Potential 2023 Project Goals

rogue · 2022-11-23, 19:01

Bass upon 2022 project goals, I have some ideas for 2023. I am less familiar with base 2 conjectures, so Gary will have to add those. I have added goals 9 thru 14 because they have millions of fast tests and will likely require coordination at the conjecture level. They might be appealing to people who want to help the project, but want to work outside of BOINC. I will contribute GPU cycles to sieving the conjectures with many thousands of k. I am already sieving (no testing) for goal 2 and will complete sieving those conjectures by the end of the year and send them to Gary when done. Of course Gary has final say on all 2023 project goals.

1. Prove 4 bases

2. Search all bases with <= 5000 k's remaining (at n=10K) to n>=25K; 36 bases

3. Search all bases with <= 20 k's remaining to n>=300K; 15 bases

4. Search all bases with 2 k's remaining to n>=500K; 29 bases

5. Search all bases < 800 with 1 k remaining to n>=600K; 35 bases

6. Find 50 top 3000 primes

7. Find 10 top 1500 primes

8. Find 5 top 500 primes

9. Search S3 to n>=50000 for k < 40e9

10. Search R3 to n>=200000 for k < 20e9

11. Search R63 to n>=50000 for k < 60e6

12. Search S63 to n>=50000 for k < 10e6

13. Search R66 to n>=50000 for k < 30e6

14. Search S66 to n>=50000

rogue · 2022-11-25, 13:59

Quote:

Originally Posted by sweety439

Search all bases...

You have no idea as to the level of effort for what you have listed.

You have contributed nearly nothing to the project, so your opinions shouldn't be considered.

VBCurtis · 2022-11-25, 16:26

Sweety-
You have had multiple garbage posts moved by moderators to your blog this week. We don't enjoy you wasting our time, and neither does this project.
Two reminders:
1. any post from you that suggests other people do work that interests you is cause for a renewal of your ban from this site.
2. when you run afoul of moderators as you have this week, your best course of action is to restrict your posting to your blog area. The more you post in math areas of this forum, the more negative attention you draw. Gary has made very clear your views and posts are not welcome in this subforum; if you can't obey the wishes of mods, you should expect more severe restrictions applied to you.

masser · 2022-11-25, 17:58

storm5510 · 2022-11-25, 18:56

Quote:

Originally Posted by VBCurtis

Sweety-
You have had multiple garbage posts moved by moderators to your blog this week. We don't enjoy you wasting our time, and neither does this project.
Two reminders:
1. any post from you that suggests other people do work that interests you is cause for a renewal of your ban from this site.
2. when you run afoul of moderators as you have this week, your best course of action is to restrict your posting to your blog area. The more you post in math areas of this forum, the more negative attention you draw. Gary has made very clear your views and posts are not welcome in this subforum; if you can't obey the wishes of mods, you should expect more severe restrictions applied to you.

The warden said to the new inmates in Cool Hand Luke: "What we have here is a failure to communicate." "Sweety" is repeatedly breaking the rules here. He/she does not believe the rules apply to him/her. A failure of communication on his/her part. Reading is one thing. Understanding and obeying is far different in this situation.

gd_barnes · 2022-11-26, 00:11

I had seen Sweety's goals very early this morning but decided to let them stand for a while to see what others thought. The responses received were of course expected and echoed most of my thoughts. Thanks for chiming in guys. My main take aways were that he has no clue how long things take because he has never actually contributed -and- he cannot seem to remember that he's been asked several times not to post such things because he doesn't contribute.

I've done a soft delete, for now, of his post since it was responded to. I will do a permanent delete on it within a day unless anyone would prefer that I leave it due to the responses.

gd_barnes · 2022-11-26, 00:31

Now...back to the task at hand...

I think we should take a step back somewhat on goals for 2023. I feel I/we may have pushed a little too hard in 2022. I've become increasingly concerned about how little double-check we've done for the project and the extremely unusual lack of primes that were coming out in 2022 only heightened that concern. For 2023, I'd definitely like to include some double-check efforts.

I'm not dismissing having goals. Just have fewer of them and make them a little less aggressive. Also since primes seem to come randomly for some reason :-), setting specific prime-finding goals may not be the best way to approach things going forward. If we make an effort to set specific search-range goals the primes will come in good time. 2022 was a very tough year for prime-finding for us but we nailed the search-range goals that were set, although it was a pretty hefty push to do so.

Mark, it's a good idea to attempt to involve the non-BOINC folks more. I'm not a fan of project-wide efforts to find thousand/millions of small primes on huge-conjectured bases, mainly due to the admin involved. But due to the age of the project, all of the easy stuff was done long ago. Doing what you are suggesting with your final 5-6 goals is not a bad way to bring in people. Maybe we will do a couple of those.

I'll keep adding more thoughts over the next few weeks as I think about things more and more input is offered here by others.

rogue · 2022-11-26, 00:51

In other words 1, 6, 7, and 8 are "luck of the draw", but the other goals are well defined.

Double-checking will require a lot of consideration. The main question is "what gets double-checked?". It will obviously require a lot of work, but I don't know how many people will want to do it. llr2 is an option for BOINC, but I do not know how easy that would be to integrate into PRPNet.

rebirther · 2022-11-26, 20:35

Quote:

Originally Posted by rogue

In other words 1, 6, 7, and 8 are "luck of the draw", but the other goals are well defined.

Double-checking will require a lot of consideration. The main question is "what gets double-checked?". It will obviously require a lot of work, but I don't know how many people will want to do it. llr2 is an option for BOINC, but I do not know how easy that would be to integrate into PRPNet.

Normally llr2 dont need a double check anymore if you can catch up all errors by a validator. If the host has a hardware issue the Gerbicz check will intervent and rerun the task max 6 times from the same last position then it will cancel the task with an error. Due this Gerbicz check you will loose 10-20% of performance / larger runtimes but can do AVX512 now.

Here is a sample which I need to rerun:

Code:

Gerbicz check failed at 47360.
Continuing from last save file.
Starting probable prime test of 10572*61^757787-1
Using FMA3 FFT length 512K, Pass1=512, Pass2=1K, clm=1, a = 3, L2 = 256*185
Gerbicz check failed at 47360.
Too much errors ; Restarting with next larger FFT length...
Continuing from last save file.
Starting probable prime test of 10572*61^757787-1
Using FMA3 FFT length 560K, Pass1=448, Pass2=1280, clm=1, a = 3, L2 = 256*185
Gerbicz check failed at 47360.
Continuing from last save file.
Starting probable prime test of 10572*61^757787-1
Using FMA3 FFT length 560K, Pass1=448, Pass2=1280, clm=1, a = 3, L2 = 256*185
Gerbicz check failed at 189440.
Continuing from last save file.
Resuming probable prime test of 10572*61^757787-1 at bit 142081 [18.74%]
Using FMA3 FFT length 560K, Pass1=448, Pass2=1280, clm=1, a = 3, L2 = 256*185
Gerbicz check failed at 189440.
Too much errors ; Restarting with next larger FFT length...
Continuing from last save file.
Resuming probable prime test of 10572*61^757787-1 at bit 142081 [18.74%]
Using FMA3 FFT length 640K, Pass1=640, Pass2=1K, clm=1, a = 3, L2 = 256*185
Gerbicz check failed at 189440.
Continuing from last save file.
Resuming probable prime test of 10572*61^757787-1 at bit 142081 [18.74%]
Using FMA3 FFT length 640K, Pass1=640, Pass2=1K, clm=1, a = 3, L2 = 256*185
Gerbicz check failed at 189440.
Continuing from last save file.
Resuming probable prime test of 10572*61^757787-1 at bit 142081 [18.74%]
Using FMA3 FFT length 640K, Pass1=640, Pass2=1K, clm=1, a = 3, L2 = 256*185
Gerbicz check failed at 189440.
Continuing from last save file.
Resuming probable prime test of 10572*61^757787-1 at bit 142081 [18.74%]
Using FMA3 FFT length 640K, Pass1=640, Pass2=1K, clm=1, a = 3, L2 = 256*185
Gerbicz check failed at 189440.
Too many errors, aborting.

rebirther · 2022-11-27, 10:19

There are a lot of 300-500k ranges with sievefiles, we could do the lower factors left and of course some of the 1k left bases

MisterBitcoin · 2022-12-18, 16:04

I think working on both S/R63 and S/R66 might be too much; it requires a lot of work & time. For S66 we already have sieve files which certainly helps, so maybe just include S/R66?

We could include S79 instead which wouldn´t be too difficult to test up to n=100K. Or bases b<100 ks-remain <250 tested up to n=100K; like R78 and R82?

2022-11-23, 19:01	#1
rogue "Mark" Apr 2003 Between here and the 2×3²×401 Posts	Potential 2023 Project Goals Bass upon 2022 project goals, I have some ideas for 2023. I am less familiar with base 2 conjectures, so Gary will have to add those. I have added goals 9 thru 14 because they have millions of fast tests and will likely require coordination at the conjecture level. They might be appealing to people who want to help the project, but want to work outside of BOINC. I will contribute GPU cycles to sieving the conjectures with many thousands of k. I am already sieving (no testing) for goal 2 and will complete sieving those conjectures by the end of the year and send them to Gary when done. Of course Gary has final say on all 2023 project goals. 1. Prove 4 bases 2. Search all bases with <= 5000 k's remaining (at n=10K) to n>=25K; 36 bases 3. Search all bases with <= 20 k's remaining to n>=300K; 15 bases 4. Search all bases with 2 k's remaining to n>=500K; 29 bases 5. Search all bases < 800 with 1 k remaining to n>=600K; 35 bases 6. Find 50 top 3000 primes 7. Find 10 top 1500 primes 8. Find 5 top 500 primes 9. Search S3 to n>=50000 for k < 40e9 10. Search R3 to n>=200000 for k < 20e9 11. Search R63 to n>=50000 for k < 60e6 12. Search S63 to n>=50000 for k < 10e6 13. Search R66 to n>=50000 for k < 30e6 14. Search S66 to n>=50000

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
2021 project goals	gd_barnes	Conjectures 'R Us	22	2022-01-01 07:29
2020 project goals	gd_barnes	Conjectures 'R Us	14	2020-12-30 15:58
2019 project goals	gd_barnes	Conjectures 'R Us	23	2019-12-07 21:12
2018 project goals	gd_barnes	Conjectures 'R Us	24	2019-01-03 09:47
Goals of the project	em99010pepe	Octoproth Search	10	2007-01-28 12:34

2022-11-25, 16:26	#3
VBCurtis "Curtis" Feb 2005 Riverside, CA 7×829 Posts	Sweety- You have had multiple garbage posts moved by moderators to your blog this week. We don't enjoy you wasting our time, and neither does this project. Two reminders: 1. any post from you that suggests other people do work that interests you is cause for a renewal of your ban from this site. 2. when you run afoul of moderators as you have this week, your best course of action is to restrict your posting to your blog area. The more you post in math areas of this forum, the more negative attention you draw. Gary has made very clear your views and posts are not welcome in this subforum; if you can't obey the wishes of mods, you should expect more severe restrictions applied to you.

2022-11-25, 17:58	#4
masser Jul 2003 Behind BB 7CA₁₆ Posts

2022-11-26, 00:11	#6
gd_barnes "Gary" May 2007 Overland Park, KS 23³ Posts	I had seen Sweety's goals very early this morning but decided to let them stand for a while to see what others thought. The responses received were of course expected and echoed most of my thoughts. Thanks for chiming in guys. My main take aways were that he has no clue how long things take because he has never actually contributed -and- he cannot seem to remember that he's been asked several times not to post such things because he doesn't contribute. I've done a soft delete, for now, of his post since it was responded to. I will do a permanent delete on it within a day unless anyone would prefer that I leave it due to the responses.

2022-11-26, 00:31	#7
gd_barnes "Gary" May 2007 Overland Park, KS 23³ Posts	Now...back to the task at hand... I think we should take a step back somewhat on goals for 2023. I feel I/we may have pushed a little too hard in 2022. I've become increasingly concerned about how little double-check we've done for the project and the extremely unusual lack of primes that were coming out in 2022 only heightened that concern. For 2023, I'd definitely like to include some double-check efforts. I'm not dismissing having goals. Just have fewer of them and make them a little less aggressive. Also since primes seem to come randomly for some reason :-), setting specific prime-finding goals may not be the best way to approach things going forward. If we make an effort to set specific search-range goals the primes will come in good time. 2022 was a very tough year for prime-finding for us but we nailed the search-range goals that were set, although it was a pretty hefty push to do so. Mark, it's a good idea to attempt to involve the non-BOINC folks more. I'm not a fan of project-wide efforts to find thousand/millions of small primes on huge-conjectured bases, mainly due to the admin involved. But due to the age of the project, all of the easy stuff was done long ago. Doing what you are suggesting with your final 5-6 goals is not a bad way to bring in people. Maybe we will do a couple of those. I'll keep adding more thoughts over the next few weeks as I think about things more and more input is offered here by others.

2022-11-26, 00:51	#8
rogue "Mark" Apr 2003 Between here and the 1C32₁₆ Posts	In other words 1, 6, 7, and 8 are "luck of the draw", but the other goals are well defined. Double-checking will require a lot of consideration. The main question is "what gets double-checked?". It will obviously require a lot of work, but I don't know how many people will want to do it. llr2 is an option for BOINC, but I do not know how easy that would be to integrate into PRPNet.

2022-11-27, 10:19	#10
rebirther Sep 2011 Germany DDE₁₆ Posts	There are a lot of 300-500k ranges with sievefiles, we could do the lower factors left and of course some of the 1k left bases

2022-12-18, 16:04	#11
MisterBitcoin "Nuri, the dragon :P" Jul 2016 Good old Germany 2³×3×37 Posts	I think working on both S/R63 and S/R66 might be too much; it requires a lot of work & time. For S66 we already have sieve files which certainly helps, so maybe just include S/R66? We could include S79 instead which wouldn´t be too difficult to test up to n=100K. Or bases b<100 ks-remain <250 tested up to n=100K; like R78 and R82?