https://hgpu.org/?p=17714
Parallel Computing for the Inverse of SPD matrix