Sie sind auf Seite 1von 2

# Tutorial 1

1. Consider two implementations of a machine, one with and one without special floating-point hardware. Consider a program P, with the following mix of operations: floating-point multiply 10%, floating-point add 15%, floating-point divide 5%, integer instructions 70%. The machine with floating-point hardware MFP requires the following number of clock cycles for each instruction class: floating-point multiply 6, floating-point add 4, floating-point divide 20, integer instructions 2. The machine with no floating-point hardware MNFP has to emulate the floating-point operations using integer instructions. The integer instructions take 2 clock cycles and the number of integer instructions needed to implement each of the floating-point operations is as follows: floating-point multiply 30, floating-point add 20, floating-point divide 50. Both machines have a clock rate of 1GHz. Find the native MIPS and MFLOPS ratings for both the machines.

2. Two programs P1 and P2 are run on two machines M1 and M2 and the following measurements were made: Program P1 P2 Time on M1 10 seconds 3 seconds Time on M2 5 seconds 4 seconds

Program P1

Instruction executed on M1
200 x 10
6

## Instruction executed on M2 160 x 106

Moreover, the clock rates of machines M1 and M2 are 200MHz and 300MHz respectively. Find the clock cycles per instruction (CPI) for program P1 on each of the two machines. Round your answers to 3 decimal places.

(b) Assuming the CPI for program P2 on each machine is the same as the CPI for program P1
computed above, find the instruction count (number of instructions) for program P2 running on each machine. (c) Machine M1 costs \$10,000 and machine M2 costs \$15,000. Now the requirement is that program P1 must be executed 200 times each hour and any remaining time is to be used for running program P2. The performance of the machine is measured by the throughout for P2. Which machine has better performance, by how much, and which one is more cost-effective? Round your answers to 3 decimal places. 3. There are four classes of instructions (A, B, C and D) in a certain instruction set. Consider two different implementations, M1 and M2, of the same instruction set. M1 has a clock rate of 500MHz. The average number of cycles for each instruction class on M1 is as follows:

Class A B C D

## CPI for this class 1 2 3 4

M2 has a clock rate of 750MHz. The average number of cycles for each instruction class on M2 is as follows: Class CPI for this class A 2 B 2 C 4 D 4 (a) Assume that peak performance is defined as the fastest rate that a machine can execute an instruction sequence chosen to maximize that rate. What are the peak performances of M1 and M2 expressed as instructions per second? (b) If the number of instructions executed in a certain program is divided equally among the classes of instructions, how much faster is M2 than M1? 4. Assume that we make an enhancement to a computer that improves some mode of execution by a factor of 10. Enhanced mode is used 50% of the time, measured as a percentage of the execution time when the enhanced mode is in use. Hint: Recall that Amdahls Law depends on the fraction of the original, unenhanced execution time that could make use of enhanced mode. Thus, we cannot directly use this 50% measurement to compute speed up with Amdahls Law. (a) What is the speedup we have obtained from fast mode? (b) What percentage of the original execution time has been converted to fast mode?