On 12/15/07, James Richard Tyrer <[EMAIL PROTECTED]> wrote:

> Are we really going to have that kind of hardware available?

I appreciate your enthusiasm.  However, one thing you're leaving out
of this is that the OGA architecture we have spec'd out is well
understood both to implement and to use.   Given our current level of
funding, we don't have the manpower to design and implement a new
architecture.  And don't forget that a more flexible architecture
requires significant software support to compile shader programs, etc.
 Regardless of whether or not a programmable architecture performs
less well or requires more hardware, we would not be able to support
it.  The fixed pipeline has been around the block a few times and will
be relatively simple to implement drivers for.  Also, OGA1 has a very
clear direction to go in, while designing a programmable GPU would
involve months of debate.  Even if programmable required _less_
hardware than fixed (which I believe is not true), I would choose
fixed for the above reasons.  [If some of you guys develop a
programmable engine and all the surrounding software before we get
around to OGA1, we'll use your design.]

In any case, who knows what's going to happen.  Speculating about GPU
pipelines is a distraction at the moment, because we really need to
get OGD1 and its baseline code finished.

And are we sure that a graphics chip is the best thing to do after
OGD1?  Maybe we should do something with shorter time to market.  This
stuff costs money to make, and we need a diverse product base to keep
the revenue up.

Endless debate does not a successful project make.  To quote Linus
Torvalds who was quoting Voltaire, perfect is the enemy of good.  Pick
something that will work and do it.  Everyone's going to want that one
extra "little" feature, but to add them all would be intractable.
Instead, we need to hit that 80% mark where everyone agrees that the
absolutely most critical features are in there and that a usable and
reasonably performing system could be made out of it.  (This is a
multicriterial optimization problem, but the optimizer has to be quick
about it.)  I don't want to squelch your creativity.  I'm a Ph.D.
student studying AI and cognitive science.  Loads to debate about in
those areas, and I really get into it.  But being too exploratory
won't work for us here.

-- 
Timothy Normand Miller
http://www.cse.ohio-state.edu/~millerti
Open Graphics Project
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)

Reply via email to