Enable Link-Time Optimization (LTO) and codegen-units = 1

[zamazan4ik]

Hi!

I noticed that in the Cargo.toml file Link-Time Optimization (LTO) for the project is not enabled. I suggest switching it on since it will reduce the binary size (always a good thing to have) and will likely improve the application's performance (due to more aggressive compiler optimizations). If you want to read more about LTO and its possible modes, I recommend starting from this Rustc documentation.

I think you can enable LTO only for the Release builds so as not to sacrifice the developers' experience while working on the project, since LTO consumes an additional amount of time to finish the compilation routine. In this case, we can create a dedicated [profile.optimized-dev] profile where LTO will be disabled (so developers experience will not be affected).

Basically, it can be enabled with the following lines to the root Cargo.toml file:

[profile.release]
codegen-units = 1
lto = true

I have made quick tests (AMD Ryzen 9 5900x, Rust 1.86, the latest version of the project at the moment, maturin build --release command) - here are the results for the resulting wheel:

• Release: 13 Mib, clean build time: 1m 12s
• Release + codegen-units = 1 + Fat LTO: 11 Mib, clean build time: 2m 44s

Thank you.

[lemorage]

Thanks for sharing! This is a very interesting suggestion. I learned a lot from the shared links, and I also saw a lot of debates on this¹²³ too. I tested on my own machine, and got the kinda same results.

Config	Build Time	Final binary Size
Without LTO	~3m 21s	~13MB
With LTO + codegen=1	~4m 34s	~11MB

I don't have strong objections to using lto, but the optimized results do not surprise me at this moment. Since cocoindex is still fairly small currently, I don’t see any huge diff benefited from this yet. But as the project grows and things get more complex, I think maybe we could see more different insights on this (as cited below, those maintainers all have different tradeoffs and choices). Anyway, just my two cents, @badmonster0 might have more words to say.

Footnotes

1. https://github.com/astral-sh/ruff/pull/9031 ↩

2. https://github.com/alacritty/alacritty/issues/6467 ↩

3. https://github.com/rust-lang/rust-analyzer/issues/4692 ↩

[badmonster0]

Thanks for the suggestions and investigations, @zamazan4ik and @lemorage !

I think we can enable it for the release workflow as @zamazan4ik suggested. The benefit of smaller binary size and faster execution will justify the extra build time during release. I'll create a PR to try it out later.