Hi,
Yesterday I have watched a video related with kinetis introduction on freescale web page.
There, they said floating point multiplication is 1 cycle, and floating point division is 3 cycles (something like that, I couldn’t remember the exact values now)
As I know, floating point multiplication and division should be same. Aren’t they? I mean dividing a value into 4 is same operation with multiplying it with 0.25? Why are their cycles different?
Thanks.
Solved! Go to Solution.
No, FP multiplication and division aren't the same. You can replace /4 by *0.25, but you can do this only for known constants. You (and not your MCU) calculate reciprocal of divider and use multiplication instead of division. But who will calculate reciprocal of variable to replace division with multiplication? Also, C compiler is not allowed to optimize any FP operation, like replacing division by constant with multiplication by constant, since these can produce not exactly the same results.
No, FP multiplication and division aren't the same. You can replace /4 by *0.25, but you can do this only for known constants. You (and not your MCU) calculate reciprocal of divider and use multiplication instead of division. But who will calculate reciprocal of variable to replace division with multiplication? Also, C compiler is not allowed to optimize any FP operation, like replacing division by constant with multiplication by constant, since these can produce not exactly the same results.
Yes you are right, thanks