Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
53 lines (41 loc) · 1.68 KB

CONTRIBUTING.md

File metadata and controls

53 lines (41 loc) · 1.68 KB
Raw

Contributing to TensorLib

Thanks for improving TensorLib. Keep changes focused, portable, and easy to review.

Build and test

The default workflow builds the library, examples, and test suite:

cmake -S . -B build -DCMAKE_BUILD_TYPE=Debug
cmake --build build --parallel
ctest --test-dir build --output-on-failure
python3 scripts/check_markdown_links.py

Before opening a pull request, also run a strict build:

cmake -S . -B build-strict \
  -DTENSORLIB_WARNINGS_AS_ERRORS=ON \
  -DTENSORLIB_ENABLE_OPENMP=OFF
cmake --build build-strict --parallel
ctest --test-dir build-strict --output-on-failure

For memory and undefined-behavior checks, enable the sanitizer option with a GCC or Clang debug build:

cmake -S . -B build-sanitize -DCMAKE_BUILD_TYPE=Debug \
  -DTENSORLIB_ENABLE_SANITIZERS=ON \
  -DTENSORLIB_ENABLE_OPENMP=OFF
cmake --build build-sanitize --parallel
ctest --test-dir build-sanitize --output-on-failure

Change guidelines

  • Follow the existing C99 style and keep compiler warnings at zero.
  • Add tests for public behavior, failure paths, and non-contiguous views.
  • Keep ownership rules explicit whenever a function accepts or returns memory.
  • Avoid machine-specific instructions unless a portable fallback and runtime dispatch are present.
  • Keep commits small and give each commit one clear purpose.
  • Update the relevant guide when public behavior or build controls change.

Performance changes

Correctness tests must pass before benchmarking. Record the compiler, flags, CPU, thread count, matrix shapes, and repeated-run variance with any performance claim. See the benchmark guide for the standard commands and reporting format.