The "conflicts goals of avoiding unnecessary point-in-triangle tests and maintaining parallelism in algorithm implementation" seems to be reflected in the difference between the traditional rendering method(just divide the screen into squares and render only one for each time) and real-time rendering method which renders the whole canvas much more slowly. The latter seems to yield better light effect(just from my experience with 3D rendering) but I don't know how yet.
The "conflicts goals of avoiding unnecessary point-in-triangle tests and maintaining parallelism in algorithm implementation" seems to be reflected in the difference between the traditional rendering method(just divide the screen into squares and render only one for each time) and real-time rendering method which renders the whole canvas much more slowly. The latter seems to yield better light effect(just from my experience with 3D rendering) but I don't know how yet.