Image Scaling

/prog/, do you have fast and good image scaling algorithm in the form of:
void resample(uint8_t *Dst, int DstW, int DstH, uint8_t *Src, int SrcW, int SrcH);

???

>>5
What moron told you floats are slow? Multiplication requires a fixed point multiplier (just like integer multiplication) for the mantissa and runs in parallel with two int adders for the exponent.
r=(M_a*2^(E_a))*(M_b*2^(E_b))<=>r=M_a*M_b*2^(E_a+E_b-bias)
It can easily be done in one clock cycle.