If I have to translate short int v[5] = {1,2,3,4,5} into assembly code how can I do it? It is ok if I'm doing something like this:
Enter 16,0
Mov ebp-4, 1
Mov ebp-8, 2
Mov ebp-12, 3
Mov ebp-16, 4
Thanks.
1 Answer
You can do that, but as @Ped7g pointed out, you need qualifiers. v[0] is going to be 20 bytes lower in memory, so you'd have to write them in reverse order.
enter 20, 0
; At this point ESP is the pointer to v[]
mov dword [bp-20], 1
mov dword [bp-16], 2
mov dword [bp-12], 3
mov dword [bp-8], 4
mov dword [bp-4], 5
The reason you have to use dword's is because imagine the contest of stack as follows;
2F 3D 17 0A 41 FF 16 18 03 22 19 0D 01 F3 D1 0C 12 02 EE 4A
using byte qualifier for byte [bp-4], 5 would only change 2F. DWORD however changes all 4 bytes and then becomes
05 00 00 00
To save program space, you could also
push bp
mov bp, sp
push 5
push 4
push 3
push 2
push 1
There are caveats to this, but because the IA32 is structured, these will extend to 32 bits. Does use a lot less code though.
Declared outside a procedure, then it would become
v: dw 1, 2, 3, 4, 5
1 Comment
Ped7g
You are defining
v: dd 1, 2, 3, 4, 5 on the stack. It's possible to define shorts array in stack memory, even if the default push/pop operates with dwords, you can still even use push, like push 0x50004 push 0x30002 push 0x10000 to have word array at original esp-10 with 2B junk at esp-11 and esp-12 (or new esp+0, esp+1, and array is at esp+2 to esp+11). If you would check the godbolt links from my comments, the C compiler indeed will produce short array, in stack (but by using mov, not push).
[]around memory references (from the operand order it looks like Intel syntax). And each assembler has subtle differences, MASM:mov word ptr [ebp-4], 1, NASM:mov word [ebp-4], 1... and you can put the values 2 bytes apart, asshort intisint16_t(probably). Also if you read those values in memory, you will end with @ebp-16 address: 4, 3, 2, 1 ... The C source was 1, 2, 3, 4, 5 ... So you are very close, but it's not the same, you got principle right, now tune details. Try it in godbolt: godbolt.org/g/7woqUa.data. Andreea: adding "-fno-omit-frame-pointer" to the gcc options in the previous link will force it to useebpnotation, closer to your original source: godbolt.org/g/qv35Tp (if you are not experienced to spot how theespcan be exploited in the first variant, this one should be more readable to you -> but I strongly suggest to copy all the variants (including -O3 ones) into debugger, and single-step over them reasoning about their differences and how each works = will teach you a lot)