gather_bgmv_grids

spark_model::layers::ops::moe_lora_grouped

Function gather_bgmv_grids 

Source 
pub fn gather_bgmv_grids(
    max_rank: u32,
    n_out: u32,
    n_slots: u32,
) -> ([u32; 3], [u32; 3])
Expand description

PURE (GPU-free, unit-tested): the exact (shrink, expand) grid triples for the decode gather-fold, given the route dims and the flat row count. Each is [ceil(out/4), n_slots, 1] with a (256,1,1) block (one 64-lane group per output, N_PER_BLOCK = 4). n_slots is a host constant per captured graph, so the shape is EXACT (no worst-case tiles) and capture-stable.