In textbooks, linear algebra expressions are often presented in a form that uses indices. Translating these expressions into efficient code quickly runs into the problem that optimized linear algebra libraries require the expressions in a form without indices. In order to still produce efficient code, we have developed LinAExternal link, which takes expressions with and without indices as input, and outputs optimized NumPy and Eigen code.
LinA is based on the publication:
- J. Klaus, M. Blacher, J. Giesen. Compiling Tensor Expressions into Einsum. Proceedings of the 23d International Conference on Computational Science (ICCS), 2023
- J. Klaus, M. Blacher, J. Giesen, P. Rump and K. Wiedom. Compiling Linear Algebra Expressions into Efficient Code. Proceedings of the 22nd International Conference on Computational Science (ICCS), (2022) 11-17