The set of allowed data formats is a bit limiting it seems and may force additional overhead when unpacking / reordering the data. The balance is now towards saving bandwidth at the cost of the CPU cycles (useful for G➝C and C➝G). Allowing wider set would allow to balance this based on the specific use-case needs (C➝C, especially for in-process handoff to a different system for further processing, can favor of CPU saving instead of bandwidth saving). Example may be simply having 2 bytes per sample.
The set of allowed data formats is a bit limiting it seems and may force additional overhead when unpacking / reordering the data. The balance is now towards saving bandwidth at the cost of the CPU cycles (useful for G➝C and C➝G). Allowing wider set would allow to balance this based on the specific use-case needs (C➝C, especially for in-process handoff to a different system for further processing, can favor of CPU saving instead of bandwidth saving). Example may be simply having 2 bytes per sample.