# NVIDIA GeForce RTX 4090, 128 SMs, sm_89, iters=20000 reps=3
mode       best TF/s   all runs
f16/f16        306.3 306.3/306.3/306.3
f16/f32        331.6 331.6/331.6/331.6
bf16/f32       330.6 329.5/330.6/329.6
s8/s32         661.4 661.4/661.3/661.3
dp4a           164.4 164.4/164.4/164.4
ffma            79.1 79.1/79.0/79.1
