S. Brawer, Introduction to Parallel Programming. Academic Press, 1989.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....and computer graphics, at a level of a single graduate level course in each topic. The reader can gain the necessary graphics background by reading one of several books: Roge85, Fole90, Fole94] Several books provide an appropriate introduction to parallel processing, including the following: [Quin87, Braw89, Alma94, Kuma94]. CHAPTER TWO BACKGROUND This chapter contains some background information for the reader. It first introduces some terminology by describing the graphics pipeline, and also the four types of parallelism possible in graphics algorithms. Additional information on these topics can be found, ....

Brawer, Steven, Introduction to Parallel Programming, Academic Press, San Diego, California, 1989.

....destination. In particular, if the mesh is p p Theta p p, then the time to multicast a message of size m from an edge processor to all the processors along the row or column perpendicular to the edge will be (t s t w m) Theta ( p p Gamma 1) The equivalent of barrier synchronization [4] can be implemented in a mesh at the cost of 4(t s t w ) p p Gamma 1) as follows. First a synchronizing message of one unit is passed along all rows left ward starting at the processors in the right most column. Then processors in the left most vertical column pass a synchronizing message ....

Steven Brawer. Introduction to Parallel Programming. Academic Press, Inc., 1250 Sixth Avenue, San Diego, CA 92101, 1989.

....fork( and wait( are low level synchronization primitives, they need not be used in the restrictive way illustrated in Figure 1.1. However, this disciplined pattern of use, or paradigm , is commonly seen in both the organization of operating systems code as well as in scientific applications [Bra89]. One important reason for the ubiquity of this pattern comes from the presence in high level programming languages of the parallel execution construct cobegin task(1) task(2) task(n) coend . A construct like this is used to specify concurrent execution of the given tasks in ....

....The significant explicit overheads in Figure 1.1 arise from the fork( calls. If a fork( call actually creates a new process on a new processor, rather than reallocating an existing one, the call might require several milliseconds on a typical 1 MIP (million instructions per second) processor [Bra89]. Experiments on the Sequent Symmetry multiprocessor indicate that process creation takes 13 14 milliseconds. Process creation involves such substantial overhead from the need to manage memory tables and create a new address space. If all needed processes are pre created and parked while ....

[Article contains additional citation context not shown here]

Brawer, S. Introduction to Parallel Programming. New York: Academic Press (1989).

....Algorithms An intriguing possibility is that of applying parallel execution of transactions to non database applications, such as the parallelization of numerical computations. Conventionally, such computations are either parallelized by the programmer using parallel programming languages [Bra89] or by so called super compilers, which translate a sequential program into a parallel one [ZC91] In here, an alternative approach is proposed: to decompose numerical algorithms into artificial transactions and subtransactions which are executed by a scheduler. As an example of how to do ....

Steven Brawer. Introduction to Parallel Programming. Academic Press, 1989.

No context found.

S. Brawer, Introduction to Parallel Programming. Academic Press, 1989.

No context found.

S. Brawer, Introduction to Parallel Programming, Academic Press, 1989.

No context found.

Brawer, S., "Introduction to Parallel Programming", Academic Press, Boston, 1989. pp.1-423 QA 76.6 B716 1989.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC