mersenneforum.org  

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

Reply
 
Thread Tools
Old 2018-11-28, 11:12   #1
SELROC
 

5·17·113 Posts
Default Trial factoring with multiple bit-level assignment

I just got 200 TF assignments, the bit-level is from 71 to 76 for each exponent, so it computes each bit-level in sequence before moving to the next assignment.
  Reply With Quote
Old 2018-11-28, 15:12   #2
SELROC
 

3·23·59 Posts
Default

Quote:
Originally Posted by SELROC View Post
I just got 200 TF assignments, the bit-level is from 71 to 76 for each exponent, so it computes each bit-level in sequence before moving to the next assignment.

photo
Attached Thumbnails
Click image for larger version

Name:	mfakto_small.png
Views:	42
Size:	705.5 KB
ID:	19330  
  Reply With Quote
Old 2018-11-28, 15:25   #3
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

52·163 Posts
Default

Quote:
Originally Posted by SELROC View Post
I just got 200 TF assignments, the bit-level is from 71 to 76 for each exponent, so it computes each bit-level in sequence before moving to the next assignment.
Yes, and if you look at what's happening in the worktodo file, it's editing the first assignment as it goes up the bit levels. That is,
Factor=aid,exponent1,71,76
becomes
Factor=aid,exponent1,72,76
then
Factor=aid,exponent1,73,76
and
Factor=aid,exponent1,74,76
finally
Factor=aid,exponent1,75,76
and then the record for exponent1 is removed from the worktodo after the last level assigned is completed.
Then the same process starts with
Factor=aid2,exponent2,71,76
etc.
I think doing it that way is much simpler than trying to sweep the worktodo file for the lowest bit level. It also gives earlier completion of the leading exponents in the worktodo file.

Last fiddled with by kriesel on 2018-11-28 at 15:26
kriesel is online now   Reply With Quote
Old 2018-11-28, 15:35   #4
Mark Rose
 
Mark Rose's Avatar
 
"/X\(‘-‘)/X\"
Jan 2013
Ͳօɾօղէօ

278910 Posts
Default

Also keep in mind that if you reserved the exponent for TF using the https://www.mersenne.org/manual_gpu_assignment/ page, as soon as you submit the first result for an exponent, you lose the assignment. So if you're getting your assignments from there, be sure to submit all your results for an exponent at the same time.
Mark Rose is offline   Reply With Quote
Old 2018-11-28, 15:46   #5
SELROC
 

47510 Posts
Default

Quote:
Originally Posted by Mark Rose View Post
Also keep in mind that if you reserved the exponent for TF using the https://www.mersenne.org/manual_gpu_assignment/ page, as soon as you submit the first result for an exponent, you lose the assignment. So if you're getting your assignments from there, be sure to submit all your results for an exponent at the same time.

no I'm still using the old page https://www.mersenne.org/manual_assignment/
the new gpu assignment page most of the times replies with an error.
I upload the results file to https://www.mersenne.org/manual_result/ , the server should ingest all results at the same time.
  Reply With Quote
Old 2018-11-28, 15:47   #6
ET_
Banned
 
ET_'s Avatar
 
"Luigi"
Aug 2002
Team Italia

129316 Posts
Default

Quote:
Originally Posted by Mark Rose View Post
Also keep in mind that if you reserved the exponent for TF using the https://www.mersenne.org/manual_gpu_assignment/ page, as soon as you submit the first result for an exponent, you lose the assignment. So if you're getting your assignments from there, be sure to submit all your results for an exponent at the same time.
I suppose it works that way even if you took your exponents from GPUto72, so be careful sending out your results
ET_ is offline   Reply With Quote
Old 2018-11-28, 16:06   #7
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

52·192 Posts
Default

Quote:
Originally Posted by ET_ View Post
I suppose it works that way even if you took your exponents from GPUto72, so be careful sending out your results
Nope. The candidates from GPU72 are "owned" by a dedicated Primenet user ("GPU Factoring"), and are "lent" out to participants to work.

Only once the system notices that the pledged work has been completed is the candidate unreserved and returned to Primenet to manage.
chalsall is online now   Reply With Quote
Old 2018-11-28, 16:19   #8
kriesel
 
kriesel's Avatar
 
"TF79LL86GIMPS96gpu17"
Mar 2017
US midwest

52×163 Posts
Default

Quote:
Originally Posted by Mark Rose View Post
Also keep in mind that if you reserved the exponent for TF using the https://www.mersenne.org/manual_gpu_assignment/ page, as soon as you submit the first result for an exponent, you lose the assignment. So if you're getting your assignments from there, be sure to submit all your results for an exponent at the same time.
That assignment page does seem a bit temperamental.
I find it works ok if I enter lower and upper bounds of a very broad exponent range and ask for lowest exponents, specify a number of assignments, and leave the rest blank. Setting the top tf bit level is optional.

There are a few ways of ensuring your TF work is not duplicated. The surest is to totally complete all assigned or needed bit levels before reporting any TF levels for the exponent. Another is to report several bit levels of an exponent, such that the usual reassignment will skip that reported exponent for a while; it creates a "higher island" than will be assigned for a while, until the factoring tide rises around it, if you've reported 3 or more levels while the usual assignment and result is a single bit level. Another way is to work on exponents so large there's little activity on them. Another is to reserve an exponent for P-1. The server will respond the TF result (result type 4) does not match the assignment type (assignment type 4!), but accept the TF reports (which can be incrementally reported as high as you like) and credit the work. The P-1 assignment will remain until a P-1 result is reported. On the chance the P-1 assignment may return a factor, TF assignment is not likely to issue while P-1 work is pending. Conversely, if TF is assigned on an exponent, P-1 should not be, while the TF assignment is pending, since the next TF level _might_ return a factor making P-1 factoring unneeded for th GIMPS purpose of finding primes.

Last fiddled with by kriesel on 2018-11-28 at 16:23
kriesel is online now   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Trial Factoring by GPU my LL assignment Luis PrimeNet 3 2016-06-28 18:41
Multiple threads per assignment? f0rteOC Hardware 3 2016-02-29 06:08
Multiple Core for 1 assignment saeres Information & Answers 10 2016-02-04 01:52
using multiple threads on an LL assignment tha Software 4 2016-02-02 13:49
option for finding multiple factors during trial factoring tha Software 24 2014-06-10 23:31

All times are UTC. The time now is 17:24.

Sat Jul 11 17:24:13 UTC 2020 up 108 days, 14:57, 1 user, load averages: 1.48, 1.27, 1.22

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.