Two computers, same assignment?
 

Is it possible to set up two computers to search for the same Mersenne prime? I have two computers working on different assignments, could they work on the same one?

Like in "work for the same one to speed up things (i.e. halve the time) - No. It is not possible, each test is quite chained, i.e. the calculation in a step depends on the result of the former step, and each step itself works with a lot of data which should be slower to be moved between computers than to be computed locally.

Like in "work for the same one, to double check the work, and avoid other people waste their time to double check" - Yes, that is possible, you have to stop/exit P95 on both computers, copy assignment lines from the "worktodo" file of the first computer to the second computer's "worktodo" file and restart P95 in both. Please note that (a) this will not short the testing time of one exponents, however some people do that (me inclusive) to ensure that their calculus is right, they watch close the intermediary results of the calculus and when it does not match, they stop both computers and re-do the last few steps, therefore saving time comparing with the case when a wrong test should be carried on to the finish, for days, and (b) this activity is generally frown upon, in local circles, because the same user double-checking his own work is not exactly in line with the "[URL="https://en.wikipedia.org/wiki/Blind_experiment#Double-blind_trials"]double blind[/URL]" concept (because the user may be a cheater or prankster) and a [U]triple check[/U] is carried on later by another user (therefore wasting more computing time of the project per total, as a group combined).

Note that the best case for you, as individual, is to have different assignments running in different computers, and most probably, different assignments running in different [U]cores[/U] of the same computer. If you computers have 4 and respective 6 cores each in its CPU, then the best for you would be to run [B][U]10[/U][/B] assignments in total, in the same time. Bust systems are different and that, you can see for yourself, if you run benchmarks from your P95 menus, in each computer. It may take you 5 days to work a single assignment in all 4 cores, that means 20 days to run four assignments, but on the other hand, it may take you 18 days to run a single assignment in a single core, which means you can do 4 assignments in 18 days, if you give each core its own work. Of course, my numbers are for exemplify only, your mileage may vary. Benchmark your systems and put the results on the benchmark thread, people may help you to chose the best config. Read around, about hyperthreads and stuff...

[edit p.s - btw, welcome to the project]

Ok thanks. Is my progress saved for others? Say I begin working on a project but quit before it's done, could someone else pick up where I started, or do they have to start all over? I've no idea what all this is FFT is all about but the numbers relevant to doing Lucas-Lehmer testing seems like something that could be saved (to the server or whatever).

I don't know anything about cores and stuff, I'll probably quit the project on this computer (which would take another two weeks at least) and just leave the old computer (with better cpu but minimal memory) running.

[QUOTE=erve02d0;435062]Ok thanks. Is my progress saved for others? Say I begin working on a project but quit before it's done, could someone else pick up where I started, or do they have to start all over? I've no idea what all this is FFT is all about but the numbers relevant to doing Lucas-Lehmer testing seems like something that could be saved (to the server or whatever).

I don't know anything about cores and stuff, I'll probably quit the project on this computer (which would take another two weeks at least) and just leave the old computer (with better cpu but minimal memory) running.[/QUOTE]

Partial works is not saved.

When you don't finish your assignment, your work is lost.
The trouble to "resume" by other user would be to great, and untrustworthy. I won't continue other user's work, unless I know he is not a cheater.

[QUOTE=LaurV;435071]When you don't finish your assignment, your work is lost.
The trouble to "resume" by other user would be to great, and untrustworthy. I won't continue other user's work, unless I know he is not a cheater.[/QUOTE]

Why is that? The program does Lucas-Lehmer testing, right? So all that needs to be saved is the iteration number $i$ and the term $s_i$. I bet a lot of people have started projects but never finished, it's a shame that their progress is lost, changing the program so that it saves $i$ and $s_i$ "for posterity" cannot possibly be a difficult thing to do?

[QUOTE=erve02d0;435073]Why is that? The program does Lucas-Lehmer testing, right? So all that needs to be saved is the iteration number $i$ and the term $s_i$. I bet a lot of people have started projects but never finished, it's a shame that their progress is lost, changing the program so that it saves $i$ and $s_i$ "for posterity" cannot possibly be a difficult thing to do?[/QUOTE]

1. It would take a lot of storage space
2. Not everyone has good upload bandwidth for frequent backups
3. Would you want to resume someone else's work that may have been done on a faulty machine?

[QUOTE=Mark Rose;435078]1. It would take a lot of storage space
2. Not everyone has good upload bandwidth for frequent backups
3. Would you want to resume someone else's work that may have been done on a faulty machine?[/QUOTE]

How much space? The iteration number would just be a few bytes; the prime95 software doesn't require a lot of memory despite constantly squaring huge numbers, can't that same concise representation be held in the server as well?

I don't know what the other stuff means...

[QUOTE=erve02d0;435083]the prime95 software doesn't require a lot of memory despite constantly squaring huge numbers, can't that same concise representation be held in the server as well?[/QUOTE]
Ask yourself these questions, as an exercise:
1. How concise is this representation? Hint: [SPOILER]see your folder where and find the size of the "p########" file. This file keeps all information that is needed to resume and it is ~8-9Mb. Bonus points if you can answer: can it be compressed?[/SPOILER]
2. Now, that's not too much indeed, but how many tests are currently running simultaneously? [SPOILER]Check the server stats (sum up [URL="http://www.mersenne.org/primenet/"]columns #11-12, here[/URL]).[/SPOILER] Multiply and estimate how much space you would need on the server. What is its cost?
3. How will data from the client get to the server? If a client simply goes away, never to return, then you cannot get this data, so are you suggesting to send the temp files to the server every few hours? If so, estimate how much network traffic this would create. What is its cost?

[QUOTE=erve02d0;435083]I don't know what the other stuff means...[/QUOTE]

There are many very smart people around here.

If it was advantageous, do you not think this would have already been done?

A joke: "Two economists are walking down the road. One says 'Is that a $100 bill laying on the ground?' 'Of course not', says the other, 'or else someone would have already picked it up.'"

[QUOTE=erve02d0;435073]Why is that? The program does Lucas-Lehmer testing, right? So all that needs to be saved is the iteration number $i$ and the term $s_i$. I bet a lot of people have started projects but never finished, it's a shame that their progress is lost, changing the program so that it saves $i$ and $s_i$ "for posterity" cannot possibly be a difficult thing to do?[/QUOTE]

But [I]s[/I][SUB]i[/SUB] is quite big.

Remember that in the Lucas-Lehmer sequence [I]s[/I][SUB]i[/SUB] ≈ [I]s[/I][SUB]i−1[/SUB] squared, so each term in the sequence doubles in size (in terms of binary digits). We don't need to retain the full-size number, just the remainder when divided by [I]M[/I][SUB]p[/SUB], but that is still a large number.

Look for the file named "p########" in the same directory where you installed the prime95 software. It is several megabytes long. That is where the work-in-progress is saved to disk, usually every half-hour.

This file stays on your computer, it does not get sent over the Internet. A few megabytes is not so much these days, but there are many thousands of computers working on the project simultaneously. Sending all of these work-in-progress files to the central database over the Internet would use a lot of bandwidth and storage space.