Hacker News new | ask | show | jobs
by ReactiveJelly 909 days ago
Yeah, slow compile loops are the worst part of Rust for me. (The 2nd-worst part is fighting the borrow checker)

I know it'll get faster as hardware marches on, but I often wish I could draw a boundary and say, "Look compiler, this program is months from being done. Let's say everything outside this boundary can be optimized and it won't change. But everything inside the boundary is quick-changing business logic that I'm actively debugging. Don't link or optimize across that boundary. Use wasm to enforce that if you need it."

In C++ I did that with Lua for a while, but the DevEx cost of FFI boundaries is pretty awful even when the performance cost is affordable. I dunno.

2 comments

> I know it'll get faster as hardware marches on

What makes you think this? Nothing else has gotten faster as hardware marches on. Almost all software is significantly slower than the equivalent (if it existed, of course) 20 years ago.

Can you not achieve this in the build system using incremental builds plus disabling LTO plus specifying that some translation units shouldn’t be optimized?