i feel like you have cheated me a little here since the area to fill is only 24 words x 380 lines x 2 planes = 18240, or 54720 fill cycles + 182400 copy cycles = 237120 in total. So i get your point but it is a closer call than you thought.
Also one need not necessarily fill the entirety of the bounding box for both planes (for instance one plane only contains faces A and B while the other only contains B and C) and it saves having to draw some lines twice. But as you say it depends on the object, a cube is a simple example but imagine an object that has many triangles in the same colour, like a 2 or 3 colour boing ball.
You always have to calculate which polygons are behind the viewer, you can't draw those in any case. There are also many calculations that you have to do per-polygon whether they are visible in the final scene or not. But if you are not doing hidden surface removal, how do you order the polygons? Simple Z-sorting is not enough for something like an L-shaped object, for example. Whereas with hidden surface removal one can often use a fixed drawing order.
When polygons are very distant it might be quicker to fill them than to do a couple of multiplies required for testing visibility, but we have to do some multiplies anyway to calculate the screen co-ordinates of all the points.
EDIT: also consider the fill operation includes an idle cycle, which can be used by CPU.
Last edited by Mrs Beanbag; 01 March 2015 at 22:04.