zenbench

Task guides for Rust benchmarks and command-line experiments.

Zenbench compares Rust functions in shuffled, serial rounds. Begin with an ordinary Cargo benchmark, then add comparisons and confirmation when needed.

Add a benchmark target

[[bench]]
name = "my_bench"
harness = false

[dev-dependencies]
zenbench = "0.1.11"

Create benches/my_bench.rs using the first benchmark tutorial, then run:

cargo bench --bench my_bench

Default Iterate runs are exploratory. Freeze your workload and candidate before using Confirm to accept a change.

Start with your task

• Write your first Rust benchmark
• Explore a code change
• Confirm a candidate or guard CI
• Compare programs or environment settings
• Confirm an external program or environment change
• Choose CPUs and work on a shared machine
• Keep results and review a small diff
• Export charts for slides and reports
• Investigate memory or CPU work
• Run named study cells and fleet observations

Choose a guide also covers installation, offline help and which commands can make an acceptance decision. These guides describe zenbench 0.1.11.

Begin with the CLI

cargo install zenbench --version 0.1.11 --locked --bin zenbench
zenbench guide
zenbench pair --help

External pair plans retain randomized fresh-process A/B blocks, correctness contracts and A/A controls. Use Iterate to explore, then external Confirm for a frozen block-level decision. For native-suite acceptance, follow confirmation.

Read further

Migrate Criterion benchmarks · Understand the methodology · See report output · API documentation · Source and issues