I tried to do a stress test on the TX1 board with running several CUDA Samples at the same time.
All four Arm cores and GPU is nearly kept at the 90-95% usage.
At first, the CUDA Samples were running normally even though the Ubuntu had a very slow response with the high load.
Unluckily most of the running CUDA Sample process are killed by the OS after approximately 36 hours of normal ruuning.
Why the Linux OS kill the CUDA Samples ? And why after 36 hours of normal ruuning?
(the /var/log/kern.log indicates that linux OS kill the CUDA Samples because of the lack of memory,and the syslog file in the attachments.)
kern.log (966 KB)