mersenneforum.org (https://www.mersenneforum.org/index.php)
-   MisterBitcoin (https://www.mersenneforum.org/forumdisplay.php?f=141)
-   -   Progress of n^41+/-1 (https://www.mersenneforum.org/showthread.php?t=23785)

 MisterBitcoin 2018-11-07 10:51

Progress of n^41+/-1

I noticed that all number of the form n^41-1 with n<6600 are fully factored. The smallest n been unfactored is n=6697 with an C128 cofactor.

I also noticed, that n^41+1 is much less factored, [STRIKE]only up to n=1100 are all numbers FF. The smallest n been unfactored is n=1109 with C122 cofactor[/STRIKE].

Between n=1000 and n=1500 are all numbers FF. Going up to n=2000.
[STRIKE]
I´ll at least pretest these numbers and list them here[/STRIKE]. From time to time I´ll take them down. Please note: Slow progress. :smile:

 axn 2018-11-07 11:33

[QUOTE=MisterBitcoin;499799]I noticed that all number of the form n^41-1 with n<6600 are fully factored. The smallest n been unfactored is n=6697 with an C128 cofactor.

I also noticed, that n^41+1 is much less factored, only up to n=1100 are all numbers FF. The smallest n been unfactored is n=1109 with C122 cofactor.

Between n=1000 and n=1500 are 18 numbers not fully factored.

I´ll at least pretest these numbers and list them here. Please note: Slow progress. :smile:[/QUOTE]

These are all SNFS numbers, so a quadcore could knock them all out in a couple of hours ?

 MisterBitcoin 2018-11-07 13:25

[QUOTE=axn;499800]These are all SNFS numbers, so a quadcore could knock them all out in a couple of hours ?[/QUOTE]

All of my resources for factoring are used for FDB (90 dd up to 99 dd). When I´m done with that range I´m able to switch to something else.

For now I´m just pretesting with one core. That takes ~4 hours/canidate (at C125).

 axn 2018-11-07 13:33

[QUOTE=MisterBitcoin;499802]For now I´m just pretesting with one core. That takes ~4 hours/canidate (at C125).[/QUOTE]

4 hours or 4 minutes? 4 hours is way too much. Like I said, these are SNFS - actual SNFS will be much faster, I think.

 MisterBitcoin 2018-11-07 13:48

[QUOTE=axn;499804]4 hours or 4 minutes? 4 hours is way too much. Like I said, these are SNFS - actual SNFS will be much faster, I think.[/QUOTE]

ECM pretest takes ~4 hours.

The C122 took 50 minutes to complete, with 1 core. :)

 VBCurtis 2018-11-07 16:53

[QUOTE=MisterBitcoin;499805]ECM pretest takes ~4 hours.

The C122 took 50 minutes to complete, with 1 core. :)[/QUOTE]

Why would you spend 5 times the SNFS factoring time on ECM? What polynomial are you using for SNFS? Something like 100 curves at B1=50k is enough.

 Batalov 2018-11-07 18:29

[QUOTE=MisterBitcoin;499805]ECM pretest takes ~4 hours.

The C122 took 50 minutes to complete, with 1 core. :)[/QUOTE]
Seconded - pretest should amount to ~10% of the cost of honest factoring or less.

Also, with a snfs polynomial this jobs will take a couple minutes, not 50! Are you factoring them as gnfs?

 MisterBitcoin 2018-11-07 19:22

1 Attachment(s)
[QUOTE=Batalov;499823]Seconded - pretest should amount to ~10% of the cost of honest factoring or less.

Also, with a snfs polynomial this jobs will take a couple minutes, not 50! Are you factoring them as gnfs?[/QUOTE]

As I said I was only using 1 thread.
The latest number (C123) took ~13 Minutes to finish using 6 threads. Log attached.

I´m using factmsieve.pl und the poly from factordb (These are not perfect, but they work :smile:).
I used a pretest level of t38, looks like its a bit to much.

 All times are UTC. The time now is 06:47.