mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Open Projects (https://www.mersenneforum.org/forumdisplay.php?f=80)
-   -   Deep Sieving 10m Digit Candidates (https://www.mersenneforum.org/showthread.php?t=10313)

lavalamp 2008-05-19 12:11

Deep Sieving 10m Digit Candidates
 
This is an extension to the [url=http://12121.vocabulate.com/27k/]2721 project[/url], in that it is sieving numbers of the form 27*2^n-1 for the range 33,219,273 <= n <= 33,554,432.

I would like to start a concentrated group effort to sieve 10 million digit candidates. My goal is to make available a relatively large group of candidates that have been sieved very deeply, ideally to 2^52. To get there I need lots of help with the sieving.

I've done a large amount of sieving already, and I have just gone past 150 T, but there's still a lot more to do. This thread will be used as a reservation system so that anyone who wants to help with the sieving can download the sieve file attached to this post, make their intentions known, and begin contributing.


I've created two tables of blocks to reserve, the first table is for blocks of width 5 T, the second is for blocks of width 10 T. For most PCs built in the last 5 years, a 5 T block will take somewhere between 10 - 15 days. Likewise, a 10 T block will take somewhere between 20 - 30 days.

Remember that sr1sieve will only use 1 core, so if you have a dual or quad core CPU and you wish to utilise all cores by sieving, you'll need to run an instance of sr1sieve on each core using the -A0, -A1, -A2 etc. switches.

If you happen to have an Intel quad core CPU, and you are only running three LLR tests due to the memory bottleneck, sieving would be a good way to use the fourth core without impacting the iteration times of the LLR tests.


To reserve a unit just make a post in the thread and I'll add you to the table. Make sure you read the last couple of posts in the thread though just in case someone has already reserved that unit and I haven't added them to the table yet.

To return factors to me please submit them to [url=http://2721.hddkillers.com/factors/]this page[/url]. It will parse the file, extract the factors, check them, and email all of the correct results to me. I don't expect to get any incorrectly reported factors since sr1sieve has a built in check to detect them, however it doesn't hurt to be extra careful. It would REALLY suck if a candidate was removed from the sieve when it was actually prime, not that we'd ever know...

[size=4]5 T blocks:[/size][code] RANGE STATUS SIEVER E.T.A. FACTORS

150 - 155 T completed lavalamp 1st June 2008 9
155 - 160 T completed lavalamp 27th May 2008 17
160 - 165 T completed lavalamp 27th May 2008 9
165 - 170 T completed CedricVonck 28th May 2008 16
170 - 175 T completed lavalamp 7th June 2008 19
175 - 180 T completed lavalamp 7th June 2008 7
180 - 185 T completed lavalamp 16th June 2008 13
185 - 190 T completed Anonymous 26th June 2008 15
190 - 195 T completed lavalamp 25th June 2008 20
195 - 200 T completed lavalamp 25th June 2008 22
200 - 205 T completed lavalamp 1st July 2008 9
205 - 210 T completed lavalamp 6th July 2008 7
210 - 215 T completed lavalamp 6th July 2008 11
215 - 220 T completed lavalamp 16th July 2008 9
220 - 225 T completed lavalamp 16th July 2008 7
225 - 230 T completed lavalamp 16th July 2008 8
230 - 235 T completed lavalamp 26th July 2008 6
235 - 240 T completed lavalamp 26th July 2008 4
240 - 245 T completed lavalamp 31st July 2008 8
245 - 250 T completed lavalamp 26th August 2008 6
250 - 255 T completed lavalamp 16th August 2008 5
255 - 260 T completed CedricVonck 17th August 2008 5
260 - 265 T completed lavalamp 3rd September 2008 8
265 - 270 T completed lavalamp 3rd September 2008 7
270 - 275 T completed lavalamp 5th September 2008 7
275 - 280 T completed lavalamp 21st September 2008 11
280 - 285 T completed lavalamp 22nd September 2008 9
285 - 290 T completed lavalamp 25th September 2008 4
290 - 295 T completed lavalamp 13th October 2008 12
295 - 300 T completed lavalamp 13th October 2008 7
300 - 305 T completed lavalamp 11th October 2008 10
305 - 310 T completed lavalamp 24th October 2008 4
310 - 315 T completed lavalamp 28th October 2008 7
315 - 320 T completed lavalamp 28th October 2008 10
320 - 325 T completed lavalamp 5th November 2008 3
325 - 330 T completed lavalamp 5th December 2008 5
330 - 335 T completed lavalamp 1st December 2008 6
335 - 340 T completed lavalamp 1st December 2008 4
340 - 345 T completed lavalamp 11th December 2008 7
345 - 350 T completed lavalamp 11th December 2008 7[/code]14,115 candidates remain after sieving to 350 T. Closed for now due to lack of interest.

ValerieVonck 2008-05-19 13:53

I will try 165T - 170 T
So if I have dual core ... I have to create 2 directories with the textfiles starting @

- 165T first one
- 168T second one

lavalamp 2008-05-19 14:41

To save you having to set up the sieve as often, you could just reserve two units.

What I do on my dual core machine is have two sieve files and two factor files, both in the same folder. I edit the starting number in the sieve files to the start of the range (so that sr1sieve thinks it's already been sieved to that point and carries on from there), and start the sieves with the following commands:[code]sr1sieve -A0 -s 5 -i "sieve1.txt" -o "sieve1.txt" -f "factors_27_160.txt" -P 160000000000000

sr1sieve -A1 -s 5 -i "sieve2.txt" -o "sieve2.txt" -f "factors_27_165.txt" -P 165000000000000[/code]And every 5 mins sr1sieve updates the first line in the sieve files with it's progress, so that if I ever restart the PC or whatever, it can carry on from where it left off.

I'd recommend putting those two commands into two batch files in the same folder as the sieve, then placing short cuts to them in your start menu folder so that it auto runs on start up.

When I started the sieve, the first lines looked like this:[code]155000000000000:M:1:2:258

160000000000000:M:1:2:258[/code]But now they look like this:[code]156334444954291:M:1:2:258

161396494308947:M:1:2:258[/code]I also installed a program called [url=http://www.iconictray.com/]Iconic Tray[/url] that allows me to minimise the windows to the system tray, since otherwise the DOS boxes take up taskbar space.

Also, I'll have to edit you into the table in a bit, I don't have the ability to edit my post, but I have asked an admin very nicely if I could be able to do that.

ValerieVonck 2008-05-19 16:14

Thanx!

I'll start asap when my composites for William are crunched.

mdettweiler 2008-05-19 16:28

lavalamp - you may want to consider asking Xyzzy for mod privileges in the Open Projects forum so you can edit the range status table in the first post of this thread.

In other news, I'm hoping to be able to help out a little here sometime soon, stay tuned. :smile:

geoff 2008-05-19 22:15

[QUOTE=lavalamp;133655]Remember that sr1sieve will only use 1 core, so if you have a dual or quad core CPU and you wish to utilise all cores by sieving, you'll need to run an instance of sr1sieve on each core using the -A0, -A1, -A2 etc. switches.[/QUOTE]

Linux (and maybe Mac?) users can add the number of cores to use to the command line with the -t switch. e.g. `sr1sieve -t4 ...' will use 4 cores. This is no faster than running 4 seperate instances, just more convenient.

mdettweiler 2008-05-20 03:13

[quote=geoff;133712]Linux (and maybe Mac?) users can add the number of cores to use to the command line with the -t switch. e.g. `sr1sieve -t4 ...' will use 4 cores. This is no faster than running 4 seperate instances, just more convenient.[/quote]
In fact, I think it's usually a little slower than running separate instances, right?

ValerieVonck 2008-05-20 18:41

Already 1 factor found: 168187471653191 | 27*2^33529390-1

ValerieVonck 2008-05-21 15:52

Total of 5 factors found & uploaded

lavalamp 2008-05-21 16:00

I appreciate your enthusiasm, but take it easy dude. You only need to upload them when it's done.

lavalamp 2008-05-27 07:38

Taking 170 - 175 and 175 - 180.


All times are UTC. The time now is 01:10.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.