The interesting thing about Crypto performance tuning is that you really have to ensure that no logical path does a different amount/kind of work than another(i.e. no short circuiting). I used to not think much of it until I saw an RSA private key recovered via acoustic analysis of capacitor whine due to a short circuit condition in a function to multiply two large numbers.(this was using a recent release of openssl) To my knowledge no other area of programming really has this pitfall
If this "I saw" has any further public details, I'd absolutely love to learn more, and I'm pretty sure others would as well.
In particular, I'm especially interested in electrical or real-world attacks - such as capacitor whine! - that can be applied a weakened security situations like asymmetric logic/branching. I vaguely recall CPU voltage fuzzing is a thing, I want to go learn more about that at some point.
I couldn't find the specific lecture/demo that I went to, but I found a video by the same guy with a similar presentation elsewhere. Coincidentally he's also one of the researchers who published the original paper on Meltdown
