Skip to content

usage.md

Latest commit

 

History

History
100 lines (74 loc) · 1.88 KB

usage.md

File metadata and controls

100 lines (74 loc) · 1.88 KB
Raw

Running Benchmarks

Part of the main README.

Run Benchmarks

Use make measure (wrapper around scripts/measure.sh):

make measure lang=go

Run all languages + all 8 core benchmarks in one combined run:

make measure

Filter language/benchmark and change profile/iterations:

make measure lang=go profile=measure
make measure lang=go profile=test
make measure lang=go,c bench=binary-trees,mandelbrot iterations=10
make measure lang=rust iterations=3

profile=measure is the default. iterations=1 is the default.

Direct Runner Script

Direct script usage (future KWA-compatible shape):

scripts/measure.sh profile=measure lang=go,c,cpp bench=binary-trees,mandelbrot iterations=10

Supported script args (key=value only):

  • profile=measure|test
  • lang=<csv>
  • bench=<csv>
  • iterations=<int>
  • gmt_dir=<path> (optional, default: ./green-metrics-tool)
  • uri=<path> (optional, default: repo root)

When lang is omitted, all supported languages are used:

  • c
  • cpp
  • csharp
  • dart
  • erlang
  • fsharp
  • go
  • haskell
  • java
  • lua
  • nodejs
  • ocaml
  • perl
  • php
  • python
  • ruby
  • rust
  • swift

When bench is omitted, these 8 core benchmarks are used:

  • binary-trees
  • fannkuch-redux
  • k-nucleotide
  • n-body
  • regex-redux
  • spectral-norm
  • fasta
  • mandelbrot

Profile behavior:

  • measure: uses canonical files benchmarks/<lang>/<benchmark>.yml and does not pass --dev-no-sleeps
  • test: uses generated files benchmarks/<lang>/<benchmark>_test.yml and passes --dev-no-sleeps

KWA Build/Run

KWA is the Go CLI that exports GMT measurements to CSV. Internals are documented in kwa/README.md.

Build KWA binary into kwa/build/kwa:

make kwa-build

Run KWA directly from source (kwa/cmd/main.go):

make kwa-run