Rectangles (for example) are currently calculating 4 axes for separating axis detection.
Rectangles should only return two axes since the parallel axes are redundant and need not be calculated.
Whatever code we use to optimize for parallel axes should probably also detect arbitrarily parallel axes like a rhombus or trapezoid polygon or even 5+ sided polygons that happen to have parallel sides.
If parallel detection is faster than separating axis detection it should be an performance optimization.
Rectangles (for example) are currently calculating 4 axes for separating axis detection.
Rectangles should only return two axes since the parallel axes are redundant and need not be calculated.
Whatever code we use to optimize for parallel axes should probably also detect arbitrarily parallel axes like a rhombus or trapezoid polygon or even 5+ sided polygons that happen to have parallel sides.
If parallel detection is faster than separating axis detection it should be an performance optimization.