Support for all recently-announced Pascal GPUs
Simplify Unified Memory programming, including support for large datasets and concurrent data on Pascal GPUs
Automated critical path analysis for easy performance optimization
I have forwarded your comment to our technical staff.
I would suggest you post your comment as a thread in the CUDA General Programming section, it will get more attention from folks who may have a solution for you.
For anyone newly installing the toolkit please be aware that there are some incompatible defaults with the RSE (Remote Services Explorer) that is used for remote dev/test and the latest versions of openssh. This was on Ubuntu 16.04 but probably is present on other platforms.
The support in openssh (by default) is both RSA and DSA (deprecated) disabled. Any attempt to do a remote services connection results in an unable to connect to sshd on ip.address.of.system. message.
So folk will need to reconfigure the /etc/ssh/sshd_config file to re-enable RSA (as DSA is deprecated anyway). In earlier versions the defaults used to work.
Please can Nvidia update the RSE level in later versions of the toolkit to support ECDSA.
How do I tell which version I have or even if it is installed? Just installed the Jetson JetPack last night, and I can’t tell what or where stuff is on the host OR the TX2.