SEARCH · Search NASA
Results for “Mathematics and Computing, Computer science”
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
Advanced Simulation and Computing (FY22 Implementation Plan Rev 0)
The DOE National Nuclear Security Administration (NNSA) Stockpile Stewardship Program (SSP) is an integrated technical program for maintaining the safety, surety, and reliability of the U.S. nuclear stockpile. The SSP incorporates nuclear test data, computational modeling and simulation, and experimental facilities to advance understanding of nuclear weapons. The suite of data analyzed comes from activities including stockpile surveillance, experimental research, and development and engineering programs. This integrated national program requires the continued use of experimental facilities and the computational capabilities to support the SSP missions. These component parts, in addition to an appropriately scaled production capability, enable NNSA to support stockpile requirements. The ultimate goal of the SSP, and thus of the Advanced Simulation and Computing (ASC) Program, is to ensure that the U.S. maintains a safe, secure, and effective strategic deterrent. Specific work activities and scope contained in this Implementation Plan (IP) represent the full-year annual operating plan for FY22. The Initial IP, effective , should be consistent with the Department’s Base Table when operating under a Continuing Resolution (CR). The final IP, effective date TBD, is consistent with the final, enacted appropriation.
Using Computational Storage Devices: OpenMP/MPI and Charliecloud [Slides]
Originally used Spark and HadoopFS. Collected interesting results, but this method had its issues: limited application, too much overhead to gauge CSDs’ raw performance. Solution? Rewrite our benchmarks without Spark using Serial Python and Serial & Parallel C++ (Combinations of OpenMP & OpenMPI). Compared to Spark and Python, C++ implementation is a lot faster. Caveat: an expert with Spark or Python would likely be able to improve the performance of those implementations. Computational power of our CSDs seem to be much lower than the host machine. Using all 4 cores of a single CSD, the job takes ~6.8x longer than using just one core on the host machine. Host also seems to scale better with increasing file size. Resulting Question: When, if ever, would it make sense to use CSDs for compute rather than a much-faster host?
Offloading Calculations to Computational Storage Devices: Spark and HDFS [Slides]
The objective is to evaluate the capabilities of multiple CSDs (provided by NDG Systems) using Hadoop Filesystem and Apache Spark. The independent variables are: number of CSDs, 0, 1, 2, 4, or 6; size of dataset, 1 GB, 5 GB, 10 GB; type of dataset, one large file with all of the data, 10 files, 100 files. The dependent variables are: job time; execution time. the constants are operations on the dataset.
FleCSI Intro 2.0 [Slides]
FleCSI is a programming system or framework for developing mulitphysics application codes. FleCSI provides two primary capabilities--runtime abstraction layer and topology data structures.
Execute BEE workflows on private cloud infrastructure (STNS01-22 BEE - FY21 P6-2)
Scope and objectives: BEE provides a portable, modular, HPC-focused workflow engine capable of managing containerized applications at scale. In FY21 BEE will expand its capabilities to provide more sophisticated handling of workflows. The ability to archive, clone, and re-run workflows will be added to BEE. The kinds of resources that BEE can use to execute workflow tasks will be expanded to include public and private clouds, such as Google Cloud Platform and OpenStack.
Execute BEE workflows on private cloud infrastructure-2.3.6.01 - LANL ATDM ST / STNS01-22 Milestone Completion Documentation (BEE-FY21 P6-2) [Slides]
This work involves the creation of the Cloud Launcher, a new subcomponent of BEE, and the extension of the BEETaskManager to run on Cloud systems. BEE will be able to interact with the Google Compute Engine and OpenStack cloud APIs to set up simple Cloud clusters for launching HPC job scripts. BEE will use existing functionality to launch jobs that previously could only be launched on HPC systems. The BEETaskManager will handle launching tasks on the Cloud cluster.
Toward Well-Provenanced Computer System Benchmarking: An Update
This report provides an update on the current design, implementation, and usage of bueno, a Python framework enabled by container technology that supports gradations of reproducibility for wellprovenanced benchmarking of sequential and parallel programs. The ultimate goal of the bueno project is to provide convenient access to mechanisms that aid in the automated generation, collection, and dissemination of data relevant for experimental reproducibility in computer system benchmarking.
Towards memristor supremacy with novel machine learning algorithms
The von Neumann architecture for current computers is slowly showing signs of limitations, due to the finite bus rate between the RAM (memory) and the CPU (computing unit). In-Memory computing is an alternative to this architecture, which however is still in the early stages of development. Memristors are an analog and classical alternative, as these components can act both as a memory, or as a simple computing unit. When wired together, these can in fact be used for a variety of applications. During the course of our research, we have developed an in-depth experimental understanding of both the limitations and the real possibilities of these devices, and developed automated algorithms for their characterization; at the theoretical level, we have incorporated the experimental features to implement novel analog computing units for specific tasks. We have shown that memristors can tackle QUBO and cubic optimization problems, and exhibit tunneling phenomena.
The Applicability of Unit Systems to High-Performance Computing Applications
Dimensional analysis is a key technique used to verify the soundness of scientific models. Most experts agree that engineering and scientific software would be made more reliable by integrating dimensional analysis in their type system. We explored how High Performance Computing (HPC) applications could integrate compile-time dimensional analysis. We started by investigating various implementation of unit systems for C++. Eventually, selecting the latest (and most advanced) one to apply to our test codes. We worked with code of increasing complexity, from a projectile trajectory calculation to the proxy-application Lulesh. This included our code, Springs-3D, which focuses on demonstrating language features while performing simple physic computations. Finally, our main contribution is a source-code analysis which extracts constraints on the dimension of all variables, functions, and constants in an application. This resulting system of equations is solved using the dimensions of a few of these objects. This analysis has the potential to greatly reduce the time spent performing dimensional analysis when refactoring application to use a representation of units.
Adversarial Perturbations Are Not So Weird: Entanglement of Robust and Non-Robust Features in Neural Network Classifiers
Neural networks trained on visual data are well-known to be vulnerable to often imperceptible adversarial perturbations. The reasons for this vulnerability are still being debated in the literature. Recently Ilyas et al. (2019) showed that this vulnerability arises, in part, because neural network classifiers rely on highly predictive but brittle “non-robust” features. In this paper we extend the work of Ilyas et al. by investigating the nature of the input patterns that give rise to these features. In particular, we hypothesize that in a neural network trained in a standard way, non-robust features respond to small, “non-semantic” patterns that are typically entangled with larger, robust patterns, known to be more human-interpretable, as opposed to solely responding to statistical artifacts in a dataset. Thus, adversarial examples can be formed via minimal perturbations to these small, entangled patterns. In addition, we demonstrate a corollary of our hypothesis: robust classifiers are more effective than standard (non-robust) ones as a source for generating transferable adversarial examples in both the untargeted and targeted settings. The results we present in this paper provide new insight into the nature of the non-robust features responsible for adversarial vulnerability of neural network classifiers.
Legacy of Metropolis continues through new materials donated to classified library collections
“From the very beginning of the Los Alamos project, it was inevitable that the Laboratory would suffer total immersion in computing,” said Nicholas Constantine Metropolis in 1976, reflecting his characteristic humor. Metropolis had succinctly summarized the prominent place of computing in Los Alamos’s mission and history. Metropolis himself played no small role in that “total immersion,” exemplified by the Lab’s supercomputing center, a postdoctoral fellowship, and the world-famous algorithm that carry his name. So does a collection of legacy materials in the National Security Research Center (NSRC). The NSRC, the Lab’s classified library, which also houses unclassified artifacts, recently received a new addition to the Metropolis Collections. This donation, 22 years after his death on October 17, 1999, provides tangible evidence of Metropolis’s continuing legacy at Los Alamos.
Partial program correctness
Discussed the preexisting problem of input dependence, interactions with other kinds of observable behavior and lack of debugger support. Also fixed (preexisting) "first" in wording based SGI review.
MTT/Gitlab CI ongoing Open MPI testing - Focus on GPU/OFI and OpenPMIx (Quadchart for ECP OMPI-X milestone STPR17-80)
Further work for this milestone confirmed that using ECP Gitlab CI does not work with PR based work flow used by the Open MPI.
BEE-FY22-1: Enhanced BEE Client 2.3.6.01 - LANL ATDM ST/STNS01-44 Milestone Completion Documentation
Develop a new graphical BEE client for managing workflows.
Non-Proprietary Companion to the Q1 CY2021 PathForward Final Assessment (WBS 2.4.1, Milestone PM-HI-1040)
The PathForward element of the Exascale Computing Project (ECP) Hardware and Integration (HI) focus area has met its goals of preparing the US industry for exascale system procurements and generally improving US competitiveness in the worldwide computing market. This report is a companion to the deliverable for the ECP milestone PM-HI-1040, Assess PathForward Impact Against Exascale Hardware Challenges, which gives a summary of the final status of each PathForward project, describes progress achieved against PathForward contract milestones, and includes a final assessment of each vendor’s progress on key exascale challenges. This companion document details the results of the PathForward research to the extent possible without disclosing proprietary information and the impact on products and US exascale systems. In addition, it captures lessons learned in order to inform future projects in general and in high-performance computing in particular.