Star on github
AI Chat

movsx

Move with sign extension

Copies a source into a register, sign-extending when the destination is wider.

386 and later

Intel's reference: MOVSX/MOVSXD

Used in
CFPFAFZFSFDFOF –––––––
  • Changes no flag

Forms

Operands

reg
The destination register. Its width is selected independently of the source width; widening uses a larger destination.
r/m
The source register or memory value, whose width may differ from the destination’s. For example, a byte source can be extended to a 64-bit register. High-byte registers such as ah cannot be used with a REX prefix.
acc
The accumulator matching the operation’s size: al, ax, eax or rax. A shorter encoding may be available.

For example, movsx rax, byte [rdi] turns a stored 0xff into rax = -1, and movsx rax, ecx extends a signed 32-bit value to 64 bits.

For widening moves, the source width is independent of the destination width: byte, word and dword memory sources read 8, 16 and 32 bits. NASM also accepts equal-width and narrowing spellings. movsx ax, bx copies 16 bits; movsx ax, ebx retains the low 16 bits. These do not create a wider signed value. A 32-bit destination still clears the upper half of its 64-bit register.

global _start

section .text
_start:
        mov rax, 12
        mov rbx, 5
        mov rcx, 3
        mov rsi, 9

        movsx rax, bl

        mov rax, 60         ; syscall 60: exit
        xor rdi, rdi
        syscall