Download Vector Models for Data-Parallel Computing (Artificial by Guy E. Blelloch PDF

By Guy E. Blelloch

Vector versions for Data-Parallel Computing describes a version of parallelism that extends and formalizes the Data-Parallel version on which the relationship computing device and different supercomputers are established. It offers many algorithms in accordance with the version, starting from graph algorithms to numerical algorithms, and argues that data-parallel types aren't basically sensible and will be utilized to a shockingly big choice of difficulties, also they are compatible for very-high-level languages and result in a concise and transparent description of algorithms and their complexity. a number of the author's rules were included into the guideline set and into algorithms presently operating at the Connection computing device. The publication contains the definition of a parallel vector computer; an intensive description of the makes use of of the experiment (also referred to as parallel-prefix) operations; the creation of segmented vector operations; parallel facts constructions for timber, graphs, and grids; many parallel computational-geometry, graph, numerical and sorting algorithms; recommendations for compiling nested parallelism; a compiler for Paralation Lisp; and information at the implementation of the test operations. man E. Blelloch is an Assistant Professor of machine technology and a imperative Investigator with the great Compiler and complex Language undertaking at Carnegie Mellon collage. Contents: advent. Parallel Vector versions. The experiment Primitives. Computational-Geometry Algorithms. Graph Algorithms. Numerical Algorithms. Languages and Compilers. Correction-Oriented Languages. knocking down Nested Parallelism. A Compiler for Paralation Lisp. Paralation-Lisp Code. The experiment Vector version. info buildings. imposing Parallel Vector types. imposing the test Operations. Conclusions. word list.

Show description

Read or Download Vector Models for Data-Parallel Computing (Artificial Intelligence Series) PDF

Similar applied mathematicsematics books

The Dead Sea Scrolls After Fifty Years: A Comprehensive Assessment

This quantity is the second one in a chain released to mark the fiftieth anniversary of the invention of the 1st scrolls at Qumran. The two-volume set incorporates a finished diversity of articles overlaying themes which are archaeological, old, literary, sociological, or theological in personality. because the discovery of the 1st scrolls in 1947 an huge variety of reviews were released.

Additional resources for Vector Models for Data-Parallel Computing (Artificial Intelligence Series)

Example text

The work on nested parallelism described in Chapter 10 greatly simplifies the translation. 3 Comparison to Circuit and Network Models In this section we compare the parallel vector models to the boolean-circuit models [117, 26, 37, 38] and to the fixed-network models. Fixed-network models are processor-oriented CHAPTER 2. PARALLEL VECTOR MODELS 28 models in which the processors are restricted to communicate to a fixed set of neighbors. These models include sets of processors connected by butterfly networks [32, 84, 11], shuffle-exchange networks [109], cube-connected-cycles [89], multidimensional grids [115], or pyramids [110].

Simulating multiple elements on each processor is important for two reasons.

1) e is the element complexity and s is the step complexity. I contend that the two limits are a cleaner notation for specifying time complexity even when using a P-RAM model. Primitive Data The P-RAM models only supply atomic values as primitive data—the memory is completely flat—and data structures are built out of these atomic values. The vector models, in contrast, supply a primitive data structure: the vector. To take advantages of the parallel vector primitives, other data structures, such as trees or graphs, should be mapped onto the primitive vectors.

Download PDF sample

Rated 4.83 of 5 – based on 47 votes