Current and Past Issues of ComputingEdge
Download the Current Issue
Dear ComputingEdge reader:
When HPC Predicts the Future: From forecasting the weather and simulating disaster scenarios to estimating climate trends and even guessing soccer scores, high-performance computing (HPC) is helping us make more-accurate predictions about the world around us. In this ComputingEdge issue, two articles from Computing in Science & Engineering describe supercomputing-enabled techniques that are leading to better predictions in biology and geology applications.
Following the Big Trends
High-performance computing (HPC) is one of the long-standing themes of computer science. It has taken some interesting turns in recent years that are highlighted in the current issue of ComputingEdge.
The issue begins with some of the traditional problems in HPC, but with twists that illustrate how the field is advancing. Katie Elyce Jones gives us some new insight into the classic problem of fluid flow in her article “Supercomputing Improves Predictions of Fluid Flow in Rock.” She shows how this problem is mapped into new high-performance graphics processing units. It is by no means an easy task. It is also essential if we are to deliver the current level of performance at an affordable price. “The analysis framework within [the model] can churn through petabytes of data per hour,” she writes. The results of this approach mean that the computer puts fewer intermediate results to disk (a task that always slows HPC) and that the “postprocessing analyses are considerably more efficient.”
Nils Heinonen twists a classic high-performance model in his article “Predicting the Risk of Cancer With Computational Electrodynamics.” At base, he is again undertaking a traditional problem, a simulation at the molecular level of a complex problem. However, his application uses this simulation to predict an unknown outcome, and, as a result, he has to think carefully about how he approaches the problem.
“Accelerators for Artificial Intelligence and High-Performance Computing,” by Dejan Milojicic, deals with a relatively new topic in HPC. We got the first hints that AI might need speed processes back in the 1980s with the Japanese Fifth Generation Project. However, the field has really expanded only in the last two decades, as we have learned to train large-scale, multi-layered neural networks with millions if not billions of parameters. This article is really a conversation among four well-regarded experts in the field. One of the great joys of reading it is watching how these four minds conceive the problems of HPC and how they forecast the next generation of these specialized machines will develop.
ComputingEdge will help you understand how these experts think. Once you grasp how they are thinking, you’ll be better able to predict how the field is advancing. It is one of the best things that we do with ComputingEdge. You would do well to take a look this issue so that you can get a glimpse of the future of HPC.
—David Alan Grier for ComputingEdge