mersenne.ca

[alpertron]

Originally Posted by James Heinrich

Not false positives, the result clearly says no factor from 2^82 to 2^83 and there is a known 82.568-bit factor. How else am I supposed to interpret this?
If the user did indeed factor from 82-83 complete and did not stop after finding the factor, the it should be reported as a completed TF-F run, with factor, not as TF-NF.

According to https://www.mersenne.ca/exponent/726064763, the factor was found on 2021-12-28T12:50:52, but the TF was done three days later. I would expect the text "skipped known factors(?)" to appear.

[James Heinrich]

Originally Posted by alpertron

the factor was found on 2021-12-28T12:50:52, but the TF was done three days later.
I would expect the text "skipped known factors(?)" to appear.

It does now, thank you.

[Zhangrc]

Thank you, James!
https://www.mersenne.ca/exponent/13291111

[LaurV]

Wow! That is a lot of work done there! Hat off and bow at James.

Some nitpicking: In the "p-1 missed" you could also show who missed the factor, as you do in the "TF missed" report. Which report, by the way, should be called "factors blah blah TF" and not "factors blah blah P-1".

Crazy ideas (because they need some more work on your side): In the user factors reports, you could transform the headers into switches (i.e. toggle the "nfs/ecm/p-1/p+1" on/off when clicked, and turn them all on when clicking "any", or something like that) - for example, if we want to see only the p-1 AND p+1 factors, but not the others (yeah, nfs are quite troublesome, because they are so big, beside of the fact that they kill the table format, because the digit strings are so long, they are demotivating too , how can I fight for a 50 digits factor when some guys got a 200 digits factor? and who care about nfs? hihi, I want to turn them off!)

[Zhangrc]

https://www.mersenne.ca/prob.php?exp...000&b2=5000000 yields 9.918380%, while https://www.mersenne.ca/prob.php?exp...000&b2=5000000 yields 9.998854%.
Purely theoretical, but B1=B2=5000000 should have a higher probability than B1=2000000, B2=5000000.
I'm aware that these formulas are based on experiments and extrapolations, but it always seems to underestimate stage 1 factoring probability.

[James Heinrich]

The probability code is not my invention, it was written by Mihai and adopted by George.

Totally unrelated, I have just become aware my graph images haven't updated for several days, am away right now, will try to fix by Monday.

[LaurV]

Originally Posted by LaurV

"TF missed" report. Which report, by the way, should be called "factors blah blah TF" and not "factors blah blah P-1".

@James: bump

[James Heinrich]

Originally Posted by LaurV

@James: bump

As you know, I've been (and still am) away for a week, will look at all these things starting tomorrow.

[James Heinrich]

Originally Posted by LaurV

Some nitpicking: In the "p-1 missed" you could also show who missed the factor, as you do in the "TF missed" report. Which report, by the way, should be called "factors blah blah TF" and not "factors blah blah P-1".

Copy-paste typos fixed. But I can't easily replicate the missed-by-whom for P-1 due to the way the data is stored and the report generated, sorry.

Originally Posted by LaurV

Crazy ideas (because they need some more work on your side): In the user factors reports, you could transform the headers into switches (i.e. toggle the "nfs/ecm/p-1/p+1" on/off when clicked, and turn them all on when clicking "any", or something like that) - for example, if we want to see only the p-1 AND p+1 factors, but not the others (yeah, nfs are quite troublesome, because they are so big, beside of the fact that they kill the table format, because the digit strings are so long, they are demotivating too , how can I fight for a 50 digits factor when some guys got a 200 digits factor? and who care about nfs? hihi, I want to turn them off!)

I haven't quite gotten around to making a pretty interface for it, but you can now specify the comma-separated list of factoring methods (from list: nfs ecm pm1 pp1) like this:
https://www.mersenne.ca/userfactors/ecm,pm1/1/bits

[kruoli]

How difficult would it be to create a report that shows who found the most factors on one exponent? Second sort criterium would be the amount of "n factors per exponent" the user found.

I would assume the first place would be no more than three or four exponents in one page.

@James: I looked at your export page and found no user records in that data to do it myself, was I overlooking something?

[James Heinrich]

Originally Posted by kruoli

How difficult would it be to create a report that shows who found the most factors on one exponent? Second sort criterium would be the amount of "n factors per exponent" the user found.

Not particularly easy. The data is primarily stored paying attention to the factor and exponent, only when you look at a specific exponent is the query performed to look up historical records as to who-did-what on that specific exponent. This works acceptably well when querying a single exponent but isn't particularly friendly to broad queries against username that would be involved in compiling such a list.