| Author |
Message |
|---|
 Nick 
Send message
Joined: 11 Jul 11
Posts: 2216
ID: 105020
Credit: 8,129,330,977
RAC: 1,539,263
                            
|
|
I think it is overdue to celebrate the 4th of July.
Celebrate the great things that day represents.
And if that is not enough, the 1st of July is when people in Darwin get to blow stuff up (fireworks). I have seen it twice. It is carnage. |
|
|
JaySend message
Joined: 27 Feb 10
Posts: 131
ID: 56067
Credit: 63,534,968
RAC: 10,618
                    
|
I think it is overdue to celebrate the 4th of July.
Celebrate the great things that day represents.
And if that is not enough, the 1st of July is when people in Darwin get to blow stuff up (fireworks). I have seen it twice. It is carnage.
Then I say lets blow up some utility bills with a month long Seventeen or Bust challenge! |
|
|
Nick
Send message
Joined: 11 Jul 11
Posts: 2216
ID: 105020
Credit: 8,129,330,977
RAC: 1,539,263
|
|
I like your thinking.
SOB seems to be a fav.
I had to PM Michael last year because I missed the joke.
I really thought that we were going to cram all the challenges into January and SOB all year after February. |
|
|
rogueVolunteer developer
Send message
Joined: 8 Sep 07
Posts: 1255
ID: 12001
Credit: 18,565,548
RAC: 0
 
|
|
A weeklong PPSlayer challenge for PPS (not mega). The intent is to get PPS to mega range so that PPS and PPS Mega can be merged. It probably won't be long enough to get that far, but it should go quite a ways to finishing off whatever PPS needs to do to get to PPS Mega range.
A weeklong PRPNet search for factorials or primorials. |
|
|
HonzaVolunteer moderator
Volunteer tester
Project scientist
Send message
Joined: 15 Aug 05
Posts: 1948
ID: 352
Credit: 6,006,997,492
RAC: 1,516,944
                               
|
A weeklong PRPNet search for factorials or primorials.
Personally, I like factorial and primirial.
But I would rather see it running on PG with double-checking...and then go for a weeklong challenge (with all the badges). This sould push it very far comparing to the progress on PRPNet.
____________
My stats |
|
|
|
|
|
I also vote for SoB.
____________
|
|
|
|
|
|
I would be ecstatic if factorial and primorial searches were moved to BOINC and then a challenge run on them.
____________
 |
|
|
|
|
I would be ecstatic if factorial and primorial searches were moved to BOINC and then a challenge run on them.
I agree.
____________
 |
|
|
Dave
Send message
Joined: 13 Feb 12
Posts: 3171
ID: 130544
Credit: 2,232,433,118
RAC: 712,029
                    
|
|
I think that requires its own official thread, as does discussion of next year's challenges. |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13951
ID: 53948
Credit: 391,787,958
RAC: 149,991
                     
|
I would be ecstatic if factorial and primorial searches were moved to BOINC and then a challenge run on them.
I agree.
We all agree on this, going back several years. Jim and I have wanted to do this for a long time, but there's a lot of difficulties involved...
It's PFGW only; LLR (and thus LLR2 as well) can not be used.
That means PFGW has to be made to work with BOINC. There's a whole bunch of technical problems right there.
But that's just the obvious stuff. The real problems occur once you succeed in getting this running out of BOINC ... because we'll completely run out of work within a week. The existing sieve files just don't go that far. On PRPNet, they'll last a long time. On BOINC? Even without TSC they won't last long, and when you factor in the almost unlimited processing power of TSC I suspect we'd run dry before all the droplets even got spooled up.
So, we'll have to do more sieving. Manual sieving? Another BOINC project? We'll need to do something. But even then, we're going to start running into limits in the software. The software that did the current sieves can't go much higher.
In short, it's *probably* doable but there's a whole lot of planning and probably a bunch of coding that would need to happen before this could occur.
____________
My lucky number is 75898524288+1 |
|
|
rogueVolunteer developer
Send message
Joined: 8 Sep 07
Posts: 1255
ID: 12001
Credit: 18,565,548
RAC: 0
|
|
psieve and mfsieve (from the mtsieve framework) support values up to 1e9# and 1e9! and have GPU counterparts.
The challenge will be sieving deeply enough. These sieves are iterative, so the time to sieve each range grows at the same rate as the size of the range. This means that it will take a long time to sieve deeply enough. My recommendation is that if PrimeGrid wants to take these searches deeper then start sieving ASAP (even if it is manual) to get a head start.
I thought there was a way to create a wrapper so that pfgw would not need to be modified to support BOINC. |
|
|
|
|
|
Of course, once a BOINC wrapper for PFGW (OpenPFGW) is in place, we would double-check the work of the preceding 10-12 years in primorial (PRS) and factorial (FPS). But I guess ten years' work will be quickly done when the full BOINC army has access to PRS/FPS?
So we would need to figure out how to sieve.
I think the original sieving ten years ago was done as manual sieving? Is it known what candidate ranges (maximal candidate attempted eliminated) and what depth (maximal potential factor considered) the old sieving did?
/JeppeSN |
|
|
|
|
psieve and mfsieve (from the mtsieve framework) support values up to 1e9# and 1e9! and have GPU counterparts.
The challenge will be sieving deeply enough. These sieves are iterative, so the time to sieve each range grows at the same rate as the size of the range. This means that it will take a long time to sieve deeply enough. My recommendation is that if PrimeGrid wants to take these searches deeper then start sieving ASAP (even if it is manual) to get a head start.
I thought there was a way to create a wrapper so that pfgw would not need to be modified to support BOINC.
Good info.
Do you know if psieve and mfsieve (from the mtsieve framework) have ever been automated via PRPNet or BOINC?
The limits 1e9# and (especially) 1e9! seem to be large enough. But if the sieving takes a long time (while still remaining relevant (below optimal depth)), maybe we should work on a system for the sieving first, on the expectation that someone will setup the wrapper for PFGW (whether it exists or is new).
/JeppeSN |
|
|
rogueVolunteer developer
Send message
Joined: 8 Sep 07
Posts: 1255
ID: 12001
Credit: 18,565,548
RAC: 0
|
psieve and mfsieve (from the mtsieve framework) support values up to 1e9# and 1e9! and have GPU counterparts.
The challenge will be sieving deeply enough. These sieves are iterative, so the time to sieve each range grows at the same rate as the size of the range. This means that it will take a long time to sieve deeply enough. My recommendation is that if PrimeGrid wants to take these searches deeper then start sieving ASAP (even if it is manual) to get a head start.
I thought there was a way to create a wrapper so that pfgw would not need to be modified to support BOINC.
Good info.
Do you know if psieve and mfsieve (from the mtsieve framework) have ever been automated via PRPNet or BOINC?
The limits 1e9# and (especially) 1e9! seem to be large enough. But if the sieving takes a long time (while still remaining relevant (below optimal depth)), maybe we should work on a system for the sieving first, on the expectation that someone will setup the wrapper for PFGW (whether it exists or is new).
/JeppeSN
I have not integrated mtsieve into PRPNet or BOINC. A wrapper would be better because adding all of the BOINC specific stuff to make programs work in that universe is a pain. I understand why it is done. It just isn't fun to do.
Since mtsieve is open source, there is nothing preventing others from grabbing the framework and trying to add the BOINC logic to it.
1e7! -> 65 million digits
1.5e7! -> 100 million digits
10e7! -> 750 million digits
13.1e7! -> 1 billion digits
pfgw seems to hang at 1e7! which is over 65 million digits in length. That could be a bug in pfgw. It will require further investigation.
I think there is value in sieving to 1e7, maybe even 1.5e7, but PRP tests for those numbers will take weeks on current hardware. It would be interesting if someone could write a generic PRP test that runs on a GPU.
In any case sieving will be a massive undertaking. It would likely take months or years to sieve deeply enough before any testing can be done. I have no idea what the optimal sieve depth would be, but I wouldn't be surprised if it were beyond 1e15.
My recommendation is to decide the max factorial to sieve, then start a manual sieving project so that users can start sieving at their convenience. With the -O and -I options of the framework it will be easy to remove factors from the file of candidates. Eventually a BOINC wrapper could be written for mfsieve/mfsievecl (if not directly built into the mtsieve framework) which would bring in more users. |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13951
ID: 53948
Credit: 391,787,958
RAC: 149,991
|
...maybe even 1.5e7, but PRP tests for those numbers will take weeks on current hardware.
With multithreading?
____________
My lucky number is 75898524288+1 |
|
|
|
|
1e7! -> 65 million digits
1.5e7! -> 100 million digits
10e7! -> 750 million digits
13.1e7! -> 1 billion digits
And primorials grow like this:
2303867# -> 1 million digits
23030549# -> 10 million digits
230277781# -> 100 million digits
2302619971# -> 1 billion digits
Of course the last one of these exceeds the 1e9# you mentioned.
/JeppeSN |
|
|
Yves GallotVolunteer developer
Project scientist
Send message
Joined: 19 Aug 12
Posts: 801
ID: 164101
Credit: 305,666,804
RAC: 5,309
|
And primorials grow like this:
2303867# -> 1 million digits
23030549# -> 10 million digits
230277781# -> 100 million digits
2302619971# -> 1 billion digits
Of course the last one of these exceeds the 1e9# you mentioned.
/JeppeSN
This is a method for calculating log 10 ;-)
|
|
|
|
|
And primorials grow like this:
2303867# -> 1 million digits
23030549# -> 10 million digits
230277781# -> 100 million digits
2302619971# -> 1 billion digits
Of course the last one of these exceeds the 1e9# you mentioned.
/JeppeSN
This is a method for calculating log 10 ;-)
Chebyshev function § Relation to primorials /JeppeSN |
|
|
rogueVolunteer developer
Send message
Joined: 8 Sep 07
Posts: 1255
ID: 12001
Credit: 18,565,548
RAC: 0
|
...maybe even 1.5e7, but PRP tests for those numbers will take weeks on current hardware.
With multithreading?
I have no idea. I haven't actually tried to test numbers of this size and haven't played much with the multi-threading features of gwnum that pfgw can take advantage of.
What I do know is that PRP testing takes longer than other forms. In a quick test I see it taking 2.5x longer than a number of the form k*b^n+c (b != 2). |
|
|
rogueVolunteer developer
Send message
Joined: 8 Sep 07
Posts: 1255
ID: 12001
Credit: 18,565,548
RAC: 0
|
|
The limit of 1e9 is arbitrary. It would be easy to bump that limit in the code. |
|
|
|
|
|
So it seems first thing required is manual sieving.
Can this be setup in the Project Staging Area and maybe later moved to be the default sieve instead of PPS_sr2sieve? This would let us get some of the tedious but necessary sieving tasks out of the way.
Meanwhile our excellent applications developers can work on the hard bit - getting the software written to perform the BOINC version of PFGW.
____________
|
|
|
streamVolunteer moderator
Project administrator
Volunteer developer
Volunteer tester
Send message
Joined: 1 Mar 14
Posts: 1022
ID: 301928
Credit: 543,195,386
RAC: 2
   
|
|
As far as I remember (correct me if I wrong), biggest problem is PFGW do not have checkpoints for primorial/factorial, this makes it useless for real Boinc projects.
|
|
|
rogueVolunteer developer
Send message
Joined: 8 Sep 07
Posts: 1255
ID: 12001
Credit: 18,565,548
RAC: 0
|
As far as I remember (correct me if I wrong), biggest problem is PFGW do not have checkpoints for primorial/factorial, this makes it useless for real Boinc projects.
pfgw does checkpoint. I just recall how frequently. |
|
|
|
|
As far as I remember (correct me if I wrong), biggest problem is PFGW do not have checkpoints for primorial/factorial, this makes it useless for real Boinc projects.
pfgw does checkpoint. I just recall how frequently.
Maybe it is only the deterministic proof (which would not be done as a BOINC task) which does not have checkpoint files. The PFGW doc file says:
D.9 Save/Resume files
Although not actually a "input" file type, PFGW does save its
processing "state" information, from time to time, while doing
PRP testing. The files saved, have "strange" names, and end
in the .pfr extension (the pfr stands for PrimeForm Resume file.)
.pfr files store the "state" of the FFT number being worked on,
along with other information, that allows PFGW to validate that
it actually IS the number. The name of the file is special
(do NOT rename it). PFGW generates this name from the number
itself. Thus, when PFGW prepares to PRP a number, it can
quickly determine what the "proper" filename for this number is,
and see if that file exists. If the file does exist, then
PFGW, can do a much more intesive check of that file, and if
it is the file for the number being processed, pfgw can reload
this file, and resume from where this test left off.
A few notes:
1. These save files are version dependend. Thus when PFGW
is upgraged (the FFT's), the save files will NOT function.
2. Only the PRP testing is save/resumed. There are just TOO
many things that would need saved for testing save/restore
to function properly.
3. The .pfr files will be deleted automatically by PFGW when
a number has been fully PRP'd.
4. PRP tests for different bases (PRP-3 vs PRP-131) can not
share the same save/resume file. The PRP base must be
the same (note PFGW will not resume if the base is different)
5. Save/Resume is only looked at for number over 2^50000. There
is some overhead in file name creation, and checking for
existance of a file. Thus for numbers under this size, it
is not time efficient to check (or to save).
6. Save/Resume is automatic. There is no user interaction.
For numbers large enough, the save file is written upon
early shutdown (^C or in WinPFGW, the stop button). Also
the save files are written every 20 minutes or so.
/JeppeSN
|
|
|
rogueVolunteer developer
Send message
Joined: 8 Sep 07
Posts: 1255
ID: 12001
Credit: 18,565,548
RAC: 0
|
Maybe it is only the deterministic proof (which would not be done as a BOINC task) which does not have checkpoint files.
That makes more sense. |
|
|
|
|
|
Anyhow, I'd strongly support a 'Conjecture Year' sort of collection of longer conjecture challenges.
____________
My lucky number is 6219*2^3374198+1
|
|
|
Nick
Send message
Joined: 11 Jul 11
Posts: 2216
ID: 105020
Credit: 8,129,330,977
RAC: 1,539,263
|
|
I wish for a challenge celebrating Thomas Sowell. |
|
|
|
|
|
Currently we need about one week to get 100000 digits further in GFN17low.
We are now at 989000 digits.
So in January we can make a GFN17low challenge with a fixed start date and an open end date, defined by the last available WU send out (no bunkering allowed, each host can get a max. of 3/5/10 (?) WU).
May be a special badge for a prime with 999999 digits?
____________
DeleteNull |
|
|
Dave
Send message
Joined: 13 Feb 12
Posts: 3171
ID: 130544
Credit: 2,232,433,118
RAC: 712,029
|
May be a special badge for a prime with 999999 digits?
To fit into th fixed pixel count of course. Good luck with that. |
|
|
|
|
|
M-1 is not difficult.
____________
DeleteNull |
|
|
|
|
May be a special badge for a prime with 999999 digits?
Quite unlikely (though possible). Would need the b to be between 10^(999998/2^17) and 10^(999999/2^17), so between 42'597'025.4 and 42'597'773.7. Expect quite few candidates in such an interval. /JeppeSN |
|
|
Nick
Send message
Joined: 11 Jul 11
Posts: 2216
ID: 105020
Credit: 8,129,330,977
RAC: 1,539,263
|
May be a special badge for a prime with 999999 digits?
Quite unlikely (though possible). Would need the b to be between 10^(999998/2^17) and 10^(999999/2^17), so between 42'597'025.4 and 42'597'773.7. Expect quite few candidates in such an interval. /JeppeSN
Then a badge could be about cracking a really small probability?
"The lucky b***ard" badge?
Could we have a luck badge?
Except we already have those.
I think what we have in place is brilliant. |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13951
ID: 53948
Credit: 391,787,958
RAC: 149,991
|
May be a special badge for a prime with 999999 digits?
Quite unlikely (though possible). Would need the b to be between 10^(999998/2^17) and 10^(999999/2^17), so between 42'597'025.4 and 42'597'773.7. Expect quite few candidates in such an interval. /JeppeSN
There are less than 100 candidates in that range. The chance of any of those being prime are not good.
____________
My lucky number is 75898524288+1 |
|
|
Nick
Send message
Joined: 11 Jul 11
Posts: 2216
ID: 105020
Credit: 8,129,330,977
RAC: 1,539,263
|
|
Terry Pratchett wrote:
"A chance in a million happens nine times out of ten"
Clever man.
He didn't say which chance in a million. |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13951
ID: 53948
Credit: 391,787,958
RAC: 149,991
|
Terry Pratchett wrote:
"A chance in a million happens nine times out of ten"
Clever man.
He didn't say which chance in a million.
If you're "one in a million", that means there's seven thousand people exactly like you!
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
Perhaps a challenge for the prime-finding projects that don't qualify for T5K? SGS and GFN15 mixed together...although that would likely lead to everyone running GFN15 since it supports GPU. Idk. Could still be interesting.
____________
|
|
|
|
|
|
Maybe PPSE to prevent it from falling out from top5000?
____________
|
|
|
|
|
Maybe PPSE to prevent it from falling out from top5000?
I'm not sure the leading edge of PPSE is capable of advancing quickly enough to prevent it from falling out of the top 5000. I think it is only a matter of a year or two before both it and GFN16 will no longer be part of the top 5000.
____________
|
|
|
|
|
|
Any updates as to the final schedule?
____________
My lucky number is 6219*2^3374198+1
|
|
|
BurVolunteer tester
Send message
Joined: 25 Feb 20
Posts: 515
ID: 1241833
Credit: 414,035,564
RAC: 40,802
               
|
|
I would like it a lot if Primorials/Factorials could be transferred into Boinc. No one is really investigating these numbers and thus progress is very slow. The last prime of the Primorial+1 form was discovered 20 years ago. That form is quite interesting though due to its link to the proof of the infinitude of primes.
But challenges to finally get PPS and GFN-17 low to merge into the Mega counterparts would be very nice as well.
____________
1281979 * 2^485014 + 1 is prime ... no further hits up to: n = 5,700,000 |
|
|
Michael GoetzVolunteer moderator
Project administrator
Send message
Joined: 21 Jan 10
Posts: 13951
ID: 53948
Credit: 391,787,958
RAC: 149,991
|
|
The tentative 2022 schedule is now visible and linked from the home page. Note that the date and name of the November challenge have changed, but the project and duration are the same.
____________
My lucky number is 75898524288+1 |
|
|
|
|
|
Thanks !
____________
"Accidit in puncto, quod non contingit in anno."
Something that does not occur in a year may, perchance, happen in a moment. |
|
|
