mersenneforum.org Report Results Here
 Register FAQ Search Today's Posts Mark Forums Read

2019-06-20, 15:04   #1156
James Heinrich

"James Heinrich"
May 2004
ex-Northern Ontario

3×1,039 Posts

Quote:
 Originally Posted by lavalamp Sorry to say this could be a while longer. Hence the large reservations.
No worries, it's been waiting 10 years, another few days/weeks won't make any difference

But on the subject of keeping your GPU fed while you might not be looking at it, I have recently simplified the automated work fetching for OBD users, and might encourage you guys to automate your workflow:
The basic batch/bash file template can be found on https://www.mersenne.ca/tf1G.php (at the bottom of the "Tips" box in the upper-right).
For OBD work you'll need to modify the work-fetching wget URL to this:
That URL will return a single assignment to <bits> (preference given to exponents with highest current TF) and will be assigned to <username> on the OBD page for 30 days.

The script runs in a loop, when it runs out of work it submits and archives the current result, fetches more work, and keeps going. If it fails to get an assignment for whatever reason it will wait 5 mins and try again.

Please give it a shot and let me know if you find any problems.

2019-06-20, 15:42   #1157
ET_
Banned

"Luigi"
Aug 2002
Team Italia

25·149 Posts

Quote:
 Originally Posted by James Heinrich No worries, it's been waiting 10 years, another few days/weeks won't make any difference But on the subject of keeping your GPU fed while you might not be looking at it, I have recently simplified the automated work fetching for OBD users, and might encourage you guys to automate your workflow: The basic batch/bash file template can be found on https://www.mersenne.ca/tf1G.php (at the bottom of the "Tips" box in the upper-right). For OBD work you'll need to modify the work-fetching wget URL to this: https://www.mersenne.ca/tf1G.php?obdtf=&obduser= That URL will return a single assignment to (preference given to exponents with highest current TF) and will be assigned to on the OBD page for 30 days. The script runs in a loop, when it runs out of work it submits and archives the current result, fetches more work, and keeps going. If it fails to get an assignment for whatever reason it will wait 5 mins and try again. Please give it a shot and let me know if you find any problems.
I am working on three different factorization projects, so I am used to manually interleave OBD exponents among other ranges and balance the worktodo.txt file, but the idea is nice.

 2019-06-20, 20:57 #1158 WGJC3107     "Will Conway" May 2018 UK 4016 Posts Not sure if this is the right place to ask but I can’t seem to find a sub-forum related to general TF1G work. How often do results get checked on TF1G? I submitted ~70 results in the 1583M range the other day and they don’t seem to have gone through. Last fiddled with by WGJC3107 on 2019-06-20 at 20:57
2019-06-20, 23:36   #1159
James Heinrich

"James Heinrich"
May 2004
ex-Northern Ontario

3×1,039 Posts

Quote:
 Originally Posted by WGJC3107 Not sure if this is the right place to ask but I can’t seem to find a sub-forum related to general TF1G work. How often do results get checked on TF1G? I submitted ~70 results in the 1583M range the other day and they don’t seem to have gone through.
There isn't a specific TF1G discussion space as far as I know, although perhaps there should be.
The results queue (which is the same queue for OBD) processes up to 1000 factor and 10000 nofactor results every minute. Every now and then there's a huge dump of hundreds of thousands of results that clog up the queue for a little while, but generally your results should be processed within a couple minutes of you submitting them. You're free to resubmit them (but if they weren't accepted the first time I don't believe they'd be handled any differently the second time), or email me your results and I'll take a look if there's something weird in your result format that's causing them to be rejected.

 2019-06-21, 14:53 #1160 WGJC3107     "Will Conway" May 2018 UK 4016 Posts I’ll just run through them again and email them to you in case they don’t come through
 2019-06-21, 18:43 #1161 WGJC3107     "Will Conway" May 2018 UK 26 Posts I've sent the results via email, my address is conwayderek44 AT gmail Last fiddled with by axn on 2019-06-22 at 01:56 Reason: email
 2019-06-21, 20:09 #1162 Dylan14     "Dylan" Mar 2017 20516 Posts Another result: Code: UID: Dylan14/desktop, no factor for M3321930229 from 2^85 to 2^86 [mfaktc 0.21 barrett87_mul32_gs] This result moves us from level 19.09 to level 19.10.
 2019-06-21, 21:08 #1163 WGJC3107     "Will Conway" May 2018 UK 26 Posts Very nice. Also, how long will the 2^86 to the 2^90 reservations by lavalamp take?
 2019-06-22, 03:48 #1164 clowns789     Jun 2003 The Computer 2×191 Posts I'm happy to see this project is being revived. The mersenne.ca site looks great. I submitted a bit of work a little while back to finally say I contributed to what I started, but I just looked again and getting one exponent from 85 to 86 bits would take 3 weeks on my GeForce 960M! I'll hold off for now, but thank everyone who is participating. I see the project going to level 25 (25 candidates to 91 bits), then branching off into wide (adding more exponents as requested by a few members) and deep (P-1/PRP when it becomes practical) searches. It looks like even at this pace we'll have several years where we can still follow the traditional format. As a final aside, the name "Operation Billion Digits" stems from Operation Iraqi Freedom, which was launched the year before this project was started. Let's hope there are no more "operations" any time soon.
2019-06-22, 13:59   #1165
lavalamp

Oct 2007
London, UK

51B16 Posts

Quote:
 Originally Posted by WGJC3107 Also, how long will the 2^86 to the 2^90 reservations by lavalamp take?
I checked in on my pooter remotely, it's just finished the first 86 to 87 bits leg and that took t = 2d 12h 34m 10.311s.

Extrapolating, I would estimate the whole 86 to 90 bits range will take 15 x t = 37ds 20h 32m 34.665s. However I'll be back home before then and finding ways to slow it down like using it and playing games.

I hope to finish the first one by the end of July still, perhaps I will run the second reservation also, at the moment it is only reserved as a backup while I am away.

Edit: Although so far, the first 2 classes in the 87-88 bit range have taken slightly over double the time of the classes in the previous range, so perhaps there is an additional small performance drop due to the bigger numbers in the powmod function. Though equally this could just be due to me browsing the machine remotely.

Last fiddled with by lavalamp on 2019-06-22 at 14:11

2019-06-22, 20:01   #1166
Dylan14

"Dylan"
Mar 2017

11·47 Posts

Quote:
 Originally Posted by lavalamp Although so far, the first 2 classes in the 87-88 bit range have taken slightly over double the time of the classes in the previous range, so perhaps there is an additional small performance drop due to the bigger numbers in the powmod function. Though equally this could just be due to me browsing the machine remotely.
I believe it’s the former reason, as now you are using the gpu kernel “BARRETT88_MUL32_GS” instead of “BARRETT87_MUL32_GS”, which I suspect involves wider multiplications. See https://www.mersenneforum.org/showpo...23&postcount=6.

 Similar Threads Thread Thread Starter Forum Replies Last Post petrw1 PrimeNet 5 2018-07-09 21:14 houding PrimeNet 15 2018-01-10 19:24 dcorbit Information & Answers 4 2012-08-13 21:53 James Heinrich Marin's Mersenne-aries 6 2004-05-27 05:47 remaker Software 4 2002-10-17 08:18

All times are UTC. The time now is 03:06.

Fri Oct 23 03:06:02 UTC 2020 up 43 days, 17 mins, 0 users, load averages: 1.66, 1.69, 1.60