[Arm-netbook] FSF-Endorseable Processor

[luke.leighton]

On Wed, Dec 5, 2012 at 12:29 PM, jm <joem at martindale-electric.co.uk> wrote:
> On Wed, 2012-12-05 at 11:22 +0000, luke.leighton wrote:
>> On Wed, Dec 5, 2012 at 8:25 AM, Simon Kenyon <simon at koala.ie> wrote:
>> > On 12/05/12 08:20, luke.leighton wrote:
>> >>   so i really really *can't* say "it's open source" - it's too broad.
>> >
>> > i've got lost in all this. what licence does it use?
>>
>>  :)   the question itself shows the source of confusion.  hope i can clarify
>>
>>  * the processor is a proprietary design.  its design (netlists) are
>> *not* available under any type of open license or free software
>> license.  much as i would like them to be.
>>
>>  * the processor can run any binaries.  it is, after all, a
>> general-purpose processor.  the license of the software that runs on a
>> general-purpose processor has absolutely no bearing, relation,
>> influence or any correlation OF ANY KIND, and this processor will be
>> no exception to that general rule.
>>
>> so... why bother at all with this exercise?
>>
>>  * because there does not exist - and i really am not kidding about
>> this - there does not exist ANYWHERE in the world a low-power embedded
>> processor that is both desirable to mass-volume manufacturers (i.e.
>> can do 1080p30 at least and modern 3D graphics) *and* is at the same
>> time FSF-Endorseable according to the criteria listed on the FSF's web
>> site.
>>
>> * it just so happens that by _not_ having any proprietary hardware
>> acceleration macros (and associated proprietary unreliable software
>> libraries) that this makes the development process a hell of a lot
>> easier.  that's the business case.
>>
>> l.
>
> This is tough one Luke.
>
> Yes we all want a CPU that is FSF endorsed.

 to clarify: i want a processor that can go into *products* that the
FSF can endorse, where those products happen to be attractive in
mass-volume.

 [ that's *not* the same as having the design of the actual CPU i.e.
its netlists being available under free software licenses ]

> There are processors like the OpenRISC processor SoC that
> has its own assembler and has its own gcc and which compiles
> and runs Linux. http://opencores.org/shop,items

 yes.  it's not multi-core, it's not been proven in 65nm, it doesn't
have a GPU, we would need to _find_ a GPU, it doesn't have a VPU, we
would have to _find_ a VPU, or get the design improved to the point
where it can.

 the number of unknowns are simply too great - so it's out.  for now.

there are very very good reasons why i picked this very specific
company to work with.

> How is FSF going to endorse a proprietary CPU against the
> the fully open OpenRisc processor without going mad?

 by doing quite a bit of work - but no more work than has to be done
for any other FSF-Endorseable product.

> Isn't it easier to say the project closed source project but that
> its very supportive of open source more than anyone else.

 i'm not interested in saying "it's supportive of open source".  i'm
*specifically* interested in attracting the resources of the
volunteers of the FSF as well, and bringing them the option of being
able to touch base with real people in mass-volume markets.

 right now they don't have that option... at all, and are in danger of
becoming irrelevant as a result.  that's not something i like, so i'm
doing something about it.

> Aims to be like AllWinner but 10x better.

 the FSF is specifically excluded from being able to use the allwinner
processors [until lima and cedarx reverse-engineering are completed]

> If eliminates FSF and it avoids having to demonize FSF if
> they felt unable to endorse.

 i would not go to all the trouble of putting a processor and/or
products in front of them that would take up their time and they found
them to be non-endorseable.

 that would be bad :)

 the point of the exercise is to go *beyond* what's already been
achieved with the allwinner a10, and to bring in the FSF volunteers as
well.

 l.