Message boards :
Number crunching :
Issue with LHC@home preferences
Message board moderation
Author | Message |
---|---|
Send message Joined: 14 Jan 10 Posts: 1268 Credit: 8,421,616 RAC: 2,139 |
I selected only Theory application. I had set 2 for Max # of jobs for this application and for Max # of CPUs for this project 1. When asking new CPU-work I don't get 2 tasks, but when I set Max # of CPUs to 2, I get a second task. |
Send message Joined: 28 Sep 04 Posts: 674 Credit: 43,161,776 RAC: 15,924 |
That seems to be same as for Atlas tasks. The Max number of CPUs limits the Max number of tasks you receive from the server. I have never understood why. |
Send message Joined: 15 Nov 14 Posts: 602 Credit: 24,371,321 RAC: 0 |
I have never understood why. You can read the official explanation: https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4686&postid=35258#35258 You can deal with it using an app_config.xml if you are using native ATLAS. (I don't think it works with the VBox version.) https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4686&postid=35121 I expect that the LHC admins would like to fix it too, but it is apparently out of their hands. |
Send message Joined: 12 Jun 18 Posts: 126 Credit: 52,457,949 RAC: 23,953 |
That explains the bizarre behavior of ATLAS but not nTheory. #WUs = Unlimited #CPU = 1, 2, 3, 4, 5, 6, 7, 8 WU max: 2, 4, 6, 8, 10, 10, 10, 10 |
Send message Joined: 15 Jun 08 Posts: 2386 Credit: 222,992,515 RAC: 136,379 |
Isn't it time to ignore the accountants and get rid of this misinterpretation of #CPUs? Experienced users find a couple of workarounds to get their multi CPU hosts saturated. Just set up multiple clients on the same box, each with only a few "CPUs". The preferences should work as an average volunteer expects them to work: Max # jobs -> sets the max number of tasks a BOINC client can hold in it's buffer Max # CPUs -> sets the number of cores/threads a task/VM will use No "special" interpretation for project X and a different one for project Y. |
©2024 CERN