Unitizing Things

1. MIPS and MFLOPS are not identical measures, even though they both measure millions of operations per second. Explain why not.
   MIPS is million instructions per second while MFLOPS is million floating-point operations. Since their counting targets are different, they are not identical measures.

2. What is the clock cycle time of a system with a 1.2MHz clock?
   Answer: Since clock cycle time = 1/clock rate, the clock cycle time is 1/(1.2 x $10^{9}$ = 0.83ns. That's right, less than a nanosecond-faster than light can move a meter. Whew!

3. Explain why an increase in performance is good, while an increase in execution time is bad, and vice versa.
   Answer: Execution time is defined as the time between the start and the completion of an event (textbook P.18). An increase in execution time is bad because this means it takes more time to finish a job.

   Performance is the reciprocal of execution time (textbook P.18). Thus, an increase in performance is good because this means an decrease in execution time.

   For example, increasing the clock rate (the number of clock cycles per second) is clearly good because you have more ``pulses'' or opportunities for the computer, a finite state machine, to change state each second. However, this increase in clock rate could be accomplished by decreasing the clock cycle time (the number of seconds between consecutive clock cycles). Thus, you need to be careful with word problems that discuss ``improving'' certain features of an architecture.

4. Express CPU performance in terms of clocks per instruction, number of instructions, and clock rate.

   Answer: From the CPU performance equation, we have

   
   

   $${\rm CPU}_{\rm perf} = \frac{\rm CR}{{\rm IC} \times {\rm CPI}}$$