README.md

Haversine Distance implementations for SimSIMD

This repository serves as a benchmark and a staging area for geospatial metrics to be included in the SimSIMD library and used in USearch. To run the benchmarks across all implementations:

bash main.sh

One can expect the following results:

• 0.py: SkLearn version. 2.9 secs for 1e5 iterations. 34.5 KOps/sec.
• 1.py: Raw Python version. 8.5 secs for 1e7 iterations. 1.18 MOps/sec.
• 2.py: NumBa version. 3.2 secs for 1e7 iterations. 3.12 MOps/sec.
• 3.c: C version. 1.1 secs for 1e8 iterations. 90.9 MOps/sec.
• 4.c: SIMD version. 2.5 secs for 1e9 iterations. 400 MOps/sec.