Many-Task Computing on Many-Core Architectures

Valero-Lara, Pedro; Nookala, Poornima; Pelayo, Fernando L.; Jansson, Johan; Dimitropoulos, Serapheim; Raicu, Ioan

doi:10.12694/scpe.v17i1.1148

Cited by 14 publications

(13 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This option is motivated by the work of Raicu and his coworkers [43] who find that an MTC application can be efficiently run on a cluster computer. The authors of [53] assess how efficiently MTC applications run on other computer architectures. A cluster computer consists of a large number of so-called personal computers (PCs) that are connected to each other through high speed interconnects.…”

Section: Methodsmentioning

confidence: 99%

Developing political-ecological theory: The need for Many-Task Computing

Haas

2019

Preprint

View full text Add to dashboard Cite

8Models of political-ecological systems can inform policies for managing ecosystems that 9 contain endangered species. One way to increase the credibility of these models is to 10 subject them to a rigorous suite of data-based statistical assessments. Doing so involves 11 statistically estimating the model's parameters, computing confidence intervals for these 12 parameters, determining the model's prediction error rate, and assessing its sensitivity to 13 parameter misspecification. 14 Here, these statistical algorithms along with a method for constructing politically fea-15 sible policies from a statistically fitted model, are coded as JavaSpaces TM programs that 16 run as compute jobs on either supercomputers or a collection of in-house workstations. 17 Several new algorithms for implementing such jobs in distributed computing environments 18 are described. 19 This downloadable code is used to compute each job's output for the management 20 challenge of conserving the East African cheetah (Acinonyx jubatus). This case study shows 21 that the proposed suite of statistical tools can be run on a supercomputer to establish the 22 credibility of a managerially-relevant model of a political-ecological system that contains 23 one or more endangered species. This demonstration means that the new standard of 24 credibility that any political-ecological model needs to meet before being used to inform 25 ecosystem management decisions, is the one given herein. 26 1 27 Keywords: social-ecological systems; ecosystem policymaking; agent-based modeling; 28 high performance computing; sensitivity analysis; statistical estimation of large models 29 1 Introduction 30There is a need to acknowledge the complexity of political-ecological systems and the signif-31 icant challenges to building theories of them [1]. Such systems lie at the interface between 32 social/political science and ecology. The complexity of each of these fields coupled with an 33 additional layer of complexity introduced by the interactions between sociological/political 34 systems and natural systems can result in highly complex system dynamics, i.e., ones that 35 are stiff, nonlinear, and possess feedback loops. For example, Schoon and Van der Leeuw [2] 36 note that systems composed of interacting sociological and ecological subsystems are quick 37 to change and rarely stay in equilibrium for long. Further, many state variables are needed 38 to describe both the decision making processes of the relevant social groups, and the func-39 tioning of the involved ecosystem. A political-ecological system is also referred to as a 40 socio-ecological system or social-ecological system (e.g., see [3]). The former term is em-41 phasized herein because those political actions and processes that drive social movements 42 are often initiated by groups seeking to gain increased political power [4]. Building such 43 models is more than an academic exercise. Indeed, the alarming decline in the planet's 44 biodiversity [5], creates a crucial need for credible politi...

show abstract

Section: Methodsmentioning

confidence: 99%

Developing political-ecological theory: The need for Many-Task Computing

Haas

2019

Preprint

View full text Add to dashboard Cite

show abstract

“…Second approach is functional classification [22]. There are three types of distributed computing systems due to the type of processing: High-throughput computing [28], high-performance computing [29] and multi-task computing [30]. However, the difference between high-performance and multi-task computing is not strict [22].…”

Section: Functional Classificationmentioning

confidence: 99%

“…The process concentrates on cooperation with the file system. Examples of the use of this type of system are: High-performance, distributed databases [22] or search engines [30].…”

mentioning

confidence: 99%

Tree-Like Distributed Computation Environment with Shapp Library

Gałecki

Daszczuk

2020

Information

View full text Add to dashboard Cite

Despite the rapidly growing computing power of computers, it is often insufficient to perform mass calculations in a short time, for example, simulation of systems for various sets of parameters, the searching of huge state spaces, optimization using ant or genetic algorithms, machine learning, etc. One can solve the problem of a lack of computing power through workload management systems used in local networks in order to use the free computing power of servers and workstations. This article proposes raising such a system to a higher level of abstraction: The use in the .NET environment of a new Shapp library that allows remote task execution using fork-like operations from Portable Operating System Interface for UNIX (POSIX) systems. The library distributes the task code, sending static data on which task force is working, and individualizing tasks. In addition, a convenient way of communicating distributed tasks running hierarchically in the Shapp library was proposed to better manage the execution of these tasks. Many different task group architectures are possible; we focus on tree-like calculations that are suitable for many problems where the range of possible parallelism increases as the calculations progress.

show abstract

“…Our code is able to compute a high number of systems (neurons) of any size in one call (CUDA kernel), using one thread per Hines system instead of one CUDA block per system. Although multiple works have explored the use of GPUs to compute multiple independent problems in parallel without transforming the data layout [14][15][16][17], the particular characteristics of the sparsity of the Hines matrices force us to modify the data layout to efficiently exploit the memory hierarchy of the GPUs (coalescing accesses to GPU memory).…”

Section: Motivationmentioning

confidence: 99%

Simulating the behavior of the Human Brain on GPUs

Valero-Lara

Martínez-Pérez

Sirvent

et al. 2018

Oil Gas Sci. Technol. – Rev. IFP Energies nouvelles

Self Cite

View full text Add to dashboard Cite

The simulation of the behavior of the Human Brain is one of the most important challenges in computing today. The main problem consists of finding efficient ways to manipulate and compute the huge volume of data that this kind of simulations need, using the current technology. In this sense, this work is focused on one of the main steps of such simulation, which consists of computing the Voltage on neurons’ morphology. This is carried out using the Hines Algorithm and, although this algorithm is the optimum method in terms of number of operations, it is in need of non-trivial modifications to be efficiently parallelized on GPUs. We proposed several optimizations to accelerate this algorithm on GPU-based architectures, exploring the limitations of both, method and architecture, to be able to solve efficiently a high number of Hines systems (neurons). Each of the optimizations are deeply analyzed and described. Two different approaches are studied, one for mono-morphology simulations (batch of neurons with the same shape) and one for multi-morphology simulations (batch of neurons where every neuron has a different shape). In mono-morphology simulations we obtain a good performance using just a single kernel to compute all the neurons. However this turns out to be inefficient on multi-morphology simulations. Unlike the previous scenario, in multi-morphology simulations a much more complex implementation is necessary to obtain a good performance. In this case, we must execute more than one single GPU kernel. In every execution (kernel call) one specific part of the batch of the neurons is solved. These parts can be seen as multiple and independent tridiagonal systems. Although the present paper is focused on the simulation of the behavior of the Human Brain, some of these techniques, in particular those related to the solving of tridiagonal systems, can be also used for multiple oil and gas simulations. Our studies have proven that the optimizations proposed in the present work can achieve high performance on those computations with a high number of neurons, being our GPU implementations about 4× and 8× faster than the OpenMP multicore implementation (16 cores), using one and two NVIDIA K80 GPUs respectively. Also, it is important to highlight that these optimizations can continue scaling, even when dealing with a very high number of neurons.

show abstract

Many-Task Computing on Many-Core Architectures

Cited by 14 publications

References 33 publications

Developing political-ecological theory: The need for Many-Task Computing

Developing political-ecological theory: The need for Many-Task Computing

Tree-Like Distributed Computation Environment with Shapp Library

Simulating the behavior of the Human Brain on GPUs

Contact Info

Product

Resources

About