CUDA

Autonomous Machines Jetson & Embedded Systems Jetson TX1
1

I install CUDA9.0 in my host. I compile a program through nvidia Nsight. I copy the program to my TX1.
I meet a problem that cudaMalloc((void **)&gpuData, sizeof(float)*size) returned CUDA driver version is insufficient for CUDA runtime version(35) at …/src/test.cu:35. How can I resolve it? I can’t find the directory that /proc/driver/nvidia/version in my TX1.

2

Hi,

Jetson GPU driver is integrated into the OS(L4T).

There are two requirements for the cross environment of Jetson:
1. Same CUDA version as the device
2. CUDA toolkit with the aarch64 cross compiler support

Currently, CUDA with cross-compiling support is only available in the JetPack.
Please use the same JetPack installer to setup your host CUDA environment.

Thanks.