TPS Anniversary Rally: April 13-18

Oddball · 2011-03-19, 08:31

Five years ago on April 13, this post started the TPS project: http://www.mersenneforum.org/showpos...02&postcount=1 . This April, we'll be having a sieve rally that begins on the 13th at 22:49 GMT, since that's when MooooMoo posted his message. The rally will be five days long: one day for each year that the project has existed. To participate, head over to the sieve reservation thread at http://www.mersenneforum.org/showthread.php?t=12260 , download the n=485K-490K file (if you don't have it already), and reserve a range. As a reference, one core of a modern CPU can complete ~2T/day.

A sieve rally has never been held before on Mersenneforum, so things might seem a bit tricky and confusing at first. But we'll make it work by making rally stats depend on when you reserve and complete your range(s). To account for different time zones, there will be a 24 hour grace period to reserve ranges and turn them in. If you finish your range in less than 5 days, you'll get credit for the entire range in the rally stats; if not, you'll only get partial (71.4%-99.9%) credit. Here are some examples:

- A 30T range reserved on April 13, 23:53 GMT and completed on April 18, 10:28 GMT will show up as 30T on the rally stats.
- A 30T range reserved on April 13, 11:00 GMT and completed on April 19, 11:00 GMT will show up as 25T on the rally stats. (5 days / 6 days * 30T = 25T)
- A 30T range reserved before April 12, 22:49 GMT or completed after April 19, 22:49 GMT will not be counted in the rally stats, but it will still show up on the stats in the sieve reservation thread.

The theme for this rally will be North America vs. the rest of the world. When you reserve a range, say what continent you're currently living in and what team (if any) you're a part of. The more you sieve, the higher you'll be on the ranking list.

Total rally stats (accounting for all work done between April 13 and April 18)

Code:

User             Amount sieved (in T)
Lennart             402
cipher              196.6
Oddball             68
gribozavr           43
mdettweiler         36.7
Flatlander          30
 
Team                          Amount sieved (in T)
Prime Search Team                 402
Everyone else                     307.6
Raiders of the Lost Primes        66.7
 
Continent           Amount sieved (in T)
Europe                 475
North America          301.3
 
Amount sieved: 776.3 T
Sieve rate: 1797 M p/sec
Factors found: ~240,000

Total rally stats (original rules):

Code:

User             Amount sieved (in T)
Lennart             402
cipher              196.6
Oddball             68
Flatlander          30
mdettweiler         15.4
 
Team                          Amount sieved (in T)
Prime Search Team                 402
Everyone else                     264.6
Raiders of the Lost Primes        45.4
 
Continent           Amount sieved (in T)
Europe                 432
North America          280
 
Amount sieved: 712 T
Sieve rate: 1648 M p/sec
Factors found: ~210,000

For reference, the stats and info from the previous rally can be found here: http://www.mersenneforum.org/showthread.php?t=14626 It'll be interesting to see who will be at the top of the rankings this time

Happy sieving!

Flatlander · 2011-03-20, 18:08

Need to be careful with people reserving ranges in a short period of time. Danger of unintentional overlap.

TimSorbet · 2011-03-20, 18:21

Quote:

Originally Posted by Flatlander

Need to be careful with people reserving ranges in a short period of time. Danger of unintentional overlap.

Simple rules can solve this: always check the post before yours before and after you post (then adjust your reservation if necessary). Avoid editing reservations posts to change ranges, but if you do edit your post, make sure that when you're done, there are no new posts that don't take yours into account (then adjust your reservation if necessary).

ellipse · 2011-04-10, 19:50

Are there any plans for a TPSieveNet to automate sieving? Primegrid had a sieve rally a few months ago which was automated, and I think it went quite well.

mdettweiler · 2011-04-10, 21:52

Quote:

Originally Posted by ellipse

Are there any plans for a TPSieveNet to automate sieving? Primegrid had a sieve rally a few months ago which was automated, and I think it went quite well.

Not at this time, no. Sieving is an entirely different animal from LLRing, so the client/server setup would likely have to be written from scratch (possibly some communication and database code could be borrowed from PRPnet, but not much beyond that). I would take a whack at it myself if I had the time...I would need to learn a lot more about programming client/server code (not to mention C++ if I was to borrow anything from PRPnet) before I could even attempt such a project.

PrimeGrid's automated sieving setup utilizes BOINC, which is very easy to set up and use on the client end, but much more complex on the server side. For a project the size of TPS, it's really not worth the effort of getting it all set up. If PrimeGrid would be willing to resume their former partnership with TPS, I'm sure they could quite easily integrate TPS sieving work; they already have a BOINC-ready tpsieve developed for their PPS Sieve subproject. The big question would be whether PrimeGrid wants to take on yet another BOINC subproject at this time, since they've already got a gazillion of them that they're stretched over.

Oddball · 2011-04-11, 05:10

Quote:

Originally Posted by ellipse

Are there any plans for a TPSieveNet to automate sieving? Primegrid had a sieve rally a few months ago which was automated, and I think it went quite well.

The closest we've got to an automated system was a website called "TwinPrimeSearch.org" that let people reserve ranges. Suppose you want to reserve 550-560, and all ranges above 500 were available. You'll then type it in, and the system will let you have that range if nobody takes it before you. However, if someone else reserved 500-600 before you submitted a range, you'll get an error message. There might even have been an option to edit ranges and an option to let the server give you the lowest available range that is, say, 1.5 G long. The system was designed for LLR reservations, but it shouldn't be that hard to adapt it to sieve reservations. Just have people submit factors found instead of primes found.

Disclaimer: What I've said regarding the TPS website may not be accurate. It was created long before I came to TPS, so I've tried to make a reasonable guess of its activity based on these posts:
http://www.mersenneforum.org/showpos...5&postcount=96
http://www.mersenneforum.org/showpos...&postcount=142
http://replay.waybackmachine.org/200...imesearch.org/

ellipse · 2011-04-13, 05:36

Quote:

Originally Posted by mdettweiler

Not at this time, no. Sieving is an entirely different animal from LLRing, so the client/server setup would likely have to be written from scratch (possibly some communication and database code could be borrowed from PRPnet, but not much beyond that). I would take a whack at it myself if I had the time...I would need to learn a lot more about programming client/server code (not to mention C++ if I was to borrow anything from PRPnet) before I could even attempt such a project.

PrimeGrid's automated sieving setup utilizes BOINC, which is very easy to set up and use on the client end, but much more complex on the server side. For a project the size of TPS, it's really not worth the effort of getting it all set up. If PrimeGrid would be willing to resume their former partnership with TPS, I'm sure they could quite easily integrate TPS sieving work; they already have a BOINC-ready tpsieve developed for their PPS Sieve subproject. The big question would be whether PrimeGrid wants to take on yet another BOINC subproject at this time, since they've already got a gazillion of them that they're stretched over.

Thanks for the info. IMO, if anything goes BOINC, it should be the LLR part and not the sieve. We're already at optimal depth for n=480K-485K and close to halfway there for n=485K-490K, but less than 2% of the range has been LLRed. Of course, the BOINC decision would be left to Oddball and the Primegrid admins.

Anyway, it's now April 13th, and there's not much activity going on in the sieve reservation thread. Anyone want to join in the rally?

cipher · 2011-04-13, 17:17

Agreed LLR needs to go to BOINC.

Oddball · 2011-04-13, 22:49

Quote:

Originally Posted by Oddball

Five years ago on April 13, this post started the TPS project: http://www.mersenneforum.org/showpos...02&postcount=1 . This April, we'll be having a sieve rally that begins on the 13th at 22:49 GMT, since that's when MooooMoo posted his message.

It's now April 13, 22:49 GMT. Happy birthday TPS!

Oddball · 2011-04-21, 19:44

The rally results are in. Lennart was first in the individual rankings, and he led Europe and the Prime Search Team to victory in the team and continent rankings as well. Overall, the project sieved 776.3T in five days and found almost a quarter million factors. At that rate, we'll get through a 10T range in about an hour and a half.

Pardon the delay in posting the results. I've bent the rules a bit to account for all sieving done from April 13 to April 18, so I waited until Max's 3136-3168T range (reserved on April 13, completed on April 21) and gribozavr's 1505-1700T range (reserved on March 29, completed on April 21) were done. Although they were finished after the grace period was over, they will get partial credit of 21.3T and 43T, respectively.

For reference, the standings with the original rules are also included in the first post. The individual, team, and continent rankings are the same with one exception: Flatlander drops from 4th place to 6th.

Looking forward, we'll be having an LLR rally in summer. Unlike previous rallies, the upcoming one will get you a spot in the top 5000 list if you find a prime, even if it's not twin. No bonus points for guessing what subproject we'll be working on then

2011-03-19, 08:31	#1
Oddball May 2010 499 Posts	TPS Anniversary Rally: April 13-18 Five years ago on April 13, this post started the TPS project: http://www.mersenneforum.org/showpos...02&postcount=1 . This April, we'll be having a sieve rally that begins on the 13th at 22:49 GMT, since that's when MooooMoo posted his message. The rally will be five days long: one day for each year that the project has existed. To participate, head over to the sieve reservation thread at http://www.mersenneforum.org/showthread.php?t=12260 , download the n=485K-490K file (if you don't have it already), and reserve a range. As a reference, one core of a modern CPU can complete ~2T/day. A sieve rally has never been held before on Mersenneforum, so things might seem a bit tricky and confusing at first. But we'll make it work by making rally stats depend on when you reserve and complete your range(s). To account for different time zones, there will be a 24 hour grace period to reserve ranges and turn them in. If you finish your range in less than 5 days, you'll get credit for the entire range in the rally stats; if not, you'll only get partial (71.4%-99.9%) credit. Here are some examples: - A 30T range reserved on April 13, 23:53 GMT and completed on April 18, 10:28 GMT will show up as 30T on the rally stats. - A 30T range reserved on April 13, 11:00 GMT and completed on April 19, 11:00 GMT will show up as 25T on the rally stats. (5 days / 6 days * 30T = 25T) - A 30T range reserved before April 12, 22:49 GMT or completed after April 19, 22:49 GMT will not be counted in the rally stats, but it will still show up on the stats in the sieve reservation thread. The theme for this rally will be North America vs. the rest of the world. When you reserve a range, say what continent you're currently living in and what team (if any) you're a part of. The more you sieve, the higher you'll be on the ranking list. Total rally stats (accounting for all work done between April 13 and April 18) Code: User Amount sieved (in T) Lennart 402 cipher 196.6 Oddball 68 gribozavr 43 mdettweiler 36.7 Flatlander 30 Team Amount sieved (in T) Prime Search Team 402 Everyone else 307.6 Raiders of the Lost Primes 66.7 Continent Amount sieved (in T) Europe 475 North America 301.3 Amount sieved: 776.3 T Sieve rate: 1797 M p/sec Factors found: ~240,000 Total rally stats (original rules): Code: User Amount sieved (in T) Lennart 402 cipher 196.6 Oddball 68 Flatlander 30 mdettweiler 15.4 Team Amount sieved (in T) Prime Search Team 402 Everyone else 264.6 Raiders of the Lost Primes 45.4 Continent Amount sieved (in T) Europe 432 North America 280 Amount sieved: 712 T Sieve rate: 1648 M p/sec Factors found: ~210,000 For reference, the stats and info from the previous rally can be found here: http://www.mersenneforum.org/showthread.php?t=14626 It'll be interesting to see who will be at the top of the rankings this time Happy sieving! Last fiddled with by Oddball on 2011-04-21 at 19:46

2011-03-20, 18:08	#2
Flatlander I quite division it "Chris" Feb 2005 England 31·67 Posts	Need to be careful with people reserving ranges in a short period of time. Danger of unintentional overlap. Last fiddled with by Flatlander on 2011-03-20 at 18:09

2011-04-21, 19:44	#10
Oddball May 2010 499 Posts	The rally results are in. Lennart was first in the individual rankings, and he led Europe and the Prime Search Team to victory in the team and continent rankings as well. Overall, the project sieved 776.3T in five days and found almost a quarter million factors. At that rate, we'll get through a 10T range in about an hour and a half. Pardon the delay in posting the results. I've bent the rules a bit to account for all sieving done from April 13 to April 18, so I waited until Max's 3136-3168T range (reserved on April 13, completed on April 21) and gribozavr's 1505-1700T range (reserved on March 29, completed on April 21) were done. Although they were finished after the grace period was over, they will get partial credit of 21.3T and 43T, respectively. For reference, the standings with the original rules are also included in the first post. The individual, team, and continent rankings are the same with one exception: Flatlander drops from 4th place to 6th. Looking forward, we'll be having an LLR rally in summer. Unlike previous rallies, the upcoming one will get you a spot in the top 5000 list if you find a prime, even if it's not twin. No bonus points for guessing what subproject we'll be working on then Last fiddled with by Oddball on 2011-04-21 at 19:50 Reason: future stuff

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Ramanujan's 125 birth anniversary	garo	Lounge	1	2011-12-30 23:57
LLRnet/PRPnet rally April 4th-11th	mdettweiler	No Prime Left Behind	55	2011-04-25 09:35
Happy 50th Anniversary, Yuri Gagarin	ewmayer	Science & Technology	2	2011-04-14 01:19
Anniversary.	Xyzzy	Lounge	4	2007-08-27 11:42
Anniversary	R.D. Silverman	Factoring	18	2007-02-19 12:55

2011-04-10, 19:50	#4
ellipse Nov 2010 2·3·5 Posts	Are there any plans for a TPSieveNet to automate sieving? Primegrid had a sieve rally a few months ago which was automated, and I think it went quite well.

2011-04-13, 17:17	#8
cipher Feb 2007 323₈ Posts	Agreed LLR needs to go to BOINC.