Enhancing Applications Performance on Intel Paragon through Dynamic Memory Allocation
نویسنده
چکیده
The Paragon operating system (OS) supports virtual memory (VM). The OS manages virtual memory by performing two services. Firstly, paging-in service pages the executable code from the service node to the compute nodes. This includes the paging-in of empty data corresponding to statically allocated arrays. Secondly, pagingout service is performed by paging the unused part of the OSF server to the boot node to make space available for the user’s executable code. These paging-in and pagingout activities take place simultaneously and drastically degrade the performance of the user code. We have investigated this problem in detail, and found that the dynamic allocation of memory completely eliminates the unnecessary and undesirable effects of paging-in empty data arrays from the service node to the compute nodes and thereby increases the performance of the applications considered in the present work by 30% to 40%.
منابع مشابه
Adaptive resource allocation for embedded parallel applications
Parallel and distributed computer architectures are increasingly being considered for application in a wide variety of computationally intensive embedded systems. Many such applications impose highly dynamic demands for resources (processors, memory, and communication network), because their computations are data-dependent, or because the applications must constantly interact with a rapidly cha...
متن کاملDOLIB: Distributed Object Library
This report describes the use and implementation of DOLIB (Distributed Object Library), a library of routines that emulates global or virtual shared memory on Intel multiprocessor systems. Access to a distributed global array is through explicit calls to gather and scatter. Advantages of using DOLIB include: dynamic allocation and freeing of huge (gigabyte) distributed arrays, both C and FORTRA...
متن کاملMolecular Dynamics for 400 Million Particles with Short-range Interactions
We report the design and performance of a computational molecular dynamics (MD) code for 400 million particles interacting through the standard pairwise 6-12 Lennard-Jones potential on a 1024-node Intel Paragon, a distributed-memory MIMD parallel computer. The initially recorded “particle-step time” was .4 microseconds. A new inter-node communication strategy ensures high parallel efficiency fo...
متن کاملOn Analyzing the Cost of Communication in Networks
s of Talks Jean-Loup Baer On the Performance of Cluster Architectures Yosi Ben-Asher The Partitioned PRAM Model, Re-Doing Basic PRAM Algorithms with Limited Resources Artur Czumaj Adaptitive Allocation Processes Ralf Diekmann Load Balancing Strategies for Scientific Computing Applications Martin Dietzfelbinger On Analyzing the Cost of Communication in Networks Pierre Fraigniaud A General Theory...
متن کاملDistributed- and shared-memory parallelizationsof assignment-based data association formultitarget tracking
To date, there has been a lack of efficient and practical distributedand shared-memory parallelizations of the data association problem for multitarget tracking. Filling this gap is one of the primary focuses of the present work. We begin by describing our data association algorithm in terms of an Interacting Multiple Model (IMM) state estimator embedded into an optimization framework, namely, ...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 1993