This document by Philippe Marguinaud describes the IO subsystem of ARPEGE/ALADIN/AROME for cycle 39t1. Only historic data IO is covered here, with a few words on Fullpos IO.

Contents :

• Organization of the code
  • Field descriptors
  • IO buffers
  • Extracting/loading fields
  • MFIOOPTS
  • Code cleaning
• Grid-point/spectral
• Traditional input/output
  • Grid-point input
  • Grid-point output
  • Grid-point output alternatives
  • Spectral fields input/output
• Extended traditional output
  • One file per NSTROUT task
  • A single file written by all tasks
  • An alternative to GATHFLNM
• Disabling traditional output
• Compacting fields with OpenMP
• SURFEX IO
  • Anatomy of a SURFEX historic file in FA
  • Conversion from/to LFI
  • Modifying SURFEX historic files
  • Enabling FA
  • The SURFEX cache
• The IO server
  • Implementation of the IO server
  • Namelist parameters
  • Field based mode
  • Gather based mode
  • Reading data produced by the IO server
• Fullpos & SURFEX
• Further developments
  • Reading input files with Multi-Threading
  • Simplifying the code
  • Removing LFI from mse project