3,637- factored, NSF grant renewed

log in

Advanced search

Message boards : News : 3,637- factored, NSF grant renewed

Author Message
Greg
Project administrator
Send message
Joined: 26 Jun 08
Posts: 582
Credit: 223,912,432
RAC: 17,546
Message 994 - Posted: 23 Sep 2012, 4:48:15 UTC

3,637- has been factored into 85-digit and 127-digit prime numbers. In other news, the National Science Foundation XSEDE grant has been renewed for another year, so we have lots of sieving to do! We can use as much computing time as you have available to donate, and your help is greatly appreciated!

Profile debrouxl
Send message
Joined: 9 Oct 11
Posts: 3
Credit: 2,012,906
RAC: 0
Message 995 - Posted: 23 Sep 2012, 7:48:43 UTC - in response to Message 994.

It's pretty good news that the grant was renewed :)

Profile Carlos Pinho [TSBTs Pirate]
Volunteer moderator
Send message
Joined: 26 Sep 09
Posts: 162
Credit: 7,723,521
RAC: 0
Message 996 - Posted: 23 Sep 2012, 9:22:48 UTC - in response to Message 995.

It's pretty good news that the grant was renewed :)


Agree.

Carlos

Speedy51
Send message
Joined: 16 Oct 09
Posts: 44
Credit: 293,342
RAC: 88
Message 999 - Posted: 27 Sep 2012, 5:18:25 UTC

Great to hear. Dose the grant give you x amount of computer time post processing?

Greg
Project administrator
Send message
Joined: 26 Jun 08
Posts: 582
Credit: 223,912,432
RAC: 17,546
Message 1000 - Posted: 27 Sep 2012, 8:13:01 UTC - in response to Message 999.

Exactly. For this year, we receive a little over 1 million CPU-hours for postprocessing.

Speedy51
Send message
Joined: 16 Oct 09
Posts: 44
Credit: 293,342
RAC: 88
Message 1004 - Posted: 27 Sep 2012, 22:12:09 UTC

Gr8 news. That's just over 167 years of CPU time. Out of interest how much work is done in an hour of postprocessing time?

tuankiet65
Send message
Joined: 25 Apr 12
Posts: 9
Credit: 112,044
RAC: 0
Message 1005 - Posted: 28 Sep 2012, 5:21:10 UTC - in response to Message 1004.

based on my calculation, TAOCP need 11114,893 hours to do the post-processing on 1 Core 2 Quad core

tuankiet65
Send message
Joined: 25 Apr 12
Posts: 9
Credit: 112,044
RAC: 0
Message 1006 - Posted: 28 Sep 2012, 10:03:39 UTC - in response to Message 1000.

Why do you have the 1000th comment Greg?

Greg
Project administrator
Send message
Joined: 26 Jun 08
Posts: 582
Credit: 223,912,432
RAC: 17,546
Message 1007 - Posted: 29 Sep 2012, 0:19:16 UTC - in response to Message 1006.
Last modified: 29 Sep 2012, 0:20:31 UTC

The amount of time required for postprocessing depends on the number. The recently completed largest number so far, 2,1061+, required about 340,000 core-hours to postprocess. I ran the code using 576 cores, so it required about 3.5 weeks of actual calculation time to complete.

I got the 1000th post simply by luck! That was not planned. :)

Speedy51
Send message
Joined: 16 Oct 09
Posts: 44
Credit: 293,342
RAC: 88
Message 1008 - Posted: 29 Sep 2012, 9:29:21 UTC

Thanks Greg. Man post processing takes along time. How many cores do you have to run on one job & whats the most core you've used at one time?

Profile Carlos Pinho [TSBTs Pirate]
Volunteer moderator
Send message
Joined: 26 Sep 09
Posts: 162
Credit: 7,723,521
RAC: 0
Message 1009 - Posted: 29 Sep 2012, 9:46:19 UTC - in response to Message 1008.
Last modified: 29 Sep 2012, 9:47:12 UTC

Thanks Greg. Man post processing takes along time. How many cores do you have to run on one job & whats the most core you've used at one time?


Here you have my timings for lasieved numbers for post-processing using a core i5 750@3.4 GHz:



Speedy51
Send message
Joined: 16 Oct 09
Posts: 44
Credit: 293,342
RAC: 88
Message 1010 - Posted: 29 Sep 2012, 11:11:39 UTC

Thanks for the chart.

Profile Carlos Pinho [TSBTs Pirate]
Volunteer moderator
Send message
Joined: 26 Sep 09
Posts: 162
Credit: 7,723,521
RAC: 0
Message 1011 - Posted: 29 Sep 2012, 12:00:25 UTC

Please don't take into consideration last column, named Total Time. I was just trying to figure how much BOINC points were equivalent for the sum of all post-processing phases. All numbers are in seconds.

Example, post processing of 5_492_plus1 took:

Filtering - 1 core - 3.300 seconds
Linear Algebra - 4 cores - 258.780 seconds
Square Root - 1 core - 1.234 seconds

So overall it took me 263.314 seconds (~73 hours) on my machine to do the post-processing of 5_492_plus1.

Greg
Project administrator
Send message
Joined: 26 Jun 08
Posts: 582
Credit: 223,912,432
RAC: 17,546
Message 1012 - Posted: 29 Sep 2012, 17:28:19 UTC - in response to Message 1008.

How many cores do you have to run on one job & whats the most core you've used at one time?

For the large numbers, I use the TACC Lonestar and SDSC Trestles clusters. Lonestar has over 20,000 cores and Trestles has over 10,000 cores. However, as you use more cores the calculation becomes less efficient. The largest I have run is 900 cores, but I typically use 576 cores in a 24 x 24 grid.

Speedy51
Send message
Joined: 16 Oct 09
Posts: 44
Credit: 293,342
RAC: 88
Message 1016 - Posted: 1 Oct 2012, 6:51:00 UTC

Thanks for the great answers. Greg have you been tempted to Sieve a job on 576? I bet it would be done fast.

Greg
Project administrator
Send message
Joined: 26 Jun 08
Posts: 582
Credit: 223,912,432
RAC: 17,546
Message 1017 - Posted: 1 Oct 2012, 8:33:46 UTC - in response to Message 1016.

Right now, NFS@Home has the rough equivalent of ~1000 cores running full time, so adding 576 cores would only reduce the overall sieve time by about 1/3.

Speedy51
Send message
Joined: 16 Oct 09
Posts: 44
Credit: 293,342
RAC: 88
Message 1019 - Posted: 9 Oct 2012, 19:49:15 UTC - in response to Message 1017.

Right now, NFS@Home has the rough equivalent of ~1000 cores running full time, so adding 576 cores would only reduce the overall sieve time by about 1/3.

Fair enough. I wasn't very clear. What I meant was if you could run 1 task on 576 cores it would be done in no time.

Message boards : News : 3,637- factored, NSF grant renewed


Home | My Account | Message Boards