On Fri, 11 Aug 2000, Jeff Smith wrote:

>The difficultly factor is high.  It is basically and Indigo2 with
>a R8000 -> SysAD torque converter (TCC) on it.  The R8K has an
>odd tlb (as soren pointed out), does not support compat KSEG space
>(so a 64 bit kernel is required), has very odd cache ops, a quirky
>graphics fifo, etc.

It's also pretty weird /looking/ with its seven pieces of silicon.

>It would be a great project to get up on NetBSD but maybe too hard.
>I have some of the docs, and a fairly good idea how it works since I
>did the IRIX port to the machine.

It would be nice for that "we support it and linux doesn't" factor.
Hehe.

>The path I would take is the same as we did at SGI.  First get Indigo2
>up on NetBSD.  Then run a 64-bit kernel on Indigo2, then ask the
>questions
>you just did  :-)

Indeed.  It's much more important to work on supporting all of the SGI
bus architectures and peripherals before we go adding new CPUs to the
mips port.  For now we have access to all of the existing mips CPU code
which should greatly decrease the amount of porting needed for sgimips.

>This was a fairly low selling machine, so not a lot of other people
>would probably have one.

In fact, the official SGI sanctioned linux/mips port doesn't plan to
support R8k's either.

Brandon D. Valentine
-- 
bandix at looksharp.net  |  bandix at structbio.vanderbilt.edu
"Truth suffers from too much analysis." -- Ancient Fremen Saying