mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Riesel Prime Search (https://www.mersenneforum.org/forumdisplay.php?f=59)
-   -   Any interest in yet another sub-megabit drive? (https://www.mersenneforum.org/showthread.php?t=15998)

Kosmaj 2011-08-31 01:53

Any interest in yet another sub-megabit drive?
 
All our drives featuring n<1M are almost over.
Is there interest in one or more new sub-megabit drives?
If so, what starting exponent do you prefer, n=700k or 800k or another one?

As far as I'm concerned I prefer testing n>1M, but of course if we start a new drive I'll test a few files.

VBCurtis 2011-08-31 03:10

I think we should continue any k from 8th or 9th drive that is below k=10000 into megabit range. I am not interested in new k's under 1M.

One new drive I would get behind (even if it starts below 1M) is adopting a contiguous range of k's below 10,000 to test together, like Peter has done with 1000-1300. I would suggest we use some of that range, but Peter is easily keeping up on those k's working alone!
-Curtis

Thomas11 2011-08-31 08:42

I would also like the idea suggested by Curtis of testing a contiguous range of k from the k<10000 pool.
According to [URL="http://www.rieselprime.de/"]http://www.rieselprime.de/[/URL] all below k=4200 are taken by others (teams or individuals), so we might start from there.

Perhaps, we could provide the test files for the sub-megabit ranges too (e.g. starting from n=700k), so that users with slower machines and/or new team members may pick from those files and others may start from n=1M.
This would also give us the chance of finding some primes before they are getting too small for the Top-5000 list...

BTW.: There are also the high weight k's from the 7th drive, we could start with from n=1M.

pb386 2011-08-31 09:09

I would be interested in something sub-megabit. A contiguous range would be a good idea and seems to be quite popular!

amphoria 2011-08-31 17:39

I am prepared to transfer k=4003-4199 to RPS if people are interested. I have searched it from n=500k-800k and it is getting close to the point where the work involved will be greater than my available resources. There are about 2900 candidates per 1000n, so a bit less work than the 9th Drive. Note that k=4153 was searched to n=1M by Beyond before he knew that I had the range reserved. k=4001 could also be added from 1M.

VBCurtis 2011-09-01 03:12

I think amphoria's range is a must-do; do we wish to extend the range above 4200? We could run 4003 up to 4300 or 4400 starting at 800k, and also start a sub-drive to get 4201-4300 or 4400 'caught up' from top5000 cutoff to 800k.

My personal opinion is that the 9th drive is a bit big, so 4003-4199 sounds just right to adopt. If it goes quickly, we can start another one on 4201-4399 without much effort- and this also has the nice effect of having drives available at multiple n-ranges like we've done for so long.

So, how about a continuation of 7th and 8th drive k's at 1M, and also 4003-4199 at wherever it is now?
-Curtis

Kosmaj 2011-09-01 07:21

Okay, thanks for your replies.

Then, we'll start another sub-megabit drive very soon.
Next, we have to decide on exponents and based on that on coefficients (k).

I think n=700k is a bit too low. PrimeGrid guys reported 30 Proth primes at around n=690k yesterday and will reach 700k soon. Here are current, approximate ranks of some primes with "milestone" exponents on Top-5000:
n=700k ... 3150
n=750k ... 2500
n=800k ... 2050
Please suggest the starting n: 700k, 750k or 800k (or another one).

k's between 4003 and 4199 are fine if we start at 800k.

Thomas11 2011-09-01 12:15

Perhaps this helps for the decision about the starting point:

For [B]k=4000-4200[/B] the FFT length changes:
at n=700k (48k --> 56k)
at n=814k (--> 64k)
at n=931k (--> 72k)

Regarding k=1400-2000 and 2000-3000, I'm not sure about NPLB's intentions. So far it seems that they are concentrating on the lower n ranges only, so that we may consider to take them too...

VBCurtis 2011-09-01 21:11

I would like to start as low as possible, as I believe there's goodness in cataloging as many primes as we can find.

re: Thomas' idea, it's a tough spot. NPLB does not have the resources to test all of 1400-3000 even to keep up with 5000th place, but Gary is nothing if not territorial and rather unlikely to quietly cede any of that range. If I am wrong about his interests and they realize they can't do it all, I'd vote for 2000-2300- that way, we'd have 5-300, 1001-1300 (peter), 2001-2300.
-Curtis

amphoria 2011-09-01 22:19

[QUOTE=VBCurtis;270598]re: Thomas' idea, it's a tough spot. NPLB does not have the resources to test all of 1400-3000 even to keep up with 5000th place, but Gary is nothing if not territorial and rather unlikely to quietly cede any of that range. If I am wrong about his interests and they realize they can't do it all, I'd vote for 2000-2300- that way, we'd have 5-300, 1001-1300 (peter), 2001-2300.[/QUOTE]

If you want to go with 2001-2300 then, in this instance, I would be surprised if Gary does anything, as he effectively give up on k>2000 in favour of prioritising k=1400-2000. The NPLB Drive that covered k>2000 was terminated several month ago. However note that I have already searched 15 k's in this range to n=1M and a few others are already reserved by RPS members.

Kosmaj 2011-09-02 10:48

We'd definitely like to avoid troubles, but I just checked Top-5000, and found that all Riesel primes with 2000 < k < 3000 posted since 2009 were reported by RPS members! Which means that nobody else is searching for them.

So, how about Ks in the 2000-2300 range, excluding already searched ones, starting at n=700k ?

k = 2235, 2265, 2295 from our 7th Drive should be excluded as well.

Dave, can you post the ones you already tested?

Thomas11 2011-09-02 11:29

I would like the idea to go with the k=2000-2300 range, starting from n=700k.

According to ([URL="http://www.rieselprime.de"]www.rieselprime.de[/URL]) the following Ks have already been tested to n=1M (or beyond):

2065, 2073, 2087, 2089, 2093, 2109, 2123, 2125, 2151, 2153,
2155, 2177, 2179, 2185, 2199 (tested by Amphoria to 1M)

2235, 2265, 2295 (part of the 7th drive, tested to 1M)

2293 (Riesel k, tested by PrimeGrid)

Other Ks obviously tested to 1M:
2145, 2175 (Curtis ?)
2187 (L. Zhou)
2203 (Gillion)

Can someone confirm this list?
Are there any other Ks to be omitted?

BTW.: Does anyone has the latest n<1M sieve file from PrimeSearch. The download links seem to be broken and the latest file I have is from May 13th...

Kosmaj 2011-09-02 12:35

My last n<1M sieve file is from April 7 :sad:

amphoria 2011-09-02 16:53

[QUOTE=Kosmaj;270635]My last n<1M sieve file is from April 7 :sad:[/QUOTE]

I have 13th June file which is the latest one that was posted.

I have just checked on the PrimeSearchTeam forum and it still appears to be available at [url]http://uwin.mine.nu/rsp/sievefile/rsp1M_20110613.7z[/url].

Note that John has just posted that the final file sieved to 100P will be posted during September.

VBCurtis 2011-09-03 05:15

Steven (sjtjung), chaos13, and I are nearly done testing the following k's to 1M:
2055, 2085, 2115, 2145, 2175.

We have small gaps to fill to reach 900k, and then about 50k block to test in the 900-1000 range. We will decide whether to merge these k's with this new drive at 1M later, once the drive nears 1M.
-Curtis

gd_barnes 2011-09-03 08:01

I saw discussion here about searching a large group of contiguous k's. I think that would be a great idea for RPS. Please note that NPLB effectively abandonded the k=2000-3400 range at n=600K because it was too large for our resources. If you guys want to work on a portion of or that entire range instead of k>4000, be our guest. Due to the smaller k's, you'd get a little faster test times for some n-ranges. That would mean that there wouldn't be a big "hole" for k=2000-3400.


Gary

gd_barnes 2011-09-03 08:08

[QUOTE=VBCurtis;270598]I would like to start as low as possible, as I believe there's goodness in cataloging as many primes as we can find.

re: Thomas' idea, it's a tough spot. NPLB does not have the resources to test all of 1400-3000 even to keep up with 5000th place, but Gary is nothing if not territorial and rather unlikely to quietly cede any of that range. If I am wrong about his interests and they realize they can't do it all, I'd vote for 2000-2300- that way, we'd have 5-300, 1001-1300 (peter), 2001-2300.
-Curtis[/QUOTE]

Territorial? OK, if you say so. :-)

Like I said in the last post, take what you want for k=2000-3400. We're still squeaking by with k=1400-2000 so we'll continue with that range for the time being.

em99010pepe 2011-09-20 13:47

I know I am late but drives at 1M and beyond should be considered not only for the size of the primes but for gathering points for the Top 5000 ranking's. For the moment I can't help because life circumstances changed :down: , I miss prime search a lot.
The only thing I can offer right now is to host or a prpnet server or a llrnet one with full remote access to Kosmaj or Curtis or Thomas...

Thomas11 2011-09-21 16:50

Just as an info:

Besides the 11th drive, Peter and I are also testing the range k=2300-3000 (in total: 315 Ks) starting from n=666666. You may have noticed our recent primes. Initial target is to get them to n=700k, which means about 300000 tests. If someone wants to join this effort, just let us know...

Edit:
The following Ks are EXCLUDED:
2325, 2355, 2415, 2445, 2505, 2547, 2565, 2591, 2655, 2685,
2715, 2741, 2745, 2775, 2797, 2835, 2895, 2925, 2927, 2929,
2931, 2933, 2935, 2937, 2939, 2941, 2943, 2945, 2947, 2949,
2951, 2953, 2955, 2983, 2985

Thomas11 2011-11-16 10:38

Due to the recent activity in the 11th drive I'd like to ask again whether there is some interest in another sub-megabit drive.

pb386 and I have tested [B]k=2300-3000[/B] for n=666666-700000 so far (see above post). Shall we open it as another team drive starting from n=700k ?

VBCurtis 2011-11-17 04:10

I would suggest waiting until the 11th passes 800k, so that we maintain as great a variety of n-levels in our drives as possible. Then I checked the status of 11th, and it's at 792k- making my suggestion moot.

Sounds like a good idea.
-Curtis


All times are UTC. The time now is 06:26.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.