arm64: cdef_dir: Preload rows to prevent stalling
After: Cortex A55 cdef_dir_8bpc_neon: 360.2 cdef_dir_16bpc_neon: 384.0
Before: Cortex A55 cdef_dir_8bpc_neon: 391.2 cdef_dir_16bpc_neon: 413.0
After: Cortex A55 cdef_dir_8bpc_neon: 360.2 cdef_dir_16bpc_neon: 384.0
Before: Cortex A55 cdef_dir_8bpc_neon: 391.2 cdef_dir_16bpc_neon: 413.0
VideoLAN code repository instance