I was looking at the PPU patent, US patent 5,125,671. I noticed that in Fig. 2 of that document, they consider the horizontal blanking period to be at the beginning of a scanline instead of at the end. I realize this is just anohter way of looking at how things work, but this way might be less confusing. Under this scheme, there no longer is "prefetching" of data for the next scanline. Instead this fetching is done during the HBLANK of the current scanline, which happens before the render. This seems like it could reduce some of the confusion involved with the pixel pipeline. Although I am uncertain how they account for scanline 1 not having an HBLANK period. Anyway, seemed interesting and so I thought I would throw it out here.