CompuServe Thread

Rather than tiles, it may be easier to implement it as rendering horizontal "slices" (whole scanline groups), or perhaps interlacing the scanlines. If you have 10 machines and you're rendering an image of 2,000×1,000 resolution, each machine could render 100 scanlines, either in chunks (machine 1 renders scan lines 1 – 100, #2 renders lines 101 – 200, etc.) or interlacing (#1 renders scan lines 1, 11, 21, 31, 41, … 981, and 991, #2 renders lines 2, 22, 32, … 992, and so on). The former would probably be easier to implement and deal with, but the latter would be more likely to equitably partition out the rendering horsepower: an image that is mostly background at the top and bottom but has lots of objects with overlapping transparency and way slow .SXPs and .AXPs (perhaps written in TSE BASIC) and Ray-Traced Shadows and other rendering-time hogs in the middle would have machines 1, 2, 9, and 10 finishing up rather quickly, while machines 4, 5, and 6 might be working all night long if the job were partitioned in scanline chunks. Also doing any sort of intra-frame partitioning would probably wreak merry havoc with Video Post and .IXPs — look at the hoops they have to jump through now just to deal with Field Rendering!