7d1b9cad87
Adds a pass for Vector to AMX operation conversion. Initially, a direct rewrite for vector contraction in packed VNNI layout is supported. Operations are expected to already be in shapes which are AMX-compatible for the rewriting to occur.