Timothy Miller wrote:
On 6/22/06, Terry Hancock <[EMAIL PROTECTED]> wrote:
> That would actually be pretty cool. It would keep X from competing
> with ordinary application tasks for resources, which is often a
> source of hang-ups or slow performance on typical single-CPU Linux
> desktop machines.
With the exception of some pathological things (like PolyPoint and
hitting against some of mi's O(n*n) or worse span rasterizing code),
most things can be quickly dumped into a DMA buffer and shipped to
the GPU while the X server sleeps. The reason for the high CPU usage
is drivers that try to render mostly using PIO. I've looked at the
Radeon driver, for instance, and it does everything but PutImage and
GetImage using PIO.
Our plan is to rely on DMA almost exclusively. With properly-written
drivers, this competition won't happen.
I don't think this is a stock OGC card that I'm imagining. It's more like
an embedded computer with OGC chipset for graphics (see other reply).
I'm just applying "separation of concerns" to making the system
more robust. With a separate CPU running X, there's less chance
of a crash in one application hanging X, or X hanging everything
else. ISTM that recovering from faults would be easier, and there'd
be fewer faults to begin with.
You can, of course, get this arrangement with an X terminal approach,
but it takes up a lot of desk space. An X terminal on a card might
have some uses.
Interesting thought -- could you run two or three of them? Obviously
you'd be constraining the terminals to be pretty close together, but
that'd be no problem in a computer-lab setting. This is basically
getting into competition with "thin clients" -- but unlike the
thin clients, it wouldn't require separate cases, nor would X have to
run through ethernet lines (communications would be on the video
bus (AGP, PCI, etc), which is very fast compared to any serial
ethernet link). IOW, it's probably cheaper than a thin client and would
have higher performance (except for the difference in production
runs, I suppose).
Cheers,
Terry
--
Terry Hancock ([EMAIL PROTECTED])
Anansi Spaceworks http://www.AnansiSpaceworks.com
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)
