profile-kernel

A CLI tool to profile CUDA kernels using NVIDIA Nsight Compute, extract source-level performance information, and automatically highlight performance bottlenecks (e.g. excessive memory access or warp stalls).

Features

• Parses Nsight Compute CSV output
• Annotates source lines with Access and Stall warnings
• Filters to only show lines with specific warnings
• Exports as a clean, annotated CSV

Installation

pip install profile-kernel

Usage

profile-kernel \
  --ncu_path /path/to/ncu \
  --exe_path /path/to/python \
  --filepath my_kernel.py \
  --output_filename output \
  [--filter access|stall|warning]  # optional

Example Output

The output CSV will contain columns like:

#, Warning Type, Warning Info, Address, SASS, ...

License

MIT License © 2025 Arjun Menon

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.