| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: fast hardware profiles switching |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 7
|
|
| Author |
|
|
_pawel_
Cruncher Joined: Apr 22, 2010 Post Count: 3 Status: Offline Project Badges: 
|
Hello,
Is it possible to fast switch (in tray menu?) between hardware profiles? for example options could be: Snooze 50% cores, 50% CPU time 100% cores, 80% CPU time 25% cores, 25% CPU time and so on |
|||
|
||||
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Yes, compile some global_prefs_override.xml files with different names, make a script to inject them into the BOINC data dir and use the boinccmd tool to read in the prefs. It's been done before so Google should be able to guide you towards previous discussions.
----------------------------------------Changing cores counts allowed for BOINC will force a benchmark test each time.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
_pawel_
Cruncher Joined: Apr 22, 2010 Post Count: 3 Status: Offline Project Badges:
|
thx :)
it seems that core count cant be even changed with global_prefs_override |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Core count is expressed in clients from 6.2.28 as percent so if a quad and wanting to disable 1 core, enter 75% as in <max_ncpus_pct>75.000000</max_ncpus_pct>
----------------------------------------for clients 5.10.45 the value was stored in different value as in number of CPUs. Then for a quad, disabling one it becomes: <max_cpus>3</max_cpus> edit: you can enter both fields, the older ignores the newer, the newer the older, with exception of 6.2.28. It will use the least of both, always round down to the nearest whole core i.e. even 99% specified for a quad is effectively 75%.
WCG
----------------------------------------Please help to make the Forums an enjoyable experience for All! [Edit 2 times, last edit by Sekerob at Apr 23, 2010 5:58:28 PM] |
||
|
|
_pawel_
Cruncher Joined: Apr 22, 2010 Post Count: 3 Status: Offline Project Badges:
|
oh ok :)
----------------------------------------max_ncpus_pct is a less obvious name ;-) are there any plans for giving an option to disable core change benchmarking? [Edit 1 times, last edit by cz3siu at Apr 23, 2010 8:28:39 PM] |
||
|
|
JmBoullier
Former Community Advisor Normandy - France Joined: Jan 26, 2007 Post Count: 3716 Status: Offline Project Badges: 
|
Benchmarks are not optional. You can force more but you cannot forbid BOINC to do one when it "thinks" it is necessary, like when processing conditions change.
----------------------------------------By the way I hope that you have enough RAM and that you can set option "Leave applications in memory when suspended" ON, otherwise when running benchmarks with older versions of BOINC you will have tasks resuming at their last checkpoint. Even if you use one of the more recent versions which do not unload tasks during benchmarks (at last) when you reduce the number of active cores/threads suspended tasks will lose the work done since their last checkpoint. Another side effect of your changing processing conditions too often will probably be that the client's scheduler will get mad, so be not surprised if it is not asking new work and reporting complete ones when you think it should, or reversely if it asks for more work than you think appropriate. |
||
|
|
Sekerob
Ace Cruncher Joined: Jul 24, 2005 Post Count: 20043 Status: Offline |
Even if you use one of the more recent versions which do not unload tasks during benchmarks (at last) when you reduce the number of active cores/threads suspended tasks will lose the work done since their last checkpoint. That bit I tested with LAIM off and logically, the tasks that were previously running on disabled cores [cores no longer allowed to be used by BOINC] unloaded after the benchmark completed. Here the log of the sequence with the 6.10.45 alpha client, confirming the [at last] feature does work, no longer running the risk of loosing the 2% spaced checkpoint progress on the DDDT2 A-Types, which could be hours:24/04/2010 11:39:55 World Community Grid Computer location: work 24/04/2010 11:39:55 General prefs: using separate prefs for work 24/04/2010 11:39:55 Reading preferences override file 24/04/2010 11:39:55 Preferences: 24/04/2010 11:39:55 max memory usage when active: 1534.63MB 24/04/2010 11:39:55 max memory usage when idle: 1841.56MB 24/04/2010 11:39:55 max disk usage: 10.00GB 24/04/2010 11:39:55 Number of usable CPUs has changed from 2 to 1. Running benchmarks. 24/04/2010 11:39:55 max CPUs used: 1 24/04/2010 11:39:55 (to change, visit the web site of an attached project, 24/04/2010 11:39:55 or click on Preferences) 24/04/2010 11:39:56 Running CPU benchmarks 24/04/2010 11:39:56 Suspending computation - running CPU benchmarks 24/04/2010 11:39:56 World Community Grid [cpu_sched] Preempting HFCC_s2_01832443_s2_0000_0 (left in memory) 24/04/2010 11:39:56 World Community Grid [cpu_sched] Preempting HFCC_s2_01840024_s2_0001_0 (left in memory) 24/04/2010 11:40:28 Benchmark results: 24/04/2010 11:40:28 Number of CPUs: 1 24/04/2010 11:40:28 1623 floating point MIPS (Whetstone) per CPU 24/04/2010 11:40:28 3285 integer MIPS (Dhrystone) per CPU 24/04/2010 11:40:28 [dcf] scaling all duration correction factors by 1.002509 24/04/2010 11:40:29 Resuming computation 24/04/2010 11:40:29 World Community Grid [cpu_sched] Resuming HFCC_s2_01832443_s2_0000_0 24/04/2010 11:40:29 World Community Grid [cpu_sched] Resuming HFCC_s2_01840024_s2_0001_0 24/04/2010 11:40:29 World Community Grid [cpu_sched] Preempting HFCC_s2_01840024_s2_0001_0 (removed from memory) 24/04/2010 11:41:01 World Community Grid General prefs: from World Community Grid (last modified 12-Apr-2010 16:28:59) 24/04/2010 11:41:01 World Community Grid Computer location: work 24/04/2010 11:41:01 General prefs: using separate prefs for work 24/04/2010 11:41:01 Reading preferences override file 24/04/2010 11:41:01 Preferences: 24/04/2010 11:41:01 max memory usage when active: 1534.63MB 24/04/2010 11:41:01 max memory usage when idle: 1841.56MB 24/04/2010 11:41:01 max disk usage: 10.00GB 24/04/2010 11:41:01 Number of usable CPUs has changed from 1 to 2. Running benchmarks. 24/04/2010 11:41:01 (to change, visit the web site of an attached project, 24/04/2010 11:41:01 or click on Preferences) 24/04/2010 11:41:01 Running CPU benchmarks 24/04/2010 11:41:01 Suspending computation - running CPU benchmarks 24/04/2010 11:41:01 World Community Grid [cpu_sched] Preempting HFCC_s2_01832443_s2_0000_0 (left in memory) 24/04/2010 11:41:33 Benchmark results: 24/04/2010 11:41:33 Number of CPUs: 2 24/04/2010 11:41:33 1622 floating point MIPS (Whetstone) per CPU 24/04/2010 11:41:33 3280 integer MIPS (Dhrystone) per CPU 24/04/2010 11:41:33 [dcf] scaling all duration correction factors by 0.999268 24/04/2010 11:41:34 Resuming computation 24/04/2010 11:41:34 World Community Grid [cpu_sched] Resuming HFCC_s2_01832443_s2_0000_0 24/04/2010 11:41:34 World Community Grid [cpu_sched] Starting HFCC_s2_01840024_s2_0001_0(resume) 24/04/2010 11:41:34 World Community Grid Restarting task HFCC_s2_01840024_s2_0001_0 using hfcc version 611 Only the task that was unloaded due core reductions shows a Restart after re-permitted the second core use. The other simply does a normal Resume.
WCG
Please help to make the Forums an enjoyable experience for All! |
||
|
|
|