climateprediction.net home page
Posts by adrianxw

Posts by adrianxw

61) Message boards : Number crunching : Iceworlds & Slowdowns hadsm3/mh - Closed - Discussion (Message 34545)
Posted 6 Aug 2008 by Profile adrianxw
Post:
I recall well the older Met Office systems at Bracknel. The Cosmos, Cray II, the ETA-10 debarcle, Cray Y-MP etc. Not terribly productive, but a lot of memories stirred!

I really quite liked programming in Fortran. I\'m a C++ jock now though, haven\'t used Fortran professionally for more then 10 years. (Carl wanted to know if I wanted a job - is how I know a little about the code base!).
62) Message boards : Number crunching : Iceworlds & Slowdowns hadsm3/mh - Closed - Discussion (Message 34539)
Posted 6 Aug 2008 by Profile adrianxw
Post:
Could be almost anything. Without a detailed knowledge of the codebase, I wouldn\'t know where to hazard a guess. I don\'t even know what compilers are used by the project, let alone options or libraries.

I know, (Carl told me), that there is a lot of Fortran in there. I also recall weirdness when, ~20 years back, I was porting a big Fortran-77 fluid dynamics application from a Gould 32/77 running MPX-32 to a VAX 11-750 under VMS. Theoretically, they were both standard compliant 32 bit systems with standard compliant compilers, but the application results differed from the same source data - we traced it to the runtime libraries and had to write our own versions of some of the mathematics functions.
63) Message boards : Number crunching : Odd looking graphic. (Message 34537)
Posted 6 Aug 2008 by Profile adrianxw
Post:
I don\'t normally look at the graphics, but as I had a couple of iceballs recently, when I see anything odd, I have started to do so. Today I found a wu at exactly 66.666% and not progressing. Now, I am aware that when it changes from phase to phase there is a period of \"different\" activity, so have not done anything about it yet, but the graphic looked odd...

64) Message boards : Number crunching : Iceworlds & Slowdowns hadsm3/mh - Closed - Discussion (Message 34466)
Posted 1 Aug 2008 by Profile adrianxw
Post:
I asked this once before and don\'t think I ever got an answer. Can I somehow, (obviously not through the regular BOINC mechanism), get EXACTLY the same wu again to run on another machine? There is a second Q6600 machine, (B4 stepping and not OC\'d), sitting 1m from the G0 Q6600. It normally does not run CPDN but there is no reason why I shouldn\'t stick it on, and simply reduce the quota on the other quad to even things out.

I am happy to run it again as, although it may fail, I get the trickles and so am not really losing anything, and it might just help.

PM me if you like.
65) Message boards : Number crunching : Iceworlds & Slowdowns hadsm3/mh - Closed - Discussion (Message 34462)
Posted 1 Aug 2008 by Profile adrianxw
Post:
I had 3 models on that quad. It has a 25% resource share, so theoretically, only 1 was running, but as LHC and Cels have no work at the moment, it gets more. The one that had been running pretty much all the time, and that which was furthest down the tree were the iceballs. The middle wu is running and trickling normally.

The machine is a Intel Q6600 and the advanced G0 stepping model. It is rated for 2.4GHz and clocked at 3.0GHz. It is in an open chasis with a huge Zalman CNPS9700 heatsink on it running at full speed. Speedfan shows the cores 38-39C but Speedfan under reports these chips by 15C so they are running at less then 55C which for a G0 is nothing. I\'ve torture tested the thing a couple of times without issue. It has crunched a good number of CPDN\'s. I don\'t think it exhibits any instability but you know what CPDN is like!

As always, if there is anything else you need, post or PM me.
66) Message boards : Number crunching : Iceworlds & Slowdowns hadsm3/mh - Closed - Discussion (Message 34442)
Posted 31 Jul 2008 by Profile adrianxw
Post:
Back to back iceballs 7537005 and 7553089. Blue globe, slow down, all the usual.
67) Message boards : Number crunching : Disappearing cobblestones (Message 34211)
Posted 3 Jul 2008 by Profile adrianxw
Post:
Our members big pluses and minuses seem to have reversed themselves today.
68) Message boards : Number crunching : Disappearing cobblestones (Message 34162)
Posted 27 Jun 2008 by Profile adrianxw
Post:
I have one team member that has picked up a -24000, and another that has +24000. Something going on that\'s for sure!
69) Message boards : Number crunching : Aborted a crawler. (Message 33875)
Posted 19 May 2008 by Profile adrianxw
Post:
This wu after 32 hours had crawled to 1.6% complete. Obviously one of the crawlers, so I\'ve aborted it.
70) Message boards : Number crunching : How can I limit climateprediction to run only one task at a time? (Message 33752)
Posted 13 May 2008 by Profile adrianxw
Post:
On my quads, I let it download some wu\'s, then set the project to \"No New Tasks\", then \"Suspend\" all but, (say), one of the wu\'s. When that wu is finished, I \"Resume\" another. So on until they are all gone, then I download some more.

That way I can have however many CPDN wu\'s running as I like. With the real long deadlines, having a few non-started wu\'s lying around for a few months is no issue.
71) Message boards : Number crunching : Task is overdue (Message 32211)
Posted 17 Jan 2008 by Profile adrianxw
Post:
IE7 is CPU hog only if one of pages contains really nasty javascript,so badly written,that it needs huge amount of proccessing.One of the more visible is \"falling snow\".If no such script runs IE7 is really inactive...

There are other reasons IE can hog resources if left running. I had found that if I used a site which required the QuickTime plugin, sometimes IE would seem to freeze but the Task Manager shows it using very lareg amounts of CPU.

Whilst I appreciate this is not an IE7 problem per-se, a casual Windows user will not necessarily comprehend the plugin concept - to a casual glance, IE7 is using the CPU.

Always remember that there are people out theer that just turn the machine on, press the button the usually press and wait for the result, they do not consider the machinations or consequences of their actions.

*** EDIT ***

Uninstall/Reinstall QuickTime fixed this particular issue.
72) Message boards : Number crunching : Trickles not showing. (Message 32191)
Posted 16 Jan 2008 by Profile adrianxw
Post:
Grabbed a new model, this one, an \"SM\", and it appears to be running as expected. It has sent a scoring trickle, which upon back calculation gives a c/hr of 23 as expected.

I will allow the model to complete and then force it to run a \"CM\" to see if the problem was specific to the original configuration of the machine, (and/or any resulting fallout), or generic.
73) Message boards : Number crunching : Trickles not showing. (Message 32190)
Posted 15 Jan 2008 by Profile adrianxw
Post:
Okay, I\'ll abort this one, which I hate to do, and get another to see if it performs better.
74) Message boards : Number crunching : Trickles not showing. (Message 32186)
Posted 15 Jan 2008 by Profile adrianxw
Post:
Following up a bit, I found an issue with the disk drive, (more specifically with the disk drive cable), which was generating a few errors in the event log. This is now fixed so I switched the model on again to see if that was the cause.

When I did, it dropped from 1.101% 78:38:20 back to 1.092% 77:38:05, (machine had been down as I was replacing the cable). I let it run for an hour and it went up to 1.101% at 78:41:02 so basically it had crunched 0.009 in just over an hour.

If the trickles come at 1.38% intervals, that gives ~153 hours per trickle, or roughly 2.2 credit per CPU hour, which is a long way from the 23 c/hr quoted earlier.

The assumption there is that the progress per unit time is a straight line - is that a fair assumption, (or within reason anyway)?

If so, I do not understand the stupifying under-performence of the wu on this box. Projects like POEM, Einstein, Rosetta, MalariaControl, SIMAP etc. are producing much the same RAC as my other Q6600.

It is not lack of resources. The task is using less physical and virtual memory then several of the other projects. The 2 machines have different MoBo\'s but one is a top notch Asus, the other an equally good MSI. RAM is the same, disks are the same, OS is the same, graphics cards are the same, heatsinks are the same, clocking is the same. Cabinets are different colours ;)

Weird stuff going on in North Sjælland.
75) Message boards : Number crunching : Trickles not showing. (Message 32044)
Posted 6 Jan 2008 by Profile adrianxw
Post:
I\'ve suspended CPDN right now. I\'ll play around with it a bit more as time permits. Some of the numbers I needed have come out of this thread anyway, so I know what I should be seeing.
76) Message boards : Number crunching : Trickles not showing. (Message 32042)
Posted 6 Jan 2008 by Profile adrianxw
Post:
I have just run several passes of the diagnostics I use, and the h/w seems to check out. What is odd is that I can\'t access the Windows EventLog. I am re-installing Windows at the moment.

When I first built it, loaded Windows and BOINC, I left it running for 3 days crunching just SIMAP. It was RAC\'ing over 2000 a day which is about right.

<fx>Shrugs</fx> Beats me what\'s up. It is ripping through Rosetta, SIMAP, MCDN and POEM. There are a couple of other projects on there with very small quotas, but even they are churning work out.
77) Message boards : Number crunching : Trickles not showing. (Message 32039)
Posted 6 Jan 2008 by Profile adrianxw
Post:
It says in the graphics that it is on step 8069 of 259248, 1.04% crunched, and 71:19:41 of CPU time to get there.
78) Message boards : Number crunching : Trickles not showing. (Message 32034)
Posted 6 Jan 2008 by Profile adrianxw
Post:
The 1GB is an unfortunate temporary situation. When I ordered the components for this system, I had ordered 2GB of Corsair 4.4.4.12 RAM. Upon building the system and testing it, Memtest-86 v3.3 showed a fault in the RAM, which moved upon swapping the modules around. It has been sent back. The result is that the machine running CPDN is running with a stick stolen from this machine, (another Q6600). At this time, the system is not overclocked, although as it is a G0 chip and has a huge Zalman on it, on a top-notch MoBo, it probably will be when the right memory arrives.

There does not appear to be any commonality between the messages preceeding the trickle-up. There are 10 projects on the machine, and most of the messages refer to regular events from these. Suspending, Restarting, Uploading - your familiar with this stuff. Just searched stdoutae.txt, no memory issues, just the confirmation of memory usage when BOINC starts and reflects the defaults for the location of the machine. No apparent errors.% done seems to be gradually increasing, not jumping about.

Is showing 25% CPU use, as expected. Similar resource figures as POEM, about half of what Rosetta is using. Remember, the model is fairly new, ISTR that the demands go up as the model progresses. Leave in memory is set as usual, no graphics running.

If, as you say, it will not generate a \"real\" trickle until it gets to 1.38% then that is probably why, since it has not. I am suprised to see it performing as poorly as that however when other members of my team with regular hardware are getting scoring trickles every day or so. If the machine trickled now and got 311 for it, that would be a credit/hour of less then 5, hugely below what it is acheiving anywhere else.

79) Message boards : Number crunching : Trickles not showing. (Message 32030)
Posted 6 Jan 2008 by Profile adrianxw
Post:
02-Jan-2008 21:17:27 [climateprediction.net] Sending scheduler request: To send trickle-up message.
02-Jan-2008 21:17:32 [climateprediction.net] Scheduler request succeeded:
03-Jan-2008 01:03:06 [climateprediction.net] Sending scheduler request: To send trickle-up message.
03-Jan-2008 01:03:15 [climateprediction.net] Scheduler request succeeded:
03-Jan-2008 17:21:23 [climateprediction.net] Sending scheduler request: To send trickle-up message.
03-Jan-2008 17:21:28 [climateprediction.net] Scheduler request succeeded:
03-Jan-2008 18:14:24 [climateprediction.net] Sending scheduler request: To send trickle-up message.
03-Jan-2008 18:14:27 [climateprediction.net] Scheduler request succeeded:
04-Jan-2008 21:22:37 [climateprediction.net] Sending scheduler request: To send trickle-up message.
04-Jan-2008 21:22:42 [climateprediction.net] Scheduler request succeeded:
06-Jan-2008 10:48:28 [climateprediction.net] Sending scheduler request: To send trickle-up message.
06-Jan-2008 10:48:34 [climateprediction.net] Scheduler request succeeded:

I have trimmed the \"got 0 new tasks\" and \"requesting...\" parts of the messages to prevent phattening.

The models have changed since I last ran CPDN. It used to run with 30% quote on a 3.2GHz Prescott HT. In that rig, it got pretty much 1 trickle per day. It is now running at 25% on a 2.4GHz Q6600, so basically has exclusive use of 1 core. I would have expected significantly more to be done per 24 hour period in this rig.

These \"I\'ve resumed\" trickles must be something new, I don\'t recall them when I was running CPDN before. The \"real\" trickles must be larger then the old. The model is at 1.022% so is probably why there are no trickles.

I can see other people getting trickles reasonably regularly however, and getting 311 credit for them. See link in previous post.

80) Message boards : Number crunching : Trickles not showing. (Message 32022)
Posted 5 Jan 2008 by Profile adrianxw
Post:
I picked up this wu a couple of days ago. I can see it sending trickle_up messages, yet the result shows \"No Trickles\".

I saw that there had been some credit problems around Christmas, but can also see that my team mates trickles are appearing normally.

Anything to check?


Previous 20 · Next 20

©2024 climateprediction.net