Parallel Processing and Super-Scalar Operation 405
the two-dimensional array. Note that these smaller processors are interconnected with their immediate
neighbours for communication with each other as well as message passing.
The resultant effect of such a processing unit may be visualized for some typical problems like nite
element analysis or heat transfer. Each processing element can compute the result of an element of a
larger continuum, and moderations at boundaries are easier. However, the need of such con guration is
no more due to the availability of main frames and super computers.
13.7 VECTOR PROCESSOR
Array processors are expensive and generally underutilized for common type of programs. While
performance-wise, almost identical, vector processors are less expensive and, therefore, more popu-
lar for implementation. Before moving to other related details, we shall have a quick discussion about
vectors. For the sake of example, let us consider a very simple case of
A = B × C
Assuming that B = 3 and C = 6, we nd that A = 18. Now, let us consider the following case of
A[i] = B[i] × C[i]
In this case, B is a one-dimensional array of numbers (integer or real) and so also is C. The result, i.e.,
A is also an array of one dimension. For each set of B and C values, we shall obtain a unique A value and
we may continue storing or printing it. In the rst case A, B and C are taken as scalar; while in the sec-
ond case they are considered as vectors. A vector processor is expert in dealing with this type of vector
Figure 13.14 Schematic of an array processor
M13_GHOS1557_01_SE_C13.indd 405M13_GHOS1557_01_SE_C13.indd 405 4/29/11 5:26 PM4/29/11 5:26 PM
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.