Wesley B. Jones

SPEComp: A New Benchmark Suite for Measuring Parallel Computer Performance

We present a new benchmark suite for parallel computers. SPEComp targets mid-size parallel servers. It includes a number of science/engineering and data processing applications. Parallelism is expressed in the OpenMP API. The suite includes two data sets, Medium and Large, of approximately 1.6 and 4 GB in size. Our overview also describes the organization developing SPEComp, issues in creating OpenMP parallel benchmarks, the benchmarking methodology underlying SPEComp, and basic performance characteristics.

SPEC HPG benchmarks for high-performance systems

In this paper, we discuss the results and characteristics of the benchmark suites maintained by the Standard Performance Evaluation Corporations (SPEC) High-Performance Group (HPG). Currently, SPECHPGhas two lines of benchmark suites for measuring performance of large-scale systems SPEC OMP and SPEC HPC2002. SPEC OMP uses the OpenMP API and includes benchmark suites intended for measuring performance of modern shared memory parallel systems. SPEC HPC2002 uses both OpenMP and MPI, and thus it is suitable for distributed memory systems, shared memory systems and hybrid systems. SPEC HPC2002 contains benchmarks from three popular application areas: chemistry, seismic and weather forecasting. Each of the three benchmarks in HPC2002 has a small and a medium data set in order to satisfy the need for benchmarking a wide range of high-performance systems. We analyse published results of these benchmark suites regarding scalability. We also present current efforts of SPEC HPG to create new releases of the benchmark suites.

IGMS: An Integrated ISO-to-Appliance Scale Grid Modeling System

This paper describes the integrated grid modeling system (IGMS), a novel electric power system modeling platform for integrated transmission-distribution analysis that co-simulates off-the-shelf tools on high performance computing platforms to offer unprecedented resolution from independent system operator (ISO) markets down to appliances and other end uses. Specifically, the system simultaneously models hundreds or thousands of distribution systems in co-simulation with detailed ISO markets and automatic generator control-level reserve deployment. IGMS uses a new message passing interface-based hierarchical co-simulation framework to connect existing sub-domain models. Our initial efforts integrate open-source tools for wholesale markets, bulk ac power flow, and full-featured distribution systems including physics-based end-use and distributed generation models (many instances of GridLAB-D). The modular IGMS framework enables tool substitution and additions for multi-domain analyses. This paper describes the IGMS tool, characterizes its performance, and demonstrates the impacts of the coupled simulations for analyzing high-penetration solar photovoltaic and price responsive load scenarios.

Predicting the electronic properties of 3D, million-atom semiconductor nanostructure architectures

The past ~10 years have witnessed revolutionary breakthroughs both in synthesis of quantum dots (leading to nearly monodispersed, defect-free nanostructures) and in characterization of such systems, revealing ultra narrow spectroscopic lines of <1meV width, exposing new intriguing effects, such as multiple exciton generation, fine-structure splitting, quantum entanglement, multi- exciton recombination and more. These discoveries have led to new technological applications including quantum computing and ultra-high efficiency solar cells. Our work in this project is based on two realizations/observations: First, that the dots exhibiting clean and rich spectroscopic and transport characteristics are rather big. Indeed, the phenomenology indicated above is exhibited only by the well-passivated defect-free quantum dots containing at least a few thousand atoms (colloidal) and even a few hundred thousand atoms (self assembled). Understanding the behavior of nanotechnology devices requires the study of even larger, million-atom systems composed of multiple components such as wires+dots+films. Second, first-principles many-body computational techniques based on current approaches (Quantum Monte-Carlo, GW, Bethe-Salpeter) are unlikely to be adaptable to such large structures and, at the same time, the effective mass-based techniques are too crude to provide insights on the many-body/atomistic phenomenology revealed by experiment. Thus, we have developed a set of methods that use an atomistic approach (unlike effective-mass based techniques) and utilize single-particle + many body techniques that are readily scalable to ~10 3

Direct enumeration of alloy configurations for electronic structural properties

We present and apply an approach to directly enumerate the band gaps and effective masses of all possible zinc blende-based alloy configurations whose unit cell contains up to a specified number of atoms. This method allows us to map the space of band gaps and effective masses versus alloy composition and atomic configuration. We demonstrate that a large number of band gaps and effective masses are available. We also discuss convergence of the method with respect to unit cell size and the combined optimization of band gap and effective mass for AlGaAs and GaInP semiconductor alloys.

Surface passivation optimization using DIRECT

We describe a systematic and efficient method of determining pseudo-atom positions and potentials for use in nanostructure calculations based on bulk empirical pseudopotentials (EPMs). Given a bulk EPM for binary semiconductor X, we produce parameters for pseudo-atoms necessary to passivate a nanostructure of X in preparation for quantum mechanical electronic structure calculations. These passivants are based on the quality of the wave functions of a set of small test structures that include the passivants. Our method is based on the global optimization method DIRECT. It enables and/or streamlines surface passivation for empirical pseudopotential calculations.

SPEC HPG Benchmarks for Large Systems

Performance characteristics of application programs onlarge-scale systems are often significantly different from those on smaller systems. In this paper, we discuss such characteristics of the benchmark suites maintained by SPEC’s High-Performance Group (HPG). The Standard Performance Evaluation Corporation’s (SPEC) High-Performance Group (HPG) has developed a set of benchmarks for measuring performance of large-scale systems using both OpenMP and MPI parallel programming paradigms. Currently, SPEC HPG has two lines of benchmark suites: SPEC OMP and SPEC HPC2002. SPEC OMP uses the OpenMP API and includes benchmark suites intended for measuring performance of modern shared memory parallel systems. SPEC HPC2002 is based on both OpenMP and MPI, and thus it is suitable for distributed memory systems, shared memory systems, and hybrid systems. SPEC HPC2002 contains benchmarks from three popular application areas, Chemistry, Seismic, and Weather Forecasting. Each of the three benchmarks in HPC2002 has small and medium data sets in order to satisfy the need for benchmarking a wide range of high-performance systems. We present our experiences regarding the scalability of the benchmark suites. We also analyze published results of these benchmark suites based on application program behavior and systems’ architectural features.

Final Technical Report: Integrated Distribution-Transmission Analysis for Very High Penetration Solar PV

Transmission and distribution simulations have historically been conducted separately, echoing their division in grid operations and planning while avoiding inherent computational challenges. Today, however, rapid growth in distributed energy resources (DERs)--including distributed generation from solar photovoltaics (DGPV)--requires understanding the unprecedented interactions between distribution and transmission. To capture these interactions, especially for high-penetration DGPV scenarios, this research project developed a first-of-its-kind, high performance computer (HPC) based, integrated transmission-distribution tool, the Integrated Grid Modeling System (IGMS). The tool was then used in initial explorations of system-wide operational interactions of high-penetration DGPV.

Time Domain Partitioning of Electricity Production Cost Simulations

Production cost models are often used for planning by simulating power system operations over long time horizons. The simulation of a day-ahead energy market can take several weeks to compute. Tractability improvements are often made through model simplifications, such as: reductions in transmission modeling detail, relaxation of commitment variable integrality, reductions in cost modeling detail, etc. One common simplification is to partition the simulation horizon so that weekly or monthly horizons can be simulated in parallel. However, horizon partitions are often executed with overlap periods of arbitrary and sometimes zero length. We calculate the time domain persistence of historical unit commitment decisions to inform time domain partitioning of production cost models. The results are implemented using PLEXOS production cost modeling software in an HPC environment to improve the computation time of simulations while maintaining solution integrity.

Exploring high-dimensional data space: Identifying optimal process conditions in photovoltaics

We demonstrate how advanced exploratory data analysis coupled to data-mining techniques can be used to scrutinize the high-dimensional data space of photovoltaics in the context of thin films of Al-doped ZnO (AZO), which are essential materials as a transparent conducting oxide (TCO) layer in CuInxGa1−xSe2 (CIGS) solar cells. AZO data space, wherein each sample is synthesized from a different process history and assessed with various characterizations, is transformed, reorganized, and visualized in order to extract optimal process conditions. The data-analysis methods used include parallel coordinates, diffusion maps, and hierarchical agglomerative clustering algorithms combined with diffusion map embedding.