Hacker News Reader

Also code produced by LLVM-based languages can be linked with low-level modules written in pure C/C++, or even assembly, if necessary. So one may quickly prototype and then slowly rewrite the software for performance, module by module. No such thing is possible for JIT-based languages or platforms.

View on HN

As an embedded engineer, you and parent have put into words something that I've known but haven't been able to articulate.

I think I encountered the sentiment (on HN) before that because CPU and memory speeds have become so far out of whack and that increasing CPU speeds have power limits that engineers will start having to worry about optimization more and more. I suppose the difference is that before, we didn't have the boat-loads of memory that we have now (in a system).

In other words, yay for LLVM IR!

Reply on news.ycombinator.com