https://hgpu.org/?p=17272
A Unified Optimization Approach for Sparse Tensor Operations on GPUs