NEON is the new black: fast JPEG optimization on ARM server

This post details the optimization of the jpegtran library for ARM processors using NEON SIMD instructions. It addresses a significant performance regression observed on ARM compared to Intel CPUs and demonstrates how applying NEON intrinsics, similar to previous SSE optimizations, can achieve over 2X speedup in JPEG optimization, bringing ARM performance closer to that of Xeon processors for the Polish image optimization service.