Work Unit 20240 is interesting in the credit result. My iMac and one of the MacPro's got credit. The second Mac Pro got zero. I can only see that the MacPro that got zero credit seems to have started and stopped the Charm application at least once more?
Good catch; we are also thinking that checkpointing might have to do something with the fact that there are (seemingly random) differences in results on macs. Richard and Michela are looking into this at the moment.
Thanks
Andre
Work Unit 20240 is interesting in the credit result. My iMac and one of the MacPro's got credit. The second Mac Pro got zero. I can only see that the MacPro that got zero credit seems to have started and stopped the Charm application at least once more?
____________
D@H the greatest project in the world... a while from now!
Good catch; we are also thinking that checkpointing might have to do something with the fact that there are (seemingly random) differences in results on macs. Richard and Michela are looking into this at the moment.
I had a simliar problem. Downloaded three WUs. One WU was computed without switching to other projects in one turn. Other two computed with some interrupts by project switching. One WU got valid.
I am not shure, but it looks like the WU without interrupts made the validation.
Andre, what do you think, should I/we retry to crunch some WUs without interrupting in one turn ?
If you have some time for this, that would be a great experiment. In the meantime, we will be doing the same in the lab (on a smaller scale since we don't have so many macs).
Thanks
Andre
Good catch; we are also thinking that checkpointing might have to do something with the fact that there are (seemingly random) differences in results on macs. Richard and Michela are looking into this at the moment.
I had a simliar problem. Downloaded three WUs. One WU was computed without switching to other projects in one turn. Other two computed with some interrupts by project switching. One WU got valid.
I am not shure, but it looks like the WU without interrupts made the validation.
Andre, what do you think, should I/we retry to crunch some WUs without interrupting in one turn ?
____________
D@H the greatest project in the world... a while from now!
Good catch; we are also thinking that checkpointing might have to do something with the fact that there are (seemingly random) differences in results on macs. Richard and Michela are looking into this at the moment.
Do you have changed anything at the Mac application ? - I have crunched a WU which was validated successfull. I have looked in the logs, the application had switched, but I had enabled the "leave application in memory while resumed" switch. I am not shure, if this was a good luck, or if the leave application in memory switch could be the solution.
Do you have changed anything at the Mac application ? - I have crunched a WU which was validated successfull. I have looked in the logs, the application had switched, but I had enabled the "leave application in memory while resumed" switch. I am not shure, if this was a good luck, or if the leave application in memory switch could be the solution.
It's possible that this setting helps alleviate the problem, but I don't know whether it would be a good solution since we would have to ask everybody to switch it on specially for D@H. Remember that this is a randomly occuring problem: it only happens when the boinc client switches projects in the middle of the checkpoint routine.
Andre
____________
D@H the greatest project in the world... a while from now!
It's possible that this setting helps alleviate the problem, but I don't know whether it would be a good solution since we would have to ask everybody to switch it on specially for D@H.
There are some projects, where the "leave application in memory" switch should be enabled by default, CPDN for example..
Remember that this is a randomly occuring problem: it only happens when the boinc client switches projects in the middle of the checkpoint routine.
As multi project cruncher, I have a heavy switching beteen the apps, so I had about 75 % invalid docking results.
I don't have a Mac, but on my Linux and XP machines I have it set to Leave applications in memory and I don't remember when I've had a WU that didn't validate. Well, there was that one where I did some hardware changes and rebooted/powered down the machine about 5 times during the first 20% of the WU, but I figured the chances of it validating were nonexistent so I aborted the WU :-)
-- David
____________
The views expressed are my own.
Facts are subject to memory error :-)
Have you read a good science fiction novel lately?