On Fri, Jan 14, 2000 at 09:28:02PM -0800, William J. Earl wrote:
> > In theory yes. In practice all firmware that I've seen so far seems to
> > be rather fragile and for some systems also very performance limited as
> > it's running from uncached memory (veeeerry slooow) or at times even
> > from 8-bit wide PROMs which is so slow that it blows every meassure for
> > slowness. Not to mention other problems. So this should really be
> > considered a better than nothing solution.
> > That being said, Ulf Carlsson has implemented a PROM console which is
> > in the CVS archive. Try it, I'm interested in reports.
> On some platforms the PROM code is actually copied to main memory
> and executed (cached) from there. If I remember correctly, O2 works that
> way. Indy is probably in the slow (possibly very slow) category.
Running from a PROM would explain why the Indy firmware crashes with
enabled L2 caches - the circuitry just wasn't designed to deal with the
type of access cycles (bursts?) as used by the l2 cache controller.
No major problem in practice.