mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > PrimeNet > GPU to 72

Reply
 
Thread Tools
Old 2012-01-24, 14:36   #1
kladner
 
kladner's Avatar
 
"Kieren"
Jul 2011
In My Own Galaxy!

28×3×13 Posts
Default P-1 discussion

Code:
Sending result to server: UID: flashjh/P1Main, M51879281 completed P-1, B1=505000, B2=10352500, E=12, We4: 3E947A63
[DIGRESS]I don't think I've ever seen anything but the mention of E=12 results. Most of mine do run the B-S extension, but only at E=6. This machine has 16GB RAM, and allocations in proportion to that for P-1, I take it?[/DIGRESS]

Last fiddled with by kladner on 2012-01-24 at 14:37
kladner is online now   Reply With Quote
Old 2012-01-24, 20:47   #2
kladner
 
kladner's Avatar
 
"Kieren"
Jul 2011
In My Own Galaxy!

270016 Posts
Default

Quote:
Originally Posted by Dubslow View Post
I have 12 GB, spread across 2-3 workers depending. It's split around 50/50 between E=6 and E=12.
I guess my allocations are not as high. I just searched results.txt for 'E=12', case sensitive, and found nothing.
kladner is online now   Reply With Quote
Old 2012-01-25, 04:59   #3
flashjh
 
flashjh's Avatar
 
"Jerry"
Nov 2011
Vancouver, WA

1,123 Posts
Default

Quote:
Originally Posted by kladner View Post
Code:
Sending result to server: UID: flashjh/P1Main, M51879281 completed P-1, B1=505000, B2=10352500, E=12, We4: 3E947A63
[DIGRESS]I don't think I've ever seen anything but the mention of E=12 results. Most of mine do run the B-S extension, but only at E=6. This machine has 16GB RAM, and allocations in proportion to that for P-1, I take it?[/DIGRESS]
My main P-1 system has 64Gb of ram. I don't limit the ram each worker can use. Exponents that require Prime95 to process 960 relative primes use around 22Gb of memory to do it all at once.

Last fiddled with by flashjh on 2012-01-25 at 05:00
flashjh is offline   Reply With Quote
Old 2012-01-25, 05:12   #4
kladner
 
kladner's Avatar
 
"Kieren"
Jul 2011
In My Own Galaxy!

28×3×13 Posts
Default

Quote:
Originally Posted by flashjh View Post
My main P-1 system has 64Gb of ram. I don't limit the ram each worker can use. Exponents that require Prime95 to process 960 relative primes use around 22Gb of memory to do it all at once.
Wow! That certainly differs from my circumstances! It gives me a better understanding of how P95 works in the P-1 realm.
kladner is online now   Reply With Quote
Old 2012-01-25, 05:18   #5
Dubslow
Basketry That Evening!
 
Dubslow's Avatar
 
"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

3·2,399 Posts
Default

In my experience, with 'vanilla' P-1 assignments via PrimeNet or GPU272, in stage two there are ~400 relative primes, and for those assignments means about ~10 GiB +- 1 would process all at once.
Code:
[Worker #2 Jan 24 22:58:39] Available memory is 3628MB.
[Worker #2 Jan 24 22:58:39] Using 3620MB of memory.  Processing 143 relative primes (0 of 384 already processed).
[Worker #1 Jan 24 22:58:39] Available memory is 4520MB.
[Worker #1 Jan 24 22:58:39] Using 4480MB of memory.  Processing 41 relative primes (359 of 480 already processed).
These correspond to
Code:
[Worker #1 Jan 24 22:58:21] Optimal P-1 factoring of M200000033 using up to 10000MB of memory.
[Worker #1 Jan 24 22:58:21] Assuming no factors below 2^77 and 2 primality tests saved if a factor is found.
[Worker #1 Jan 24 22:58:21] Optimal bounds are B1=1850000, B2=43937500
[Worker #1 Jan 24 22:58:21] Chance of finding a factor is an estimated 4.47%
(note that's a 200M expo) and
Code:
[Worker #2 Jan 24 22:58:26] Optimal P-1 factoring of M54654059 using up to 10000MB of memory.
[Worker #2 Jan 24 22:58:26] Assuming no factors below 2^72 and 2 primality tests saved if a factor is found.
[Worker #2 Jan 24 22:58:26] Optimal bounds are B1=540000, B2=11610000
[Worker #2 Jan 24 22:58:26] Chance of finding a factor is an estimated 4.12%
For the 'vanilla' assignment, the bounds are a bit lower because of the TF=72, so if you were to reserve one from PrimeNet, you'd probably have > 400 relative primes. (Note for the large P-1 [200M] that the relative prime count is roughly the same [therefore I guess that it's B2/B1] but the memory requirement per relative prime goes up with the B2, which means it goes it with the expo.)

Edit: 64 GiB is a lot of RAM

Last fiddled with by Dubslow on 2012-01-25 at 05:19
Dubslow is offline   Reply With Quote
Old 2012-01-25, 05:29   #6
flashjh
 
flashjh's Avatar
 
"Jerry"
Nov 2011
Vancouver, WA

100011000112 Posts
Default

Quote:
Originally Posted by Dubslow View Post
In my experience, with 'vanilla' P-1 assignments via PrimeNet or GPU272, in stage two there are ~400 relative primes, and for those assignments means about ~10 GiB +- 1 would process all at once.
Code:
[Worker #2 Jan 24 22:58:39] Available memory is 3628MB.
[Worker #2 Jan 24 22:58:39] Using 3620MB of memory.  Processing 143 relative primes (0 of 384 already processed).
[Worker #1 Jan 24 22:58:39] Available memory is 4520MB.
[Worker #1 Jan 24 22:58:39] Using 4480MB of memory.  Processing 41 relative primes (359 of 480 already processed).
These correspond to
Code:
[Worker #1 Jan 24 22:58:21] Optimal P-1 factoring of M200000033 using up to 10000MB of memory.
[Worker #1 Jan 24 22:58:21] Assuming no factors below 2^77 and 2 primality tests saved if a factor is found.
[Worker #1 Jan 24 22:58:21] Optimal bounds are B1=1850000, B2=43937500
[Worker #1 Jan 24 22:58:21] Chance of finding a factor is an estimated 4.47%
(note that's a 200M expo) and
Code:
[Worker #2 Jan 24 22:58:26] Optimal P-1 factoring of M54654059 using up to 10000MB of memory.
[Worker #2 Jan 24 22:58:26] Assuming no factors below 2^72 and 2 primality tests saved if a factor is found.
[Worker #2 Jan 24 22:58:26] Optimal bounds are B1=540000, B2=11610000
[Worker #2 Jan 24 22:58:26] Chance of finding a factor is an estimated 4.12%
For the 'vanilla' assignment, the bounds are a bit lower because of the TF=72, so if you were to reserve one from PrimeNet, you'd probably have > 400 relative primes. (Note for the large P-1 [200M] that the relative prime count is roughly the same [therefore I guess that it's B2/B1] but the memory requirement per relative prime goes up with the B2, which means it goes it with the expo.)

Edit: 64 GiB is a lot of RAM
I've never P-1d anything that high. I've been working in the G272 assignments range, mostly stuff TFd to 72, but some a little lower.
flashjh is offline   Reply With Quote
Old 2012-01-25, 05:32   #7
Dubslow
Basketry That Evening!
 
Dubslow's Avatar
 
"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

160358 Posts
Default

How do you get 900 relative primes then? What's your B2/B1 ratio?
Dubslow is offline   Reply With Quote
Old 2012-01-25, 05:49   #8
flashjh
 
flashjh's Avatar
 
"Jerry"
Nov 2011
Vancouver, WA

1,123 Posts
Default

Quote:
Originally Posted by Dubslow View Post
How do you get 900 relative primes then? What's your B2/B1 ratio?
Here's an example I'm running right now:
Code:
[Jan 24 22:43] Setting affinity to run worker on logical CPU #9
[Jan 24 22:43] Optimal P-1 factoring of M50521529 using up to 58967MB of memory.
[Jan 24 22:43] Assuming no factors below 2^71 and 2 primality tests saved if a factor is found.
[Jan 24 22:43] Optimal bounds are B1=525000, B2=11943750
[Jan 24 22:43] Chance of finding a factor is an estimated 4.53%
[Jan 24 22:43] Setting affinity to run helper thread 1 on logical CPU #10
[Jan 24 22:43] Setting affinity to run helper thread 2 on logical CPU #11
[Jan 24 22:43] Using Core2 type-3 FFT length 2688K, Pass1=896, Pass2=3K, 4 threads
[Jan 24 22:43] Setting affinity to run helper thread 3 on logical CPU #12
[Jan 24 22:44] Available memory is 58967MB.
[Jan 24 22:44] Using 20779MB of memory.  Processing 960 relative primes (0 of 960 already processed).
I have this worker doing all 960 in one pass. This usually doesn't happen because I have MaxHighMemWorkers=5 in my local.txt, so sometimes another worker will finish stage 1 and 'take' some ram from another worker. I have left MaxHighMemWorkers set to 4, but then a lot of ram goes unused from time to time.

Edit1: Like I said, I've never run an exponent in the 200,XXX,XXX range, I may reserve one just to see the numbers.

Edit2: Just to see I reserved a 400,000,000 exponent. I'll post the stage 2 info when the system gets there.

Last fiddled with by flashjh on 2012-01-25 at 06:12
flashjh is offline   Reply With Quote
Old 2012-01-25, 14:39   #9
flashjh
 
flashjh's Avatar
 
"Jerry"
Nov 2011
Vancouver, WA

1,123 Posts
Default

Stage 1, P-1 on my machine for 400000000 exponent is ~6.5 days.
flashjh is offline   Reply With Quote
Old 2012-01-25, 16:04   #10
kladner
 
kladner's Avatar
 
"Kieren"
Jul 2011
In My Own Galaxy!

28·3·13 Posts
Default

Thanks for the examples, Dubslow and Jerry. I had wondered about the number of relative primes, too.

I've been running 4 P95 workers, 3 GB day, 5 GB night. RP's are almost always 192. Typical B1=~520,000 and B2=~10,790,000. Yesterday I changed MaxHighMemWorkers from 3 to 2 to see what difference, if any, that makes. I don't think that change has had time to work through, yet.

EDIT: The above are all on GPU272 assignments which range from the mid 40M's to the mid-to-upper 50M's.

Last fiddled with by kladner on 2012-01-25 at 16:07
kladner is online now   Reply With Quote
Old 2012-01-25, 16:15   #11
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

2·4,643 Posts
Default

Quote:
Originally Posted by flashjh View Post
Here's an example I'm running right now:
Wow!!!

What are the CPU specs on that beast?
chalsall is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Discussion about dates Flatlander Twin Prime Search 12 2011-11-17 09:40
10,375- LA discussion Raman Cunningham Tables 27 2008-12-04 21:17
P-1 discussion AntonVrba Prime Cullen Prime 5 2007-04-04 04:59
factexcl.txt discussion hhh Prime Sierpinski Project 5 2006-11-22 17:50
New .dat discussion VJS Prime Sierpinski Project 7 2006-07-25 14:31

All times are UTC. The time now is 02:43.

Wed Oct 21 02:43:18 UTC 2020 up 40 days, 23:54, 0 users, load averages: 1.70, 1.72, 1.64

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.