Implementing software division in DSP

[kirgizz]

l_sub(l_num,l_denom);

Hi,

could someone give me a tip or a pair of approximation formulas for implementing division in software? (-> using assembler)

About application:
I need to acquire the frequency of an input TTL-signal. I use a timer and measure period of the input signal. Then the measured value is divided by reference value (DSP clock). My design is using 32 bit fixed point data. Measured value and the divider are unsigned integers.
I think, my DSP has a ROM-based table of reciprocal values, but it's not accurate enough. I suppose, it can be implemented by shifting,addition and multiplication.
Does it???

Thank you in advance

 

[Akorostel]

Hi!

There are a lot of software division algoritms over the internet.
For example, try following references:
**broken link removed**
https://www.sccs.swarthmore.edu/users/03/jcorder/CS23/lab3/extensions.html

The examples are in C, but they will let you understand how to implement it in assembly, because the algorithms are quite easy.

Best regards!

 

[kirgizz]

Thank you Akorostel,

it was useful. I know the right direction now.

 

[Eric Lin]

See this c code.

Word32 L_div_s(Word32 var1, Word32 var2)
{
Word32 var_out = 0;
Word32 iteration;
Word32 L_num;
Word32 L_denom;

if ((var1 > var2) || (var1 < 0) || (var2 < 0))
{
printf("Division Error var1=%d var2=%d\n",var1,var2);
exit(0);
}

if (var2 == 0)
{
printf("Division by 0, Fatal error \n");
exit(0);
}

if (var1 == 0)
{
var_out = 0;
}
else
{
if (var1 == var2)
{
var_out = MAX_32;
}
else
{
L_num = var1;
L_denom = var2;

for(iteration=0;iteration<31;iteration++)
{
var_out <<=1;
L_num <<= 1;

if (L_num >= L_denom)
{
L_num = L_sub(L_num,L_denom);
var_out = L_add(var_out,1);
}
}
}
}

return(var_out);
}

Word32 L_sub(Word32 L_var1, Word32 L_var2)
{
Word32 L_var_out;

L_var_out = L_var1 - L_var2;

if (((L_var1 ^ L_var2) & MIN_32) != 0)
{
if ((L_var_out ^ L_var1) & MIN_32)
{
L_var_out = (L_var1 < 0L) ? MIN_32 : MAX_32;
Overflow = 1;
}
}
return(L_var_out);
}

Word32 L_add(Word32 L_var1, Word32 L_var2)
{
Word32 L_var_out;

L_var_out = L_var1 + L_var2;

if (((L_var1 ^ L_var2) & MIN_32) == 0)
{
if ((L_var_out ^ L_var1) & MIN_32)
{
L_var_out = (L_var1 < 0) ? MIN_32 : MAX_32;
Overflow = 1;
}
}
return(L_var_out);
}