Search code examples
assemblyadditionx86-64att

addq with 64bit immediates?

I try to add some values in assembler (AMD AT&T syntax). But I have problems with big immediate values (> 32bit).

    .globl f
    .type f, @function
f:
    movq %rdi, %r10
    addq $0x0000000000000008, %r10 # no problem
    addq $0x0000000122446688, %r10 # error
    movq %r10, %rax
    ret

The marked line returns the following gcc-error:

Error: operand type mismatch for `add'

But addq should be able to handle this number:

0000000122446688(hex) = 4869875336(dec)

and log2(4869875336) ~ 33bit (which is much smaller than 64bit).

Solution

  • ADD doesn't take a 64 bit immediate, there is only a version that accepts a 32 bit sign extended one and your second constant doesn't fit that constraint. See the instruction set reference.

    Note: the q suffix specifies operation size, not the size of the immediate. It is a 64 bit addition, but the operand is encoded as 8 or 32 bit sign-extended immediate.