> Dave Daniels wrote:
> > 
> >   IS> Provide memory priorities, as you can already, probably, and
> >   IS> ignore memory that is much too slow. e.g., on A3000 and
> >   IS> A4000 Z2 memory should be completely ignored ;-)
> > 
> > The problem on a machine like the RiscPC is that there is no
> > concept of fast and slow memory. You would have to make
> > assumptions about which area of memory is which. On the
> > other hand, given that the RiscPC is very much a minority machine
> > today it probably would not matter too much if you kludged it.
> 
> I believe that the kinetic card info was given to aleph 1, partly for
> the pc card stuff, but also for arm linux support, anyone considered
> asking them for the technical info?  In this more open acorn world they
> may well give it. I imagine there should be some way to detect that it's
> a kinetic and if there's not we'll have to use a boot option.
> 
> To kludge sounds wrong, rpc maybe a minority machine, but better it's
> still better to do it right if we can.  Linux is starting to catch on,
> but so might netbsd on the RPC, armlinux is a very dodgey hack to get
> the newest version (debian based one) installed, but aleph 1 are working
> on this.

Maybe (and it's only maybe -- I haven't looked at the source), it would be 
worth expanding the bootstrap header for the memory regions reported to 
add a "speed/priority" field.  Any kludging (read: built-in knowledge of 
the memory speeds of each segment) would then go in the bootstrap program 
rather than the main kernel.

R.