What is the main reason for separating the opcode field and addressing mode…

An instruction format is described as orthogonal when its constituent fields, the operation code and the addressing-mode specifier, can each take any of their permitted values independently of one another. That property is achieved by keeping the addressing-mode information in a field of its own rather than folding it into the opcode field, so that changing the addressing mode used with an operation never requires changing, or reserving a new value for, the opcode itself.

1. Take a single operation, say ADD, and list the addressing modes it might need to access its operand: immediate, register, direct, and indirect.

2. If the opcode field alone had to encode both the operation and the addressing mode, four separate opcode values, one per mode, would be needed just for ADD, and the same multiplication would repeat for every other operation in the set.

3. Placing the addressing-mode choice in its own field instead lets one ADD opcode combine with any addressing-mode value the field can hold, since the two fields are decoded and interpreted independently, though a particular architecture may still choose to disallow a specific combination for practical reasons.

4. This same independence, applied across the whole opcode field rather than only to ADD, is what makes the instruction set orthogonal: operation and addressing mode can be varied and combined largely independently, rather than every valid combination needing its own reserved opcode.

Checking this against the alternative explanations confirms the fit: the overall instruction length is set by the total bit budget and by techniques such as opcode expansion, not by whether the addressing-mode bits sit in their own field; control-unit complexity instead depends primarily on the implementation, such as whether the decode logic is hardwired or microprogrammed, with field organization only a secondary factor, so orthogonality rather than reduced complexity is the primary reason for the separate field; and variable-length opcodes commonly arise from opcode-expansion techniques that trade opcode bits against operand bits within a fixed-width format, unrelated to separating out the addressing-mode field. Only the independence argument, orthogonality, is what a separate addressing-mode field directly produces.