[quote=Mini-Geek;194600]At least now I know to remove k's like that. (it [I]is[/I] that k is a multiple of the base and k-1 is composite, right?) I vaguely knew about multiple of base, but guessed that the 'modular' script calls would take care of that.

Hmm...well, I did make one modification that [I]might[/I] make a difference (don't think so). Karsten's do3.bat was:
[code]@echo off
call doranges 3 1 1
call modular 3 9 0
call modular 3 27 0
call modular 3 81 0
call modular 3 243 0
call modular 3 729 0
call modular 3 2187 0
call modular 3 6561 0
call modular 3 19683 0
call modular 3 59049 0
call modular 3 177147 0
call modular 3 531441 0
call modular 3 1594323 0
call modular 3 4782969 0
call modular 3 14348907 0
call modular 3 43046721 0
call doranges 3 2 1000
[/code]I got rid of all the 'modular' lines after the first one, for when I did 571-575, since when doing 570-571 it removed none after the first line. I supposed it's because 27, 81, ... are all multiples of 9 and so they're redundant. AFAICT nothing there was supposed to remove k's where b is divisible by k and k-1 is composite.
Removing all k's divisible by 9 wasn't wrong, was it? I just assumed Karsten's files would be set up to work right.[/quote]

Automatically removing all k's that are divisible by 9 is incorrect. I don't understand this script so I can't tell you what Karsten was trying to do. You can't ever just automatically remove multiples of anything. k-1 (for Riesel) or k+1 (for Sierp) must first be tested for compositness before removal.

The problem is that it is possible that you have removed a k that should be remaining if that k is divisible by 9 and k-1 is prime. If I understand you correctly, that is a problem and the range needs to be rerun.

Suggestion: Using Willem's automated script that has just now been posted, please rerun the range and confirm that there are 17 k's remaining.

Sorry.

I'll reiterate to everyone again here: If you are not clear on multiples of the base as it relates to removing k's, base 3 is the most difficult possible base that you can choose to run. Please consider a higher base where k's that are multiples of the base are far less common until you are 100% clear on how it works.

Personally, I wish base 3 would just go away. It's just a big problem child. lol


Gary

[quote=gd_barnes;194758]The problem is that it is possible that you have removed a k that should be remaining if that k is divisible by 9 and k-1 is prime. If I understand you correctly, that is a problem and the range needs to be rerun.[/quote]
I see...would it be sufficient to rerun k's divisible by 9 where k-1 is prime? If I understand this all correctly, it ought to be.

I'm most likely not the only one that was confused by base 3, considering Karsten's scripts, by default, do exactly what I did. If many people used Karsten's scripts, or their own wrong-in-similar-or-different-ways scripts, then potentially a very large portion of base 3 is missing k's divisible by 9 with k-1 prime.

[quote=Mini-Geek;194805]I see...would it be sufficient to rerun k's divisible by 9 where k-1 is prime? If I understand this all correctly, it ought to be.[/quote]
I've done this, and there were no missed Riesel numbers in my range (i.e. all k's matching the above criteria had a prime before 25k). I've e-mailed primes and the details of how I got my list of k's (just in case something was wrong) to Gary.

the script you mentioned in post #258 i gave was incorrect (first given here: [url]http://www.mersenneforum.org/showpost.php?p=148625&postcount=22[/url])

i warned in a later post not to use that script (see [url]http://www.mersenneforum.org/showpost.php?p=148815&postcount=26[/url]).

the new script in post #259 was done later when i corrected the issues. (see [url]http://www.mersenneforum.org/showpost.php?p=185103&postcount=58[/url])


so to all:

read carefully the whole thread of those scripts here: [url]http://www.mersenneforum.org/showthread.php?t=10864[/url]

I remember the issue. I'm quite confident that we reran or double-checked k's ran with the incorrect script.

Tim, what you did now sounds good and I got your Email of additional primes. I'm sorry you ended up with an older script there causing you extra work.

For my own comfort level, I'm rerunning the k-range up to n=5000 using a good old standard PFGW script and nothing automated. That should take a little < 1 day. I'll then take the k's remaining and manually remove appropriate multiples of the base, use your primes for n>5000 to remove more k's, and see if what is remaining is what you've posted here.


Gary

[QUOTE]Attached is my script. It can be used for any base, it filters out the multiple of that base. You can set the range for the k and the depth for the n. It gives two output files, one with all the primes/prps found (can be surpressed) and one output file with the remaining k.[/QUOTE]

Willem, I have a question about your rieselator.txt.

In the sequence:

# Can k be divided by the base?
IF (k % base > 0) THEN GOTO Prime_k_test
PRP k-1
IF !(ISPRIME) THEN GOTO Next_k

Why is the k being skipped if k-1 is prime? I thought that if the k is divisible by the base and k-1 (+1 for Sierp) is prime, then the k needs to be further tested.

Shouldn't it read IF!(ISPRIME) THEN GOTO Set ELSE GOTO Next-k

Thanks, Ian

I'll take 575M-580M. Now that I know what can be eliminated and what can't, I'm much more confident that I can make sure everything's doing what it's supposed to. :smile:

[QUOTE=MyDogBuster;194931]Willem, I have a question about your rieselator.txt.

In the sequence:

# Can k be divided by the base?
IF (k % base > 0) THEN GOTO Prime_k_test
PRP k-1
IF !(ISPRIME) THEN GOTO Next_k

Why is the k being skipped if k-1 is prime? I thought that if the k is divisible by the base and k-1 (+1 for Sierp) is prime, then the k needs to be further tested.

Shouldn't it read IF!(ISPRIME) THEN GOTO Set ELSE GOTO Next-k

Thanks, Ian[/QUOTE]

Are you troubled by double negatives? It skips when k-1 is not prime.

# Can k be divided by the base?
IF (k % base > 0) THEN GOTO Prime_k_test
PRP k-1
IF !(ISPRIME) THEN GOTO Next_k

in plain text:
1) If k is not a multiple of base then it needs to be tested
2) If k is a multiple of base then k = bm and kb^n = bmb^n = mb^(n+1)
3) because we are testing k in ascending order m has already been tested
4) so we must have found mb^x-1 = kb^(x-1)-1 = prime
5) We discard tests for n = 0
6) combining 4) and 5) means that when mb^x-1 with x = 1 equals k-1 = prime.
7) then we can not rely on the previous value and a new one is needed.

Helpfull? Willem.

Thanks Willem, the double negative threw me off.

Done, 19 k's remaining.
[CODE]575375854
575501228
575679536
575782156
576836434
577207216
577302326
577490078
577707926
577756162
577945388
578049356
578270176
578322182
578432578
579044272
579374758
579402968
579796304[/CODE]

Status update :)
 

Regarding my Riesel base 3 reservation for k=3677878, my status as of today is n=~890k. There is less than 1800 testings remaining and as of today, I've begun preparing and sieving a NewPGen format sievefile going from n=1M to n=1G. It appears that sieving to optimal sievedepth once the 1M reservation completes, will leave less than 20K k/n pairs per million n's.

I was considering to sieve and test the remaining k's from k=200M to k<=700M (2355 total), from n=25K to n=100K. However sieving goes incredibelly slow, since I'm not able to make srsieve write the factors it finds for a given range to a factor file, which means that I can only sieve on 1 core. So can anyone tell me how to make the latest version of srsieve write a factor file? I should add that I've used the "-f" flag, which doesn't cause any mistakes, and doesn't produce a factor file. Using "-f factors.txt" gives an error were srsieve tells me that it cannot open input file. So can anyone give me a plain and simple example on how to get srsieve to crunch some numbers and write the found factors to a factors file?

If I recieve something helpfull knowledge, I will reconsider doing the 2355 k's and take them from n>25K to n<=100K else I'll continue my sieve range project for k=3677878. I btw thinks it's easier if just 1 person crunches the entire range, since it will produce a lot of primes and no doubleprimed k's with the "for most 1 primes" function activated.

On a side note. Sierp base 63 is about 33 % complete on the Dual core, and Sierp base 82 is complete to n=25K, with 91 individual k's remaining at n=25K. I also btw had 230 k's remaining at n=2500 so no mistakes to report :smile:

Hope to hear from someone with the prober knowledge. Should btw add, that the Sierp base 82 k's is going to be send together with the Riesel base 3 completed range once it gets to n=1M in just a few weeks.

Regards

KEP