unsigned short to custom float 16位

Question

I have some big floats array, but I dont need all the 32bit precision and need save ram, so I use unsigned short type, since I only need 2 decimals an max dont excedd 654.

So the only thing I have to do for each ushort is to multiply with 0.01;

Is there a way I can create a type for this, using typedef or union, or struct, without wasting ram and save maximum cpu?

what I have now is this, not cpu friendly I guess

vector<unsigned short> vect(bigvalue);
for(int i=0;i<bigvalue;i++)(vect[i]*0.01)*some expression...

What I want is some type, wich return already multiplied by 0.01 not wasting ram

vector<float16bit> vect(bigvalue);
for(int i=0;i<bigvalue;i++)(vect[i])*some expression...

Is there a better way?

Thanks

Answer 1

You can create class (witch stores unsigned short number) with overloaded arithmetic operators (+ - * /, etc). These operators should consider factor of 100 while doind their job.

class FastFloat
{
    unsigned short n;

    ...

    FastFloat(const unsigned short n): n(n){}

    FastFloat operator+(const FastFloat &ff)
    {
        return FastFloat(this->n + ff.n);
    }

    // same for other operators
};