Factoring database

[Andi47]

Does the database check if reported factors are composite?

[10metreh]

Originally Posted by Andi47

Does the database check if reported factors are composite?

It does a probable primality test, yes. I haven't found any composite factors for the database yet, but all the other numbers are tested, so I presume reported factors will as well.

[Andi47]

The workers seem to have problems with highly composite (and very smooth) (co-)factors that contain lots of p4 factors - it seems that "ecm to low limits" finds the whole lot of tiny factors and thus can't split the number. Perhaps the trial factoring limit should be raised to 10000.

[jasonp]

Msieve trial factors to 100k...

[10metreh]

Originally Posted by Andi47

The workers seem to have problems with highly composite (and very smooth) (co-)factors that contain lots of p4 factors - it seems that "ecm to low limits" finds the whole lot of tiny factors and thus can't split the number. Perhaps the trial factoring limit should be raised to 10000.

I'd say 1e6. The ECM system is odd as well. I'd just do TF to 1e6, then do ECM to 15, 20, 25 etc. digits followed by QS/NFS.

BTW, is there a factor table for numbers near a googol?

[Andi47]

Sometimes when I report a (valid) factor, I get an error message:

Error: Expression (^ is infix)

In most cases the factor is accepted anyway, but it's annoying when I get the error message every time when I import a factor of M10080 (and for some other Mersennes).

[Andi47]

Originally Posted by henryzz

have you added looking for algebraic factorizations

Originally Posted by Syd

Not yet, maybe later.

I extend this question to Aurifeuillians. And the database does not recognize them:

2^1925-2^963+1

Short forms
No short forms found

[10metreh]

Syd, I suggest looking at the code for Alpertron's applet and finding the part which deals with algebraics and Aurifeuillians. It might help.

Line 3 in post #27 (near-Cunninghams) has gone unanswered.

[Syd]

Thank you very much for your comments, they show up a lot more bugs than i expected.

First - the one i accidently missed:

How much are you doing on near-Cunningham numbers (numbers of the form ?

The template for these numbers exists, thats all so far. These numbers are the next ones to import once the other bugs are fixed.

I got a similar problem earlier. When I submitted a factor that finished the number, the submitted factors appeared in the form "Factor submitted: XXX" and in the factorization but the number was still "Composite, factors known" rather than "Composite, fully factored". If I searched for that number again the problem would be sorted.

Currently the number status is not updated during the page request itself, but right after it. This is indeed confusing, will change it soon.

Does the database check if reported factors are composite?

Yes, they are tested if <= 1000 digits.

Error: Expression (^ is infix)

Thats a wired one, couldn´t track it down yet.

I extend this question to Aurifeuillians. And the database does not recognize them:

It wont recognize most of them yet. Thats another todo point.


The factoring limits: Will add this to the workers


Thank you again

Syd

[10metreh]

Is there a factor table for the near-Cunninghams? I couldn't find one.

When "assigned to worker X (sieve)" appears, does this mean SIQS and what programs are you using? I saw it appear on a C97 that was far easier by SNFS.

[Syd]

Originally Posted by 10metreh

When "assigned to worker X (sieve)" appears, does this mean SIQS and what programs are you using? I saw it appear on a C97 that was far easier by SNFS.

I use msieve (MPQS) for sieving. The "old" code was quite buggy, it sometimes even assigned C>100 to msieve, without doing ECM first. The limit should be at about 80 digits, maybe its already working now.

Syd