mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > PrimeNet > GPU to 72

Reply
 
Thread Tools
Old 2017-07-27, 21:25   #1
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

23·1,103 Posts
Default What do we do next?

So, after several years of work we have finally pulled ahead of the LL'ers and P-1'ers.

Extreme coolness!

My question is: what do people want to do?

Do we continue to go depth first (read: taking 85M etc to 76 bits), or do we work the higher ranges breadth-first (each range from 71 bits to 72 bits, and then 72 to 73, etc.

I ask this for two reasons:

1. TF'ing is very far ahead of the LL'ers and the P-1'ers.

2. There was a recent incident where an independent TF'er was working in (and interfered with) a range we were working.

I would appreciate feedback on what people would like to do.
chalsall is offline   Reply With Quote
Old 2017-07-27, 21:44   #2
Mark Rose
 
Mark Rose's Avatar
 
"/X\(‘-‘)/X\"
Jan 2013
Ͳօɾօղէօ

22·5·139 Posts
Default

I think depth-first still makes sense. If we get a surge in LL or P-1, it will be nice to have that buffer. It's also nice to "finish" work (until the balance shifts).

Those who want to go shallow, like I do on my slow cards, can set a lower bit commitment.
Mark Rose is offline   Reply With Quote
Old 2017-07-27, 23:09   #3
petrw1
1976 Toyota Corona years forever!
 
petrw1's Avatar
 
"Wayne"
Nov 2006
Saskatchewan, Canada

10AB16 Posts
Default

Quote:
Originally Posted by chalsall View Post
So, after several years of work we have finally pulled ahead of the LL'ers and P-1'ers.

Extreme coolness!

My question is: what do people want to do?

Do we continue to go depth first (read: taking 85M etc to 76 bits), or do we work the higher ranges breadth-first (each range from 71 bits to 72 bits, and then 72 to 73, etc.

I ask this for two reasons:

1. TF'ing is very far ahead of the LL'ers and the P-1'ers.

2. There was a recent incident where an independent TF'er was working in (and interfered with) a range we were working.

I would appreciate feedback on what people would like to do.

YAY!!!!

Would there be an option 3 to get a little more aggressive on the bit levels?
That is, go a bit deeper a little ahead of the chart recommendations?

Over these several years the project went from barely and sometimes not keeping ahead of the leading edge to having a healthy buffer. It can only get bigger as exponents increase and LL times accordingly.
petrw1 is offline   Reply With Quote
Old 2017-07-27, 23:13   #4
GP2
 
GP2's Avatar
 
Sep 2003

2·1,289 Posts
Default

When TF takes twice as long every time you increase the depth by one bit, at some point it hits diminishing returns. I don't do even token TF anymore because GPU pricing on the cloud is so out of whack, so I don't know if you guys have reached that point yet, but thinking out-of-the-box, there's a big fat 9–10 year backlog for DC LL tests, just sayin'.
GP2 is offline   Reply With Quote
Old 2017-07-27, 23:45   #5
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

227816 Posts
Default

Quote:
Originally Posted by GP2 View Post
...there's a big fat 9–10 year backlog for DC LL tests, just sayin'.
Yeah, I hear you loud and clear. That's why I asked this question.

This is the "Great Internet Mersenne PRIME Search". Our goal is to find primes.

Finding factors ahead is useful, but it's a also a bit like masturbation (I hope that is taken the way it was intended)....
chalsall is offline   Reply With Quote
Old 2017-07-27, 23:52   #6
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

23·1,103 Posts
Default

Quote:
Originally Posted by petrw1 View Post
Would there be an option 3 to get a little more aggressive on the bit levels? That is, go a bit deeper a little ahead of the chart recommendations?
That is an option. Would take me about 30 seconds to update the SQL and the Perl.

The question is would those who actually do the work want to? Are you suggesting taking 85M to 77 bits?

As an aside guys, thanks for all the work.

I remember being attacked by a certain person for daring to try to go to 72, and now we're debating if we should go to 77.

How times change!
chalsall is offline   Reply With Quote
Old 2017-07-28, 00:00   #7
masser
 
masser's Avatar
 
Jul 2003

33×47 Posts
Default

Quote:
Originally Posted by chalsall View Post

2. There was a recent incident where an independent TF'er was working in (and interfered with) a range we were working.
Curious. How did this happen?
masser is offline   Reply With Quote
Old 2017-07-28, 00:13   #8
chalsall
If I May
 
chalsall's Avatar
 
"Chris Halsall"
Sep 2002
Barbados

23·1,103 Posts
Default

Quote:
Originally Posted by masser View Post
Curious. How did this happen?
Please see this post.

I barely have the time to say things once, please do try to keep up. It's a little bit like trying to teach a snail physics.

(Just to share, I actually said that to my girlfriend recently. She was not amused....)
chalsall is offline   Reply With Quote
Old 2017-07-28, 17:08   #9
petrw1
1976 Toyota Corona years forever!
 
petrw1's Avatar
 
"Wayne"
Nov 2006
Saskatchewan, Canada

17·251 Posts
Default

Quote:
Originally Posted by GP2 View Post
When TF takes twice as long every time you increase the depth by one bit, at some point it hits diminishing returns. I don't do even token TF anymore because GPU pricing on the cloud is so out of whack, so I don't know if you guys have reached that point yet,
Agreed but some of the newest GPUs are bloody fast at TF; I don't actually know though if they are just as bloody fast at LL/DC.

Quote:
Originally Posted by GP2 View Post
but thinking out-of-the-box, there's a big fat 9–10 year backlog for DC LL tests, just sayin'.
Are you maybe hinting that the GPU'ers consider running something like CUDALucas and work on the DC LL backlog? Not a bad idea.
petrw1 is offline   Reply With Quote
Old 2017-07-28, 18:12   #10
Mark Rose
 
Mark Rose's Avatar
 
"/X\(‘-‘)/X\"
Jan 2013
Ͳօɾօղէօ

22×5×139 Posts
Default

Quote:
Originally Posted by petrw1 View Post
Agreed but some of the newest GPUs are bloody fast at TF; I don't actually know though if they are just as bloody fast at LL/DC.
They are fast at LL/DC, but they're much faster at TF.
Mark Rose is offline   Reply With Quote
Old 2017-07-28, 18:23   #11
Dubslow
Basketry That Evening!
 
Dubslow's Avatar
 
"Bunslow the Bold"
Jun 2011
40<A<43 -89<O<-88

3·29·83 Posts
Default

Quote:
Originally Posted by chalsall View Post

I remember being attacked by a certain person for daring to try to go to 72, and now we're debating if we should go to 77.
Dubslow is offline   Reply With Quote
Reply

Thread Tools


All times are UTC. The time now is 10:03.

Tue Mar 31 10:03:28 UTC 2020 up 6 days, 7:36, 0 users, load averages: 1.59, 1.34, 1.25

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.