Performance of Compound Shape vs Breakable Constraint

[wormszer]

Which will give better performance?

Given the same number of objects in a compound shape vs constrained together.
I am thinking the number of objects <100.

I am thinking compound shape will since it has less active objects in the world.
Given a decent implementation of the extra code to needed to break them apart, calculate breaking, insert new bodies etc.

But having breakable constraints seems simpler overall than having to manage the removal of shapes from the compound shape.

Any thoughts, experience or benchmarks anyone has?

Thanks