Learnings and experimentation with GPU programming
- cuda
- Triton
The easiest way to set the environment is
- Have a Linux machine with a GPU
- Ensure
nvidia-smi
is installed- Note the CUDA version
- Install
conda
- Install the relevant CUDA toolkit version (from 2.1) using
conda
. Reference - Use
nvcc --version
to verify the installation - Run
nvcc -o <output file name> <file to compile>