cuprate-hinto-janai/benches
hinto.janai b538272f4d
Some checks failed
Audit / audit (push) Has been cancelled
Deny / audit (push) Has been cancelled
readme, basic examples
2024-06-26 17:03:17 -04:00
..
criterion/cuprate-json-rpc readme, basic examples 2024-06-26 17:03:17 -04:00
harness readme, basic examples 2024-06-26 17:03:17 -04:00
README.md readme, basic examples 2024-06-26 17:03:17 -04:00

Benches

This directory contains Cuprate's benchmarks and benchmarking utilities.

1. File layout and purpose

This directory is sorted into 4 important categories:

Sub-directory Purpose
harness/src Cuprate's custom benchmarking harness binary
harness/lib Cuprate's custom benchmarking harness library
harness/* Macro-benchmarks for whole crates or sub-systems (using Cuprate's custom benchmarking harness)
criterion/* Micro-benchmarks for crates (e.g. timings for a single function)

2. Harness

The harness is:

  • cuprate-harness; the actual binary crate ran
  • cuprate-harness-lib; the library that other crates hook into

The purpose of the harness is very simple:

  1. Set-up the benchmark
  2. Start timer
  3. Run benchmark
  4. Output data

The harness runs the benchmarks found in harness/.

The way benchmarks "plug-in" to the harness is simply by implementing cuprate_harness_lib::Benchmark.

See cuprate-harness-lib crate documentation for a user-guide:

cargo doc --open --package cuprate-harness-lib

2.1 Creating a harness benchmark

  1. Create a new crate inside benches/harness (consider copying benches/harness/test as a base)
  2. Pull in cuprate_harness_lib as a dependency
  3. Implement cuprate_harness_lib::Benchmark
  4. Add a feature inside cuprate_harness for your benchmark

2.2 Running a harness benchmark

After your benchmark is implemented, run this command:

cargo run --release --package cuprate-harness --features $YOUR_BENCHMARK_CRATE_FEATURE

For example, to run the test benchmark:

cargo run --release --package cuprate-harness --features test

3. Criterion

Each sub-directory in here is a crate that uses Criterion for timing single functions and/or groups of functions.

They are generally be small in scope.

See criterion/cuprate-json-rpc for an example.

3.1 Creating a Criterion benchmark

  1. Copy criterion/test as base
  2. Read the Getting Started section of https://bheisler.github.io/criterion.rs/book
  3. Get started

3.1 Running a Criterion benchmark

To run all Criterion benchmarks, run this from the repository root:

cargo bench

To run specific package(s), use:

cargo bench --package $CRITERION_BENCHMARK_CRATE_NAME

For example:

cargo bench --package cuprate-criterion-json-rpc