mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Hardware > GPU Computing

Reply
 
Thread Tools
Old 2019-11-05, 06:01   #1
Rodrigo
 
Rodrigo's Avatar
 
Jun 2010
Pennsylvania

92710 Posts
Default Unable to get manual TF assignments for GPU

Not sure if this or the PrimeNet subforum is the right place for this topic, but here goes...

Over the past several days, more than once I have been unable to obtain most types of manual GPU assignments. Of the preferred work range categories available in the drop-down menu, only "TF for 100M digit Mersenne numbers" works. The rest meet with a result like this one:

Code:
aError code: 40 
Error text: No assignment available for GPU trial factoring, cpu_id:
Is this a glitch, or an intended result?

Thanks for any info on this!
Rodrigo is offline   Reply With Quote
Old 2019-11-05, 07:42   #2
axn
 
axn's Avatar
 
Jun 2003

10010010111002 Posts
Default

I got some work in the 100m bits (not digits) range by giving:

Preferred work range: TF for first time LL
Exponent rage: 95m to 105m
Work preference: Lowest exponent
Bit Level: 76
axn is offline   Reply With Quote
Old 2019-11-05, 14:12   #3
Prime95
P90 years forever!
 
Prime95's Avatar
 
Aug 2002
Yeehaw, FL

11100000000112 Posts
Default

A little history. This page was originally written to replace the heuristics GPU72 uses. GPU72 would still exist, but would basically relay requests to this page and return the assignments. GPU72 would no longer reserve huge blocks of TF assignments.

Given that goal, the server tracks what range of exponents it expects to be handing out for LL in the next 90(?) days. The GPUs would then TF those ranges as deep as possible - with assignments always given out to maximize the LL-time-saved vs. chance-of-finding-a-factor.

Fast forward to today. GPU72 never changed over to this system. That is fine, Chris does a great job monitoring the various wavefronts, probably better than an algorithm I came up with 5 years ago. This page is now used only by a few non-GPU72 users. What has happened is GPU72 has reserved pretty much all of the next-90-day ranges (displayed on the bottom the web page).

What you need to do is manually extend the top end of the LL range.
Prime95 is online now   Reply With Quote
Old 2019-11-05, 14:23   #4
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

4,457 Posts
Default

Quote:
Originally Posted by Prime95 View Post
A little history. This page was originally written to replace the heuristics GPU72 uses. GPU72 would still exist, but would basically relay requests to this page and return the assignments. GPU72 would no longer reserve huge blocks of TF assignments.

Given that goal, the server tracks what range of exponents it expects to be handing out for LL in the next 90(?) days. The GPUs would then TF those ranges as deep as possible - with assignments always given out to maximize the LL-time-saved vs. chance-of-finding-a-factor.

Fast forward to today. GPU72 never changed over to this system. That is fine, Chris does a great job monitoring the various wavefronts, probably better than an algorithm I came up with 5 years ago. This page is now used only by a few non-GPU72 users. What has happened is GPU72 has reserved pretty much all of the next-90-day ranges (displayed on the bottom the web page).

What you need to do is manually extend the top end of the LL range.
Thanks for the info.
And please consider moving the upper limit of the gpu manual assignments page, to a level that will reduce or eliminate error 40. It used to be a rare occurrence, but lately it seems to be the default response. Currently the upper limit is 96M but needs to be manually increased to 101M every time to obtain any new manual TF assignments. Also lowest available exponent choice (instead of highest bits) I'd like saved from one assignment fetch to the next.
kriesel is online now   Reply With Quote
Old 2019-11-05, 16:21   #5
Rodrigo
 
Rodrigo's Avatar
 
Jun 2010
Pennsylvania

39F16 Posts
Default

Thanks everyone for the interesting background and the helpful tips.
Rodrigo is offline   Reply With Quote
Old 2020-04-22, 14:20   #6
leonardyan96
 
leonardyan96's Avatar
 
"Hazawa Tsugumi"
May 2017
Northern China

2×7 Posts
Default

Quote:
Originally Posted by Prime95 View Post
A little history. This page was originally written to replace the heuristics GPU72 uses. GPU72 would still exist, but would basically relay requests to this page and return the assignments. GPU72 would no longer reserve huge blocks of TF assignments.

Given that goal, the server tracks what range of exponents it expects to be handing out for LL in the next 90(?) days. The GPUs would then TF those ranges as deep as possible - with assignments always given out to maximize the LL-time-saved vs. chance-of-finding-a-factor.

Fast forward to today. GPU72 never changed over to this system. That is fine, Chris does a great job monitoring the various wavefronts, probably better than an algorithm I came up with 5 years ago. This page is now used only by a few non-GPU72 users. What has happened is GPU72 has reserved pretty much all of the next-90-day ranges (displayed on the bottom the web page).

What you need to do is manually extend the top end of the LL range.
So, does this mean, if an exponent is reported to be assigned to someone named "GPU Factoring", it has actually gone to GPU72? This is a long-time question of me because I have never seen this user to publish any result.
leonardyan96 is offline   Reply With Quote
Old 2020-04-22, 14:23   #7
kladner
 
kladner's Avatar
 
"Kieren"
Jul 2011
In My Own Galaxy!

234348 Posts
Default

Quote:
Originally Posted by leonardyan96 View Post
So, does this mean, if an exponent is reported to be assigned to someone named "GPU Factoring", it has actually gone to GPU72? This is a long-time question of me because I have never seen this user to publish any result.
The results show up for the individual users who got assignments from GPU72.
kladner is offline   Reply With Quote
Old 2020-04-22, 14:40   #8
leonardyan96
 
leonardyan96's Avatar
 
"Hazawa Tsugumi"
May 2017
Northern China

168 Posts
Default

Quote:
Originally Posted by kladner View Post
The results show up for the individual users who got assignments from GPU72.
I have always been thinking "GPU Factoring" is just a troll that reserves large amount of tasks without actually running the computation.
leonardyan96 is offline   Reply With Quote
Old 2020-04-22, 14:49   #9
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

2·4,657 Posts
Default

Quote:
Originally Posted by leonardyan96 View Post
I have always been thinking "GPU Factoring" is just a troll that reserves large amount of tasks without actually running the computation.
Nope. The account "GPU Factoring" was created before GPU72 was even named thusly.

And, so you know, while no assignments are actually completed by this account, what happens is GPU72 "lends" out assignments to TF'ers, who submit the results back to Primenet (and so get the credits).

This repeats (to ever higher bit-levels) until the candidate has either been factored, or it's been TF'ed as high as makes sense. They're then released back to Primenet to assign for either P-1'ing or FC'ing.
chalsall is offline   Reply With Quote
Old 2020-04-22, 15:03   #10
leonardyan96
 
leonardyan96's Avatar
 
"Hazawa Tsugumi"
May 2017
Northern China

2×7 Posts
Default

Quote:
Originally Posted by chalsall View Post
Nope. The account "GPU Factoring" was created before GPU72 was even named thusly.

And, so you know, while no assignments are actually completed by this account, what happens is GPU72 "lends" out assignments to TF'ers, who submit the results back to Primenet (and so get the credits).

This repeats (to ever higher bit-levels) until the candidate has either been factored, or it's been TF'ed as high as makes sense. They're then released back to Primenet to assign for either P-1'ing or FC'ing.
Thanks for your answer. That makes me relieved because now I'm sure all these exponents are taken care of.
leonardyan96 is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
ECM manual assignments and # of curves hansl Information & Answers 5 2019-07-23 16:38
Manual assignments cpu/generic vs gpu TF kriesel PrimeNet 1 2019-06-04 15:00
unable to fetch assignments dragonbud20 MISFIT 2 2016-10-18 08:01
On getting manual assignments Qubit PrimeNet 2 2014-07-30 12:30
Unable to migrate manual assignment to Primenet 5. xode PrimeNet 14 2009-05-25 20:48

All times are UTC. The time now is 19:57.

Wed Sep 30 19:57:22 UTC 2020 up 20 days, 17:08, 0 users, load averages: 1.88, 1.91, 1.85

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.