## Other

drummers-lowrise

Message boards : Sieving : PPS sieve depth

 Subscribe SortOldest firstNewest firstHighest rated posts first
Author Message
rebirther

Joined: 10 Aug 05
Posts: 783
ID: 85
Credit: 175,774,608
RAC: 0

Message 57830 - Posted: 26 Sep 2012 | 14:37:28 UTC

Whats the optimal sieve depth or where we will finished it?

[DPC]Charley

Joined: 15 Apr 11
Posts: 803
ID: 95137
Credit: 94,993,523
RAC: 0

Message 57944 - Posted: 1 Oct 2012 | 21:15:33 UTC

Good question :)

As I don't have any numbers handy that show me how long it currently takes to eliminate 1 candidate by sieving, I can't give you a really definitive answer.

However, I'm willing to hazard a guess that we won't be stopping any time soon as the lower n's with high k's still take multiple hours on recent AVX enabled CPU's whereas it only takes minutes to sieve, though it's unclear how many candidates are actually being removed per work unit.

TPTB however can probably give an answer in which actual numbers are involved. (I hope)
____________
PrimeGrid Challenge Overall standings --- Last update: From Pi to Paddy (2016)

Mr. Cool

Joined: 20 Jan 11
Posts: 160
ID: 82236
Credit: 758,564,854
RAC: 135,223

Message 69660 - Posted: 6 Oct 2013 | 0:26:15 UTC

Michael, would you please calculate optimal sieve depths for the PPS sieve the same like you've done for the GFN Sieve?

Michael Goetz
Volunteer moderator

Joined: 21 Jan 10
Posts: 14011
ID: 53948
Credit: 435,627,755
RAC: 870,396

Message 69666 - Posted: 6 Oct 2013 | 3:24:07 UTC - in response to Message 69660.

Michael, would you please calculate optimal sieve depths for the PPS sieve the same like you've done for the GFN Sieve?

I'll leave that to others who are more familiar with it.
____________
My lucky number is 75898524288+1

cncr04s

Joined: 17 Sep 05
Posts: 139
ID: 797
Credit: 3,990,128,691
RAC: 2

Message 69667 - Posted: 6 Oct 2013 | 3:24:35 UTC - in response to Message 69660.

Whats the optimal sieve depth or where we will finished it?

PPS Sieve is one of the most worked on sub projects here.
While, at some point the sieve becomes un-optimal, there would be a massive piece of computing power lost.
Any sort of sieve depth calculation should also take in account this loss of power, as there is no gpu pps llr for people to switch to once the sieving ends.
____________

JimB
Honorary cruncher

Joined: 4 Aug 11
Posts: 920
ID: 107307
Credit: 989,270,184
RAC: 150,909

Message 69671 - Posted: 6 Oct 2013 | 4:09:00 UTC

There will always be PPS sieving. Think about it this way:

Originally we sieved n=0-3M. I don't know if that started as manual sieving, but before I came here it was on BOINC. While that was on BOINC, at some point manual sieving was started for n=3M-6M.

When it seemed that 0-3M had been sieved far enough, sieving on 0-3M was suspended, 3M-6M moved to BOINC, and manual sieving started on 6M-9M. That's where we are right now.

What will happen next is that sieving on 3M-6M will be retired, 6M-9M will move to BOINC and 9M-12M will start on manual sieving.

I create new sieve files about once a month. At that time Lennart and I discuss whether it's time to change over yet. So far we've decided the time hasn't yet come.

Mr. Cool

Joined: 20 Jan 11
Posts: 160
ID: 82236
Credit: 758,564,854
RAC: 135,223

Message 69681 - Posted: 6 Oct 2013 | 5:57:41 UTC - in response to Message 69671.

JimB, can you please share how many factors are found on average per one PPS Sieve wu at the moment?

JimB
Honorary cruncher

Joined: 4 Aug 11
Posts: 920
ID: 107307
Credit: 989,270,184
RAC: 150,909

Message 69689 - Posted: 6 Oct 2013 | 12:14:29 UTC - in response to Message 69681.

JimB, can you please share how many factors are found on average per one PPS Sieve wu at the moment?

Short answer: 2.26 factors per workunit on average.

Better answer: Factors found is only half of the story. The percentage of those factors that result in new removals from the sieve is just as important. I don't have those numbers at my fingertips except when I'm creating sieves. I don't keep any of that data because 1) I've never had any reason to revisit it and 2) it gets outdated very quickly. The last sieves were created on September 23rd, the next ones will be near the end of October. Visiting the link shows you how many candidates are left in the RSP factor files. We don't release the PPS factor files, but the numbers are similar. Keep in mind that the PPS Sieving affects six different sieve files: PPS4M, PPS5M, PPS6M, RSP4M, RSP5M and RSP6M. Manual sieving affects the same files in the 7M-9M range (another six sieves).

I can also tell you that we're getting close to the point of changing over to the next ranges. I just don't know offhand how close right this moment.

Dave

Joined: 13 Feb 12
Posts: 3209
ID: 130544
Credit: 2,288,183,124
RAC: 754,139

Message 70120 - Posted: 17 Oct 2013 | 12:05:41 UTC

Hi

Interesting. Will the next range have any impact on amount of processing over time i.e RAC change?

JimB
Honorary cruncher

Joined: 4 Aug 11
Posts: 920
ID: 107307
Credit: 989,270,184
RAC: 150,909

Message 70122 - Posted: 17 Oct 2013 | 14:36:56 UTC - in response to Message 70120.

Hi

Interesting. Will the next range have any impact on amount of processing over time i.e RAC change?

I don't really know. If so, we'll make an adjustment in the credit. Remember, this has only happened once before at PrimeGrid and that was before my time.

edit: typo

Ken_g6
Volunteer developer

Joined: 4 Jul 06
Posts: 940
ID: 3110
Credit: 261,913,874
RAC: 8,863

Message 72686 - Posted: 25 Jan 2014 | 21:04:14 UTC - in response to Message 70120.

Hi

Interesting. Will the next range have any impact on amount of processing over time i.e RAC change?

The first WUs will be relatively slow. Once the size of prime factors being tested gets over about 80T, the speed should jump back up close to where it was.* It won't be exactly the same, though, because there are more prime numbers to try factoring with in the 80T range than in the ~350P range where we are now. I estimate WUs will be about 26% longer at that point.

* This doesn't apply to the CPU apps, which will be much slower to come back to their current speed.
____________

Message boards : Sieving : PPS sieve depth