Partial Differential Equations (PDE) are the heart of most simulations in many scientific fields, from Fluid Mechanics to Astrophysics. One the most popular mathematical schemes to solve a PDE is Finite Difference (FD...
详细信息
Partial Differential Equations (PDE) are the heart of most simulations in many scientific fields, from Fluid Mechanics to Astrophysics. One the most popular mathematical schemes to solve a PDE is Finite Difference (FD). In this work we map a PDE-FD algorithm called Reverse Time Migration to a GPU using CUDA. This seismic imaging (Geophysics) algorithm is widely used in the oil industry. GPUs are natural contenders in the aftermath of the clock race, in particular for High-performance Computing (HPC). Due to GPU characteristics, the parallelism paradigm shifts from the classical threads plus SIMD to Single Program Multiple Data (SPMD). The NVIDIA GTX 280 implementation outperforms homogeneous CPUs up to 9x (Intel Harpertown E5420) and up to 14x (IBM PPC 970). These preliminary results confirm that GPUs are a real option for HPC, from performance to programmability.
For understanding the performance differences caused by different computer architectures and optimization technologies, the paper characterizes performance of SPEC CPU2006 benchmarks both on Intel and AMD platform. Us...
详细信息
For understanding the performance differences caused by different computer architectures and optimization technologies, the paper characterizes performance of SPEC CPU2006 benchmarks both on Intel and AMD platform. Using the performance events collected by Performance Monitor Unit, we compare their performance differences and analyze the causes from the features of architectures and the optimization technologies. Our results and analysis help the programmers better understand the memory performance and branch prediction performance which are caused by architectures and guide them in the optimization. The analysis also help the processor architects better balance among the key architecture factors such as memory disambiguation and hardware prefetch when doing architecture design decision.
作者:
Robin LampertDavid HamiltonElliot SolowayYale University
Department of Computer Science Cognition and Programming Project P. O. Box 2158 Yale Station New Haven CT 06520 IBM
Systems Integration Division 3700 Bay Area Blvd. Houston TX 77058 U.S.A.
In an effort to learn more about how testers test code, we observed several testers while they designed tests to check a change which had been made to code. Using a case study methodology, we gathered empirical data f...
In an effort to learn more about how testers test code, we observed several testers while they designed tests to check a change which had been made to code. Using a case study methodology, we gathered empirical data from the ‘real world’—professional testers, and actual software products. We found that testers do share some common work patterns. These patterns can be seen in their information gathering, their use of heuristics and their construction of mental models. This work is extremely knowledge intensive, experience appears to have a useful effect. In this paper we will discuss how we collected and analysed our data. Then we will present some of our observations about how the testers gathered information, used heuristics, formed mental models and were affected by their previous experience in the course of designing their test scenarios. Based on these observations we comment on training and tools for testers.
The present paper contains a study regarding importance stability in the mechanical systems, mechanical systems are emerging under mechanical vibrations action, it referring especially to the human hand–arm system. M...
Different researches have shown that long–term exposure to whole–body vibration can induce different injuries like back pain, injuries of the different part of the body, disturbing the physical and intellectual acti...
Two types of tin oxide ceramic anodes with 3 different bulk densities are prepared and tested in a laboratory aluminium reduction cell. During normal electrolysis, the regular chemical dissolution process of anode mat...
This study investigates the computer perceptions of students at home, at school and at internet cafe, and the computer applications they preferred. For this purpose, the frequency of computer usage, the applications t...
详细信息
The automatic programming system has been considered by means of which it becomes easier to carry out traditional programming stages. There is discussed both recursive forms: parallel, interrecursion and recursion of ...
详细信息
The automatic programming system has been considered by means of which it becomes easier to carry out traditional programming stages. There is discussed both recursive forms: parallel, interrecursion and recursion of high level that exist for functional programming languages and induction methods for the purpose of their verification. The way how to present imperative languages easy and double cycles by means of recursion forms is shown, the possibility of verification has been studied for each recursion form.
Pervasive computing in mobile ad hoc networks requires that applications query their network environment for services and react to a plethora of events fired by other devices in that network responding to such queries...
详细信息
ISBN:
(纸本)9781605583648
Pervasive computing in mobile ad hoc networks requires that applications query their network environment for services and react to a plethora of events fired by other devices in that network responding to such queries. Current context-aware and event-driven architectures require the programmer to react to these events via a carefully crafted network of observers and event handlers, while inherently introducing complex concurrency issues. This paper proposes the integration of two techniques to solve these problems: ambient references and reactive programming. Ambient references are object-oriented communication abstractions that represent nearby remote objects in the mobile network and that make it possible to generate the events mentioned above. The reactive programming paradigm provides an abstraction over events such that the application can be written in a conventional programming style with explicit control flow. Copyright 2008 ACM.
Extensions to the actor model have been proposed to ease the development of mobile ad hoc applications. However, programming in the actor model is still difficult as it does not provide abstractions to synchronously c...
详细信息
ISBN:
(纸本)9781605583648
Extensions to the actor model have been proposed to ease the development of mobile ad hoc applications. However, programming in the actor model is still difficult as it does not provide abstractions to synchronously coordinate multiple actors. Thus, when programmers want to coordinate two or more actors they have to do it by hand. Additionally, programmers need to write failure recovery code for failures caused by the concurrent access to the actors they are using. Coding this manually is error prone and might even be not feasible. We propose a language abstraction based on software transactional memory to coordinate actors. We show that the integration of software transactional memory with the actor model makes writing failure recovery code superuous. Moreover, we show that our system even handles those cases where manually writing the failure recovery code is not feasible. Copyright 2008 ACM.
暂无评论