Measure the drawing speed

Drawing writes straight into ratatui’s buffer and allocates nothing; a counting allocator in the tests holds that. The bench draws 20,000 frames into a 150×40 buffer, each with twelve hints and a notice, a status row of five segments, the hints wrapped at 60 columns, a confirm bar, an input bar holding diacritics and wide glyphs, and a masked input bar, and times each part:

git clone https://github.com/gmrdad82/pito-footer && cd pito-footer
cargo run --release --example bench

A number after -- sets the frames: cargo run --release --example bench -- 100000. It prints the mean and the worst frame, then the mean of each part:

pito-footer bench: 20000 frames at 150x40: mean 38.5 µs, worst … µs
  hints and notice: 8.37 µs
  status row: 7.61 µs
  wrapped hints: 5.85 µs
  confirm bar: 3.06 µs
  input bar: 8.08 µs
  masked input bar: 5.52 µs

Those are the numbers measured once on 2026-10-14 on an AMD Ryzen 9 9950X under Linux with Rust 1.97.1, release profile, on the code released as 1.0.0, while other renders ran on the machine. A frame of all six parts takes 0.23% of a 60 fps frame. The worst frame depends on the machine’s load, so it is left out here.

Frame of all six parts, mean	38.5 µs
Allocations while drawing	0
Direct dependencies	3 (4 with crossterm)
Crates in the build, the crate included	42
Lines of Rust in src/, blank lines left out	2,078
Rebuild of the crate after a change, release	1.5 s
Clean release build with every dependency	32.8 s

The build times are one run each on the same machine, with cargo’s default jobs. The chart is drawn with pito-compare from the repository’s render/bench.toml.