mersenneforum.org Sad Milestone: 1200 (now 1800+) days since last prime
 Register FAQ Search Today's Posts Mark Forums Read

2013-07-09, 19:17   #34
Citrix

Jun 2003

1,609 Posts

Quote:
 Originally Posted by ltd Finally all received manual ranges imported. Next task give the DC server something to do again.
Excellent! Could you also update the primegrid reservations when you get a chance. Also I am interested in doing some sieving work on the side, could you build a .dat file for me.Thanks.

 2013-07-09, 20:04 #35 ltd     Apr 2003 30416 Posts I have already requested the latest results from Primegrid. The latest valid sieve file was this one: http://www.psp-project.de/test/sievecomb.zip That was the combined sieve with SOB. To make a PSP only file remove the SOB blocks and change the number in the first line. Last fiddled with by ltd on 2013-07-09 at 20:06
 2013-07-09, 20:09 #36 Citrix     Jun 2003 31118 Posts Is it possible to create a .dat (PSP only) with only the untested values (once you update the Database with results from primegrid) (if this is too much work, I would not spend too much time on it for now.) Last fiddled with by Citrix on 2013-07-09 at 20:16
2013-07-09, 20:20   #37
opyrt

Apr 2008
Oslo, Norway

7·31 Posts

Quote:
 Originally Posted by Citrix Is it possible to create a .dat (PSP only) with only the untested values (if this is too much work, I would not spend too much time on it for now.)
Hi Citrix,

The speedup from doing this is very, very small. I did a few tests on it back when we were still sieving. And it's a very good thing to do the combined PSP/SoB sieving as our project isn't done if their's isn't. It also makes chances of finding factors better, although at the sieve limit we're at they're quite slim either way. :)

-Kai

2013-07-09, 20:29   #38
Citrix

Jun 2003

1,609 Posts

Quote:
 Originally Posted by opyrt Hi Citrix, The speedup from doing this is very, very small. I did a few tests on it back when we were still sieving. And it's a very good thing to do the combined PSP/SoB sieving as our project isn't done if their's isn't. It also makes chances of finding factors better, although at the sieve limit we're at they're quite slim either way. :) -Kai
I agree with you. I am not doing any large scale sieve etc.
Just need the candidates to test something.

 2013-07-11, 16:59 #39 Citrix     Jun 2003 1,609 Posts Are all ranges under 14M done? What is remaining... we should finish it as high priority!
2013-07-11, 17:13   #40
opyrt

Apr 2008
Oslo, Norway

21710 Posts

Quote:
 Originally Posted by Citrix Are all ranges under 14M done? What is remaining... we should finish it as high priority!
Assuming you mean first pass tests, PrimeGrid has n=14489501 as lowest untested value. Other than the PG tests we only have the ones in the "pick a range" thread left under 15M.

Edit: And a few tests in progress on the prpnet server. They are all tests that had timed out:
prpnetnew13000000.txt = 2
prpnetnew13300000.txt = 13
prpnetnew13350000.txt = 19
prpnetnew13400000.txt = 7

-Kai

Last fiddled with by opyrt on 2013-07-11 at 17:19

 2013-07-13, 04:13 #41 Citrix     Jun 2003 1,609 Posts Kai, Have you tested running ppsieve GPU version on the dat file? From what I understand the runtime will be proportional to the nmax-nmin range size. So we can just sieve a small n range (eg. the PRP tests that we are currently doing) and save some PRP tests. I think this will be faster than doing PRP tests. I don't have a GPU myself to do any testing. Do you have one? https://sites.google.com/site/kenscode/prime-programs Last fiddled with by Citrix on 2013-07-13 at 04:25
2013-07-13, 16:47   #42
opyrt

Apr 2008
Oslo, Norway

7×31 Posts

Quote:
 Originally Posted by Citrix I don't have a GPU myself to do any testing. Do you have one? https://sites.google.com/site/kenscode/prime-programs
Ungelovende on my team an I did some tests now, and ppsieve is not performing well on this. This is from the ATI version:
Code:
p=83000000021757953, 362.6K p/sec, 0.06 CPU cores, 0.0% done.
p=83000000042467329, 345.2K p/sec, 0.05 CPU cores, 0.0% done.
This is from sr2sieve:
Code:
p=830001203503861, 10076587 p/sec, 0 factors, 0.0% done
So sr2sieve is running in circles around ppsieve. We only tested on one k as ppsieve is meant for ranges of k's, not separate k's.

-Kai

2013-07-13, 17:26   #43
Citrix

Jun 2003

1,609 Posts

Quote:
 Originally Posted by opyrt Ungelovende on my team an I did some tests now, and ppsieve is not performing well on this. This is from the ATI version: Code: p=83000000021757953, 362.6K p/sec, 0.06 CPU cores, 0.0% done. p=83000000042467329, 345.2K p/sec, 0.05 CPU cores, 0.0% done. This is from sr2sieve: Code: p=830001203503861, 10076587 p/sec, 0 factors, 0.0% done So sr2sieve is running in circles around ppsieve. We only tested on one k as ppsieve is meant for ranges of k's, not separate k's. -Kai
It will definitely be slower for the whole 50M. How about the speeds for 100K-1M n.
What is the 0.06 CPU mean?
I think it might be slower for PSP as ppsieve does not take into account the low weight of PSP numbers compared to srsieve. This is something we can fix.

Last fiddled with by Citrix on 2013-07-13 at 17:45

2013-07-13, 17:56   #44
opyrt

Apr 2008
Oslo, Norway

7×31 Posts

Quote:
 Originally Posted by Citrix It will definitely be slower for the whole 50M. How about the speeds for 100K-1M n.
I didn't test that, as we are at n=15M+ now.

Quote:
 What is the 0.06 CPU mean?
I believe it is how much CPU that is used to feed the GPU.

 Similar Threads Thread Thread Starter Forum Replies Last Post aketilander Data 16 2016-01-19 23:32 cuBerBruce Puzzles 3 2014-12-01 18:15 tcharron PrimeNet 3 2013-08-29 06:44 tha Data 526 2010-11-23 00:09 opyrt Prime Sierpinski Project 65 2010-10-06 13:18

All times are UTC. The time now is 18:26.

Mon Mar 20 18:26:20 UTC 2023 up 214 days, 15:54, 0 users, load averages: 2.14, 1.75, 1.43