关于X86 SDM中ADD指令格式与8086差异及04指令码的技术问询

Great question—this is a super common confusion when transitioning from learning the original 8086 instruction set basics to diving into the full x86 Architecture Software Developer Manual (SDM). Let’s break this down clearly:

1. Your 8086 "general format" isn’t the only 8086 format #

The OPCODE | DIRECTION | WIDTH | MOD | REG | R/M structure you’re thinking of is the general-purpose instruction format used for operations that involve registers, memory, or both (e.g., ADD AX, BX or ADD [BX], CL). But the 8086 instruction set also includes short-form, specialized opcodes designed to save space and speed up common operations—like immediate-value additions to the accumulator register (AL/AX).

2. Breaking down 04 ib #

The 04 ib entry in the SDM refers to one of these specialized 8086-era instructions:

• 04: This is the single-byte opcode that explicitly encodes the operation AL = AL + imm8 (8-bit addition of an immediate value to the AL register). No MOD/RM fields are needed here because the destination is fixed as AL, and the source is an immediate byte.
• ib: The suffix ib stands for immediate byte—this is exactly what you guessed: the second byte in the instruction is the 8-bit immediate value to add to AL.

Example in action #

If you write the assembly instruction:

ADD AL, 0x15

The corresponding machine code will be 04 15—where 04 is the opcode, and 15 is the ib immediate byte.

3. Why this doesn’t match your general format #

The general format you learned applies to instructions where the CPU needs to decode which registers/memory locations to use. For accumulator-specific immediate operations, the opcode itself already defines the destination (AL/AX/EAX/RAX, depending on mode), so there’s no need for the MOD/RM fields. This was a deliberate design choice in 8086 to optimize common operations (since the accumulator was heavily used in early assembly code).

For completeness, there’s also a 16/32/64-bit version of this instruction:

• 05 iw (8086): AX = AX + imm16
• In 32-bit mode, 05 id becomes EAX = EAX + imm32
• In 64-bit mode, 05 id extends to RAX = RAX + imm32 (sign-extended to 64 bits)

Wrap-up #

The SDM’s instruction format descriptions include all x86 instructions—both the general-purpose ones with MOD/RM fields and the specialized short-form ones like 04 ib. This is why it looks different from the single general format you initially learned: you’re looking at a purpose-built opcode for accumulator immediate additions, which was part of the 8086 set from day one.

内容的提问来源于stack exchange，提问作者mahmood