Thanks for sharing, TIL about interaction nets. I found the article linked at the bottom easier to understand[0].
> Another issue with multi-core CPUs is the difficulty of programming them. Debuggability is a major issue, and entirely new classes of issues arise, such as race conditions. However, in interaction nets, all active pairs are independent, and so can actually be executed in any order. This property is known as confluence, and means that no matter the order in which things are executed, the result will be the same. This makes debugging straightforward and eliminates issues like race conditions. This allows Vine to be parallel out of the box; the programmer does not have to do anything special with threads, processes, CUDA kernels, etc. to get parallel execution.
I read through the first half, but the graph operations are really confusing me. What's the difference between a principal port and auxiliary ports? A lot of the rewrites are swapping them around so I'm confused what their significance is.
I was a bit surprised by Vine's syntax. I expected something a lot more like Lisp.
I'd be interested to see how this would pan out.
Seems to describe something analogous instruction level parallelism, which traditional CPUs made this millennium have as well via instruction pipelines, branch prediction, and speculative execution. Some of which turned out to not work too well so now we generally run hobbled CPUs with Spectre/Meltdown mitigation enabled.
The reason why we tend to reach for difficult abstractions for core level parallelism is that most automatic/safe approaches tend to stumble pretty hard on Amdahl's law, where most parallel code is consequently almost entirely designed around minimizing critical sections.
I don't fully see how inets would resolve Amdahl's law.