mersenneforum.org  

Go Back   mersenneforum.org > New To GIMPS? Start Here! > Homework Help

Reply
 
Thread Tools
Old 2011-06-28, 06:53   #1
JimboPrimer
 

486410 Posts
Question Dual Core to process single work unit?

I have a Core 2 Duo MacBook. In the past I have been automatically given 2 work units assignment by GIMPS being worked on by Worker1 and Worker2.

I have changed my preference to work on the LL100M+ and was given one work unit. Accordingly, I changed my worker configuration as follows:

Worker#1 -- Whatever makes the most sense -- Smart assignment
Worker#2 -- 100,000,000 digit numbers to test -- Smart assignment

My questions:
1) Is it possible to make both workers process a single work unit and not two separate Mersenne number candidates?
2) If so, what settings do I change?
3) Just a thought: Would changing Worker#1 setting to "100,000,000 digit numbers to test" do the trick? Or will it be automatically assigned another number to test? I'd prefer testing only one because I don't want my computer to be overwhelmed.

Thanks for any help
  Reply With Quote
Old 2011-06-28, 15:33   #2
petrw1
1976 Toyota Corona years forever!
 
petrw1's Avatar
 
"Wayne"
Nov 2006
Saskatchewan, Canada

5·953 Posts
Default

Quote:
Originally Posted by JimboPrimer View Post
I have a Core 2 Duo MacBook. In the past I have been automatically given 2 work units assignment by GIMPS being worked on by Worker1 and Worker2.

I have changed my preference to work on the LL100M+ and was given one work unit. Accordingly, I changed my worker configuration as follows:

Worker#1 -- Whatever makes the most sense -- Smart assignment
Worker#2 -- 100,000,000 digit numbers to test -- Smart assignment

My questions:
1) Is it possible to make both workers process a single work unit and not two separate Mersenne number candidates?
2) If so, what settings do I change?
3) Just a thought: Would changing Worker#1 setting to "100,000,000 digit numbers to test" do the trick? Or will it be automatically assigned another number to test? I'd prefer testing only one because I don't want my computer to be overwhelmed.

Thanks for any help
For off I just want you to be aware that your PC will take well over one year to complete a single 100,000,000 test even is you assign both cores to one assignment

1) Yes

2) I can only speak for the WINDOWS GUI ... not sure if MAC menus are different:
- Test ... Stop (All workers)
- Test ... Worker Windows
--- Number of workers to run: 1
--- Worker Number: All workers OR Worker #1
--- Type of work to get: 100,000,000 Digit numbers to test
--- CPU Affinity: Smart Assignment
--- CPUs to use: 2
--- OK
- Test ... Start

3) If all you do is change the work type that worker will continue to completion whatever is currently assigned and then start the new work type.

If Worker #2 already has a 100,000,000 in progress and you want to move it to Worker #1
- Test ... Stop (All workers)
- Test ... Exit
- Find and edit your worktodo.txt file (or whatever MAC names it)
- physically cut and paste the appropriate lines from below Worker #2 to below Worker #1
- Save it
- Execute Prime95

You say you "...don't want my computer to be overwhelmed..."
Whether you have each core processing their own test or both cores on one your PC's CPU will be virtually 100% busy all the time and without good cooling and ventilation could get quite hot. The vast majority of us run all our PCs 100% all the time and rarely have any issues. There is a bigger risk if your PC's vents get clogged with dust OR if you OverClock a lot. If you are concerned about overwhelming your PC you can set up to have 1 worker window with 1 CPU realizing that your PC will then be 50% utilized by Prime95 and all tests will take twice as long of you will complete half as many.

If you want to abandon whatever Worker #1 is currently working on so it can immediately continue the 100,000,000 test. Find the exponent of that assignment
- Test ... Status
- Note the exponent number only (not the preceeding M)
- Advanced ... Unreserve Exponent

OR do it via the Server noting that your PC won't actually stop that assignment until the next scheduled check in date/time OR until you:
- Advanced ... Manual Communication
- Send new end dates
petrw1 is offline   Reply With Quote
Old 2011-06-30, 01:39   #3
James Heinrich
 
James Heinrich's Avatar
 
"James Heinrich"
May 2004
ex-Northern Ontario

2·17·103 Posts
Default

Quote:
Originally Posted by petrw1 View Post
For off I just want you to be aware that your PC will take well over one year to complete a single 100,000,000 test even is you assign both cores to one assignment
Adding more specific numbers: A 100M-digit test is roughly 5000GHz-days of work. A moderately fast Core2 Duo notebook CPU (e.g. P8700) will complete around 3GHz-days/day of work. Assuming it does nothing else (but it presumably will), runs 24/7, doesn't overheat (quite likely with a notebook), and you get 100% efficiency spreading across both cores (you won't), it should take around 2.3 years to complete a single test. Remembering that's a best-case scenario (and 4+ years is realistic on that hardware) you'd probably be better off sticking with "whatever makes sense", which will likely be "normal" L-L tests. Double-checking L-L worktype would also be a good choice.
James Heinrich is online now   Reply With Quote
Old 2011-07-03, 07:50   #4
davieddy
 
davieddy's Avatar
 
"Lucan"
Dec 2006
England

2·3·13·83 Posts
Default More cold water...

A 100M digit test is ~ 10x less likely to find a prime that a 10M digit one.
davieddy is offline   Reply With Quote
Old 2011-07-03, 10:51   #5
James Heinrich
 
James Heinrich's Avatar
 
"James Heinrich"
May 2004
ex-Northern Ontario

2×17×103 Posts
Default

Quote:
Originally Posted by davieddy View Post
A 100M digit test is ~ 10x less likely to find a prime that a 10M digit one.
I have to disagree with you there. Working in that range is less likely to find a prime due to longer work times, therefore fewer exponents processed, therefore lower cumulative chance of finding a prime; but wouldn't each exponent have a similar (very low) individual chance of being prime?
James Heinrich is online now   Reply With Quote
Old 2011-07-03, 13:02   #6
Christenson
 
Christenson's Avatar
 
Dec 2010
Monticello

5×359 Posts
Default

Quote:
Originally Posted by James Heinrich View Post
I have to disagree with you there. Working in that range is less likely to find a prime due to longer work times, therefore fewer exponents processed, therefore lower cumulative chance of finding a prime; but wouldn't each exponent have a similar (very low) individual chance of being prime?
We need to get precise as to what we mean by "chance of finding a prime". There are, as you note, two factors at work.

First, if you do a 100M digit test, it will take much (36x) longer than the current M(50M) LL tests, and that has a significant opportunity cost in fewer factors that can be checked. (That is, if the OP even has enough patience and persistence to complete the one test, no surprise or blame intended if not, you are an absolute saint if you do -- that probability is much less than unity)

Second, I don't think anyone expects Mersenne Primes to get any more common as the exponents get bigger. I would expect them to get less common, on the basis that the probability for a "random" number of the larger size to be prime decreases. But I don't think this is by a factor of 5 or 10, or even 2 -- I'd have to go look up the formula for the number of primes below a given point, and I think we'll get something like 22/26 or so.
Christenson is offline   Reply With Quote
Old 2011-07-03, 13:42   #7
axn
 
axn's Avatar
 
Jun 2003

10100000110012 Posts
Default

Quote:
Originally Posted by James Heinrich View Post
but wouldn't each exponent have a similar (very low) individual chance of being prime?
Quote:
Originally Posted by Christenson View Post
I would expect them to get less common, on the basis that the probability for a "random" number of the larger size to be prime decreases. But I don't think this is by a factor of 5 or 10, or even 2 -- I'd have to go look up the formula for the number of primes below a given point, and I think we'll get something like 22/26 or so.
No. A 10M digit number is 10x more likely than a 100M digit number to be prime (all else being the same). I think you have taken one log too many to come up with 22/26.
axn is offline   Reply With Quote
Old 2011-07-03, 13:58   #8
Christenson
 
Christenson's Avatar
 
Dec 2010
Monticello

5·359 Posts
Default

Indeed I have taken one log too many...LL-Ds deal with 10M digit #s, LLs deal with 20M digit #s, and a 100M digit test deals with a 100M digit #.
Christenson is offline   Reply With Quote
Old 2011-07-04, 03:19   #9
davieddy
 
davieddy's Avatar
 
"Lucan"
Dec 2006
England

2×3×13×83 Posts
Default

Quote:
Originally Posted by Christenson View Post
Indeed I have taken one log too many...LL-Ds deal with 10M digit #s, LLs deal with 20M digit #s, and a 100M digit test deals with a 100M digit #.
Can't resist:
I caught WBLipp out with this last year.
I asserted that a 100M digit test was 10x less likely
to find a prime than a 10M one.
"No. The probability goes as 1/logN. It's called the "Prime Number Theorem".

I just grinned.

Confusing as the terminology is, the "exponent" is
log2 (the Mersenne number).
# digits is log10 (the Mersenne number)

Last fiddled with by davieddy on 2011-07-04 at 03:41
davieddy is offline   Reply With Quote
Old 2011-07-04, 03:58   #10
davieddy
 
davieddy's Avatar
 
"Lucan"
Dec 2006
England

2·3·13·83 Posts
Default

Quote:
Originally Posted by axn View Post
No. A 10M digit number is 10x more likely than a 100M digit number to be prime (all else being the same). I think you have taken one log too many to come up with 22/26.
To look at it another way, Wagstaff expects 1 MP with exponent
between 33M and 50M, and one with exponent between 330M and 500M.

It is the density of prime exponents which doesn't differ much (exercise*)
so we expect to have to do nearly 10x as many LL tests.

David

* I reckon 7/8 th

Last fiddled with by davieddy on 2011-07-04 at 04:45
davieddy is offline   Reply With Quote
Old 2011-07-30, 12:53   #11
joblack
 
joblack's Avatar
 
Oct 2008
n00bville

10110110002 Posts
Default

Quote:
Originally Posted by davieddy View Post
Can't resist:
I caught WBLipp out with this last year.
I asserted that a 100M digit test was 10x less likely
to find a prime than a 10M one.
"No. The probability goes as 1/logN. It's called the "Prime Number Theorem".

I just grinned.

Confusing as the terminology is, the "exponent" is
log2 (the Mersenne number).
# digits is log10 (the Mersenne number)
It's not less likely to find a Mersenne prime in the 100M area. You only have to invest more time to be certain.
joblack is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Single vs Dual channel memory TObject Hardware 5 2014-12-24 05:58
Dual Core to Quad Core Upgrade Rodrigo Hardware 6 2010-11-29 18:48
Why factoring is single-core designed? otutusaus Software 33 2010-11-20 21:05
exclude single core from quad core cpu for gimps jippie Information & Answers 7 2009-12-14 22:04
Importance of dual channel memory for dual core processors patrik Hardware 3 2007-01-07 09:26

All times are UTC. The time now is 18:46.


Sun Oct 17 18:46:19 UTC 2021 up 86 days, 13:15, 0 users, load averages: 1.31, 1.49, 1.40

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.