CUDA and TensorRT ramp up and snippets

Resources

Programming Massively Parallel Processors

CUDA by Example
Source code

NVIDIA

CUDA C++ Programming Guide
CUDA samples
CUDA Toolkit
Tensor RT

Local or cloud machine

GTX 1650 - entry level
Alternatives
Koyeb
Vast.ai

Local debugging

Set CMAKE_CUDA_ARCHITECTURES to the machine graphic card. For GeForce RTX 5070 the architecture is 70.

On CLion add cuda-gdb as a custom debugger in Settings | Build, Execution... | Toolchain and set Debugger to /usr/bin/cuda-gdb

This works in cuda-gdb but not in CLion

nvcc -g -G -o cmake-build-debug-cuda-debug/hello_main hello_main.cu
cuda-gdb ./cmake-build-debug-cuda-debug/hello_main