The really old TI-83's actually used a physical Z80, coupled with some fancy memory mapping hardware to let it access all of it's "ROM". While this doesn't really apply to our example since the memory isn't Flash, the concept is similar. TI just found cheaper ways to implement these schemes. They kinda ... stuck with us.

I figure that this information is only valid for your good ol' Ti-73's, TI-83+/84+(SE)s and perhaps a select few other calculators. You'd have to ask around for how the Ti-89 calcs work, though I think they work in a similar way. The Nspire is completely different and ... well. I'm not the person to ask.

I don't know whether they used a physical Z80 or not, but the TI-85 and 86 also used similar paging schemes. The 85, IIRC, used paged ROM (RAM was just 32K and not paged), and the 86 used paged ROM and RAM. The 81, being the first TI graphing calculator and very primitive compared to its successors, had no paging at all—it was just straight 32K ROM + 8K RAM, IIRC.

I'm not for sure how the 68K calcs are mapped, but I believe that the whole memory map (RAM, ROM, and ports) is “flat” without paging since the address size is much larger. (At least, I didn't have to deal with bank switching when writing programs like Flash Archive Undelete, which scans the whole archive portion of flash, so that at least isn't bank-switched). Someone more knowledgeable would have to answer this for sure, though.

From what I remember, platforms like the Nspire and modern PCs tend to copy ROM to RAM because RAM is faster, but my understanding is that the Z80 and 68K calcs execute most code directly from flash ROM (though archived variables and programs are temporarily copied to RAM on the 89/92+/V200; probably the same with Z80s but I'm not really a Z80 person). I figure the processors in those calculators are not fast enough for the speed difference between RAM and ROM to have a significant impact on performance.

The TI-89 and other 68k calcs have 24-bit memory addressing, so there is no need for swapping any memory cause that's 16MB of space to use. The OS is run from the Flash ROM (I think the Flash is fast enough to handle it, since the processor runs at ~15MHz or less; I don't know the exact details but different models have different processor speeds). Normal programs are loaded into RAM. I have no idea what happens with applications.

I believe it pushes everything to a different point in RAM (near the end) and then pops the pogram to $9D95.  Not to be confused with the stack -- I just like using the words push and pop

This is why programs never overwrite each other -- TI did some good programming work on their calculators

Never heard of it, but sounds like something TI would pull jk

they might have a purpose such as wasting time so that the LCD won't garble images if you send them too quick or something

But even then, I think you can do it much easier (and efficiently) with some NOPs or loading the buffer methods

How many T states does it waste exactly?

The TI-89 and other 68k calcs have 24-bit memory addressing, so there is no need for swapping any memory cause that's 16MB of space to use. The OS is run from the Flash ROM (I think the Flash is fast enough to handle it, since the processor runs at ~15MHz or less; I don't know the exact details but different models have different processor speeds). Normal programs are loaded into RAM. I have no idea what happens with applications.

Hmm I was sure the 89T had 2.7 MB of flash or something O.o. Does it really have that much hidden memory?