3 ms·
The Rust Compilation Model Calamity
- atombender 7y agoGlad lots of bright minds are working on this, and the problem isn't being downplayed by the community. This is one instance where I think we can learn from Go's "ergonomics first" principles, even though that strategy has cost it a place in the benchmark rankings. I'm very curious about how Jonathan Blow's Jai language will turn out, as it supposedly has a dedicated backend for debug builds which is naive and doesn't optimize, and produces (apparently; the compiler isn't public, so there's no way to verify this) compilation times on par with Go. Anyone know what happened around 2019-09-01, when all the compile times dipped quite a bit in the graph?
- est31 7y ago> Anyone know what happened around 2019-09-01, when all the compile times dipped quite a bit in the graph? The dip is roughly in the center between 2019-07-01 and 2019-09-01, so I'd say it was somewhere in the start of August. As the usual suspect for speed improvements nnethercote seems to have gotten no merged PRs in that date range, I suspect it's the impact of pipelined compilation [1]. The PR that enabled it for rust's nightlies [2] was merged on Aug 1 so the date would fit quite well. [1]: https://github.com/rust-lang/cargo/pull/7143 https://github.com/rust-lang/cargo/pull/7143 [2]: https://github.com/rust-lang/rust/pull/63181 https://github.com/rust-lang/rust/pull/63181
- atombender 7y agoThanks. I was referring to the dip that ends 2019-09-01 and then reverts to the earlier value, i.e. an apparent regression.
- est31 7y agoNo idea why the dip ended at that day.