mersenneforum.org Any way to run trial factoring on a SPECIFIC Exponent?
 Register FAQ Search Today's Posts Mark Forums Read

 2021-12-25, 10:12 #1 IndrajithGamage     "Indrajith Gamage" Dec 2021 Sri Lanka 810 Posts Any way to run trial factoring on a SPECIFIC Exponent? Hi, so I have reserved (as in Assigned) the exponent 384,989,999 for PRP. As I have said earlier, I have significant reason to say that this will be the first ever 100M prime to be confirmed. (said here - https://mersenneforum.org/showthread.php?t=27427&page=2) My reserved exponent is here. https://www.mersenne.org/report_expo...4989999&full=1 As you can see at above link, I was running the trial factoring from 2^73 upto 2^76 and confirmed NF in those. Since I'm doing this on my own laptop and had other work, I stopped it there. However, when I came to check again, a user named @ViliamF has continued the job and completed 2^77 to 2^82 and confirmed NF. I still want to continue some trial factoring beyond 2^82, since it's quite fast. (perhaps upto 2^100 or so) Is there a way to get that assignment? I tried from the forum but I'm not getting the same exponent. I don't care about others because I have a very good reason to believe that this will be the first ever 100M prime and want to prove it. I'm running the PRP as well but evidently that's taking quite a long time. So I would like to do some TF and P-1 etc in the mean time to double confirm the results with smaller cost. Thank you in advance. Last fiddled with by IndrajithGamage on 2021-12-25 at 10:29
 2021-12-25, 10:41 #2 Zhangrc   "University student" May 2021 Beijing, China 3·67 Posts Maybe... consider buying an RTX3090 first. It doesn't make much sense to do higher bit levels of trial factoring. If you want to TF it to 2^100, you need about 6.67e+9 GHzDays, which could run roughly 300 years on a single RTX3090. Maybe you want to run a P-1 first. that's 200 GHzDays for B1=2500000, B2=100000000. Allocate enough RAM before running this. Pminus1=1,2,384989999,-1,2500000,100000000,82 (However, if you are so sure that it is a prime, why do you want to find a factor?) Last fiddled with by Zhangrc on 2021-12-25 at 10:46
2021-12-25, 10:58   #3
IndrajithGamage

"Indrajith Gamage"
Dec 2021
Sri Lanka

23 Posts

Quote:
 Originally Posted by Zhangrc Maybe... consider buying an RTX3090 first. It doesn't make much sense to do higher bit levels of trial factoring. If you want to TF it to 2^100, you need about 6.67e+9 GHzDays, which could run roughly 300 years on a single RTX3090. Maybe you want to run a P-1 first. that's 200 GHzDays for B1=2500000, B2=100000000. Allocate enough RAM before running this. Pminus1=1,2,384989999,-1,2500000,100000000,82 (However, if you are so sure that it is a prime, why do you want to find a factor?)

Seriously? I thought the numbers were linear. My laptop completed 73 to 76 in like 6 hours.

That's cool then. Yeah I'm running the PRP now. Didn't think factoring takes longer than that.

Either way, is it possible to just put that on worktodo and restart the program for P-1? Don't I need some key from the server?

Last fiddled with by IndrajithGamage on 2021-12-25 at 11:01

2021-12-25, 13:32   #4
kriesel

"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

10111110101102 Posts

Quote:
 Originally Posted by IndrajithGamage Seriously? I thought the numbers were linear.
Possible factors are f=2kp+1 for Mp=2p-1, f=1 or 7 mod 8, f prime. For fixed p, how many k fit f between 273 and 274? How many between 283 and 284? Use that math degree you made sure everyone knew of. Or use the reference info. Do the GIMPS beginners' reading described there. SENSIBLE limit TF or SENSIBLE bounds P-1 takes of order 1/40 of a primality test duration.
https://www.mersenne.org/report_expo...exp_hi=&full=1 shows it has had enough TF, but woefully little bounds on P-1. Check that you have set allowed memory to as much as you can safely allocate (reducing slightly if thrashing to page file occurs), not the mere ~0.3 GB that is prime95's installation default. Then run P-1 with adequate bounds as shown on https://www.mersenne.ca/exponent/384989999. (I recommend the larger GPU72 row bounds.) If you give prime95 enough ram to work with it can choose optimal bounds itself, as when a PRP assignment is received that indicates a need for more P-1, or a PFactor assignment.
See the attached prime95 example, for my main laptop equipped with 16 GiB and many other tasks running. Mprime on Linux has equivalent controls. Both have excellent documentation files.
I hope you're doing this PRP on a CPU, or AMD or other GPU with relatively strong DP performance (not an RTX30xx or similar new NVIDIA GPU which have weak DP performance compared to SP so is much more productive doing TF, much less productive doing P-1 or primality testing).

There are also other ways of accomplishing it, but perhaps the conceptually simplest is:
Stop and exit prime95, then paste the Pminus1 line given in a previous post in front of the PRP line in your prime95 worktodo file and save the change, exit the editor, and restart the prime95 program.
(Do P-1 to adequate / optimal bounds, once. Running smaller first is wasteful of computing time.)
Quote:
 Originally Posted by Zhangrc (However, if you are so sure that it is a prime, why do you want to find a factor?)
Doing TF high enough, P-1 to optimal bounds, then PRP/proof, is good practice.

Quote:
 Originally Posted by IndrajithGamage Is there a way to get that assignment? I tried from the forum but I'm not getting the same exponent. I don't care about others because I have a very good reason to believe that this will be the first ever 100M prime and want to prove it.
The forum does not issue assignments. mersenneforum.org and mersenne.org are different domains.
You've posted repeatedly about 314159257 and subsequently 384989999, "confident", etc. but have given no concrete or testable reasons for your confidence, expectations, etc. Getting NF for TF to the puny (even on a log scale) feasible levels we normally run, or practical P-1 bounds, is consistent both with the exceedingly rare occurrence of Mersenne primes, and consistent with the 999,999+ ppm occurrence of Mersenne composites of prime exponent of current interest, so does not qualify as a reason to think a particular Mersenne is prime. What is your reason? Would R. Gerbicz or another competent number theorist agree it is a plausible reason?
(not log 82 = 1.914 while log (314159257/2) = 8.196, while what we should be comparing is 82 vs 314159257/2 = log2(tested factors) vs log2(possible smallest prime factor range) That is, throughput-optimal TF is only ~82*2/314159257 ~ 5.22x10-5% ON A LOG SCALE)
Attached Thumbnails

Last fiddled with by kriesel on 2021-12-25 at 14:15

2021-12-25, 14:59   #5
IndrajithGamage

"Indrajith Gamage"
Dec 2021
Sri Lanka

23 Posts

Quote:

Hey, thanks for the clear and precise reply.

So I did what you asked. Here's the screenshots. (seems like I can't attach files, so these are hosted links) I have only 8GB of ram so gave it 4GB. Reduced the upper P-1 bound to 50M as well due to this. I hope those amounts are okay.

https://i.ibb.co/0ZZHLzZ/Annotation-...-25-201932.png
https://i.ibb.co/fvrnwmJ/Annotation-...-25-202033.png
https://i.ibb.co/tbnTgCk/Annotation-...-25-202151.png

However, I think it's still running the PRP, and waiting for that to end before P-1 is run. How can I correct that? Perhaps break down to 2 workers or something?

Also, my Laptop is an HP 455G7 with an AMD Ryzen 4500U and Radeon RX Vega 6 GPU, if that helps. Yes, I tried using GPUowl, but that seems quite confusing with all the builds etc. So I'm just running Prime net on the CPU. It seems straightforward, just install, give user ID and run.

As for the reasons for my confidence, sadly its not scientifically rigorous. I guess that's fine if I can run a PRP and prove it's Prime right? So that's what I'm doing now.

Last fiddled with by IndrajithGamage on 2021-12-25 at 15:03

2021-12-25, 15:02   #6
Dr Sardonicus

Feb 2017
Nowhere

2·52·107 Posts

Quote:
 Originally Posted by IndrajithGamage Hi, so I have reserved (as in Assigned) the exponent 384,989,999 for PRP. As I have said earlier, I have significant reason to say that this will be the first ever 100M prime to be confirmed.
No, that's not what you said. You didn't say anything about a "reason," significant or otherwise. What you did say was,
Quote:
 Originally Posted by IndrajithGamage P.S. Here's a personal favor for you, in case waiting is not that great for you. The first 100M+ digit prime will be M384,989,999. Dedicate some time to that, just for the sake of believing a total stranger from a foreign country. Once you do that and win the price money, hopefully you'll feel inclined to believe that I'm not spewing uneducated rubbish here. (no, I don't want a cent from that) Then you can hopefully see for yourself why "Pi Exponent" is indeed a prime.
I note that, following up on the first sentence of the quote from your post to this thread,
Quote:
 I'm running the PRP as well but evidently that's taking quite a long time.
This raises a question: Why are you doing a PRP test on your latest I-proclaim-it-will-yield-a-Mersenne-prime exponent? In the second quote (from your earlier post), you are still insisting that the exponent 314159257 will yield a Mersenne prime, despite the fact that a verified PRP test had already proven M314159257 to be composite six months before you posted your claim. In the same earlier post quoted above, your reaction to this fact was,
Quote:
 M314159257 is prime, and if any methods currently used are saying otherwise, I can only think that some sort of error is causing that.
So, I have significant reason to say that, if your PRP test proves the number to be composite, you will simply refuse to accept the result.

Last fiddled with by Dr Sardonicus on 2021-12-25 at 15:36 Reason: Correct a misstatement

2021-12-25, 15:33   #7
IndrajithGamage

"Indrajith Gamage"
Dec 2021
Sri Lanka

23 Posts

Quote:
 Originally Posted by Dr Sardonicus No, that's not what you said. You didn't say anything about a "reason," significant or otherwise. What you did say was,I note that, following up on the first sentence of the quote from your earlier post,This raises a question: Why are you doing a PRP test on your latest I-proclaim-it-will-yield-a-Mersenne-prime exponent? In the second quote (from your earlier post), you are still insisting that the exponent 314159257 will yield a Mersenne prime, despite the fact that a verified PRP test had already proven M314159257 to be composite six months before you posted your claim. In the same earlier post quoted above, your reaction to this fact was, So, I have significant reason to say that, if your PRP test proves the number to be composite, you will simply refuse to accept the result.

Hi, I thought I'll be saying this to the media first, but since you asked, I'll clarify why.

Yes, I still think/know that Pi Exponent is indeed Prime, but looks like any hopes of convincing anyone so is dwindled. No one will believe when I say that some sort of glitch may have given an erroneous composite result, because I'm just another stranger right? So my first thought was to indeed run an LL on that and prove it.

However, I checked and it's really just barely below 100M digits long. So as long as I'm running something, why not do it on the biggest prize of 100M digits? Besides, it carries a huge monetary prize also if I read right, which is obviously a far more fruitful way to get compensated for my effort. I figured that when I prove the very first 100M Prime (and probably win the prize money as well), I'll have enough credential behind me when I ask GIMPS to just run an LL for the Pi exponent as well. More people will be inclined to try that under the circumstance right? I guess you see the logic here?

I certainly like the higher prize money attached to M384989999, but my emotional attachment is indeed with the Pi exponent M314159257. I figured doing the most sought after one (as well as hitherto not "proven" as composite) will be more convincing to all of you. I'm equally adamant that both of these numbers are Prime. However, as things stand, I'll have better luck convincing GIMPS by doing M384989999 first.

I know I sound quite non-nonsensical right now, but that's okay. You don't have to believe me. Help me run the P-1 and PRP on the M384989999 for now. After it's proven, we can discuss things further.

Last fiddled with by IndrajithGamage on 2021-12-25 at 15:43

2021-12-25, 15:46   #8
mathwiz

Mar 2019

DA16 Posts

Quote:
 Originally Posted by IndrajithGamage I know I sound quite non-nonsensical right now, but that's okay. You don't have to believe me. Help me run the P-1 and PRP on the M384989999 for now. After it's proven, we can discuss things further.
It's not really OK. You're flooding our forum with delusional BS and ignoring actual math. Please stop posting this garbage here.

2021-12-25, 15:50   #9
kriesel

"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

10111110101102 Posts

PRP/GEC/proof (successful cert) is accepted by capable mathematicians as proof positive the Mersenne number is composite when it indicates composite. M314159257 has been proven composite. Accept that. Or persist in sounding like a kook by denying it, and thereby giving us reason to not take your claims seriously. No matter how "confident" you are in believing it, 2+2 does not equal 5 or 3.3. The disbelief is not because you are a stranger to us, it is because you're posting demonstrable falsehoods.

As explained before, the Lucas Lehmer test is definitive for a prime, but the catch is, that the error checking possible is not nearly as strong, and so about 20% of tests for roughly that exponent size will be wrong in practice, and so it takes several LL tests to confirm a prime discovery using different software and hardware.
PRP/GEC has in practice, only ~24ppm error rate, and that is inflated from before the code outside the Gerbicz error check was hardened against possible bit errors. So PRP is now at least thousands of times more reliable in proving composites than the LL you are so fond of. Plus the VDF allows proving the completion and correctness of the entire PRP run, at a small fraction of the cost of an additional full LL or PRP run. (traditional double-check) There is no equivalent available for LL. There is no justification for wasting cycles on LL of numbers already proven composite by PRP & PRPDC or proof.

https://quoteinvestigator.com/2010/05/17/remain-silent/

Please post less, and read more, learn, and understand the reference info already linked above, Knuth, Riesel, or other good reference books. Raise your level of understanding to where you can begin to understand what it is you don't yet know about this very specialized area, which is apparently a great deal, and ask questions to learn after doing your own due diligence studying the available references and documentation. Expect to spend years of spare time doing it. That's what I've done, and I'm still a long ways from being able to do what the crack programmers or number theorists are doing. I have lots of company.

https://en.wikipedia.org/wiki/Dunnin...3Kruger_effect

I recommend tackling the learning curve and ceasing the posting of nonsensical content before you get banned from the forum by frequently annoying the moderators and others with provably false silly claims too often.
If you study hard and post little you might someday become a real asset to the group, beyond contributing compute cycles from a laptop on quixotic pursuits. Are you willing to do the long hard work that takes? In this specialized area, you appear to be at stage 1: https://hypnosistrainingacademy.com/...petence-model/
Quote:
 Originally Posted by IndrajithGamage I think it's still running the PRP, and waiting for that to end before P-1 is run. How can I correct that?
I already gave you step by step instructions. Paragraph 2 which you quoted, and then asked for it!

Oh, and all 3 of the posted png files produce this when I attempted to access them:
Code:
Secure Connection Failed

An error occurred during a connection to i.ibb.co. PR_CONNECT_RESET_ERROR

The page you are trying to view cannot be shown because the authenticity of the received data could not be verified.

Learn more…

Last fiddled with by kriesel on 2021-12-25 at 16:32

2021-12-25, 16:17   #10
Dr Sardonicus

Feb 2017
Nowhere

2·52·107 Posts

Quote:
 Originally Posted by IndrajithGamage No one will believe when I say that some sort of glitch may have given an erroneous composite result, because I'm just another stranger right?
Wrong.

No one will believe you because in the first place, you gave absolutely no justification for your claim that M314159257 is prime, in the second place because you displayed your ignorance of, e.g. the fact that a PRP test can prove a number to be composite, and in the third place you have not given any reason to question the result of the PRP test on M314159257.

And, you're only questioning this one specific test result, and only because it falsifies your claim. That certainly does nothing for your credibility.

What it does do is, is to reconfirm that you are a crank.

BTW, Ramanujan had a small library of books by circle-squarers and other cranks. (source: The World of Mathematics, Volume 1)

2021-12-25, 16:44   #11
IndrajithGamage

"Indrajith Gamage"
Dec 2021
Sri Lanka

108 Posts

Quote:
 Originally Posted by kriesel PRP/GEC/proof (successful cert) is accepted by capable mathematicians as proof positive the Mersenne number is composite when it indicates composite. M314159257 has been proven composite. Accept that. Or persist in sounding like a kook by denying it, and thereby giving us reason to not take your claims seriously. No matter how "confident" you are in believing it, 2+2 does not equal 5 or 3.3. The disbelief is not because you are a stranger to us, it is because you're posting demonstrable falsehoods. As explained before, the Lucas Lehmer test is definitive for a prime, but the catch is, that the error checking possible is not nearly as strong, and so about 20% of tests for roughly that exponent size will be wrong in practice, and so it takes several LL tests to confirm a prime discovery using different software and hardware. PRP/GEC has in practice, only ~24ppm error rate, and that is inflated from before the code outside the Gerbicz error check was hardened against possible bit errors. So PRP is now at least thousands of times more reliable in proving composites than the LL you are so fond of. Plus the VDF allows proving the completion and correctness of the entire PRP run, at a small fraction of the cost of an additional full LL or PRP run. (traditional double-check) There is no equivalent available for LL. There is no justification for wasting cycles on LL of numbers already proven composite by PRP & PRPDC or proof. https://quoteinvestigator.com/2010/05/17/remain-silent/ Please post less, and read more, learn, and understand the reference info already linked above, Knuth, Riesel, or other good reference books. Raise your level of understanding to where you can begin to understand what it is you don't yet know about this very specialized area, which is apparently a great deal, and ask questions to learn after doing your own due diligence studying the available references and documentation. Expect to spend years of spare time doing it. That's what I've done, and I'm still a long ways from being able to do what the crack programmers or number theorists are doing. I have lots of company. https://en.wikipedia.org/wiki/Dunnin...3Kruger_effect I recommend tackling the learning curve and ceasing the posting of nonsensical content before you get banned from the forum by frequently annoying the moderators and others with provably false silly claims too often. If you study hard and post little you might someday become a real asset to the group, beyond contributing compute cycles from a laptop on quixotic pursuits. Are you willing to do the long hard work that takes? In this specialized area, you appear to be at stage 1: https://hypnosistrainingacademy.com/...petence-model/ I already gave you step by step instructions. Paragraph 2 which you quoted, and then asked for it! Oh, and all 3 of the posted png files produce this when I attempted to access them: Code: Secure Connection Failed An error occurred during a connection to i.ibb.co. PR_CONNECT_RESET_ERROR The page you are trying to view cannot be shown because the authenticity of the received data could not be verified. Please contact the website owners to inform them of this problem. Learn more…
Thanks, and I'll stop replying to those who post things irrelevant to this. That's better. I appreciate if we can all keep the conversation limited to what's asked here.

So here are the said pngs. My question was/is, is that worktodo accurate. I specifically want to pause the PRP test, and do the P-1 at upper bound 50,000,000. I also want to know if the 4GB ram amount and other variables is sufficient. (if not, what are the best bounds for the Ram amount)

My PC is AMD Ryzen 4500U with 6 cores at 4GHz so conservatively 15-20GHz can be given. Ram is 8GB and can allocate like 200GB from my SSD (NVMe if that's important).
Attached Thumbnails

Last fiddled with by Dr Sardonicus on 2021-12-26 at 02:00 Reason: Remove problematic attachment

 Similar Threads Thread Thread Starter Forum Replies Last Post king Information & Answers 5 2018-02-21 18:15 Karl M Johnson Software 12 2015-10-12 15:56 Qubit Software 4 2014-01-15 05:04 137ben Software 9 2012-05-26 19:16 Unregistered Information & Answers 19 2008-05-30 08:34

All times are UTC. The time now is 08:35.

Fri Jan 21 08:35:25 UTC 2022 up 182 days, 3:04, 0 users, load averages: 1.67, 1.81, 1.57