Period 4. Max throughput. I'm still 100% sure that period 3 is possible, though it would be very different; your machine would actually be slower since you'd take 3 cycles between input grabs instead of 2, seemingly giving you more time, but 3 instructions per component is extremely restrictive, as explained in my notes for my first submission. This would likely require pipelines that are very different than the ones I've used so far. Still, I'm 100% sure that someone like biggie could easily hit 4h p3. I've discussed P2 in my notes for my first showcase solution already, but I don't have anything new to add here. To players who have just tuned in and haven't heard said notes on P2, I'll summarize saying "3i but everything is wanded".
Either way, I had lots of fun optimizing for (throughput,period), and stuff like this is why I wish "choose your own metric" was viable in a tournament.