Paul S. Crozier

Improving performance via mini-applications.

Application performance is determined by a combination of many choices: hardware platform, runtime environment, languages and compilers used, algorithm choice and implementation, and more. In this complicated environment, we find that the use of mini-applications - small self-contained proxies for real applications - is an excellent approach for rapidly exploring the parameter space of all these choices. Furthermore, use of mini-applications enriches the interaction between application, library and computer system developers by providing explicit functioning software and concrete performance results that lead to detailed, focused discussions of design trade-offs, algorithm choices and runtime performance issues. In this paper we discuss a collection of mini-applications and demonstrate how we use them to analyze and improve application performance on new and future computer platforms.

Capture of Volatile Iodine, a Gaseous Fission Product, by Zeolitic Imidazolate Framework-8

Here we present detailed structural evidence of captured molecular iodine (I(2)), a volatile gaseous fission product, within the metal-organic framework ZIF-8 [zeolitic imidazolate framework-8 or Zn(2-methylimidazolate)(2)]. There is worldwide interest in the effective capture and storage of radioiodine, as it is both produced from nuclear fuel reprocessing and also commonly released in nuclear reactor accidents. Insights from multiple complementary experimental and computational probes were combined to locate I(2) molecules crystallographically inside the sodalite cages of ZIF-8 and to understand the capture of I(2) via bonding with the framework. These structural tools included high-resolution synchrotron powder X-ray diffraction, pair distribution function analysis, and molecular modeling simulations. Additional tests indicated that extruded ZIF-8 pellets perform on par with ZIF-8 powder and are industrially suitable for I(2) capture.

Atomistic simulations of biologically realistic transmembrane potential gradients.

We present all-atom molecular dynamics simulations of biologically realistic transmembrane potential gradients across a DMPC bilayer. These simulations are the first to model this gradient in all-atom detail, with the field generated solely by explicit ion dynamics. Unlike traditional bilayer simulations that have one bilayer per unit cell, we simulate a 170 mV potential gradient by using a unit cell consisting of three salt-water baths separated by two bilayers, with full three-dimensional periodicity. The study shows that current computational resources are powerful enough to generate a truly electrified interface, as we show the predicted effect of the field on the overall charge distribution. Additionally, starting from Poissons equation, we show a new derivation of the double integral equation for calculating the potential profile in systems with this type of periodicity.

Towards De Novo Design of Deoxyribozyme Biosensors for GMO Detection

Hybrid systems that provide a seamless interface between nanoscale molecular events and microsystem technologies enable the development of complex biological sensor systems that not only detect biomolecular threats, but also are able to determine and execute a programmed response to such threats. The challenge is to move beyond the current paradigm of compartmentalizing detection, analysis, and interpretation into separate steps. We present methods that will enable the de novo design and development of customizable biosensors that can exploit deoxyribozyme computing (Stojanovic and Stefanovic, 2003) to concurrently perform in vitro target detection, genetically modified organism detection, and classification.

The development of Mellanox/NVIDIA GPUDirect over InfiniBand--a new model for GPU to GPU communications

The usage and adoption of General Purpose GPUs (GPGPU) in HPC systems is increasing due to the unparalleled performance advantage of the GPUs and the ability to fulfill the ever-increasing demands for floating points operations. While the GPU can offload many of the application parallel computations, the system architecture of a GPU-CPU-InfiniBand server does require the CPU to initiate and manage memory transfers between remote GPUs via the high speed InfiniBand network. In this paper we introduce for the first time a new innovative technology—GPUDirect that enables Tesla GPUs to transfer data via InfiniBand without the involvement of the CPU or buffer copies, hence dramatically reducing the GPU communication time and increasing overall system performance and efficiency. We also explore for the first time the performance benefits of GPUDirect using Amber and LAMMPS applications.

Simulations of single grafted polyelectrolyte chains: ssDNA and dsDNA

The structure of a single, grafted polyelectrolyte, DNA, is investigated by molecular dynamics simulations. The polyelectrolyte is treated as a bead–spring model with explicit charges using parametrizations of both flexible (ssDNA) and stiff (dsDNA) polyelectrolytes. In this single chain limit with no added salt, the flexible ssDNA is always highly extended. Counterion condensation on both molecules is found to be chain length dependent. The counterion distribution is not localized to the chain volume as in related polyelectrolyte brush states. Even at large chain lengths, where the majority of counterions are condensed, a significant fraction of counterions reside far from the chain. The distributions of positions of the nongrafted end monomer for ssDNA and dsDNA differ significantly, indicating a possibility for distinguishing the two states in DNA array technologies.

How a small change in retinal leads to G-protein activation: initial events suggested by molecular dynamics calculations.

Rhodopsin is the prototypical G‐protein coupled receptor, coupling light activation with high efficiency to signaling molecules. The dark‐state X‐ray structures of the protein provide a starting point for consideration of the relaxation from initial light activation to conformational changes that may lead to signaling. In this study we create an energetically unstable retinal in the light activated state and then use molecular dynamics simulations to examine the types of compensation, relaxation, and conformational changes that occur following the cis–trans light activation. The results suggest that changes occur throughout the protein, with changes in the orientation of Helices 5 and 6, a closer interaction between Ala 169 on Helix 4 and retinal, and a shift in the Schiff base counterion that also reflects changes in sidechain interactions with the retinal. Taken together, the simulation is suggestive of the types of changes that lead from local conformational change to light‐activated signaling in this prototypical system. Proteins 2007.

An energy-conserving two-temperature model of radiation damage in single-component and binary Lennard-Jones crystals

Two-temperature models are used to represent the interaction between atoms and free electrons during thermal transients such as radiation damage, laser heating, and cascade simulations. In this paper, we introduce an energy-conserving version of an inhomogeneous finite reservoir two-temperature model using a Langevin thermostat to communicate energy between the electronic and atomic subsystems. This energy-conserving modification allows the inhomogeneous two-temperature model to be used for longer and larger simulations and simulations of small energy phenomena, without introducing nonphysical energy fluctuations that may affect simulation results. We test this model on the annealing of Frenkel defects. We find that Frenkel defect annealing is largely indifferent to the electronic subsystem, unless the electronic subsystem is very tightly coupled to the atomic subsystem. We also consider radiation damage due to local deposition of heat in two idealized systems. We first consider radiation damage in a large face-centered-cubic Lennard-Jones (LJ) single-component crystal that readily recrystallizes. Second, we consider radiation damage in a large binary glass-forming LJ crystal that retains permanent damage. We find that the electronic subsystem parameters can influence the way heat is transported through the system and have a significant impact on the number of defects after the heat deposition event. We also find that the two idealized systems have different responses to the electronic subsystem. The single-component LJ system anneals most rapidly with an intermediate electron-ion coupling and a high electronic thermal conductivity. If sufficiently damaged, the binary glass-forming LJ system retains the least permanent damage with both a high electron-ion coupling and a high electronic thermal conductivity. In general, we find that the presence of an electronic gas can affect short and long term material annealing.

Optimal Utilization of Heterogeneous Resources for Biomolecular Simulations

Biomolecular simulations have traditionally benefited from increases in the processor clock speed and coarse-grain inter-node parallelism on large-scale clusters. With stagnating clock frequencies, the evolutionary path for performance of microprocessors is maintained by virtue of core multiplication. Graphical processing units (GPUs) offer revolutionary performance potential at the cost of increased programming complexity. Furthermore, it has been extremely challenging to effectively utilize heterogeneous resources (host processor and GPU cores) for scientific simulations, as underlying systems, programming models and tools are continually evolving. In this paper, we present a parametric study demonstrating approaches to exploit resources of heterogeneous systems to reduce time-to-solution of a production-level application for biological simulations. By overlapping and pipelining computation and communication, we observe up to 10-fold application acceleration in multi-core and multi-GPU environments illustrating significant performance improvements over code acceleration approaches, where the host-to-accelerator ratio is static, and is constrained by a given algorithmic implementation.

Performance modeling of microsecond scale biological molecular dynamics simulations on heterogeneous architectures

Performance improvements in biomolecular simulations based on molecular dynamics (MD) codes are widely desired. Unfortunately, the factors, which allowed past performance improvements, particularly the microprocessor clock frequencies, are no longer increasing. Hence, novel software and hardware solutions are being explored for accelerating performance of widely used MD codes. In this paper, we describe our efforts on porting, optimizing and tuning of Large‐scale Atomic/Molecular Massively Parallel Simulator, a popular MD framework, on heterogeneous architectures: multi‐core processors with graphical processing unit (GPU) accelerators. Our implementation is based on accelerating the most computationally expensive non‐bonded interaction terms on the GPUs and overlapping the computation on the CPU and GPUs. This functionality is built on top of message passing interface that allows multi‐level parallelism to be extracted even at the workstation level with the multi‐core CPUs and allows extension of the implementation on GPU‐enabled clusters. We hypothesize that the optimal benefit of heterogeneous architectures for applications will come by utilizing all possible resources (for example, CPU‐cores and GPU devices on GPU‐enabled clusters). Benchmarks for a range of biomolecular system sizes are provided, and an analysis is performed on four generations of NVIDIAs GPU devices. On GPU‐enabled Linux clusters, by overlapping and pipelining computation and communication, we observe up to 10‐folds application acceleration in multi‐core and multi‐GPU environments illustrating significant performance improvements. Detailed analysis of the implementation is presented that allows identification of bottlenecks in algorithm, indicating that code optimization and improvements on GPUs could allow microsecond scale simulation throughput on workstations and inexpensive GPU clusters, putting widely desired biologically relevant simulation time‐scales within reach of a large user community. In order to systematically optimize simulation throughput and to enable performance prediction, we have developed a parameterized performance model that will allow developers and users to explore the performance potential of future heterogeneous systems for biological simulations. Copyright