17.8 a. Consider a uniprocessor with separate data and instruction caches, with hit ratios of
and respectively. Access time from processor to cache is cclock cycles, and
transfer time for a block between memory and cache is bclock cycles. Let be the
fraction of memory accesses that are for instructions, and is the fraction of dirty
lines in the data cache among lines replaced.Assume a write-back policy and deter
mine the effective memory access time in terms of the parameters just defined.
b. Now assume a bus-based SMP in which each processor has the characteristics of
part (a). Every processor must handle cache invalidation in addition to memory
reads and writes. This affects effective memory access time. Let be the fraction
of data references that cause invalidation signals to be sent to other data caches.
The processor sending the signal requires tclock cycles to complete the invalida
tion operation. Other processors are not involved in the invalidation operation.
Determine the effective memory access time.
17.9 What organizational alternative is suggested by each of the illustrations in Figure 17.24?
17.10 In Figure 17.8, some of the diagrams show horizontal rows that are partially filled. In
other cases, there are rows that are completely blank. These represent two different
types of loss of efficiency. Explain.
17.11 Consider the pipeline depiction in Figure 12.13b, which is redrawn in Figure 17.25a,
with the fetch and decode stages ignored, to represent the execution of thread A. Fig
ure 17.25b illustrates the execution of a separate thread B. In both cases, a simple
pipelined processor is used.
a. Show an instruction issue diagram, similar to Figure 17.8a, for each of the two threads.
b. Assume that the two threads are to be executed in parallel on a chip multiproces
sor, with each of the two processors on the chip using a simple pipeline. Show an