An OpenCL micro-benchmark suite for GPUs and CPUs

Open computing language (OpenCL) is a new industry standard for task-parallel and data-parallel heterogeneous computing on a variety of modern CPUs, GPUs, DSPs, and other microprocessor designs. OpenCL is vendor independent and hence not specialized for any particular compute device. To develop efficient OpenCL applications for the particular platform, we still need a more profound understanding of architecture features on the OpenCL model and computing devices. For this purpose, we design and implement an OpenCL micro-benchmark suite for GPUs and CPUs. In this paper, we introduce the implementations of our OpenCL micro benchmarks, and present the measuring results of hardware and software features like performance of mathematical operations, bus bandwidths, memory architectures, branch synchronizations and scalability, etc., on two multi-core CPUs, i.e. AMD Athlon II X2 250 and Intel Pentium Dual-Core E5400, and two different GPUs, i.e. NVIDIA GeForce GTX 460se and AMD Radeon HD 6850. We also compared the measuring results with existing benchmarks to demonstrate the reasonableness and correctness of our benchmark suite.     

Performance,optimization, and fitness: Connecting applications to architectures

Recent trends involving multicore processors and graphical processing units (GPUs) focus on exploiting task‐ and thread‐level parallelism. In this paper, we have analyzed various aspects of the performance of these architectures including NVIDIA GPUs, and multicore processors such as Intel Xeon, AMD Opteron, IBM's Cell Broadband Engine. The case study used in this paper is a biological spiking neural network (SNN), implemented with the Izhikevich, Wilson, Morris–Lecar, and Hodgkin–Huxley neuron models. The four SNN models have varying requirements for communication and computation making them useful for performance analysis of the hardware platforms. We report and analyze the variation of performance with network (problem size) scaling, available optimization techniques and execution configuration. A Fitness performance model, that predicts the suitability of the architecture for accelerating an application, is proposed and verified with the SNN implementation results. The Roofline model, another existing performance model, has also been utilized to determine the hardware bottleneck(s) and attainable peak performance of the architectures. Significant speedups for the four SNN neuron models utilizing these architectures are reported; the maximum speedup of 574x was observed in our GPU implementation. Our results and analysis show that a proper match of architecture with algorithm complexity provides the best performance. Copyright © 2010 John Wiley & Sons, Ltd.     

并行时空处理模型下的快速N-body算法

图形处理器(graphic processing unit,GPU)的最新发展已经能够以低廉的成本提供高性能的通用计算。基于GPU的CUDA(compute unified device architecture)和OpenCL(open computing language)编程模型为程序员提供了充足的类似于C语言的应用程序接口(application programming interface,API),便于程序员发挥GPU的并行计算能力。采用图形硬件进行加速计算,通过一种新的GPU处理模型——并行时间空间模型,对现有GPU上的N-body实现进行了分析,从而提出了一种新的GPU上快速仿真N-body问题的算法,并在AMD的HD Radeon 5850上进行了实现。实验结果表明,相对于CPU上的实现,获得了400倍左右的加速;相对于已有GPU上的实现,也获得了2至5倍的加速。     

在Intel Knights Corner和NVIDIA Kepler架构上OpenACC的性能可移植性分析

OpenACC是一套基于指导语句方式的并行编程语言标准.编程者可以通过在代码中添加符合该标准的指导语句,经OpenACC编译器的编译,将串行代码并行化地移植到加速器或者协处理器上,进而获得异构加速器所带来的加速效果.OpenACC与CUDA和OpenCL这类异构并行编程技术的不同之处在于,它的目的是使编程者在应用移植过程中不需要考虑加速器或协处理器的底层硬件架构,从而降低编程难度.同时它也具有仅需维护一套代码便可在不同硬件平台上运行的优良跨平台性.因此,OpenACC是一个值得研究的并行编程标准.如今的异构加速硬件设备呈现出多元化趋势.在2013年11月的Top500榜单上排名第一的“天河二号”使用了48000块构建在IntelKnights Corner架构之上的协处理器.与此同时,发布不久的NVIDIA公司最新的Kepler架构GPU产品由于多年来的GPU市场积累也迅速形成了可观的用户群体.对于并非追求性能极限的应用移植者而言,寻求应用性能和移植简易性之间的平衡是相当重要的议题.只需要编写一套代码便可运行在这两种硬件平台上的OpenACC正迎合了用户在移植简易性上的需求.解决了移植的简易性之后,同一个应用在不同硬件平台上的性能表现便成了用户最想了解的问题.通过实验和构建性能模型向读者展示使用OpenACC移植的应用在Intel Knights Corner和NVIDIA Kepler架构硬件上的性能可移植性.     

Optimized OpenCL implementation of the Elastodynamic Finite Integration Technique for viscoelastic media

Development of parallel codes that are both scalable and portable for different processor architectures is a challenging task. To overcome this limitation we investigate the acceleration of the Elastodynamic Finite Integration Technique (EFIT) to model 2-D wave propagation in viscoelastic media by using modern parallel computing devices (PCDs), such as multi-core CPUs (central processing units) and GPUs (graphics processing units). For that purpose we choose the industry open standard Open Computing Language (OpenCL) and an open-source toolkit called PyOpenCL. The implementation is platform independent and can be used on AMD or NVIDIA GPUs as well as classical multi-core CPUs. The code is based on the Kelvin–Voigt mechanical model which has the gain of not requiring additional field variables. OpenCL performance can be in principle, improved once one can eliminate global memory access latency by using local memory. Our main contribution is the implementation of local memory and an analysis of performance of the local versus the global memory using eight different computing devices (including Kepler, one of the fastest and most efficient high performance computing technology) with various operating systems. The full implementation of the code is included.     

Stencil computations on heterogeneous platforms for the Jacobi method: GPUs versus Cell BE

We are witnessing the consolidation of the heterogeneous computing in parallel computing with architectures such as Cell Broadband Engine (Cell BE) or Graphics Processing Units (GPUs) which are present in a myriad of developments for high performance computing. These platforms provide a Software Development Kit (SDK) to maximize performance at the expense of dealing with complex and low-level architectural details which makes the software development a daunting task. This paper explores stencil computations in several heterogeneous programming models like Cell SDK, CellSs, ALF and CUDA to optimize the Jacobi method for solving Laplace??s differential equation. We describe the programming techniques to extract the maximum performance on the Cell BE and the GPU, and compare their computing paradigms. Experimental results are shown on two Nvidia Teslas and one IBM BladeCenter QS20 blade which incorporates two 3.2?GHz Cell BEs v?5.1. The speed-up factor for our set of GPU optimizations reaches 3?C4×, and the execution times defeat those of the Cell BE by an order of magnitude, also showing great scalability when moving towards newer GPU generations and/or more demanding problem sizes.     

Evaluations of OpenCL-written tsunami simulation on FPGA and comparison with GPU implementation

When a tsunami occurred on a sea area, prediction of its arrival time is critical for evacuating people from the coastal area. There are many problems related to tsunami to be solved for reducing negative effects of this serious disaster. Numerical modeling of tsunami wave propagation is a computationally intensive problem which needs to accelerate its calculations by parallel processing. The method of splitting tsunami (MOST) is one of the well-known numerical solvers for tsunami modeling. We have developed a tsunami propagation code based on MOST algorithm and implemented different parallel optimizations for GPU and FPGA. In the latest study, we have the best performance of OpenCL kernel which is implemented tsunami simulation on AMD Radeon 280X GPU. This paper targets on design and evaluation on FPGA using OpenCL. The performance on FPGA design generated automatically by Altera offline compiler follows the results of GPU by several kernel modifications.     

基于OpenCL的连续数据无关访存密集型函数并行与优化研究

连续的数据无关是指计算目标矩阵连续的元素时使用的源矩阵元素之间没有关系且也为连续的,访存密集型是指函数的计算量较小,但是有大量的数据传输操作。在OpenCL框架下,以bitwise函数为例,研究和实现了连续数据无关访存密集型函数在GPU平台上的并行与优化。在考察向量化、线程组织方式和指令选择优化等多个优化角度在不同的GPU硬件平台上对性能的影响之后,实现了这个函数的跨平合性能移植。实验结果表明,在不考虑数据传输的前提下,优化后的函数与这个函数在OpenCV库中的CPU版本相比,在AMD HD 5850 GPU达到了平均40倍的性能加速比;在AMD HD 7970 GPU达到了平均90倍的性能加速比;在NVIDIA Tesla 02050 CPU上达到了平均60倍的性能加速比;同时,与这个函数在OpenCV库中的CUDA实现相比,在NVIDIA Tesla 02050平台上也达到了1.5倍的性能加速。     

Graphics Processing Units and Open Computing Language for parallel computing

Graphics Processing Units (GPUs) have become increasingly powerful over the last decade. Programs taking advantage of this architecture can achieve large performance gains and almost all new solutions and initiatives in high performance computing are aimed in that direction. To write programs that can offload the computation onto the GPU and utilize its power, new technologies are needed. The recent introduction of Open Computing Language (OpenCL), a standard for cross-platform, parallel programming of modern processors, has made a step in the right direction. Code written with OpenCL can run on a wide variety of platforms, adapting to the underlying architecture. It is versatile yet easy to learn due to similarities with the C programming language. In this paper, we will review the current state of the art in the use of GPUs and OpenCL for parallel computations. We use an implementation of the n-body simulation to illustrate some important considerations in developing OpenCL programs.     

基于Chan-Vese模型的面向多核CPU和GPU的人脸轮廓提取并行算法

针对人脸轮廓提取中Chan-Vese模型计算量大、分割速度缓慢等问题,采用开放计算语言(OpenCL)并行编程模型,提出了一种基于图形处理器(GPU)和多核CPU加速的并行算法。该算法首先将模型的框架进行重构,消除模型中的数据依赖关系;然后,利用开放计算语言对算法进行并行化以及相应的优化。实验结果表明,与单线程算法相比,在NVIDIA GTX660和AMD FX-8530下达到了较高的加速比。     

dOpenCL: Towards uniform programming of distributed heterogeneous multi-/many-core systems

Modern computer systems become increasingly distributed and heterogeneous by comprising multi-core CPUs, GPUs, and other accelerators. Current programming approaches for such systems usually require the application developer to use a combination of several programming models (e.g., MPI with OpenCL or CUDA) in order to exploit the system’s full performance potential. In this paper, we present dOpenCL (distributed OpenCL)—a uniform approach to programming distributed heterogeneous systems with accelerators. dOpenCL allows the user to run unmodified existing OpenCL applications in a heterogeneous distributed environment. We describe the challenges of implementing the OpenCL programming model for distributed systems, as well as its extension for running multiple applications concurrently. Using several example applications, we compare the performance of dOpenCL with MPI + OpenCL and standard OpenCL implementations.     

Portable Solution for Modeling Compressible Flows on All Existing Hybrid Supercomputers

A variant of a numerical algorithm for simulating viscous gasdynamic flows on unstructured hybrid grids and its software implementation for heterogeneous computations is described. The system of Navier–Stokes equations is approximated by the finite-volume method of an increased approximation order with the values of the variables being defined at the mass centers of the grid elements. The distributed software implementation of the numerical algorithm is adapted to running on hybrid computer systems of various architectures. Comparative implementations were created using the MPI, OpenMP, CUDA, and OpenCL software models permitting the use of multicore processors and various types of accelerators, including NVIDIA and AMD graphics processors, and Intel Xeon Phi multicore coprocessors. The data exchange between MPI processes and between processors and accelerators is carried out simultaneously with the execution of calculations (both in MPI + OpenMP mode and when using CUDA or OpenCL). The indicators of parallel efficiency and performance on systems with different types of computing devices are studied in detail. In the tests, up to 260 GPUs were successfully used.     

OpenCL performance portability for general‐purpose computation on graphics processor units: an exploration on cryptographic primitives

The modern trend toward heterogeneous many‐core architectures has led to high architectural diversity in both high performance and high‐end embedded systems. To effectively exploit the computational resources of such a wide range of architectures, programming languages and APIs such as OpenCL have become increasingly popular. Although OpenCL provides functional code portability and the ability to fine tune the application to the target hardware, providing performance portability is still an open problem. Thus, many research works have investigated the optimization of specific combinations of application and target platform. In this paper, we aim at leveraging the experience obtained in the implementation of algorithms from the cryptography domain to provide a set of guidelines for modern many‐core heterogeneous architecture performance portability and to establish a base on which domain‐specific languages and compiler transformations could be built in the near future. We study algorithmic choices and the effect of compiler transformations on three representative applications in the chosen domain on a set of seven target platforms. To estimate how well the application fits the architecture, we define a metric of computational intensity both for the architecture and the application implementation. Besides being useful to compare either different implementation or algorithmic choices and their fitness to a specific architecture, it can also be useful to the compiler to guide the code optimization process. Copyright © 2014 John Wiley & Sons, Ltd.     

Introducing and Implementing the Allpairs Skeleton for Programming Multi-GPU Systems

Algorithmic skeletons simplify software development: they abstract typical patterns of parallelism and provide their efficient implementations, allowing the application developer to focus on the structure of algorithms, rather than on implementation details. This becomes especially important for modern parallel systems with multiple graphics processing units (GPUs) whose programming is complex and error-prone, because state-of-the-art programming approaches like CUDA and OpenCL lack high-level abstractions. We define a new algorithmic skeleton for allpairs computations which occur in real-world applications, ranging from bioinformatics to physics. We develop the skeleton’s generic parallel implementation for multi-GPU Systems in OpenCL. To enable the automatic use of the fast GPU memory, we identify and implement an optimized version of the allpairs skeleton with a customizing function that follows a certain memory access pattern. We use matrix multiplication as an application study for the allpairs skeleton and its two implementations and demonstrate that the skeleton greatly simplifies programming, saving up to 90 % of lines of code as compared to OpenCL. The performance of our optimized implementation is up to 6.8 times higher as compared with the generic implementation and is competitive to the performance of a manually written optimized OpenCL code.     

Many-Core vs. Many-Thread Machines: Stay Away From the Valley

We study the tradeoffs between many-core machines like Intelpsilas Larrabee and many-thread machines like Nvidia and AMD GPGPUs. We define a unified model describing a superposition of the two architectures, and use it to identify operation zones for which each machine is more suitable. Moreover, we identify an intermediate zone in which both machines deliver inferior performance. We study the shape of this ldquoperformance valleyrdquo and provide insights on how it can be avoided.     

基于OpenCL的拉普拉斯图像增强算法优化研究

OpenCL是面向异构计算平台的通用编程框架,然而由于硬件体系结构的差异,如何在平台间功能移植的基础上实现性能移植仍是有待研究的问题。当前已有算法优化研究一般只针对单一硬件平台,它们很难实现在不同平台上的高效运行。在分析了不同GPU平台底层硬件架构的基础上,从Global Memory的访存效率、GPU计算资源的有效利用率及其硬件资源的限制等多个角度考察了不同优化方法在不同GPU硬件平台上对性能的影响;并在此基础上实现了基于OpenCL的拉普拉斯图像增强算法。实验结果表明,优化后的算法在不考虑数据传输时间的前提下,在AMD和NVIDIA GPU上都取得了3.7～136.1倍、平均56.7倍的性能加速,优化后的kernel比NVIDIA NPP库中相应函数也取得了12.3%～346.7%、平均143.1%的性能提升,验证了提出的优化方法的有效性和性能可移植性。     

Sailfish: A flexible multi-GPU implementation of the lattice Boltzmann method

We present Sailfish, an open source fluid simulation package implementing the lattice Boltzmann method (LBM) on modern Graphics Processing Units (GPUs) using CUDA/OpenCL. We take a novel approach to GPU code implementation and use run-time code generation techniques and a high level programming language (Python) to achieve state of the art performance, while allowing easy experimentation with different LBM models and tuning for various types of hardware. We discuss the general design principles of the code, scaling to multiple GPUs in a distributed environment, as well as the GPU implementation and optimization of many different LBM models, both single component (BGK, MRT, ELBM) and multicomponent (Shan–Chen, free energy). The paper also presents results of performance benchmarks spanning the last three NVIDIA GPU generations (Tesla, Fermi, Kepler), which we hope will be useful for researchers working with this type of hardware and similar codes.     

Emerging Architectures Enable to Boost Massively Parallel Data Mining Using Adaptive Sparse Grids

Gaining knowledge out of vast datasets is a main challenge in data-driven applications nowadays. Sparse grids provide a numerical method for both classification and regression in data mining which scales only linearly in the number of data points and is thus well-suited for huge amounts of data. Due to the recursive nature of sparse grid algorithms and their classical random memory access pattern, they impose a challenge for the parallelization on modern hardware architectures such as accelerators. In this paper, we present the parallelization on several current task- and data-parallel platforms, covering multi-core CPUs with vector units, GPUs, and hybrid systems. We demonstrate that a less efficient implementation from an algorithmical point of view can be beneficial if it allows vectorization and a higher degree of parallelism instead. Furthermore, we analyze the suitability of parallel programming languages for the implementation. Considering hardware, we restrict ourselves to the x86 platform with SSE and AVX vector extensions and to NVIDIA’s Fermi architecture for GPUs. We consider both multi-core CPU and GPU architectures independently, as well as hybrid systems with up to 12 cores and 2 Fermi GPUs. With respect to parallel programming, we examine both the open standard OpenCL and Intel Array Building Blocks, a recently introduced high-level programming approach, and comment on their ease of use. As the baseline, we use the best results obtained with classically parallelized sparse grid algorithms and their OpenMP-parallelized intrinsics counterpart (SSE and AVX instructions), reporting both single and double precision measurements. The huge data sets we use are a real-life dataset stemming from astrophysics and artificial ones, all of which exhibit challenging properties. In all settings, we achieve excellent results, obtaining speedups of up to 188 × using single precision on a hybrid system.     

Modular Neural Tile Architecture for Compact Embedded Hardware Spiking Neural Network

Biologically-inspired packet switched network on chip (NoC) based hardware spiking neural network (SNN) architectures have been proposed as an embedded computing platform for classification, estimation and control applications. Storage of large synaptic connectivity (SNN topology) information in SNNs require large distributed on-chip memory, which poses serious challenges for compact hardware implementation of such architectures. Based on the structured neural organisation observed in human brain, a modular neural networks (MNN) design strategy partitions complex application tasks into smaller subtasks executing on distinct neural network modules, and integrates intermediate outputs in higher level functions. This paper proposes a hardware modular neural tile (MNT) architecture that reduces the SNN topology memory requirement of NoC-based hardware SNNs by using a combination of fixed and configurable synaptic connections. The proposed MNT contains a 16:16 fully-connected feed-forward SNN structure and integrates in a mesh topology NoC communication infrastructure. The SNN topology memory requirement is 50 % of the monolithic NoC-based hardware SNN implementation. The paper also presents a lookup table based SNN topology memory allocation technique, which further increases the memory utilisation efficiency. Overall the area requirement of the architecture is reduced by an average of 66 % for practical SNN application topologies. The paper presents micro-architecture details of the proposed MNT and digital neuron circuit. The proposed architecture has been validated on a Xilinx Virtex-6 FPGA and synthesised using 65 nm low-power CMOS technology. The evolvable capability of the proposed MNT and its suitability for executing subtasks within a MNN execution architecture is demonstrated by successfully evolving benchmark SNN application tasks representing classification and non-linear control functions. The paper addresses hardware modular SNN design and implementation challenges and contributes to the development of a compact hardware modular SNN architecture suitable for embedded applications     

Enabling OpenCL support for GPGPU in Kernel‐based Virtual Machine

The importance of heterogeneous multicore programming is increasing, and Open Computing Language (OpenCL) is an open industrial standard for parallel programming that provides a uniform programming model for programmers to write efficient, portable code for heterogeneous computing devices. However, OpenCL is not supported in the system virtualization environments that are often used to improve resource utilization. In this paper, we propose an OpenCL virtualization framework based on Kernel‐based Virtual Machine with API remoting to enable multiplexing of multiple guest virtual machines (guest VMs) over the underlying OpenCL resources. The framework comprises three major components: (i) an OpenCL library implementation in guest VMs for packing/unpacking OpenCL requests/responses; (ii) a virtual device, called virtio‐CL, that is responsible for the communication between guest VMs and the hypervisor (also called the VM monitor); and (iii) a thread, called CL thread, that is used for the OpenCL API invocation. Although the overhead of the proposed virtualization framework is directly affected by the amount of data to be transferred between the OpenCL host and devices because of the primitive nature of API remoting, experiments demonstrated that our virtualization framework has a small virtualization overhead (mean of 6.8%) for six common device‐intensive OpenCL programs and performs well when the number of guest VMs involved in the system increases. These results indirectly infer that the framework allows for effective resource utilization of OpenCL devices.Copyright © 2012 John Wiley & Sons, Ltd.