The driver should be loaded at boot time. You can automate this on boot by editing:

/etc/modules:

 ...
#
nvidia-uvm
 ...

Athena (currently) accesses the ethernet port using a raw socket.
A raw socket can only be opened as root, a program suid root, or by a program sudo setcap cap_ipc_lock,cap_net_raw=+pe
A setcap/suid root program won't run with links to shared libs that ldconfig doesn't know about. So you MUST create a file named:

/etc/ld.so.conf.d/libcuda.conf:

# path to cuda libs, etc.
/usr/local/cuda/lib64

Then run ldconfig as root:

$ sudo ldconfig

Athena now is in the development stage, and uses modified anan or hermes boards with experimental FPGA firmware while Minerva design is completed. Minerva will be a pciE based board, and should be able to run both ADCs at full sample rate. But for now, we use the modifed hardware thru their ethernet ports.
Considerations:

• Even at a rate of 1GBit per second, we can only move samples from 1/2 of 1 ADC (922MBit per second consumed).
• For transmit this means moving data in both directions at this rate.
• Thus it is desirable (mandatory for transmit) to use a 2nd dedicated NIC for the connection between the anan/hermes and the Athena box.

For the following ethernet setup examples assume the following. You system will probably use different names for the specific NIC cards:

• Primary NIC used for general traffic and Athena <> Thetis communication: eno1
• Second NIC used for dedicated hardware <> Athena communication: enp18s0

The NICs must be configured properly for boot.

• The routes must be configured.
• The radio interface MTU must be set to 9000.

/etc/rc.local:

 ...
# use 1st NIC for general traffic:
/sbin/route add default gw Router eno1

# use jumbo packets between hardware and Athena on 2nd NIC:
/sbin/ifconfig enp18s0 mtu 9000
 ...

/etc/network/interfaces:

 ...
auto eno1
    iface eno1 inet static
        mtu 9000
        address 192.168.0.2
        netmask 255.255.255.0
        gateway 192.168.0.1

auto enp18s0
    iface enp18s0 inet static
        mtu 9000
        address 192.168.1.2
        netmask 255.255.255.0
 ...

You MUST increase the size of raw socket buffers used by the network stack, 32MB seems to work, I use 64MB for grins:

$ sudo sysctl -w net.core.rmem_max=67108864
$ sudo sysctl -w net.core.wmem_max=67108864
$ sudo sysctl -w net.core.rmem_default=67108864
$ sudo sysctl -w net.core.wmem_default=67108864

You can automate this on boot by editing:

/etc/sysctl.conf:

#######################################################
# Large socket buffer for dfcSDR
#
net.core.rmem_max=67108864
net.core.wmem_max=67108864
net.core.rmem_default=67108864
net.core.wmem_default=67108864

Advanced Configuration Examples