mersenne.ca

[Gordon]

Originally Posted by James Heinrich

[snip]and someone has done real TF from 67 onwards [snip]

Yes, that's me, systematically working through everything back down to 100k. At that point, mfaktc won't go any further down.

[masser]

See attached screenshot for M30759907.

300M GhzD for a 2006 TF effort?


EDIT: Ah, I think I see what happened. Composite factor submitted and user got TF credit for unfactored composite.

[James Heinrich]

I have edited the 2006 record to show two prime factors found instead of the composite for clarity.
We're not even certain that this was a TF effort (such detail was not recorded back then) but it seems likely given the size of the factors. Odd though that he found the already-known 68.1-bit factor, discovered a new 69.6-bit factor, but missed the 68.6-bit factor you just discovered (16 years later).

I have also backdated the discovery date of the 69.6-bit factor from 2008-03-05 to 2006-10-01

[axn]

Considering the fact that a) TF doesn't find composite factors (outside the bit depth), b) the factors are P-1 smooth, and c) missed another factor in the range, this is almost certainly a P-1 find.

[James Heinrich]

Originally Posted by axn

TF doesn't find composite factors

I think (but I could be mistaken) that Prime95 may report a composite TF factor if it finds multiple TF factors under certain circumstances (same class, perhaps?). A rare occurrence at best, of course.

Originally Posted by axn

this is almost certainly a P-1 find.

I agree, and have adjusted accordingly.

[Uncwilly]

Can the worktodo balancer be extended to general PRP tests to?
Someone was asking for a script to produce worktodo sections, I pointed to your balancer. They said it did not work for things like:
PRP=718849,2,1564054,1
I verified it and tried with an "N/A," added in.

[axn]

Originally Posted by James Heinrich

I think (but I could be mistaken) that Prime95 may report a composite TF factor if it finds multiple TF factors under certain circumstances (same class, perhaps?). A rare occurrence at best, of course.

Not that I'm aware of. I am aware of one condition under which it outputs composite factor, which is when the composite factor is of the correct size. Eg:- Suppose a number has a 35-bit and 40-bit factor. Then their product would be either 74- or 75-bit. Accordingly, TF from 73-74 (resp. 74-75) will find the composite factor and report it. This can happen for mfaktx as well. Otherwise, these programs outputs one line per factor even if there are multiple found during a class.

[James Heinrich]

Originally Posted by Uncwilly

Can the worktodo balancer be extended to general PRP tests to?
PRP=718849,2,1564054,1

It parses PRP lines in the form:
PRP=[AID,]k,b,n,c[,how_far_factored,tests_saved[,base,residue_type]][,known_factors]
But it currently expects only k=1,b=2,c=-1
I can easily make it accept other values but the GHz-days for the assignment would be calculated as if k=1,b=2,c=-1. If all the assignments use the same values then it's probably not an issue, otherwise the "balancing" part would likely be wrong.

[James Heinrich]

Originally Posted by axn

programs outputs one line per factor even if there are multiple found during a class.

I tried to test this with the easiest example I could find: Factor=409331239,72,73 but my Prime95 seemed to freeze partway through. Not sure if it's a Prime95 bug or just my computer. Both factors should be found in class 3077.

[James Heinrich]

Originally Posted by James Heinrich

I tried to test this with the easiest example I could find: Factor=409331239,72,73 but my Prime95 seemed to freeze partway through. Not sure if it's a Prime95 bug or just my computer. Both factors should be found in class 3077.

Tried it again on my other computer. At least with current versions of Prime95 it does report factors separately (presumably it was the same in 2006 but I didn't bother to dig out such an old version).

UID: JamesHeinrich/i3-8100, M409331239 has a factor: 7461639505688141994167 (TF:72-73)
UID: JamesHeinrich/i3-8100, M409331239 has a factor: 7761132559942730699447 (TF:72-73)

[James Heinrich]

Originally Posted by LaurV

In the "p-1 missed" you could also show who missed the factor, as you do in the "TF missed" report.

I said no back then, but I've changed my mind. You can now see who missed the P-1 factor and when.

You can also look at the list by user and by year, and see that both LaurV and George have missed a P-1 factor.