numpy>=1.24
xdsl>=0.62

[benchmark]
jax[cuda12]

[gpu]
cuda-tile>=1.2
cupy-cuda13x

[test]
pytest>=7.0
