Initialize a dense matrix at GPU shared memory using COO index

Question

All

I want to use the sparse matrix information (e.g., COO) to initialize a dense matrix. I find my CUDA program\'s major performance bottleneck here.

__sh