Table 3.
Results from benchmarks comparing performance in Exdir with h5py.
| Library | h5py | Exdir | h5py | Exdir |
|---|---|---|---|---|
| OS | Linux | Linux | Windows | Windows |
| Add 5 attributes | 0.002 s | 0.010 s | 0.002 s | 0.030 s |
| Add 200 attributes (one by one) | 0.066 s | 3.6 s | 0.068 s | 5.5 s |
| Add 200 attributes (single operation) | N/A | 0.030 s | N/A | 0.049 s |
| Add dataset with 106 64-bit floats | 0.009 s | 0.013 s | 0.019 s | 0.040 s |
| Add dataset with 108 64-bit floats | 0.38 s | 0.54 s | 4.4 s | 0.83 s |
| Create 5,000 groups (thorough validation) | 0.26 s | 8.1 s | 0.36 s | 8.9 s |
| Create 5,000 groups (minimal validation) | 0.26 s | 1.1 s | 0.36 s | 8.4 s |
| Create tree (3 groups × 5 levels) | 0.14 s | 0.34 s | 0.14 s | 2.1 s |
| Write 3D slice (100 × 300 × 100) | 0.0033 s | 0.0048 s | 0.031 s | 0.029 s |
A 2GB RAM disk was used as virtual hard drive for the tests. Filename validation was disabled for Exdir in all tests. Software used: Python 3.6, NumPy 1.13.1, Ubuntu 16.04, Windows 7. Hardware used: Linux: Intel Core i7-5820K 3.30 GHz, 32 GB RAM, Intel 535 512 GB SSD. Windows: HP EliteBook 8570p, Intel Core i7-3520M, 2.90 GHz, 8 GB RAM. Samsung 128 GB SSD.