Can CUDA Core and Tensor Core in one SM execute concurrently?

For example, when cuda cores is executing a kernel, can tensor core execute another task at the same time? If not, what about different SMs?

Yes, at the SASS level, in a particular clock cycle, a CUDA core instruction (e.g. FFMA) can be issued in one SMSP while in the same clock cycle, in another SMSP, a tensor core instruction can be issued.