Scientific Visualization a Necessary Chore the Eighties

F or the past 30 years, I’ve focused my research activities on gaining a better understanding of the nonlinear development of gravitationally driven instabilities in astrophysical fluid flows. The computational fluid dynamics (CFD) techniques that my group at Louisiana State University (LSU) has employed to model various astrophysical phenomena are, broadly speaking, similar to those the engineering community uses to model compressible, supersonic flows in various terrestrial environments. Currently, in collaboration with Juhan Frank and Patrick Motl at LSU, my students and I are using these techniques to simulate mass-transfer instabilities in binary star systems for which Newton’s law provides an adequate description of the time-varying gravitational field;1,2 through a collaboration with Luis Lehner (also at LSU), my group is expanding the capabilities of our simulation tools to model compact binary star systems, which require a general relativistic treatment of gravity.3 Not surprisingly, over the years, my research has benefited from the rapidly advancing industry of highperformance computing (HPC) and communications. It has been exciting to be involved in large-scale numerical simulation development for the past three decades, as the raw capabilities of HPC resources have routinely doubled according to Moore’s law. However, continuously upgrading my group’s software simulation algorithms to keep pace with ever-improving HPC hardware has been a tremendous challenge, to say the least. The concurrent development of hardware and software tools to facilitate scientific visualization has been extraordinary as well. With ever increasing degrees of sophistication, these tools have provided absolutely critical support for our efforts to diagnose the results of complex CFD simulations. But this, too, has come with a price. As the students who have honed their research skills in my group will attest, roughly one-third of the person-hours invested in research every year have been directed at improving our abilities to effectively visualize the results of our 3D, timedependent CFD simulations. Most recently, my group has begun collaborating with Claudio Silva and his colleagues from the Scientific Computing Institute (SCI) at the University of Utah to explore how our scientific visualization efforts can benefit from VisTrails’ provenancetracking capabilities. Over the next few years, Claudio and I plan to use the CiSE Visualization Corner and attending Web technologies as a venue for sharing highlights of our collaboration in hopes of promoting the idea of reproducible visualizations (see the “Reproducability and Sharing Data and Processes for the Vizualization Corner” sidebar in last issue’s column4). Before embarking on this new adventure, however, it’s useful—and certainly therapeutic—for me to recount key events from the past couple of decades that have strongly influenced my group’s past and present scientific visualization activities. Here, I illustrate how one astrophysicist with an interest in scientific visualization but, until recently, little access to inhouse visualization expertise, “made do” with the tools available at the time. I suspect that many computational scientists will identify with the struggles as well as the successes associated with these recounted events, at least in spirit if not in detail.