20060903, 21:23  #1 
Dec 2004
100101011_{2} Posts 
Curious how sieve points are calculated
Just curious how the sieve points are calculated for PSP?
Is there a link? Sorry to lazy today to seach.... 
20060903, 21:38  #2 
Jun 2003
1,579 Posts 

20060904, 14:40  #3 
Dec 2004
453_{8} Posts 
This answers my question somewhat:
Sieving: Points given take into account the n value removed and the factor found. You get more points for higher n and for larger factors. The maxmimum points you can get per factor is 6000. This was done due to the fact that else a lucky "p1" factor would give you far to many points. There is a possibility that the maximum points will be changed in the future. This comes out of two reasons. The 6000 was set when our maximum n was 20m. The second reason will be a new siever that allows higher sieving then ~1000T. Just curious on the max of p and n as well as the equation. I like the idea that factors score as soon as they are submitted, and limiting p and n are great ideas. 
20060904, 16:03  #4 
Apr 2003
2^{2}×193 Posts 
The formula used at the moment is the following:
sievepoints=least(6000,pvalue/10000000000000.0*nvalue/1000000.0) This was my first shoot. As i had no clue how the distribution of factors changes over time there are no special measurements implemented. If somebody has suggestions how to make the stats better please inform me. It is technicaly very easy to change but be warned you must have very very very good arguments else i will not change anything. I have learned my lessons from the time where i made some adjustements to the PRP stats and we lost some users due to the fact that i was unable to describe good enough why i made the changes and that the stats are fairer now. Lars 
20060904, 17:27  #5 
Dec 2005
313 Posts 
Hi Lars, :)
The conversation came out of a discussion we were having on the ARS psp forum thread and was more related to the issue of trying to figure out what was better to crunch in relation to points awarded for X amount of work. VJS asked me if I knew, and I said that you had told me once but in my old feeble mind I'd forgotten. :) So this wasn't driven by a desire to change the factor point scoring method. (at least as far as I'm aware.) Bruce 
20060904, 22:01  #6  
Dec 2004
299_{10} Posts 
I'm not trying to get you to change anything I think it works just fine currently the questions just came out as brucifer described.
At the bottom of this post is the way SoB scores factors you will see that it's different compared to PSP. Personally see advantages to both ways. I like the idea of factors scoring as soon as they are submitted, sometimes one sieves and sieves for no points in SoB. The user will leave for a few months and come back later. They don't understand how their 500 points turned into 2 million. (The whole window scoring concept). Quote:
http://www.aooq73.dsl.pipex.com/2006/scores_p.htm It's not obvious, but you have to also consider that the number of factors is decreasing b/c of the removal of k's due to primes. Under Current sieve points (for factoring) Basically one will find the same number of factors in the range of 2^452^46 as 2^462^47 although the range of p is twice as large.  p < 40T, score = p/1T p > 40T, in ‘main active' window, 0 PRP tests performed, score = (n/1M ^ 2) * 125 * bias p > 40T, between ‘DC active window’ and ‘n upper bound’, 0 PRP tests performed, score = (n/1M ^ 2) * 125 * bias (then score will not increase further). p > 40T, in 'DC active' window, 0 PRP tests performed, score = (n/1M ^ 2) * 125 * bias p > 40T, in 'DC active' window, 1 PRP tests performed, score = (n/1M ^ 2) * 125 * 0.6 * bias p > 40T, in 'DC active' window, 2 PRP tests performed, score = (n/1M ^ 2) * 125 * 0.2 * bias p > 40T, in ‘completed' window, score = (n/1M ^ 2) * 125 * 0.2 p > 40T, k=prime; as above then frozen when a prime is found, factors found after the prime score 0 none of the above, score = (as duplicate, see below) bias = p/40T or current 90% sieve point/40T, whichever is lower. A duplicate factor will score as follows: score = p/100T, capped at 35, or the score (above) as if it were unique, whichever is lower; when a prime is found all duplicates ever found for that k score 0. A unique factor is the first factor found for a candidate. Scores for each unique factor are remembered. Scores can go up (as an 'active' window moves to cover a factor that was above the window), but cannot go down (as a factor moves out of a window). Scores with 0 PRP tests exiting the main active window will receive no further increase (i.e. no increase when they reach the DC active window). The 'main active' windows is (<next candidate>) < n < (<next candidate> + 200K). The ‘DC active' windows is (<next double check candidate>) < n < (<next double check candidate> + 200K) A second ‘DC active' windows is (<next double check candidate2>) < n < (<next double check candidate2> + 10K) The 'completed' windows is 0 < n < (<next double check candidate>) ‘n upper bound’ is the lowest of the ‘n bound (upper)’ described here Excluded factors (those factors not present after sieving 100<n<20M to p=1G) do not score. Before 21July2003 scores were calculated as follows: n < 300K, score = p/1T * ((n*n)/(300K * 300K)) 300K < n < 20, score = p/1T n > 20M, score = p/1T *0.05 Duplicates, score = score * 0.01; when a prime is found all duplicates ever found for that k score 0. 

20060905, 14:11  #7 
Dec 2004
13×23 Posts 
I just wanted to add my 0.02 regarding sieve and P1.
As everyone knows using P1 for PSP is not benifital at this point. Regardless I'll make the comment for future reference. Part of the problem with P1 is the scoring... project wise, this issue generally comes down to a decent method of calculating the score for the factor in regards to sieved factors. When the time comes personally I would actually not add the points scored by factors from P1 towards sieve points. I realize it's factoring but its not sieve. Personally I would put P1 factor scores into the LLR section. Why? I believe it makes the most sence to put them here since P1 basically tests or removes those n about to be tested for primes. (which is more LLR based) As for the scoring, I would score those factors as 1.5x a llr test since factored k/n pairs do not have to be doublechecked. 
20060910, 10:09  #8 
Feb 2003
Istanbul
2^{2}×13 Posts 
What do the others think about P1 scoring?
I'm working on P1 tests. When I have some data I'll share with you. By the way any advices about P1 tests will be appreciated. 
20060910, 18:45  #9 
Jun 2005
373 Posts 
Right now, P1 is not worth the cycles, and if we keep up sieving like this,it won't become efficient before a long time. That's why I don't worry so much about it (for PSP). H.

20060910, 19:02  #10 
Dec 2004
453_{8} Posts 
Agreed with HHH, P1 is a waste of cycles right now. If you don't want to prp with your machine, sieve even if it's less efficient on that CPU.
Perhaps around 5M P1 will become efficient but I would not start thinking about that until 4M. Who knows sieve could reach 1000T by then, its not impossible. I believe LTD (Lars) does calculations and tests every 1M just to check if P1 is worth it. In any regards the question still remains about the points... I think P1 points should count towards PRP points..., but again I wouldn't worry about that until we start P1'ing. Last fiddled with by VJS on 20060910 at 19:04 
Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
Let y^2=xzx^2+1, and if value of x is known, can y and z be directly calculated? Given all variable  MARTHA  Number Theory Discussion Group  11  20180302 15:55 
How are GHzDays calculated?  Unregistered  Information & Answers  10  20130517 14:26 
Two points  devarajkandadai  Math  19  20121215 14:15 
how is the throughput calculated?  ixfd64  PrimeNet  5  20080521 13:39 
More points for PRP?  Mystwalker  Prime Sierpinski Project  6  20060103 23:32 