[ Up ] [ Materials Science ] [ Solving QCD ] [ Teaching Physics ] [ LASSP ] [ Chemistry ] [ Physics Training ] [ Mathematics ]

Solving QCD On Intel Workstations

We propose to use a small number of Intel’s most powerful workstations to demonstrate that QCD, the fundamental theory of proton substructure, can be solved numerically on Intel workstations. Although such calculations traditionally employ only the largest of supercomputers, algorithmic innovations pioneered at Cornell have dramatically reduced the computational cost, making state-of-the-art work possible on Intel’s powerful desktop systems. This project will demonstrate this point by completing, on Intel hardware, a series of forefront calculations directly relevant to current experimental work in particle physics. We will also make our simulation software freely available via the Web, thereby jump-starting research in numerical QCD by new groups throughout the United States and abroad.

QCD, the fundamental theory of quarks and gluons that explains the internal structure and interactions of protons and neutrons, can only be solved numerically, using Monte Carlo simulations. Until two years ago it was universally accepted that such simulations were possible only on massively parallel supercomputers capable of teraflops-scale calculations; tens of millions of dollars have been invested in the US, Europe and Japan to develop and acquire machines for this purpose. Nevertheless, in early 1996, Prof. Lepage gave a colloquium at Cornell during which he did a complete state-of-the-art calculation using a single, relatively slow laptop computer.

Lepage and his collaborators recently began investigating new algorithms for QCD simulations. They have now shown that these will reduce the cost of simulations by factors of 1,000 to 1,000,000. This development is revolutionizing the study of strong interaction physics, a key problem in theoretical physics for the past fifty years. During the past decade, numerical QCD has come to be dominated by a small number of large international collaborations built around very large and costly computing facilities. Projects that once required months of supercomputer time can be tackled in days or even hours on a workstation. This will result in far greater output from numerical QCD, but also will liberate researchers to be far more daring in what they try. These new developments have generated considerable interest within physics (see, for example, Gary Taubes’ news report in the 15 Dec. 1995 issue of Science), but the large QCD collaborations are very conservative and have yet to capitalize fully on the new simulation technology. The goals of this project are:

Participants

Last modified on: 07/30/99