It's definetely late to resurrect this discussion, but maybe the most tech-minded of you will appreciate this ;-)

No you are correct. But he does have a point in that if shadows overlapped with a lot of detail it would look strange. The shadows aren't of course, semi-transparent because you had to do a lot of wizardry to get tints and active transparencies in the Amiga, and it either took AGA chipset or master coder, or both.

Actually in TCE adding real transparencies would have been extremely easy. As some of you have noted, the in-game area (entirely separated from the status bar) has only 16 colors, which means that the BB only used 4 bitplanes. On OCS/ECS 5 planes were perfectly possible, so it would have been just a matter of adding one more plane, setting the palette so that the colors 16-31 were the darker versions of colors 0-15 and blitting the shadow masks on the 5th bitplane. Indeed, almost the same could have been possible even if the game used 5 planes, as OCS/ECS handled also 6 bitplanes: the only restriction, in that case, is that the colors 32-63 were forced to reflect the colors 0-31 with halved brightness (hence, the EHB = Extra Half Brite mode). So, now you'd be probably and rightfully wondering why the BB did not do go for such a simple solution: well, I guess that the problem was that just activating the 5th bitplane would steal more DMA time, making everything (occasionally) slower.

Look at Lionheart.

Uhm... are you thinking of the water in the first levels? If so, in that case the trick is much different: at a certain Y of the screen the Copper (a very simple 3-instruction coprocessor that had access to the chipset) was instructed to change the palette on the fly - that's basically the same mechanism that allowed to have so many hues for the sky (despite the background had only 8 colors) and the line-by-line parallax. This means that the same effect could not have been used to achieve shadows.


Besides this, somewhere in the thread somebody wonders about full whites/grays... well, I've always been coding for AGA so I don't really know, but, anyway, the AGA chipset for compatibility automatically copies what gets written in the upper 12 bits of color registers to the lower 12 bits, so I'd say that 0xFFF on OCS/ECS equals 0xFFFFFF on AGA (and not 0xF0F0F0).

One final thing: I don't think that the flat tiles were used for performance reasons, because that would have made the tile engine much more complicated for a very little gain; moreover, the gain would not have been constant as it would have depended on the graphics shown, meaning that in certain occasions the engine would have performed faster than in others, which certainly is not as desireable as a steady refresh rate (which, from what I've seen, is what TCE has). I agree with those that say that the flat colors and dithering were chosen to give a better idea of the different levels.

Congratulations to everybody for the nice analysis carried out here!

saimo