Text archives Help

---

Definitely keep in touch, because we have done a few things like you described and we are looking at parallel/MPI versions of Manta. It would definitely be great to have someone else looking at it!
Steve


On Mar 22, 2007, at 3:01 PM, Shreekanth Pavar wrote:

Quoting "Steven G. Parker" <sparker@cs.utah.edu>:

Technically, Manta would be considered sort-last, but as Abe points =20
out it does not really do any sorting. The sort- first,middle,last =20
is really more suited for a rasterization pipeline.   For a ray-=20
tracer, the two main possibilities are image-space parallel or object-=20=

space parallel.   Manta is the former.

Steve

Abe and Steve

Thankyou both for your replies which have cleared up some confusion on
my side.
I'm slowly understanding more about both Manta and Ray Tracing with each email
exchange :)

The paper that Abe referred me to:

D.E. DeMarle, C.P. Gribble, S. Boulos, S.G. Parker. �Memory Sharing for
Interactive Ray Tracing on Clusters,� In Parallel Computing,  Vol.
31, No. 2,
pp. 221--242. 2005

refers to work done in running Manta using distributed shared memory over a PC
cluster. This is similar to what we're intending to do, so thankyou for this
reference - we don't want to duplicate any work already done !

My intention is to follow up on a suggestion made by Hansong Zhang at SGI to
port Manta to run on a fat-node (i.e. multicore) cluster. As far as I know,
this has not yet been tried. This would form the bulk of the implementation
work for my PhD.

I was thinking of doing this in two stages:

(1) Get Manta running in single-threaded mode on multiple 'nodes' in a PC
cluster (using MPI to pass data between nodes)

(2) Add (i.e. re-introduce) the multi-threading onto each node in the above
cluster to take advantage of multicore nodes

To keep things simple to start with, I plan to the make the whole dataset
available to each node. This restricts the size of the datasets which can be
visualised, so I eventually plan to also introduce a data decomposition
scheme.
With your permission, this could be the one which you used in the above paper.

I'd welcome any comments/suggestions about this approach and will keep you
updated with my progress.

Regards

Shree

---

• [MANTA] MANTA : Questions about how the rendering stack works, Shreekanth Pavar, 03/20/2007
  • Re: [MANTA] MANTA : Questions about how the rendering stack works, Steven G. Parker, 03/20/2007
    • Re: [MANTA] MANTA : Questions about how the rendering stack works, Shreekanth Pavar, 03/21/2007
      • Re: [MANTA] MANTA : Questions about how the rendering stack works, Abe Stephens, 03/21/2007
        • Re: [MANTA] MANTA : Questions about how the rendering stack works, Steven G. Parker, 03/22/2007
          • Re: [MANTA] Running on a fat-node cluster, Shreekanth Pavar, 03/22/2007
            • Re: [MANTA] Running on a fat-node cluster, Steven G. Parker, 03/22/2007
            • Re: [MANTA] Running on a fat-node cluster, Abe Stephens, 03/22/2007
              • Re: [MANTA] Running on a fat-node cluster, Solomon Boulos, 03/22/2007
              • Re: [MANTA] Running on a fat-node cluster, Christiaan Paul Gribble, 03/25/2007

---

Archive powered by MHonArc 2.6.16.