I'm interested to see how a compiler for the Mill CPU would handle such large instruction widths. Whilst it doesn't have the restriction VLIW had of having a fixed instruction width, I'm not entirely convinced we would often see 30+ operations packed in the same instruction.