1.5a) P1: 3GHz/1.5 = 2*10^9 instructions per second P2: 2.5GHz/1.0 = 2.5*10^9 instructions per second P3: 4.0GHz/2.2 = 1.82*10^9 instructions per second Therefore, P2 has the highest performance expressed in instructions per second.b) Number of cycles P1: 3GHz*10s = 3*10^10 cycles P2: 2.5GHz*10s = 2.5*10^10 cycles P3: 4.0GHz*10s = 4.0*10^10 cycles Number of instructions P1: 2*10^9*10 = 2*10^10 instructions P2: 2.5*10^9*10 = 2.5*10^10 instructions P3: 1.82*10^9*10 = 1.82*10^10 instructionsc) ET = IC*CPI/CR CR = IC*CPI/ET 0.7ET = IC*1.2*CPI/CR_new CR_new = IC*CPI*1.71/ET = (IC*CPI/ET)*1.71 = CR*1.71 New clock rate P1: 3*1.71 = 5.13GHz P2: 2.5*1.71 = 4.28GHz P3: 4.0*1.71 = 6.84GHz1.6 P1 CT = 1/CR = 1/(2.5*10^9) = 400ps ET = IC*CPI*CT = 1*10^6*(1*0.1+2*0.2+3*0.5+3*0.2)*400*10^-12 = 1.04*10^-4 seconds P2 CT = 1/CR = 1/(3.0*10^9) = 333ps ET = IC*CPI*CT = 1*10^6*2*333*10^-12 = 6.66*10^-5 seconds P2 is fastera) Global CPI P1: 1*0.1+2*0.2+3*0.5+3*0.2 = 2.6 P2: 2b) P1: 1*10^6*2.6 = 2.6*10^6 cycles P2: 1*10^6*2.0 = 2.0*10^6 cycles1.7a)ET = IC*CPI*CT Compiler A 1.1 = 1*10^9*CPI*1*10^-9 CPI = 1.1 Compiler B 1.5 = 1.2*10^9*CPI*1*10^-9 1.2*CPI = 1.5 CPI = 1.25b) ET = IC*CPI/CR CR = IC*CPI/ET For same ET CR_B/CR_A = (IC_B*CPI_B)/(IC_A*CPI_A) = (1.2*10^9*1.25)/(1*10^9*1.1) = 1.37 Compiler B is 1.37 times faster than Compiler Ac) For same CR/CT ET_A/ET_New = (IC_A*CPI_A)/(IC_New*CPI_New) = (1.0*10^9*1.1)/(6.0*10^8*1.1) = 1.67 ET_B/ET_New = (IC_B*CPI_B)/(IC_New*CPI_New) = (1.2*10^9*1.25)/(6.0*10^8*1.1) = 2.271.131.13.1 0.2*70s = 14s 14s/250s*100% = 5.6%1.13.2 70s+40s+85s = 195s INT = 250s-195s = 55s 250s*0.8 = 200s 70+40+85+55x = 200 55x = 5 x = 1/11 = 9.09% Therefore, INT operations are reduced by 90.9%1.13.3 No because 200 = 40x+210 is
View Full Document