srsieve/sr2sieve enhancements - Page 8

rogue · 2013-03-30, 00:40

Quote:

Originally Posted by henryzz

I don't think multithreading is possible on windows.
From memory it uses pthreads which I think is officially linux only. It does seem to be possible to build on win32 now and there seem to be claims of having it working for 64-bit. You would have to compile sr1sieve yourself to take advantage of this on windows.

I build with mingw, not cygwin. I believe that pthreads exist in cygwin, but not mingw.

henryzz · 2013-03-30, 01:20

Quote:

Originally Posted by rogue

I build with mingw, not cygwin. I believe that pthreads exist in cygwin, but not mingw.

http://sourceforge.net/apps/trac/min...ile%20pthreads
This is the link I came across. That is mingw-w64.

pepi37 · 2013-03-30, 11:10

To simulate multithreads option can I on my I7 ( HT on) open 8 cmd windows and in each one put sr2sieve -i sr_10.abcd -pxx -Pxxx
and than distribute range xx-xxx and after process is over join all factors and then remove it from abcd file?

henryzz · 2013-03-30, 11:28

Quote:

Originally Posted by pepi37

To simulate multithreads option can I on my I7 ( HT on) open 8 cmd windows and in each one put sr2sieve -i sr_10.abcd -pxx -Pxxx
and than distribute range xx-xxx and after process is over join all factors and then remove it from abcd file?

That is the standard method. srfile is used to remove the factors in factors.txt.

pepi37 · 2013-03-30, 11:46

WOW :)
You can tell me whatever you wont, but this is discovery for me: with 12 cores ( 4+4, and 4) on second CPU I am capable of doing much deeper sieving in fraction of time :) I was "afraid" how will ABCD file tolerate to be accesed from few process in same time, but my fear is / was unnecessary
And much deeper sieve is less llr-ing :)
P.S It looks like that in this method must be same number of directories as cores: if you run if from same directory only last sieve process write checkpoint:and that is not practical in case of lets say power failure...

rogue · 2013-03-30, 14:41

Quote:

Originally Posted by pepi37

To simulate multithreads option can I on my I7 ( HT on) open 8 cmd windows and in each one put sr2sieve -i sr_10.abcd -pxx -Pxxx
and than distribute range xx-xxx and after process is over join all factors and then remove it from abcd file?

I use Console2. One window, eight tabs with a command prompt.

Puzzle-Peter · 2013-03-30, 16:12

Running multiple instances in one directory can result in the messing up of checkpoint files because all instances try to write to the same checkpoint file. At least under Linux that's the case. Don't know about windows. That's why I run every instance in a seperate directory.

I should add that I stop and restart very often because I need the machines for work during the day. If you finish in one go it's probably no problem.

pepi37 · 2013-03-30, 16:43

Quote:

Originally Posted by rogue

I use Console2. One window, eight tabs with a command prompt.

Great tip- very useful info! I immidiatelly start to use it on LLR :)

pepi37 · 2013-03-30, 16:44

Quote:

Originally Posted by Puzzle-Peter

Running multiple instances in one directory can result in the messing up of checkpoint files because all instances try to write to the same checkpoint file. At least under Linux that's the case. Don't know about windows. That's why I run every instance in a seperate directory.

I should add that I stop and restart very often because I need the machines for work during the day. If you finish in one go it's probably no problem.

I noticed this, but was small test sieve.
No I was running from different directories :)

f1pokerspeed · 2013-03-30, 19:12

I wish there was a way that I could automatically run the sieve with a batch file that does all of the ranges for you.

An algorithm something like this:

○ Console opens
○ Input range increment, start, end, number of cores to run on
○ Split sieve work
○ Start instances of sieve

○ Read from file in arbitrary location via temp file?

○ Output to console with state of combined sieve every 60 sec.

Something like this could work quite well, I think. Would this be difficult to program, or is it a trivial task? The instances could be started via a customized batch file or something like that.

Thoughts?

pepi37 · 2013-03-30, 22:09

Can someone explain better what is this options for?

-B --baby METHOD Use METHOD for baby step mulmods.
-G --giant METHOD Use METHOD for giant step mulmods.
--ladder METHOD Use METHOD for ladder mulmods.

Is any of those methods faster or slower ( and in what cases)?

2013-03-30, 11:46	#82
pepi37 Dec 2011 After 1.58M nines:) 1,699 Posts	WOW :) You can tell me whatever you wont, but this is discovery for me: with 12 cores ( 4+4, and 4) on second CPU I am capable of doing much deeper sieving in fraction of time :) I was "afraid" how will ABCD file tolerate to be accesed from few process in same time, but my fear is / was unnecessary And much deeper sieve is less llr-ing :) P.S It looks like that in this method must be same number of directories as cores: if you run if from same directory only last sieve process write checkpoint:and that is not practical in case of lets say power failure... Last fiddled with by pepi37 on 2013-03-30 at 12:11

2013-03-30, 16:12	#84
Puzzle-Peter Jun 2009 2²×5²×7 Posts	Running multiple instances in one directory can result in the messing up of checkpoint files because all instances try to write to the same checkpoint file. At least under Linux that's the case. Don't know about windows. That's why I run every instance in a seperate directory. I should add that I stop and restart very often because I need the machines for work during the day. If you finish in one go it's probably no problem. Last fiddled with by Puzzle-Peter on 2013-03-30 at 16:14

2013-03-30, 19:12	#87
f1pokerspeed Jun 2012 2·53 Posts	I wish there was a way that I could automatically run the sieve with a batch file that does all of the ranges for you. An algorithm something like this: ○ Console opens ○ Input range increment, start, end, number of cores to run on ○ Split sieve work ○ Start instances of sieve ○ Read from file in arbitrary location via temp file? ○ Output to console with state of combined sieve every 60 sec. Something like this could work quite well, I think. Would this be difficult to program, or is it a trivial task? The instances could be started via a customized batch file or something like that. Thoughts?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Sieving twins with srsieve	henryzz	Twin Prime Search	0	2014-03-18 12:44
Intel announces multi-core enhancements for Haswell chips	ixfd64	Hardware	8	2012-02-10 20:32
LLRnet enhancements	kar_bon	No Prime Left Behind	10	2008-03-28 11:21
TODO list and suggestions/comments/enhancements	Greenbank	Octoproth Search	2	2006-12-03 17:28
Suggestions for future enhancements	Reboot It	Software	16	2003-10-17 01:31

2013-03-30, 11:10	#80
pepi37 Dec 2011 After 1.58M nines:) 1,699 Posts	To simulate multithreads option can I on my I7 ( HT on) open 8 cmd windows and in each one put sr2sieve -i sr_10.abcd -pxx -Pxxx and than distribute range xx-xxx and after process is over join all factors and then remove it from abcd file?

2013-03-30, 22:09	#88
pepi37 Dec 2011 After 1.58M nines:) 1699₁₀ Posts	Can someone explain better what is this options for? -B --baby METHOD Use METHOD for baby step mulmods. -G --giant METHOD Use METHOD for giant step mulmods. --ladder METHOD Use METHOD for ladder mulmods. Is any of those methods faster or slower ( and in what cases)?