Taylor series expotential function in assembly

Question

I am trying to implement expotential function algorithm in assembly according to this equation: http://upload.wikimedia.org/math/4/5/9/4597c1e758b3aeb83adcb03d3f75d00e.png

My code in assembly:

.data
x:
    .space 8
result:
    .space 8
counter:
    .space 8
factorial:
    .space 8
n:
    .space 8
xn:
    .space 8
.text
.global expot
expot:
    pushl   %ebp
    movl    %esp, %ebp
    flds    8(%ebp)     # wczytaj x
    fstps   x
    flds    12(%ebp)    # wczytaj precyzje
    fstps   n
    fldz    
    fstps   result
    fld1
    fstps   factorial
    fld1
    fstps   counter
    fld1
    fstps   xn

loop:
    flds    factorial
    flds    xn
    fdiv    %st(1)
    fldl    result
    fadd    %st(1), %st(0)
    fstps   result

    flds    counter
    fcom    n
je end

    flds    xn
    fmul    x
    fstps   xn

    fld1
    flds    counter
    fadd    %st(1)
    fstps   counter

    flds    counter
    flds    factorial
    fmul    %st(1)
    fstps   factorial
jmp loop

end:
leave
ret

My code in C:

#include <stdio.h>

extern float expot (float x, float n);

int main(void)
{
    float x = expot (1, 4);
    printf ("%f\n", x);
    return 0;
}

My problem is that this function doesn't return anything. When I am debugging it with gdb I see that in second iteration, when it should calculate x^n, the result is -nan and then when it should calculate 2! it returns -nan again. I completly don't know what is wrong with this code.

Thanks for all your responses.

Answer 1

The FPU stack isn't endless. In your loop you push (fld) more values than you pop (fstp). So the stack becomes full and every new push produces a NaN.