mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Lounge (https://www.mersenneforum.org/forumdisplay.php?f=7)
-   -   Predict M48... (https://www.mersenneforum.org/showthread.php?t=12001)

NBtarheel_33 2013-01-30 12:18

[QUOTE=Batalov;326514]On the bright side, [URL="http://primes.utm.edu/mersenne/LukeMirror/mersenne.htm"]he wasn't an Inquisitor[/URL], some people say.[/QUOTE]

*Some* people say. The ones that got away. As opposed to those who got strapped to the comfy chair to listen to the voice of Ben Stein recite every single digit of 2^43,112,609-1. (Yes, little did Mersenne know that his favorite numerical torture implement was itself a Mersenne prime!)

[QUOTE]Indeed, he was probably quite a pleasant fellow.[/QUOTE]

Actually, from what we are learning in this thread, he was quite a crank, probably scrawling cmd-like messages and random claims of primality on the walls of the abbey. What if...he was Blazys-esque?!?

[QUOTE]Nobody expects...

Spam, spam, spam, spam, spam, spam! Glorious spam![/QUOTE]

THE SPANISH INQUISITION! Our weapons are trial factoring...trial factoring and LL testing...no wait, our two main weapons are trial factoring and LL testing...and double-checking...no wait, our three main weapons are...Cardinal Biggles, you'll have to say it!

LaurV 2013-01-30 13:10

[QUOTE=NBtarheel_33;326639]Double the number of GIMPSters and you cut this to a five-year drought. Increase by a factor of ten and you cut this to a year's drought![/QUOTE]
Multiply the number of gimpsters by 86400, and you cut this to one hour drought! Yeaaaaa!
(so easy to say...)

davieddy 2013-01-30 13:23

[QUOTE=LaurV;326646]Multiply the number of gimpsters by 86400, and you cut this to one hour drought! Yeaaaaa!
(so easy to say...)[/QUOTE]Getting China, India ... on board would help a bit.

D

rcv 2013-01-30 14:16

[QUOTE=Batalov;326514]On the bright side, [URL="http://primes.utm.edu/mersenne/LukeMirror/mersenne.htm"]he wasn't an Inquisitor[/URL], some people say. Indeed, he was probably quite a pleasant fellow.[/QUOTE]
A monk or a friar? Perhaps, if the recent discovery is confirmed, the Religious Studies Department of UCMO can elighten us.

petrw1 2013-01-30 14:37

[QUOTE=davieddy;326648]Getting China, India ... on board would help a bit.

D[/QUOTE]

Back in the day when I bought my first PCs they came prebundled with a bunch of "free" software. Maybe George could partner with Dell OR Gateway Or ... And bundle in Prime95. Then just imagine the thruput ... And the plethora of ANONYMOUS accounts!

Dubslow 2013-01-30 18:46

Quite honestly, the easiest way to drum up support would be to port it to BOINC -- but there are a [i]huge[/i] host of issues with that.

Uncwilly 2013-01-31 00:49

[QUOTE=petrw1;322036]49532123 - January 12, 2013[/QUOTE]

Date-wise this is the closest to the actual discovery date.

cheesehead 2013-01-31 02:22

[QUOTE=Dubslow;326696]Quite honestly, the easiest way to drum up support would be to port it to BOINC -- but there are a [I]huge[/I] host of issues with that.[/QUOTE]Can you describe some of those issues for us?

Dubslow 2013-01-31 04:01

[QUOTE=cheesehead;326783]Can you describe some of those issues for us?[/QUOTE]

A lot of the things that are built into Prime95, such as internet communication, work unit management, affinties/threads per worker, HighMemWorkers, PauseWhenRunning, system priority, etc., are all handled by BOINC itself in BOINC applications. It was designed to be as easy as possible to just plug in computational code and let the wrapper do the rest. I'm not sure how easy it would be to separate just the computational code out from Prime95, but I doubt it would be easy.

Additionally, the vast majority of BOINC projects have computational units that last a few hours or perhaps a few days, tops. The entire BOINC eco-system is built around this assumption, which just plain is not true for anything GIMPS related, except for GPU TF, and CPU P-1 perhaps at a stretch. (Now there's an idea!) Examples of this assumption include the widely used feature to run multiple projects at once: if the user has chosen three projects to be run at 50%, 25%, and 25% of total time respectively, BOINC will spend 4 hours on the first, 2 hours on the second, 2 hours on the third, and then repeat itself. The only viable way to mesh an LL test with that sort of heterogeneity would be to use one core for GIMPS all the time, which I'm not sure is possible to be specified via the API.

Additionally, we would need a BOINC server and some way to allocate work units, but chalsall has in the past volunteered to take care of that end (à la GPU272). Even with that part of it taken care of (with no small amount of work on chalsall's part) it would still be difficult to implement.



[B]Here's an idea:[/B] we could implement a BOINC project that runs only GPU TF, GPU LL (would be a stretch) and CPU P-1. The first should be a lot easier than any other suggestions, because mfakt* are relatively independent programs (i.e. paragraph one does not apply) and because the "work units" are of the sort that BOINC was designed around (so paragraph two doesn't apply). GPU LL would be similar, except that the work units would be rather long, but it might be doable; the last would be hardest, because paragraph one still applies, while the second does not.

axn 2013-01-31 05:41

[URL="http://www.mersenneforum.org/showthread.php?t=16850"]This[/URL] thread had some useful discussions regarding BOINCifying GIMPS (towards the end). Perhaps it is high time to start some focused discussion on making this into reality? It doesn't cost much to think about these things. Who knows, if something feasible does come out, we can get it implemented.

NBtarheel_33 2013-01-31 09:15

[QUOTE=LaurV;326646]Multiply the number of gimpsters by 86400, and you cut this to one hour drought! Yeaaaaa!
(so easy to say...)[/QUOTE]

730,000 * 86,400 = 63,072,000,000 CPUs. A nine-core system for every human being on Earth. OK, let's do it! As the physicists say, we'll make some simplifying assumptions that (1) electricity is free and infinite, and (2) the carbon emissions of 6.3E10 CPUs are negligible. OK, we're all set. :smile:

...But seriously, I really do believe that we ought to be able to increase our user base significantly (I'm talking at least double) by proactively contacting heads of math departments, math clubs, etc. and encouraging them to join up. How about if MIT joined GIMPS? How about if every high school in New York City joined GIMPS? Remember, you might get more than just the school's computers...you might also get student computers, their families' computers, their friends' computers, etc. etc.

In the language of Web 2.0, it's called "going viral".


All times are UTC. The time now is 06:00.

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