Perpetual ECM factoring challenge thread...

[geoff]

Originally Posted by Mystwalker

I started some ECMing on M2018 yesterday

Note that 2,2018L and 2,2018M are also unfactored, and in fact have no reported ECM effort yet. You could work on all four numbers by doing curves on M4036.

P1123 suggested by Bob Silverman earlier in this thread is also one of these cases: M1123 P1123 2,2246L and 2,2246M are all unfactored.

I copied the factors of P1009 and M1009 into a M2018 entry in the lowm.txt file

That is the way to do it. If you wanted to work on M4036 these are the known factors of 2,2018L and 2,2018M:
M( 4036 )C: 5
M( 4036 )C: 12109
M( 4036 )C: 242161
M( 4036 )C: 952957298276617

[garo]

Originally Posted by garo

If I have time, and I'm not promising anything so do not hold me to it, I may put up such a page in the near future i.e. in a month or so after I am back from vacation.

It won't be snazzy but it will have all the required info.

I am back from vacation and my thoughts have turned to the pressing need for a file or web page that shows the amount of ECM work done on Cunningham numbers other than 2+ and 2-.

I know that Paul is definitely someone whose input/help I need and I think the same goes for Bob. My starting point is the composites file on Paul Zimmerman's page mentioned above by Bob. This file is rather unclean and has a lot of numbers for which factors have been found. Paul had offered to clean it up. So if Paul, you've got some scripts that will clean it up quick, I'd be much obliged. Otherwise I'll likely write some awk scripts to clean the file up and then try to get the curves done info organized.

So I'd like to ask anyone and everyone who can help clean this file and/or point me to additional information about curves done to post in this thread.

Let's hope this thread and the effort does not die off.

[Xyzzy]

I can provide a database, a web page and bandwidth... I'm also willing to maintain the page if someone shows me how to understand the numbers and stats...

[R.D. Silverman]

Originally Posted by garo

I am back from vacation and my thoughts have turned to the pressing need for a file or web page that shows the amount of ECM work done on Cunningham numbers other than 2+ and 2-.

I know that Paul is definitely someone whose input/help I need and I think the same goes for Bob. My starting point is the composites file on Paul Zimmerman's page mentioned above by Bob. This file is rather unclean and has a lot of numbers for which factors have been found. Paul had offered to clean it up. So if Paul, you've got some scripts that will clean it up quick, I'd be much obliged. Otherwise I'll likely write some awk scripts to clean the file up and then try to get the curves done info organized.

So I'd like to ask anyone and everyone who can help clean this file and/or point me to additional information about curves done to post in this thread.

Let's hope this thread and the effort does not die off.

Here's some statistics on numbers I ran

Number 3M 11M 43M
------ -- --- ---
2,1294L 700
2,1322M 700
2,1334L 700
2,1342L 700
2,1342M 700 200 100
2,1346L 700
2,1366L 700
2,1366m 700
2,1378M 700
2,1406L 700
2,1406M 700
2,1418M 700
2,1426L 700
2,1426L 700
2,1430L 700 200 100
2,1450L 700 200 100
2,1454L 700
2,1478L 700
2,1490L 700
2,1526L 700
2,1534L 700
2,1534M 700 200 100
2,1550L 700
2,1582L 700
2,1586L 700
2,1586M 700
2,1598L 817
2,1606M 700
2,1610L 700 200 100
2,1630M 700
2,1634L 700 200 100
2,1642L 700
2,1654L 700
2,1658L 700 200 100
2,1658M 700
2,1666L 700 200 100
2,1678L 700
2,1686L 700
2,1690M 700
2,1694L 700
2,1694M 700
2,1714L 700
2,1714M 700
2,1726M 700 200 100
2,1734L 700
2,1758L 700
2,1782M 700 200 100
2,1790M 700
2,1794M 700 200 100
2,1798M 700
2,1806M 700 200 100
2,1826M 700
2,1862L 700
2,1870L 700
2,1870M 700
2,1910M 700
2,1914M 700
2,1926M 700
2,1946L 700
2,1962M 700
2,1970L 700
2,1990M 700
2,2002M 700
2,2034L 700
2,2046L 700 200 100
2,2050L 700
2,2070L 700
2,2074M 700
2,2090M 700
2,2106L 700
2,2110M 700
2,2114M 700
2,2118L 700
2,2154M 700
2,2170L 700
2,2190L 700
2,2214L 700
2,2226L 700
2,2226M 700 200 100
2,2238M 700
2,2250L 700 200 100
2,2262L 700
2,2334L 700
2,2346L 700
2,2358M 700
2,2370L 700
2,2370M 700 200 100
2,2382L 700
2,2394M 700

Bob

[Xyzzy]

Bob, what happened to your old account? Email me if you are locked out of it or something...

xyzzy@mersenneforum.org

[garo]

Thanks Mike for your offer. I'll take it up. I'm waiting to hear from xilman before proceeding because I'm sure he has a few tricks that will save me time.

[xilman]

Originally Posted by garo

Thanks Mike for your offer. I'll take it up. I'm waiting to hear from xilman before proceeding because I'm sure he has a few tricks that will save me time.

I'll see what I can do. It may take me a couple of days as I'm rather busy today.

I am assuming that we will only be interested in the statistics for the composite numbers. That is, items only of historical interest should not be included. If this assumption is wrong, please get back to me quickly.

Paul

[garo]

Absolutely! What I had in mind is a page which shows all the composite numbers with fields for the original number say 2^1033L; Number of digits in the composite: 258; and ECM effort for the 35, 40, 45 and the 50 digit level.

So historical records are not reuired. If a number has been completely factored, it will not be in the table. The page is essentially intended to be a guide for people interested in doing ECM on Cunningham tables but with no easy one stop page showing how much effort has already been performed.

[xilman]

Originally Posted by garo

Absolutely! What I had in mind is a page which shows all the composite numbers with fields for the original number say 2^1033L; Number of digits in the composite: 258; and ECM effort for the 35, 40, 45 and the 50 digit level.

So historical records are not reuired. If a number has been completely factored, it will not be in the table. The page is essentially intended to be a guide for people interested in doing ECM on Cunningham tables but with no easy one stop page showing how much effort has already been performed.

Ok. I've made a small start but it is turning out to be more effort than I hoped.

I'll deal with each table (2+, 2-, 3+, 3-, ..., 11+, 11-, 12+, 12-) separately.

Current state of play is that I've extracted all records of composites in my tables. I've reformated Paul Zimmermann's table to make it easier to manipulate by machine and I've stripped out a lot of the unneeded information. Paul's data is now merged with my records.

Still to do is to remove the fully factored records from Paul's data (not easy to do in isolation, but straightforward, if tedious, in the merged data); check for consistency (e.g. that a C245 really does have 245 digits) and then get the ECM data into a consistent format. This is going to take some time, based on the progress made so far with the 7- table.

I propose presenting the data as a table with these headings:

base, index[+-LM], Size, #1M, #3M, #11M, #43M, #110M, decimal

An example record would be:

7,253-,C177,1800,4500,523,0,0,837361951192095087939641166818148470141501362644041264056984651004465207725218545935200829096821327274332372634642510557134137527272901297505888633017384270685685057761924580483443373548537463

The decimal representation is put last on the line so it's available for cut and paste, but doesn't push the other information over off the right side of the screen


Paul

[garo]

Great! Paul, the idea of doing it on a per exponent basis is a good one. If you wish, I can do the tedious work for some of the exponents and we could then cross-check each other's work. If you have any scripts/tricks that you'd like to share they'd be appreciated.

I am not sure if adding the decimal representation is a good idea as it may sacrifice the readability of the page. Moreover, composite information is always available at http://www.cerias.purdue.edu/homes/ssw/cun/prime.php . So linking to this page may be sufficient. But we can leave the info in for the moment. Stripping it out later will be trivial.

[xilman]

Originally Posted by garo

Great! Paul, the idea of doing it on a per exponent basis is a good one. If you wish, I can do the tedious work for some of the exponents and we could then cross-check each other's work.

Garo, please mail me (paul@leyland.vispa.com) so that I can send you various files. I don't want to clutter up the forum with the data.

I now have the 7- table in both CSV (comma separated variable) and HTML format. There are a couple of disclaimers that need to be added though. The first is that the data represents the minimum amount of work performed. It's quite clear that more than this has been done on many if not all of the numbers. Where I have no data, I record that no curves have been run. This is almost certainly wrong, but it's the best (i.e. most conservative) I can do. The second disclaimer is that sometimes PaulZ's data is hard to interpret. Again, I've made the conservative estimate in these cases and so may have not recorded work that has been done.

Given that I'm putting a lot of effort into this, I may well put the tables on my own Cunningham pages which currently reside at http://www.leyland.vispa.com/numth/f...ngham/main.htm

Paul