WARNING: A bug in commit f51a163 causes this release to miss some duplicates. Use v1.2.1 or newer instead!
This release primarily contains algorithm changes that will improve performance on systems with smaller CPUs. Also included is an experimental tree rebalancing algorithm if you're inclined to compile and play with it; while it does a fantastic job of reducing the maximum file tree depth, it doesn't improve performance in benchmarks and occasionally makes it worse. Regardless, it's available for testing with the -DUSE_TREE_REBALANCE directive passed to the compiler.
Overall cache misses on cachegrind benchmarks went down somewhat in benchmark tests; simulating v2.1 on the L1 caches of an Intel Atom D2700 showed 5% less L1 cache misses than v2.0.2.