Structure of Parallel Computers-Parallel Processing-Handout, Exercises of Parallel Computing and Programming

Download Structure of Parallel Computers-Parallel Processing-Handout and more Exercises Parallel Computing and Programming in PDF only on Docsity!

4 Structure of Parallel Computers In the last chapter we saw how parallelism available at the instruction level can be used extensively in designing processors. This type of parallelism is known in the literature as fine grain parallelism as the smallest grain is an instruction in a program. In this chapter we will examine how processors may be interconnected to build parallel computers which use a coarser grain, for example, threads or processes executing in parallel. 4.1 A GENERALIZED STRUCTURE OF A PARALLEL COMPUTER A parallel computer is defined as: an interconnected set of Processing Elements (PEs) which cooperate by communicating with one another to solve large problems fast. —_ We see from this definition that the keywords which define the structure of a parallel computer are PEs, communication and cooperation. Figure 4.1 shows a generalized structure of a parallel Collection of Processing Elements DM bo Go Go ba ae [re] [re] [Pe] [Pe Joee [re] fo] fe] fo] fe fey COMMUNICATION NETWORK PE : Processing Element CL: Communication Interface VO : InpuvOutput M: PE’s local memory Remarks: PE may have a private memory in which case it is called a Computing Element (CE) Fig. 4.1 A generalized structure of a parallel computer. 97 docsity.co