Parallel run-time for CO-OPN

Dissertação para obtenção do Grau de Mestre em Engenharia Informática

A system’s approach to cache hierarchy-aware decomposition of data-parallel computations

Dissertação para obtenção do Grau de Mestre em Engenharia Informática

Including the workload effect in the parallel program signature

Performance prediction and application behavior modeling have been the subject of exten- sive research that aim to estimate applications performance with an acceptable precision. A novel approach to predict the performance of parallel applications is based in the con- cept of Parallel Application Signatures that consists in extract an application most relevant parts (phases) and the number of times they repeat (weights). Executing these phases in a target machine and multiplying its exeuction time by its weight an estimation of the application total execution time can be made. One of the problems is that the performance of an application depends on the program workload. Every type of workload affects differently how an application performs in a given system and so affects the signature execution time. Since the workloads used in most scientific parallel applications have dimensions and data ranges well known and the behavior of these applications are mostly deterministic, a model of how the programs workload affect its performance can be obtained. We create a new methodology to model how a program’s workload affect the parallel application signature. Using regression analysis we are able to generalize each phase time execution and weight function to predict an application performance in a target system for any type of workload within predefined range. We validate our methodology using a synthetic program, benchmarks applications and well known real scientific applications.

Data-parallel computation in parallel and distributed environments

The past few decades have seen a considerable increase in the number of parallel and distributed systems. With the development of more complex applications, the need for more powerful systems has emerged and various parallel and distributed environments have been designed and implemented. Each of the environments, including hardware and software, has unique strengths and weaknesses. There is no single parallel environment that can be identified as the best environment for all applications with respect to hardware and software properties. The main goal of this thesis is to provide a novel way of performing data-parallel computation in parallel and distributed environments by utilizing the best characteristics of difference aspects of parallel computing. For the purpose of this thesis, three aspects of parallel computing were identified and studied. First, three parallel environments (shared memory, distributed memory, and a network of workstations) are evaluated to quantify theirsuitability for different parallel applications. Due to the parallel and distributed nature of the environments, networks connecting the processors in these environments were investigated with respect to their performance characteristics. Second, scheduling algorithms are studied in order to make them more efficient and effective. A concept of application-specific information scheduling is introduced. The application- specific information is data about the workload extractedfrom an application, which is provided to a scheduling algorithm. Three scheduling algorithms are enhanced to utilize the application-specific information to further refine their scheduling properties. A more accurate description of the workload is especially important in cases where the workunits are heterogeneous and the parallel environment is heterogeneous and/or non-dedicated. The results obtained show that the additional information regarding the workload has a positive impact on the performance of applications. Third, a programming paradigm for networks of symmetric multiprocessor (SMP) workstations is introduced. The MPIT programming paradigm incorporates the Message Passing Interface (MPI) with threads to provide a methodology to write parallel applications that efficiently utilize the available resources and minimize the overhead. The MPIT allows for communication and computation to overlap by deploying a dedicated thread for communication. Furthermore, the programming paradigm implements an application-specific scheduling algorithm. The scheduling algorithm is executed by the communication thread. Thus, the scheduling does not affect the execution of the parallel application. Performance results achieved from the MPIT show that considerable improvements over conventional MPI applications are achieved.

On-the-Fly Maintenance of Series-Parallel Relationships in Fork-Join Multithreaded Programs

A key capability of data-race detectors is to determine whether one thread executes logically in parallel with another or whether the threads must operate in series. This paper provides two algorithms, one serial and one parallel, to maintain series-parallel (SP) relationships "on the fly" for fork-join multithreaded programs. The serial SP-order algorithm runs in O(1) amortized time per operation. In contrast, the previously best algorithm requires a time per operation that is proportional to Tarjan’s functional inverse of Ackermann’s function. SP-order employs an order-maintenance data structure that allows us to implement a more efficient "English-Hebrew" labeling scheme than was used in earlier race detectors, which immediately yields an improved determinacy-race detector. In particular, any fork-join program running in T₁ time on a single processor can be checked on the fly for determinacy races in O(T₁) time. Corresponding improved bounds can also be obtained for more sophisticated data-race detectors, for example, those that use locks. By combining SP-order with Feng and Leiserson’s serial SP-bags algorithm, we obtain a parallel SP-maintenance algorithm, called SP-hybrid. Suppose that a fork-join program has n threads, T₁ work, and a critical-path length of T[subscript ∞]. When executed on P processors, we prove that SP-hybrid runs in O((T₁/P + PT[subscript ∞]) lg n) expected time. To understand this bound, consider that the original program obtains linear speed-up over a 1-processor execution when P = O(T₁/T[subscript ∞]). In contrast, SP-hybrid obtains linear speed-up when P = O(√T₁/T[subscript ∞]), but the work is increased by a factor of O(lg n).

Autonomous execution of tasks by swarm carrier agents in swarm-array computing

How can a bridge be built between autonomic computing approaches and parallel computing system? The work reported in this paper is motivated towards bridging this gap by proposing swarm-array computing, a novel technique to achieve autonomy for distributed parallel computing systems. Among three proposed approaches, the second approach, namely 'Intelligent Agents' is of focus in this paper. The task to be executed on parallel computing cores is considered as a swarm of autonomous agents. A task is carried to a computing core by carrier. agents and can be seamlessly transferred between cores in the event of a pre-dicted failure, thereby achieving self-ware objectives of autonomic computing. The feasibility of the proposed approach is validated on a multi-agent simulator.

A parallel formulation for the simulation of a generic branch predictor

A parallel formulation for the simulation of a branch prediction algorithm is presented. This parallel formulation identifies independent tasks in the algorithm which can be executed concurrently. The parallel implementation is based on the multithreading model and two parallel programming platforms: pthreads and Cilk++. Improvement in execution performance by up to 7 times is observed for a generic 2-bit predictor in a 12-core multiprocessor system.

Application execution management on the InteGrade opportunistic grid middleware

The InteGrade project is a multi-university effort to build a novel grid computing middleware based on the opportunistic use of resources belonging to user workstations. The InteGrade middleware currently enables the execution of sequential, bag-of-tasks, and parallel applications that follow the BSP or the MPI programming models. This article presents the lessons learned over the last five years of the InteGrade development and describes the solutions achieved concerning the support for robust application execution. The contributions cover the related fields of application scheduling, execution management, and fault tolerance. We present our solutions, describing their implementation principles and evaluation through the analysis of several experimental results. (C) 2010 Elsevier Inc. All rights reserved.

Parallel Cyclostationarity-Exploiting Algorithm for Energy-Efficient Spectrum Sensing

The evolution of wireless communication systems leads to Dynamic Spectrum Allocation for Cognitive Radio, which requires reliable spectrum sensing techniques. Among the spectrum sensing methods proposed in the literature, those that exploit cyclostationary characteristics of radio signals are particularly suitable for communication environments with low signal-to-noise ratios, or with non-stationary noise. However, such methods have high computational complexity that directly raises the power consumption of devices which often have very stringent low-power requirements. We propose a strategy for cyclostationary spectrum sensing with reduced energy consumption. This strategy is based on the principle that p processors working at slower frequencies consume less power than a single processor for the same execution time. We devise a strict relation between the energy savings and common parallel system metrics. The results of simulations show that our strategy promises very significant savings in actual devices.

Parallel implementation of a robust optical flow technique

[EN] The accuracy and performance of current variational optical ow methods have considerably increased during the last years. The complexity of these techniques is high and enough care has to be taken for the implementation. The aim of this work is to present a comprehensible implementation of recent variational optical flow methods. We start with an energy model that relies on brightness and gradient constancy terms and a ow-based smoothness term. We minimize this energy model and derive an e cient implicit numerical scheme. In the experimental results, we evaluate the accuracy and performance of this implementation with the Middlebury benchmark database. We show that it is a competitive solution with respect to current methods in the literature. In order to increase the performance, we use a simple strategy to parallelize the execution on multi-core processors.

Performance evaluation of a parallel algorithm for simultaneous untangling and smoothing of tetrahedral meshes

[EN]A new parallel algorithm for simultaneous untangling and smoothing of tetrahedral meshes is proposed in this paper. We provide a detailed analysis of its performance on shared-memory many-core computer architectures. This performance analysis includes the evaluation of execution time, parallel scalability, load balancing, and parallelism bottlenecks. Additionally, we compare the impact of three previously published graph coloring procedures on the performance of our parallel algorithm. We use six benchmark meshes with a wide range of sizes. Using these experimental data sets, we describe the behavior of the parallel algorithm for different data sizes. We demonstrate that this algorithm is highly scalable when it runs on two different high-performance many-core computers with up to 128 processors...

Parallel vector processing of multidimensional orthogonal transforms for digital signal processing applications

The performance of the parallel vector implementation of the one- and two-dimensional orthogonal transforms is evaluated. The orthogonal transforms are computed using actual or modified fast Fourier transform (FFT) kernels. The factors considered in comparing the speed-up of these vectorized digital signal processing algorithms are discussed and it is shown that the traditional way of comparing th execution speed of digital signal processing algorithms by the ratios of the number of multiplications and additions is no longer effective for vector implementation; the structure of the algorithm must also be considered as a factor when comparing the execution speed of vectorized digital signal processing algorithms. Simulation results on the Cray X/MP with the following orthogonal transforms are presented: discrete Fourier transform (DFT), discrete cosine transform (DCT), discrete sine transform (DST), discrete Hartley transform (DHT), discrete Walsh transform (DWHT), and discrete Hadamard transform (DHDT). A comparison between the DHT and the fast Hartley transform is also included.(34 refs)

Using P-GENESIS for Parallel Simulation of GENESIS Models

P-GENESIS is an extension to the GENESIS neural simulator that allows users to take advantage of parallel machines to speed up the simulation of their network models or concurrently simulate multiple models. P-GENESIS adds several commands to the GENESIS script language that let a script running on one processor execute remote procedure calls on other processors, and that let a script synchronize its execution with the scripts running on other processors. We present here some brief comments on the mechanisms underlying parallel script execution. We also offer advice on parallelizing parameter searches, partitioning network models, and selecting suitable parallel hardware on which to run P-GENESIS.

Improving the efficiency of nondeterministic indepemndent and-parallel systems

We present the design and implementation of the and-parallel component of ACE. ACE is a computational model for the full Prolog language that simultaneously exploits both or-parallelism and independent and-parallelism. A high performance implementation of the ACE model has been realized and its performance reported in this paper. We discuss how some of the standard problems which appear when implementing and-parallel systems are solved in ACE. We then propose a number of optimizations aimed at reducing the overheads and the increased memory consumption which occur in such systems when using previously proposed solutions. Finally, we present results from an implementation of ACE which includes the optimizations proposed. The results show that ACE exploits and-parallelism with high efficiency and high speedups. Furthermore, they also show that the proposed optimizations, which are applicable to many other and-parallel systems, significantly decrease memory consumption and increase speedups and absolute performance both in forwards execution and during backtracking.

Towards execution time estimation in abstract machine-based languages

Abstract machines provide a certain separation between platformdependent and platform-independent concerns in compilation. Many of the differences between architectures are encapsulated in the speciflc abstract machine implementation and the bytecode is left largely architecture independent. Taking advantage of this fact, we present a framework for estimating upper and lower bounds on the execution times of logic programs running on a bytecode-based abstract machine. Our approach includes a one-time, programindependent proflling stage which calculates constants or functions bounding the execution time of each abstract machine instruction. Then, a compile-time cost estimation phase, using the instruction timing information, infers expressions giving platform-dependent upper and lower bounds on actual execution time as functions of input data sizes for each program. Working at the abstract machine level makes it possible to take into account low-level issues in new architectures and platforms by just reexecuting the calibration stage instead of having to tailor the analysis for each architecture and platform. Applications of such predicted execution times include debugging/veriflcation of time properties, certiflcation of time properties in mobile code, granularity control in parallel/distributed computing, and resource-oriented specialization.