I am super excited that our MLIR sparse compiler journey that started in Nov 2020 is where it is today, with a sound sparse ecosystem that includes sparse tensor types as first class citizens and a reference implementation for sparsification that runs a sparse language like PyTACO end-to-end or can be used to exhaustively explore all possible sparse storage schemes and optimizations to find the best implementation for a particular sparse kernel. Our hope is that the open-source nature of the project benefits researchers that are new to this domain and also solicits contributions from experts in the field.
You can read all about this in our paper:
Compiler Support for Sparse Tensor Computations in MLIR
Aart J.C. Bik, Penporn Koanantakool, Tatiana Shpeisman, Nicolas Vasilache, Bixia Zheng, and Fredrik Kjolstad
available as pre-print on arXiv, [2202.04305] Compiler Support for Sparse Tensor Computations in MLIR, February 9, 2022.