study-triton

This repository is for studying OpenAI's Triton.

Contents

• 1_hello_triton: a kernel that prints "Hello Triton Kernel!". explains func[grid]() syntax.
• 2_inout_tensor: how to input/output tensors to/from a kernel
• 3_measuring_performance: measure the performances of copy methods in PyTorch in terms of elapsed time using triton.testing library.
• 4_2d_grid_and_matmul: a kernel that computes matrix multiplication.
• 5_out_of_index: how to handle out of index problem with an example of matrix multiplication.
• 6_random: using random seed in triton.jit'd function with an example of random tensor initialization.
• 7_pybind11: how to use C++ code in Python

To be writen