new sieving software - Page 7

geoff · 2007-02-18, 01:56

Quote:

Originally Posted by Flatlander

Guys, help me out here!
I have flu

and not much is making sense. What is the quickest way of sieving 6 k and the same time ie.

uuuu , vvvv , wwww , xxxx , yyyy , zzzz from, say, 270,000 < n < 500,000.
Athlon XP, verbose, save every 10 mins, cache sizes of 128 and 512.
Then feeding the output into LLR.

btw All k have core() <16.

Use srsieve to get the sieve started. if zzzz is your largest k then you need to sieve at least up to p=zzzz, but on 32-bit machines it is often faster to sieve all the way up to 2^32 with srsieve:

Code:

  srsieve -s 10 -v -a -P 4e9 -n 270e3 -N 500e3 "uuuu*2^n-1" "vvvv*2^n-1" ...

The output sieve file will be called sr_2.abcd, copy this to sr2data.txt. Then if you want to sieve up to 1000e9, create a file sr2work.txt containing one line

Code:

4,1000

then run:

Code:

  sr2sieve -v -l128 -L512

This will write new factors to a file called factors4.txt. You can remove the factors from sr_2.abcd with the command

Code:

  srfile -a -k factors4.txt sr_2.abcd

then convert the file to a format PRP or LLR will understand:

Code:

  srfile -G sr_2.abcd

The PRP/LLR input file will be called t17_b2.prp

Note that sr2sieve doesn't actually save the sieve file, it just creates a small checkpoint file every 5 minutes. This is all you need to resume provided you keep a copy of the original sieve file and the found factors.

I know this is complicated. sr2sieve was really designed for distributed sieving where an admisistrator would handle all this stuff and the user just downloads the sr2data.txt and reports back the factorsNNN.txt.

Cruelty · 2007-02-18, 08:18

Quote:

Originally Posted by Flatlander

Athlon XP, verbose, save every 10 mins, cache sizes of 128 and 512.

Actually Athlon XP has 64kB (data) + 64kB (instruction) L1 cache... I'm guessing that "--L1-cache" switch should be set to 64 instead of 128 then... or am I wrong?

geoff · 2007-02-19, 05:06

Quote:

Originally Posted by Cruelty

Actually Athlon XP has 64kB (data) + 64kB (instruction) L1 cache... I'm guessing that "--L1-cache" switch should be set to 64 instead of 128 then... or am I wrong?

That is right, only the size of the data cache that matters here. It only affects the size of the hashtable used, but to get the last drop of performance you can experiment with setting the hashtable size manually via the -H switch.

geoff · 2007-02-27, 20:31

This version should be a bit faster for most sequences (1-8% faster than version 1.0.8 for the few sequences I have tested). Also the process priority is now lowered to idle by default, new command line switched are:

Code:

 -z --idle             Run at idle priority. (Default)
 -Z --no-idle          Don't adjust priority.

VBCurtis · 2007-05-27, 16:32

In case you all don't read other forums, Geoff's sr1 and sr2 sieve programs have gone through a fairly major update, good for 5-10% on most architectures. He built in a quick benchmark for 32-bit code that picks the fastest methods for each CPU, rather than having us guess which program to run on which CPU.

I got about 8% improvement on a TurionX2.

Go get it!
-Curtis

Kosmaj · 2007-06-02, 18:12

Curtis, thanks.

Just to repeat, srsieve is available here

Follow links sr1, or sr2. The latest versions appear to be:
sr1sieve ... 1.1.3 Accepts NewPGen output file.
sr2sieve ... 1.5.5 Requires a file in Riesel.dat format (used by rieselsieve and prothsieve).

Thanks to geoff for his hard work!

monst · 2007-06-02, 20:37

I just downloaded sr1sieve (version 1.1.3) but am having difficulties getting it to recognize a NewPGen file...

Here is the error I get...
ERROR: Line 1: Wrong sieve type in input file `input.txt'.

The first line of input.txt is...
1000000000000:P:0:2:1

This is directly from NewPGen and is the standard k.b^n+1 type.

Any help would be appreciated. Thanks.

VBCurtis · 2007-06-02, 23:28

What is the format of the next line? Is it kkkkk nnnnnnn?

Are there any malformed lines later in the file? I once had a bad HD cache-controller, which corrupted a line or two thousands of lines into the file; it took me quite a while to figure out where to look.

Skim the rest of the file to make sure it is all of that form, with a single k-value; sr1sieve only works on single-k files.
-Curtis

Kosmaj · 2007-06-03, 02:23

1) Make sure it's a fixed k sieve as Curtis said.

2) If it is, but you still have problems then try to change the first line into:
1000000000000:P:1:2:257

monst · 2007-06-03, 16:26

Thanks for the info. I was not using a fixed-k file.

rogue · 2011-06-14, 14:10

If anyone is interested, I have a Win64 build of sr2sieve. It is twice as fast as a Win32 build. You must have a 64-bit version of Windows to run. I had to monkey with the code a little to get it to build, but that was just to address external symbols and none of the binary code. Send me a PM with your e-mail address if you are interested. I could probably do the same with Geoff's other sieving programs, but I'm going to hold off for now. I sent an e-mail to Geoff hoping that he will provide the release himself.

2007-02-27, 20:31	#70
geoff Mar 2003 New Zealand 13·89 Posts	sr1sieve 1.0.11 This version should be a bit faster for most sequences (1-8% faster than version 1.0.8 for the few sequences I have tested). Also the process priority is now lowered to idle by default, new command line switched are: Code: -z --idle Run at idle priority. (Default) -Z --no-idle Don't adjust priority.

2007-06-02, 18:12	#72
Kosmaj Nov 2003 2×1,811 Posts	Curtis, thanks. Just to repeat, srsieve is available here Follow links sr1, or sr2. The latest versions appear to be: sr1sieve ... 1.1.3 Accepts NewPGen output file. sr2sieve ... 1.5.5 Requires a file in Riesel.dat format (used by rieselsieve and prothsieve). Thanks to geoff for his hard work! Last fiddled with by Kosmaj on 2007-06-02 at 18:19

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Generalized Cullen/Woodall Sieving Software	rogue	And now for something completely different	13	2014-12-29 19:11
Suggestion for new sieving software	ATH	Factoring	3	2012-04-04 13:03
Line sieving vs. lattice sieving	JHansen	NFSNET Discussion	9	2010-06-09 19:25
Sieving Software	lavalamp	Software	10	2007-10-20 23:07
Software	TTn	PSearch	0	2004-05-04 13:16

2007-05-27, 16:32	#71
VBCurtis "Curtis" Feb 2005 Riverside, CA 4,861 Posts	In case you all don't read other forums, Geoff's sr1 and sr2 sieve programs have gone through a fairly major update, good for 5-10% on most architectures. He built in a quick benchmark for 32-bit code that picks the fastest methods for each CPU, rather than having us guess which program to run on which CPU. I got about 8% improvement on a TurionX2. Go get it! -Curtis

2007-06-02, 20:37	#73
monst Mar 2007 179 Posts	I just downloaded sr1sieve (version 1.1.3) but am having difficulties getting it to recognize a NewPGen file... Here is the error I get... ERROR: Line 1: Wrong sieve type in input file `input.txt'. The first line of input.txt is... 1000000000000:P:0:2:1 This is directly from NewPGen and is the standard k.b^n+1 type. Any help would be appreciated. Thanks.

2007-06-02, 23:28	#74
VBCurtis "Curtis" Feb 2005 Riverside, CA 11375₈ Posts	What is the format of the next line? Is it kkkkk nnnnnnn? Are there any malformed lines later in the file? I once had a bad HD cache-controller, which corrupted a line or two thousands of lines into the file; it took me quite a while to figure out where to look. Skim the rest of the file to make sure it is all of that form, with a single k-value; sr1sieve only works on single-k files. -Curtis

2007-06-03, 02:23	#75
Kosmaj Nov 2003 111000100110₂ Posts	1) Make sure it's a fixed k sieve as Curtis said. 2) If it is, but you still have problems then try to change the first line into: 1000000000000:P:1:2:257

2007-06-03, 16:26	#76
monst Mar 2007 179 Posts	Thanks for the info. I was not using a fixed-k file.

2011-06-14, 14:10	#77
rogue "Mark" Apr 2003 Between here and the 11×577 Posts	If anyone is interested, I have a Win64 build of sr2sieve. It is twice as fast as a Win32 build. You must have a 64-bit version of Windows to run. I had to monkey with the code a little to get it to build, but that was just to address external symbols and none of the binary code. Send me a PM with your e-mail address if you are interested. I could probably do the same with Geoff's other sieving programs, but I'm going to hold off for now. I sent an e-mail to Geoff hoping that he will provide the release himself.