This is a follow up from thread Conversion from ASCII to Floating Point.

As some in the group (me included) are curious about what performance gain you can have using the math coprocessor on the infineon C517 I have done the following little test with the Raisonance compiler.

Here is the code (basic operations on long variables):

long a,b,c,d;
void main()
{
a=120;
b=310;
c=876488777;
d=3495;

a=b+c;
c=a/d;
d=a*b;
b=c-d;
}

without using the coprocessor: size=627, speed=3094
with the coprocessor: size=567, speed=1173

The gain in code size may somewhat be due to the fact that I had to turn off optimizations to get meaningful numbers from such a simple program, but the gain in execution time (expressed in milliseconds) is very relevant and should remain more or less constant vs program size (provided that the program does a lot of math).

Regards,

Luca