Get the latest tech news
Hashed sorting is typically faster than hash tables
Benchmarks and theoretical explanation of why and when hashed radix sort beats hash tables.
Once the dataset outgrows CPU caches (often around ~1 MiB on a single core, CPU-dependent), both hashing and sorting become limited by cache-line fetch bandwidth to main memory. In some scenarios, such as hash consing, common subexpression elimination, or parser lookup tables, the requirement for batching is a dealbreaker: sorting isn’t viable. This is because hash tables use O(unique keys) memory, while radix sort uses O(accesses); fitting into a smaller footprint often yields better memory-system performance.
Or read this on Hacker News