mersenneforum.org  

Go Back   mersenneforum.org > Factoring Projects > Msieve

Reply
 
Thread Tools
Old 2016-03-14, 13:32   #12
prss
 
prss's Avatar
 
Jun 2011
Brazil

2·11 Posts
Default

Right jasonp, but how i choose A and B for spend less time in search?
Has some formula to choose A and B in function of N ?
prss is offline   Reply With Quote
Old 2016-03-14, 13:56   #13
jasonp
Tribal Bullet
 
jasonp's Avatar
 
Oct 2004

354110 Posts
Default

Not really; all the guidance on choosing the range is pretty anecdotal.

A small range finishes quickly, and small A and B find polynomials that are very highly skewed, which maximizes the search space for stage 2 to find a good polynomial. If you only have a little time to search, A=1 is a good idea; but the best polynomials have very good size and very good root properties, whereas limiting the A-B range to small values will find excellent root properties but will not do much to find a polynomial with good size.

Polynomial selection is a process where the only way to find the best results is to spend a lot of time looking. Since we often do not have a lot of time, there is a tradeoff we all have to make based on how long the sieving will take and how good the results found so far have been. If you have an Amazon compute farm ready to do the sieving you do not need a perfect polynomial, and if you are factoring very large integers you can spend weeks looking for a polynomial.
jasonp is offline   Reply With Quote
Old 2016-03-14, 15:37   #14
Dubslow
Basketry That Evening!
 
Dubslow's Avatar
 
"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

3×29×83 Posts
Default

Quote:
Originally Posted by prss View Post
Right jasonp, but how i choose A and B for spend less time in search?
Has some formula to choose A and B in function of N ?
If you have the answer, you should definitely consider publishing it in a peer reviewed journal. (Read: as jason implies, there is no real good answer. NFS is as much more so guess work, smoke and mirrors than a proper A to B algorithm.)
Dubslow is offline   Reply With Quote
Old 2016-03-14, 20:29   #15
prss
 
prss's Avatar
 
Jun 2011
Brazil

2·11 Posts
Default

As I'm not have a Amazon cluster with 64GB+ of memory, i might consider to sieving and execute linear algebra phase storing data in a SSD disk, solving matrix in parts?
prss is offline   Reply With Quote
Old 2016-03-16, 03:05   #16
prss
 
prss's Avatar
 
Jun 2011
Brazil

2·11 Posts
Default

Quote:
Originally Posted by jasonp:
Sort the lines in the file and run the root optimization part of stage 2 (with -npr) only on the lines of the .ms file whose last field is the smallest, maybe the smallest 100 or so entries.
By which term I need to order? The line have that format:
12 120274200 -1925450320035100 -15282382788549599 128117774098336313505596 64799414552419869488988 357477963239 -452243407311138901133651 -0.74 7.115460e+015
prss is offline   Reply With Quote
Old 2016-03-16, 04:36   #17
Dubslow
Basketry That Evening!
 
Dubslow's Avatar
 
"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

3·29·83 Posts
Default

He said "last field", so 7.115460e+015.
Dubslow is offline   Reply With Quote
Old 2016-03-20, 14:45   #18
prss
 
prss's Avatar
 
Jun 2011
Brazil

2×11 Posts
Default

Sorry Dubslow, it was my lack of attention.
Thanks by clarifications.

I have the following GPU:
Code:
[ CUDA: GeForce GT 630M ]
 
Device Properties:
Device Name						GeForce GT 630M
PCI Domain / Bus / Device		0 / 1 / 0
Clock Rate						950 MHz
Asynchronous Engines			1
Multiprocessors / Cores			2 / 96
L2 Cache						128 KB
Max Threads Per Multiprocessor	1536
Max Threads Per Block			1024
Max Registers Per Block			32768
Max Instructions Per Kernel		512 million
Warp Size						32 threads
Max Block Size					1024 x 1024 x 64
Max Grid Size					65535 x 65535 x 65535
How many threads result in better speedup in polynomial selection using msieve?
prss is offline   Reply With Quote
Old 2016-03-22, 11:26   #19
prss
 
prss's Avatar
 
Jun 2011
Brazil

101102 Posts
Default

After of some tests i see that greater speedup happen when use number of threads equal to Warp Size (32 in my case).

My GPU with 96 cuda cores was 5x faster than core i5 2,5 Ghz.
prss is offline   Reply With Quote
Old 2016-03-22, 12:47   #20
jasonp
Tribal Bullet
 
jasonp's Avatar
 
Oct 2004

3,541 Posts
Default

The number of CPU threads should have nothing to do with the size of a CUDA warp, and in fact the latest SVN limits the number of GPU stage 1 threads to 4.

To repeat, this all becomes a moot point when you are not solving very small problems. Once you start factoring 140+ digit integers, forget about the number of threads.
jasonp is offline   Reply With Quote
Old 2016-08-20, 07:48   #21
cgy606
 
Feb 2012

32·7 Posts
Default

I'm new to using msieve by itself (I use yafu). I would like to use msieve for polynomial selection. I run the following command on a C141:

echo "(134!-5)/2844729212424914765373584725399627828496565276362417868278840579676885183018079137082195" | msieve152.exe -g 0 -t 2 -v -np1

I am using a windows 10 machine with a nvidia graphics card and I keep getting the error: "The programs can't start because pthreadGC2.dll" is missing". Does anybody know how to fix this?
cgy606 is offline   Reply With Quote
Old 2016-08-20, 11:58   #22
xilman
Bamboozled!
 
xilman's Avatar
 
"π’‰Ίπ’ŒŒπ’‡·π’†·π’€­"
May 2003
Down not across

10,753 Posts
Default

Quote:
Originally Posted by cgy606 View Post
I'm new to using msieve by itself (I use yafu). I would like to use msieve for polynomial selection. I run the following command on a C141:

echo "(134!-5)/2844729212424914765373584725399627828496565276362417868278840579676885183018079137082195" | msieve152.exe -g 0 -t 2 -v -np1

I am using a windows 10 machine with a nvidia graphics card and I keep getting the error: "The programs can't start because pthreadGC2.dll" is missing". Does anybody know how to fix this?
Yes. Even though I don't do Windoze, I can read an error message.

You need to track down a copy of the file "pthreadGC2.dll" and install it in the correct place. I suggest that you follow this link, as I did, and examine the results.

https://duckduckgo.com/?t=ouk&q=pthreadGC2.dll&ia=web



(Though I prefer duckduckgo to Google for privacy reasons.)

Last fiddled with by xilman on 2016-08-20 at 12:00 Reason: Fix minor tyop.
xilman is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
error when running msieve 1.53 with cuda aein Msieve 9 2019-02-25 14:09
Can anyone help me about msieve 1.53 with CUDA? Seto Msieve 8 2016-09-24 12:54
Problem in running msieve with CUDA mohamed Msieve 20 2013-08-01 08:27
Help with msieve CUDA (C1060) toorandom Msieve 7 2012-11-28 17:37
CUDA_ERROR_LAUNCH_OUT_OF_RESOURCES when running msieve 1.5.0 with CUDA ryanp Msieve 3 2012-06-12 03:27

All times are UTC. The time now is 00:46.


Sat Jul 17 00:46:29 UTC 2021 up 49 days, 22:33, 1 user, load averages: 1.49, 1.42, 1.35

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.