Repository containing the code for my Bachelor Thesis 'Redesign of VIGRA Based on Xtensor'

This project is build on CMake with the vcpkg toolchain.

Project structure:

• benchmarks: C++ code of the performed benchmarks
• benchmarks-result: Contains JSON files with the recorded benchmark data and rendered tables / graphs used in the thesis
• include
  • raw: utility header to access flattened arrays as strided n-dimensional view for benchmarking
  • xvigra: production code of this work
  • xvigra_legacy: contains different variants of the convolution implementation which are used for the benchmarks
• py_impl: The Python reference implementation of the explicit convolution
• resources: Image files needed for demos and tests
• scripts: Python scripts to generate test data, plot benchmark data and increase convience of the build process
• src: Some small demos to better understand the xtensor / xvigra API
• tests: Unit tests used to ensure the correct behaviour of the Explicit and Separable Convolution

different build files for Windows and Linux with a high-level dual build based on the WSL.