mersenneforum.org  

Go Back   mersenneforum.org > Factoring Projects > Aliquot Sequences

Reply
 
Thread Tools
Old 2010-08-08, 02:40   #1
Greebley
 
Greebley's Avatar
 
May 2009
Dedham Massachusetts USA

3·281 Posts
Default Minor changes to a lot of sequences

I am trying to get up to date results from the database. It can mean doing the quick ecm on seqs, and I may need to advance several sequence a few indices to see if a sequence continues past a possible trouble spot.

So hopefully I won't step on anyones toes, but if you see small changes to a sequence that is reserved, it may be me. I won't make long runs on a sequence without reserving it.

I just don't want anyone to worry that someone else is competing on their sequence.

If there are issues, then post here so I can try to avoid them.

---

As an aside there are more errors in the database than I hoped that are hard to find. It may take a good while to compile a full set of sequences and I am not going to be as sure as I would like that I have all the known data. Mostly this project is insurance against a (more) serious problem with the db. I am hoping things get fixed.
Greebley is offline   Reply With Quote
Old 2010-08-08, 18:32   #2
Andi47
 
Andi47's Avatar
 
Oct 2004
Austria

2·17·73 Posts
Default

Quote:
Originally Posted by Greebley View Post
I am trying to get up to date results from the database. It can mean doing the quick ecm on seqs, and I may need to advance several sequence a few indices to see if a sequence continues past a possible trouble spot.

So hopefully I won't step on anyones toes, but if you see small changes to a sequence that is reserved, it may be me. I won't make long runs on a sequence without reserving it.

I just don't want anyone to worry that someone else is competing on their sequence.

If there are issues, then post here so I can try to avoid them.

---

As an aside there are more errors in the database than I hoped that are hard to find. It may take a good while to compile a full set of sequences and I am not going to be as sure as I would like that I have all the known data. Mostly this project is insurance against a (more) serious problem with the db. I am hoping things get fixed.
I am currently GNFSing the c139* of alq10212 - if it turns out that the next step(s) are a bit easier to factor, I will (re-)reserve it.

* this might take a bit longer - due to holidays.
Andi47 is offline   Reply With Quote
Old 2010-08-08, 22:13   #3
schickel
 
schickel's Avatar
 
"Frank <^>"
Dec 2004
CDP Janesville

2·1,061 Posts
Default

Quote:
Originally Posted by Greebley View Post
If there are issues, then post here so I can try to avoid them.
No problem here. I just double-checked mine and uploaded everything to the present...
Quote:
---

As an aside there are more errors in the database than I hoped that are hard to find. It may take a good while to compile a full set of sequences and I am not going to be as sure as I would like that I have all the known data. Mostly this project is insurance against a (more) serious problem with the db. I am hoping things get fixed.
After you're done, Wolfgang or Clifford would be the ones to check with wrt all the currently active work we've been doing. Who knows, we might find a couple of things that got missed somehow. All of my status files are so out of date they wouldn't even be worth checking against....

If you need a place to host everything, I can set up a public FTP site here on my end to store things.
schickel is offline   Reply With Quote
Old 2010-08-09, 15:11   #4
Greebley
 
Greebley's Avatar
 
May 2009
Dedham Massachusetts USA

3×281 Posts
Default

Previously I uploaded to sendspace, but that has the problem of being very temporary. It might be nice to have the sequences available for longer.

I think I have figured out a way to know if I have the full sequence - the final entry should always have a large cofactor that the database doesn't have the factors for it, and it has no small factors. I don't think I have seen the database break such that it 'forgets' the factorization of a large cofactor of this form (it can forget the factorization of a full entry but that will be divisible by 2). I will need to come up with a program to check this as checking 10k by hand is too much.
Greebley is offline   Reply With Quote
Old 2010-08-09, 20:17   #5
kar_bon
 
kar_bon's Avatar
 
Mar 2006
Germany

55018 Posts
Default

Quote:
Originally Posted by Greebley View Post
As an aside there are more errors in the database than I hoped that are hard to find. It may take a good while to compile a full set of sequences and I am not going to be as sure as I would like that I have all the known data. Mostly this project is insurance against a (more) serious problem with the db. I am hoping things get fixed.
I've changed my script for checking and finding errors in the DB.
I've checked all sequences for n<100000 and found these errors:

13596: index 835 -> P91 is composite!
76584: index 243 -> P43 is composite!

I have to change my script because all seqs with transitions like
276 index 13: "3^9+1" instead of "19628"
will not be checked at all! (There're 29 of them for n<100000).

This script downloaded/checked all 905 seqs in about 1h 40min!
So the others ranges may take a while (after I've changed the script).
kar_bon is offline   Reply With Quote
Old 2010-08-11, 00:17   #6
Greebley
 
Greebley's Avatar
 
May 2009
Dedham Massachusetts USA

84310 Posts
Default

5148 should be coming up as a broken sequence.
Greebley is offline   Reply With Quote
Old 2010-08-12, 12:24   #7
kar_bon
 
kar_bon's Avatar
 
Mar 2006
Germany

55018 Posts
Default

Ok, here is it!

I've wrote some small scripts to download a bunch of seqs, check them and (after some manually work) patch them automatically!

How it works:

1. Giving a range of seqs to download, a script determines them and download all to the local PC.

2 Running another script will create a checking-script (creating, so it can be used more than once) and will result in a file with errors found.

3. The manually step is to create a patch-file which will eliminate the small-primes issue and other things.

4. After 3. is done, apply the patches to the downloaded seqs and run step 2 again.

This will give those seqs in the desired range without any error!

I've done this for all seqs < 100000 (902 currently) which give about 172 MB of ELF-files all over!

Timings:
Step 1 takes about 1h 40min for those 900 seqs.
Step 2 only a few minutes.
Step 3 is timeconsuming, because much manually work but this has to be done only once!
It took me several hours to bring the FactorDB and all ELFs to the format needed.

Patching:
I've patched the program "chl.exe" from C.Stern for checking ELF-files:
If an incomplete factored line (only small primes present) in the last line, chk.exe writes "Computation error" and give the line.
I've delete the text and only display the last line (see below why to do so!).

I'm using several UnxUtils to create patches.

The program "diff.exe" is able to create a file with differences of 2 text-files.
Example:
Code:
--- 276_.elf	Wed Aug 11 09:12:27 2010
+++ 276.elf	Wed Aug 11 09:12:58 2010
@@ -14 +14 @@
-13.   3^9+1 = 2^2 * 7 * 19 * 37
+13.   19684 = 2^2 * 7 * 19 * 37
Here the two files "276.elf" and "276_.elf" differ at line 13 -> The short form of the line which the FactorDB sometimes display!

The program "patch.exe" (must rename it "pat.exe" because WIN has a "patch.exe", too) can use this small file and apply the patch automatically!

I've collected all patches in one file and calling "pat.exe -p1 <all.pat" will patch all downloaded ELF-files without any further manually work.

Notes:
It's highly recommended, that the FactorDB do not contain an unfactored line! Only one call of "Quick ECM" there is needed (-> small factors present).
But this is in many cases not available, bacause of the automatism of calculating up to say 110 digits and submitting results.

If anyone is interested, I can give those scripts.

PS: The patched files eleiminates also the errors of the 2 broken seqs given in post #5 so everyone can continue with them!

Last fiddled with by kar_bon on 2010-08-12 at 12:25
kar_bon is offline   Reply With Quote
Old 2010-08-12, 12:56   #8
Andi47
 
Andi47's Avatar
 
Oct 2004
Austria

2×17×73 Posts
Default

Quote:
Originally Posted by kar_bon View Post
Ok, here is it!

I've wrote some small scripts to download a bunch of seqs, check them and (after some manually work) patch them automatically!

How it works:

1. Giving a range of seqs to download, a script determines them and download all to the local PC.

2 Running another script will create a checking-script (creating, so it can be used more than once) and will result in a file with errors found.

3. The manually step is to create a patch-file which will eliminate the small-primes issue and other things.

4. After 3. is done, apply the patches to the downloaded seqs and run step 2 again.

This will give those seqs in the desired range without any error!
I guess, you mean that it gives the error-free seqs in a local copy - or do you know any method to eliminate the "small composites" issue in the DB?
Andi47 is offline   Reply With Quote
Old 2010-08-12, 13:24   #9
kar_bon
 
kar_bon's Avatar
 
Mar 2006
Germany

55018 Posts
Default

Quote:
Originally Posted by Andi47 View Post
I guess, you mean that it gives the error-free seqs in a local copy - or do you know any method to eliminate the "small composites" issue in the DB?
Oh, I thought this was clear! Sure only local error-free ELF-files!
kar_bon is offline   Reply With Quote
Old 2010-08-12, 14:42   #10
Greebley
 
Greebley's Avatar
 
May 2009
Dedham Massachusetts USA

3·281 Posts
Default

One thing to watch out for is that the database has a completely factored line, but there is no next line of the sequence. Certain sequences have a lot of these. If you get a sequence from the database and all terms are factored you need to compute the next value of the aliquot sequence and then get the aliquot sequence for the new next value. This may have to be repeated 10 or more times for the worst offenders.
Greebley is offline   Reply With Quote
Old 2010-08-12, 14:57   #11
kar_bon
 
kar_bon's Avatar
 
Mar 2006
Germany

55018 Posts
Default

Quote:
Originally Posted by Greebley View Post
One thing to watch out for is that the database has a completely factored line, but there is no next line of the sequence. Certain sequences have a lot of these. If you get a sequence from the database and all terms are factored you need to compute the next value of the aliquot sequence and then get the aliquot sequence for the new next value. This may have to be repeated 10 or more times for the worst offenders.
Therefore I've collected all patches needed in one file. This can be extended with new ones if other errors occur!

So this has to be done only once and could be offered for all users here!

I've uploaded a test with only the seq for 276 here; it's about 400kB so to big for an attachment here.

Only call "do.bat" (WIN) and the ELF-file will be downloded from DB and patched!

If all is done and ok, the file "diff.txt" only contains the seq-number (here "276"); for others errors, there's more text.
kar_bon is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Any interest in all sequences/open sequences? Greebley Aliquot Sequences 6 2012-04-07 10:06
A minor typo Chuck GPU to 72 1 2011-12-12 13:36
Minor GMP-ECM bug jasonp GMP-ECM 2 2007-11-25 18:40
Minor bug PhilF Software 1 2006-03-22 01:04
A minor bug in PRP-24.14 Kosmaj 15k Search 3 2005-08-29 20:28

All times are UTC. The time now is 03:59.

Wed Apr 14 03:59:24 UTC 2021 up 5 days, 22:40, 0 users, load averages: 2.55, 2.12, 2.06

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

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.