Function qsa_score_rows_tc 

pub fn qsa_score_rows_tc(
    gpu: &dyn GpuBackend,
    kernel: KernelHandle,
    q: DevicePtr,
    block_keys: DevicePtr,
    scores: DevicePtr,
    rows: u32,
    n_blocks_max: u32,
    first_pos: u32,
    score_stride: u32,
    ratio: u32,
    stream: u64,
) -> Result<()>

2026-09-25: Tensor-core qsa_score_rows: one block of 8 warps covers 16 rows by 64 blocks instead of one block per (row, block). q is split into BF16 high and low halves for the MMA, so the scores are not bit-identical to qsa_score_rows; they feed a top-k.