All benchmarks measure computation time from JS and Python, respectively. To learn more, check out benchmark methodology.
Benchmark Summary
- Average speedup: 2.07x vs NumPy
- Best case: 63.03x
- Worst case: 0.08x
- Total benchmarks: 2390
- Machine: Apple M4 Max (16 cores, 128 GB, arm64)
- numpy-ts version: 1.4.0
Performance by Category
| Category | Avg Speedup | Count | Faster | Slower |
|---|---|---|---|---|
| creation | 2.42x | 213 | 189 | 24 |
| arithmetic | 2.65x | 295 | 282 | 13 |
| math | 0.91x | 125 | 50 | 75 |
| trig | 1.10x | 216 | 119 | 97 |
| gradient | 7.54x | 22 | 22 | 0 |
| linalg | 3.19x | 269 | 230 | 39 |
| reductions | 1.77x | 413 | 328 | 85 |
| manipulation | 2.25x | 231 | 183 | 48 |
| io | 2.96x | 66 | 51 | 15 |
| indexing | 1.29x | 115 | 52 | 63 |
| bitwise | 2.02x | 10 | 10 | 0 |
| sorting | 1.32x | 75 | 45 | 30 |
| logic | 4.18x | 142 | 130 | 12 |
| statistics | 4.00x | 26 | 23 | 3 |
| sets | 2.46x | 33 | 24 | 9 |
| random | 1.67x | 46 | 41 | 5 |
| polynomials | 4.36x | 27 | 21 | 6 |
| fft | 1.24x | 66 | 30 | 36 |
Performance by DType
| DType | Avg Speedup | Median Speedup | Count |
|---|---|---|---|
| float64 | 1.90x | 1.73x | 288 |
| float32 | 2.21x | 2.02x | 238 |
| float16 | 2.25x | 2.28x | 211 |
| int64 | 1.54x | 1.36x | 196 |
| uint64 | 1.48x | 1.35x | 188 |
| int32 | 2.00x | 2.05x | 204 |
| uint32 | 2.03x | 1.99x | 189 |
| int16 | 2.32x | 2.69x | 185 |
| uint16 | 2.30x | 2.72x | 184 |
| int8 | 2.74x | 2.88x | 185 |
| uint8 | 2.68x | 2.79x | 186 |
| complex128 | 1.88x | 1.68x | 68 |
| complex64 | 1.77x | 1.56x | 68 |