Compares the performance of JNA interface mapping, direct mapping, and raw dlopen()/dlsym().
Simply execute make to run the benchmarks.
make [NUM_SAMPLES=<# of samples>] [NUM_INVOCATIONS=<# of invocations in each sample>]
Please see here for the workflow run.
| Windows | macOS | Ubuntu |
|---|---|---|
 |
 |
 |
| OS | Interface Mapping (seconds) | Direct Mapping (seconds) | Raw dlopen()/LoadLibrary() |
|---|---|---|---|
| Windows | 0.0514090 | 0.0129870 (3.958x faster) | 0.0001917 (268.114x faster) |
| macOS | 0.0273280 | 0.0063170 (4.326x faster) | 0.0001028 (265.765x faster) |
| Ubuntu | 0.0440980 | 0.0111250 (3.964x faster) | 0.0002869 (153.705x faster) |