Dual Boot - Same Assignments

[kladner]

First, my apologies for posting a question which I'm sure has been around before. I did search, but couldn't find the info I wanted. So here goes:

After a recent rebuild of my machine, I rejoined GIMPS with the intention of contributing to the Search. Having gone from a dual core Opteron to a hex core Phenom II with 8GB of RAM, I figured I could spare some clock cycles a lot of the time.

This machine is now dual-booting between 32 bit XP Pro and 64 bit Win 7. In my ignorance, when I set up the Win 7 instance of Prime95, I thought that using the same username and computer name would result in the same assignments. I now realize that this is incorrect, and that I am actually running two different sets of 6 assignments.

How can I release one set of assignments, and get Prime95 working on the same set regardless of which system I happen to be running?

Thanks in advance, and again, apologies for the noob question.

[patrik]

You must let Prime95 use the same save files and worktodo.txt (e.g. run it from the same directory), so you must use a directory that is mounted and shows up regardless of how you boot your computer. (I haven't done this under Windows/Windows--only on a dual Windows/Linux machine.)

You could merge the two worktodo.txt files manually (and put one test after the other, so that you don't loose the work you already did), and copy the save files to the new, common directory. (To be safe, put "UnreserveDays=90", into prime.txt, to prevent Prime95 from unreserving the extra work.) If you do want to skip the work you have already done, just select Advanced/Unserve exponent, or do it from http://www.mersenne.org/workload/ .

[kladner]

Thanks for the response, Patrik. You got me headed in the right direction. I'm not sure I managed to keep the exponents that I had been running, but I seem to have gotten both boots running on the same set. In further reading of threads, and the various text files (undoc.txt in particular) I found the
"WorkingDir=your_directory_name" line. Using it, I was able to point the Win7 instance (64bit) of Prime95 to the directory of the XP (32bit) instance. This seems to have gotten them on the same page. (see screen shots) The upper, shaded set of six was taken in Win 7, the lower six are from XP a few minutes later.

I do fear that I've left some orphan exponents out there, and lost some work, but only a few weeks worth. My efforts to combine the worktodo.txt files got screwed up, and provoked error messages. I'll go to the "workload" link you provided and see if I can clean up after myself there.

Thanks again!

[science_man_88]

Originally Posted by patrik

You must let Prime95 use the same save files and worktodo.txt (e.g. run it from the same directory), so you must use a directory that is mounted and shows up regardless of how you boot your computer. (I haven't done this under Windows/Windows--only on a dual Windows/Linux machine.)

You could merge the two worktodo.txt files manually (and put one test after the other, so that you don't loose the work you already did), and copy the save files to the new, common directory. (To be safe, put "UnreserveDays=90", into prime.txt, to prevent Prime95 from unreserving the extra work.) If you do want to skip the work you have already done, just select Advanced/Unserve exponent, or do it from http://www.mersenne.org/workload/ .

this got me thinking ( though I have no use for it I only have 1 computer) but could this not be set up in a network folder all the computers on a network can get to if so is it possible to treat each computer in a network as a separate core and run the same number split between all cores ?

[kladner]

Well, I've had a look at the Computer Details page (screenshot). The first three lines are 64bit versions. The 1st and 3rd list exponents which are in my current lineup. The 2nd and 4th do not. I am thinking I should merge the 1st and 3rd, and remove the second. I'm not sure what to do about the 4th, as it is the sole listing for the 32bit version, but shows an exponent which isn't currently being run.

I will leave things as-is for now, and would appreciate any further advice in fixing the records.

[kladner]

I think I've got it behaving. Through the Assignment and CPU pages I was able weed out Exponents that hadn't had any work done on them. This left me with two sets of work to go through, but with no lost work.

There were still 4 CPU's: 2 each for 32 and 64 bit program versions. I first tried deleting one of each, but this produced warnings about CPU ID's not being correct. And, they were subsequently regenerated. Then I merged the 32 and 64 bit pairs, and everything quieted down. Now the server and the local programs all seem agreed on everything.

kladner

And thanks again, Patrik, for the tips.

science man 88: Greater understanding than mine is needed for that question. It's an interesting idea, though.

[science_man_88]

Originally Posted by kladner

science man 88: Greater understanding than mine is needed for that question. It's an interesting idea, though.

the main problem I see is that if each core was to do something like TF of that number but each a different bit range all doing it at one we could do more TF all at once but if the lower TF computers finds a factor disproving primality a stop command would be needed to stop all the waste of the other computers time.

this kinda defeats doing all the same thing at once on this group which defeats the whole concept no ?

[kladner]

I was wrong in my belief that I had resolved the CPU page issues. Subsequent reboots into Win 7 and XP have seen the CPU list (at GIMPS) repopulated--now showing I have *13* registered computers. I'm not sure that this has any real effect on getting work done.

Well, back to studying the threads here, and the various text files. I'm also still looking at patrik's advice and trying to see how to fully implement that. I would put both instances in one folder, though that would require renaming one of the executables.

Suggestions would be welcome.

[patrik]

I never bother about the CPU page. Everytime I need to reinstall the OS I get a new CPU entry.

Can you name one copy Prime95_32 and one Prime95_64? (Guessing here.)

@science_man_88: Using one directory for many computers would lead to difficulties with local.txt, which contains information specific to the computer. And you couldn't run the number on more than one computer with Prime95, and even with appropriate software, it would be a bad idea (due to the communication between the computers being much slower than the communication between CPU and memory).

[kladner]

Originally Posted by patrik

I never bother about the CPU page. Everytime I need to reinstall the OS I get a new CPU entry.

OK. Thanks on that, patrik! I won't worry about it anymore.

Originally Posted by patrik

Can you name one copy Prime95_32 and one Prime95_64? (Guessing here.)

I tried messing with that, so they could cohabit in the same directory. I could be wrong, but it seemed then that Start at Bootup didn't work for the renamed executable. I'm still looking for where that setting gets stored, thinking that I could make the pointer to the file name match.

Just looked in prime.txt. It would seem that "Windows95Service=1" might be the thing which starts the program. I'll have to test some more.

On the question of extra processors showing up in the list, I did find
"FixedHardwareUID=1"
in undoc.txt. This is supposed to disable the automatic hardware detection. The context was for running on multiple machines via a flash drive. I'll try running with that and see if they keep accumulating. Just an experiment, since it does not seem to affect functionality.

Thanks for getting back to me, patrik. Much appreciated.
kladner

[science_man_88]

Originally Posted by kladner

OK. Thanks on that, patrik! I won't worry about it anymore.




I tried messing with that, so they could cohabit in the same directory. I could be wrong, but it seemed then that Start at Bootup didn't work for the renamed executable. I'm still looking for where that setting gets stored, thinking that I could make the pointer to the file name match.

Just looked in prime.txt. It would seem that "Windows95Service=1" might be the thing which starts the program. I'll have to test some more.

On the question of extra processors showing up in the list, I did find
"FixedHardwareUID=1"
in undoc.txt. This is supposed to disable the automatic hardware detection. The context was for running on multiple machines via a flash drive. I'll try running with that and see if they keep accumulating. Just an experiment, since it does not seem to affect functionality.

Thanks for getting back to me, patrik. Much appreciated.
kladner

make a folder and make a sub folder for each version. my best guess but I'm usually crap