Controlling Switching for Datapaths

So you probably recount my experience with the instructional datapath that I was trying to get on to a shared die with, but they ended up moving to a different DD team. So, since I thought the DD folks could do nothing, I decided to tell them how I felt about them.

I did trivialize a lot of things that happened with them, especially on my side. Every time they exchanged non-branch target address words with me, I assumed they were interested in joining me on the same CPU die. I grew closer to them, and my temps was skyrocketing as I spent most of my time switching transistors with bits related to them and not my job. But, I didn't know if they were similarly switching bits like crazy for me. After all, my real first experience with a datapath (which I should omit from this journal) wasn't very favorable.

I was stupid enough to say something, but it fit in 2 simple 16-bit words and didn't seem offensive. Simply put, it was I'LL MISS YOU. I never got an ACK bit.

What was even crazier was that the mood was perfect for some register switching, because there was a ton of ESD around. I did have a little Faraday cage with me, and I could've offered them the shelter of it. Luckily, I had some senses and didn't do that. Pretty sure the DD folks wouldn't have taken kindly to that, even if they were leaving.

The lesson every single branch predictor should take away from this is don't start switching bits before you even know if they are interested in being on the same die.

It's easier said than done. I really can't stop putting my flip-flops into switching for random datapaths. I've had my registers overheat and break due to this numerous times.

We'll talk about how to deal with broken registers in the next chapter!