“Processing the input file in parallel provides a huge win over r1, taking the time from 1 minute 45 seconds to 22.6 seconds. For comparison, the previous “optimised non-parallel” version, solution 7, took 22.1 seconds.”
Seven rounds of single-threaded tuning got Hoyt to 22.1 seconds. Throwing cores at the naive version got 22.6. Combining both hit 3.4 seconds with only the standard library, no unsafe and no mmap. The fastest Java entry still runs almost 4x faster, written by the founder of GraalVM. Profile first, then parallelise. Most teams do neither.