Tried reinstalling BOINC Manager version but still flaky. Uninstalled that and installed the simple GUI which is really just a pretty face for the CLI. Hopefully that will be more stable while I'm gone.
I don't know if it's a BOINC Manager problem, or an Einstein problem, so I'll report it here first. I watched BOINC Manager tell Einstein to pause and SETI to resume running, but Einstein didn't give up the cpu - both ended up fighting for control and each was getting around 45% of the cpu cycles. I quit BOINC Manager and restarted it - same situation with both trying to run. I took a screen shot that shows Einstein paused in the background, but Activity Monitor is showing it still running.
I next tried telling Einstein to suspend; had no effect. I took another screen shot where BOINC Manager shows SETI running, Einstein suspended, and yet Activity Monitor is showing the two clients running at the same time with each having about 45% of the cpu. If it will help, I can email the screen shots to someone...
I actually saw the partially completed unit resume, with zero CPU time.
Does the setting of "remove from memory when suspended" make a difference to that?
BM
Doesn't seem to. I've tried it both ways. I think there is a problem with the BOINC Manager.
I suspect it is a problem in the BOINC App API library not in the BOINC Manager. Could you please see if the client_state.xml file contains the correct CPU time?
I just returned my first two work units and the first one is "invalid", the other one is still "initial". The link to the result is http://einsteinathome.org/task/5684527. Does this mean that there is a problem with the einstein altivec code? The WU seems to be processed and sent correctly. Is this a known problem? I am using an iBook G4 / 933 and MacOS X 10.3.9 with Boinc 4.43 (adv).
Thanks for help!
I tried to reproduce your problem. Running einstein only I couldn't see anything unusual, so I attached to SETI as well. I found that neither einstein nor SETI took any suspension, both were (more or less happily) sharing the CPU time, regardless of what the client tells them to do, either because of the normal scheduling or because I manually told him to suspend a project or result. Very strange. There's definitely some bug left in the boinc (App) lib. Asymmetries (like you see with Predictor and Einstein) are probably caused by the different library version the Apps are using.
Working fine for me,
average of CPU per WU is 13k sec instead of 22k sec (PM G5 dual 2,5)
I'm using BOINC Menubar (simple GUI) v4.43
I had noticed that swap with other BOINC applications worked better with settings "remove from memory when suspended"
[quote}
I suspect it is a problem in the BOINC App API library not in the BOINC Manager. Could you please see if the client_state.xml file contains the correct CPU time?
Looks like something is stuck. Hasn't returned an Einstein work unit since the 23rd, but continues to return predictor work units. Appears to be stuck on predictor and not switching. I'll check the logs when I get back, but that won't be for another 13 days. Machine is running, BOINC is running, but Einsten isn't.
The new client worked well until BOINC contacted the server to get more work. After that, the client stopped computing. I can still see the process using top; it is using 0% CPU. I have to exit BOINC and restart to resume computation. It could be due to BOINC too. I am using BOINC version 4.44 from the Team MacNN webpage. I did not have this problem while running the older Einstein client though.
Tried reinstalling BOINC
)
Tried reinstalling BOINC Manager version but still flaky. Uninstalled that and installed the simple GUI which is really just a pretty face for the CLI. Hopefully that will be more stable while I'm gone.
-David
Team MacNN - The best Macintosh team ever.
I don't know if it's a BOINC
)
I don't know if it's a BOINC Manager problem, or an Einstein problem, so I'll report it here first. I watched BOINC Manager tell Einstein to pause and SETI to resume running, but Einstein didn't give up the cpu - both ended up fighting for control and each was getting around 45% of the cpu cycles. I quit BOINC Manager and restarted it - same situation with both trying to run. I took a screen shot that shows Einstein paused in the background, but Activity Monitor is showing it still running.
I next tried telling Einstein to suspend; had no effect. I took another screen shot where BOINC Manager shows SETI running, Einstein suspended, and yet Activity Monitor is showing the two clients running at the same time with each having about 45% of the cpu. If it will help, I can email the screen shots to someone...
C
Join Team MacNN
RE: RE: RE: I actually
)
I suspect it is a problem in the BOINC App API library not in the BOINC Manager. Could you please see if the client_state.xml file contains the correct CPU time?
Director, Einstein@Home
I just returned my first two
)
I just returned my first two work units and the first one is "invalid", the other one is still "initial". The link to the result is http://einsteinathome.org/task/5684527. Does this mean that there is a problem with the einstein altivec code? The WU seems to be processed and sent correctly. Is this a known problem? I am using an iBook G4 / 933 and MacOS X 10.3.9 with Boinc 4.43 (adv).
Thanks for help!
Shaktai, I tried to
)
Shaktai,
I tried to reproduce your problem. Running einstein only I couldn't see anything unusual, so I attached to SETI as well. I found that neither einstein nor SETI took any suspension, both were (more or less happily) sharing the CPU time, regardless of what the client tells them to do, either because of the normal scheduling or because I manually told him to suspend a project or result. Very strange. There's definitely some bug left in the boinc (App) lib. Asymmetries (like you see with Predictor and Einstein) are probably caused by the different library version the Apps are using.
BM
BM
Working fine for me so
)
Working fine for me so far:
http://einsteinathome.org/host/205881/tasks
CPU time per WU dropped by almost 50%, from ~79k sec to ~43k sec.
Machine info:
http://einsteinathome.org/host/205881
It's even pausing/resuming properly when BOINC swaps over to the SETI@Home Altivec-optimized application (setiathome 4.70).
Are you a musician? Join the Musicians team.
Meet the Musicians
Working fine for me, average
)
Working fine for me,
average of CPU per WU is 13k sec instead of 22k sec (PM G5 dual 2,5)
I'm using BOINC Menubar (simple GUI) v4.43
I had noticed that swap with other BOINC applications worked better with settings "remove from memory when suspended"
[quote} I suspect it is a
)
[quote}
I suspect it is a problem in the BOINC App API library not in the BOINC Manager. Could you please see if the client_state.xml file contains the correct CPU time?
Team MacNN - The best Macintosh team ever.
Looks like something is
)
Looks like something is stuck. Hasn't returned an Einstein work unit since the 23rd, but continues to return predictor work units. Appears to be stuck on predictor and not switching. I'll check the logs when I get back, but that won't be for another 13 days. Machine is running, BOINC is running, but Einsten isn't.
Guess Bruce's API argument makes sense.
http://einsteinathome.org/host/265788/tasks
Team MacNN - The best Macintosh team ever.
The new client worked well
)
The new client worked well until BOINC contacted the server to get more work. After that, the client stopped computing. I can still see the process using top; it is using 0% CPU. I have to exit BOINC and restart to resume computation. It could be due to BOINC too. I am using BOINC version 4.44 from the Team MacNN webpage. I did not have this problem while running the older Einstein client though.