During 0.1.x:

  - complete simulation of 8515 device.

  - external peripheral device connection interface.

  - write, read and access watchpoints

  - documentation

During 0.2.x:

  - add support for all at90s**** devices (Is this too ambitious?)

  - speed improvements (optimization)

During 0.3.x:

  - add support for megas and tiny's (This might require some redesign
    work to get the atmega128 like devices right)

Misc TODO's:

The conversion to autoconf/automake configuration is working, but still needs
to have the actual porting issues worked into the project. Also, the autoconf
macros (the config/*.m4 files) could be robustified. For instance, the
TROTH_PROG_AVR_CC macro really should do more than just checking if avr-gcc is
in yout $PATH. It should also make sure that avr-gcc works. Like wise for the
other macros. It's a start at least.

Build the virtual hardware harness and gui. I'm thinking it be able to plug a
device into something like the Atmel stk200 board thus giving the user LEDs
for visual feed back and push buttons for supplying inputs to the ports. Also,
I'd like to simulate the analog comparators and ADCs if possible. It might
also be interesting to simulate a popular LCD display.  More interesting might
be the potential for custom user designed virtual peripherals which could be
plugged into the virtual board. I wonder if I can use the same interface as
gpsim, thus allowing the sharing of modules between the two projects?

Make a python module out of the library.

Need to write functions for loading/dumping eeprom from/to files.

Add support for ihex, srec, elf, and whatever else input file formats.

The interrupt code looks like it needs a major re-thinking. It doesn't handle
the interrupt flag mechanism and doesn't support automatic vectoring into ISRs
that occurred when the interrupt occured and was disabled either globally or
by the specific control register and then the interrupt was subsequently
enabled. This looks like it might be yucky. I'm seriously thinking that
interrupts need to be handled via some asyncronous mechanism like unix
signals. Also, some way to implement the interrupt response time needs to be
thought out.

Add code to output number of instructions executed (and per second) and total
number of clocks (and per second). Should go in main.c.

There's a bunch of FIXME comments in the code in places where someone needs to
give more thought to the logic.

These io registers still need to be implemented:
  GIMSK
  GIFR
  All the timer/counter 1 stuff.
  Lots of others too.