[openib-general] basic IB doubt

[Caitlin Bestler]

openib-general-bounces at openib.org wrote:
> On 8/24/06, Woodruff, Robert J <robert.j.woodruff at intel.com> wrote:
>> If the feature gives them a huge advantage in performance (and it
>> does) and all of the hardware vendors that they deal with already
>> implement it, then yes, they will force, by defacto standard that all
>> other newcomers implement it or face the fact that no one will buy
>> their hardware. It seems like that is what is happening in this case.
> 
> Actually, if a hardware implementation provided the same
> performance (in this case latency) by polling on a CQ as one
> where polling on memory was garanteed to work, the customer
> may actually prefer the "standard" implementation.
> 

Exactly. The correct solution is to rely on a CQE to signal
a completion. That is why the standards are written the way
they are.

For iWARP there are network performance reasons why in-order
memory writes will never be guaranteed.

Now any time an *application* wants to write something that 
is vendor dependent then it is really up to that application
developer to decide if the benefit is worthwhile. And in my
opinion the "in order write" solution is not a feature, it 
is a work around to a slow CQ. So I certainly would not want
to encourage that workaround, but applications are always
free to do device or vendor specific workaround in their
application code. It's their code.

But I would hope that we would agree that no code that is
part of the project should rely on this "feature".