Real measured performance data from the What Framework test suite. All benchmarks run on Node.js v22.23.1.
Last updated · v0.11.6 · linux
| Operation | Ops/Second | Time per Op |
|---|---|---|
| Signal create | 1.8M ops/sec | 0.6µs |
| Signal read (100x per iteration) | 1.7M ops/sec | 0.6µs |
| Signal write (no subscribers) | 854K ops/sec | 0.0012ms |
| Signal write (1 subscriber) | 322K ops/sec | 0.0031ms |
| Signal write (10 subscribers) | 111K ops/sec | 0.0090ms |
| Signal peek (untracked read) | 2.1M ops/sec | 0.5µs |
Subscriber overhead is expected — effects run on every write.
| Operation | Ops/Second | Time per Op |
|---|---|---|
| Computed create + read | 506K ops/sec | 0.0020ms |
| Computed chain (depth 5) | 139K ops/sec | 0.0072ms |
| Computed diamond dependency | 237K ops/sec | 0.0042ms |
| Operation | Ops/Second | Time per Op |
|---|---|---|
| Effect create + dispose | 1.3M ops/sec | 0.8µs |
| Effect with 10 signal deps | 381K ops/sec | 0.0026ms |
| Operation | Ops/Second | Time per Op |
|---|---|---|
| Batch 100 writes, 1 effect | 192K ops/sec | 0.0052ms |
| Batch 10 signals, 10 writes each | 164K ops/sec | 0.0061ms |
Batching prevents intermediate computations, making 100 writes nearly as fast as 1.
| Operation | Ops/Second | Time per Op |
|---|---|---|
| Single element creation | 3.7M ops/sec | 0.3µs |
| Nested (3 levels) | 2.9M ops/sec | 0.3µs |
| List of 100 items | 34K ops/sec | 0.0298ms |
| Component call | 5.7M ops/sec | 0.2µs |
| Operation | Ops/Second | Time per Op |
|---|---|---|
| SSR simple element | 414K ops/sec | 0.0024ms |
| SSR nested tree | 231K ops/sec | 0.0043ms |
| SSR list of 100 items | 7K ops/sec | 0.1403ms |
| SSR component tree (50 items) | 28K ops/sec | 0.0363ms |
| Package | Size (gzipped) |
|---|---|
| what-framework (core) | ~4kB |
Measured using gzip -9 compression.
All benchmarks run on Node.js v22.23.1 with the following approach:
• Warmup: 100–1000 iterations discarded before measurement
• Iterations: 500–10,000 per test (varies by operation cost)
• Timing: High-resolution performance.now()
• Calculation: Trimmed mean (discard top/bottom 10%)
• Environment: linux, Node.js v22.23.1
• Automated: Benchmarks regenerate every 3 days via CI
Source code available in /benchmark directory. Run with npm run bench.