use64
org 0x40000000000
xor eax,eax
virtual at 0
abc:
end virtual
mov rax,[gs:abc]
    

test.asm [7]:
mov rax,[gs:abc]
error: value out of range.
    

mov rax,[gs:qword abc]    

By default fasm uses RIP relative addressing, and you can't reach 0x0 from 0x40000000000 with a 31 bit offset. To use absolute addressing you can use this:

Code:

mov rax,[gs:qword abc]    

By default fasm uses RIP relative addressing, and you can't reach 0x0 from 0x40000000000 with a 31 bit offset. To use absolute addressing you can use this:

Code:

mov rax,[gs:qword abc]    

but this type of code generate more binary byte, ...

... a better way to addressing cpu self data struct?

funggong wrote:

but this type of code generate more binary byte, ...

Than what? Even so, is byte count really a problem for your application? Are you in a memory constrained environment? Do you have cache thrashing problems?

funggong wrote:

... a better way to addressing cpu self data struct?

"better" depends upon what you are doing. It would probably be "better" IMO to firstly get the code working, and only later worry abut things like byte counts and performance problems once you have identified where any problems actually occur.