For parts A and B of this problem, consider the performance of two x86 processors executing the same program to completion. Processor A, produced by AMD, has a clock rate of 2.5 GHz, and has an average CPI for the given program of 2. Processor B, produced by Intel, has a clock rate of 3 GHz, and has an average CPI for the given program of 1.5.

A. What is the speedup of processor A versus B?

B. If both machines cost the same amount and you are primarily interested in running this program, which processor should you buy?

For the next four parts to this problem consider a program with the following instruction makeup and CPI on a given machine:

C. What is the overall average CPI of the program?

D. If the program is 1,000,000 instructions long (dynamic instruction count), and the clock frequency is 2GHz, how long will it take to execute?

E. Imagine you were to design a new processor to execute the given program. You have the option of either spending effort to improve the memory operation performance by 1 cycle, or to improve the performance of control flow instructions by 3 cycles. Which would be the better option? Use speedup to justify your answer.

F. Imagine you could chose to spend effort improving one of the instruction times above (ie. Simple ALU, Complex ALU, etc. ). Which instruction class would you choose to improve? What is the maximum performance gain (speedup) to be had (imagine they took 0 time)?

Instruction Type Simple ALU (ADD, SUB, ORR, AND, etc.) Complex ALU (MUL, DIV, FP) Memory operations (LDUR, STUR, etc.) Control flow (CBZ, CBNZ, B, BL, BR, etc.) Instruction Frequency | CPI 259 15% 30% 30%