# NVIDIA GeForce RTX 4090, 128 SMs, sm_89, iters=40000 reps=3
mode       best TF/s   all runs
f16/f16        331.6 331.6/331.6/330.9
f16/f32        330.8 330.8/330.8/330.8
bf16/f32       330.3 330.3/328.3/328.2
s8/s32         659.5 659.5/659.4/659.4
dp4a           164.1 164.1/164.0/164.1
ffma            79.1 79.0/79.1/79.1
