- Let's assume PCIe 1x (the answer to the alternatives is basically
the same). How are you going to connect that to a processing element?
I think there are really two ideas that are floating around. The
first is just as
you are suggesting - attaching the card locally to the processor and
using the
added cpu to decode video, 3D acceleration, etc.
The other is a remote box with a small, low power form factor. It
attaches via
ethernet and streams video, and possibly OpenGL or X.
- What kind of processing element are you going to use for basic
graphics stuff? (DSP? something else?)
While it would be nice to have a cpu (since AFAICT they are easier to
program) in
the media box, we will almost certainly need a DSP for the processing
power
it provides, either that or FPGA our own.
- What kind of processing is your primary focus? Graphics rendering?
2D? 3D? Video decoding?
Video decoding and possibly Xrender acceleration should be nice.
- How are you going to handle things like graphics memory? That is,
how is the processing element connected to the memory?
Many DSPs have built in memory controllers, or a small FPGA.
- How do you access your video framebuffer and scan it out to a
monitor, regardless of the format?
I have been thinking about this, since it essentially requires dual port
DRAM across multiple rows in the same bank. One way is to simply have
two banks in each DRAM duplicate the other two banks and use twice
the RAM,
which gives us free double buffering. There has to be a better way.
Scanning it out to the monitor is as simple as perfecting the video
output module
Simon and I wrote.
but it appears that some of you do not have much
hardware design experience.
Very true, sadly. See below.
Is it time to create an ogml-advocacy list that intended for rather
noisier discussions?
That seems reasonable to me. Then the ogml list is dedicated to
perfecting features
that have been decided on.
Also, before much more of this is discussed, I strongly recommend that
some people volunteer or be elected to join gEDA-user and become
proficient in PCB design. No design can be realized if you have no
idea how to realize it. The high-level dreaming is healthy, but if
you want to achieve something, you have to have a healthy dose of the
practical and keep in mind that every finished design is 99.9%
practicality.
I would like to do this. If we design the boards, you have access to
fabs, correct?
I think, as a start, we should decide on something to build. Maybe a
DSP and FPGA
interfaced with a few megs of DRAM. Then we can fab a few, give it to
some developers,
and get all the performance we possibly can out of it.
In terms of DSPs, a good choice is one of the TI
I am personally most interested in actually designing the board - I
missed that chance on ODG.
nicholas
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)
