README.md

Structure of the repository

1. src implementation directory
   1. impl folder stores compressed implementations done during thesis
      1. convolution archive stores version of the TNL git repository checked out at branch with the convolution benchmarks
      2. grid archive stores version of the TNL git repository checked out at branch with grid improvements and heat benchmarks
   2. vis folder stores visualisation of the results
      1. Evaluation.ipynb the python notebook with the visualisations
      2. data data of the measurements.
      3. requirements.txt represents the environment description for the visualisation notebook
   3. thesis folder stores the source code of thesis in LaTeX
2. text text of the work

Visualization

To start visualization notebook you need to install libraries used in the notebook. The suggested way is to use miniconda and create the environment in the following way from the vis directory.

conda create --name <env_name> --file requirements.txt

Then the environment should be activated with the following command:

source activate <env_name>

Finally, to start visualization notebook launch the following command

jupyter lab Evaluation.ipynb

Compilation

The improvements in TNL library are compiled with the TNL build system. Before compilation you may want to ensure, that you have at least NVCC 11.6 and GCC 11.2.0, because for older versions it may not compile. To compile the benchmarks unzip the required archive and execute the following commands in the root folder of repository.

./build --with-cuda=yes|no --with-mpi=yes|no benchmarks

The same can be done for tests.

./build --with-cuda=yes|no --with-mpi=yes|no tests