Try multiplication on R3 (especially radix sort)

[furkansimsekli]

We know that division operations are not handled at cpu level, instead the compiler calculates it by hand. It's an uneducated approach maybe, but I will try multiplying with 1/X to see how power consumption changes.

[div72]

Considering we sort numbers in [-300, 300], 13 * n >> 7 is a good enough approximation for n / 10.

[div72]

This works for the entire range of int on R3: (The loop will terminate in 5 steps max)

def div10(n):
    n >>= 1
    while True:
        cand = (n * 0x4ccd) & 0x7FFF
        if cand < 0x199a:
            return cand
        n -= 1