I just changed how much memory it could use to 4000MB
Checked prime.txt and saw my username missing, stopped, added my name back in, and started the client up.
I see the stats showing up at SOB
Indeed, here is what the web page shows.
I'm at 16% across the board... and doing iterations of .025 secPHP Code:
ID Number Assigned Last Reported Progress Percent cEMs/s
970360 24737•2^13286191+1 Sun Feb 22 03:32:32 2009 Sun Feb 22 20:37:42 2009 16 0 % 6158527
But, looking good so far!
Using VISTA 64 bit
If I am reading your stats correctly, then the 16 percent complete is getting put under the progress column? That would explain my 2's and 1's that show up in my progress column, since the last time Prime95 sent updates, I was at 2 and 1% on all my tests...
Yes, the stats should even show up before it is completed now. And I fixed the stats so progress will have the ( * n / 100.0) term correctly added. The client doesn't report the exact iteration like SB but that estimate should be good and give proper percents now for unfinished tests.
Are the cEMs really wrong? I copied the cEM calculation code out of the current SB server. Let me know if I should look into it further or if it was just confusion with the progress numbers being wrong.
Cheers,
Louie
Thanks for all your testing. First off, your config files are perfect and so this isn't your fault.
But, the linux client is violating my assumption that a client always sends a computer update before requesting work... an assumption which was validated by all other clients. I thought I had code in place to force the client to send this update in the event the guid / username wasn't in the database, but looking at my code now, it looks like that code would only execute if the database query failed, not if it were empty... doh!
So I've fixed that. You should be able to just erase your worktodo file and restart it. Please email your full prime.log file after you try this whether it works or fails. lhelm (at) seventeenorbust.com
Cheers,
Louie
No problem, I like to help
But the good news is, what you did worked, the test that were downloaded are now assigned to my username at the pending test page.
Can a few of you let your tests complete? Since these are all second pass tests, I want to match up some residues and make sure they correlate with past results and are being stored well.
Keep up feedback on benchmarks and errors too. It's interesting to learn what the best settings for multi-core machines are.
Cheers,
Louie
I wouldn't mind finishing a couple of tests, but this poor laptop has heat dissipation issues, so I have to use the Throttle=50 setting in prime.txt to keep it running. It could take some time to finish. Currently at 5.27% and 3.27%. Also note that I'm not seeing any pending tests, I suppose I reserved too early. Running linux 64 bit. Should I release my current tests and talk to the server again?
When I am using SoB (the old client) I normally use SBqueue to give work to them (this because of offline pc's)
Does mprime has its own way to buffer in offline (which I suppose)
Just a question that popped into my head because I suppose mprime asks in a different way "tests" than old SoB client does.
I understand that maybe this doesn't have priority, but just to mention it, in case of later development.
Last edited by woolysheep; 02-23-2009 at 12:56 PM.
mprime/prime95 queues up as many days of work as you tell it to.
if you run ./mprime -m and choose 14. Options/Preferences you can choose, among other things, "days of work to queue up".
Ah oke, that's good, thanks.
But what if I have an offline PC, and I want to export work from an online client to an offline client. Is that possible?
This is perhaps what you need.Originally Posted by undoc.txt
Louie,
If you haven't already, you can check my first test, 10223*2^13272329+1
On my i7, it looks like running 8 workers, 1 cpu each, I get an average of 0.053 sec/iter each (0.0066 sec/iter overall), which is almost exactly the same as 4 workers, 2 cpu's each (0.0065 sec/iter overall).
Any idea how to run it as service under windows ? The readme says:
But i cannot find the option in the app.Checking the "Start at Bootup" menu choice will run prime95 as a service
that starts when your computer boots up
Anyhow.. thanks for the update, I was realy scared that this project was heading to a slow death..
There still seems to be some problem with the PrimeNet user id being erased when doing a manual communication (and, I assume, the daily automatic communication). It does not happen every time, maybe 1/2 the time or so.
If you check your logs around 15:45 AKST (which I think would be 18:45 your time if you are in Texas) you should see a bunch of completion date updates from me, which may shed some light on the problem. The is no indication in my Prime95 logs of any problem... Let me know if I can do anything else to help track down this error.
You need to make sure you have admin rights to see the "Start at Bootup" features. If you are using Vista, that option no longer works, and you need to download one of the special NT service versions:
32bit: ftp://mersenne.org/gimps/winnt258.zip
64bit: ftp://mersenne.org/gimps/win64nt258.zip
Why not just stick a shortcut in your Startup menu?
And the user id problem seems to be fixed now, Louie. Thank you.
IronBits:
It looks like the time between updates is hard-coded to be in days in the source code, so the smallest unit of time you can specify is 1 day.
If we end up forking Prime95 for use with SoB, it would easy to change it to use minutes or hours instead of days (or simply modify it to allow something like 0.1 days).
I' ve done some benchmark-tests. Some values are not logical. Perhaps someone have an idea. I run each test about 15 minutes on my Core2 Quad CPU Q6600 2.40 GHz, Vista 32 bit. Are 15 minutes too short to get comparable values?
SB v.2.5.0, 1 client, transmit intermediate blocks: yes
n = 12620926
19:02:51 n.high = 1569
19:17:09 n.high = 26673 ==> 0,0366 sec/Iter
SB v.2.5.08, 1 client, transmit intermediate blocks: yes
n = 12758729
19:25:15 n.high = 1535
19:39:58 n.high = 27630 ==> 0,0338 sec/Iter
SB v.2.5.08, 1 client, transmit intermediate blocks: no
n = 12758729
19:49:03 n.high = 7375
20:06:03 n.high = 32450 ==> 0,0407 sec/Iter
.......slower than "transmit intermediate blocks: yes"? So I repeated the test:
SB v.2.5.08, 1 client, transmit intermediate blocks: no
n = 13316585
20:54:04 n.high = 1409
21:09:29 n.high = 25362 ==> 0,0386 sec/Iter
SB v.2.5.08, 4 clients, transmit intermediate blocks: yes
n = 13016525
20:09:30 n.high = 1475
20:24:06 n.high = 19175 ==> 0,0495 sec/Iter
n = 13309097
20:10:07 n.high = 1411
20:24:23 n.high = 14110 ==> 0,0674 sec/Iter
n = 13315711
20:10:29 n.high = 1409
20:23:45 n.high = 12681 ==> 0,0706 sec/Iter
n = 13315774
20:10:37 n.high = 1409
20:23:45 n.high = 12681 ==> 0,0699 sec/Iter
.......=> Overall rate: (0,0495+0,0674+0,0706+0,0699)/4 /4 = 0,0161 sec/Iter
SB v.2.5.08, 4 clients, transmit intermediate blocks: no
n = 13316285
20:30:54 n.high = 1409
20:48:16 n.high = 22544 ==> 0,0493 sec/Iter
n = 13316350
20:31:06 n.high = 1409
20:47:37 n.high = 21135 ==> 0,0502 sec/Iter
n = 13316527
20:31:19 n.high = 1409
20:47:31 n.high = 19726 ==> 0,0531 sec/Iter
n = 13316554
20:31:29 n.high = 1409
20:47:46 n.high = 19726 ==> 0,0533 sec/Iter
.......=> Overall rate: (0,0493+0,0502+0,0531+0,0533)/4 /4 = 0,0129 sec/Iter
Prime 95 v.25.9 build 3, 1 worker, 1 CPU/worker, Run on any CPU
n = 13316787 .......... ==> 0,0380 sec/Iter
Prime 95 v.25.9 build 3, 1 worker, 4 CPU/worker, Run on any CPU
n = 13316863 .......... ==> 0,0220 sec/Iter
Prime 95 v.25.9 build 3, 4 workers, 1 CPU/worker, Smart assignment
n = 13317068 .......... ==> 0,0473 sec/Iter
n = 13317106 .......... ==> 0,0488 sec/Iter
n = 13317449 .......... ==> 0,0470 sec/Iter
n = 13317511 .......... ==> 0,0477 sec/Iter
.......=> Overall rate: (0,0473+0,0488+0,0470+0,0477)/4 /4 = 0,0119 sec/Iter
Prime 95 v.25.9 build 3, 4 workers, 1 CPU/worker, Run on any CPU
n = 13317607 .......... ==> 0,0478 sec/Iter
n = 13317629 .......... ==> 0,0467 sec/Iter
n = 13317703 .......... ==> 0,0472 sec/Iter
n = 13317737 .......... ==> 0,0464 sec/Iter
.......=> Overall rate: (0,0478+0,0467+0,0472+0,0464)/4 /4 = 0,0118 sec/Iter
jimmy
More good news. I figured out how to assign P-1 factoring work through our custom primenet server.
After a few kinks were worked out, I even have it storing the results in the database when they are completed too! And it gives credit for completing 2 tests in the form of cems now.
There are a couple minor bumps in the road to rolling this out as mainstream though. There is a one-line bug in the Prime95 code that prevents optimal P-1 factoring from being executed before a PRP test. It's simple to fix and I'm sure George Woltman will want to correct it once I show him. We've been corresponding and he's busy til March so this can wait til then.
But if this were changed in (a possible) v26.0 or v29.55 of Prime95, that would be sweet. Then we could automatically have the P-1 tests done right before a test and the need to do separate P-1 testing would be diminished. I realize we'd be giving up a little bit by not always having computers with high mem to do it, but that's a small edge to give up in exchange for always doing at least some P-1 testing.
System wide, this would give us a 3% efficiency boost or so.
Cheers,
Louie
Excellent news!
I finished those work units and it rolled right on into the next without any problems.
It would be nice if this table could be updated: http://www.seventeenorbust.com/stats/byOS.mhtml
Adding in XP 64bit as well as Vista 32 and 64bit.
And while you are there, you might as well plug in Windows 7
That's great news! Perhaps he would be willing to also modify it to allow < 1 day update intervals. I believe all it would take is changing a couple lines in commonc.c and commonc.h to change the DAYS_BETWEEN_CHECKINS variable to a float and one line to allow it to have values < 1.
I also finished 4 tests this morning using Prime95. (None prime - shucks!) Entires from results.txt are below.
UID: enderak, 55459*2^13285954+1 is not prime. RES64: C0225A46C9E9434A. Wd4: A948BD95,00000000, AID: 000000000000000000000000000ECE72
UID: enderak, 55459*2^13285786+1 is not prime. RES64: AD3FF9FBC307A98C. Wd4: A7F8BC45,00000000, AID: 000000000000000000000000000ECE70
UID: enderak, 55459*2^13285714+1 is not prime. RES64: 3C78FD863D818BDC. Wd4: A768BBB5,00000000, AID: 000000000000000000000000000ECE6E
UID: enderak, 24737*2^13285471+1 is not prime. RES64: 7E209083A58982B0. Wd4: A582B9CF,00000000, AID: 000000000000000000000000000ECE6B
2 updates:
1) I added new first-pass tests into SB's queue. The test server for the new client is going to pull tests from this too, although, they will be a bit higher n values than the rest of the network is starting at.
2) I enabled a nifty feature that allows the primenet server to signal clients that work is no longer needed. This will be useful for when we find the next prime... I'll be able to immediately cancel out all work for that k value, even ongoing tests.
Anyway, to test #1 and #2, instead of waiting for a prime, I'm just telling all your clients to no longer test your current assignments and to get new, first-pass tests right now. Someone post me a log report of this happening to your client. Or let me know if anything goes haywire.
I think change #1 is going to greatly increase our chances of finding a prime soon and #2 is going to be a nice bonus when it happens.
Cheers,
Louie
Worked here too. Log is below. Had to manually expire the old tests on the website - not sure if this little thing is worth bothering with though.
PHP Code:
[Sun Mar 01 08:09:19 2009 - ver 25.9]
Updating computer information on the server
URL: http://www.seventeenorbust.com/v5server/?v=0.95&px=GIMPS&t=uc&g=13eb7d72568d7545a05242d8510cad54&hg=f265e46cf94178f7d658297482f04166&wg=acbc0c3d4c0859d102ae38294b75086c&a=Windows64,Prime95,v25.9,build+3&c=Intel(R)+Core(TM)+i7+CPU+++++++++965++@+3.20GHz&f=Prefetch,SSE,SSE2,SSE4&L1=32&L2=256&np=4&hp=2&m=12285&s=3238&h=24&r=1024&L3=8192&ss=41&sh=0267B8AC07F993B5B1DEE0FA38DB3A0A
RESPONSE:
pnErrorResult=0
pnErrorDetail=SUCCESS
g=13eb7d72568d7545a05242d8510cad54
u=enderak
un=enderak
=END=
Sending expected completion date for 10223*2^13285529+1: Mar 03 2009
URL: http://www.seventeenorbust.com/v5server/?v=0.95&px=GIMPS&t=ap&g=13eb7d72568d7545a05242d8510cad54&k=000000000000000000000000000ECE6C&stage=PRP&c=0&p=60.4786&d=86400&e=184938&ss=18467&sh=F4FB4AFECB268C579ED64EE873CA7AD0
RESPONSE:
pnErrorResult=47
pnErrorDetail=ERROR_WORK_NO_LONGER_NEEDED
=END=
PrimeNet error 47: Unknown error code
ERROR_WORK_NO_LONGER_NEEDED
Sending expected completion date for 55459*2^13285594+1: Mar 06 2009
URL: http://www.seventeenorbust.com/v5server/?v=0.95&px=GIMPS&t=ap&g=13eb7d72568d7545a05242d8510cad54&k=000000000000000000000000000ECE6D&c=0&p=0.0000&d=86400&e=423024&ss=6334&sh=DDCB1AF77DE1DCBDC20C2A5C7195AB8E
RESPONSE:
pnErrorResult=47
pnErrorDetail=ERROR_WORK_NO_LONGER_NEEDED
=END=
PrimeNet error 47: Unknown error code
ERROR_WORK_NO_LONGER_NEEDED
Getting assignment from server
URL: http://www.seventeenorbust.com/v5server/?v=0.95&px=GIMPS&t=ga&g=13eb7d72568d7545a05242d8510cad54&c=0&ss=26500&sh=C5ECFAB5AD716980EEBCA8EF3251639D
RESPONSE:
pnErrorResult=0
pnErrorDetail=SUCCESS
k=000000000000000000000000000ED3DF
w=150
A=24737
b=2
n=16000423
c=1
sf=54.0
saved=2.5
=END=
Got assignment 000000000000000000000000000ED3DF: PRP 24737*2^16000423+1
Sending expected completion date for 24737*2^16000423+1: Mar 07 2009
URL: http://www.seventeenorbust.com/v5server/?v=0.95&px=GIMPS&t=ap&g=13eb7d72568d7545a05242d8510cad54&k=000000000000000000000000000ED3DF&c=0&p=0.0000&d=86400&e=563569&ss=19169&sh=1E88E88AF7DF85E4E3ECA95AD26F54D2
RESPONSE:
pnErrorResult=0
pnErrorDetail=SUCCESS
=END=
Sending expected completion date for 21181*2^13285772+1: Mar 03 2009
URL: http://www.seventeenorbust.com/v5server/?v=0.95&px=GIMPS&t=ap&g=13eb7d72568d7545a05242d8510cad54&k=000000000000000000000000000ECE6F&stage=PRP&c=1&p=64.0264&d=86400&e=152179&ss=15724&sh=A20DEEDE3EA3033EF4AEF4EDB0BF81D1
RESPONSE:
pnErrorResult=47
pnErrorDetail=ERROR_WORK_NO_LONGER_NEEDED
=END=
PrimeNet error 47: Unknown error code
ERROR_WORK_NO_LONGER_NEEDED
Getting assignment from server
URL: http://www.seventeenorbust.com/v5server/?v=0.95&px=GIMPS&t=ga&g=13eb7d72568d7545a05242d8510cad54&c=1&ss=11478&sh=A319A5625A9D460108CCD077714B9A77
RESPONSE:
pnErrorResult=0
pnErrorDetail=SUCCESS
k=000000000000000000000000000ED3E0
w=150
A=10223
b=2
n=16000457
c=1
sf=54.0
saved=2.5
=END=
Got assignment 000000000000000000000000000ED3E0: PRP 10223*2^16000457+1
Sending expected completion date for 10223*2^16000457+1: Mar 07 2009
URL: http://www.seventeenorbust.com/v5server/?v=0.95&px=GIMPS&t=ap&g=13eb7d72568d7545a05242d8510cad54&k=000000000000000000000000000ED3E0&c=1&p=0.0000&d=86400&e=563570&ss=29358&sh=263B1DF155A2BDBB1820F9D3DF1210F8
RESPONSE:
pnErrorResult=0
pnErrorDetail=SUCCESS
=END=
Sending expected completion date for 24737*2^13285903+1: Mar 03 2009
URL: http://www.seventeenorbust.com/v5server/?v=0.95&px=GIMPS&t=ap&g=13eb7d72568d7545a05242d8510cad54&k=000000000000000000000000000ECE71&stage=PRP&c=2&p=63.2127&d=86400&e=155622&ss=26962&sh=A83F28A500F076ADD56038343B61E4DB
RESPONSE:
pnErrorResult=47
pnErrorDetail=ERROR_WORK_NO_LONGER_NEEDED
=END=
PrimeNet error 47: Unknown error code
ERROR_WORK_NO_LONGER_NEEDED
Getting assignment from server
URL: http://www.seventeenorbust.com/v5server/?v=0.95&px=GIMPS&t=ga&g=13eb7d72568d7545a05242d8510cad54&c=2&ss=24464&sh=1AA0F3A6B0100859328034157AAFB815
RESPONSE:
pnErrorResult=0
pnErrorDetail=SUCCESS
k=000000000000000000000000000ED3E1
w=150
A=22699
b=2
n=16000534
c=1
sf=54.0
saved=2.5
=END=
Got assignment 000000000000000000000000000ED3E1: PRP 22699*2^16000534+1
Sending expected completion date for 22699*2^16000534+1: Mar 07 2009
URL: http://www.seventeenorbust.com/v5server/?v=0.95&px=GIMPS&t=ap&g=13eb7d72568d7545a05242d8510cad54&k=000000000000000000000000000ED3E1&c=2&p=0.0000&d=86400&e=563573&ss=5705&sh=83B70DA8BAFB90682C2001CC181C7757
RESPONSE:
pnErrorResult=0
pnErrorDetail=SUCCESS
=END=
Sending expected completion date for 10223*2^13285961+1: Mar 03 2009
URL: http://www.seventeenorbust.com/v5server/?v=0.95&px=GIMPS&t=ap&g=13eb7d72568d7545a05242d8510cad54&k=000000000000000000000000000ECE73&stage=PRP&c=3&p=61.4428&d=86400&e=180432&ss=28145&sh=B0BDD0114C9DCE63059942A187EDB07F
RESPONSE:
pnErrorResult=47
pnErrorDetail=ERROR_WORK_NO_LONGER_NEEDED
=END=
PrimeNet error 47: Unknown error code
ERROR_WORK_NO_LONGER_NEEDED
Getting assignment from server
URL: http://www.seventeenorbust.com/v5server/?v=0.95&px=GIMPS&t=ga&g=13eb7d72568d7545a05242d8510cad54&c=3&ss=23281&sh=EA25579F78B4C74EE87AC1285A493B91
RESPONSE:
pnErrorResult=0
pnErrorDetail=SUCCESS
k=000000000000000000000000000ED3E2
w=150
A=24737
b=2
n=16000591
c=1
sf=54.0
saved=2.5
=END=
Got assignment 000000000000000000000000000ED3E2: PRP 24737*2^16000591+1
Sending expected completion date for 24737*2^16000591+1: Mar 07 2009
URL: http://www.seventeenorbust.com/v5server/?v=0.95&px=GIMPS&t=ap&g=13eb7d72568d7545a05242d8510cad54&k=000000000000000000000000000ED3E2&c=3&p=0.0000&d=86400&e=563575&ss=16827&sh=1922280242BAE358DACF839EEAA0932B
RESPONSE:
pnErrorResult=0
pnErrorDetail=SUCCESS
=END=
Under the Options>Preferences menu, what is your "Days of work to queue up" set to? The default I think is 5, but I set mine at 1 to prevent checking out tests way ahead of time. I think you could also set it at 0 to not check any out until you submit your current tests.
I've put some linux x64 systems systems online (in total 0.2THz) all seems to work fine. Hope they update this evening so I can see some stats soon .