mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > PrimeNet

Reply
 
Thread Tools
Old 2020-07-16, 23:58   #23
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

221278 Posts
Default

Quote:
Originally Posted by Rodrigo View Post
That said, it may be time to consider more automated methods for TFing on GPUs.
Never send a human to do a machine's job...

Check out primetools for Linux and MISFIT for Wincrows.

They're literally fire-and-forget.
chalsall is offline   Reply With Quote
Old 2020-07-17, 00:14   #24
Rodrigo
 
Rodrigo's Avatar
 
Jun 2010
Pennsylvania

16378 Posts
Default

Quote:
Originally Posted by chalsall View Post
Never send a human to do a machine's job...

Check out primetools for Linux and MISFIT for Wincrows.

They're literally fire-and-forget.
Yeah, I may need to finally do that and find other ways to stay engaged with the action, such as checking my account's progress (or lack thereof) on the leader boards every few days.

Thanks for the primetools and MISFIT suggestions.
Rodrigo is offline   Reply With Quote
Old 2020-07-17, 00:36   #25
storm5510
Random Account
 
storm5510's Avatar
 
Aug 2009
U.S.A.

1,499 Posts
Default

Quote:
Originally Posted by Rodrigo View Post
..I've greatly enjoyed doing the GPU manual assignment + reporting thing for all these years. Keeps me engaged with the action IMHO. That said, it may be time to consider more automated methods for TFing on GPUs.
There is one GPU automated site I know of: https://www.mersenne.ca/

There is a batch file for Windows, a bash file for Linux, and, a Python script. Just look for "Trial factoring assignments > 1000M." It is in the list on the right side, third down from the top. I ran it for a long time. It takes a little setup to get started. After that, it takes care of itself.
storm5510 is offline   Reply With Quote
Old 2020-07-17, 13:34   #26
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

2·4,297 Posts
Default

Quote:
Originally Posted by storm5510 View Post
There is a batch file for Windows, a bash file for Linux, and, a Python script. Just look for "Trial factoring assignments > 1000M."
The but if you want to work in the area that is currently useful, >1000M is not it. TF that is currently useful is ≤20,000,000 ahead of the current DC and FT wave fronts.
Uncwilly is offline   Reply With Quote
Old 2020-07-18, 15:33   #27
storm5510
Random Account
 
storm5510's Avatar
 
Aug 2009
U.S.A.

1,499 Posts
Default

Quote:
Originally Posted by Uncwilly View Post
The but if you want to work in the area that is currently useful, >1000M is not it. TF that is currently useful is ≤20,000,000 ahead of the current DC and FT wave fronts.
Both are true, and there is no credit given for exponents >1000M.

Misfit: What a nightmare this created for me. I did not properly configure it and I ended up with nearly a thousand TF's to run, including a lot of duplicates. At the time, this was for GPUto72. It took quite a while to clean those up with a GPU. I could have "unreserved" a lot of them, but I did not want to. I stopped using it. I held on to one of its fetch files. I was able to tailor a batch file to use that and get a fixed number of assignments on each pass. It worked quite well even though I had to submit the results manually.
storm5510 is offline   Reply With Quote
Old 2020-07-23, 03:35   #28
LOBES
 
Mar 2019
USA

24 Posts
Default

When I get assignments directly from the Manual GPU Assignment page, I will usually extend the end factor. For example, I got exponent 995397037, and it assigned it to me (LOBES), and was expecting 2^71 to 2^72. In my worktodo.txt, I changed the 72 to a 78, and started it up. In this case, I am under the impression that this exponent is "assigned" to me, and won't be assigned to someone else until my TF expires or is submitted, so I assume it is ok to extend that limit. Am I correct? I would feel terrible if I was impacting someone else. Granted, it only takes my GPU about an hour to go from 71 to 78, so the chances of me stepping on toes is probably low.

However, when I get exponents from GPU72, I never change the end factor, since I think it is possible to pledge a higher factor on the same exponent. I could be wrong, but I didn't want to take any chances.
LOBES is offline   Reply With Quote
Old 2020-07-23, 04:08   #29
Uncwilly
6809 > 6502
 
Uncwilly's Avatar
 
"""""""""""""""""""
Aug 2003
101×103 Posts

2×4,297 Posts
Default

If you don't submit the results until you submit the whole run, up to 78, you are fine. If you submit them as they complete it is possible (but unlikely) that you could turn it in and have someone pick it up for the higher bits.

With GPU72 your pledge level is what you should stick to.

GPU72 is the best option, because it is more closely managed to maximize the benefit to GIMPS. And with your GPU, it would be most beneficial to GIMPS for you to work down in the lower exponents. We won't get to testing the 995,xxx,xxx range for a decade or more.
Uncwilly is offline   Reply With Quote
Old 2020-07-23, 04:13   #30
LOBES
 
Mar 2019
USA

24 Posts
Default

Quote:
Originally Posted by Uncwilly View Post
GPU72 is the best option, because it is more closely managed to maximize the benefit to GIMPS. And with your GPU, it would be most beneficial to GIMPS for you to work down in the lower exponents. We won't get to testing the 995,xxx,xxx range for a decade or more.
What about choosing "Trial Factoring" from the regular manual assignments page (as opposed to the GPU assignments page) and just letting it choose the exponent? Will it pick the most beneficial lower exponent to work on?

Or would GPU72 still be the best option?
LOBES is offline   Reply With Quote
Old 2020-07-23, 04:35   #31
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

3×13×113 Posts
Default

Try the settings in the attachment.
Attached Thumbnails
Click image for larger version

Name:	gpu tf.png
Views:	45
Size:	73.6 KB
ID:	22859  
kriesel is offline   Reply With Quote
Old 2020-07-23, 16:27   #32
storm5510
Random Account
 
storm5510's Avatar
 
Aug 2009
U.S.A.

101110110112 Posts
Default

Quote:
Originally Posted by Uncwilly View Post
...With GPU72 your pledge level is what you should stick to.

GPU72 is the best option, because it is more closely managed to maximize the benefit to GIMPS. And with your GPU, it would be most beneficial to GIMPS for you to work down in the lower exponents. We won't get to testing the 995,xxx,xxx range for a decade or more.
I never changed the pledge level in the config file. I left it at 72. Over time, the assignments went past that. I had some which were 2^75. I understand the reason. Staying closer to the wavefront. There are probably members who do not wish to factor this high using slower machines without GPU's. If a person elects to stay at 72, then do not assign them any exponents beyond that, or indicate there is nothing available at this level. If the latter happens to be true, then they can go to mersenne.org and pick what they want as manual assignments.
storm5510 is offline   Reply With Quote
Old 2020-07-23, 16:47   #33
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

3·7·443 Posts
Default

Quote:
Originally Posted by storm5510 View Post
There are probably members who do not wish to factor this high using slower machines without GPU's. If a person elects to stay at 72, then do not assign them any exponents beyond that...
That's why GPU72 has various options.

"What Makes Sense" will never assign work beyond the requested pledge level (although it will sometimes assign work to a lower pledge level).

"Let GPU72 Decide", on the other hand, will set the TF'ing level to what makes the most sense at that paricular time. (Yeah, I know... Historical reasons for this somewhat backward policy.)

The other options (Lowest Exponent, Lowest TF Depth, etc) will also never assign work past the pledge level.
chalsall is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Poaching blip Data 8 2016-01-30 01:59
Poaching Kevin Lounge 121 2010-10-26 06:49
Poaching davieddy Lounge 6 2010-10-16 12:31
Poaching and v5 PrimeCruncher PrimeNet 6 2004-04-05 19:17
New fashion poaching (???) lycorn Lounge 6 2003-01-31 08:33

All times are UTC. The time now is 10:22.

Sat Sep 19 10:22:27 UTC 2020 up 9 days, 7:33, 0 users, load averages: 1.06, 1.32, 1.48

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.