Function k2_ssd_bmm 

pub fn k2_ssd_bmm(c_bnlhr: Tensor<5>, b_bnlhr: Tensor<5>) -> Tensor<5>

Based on the Kernel 2 Triton reference _bmm_chunk_fwd_kernel (ssd_bmm.py).

Returns:

• cb_bnhll [used in K5][!].