An Open Request for Sieve Help!

[jmblazek]

As PrimeGrid continues to add projects, more pressure is placed on the sieving aspect to keep up with growth. Therefore, PrimeGrid is in a continuous need for sievers...both manual and BOINC.

Currently, there are seven sieving efforts underway...including both BOINC and manual. Some are long term sieves like the combined Prime Sierpinski Project and Seventeen or Bust sieve. Others are shorter, like Proth Sieve, having ending dates within the next several months. However, all of them need help.

Currently, a few of the sieves have been prioritized depending on the demand of primality testing. The top three priorities are:

1. Proth Sieve - BOINC
2. 321 Sieve - Manual (manual credit/cobblestones available)
3. Prime Sierpinski Project (Seventeen or Bust) Sieve - BOINC and Manual

If anyone is interested, please see this post for more details: Calling All Sievers (http://www.primegrid.com/forum_thread.php?id=1040&nowrap=true#10604)

Thank you for your consideration.

[jasong]

As PrimeGrid continues to add projects, more pressure is placed on the sieving aspect to keep up with growth. Therefore, PrimeGrid is in a continuous need for sievers...both manual and BOINC.

Currently, there are seven sieving efforts underway...including both BOINC and manual. Some are long term sieves like the combined Prime Sierpinski Project and Seventeen or Bust sieve. Others are shorter, like Proth Sieve, having ending dates within the next several months. However, all of them need help.

...

If anyone is interested, please see this post for more details: Calling All Sievers (http://www.primegrid.com/forum_thread.php?id=1040&nowrap=true#10604)

Thank you for your consideration.
I just increased my share on my laptop and enabled all the sieves at the site(apparently, BAM doesn't do anything to the sub-project selection at PrimeGrid)

[jmblazek]

Good progress has been made towards the top 3 priorities. A summary is below:

1. Proth Sieve (p=106T) - goal, to reach a sieve depth of 500T initially and 850T long term
2. 321 Sieve (p=700T) - goal, to reach a sieve depth of 1000T initially
3. Prime Sierpinski Project (Seventeen or Bust) Sieve (p=11.1P) - goal, to reach p=25P before 13 October 2009 (2 year anniversary)

As you can see, #2 & #3 are doing quite well. However, #1 could still use some help. It's only available for 64 bit Linux and Windows. WU's are 2 to 2 1/2 hours long. The sieve file is 380MB.

If anyone is interested, please see this post for more details: Calling All Sievers (http://www.primegrid.com/forum_thread.php?id=1040&nowrap=true#10604). If you are interested specifically in Proth Sieve, please see this post: PPS Sieve (http://www.primegrid.com/forum_thread.php?id=973&nowrap=true#9505)

Thank you for your consideration.

[em99010pepe]

Explain here for "n up to 5M, "optimal" sieve depth is approximately 850T" for both projects, 321 Sieve and Proth Sieve. Thanks.

[jmblazek]

Explain here for "n up to 5M, "optimal" sieve depth is approximately 850T" for both projects, 321 Sieve and Proth Sieve. Thanks.
The current 321 sieve is a multi-sequence (2) sieve and is for 5M-25M so the 1P initial goal is just a milestone. We'll be sieving much deeper.

The 850T depth reached for the first 321 single sequence sieve (1<n<5M) had a factor/second rate that was almost a double pass LLR test at n~4M using a C2Q 6600 at 2.4GHz. We used 64 bit sr1sieve v1.3.4 and LLR v3.7.1c for timings.

For Proth Sieve, the p=850T is just a guideline. Since the sieve includes 598 sequences, we are not looking at depths for individual k's...it's an overall picture that we're interested in right now. This is our largest multi-sequence sieve so we'll monitor and fine tune it as it advances. I'm sure conditions will change as we get deeper.

One thing for sure, all k's will be sieved minimally 1.5X a first pass LLR rate at n~4M. We are currently sieving with 64 bit sr2sieve v1.8.2.

As for "optimal" sieve depth, there will be "optimal" depths for each CPU type and speed. And as software/hardware improves, depths will change. Again, we just look at it in general terms and make adjustments accordingly.

[em99010pepe]

About the several llrnet servers, what's the maximum allowed time for a user to complete a job?

[jmblazek]

About the several llrnet servers, what's the maximum allowed time for a user to complete a job?

4 * 24 * 3600 (4 days) for all ports

[jmblazek]

We'd like to do a PUSH in PPS LLR to get it up to n=346,000 which is Top 5000 level in Chris Caldwell's The Largest Known Primes Database! (http://primes.utm.edu/primes/)

However, the sieve is still more important right now. We have reached the 2T/day level but really need to be at the 4T/day level. With this rate, we'd be advancing the sieve at over 100T per month. This would give us still almost a year of sieving but also allow us to push the lower levels of PPS LLR.

Therefore, PPS Sieve is STLL IN NEED of 64 bit sievers (BOINC). If you are interested, please PM me or Lennart and you'll be added to the project. If you know of anyone who might be interested, please direct them to this thread (http://www.primegrid.com/forum_thread.php?id=973).

Thank you for your help!

Basic requirements:

1. 64 bit OS only (Linux or Windows) - 64 bit apps
2. 500 MB RAM per core
3. 2+ GHZ