Factor found that should have been found by P-1

[tha]

One of my machines found a factor through regular factoring, while that factor should have been found by P-1 before:

http://www.mersenne.ca/exponent/12067273

[TheMawn]

At first glance I thought you were wrong because there wasn't any record of P-1 on Primenet but there appears to be some on mersenne.ca (the link you provided). The P-1 should have found that factor during stage 2 according to the work submitted by the unknown user.

[VictordeHolland]

P-1 misses are not so rare, there are some 200+ known:
http://www.mersenne.ca/p1missed.php

Probably a too small FFT size was used or faulty hardware.

[Madpoo]

Originally Posted by VictordeHolland

P-1 misses are not so rare, there are some 200+ known:
http://www.mersenne.ca/p1missed.php

Probably a too small FFT size was used or faulty hardware.

Yup, no record on mersenne.org of a P-1 being run on it. At least, the clients assigned that exponent never checked anything in indicating that P-1 had been done at all.

[James Heinrich]

Originally Posted by Madpoo

Yup, no record on mersenne.org of a P-1 being run on it. At least, the clients assigned that exponent never checked anything in indicating that P-1 had been done at all.

But I didn't make up those bounds. Before Tha found that factor the mersenne.org report would have said no-factor-to=66, and P-1 bounds (like this). Unfortunately the mersenne.org report suppresses the no-factor-to and max-p1-bounds data if a factor is known. Nothing shows up on the report because mersenne.org doesn't have a specific record of who did the P-1 and when, just what bounds it was done to. The oldest P-1 run I have on file is for M18,752,249 from 03-Jan-2000, so presumably a 12M exponent would have had its P-1 done several months/years earlier than that, and no records of who-did-what-when were kept at the time (or were kept and have since been lost).

[Prime95]

Originally Posted by Madpoo

Yup, no record on mersenne.org of a P-1 being run on it. At least, the clients assigned that exponent never checked anything in indicating that P-1 had been done at all.

It *may* be possible to retrieve some of that old data. The Primenet server has the old v4 database on it. With some effort, we might be able to create a history entry for each exponent: "Imported from v4 DB: TF to 2^56, P-1 to 150000 / 3000000" or some such. Ditto for LL results.

Personally, I'm not sure that is worth the effort.

[tha]

Just the 'sannerud | laptop' data from the v4 server, if any, would be something. Is there a way to download the v4 data or request it per dvd?

[Prime95]

Originally Posted by tha

Just the 'sannerud | laptop' data from the v4 server, if any, would be something. Is there a way to download the v4 data or request it per dvd?

The v4 server did not keep the detailed log that v5 server does. So, no, there is no way to tell which user found a factor, returned a no-factor result, did the P-1, etc. The LL results table did record the userid and user-provided computer name, but did not record the date of the result (so no help in rediscovering milestone dates).

The reason I brought up the v4 server data, is that it could be used to find P-1 bounds of exponents that have later had a factor found. The v5 server deletes that information from the "exponents with no known factor" table, but is often still available in the detailed results log.

[James Heinrich]

Originally Posted by Prime95

The v5 server deletes that information from the "exponents with no known factor" table

Must it do that? Can we not keep that data for reference?

[Prime95]

Originally Posted by James Heinrich

Must it do that? Can we not keep that data for reference?

It would be tricky to change the current table layout. Numerous stored procedures and PHP code expect an exponent to be in only one of three tables (exponents with known factors, exponents with no known factors, mersenne primes).

I see three options: 1) Fix all the code that expects an exponent to be in only one of the three tables above. 2) Create a new table that stores the data we are currently deleting when a factor is found. 3) Do nothing, live with the current shortcomings.

Option 1 is the cleanest table layout, option 2 is the safest, option 3 is the easiest!

[James Heinrich]

I'm fine with #2, especially if someone else does it