plottypus self-cost measurements (perf-smoke, M5 Pro, macOS 27)

What plottypus costs to leave open in a terminal, and what btop 1.4.7 and macmon 0.8.2 cost on the same machine, measured with perf-smoke: each tool runs under a pseudo-terminal at 160x45 with a fresh config, a 5 s warm-up and a 20 s window, three times, and the median of each metric is reported with its range. Metrics include CPU as a percentage of one core, the kernel's per-task energy estimate in watts, draws and idle draws per second, wakeups, syscalls, context switches, RSS, memory footprint, first frame and snapshot period. Machine: MacBook Pro Mac17,8 (Apple M5 Pro), macOS 27.0, AC power. Each row carries the 1-minute load average and a load class: "light" rows (load 1.2 to 3.8 on 18 CPUs, other builds running) come from the documented runs of 2026-09-23; "saturated" rows (load 56 to 143) are a re-run of the shipped 1.0.0 binary and the two competitors on 2026-09-24. The figures the site prints (about 1.5% of one core at 500 ms, about 1.0% at 1 s, about 0.007 W) are the light-load "perf-pass" rows.

Captured 2026-09-23 to 2026-09-24. Method: perf-smoke: pty at 160x45, 5 s warm-up, 20 s window, median of 3 runs; proc_pid_rusage RUSAGE_INFO_V6 (CPU time, ri_energy_nj, wakeups, footprint); Synchronized-update markers (ESC[?2026h) counted as draws. Full notes: README.md.

Used by: /.

cpu_pct (% of one core)

User plus system CPU time over the window.

energy_w (W)

Kernel per-task energy estimate (ri_energy_nj) over wall time.

idle_draws_per_s (per second)

Frames that changed nothing on screen.

wakeups_per_s (per second)

Interrupt wakeups of the process.

rss_mb (MB)

Resident memory at the end of the window.

Licensed CC BY 4.0: reuse it anywhere, with credit.

Sheth, L. (2026). plottypus self-cost measurements (perf-smoke, M5 Pro, macOS 27) (Version 1.0) [Data set]. plottypus.com. https://plottypus.com/data/plottypus-self-cost