| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Active Research
Forum: Mapping Cancer Markers Forum
Thread: ~35% sys calls, ok? |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 33
|
|
| Author |
|
|
!evil
Cruncher Joined: Aug 3, 2023 Post Count: 4 Status: Offline |
Hi,
I noticed that MCM spends a lot of time in kernel space and `strace` shows enourmous amounts of ... times({tms_utime=133602 /* 1336.02 s */, tms_stime=71663 /* 716.63 s */, tms_cutime=0, tms_cstime=0}) = 1431858097 times({tms_utime=133602 /* 1336.02 s */, tms_stime=71663 /* 716.63 s */, tms_cutime=0, tms_cstime=0}) = 1431858097 times({tms_utime=133602 /* 1336.02 s */, tms_stime=71663 /* 716.63 s */, tms_cutime=0, tms_cstime=0}) = 1431858097 times({tms_utime=133602 /* 1336.02 s */, tms_stime=71663 /* 716.63 s */, tms_cutime=0, tms_cstime=0}) = 1431858097 times({tms_utime=133602 /* 1336.02 s */, tms_stime=71663 /* 716.63 s */, tms_cutime=0, tms_cstime=0}) = 1431858097 times({tms_utime=133602 /* 1336.02 s */, tms_stime=71663 /* 716.63 s */, tms_cutime=0, tms_cstime=0}) = 1431858097 times({tms_utime=133602 /* 1336.02 s */, tms_stime=71663 /* 716.63 s */, tms_cutime=0, tms_cstime=0}) = 1431858097 times({tms_utime=133602 /* 1336.02 s */, tms_stime=71663 /* 716.63 s */, tms_cutime=0, tms_cstime=0}) = 1431858097 times({tms_utime=133602 /* 1336.02 s */, tms_stime=71663 /* 716.63 s */, tms_cutime=0, tms_cstime=0}) = 1431858097 times({tms_utime=133602 /* 1336.02 s */, tms_stime=71663 /* 716.63 s */, tms_cutime=0, tms_cstime=0}) = 1431858097 ... Is this intended? |
||
|
|
hchc
Veteran Cruncher USA Joined: Aug 15, 2006 Post Count: 806 Status: Offline Project Badges: 
|
This question is interesting but also highly technical. You'd probably have better luck finding out who developed the science application that MCM1 uses and contact the developer(s) directly (e-mail, even phone or snail mail).
----------------------------------------
|
||
|
|
thunder7
Senior Cruncher Netherlands Joined: Mar 6, 2013 Post Count: 232 Status: Offline Project Badges: 
|
Are you able to code in c? There is the ld_preload mechanism, where you can override libc for a single function call. I wonder what would happen if you created a local times() function call that used a 1-1000 loop, calling the real times() the first time, and returned -1 the other 999 times. Might that speed up things, of will boinc error out on receiving a -1?
|
||
|
|
!evil
Cruncher Joined: Aug 3, 2023 Post Count: 4 Status: Offline |
I just tried but don't know how to inject the env variable. Setting it for boinc makes no difference. If anyone wants to spend more time on this, here is the code (compiled with gcc -fPIC -shared -o /path/to/file.so /path/to/file.c)
#include <sys/times.h> #include <dlfcn.h> typedef clock_t (*real_times_t)(struct tms*); struct tms previous_tms; clock_t previous_clock; int i = 1000; clock_t times(struct tms *buf) { if (i == 1000) { previous_clock = ((real_times_t)dlsym(RTLD_NEXT, "times"))(&previous_tms); i = 1; } else { i = i + 1; } *buf = previous_tms; return previous_clock; } I don't expect a solution, its just a technical note which might help the MCM Team. And they should keep half an eye on the forum for things like this imo. |
||
|
|
AgrFan
Senior Cruncher USA Joined: Apr 17, 2008 Post Count: 376 Status: Offline Project Badges: 
|
See this thread for more details re: performance degradation between Windows and Linux ...
----------------------------------------https://www.worldcommunitygrid.org/forums/wcg...ead,44664_offset,0#681380 Right now, it's best to run MCM on Windows and SCC on Linux using separate device profiles. SCC uses VINA which runs quicker on Linux than Windows. [Edit 6 times, last edit by AgrFan at Aug 15, 2023 12:21:32 AM] |
||
|
|
!evil
Cruncher Joined: Aug 3, 2023 Post Count: 4 Status: Offline |
Ah, I see. Thanks for pointing there! Then I discourage trying the LD_PRELOAD stuff xD
|
||
|
|
thunder7
Senior Cruncher Netherlands Joined: Mar 6, 2013 Post Count: 232 Status: Offline Project Badges:
|
I don't have any of the longer MCM tasks at the moment (which took >4 hours on an aging Xeon V4 with 2.2 GHz).
----------------------------------------But I suspect the place to add that LD_PRELOAD variable is the wcg* files in /var/lib/boinc/projects/www.worldcommunitygrid.org. Those are actully running the work, so exchanging one of them for a bash-wrapper that sets LD_PRELOAD and then calls the original executable and passes all commandline arguments might work. But I agree getting a new/better version of this seemingly inefficient executable would be better. It doesn't seem to be open source, to more than offer to test it I can't do, I'm afraid. [Edit 2 times, last edit by thunder7 at Aug 15, 2023 11:42:20 AM] |
||
|
|
Speedy51
Veteran Cruncher New Zealand Joined: Nov 4, 2005 Post Count: 1297 Status: Offline Project Badges:
|
See this thread for more details re: performance degradation between Windows and Linux ... https://www.worldcommunitygrid.org/forums/wcg...ead,44664_offset,0#681380 Right now, it's best to run MCM on Windows and SCC on Linux using separate device profiles. SCC uses VINA which runs quicker on Linux than Windows. I agree with you if you are wanting to return results as fast as possible. My thoughts are with the limited amount of work there seems to be at the moment the quicker we return it the quicker we may run out of work in regards to speeding MCM up I am sure if it can be done & the scientists want done it would have been done a considerable time ago. I will only run applications issued to me by World Community Grid Have a great day  |
||
|
|
Sgt.Joe
Ace Cruncher USA Joined: Jul 4, 2006 Post Count: 7687 Status: Offline Project Badges: 
|
I agree with you if you are wanting to return results as fast as possible. My thoughts are with the limited amount of work there seems to be at the moment the quicker we return it the quicker we may run out of work It is not a question of returning results as soon as possible, but of trying to run the work units where they are the most efficient. I would like to deploy my electricity where it does the most work per watt. Cheers
Sgt. Joe
*Minnesota Crunchers* |
||
|
|
AgrFan
Senior Cruncher USA Joined: Apr 17, 2008 Post Count: 376 Status: Offline Project Badges:
|
I agree with you if you are wanting to return results as fast as possible. My thoughts are with the limited amount of work there seems to be at the moment the quicker we return it the quicker we may run out of work It is not a question of returning results as soon as possible, but of trying to run the work units where they are the most efficient. I would like to deploy my electricity where it does the most work per watt. Cheers +1 |
||
|
|
|