G'Day all,
I'm at that stage in my physics engine where I need to get the Penetration Depth of two colliding polyhedra. So, I've written a GJK implementation based on Casey's geometric simplification (along with suggestions from Rory, etc, in this thread)
Now, I am using Jay's method of finding the penetration depth along the relative motion vector (as described here and here), but this fails when there is no relative velocity between the bodies. So, I am falling back on EPA.