I have been running MT on different machines to check the validation. The delay in the eventual validation seems to be coming from some of the longer running work units taking time until there are enough units to validate against in the system. For the gpu separation code the units are processed at the same relative rate that the system flow fairly quickly. I am still monitoring it and trying to get a more coherent picture of the issue. If you could post the work units that are failing I could look at them and try to run them on my machines to see if I can see why it was failing.

I haven't had an error on the MTs on my 10.7 Os X or 10.8 Os X I haven't checked our 10.6 machine for errors lately so any details on system specifics (Os version / Boinc Version) may help isolate an error. Though the binaries for the Unix based operating systems are much more self contained than the windows systems.


jeff

If you could post the work units that are failing I could look at them and try to run them on my machines to see if I can see why it was failing.

Thanks for the quick response!

OK. Here are the N-Body tasks which are still on my Tasks page as I write:

http://milkyway.cs.rpi.edu/milkyway/results.php?userid=156736&offset=0&show_names=0&state=0&appid=7

1 validated, 1 inconclusive, 3 errors. All MT.

I run CPU apps only, the machine has a 2-core processor, and my BOINC setting is to use 100% of them.

NG

The errors seem to be all related to when it is checkpointing.

I will try the parameters on a test machine and see what we get.


No details outside of this right now.

Jeff

If you run Dependency Walker against the main N-Body executable, it will tell you which DLLs are linked and whether they are present on your computer. It will probably flag up libgomp_64-1.dll and pthreadGC2_64.dll (because the distribution names are different), and you usually get warnings about late-loading dependencies too - they can be ignored. But it helps to narrow the list of suspects down.

Finally had time to look at this again. libgomp_64-1.dll and pthreadGC2_64.dll did indeed show as unresolved. So I copied libgomp_64-1_nbody_1.36.dll to just libgomp_64-1.dll, and similarly for pthread. I reactivated nbody on the machine, pulled some down. Runs like a champ on both the MT and non-MT versions.

Resolves the issue on this machine. Hope it provides some help to others

Bob