits (1,022,562)SoB LLR Ruby: Earned 2,000,000 credits (2,015,539)SR5 LLR Ruby: Earned 2,000,000 credits (2,000,481)SGS LLR Ruby: Earned 2,000,000 credits (2,000,014)TRP LLR Ruby: Earned 2,000,000 credits (3,000,865)321 Sieve Ruby: Earned 2,000,000 credits (2,000,357)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,000,096)PPS Sieve Emerald: Earned 50,000,000 credits (50,000,422)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,000,500)AP 26/27 Sapphire: Earned 20,000,000 credits (20,000,721)GFN Sapphire: Earned 20,000,000 credits (20,000,117)PSA Emerald: Earned 50,000,000 credits (50,000,198)
Message 67343 - Posted: 26 Jun 2013 | 3:53:44 UTC - in response to Message 67342.

And what punitive measures we will take for them?
Public reprimand?.. Flogging?.. Guillotine?..
)))

Give a 'cheater' badge!

Profile x3mEn
Avatar
Send message
Joined: 21 Jul 10
Posts: 344
ID: 64131
Credit: 55,738,523
RAC: 25,545
321 LLR Amethyst: Earned 1,000,000 credits (1,137,410)Cullen LLR Amethyst: Earned 1,000,000 credits (1,005,544)ESP LLR Amethyst: Earned 1,000,000 credits (1,529,595)Generalized Cullen/Woodall LLR Amethyst: Earned 1,000,000 credits (1,020,167)PPS LLR Ruby: Earned 2,000,000 credits (2,023,233)PSP LLR Ruby: Earned 2,000,000 credits (2,065,730)SoB LLR Ruby: Earned 2,000,000 credits (2,036,592)SR5 LLR Ruby: Earned 2,000,000 credits (2,008,007)SGS LLR Ruby: Earned 2,000,000 credits (2,001,244)TRP LLR Ruby: Earned 2,000,000 credits (2,002,759)Woodall LLR Amethyst: Earned 1,000,000 credits (1,017,997)321 Sieve Ruby: Earned 2,000,000 credits (2,005,104)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,000,304)Generalized Cullen/Woodall Sieve Ruby: Earned 2,000,000 credits (2,087,920)PPS Sieve Jade: Earned 10,000,000 credits (11,092,841)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,313,121)TRP Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,284,804)AP 26/27 Amethyst: Earned 1,000,000 credits (1,750,619)GFN Turquoise: Earned 5,000,000 credits (5,599,058)PSA Turquoise: Earned 5,000,000 credits (8,756,474)
Message 67345 - Posted: 26 Jun 2013 | 4:17:50 UTC
Last modified: 26 Jun 2013 | 4:27:00 UTC

Actually we already thought about this trick, and it makes sense during a LLR challenge, but we didn't applied them yet.

But in other ordinar time it has significant deficiencies.
- you are forced to babysit and abort tasks online
- or you have to keep a large enough cache to abort long-time tasks 1 or 2 times a day. In this case your chances to be an initial finder tend to zero
- or you have to make some kind of "smart machine" which will kill / crash "high k value" tasks in an automatic mode. It's possible, I guess, but is it worth it? Badge? Credits? Cheater glory?
____________

Profile sergeProject donor
Avatar
Send message
Joined: 21 Jun 12
Posts: 110
ID: 144858
Credit: 199,518,009
RAC: 104,661
Eliminated 2 conjecture "k"s321 LLR Turquoise: Earned 5,000,000 credits (5,342,838)Cullen LLR Silver: Earned 100,000 credits (158,983)ESP LLR Gold: Earned 500,000 credits (505,527)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,268,642)PPS LLR Sapphire: Earned 20,000,000 credits (23,395,331)PSP LLR Silver: Earned 100,000 credits (120,541)SoB LLR Amethyst: Earned 1,000,000 credits (1,280,028)SR5 LLR Ruby: Earned 2,000,000 credits (2,649,390)SGS LLR Gold: Earned 500,000 credits (518,356)TRP LLR Silver: Earned 100,000 credits (141,131)Woodall LLR Silver: Earned 100,000 credits (116,405)Cullen/Woodall Sieve (suspended) Silver: Earned 100,000 credits (235,088)PPS Sieve Amethyst: Earned 1,000,000 credits (1,031,526)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Gold: Earned 500,000 credits (500,445)TRP Sieve (suspended) Gold: Earned 500,000 credits (507,938)GFN Double Bronze: Earned 100,000,000 credits (160,454,410)PSA Silver: Earned 100,000 credits (291,533)
Message 67348 - Posted: 26 Jun 2013 | 8:21:18 UTC
Last modified: 26 Jun 2013 | 8:22:01 UTC

To any interested reader-

Firstly, I find personal attacks a) uninteresting, b) unimportant for the common goal.
Thick skin helps and I am not a $100 bill to enjoy everyone's unanimous love.

Second, "there's nothing easier and more rewarding than telling the truth" (c).

Now, listen carefully to the following mathematical argument.

Lemma 1. Presieved to the same level, candidates (in this context, k*b^n+c) have the probability to be prime dependent on n, but almost not on k:
P(k,n) ~ O(1 / (n*log(b) + log(k))) ~ O(1/n), with n>>log(k).
Well known. Left as an easy exercise to the reader.

Lemma 2. The function T(k,n) of time spent on working on a candidate is monotonous on both k and n. For constant n, it has a range of 1.5 in extreme values in the working domain.
Monotonicity is easily demonstrated (both the algorithm and the LLR implementation are open; “use the source, Luke, use the source!”). The range of k dependence is easily quantified, -- it is an easy, technical problem. Boring!

Exercise 1. Given an objective function max P(k,n)/T(k,n), define the best strategy to find most primes per unit of time, under constant resources.
Spoiler: run the smallest k's until its speed is negated by next FFT size; then proceed with the 2nd smallest k, etc.
Exercise 2 (for the birds). Given an objective function "max credit/time", define the best strategy.
Spoiler: run the smallest k's until its speed is negated by next FFT size.

Bonus question: why is it the same strategy?

(In other words: can you tell, which of the two functions someone is maximizing? …and if you now realize that you actually can't, is that because one of the credit function is currently implemented wrongly? The time cost function doesn't have to be implemented, it simply presents itself. Is the solution (ORLY?) “wide range of punitive actions” against those who know math? Really, how dare they!)

Exercise 3 (sort of most important): In the absence of constraint of constant resources, given the open system of crunchers, how do you maximize throughput?

Here the answer is not so easy and it is not mathematical per se. To address it properly, another domain of knowledge is need - social psychology.

Here's a sketch of some strategies:
1. A short-term gain, followed by possible catastrophic loss: Debase the currency, and arbitrarily double credits. Independent of work units, or anything. Detach the concept of credits from actual computational complexity of done work. This is, in fact, happening in the other thread.) => Initially, the will be a swing of users from other projects - then they will retaliate (inflate their currency too if necessary), but this project will also pay in reputation (see #4).
2. Credit not for actual compute cycles spent, but for "effort" (for example, wall-clock-time spent). Variations are possible, e.g. to quote certain obscure sources, "здесь мерилом работы считают усталость". (If you cannot google-translate that, you can safely skip this last illustration.) This credit variation has been done, and we can safely agree that the result was not so good.
3. Credit not for actual compute cycles spent, but for loyalty ...or its covariates, age in project... etc etc etc
-OR-
4. Credit as fair as you can estimate. Don't take this option lightly. There are known demonstrations that this is a very important consideration even if irrational:
see http://en.wikipedia.org/wiki/Ultimatum_game . People will not take something for nothing in the Ultimatum game (which is rationally always a win) if they perceive that they were treated unfairly. Irrational but true.

I believe that if you will maximize fairness of treating your contributors, then you will maximize the throughput of the project. (And vice versa, duh!)

Opinions, questions?


P.S. Estimating T(k,n) as a smooth function of both k and n (currently a smooth function of n only), is quite easy (forget about about block jumps – and you already did for n; why not for k? Simple bivariate regression, too hard?)

I can give T(k,n) (normalized to current credit units) to you and you can verify its correctness, if you are too busy for that. Interested?

Profile j.sheridanProject donor
Send message
Joined: 21 Mar 11
Posts: 437
ID: 91622
Credit: 679,433,675
RAC: 0
321 LLR Ruby: Earned 2,000,000 credits (3,002,360)Cullen LLR Ruby: Earned 2,000,000 credits (3,040,127)PPS LLR Ruby: Earned 2,000,000 credits (4,146,102)PSP LLR Turquoise: Earned 5,000,000 credits (5,432,836)SoB LLR Turquoise: Earned 5,000,000 credits (5,623,361)SR5 LLR Gold: Earned 500,000 credits (640,957)SGS LLR Ruby: Earned 2,000,000 credits (4,053,556)TRP LLR Ruby: Earned 2,000,000 credits (3,826,598)Woodall LLR Ruby: Earned 2,000,000 credits (3,123,627)321 Sieve Silver: Earned 100,000 credits (237,768)Cullen/Woodall Sieve (suspended) Double Silver: Earned 200,000,000 credits (265,102,350)PPS Sieve Double Silver: Earned 200,000,000 credits (292,642,933)TRP Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,801,812)GFN Emerald: Earned 50,000,000 credits (82,759,285)
Message 67476 - Posted: 30 Jun 2013 | 20:37:43 UTC - in response to Message 67348.


Opinions?

yes, but you probably don't want or care to hear them and even if you did I'd probably get banned for expressing them here.

Profile DaveBProject donor
Avatar
Send message
Joined: 20 Jun 09
Posts: 351
ID: 42198
Credit: 11,898,570
RAC: 0
321 LLR Gold: Earned 500,000 credits (547,769)Cullen LLR Silver: Earned 100,000 credits (158,867)PPS LLR Silver: Earned 100,000 credits (162,534)PSP LLR Silver: Earned 100,000 credits (134,373)SoB LLR Gold: Earned 500,000 credits (510,012)SR5 LLR Silver: Earned 100,000 credits (266,757)SGS LLR Gold: Earned 500,000 credits (510,560)TRP LLR Silver: Earned 100,000 credits (242,439)Woodall LLR Gold: Earned 500,000 credits (508,989)321 Sieve Silver: Earned 100,000 credits (200,123)Cullen/Woodall Sieve (suspended) Silver: Earned 100,000 credits (229,801)PPS Sieve Ruby: Earned 2,000,000 credits (4,317,221)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Bronze: Earned 10,000 credits (21,147)TRP Sieve (suspended) Silver: Earned 100,000 credits (333,494)AP 26/27 Bronze: Earned 10,000 credits (24,117)GFN Ruby: Earned 2,000,000 credits (3,730,320)
Message 67479 - Posted: 30 Jun 2013 | 22:28:11 UTC - in response to Message 67348.

Interested?


No
____________
Member team AUSTRALIA
My lucky number is 9291*2^1085585+1

Profile Michael GoetzProject donor
Volunteer moderator
Project administrator
Project scientist
Avatar
Send message
Joined: 21 Jan 10
Posts: 12675
ID: 53948
Credit: 184,626,545
RAC: 147,034
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,063,182)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Ruby: Earned 2,000,000 credits (4,146,403)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Ruby: Earned 2,000,000 credits (2,773,744)PSP LLR Ruby: Earned 2,000,000 credits (2,632,269)SoB LLR Sapphire: Earned 20,000,000 credits (34,158,496)SR5 LLR Turquoise: Earned 5,000,000 credits (8,293,415)SGS LLR Ruby: Earned 2,000,000 credits (2,012,222)TRP LLR Ruby: Earned 2,000,000 credits (2,737,347)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Turquoise: Earned 5,000,000 credits (5,046,112)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (20,110,788)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Turquoise: Earned 5,000,000 credits (7,090,096)GFN Emerald: Earned 50,000,000 credits (64,594,991)PSA Jade: Earned 10,000,000 credits (10,305,150)
Message 67692 - Posted: 14 Jul 2013 | 14:12:04 UTC - in response to Message 67341.

If you don't put k dependency in your credit formula, you may see high k's WUs aborted systematically by impatient users: there's no reason for them to run a 1.5 time longer job for the same credit as a short one with low k.


We're not going to have 100 different credit calculations for 1 project, and we're not going to run 100s of tests to find out where all the FFT shift points are. The shift points vary by processor type, because different processors will use different FFTs. Even if we wanted to do this -- which we definitely do not -- it wouldn't work anyway because the behavior and timing will vary from system to system.

So, plain and simple, what you're suggesting isn't going to happen.

If we find that people are trying to game the system, there's no shortage of punitive strategies that can be employed to discourage such behavior.


It's been a while, so an update is in order.

The basic problem here is that on some projects, notably SR5 but also TRP, the variation in k is sufficient to produce significant changes in the run time which aren't accounted for in the credit formula.

The problem is due to different FFT sizes being used in different tasks. The credit formula is adjusted to work on a single FFT size for each project, or an average of all the FFT sizes on a project.

Because of the varying FFT sizes, there's noticeable differences in the credit per hour you will get between different tasks in those projects.

That's not good. It's not nearly as bad as the old credit formula, but it's still something we'd prefer to remove.

We have a revised credit formula that will compensate for this problem, treating each FFT size differently. It completely fixes the problem.

The reason we haven't implemented it is because it would introduce a new problem, and it's not clear if the new problem would be worse than the old problem.

The underlying problem with incorporating FFT size into the time estimate (and hence the credit) is that different CPUs will use different FFT sizes. Since the only way to know what FFT size is being used is to look at the output from LLR, this means we have to look at the result coming back from each LLR task to see which FFT size was used.

Using the FFT size reported by host computers is sort of a step backwards, in that it relies on data from the host to determine credit, much the way the original BOINC "benchmark+cpu_time" method did. This means that two different computers could request different amounts of credit for the same task because they used different FFTs.

We'd be back in the situation where your wingman would have an effect on how much credit you receive. This is the new problem that this system would introduce.

We haven't yet decided what to do. We might do nothing and leave things the way they are, we might go with the FFT system and live with having the CPU type affect, somewhat, the amount of credit each task gets, or we might do something else entirely.

If we leave things the way they are, please understand that systematically aborting tasks merely to get those that provide more credit should be considered to be against our policies and we may stop sending additional tasks to users who do this. That's what I meant by "punitive".
____________
Please do not PM me with support questions. Ask on the forums instead. Thank you!

My lucky number is 75898524288+1

Profile Michael GoetzProject donor
Volunteer moderator
Project administrator
Project scientist
Avatar
Send message
Joined: 21 Jan 10
Posts: 12675
ID: 53948
Credit: 184,626,545
RAC: 147,034
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,063,182)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Ruby: Earned 2,000,000 credits (4,146,403)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Ruby: Earned 2,000,000 credits (2,773,744)PSP LLR Ruby: Earned 2,000,000 credits (2,632,269)SoB LLR Sapphire: Earned 20,000,000 credits (34,158,496)SR5 LLR Turquoise: Earned 5,000,000 credits (8,293,415)SGS LLR Ruby: Earned 2,000,000 credits (2,012,222)TRP LLR Ruby: Earned 2,000,000 credits (2,737,347)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Turquoise: Earned 5,000,000 credits (5,046,112)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (20,110,788)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Turquoise: Earned 5,000,000 credits (7,090,096)GFN Emerald: Earned 50,000,000 credits (64,594,991)PSA Jade: Earned 10,000,000 credits (10,305,150)
Message 67693 - Posted: 14 Jul 2013 | 20:43:13 UTC

From the "be careful what you ask for, you just might get it" department:

We could force LLR to use the same, larger, LLR size for all SR5 tasks. That would effectively slow down the low-k tasks to the same speed as the high-k tasks, and make the credit flat amongst all the tasks.

And no, we won't really do that.

(That wasn't my idea. I'll refrain from saying whose idea it was in order to protect the guilty.)
____________
Please do not PM me with support questions. Ask on the forums instead. Thank you!

My lucky number is 75898524288+1

Profile sergeProject donor
Avatar
Send message
Joined: 21 Jun 12
Posts: 110
ID: 144858
Credit: 199,518,009
RAC: 104,661
Eliminated 2 conjecture "k"s321 LLR Turquoise: Earned 5,000,000 credits (5,342,838)Cullen LLR Silver: Earned 100,000 credits (158,983)ESP LLR Gold: Earned 500,000 credits (505,527)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,268,642)PPS LLR Sapphire: Earned 20,000,000 credits (23,395,331)PSP LLR Silver: Earned 100,000 credits (120,541)SoB LLR Amethyst: Earned 1,000,000 credits (1,280,028)SR5 LLR Ruby: Earned 2,000,000 credits (2,649,390)SGS LLR Gold: Earned 500,000 credits (518,356)TRP LLR Silver: Earned 100,000 credits (141,131)Woodall LLR Silver: Earned 100,000 credits (116,405)Cullen/Woodall Sieve (suspended) Silver: Earned 100,000 credits (235,088)PPS Sieve Amethyst: Earned 1,000,000 credits (1,031,526)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Gold: Earned 500,000 credits (500,445)TRP Sieve (suspended) Gold: Earned 500,000 credits (507,938)GFN Double Bronze: Earned 100,000,000 credits (160,454,410)PSA Silver: Earned 100,000 credits (291,533)
Message 67694 - Posted: 14 Jul 2013 | 21:03:31 UTC - in response to Message 67692.

Michael-

It's great that we are more on the same page as could have appeared from the previous impression.

Trying to use the FFT size report from the client is an interesting technical solution, and quite accurate. Based on your initial reaction, I didn't expect this to be even on the menu, but it is great that you considered that. Yes, some decisions will have to be made: for example, base all credit on the most abundant CPU type in the project (call it ref_CPU, an average SB, or is it IB already? ...sometime later Haswells may throw a monkey wrench in this construction, too). Then the credit function would be roughly N*T_per_iter(ref_CPU, self_reported_FFT). Notes: 1. linear dependence on N is always in effect, so we'll take it out of the fit right away; 2. The app binaries are already authenticated against forgeries, so the server can trust that self_reported_FFT is a true value.

I've run a different fit back then, because I thought that you aimed for simplicity (no stepwise functions). It was as follows:
- prepare LLR binary that quits after the very first report about T_per_iter (that's after 10,000 iterations)
- run while(1) { pick a random WU from the sieve_2M_5M; run ./llr -d a.npg; save output and cleanup z* files }
- After collecting a cloud of dots, I've run a few fits: on N only, on N and k, on the first PCA (which is unsurprisingly, log(N+c logK/K_av), with some c).

The best smooth fit is not great, but nevertheless removes half of the residual from the fit on N alone. The fit is guided by the known theoretical complexity. The final output matches the intuition: you can simply plug N+c logK/K_av into the already employed credit formula for T_per_iter. This doesn't take any feedback from the client software. If K_av is taken to be average(k) over the WU space, this formula will stitch with the old one. I can lookup my notes that I saved somewhere, what the c value was. Off the top of my head it was about ~10^5 (not tiny wrt N).

Your approach to use self_reported_FFT is an even better one. It will be interesting to see it in action.

Well, there is still a huge disparity between credit for different classes of tasks (up to 2x, 3x?) This makes all of these accuracy improvements rather like (if PrimeGrid was a house in need of TLC as they say) polishing a door handle when the whole kitchen may need remodeling. But that's a totally different story which I haven't even suspected before I've read Randy's observations in the other thread.

Anyway, there's nothing wrong with improving by one step at a time.

Yours looks like a good plan forward.
Cheers,
-S
____________
My lucky number is Phi(4, 2^2396029-1)/2.

Profile Michael GoetzProject donor
Volunteer moderator
Project administrator
Project scientist
Avatar
Send message
Joined: 21 Jan 10
Posts: 12675
ID: 53948
Credit: 184,626,545
RAC: 147,034
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,063,182)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Ruby: Earned 2,000,000 credits (4,146,403)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Ruby: Earned 2,000,000 credits (2,773,744)PSP LLR Ruby: Earned 2,000,000 credits (2,632,269)SoB LLR Sapphire: Earned 20,000,000 credits (34,158,496)SR5 LLR Turquoise: Earned 5,000,000 credits (8,293,415)SGS LLR Ruby: Earned 2,000,000 credits (2,012,222)TRP LLR Ruby: Earned 2,000,000 credits (2,737,347)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Turquoise: Earned 5,000,000 credits (5,046,112)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (20,110,788)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Turquoise: Earned 5,000,000 credits (7,090,096)GFN Emerald: Earned 50,000,000 credits (64,594,991)PSA Jade: Earned 10,000,000 credits (10,305,150)
Message 67695 - Posted: 14 Jul 2013 | 21:24:38 UTC - in response to Message 67694.

Yes, some decisions will have to be made: for example, base all credit on the most abundant CPU type in the project (call it ref_CPU, an average SB, or is it IB already? ...sometime later Haswells may throw a monkey wrench in this construction, too).


The majority of CPU's are NOT AVX capable. That includes pre-SB Intel as well as AMD. I prefer to use a Core2 CPU as the benchmark machine because it's the most modern Intel CPU that does not have Turboboost or hyperthreading, both of which can make repeatable benchmarks more challenging.

The app binaries are already authenticated against forgeries, so the server can trust that self_reported_FFT is a true value.


That's not true, because there's nothing guaranteeing that the Boinc client itself is genuine. The signature system is designed to assure the user that they are running our authentic software; it is not designed to guarantee to the server that the user is using our software.


____________
Please do not PM me with support questions. Ask on the forums instead. Thank you!

My lucky number is 75898524288+1

Profile sergeProject donor
Avatar
Send message
Joined: 21 Jun 12
Posts: 110
ID: 144858
Credit: 199,518,009
RAC: 104,661
Eliminated 2 conjecture "k"s321 LLR Turquoise: Earned 5,000,000 credits (5,342,838)Cullen LLR Silver: Earned 100,000 credits (158,983)ESP LLR Gold: Earned 500,000 credits (505,527)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,268,642)PPS LLR Sapphire: Earned 20,000,000 credits (23,395,331)PSP LLR Silver: Earned 100,000 credits (120,541)SoB LLR Amethyst: Earned 1,000,000 credits (1,280,028)SR5 LLR Ruby: Earned 2,000,000 credits (2,649,390)SGS LLR Gold: Earned 500,000 credits (518,356)TRP LLR Silver: Earned 100,000 credits (141,131)Woodall LLR Silver: Earned 100,000 credits (116,405)Cullen/Woodall Sieve (suspended) Silver: Earned 100,000 credits (235,088)PPS Sieve Amethyst: Earned 1,000,000 credits (1,031,526)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Gold: Earned 500,000 credits (500,445)TRP Sieve (suspended) Gold: Earned 500,000 credits (507,938)GFN Double Bronze: Earned 100,000,000 credits (160,454,410)PSA Silver: Earned 100,000 credits (291,533)
Message 67696 - Posted: 14 Jul 2013 | 21:44:06 UTC - in response to Message 67695.

The app binaries are already authenticated against forgeries, so the server can trust that self_reported_FFT is a true value.


That's not true, because there's nothing guaranteeing that the Boinc client itself is genuine. The signature system is designed to assure the user that they are running our authentic software; it is not designed to guarantee to the server that the user is using our software.

Hmm, this sounds bad, and it would be even worse to go over details here, but I know what you mean. I am sure that you guys are thinking about fixing that. The usual argument that no one would care to break in because there's nothing to be gained was apparently recently demonstrated to be a weak one (even if the attack was not necessarily by forging the client); some people apparently care to break things.

If it is any consolation, I can assure you that existential forgery attacks against GIMPS were "successful" as well. For a white hat exercise and having spent only a couple hours tinkering, I've had submitted some fairly bogus results to the GIMPS server: "the largest ever Lucas Lehmer test", followed by "the largest ever Lucas Lehmer test double-checked (with a different shift value)", both accepted by the server. The result was of course cleared from the database after half a day, but I am not sure whether they updated their security as a follow-up. (The bug report, i.e. the full description of the attack, was of course filed to GW and SK.)

Profile Michael GoetzProject donor
Volunteer moderator
Project administrator
Project scientist
Avatar
Send message
Joined: 21 Jan 10
Posts: 12675
ID: 53948
Credit: 184,626,545
RAC: 147,034
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,063,182)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Ruby: Earned 2,000,000 credits (4,146,403)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Ruby: Earned 2,000,000 credits (2,773,744)PSP LLR Ruby: Earned 2,000,000 credits (2,632,269)SoB LLR Sapphire: Earned 20,000,000 credits (34,158,496)SR5 LLR Turquoise: Earned 5,000,000 credits (8,293,415)SGS LLR Ruby: Earned 2,000,000 credits (2,012,222)TRP LLR Ruby: Earned 2,000,000 credits (2,737,347)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Turquoise: Earned 5,000,000 credits (5,046,112)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (20,110,788)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Turquoise: Earned 5,000,000 credits (7,090,096)GFN Emerald: Earned 50,000,000 credits (64,594,991)PSA Jade: Earned 10,000,000 credits (10,305,150)
Message 67697 - Posted: 14 Jul 2013 | 22:04:23 UTC - in response to Message 67696.

The app binaries are already authenticated against forgeries, so the server can trust that self_reported_FFT is a true value.


That's not true, because there's nothing guaranteeing that the Boinc client itself is genuine. The signature system is designed to assure the user that they are running our authentic software; it is not designed to guarantee to the server that the user is using our software.

Hmm, this sounds bad, and it would be even worse to go over details here, but I know what you mean. I am sure that you guys are thinking about fixing that. The usual argument that no one would care to break in because there's nothing to be gained was apparently recently demonstrated to be a weak one (even if the attack was not necessarily by forging the client); some people apparently care to break things.

If it is any consolation, I can assure you that existential forgery attacks against GIMPS were "successful" as well. For a white hat exercise and having spent only a couple hours tinkering, I've had submitted some fairly bogus results to the GIMPS server: "the largest ever Lucas Lehmer test", followed by "the largest ever Lucas Lehmer test double-checked (with a different shift value)", both accepted by the server. The result was of course cleared from the database after half a day, but I am not sure whether they updated their security as a follow-up. (The bug report, i.e. the full description of the attack, was of course filed to GW and SK.)


One of my hobbies is running MMORPG games (and playing them), and a perpetual problem in that industry is that, fundamentally, you can't control what's running on the user's computer. There are smart people out there, and some of them have way too much time on their hands, and they WILL find a way to cheat. Or steal.

Preventing users from intentionally running forged clients is probably impossible. One drastic step taken by Blizzard to protect the integrity of their games, most notably World of Warcraft, is to have a background task running that scans all the processes on the computer, looking for specific cheating software. It's not 100% effective, AND it most certainly rubs some people the wrong way.

The BOINC approach to this is to not worry if users want to run their own software. It is, in fact, encouraged in order to allow operations on platforms that aren't officially supported.

Security is provided not by forcing people to run our software, but by validating the results that come back.
____________
Please do not PM me with support questions. Ask on the forums instead. Thank you!

My lucky number is 75898524288+1

Profile Michael GoetzProject donor
Volunteer moderator
Project administrator
Project scientist
Avatar
Send message
Joined: 21 Jan 10
Posts: 12675
ID: 53948
Credit: 184,626,545
RAC: 147,034
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,063,182)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Ruby: Earned 2,000,000 credits (4,146,403)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Ruby: Earned 2,000,000 credits (2,773,744)PSP LLR Ruby: Earned 2,000,000 credits (2,632,269)SoB LLR Sapphire: Earned 20,000,000 credits (34,158,496)SR5 LLR Turquoise: Earned 5,000,000 credits (8,293,415)SGS LLR Ruby: Earned 2,000,000 credits (2,012,222)TRP LLR Ruby: Earned 2,000,000 credits (2,737,347)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Turquoise: Earned 5,000,000 credits (5,046,112)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (20,110,788)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Turquoise: Earned 5,000,000 credits (7,090,096)GFN Emerald: Earned 50,000,000 credits (64,594,991)PSA Jade: Earned 10,000,000 credits (10,305,150)
Message 69651 - Posted: 5 Oct 2013 | 20:30:32 UTC

The credit formula has been adjusted to compensate for the effect of different FFT sizes on computation time.

Please see this thread for more information.
____________
Please do not PM me with support questions. Ask on the forums instead. Thank you!

My lucky number is 75898524288+1

Message boards : Sierpinski/Riesel Base 5 Problem : Strange Times

[Return to PrimeGrid main page]
DNS Powered by DNSEXIT.COM
Copyright © 2005 - 2019 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 1.01, 1.07, 1.07
Generated 19 Aug 2019 | 6:10:40 UTC