Question

I have a 3-vector, let’s say

v = vec3(-4, 2, 3)

I would like to do a max on the absolute values of the components, so the equivalent of:

max(abs(v[0]), max(abs(v[1]), abs(v[2]))) == 4

However, I have a requirement that I need to preserve the sign. So for example:

magic_max(v[0], magic_max(v[1], v[2])) == -4.

It’s a trivial problem if I use conditional branching, but I’m trying to do this in as few operations as possible, and avoid branching. Any ideas on where to look? Maybe there’s some bit-shifting magic that can be done?

Answer 1

I would determ the max AND the min of all values, and then decide what is abs larger

ma = max(v[0], max(v[1], v[2]));
mi = min(v[0], min(v[1], v[2]));

res = abs(mi) > ma ? mi : ma;

If you want to get the sign, replace the last line with an if

if (abs(mi) > ma) {
  sign = -1;
  res = mi;
} else {
  sign = +1;
  res = ma;
}

However, what should happen on (0, 0, 0)? no sign?