I was just writing and evaluating my baby’s first comparison tests between generating a plane from the join of three PGA points versus a conventional Vector library which constructs a plane from two cross products and a dot from a point. When using the initial points on the standard plane equation to solve for 0, I found that the PGA representation has a small error (about xx.e-05) while the vector representation is purely zero.
I’m using the c++ implementation generated from ganja so they’re all just floats. All of the inital point values fall between the range [-10.f, 10.f]. I’m wondering if anyone else has encountered this or if there are known solutions to get comparable errors with traditional solutions.