I was wondering how people handle the pure pixel core of their ppus. Since the instructions within this core are executing more than 3,500 times a millisecond, it is imperative they run as fast as possible. I imagine the fastest one could hope to do their pixel writing would be:

get 1st pixel color from the tables and shift it left by one
get 2nd pixel color from the tables and add it to the 1st
add the total to the color for the 8x8 pixel square the pixel is in
Get the NES color by going into memory and taking Memory[$3F00 + Color]
Index this color into an RGB Table
Set a screen pixel buffer to the Integer RGB Table value

I was wondering if anyone had any tricks or such they use to speed up one of these steps significantly, with assembly or otherwise. Is it possible to do these at any time before rendering of the pixels to speed up the drawing?

Thanks guys. :)