An assembly program I'm writing will not compile with the y assembler, citing:
error: invalid size for operand 2
On the following line:
movsx rbx, dword [rsi+4*rcx] ; Copy double-word and sign extend.
However, I can't find any reason why dword should not work. I want to move a double-word (4 bytes) at address rsi+4*rcx into 8 byte register rbx. So there remain 32 bits to be "padded" after copying it into what is effectively ebx. If I change the size to byte, I get no error. But this is not what I want.
There is a question with a similar title here. However, the poster had forgotten to include any size operands whatsoever, and the answer to the question did not resolve my problem.
Edit: I've added the full program below in case the particular syntax I've copied here is not the culprit.
segment .data
a:
dd 1
dd 3
dd 0
dd 1
dd 7
dd 9
dd 5
dd 2
b:
dd 8
dd 3
dd 3
dd 9
dd 6
dd 4
dd 1
dd 1
p dq 0
segment .text
global main
main:
xor rax, rax ; Set sum to 0.
xor rcx, rcx ; Set counter to 0.
lea rsi, [a] ; Set source 1.
lea rdi, [b] ; Set source 2.
dot:
movsx rbx, dword [rsi+4*rcx] ; Copy in double-word.
movsx rdx, dword [rdi+4*rcx] ; Copy in other double-word.
imul rbx, rdx ; Multiply the two double-words.
add rax, rbx ; Sum product so far.
inc rcx
cmp rcx, 8
jz done
jmp dot
done:
mov [p], rax
xor rax, rax
ret
The assembler calls the desired instruction movsxd for some reason:
movsxd rbx, dword [rsi+4*rcx]
This should work.