PARI's commands

[axn]

Originally Posted by 3.14159

I guess I'll try once again..

.. No luck. It doesn't work for anything > 100 digits.

http://en.wikipedia.org/wiki/Hasty_generalization

[3.14159]

Originally Posted by axn

http://en.wikipedia.org/wiki/Hasty_generalization

In that case, I respond to you with:

"Insanity: Trying the same thing over and over and expecting a different result."

A nice addition to YAFU would be an addition of NFS to it. (IMO.)

[kar_bon]

Originally Posted by 3.14159

In that case, I respond to you with:

"Insanity: Trying the same thing over and over and expecting a different result."

I've done it more than once without any probs.

[3.14159]

Originally Posted by kar_bon

I've done it more than once without any probs.

.. It must be the bastard computer, then..

Oh, not really..

Originally Posted by YAFU documentation

post-
processing courtesy of Jason Papadopoulos's msieve block Lanczos implementation.
Useful for inputs < 105 digits.

100-105 is the best I can hope to do, in that case.

[mdettweiler]

Originally Posted by 3.14159

.. It must be the bastard computer, then..

Oh, not really..



100-105 is the best I can hope to do, in that case.

Note that useful != possible. While YAFU becomes not particularly useful over 105 digits (because the time it takes to factor a number with QS at that size grows quite a bit more quickly than that for GNFS), it can still be used if you really want to. As an example, just for grins the author of YAFU set a QS record a year or so back by using it to factor a 117 digit (size IIRC) number.

BTW, I'm assuming that you never were able to get GGNFS working? If you'd like, I can zip up my working copy of it and send it to you; I have a 32-bit Windows system so it should work for you as-is.

[3.14159]

Originally Posted by mdettweiler

Note that useful != possible. While YAFU becomes not particularly useful over 105 digits (because the time it takes to factor a number with QS at that size grows quite a bit more quickly than that for GNFS), it can still be used if you really want to. As an example, just for grins the author of YAFU set a QS record a year or so back by using it to factor a 117 digit (size IIRC) number.

BTW, I'm assuming that you never were able to get GGNFS working? If you'd like, I can zip up my working copy of it and send it to you; I have a 32-bit Windows system so it should work for you as-is.

The former; The thing crashes when I try a number that size!

The latter; It would be dearly appreciated.

I just made a nice text file of composites (What YAFU would class as RSA, ranging from 80-digit numbers to 160-digit numbers.)

Maybe I can zip up the composites and send them to you, in case you're interested in doing a bit of work. (Be warned; It's 8400, 8600, or 8800 composites, and surprisingly, it fit into less than 1 MB.)

About GGNFS; Let's say an n-digit number takes a certain amount of time, x. How many digits would need to be added to double the amount of time it takes to factor?

I'm guessing.. about 3-5.

[mdettweiler]

Originally Posted by 3.14159

The former; The thing crashes when I try a number that size!

The latter; It would be dearly appreciated.

Cool, will do. I'll put the package together and send it to you (via PM as a sendspace link) as soon as I can get the chance.

I just made a nice text file of composites (What YAFU would class as RSA, ranging from 80-digit numbers to 160-digit numbers.)

Maybe I can zip up the composites and send them to you, in case you're interested in doing a bit of work. (Be warned; It's 8400, 8600, or 8800 composites, and surprisingly, it fit into less than 1 MB.)

No thanks--my CPUs are mostly booked to the end of the year. Also, while factoring random numbers can have an appeal when you're learning how to do factoring, it's sufficiently "old hat" to me by now that if I'm going to factor a number, I'd prefer it to be one that's going to be useful somewhere.

About GGNFS; Let's say an n-digit number takes a certain amount of time, x. How many digits would need to be added to double the amount of time it takes to factor?

I'm guessing.. about 3-5.

Just about right on. The rule of thumb is that factoring time doubles every 4 digits. (In reality it's a little more complex than that, but it works pretty well for an approximation.)

[3.14159]

Originally Posted by Max

No thanks--my CPUs are mostly booked to the end of the year. Also, while factoring random numbers can have an appeal when you're learning how to do factoring, it's sufficiently "old hat" to me by now that if I'm going to factor a number, I'd prefer it to be one that's going to be useful somewhere.

What are you actively working on? I don't remember all of the active projects..

Originally Posted by Max

Cool, will do. I'll put the package together and send it to you (via PM as a sendspace link) as soon as I can get the chance.

Alright, seems fair enough.

Woohoo. A prime-numbered post. (1499)

[bsquared]

Originally Posted by mdettweiler

Note that useful != possible. While YAFU becomes not particularly useful over 105 digits (because the time it takes to factor a number with QS at that size grows quite a bit more quickly than that for GNFS), it can still be used if you really want to. As an example, just for grins the author of YAFU set a QS record a year or so back by using it to factor a 117 digit (size IIRC) number.

It was a C130, back in feburary '09. That was when I had access to a university opteron cluster.

Originally Posted by 3.14159

The former; The thing crashes when I try a number that size!

That would be news to me... but I haven't tested number that big in quite a while, so its possible something broke. I'll have a look.

[3.14159]

Originally Posted by bsquared

That would be news to me... but I haven't tested number that big in quite a while, so its possible something broke. I'll have a look.

It's probably just me.. The number I intended to split was 54226171049950775045015304584853623265770119570639252888707467753514344661290463063582064327739908372829. (104 digits)

I already collected all the relations (140700 relations); It's the post-processing which fails.

Also; Wasn't the record for QS a 135-digit split, into a p66 and p69? I think it was for a cofactor of 2⁸⁰³ - 2⁴⁰² + 1.

[mdettweiler]

Originally Posted by 3.14159

What are you actively working on? I don't remember all of the active projects..

Mostly No Prime Left Behind (searching for general primes of the form k*2^n-1) and Conjectures 'R Us (proving Sierpinski/Riesel conjectures for bases <=1030). I have two machines, a quad and a dualcore; the quad is not as readily accessible, so on it I do strictly primality-testing work via PRPnet (currently for CRUS). The dualcore is my personal machine, so I do "special" jobs on it--currently one core is doing sieving for the Twin Prime Search and the other is running BOINC for PrimeGrid attempting to collect a full set of score badges.