While I was a bit working on some ASM for mr womp womp's and my project, I thought of a *good* way to have 1 ASM program, while still keeping the compability with the CSE and the CE. The main difference between that too, is the 2- or 3-byte registers. So I thought of a way to detect whether it is an CE or an CSE.
Code:

Code:

When running this on the CSE, it will be parsed as this:

Code:

Since 2+$FFFF > $FFFF, it will always jump to address XXXX. In the meantime, when running at the CE, it will look like this:

Code:

But since 2+$FFFF00 < $FFFFFF, this code will never jump to address XXXX00, so that address may be *random*. Now let's practice. Let's say I want to display HL where HL is equal to $1234.
The code for this is
Code:

When compiling for the CE, this would be
Code:
or something like that. For the CSE, it will look like this:

Code:

Now let's combine this into 1 program. Our frame was this:

Code:

Now let's expand this. Since the calc will jump when overflow, thus when executing from the CSE, the code after the jump (NOT the jump-location) should be for the CE. So it would look like this:

Code:

The last task is to figure out what the address of _Lbl_XXXX is. Just count the bytes of the code for the CE, and the bytes of the frame, and at that to _UserMem, or $0BA6, if I'm right. In my case, it would be this:
XXXX = $A623. Thus, my final code would look like this:

Code:

Now, let's final running this on the CSE, would look like this: http://www.rafb.me/results/w8XJmL37.html
And at the CE: http://www.rafb.me/results/M1izML20.html
Note that the stuff that can crash, will never happen.

Hope that this helps to develop cross-compatible games! 😁

I think of an easier manner:

Code:

And let's not forget the header is different too 🙂

Very interesting! This would not be too practical for many programs, but this could be used to create something like a universal Doors installer, where the user downloads one very large program which determines the calculator and installs the correct version. Again, not too practical for most programs as they would need to be optimized for two different processors with different instruction speeds, and any values loaded into address registers would need the upper bytes to be zero. This is essentially a way of either packaging two separate programs in one, or creating small utility programs that only need to differentiate slightly. Either way, nice concept!

Yes, this damned header remains the main trouble ☹️