The instruction div
, divides the value (unsigned integer) stored in the pair of registers edx:eax
(dividend - 64 bits) by the target operator (can be a register or memory location - 32 bits), and stores the quotient in eax
and the rest in edx
.
As you are storing the value of op1
in edx
, and dividing edx:eax
by edx
(which is part of the dividend), the result of the division is probably greater than 32 bits and this generates the error Floating point exception (or divide error as Intel's manual ).
One possible solution to the problem is to zero the register
edx
and use a register other than
edx
and
eax
to store the divisor (if the division you are trying to do is 32 bits even and not 64 ).
Here is an example of your changed code, using the ebx
register to store the divisor:
Conta:
push %ebp
movl %esp, %ebp
movl op2, %ebx # armazena o divisor em ebx
xorl %edx, %edx # zera edx (parte alta do dividendo)
movl op1, %eax # armazena o dividendo (parte baixa) em eax
divl %ebx, %eax # divide edx:eax por ebx
movl %eax, resultado
movl %ebp, %esp
pop %ebp
ret
Example run: 13-by-3 split:
Conta () at teste.a:30
30 movl op2, %ebx
(gdb)
31 xorl %edx, %edx
(gdb)
32 movl op1, %eax
(gdb)
33 divl %ebx, %eax
(gdb) info registers
eax 0xd 13 <----- parte "baixa" do dividendo
ecx 0x0 0
edx 0x0 0 <----- parte "alta" do dividendo
ebx 0x3 3 <----- divisor
...
(gdb) s
34 movl %eax, resultado
(gdb) info registers
eax 0x4 4 <----- quociente
ecx 0x0 0
edx 0x1 1 <----- resto
ebx 0x3 3
...