[Paraview-developers] IceT compositing in paraview

Biddiscombe, John A. biddisco at cscs.ch
Wed Jul 30 09:26:54 EDT 2014


Dear List, but hopefully/probably for the attention of Ken

Our custom splotch representation generates an image on each rank, but to get certain effects right, we pass all the images back from ranks>0 to rank=0 and perform a post render step on rank 0 (pixel normalization). Then  rank0 returns the final image to paraview and all ranks N>0 return an empty (black) image.

This means that the compositing step we do (inefficient because it is using MpiReduce and not binary swap or direct send - or any other nice stuff in IceT) is performed once, and then  paraview does its own compositing step as well.

I'd like to remove the duplicated compositing phase, but our image has no depth buffer information and so I need to tell paraview/IceT to do a composition of the images using a SUM operation rather than a z-buffer test.

is it possible to do this? (from a representation?)

If it is not possible to override the compositing step from paraview, could I implement a compositing step like allReduce(MPI::SUMP_OP), but using iceT instead of just raw MPI as it might improve performance.

We'd like to use>1024 GPUs to render some really big cosmological data and so the compositing will be expensive and doing it twice is naturally undesirable.

many thanks for any clues (I may have asked this question a couple of years ago, but have forgotten the answer).

yours

JB


--
John Biddiscombe,                        email:biddisco @.at.@ cscs.ch
http://www.cscs.ch/
CSCS, Swiss National Supercomputing Centre  | Tel:  +41 (91) 610.82.07
Via Trevano 131, 6900 Lugano, Switzerland   | Fax:  +41 (91) 610.82.82

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/paraview-developers/attachments/20140730/e372d371/attachment-0001.html>


More information about the Paraview-developers mailing list