Reply With QuoteCan't speak for the whole team, but think most of the DPC disagree.
If I can choose between a project with realtime/multiple updates.. or one with only once a day, it's an easy choice for me.
I think the increased production by NOT sending an update for every completed block is more than enough reason to update. I missed the up-to-the-second stats at first too, but after using the new client for a while now, I think I'm in agreement with Louie that once a day is fine. Besides, once a day updates doesn't seem to discourage people from contributing to GIMPs...
Can't speak for the whole team, but think most of the DPC disagree.
If I can choose between a project with realtime/multiple updates.. or one with only once a day, it's an easy choice for me.
You can send a manual update any time you want. That's what I do when I want to up-to-the minute stats on my user page.
Advanced>Communication>Send new expected completion dates to server
"Today entirely the maniac there is no excuse with the article." Get free DOS, OS/2 and Windows games at RGB Classic Games
Couple of questions.
Is "DaysBetweenCheckins=" the setting that connects to the server and uploads the data ?
and is that setting only between 1 and 7 ?
Also is there a setting for the size of the prime.log file or will this thing grow forever ?
Yes, and only 1 to 7 is valid. In the next release I'll allow values between 0.04 and 7 (though you'll only be able to set it to an integer using the GUI interface).Originally Posted by NicColt
See MaxLogFileSize in undoc.txt. Default is 250K.
So once every hour.. that's perfect
It would be sweet if the next release is sooner.Of course you would let us all know, correct??
Thanks prime95 for your support.
Question: I have completed 2 tests on an offline machine in the lab. I got a results.txt file with 2 lines like this
UID: ML/Kirk, k*2^n+1 is not prime. RES64: CB1165C1C63D7DAD. Wd1: D70B9C4F,00000000, AID: 000000000000000000000000000ECA84 (changed some numbers to prevent abuse)
When i copied these files to my online pc and started prime95 it just fetched 2 new wu's but it didnt report the finished one's
Can anyone explain how to report these?
Being able to complete work offline is relevant to my interests, if someone could point me at some docs for that I would be grateful.
Did you copy the entire folder, or just certain files? I've never done it myself since all my machines have internet access, but it sounds like maybe you didn't copy the worktodo.txt file...?
I am pretty sure you can just copy the entire folder onto a USB stick and run prime95 from the stick on the internet-connected machine, go to Advanced>Manual Communication, and it will report just fine and get new work that you can then copy back to the offline computer.
Try copying over the prime.spl (this assumes you did not uncheck the "Use Primenet...." checkbox in the Test/Primenet dialog box).
Also, be careful not to overwrite an existing prime.spl file!
Hmm I'm afraid when trying several options i did uncheck and check the "Use Primenet..." checkbox
So there is no way to report these tests now and the only option is to run them again?
Today another machine finished 2 tests, I copied the whole sobgimps folder to an usb stick and start prime95 at home...it only updated computerinformation on the server but didnt report any results...
So I now have 4 finshed tests but i cant seem to report them
The prime.log has a history of all the exact URLs it opens on the server to communicate results.Today another machine finished 2 tests, I copied the whole sobgimps folder to an usb stick and start prime95 at home...it only updated computerinformation on the server but didnt report any results...
So I now have 4 finshed tests but i cant seem to report them
So you can recover the tests by:
1) Open your prime.log file
2) Locate the lines where it tried to contact the server with final results.
3) Copy the full URL(s)
4) Paste them into a web browser and open them
5) It should say "SUCCESS" in your browser if the key matches and it updates the record
Cheers,
Louie
Frentik, I found this while browsing through the undocumented features readme. I am not sure if this will help with your problem or not, but it's worth a try.
Thx guys for the replies, i will definitely start using the FixHardwareUid option next time..
@louie:
I did have a look in the log file but there are many connections with the server
are there specific lines/options i have to search for?
Search for "Sending result to server" and then the line under that should have the URL.
Hmm I dont have a line like that. First I have a line on march 24: "getting assignment from server" (duh), then a line on march 25: "updating computer information on the server" when i started the client on the offline pc. This command failed because its not online: "CURL library error: Could not resolve host: www.seventeenorbust.com; No data record of requested type"
Then another line on 6 april when i started the client again on my online pc, but it only updated the computerinformation again because i switched back. It never reported the completed tests....
I'm making an offline submission page in the next few days that will allow you to directly submit the results.txt file for simpler offline processing in the future.
I'll let you know in this thread when it is available. Thanks for the feedback Frentik!
Cheers,
Louie
Great
Oh another thing I noticed: The offline PC also had two new tests in the worktodo.txt file but because it couldn't report the old ones it didn't start the new tests and wasted precious computing time
I also tried to contact K. Brazier because I wanted to ask if he could update SBqueue so it can handle the new gimps requests, but i cant find an email adres at his sbqueue homepage. So if anybody does know his email or another way to contact him, feel free
The offline submission page is up:
http://www.seventeenorbust.com/offline/
I've tested it pretty heavily. It will give you proper credit for primes, composites, and even factors (in preparation for when some tests will be avoided via P-1).
So yeah, try it out and let me know how it works.
Just post the entire contents of your results.txt in the box.
Cheers,
Louie
Regarding factoring... Will factoring be done by all participating in the PRP effort (before the PRP test), or will it be possible to choose only factoring from within the client as it is in GIMPS?
The reason I'm wondering is because some computers are much better at one or the other. Like with a 64 bit computer with 24GB RAM I would prefer to only do factoring.
I understand the benefit of more ram for the second stage of P-1, but I'm not planning to offer a factor only option.
Unlike GIMPS where all factoring is eventually useful, we don't need to factor any further k/n pairs in a k-series once a prime is found, so dedicated factoring will always produce "wasted work" when a prime is finally found because several k/n tests we will never PRP will have had factoring attempted. Because the efficiency gain from more RAM is so small, it's a net loss to do dedicated factoring... due to too many tests inevitably being run.
For our project, delaying the P-1 tests as long as possible is the right choice.
Cheers,
Louie
OK, thanks for the answer. :-)
GIMPS have an option called "GIMPS - What makes sense", if SoB had the same option, would it not be possible to optimize when factorisation is done?
I would like that as well, I don't care much for primality tests. I prefer to only do factoring. In fact, I assumed untill I read the above posts that factoring would be separate from PRP, as it is now, but handled through the new server. I have set all my machines that were doing sieving and factoring to workpreference 4 (P-1 factoring),running mprime-sb, in anticipation of firstpass reaching 17.2M.
Let me also add that Mike's sieving/factoring stats system was the reason I started sieving, and later P-1 factoring, for this project in the first place. It would be a shame to make it obsolete, it's IMO the best stats system of any DC project I have tested.
I used to run 3 of the old client as they completed I replaced the first 2 with mprime clients.
The last old (2nd pass test) expired this morning my time I expired the test it grabbed and didn't start or replace it with another process.
One of the mprime clients was only running one test and now the old client has gone offline it has decided to grab another piece of work to take up the slack.
Is this a correct interpretation of the mprime client behavior?
Would I be better off just running one process and mprime deciding how much work to grab? Ie what would give me biggest throughput?
Thanks
Is this on the same computer? If so, you only need to run one instance of mprime and it will automatically use an appropriate number of tests for your CPU (or you can tell it what you how many to run and on which cores if you don't like what it chooses)
Just a single computer [Pentium D 805 Core 2 Duo running 32bit ubuntu linux]
So what is the best way to reduce to a single instance let on mprime client finish the started tests and then kill it never to be seen again and let the remaining process pick up? Or can I move the partially completed tests into the other directory?
If the test is close to finishing and you are able to monitor it, then that should work.
But, if you want to move it and not worry about waiting for it to finish, you should be able to just move the appropriate p<k>_<n> files and copy the entry from the old client's worktodo file into the other worktodo file and it will pick it up where the other client left off just fine.
Is there a way to set the new client to flush at a certain hour each day? like, 8pm, go flush?
I just tried, this is a copy of the output (xxxx is me to hide actual values):So no points so farResults
[Tue Mar 17 23:22:10 2009]
10223*2^16017341+1 does not need P-1 factoring.
22699*2^16017454+1 does not need P-1 factoring.
[Mon Apr 06 07:58:03 2009]
UID: ML, 10223*2^16017341+1 is not prime. RES64: B573E5C251C19xxx. Wd1: D16186A4,00000000, AID: 00000000000000000000000xxxxxxxx
[Mon Apr 06 08:05:46 2009]
UID: ML, 22699*2^16017454+1 is not prime. RES64: 9AE877ED2A9F5xxx. Wd1: D2438786,00000000, AID: 0000000000000000000000000xxxxxxx
Result Totals
0 valid factors were found in the results and added to the database.
0 completed tests were updated in the database.
hmmm... email those lines to me Frentik. It will help me debug and make it work better. I'll make sure you still get credit.
lhelm (at) seventeenorbust.com
Cheers,
Louie
Done!hmmm... email those lines to me Frentik. It will help me debug and make it work better. I'll make sure you still get credit.
Cheers,
Louie
Thanks Frentik! Your data allowed me to fix the bug that caused your submissions to be denied. Offline submissions should be working again.
Cheers,
Louie
I got some more results so i'll be doing another test run
update:
Yep, al working fine now[Thu Apr 02 06:29:21 2009]
UID: ML/Kirk, 55459*2^16018066+1 is not prime. RES64: CD1165C1C63D7DAD. Wd1: D70B8C4E,00000000, AID: 000000000000000000000000000EDA85
[Thu Apr 02 10:04:12 2009]
UID: ML/Kirk, 10223*2^16017965+1 is not prime. RES64: 06F7EA224DE5A9EB. Wd1: D6418B84,00000000, AID: 000000000000000000000000000EDA84
Result Totals
0 valid factors were found in the results and added to the database.
2 completed tests were updated in the database.
I have a dual boot system running Ubuntu 9.04 x64 and XP-64. There are times where I run Ubuntu and wish to run mprime in the same folder as is in XP. I get the following error:
libcurl.so.4: cannot open shared object file: No such file or directory
What am I doing wrong? I must admit that I am a noob when it comes to Linux. Any help would be appreciated.
Disregard the last post. I had been looking for a solution to this and the dual boot question for about an hour and in haste posted it. I found the answer in the end.
sudo apt-get install curl
Copy mprime in the same folder as that in Win and run
./mprime -m
For those noobs like me who may be looking for the answer.
Posting Permissions
