Re: as/400 / linux / lpar

On Fri, 31 May 2002, Leif Svalgaard wrote:

> Seriously, none of this, not faster floating point Vern, are needed
> for fractional LPARs. But, as long as nobody will tell us the real
> truth behind this, this whole discussion is at best amusing.
> Where did I hear years ago, that "interactive feature" was
> hardware? even came on or needed a special "decelerator"
> hardware card...  :-)

The only way I can figure that linux running in an LPAR would really
require special hardware is if that special hardware enables some fraction
of the CPU to work independantly of some other fraction of the CPU, and
that fraction can only access some fraction of main memory, and requests
to the various points of i/o can be tagged as to which part of the CPU
requested the i/o.  You wouldn't want your os/400 to write to your linux
parts of the disk or vice versa.  You certainly wouldn't want one os to
change some memory that the other needs to access.

A special CPU isn't going to enable all this.  You need special everything
else, too.  That's why this stuff is done in software with a host system
and a guest system.  If the hardware really could be special enough that
it allowed linux to run on bare metal then you could just turn of the
os/400 part of it.  But (as I understand it) you can't do that.

James Rich
james@eaerich.com