The same manufacturers introduce new quad core versions of P1 and P2 processors.

Question

The same manufacturers introduce new quad core versions of P1 and P2 processors. NewP1 runs at 3.4GHz and costs $170. NewP2 runs at 2.6GHz and costs $170. We can use two different compilers for these processors. For a specific benchmark program, compiler A generates 2250 instructions per core with an effective CPI of 1.5 for NewP1 and 625 instructions/core and CPI of 3.2 for NewP2. Compiler B generates 1750 instructions/core with a CPI of 1.8 for NewP1 and 500 instructions/core with a CPI of 5.0 for NewP2.
Which processor-compiler combination is the best?
Using the metric Cycles per Second times Cores Used divided by Dollars Squared, which is the best machine? The same manufacturers introduce new quad core versions of P1 and P2 processors. NewP1 runs at 3.4GHz and costs $170. NewP2 runs at 2.6GHz and costs $170. We can use two different compilers for these processors. For a specific benchmark program, compiler A generates 2250 instructions per core with an effective CPI of 1.5 for NewP1 and 625 instructions/core and CPI of 3.2 for NewP2. Compiler B generates 1750 instructions/core with a CPI of 1.8 for NewP1 and 500 instructions/core with a CPI of 5.0 for NewP2.
Which processor-compiler combination is the best?
Using the metric Cycles per Second times Cores Used divided by Dollars Squared, which is the best machine?
Which processor-compiler combination is the best?
Using the metric Cycles per Second times Cores Used divided by Dollars Squared, which is the best machine?

Explanation / Answer

For processor NewP1 and compiler A:

For processor NewP1 and compiler A:

no of instructions/core=2250

CPI=1.5

Cycle time=1/3.4*109 = 294.12 ps(pico seconds)

CPU time=(number of instructions)*(cycles per instruction)*(cycle time)=2250*1.5*294.12=992.65 ns(nano seconds)

For processor NewP2 and compiler A:

no of instructions/core=625

CPI=3.2

Cycle time=1/2.6*109 = 384.62 ps(pico seconds)

CPU time=(number of instructions)*(cycles per instruction)*(cycle time)=625*3.2*384.62=769.23 ns(nano seconds)

For processor NewP1 and compiler B:

no of instructions/core=1750

CPI=1.8

Cycle time=1/3.4*109 = 294.12 ps(pico seconds)

CPU time=(number of instructions)*(cycles per instruction)*(cycle time)=1750*1.8*294.12=926.47 ns(nano seconds)

For processor NewP2 and compiler B:

no of instructions/core=500

CPI=5.0

Cycle time=1/2.6*109 = 384.62 ps(pico seconds)

CPU time=(number of instructions)*(cycles per instruction)*(cycle time)=500*5.0*384.62=961.53 ns(nano seconds)

Of all the combinations Processor NewP2 and compiler A has low CPU time so this combination is the best which have better performance.

Using the metric (Cycles/s)*Cores/(Dollars2 )

NewP1: ((3.4 × 109 ) 4 )/ 1702 = 4.71 × 105 cycles*core seconds*dollars2

NewP2: ((2.6 × 109 ) 4 )/ 1702 = 3.60 × 105 cycles*core seconds*dollars2

Using this metric, NewP1 has the highest value. However, this metric is not very useful because the performance problem(in 1st question) above shows that the NewP2 can run programs faster with the same number of cores and cost. It is not indicative of a good computer for its price.

Related Questions

Navigate

• Browse (All)
• Browse T
• Subjects

---

---

Integrity-first tutoring: explanations and feedback only — we do not complete graded work. Learn more.