Small Business Innovation Research/Small Business Tech Transfer
High-Quality Random Number Generation Software for High-Performance Computing
Project Description
Random number (RN) generation is the key software component that permits random sampling. Software for parallel RN generation (RNG) should be based on RNGs that are good serial generators, but also are suitable for concurrent execution, i.e. RNs produced by different concurrent processes must be statistically independent of one another. With dependence, the work done on one concurrent process will be partially redone on another process, thus reducing the efficiency of the concurrent computation and defeating the purpose of concurrency. Although parallel RNGs exist for older parallel computing paradigms such as clusters, modern advances in computer architecture, specifically the hardware support for multithreading via multicore and other architectures, and widely available computational coprocessor technologies, such as the General Purpose Graphics Processing Unit (GPGPU) have created the need for high-quality RNG software to support these new architectural features. Thus, this Phase I research project addresses these new architectures from a mathematically rigorous stand-point of preventing statistical dependence and allowing for reproducibility in modeling and simulation as well as other applications. In addition, it investigates applications to small memory technologies such as embedded systems and radio-frequency identification (RFID) chips.
More »
Anticipated Benefits
The main beneficiary of the proposed tool is NASA Ames Research Center's Advanced Supercomputing Division (http://www.nas.nasa.gov/) which provides tools and resources for supercomputing at NASA. Random number generation is important in any project with a stochastic component and the advantage of a GPGPU approach is a major saving in money for computational power. A number of other NASA centers rely on detailed simulations of stochastic processes on supercomputers. The NASA Center for Climate Simulation (http://www.nccs.nasa.gov/) provides climate simulation tools and high-performance computing resources for NASA projects. Such simulations would benefit from GPGPU random number generation tools. NASA is a key participant in the federal High Performance Computing and Communications (HPCC) Program (http://www.hq.nasa.gov/hpcc/mission.htm) to extend U.S. technological leadership in high-performance computing and communications for the benefit of NASA stakeholders: the U.S. aeronautics, Earth and space sciences, and spaceborne research communities. The projects NASA HPCC supports would also benefit from the proposed tools.
The market for parallel random number generators is large and expanding. Virtually all scientific computing applications have a stochastic component requiring random numbers and any parallelization requires a parallel RNG to avoid coherence. The three largest RNG-consuming applications are (1) discrete-event simulation including war gaming, many systems-level simulation tools, and video games (e.g., Army training simulations); (2) particle Monte Carlo for simulating nuclear weapons and other large nuclear processes (astrophysics) and for planning treatments for radiology; and (3) finance where MC based computations are mandated in many nations under the Basel II accord for banking supervision; insurance companies, hedge funds, and investment firms rely on MC to price derivative products. Other consumers are intelligence agencies doing cryptanalysis and cryptography, aerospace and defense contractors doing modeling and simulation, and universities and federally funded research labs/national labs that do noise modeling and statistical analysis. More »
The market for parallel random number generators is large and expanding. Virtually all scientific computing applications have a stochastic component requiring random numbers and any parallelization requires a parallel RNG to avoid coherence. The three largest RNG-consuming applications are (1) discrete-event simulation including war gaming, many systems-level simulation tools, and video games (e.g., Army training simulations); (2) particle Monte Carlo for simulating nuclear weapons and other large nuclear processes (astrophysics) and for planning treatments for radiology; and (3) finance where MC based computations are mandated in many nations under the Basel II accord for banking supervision; insurance companies, hedge funds, and investment firms rely on MC to price derivative products. Other consumers are intelligence agencies doing cryptanalysis and cryptography, aerospace and defense contractors doing modeling and simulation, and universities and federally funded research labs/national labs that do noise modeling and statistical analysis. More »
Primary U.S. Work Locations and Key Partners
| Organizations Performing Work | Role | Type | Location |
|---|---|---|---|
| DANIEL H. WAGNER ASSOCIATES, INC. | Lead Organization | Industry | Exton, Pennsylvania |
Ames Research Center
(ARC)
|
Supporting Organization | NASA Center | Moffett Field, California |
Primary U.S. Work Locations
-
California
-
Pennsylvania
Suggest an Edit
Recommend changes and additions to this project record.
This is a historic project that was completed before the creation of TechPort on October 1, 2012. Available data has been included. This record may contain less data than currently active projects.