![]() |
|
|
#1 |
|
"GIMFS"
Sep 2002
Oeiras, Portugal
158210 Posts |
Is it possible to run part (say from 0 to 95%) of a LL test on a exponent that is assigned by Primenet to a certain machine in another one (not connected to the Internet), then copying the pxxxxxxx and qxxxxxxx files to the original machine and finish off the exponent there? Or the client detects it´s been run on different machines and produce some form of error?
This may be useful if you have problems with the machine assigned to Primenet and have a spare one that for some reason can´t be connected. while the first one is under repair, work on the exponent will progress. |
|
|
|
|
|
#2 |
|
"Richard B. Woods"
Aug 2002
Wisconsin USA
22·3·641 Posts |
Seems okay to me, as long as you've successfully run the torture tests on each system before LL testing.
|
|
|
|
|
|
#3 |
|
Sep 2002
32×13 Posts |
I think that perhaps you will not recieve credit for the exponent. I've shifted around files like that in the past, and I have 3 or 4 8M DCs that I have under my name on primenet, but which haven't been added to my total. It's very possible, just undesirable if you're a stats whore like me. :(
|
|
|
|
|
|
#4 |
|
"GIMFS"
Sep 2002
Oeiras, Portugal
158210 Posts |
Reckon we all are stats whores in the GIMPS community... 8)
Anyway, I can´t figure out why you haven´t receive credit for the exponents, once the system has accepted them with "no complaints". Could it be the case that your results were found to be incorrect by triple checking? I know that in that case we loose the credit :( |
|
|
|
|
|
#5 |
|
"Richard B. Woods"
Aug 2002
Wisconsin USA
22·3·641 Posts |
I'm not a stats whore, but I could portray one on TV. :)
When PrimeNet accepts a result, it credits you only if it thinks you're the one to whom it assigned that test. If your ID doesn't match what PrimeNet has on record for that assignment, it simply doesn't credit anyone. |
|
|
|
|
|
#6 |
|
Aug 2002
Portland, OR USA
2×137 Posts |
Hmmm,
If you set up the second machine by recreating the entire prime95 directory from the original machine, both will have the same id, won't they? There would be no way for prime95 to detect the switch. (Even if it does some cpu/timing/mem tests and makes adjustments, it will assume you just upgraded, or downgraded, the one machine.) No, IMHO, If you're not being given credit, I doubt it has anything to do with the id of the machine that did the work vs. the one that submitted the results. As cheesehead said, The server only cares about the match between the assigned id and the submitters id. |
|
|
|
|
|
#7 |
|
"GIMFS"
Sep 2002
Oeiras, Portugal
2·7·113 Posts |
OK, so I assume that in the situation I described in my first post I´ll get the credit, because the machine that is actually assigned to Primenet is the same that reports the result of the LL test. The second one is just used to compute part of the result and it is never connected to the server, so its ID (if it has any assigned) never shows up.
I´ll give it a go one of these days then let you know about it... |
|
|
|
|
|
#8 |
|
Aug 2002
110010002 Posts |
I borrow computers from coworkers evenings and weekends, so I'm constantly shuffling partially completed tests around among machines. I have had no problem getting credit for the results.
|
|
|
|
|
|
#9 |
|
Aug 2002
2·32·13·37 Posts |
Depending on how crazy you swap work and the client you use, you might have to watch to make sure the save files are written in the same bit-endian (Is that a word?) form...
Ernst mentions this somewhere on his web page... Of course, by running anything other than Prime95 you sacrifice PrimeNet credit anyways, but in the big scheme (Finding a prime!) PrimeNet credit is a bit silly... |
|
|
|
|
|
#10 | |
|
Oct 2002
Lost in the hills of Iowa
26×7 Posts |
Quote:
AFAIK, the machine those intermediate (and all OTHER applicable files - I transfered the whole directory except for prime itself) results got transfered TO should get the credit. I'll be able to let you know for sure a little after mid-month, after the P-IV machine I transfered *to* completes that LL test. I seriously doubt you would want to transfer between different versions of the client (LINIX to Win, for example) - dunno if different client version *numbers* of the same client would be a problem, but kinda doubt it on "after-the-dot" revisions. |
|
|
|
|
|
|
#11 | |
|
Aug 2002
Termonfeckin, IE
24·173 Posts |
Quote:
What you CANNOT do is swap files between say Mlucas (for Alpha, Solaris etc.) and Prime95. Or Glucas and Mlucas and Prime95. You get the idea.... |
|
|
|
|
![]() |
Similar Threads
|
||||
| Thread | Thread Starter | Forum | Replies | Last Post |
| Odd scaling of test times between two machines | mdettweiler | Hardware | 3 | 2014-07-28 16:35 |
| Running msieve LA with openmpi - do all machines need to be same/similar | EdH | Msieve | 32 | 2013-11-08 17:57 |
| Dual Core and running the same test..? | Unregistered | Information & Answers | 1 | 2008-03-01 15:02 |
| strange problem with torture test on 16core machines | TheJudger | Hardware | 5 | 2006-04-08 11:20 |
| Running on multiple machines | Helfire | Software | 8 | 2004-01-14 00:09 |