https://hgpu.org/?p=16447
MAGMA Batched: A Batched BLAS Approach for Small Matrix Factorizations and Applications on GPUs