> ## Documentation Index
> Fetch the complete documentation index at: https://brushysuite.gfrancodev.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Benchmarks

> Comparative Tier 1 benchmarks for @brushy/di-core vs tsyringe, InversifyJS, and awilix.

# Benchmarks

**@brushy/di-core ranked #1** among DI libraries (tsyringe, InversifyJS, awilix) in **8/8 scenarios**. Baseline (`new` direct) is measured separately as overhead reference, not ranked against DI runtimes.

## Environment

| Field    | Value                                   |
| -------- | --------------------------------------- |
| Date     | 2026-09-12                              |
| Node     | v22.23.2                                |
| Platform | linux x64                               |
| CPUs     | 16 × AMD Ryzen 7 5700X 8-Core Processor |
| Config   | `time=3000ms`, `runs=5` (full suite)    |

<Note>
  Results are indicative. Run benchmarks on an idle machine for stable numbers. They are not a substitute for profiling your own app.
</Note>

<Note>
  Published numbers come from the full suite (`time=3000ms`, `runs=5`). CI on `main` refreshes `BENCHMARK.md` weekly via [`.github/workflows/benchmark.yml`](https://github.com/brushysuite/brushy-librarys/blob/main/.github/workflows/benchmark.yml).
</Note>

## Speed advantage over best competitor

```mermaid theme={null}
xychart-beta
    title "Brushy throughput advantage vs best competitor (%)"
    x-axis ["cold", "warm", "transient", "deep", "wide", "factory", "register", "request"]
    y-axis "% faster" 0 --> 650
    bar [77, 25, 18, 38, 50, 26, 69, 583]
```

## Throughput (p50) by scenario

| Scenario        | @brushy/di-core | Best competitor    | Faster    | Throughput (relative)        |
| --------------- | --------------- | ------------------ | --------- | ---------------------------- |
| singleton\_cold | 4.52M/s         | tsyringe 2.56M/s   | **+77%**  | `██████████` vs `██████░░░░` |
| singleton\_warm | 8.33M/s         | awilix 6.67M/s     | **+25%**  | `██████████` vs `████████░░` |
| transient       | 6.21M/s         | inversify 5.26M/s  | **+18%**  | `██████████` vs `████████░░` |
| deep\_graph     | 12.50M/s        | awilix 9.09M/s     | **+38%**  | `██████████` vs `███████░░░` |
| wide\_graph     | 12.50M/s        | awilix 8.33M/s     | **+50%**  | `██████████` vs `███████░░░` |
| factory\_deps   | 8.33M/s         | awilix 6.62M/s     | **+26%**  | `██████████` vs `████████░░` |
| register\_batch | 648.09K/s       | tsyringe 382.41K/s | **+69%**  | `██████████` vs `██████░░░░` |
| request\_scope  | 3.83M/s         | awilix 560.85K/s   | **+583%** | `██████████` vs `█░░░░░░░░░` |

`request_scope`: tsyringe and inversify have no native request scope (N/A). Bars are scaled per row (@brushy/di-core = 10 blocks).

## Full results and methodology

* [BENCHMARK.md](https://github.com/brushysuite/brushy-librarys/blob/main/packages/di-bench/results/BENCHMARK.md): executive summary, rankings, and detailed metrics
* [METHODOLOGY.md](https://github.com/brushysuite/brushy-librarys/blob/main/packages/di-bench/METHODOLOGY.md): scenario definitions, adapters, and CI refresh policy
* [latest.json](https://github.com/brushysuite/brushy-librarys/blob/main/packages/di-bench/results/latest.json) / [latest.csv](https://github.com/brushysuite/brushy-librarys/blob/main/packages/di-bench/results/latest.csv): machine-readable exports

## Run locally

Build `@brushy/di-core`, then run the benchmark suite from the project root:

```bash theme={null}
npm run build-packages
npm run bench:quick
```

For the full CI-equivalent suite (5 runs × 3s per task, updates published artifacts):

```bash theme={null}
npm run bench:report
```

Partial runs write gitignored `partial-*` files under `packages/di-bench/results/`. Only full runs update `latest.*` and `BENCHMARK.md`.