mersenneforum.org  

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

Reply
 
Thread Tools
Old 2021-08-09, 00:17   #650
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

23·5·251 Posts
Default

OK... I spent some "quality time" with the GPU72 codebase, getting the DCTF code paths to be better "aware" of the TwoK sub-project.

For anyone interested in helping Wayne et al, the DC TF assignment form has been refactored to properly make available (and preview) the candidates GPU72 is currently "curating".

The Colab Notebook's Instance Worktypes will now give out candidates to TF, and specific 2xM ranges which have been poorly P-1'ed (TF'ed to at least 73 first). The default "Let GPU72 Decide" chooses these work types; anyone who wants to continue doing FTC work has that option by simply setting an explicit work type.

Please let me know if anyone sees any SPEs, or if there are any ranges anyone wants to work on by themselves (and I'll remove them from the database).

Or... If anyone has a particular range they'd like to P-1, some TF'ing first can soften things up a bit...
chalsall is online now   Reply With Quote
Old 2021-08-09, 08:45   #651
LaurV
Romulan Interpreter
 
LaurV's Avatar
 
"name field"
Jun 2011
Thailand

24·613 Posts
Default

Good! Well done!

Next step: offer P-1 assignments to colab instances...
(and we don't need to reserve, neither report results by hand, which take us some precious time, hehe).

Last fiddled with by LaurV on 2021-08-09 at 08:45
LaurV is offline   Reply With Quote
Old 2021-08-09, 14:23   #652
tha
 
tha's Avatar
 
Dec 2002

827 Posts
Default

Quote:
Originally Posted by chalsall View Post
Or... If anyone has a particular range they'd like to P-1, some TF'ing first can soften things up a bit...
I am still active in the entire 9M range doing P-1 with B1=2M and B2=162M

Parts have extensive TF on them and other parts a lot less. I prefer large ranges done to the same standard, rather than quit when a certain amount of factors have been found in that range.
tha is offline   Reply With Quote
Old 2021-08-09, 23:43   #653
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

273816 Posts
Default

Quote:
Originally Posted by tha View Post
I am still active in the entire 9M range doing P-1 with B1=2M and B2=162M. Parts have extensive TF on them and other parts a lot less.
OK. As a sanity check of the various code paths, I've prioritized fifty (50#) candidates currently at 69 "bits" in 9.2M (where I'm inferring you are currently working) to be taken up to 73 by the DCTF Colab workers. Sorted by B2 accending; this work is underway right now.

Quote:
Originally Posted by tha View Post
I prefer large ranges done to the same standard, rather than quit when a certain amount of factors have been found in that range.
That is a common approach by many. I personally like to see things neat (OCD).

To provide better "situational awareness", please see the updated TwoK Report. Hopefully, this will make it clearer what's going on. Suggestions for additional data are welcomed.

Also, would people be interested in being able to get TwoK P-1 work from GPU72 by way of the forms (in addition to the Colab CPU worktypes)?
chalsall is online now   Reply With Quote
Old 2021-08-10, 00:30   #654
petrw1
1976 Toyota Corona years forever!
 
petrw1's Avatar
 
"Wayne"
Nov 2006
Saskatchewan, Canada

22·3·13·31 Posts
Default

Quote:
Originally Posted by chalsall View Post
To provide better "situational awareness", please see the updated TwoK Report. Hopefully, this will make it clearer what's going on. Suggestions for additional data are welcomed.
I like

Quote:
Originally Posted by chalsall View Post
Also, would people be interested in being able to get TwoK P-1 work from GPU72 by way of the forms (in addition to the Colab CPU worktypes)?
The following will probably confuse more than enlighten ... sorry.
I can get more elaborate if required.

The assignments need to have B1/B2 appropriate to encourage completion of the range.
The difficulty is not knowing the balance of TF vs P1 that is/will be available.
The determines how aggressive P1 needs to be vs how deep TF will go and hence what percentage of the remaining factors each method will find.

This is probably best served if you can allow us to specify either:
- our own B1/B2
- required percentage: either actual or increase required
- number of LL tests saved parm ... which dictates required B1/B2
OR if it defaults to relatively aggressive B1/B2
petrw1 is online now   Reply With Quote
Old 2021-08-10, 02:24   #655
axn
 
axn's Avatar
 
Jun 2003

5,179 Posts
Default

Quote:
Originally Posted by petrw1 View Post
OR if it defaults to relatively aggressive B1/B2
OR if it scan's the exponent's history of past P-1, TF, ECM, etc. and determines the best B1/B2 appropriate for the exponent/range.

Just sayin'
axn is offline   Reply With Quote
Old 2021-08-10, 08:13   #656
lycorn
 
lycorn's Avatar
 
"GIMFS"
Sep 2002
Oeiras, Portugal

11×137 Posts
Default

Quote:
Originally Posted by chalsall View Post
Thanks for checking in. It's easy to inadvertently "step on toes" when everyone's working "off-the-books".


Please see this report for some context
What is the meaning of "owned" in the report?

I´ve been working in the 14.8 and 14.9 ranges (and had posted that in this thread couple of weeks ago). Yesterday I sent some results to the server only to find out that they were no longer needed as SRBASE had already returned them. So I assume it is better to stay away from 0 - 20M ranges for the time being.
Or is it safe to work on 16.9M from 70 to 71, although all remaining 2096 exponents are listed as "owned" in the report?

Last fiddled with by lycorn on 2021-08-10 at 08:14
lycorn is online now   Reply With Quote
Old 2021-08-10, 10:28   #657
LaurV
Romulan Interpreter
 
LaurV's Avatar
 
"name field"
Jun 2011
Thailand

24×613 Posts
Default

Hey Chris, let me be the first to congratulate you for changing the format of the twok table, it looks nicer and it has more comprehensive information. I think you did a little bit of work there! On the other hand, you just hit me in the balls so hard I was using your table as input for my own table, and separation in ranges, as well as substituting 4400000 with 4.4M, etc, albeit easier to read, it screws up my format. Now I will have to work a bit more to maintain it, and I think I will give up, anyhow your new format has all the data inside, and straw-fire-SRBase joining the fray impede me to boast about the fact that I was producing 80% of the factors (which was the case till few days ago, but currently after SRBase started raining the factors, or better, fireworks with factors, they shot factors in all directions , I found out that I only produced about 16% of the total factors produced since I joined ~63 days ago; anyhow the "fire-from-straws" will burn fast and be gone and I will stay a bit longer till I'll finish 11M, recovering a part of the "lost face" later, hehe).

On the more positive note, let me congratulate you for reaching 200 ranges - one more and we can say "less than 200 for less than 2000" hehe, that is because 11.1M is done (your table will catch it in an hour or two) and due to the help from SRBase, about 38% of the 11.1M range remained not searched with high P-1 bounds. As good people here suggested, I will delete the rest of the assignments and move back to 11.4, letting some low hanging fruits for whoever will want to exercise their GPUs in the future.

For records, there are 2 factors found by Dylan Delgado, 21 found by SRBase TF to 70 bits, and the rest to 97 or 98 (96 were needed, but I lost the track again and found some more) were found by me, with high-bound P-1 on ~62% of the range. There was no P+1 done by me (don't know for other people).

200 sub-ranges to go.

Last fiddled with by LaurV on 2021-08-10 at 11:39
LaurV is offline   Reply With Quote
Old 2021-08-10, 11:32   #658
ZacHFX
 
ZacHFX's Avatar
 
Mar 2017
Halifax, NS

3810 Posts
Default

Like Lycorn's post above, I tried to submit 70-71 bits in the 12.3M range today, and found that SRBase had completed to 70 bits already.
ZacHFX is offline   Reply With Quote
Old 2021-08-10, 11:45   #659
lycorn
 
lycorn's Avatar
 
"GIMFS"
Sep 2002
Oeiras, Portugal

11·137 Posts
Default

Yes, it seems that our posts were overlooked.
Not funny.
lycorn is online now   Reply With Quote
Old 2021-08-10, 13:55   #660
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

23×5×251 Posts
Default

Quote:
Originally Posted by lycorn View Post
Yes, it seems that our posts were overlooked. Not funny.
My sincere apologies for the oversight -- the whole reason I do this is to avoid toe stepping!

Please... Anyone working directly with Primenet, take a look at the updated report, and let me know if GPU is "curating" a range you're working on. More than happy to leave ranges to others to work on their own.
chalsall is online now   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Thinking of Joining GPU to 72 jschwar313 GPU to 72 3 2016-01-31 00:50
Thinking about lasieve5 Batalov Factoring 6 2011-12-27 22:40
Thinking about buying a panda jasong jasong 1 2008-11-11 09:43
Loud thinking on irregular primes devarajkandadai Math 4 2007-07-25 03:01
Question on unfactored numbers... WraithX GMP-ECM 1 2006-03-19 22:16

All times are UTC. The time now is 00:54.


Mon Nov 29 00:54:15 UTC 2021 up 128 days, 19:23, 0 users, load averages: 1.14, 1.07, 1.11

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, 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.