On 9/22/2019 2:55 PM, Justin Cordesman wrote: > > Forgive me if this is infeasible since I don’t know the architecture > well enough, but what about using the MMU’s privileged position to > watch for a specific peek or poke or both pattern and only then expose > addresses? Assuming that is doable, you could even have alternate > patterns that would expose the MMU at different addresses so that it > was run time selectable. > > Justin > > > On Sep 22, 2019, at 13:37, Mia Magnusson <mia_at_plea.se> wrote: > > > It is feasible, and planned. When powered on, the MMU will be hidden from the memory map. If the programmer reads a specific sequence of bytes in a specific area in succession, the registers will be mapped in. So, that's fine. But, after the registers are mapped in, where is the best place to put them that keeps them from "being in the programmer's road" all the time. Assuming you get writes under IO registers, the issue there is mirrored locations, so that's out. writes do belled though under ROM, so that would work with writes, but reads require you to map the ROMs out. Maybe not a horrible deal, but wanting comment on that. Using a space that has RAM alreayd present in the normal map (0-9fff, c000-cfff) works, but also could be a problem for some apps. If we go that route, it's best not to put the registers in the middle of a RAM bank ($7000 would be bad, for example, as it breaks up a long run of contiguous RAM). putting near top of c000 works, for example. Jim -- Jim Brain brain_at_jbrain.com www.jbrain.comReceived on 2020-05-29 22:40:34
Archive generated by hypermail 2.3.0.