Repository navigation
- Substantially improve performance and memory consumption on large datasets (>10 GB). On one ~30GB dataset, there was an 80x performance increase and dozens of GB less memory consumed while the algorithm was running.
- Prevented surface area falsely computing as zero due to all RAM being consumed