[fused moe kernel] bug for column major?

[sleepwalker2017]

https://github.com/pytorch-labs/applied-ai/blob/main/kernels/triton/inference/col_major_moe_gemm/v2_moe_fused.py

    pid_m = (pid % grid_n)
    pid_n = pid // grid_m

    return pid_m, pid_n

The column major seems to give wrong result. In fact, a lot of blocks doesn't do any computation because they just return because of incorrect value. So the kernel seems much faster, but it's wrong. right?