Re: [openrisc] Re: OR1000 16 bit instruction set

["Marko Mlinar" <markom@opencores.org>]

Hi Andreas!

> Since my mails again seem to be lost without error, I'm resending...
I suspect the problem is on your side, since I (or anybody else I know) have
never
experienced any email loss to any oc list.

> If we would never implement OR16 and OR32 on the same core this would
> also change the design goal of OR16.  It wouldn't need to be designed to
> be compatible with the OR32 hardware and thus not necessarily be a
> subset of OR32's operations.
yes; The one of the major points for not switching and non compatibility
is power consumption. E.g. ARM's thumb adds additional pipeline stage
to decode thumb instruction into 32bit one.

> Which would make it a totally separate project, but still worthwile as a
> control CPU (no vector and FPU options like OR32).  That's where
> Altera's NIOS seems to be placed.
yes, but like Damjan said this is not such high priority, maybe it would
be better to finish or32 architecture first - there is still floating point
and
some other modules to do. We somehow agree that it is better to have full
support at start and then add application specific inprovements.

best wishes,

Marko


--
To unsubscribe from openrisc mailing list please visit http://www.opencores.org/mailinglists.shtml