This repository is dedicated to recording and maintaining a set of commonly used CUDA kernels.
The goal is to provide simple and easy-to-understand implementations for reference and practice.
- Element Wise Add
- Inclusive Prefix Sum
- Softmax
- ReLu
- Sgemm
This list will be continuously updated as more kernels are added.
cd CUDA-KERNELS
pip install -e .
cd kernels/$KERNEL_FOLDER
python setup.py install
python test.py