> graphs, compare, dashboard, bootstrap, status.

Warning: although measurements known to have high variation are marked with '?'/'??', this does not mean that unmarked measurements are guaranteed to have low variation. Verify the measurement against the last "noise run" which shows the perf difference of a non-functional change.

Cache states

Most benchmarks have at least 4 cache states for which we collect data:

• full: a non-incremental full build starting with empty cache
• incr-full: an incremental build starting with empty cache
• incr-unchanged: an incremental build starting with complete cache, and unchanged source directory -- the "perfect" scenario for incremental.
• incr-patched: an incremental build starting with complete cache, and an altered source directory. The typical variant of this is "println" which represents the addition of a `println!` macro somewhere in the source code.