Factoring database

[RichD]

Originally Posted by RichD

I think the Quick ECM logic has changed again. You may not necessarily get all the factors within 4 seconds. Once it finds a factor (say > 2000?) it stops. You have to submit again on the remaining composite. But, if it can't find a single factor quickly, you do get more time than 4 seconds. In fact, quite a bit more it appears.

OK, someone made a liar out of me again. :-)

The Quick ECM just pulled a p8, p15, p18, p21 out of a c61 in a single pass.

[Mini-Geek]

Originally Posted by RichD

Ah, you might have a point Mini-Geek. I am logged in and the trial I was performing had all the other factors listed above the Quick ECM. In another words, I was going down the line picking primes off but left the c14 for last.

I've also noticed that it often fails on very small inputs, below about 20 digits. I'm not sure why. I'm sure that msieve can break them ridiculously easily. Maybe there's some problem with finding the factors in the msieve output when the number crunches too fast?

Originally Posted by RichD

OK, someone made a liar out of me again. :-)

The Quick ECM just pulled a p8, p15, p18, p21 out of a c61 in a single pass.

I've also noticed that it usually just pulls one factor at a time, but this could just be due to the nature of ECM, and more particularly the parameters Syd uses.

[10metreh]

Originally Posted by Mini-Geek

I've also noticed that it often fails on very small inputs, below about 20 digits. I'm not sure why. I'm sure that msieve can break them ridiculously easily. Maybe there's some problem with finding the factors in the msieve output when the number crunches too fast?

error: tiny factoring failed

If you have ever encountered this error, you will know that it only occurs on numbers below about 26 digits. It's rare, but it does occur. This may explain the problem.

[henryzz]

is there a problem when msieve finds a prime factor as apposed to a prp factor?
that will only happen with factors less than the trial factoring bound squared

[RichD]

Originally Posted by jasonp

Msieve uses trial division and pollard rho first, no 6-digit factors would survive that. Of course if the input is not factored msieve will then happily spend huge amounts of time trying to make further progress...

Originally Posted by henryzz

is there a problem when msieve finds a prime factor as apposed to a prp factor?
that will only happen with factors less than the trial factoring bound squared

huh??

[Andi47]

Originally Posted by RichD

huh??

When msieve finds a probable prime, the output says something like "prp = xx".
when it finds a (proved) prime, the output says something like "p = xx"

These are two different things, especially if you have to parse the line.

[RichD]

Originally Posted by Andi47

When msieve finds a probable prime, the output says something like "prp = xx".
when it finds a (proved) prime, the output says something like "p = xx"

These are two different things, especially if you have to parse the line.

I guess I am from the old schooling where memory (common bank) transfers were the norm for critical results presentation.

[gd_barnes]

Aliquot sequence 10^53 has the squared line bug.

I tried the "repair sequence" link many times over many days and it has failed to correct the problem.

Can someone look into this?

BTW, the "quick ECM" button is really cool. Although it's quirky at times, sometimes it takes several attempts to crack a small factor, there are other times it will find factors > 30 digits. I've gotten to where I automatically hit it anywhere from 3-5 times if it hasn't found all of the factors. Frequently it will find a 20-30 digit factor on the 2nd or 3rd attempt.

Is there a time frame when we're going to get some more workers so that we can do some larger curves online?


Thanks,
Gary

[dekcarki]

Originally Posted by gd_barnes

Aliquot sequence 10^53 has the squared line bug.

I tried the "repair sequence" link many times over many days and it has failed to correct the problem.

Can someone look into this?

- go to the last merge 228805106151339279520
- "repair sequence"

This should work with every sequence which merges with another seq.

[akruppa]

Paul Zimmermann told me that he has entered all factors of Aliquot sequences he has (partly listed on http://www.loria.fr/~zimmerma/records/aliquot.html) to the factor database.

Alex

[Andi47]

Just noticed: When I click "Quick ECM", the ECM curves don't appear under "work done".