Okay, we don't really need to know the hardware, per SE.

For a TI emulator, you don't need ti83plus.inc. It's very abstract in it's implementation. An emulator for TI calcs in essence emulates things like ports and such, not memory and bcalls.  The OS handles all bcalls and such on it's own with it's own code.  The code is gotten from a ti-calc based on a ROM.

For a prizm emulator, we need to know screen size, and low-level how it talks to it.  That means we need some sort of disassembly of the OS so we can analyze that stuff.

I'm pretty sure we could run it with just the OS.  I don't know how picky Casio is with roms, but I'm sure it will be alright for personal use or something like this. Now, we need to be careful to see any signs of no-hacking policies. I don't like messing with stuff where it could be legally bad, and wouldn't write an emulator at all in that case.  Does anyone know of a disassembler for the SH3? Or at least an instruction set somewhere.

Necro, 'cause I never saw this...

Since the last posts, the situation has changed somewhat. Yes, there are multiple disassemblers available and the OS can be probably run without a ROM image. The hardware documentation is available in the previous post.

I have a feeling that pretty soon we will be capable of building a full Prizm emulator. Even though we haven't fully disassembled the OS yet and the ports are still unknown, an emulator can at least be started. The first thing that can be done is the SH3 emulation core. With this being already fully documented, emulating the instructions is possible and I was actually going to start writing the insructions soon too. This will be in Java for several reasons that I will explain later. Next thing to be emulated will have to be the memory. Now for Ti-83+ emulators you could just make a 64 kb array and call that memory (ignoring flash and ram pages at the moment). Now with a 32 bit proc this is a lot harder. So if anyone is good with writing memory management routines in java I could use some help. Then there is also the flash to be emulated which I believe could just be stored in a file and then accsessed from there. I'm not yet sure if the flash is set up in pages or if it just addresed similar to the ram.

In other hardware the keyboard shouldn't be that hard and the screen will be easy once we find where it is mapped to. usb could be hard to work with, but that shouldn't be nessecary in the initial version of the software. There is also the serial port, but it acts diffrently than the 83+ one because it has input and output buffers. The last hardware I can think of is the clock and that should not be hard to emulate either.

Now I want to go over some project details. Even though I'm mainly a C++ programmer, I wanted to do this in java, one because of portability. Two because my idea was to make each of the hardware devices its own thread so they can operate simultaneously. (this feature being very easy to implement in java) I also wanted to add full debugging software just like in wabbit. This project is also quite ambitious so I will need lots of help from everyone. Once I get started on this I will post updates and also send out help requests. For the past few weeks I have been very busy (especially because I have a job now), but I will have more free time in the near future. Lastly my current priority is to get that app signer out, to make Prizm programming much easier. It's just been I've just been very busy as I previously stated.

I can help if you need conceptual level help with any of the hardware processes or opcodes. I'm still working on all of the I/O, excluding keyboard (which I'm not even going to touch until I can at least use the screen), so that might be a bit too ambitious to emulate. But the CPU core is definitely possible. Just be careful with the arithmetic instructions and read the instruction descriptions. Some of them have very important but slightly subtle effects.

On a side note, the Renesas documentation has C code for all of the instructions, so that might be easier in C rather than Java.

Or I could just port the C codes to java

Well I already figured out a memory model which a will give a brief description of. How it will work is when an instruction is passed for a memory location, the emulator then converts the physical adress into a virtual one which can be stored into a dynamic array. So in order of execution, a program on the Prizm passes a virtual address to the OS, which then converts that into a physical address, then the emulator converts that into a virtual adress, then the jre converts that into a different virtual address, and then converts that into a physical address. Wow this makes my head hurt. 

I'm glad to hear an emu might be eventually doable. Good luck to whoever starts working on this.