My bad - I misread a doc recently which implied otherwise, albeit that GCN used shorter ones. Just checked AMD docs straight and it was indeed normal scalar ins

My bad - I misread a doc recently which implied otherwise, albeit that GCN used shorter ones. Just checked AMD docs straight and it was indeed normal scalar instructions. AIE and AIE-ML from AMD do use VLIW btw