Performance

Rune runs about 1.3 compiler processes per core (-j, build.jobs) and hands every rustc and build script the same GNU jobserver, so a compiler that is alone at the end of a build can take the idle slots for its codegen threads while the total never exceeds the budget. A crate starts the moment the .rmeta of each dependency exists; a build script is its own step that starts once its build-dependencies are built, overlapping the rest of its package's dependencies. Before starting another compiler Rune checks available memory (Linux MemAvailable, macOS memory pressure) and, while it is short, lets running units finish first — a machine with many cores and little memory neither swaps nor loses a rustc to the OOM killer. Each compiler's peak memory is measured (wait4) and remembered, so a crate known to need 2 GiB waits until 2 GiB are free instead of starting next to three others like it. Starting a job never waits on the jobserver pipe either: rustc's codegen threads borrow idle tokens, and a job that starts while they hold them all simply owes one, repaid as they return — the critical path never waits behind another crate's code generation. build.low-priority = true runs every compiler at a lower scheduling priority so the rest of the machine stays responsive (rune watch always does).

What stable Rust cannot parallelize is a chain of crates that must be compiled one after another, and a single crate's front end. A build longer than 10 s that kept fewer than 60 % of the cores busy ends with one line naming that chain; rune build --timings (or RUNE_TIMINGS=1) prints it in full, with how long each phase of the command took, and writes .rune/timings/rune-timing.html — a timeline of every step with the chain highlighted and how many ran at once — plus a Chrome trace in .rune/fingerprints/trace.json (https://ui.perfetto.dev). Splitting the largest crate of that chain is usually what helps. On a nightly compiler, build.frontend-threads = "auto" gives crates that start while cores are idle a parallel front end.