How to work with negative numbers in assembler?

Question

assembler

Goal: Compute (2*a/b-1)/(a-28+c) in an assembler function called from a C++ file.
Problem: an assembler function from time to time perceives a certain number not as negative in the additional code or as it is there, but as positive.
This is how the parameters are declared in the C++ file:

extern "C"
{
    signed __int8 a = 0, b = 1, c = 29;
    signed __int8 numerator = 0;
    signed __int16 denominator = 0;
    signed __int8 result = 0;
    void ASM_count();
}

This is what the assembler file looks like:

;.586
;.model large, C
.data
extern a:sbyte 
extern b:sbyte
extern c:sbyte
extern numerator:sword
extern denominator:sbyte
extern result:sword
;(2*a/b-1)/(a-28+c)
.code
  ASM_count PROC
;///////////////////////////////numerator///////////////////////////////
xor eax, eax
mov al, 2
imul a
;cbw
idiv b
dec ax
mov numerator, ax

;///////////////////////////////denominator///////////////////////////////
xor eax, eax
mov al, a
sub al, 28
add al, c
mov denominator, al
;///////////////////////////////result///////////////////////////////
xor eax, eax
mov ax, numerator
idiv denominator
mov result, ax
ret
  ASM_count endp
end

This is how the problem manifests itself:
the numbers are calculated correctly, but they are displayed incorrectly.



Moreover, any value out of three can be displayed incorrectly.

Answer 1

signed __int8 numerator = 0;
signed __int16 denominator = 0;

extern numerator:sword
extern denominator:sbyte
...
mov denominator, al

Do you see that the types in C and in assembler do not match?
In a good way, you need to replace __int8 numeratorwith __int16 numerator, extern denominator:sbytewith extern denominator:swordand mov denominator, alwith mov denominator, ax, so that it becomes as it should.
Everything else looks good.