However, they violate the general rule that an 805x has only 256 bytes of IRAM, given that IRAM is NOT the same as XRAM contained on-chip. Frankly, I've got no idea how those variants that claim to have 512 bytes of IRAM address the extra 256 bytes of IRAM, nor do I know of how the mechanism works.

Many, well, quite a few, variants have XRAM on-chip, but that doesn't violate the basic rules, since XRAM is still addressed as XRAM. Internal RAM, however, namely that addressed indirectly using R0..R7, or directly addressed, doesn't fit that model.

I've yet to see an MCU that allows external devices to take DMA to internal XRAM, but that would be really useful, provided, of course, that there's enough of it. This week, I'd really like an MCU that has 256KB of on-chip XRAM that I can access from an external device, say, during the first half of the ALE cycle, when we know the addressed aren't yet valid. If only there were a way ... <sigh> ... but then, that's what gives rise to such wierdneses, isn't it? Sadly, they haven't yet put enough XRAM on-chip yet.

Strict adherence to the Harvard Architecture doesn't permit writes of the code space, nor does it permit execution of code residing in data space.

RE