[QUOTE=lorgix;244575]I sometimes compute more than I think.

In neither of the above cases was the input expression used the number stated above.

In the case of 18*74^74+1 the input expression was probably
(18*74^74+1)/9332294972008202483366334827

Because that factor was already known.


I'm in the process of learning these things. Which is why I ran curves on a c98 that would probably be efficiently factored gnfs (right/wrong?).

In short; I don't have all the math down. How hardware executes the algorithms I know even less about.

Sorry if I caused unnecessary confusion.[/QUOTE]

Some amount of ECM is warranted on a c98 prior to gnfs or siqs, just probably not an amount wherein you'd be likely to find a p56 or even a p42 factor. A hundred curves or so at B1=1M would probably suffice. The point of yafu's factor() function is to reduce/eliminate the guesswork involved in this balance, especially for newcomers.

What's gotten everyone confused is that the sigma parameter allows one to directly reproduce an ECM factorization. For example in yafu (which it appears that you are using), you could do

[CODE]./yafu "ecm((2*58^58+1)/27/121/41)" -B1ecm 6000000 -sigma 2213424655 [/CODE]

and have the factors more or less immediately pop out. But it doesn't. So I, for one, am left wondering if there is a bug in yafu, or if that sigma has been incorrectly posted, or if there is a discrepancy between that sigma and the number it was actually used to factor, or something else entirely...

Can you email me the entire log file, if its still available and not too large (and if you are in fact using yafu to do this ecm work)?

Speaking of which;

I have found at least one factor through Suyama's extension.


P-1 found a factor in stage #2, B1=80000, B2=1500000.
M2139539 has a factor: 50417810253342889667777

50417810253342889667776=2^6*719*46747*2139539*10954717

[B]10954717[/B] > [B]1500000[/B]

Yes, I did use YAFU.

The only log I have is factor.txt.

And the only two lines that relate to this read exactly like this;

[CODE]01/04/11 00:01:47 v1.21 @ PERIMETROS, prp56 = 64254289193526285238781279369796423284005558353861395881 (found in stg2 of curve 103 (thread 0) with sigma = 2213424655)
01/04/11 00:01:47 v1.21 @ PERIMETROS, Finished 103 curves using Lenstra ECM method on C42 input, B1 = 6000000, B2 = 2000000000[/CODE]Should that be interpreted as something other than a p56 being found? Going by what I [I]do[/I] know it [I]is[/I] surprising.

Also strange that it calls the cofactor 'C42'.

[QUOTE=lorgix;244582]Yes, I did use YAFU.

The only log I have is factor.txt.

And the only two lines that relate to this read exactly like this;

[CODE]01/04/11 00:01:47 v1.21 @ PERIMETROS, prp56 = 64254289193526285238781279369796423284005558353861395881 (found in stg2 of curve 103 (thread 0) with sigma = 2213424655)
01/04/11 00:01:47 v1.21 @ PERIMETROS, Finished 103 curves using Lenstra ECM method on C42 input, B1 = 6000000, B2 = 2000000000[/CODE]Should that be interpreted as something other than a p56 being found? Going by what I [I]do[/I] know it [I]is[/I] surprising.

Also strange that it calls the cofactor 'C42'.[/QUOTE]

There *is* a bug in yafu for 64 bit windows systems. GMP-ECM's sigma value is not reported back to the main part of the program correctly. Are you using 64 bit windows, by any chance?

Yes, sir. 64bit Windows 7.

Here are two other sigma/factor-pairs that should be faulty for the same reason;

[CODE]46*73^73+1 5814808224153889249145425916021 sigma = 3059436187 B1 = 1000000, B2 = 100000000

30*75^75+1 40473625354244197601195075450406761 sigma = 2128358885 B1 = 1430000, B2 = 367000000[/CODE]

Ben, is the true sigma value decypherable from what you saw in the code?

[QUOTE=Batalov;244619]Ben, is the true sigma value decypherable from what you saw in the code?[/QUOTE]

No, unfortunately the true sigma value was just thrown away (again, just for win64 systems). :blush:

I'll get the fix out tonight. Sorry about this folks.

[SIZE="1"](but in an effort to make myself feel better, this must not be causing too many issues if I'm only now hearing about it... this bug has been in there a while).[/SIZE]

Would it make sense for posts 100, and 102 through this one, to be moved to the yafu thread?

Also, for what's its worth, I believe that the p56 was really found, even if it can't be proven. I owe you several virtual beers, lorgix, if this prevents you from making it on the top 10 list (however briefly). If you ever find yourself near Rochester, MN, USA, feel free to look me up to make those virtual beers real ones :)

[QUOTE=bsquared;244624]Also, for what's its worth, I believe that the p56 was really found, even if it can't be proven. I owe you several virtual beers, lorgix, if this prevents you from making it on the top 10 list (however briefly). If you ever find yourself near Rochester, MN, USA, feel free to look me up to make those virtual beers real ones :)[/QUOTE]

Don't worry about it.

YAFU is under development, right? Is [I]anyone[/I] using it without knowing that?

I do like beer though.


Thanks for all the effort you're putting into YAFU!

I downloaded 1.22.

Fun fact: We're ~4600miles apart. But if I'm ever in the states I'd probably like the climate in MN; don't like heat very much.

Thanks for your understanding.

Yesterday morning when I got up, it was -22C outside. Maybe that contributes to why I like factoring - the computer doubles as a nice space heater for most of the year :smile: