X64汇编中0x40 REX前缀opcode的作用是什么？

Great question—this is a common point of confusion when diving into x86-64 instruction encoding. Let's break this down clearly:

What Does the 0x40 REX Prefix Actually Do? #

First, let's recall the structure of REX prefixes (which range from 0x40 to 0x4F). A REX prefix uses 4 bits to extend instruction fields:

• REX.W (bit 3): Extends operand width to 64 bits (when set)
• REX.R (bit 2): Extends the reg field in ModRM bytes
• REX.X (bit 1): Extends the index field in SIB bytes
• REX.B (bit 0): Extends the r/m field in ModRM or base field in SIB

The 0x40 prefix is 01000000 in binary—all four extension bits are 0. That means it doesn't modify any part of the following instruction. So 40 53 behaves exactly like 53 (both encode push rbx) because the prefix adds no functional change.

Why Would Compilers/Assemblers Generate This Redundant Prefix? #

Even though it's functionally useless, there are a few practical reasons you might see this in compiled code like Kernel32.dll:

• Simplified Code Generation: Many compilers' code generators avoid complex conditional checks by adding a REX prefix uniformly for 64-bit register operations, even when the instruction doesn't require it. For example, instead of checking if each instruction already supports 64-bit registers natively (like push rbx), the generator just slaps on a REX prefix to cover all cases—this reduces code complexity in the compiler itself.
• Instruction Alignment: CPUs perform better when instructions are aligned to certain boundaries (e.g., 2-byte, 4-byte, or 16-byte). If inserting a 1-byte redundant REX prefix helps align a subsequent critical instruction to an optimal address, the compiler will do it to improve performance.
• Legacy/Compatibility Reasons: Some older code generation tools or assemblers have retained this habit, or it might be used to ensure compatibility with specific debuggers, emulators, or binary analysis tools that expect REX prefixes in certain contexts.

Key Takeaway #

The 0x40 REX prefix is completely redundant in cases like push rbx—it doesn't change the instruction's behavior, but it's a valid (if unnecessary) part of x86-64 encoding. Compilers generate it for practical, implementation-specific reasons rather than functional ones.

内容的提问来源于stack exchange，提问作者c00000fd