Riesel base 3 reservations/statuses/primes - Page 13

michaf · 2008-11-23, 13:47

While testing my 80M range (520M-600M)

I first tried pfgw to 1000 using Karsten's scripts,
then sieve it, and tried Karsten's 'high-n' script

That took about 4 minutes writing time when a prime was found, so it was waaaay too slow :)

second,I tried to n=2000 with pfgw, sieved, and then high-n'd it, which was way better (30 sec/write). I'm finishing that now as it is quick enough.

I think that running pfgw to 2500 or even 3000 is the most efficient.
I will test on following ranges what is best (I'll go for 100M ranges then, and will note total time taken. expect an update in about 10 weeks :) )

Flatlander · 2008-11-23, 14:14

Quote:

Originally Posted by gd_barnes

Chris, can you double-check the k=500M-505M range? I messed around with it for a bit but don't feel I can spare the resources at the moment...too many other efforts going on.

Thanks,
Gary

No problem. I'll start when the NPLB rally finishes.

gd_barnes · 2008-11-24, 11:38

I have done some preliminary double-check work on the k=500M-506M range. After getting primes from Micha for n=1004-25000, I decided to just run the entire range up to n=1005 to have an n=2 overlap.

Then using the primes for n=1004-25000 that Micha sent me and my remaining k's after my testing, I was able to determine the k's that should be remaining at n=25K. To my surprise, they are correct as listed by Micha. (No offense!

) I'm just amazed that there are only 8 k's remaining at n=25K for k=500M-505M!

I did find a few problems with either Micha's testing or Karsten's scripts though. For the range of n=1001-1004, I found 6 primes that Micha had as remaining when starting the 2nd part of Karsten's automated processing at n=1004 (n=1005 for some k's it seems). What that means is that he effectively "missed" those primes, which could have caused k's to be remaining that weren't. Fortunately all of the k's eventually yielded primes before n=2100.

Below are n=1001-1004 primes that were missed when starting the 2nd part of the automated processing at n=1004 (or n=1005). The prime on the left is the one that I found, the prime on the right in paranthesis is the one found by Micha.

500656302*3^1003-1 is prime (prime found at n=1300)
502183398*3^1004-1 is prime (prime found at n=1758)
502572138*3^1002-1 is prime (prime found at n=1785)
503150118*3^1002-1 is prime (prime found at n=1909)
505353468*3^1002-1 is prime (prime found at n=1620)
505663726*3^1001-1 is prime (prime found at n=2091)

Note that the automated process had some n=1004 primes in it so it is very confusing that it did not find the k=502183398 prime at n=1004. My overlap up to n=1005 confirmed the 3 n=1005 primes found by the automated process.

Please be careful when making the transition from one part of the automated process to the next. It would be easy to drop a k that had no prime or include a k that already had a prime. In this case, we included 6 k's that already had primes.

Chris, please continue your double-check process also. Testing to n=1000 is all that I felt I had time for. Your double-check process is needed for checking up to n=25K. This kind of detailed checking is needed for a new automated process such as this.

Thanks,
Gary

michaf · 2008-11-24, 21:26

Hmm.. no offense taken :)

I find it troubling that there are some misses though.

another thing: I just found out the hard way that doing large ranges just isn't doable.
The harddrive thrashes, and just keeps on writing/reading, while the processor is out of work...
I will do all the ranges by 10M now :)

gd_barnes · 2008-11-25, 03:12

Quote:

Originally Posted by michaf

Hmm.. no offense taken :)

I find it troubling that there are some misses though.

another thing: I just found out the hard way that doing large ranges just isn't doable.
The harddrive thrashes, and just keeps on writing/reading, while the processor is out of work...
I will do all the ranges by 10M now :)

Since the missing primes are all n=1001-1004, my impression is that you stopped testing your "1st run" at n=1000 but started testing your "2nd run" at n=1004 with the k's that were remaining at n=1000. Perhaps it was something to do with having to re-run the "2nd run" as a result of only doing the k=500M-502M range the 1st time around. Although that still doesn't explain why the 2nd run found 2 primes for n=1004 but missed 1 other prime for n=1004.

Can you check where you started and stopped your 2 runs? We need to know if there is a bug in Karsten's process or if it was just a "user error".

Gary

henryzz · 2008-11-25, 07:12

Quote:

Originally Posted by henryzz

unfortunately this has not worked
it produced a file with n values in the order

Code:

etc.
which is not very easy to split

once i have a solution for this i will be able to send my results to gary

Flatlander · 2008-11-25, 15:32

Gary, either everyone here thinks you are infallible or nobody reads your posts!

Quote:

Something does smell a little bit fishy here though:

Code:

k-range   k's remaining
500M-505M    8
505M-510M   27
510M-515M   15
515M-520M   30

Er, there weren't 8 ks between 500M and 505M in michaf's post, there were 17! I can't belive no-one noticed! lol

I have just confirmed them:

Code:

This is a very reassuring double check.

I'll stop laughing soon.

Flatlander · 2008-11-25, 16:48

Quote:

Originally Posted by gd_barnes

...
I double-checked the removal of multiples of the base (MOB) with your site. Looks great! There were no k's remaining that are divisible by 3. A total of 12 were eliminated leaving 35 k's remaining for the range. Nice work!

Gary

Looks like the joke's on me. lol

michaf · 2008-11-25, 17:28

Quote:

Originally Posted by gd_barnes

Can you check where you started and stopped your 2 runs? We need to know if there is a bug in Karsten's process or if it was just a "user error".

Gary

The sieve started at n=1000, but that doesn't imply an error in the script.
I have had some very hard working days, so any user error that is possible to make, will most likely be made :)

gd_barnes · 2008-11-26, 03:38

Quote:

Originally Posted by Flatlander

Gary, either everyone here thinks you are infallible or nobody reads your posts!

Er, there weren't 8 ks between 500M and 505M in michaf's post, there were 17! I can't belive no-one noticed! lol

I have just confirmed them:

Code:

This is a very reassuring double check.

I'll stop laughing soon.

I am infallible. You didn't understand. The # of k's remaining is AFTER removing k's that are divisible by the base that don't need to be searched. Just look at Kenneth's website and you'll see!

Divide all the k's by 3. If a k is divisible by 3, subtract 1 and see if it is composite. If so, remove it; if not keep it. In almost all cases, you will remove it because k/3^q is already being searched or already has a prime.

Ha, ha, ha. Now I laugh times 3!

Gary

Edit: I just now saw that you said the "joke's on me". I can delete your posts and this post if you want but I have to admit I couldn't resist getting in a dig on you. BTW, both Kenneth and Max have caught errors on my web pages before. :-)

Flatlander · 2008-11-26, 14:19

Quote:

Originally Posted by gd_barnes

...
Edit: I just now saw that you said the "joke's on me". I can delete your posts and this post if you want but I have to admit I couldn't resist getting in a dig on you. BTW, both Kenneth and Max have caught errors on my web pages before. :-)

No, I think my public humilation should stand. It might teach me to keep my big mouth shut!

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Bases 251-500 reservations/statuses/primes	gd_barnes	Conjectures 'R Us	2502	2023-01-28 17:08
Bases 33-100 reservations/statuses/primes	Siemelink	Conjectures 'R Us	1751	2023-01-21 20:37
Bases 101-250 reservations/statuses/primes	gd_barnes	Conjectures 'R Us	1010	2023-01-20 17:22
Bases 6-32 reservations/statuses/primes	gd_barnes	Conjectures 'R Us	1420	2022-12-30 17:20
Sierp base 3 reservations/statuses/primes	gd_barnes	Conjectures 'R Us	423	2022-11-24 15:47

2008-11-23, 13:47	#133
michaf Jan 2005 479 Posts	While testing my 80M range (520M-600M) I first tried pfgw to 1000 using Karsten's scripts, then sieve it, and tried Karsten's 'high-n' script That took about 4 minutes writing time when a prime was found, so it was waaaay too slow :) second,I tried to n=2000 with pfgw, sieved, and then high-n'd it, which was way better (30 sec/write). I'm finishing that now as it is quick enough. I think that running pfgw to 2500 or even 3000 is the most efficient. I will test on following ranges what is best (I'll go for 100M ranges then, and will note total time taken. expect an update in about 10 weeks :) )

2008-11-24, 11:38	#135
gd_barnes "Gary" May 2007 Overland Park, KS 2×5×1,181 Posts	I have done some preliminary double-check work on the k=500M-506M range. After getting primes from Micha for n=1004-25000, I decided to just run the entire range up to n=1005 to have an n=2 overlap. Then using the primes for n=1004-25000 that Micha sent me and my remaining k's after my testing, I was able to determine the k's that should be remaining at n=25K. To my surprise, they are correct as listed by Micha. (No offense! ) I'm just amazed that there are only 8 k's remaining at n=25K for k=500M-505M! I did find a few problems with either Micha's testing or Karsten's scripts though. For the range of n=1001-1004, I found 6 primes that Micha had as remaining when starting the 2nd part of Karsten's automated processing at n=1004 (n=1005 for some k's it seems). What that means is that he effectively "missed" those primes, which could have caused k's to be remaining that weren't. Fortunately all of the k's eventually yielded primes before n=2100. Below are n=1001-1004 primes that were missed when starting the 2nd part of the automated processing at n=1004 (or n=1005). The prime on the left is the one that I found, the prime on the right in paranthesis is the one found by Micha. 5006563023^1003-1 is prime (prime found at n=1300) 5021833983^1004-1 is prime (prime found at n=1758) 5025721383^1002-1 is prime (prime found at n=1785) 5031501183^1002-1 is prime (prime found at n=1909) 5053534683^1002-1 is prime (prime found at n=1620) 5056637263^1001-1 is prime (prime found at n=2091) Note that the automated process had some n=1004 primes in it so it is very confusing that it did not find the k=502183398 prime at n=1004. My overlap up to n=1005 confirmed the 3 n=1005 primes found by the automated process. Please be careful when making the transition from one part of the automated process to the next. It would be easy to drop a k that had no prime or include a k that already had a prime. In this case, we included 6 k's that already had primes. Chris, please continue your double-check process also. Testing to n=1000 is all that I felt I had time for. Your double-check process is needed for checking up to n=25K. This kind of detailed checking is needed for a new automated process such as this. Thanks, Gary

2008-11-24, 21:26	#136
michaf Jan 2005 479 Posts	Hmm.. no offense taken :) I find it troubling that there are some misses though. another thing: I just found out the hard way that doing large ranges just isn't doable. The harddrive thrashes, and just keeps on writing/reading, while the processor is out of work... I will do all the ranges by 10M now :)