Distributed Volume Renderer

 

DVR is designed to render larger-than-memory volumetric datasets at interactive framerates by using Beowulf-style clusters.

It is:

  • fast - uses the modern shear-warp algorithm,
  • easy - just load your dataset and drag the mouse to move around the volume,
  • scalable - runs in parallel on clusters, and
  • flexible - can be scripted via Tcl.

DVR was developed at The Centre for Astrophysics and Supercomputing, at Swinburne University of Technology, funded by an Expertise Grant (EPPNSW049) from the Victorian Partnership for Advanced Computing (VPAC). The principal investigators on the grant were David Barnes and Paul Bourke, and the design and coding were accomplished by Brett Beeson. We gratefully acknowledge J├╝rgen Schulze for providing the core rendering code ( Virvo) on which we built the distributed system.

A paper describing DVR has been published (Beeson, Barnes & Bourke, 2003, "A distributed-data implementation of the perspective shear-warp volume rendering algorithm for visualisation of large astronomical cubes," Publications of the Astronomical Society of Australia. 20, 300). The accepted version is available here in pdf format. The sample animations referenced in the paper are available in the table below.

Sample content

Low quality AVI format (size in MB)

High quality QuickTime format (size in MB)

N-body cosmology

nbodycos.avi (5.5)

nbodycos.mov (67)

N-body galaxy formation and evolution

nbodygfe.avi (3.8)

nbodygfe.mov (23)

DVR is available for download! It will build on most Linux systems, as well as the Darwin (Mac OS X) operating system, and Sun Solaris. X-Windows (eg. XFree86, XDarwin, ...) is required, as well as a recent version of Tcl and the Tk Toolkit (versions 8.3 and 8.4 are known to work). The GNU compiler collection version 3.1.0 or upwards are required. As DVR is oriented towards astronomy data, the CFITSIO and CCfits libraries are necessary to support standard astronomy FITS files. We have recently added limited support for HDF format files using HDF v. 4.1r5.

DVR can be run on a single machine, or on a cluster with rsh or ssh installed. No support is guaranteed, but you may submit questions regarding DVR to Brett Beeson or David Barnes, and we may be able to assist.

DVR code version

Date released

Size (MB)

Download

v0_91

2003 May 07

2.0

dvr.v0_91.tar.gz

v0_95

2003 June 12

2.0

dvr.v0_95.tar.gz

v0_96

2003 June 13

2.0

dvr.v0_96.tar.gz

v0_98

2003 June 30

2.0

dvr.v0_98.tar.gz

v1_00

2003 December 1

2.2

dvr.v1_00.tar.gz

Extract the downloaded file and read the top-level README file for brief installation instructions. Please note that the docs directory contains information which may be out-of-date or of questionable value.

Some screenshots and example renderings from DVR are available on this external site.

DVR has also been deployed across a private compute and data grid (Rixon et al., 2004, "VIsualizing Data Cubes on the Grid," Astronomical Data Analysis Software and Systems XII, submitted). An interface to DVR also exists for the AIPS++ viewer.

The Fornax cluster of galaxies (above): dvr was used to volume render this volume of data whose axes are longitude and latitude on the sky and line-of-sight velocity of neutral hydrogen hyperfine emission. The voxel values record incident flux density. The volume was produced from observations with the Parkes 64m radiotelescope using the 21cm Multibeam receiver (Waugh et al., in prep.).