Join PrimeGrid
Returning Participants
Community
Leader Boards
Results
Other
drummers-lowrise
|
Message boards :
Project Staging Area :
2015 PRPNet June Challenge
| Author |
Message |
|---|
 RogerVolunteer developer
Volunteer tester
Send message
Joined: 27 Nov 11
Posts: 1138
ID: 120786
Credit: 268,668,824
RAC: 0
                    
|
|
GFN65536 - Solstice Challenge - 20-27th June 2015
Announced is a one week GFN65536 (Generalized Fermat Number Prime Search N=65536) challenge from the 20th until the 27th of June. I suggest we start at 12:00 UTC and end at 12:00 UTC.
Around the 21st of June, the Sun is at its most northerly declination (+23.5 degrees). This corresponds to the northern summer solstice and marks the longest day of the year for the northern hemisphere. In contrast, this is the date of the southern winter solstice and marks the shortest day of the year for the southern hemisphere.
The largest prime found in this sub-project so far is 3553248^65536+1 by Scott Brown on the 8th of April 2015. Search is currently completed to b=3,574,500. We expect 4.8 primes between here and b=4,000,000.
Source: http://yves.gallot.pagesperso-orange.fr/primes/stat.html
More about the Generalized Fermat Number Search can be found here. News and infos about the PRPNet client can be found here.
To take part, you have to activate the following lines in prpclient.ini:
server=GFN65536:100:1:prpnet.primegrid.com:12003
The Genefer80 transform will be used exclusively at N=65536. Genefer automatically uses this transform so no need to do anything other than make sure the geneferexe=genefer64.exe (or genefer32.exe if your running 32 bit OS) line is uncommented in the master_prpclient.ini. Other geneferexe lines should be commented out as ocl and cuda are beyond their b-limits. Remember you also need to run the "install", "update" and "start" batch file for the appropriate number of cores on your PC. Give it a go before the challenge to iron out any problems.
Stats will be available 15 minutes after the start at the well known place here.
All previous PRPNet challenge stats can be found here.
Good luck!
____________
 | |
|
Sysadm@Nbg  Volunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 5 Feb 08
Posts: 1283
ID: 18646
Credit: 1,120,584,411
RAC: 145,410
                        
|
|
Congrats to Scott_Brown for being the "Early Bird" :)
Stats are running, happy challenging!
____________
Sysadm@Nbg
my current lucky number: 5195*2^2241611+1
PSA-PRPNet-Stats-URL: http://u-g-f.de/PRPNet/
 | |
|
RafaelVolunteer tester
Send message
Joined: 22 Oct 14
Posts: 994
ID: 370496
Credit: 963,442,412
RAC: 343,781
                            
|
GFN65536 - Solstice Challenge - 20-27th June 2015
Announced is a one week GFN65536 (Generalized Fermat Number Prime Search N=65536) challenge from the 20th until the 27th of June. I suggest we start at 12:00 UTC and end at 12:00 UTC.
Around the 21st of June, the Sun is at its most northerly declination (+23.5 degrees). This corresponds to the northern summer solstice and marks the longest day of the year for the northern hemisphere. In contrast, this is the date of the southern winter solstice and marks the shortest day of the year for the southern hemisphere.
The largest prime found in this sub-project so far is 3553248^65536+1 by Scott Brown on the 8th of April 2015. Search is currently completed to b=3,574,500. We expect 4.8 primes between here and b=4,000,000.
Source: http://yves.gallot.pagesperso-orange.fr/primes/stat.html
More about the Generalized Fermat Number Search can be found here. News and infos about the PRPNet client can be found here.
To take part, you have to activate the following lines in prpclient.ini:
server=GFN65536:100:1:prpnet.primegrid.com:12003
The Genefer80 transform will be used exclusively at N=65536. Genefer automatically uses this transform so no need to do anything other than make sure the geneferexe=genefer64.exe (or genefer32.exe if your running 32 bit OS) line is uncommented in the master_prpclient.ini. Other geneferexe lines should be commented out as ocl and cuda are beyond their b-limits. Remember you also need to run the "install", "update" and "start" batch file for the appropriate number of cores on your PC. Give it a go before the challenge to iron out any problems.
Stats will be available 15 minutes after the start at the well known place here.
All previous PRPNet challenge stats can be found here.
Good luck!
So, uh, I kinda want to participate, but I'm a little confused. A couple quick questions:
1- Is it GFN (short), GFN WR, or neither? How do I set my preference pages accordingly?
2- Can you explain that code tweaking a bit further? Now, I have tinkered with the cc_config and app_config files, as well as moving files locally before, but in those cases, I knew what a I wanted and where to find it. This prpclient.ini file and those tags are unknown to me.
3- Do I need client restart, read config files, or just project update (or maybe even nothing at all)? | |
|
Sysadm@Nbg Volunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 5 Feb 08
Posts: 1283
ID: 18646
Credit: 1,120,584,411
RAC: 145,410
|
More about the Generalized Fermat Number Search can be found here. News and infos about the PRPNet client can be found here.
So, uh, I kinda want to participate, but I'm a little confused. A couple quick questions:
1- Is it GFN (short), GFN WR, or neither? How do I set my preference pages accordingly?
2- Can you explain that code tweaking a bit further? Now, I have tinkered with the cc_config and app_config files, as well as moving files locally before, but in those cases, I knew what a I wanted and where to find it. This prpclient.ini file and those tags are unknown to me.
3- Do I need client restart, read config files, or just project update (or maybe even nothing at all)?
It is not a BOINC-Challenge! Forget about the BOINC-things lik cc_config et al.
You had to set up the PRPNet-Client and choose there in the client-configuration the GFN65536-server-port
Links in the quote above (yours shortend to the relevant links)
____________
Sysadm@Nbg
my current lucky number: 5195*2^2241611+1
PSA-PRPNet-Stats-URL: http://u-g-f.de/PRPNet/
| |
|
RafaelVolunteer tester
Send message
Joined: 22 Oct 14
Posts: 994
ID: 370496
Credit: 963,442,412
RAC: 343,781
|
More about the Generalized Fermat Number Search can be found here. News and infos about the PRPNet client can be found here.
So, uh, I kinda want to participate, but I'm a little confused. A couple quick questions:
1- Is it GFN (short), GFN WR, or neither? How do I set my preference pages accordingly?
2- Can you explain that code tweaking a bit further? Now, I have tinkered with the cc_config and app_config files, as well as moving files locally before, but in those cases, I knew what a I wanted and where to find it. This prpclient.ini file and those tags are unknown to me.
3- Do I need client restart, read config files, or just project update (or maybe even nothing at all)?
It is not a BOINC-Challenge! Forget about the BOINC-things lik cc_config et al.
You had to set up the PRPNet-Client and choose there in the client-configuration the GFN65536-server-port
Links in the quote above (yours shortend to the relevant links)
Meh.... Well, I guess ty for the info anyway.
| |
|
|
|
|
May as well of been talking Chinese to me...this thread has made me laugh literally...very funny indeed...obviously I am brain dead so I will just leave this alone once again and go back to my box and continue with the usual stuff...lolol | |
|
Grebuloner Volunteer tester
Send message
Joined: 2 Nov 09
Posts: 658
ID: 49572
Credit: 5,081,727,882
RAC: 1,036,629
                                   
|
|
#3! Though I don't expect that to last...
Interesting observation on the stats this morning, since there are so few current participants at the moment, it's one user per team. So why are some of the credit numbers different between the user stats and team stats for the same work?
1 Lumiukko 103 190,088
2 Ross* 86 158,727
3 Grebuloner 60 110,740
1 PrimeSearchTeam 103 190,087
2 Sicituradastra. 86 158,727
3 The_Knights_Who_Say_Ni! 60 110,739
I know it's just one silly point here and there, but I'm still curious.
____________
Eating more cheese on Thursdays. | |
|
|
|
|
My second computer's genefer64 && 32 bits are crashing right after the Using AVX(intel) Transform is written to the console.
CPU genefer works fine in boinc but not in PRPnet, using windows 7 64 bit. Any ideas what may cause it?
Also could we use gpus for this by any chance? :) | |
|
Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14636
ID: 53948
Credit: 1,016,504,780
RAC: 716,935
                
|
My second computer's genefer64 && 32 bits are crashing right after the Using AVX(intel) Transform is written to the console.
CPU genefer works fine in boinc but not in PRPnet, using windows 7 64 bit. Any ideas what may cause it?
Also could we use gpus for this by any chance? :)
GFN 32768 and GFN 65536 are well beyond the B range where anything other than the x87 transform can work. No chance whatsoever of a GPU version of Genefer working on current (or project future) GPUs.
Furthermore, the GPUs are less efficient on smaller Ns -- they spend proportionally more time setting up each calculation and less time actually calculating, so efficiency would be significantly lower even if running the GPU apps were possible.
____________
My lucky number is 75898524288+1 | |
|
|
|
My second computer's genefer64 && 32 bits are crashing right after the Using AVX(intel) Transform is written to the console.
CPU genefer works fine in boinc but not in PRPnet, using windows 7 64 bit. Any ideas what may cause it?
Also could we use gpus for this by any chance? :)
GFN 32768 and GFN 65536 are well beyond the B range where anything other than the x87 transform can work. No chance whatsoever of a GPU version of Genefer working on current (or project future) GPUs.
Furthermore, the GPUs are less efficient on smaller Ns -- they spend proportionally more time setting up each calculation and less time actually calculating, so efficiency would be significantly lower even if running the GPU apps were possible.
Thanks for this info, as a current student it's something that I'll have to explore as it's pretty neat. | |
|
|
|
|
Out of interest what do people use as the machine ID for the PRP client? Does the client write checkpoints if so how often? | |
|
RogerVolunteer developer
Volunteer tester
Send message
Joined: 27 Nov 11
Posts: 1138
ID: 120786
Credit: 268,668,824
RAC: 0
|
Out of interest what do people use as the machine ID for the PRP client? Does the client write checkpoints if so how often?
machineid differentiates clients using the same e-mail ID.
You can set it to anything you like. There is no wrong answer.
I set mine to give info to others about my machine, i.e. X61100HD7970 = X6 1100T AMD CPU & HD7970 AMD GPU
If you find a prime the machineid is stored with the prime on the server.
I started using BOINC Computer ID for my machineid but moved to something more tangible long term as Computer ID changes every time you have to reinstall the OS.
Looking at the genefer.ckpt file, seems to be re-written at least every minute. | |
|
Sysadm@Nbg Volunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 5 Feb 08
Posts: 1283
ID: 18646
Credit: 1,120,584,411
RAC: 145,410
|
#3! Though I don't expect that to last...
Interesting observation on the stats this morning, since there are so few current participants at the moment, it's one user per team. So why are some of the credit numbers different between the user stats and team stats for the same work?
1 Lumiukko 103 190,088
2 Ross* 86 158,727
3 Grebuloner 60 110,740
1 PrimeSearchTeam 103 190,087
2 Sicituradastra. 86 158,727
3 The_Knights_Who_Say_Ni! 60 110,739
I know it's just one silly point here and there, but I'm still curious.
I 've seen this too; no fact on my side;
I compute the number now ./. the number at start; the numbers come from the official stats pages at http://prpnet.primegrid.com:12003/
for workunits the results are equal between team and member
for score the results differ by 1 (but not at all user/team comparison);
about the cause I only can speculate; maybe some sort of rounding error or (not seen) summing of decimal places
____________
Sysadm@Nbg
my current lucky number: 5195*2^2241611+1
PSA-PRPNet-Stats-URL: http://u-g-f.de/PRPNet/
| |
|
RogerVolunteer developer
Volunteer tester
Send message
Joined: 27 Nov 11
Posts: 1138
ID: 120786
Credit: 268,668,824
RAC: 0
|
|
After 47 hours we've done 9322 WUs!
Leading edge now at 3,659,394.
Number of GF primes expected between 3,574,500 and 3,659,394:
Chance of no prime: 38.28% (100%)
Chance of 1 prime_: 36.76% (62%)
Chance of 2 primes: 17.65% (25%)
Chance of 3 primes: 5.65% (7%)
Chance of 4 primes: 1.36% (2%)
Chance of 5 primes: 0.26% (0%)
Expected number of GF primes = 0.9571
=0.3676+2*0.1765+3*0.0565+4*0.0135+5*0.0026 | |
|
|
|
Out of interest what do people use as the machine ID for the PRP client? Does the client write checkpoints if so how often?
machineid differentiates clients using the same e-mail ID.
You can set it to anything you like. There is no wrong answer.
I set mine to give info to others about my machine, i.e. X61100HD7970 = X6 1100T AMD CPU & HD7970 AMD GPU
If you find a prime the machineid is stored with the prime on the server.
I started using BOINC Computer ID for my machineid but moved to something more tangible long term as Computer ID changes every time you have to reinstall the OS.
Looking at the genefer.ckpt file, seems to be re-written at least every minute.
Thanks Roger. Things are ticking along nicely. I'm not sure with the checkpoint file is located however I certainly know it is working when I close the command window using Ctrl + C. For the purpose of this challenge I have also used my Boinc ID= Speedy | |
|
RogerVolunteer developer
Volunteer tester
Send message
Joined: 27 Nov 11
Posts: 1138
ID: 120786
Credit: 268,668,824
RAC: 0
|
|
Leading edge has now passed 3,700,000
Chance of no prime: 24.19% | |
|
RogerVolunteer developer
Volunteer tester
Send message
Joined: 27 Nov 11
Posts: 1138
ID: 120786
Credit: 268,668,824
RAC: 0
|
|
Form of prime we're searching is: b*65536+1. Lets take a closer look at the chance of prime.
The primes found are listed on the PRPNet server: http://prpnet.primegrid.com:12003/user_primes.html
and earlier ones on OEIS: https://oeis.org/A251597
In the graph below you'll see the b-gap between prime finds ranked by size of the gap with an exponential line of best fit.
Leading edge now ~3,733,118. Gap from the previous prime find ~179,870 and growing, shown as red in the graph.
Certainly looks like a big jump from b-gap 178,522 to 254,584 either side of the red column.
Maybe we'll find a prime before the b-gap grows to 254,584 and help fill in this distribution, i.e. b < 3,807,832.
Note that the current gap is ranked 39 out of 44, so 86% of gaps (found so far) are smaller than the current one. | |
|
|
|
|
Note, however, that it is expected that the typical gap size (i.e. the difference between consecutive b values such that b^65536+1 are prime) is expected to increase (quite slowly!) as we move towards higher and higher numbers.
What does the graph look like if you do not "sort" the columns? Is this increasing gap trend visible yet?
/JeppeSN | |
|
RogerVolunteer developer
Volunteer tester
Send message
Joined: 27 Nov 11
Posts: 1138
ID: 120786
Credit: 268,668,824
RAC: 0
|
What does the graph look like if you do not "sort" the columns? Is this increasing gap trend visible yet?
See below graph of the gap size ranked by b, low to high, i.e. not sorted.
You can see on average for each prime, gap size increases by 980.
So yes, increasing gap trend is quite visible.
You can use this to correct the prior series.
Should get a distribution better fitting the exponential curve. | |
|
axnVolunteer developer
Send message
Joined: 29 Dec 07
Posts: 285
ID: 16874
Credit: 28,027,106
RAC: 0
           
|
|
Pulling the data from Top 5000, I get an average gap of ~81000 and a std deviation of ~88000 which is in excellent agreement with a Poisson process (average = std dev). | |
|
|
|
|
It looks like congratulations are in order for Andrew Hughes! | |
|
RogerVolunteer developer
Volunteer tester
Send message
Joined: 27 Nov 11
Posts: 1138
ID: 120786
Credit: 268,668,824
RAC: 0
|
|
Andrew is a first time PRPNet user, good to see him get a prime. | |
|
Sysadm@Nbg Volunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 5 Feb 08
Posts: 1283
ID: 18646
Credit: 1,120,584,411
RAC: 145,410
|
|
\o/
GFN65536: User Andrew_Hughes has found 1 prime and 0 PRPs, 1 is hidden
I 'm curious, unhide 'em :o)
____________
Sysadm@Nbg
my current lucky number: 5195*2^2241611+1
PSA-PRPNet-Stats-URL: http://u-g-f.de/PRPNet/
| |
|
Michael Goetz Volunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 14636
ID: 53948
Credit: 1,016,504,780
RAC: 716,935
|
|
The prime is now verified, unhidden, and reported: http://primes.utm.edu/primes/page.php?id=120036.
Congratulations!
____________
My lucky number is 75898524288+1 | |
|
|
|
|
Congrats on the Prime, Andrew!
- I'm hot on yr heels in the Challenge Rankings.. But will there be enough time to catch you?? - Time will Tell...
____________
 | |
|
|
|
|
And now one for Ross... | |
|
|
|
|
Thanks guys! And @steverc: We'll see, just put another computer online :D | |
|
JimB Honorary cruncher
Send message
Joined: 4 Aug 11
Posts: 921
ID: 107307
Credit: 1,001,476,823
RAC: 0
      
|
|
Ross*'s prime has been confirmed, reported and is now unhidden. http://primes.utm.edu/primes/page.php?id=120037 | |
|
|
|
|
It is possible to "hack" the search function on Caldwell's Top 5000 site so that you can link to a search query. A search like Description=.^65536+1, Comment=generalized fermat, OnList=verified, Number=200, Style=HTML gives all known GFN65536 primes, including the two we found in this challenge up to now. | |
|
RogerVolunteer developer
Volunteer tester
Send message
Joined: 27 Nov 11
Posts: 1138
ID: 120786
Credit: 268,668,824
RAC: 0
|
|
3803826^65536+1 is prime!
Is shown on PRPNet server as PRP.
The other PRP on this port is also prime:
http://primes.utm.edu/primes/page.php?id=119309
3503138^65536+1 is prime!
Can the PRPNet server and PRPNet-Stats be changed to show them as prime rather than PRP? | |
|
|
|
Can the PRPNet server and PRPNet-Stats be changed to show them as prime rather than PRP?
For all of the PRPNet projects, it appears that the designation PRP or Prime is rather arbitrary. The user client might use a program (like genefer) which only proves the number is a probable prime. But in each of those cases, a check has been done (later) with another program which proves primality in a non-deterministic way. So all "PRP" here are really prime.
Maybe "PRP" indicates that the user had a desirable configuration where he prevented expensive extra testing from being performed locally before submitting the find to the server!
I see no subproject where we do not perform a full primality test eventually.
(Of course, if we found a PRP which turned out not to be prime, that would be sensational.)
Why was PRPNet called PRPNet?
/JeppeSN
| |
|
Scott Brown Volunteer moderator
Project administrator
Volunteer tester
Project scientist
Send message
Joined: 17 Oct 05
Posts: 2535
ID: 1178
Credit: 29,395,962,221
RAC: 9,137,904
                                                             
|
Maybe "PRP" indicates that the user had a desirable configuration where he prevented expensive extra testing from being performed locally before submitting the find to the server!
/JeppeSN
This is exactly what happens in the case of the GFN primes. Genefer is only a PRP test. If other prime finding programs are also uncommented (i.e., LLR, PFGW), then one of those tests are run on a PRP find. Otherwise, the PRP is returned to the server and internal prime verification is done.
Leaving the other tests on for the small GFN ports is not really that "expensive" in extra testing (e.g., the GFN65536 port would primality test a PRP in about the same time one can do a PPSE test on BOINC). For the two larger ports (n=18 and n=19), the extra tests do take substantially extra time.
| |
|
|
|
Thanks guys! And @steverc: We'll see, just put another computer online :D
"We need more power, Scotty! They're getting away!"
"I'm givin' 'er all she's got Cap'n! She cannae tak' any more!"
... I guess 10th place will have to do...
____________
| |
|
JimB Honorary cruncher
Send message
Joined: 4 Aug 11
Posts: 921
ID: 107307
Credit: 1,001,476,823
RAC: 0
|
|
We've got another prime found by pabliedung. 3872228^65536+1 has been confirmed by both LLR and PFGW and reported to the Top 5000 Primes site: http://primes.utm.edu/primes/page.php?id=120042.
Congratulations pabliedung! | |
|
RogerVolunteer developer
Volunteer tester
Send message
Joined: 27 Nov 11
Posts: 1138
ID: 120786
Credit: 268,668,824
RAC: 0
|
|
Congratulations to the prime finders:
Andrew_Hughes for 3785880^65536+1,
Ross* for 3803826^65536+1 and
pabliedung for 3872228^65536+1, even if it was found after the challenge.
Challenge Stats are final. Congratulation to zunewantan and Aggie_The_Pew!
The top of the challenge rankings is as follows:
top users
1 zunewantan 18,437,318
2 Ross* 9,407,806
3 hiroaki 5,494,879
4 Scott_Brown 5,434,941
5 Grebuloner 4,608,976
top teams
1 Aggie_The_Pew 22,647,603
2 Sicituradastra. 12,748,591
3 The_Knights_Who_Say_Ni! 7,471,387
4 Team_JPN 5,494,879
5 Duke_University 5,434,941
Collectively we've done 40,226 WUs, had 47 competitors from 22 Teams and advanced the GFN65536 leading edge past b=3,890,506.
Remember to make a suggestion at the 2015 PRPNet Challenges - Discussion and Dates thread.
See you at the next challenge!
____________
| |
|
Message boards :
Project Staging Area :
2015 PRPNet June Challenge |
