>If anything level1 should be $3F3F3F3F3F3F3F3F.

You're right. Also, the one that says $8080... should actually be all FF's (At first, I thought this instruction did unsigned comparisons, which is why I wrote it the way I did. Later I realized that it was a signed comparison, but forgot to update the code).


> In addition, I believe pcmpgtpb is the older version of the instruction pcmpgtb, since pcmpgtpb does not compile. Weirdly enough, intel's homepage has pcmpgtpb, but directly below it it uses pcmpgtb, as though they are exactly the same. Most likely they are.

As far as I know, there are only 3 compare for greater than instructions, one for bytes, words, and doubles.


>The overhead of dealing with 8 bit clipping and fine horizontal scrolling complicates matters more

For clipping only PF, it's probably easiest just to render the transparency color overtop the PF margin before rendering OBJs.
For clipping PF+OBJ, follow the same step as mentioned, except do it after OBJs are updated.
For clipping only OBJ, simply flag all the bits in each pixel in the PF margin so that they're marked as OBJ pixels. Then when you go to draw OBJs, your OBJ processing algorithm will avoid drawing any OBJs there, because it thinks that OBJs are already there.


>After implementing I greatly increased the speed of my ppu (speed increased by about 25%, with all thing implemented), thanks for your tips!

I'm glad to hear that it was successful.