Sieving drive for k=1005-2000 n=200K-500K

[gd_barnes]

This is a sieving drive for k=1005-2000 and n=200K-500K. If you can spare the machines, we are asking that people dedicate 15-20% of their CPU resources to the effort so that we can complete it in ~3 weeks. I will personally be putting 2 of my newer 64-bit Intel quads on it.

The reason for the priority is because the 5000th-place prime on the top-5000 site is rising rapidly, largely as a result of us finding so many primes on our 1st drive and now the 5th/6th/7th drives. We would like to begin LLRing the effort before the 5000th place prime reaches n=350K, which at the current rate will be in about 3 weeks.

Bruce started the sieve on n=50K-500K several months ago and got it up to P=160G. Ian has accelerted the pace in the last 2 months. Our current sieve depth is P=1400G and Ian is currently working on P=1400G-1800G. I have determined the optimal sieve depth for the n=50K-200K range to be about P=1600G so we will break off that range and not include it in this effort. In the mean time, we need to continue sieving n=200K-500K starting from P=1800G.

As you can guess, even for the smaller range of n=200K-500K, the file is huge. Even in ABCD format, the entire n=50K-500K range is 32 MB and the n=200K-500K range that we will be working on is 21 MB after removing all factors up to P=1400G. Zipped, it should be about 6 MB. I provide a link to the zipped file below. If it is a problem for you to pull it off of the link, let me know and I'll Email it to you.

We will likely sieve to about P=6T (6000G) and then break off n=200K-350K. We'll then need to continue sieving n=350K-500K. A rough estimate for the optimum sieve depth for that final range is P=9T-10T (9000G-10000G).

Ian has been sieving P=200G per core and it has been taking him ~10-11 days using 64-bit sieving. Assuming that you have new cores and a 64-bit O.S., I would suggest taking P=100G ranges per core to start with to guage the sieving speed of your machine. If his machines are a guage, that would take you 5-6 days for each core. For slower/older machines or if you. have a 32-bit O.S., I would suggest half of that.

sr2sieve is what we will use. Let us know if you need the executable or more detailed instructions on using it.

I prefer that we use the more "modern" method of sieving by specifying the ranges and file name at the command prompt as in:

sr2sieve -p 1800e9 -P 2000e9 -i sieve1005-2000-200K-500K.txt

The above would be if you were sieving P=1800G-2000G. The file is listed after the "-i" command and is the actual file name that will be in the zipped file at googlepages. Feel free to name it something shorter if you want or use the "srwork" older convention where you don't have to specify a file name.

When complete, you should have a factors.txt file. Just post the file here in this thread or if it is too big, please Email the file to me at:
gbarnes017 at gmail dot com

Sieve file:
1005-2000-200K-500K [factors removed up to P=3800G]


Reservations:

  P-range     reserved by     Status
n=50K-500K:
   0G- 160G   Brucifer        complete
 160G-1800G   MyDogBuster     complete
 
n=200K-500K:
1800G-3400G   gd_barnes       complete
3400G-3800G   Flatlander      complete
3800G-4000G   MyDogBuster     complete
4000G-4050G   Mini-Geek       complete
4050G-5850G   MyDogBuster     complete
5850G-5900G   mdettweiler     complete
5900G-7100G   gd_barnes       complete
7100G-7200G   Mini-Geek       complete
7200G-8200G   IronBits        complete
8200G-8600G   gd_barnes       complete
8600G-8800G   glennpat        complete
8800G-10000G  gd_barnes       complete
It won't be neccesary to reserve ranges for P>10000G.

After Ian is complete to P=1800G, we will start an 8th drive for n=50K-350K beginning with the range of n=50K-200K. After we have completed all sieving, we will start a 9th drive for n=350K-500K. n=200K-350K will be done after n=50K-200K in the 8th drive.

There's a lot of work to do so let's get crackin'. Once this is done, we'll have a drive where we'll likely be finding a gob of top-5000 primes each day!


Gary

[gd_barnes]

Reserving P=1800G-3400G to put on 8 cores. The sieving speed on my Intel machines is a bit faster than the speed of Ian's sieve due to removing the n=50K-200K range and it being a higher P-range. Sr2sieve gives the ETA for the P=1800G-2000G range on one core as Jan. 5th so I'm loading all 8 of 'em up for 10 days! :-)

2 quads are coming off the 5th drive, which has been good to me so far. Boo-hoo! lol For the first time in 2+ months, I'll have no cores on port 400. 4+ are still on port 5000 though.


Gary

[Flatlander]

Taking 3400-3600G.

Another one for the book:
"Never knowingly undersieved."

I'm running the above on two cores, 32 bit Windows, 3.15GHz. ETA 1st Jan.


This would be an ideal opportunity for me to try out the 64bit Ubuntu I have installed on my C2Duo.
I don't really want to start learning to speak fluent Linux just yet, but perhaps someone could post instructions here for me to get two instances of sr2sieve running? (Or maybe send me a PM.)
The instructions would need to be very specific. It would defeat the object if the PC was idle for hours while I figure things out.

[nuggetprime]

Why are we not using srsieve? For that lot of k's it should be faster.

--nugget

[mdettweiler]

Originally Posted by Flatlander

Taking 3400-3600G.

Another one for the book:
"Never knowingly undersieved."

I'm running the above on two cores, 32 bit Windows, 3.15GHz. ETA 1st Jan.


This would be an ideal opportunity for me to try out the 64bit Ubuntu I have installed on my C2Duo.
I don't really want to start learning to speak fluent Linux just yet, but perhaps someone could post instructions here for me to get two instances of sr2sieve running? (Or maybe send me a PM.)
The instructions would need to be very specific. It would defeat the object if the PC was idle for hours while I figure things out.

Well, sr2sieve under Linux, both for 32-bit and 64-bit, is in fact quite identical in almost every way to the Windows version. First, you need to download the latest 64-bit Linux version of sr2sieve, here. Make a directory somewhere in a convenient place for sr2sieve, and extract the file you just downloaded there. Then, download the sieve file from this thread and put it in that folder too.

Now open a terminal window and navigate to your sr2sieve folder using the "cd" command. This is very similar to cd/chdir in DOS; the only difference is that capitalization DOES matter, and that you should use forward slashes instead of backslashes. For example, if sr2sieve is in the directory "sr2sieve" on your desktop, type the following command:
cd /home/username/Desktop/sr2sieve
Replavce "username" with the username that you use to log in (that is, the lowercase, short one, not the full name that Ubuntu also has you enter when setting up).

Once you've gotten to your sr2sieve directory, you can run sr2sieve pretty much the same way you do under Windows. Th eonly difference is that you have to preface the commands with "./" (without quotes). For example, if under Windows you would have executed the following command:
sr2sieve -p 1800e9 -P 2000e9 -i sieve1005-2000-200K-500K.txt
under Linux you would run the following instead:
./sr2sieve -p 1800e9 -P 2000e9 -i sieve1005-2000-200K-500K.txt

If you'd like to run multiple instances, simply do it whichever way you were doing it under Windows. Both the -u method and running from separate folders will work just as well under either operating system.

Hope this helps! Feel free to drop me a PM if you've got any questions.

Max

[mdettweiler]

Originally Posted by nuggetprime

Why are we not using srsieve? For that lot of k's it should be faster.

--nugget

I think srsieve *used* to be faster for this many k's, but modifications introduced in recent versions of sr2sieve have in fact made sr2sieve faster. Feel free to try it for yourself and compare the p/sec. time to see which one's faster on your machine.

[gd_barnes]

Originally Posted by nuggetprime

Why are we not using srsieve? For that lot of k's it should be faster.

--nugget

No. With the newer versions of sr(x)sieve, sr2sieve is definitely faster.

Originally Posted by mdettweiler

I think srsieve *used* to be faster for this many k's, but modifications introduced in recent versions of sr2sieve have in fact made sr2sieve faster. Feel free to try it for yourself and compare the p/sec. time to see which one's faster on your machine.

Yes, nugget, feel free to give both a try if you want. I think the latest version of srsieve is 0.6.10 and for sr2sieve is 1.7.15. Be sure and use these latest 64-bit versions when comparing.


Gary

[gd_barnes]

Originally Posted by Flatlander

Taking 3400-3600G.

Another one for the book:
"Never knowingly undersieved."

I'm running the above on two cores, 32 bit Windows, 3.15GHz. ETA 1st Jan.


This would be an ideal opportunity for me to try out the 64bit Ubuntu I have installed on my C2Duo.
I don't really want to start learning to speak fluent Linux just yet, but perhaps someone could post instructions here for me to get two instances of sr2sieve running? (Or maybe send me a PM.)
The instructions would need to be very specific. It would defeat the object if the PC was idle for hours while I figure things out.

I'm still pretty much a Linux newb and using Max's instructions several months back was easily able to get 64-bit sr(x)sieve running on my Linux machines.

You just have to have Ubuntu version 8.04 or later installed on your machines to get the increased speed of 64-bit processing. Correct me if I'm wrong there Max.

Come on Max, Carlos, and co., let's get some sieving reservations going here. Ian, can you add another quad to the mix also? It's OK if some of your reservations here are delayed by a couple of weeks.


Gary

[Flatlander]

Taking/running 3600-3800G on two cores in Ubuntu 64bit.


I am getting little/no speed increase over 32bit Windows. Where can I get details of CPU speed/temps etc? How do I assign affinity to each core?

[em99010pepe]

How many k's are we talking about? How many are we double checking?

Carlos

[MyDogBuster]

I'm taking 3800 to 4200G