At Fri, 2 May 2003 05:30:18 +0900,
Izumi Tsutsui wrote:
> Ok, then are there critical problem to move these CPU dependent code
> into arch/mips? Does it require any changes of API or some structures?
> Or it's just "a bunch of work"?

The information from arch/hpc/include/platid.h is widely used in
arch/hpcmips/vr/* and arch/hpdcmips/tx/*.  Any code under arch/mips
should not have any knowledge about platid.  It is too specific
information from the arch/mips's point of view.  So we cannot "just
move files from hpcmips/ to mips/".  Instead, we have to redesign how
to support for VR41xx/TX39xx.  As a result of the redesign, it might
require drastic changes of API and structures.

> > Definitely this.  arch/mips/nec and arch/mips/toshiba?
> 
> Currently hpcmips port has hpcmips/vr and hpcmips/tx,
> but I prefer mips/vr41 and mips/tx39 (or mips/vr41xx, mips/tx39xx)
> because both vendors have many other own mips processors,
> Vr4300,5000,7700 by NEC and TX49,79 by Toshiba:

"arch/mips/nec" is little bit too generic.

For example, VR41xx and VR5000 have many different point:

	- VR41xx is designed for PDA.
	  VR5000 is designed for workstation.
	- VR41xx has integrated I/O (UART, LCD, touch pad, audio, etc...)
	  VR5000 has not.
	- VR41xx has no secondary cache.
	  VR5000 has.

VR41xx and VR5000 are different CPUs :-)

--
Naoto Shimazaki