LL Tests Over Internet

JuanTutors · 2004-08-13, 13:55

I'm wondering if it would be possible to set up a regular webpage (like a homepage) that sends one to three iterations of a LL test for a visitor to do, and does this until an entire LL test is completed.

For example, if I set a page to store S[1]=4 of the LL test for M_p, p some prime. Then someone visits my page, the page tells their computer what M_p and S[1] are, calculates up to S[4], and sends back that result. Then the next person to visit the page gets sent M_p and S[4], etc.

I understand something like this would take a LONG time. But I'm thinking something separate from GIMPS, working with different Mersenne numbers, that won't affect the program. Most of the logistical reasons I can think of can be avoided with some effort. (Say, like calculation errors, simultaneous or a high volume of visits, quantity of data that must be transfered, etc.) If a page has a high volume of visits, some of the time issue can also be avoided by working on multiple Mersennes.

wblipp · 2004-08-13, 14:28

How are you going to perform the calculations?

Prime95 has hand tweaked assembler with optimization branches for different processors and caches. Are you going to replace that with Java? Too slow.

How big is the residue you need to transfer each time? A bit bigger than the last prime we found. Too much bandwidth.

I don't think it's a promising approach.

William

ET_ · 2004-08-13, 18:32

Quote:

Originally Posted by wblipp

How are you going to perform the calculations?

Prime95 has hand tweaked assembler with optimization branches for different processors and caches. Are you going to replace that with Java? Too slow.

How big is the residue you need to transfer each time? A bit bigger than the last prime we found. Too much bandwidth.

I don't think it's a promising approach.

William

Maybe it could be feasible for factoring purposes... you only have to send and receive k and p

I'll put it in my (too long) worktodo list

Luigi

JuanTutors · 2004-08-13, 23:19

Quote:

Originally Posted by ET_

Maybe it could be feasible for factoring purposes... you only have to send and receive k and p

I'll put it in my (too long) worktodo list

Luigi

Awesome. I think that would be an awesome way to distribute computing, especially for factoring. Would you even need that much more than the average page counter? I wouldn't mind putting something like that on some random page of mine.

I wonder how long it would take a page like this to find a factor of, or even completely factor the average 10M Mersenne...

ET_ · 2004-08-14, 11:13

Quote:

Originally Posted by dominicanpapi82

Awesome. I think that would be an awesome way to distribute computing, especially for factoring. Would you even need that much more than the average page counter? I wouldn't mind putting something like that on some random page of mine.

I wonder how long it would take a page like this to find a factor of, or even completely factor the average 10M Mersenne...

I don't think that completely factor a Mersenne using DC pages would be a good idea... it would take too long, but as in ECM, you may find factors working on average exponents

Maybe a good start could be reserving a bunch of (say) 1000 exponent that are not in Primenet, and create a web prorgam to release and work on it to rise each one to the next bit of factorization. The 1000 exponents could be distributed using a circular queue through a MySQL database that had a table with the exponent itself, its status (free, reserved, factor-found, done), actual k and final k (related to their bit-depth). Another table in the same database would hold factors found for each exponent. A final table would hold number of actual partecipants, their achievements (it's easy to track how many k's each of them has computed) and offer a small stats page.

The project is ready. I'm only waiting for my ISP OK to set it up

Luigi

JuanTutors · 2004-08-17, 02:22

Luigi, Have you been able to test it with some known values yet?

ET_ · 2004-08-17, 12:09

Quote:

Originally Posted by dominicanpapi82

Luigi, Have you been able to test it with some known values yet?

Not yet... It is hot in Rome, and I just can't stay too long at the PC

But I know a PHP translation of my original routine may be used; then it's only a matter of communication.

2004-08-13, 13:55	#1
JuanTutors "Juan Tutors" Mar 2004 1060₈ Posts	LL Tests Over Internet I'm wondering if it would be possible to set up a regular webpage (like a homepage) that sends one to three iterations of a LL test for a visitor to do, and does this until an entire LL test is completed. For example, if I set a page to store S[1]=4 of the LL test for M_p, p some prime. Then someone visits my page, the page tells their computer what M_p and S[1] are, calculates up to S[4], and sends back that result. Then the next person to visit the page gets sent M_p and S[4], etc. I understand something like this would take a LONG time. But I'm thinking something separate from GIMPS, working with different Mersenne numbers, that won't affect the program. Most of the logistical reasons I can think of can be avoided with some effort. (Say, like calculation errors, simultaneous or a high volume of visits, quantity of data that must be transfered, etc.) If a page has a high volume of visits, some of the time issue can also be avoided by working on multiple Mersennes.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Internet of Bings	Nick	Soap Box	425	2022-05-14 03:08
Internet down question	ctteg	Information & Answers	2	2018-03-13 17:57
Politics on the Internet for Dummies	jasong	jasong	3	2013-09-19 05:05
PCs not connected to internet	kdelisle2005	Software	3	2011-06-22 03:09
Too Much Internet Access.	M0CZY	Software	3	2005-10-17 15:41

2004-08-13, 14:28	#2
wblipp "William" May 2003 New Haven 2,371 Posts	How are you going to perform the calculations? Prime95 has hand tweaked assembler with optimization branches for different processors and caches. Are you going to replace that with Java? Too slow. How big is the residue you need to transfer each time? A bit bigger than the last prime we found. Too much bandwidth. I don't think it's a promising approach. William

2004-08-17, 02:22	#6
JuanTutors "Juan Tutors" Mar 2004 230₁₆ Posts	Luigi, Have you been able to test it with some known values yet?