# In case of NVIDIA GeForce RTX 3080 Laptop GPU
$ yes |  MATMUL_VERSION=8  make  | grep 'Kernel version\|GFLOPS'
[info] Dispatching Kernel version 8: 2D blocktiling with loop unrolling and vectorization, 30 iterations ...
113.3 milliseconds / dispatch ~ 2426.99 GFLOPS
$ yes |  MATMUL_VERSION=9  make  | grep 'Kernel version\|GFLOPS'
[info] Dispatching Kernel version 9: 2D blocktiling with loop unrolling, vectorization and transpose, 30 iterations ...
74.7 milliseconds / dispatch ~ 3680.25 GFLOPS

# In case of M2 pro
$ yes |  MATMUL_VERSION=8  make  | grep 'Kernel version\|GFLOPS'
[info] Dispatching Kernel version 8: 2D blocktiling with loop unrolling and vectorization, 30 iterations ...
164.3 milliseconds / dispatch ~ 1672.82 GFLOPS
$ yes |  MATMUL_VERSION=9  make  | grep 'Kernel version\|GFLOPS'
[info] Dispatching Kernel version 9: 2D blocktiling with loop unrolling, vectorization and transpose, 30 iterations ...
160.7 milliseconds / dispatch ~ 1710.18 GFLOPS