Re: New CBM-II discoveries

smf wrote:

> If someone wanted to port a different OS (DOSPLUS for example) without
> forcing to use a different ROM 

You don't have to force anyone into using a specific ROM. What my PC
emulation software does (if it's not resident in the ROM, or the ROM
contains and older version), is it loads itself into the highest portion
of the RAM and then lies to the underlying OS that the computer has a
little less memory, so that the OS does not overwrite it.

> and not wanting to use the exact same
> mechanism as the current ROM, they may try to access the 6525 directly.

But again - why? What's wrong with the IPC mechanism in the current ROM?
It implements the data transfer between two CPUs, and does it well
enough. Also, the communication protocol is baked into the KERNAL so if
you want to change it, you would also have to rewrite the relevant
portion of the KERNAL and fit it somehow into the small amount of RAM
that you have in bank 15.

Supposedly though, you came up with a better IPC protocol that allows
you to exchange bytes 5% faster. So you want to program the chip
directly. This would be akin to bypassing the KERNAL's IEEE routines and
writing your own fastloader for the IEEE drive. If you do that, you will
need to take into account that you _will_ need different fastloaders for
the 8050, for the 2031 and so on. Similarly, if you really want to
program the bare metal on the card for some reason, you would have to
take into account that there can be either a 6523 or 8255 there. Not a
big deal.

> Maybe the 8088 card is a lost cause and there has been an unfixable
> fracture, I don't see that copying that over to a graphics card is a
> good idea. It's likely to be even more useful to talk directly to the
> card, rather than going through a ROM interface that might not be optimal.

There are of course differences between the graphic card and the 8088 card:

* In the graphic card, the 6523 is an essential part of accessing the
video memory. Whatever you want to do with the card, you have to
interact with the 6523 directly. Whereas in the 8088 card, the 6523 is
relegated to a secondary role of being the middleman between the two
CPUs, and the normal software running on either CPU has no need to
interact with it directly.

* The software for the graphic card does not (yet) exist. Whereas the
software for the 8088 card does exist, in the form of the card's BIOS
which hides all the IPC complexity from the user. And because this BIOS
is tied to a specific card anyway, you can hide all the differences in
the IPC hardware there and all the other software does not need to know
anything about them.

Regards,
Michau