Redis 8.8 performance improvements: Faster string, hash, streams, SCAN & more

This post details performance improvements in Redis 8.8, focusing on MGET/MSET (up to 68% faster with pipelined I/O threads via memory prefetching), HGETALL (up to 25% faster on large hashes via broader prefetch and batched dict bucket prefetch), SCAN family (up to 40% faster via stack-allocated vector containers replacing linked lists), Streams (up to 83% faster XREADGROUP via specialized radix tree optimizations), Sorted Sets (up to 74% faster ZADD/ZINCRBY/ZRANGEBYSCORE via pure-C fast_float implementation and widened Clinger fast path), and Bitmap operations (up to 28% faster on x86 via AVX-512 vectorization). Persistence and replication also see speedups, including 60% faster diskless full sync and reduced BGSAVE overhead through MADV_DONTNEED and disabling memory tracking in the fork child. Replication feed path is optimized, and I/O threads now service primary/replica clients.