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

Abstract

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 NVIDIA's 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.workload, off-loading the non-bonded calculations onto the GPUs provides the best utilization of the resources. On the other hand, the case of the PME method requires the use of FFTs, which results in significant communication between the concurrent elements as well as a slightly smaller requirement of the FP arithmetic. Overall, the results show that a single workstation with a total of eight Figure 2. Improvement in time-to-solution with GPU-enabled LAMMPS for the cut-off based MD simulations. Tests were performed with Tesla C1060 GPUs and hardware configuration B (Table II); all tests were performed for 1000 steps with 2 fs time-step (time-to-solution for CPU-only version in the inset). 1364 P. K. AGARWAL ET AL.CPU-cores (two quad-core processors) and one, two and four...

High performance computing and simulation: architectures, systems, algorithms, technologies, services, and applications

“…Two papers in this category present research concerned with computational biology: the first one uses Field‐Programmable Gate Arrays (FPGAs) and Application‐Specific Integrated Circuits (ASICs) for DNA sequence alignment , and the second solves large plant motif problems with GPU computing . Still, with hybrid computing porting of applications for GPU, accelerators have been achieved for a molecular dynamic framework and for the dynamics part of a weather forecasting package . The second group has two papers that deal with methodologies and algorithms : in , a cluster‐based algorithm helps determine the catchment basin of rivers in large digital elevation models (DEMs), whereas the second paper proposes an interesting speedup‐test protocol on the basis of well‐known statistical tests .…”

On the benefits of the remote GPU virtualization mechanism: The rCUDA case

“…To that end, although programmers must specify which parts of the application are executed on the CPU and which parts are off-loaded to the GPU, the existence of libraries and programming models such as CUDA (Compute Unified Device Architecture) [1] noticeably ease this task. In this context, GPUs significantly reduce the execution time of applications from domains as different as Big Data [2], chemical physics [3], computational algebra [4], image analysis [5], finance [6], and biology [7], for instance.…”

Remote GPU Virtualization: Is It Useful?

“…To that end, although programmers must specify which parts of the application are executed on the CPU and which parts are off-loaded to the GPU, the existence of libraries and programming models such as CUDA (Compute Unified Device Architecture) [11] noticeably ease this task. In this context, GPUs significantly reduce the execution time of applications from domains as different as Big Data [28], chemical physics [21], computational algebra [29], image analysis [17], finance [26], and biology [1] for instance.…”