Changes to "The plan"

Changes to "The plan"

[akuster]

This is an addition to Paulus' plan back in November regarding
/platforms restructure with a change for 4xx I am thinking of for 2.5
  "http://lists.linuxppc.org/linuxppc-dev/200111/msg00083.html"


Paulus' plan mentioned a possibility of adding subdirectories within
/platforms if the number of files for a given platform was large enough
and I think 4xx qualifies.  I have a local copy working with a
/platforms/4xx structure were I moved 49 files to.  I am not sure if
this was the intent of "The plan" or not but it does make things look
cleaner.

Since I was in the mode of cleaning house I moved most of the 4xx config
options to /platforms/4xx/Config.in and have arch/ppc/config.in source it.

Also I think there  may a few other files that may need to move to/from
platforms/4xx to/from /kernel such as ibm405.h or ibm_ocp.h

so platforms/4xx looks like this:

Config.in
Makefile
ash.c
ash.h
beech.c
beech.h
ceder.c
ceder.h
cpci405.c
cpci405.h
ebony.c
ebony.h
ep405.c
ep405.h
ibm405.h
ibm405gp.c
ibm405gp.h
ibm405lp.c
ibm405lp.h
ibm440gp.c
ibm440gp.h
ibm_ocp.h
ibmnp405h.c
ibmnp405h.h
ibmnp405l.c
ibmnp405l.h
ibmnp4gs.c
ibmnp4gs.h
ibmstb3.c
ibmstb3.h
ibmstb4.c
ibmstb4.h
ibmstbx25.c
ibmstbx25.h
oak.h
oak_setup.c
oak_setup.h
rainier.c
rainier.h
redwood.c
redwood.h
redwood5.c
redwood5.h
redwood6.c
redwood6.h
walnut.c
walnut.h
xilinx_ml300.c
xilinx_ml300.h
xilinx_ocp/ -> 19 files
xilinx_ocp.h

comments, feedback or suggestions

armin


** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

Re: Changes to "The plan"

[Ralph Blach]

Armin

It has always been my contention the the 4xx should be organized in the
following structure
Core
Chip
board

This would tree would then reflect the reallity of chip design methodology.
the 405 now has many variants
so it would be
405 Core
   Chip
      405CR
      405GP
      Board
                 Walnut
                  .
                  .
                  .
      NPE405L
       .
       .
       .


Chip


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Changes to "The plan"

[Dan Malek]

Ralph Blach wrote:

> It has always been my contention the the 4xx should be organized in the
> following structure

I know, and like I have said in the past that doesn't fit into the rest
of the PowerPC strucuture very well.  We have processor parts and we have
boards (or platforms), that we can mix and match in a variety of ways.
There are other board vendors that want the exact opposite, since they
have a single board that can have a variety of different processors.
We currently separate the processors from the boards, so either of the
views will work.  The configuration scripts determine how we "view"
the underlying structure.  Currently, you select a processor type, and
then we provide a board selection to go with that, which is exactly
what you are asking for.


> This would tree would then reflect the reallity of chip design methodology.

The reality of chip design is they are becoming more integrated, things that
used to be on a board are now on a chip.  Logically no different to a
system configurator.  Our current configuration method simply asks for a
board type to reduce the number of configuration questions to be answered.
We could just as easily have a configurator that asks lots of questions
about the peripherals that need to be supported, and the underlying
directory structure wouldn't have to change.


> the 405 now has many variants

Doesn't matter to us, we can deal with that today.  There are lots
of variants of other PowerPC chips as well, and we found ways to
make them look _less_ different, requiring fewer configuration options
and a less complex directory structure.  You may want to consider
doing the same.  Please don't confuse the marketing enjoyment of
lots of variants with the software necessary to support them.

Thanks.


	-- Dan


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Changes to "The plan"

[Paul Mackerras]

Ralph Blach writes:

> It has always been my contention the the 4xx should be organized in the
> following structure
> Core
> Chip
> board
>
> This would tree would then reflect the reallity of chip design methodology.
> the 405 now has many variants
> so it would be
> 405 Core
>    Chip
>       405CR
>       405GP
>       Board
>                  Walnut

Wouldn't that make it more difficult to take advantage of the fact
that (for instance) the MAL and ENET are virtually identical on the
405GP and the 440GP?

Paul.

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Changes to "The plan"

[David Gibson]

On Tue, Jul 30, 2002 at 08:02:42AM -0400, Ralph Blach wrote:
>
> Armin
>
> It has always been my contention the the 4xx should be organized in the
> following structure
> Core
> Chip
> board
>
> This would tree would then reflect the reallity of chip design methodology.
> the 405 now has many variants
> so it would be
> 405 Core
>    Chip
>       405CR
>       405GP
>       Board
>                  Walnut
>                   .
>                   .
>                   .
>       NPE405L
>        .
>        .
>        .

It's not clear to me what you're proposing.  Are you talking about the
layout of the Config.in options, the layout of the .c files in various
directories or the call structure of the platform initialization code?

--
David Gibson			| For every complex problem there is a
david@gibson.dropbear.id.au	| solution which is simple, neat and
				| wrong.
http://www.ozlabs.org/people/dgibson

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Changes to "The plan"

[akuster]

David Gibson wrote:
> On Tue, Jul 30, 2002 at 08:02:42AM -0400, Ralph Blach wrote:
>
>>Armin
>>
>>It has always been my contention the the 4xx should be organized in the
>>following structure
>>Core
>>Chip
>>board
>>
>>This would tree would then reflect the reallity of chip design methodology.
>>the 405 now has many variants
>>so it would be
>>405 Core
>>   Chip
>>      405CR
>>      405GP
>>      Board
>>                 Walnut
>>                  .
>>                  .
>>                  .
>>      NPE405L
>>       .
>>       .
>>       .
>>
>
> It's not clear to me what you're proposing.  Are you talking about the
> layout of the Config.in options, the layout of the .c files in various
> directories or the call structure of the platform initialization code?
>
>

Well, the lastest 2.5 model working after having some local discusions is:

board files are in /platforms/4xx such as walnut.*, ebony.*

implimentation files are in /kernel/4xx such as ibm405gp.*, ibm44gp.*,
ppc4xx_* & ppc405_*, I guess head should go there too.

the core files like ibm4xx, ibm405.h & ibm440.h are in include/asm-ppc
as well as ibm_ocp.h

Things like ocp.c and ocp_proc will most likely end up in drivers/ocp
since ocp can support more than just ppc.

comments
armin


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/