Firefly and PC GAMESS-related discussion club
Learn how to ask questions correctly
Firefly Bug
alex
yakovenko.alexander@gmail.com
I am reporting about firefly instabilities of parallel run with 4 threads on Intel CPUs. With MP2 single-point calculations I have observed instability that happens on average once per 10-15 successful runs. The bug is non-reproducing and probably connected to semaphores allocation/initialization. If bug then all copies of Firefly works with 100% CPU but there is no more output after ‘YOU MAY CONSIDER DROPPING ONE OR MORE AOS, ...’. The problem might occur at n-th sequential run on the same node and at the first run as well. I have never met the bug on my 2-CPU workstation but have such experience with sequential runs on some SMP machines of our cluster.
P.S. killing of dead-locked firefly is a cure for pack of runs jobs supplied within a script.
Regards,
Alex.
![[ Previous ]](/ceilidh-icons/ceil_prv.gif)
![[ Next ]](/ceilidh-icons/ceil_nxt.gif)
![[ Index ]](/ceilidh-icons/ceil_ind.gif)
Sat Jan 30 '10 0:59am
![[ Reply ]](/ceilidh-icons/ceil_rep.gif)
![[ Edit ]](/ceilidh-icons/ceil_edt.gif)
![[ Delete ]](/ceilidh-icons/ceil_del.gif)
This message read 1495 times