DENSE_GEMM_M16_BF16_K_STEP

metrale_model_layers::layers::ops

Constant DENSE_GEMM_M16_BF16_K_STEP 

Source 
pub const DENSE_GEMM_M16_BF16_K_STEP: u32 = 64;
Expand description

2026-09-25: K granularity: the kernel’s 4-stage cp.async pipeline advances 64 elements per step (DGM16_K_STEP), which also keeps every weight row 16-byte aligned.