# NVIDIA GeForce RTX 4090, 128 SMs, sm_89, iters=20000 reps=3
mode       best TF/s   all runs
f16/f16        325.7 306.3/306.3/325.7
f16/f32        165.6 165.6/165.6/165.6
bf16/f32       165.5 165.4/165.5/165.5
s8/s32         661.6 661.5/661.6/661.4
dp4a           164.5 164.5/164.5/164.5
ffma            79.1 79.1/79.1/79.1
