gaspi_proc_term sometimes hangs after otherwise successful program execution

[krzikalla]

Configuration: hybrid MPI/GASPI application, TCP device, four processes on one node (hence local communication only)

Sometimes during gaspi_proc_term some processes hang in pthread_join (tcp_dev_stop_device).
The tcp_virt_dev thread (for which the join waits) hangs in a read (can't say which one).

[krzikalla]

Further investigation: it happens at tcp_device.c:1351. The read blocks and waits until data is available. However, for some reason (unknown to me) there seems to be no more data. Some processes are already done with gaspi_proc_term (apparently always proc 0, sometimes also some other), but some others are hanging.

[rumach]

Hi there. Do you have a simple program which reproduces this behaviour, more or less, consistently?