Showing papers by "Steven J. Plimpton published in 2016"

Resistive memory device requirements for a neural algorithm accelerator

Abstract: Resistive memories enable dramatic energy reductions for neural algorithms. We propose a general purpose neural architecture that can accelerate many different algorithms and determine the device properties that will be needed to run backpropagation on the neural architecture. To maintain high accuracy, the read noise standard deviation should be less than 5% of the weight range. The write noise standard deviation should be less than 0.4% of the weight range and up to 300% of a characteristic update (for the datasets tested). Asymmetric nonlinearities in the change in conductance vs pulse cause weight decay and significantly reduce the accuracy, while moderate symmetric nonlinearities do not have an effect. In order to allow for parallel reads and writes the write current should be less than 100 nA as well.

Direct simulation Monte Carlo investigation of the Rayleigh-Taylor instability

Abstract: The Rayleigh-Taylor instability (RTI) is investigated using the direct simulation Monte Carlo method of molecular gas dynamics. Fully resolved two-dimensional simulations are performed to quantify the growth of flat and single-mode perturbed interfaces between two atmospheric-pressure monatomic gases as a function of the Atwood number and the gravitational acceleration. Future simulations on more extreme computational platforms will enable investigation of the RTI in greater detail.

Oxygen Modulates the Effectiveness of Granuloma Mediated Host Response to Mycobacterium tuberculosis: A Multiscale Computational Biology Approach.

Abstract: Mycobacterium tuberculosis associated granuloma formation can be viewed as a structural immune response that can contain and halt the spread of the pathogen. In several mammalian hosts, including non-human primates, Mtb granulomas are often hypoxic, although this has not been observed in wild type murine infection models. While a presumed consequence, the structural contribution of the granuloma to oxygen limitation and the concomitant impact on Mtb metabolic viability and persistence remains to be fully explored. We develop a multiscale computational model to test to what extent in vivo Mtb granulomas become hypoxic, and investigate the effects of hypoxia on host immune response efficacy and mycobacterial persistence. Our study integrates a physiological model of oxygen dynamics in the extracellular space of alveolar tissue, an agent-based model of cellular immune response, and a systems biology-based model of Mtb metabolic dynamics. Our theoretical studies suggest that the dynamics of granuloma organization mediates oxygen availability and illustrates the immunological contribution of this structural host response to infection outcome. Furthermore, our integrated model demonstrates the link between structural immune response and mechanistic drivers influencing Mtbs adaptation to its changing microenvironment and the qualitative infection outcome scenarios of clearance, containment, dissemination, and a newly observed theoretical outcome of transient containment. We observed hypoxic regions in the containment granuloma similar in size to granulomas found in mammalian in vivo models of Mtb infection. In the case of the containment outcome, our model uniquely demonstrates that immune response mediated hypoxic conditions help foster the shift down of bacteria through two stages of adaptation similar to the in vitro non-replicating persistence (NRP) observed in the Wayne model of Mtb dormancy. The adaptation in part contributes to the ability of Mtb to remain dormant for years after initial infection.

Increasing molecular dynamics simulation rates with an 8-fold increase in electrical power efficiency

Abstract: Electrical power efficiency is a primary concern in designing modern HPC systems. Common strategies to improve CPU power efficiency rely on increased parallelism within a processor that is enabled both by an increase in the vector capabilities within the core and also the number of cores within a processor. Although many-core processors have been available for some time, achieving power-efficient performance has been challenging due to the offload model. Here, we evaluate performance of the molecular dynamics code LAMMPS on two new Intel® processors including the second generation many-core Intel® Xeon Phi™ processor that is available as a bootable CPU. We describe our approach to measure power consumption out-of-band and software optimizations necessary to achieve energy efficiency. We analyze benefits from Intel® Advanced Vector Extensions 512 instructions and demonstrate increased simulations rates with over 9X the CPU+DRAM power efficiency when compared to the unoptimized code on previous generation processors.

Validation simulations of the DSMC code SPARTA

Abstract: The Direct Simulation Monte Carlo (DSMC) method has been used for over 50 years to simulate rarified gas flows for a wide range of applications where continuum assumptions fail. Extensive efforts to validate and verify the DSMC method are essential to establishing confidence in the technique and to promoting its broader application. Here, validation results for two cases are presented using a recently developed, open-source DSMC code called SPARTA.

Direct Simulation Monte Carlo Investigation of Hydrodynamic Instabilities in Gases.

Abstract: The Rayleigh-Taylor instability (RTI) is investigated using the Direct Simulation Monte Carlo (DSMC) method of molecular gas dynamics. Here, two-dimensional and three-dimensional DSMC RTI simulations are performed to quantify the growth of flat and single-mode-perturbed interfaces between two atmospheric-pressure monatomic gases. The DSMC simulations reproduce all qualitative features of the RTI and are in reasonable quantitative agreement with existing theoretical and empirical models in the linear, nonlinear, and self-similar regimes. At late times, the instability is seen to exhibit a self-similar behavior, in agreement with experimental observations. For the conditions simulated, diffusion can influence the initial instability growth significantly.

Software from "Increasing Molecular Dynamics Simulation Rates with an 8-Fold Increase in Electrical Power Efficiency"

Abstract: Electrical power efficiency is a primary concern in designing modern HPC systems. Common strategies to improve CPU power efficiency rely on increased parallelism within a processor that is enabled both by an increase in the vector capabilities within the core and also the number of cores within a processor. Although many-core processors have been available for some time, achieving power-efficient performance has been challenging due to the offload model. Here, we evaluate performance of the molecular dynamics code LAMMPS on two new Intel® processors including the second generation many-core Intel® Xeon Phi™ processor that is available as a bootable CPU. We describe our approach to measure power consumption out-of-band and software optimizations necessary to achieve energy efficiency. We analyze benefits from Intel® Advanced Vector Extensions 512 instructions and demonstrate increased simulations rates with over 9X the CPU+DRAM power efficiency when compared to the unoptimized code on previous generation processors.