Science at 20 petaflops
The Center for Accelerated Application Readiness is working with six world-class applications that are representative of the Oak Ridge Leadership Computing Facility computing workload to show how GPUs can revolutionize computational science when they are combined with CPUs.
Combustion with S3D – Three-quarters of the fossil fuel used in the United States goes to cars and trucks, which produce a quarter of the country's greenhouse gases. Advances in fuel efficiency and alternative-fuel engines benefit national security and the environment. The S3D application simulates fuel burning under highly turbulent conditions. At 20 petaflops, it will move beyond simple fuels to tackle complex, larger-molecule hydrocarbon fuels such as isooctane (a surrogate for gasoline), commercially important oxygenated alcohols such as ethanol and butanol, and biofuel surrogates (blends of methyl butanoate, methyl decanoate and n-heptane.)
Magnetic systems with WL-LSMS – Magnetism at the atomic scale plays an important role in many industrial materials, including steels and iron-nickel alloys; and lightweight yet strong permanent magnets are important components in highly efficient electric motors and generators. Small improvements in the performance of these materials will result in more competitive industries and greater energy efficiency. WL-LSMS combines two methods, known as locally self-consistent multiple scattering and Wang- Landau. It analyzes magnetic materials at the nanoscale, allowing researchers to directly and accurately calculate, for example, the temperature above which a material loses its magnetism. At 20 petaflops, WL-LSMS will improve calculations of a material's thermodynamics or calculate the underlying magnetic states with greatly reduced margins of error.
Biophysical science with LAMMPS – Biofuels are among the most promising approaches to alternative energy production, but the process of turning woody plants into fuel is laborious and expensive. LAMMPS—or Large-scale Atomic/Molecular Massively Parallel Simulator—explores bioenergy using molecular dynamics, modeling problems such as membrane fusion, large biomolecular simulations for proteins and lignocellulose for biofuels. At 20 petaflops, it will be able to overcome size limitations on systems with charged particles, expanding to millions of atoms.
Nuclear reactors with Denovo – Nuclear power provides abundant, reliable and emission-free electricity. To be effective, though, safety must be guaranteed, and the volume of radioactive waste must be reduced (e.g., by burning the fuel longer in the reactor). Denovo is a powerful tool for ensuring the safe and efficient operation of today's nuclear power plants. At 20 petaflops it will take only about 13 hours to simulate a fuel rod through one round of use in a reactor core. The same simulation took 60 hours on Jaguar.
Climate change with CAM-SE – Improved atmospheric modeling will help climate researchers better understand future air quality, as well as the effect of particles suspended in the air—a large source of uncertainty regarding the climate's response to natural and human effects. The Community Atmosphere Model–Spectral Element simulates long-term global climate to inform public policy and improve scientific understanding of climate changes. At 20 petaflops it will be able to increase the simulation speed to between one and five years per computing day. The increase in speed is needed to make ultra-high-resolution, full-chemistry simulations feasible over decades and centuries and would allow researchers to quantify uncertainties by running multiple simulations.
Radiation transport and advanced algorithms with NRDF – The Non-Equilibrium Radiation Diffusion application models the journey of noncharged particles. It is also being used to develop advanced computing techniques that allow applications to solve larger problems by focusing computing power only on critical areas of a simulated system. NRDF has applications in areas such as astrophysics, nuclear fusion, and atmospheric radiation; the algorithms being developed for it should prove valuable in many other areas, such as fluid dynamics, radiation transport, groundwater transport, nuclear reactors and energy storage. —Leo Williams