Stanford and MIT's TACO compiler outpaces MATLAB's Tensor Toolbox by orders of magnitude
Taco: The Tensor Algebra Compiler
TACO is a compiler that generates optimized kernels for sparse and dense linear and tensor algebra, from sparse matrix-vector multiplication to MTTKRP on higher-order tensors. It supports formats like CSR, COO, and CSF, offers C++ and Python APIs, and runs in the browser as a code generator. Its generated kernels beat the MATLAB Tensor Toolbox by up to several orders of magnitude and rival Eigen, Intel oneMKL, and SPLATT.
Under the hood, TACO employs a novel compiler-based technique to generate kernels that are optimized for the computations you want to perform.