Friday

Carlos O'Donell Jr. carlos at baldric.uwo.ca
Wed Mar 20 11:49:33 PST 2002


Beowulf-ers,

I recently posted a quick idea about rehasing the
concept of tiled display + beowulf.

My initial thoughts were along the line of:
http://www.baldric.uwo.ca/~carlos/pdisplay.html

Some people were quick to point out (as was I myself)
that it seems rather insane to render and _then_ transfer
the results to the nodes.

However, my points are the following:

- Division of rendering protocols is incredibly complex.
	= WireGL has only a subset of the openGL implementation.

- Division of labour at the X server level requires a lot
  of resources per node.
	= x2x and VNC derivatives are included here.
	= What about micro-X servers?

What's left? Push the division into the lower layers.

- Distributed framebuffer device
	= Similar attributees to VNC systems.
	= Less resources per node
	= Increased network traffic
		- Possibly insane amounts of BW.

Thoughts?

Cheers,
Carlos.

Link List:

http://www.ncsa.uiuc.edu/TechFocus/Deployment/DBox/documentation.html
http://graphics.stanford.edu/~humper/chromium_documentation/
http://www-fp.mcs.anl.gov/fl/publications-electronic-files/intro-to-buildling-projection-822.pdf
http://www.nat.vu.nl/~tvdscha/okf2001/
http://www-unix.mcs.anl.gov/~jones/
http://evlweb.eecs.uic.edu/eric/viscluster/
http://eightway.arc.ucla.edu/projects/outcome_multimonitortiled.htm
http://www.khronos.org/frameset.htm
http://evlweb.eecs.uic.edu/cavern/teranode/teravision.html
http://graphics.stanford.edu/papers/mural_design/mural_design.pdf
http://graphics.stanford.edu/~humper/sc2000_talk/sc2000_4up.pdf
http://www.research.att.com/~csilva/papers/cga2000cr.pdf
http://www.cs.unc.edu/~gotz/Resume/tr01-025.pdf
http://www.uwsg.iu.edu/hypermail/linux/kernel/9808.1/1201.html (wow.... 1998!)
http://www.cs.brown.edu/software/xmx/
http://www-unix.mcs.anl.gov/~jones/Insts/vncwall_inst.html
http://visbench.ncsa.uiuc.edu/DisplayWall/
http://pmatccs.monmouth.army.mil/pages/divisions/chs/pdffiles/tileddisplayarray.pdf
http://www.cs.vu.nl/~renambot/vr/papers/parallel-api.pdf
http://www.ticam.utexas.edu/~wjb/Metabuffer.htm




More information about the Beowulf mailing list