Join PrimeGrid
Returning Participants
Community
Leader Boards
Results
Other
drummers-lowrise
|
Message boards :
Number crunching :
Primegrid running even when it isn\'t?
| Author |
Message |
|---|
|
|
|
I have the leave in memory when suspended option on. PrimeGrid is the only project to do this. Even when it says waiting to run, it uses CPU time and the time in the boinc manager increases. I only have two CPU cores. What's up?
Screenshot
____________
| |
|
|
|
I have the leave in memory when suspended option on. PrimeGrid is the only project to do this. Even when it says waiting to run, it uses CPU time and the time in the boinc manager increases. I only have two CPU cores. What's up?
Screenshot
I was going to bring this up, I get a lot of funny behaviour like this from PrimeGrid. | |
|
RytisVolunteer moderator
Project administrator
Send message
Joined: 22 Jun 05
Posts: 2653
ID: 1
Credit: 110,106,285
RAC: 48,083
                     
|
|
It is because LLR application behaves badly when running under BOINC. As you may know, it is actually a standalone application, which is run under BOINC using BOINC wrapper application, which handles suspend/resume/abort messages from BOINC.
Now comes the tech stuff :) LLR application is actually multithreaded. When the application is launched, we attempt to find out the thread ID of the computation thread, and use it later for message handling. However, this doesn't always work (for unknown reasons). In such case wrapper cannot suspend correct application, and "runaway" situation occurs.
[edit] While sieve applications also use wrapper, they don't have the issue.
____________
| |
|
|
|
In such case wrapper cannot suspend correct application, and "runaway" situation occurs.
The wrapper has been modified. See [trac]changeset:14862[/trac] and [trac]changeset:14863[/trac].
____________
Jord.
Would Ancient Alien Theorists surmise that primes were introduced by aliens for communication purposes? | |
|
RytisVolunteer moderator
Project administrator
Send message
Joined: 22 Jun 05
Posts: 2653
ID: 1
Credit: 110,106,285
RAC: 48,083
|
In such case wrapper cannot suspend correct application, and "runaway" situation occurs.
The wrapper has been modified. See [trac]changeset:14862[/trac] and [trac]changeset:14863[/trac].
Since PG is using a completely different wrapper, the change will have to be ported over. Should not take long though.
____________
| |
|
|
|
|
Just curious if the memory and complete issue has been fixed? | |
|
|
|
Just curious if the memory and complete issue has been fixed?
This thread is from March, here it is August, has this been fixed yet? I think I noticed the same problem.
Running BOINC 5.10.45, and just started running PrimeGrid (and a bunch of other apps, had been running some of the older apps for years). Here are my startup messages (ie. running only a single processor, etc.)
8/10/2008 10:56:19 PM||Starting BOINC client version 5.10.45 for windows_intelx86
8/10/2008 10:56:19 PM||log flags: task, file_xfer, sched_ops
8/10/2008 10:56:19 PM||Libraries: libcurl/7.18.0 OpenSSL/0.9.8e zlib/1.2.3
8/10/2008 10:56:19 PM||Data directory: X:\Program Files\Boinc
8/10/2008 10:56:20 PM||Processor: 1 GenuineIntel Intel(R) Pentium(R) M processor 1.86GHz [x86 Family 6 Model 13 Stepping 8]
8/10/2008 10:56:20 PM||Processor features: fpu tsc pae nx sse sse2 mmx
8/10/2008 10:56:20 PM||OS: Microsoft Windows XP: Professional Edition, Service Pack 2, (05.01.2600.00)
What I do is during the day, I run shorter (timewise) applications & wu's (that complete in an hour or two), and I save the wu's that need 100's of hours to run nights/weekends when I'm away from work.
On this WU:
8/11/2008 6:36:56 PM|PrimeGrid|Restarting task llrCUL_17412355_1 using llrCUL version 508
I placed it into "Task Suspended by User" this morning, and mid-afternoon I happened to notice that the CPU Time & Time to Completion numbers were changing. I opened the Win XP Task Manager, and noticed that PG was taking about 50% of my CPU, and sharing the other 50% with either the SHA-1 Collision Search or Spinhenge Metropolis applications. When both of those completed, and uFluids became the active application, I noticed for over 30 minutes that PG was taking 99% of the processor, and uFluid wasn't getting any. Then at some point, uFluids started getting everything.
Example, while project was in "Task Suspended by User"
Wall Clock CPU time Completion Time to Complete
2:28 PM 43:22:36 49.080% 87:17:47 Task Suspended by User
3:45 PM 44:07:37 49.930% 85:06:48 Task Suspended by User
5:00 PM 44:23:17 50.180% 84:30:12 Task Suspended by User
6:30 PM 44:23:17 50.180% 84:30:12 Task Suspended by User | |
|
|
|
|
I definitely have this problem and I'm new around here, which I expect means that the wrapper application is the latest.
Is there a manual work-around?
____________
Warped
  | |
|
|
|
|
As a post-script, I went to task manager and killed the LLR applications there. All seems to be OK. I hope the applications do not start from scratch again when it's their turn.
____________
Warped
| |
|
Message boards :
Number crunching :
Primegrid running even when it isn\'t? |
