Power (pow) returns incorrect result

[mnfisher]

Just spent a while chasing a bug in my code - but it turns out 'pow' returns an incorrect value.

It actually returns the result - 1 (except for pow(2, 0) and pow(2, 1))

The attached demonstrates this behaviour (on an Arduino - other devices may differ?):

2**0 = 1
2**1 = 2
2**2 = 3
2**3 = 7
2**4 = 15
2**5 = 31
2**6 = 63
2**7 = 127
2**8 = 255 // The example in the WIKI should = 256!

Martin

[mnfisher]

It might be because it uses floating point...

pow(2, flt_fromi(FCL_I));

- not sure why it doesn't also do flt_fromi(2)?

However - it shouldn't make it so wrong ?

Wrote my own integer

[Steve-Matrix]

Thanks, Martin.

I've just checked and confirmed this. But on an 8-bit PIC it's fine.

If I strip out the flt_* functions in C code, then there are some oddities:

Code: Select all

FCV_F = pow(2, 4);   //this produces 16.00000

//but...

FCV_I = 4;
FCV_F = pow(2, FCV_I);   //this produces 15.999999

FCV_I = pow(2, FCV_I);   //this produces 15

(FCV_F is a float and FCV_I is an int).

So there is a truncation issue.

Perhaps the "pow" function in the Arduino is using some algorithm that makes the power calculation quicker, but introduces this truncation issue.

We'll continue to investigate.

[mnfisher]

This might give a solution:

Code: Select all

#include <math.h>
#include <stdio.h>

int main()
{
    int a, b;

    // Using typecasting for
    // integer result
    a = (int)(pow(5, 2) + 1e-9);
    b = round(pow(5,2));
    printf("%d \n%d", a, b);

    return 0;
}

Which looks to give the correct result on both cases (from stackoverflow)

[Steve-Matrix]

More on this. It's not a bug as such. It's the way the pow() function works in the C code for Arduino/AVR.

The pow() function returns a float. And if a float is casted to an int then this is a truncation rather than a rounding. So 15.999999 becomes 15 and not 16.

So in your code it's probably wise to use the round() function if you are definitely expecting an integer rather than relying on C casting.

Code: Select all

myInt = round( pow(2, x) )