rustc-timings

Profile rustc compilation using nightly Rust with the -Zself-profile flag to generate timing data. Process the output with measureme tools (summarize and crox), convert to Chrome profiler format, then analyze with DuckDB to identify bottlenecks in trait resolution, monomorphization, and LLVM backends.

Slow builds typically stem from trait resolution complexity, excessive monomorphization, or LLVM backend overhead. Self-profiling with -Zself-profile captures per-crate timing and monomorphization counts, revealing which components consume the most compilation time.

Yes. Self-profiling works across nightly Rust toolchains, allowing you to compare compilation metrics and identify performance regressions. DuckDB-based analysis aggregates and filters timing data to spot differences between toolchain versions.

You need nightly Rust for -Zself-profile support, measureme utilities (summarize to process profiling output and crox to convert to Chrome format), and DuckDB to query and analyze timing distributions and per-crate metrics.

Use the summarize tool to process self-profile output, then crox to convert the result to chrome_profiler.json format, which can be visualized in Chrome DevTools or analyzed with DuckDB for structured queries on timing data.

Self-profiling captures per-crate compilation time, monomorphization counts, trait resolution steps, and LLVM backend duration. DuckDB analysis extracts hotspots and time distribution across components to pinpoint optimization opportunities.