Skip to main content

emit_tiled_dp4a_bench

logicaffeine_compile::codegen_cuda::gemm_tile

Function emit_tiled_dp4a_bench 

Source 
pub fn emit_tiled_dp4a_bench(
    tile: &CudaTile,
    m: usize,
    k: usize,
    n: usize,
) -> String
Expand description

Emit a TIMING bench for the tiled kernel at M×K×N: fills A/B on device, warms up, times 30 launches with CUDA events, and prints GOPS <value> (2·M·N·K / seconds). Used by the autotuner to measure each lattice point on the real device and lock the fastest.