Bugfix Release

Message boards : News : Bugfix Release

To post messages, you must log in.

AuthorMessage
Timo Strunk
Volunteer moderator
Project administrator

Send message
Joined: 6 Sep 07
Posts: 560
Credit: 10,729,563
RAC: 0
Message 5241 - Posted: 13 Jan 2012, 16:01:07 UTC

There has been a short bugfix release for linux. OpenSUSE users will hopefully not segfault anymore. The problem was the upx compression of the executable.
Es gab einen kleinen Bugfix Release für Linux. OpenSUSE user sollten nun keine Probleme mehr mit Segfaults haben. Das Problem war die UPX Kompression der Files.
ID: 5241 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Stef42

Send message
Joined: 4 Jan 12
Posts: 11
Credit: 5,852,568
RAC: 0
Message 5242 - Posted: 13 Jan 2012, 16:25:31 UTC - in response to Message 5241.  

Can you tell something about the development of the OpenCL version? (fix for nvidia, speed improvement, etc.)
ID: 5242 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Nolion

Send message
Joined: 29 Nov 11
Posts: 2
Credit: 62,278,552
RAC: 0
Message 5254 - Posted: 14 Jan 2012, 2:38:57 UTC - in response to Message 5241.  

Anyone else seeing this? My credit/second is up by roughly a factor of 6 on the new application (Ubuntu 11.04 x86_64, HD4850). Not that I'm complaining! These were the 10 most recent WUs on this host when I checked:


Task Work unit Run time(sec) CPU time(sec) Credit Application
61433919 56713093 6,186.60 4,055.59 28,687.19 POEM++ OpenCL version v0.02 (opencl_ati_100)
61433915 56713090 6,076.56 4,012.73 28,176.92 POEM++ OpenCL version v0.02 (opencl_ati_100)
61433911 56713086 6,264.02 4,222.96 29,046.19 POEM++ OpenCL version v0.02 (opencl_ati_100)
61433908 56713083 5,265.54 4,070.52 24,416.23 POEM++ OpenCL version v0.02 (opencl_ati_100)
61433897 56713077 6,427.80 4,328.41 29,805.62 POEM++ OpenCL version v0.02 (opencl_ati_100)
61426769 56707521 6,935.93 4,841.34 4,987.44 POEM++ OpenCL version v0.01 (opencl_ati_100)
61426760 56707512 6,573.06 4,473.11 4,778.46 POEM++ OpenCL version v0.01 (opencl_ati_100)
61426747 56707499 6,418.88 4,287.75 4,716.40 POEM++ OpenCL version v0.01 (opencl_ati_100)
61426740 56707492 6,207.01 4,022.81 4,417.70 POEM++ OpenCL version v0.01 (opencl_ati_100)
61426733 56707485 6,226.32 4,149.62 4,393.48 POEM++ OpenCL version v0.01 (opencl_ati_100)
ID: 5254 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile mmstick

Send message
Joined: 7 Jan 12
Posts: 44
Credit: 34,755,949
RAC: 0
Message 5258 - Posted: 14 Jan 2012, 4:21:56 UTC

I wish I could at least get a Work Unit, I haven't gotten a POEM++ or OpenCL work unit in the last day with my two 6850s and 4870 since the servers always out saying "not reporting or requesting tasks"
ID: 5258 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile kashi

Send message
Joined: 22 Dec 09
Posts: 65
Credit: 76,582,573
RAC: 0
Message 5261 - Posted: 14 Jan 2012, 6:15:27 UTC - in response to Message 5258.  
Last modified: 14 Jan 2012, 6:28:17 UTC

I had trouble getting a decent size cache when I was using BOINC 7.0.3. Would only receive "not reporting or requesting tasks" message and no new tasks were requested until the cache of unstarted tasks reduced to zero.

That's why I switched back to 7.0.2. I prefer to keep a cache of a few days tasks on non quorum projects when it has been raining heavily as this sometimes causes the internet connection to fail. With quorum projects I usually limit my cache to 1 day or less so I'm not causing any pending tasks delay for wingmen.

Also you won't receive any POEM++ CPU tasks unless you have included that application in your app_info.xml file.
ID: 5261 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile mmstick

Send message
Joined: 7 Jan 12
Posts: 44
Credit: 34,755,949
RAC: 0
Message 5262 - Posted: 14 Jan 2012, 8:53:27 UTC

What is the proper syntax for including the CPU version in app_info.xml?
ID: 5262 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile kashi

Send message
Joined: 22 Dec 09
Posts: 65
Credit: 76,582,573
RAC: 0
Message 5263 - Posted: 14 Jan 2012, 10:03:47 UTC - in response to Message 5262.  
Last modified: 14 Jan 2012, 10:50:43 UTC

Not sure, I can adjust parameters without mucking things up but I'm not confident about composing an app_info.xml file without using the information in client_state.xml as a template, possibly add something like this to your GPU OpenCL app_info.xml file for Windows:

<app>
<name>poempp</name>
<user_friendly_name>POEM++</user_friendly_name>
</app>
<file_info>
<name>poempp_0.8_windows_intelx86</name>
<executable/>
</file_info>
<app_version>
<app_name>poempp</app_name>
<version_num>8</version_num>
<file_ref>
<file_name>poempp_0.8_windows_intelx86</file_name>
<main_program/>
</file_ref>
</app_version>

and perhaps for Linux 64 you would substitute poempp_0.9_x86_64-pc-linux-gnu for <name> and <file_name> and change <version_num> value to 9.

Applications can be downloaded from http://boinc.fzk.de/poem/download/.
ID: 5263 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Christopher Herr
Avatar

Send message
Joined: 13 Mar 10
Posts: 97
Credit: 1,505,265,234
RAC: 0
Message 5383 - Posted: 23 Jan 2012, 11:15:05 UTC

While we are on the topic of bugfixing:

Again there is no GPU work, but something else caught my eye more than once since the GPU apps have been distributed and I wonder if no work available is one possible result of a scheduler going "berserk" due to this:

Take a look at the server status page and tell me I am crazy, but the server clock is is fifteen to twenty minutes ahead, is it not? I have no clue about the server side code of BOINC, but that is not normal and the Karlsruher Institute of Technology (KIT) has for sure its own NTP server, but then that is totally going wrong, too...
ID: 5383 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
waitingForTheMiracle

Send message
Joined: 27 Nov 11
Posts: 21
Credit: 20,120,682
RAC: 0
Message 5384 - Posted: 23 Jan 2012, 12:16:01 UTC

Yes, the server clock has been about 15 minutes off since I came here (early Jan). But I doubt it has more than a minor impact.
The main issue seems to be that we have more GPU crunching power than the project knows what to do with.

According to boincstats the project as a whole generated about a million points per day prior to the release of the GPU app.
Four weeks later, it's more than 30 million points on a good day (a day without GPU work scarcity).
If 90% of those points are from GPUs, at 3k points per WU, we need close to 10k WUs every day just to feed the GPUs that already signed up.


(Care to guess when the GPU app was released, and which days the project was running low on WUs?)

Looks like a runaway success to me, probably enough of a success to create issues on the backend. It takes time to scale infrastructure (generating work, storing and analyzing results) to deal with order-of-magnitude growth.

So I guess that "out of GPU work" is simply a symptom of bottlenecks elsewhere. I imagine it will stay that way for a while.
ID: 5384 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Timo Strunk
Volunteer moderator
Project administrator

Send message
Joined: 6 Sep 07
Posts: 560
Credit: 10,729,563
RAC: 0
Message 5386 - Posted: 23 Jan 2012, 14:44:37 UTC

Hi everybody,

totally true, but the bottleneck is the team at the moment. The GPU workunits take time to postprocess. I will start new ones today. Hopefully this shortage will be out for some time.

Best,
Timo
ID: 5386 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Christopher Herr
Avatar

Send message
Joined: 13 Mar 10
Posts: 97
Credit: 1,505,265,234
RAC: 0
Message 5388 - Posted: 23 Jan 2012, 15:02:29 UTC - in response to Message 5386.  
Last modified: 23 Jan 2012, 15:04:55 UTC

Hi everybody,

totally true, but the bottleneck is the team at the moment. The GPU workunits take time to postprocess. I will start new ones today. Hopefully this shortage will be out for some time.

Best,
Timo

Ah okay, thank you for this honest answer and for keeping the supply coming! :-)

Though I do not think you all hope for less GPU power, right? ;-)

The first release of the Windows GPU app is working out nicely at least for me, so massive kudos here! :-D

Something completely else: when will the first GPU results be put into some form of a visible scientific result like a paper or an article in a magazine?
I am sorry, if I am getting ahead of myself and I am asking foolish questions!
ID: 5388 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
BarryAZ

Send message
Joined: 12 Nov 08
Posts: 94
Credit: 171,936,261
RAC: 0
Message 5396 - Posted: 24 Jan 2012, 1:23:27 UTC

Looks like you are already out of disk space - I've some CPU uploads that are failing for lack of space.


ID: 5396 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
pvh

Send message
Joined: 6 Mar 10
Posts: 31
Credit: 459,381,173
RAC: 0
Message 5397 - Posted: 24 Jan 2012, 5:34:27 UTC - in response to Message 5261.  

I had trouble getting a decent size cache when I was using BOINC 7.0.3. Would only receive "not reporting or requesting tasks" message and no new tasks were requested until the cache of unstarted tasks reduced to zero.


The solution for this problem is in this thread

http://boinc.fzk.de/poem/forum_thread.php?id=659

Works like a charm. I have 43 WUs in my cache now! This is with BOINC 7.0.8.
ID: 5397 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote

Message boards : News : Bugfix Release


Copyright © 2017 KIT-INT