We advocate the Loop-of-stencil-reduce pattern as a means of simplifying the implementation of data-parallel programs on heterogeneous multi-core platforms. Loop-of-stencil-reduce is general enough to subsume map, reduce, map-reduce, stencil, stencil-reduce, and, crucially, their usage in a loop in both data-parallel and streaming applications, or a combination of both. The pattern makes it possible to deploy a single stencil computation kernel on different GPUs. We discuss the implementation of Loop-of-stencil-reduce in FastFlow, a framework for the implementation of applications based on the parallel patterns. Experiments are presented to illustrate the use of Loop-of-stencil-reduce in developing data-parallel kernels running on heterogeneous systems.
Spatial computing devices have been shown to significantly accelerate stencil computations, but have so far relied on unrolling the iterative dimension of a single stencil operation to increase temporal locality. This work considers the general case of mapping directed acyclic graphs of heterogeneous stencil computations to spatial computing systems, assuming large input programs without an iterative component. StencilFlow maximizes temporal locality and ensures deadlock freedom in this setting, providing end-to-end analysis and mapping from a high-level program description to distributed hard…
Block iterative methods are extremely important as smoothers for multigrid methods, as preconditioners for Krylov methods, and as solvers for diagonally dominant linear systems. Developing robust and efficient algorithms suitable for current and evolving GPU and multicore CPU systems is a significant challenge. We address this issue in the case of constant-coefficient stencils arising in the solution of elliptic partial differential equations on structured 3D uniform and adaptively refined grids. Robust, highly parallel implementations of block Jacobi and chaotic block Gauss-Seidel algorithms …
The increase in HPC systems size and complexity, together with increasing on-chip transistor density, power limitations, and number of components, render modern HPC systems subject to soft errors. Silent data corruptions (SDCs) are typically caused by such soft errors in the form of bit-flips in the memory subsystem and hinder the correctness of scientific applications. This work addresses the problem of protecting a class of iterative computational kernels, called stencils, against SDCs when executing on parallel HPC systems. Existing SDC detection and correction methods are in general either…
Stencils represent a class of computational patterns where an output grid point depends on a fixed shape of neighboring points in an input grid. Stencil computations are prevalent in scientific applications engaging a significant portion of supercomputing resources. Therefore, it has been always important to optimize stencil programs for the best performance. A rich body of research has focused on optimizing stencil computations on almost all parallel architectures. Stencil applications have regular dependency patterns, inherent pipeline-parallelism, and plenty of data reuse. This makes these …
Stencil computations are widely used in HPC applications. Today, many HPC platforms use GPUs as accelerators. As a result, understanding how to perform stencil computations fast on GPUs is important. While implementation strategies for low-order stencils on GPUs have been well-studied in the literature, not all of proposed enhancements work well for high-order stencils, such as those used for seismic modeling. Furthermore, coping with boundary conditions often requires different computational logic, which complicates efficient exploitation of the thread-level parallelism on GPUs. In this paper…
The radiation transport equation is a mesoscopic equation in high dimensional phase space. Moment methods approximate it via a system of partial differential equations in traditional space-time. One challenge is the high computational intensity due to large vector sizes (1600 components for P39) in each spatial grid point. In this work, we extend the calculable domain size in 3D simulations considerably, by implementing the StaRMAP methodology within the massively parallel HPC framework NAStJA, which is designed to use current supercomputers efficiently. We apply several optimization technique…
Stencil computation is one of the most important kernels in various scientific and engineering applications. A variety of work has focused on vectorization and tiling techniques, aiming at exploiting the in-core data parallelism and data locality respectively. In this paper, the downsides of existing vectorization schemes are analyzed. Briefly, they either incur data alignment conflicts or hurt the data locality when integrated with tiling. Then we propose a novel transpose layout to preserve the data locality for tiling and reduce the data reorganization overhead for vectorization simultaneou…
The growth of data to be processed in the Oil & Gas industry matches the requirements imposed by evolving algorithms based on stencil computations, such as Full Waveform Inversion and Reverse Time Migration. Graphical processing units (GPUs) are an attractive architectural target for stencil computations because of its high degree of data parallelism. However, the rapid architectural and technological progression makes it difficult for even the most proficient programmers to remain up-to-date with the technological advances at a micro-architectural level. In this work, we present an extension …
We present an efficient approach for writing architecture-agnostic parallel high-performance stencil computations in Julia, which is instantiated in the package ParallelStencil.jl. Powerful metaprogramming, costless abstractions and multiple dispatch enable writing a single code that is suitable for both productive prototyping on a single CPU thread and production runs on multi-GPU or CPU workstations or supercomputers. We demonstrate performance close to the theoretical upper bound on GPUs for a 3-D heat diffusion solver, which is a massive improvement over reachable performance with CUDA.jl …
Imperial College London（帝国理工学院）；Argonne National Laboratory（阿贡国家实验室）
Stencil loops are a common motif in computations including convolutional neural networks, structured-mesh solvers for partial differential equations, and image processing. Stencil loops are easy to parallelise, and their fast execution is aided by compilers, libraries, and domain-specific languages. Reverse-mode automatic differentiation, also known as algorithmic differentiation, autodiff, adjoint differentiation, or back-propagation, is sometimes used to obtain gradients of programs that contain stencil loops. Unfortunately, conventional automatic differentiation results in a memory access p…
Stencil computation is one of the most important kernels in various scientific and engineering applications. A variety of work has focused on vectorization techniques, aiming at exploiting the in-core data parallelism. Briefly, they either incur data alignment conflicts or hurt the data locality when integrated with tiling. In this paper, a novel transpose layout is devised to preserve the data locality for tiling in the data space and reduce the data reorganization overhead for vectorization simultaneously. We then propose an approach of temporal computation folding designed to further reduce…
Modern compute nodes in high-performance computing provide a tremendous level of parallelism and processing power. However, as arithmetic performance has been observed to increase at a faster rate relative to memory and network bandwidths, optimizing data movement has become critical for achieving strong scaling in many communication-heavy applications. This performance gap has been further accentuated with the introduction of graphics processing units, which can provide by multiple factors higher throughput in data-parallel tasks than central processing units. In this work, we explore the com…
In this paper, we address the problem of designing a distributed application meant to run both classical and asynchronous iterations. MPI libraries are very popular and widely used in the scientific community, however asynchronous iterative methods raise non-negligible difficulties about the efficient management of communication requests and buffers. Moreover, a convergence detection issue is introduced, which requires the implementation of one of the various state-of-the-art termination methods, which are not necessarily highly reliable for most computational environments. We propose here an …
We present a scalability study of Golub-Kahan bidiagonalization for the parallel iterative solution of symmetric indefinite linear systems with a 2x2 block structure. The algorithms have been implemented within the parallel numerical library PETSc. Since a nested inner-outer iteration strategy may be necessary, we investigate different choices for the inner solvers, including parallel sparse direct and multigrid accelerated iterative methods. We show the strong and weak scalability of the Golub-Kahan bidiagonalization based iterative method when applied to a two-dimensional Poiseuille flow and…
The advent of asynchronous iterative scheme gives high efficiency to numerical computations. However, it is generally difficult to handle the problems of resource management and convergence detection. This paper uses JACK2, an asynchronous communication kernel library for iterative algorithms, to implement both classical and asynchronous parareal algorithms, especially the latter. We illustrate the measures whereby one can tackle the problems above elegantly for the time-dependent case. Finally, experiments are presented to prove the availability and efficiency of such application.
A parallel implementation of a compatible discretization scheme for steady-state Stokes problems is presented in this work. The scheme uses generalized moving least squares to generate differential operators and apply boundary conditions. This meshless scheme allows a high-order convergence for both the velocity and pressure, while also incorporates finite-difference-like sparse discretization. Additionally, the method is inherently scalable: the stencil generation process requires local inversion of matrices amenable to GPU acceleration, and the divergence-free treatment of velocity replaces …
This article describes a method for creating applications for cluster computing systems using the parallel BSF skeleton based on the original BSF (Bulk Synchronous Farm) model of parallel computations developed by the author earlier. This model uses the master/slave paradigm. The main advantage of the BSF model is that it allows to estimate the scalability of a parallel algorithm before its implementation. Another important feature of the BSF model is the representation of problem data in the form of lists that greatly simplifies the logic of building applications. The BSF skeleton is designed…
The paper is devoted to the development of a methodology for evaluating the scalability of compute-intensive iterative algorithms used in simulating complex physical processes on supercomputer systems. The proposed methodology is based on the BSF (Bulk Synchronous Farm) parallel computation model, which makes it possible to predict the upper scalability bound of an iterative algorithm in early phases of its design. The BSF model assumes the representation of the algorithm in the form of operations on lists using high-order functions. Two classes of representations are considered: BSF-M (Map BS…
This paper examines a new parallel computation model called bulk synchronous farm (BSF) that focuses on estimating the scalability of compute-intensive iterative algorithms aimed at cluster computing systems. In the BSF model, a computer is a set of processor nodes connected by a network and organized according to the master/slave paradigm. A cost metric of the BSF model is presented. This cost metric requires the algorithm to be represented in the form of operations on lists. This allows us to derive an equation that predicts the scalability boundary of a parallel program: the maximum number …
To design efficient parallel algorithms, some recent papers showed that many sequential iterative algorithms can be directly parallelized but there are still challenges in achieving work-efficiency and high-parallelism. Work-efficiency can be hard for certain problems where the number of dependences is asymptotically more than optimal sequential work bound. To achieve high-parallelism, we want to process as many objects as possible in parallel. The goal is to achieve $\tilde{O}(D)$ span for a problem with the deepest dependence length $D$. We refer to this property as round-efficiency. In this…
Recurrent neural networks (RNN) have been successfully applied to various sequential decision-making tasks, natural language processing applications, and time-series predictions. Such networks are usually trained through back-propagation through time (BPTT) which is prohibitively expensive, especially when the length of the time dependencies and the number of hidden neurons increase. To reduce the training time, extreme learning machines (ELMs) have been recently applied to RNN training, reaching a 99\% speedup on some applications. Due to its non-iterative nature, ELM training, when paralleli…
On modern parallel architectures, the cost of synchronization among processors can often dominate the cost of floating-point computation. Several modifications of the existing methods have been proposed in order to keep the communication cost as low as possible. This paper aims at providing a brief overview of recent advances in parallel iterative methods for solving large-scale problems. We refer the reader to the related references for more details on the derivation, implementation, performance, and analysis of these techniques.
Center for Computation；Technology Louisiana State University The STE || AR Group, http://stellar-group.org Baton Rouge, LA, USA；Technology Louisiana State University The STE || AR Group, http://stellar-group.org
This paper proposes a model for specifying data flow based parallel data processing programs agnostic of target Big Data processing frameworks. The paper focuses on the formal abstract specification of non-iterative and iterative programs, generalizing the strategies adopted by data flow Big Data processing frameworks. The proposed model relies on monoid AlgebraandPetri Netstoabstract Big Data processing programs in two levels: a high level representing the program data flow and a lower level representing data transformation operations (e.g., filtering, aggregation, join). We extend the model …
LAAS-CNRS（LAAS-法国国家科学研究中心）；Université de Toulouse（图卢兹大学）；CNRS（法国国家科学研究中心）
We describe several features of parallel or distributed asynchronous iterative algorithms such as unbounded delays, possible out of order messages or flexible communication. We concentrate on the concept of macroiteration sequence which was introduced in order to study the convergence or termination of asynchronous iterations. A survey of asynchronous iterations for convex optimization problems is also presented. Finally, a new result of convergence for parallel or distributed asynchronous iterative algorithms with flexible communication for convex optimization problems and machine learning is…
