fond of a factor? Urn yourself to become remains

[LaurV]

Originally Posted by James Heinrich

Unfortuantely (sic!) that's not at all uncommon, due in no small part to a buggy P-1 implementation in early versions of Prime95.

Scrolling (page changing) there doesn't work. Also: many old/small stuff was not found by P-1 because it was already found by TF to low limits (in spite of the fact that the middle of the table is not visible, the beginning and the end is, because the sorting by size/whatever, works well).

[James Heinrich]

Originally Posted by LaurV

Scrolling (page changing) there doesn't work.

Thanks, fixed now.

Originally Posted by LaurV

Also: many old/small stuff was not found by P-1 because it was already found by TF to low limits

No, this report only includes exponents where a NF-PM1 result was reported and then a factor was found to be in the range that the previous P-1 should have found but didn't.
For an example: M19,998,217 has a NF-PM1 reported on 2003-10-08, LL completed 2003-11-08, DC-LL completed 2007-12-03 (both a waste of time) and then 10 years after the failed P-1 a factor was found by TF on 2013-08-14.
When I generate data for this report I explicitly look for factors found after the P-1 was done.

[Ensigm]

Originally Posted by James Heinrich

data for this report I explicitly look for factors found after the P-1 was done.

Are you the maintainer of mersenne.ca? I think some exponents in the list are actually results of glitches in data conversion, such as 50077721. It might be better to exclude those cases where P-1 date is unknown.


i.e., if a P-1 date is unknown, we treat it as done after the factor discovery. If a factor discovery date is unknown, we treat it as discovered during or before the P-1 run.

[James Heinrich]

Originally Posted by Ensigm

I think some exponents in the list are actually results of glitches in data conversion, such as 50077721

Thanks for highlighting a specific example. There was an illogic in part of my code that got the P-1 run and P-1 bounds disconnected. I have patched that and fixed it so it shouldn't happen any further. And that eliminates about 14k/16 exponents that were in the "missed by P-1" list.

[storm5510]

From August 23rd. It may already be here somewhere. Anything shorter than 30 digits, I do not bother with...

M100091029 has a 101.349-bit (31-digit) factor: 3228091182977790599237506837961 (P-1,B1=650000,B2=22000000)

[firejuggler]

ryan did it again

Ryan Propper
F-ECMFactor:  

1270133764632902720778602923087552962031274925622407641153 / (ECM curve  1, B1=110000000, B2=900514153782, Sigma=11360005842630690070)

as well as for 5231 and 5351 ( but slightly less impressive)

[LaurV]

As we were talking about ECM records, I was looking to see if my memory still serves me right about those 83 digits, and found that there are some misconcordances between this table and this table. We have all their lines, but sometimes with different names (usually, anonymous, when the right names should be provided, even if they are not members of gimps, that would be fair for the discoverers), while they do NOT have all our lines, probably some of our records were not reported to them (like Ryan's M2671 factor). Somebody may report our records to them, so they become part of the "all times/all kinds" record history (not me, the discoverer should do it, but if they are not interested, maybe James?).

[Jwb52z]

P-1 found a factor in stage #2, B1=723000, B2=19528000.
UID: Jwb52z/Clay, M100310531 has a factor: 54958127519802952206597652367 (P-1, B1=723000, B2=19528000),

95.472 bits.

[axn]

Another big one. 42 digits, 138 bits

M3667249 has a factor: 343439538302947378252521531857403032501081 (P-1, B1=30000000, B2=600000000, E=12)

Not my biggest, but still a top-40

[James Heinrich]

Originally Posted by axn

Another big one. 42 digits, 138 bits. Not my biggest, but still a top-40
M3667249 has a 137.979-bit (42-digit) factor: 343439538302947378252521531857403032501081 (P-1,B1=30000000,B2=600000000,E=12)

Wow, impressive, congrats!

[storm5510]

I was goofing with the GPU functions of GMP-ECM. The input cannot be larger than 2^1018. I was pulling test values out of the air when I came up with this:

Composite cofactor ((((2^943-1)/41884514890068404473)/21698431)/62209711)/164511353 has 241 digits

I am not sure how anyone would convert this to a single large number. Consider it a curiosity, amusement, or whatever.


Edit: I thought about Yafu, so I ran this:

yafu (2^^943-1)*41884514890068404473*21698431*62209711*164511353

And came up with this:

691547141158611815988660322245152254658444810391024156084685520265851892528698195573824658968142416474365688635818459535379264569959609451020315792308103892844793652170248778874582185260236045114390228299259879386132422860579103312328382466032643920718110437396847181215829653041151849017588990719779703721487849925242727519103