KSOME (Kinetic Simulations of Microstructural Evolution), An object Kinetic Monte Carlo Code
KSOME (kinetic simulations of microstructural evolution) is an object Kinetic Monte Carlo (OMKC) simulation code developed to simulate microstructural evolution in materials under irradiation. KSOME is developed from the ground up at PNNL. KSOME was developed because of the need for a flexible OKMC code to perform high fidelity simulations of microstructural evolution in materials under irradiation. During the development of KSOME, priority was given to the flexibility, ease of upgradability and computational Efficiency, in that order. Feature of KSOME, some of which will discussed in detail later are No Limit on the number of defect types or defect sub-types No Limit on the number of parameters that can used to capture atomistic details of a defect - Defects of same type with different characteristics can distinguished, allowing for the inclusion of corresponding diffusion-reaction events precisely - Eg.1: A vacancy loop or spherical vacancy clusters or void can be distinguished. Eg.2: (110) and (111) type interstitial defects can be distinguished Ability to incorporate a wide variety of diffusion-reaction properties of irradiation produced defects such as Vacancies (V), Interstitials (I), Transmutants (in case of neutron irradiation) Ability to incorporate pre-existing such as grain Boundaries and dislocations (also called extended defects), alloying and other Impurities Only the diffusion-reaction categories and data-management scheme for efficient execution of the KMC algorithm are hardwired. - Rules for the execution of individual diffusion-reaction processes are provided via input files. - Hence simulations are designed via input files for a particular set of conditions. Gives the user ability to perform high fidelity KMC simulations of radiation damage in materials. Ability to parse mathematical expression via a regular text file. - Eliminates the need for source code modification every time a new mathematical expression to calculate a defect property needs to incorporated into a simulation. - This ability also lets ability to include diffusion-reaction processes that are influenced by long-range interaction between defects
QASMBench: A Low-level QASM Benchmark Suite for NISQ Evaluation and Simulation
We propose a low-level benchmark suite called QASMBench based on the OpenQASM quantum assembly language. It collects commonly seen quantum algorithms and routines from a variety of domains, including chemistry, simulation, linear algebra, searching, optimization, quantum arithmetic, machine learning, fault tolerance, cryptography, etc. from public sources.
DM-Sim: Density Matrix Quantum Circuit Simulation on Modern GPU Clusters
We build a density-matrix quantum circuit simulator accelerated by multi-GPUs. We propose a new simulation approach that significantly reduces communication overhead. We build the tool-chain for supporting high-level quantum programming language, synthesize testing quantum circuits, and enabling ultra-deep quantum simulation. Our simulator has been tested on NVIDIA DGX-1 and DGX-2 systems. It has also been tested on ORNL summit supercomputer using 1024 GPUs.
qFeature
This package contains statistical routines for extracting features from multivariate time-series data which can then be used for subsequent multivariate statistical analysis to identify patterns and anomalous behavior. It calculates local linear or quadratic regression model fits to moving windows for each series and then summarizes the model coefficients across user-defined time intervals for each series. These methods are domain agnostic—but they have been successfully applied to a variety of domains, including commercial aviation and electric power grid data.
DeepMPC for FlexyAir
This is a trained machine learning model written in Pytorch for automated control of the FlexyAir (https://ocl.sk/hardware/flexyair/) dynamical system control educational device.
Diet Parselantro: a widget to support regular expression construction for text organization
Diet Parselantro is a data annotation tool that helps users parse, label, and organize textual data into a taxonomy. It integrates with the python Jupyter notebook environment, a web-based development environment commonly used by data scientists, researchers, and analysts. Diet Parselantro uses regular expressions (regex's) to parse and label data. Users can define new categories and specify a corresponding regex for each. The tool automatically categorizes the textual data based on which regexes they match. Categories and their corresponding regexes can be hierarchical and deeply nested. These hierarchies are visualized in an icicle plot, allowing the user to quickly overview, navigate and refine their categories. The final output of the Diet Parselantro tool is a dataset where each data row is labeled with the category (or categories) they match with. This dataset can be downloaded and used by the data analyst in further analysis tasks.
S-BLAS: A Scalable Sparse-BLAS Kernel Library for Multi-GPU HPC Platforms
In this software package, we have implemented scalable multi-GPU kernel designs for four widely-used BLAS kernels: SpMV, SpTrsv, SpTrans and SpMM on NVIDIA DGX-1 and DGX-2 type of scale-up system, which is the typical configuration for state-of-the-art HPC like OLCF Summit and LLNL Sierra supercomputers.