Reference

Benchmark

GitHub Edit page

Benchmark

A WASM codec is only useful if it keeps up with realtime. The repo ships a benchmark that pits libopus-wasm against the native @discordjs/opus addon on the same frames.

#Run it

pnpm benchmark

This builds the library and then runs scripts/benchmark-native.mjs. The native comparison requires @discordjs/opus to build on the host (a dev dependency), so a working node-gyp toolchain is needed for the benchmark — though never for using libopus-wasm itself.

#Sample result

Apple Silicon, Node 26, 20,000 iterations, 48 kHz stereo, 20 ms frames:

wasm encode:   15,304 ops/sec
native encode: 15,741 ops/sec
wasm decode:   38,416 ops/sec
native decode: 41,280 ops/sec

Encode runs within a few percent of native; decode is close behind. For 20 ms frames that is far above the ~50 frames/sec a single realtime stream needs, with ample headroom for many concurrent streams.

These numbers are a regression check, not a portable score. Absolute throughput depends on CPU, Node version, and Emscripten flags, so compare runs on the same machine rather than against the figures above.

#In CI

The GitHub Actions workflow exposes a manual Benchmark job (workflow_dispatch) so you can capture numbers on the CI runner on demand, separate from the per-commit test run.

#Next