The screen you are working with is a array of pixels

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: stores local variables in registers, and thus references to local variables do not require any load or store instructions. Figure 6.46 summarizes the results of our inner loop analysis. Notice that the six versions pair up into three equivalence classes, which we denote by the pair of matrices that are accessed in the inner loop. For example, versions and are members of Class because they reference arrays and (but not ) in their innermost loop. For each class, we have counted the number of loads (reads) and stores (writes) in each inner loop iteration, the number of references to , , and that will miss in the cache in each loop iteration, and the total number of cache misses per iteration. The inner loops of the Class routines (Figure 6.45(a) and (b)) scan a row of array with a stride of 1. Since each cache block holds four doublewords, the miss rate for is 0.25 misses per iteration. On the other hand, the inner loop scans a column of with a stride of Ò. Since Ò is large, each access of array results in a miss, for a total of ½ ¾ misses per iteration. The i...
View Full Document

This note was uploaded on 09/02/2010 for the course ELECTRICAL 360 taught by Professor Schultz during the Spring '10 term at BYU.

Ask a homework question - tutors are online