Riesel base 3 reservations/statuses/primes

[rebirther]

Reserving R3 from n=225-300k (1-2.147G) for BOINC

[rogue]

Originally Posted by rogue

I didn't check the reservation page and I did not look back far enough on this thread. I'll do from 60G to 63G then. I'll assume that your results for > 63G are good.

I am almost done to n=15000 for the range. It is taking much longer than expected.

[VBCurtis]

Originally Posted by VBCurtis

All 3 sieves that I'm running at present (1G-2^31, 4-5G, 5-6G) are sieving to 500k. Reb is testing chunks of these sieves as sieve depth reaches the correct level for those chunks.

This winter, I'll fill the gap from 2^31 = 2.147G to 4G via srsieve. In the long run, I'd like to sieve 100k-500k for any 1G range that individuals get to 100k, but that depends to some extent on how popular this base gets. I do plan to use sr2sieve for 0-2^31 from 500k up to get some top-5000 work available for this drive by Spring.

-Curtis

Quarterly update:
4-6G are now sieving together 250-500k. 250-300 will be sent to Reb in a few weeks, then 300-500 at once some time later.
1G-2^31 is testing 225-300, with 300-500 left in the sieve. sr2 allows up to 8 threads, so the time delay from Reb sending me primes to a new chunk being ready for BOINC is shorter.
2^31-4G has commenced sieving.

Once 1G-2^31 completes to 500k, I will sieve 0-2^31 together from 500k to 3M. Unless someone figures out which k's from 2^31 to 2^32 work with sr2, this is the largest k-range that will be done with sr2, so this block will sieve together forever. This will be a substantial undertaking, but I just acquired a 32-core Xeon server, so I'll feed 16 cores with sr2 and get a block ready for BOINC within a month.

[gd_barnes]

The main priority would be sieving k=2^31-4G for n=100K-500K to get it "caught up" with the other ranges.

[KEP]

Range 33G-43G is close to completion. Everything will be handed over in one mail with various drive links.

Reserving for R3 as new, the range 43G-49G to n=25K using srbsieve.

[rebirther]

R3 tested to n=300k (225-300k) (1-2.147G)

191 primes found, 760 remain in this range

Results emailed, Base released

[KEP]

I'm reserving for sieving for R3, k=13G to k=21G from n=25001 to n=100000 for my Sandy Bridge.

[VBCurtis]

Originally Posted by KEP

I'm reserving for sieving for R3, k=13G to k=21G from n=25001 to n=100000 for my Sandy Bridge.

You should consider doing these in 1G increments, or even smaller. After 1000 or so k values, srsieve isn't more efficient doing them all at once versus in separate sieves. Also, since srsieve isn't multi-threaded, you can do 8 different sieves on 8 hyperthreads and acheive happiness with less fiddling-with-factor-files.

Rebirther- If KEP tests 25-50k on his own, will you take 50-100k on boinc? Tons of tiny tests, might not be worth the admin time...

[rebirther]

Originally Posted by VBCurtis

Rebirther- If KEP tests 25-50k on his own, will you take 50-100k on boinc? Tons of tiny tests, might not be worth the admin time...

Yes, no problem.

[KEP]

Originally Posted by VBCurtis

You should consider doing these in 1G increments, or even smaller. After 1000 or so k values, srsieve isn't more efficient doing them all at once versus in separate sieves. Also, since srsieve isn't multi-threaded, you can do 8 different sieves on 8 hyperthreads and acheive happiness with less fiddling-with-factor-files.

Rebirther- If KEP tests 25-50k on his own, will you take 50-100k on boinc? Tons of tiny tests, might not be worth the admin time...

Well, I'm running this way:

13G-15G 1 core
15G-17G 1 core
17G-19G 1 core
19G-21G 1 core

I'm going to sieve this range on my Sandy Bridge, untill R3 gets completely tested to n=25K and after that I plan on testing with my i5-4670 the range of 13G-21G to n=100K. Personally even if Reb wouldn't mind, I do not recommend due to the short testing time to use BOINC. An n=100K test will take on the previously mentioned i5 about 9 sec per test, so a lot of time will be wasted on compressing and uploading results. Just my thoughts. If Reb still likes to use BOINC, I may as well stop at n=50K and hand Reb the rest

[rebirther]

Originally Posted by KEP

I'm going to sieve this range on my Sandy Bridge, untill R3 gets completely tested to n=25K and after that I plan on testing with my i5-4670 the range of 13G-21G to n=100K. Personally even if Reb wouldn't mind, I do not recommend due to the short testing time to use BOINC. An n=100K test will take on the previously mentioned i5 about 9 sec per test, so a lot of time will be wasted on compressing and uploading results. Just my thoughts. If Reb still likes to use BOINC, I may as well stop at n=50K and hand Reb the rest

Let me know how many tests are total for this range (9s are ok but not less). Iam still waiting to rebuilt my VM on Saturday for more RAM, I have hit the limit of 32GB at the moment.