mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Data (https://www.mersenneforum.org/forumdisplay.php?f=21)
-   -   Thinking out loud about getting under 20M unfactored exponents (https://www.mersenneforum.org/showthread.php?t=22476)

petrw1 2017-07-27 05:06

I think I'll start somewhere in this range....some day soon.
This is the current DC range so it can help there too.

[CODE]
100K ToGo 2^N
40.0 2059 72
40.1 2116 72
40.2 2046 72
40.3 2096 72
40.4 2032 72
40.5 2077 72
40.6 2023 72
40.7 2033 72
40.8 2066 72
40.9 2004 72
41.0 2063 72
41.1 2093 72
41.2 2012 72
41.3 2041 72
41.5 2035 72
41.6 2049 72
41.7 2099 72
41.8 2051 72
41.9 2047 72
42.0 2021 72
42.1 2017 72
42.2 2014 72
42.3 2045 72
42.4 2022 72
42.5 2003 72
42.6 2129 72
42.7 2035 72
42.8 2015 72
42.9 2067 72
43.0 2133 72
43.1 2046 72
43.2 2064 72
43.3 2097 72
43.4 2102 72
43.6 2021 72
43.7 2025 72
43.8 2049 72
43.9 2028 72
44.0 2017 72
44.1 2012 72
44.2 2024 72
44.6 2098 72
44.7 2060 72
44.8 2009 72
44.9 2034 72
45.0 2015 72
45.2 2046 72
45.3 2090 72
45.4 2014 72
45.5 2096 72
45.6 2024 72
45.7 2031 72
45.8 2093 72
46.0 2042 72
46.1 2005 72
46.2 2028 72
46.3 2097 72
46.5 2061 72
46.6 2054 72
46.7 2044 72
46.8 2030 72
47.0 2025 72
47.1 2060 72
47.2 2038 72
47.5 2065 72
47.6 2019 72
47.7 2046 72
47.8 2074 72
48.0 2028 72
48.1 2026 72
48.2 2053 72
48.3 2035 72
48.4 2113 72
48.6 2030 72
48.7 2029 72
48.9 2023 72
49.1 2004 72
49.3 2049 72
49.4 2081 72
49.5 2056 72
49.6 2121 72
49.8 2072 72
49.9 2008 72
50.2 2053 73
50.3 2034 73
50.4 2019 73
50.6 2005 73
50.8 2012 73
50.9 2017 73
51.3 2000 73
51.7 2004 73
52.2 2003 73
52.3 2057 73
52.4 2005 73
52.9 2045 73
53.2 2034 73
53.3 2002 73
53.5 2029 73
53.9 2053 73
54.0 2018 73
54.2 2019 73
54.6 2009 73
55.0 2009 73
55.4 2011 73
55.6 2016 73
55.9 2011 73
56.0 2002 73
56.1 2010 73
56.3 2032 73
56.5 2032 73
56.6 2043 73
56.7 2027 73
56.8 2061 73
57.4 2009 73
57.6 2030 73
57.7 2008 73
57.8 2004 73
58.1 2032 73
58.7 2067 73
58.8 2019 73
59.4 2065 73
59.6 2031 73
59.9 2004 73[/CODE]

A lot of these will have poor P-1; might be more efficient to start there?

petrw1 2017-08-31 02:54

So here's my plan....
 
I will be starting my little pet project within the next week with 2 GPUs and adding about 30 Cores a few weeks later.

I'll be working in the extremes ranges.
The bigger horses in the 40M and 50M ranges.
The side benefit is that this will reduce the number requiring DC.
The smaller ponies in the 2M range.

For the GPUs there are about 16,000 Assignments in the high 40's factored below the yellow bar: [url]http://www.mersenne.ca/status/tf/0/0/3/4000[/url]
I expect to find about 200 factors there.

For the faster cores with adequate RAM I will be P-1'ing exponents in that same range that were done with B1=B2. These seem to give better odds at finding a factor than doing P-1 where the B1 and B2 bounds were less than adequate.

I'll let a few slower cores dabble at the low end (2M) doing ECM.

=================

If anyone is intrigued you are more than welcome to join in.
Keep me posted so we don't step on each other's toes.

See the first post for the details on this little initiative.

Wayne

chalsall 2017-09-04 14:55

[QUOTE=petrw1;466721]For the GPUs there are about 16,000 Assignments in the high 40's factored below the yellow bar: [url]http://www.mersenne.ca/status/tf/0/0/3/4000[/url] I expect to find about 200 factors there.

If anyone is intrigued you are more than welcome to join in. Keep me posted so we don't step on each other's toes.[/QUOTE]

After conferring with Wayne, I have re-enabled DCTF on GPU72 to take the candidates from 48.8M to 50M up to (at least) 73 bits (the optimal economic cross-over point for contemporary GPUs). This is to help avoid any "toes being stepped on".

If anyone is interested, the DCTF manual assignments page has been re-enabled, and the spiders (MISFIT et al) can simply ask for DCTF work again.

tha 2017-09-07 07:27

For what it is worth, if been redoing P-1 on exponents that had their P-1 done with B1=B2 (usually on machines with no spare memory to do the higher B2 level). Almost all of it in the 10M to 20M range. Mostly with B1=300000 and B2=6000000. About one in 38 candidates yields a new found factor on a previously unfactored exponent.

petrw1 2017-09-07 15:02

[QUOTE=tha;467296]For what it is worth, if been redoing P-1 on exponents that had their P-1 done with B1=B2 (usually on machines with no spare memory to do the higher B2 level). Almost all of it in the 10M to 20M range. Mostly with B1=300000 and B2=6000000. About one in 38 candidates yields a new found factor on a previously unfactored exponent.[/QUOTE]

Thanks....I'll note that and try not to step on your toes.

I realize my little sub-project is actually quite a massive undertaking that I couldn't dream of finishing in my lifetime.

But I understand that there are always people dabbling at factoring in these ranges:
- Some are doing TF in the 20M or 30M ranges; in fact most ranges.
- There is a steady stream of ECM in the 0M - 20M range.
- Others are doing P-1 here and there.

So we'll see where we are at end of 2018.

To compare: since the beginning of 2017 there have been 4,082 exponents newly factored under 60M.
- Over half of these in the 20-30M range.
- And another 30% under 10M

-W-

tha 2017-09-07 21:11

[QUOTE=petrw1;467318]Thanks....I'll note that and try not to step on your toes.

/QUOTE]

Oh, thanks, but I am not claiming any range, do go ahead.

petrw1 2017-09-16 22:21

Almost 8 weeks later...an update.
 
698 (1.25%) exponents factored in the ranges of interest....0 - 59.9M

1 range cleared (now below 2,000): 22.2M
Several with significant progress.

Gordon 2017-09-16 22:43

I am doing a lot of the TF in the under 10m block.

Have already done to 67 bits

3,5,6,7,8 & 9m to 67 bits. Someone else took 9m to 68 bits

2m is 2/3 the way to 67 bits. So about 100k'ish exponents so far.

Did quite a bit in the sub 1m range to lift them to 64 bits, but mfaktc doesn't work below 100k so would need to use the cpu - which is currently tied up doing the ecm stage 1's to keep the deep stage 2's going on M4007.

GP2 2017-09-16 23:54

[QUOTE=Gordon;467921]
Did quite a bit in the sub 1m range to lift them to 64 bits, but mfaktc doesn't work below 100k so would need to use the cpu - which is currently tied up doing the ecm stage 1's to keep the deep stage 2's going on M4007.[/QUOTE]

Below 100k all exponents have already been ECM'd to t=30 (about 100 bits), so only about a 37% chance of finding a factor smaller than 100 bits. So TF to 64 bits would have a very very low probability of finding any new factors.

Below 656k all exponents have already been ECM'd to t=25 (about 83 bits). Poor prospects for finding factors of 64 bits or smaller.

The above is for all exponents, including the ones that already have at least one known factor.

If we consider only exponents with no known factors, then everything below 1M has already been ECM'd to t=25.


Edit: also [URL="http://www.mersenneforum.org/showthread.php?t=19014&p=461679"]user TJAOI has covered everything up to 64 bits[/URL].

petrw1 2017-09-17 04:05

[QUOTE=GP2;467922]... everything below 1M has already been ECM'd to t=25. ...
[/QUOTE]

Good to know...however.

The lowest range of interest to my little subproject (getting all 100K ranges below 2,000 unfactored) is 1.8M

petrw1 2017-09-17 04:59

[QUOTE=Gordon;467921]I am doing a lot of the TF in the under 10m block.

Have already done to 67 bits

3,5,6,7,8 & 9m to 67 bits. Someone else took 9m to 68 bits

2m is 2/3 the way to 67 bits. So about 100k'ish exponents so far.

Did quite a bit in the sub 1m range to lift them to 64 bits, but mfaktc doesn't work below 100k so would need to use the cpu - which is currently tied up doing the ecm stage 1's to keep the deep stage 2's going on M4007.[/QUOTE]

Thanks....


All times are UTC. The time now is 05:58.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.