• added various ARM optimization (thanks @georges-arm)
• added a DASH optimized decoder refresh mode (idr_no_radl)
• added a GOP adaptive QP cascade (only for non-QPA case, which already does it on GOP/frame/block level)
• improved rate control and QPA haromization, especially for preset fast
• added experimental film grain analysis
• memory reductions for presets faster/fast/medium
• preset tuning
  • 5% speedup for medium and 2% for slow/slower
• other improvements, cleanups and fixes