Parallel implementations of distributed global optimization (DGO) [13] on MP-1 and NCUBE parallel computers revealed an approximate O(n) increase in the performance of this algorithm. Therefore, the implementation of the DGO on parallel processors can remedy the only draw back of this algorithm which is the O(n2) of execution time as the number of the dimensions increase. The speed up factor of the parallel implementations of DGO is measured with respect to the sequential execution time of the identical problem on SPARC IV computer. The best speed up was achieved by the SIMD implementation of the algorithm on the MP-1 with the total speedup of 126 for an optimization problem with n = 9. This optimization problem was distributed across 128 PEs of Mas-Par.
A new technique of global optimization and its applications in particular to neural networks are presented. The algorithm is also compared to other global optimization algorithms such as Gradient descent (GD), Monte Carlo (MC), Genetic Algorithm (GA) and other commercial packages. This new optimization technique proved itself worthy of further study after observing its accuracy of convergence, speed of convergence and ease of use. Some of the advantages of this new optimization technique are listed below: 1. Optimizing function does not have to be continuous or differentiable. 2. No random mec…
Recent years have witnessed a surge of interest in parallel and distributed optimization methods for large-scale systems. In particular, nonconvex large-scale optimization problems have found a wide range of applications in several engineering fields. The design and the analysis of such complex, large-scale, systems pose several challenges and call for the development of new optimization models and algorithms. The major contribution of this paper is to put forth a general, unified, algorithmic framework, based on Successive Convex Approximation (SCA) techniques, for the parallel and distribute…
This paper considers parallel Gröbner bases algorithms on distributed memory parallel computers with multi-core compute nodes. We summarize three different Gröbner bases implementations: shared memory parallel, pure distributed memory parallel and distributed memory combined with shared memory parallelism. The last algorithm, called distributed hybrid, uses only one control communication channel between the master node and the worker nodes and keeps polynomials in shared memory on a node. The polynomials are transported asynchronous to the control-flow of the algorithm in a separate distribute…
Loop scheduling techniques aim to achieve load-balanced executions of scientific applications. Dynamic loop self-scheduling (DLS) libraries for distributed-memory systems are typically MPI-based and employ a centralized chunk calculation approach (CCA) to assign variably-sized chunks of loop iterations. We present a distributed chunk calculation approach (DCA) that supports various types of DLS techniques. Using both CCA and DCA, twelve DLS techniques are implemented and evaluated in different CPU slowdown scenarios. The results show that the DLS techniques implemented using DCA outperform the…
In this paper we present and evaluate a parallel algorithm for solving a minimum spanning tree (MST) problem for supercomputers with distributed memory. The algorithm relies on the relaxation of the message processing order requirement for one specific message type compared to the original GHS (Gallager, Humblet, Spira) algorithm. Our algorithm adopts hashing and message compression optimization techniques as well. To the best of our knowledge, this is the first parallel implementation of the GHS algorithm that linearly scales to more than 32 nodes (256 cores) of Infiniband cluster.
With the advent of multi-core processors and their fast expansion, it is quite clear that {\em parallel computing} is now a genuine requirement in Computer Science and Engineering (and related) curriculum. In addition to the pervasiveness of parallel computing devices, we should take into account the fact that there are lot of existing softwares that are implemented in the sequential mode, and thus need to be adapted for a parallel execution. Therefore, it is required to the programmer to be able to design parallel programs and also to have some skills in moving from a given sequential code to…
Distributed deep learning is becoming increasingly popular due to the expanding demand for computing resources for deep learning models with a larger amount of parameters. Different from traditional training approaches, data-parallel training allows multiple compute nodes to train large deep learning models simultaneously in order to boost the training efficiency. In this paper, we present and compare six strategies for data-parallel training using PyTorch on the language model GPT-2 with 100M parameters using a qualitative approach. These strategies are Single GPU, Single Parameter Server, Di…
We present four high performance hybrid sorting methods developed for various parallel platforms: shared memory multiprocessors, distributed multiprocessors, and clusters taking advantage of existence of both shared and distributed memory. Merge sort, known for its stability, is used to design several of our algorithms. We improve its parallel performance by combining it with Quicksort. We present two models designed for shared memory MIMD (OpenMP): (a) a non-recursive Merge sort and (b) a hybrid Quicksort and Merge sort. The third model presented is designed for distributed memory MIMD (MPI) …
Collaborative filtering is amongst the most preferred techniques when implementing recommender systems. Recently, great interest has turned towards parallel and distributed implementations of collaborative filtering algorithms. This work is a survey of the parallel and distributed collaborative filtering implementations, aiming not only to provide a comprehensive presentation of the field's development, but also to offer future research orientation by highlighting the issues that need to be further developed.
Sorting has been one of the most challenging studied problems in different scientific researches. Although many techniques and algorithms have been proposed on the theory of having efficient parallel sorting implementation, however achieving desired performance on different types of the architectures with large number of processors is still a challenging issue. Maximizing parallelism level in applications can be achieved by minimizing overheads due to load imbalance and waiting time due to memory latencies. In this paper, we present a distributed sorting algorithm implemented in PGX.D, a fast …
Graphs and their traversal is becoming significant as it is applicable to various areas of mathematics, science and technology. Various problems in fields as varied as biochemistry (genomics), electrical engineering (communication networks), computer science (algorithms and computation) can be modeled as Graph problems. Real world scenarios including communities their interconnections and related properties can be studied using graphs. So fast, scalable, low-cost execution of parallel graph algorithms is very important. In this implementation of parallel breadth first search of graphs, we impl…
The implementation of a vast majority of machine learning (ML) algorithms boils down to solving a numerical optimization problem. In this context, Stochastic Gradient Descent (SGD) methods have long proven to provide good results, both in terms of convergence and accuracy. Recently, several parallelization approaches have been proposed in order to scale SGD to solve very large ML problems. At their core, most of these approaches are following a map-reduce scheme. This paper presents a novel parallel updating algorithm for SGD, which utilizes the asynchronous single-sided communication paradigm…
Considering the diverse nature of real-world distributed applications that makes it hard to identify a representative subset of distributed benchmarks, we focus on their underlying distributed algorithms. We present and characterize a new kernel benchmark suite (named IMSuite) that simulates some of the classical distributed algorithms in task parallel languages. We present multiple variations of our kernels, broadly categorized under two heads: (a) varying synchronization primitives (with and without fine grain synchronization primitives); and (b) varying forms of parallelization (data parall…
Triangle counting is a fundamental graph analytic operation that is used extensively in network science and graph mining. As the size of the graphs that needs to be analyzed continues to grow, there is a requirement in developing scalable algorithms for distributed-memory parallel systems. To this end, we present a distributed-memory triangle counting algorithm, which uses a 2D cyclic decomposition to balance the computations and reduce the communication overheads. The algorithm structures its communication and computational steps such that it reduces its memory overhead and includes key optim…
Distributed Transactional Memory (DTM) is an emerging approach to distributed synchronization based on the application of the transaction abstraction to distributed computation. DTM comes in several system models, but the control flow model (CF) is particularly powerful, since it allows transactions to delegate computation to remote nodes as well as access shared data. However, there are no existing CF DTM systems that perform on par with state-of-the-art systems operating in other models. Hence, we introduce a CF DTM synchronization algorithm, OptSVA-CF. It supports fine-grained pessimistic c…
We present DASH, a C++ template library that offers distributed data structures and parallel algorithms and implements a compiler-free PGAS (partitioned global address space) approach. DASH offers many productivity and performance features such as global-view data structures, efficient support for the owner-computes model, flexible multidimensional data distribution schemes and inter-operability with STL (standard template library) algorithms. DASH also features a flexible representation of the parallel target machine and allows the exploitation of several hierarchically organized levels of lo…
This paper presents the design, implementation, and evaluation of the PyTorch distributed data parallel module. PyTorch is a widely-adopted scientific computing package used in deep learning research and applications. Recent advances in deep learning argue for the value of large datasets and large models, which necessitates the ability to scale out model training to more computational resources. Data parallelism has emerged as a popular solution for distributed training thanks to its straightforward principle and broad applicability. In general, the technique of distributed data parallelism re…
This paper presents our recent efforts, zenLDA, an efficient and scalable Collapsed Gibbs Sampling system for Latent Dirichlet Allocation training, which is thought to be challenging that both data parallelism and model parallelism are required because of the Big sampling data with up to billions of documents and Big model size with up to trillions of parameters. zenLDA combines both algorithm level improvements and system level optimizations. It first presents a novel CGS algorithm that balances the time complexity, model accuracy and parallelization flexibility. The input corpus in zenLDA is…
A well known golden rule of journalism (and many other fields too) is that if you want to know the full story about something you have to answer all the five W's (Who, What, When, Where, Why) and the H (How). This extended abstract is about what is missing in parallel and distributed simulation and how this affects its popularity.
In this paper, we study a class of set cover problems that satisfy a special property which we call the {\em small neighborhood cover} property. This class encompasses several well-studied problems including vertex cover, interval cover, bag interval cover and tree cover. We design unified distributed and parallel algorithms that can handle any set cover problem falling under the above framework and yield constant factor approximations. These algorithms run in polylogarithmic communication rounds in the distributed setting and are in NC, in the parallel setting.
We present a novel distributed union-find algorithm that features asynchronous parallelism and k-d tree based load balancing for scalable visualization and analysis of scientific data. Applications of union-find include level set extraction and critical point tracking, but distributed union-find can suffer from high synchronization costs and imbalanced workloads across parallel processes. In this study, we prove that global synchronizations in existing distributed union-find can be eliminated without changing final results, allowing overlapped communications and computations for scalable proce…
Graph coloring is often used in parallelizing scientific computations that run in distributed and multi-GPU environments; it identifies sets of independent data that can be updated in parallel. Many algorithms exist for graph coloring on a single GPU or in distributed memory, but to the best of our knowledge, hybrid MPI+GPU algorithms have been unexplored until this work. We present several MPI+GPU coloring approaches based on the distributed coloring algorithms of Gebremedhin et al. and the shared-memory algorithms of Deveci et al. . The on-node parallel coloring uses implementations in Kokko…
CoInDiVinE is a tool for parallel distributed model checking of interactions among components in hierarchical component-based systems. The tool extends the DiVinE framework with a new input language (component-interaction automata) and a property specification logic (CI-LTL). As the language differs from the input language of DiVinE, our tool employs a new state space generation algorithm that also supports partial order reduction. Experiments indicate that the tool has good scaling properties when run in parallel setting.
Self-stabilizing algorithms are an important because of their robustness and guaranteed convergence. Starting from any arbitrary state, a self-stabilizing algorithm is guaranteed to converge to a legitimate state.Those algorithms are not directly amenable to solving distributed graph processing problems when performance and scalability are important. In this paper, we show the "Abstract Graph Machine" (AGM) model that can be used to convert self-stabilizing algorithms into forms suitable for distributed graph processing. An AGM is a mathematical model of parallel computation on graphs that add…
This paper describes the architectural design as well as key implementation details of the Open Source popt4jlib library (https://githhub.org/ioannischristou/popt4jlib) that contains a fairly large number of meta-heuristic and other exact optimization algorithms parallel/distributed Java implementations. Although we report on speedup and efficiency issues on some of the algorithms in the library, our main concern is to detail the design decisions for the key parallel/distributed infrastructure built into the library, so as to make it easier for developers to develop their own parallel implemen…
Universidad Complutense de Madrid（马德里康普顿斯大学）；The MITRE Corporation（米特雷公司）
Cloud simulation environments today are largely employed to model and simulate complex systems for remote accessibility and variable capacity requirements. In this regard, scalability issues in Modeling and Simulation (M\&S) computational requirements can be tackled through the elasticity of on-demand Cloud deployment. However, implementing a high performance cloud M\&S framework following these elastic principles is not a trivial task as parallelizing and distributing existing architectures is challenging. Indeed, both the parallel and distributed M\&S developments have evolved following sepa…
This paper introduces a novel approach to automatic ahead-of-time (AOT) parallelization and optimization of sequential Python programs for execution on distributed heterogeneous platforms. Our approach enables AOT source-to-source transformation of Python programs, driven by the inclusion of type hints for function parameters and return values. These hints can be supplied by the programmer or obtained by dynamic profiler tools; multi-version code generation guarantees the correctness of our AOT transformation in all cases. Our compilation framework performs automatic parallelization and sophis…
This paper describes the architecture of MOSE (My Own Search Engine), a scalable parallel and distributed engine for searching the web. MOSE was specifically designed to efficiently exploit affordable parallel architectures, such as clusters of workstations. Its modular and scalable architecture can easily be tuned to fulfill the bandwidth requirements of the application at hand. Both task-parallel and data-parallel approaches are exploited within MOSE in order to increase the throughput and efficiently use communication, storing and computational resources. We used a collection of html docume…
Parallel Collaborative ADMM Privacy Computing and Adaptive GPU Acceleration for Distributed Edge Networks
College of Cyber Security, Jinan University（暨南大学网络空间安全学院）；Department of Electrical and Computer Engineering, Western University（韦仕敦大学电气与计算机工程系）；School of Computer Science, Guangzhou University（广州大学计算机科学与工程学院）
Current high-performance computer systems used for scientific computing typically combine shared memory computational nodes in a distributed memory environment. Extracting high performance from these complex systems requires tailored approaches. Task based parallel programming has been successful both in simplifying the programming and in exploiting the available hardware parallelism for shared memory systems. In this paper we focus on how to extend task parallel programming to distributed memory systems. We use a hierarchical decomposition of tasks and data in order to accommodate the differe…
As computer clusters become more common and the size of the problems encountered in the field of AI grows, there is an increasing demand for efficient parallel inference algorithms. We consider the problem of parallel inference on large factor graphs in the distributed memory setting of computer clusters. We develop a new efficient parallel inference algorithm, DBRSplash, which incorporates over-segmented graph partitioning, belief residual scheduling, and uniform work Splash operations. We empirically evaluate the DBRSplash algorithm on a 120 processor cluster and demonstrate linear to super-…
In this tutorial paper, we will firstly review some basic simulation concepts and then introduce the parallel and distributed simulation techniques in view of some new challenges of today and tomorrow. More in particular, in the last years there has been a wide diffusion of many cores architectures and we can expect this trend to continue. On the other hand, the success of cloud computing is strongly promoting the everything as a service paradigm. Is parallel and distributed simulation ready for these new challenges? The current approaches present many limitations in terms of usability and ada…
Heterogeneous clusters with nodes containing one or more accelerators, such as GPUs, have become common. While MPI provides inter-address space communication, and OpenCL provides a process with access to heterogeneous computational resources, programmers are forced to write hybrid programs that manage the interaction of both of these systems. This paper describes an array programming interface that provides users with automatic and manual distributions of data and work. Using work distribution and kernel def and use information, communication among processes and devices in a process is perform…
A viable approach for building large-scale quantum computers is to interlink small-scale quantum computers with a quantum network to create a larger distributed quantum computer. When designing quantum algorithms for such a distributed quantum computer, one can make use of the added parallelization and distribution abilities inherent in the system. An added difficulty to then overcome for distributed quantum computing is that a complex control system to orchestrate the various components is required. In this work, we aim to address these issues. We explicitly define what it means for a quantum…
Generative adversarial networks (GANs) are widely used to learn generative models. GANs consist of two networks, a generator and a discriminator, that apply adversarial learning to optimize their parameters. This article presents a parallel/distributed implementation of a cellular competitive coevolutionary method to train two populations of GANs. A distributed memory parallel implementation is proposed for execution in high performance/supercomputing centers. Efficient results are reported on addressing the generation of handwritten digits (MNIST dataset samples). Moreover, the proposed imple…
A novel approach is presented to teach the parallel and distributed computing concepts of synchronization and remote memory access. The single program multiple data (SPMD) partitioned global address space (PGAS) model presented in this paper uses a procedural programming language appealing to undergraduate students. We propose that the amusing nature of the approach may engender creativity and interest using these concepts later in more sober environments. Specifically, we implement parallel extensions to LOLCODE within a source-to-source compiler sufficient for the development of parallel and…
Deep Neural Networks (DNNs) are becoming an important tool in modern computing applications. Accelerating their training is a major challenge and techniques range from distributed algorithms to low-level circuit design. In this survey, we describe the problem from a theoretical perspective, followed by approaches for its parallelization. We present trends in DNN architectures and the resulting implications on parallelization strategies. We then review and model the different types of concurrency in DNNs: from the single operator, through parallelism in network inference and training, to distri…
As Grids are loosely-coupled congregations of geographically distributed heterogeneous resources, the efficient utilization of the resources requires the support of a sound Performance Prediction System (PPS). The performance prediction of grid resources is helpful for both Resource Management Systems and grid users to make optimized resource usage decisions. There have been many PPS projects that span over several grid resources in several dimensions. In this paper the taxonomy for describing the PPS architecture is discussed. The taxonomy is used to categorize and identify approaches which a…
For parallel breadth first search (BFS) algorithm on large-scale distributed memory systems, communication often costs significantly more than arithmetic and limits the scalability of the algorithm. In this paper we sufficiently reduce the communication cost in distributed BFS by compressing and sieving the messages. First, we leverage a bitmap compression algorithm to reduce the size of messages before communication. Second, we propose a novel distributed directory algorithm, cross directory, to sieve the redundant data in messages. Experiments on a 6,144-core SMP cluster show our algorithm o…
The ability to express a program as a hierarchical composition of parts is an essential tool in managing the complexity of software and a key abstraction this provides is to separate the representation of data from the computation. Many current parallel programming models use a shared memory model to provide data abstraction but this doesn't scale well with large numbers of cores due to non-determinism and access latency. This paper proposes a simple programming model that allows scalable parallel programs to be expressed with distributed representations of data and it provides the programmer …
The 8th International Workshop on Parallel and Distributed Methods in verifiCation (PDMC 2009) took place on November 4, 2009 at the Eindhoven University of Technology, in conjunction with Formal Methods 2009 and other related events for the first time under the heading of Formal Methods Week. This volume contains the final workshop proceedings.
A future is a programming construct designed for concurrent and asynchronous evaluation of code, making it particularly useful for parallel processing. The future package implements the Future API for programming with futures in R. This minimal API provides sufficient constructs for implementing parallel versions of well-established, high-level map-reduce APIs. The future ecosystem supports exception handling, output and condition relaying, parallel random number generation, and automatic identification of globals lowering the threshold to parallelize code. The Future API bridges parallel fron…
MixServe: An Automatic Distributed Serving System for MoE Models with Hybrid Parallelism Based on Fused Communication Algorithm
School of Computer Science and Engineering, Southeast University（东南大学计算机科学与工程学院）；Department of Computer Science and Engineering, The Chinese University of Hong Kong（香港中文大学计算机科学与工程系）
Scripting languages such as Python and R have been widely adopted as tools for the productive development of scientific software because of the power and expressiveness of the languages and available libraries. However, deploying scripted applications on large-scale parallel computer systems such as the IBM Blue Gene/Q or Cray XE6 is a challenge because of issues including operating system limitations, interoperability challenges, parallel filesystem overheads due to the small file system accesses common in scripted approaches, and other issues. We present here a new approach to these problems…
Parallel and distributed application design is a major area of interest in the domain of high performance scientific and industrial computing. Over the years, various approaches have been proposed to aid parallel program developers to modeling their applications. In this paper it will be used some concepts from agile development methodologies and Unified Modeling Language (UML) to modeling parallel and distributed applications. The UML-based approach of this paper describes through different artifacts and graphs the main flows of events in the development of parallel and high performance appli…
Big graphs (networks) arising in numerous application areas pose significant challenges for graph analysts as these graphs grow to billions of nodes and edges and are prohibitively large to fit in the main memory. Finding the number of triangles in a graph is an important problem in the mining and analysis of graphs. In this paper, we present two efficient MPI-based distributed memory parallel algorithms for counting triangles in big graphs. The first algorithm employs overlapping partitioning and efficient load balancing schemes to provide a very fast parallel algorithm. The algorithm scales …
New algorithms and optimization techniques are needed to balance the accelerating trend towards bandwidth-starved multicore chips. It is well known that the performance of stencil codes can be improved by temporal blocking, lessening the pressure on the memory interface. We introduce a new pipelined approach that makes explicit use of shared caches in multicore environments and minimizes synchronization and boundary overhead. For clusters of shared-memory nodes we demonstrate how temporal blocking can be employed successfully in a hybrid shared/distributed-memory environment.
In this paper, we derive and investigate approaches to dynamically load balance a distributed task parallel application software. The load balancing strategy is based on task migration. Busy processes export parts of their ready task queue to idle processes. Idle--busy pairs of processes find each other through a random search process that succeeds within a few steps with high probability. We evaluate the load balancing approach for a block Cholesky factorization implementation and observe a reduction in execution time on the order of 5\% in the selected test cases.
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…
We present a set of programming tools (classes and functions written in C++ and based on Message Passing Interface) for fast development of generic parallel (and non-parallel) lattice simulations. They are collectively called MDP 1.2. These programming tools include classes and algorithms for matrices, random number generators, distributed lattices (with arbitrary topology), fields and parallel iterations. No previous knowledge of MPI is required in order to use them. Some applications in electromagnetism, electronics, condensed matter and lattice QCD are presented.
