Comments by "Mikko Rantalainen" (@MikkoRantalainen) on "The Truth About The Fast Inverse Square on N64 | Prime Reacts" video.

20 µs is quite important time even for web programming if you're computing JSON response that consist of 500 elements. If you take 20 µs per element, it will be 10 ms for the whole list already.
5
@MrHamof It's more like debug code has more stuff in RAM and off-by-one errors have higher probability of overwriting some debug information instead of real data so buggy code doesn't crash the system equally often. Of course, when your system is memory bandwidth limited in the first place, having to use more RAM to keep debug stuff in the RAM is not a great plan for high performance.
4
5:00 100% agreed!
1
One could also explain N64 as having no RAM at all but being able to run CPU instructions from VRAM if GPU is stopped while that's happening. And there's 16 kB cache in the CPU itself so as long as you can keep everything in that space, the GPU doesn't need to be stopped.
1
The second column is logically "how many idle cycles we cause for the GPU to wait if we run these instructions on CPU (as a result of using shared memory bus)". But you can allow the GPU to keep running if your code only uses CPU cache instead of RAM.
1
@n00blamer Typically XOR would use syntax a^=b because usually % means remainder for a division.
1
I mostly agree but if you make everything super slow then your code ends up really bad until one single part overwhelms all the rest. I think it's better to decide the required latency at the start and then do whatever it takes to keep the final performance at least as good as your decided latency. For a game, the decided latency could be 1/30 or 1/60 seconds to match typical displays. For a web service, the target latency could be 50 or 100 ms. Then you know when you have to start optimizing: whenever your existing code cannot meet the latency requirement. What to optimize? The part that overwhelms everything else at that time.
1