ive decided to drop the 32 bit mcu atm becuase it doesn't have a full (8bit) io port,  using two atmega 8515 @ 6 times the tia clock (21.47727 mhz) one used as a gameboy like MBC (you can bankswitch[32k ram/flash/gpio etc] and by writing to the data bus, using the upper address bits of the 6502, could be used as a co-processor aswell imho) , the other soley used to drive the TIA, with it having its own 32k of  "display" ram,  and would be accessed also with the extra address bits from the 6502, this chip will function similar to an antic cpu, to offset the display to the atmega, and do your logic with the 6502 cpu