[ogb-discuss] Creating a place for AMD-related work

[Garrett D'Amore]

Herman, George wrote:
> I can't speak for Intel, but AMD would have no issue in helping to
> enable features for Sparc, as long as that enablement also included AMD.
> We could learn a lot, together, if we did. (Isn't that in the spirit of
> open-source?) AMD just wants a level playing field that allows us to
> compete fairly on the benefits of our products.
>   

Yes, and I think most of the enabling features from the other vendors 
are advertised in such a way that the level playing field works pretty 
well... Intel and Sun at least have been pretty open about their recent 
hardware, or at least how to talk to that hardware from software. (There 
are a few exceptions, but in the general....)

(I could beat AMD about the head wrt ATI graphics right now, but I 
*hope* that situation will improve shortly...)

> It certainly would seem to be a problem if an engineer used NDA
> information to develop features for a competitor of that information.
> However, there should be no issue if that same engineer generated code
> to support each of the vendors' features using each of the applicable
> vendor's NDA material. (That is the way it works at most/all of our
> other ISV/OSVs.)
>   

At Sun the teams working on AMD features are largely independent of the 
engineers working on Intel features, which are also largely independent 
of the engineers working on SPARC. The high level abstract ideas (the 
idea of an IOMMU for example) is sufficiently broad that engineers *can* 
work on a generalized IOMMU framework, but the folks working on the 
details of a SPARC IOMMU probably shouldn't be working on the details of 
an AMD IOMMU. At least that is what I was given to understand by the Sun 
policy training we were asked to attend.

> BTW... Most of the engineers now assigned to working on Intel NDA
> enhancements were already exposed to AMD NDA information. If your
> understanding of Sun policy is correct... it would seem that Sun would
> have had to hire new engineers to work on Intel NDA features to comply
> to that policy.
>   

I think they can, just *not at the same time*. In the past Intel has had 
some very, very restrictive agreements about their technology NDAs, in 
some cases with multimonth or even multiyear restrictions on what 
exposed folks could work on. These days, things are much much better, as 
most of the information that Intel disseminates is done in open forums, 
with only a little NDA lead time, if any.

-- Garrett
> -George
>
> -----Original Message-----
> From: Garrett D'Amore [mailto:garrett at damore.org] 
> Sent: Thursday, October 18, 2007 8:31 AM
> To: Gavin Maltby
> Cc: James Carlson; Ostrovsky, Boris; ogb-discuss at opensolaris.org;
> Herman, George
> Subject: Re: [ogb-discuss] Creating a place for AMD-related work
>
> Gavin Maltby wrote:
>   
>> On 10/16/07 16:46, Garrett D'Amore wrote:
>>     
>>> Just a quick note, for those not in the know. Sun policy requires 
>>> (and for good reason!) that engineers with access to NDA materials 
>>> from either AMD or Intel work only on the platform (either AMD or 
>>> Intel) for which they got the NDA. Its too bad at some level, because
>>>       
>
>   
>>> there could 
>>>       
>> I thought it was more a question of funding: if your headcount is
>>     
> funded
>   
>> by the Intel deal then you shouldn't work on AMD-specific features,
>>     
> etc.
>   
>> Otherwise put: I know of a number of exceptions to the above
>>     
>
> There was an internal training about this, and no, its not just 
> funding.  Its legal CYA.  If someone writes a feature to support AMD 
> chips, using NDA information from Intel (and given the large overlap 
> between the product lines this is not as insane as it first sounds) it 
> leaves Sun open to a lawsuit from Intel.  And the reverse is true.
>
> And of course, neither Intel nor AMD want to enable SPARC design,
> either.
>
>     -- Garrett
>
>
>
>
>
>