mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Proth Prime Search (https://www.mersenneforum.org/forumdisplay.php?f=109)
-   -   Proth primes with k = 1281979 (https://www.mersenneforum.org/showthread.php?t=26482)

bur 2021-05-03 07:40

No new primes for n < 3,600,000


Largest n in progress = 3,875,000
# of digits = 1,166,500 (rank: ~ 325)
FFT size = 400k
Avg. time = 4970 s

Smallest n in progress = 3,600,000
# of digits = 1,085,000 (rank: ~ 395)
FFT size = ---k
Avg. time = ---- s


Around July all sieved candidates of n < 4,100,000 should have been tested. I'm not so optimistic anymore that a prime will turn up...

bur 2021-05-29 07:06

Things progressed much faster than I expected, the last batch 4,095e6 < n < 4,100e6 will be assigned to a core today and everything should be finished until June, 10th.


If there will be no prime in these last few numbers, I'll continue by sieving the proth side to higher n and doing LLR on-the-fly depending on sieve/LLR times as LaurV suggested.


According to prime95 I'll be able to continue single-threaded until an FFT size of 640k is reached which will be around n = 5.6e6, which agrees nicely with the 5.3 MB of L3 each core has. I just hope a prime will turn up before testing becomes really slow. I'm set on finding a mega prime for this k though. :D

bur 2021-06-11 10:59

All n < 4,100,000 have been tested, no new primes other than the ones already posted.


Currently I'm sieving in the range of 4,100,000 < n < 10,000,000.

Smallest p-P range in progress: 87E12-107E12
Avg. time per factor: 810 s

Largest p-P range in progress: 310E12-330E12
Avg. time per factor: 3160 s

Remaining candidates: 100,936 / 5,900,000 (1.71%)


Est. time per LLR test of smallest candidate: 5600 s (n = 4,100,014)

bur 2021-07-20 17:31

Quick update, no new primes...


All n < 4,300,000 have been tested.

n < 10,000,000 sieved up to 570E12 (10,000 s / factor)

Approximately 90,000 candidates remain.

Largest LLR-test currently running:
n = 4,450,000
FFT = 448k
6200 s / test
Caldwell entry rank: 290


If the LLR-runtime would increase strictly quadratic, then it'll take to around n = 5,600,000 until an LLR test takes again as long as sieving to eliminate a candidate. As a very rough estimate that should take about 7 months on the 10 core.

And when will the next prime appear? Judging from Primegrid's results on other prime k values it can be a looong time.

bur 2021-08-27 17:14

To keep up with the roughly monthly updates and since it's been one year now since I started searching this k, here are the current stats:

As always, no new prime.


All n < 4,600,000 have been tested.

Largest LLR-test currently running:
n = 4,790,000
FFT = 512k
7500 s / test
Caldwell entry rank: 274

bur 2021-11-22 16:32

Once again, a status update with no new prime:

[B]Sieving[/B]
All n < 10,000,000 have been sieved to 800e12.
Last reported time per factor: approx. 12,000 s

76,167 candidates (1.6%) are left in the range 5,250,000 < n < 10,000,000.


[B]LLR[/B]
All n < 5,250,000 have been tested.

Smallest LLR-test currently running:
n = 5.25M
FFT = 560k
duration = 9975 s / test
digits = 1.58M
Caldwell entry rank: 254

Largest LLR-test currently running:
n = 5.36e6
FFT = 560k
duration = 10100 s / test
digits = 1.61M
Caldwell entry rank: 252

rogue 2021-11-22 17:09

If you have a GPU, use srsieve2cl for sieving. Depending upon the GPU, it might be faster than sr1sieve. I have a mid-range GPU and it is 5x faster than sr1sieve on the same CPU.

dannyridel 2021-11-24 00:51

[QUOTE=rogue;593620]If you have a GPU, use srsieve2cl for sieving. Depending upon the GPU, it might be faster than sr1sieve. I have a mid-range GPU and it is 5x faster than sr1sieve on the same CPU.[/QUOTE]

Where can srsieve2cl be obtained?

rogue 2021-11-24 15:23

[QUOTE=dannyridel;593736]Where can srsieve2cl be obtained?[/QUOTE]

It is part of mtsieve and can be found here: [url]https://sourceforge.net/projects/mtsieve/[/url]

dannyridel 2021-11-25 00:17

[QUOTE=rogue;593758]It is part of mtsieve and can be found here: [url]https://sourceforge.net/projects/mtsieve/[/url][/QUOTE]

Great! Thanks.

Happy5214 2021-11-26 00:05

[QUOTE=rogue;593620]If you have a GPU, use srsieve2cl for sieving. Depending upon the GPU, it might be faster than sr1sieve. I have a mid-range GPU and it is 5x faster than sr1sieve on the same CPU.[/QUOTE]

Not to hijack this thread, but based on personal experience, I've had trouble getting my GeForce RTX 2060 to sieve faster with srsieve2cl than my Intel Core i7-10875H running 8 threads with sr1sieve. In fact, it's not even close with my Riesel data (at least 2-to-1 IIRC).


All times are UTC. The time now is 05:29.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.