4. In this question, we consider the relative performance obtained by running a particular program with various caches installed between the memory and the CPU. Consider the following C function: char...

Extracted text: 4. In this question, we consider the relative performance obtained by running a particular program with various caches installed between the memory and the CPU. Consider the following C function: char a [512], b[512] add_arrays () { register int i; for (i = 0; i < 511;="" i="i+1)" {="" a[i+1]="a" [i]="" +="" b[i];="" work!!="" }="" 9="" }="" (the="" register="" keyword="" tells="" the="" c="" compiler="" to="" reserve="" a="" register="" for="" the="" variable="" i;="" let's="" assume="" it="" keeps="" i="" in="" ro.="" what="" does="" this="" mean="" for="" us?="" the="" variable="" i="" is="" not="" stored="" in="" main="" memory-="" it's="" stored="" in="" the="" cpu;="" this="" mean="" that="" it="" never="" has="" to="" be="" pulled="" from="" main="" memory="" into="" the="" cache="" when="" the="" cpu="" accesses="" it.="" the="" char="" keyword="" allocates="" character="" arrays="" at="" fixed="" locations="" in="" memory,="" using="" 1="" byte="" per="" character.)="" recall="" that="" in="" c,="" an="" array="" is="" a="" contiguous="" chunk="" of="" memory.="" therefore,="" in="" this="" code,="" array="" a="" starts="" at="" some="" address="" (say,="" 0x00a),="" and="" ends="" at="" (start="" address)="" +="" n-sizeof="" (char)="" address,="" where="" n="" is="" the="" number="" of="" elements="" and="" sizeof="" (char)="" is="" the="" number="" of="" bytes="" a="" char="" takes="" in="" memory="" (and="" in="" this="" example,="" we're="" using="" char="" because="" it's="" 1="" byte="" and="" makes="" calculating="" simple!).="" for="" the="" following="" questions,="" assume="" that="" the="" memory="" can="" read="" or="" write="" 1="" byte="" at="" a="">


Extracted text: (f) Suppose we use a 2-way, 256-set (that is, 512-byte) set-associative cache with a block size of 1 byte. The timing specifications are the same as in Part (c). A quasi-LRU replacement strategy is used, replacing the least recently used element within the set. Does this cache help any? What is the average memory overhead now for this instruction? Explain.