A Parallel Volume Rendering System for Scientific Visualization

Approach: A parallel volume rendering Application Programming Interface (API) based on the splatting algorithm was designed and implemented. The API renders 3-D scalar and vector fields in structured grids, in both time and space. An X-window based GUI with multiple control panels was also designed and implemented for interactive control of visualization parameters and trackball control of viewing positions. A network interface was built in between the GUI and the renderer API that supports image compression over low-speed network and high-resolution displays such as HiPPI framebuffer and Powerwall over high-speed network. All together, the distributed volume rendering system is called the ParVox (Parallel Voxel Renderer).

Accomplishments: The ParVox was first demonstrated in October 1997. In FY'97, we continue the development and implementation of the ParVox system. The major accomplishments include:

1. Vector visualization -- we have tried several ways for visualizing vectors, including Line Integral Convolution (LIC) and magnitudes. Both methods convert vectors into scalars, then using scalar rendering to look at the data. We found out LIC is not a very intuitive way to visualize 3-D volumes; it is more effective if the vectors can be mapped to a predefined surface. We also found out that magnitude is a good representation for certain datasets. By adjusting the voxel opacity based on the magnitude values, we are able to visualize the eddy-resolving at Atlantic Ocean in the ocean velocity datasets from the Ocean Climate Modelling Project. The data was provided Yi Chao, using the Parallel Ocean Program (POP). A five-year ocean circulation animation was generated using ParVox on the JPL Cray T3D. One still image from the animation is shown in Figure 1.

   

2. Slice visualization -- Plotting 2-D slices is a common way to look at 3-D datasets. We added a new feature in ParVox to enable visualization of multiple slices along three major axes. A new control panel was built in GUI to allow users to define up to 10 slices with designated opacities. A multiple-slice image rendered using data from a 3D Thermal Convective Flow Model is shown in Figure 2. Data was provided by Ping Wang.

3. Color and opacity editing -- We integrated a color editing software, Icol, developed by Army High Performance Computing Research Center (AHPCRC), in the ParVox GUI. It is a useful tool for interactive editing and sketching color maps and opacity maps that are interpolated between key points.

   In addition, we wrote a technical paper, ParVox -- A Parallel Splatting Volume Rendering System for Distributed Visualization. This paper will be published in the IEEE Proceedings of Parallel Rendering Symposium, October 1997. A ParVox home page will be available in the near future.

Status/Plans: The major plan for FY'98 is to port ParVox to MPI2.0. ParVox uses asynchronous one-way communication routines in Cray's shmem library to interleave the rendering and compositing, thus hiding the communication overhead and improving the parallel algorithm's efficiency. The MPI2.0 specification, announced in July '97, includes a one-way communication API that is very similar to Cray's shmem library. To minimize the porting effort and optimize the memory access performance of the shared-memory MPP, which is becoming the dominant architecture for the parallel supercomputers, we decide to port ParVox to the MPI2.0. Currently, none of the supercomputer vendors are supporting MPI2.0 yet, but we expect it will become available soon. The second plan for the FY'98 is to provide functional pipelining by separating the compression module from the rendering module and building interface between the application program and the ParVox. A fully pipelined ParVox distributed visualization system is illustrated in Figure 3.

Point of Contact:
P. Peggy Li
Jet Propulsion Laboratory
(818)354-1341