# NVIDIA GeForce RTX 4080, 76 SMs, sm_89, iters=20000 reps=3
mode       best TF/s   all runs
f16/f16        213.9 193.2/213.9/213.7
f16/f32        213.7 213.7/213.6/213.7
bf16/f32       213.6 213.6/212.7/212.7
s8/s32         425.7 425.6/425.7/425.6
dp4a           105.8 105.8/105.8/105.7
ffma            50.7 50.7/50.7/50.7
