Google+ The STAPL pList | Educating the Next Generation of Leaders in Nuclear Security Sciences
Skip navigation

Citation:

X. Xu, "The STAPL pList," M.S. Thesis, Department of Computer Science and Engineering, Texas A&M University, College Station, TX (2010).

Abstract:

We present the design and implementation of the Standard Template Adaptive Parallel Library (stapl) pList, a parallel container that has the properties of a sequential list, but allows for scalable concurrent access when used in a parallel program. The stapl is a parallel programming library that extends C++ with support for parallelism. stapl provides a collection of distributed data structures (pContainers) and parallel algorithms (pAlgorithms) and a generic methodology for extending them to provide customized functionality. stapl pContainers are thread-safe, concurrent objects, providing appropriate interfaces (pViews) that can be used by generic pAlgorithms.

The pList provides Standard Template Library (stl) equivalent methods, such as insert, erase, and splice, additional methods such as split, and efficient asynchronous (non- locking) variants of some methods for improved parallel performance. List related algorithms such as list ranking, Euler Tour (ET), and its applications to compute tree based functions can be computed efficiently and expressed naturally using the pList.

Lists are not usually considered useful in parallel algorithms because they do not allow random access to its elements. Instead, they access elements through a serializing traversal of the list. Our design of the pList, which consists of a collection of distributed lists (base containers), provides almost random access to its base containers. The degree of parallelism supported can be tuned by setting the numberiv of base containers. Thus, a key feature of the pList is that it offers the advantages of a classical list while enabling scalable parallelism. 

We evaluate the performance of the stapl pList on an IBM Power 5 cluster and on a CRAY XT4 massively parallel processing system. Although lists are generally not considered good data structures for parallel processing, we show that pList methods and pAlgorithms, and list related algorithms such as list ranking and ET technique operating on pLists provide good scalability on more than 16, 000 processors. We also show that the pList compares favorably with other dynamic data structures such as the pVector that explicitly support random access.

See Document

Associated Project(s):

  • SHIELD (Smuggled HEU Interdiction through Enhanced anaLysis and Detection): A Framework for Developing Novel Detection Systems Focused on Interdicting Shielded HEU

  • View Sitemap