Online Source Code Repositories

Sometimes, I like looking up the definition of classes or functions from libraries or tools I use to understand how they are actually implemented. Back in the day, this meant figuring out how to get the source package of that library, downloading it and then searching inside its source files. But now that almost all libraries share their code on Github (or similar services), it has become quite convenient to look up source code repositories. The recent addition of the ability to open a Github repository using VS Code running in the browser is the icing on the cake!

Here are the source code repositories that I most frequently consult:

• C/C++
  • GCC: Mirror of GCC. I use this to consult the C headers and C++ STL that ships with Linux.
  • GLibC: Mirror of the GNU C Library source code.
  • Microsoft STL: Microsoft’s implementation of the C++ STL.
  • Musl: Mirror of musl library repo. Minimal implementation of C standard library.
• Linux
  • Linux kernel: Linus Torvalds’ fork of the Linux kernel source code.
• Python
  • CPython: The implementation of the standard distribution of Python is implemented in C.
• Misc
  • Calibre: Ebook management tool.
  • Vim: Source code of Vim.
• NVIDIA
  • CUDA Samples: Easiest way to check usage of CUDA API is to check how it is used in CUDA samples.
  • CUTLASS: GEMM kernels created using C++ template abstractions.
  • Data Loading Library (DALI): Data loading and pre-processing to accelerate DL applications.
  • Deep Learning Accelerator SW: DLA
  • Open GPU Kernel Modules: NVIDIA’s open source Linux kernel driver modules
  • PyCUDA: Good way to check out actual use of CUDA APIs.
  • TensorRT: OSS components of TensorRT.
  • Torch-TensorRT
  • Triton Inference Server
• Deep Learning
  • PyTorch
  • TensorFlow: tensorflow/compiler/tf2tensorrt subdirectory for TF-TRT.