September 2, 2010
It’s a feature of Thomas Jefferson National Accelerator Laboratory that it has at least two other names, including Jefferson Lab and JLab. Similarly, parts of our organization go by different names - the Theory group, the Theory Department, the Theory Center and the Center for Theoretical and Computational Physics. But a new name might be “Game Center.” Let me explain.
Large-scale computing has been a major deal for the Department of Energy for many years. Many have read of Feynman’s calculating efforts in the Manhattan Project. Later, the so-called supercomputers of the ’60s, ’70s, and ’80s with CDC, Cray, IBM and Amdahl were installed with restricted access in the weapons laboratories.
The particle and nuclear physics labs across the world also needed large-scale computational capability. The special attributes of the computations started to play a major role, and the nature of the physics interaction data, with thousands or even millions of events per experiment, started to influence things. Each event requires essentially the same treatment, and this led to the adoption of farms of small computers all operating at the same time, each on a different event to make the computational problem of data analysis tractable without breaking the bank.
In approximately the same era, it was recognized that a collaboration between small computers connected as in a lattice could be used to calculate, albeit approximately, the solution of several problems in quantum chromodynamics, the theory of strong interactions. So, the theorists came to play. Some developed their own devices. The Columbia group and collaborators built several generations of special processors, some in collaboration with IBM. The QCDOC (Quantum ChromoDynamics On a Chip) evolved into the Blue Gene series of computers. There were analogous developments in Europe, particularly in Italy with the APE series of projects. But the interconnection of commodity computers turned out to be a competitive path. In the late ’80s, an in-house development at Fermilab indicated the way, and by the turn of the century, large clusters of personal computers with fast connections were installed at Fermilab and Jefferson Lab.
Now when we go and buy a PC in the store, we usually expect that we don’t need to talk directly to the machine; there is an already developed language, such as Basic, or C(C++), Fortran, Cobol. Support for the development of analogous tools, for the theorists and others, was provided by the SciDAC (Scientific Discovery with Advanced Computing) program in the Office of Science. These developments broadened the use enormously and made life easier for the practitioners. Today, the best of this software can run on many platforms and enables computations and collaborations on a nationwide and even worldwide scale.
In many ways this story is one of a revolution. Computers penetrated the heart of our science on the leading edge of the same wave that brought computing into our everyday lives. Imagine how much faster the Manhattan Project could have gone if they had had the computing now installed in your home appliances, or in your toys!
In your toys, now there’s a thought! Over the past couple of years our own Jefferson Lab team has been working on tools to allow theorists to talk to our toys, or at least to the Graphical Processing Units, GPUs, in the game boxes. So, when our theorists received $5 million of ARRA funding, they bought computers with GPUs attached. This is the origin of their recent smiles. The team members from Jefferson Lab and several collaborating institutions, such as Boston and Harvard universities, as well as the local groups at Old Dominion and the College of William and Mary, are realizing power at a level in excess of that anticipated by factors up to 10. We showed them off to Bill Brinkman and to Steve Koonin, and they have had articles on the DOE websites, and also on our own website. As a result, there have been publications with predictions of new states, which has the Hall D GlueX team, whose experiment is being built as part of the 12 GeV Project, salivating in anticipation of getting their hands on some data.
So, yes! Center for Theoretical and Computational Physics, but “Game Center” would also serve!!