Constant W8A16_GEMM_M16_N_TILE 

pub const W8A16_GEMM_M16_N_TILE: u32 = 32;

2026-09-25: N columns one CTA owns on the default instantiation. The launch grid and m16_tc_kernel (dense_ffn_m16_tc.rs) both use it; the kernel’s M16_N_TILE must equal it.