x264's SEA exhaustive search

[Sagekilla]

From what I read, ESA (SEA) in x264 losslessly removes candidates from even showing up for the SAD search. Is there any way the other ME algos could take advantage of this, or is something like this already in place?

Just curious :)

[akupenguin]

The iterative searches lossily remove candidates. Although there may be a few candidates that would be removed by SEA (due to their DC) that aren't removed by UMH (due to their position), in general UMH is much more aggressive than SEA at removing candidates without even looking at them. Which is precisely why it's, y'know, faster than ESA. Furthermore, SEA uses SIMD, but UMH's candidates aren't consecutive, so it would have to fall back to slower scalar math.