x86-64汇编程序运行终端无输出及寄存器替换后失效的问题咨询

Hey there! Let's work through your two issues one by one—first figuring out why your assembly program isn't producing any output, then explaining why swapping rcx with rsp breaks everything.

Your code has a few critical bugs that are preventing it from working as expected:

• You're comparing variable addresses, not their values
  When you do mov rax, first_number, you're loading the memory address of the first_number variable into rax, not the value it stores. Same with mov rdi, second_number—you're comparing the fixed compile-time addresses of the two variables instead of their contents. If you wanted to compare numeric values, define them as numeric constants (e.g., first_number: dq 50 instead of dq '50') and load their values with mov rax, [first_number].

• Invalid memory access in the write syscall
  When you do mov rcx, '1', you're putting the ASCII value of '1' (0x31) into rcx. Then mov rsi, [rcx] tries to read from memory address 0x31—which is in a reserved kernel memory region your user-space program can't access. This either causes a crash or feeds garbage data to the write syscall, resulting in no visible output.
  Additionally, your write parameters are wrong: you're telling it to output 8 bytes (rdx=8), but you only want to print a single character. You should set rdx=1, and point rsi to a valid memory location containing your output character (like a variable in the .data section).

• Syntax errors from full-width quotes
  Your code uses Chinese full-width quotes (‘1’, ‘0‘) instead of English single quotes ('1', '0'). NASM won't recognize these characters and will throw compilation errors—make sure to use standard ASCII quotes.

Fixed Example Code #

Here's a corrected version that compares numeric values and prints the expected result:

section .data
first_number: dq 50       ; Store numeric value 50, not ASCII string
second_number: dq 40      ; Store numeric value 40
result_true: db '1', 0xA  ; '1' plus newline for clean output
result_false: db '0', 0xA ; '0' plus newline

section .text
global _start
_start:
    mov rax, [first_number]  ; Load actual value of first_number
    mov rdi, [second_number] ; Load actual value of second_number
    cmp rax, rdi
    jae .true
    jb .false

.true:
    mov rsi, result_true     ; Point to the '1' string
    mov rdx, 2               ; Length: '1' + newline = 2 bytes
    jmp .execution

.false:
    mov rsi, result_false    ; Point to the '0' string
    mov rdx, 2               ; Length: '0' + newline = 2 bytes
    jmp .execution

.execution:
    mov rax, 1               ; Syscall number for write
    mov rdi, 1               ; File descriptor (stdout)
    syscall                  ; Invoke write

    mov rax, 60              ; Syscall number for exit
    xor rdi, rdi             ; Exit code 0
    syscall                  ; Invoke exit

Compile and run it with your existing command:

nasm -felf64 example.asm -o example.o && ld example.o -o example && ./example

The rsp register is the stack pointer—a special register that tracks the top of your program's stack. The stack is used for critical operations like storing return addresses, preserving register values, and managing function calls.

• When you do mov rsp, '1', you're overwriting the stack pointer with the ASCII value 0x31. This completely breaks the stack's structure—your program can no longer handle stack operations correctly, which will almost certainly cause a crash before it reaches the write syscall.
• Even if the program didn't crash, mov rsi, [rsp] would still try to read from address 0x31, which is invalid user-space memory—repeating the same invalid access bug as with rcx.

Never overwrite rsp with arbitrary values unless you're explicitly managing stack space (e.g., allocating room for local variables with sub rsp, 16).

内容的提问来源于stack exchange，提问作者Ruslan