there’s still work to do. the biggest gaps are SIMD prefilters for non-literal patterns - the dotnet version gets a lot of mileage from vectorized character class membership tests that we don’t have yet - and the bidirectional SIMD routines needed for our right-to-left scanning. pattern-specific optimizations like start-set inference and alternation simplification are also on the list. there’s also a lot of other low-hanging fruit - we don’t have any statistical optimizations yet, the Unicode classes could be baked in instead of constructing them while parsing, the memory usage could also be improved, there are many pattern-specific shortcuts we could add, and so on. but i hope the benchmarks show that it’s already competitive and useful in its current state.
В рыболовной сети нашли 15-метровую тушу редкого кита20:45
。关于这个话题，新收录的资料提供了深入分析
Every minute you aren't running 69 agents, you are falling behind
interested in this work: https://akselmo.dev/posts/breeze-and-union-preparing/
依法开展检察侦查。对司法工作人员利用职权实施的徇私枉法、滥用职权、刑讯逼供等犯罪，依法立案侦查1372人，惩治司法腐败。对公安机关管辖的国家机关工作人员利用职权实施的重大犯罪案件，确需检察机关直接受理的，依法立案侦查69人。