CPU Performance Evaluation: Cycles Per Instruction (CPI)

Most computers run synchronously utilizing a CPU clock running at a constant clock rate: Or clock frequency: f Clock cycle
where:

Clock rate = 1 / clock cycle

f = 1 /C
cycle 1 cycle 2 cycle 3

The CPU clock rate depends on the specific CPU organization (design) and hardware implementation technology (VLSI) used. A computer machine (ISA) instruction is comprised of a number of elementary or micro operations which vary in number and complexity depending on the the instruction and the exact CPU organization (Design). A micro operation is an elementary hardware operation that can be performed during one CPU clock cycle. This corresponds to one micro-instruction in microprogrammed CPUs. Examples: register operations: shift, load, clear, increment, ALU operations: add , subtract, etc. Thus: A single machine instruction may take one or more CPU cycles to complete termed as the Cycles Per Instruction (CPI). Instructions Per Cycle = IPC = 1/CPI Average (or effective) CPI of a program: The average CPI of all instructions executed in the program on a given CPU design.
Cycles/sec = Hertz = Hz MHz = 106 Hz GHz = 109 Hz

4th Edition: Chapter 1 (1.4, 1.7, 1.8) 3rd Edition: Chapter 4

EECC550 - Shaaban
#1 Lec # 3 Winter 2009 12-8-2009

Generic CPU Machine Instruction Processing Steps

Instruction Fetch Instruction Decode Operand Fetch Execute Result Store Next Instruction Obtain instruction from program memory
The Program Counter (PC) points to the instruction to be processed

Determine required actions and instruction size Locate and obtain operand data
From data memory or registers

Compute result value or status Deposit results in storage (data memory or register) for later use Determine successor or next instruction
(i.e Update PC to fetch next instruction to be processed)

EECC550 - Shaaban
#2 Lec # 3 Winter 2009 12-8-2009

 For a specific program compiled to run on a specific machine (CPU) A, has the following parameters:
I The total executed instruction count of the program. The average number of cycles per instruction (average CPI). Clock cycle of machine A C Or effective CPI CPI

Computer Performance Measures: Program Execution Time

How can one measure the performance of this machine (CPU) running this program? Intuitively the machine (or CPU) is said to be faster or has better performance running this program if the total execution time is shorter. Thus the inverse of the total measured program execution time is a possible performance measure or metric:
Seconds/program

Programs/second

PerformanceA = 1 / Execution TimeA

How to compare performance of different machines? What factors affect performance? How to improve performance?

EECC550 - Shaaban
#3 Lec # 3 Winter 2009 12-8-2009

Comparing Computer Performance Using Execution Time

To compare the performance of two machines (or CPUs) A, B running a given specific program: PerformanceA = 1 / Execution TimeA PerformanceB = 1 / Execution TimeB Machine A is n times faster than machine B means (or slower? if n < 1) :
Speedup = n = PerformanceA PerformanceB = Execution TimeB Execution TimeA

(i.e Speedup is ratio of performance, no units)

Example: For a given program: Execution time on machine A: ExecutionA = 1 second Execution time on machine B: ExecutionB = 10 seconds PerformanceA / PerformanceB = Execution TimeB / Execution TimeA = 10 / 1 = 10
The performance of machine A is 10 times the performance of machine B when running this ISAs provided The two CPUs may target different program, or: Machine A is said to be 10 EECC550 - Shaaban thetimes fasterwritten in a high level language (HLL)program. program is than machine B when running this
#4 Lec # 3 Winter 2009 12-8-2009

Speedup=

CPU Execution Time: The CPU Equation

A program is comprised of a number of instructions executed , I Measured in: instructions/program The average instruction executed takes a number of cycles per instruction (CPI) to be completed. Or Instructions Per Cycle (IPC): Measured in: cycles/instruction, CPI IPC = 1/CPI CPU has a fixed clock cycle time C = 1/clock rate Measured in: seconds/cycle
C = 1/f

CPU execution time is the product of the above three parameters as follows: Executed
CPU time CPU time = Seconds = Seconds Program Program = Instructions x Cycles x Seconds = Instructions x Cycles x Seconds Program Instruction Cycle Program Instruction Cycle

T =
execution Time per program in seconds

I x
Number of instructions executed

CPI

C
CPU Clock Cycle

Average CPI for program

(This equation is commonly known as the CPU performance equation)

EECC550 - Shaaban
#5 Lec # 3 Winter 2009 12-8-2009

CPU Average CPI/Execution Time

For a given program executed on a given machine (CPU): CPI = Total program execution cycles / Instructions count
(i.e average or effective CPI)

CPU clock cycles = Instruction count x CPI

(executed)

Executed

CPU execution time =

T
execution Time per program in seconds

= CPU clock cycles x Clock cycle = Instruction count x CPI x Clock cycle = I x Average x CPI C
Number of instructions executed or effective CPI for program

CPU Clock Cycle

(This equation is commonly known as the CPU performance equation)

CPI = Cycles Per Instruction

EECC550 - Shaaban
#6 Lec # 3 Winter 2009 12-8-2009

CPU Execution Time: Example

A Program is running on a specific machine (CPU) with I the following parameters:
Total executed instruction count: 10,000,000 instructions Average CPI for the program: 2.5 cycles/instruction. CPU clock rate: 200 MHz. (clock cycle = C = 5x10-9 seconds)

What is the execution time for this program:

CPU time CPU time = Seconds = Seconds Program Program

i.e 5 nanoseconds

= Instructions x Cycles x Seconds = Instructions x Cycles x Seconds Program Instruction Cycle Program Instruction Cycle

CPU time = Instruction count x CPI x Clock cycle = = = 10,000,000 10,000,000 0.125 seconds x 2.5 x 1 / clock rate x 2.5 x 5x10-9
EECC550 - Shaaban
#7 Lec # 3 Winter 2009 12-8-2009

Nanosecond = nsec =ns = 10-9 second

T = I x CPI x C

Aspects of CPU Execution Time

CPU Time = Instruction count executed x CPI x Clock cycle
T = I x CPI x C
Depends on: Program Used Compiler ISA

Instruction Count I
(executed)

Depends on: Program Used Compiler ISA CPU Organization

CPI
(Average CPI)

Clock Cycle C

Depends on: CPU Organization Technology (VLSI)

EECC550 - Shaaban
#8 Lec # 3 Winter 2009 12-8-2009

Factors Affecting CPU Performance

CPU time CPU time = Seconds = Seconds Program Program = Instructions x Cycles x Seconds = Instructions x Cycles x Seconds Program Instruction Cycle Program Instruction Cycle

I x CPI x C Instruction Cycles per Clock Rate Count (1/C) Instruction

Program Compiler Instruction Set Architecture (ISA) Organization

(CPU Design)

Technology
(VLSI)

T = I x CPI x C

EECC550 - Shaaban
#9 Lec # 3 Winter 2009 12-8-2009

Performance Comparison: Example

From the previous example: A Program is running on a specific machine (CPU) with the following parameters: Total executed instruction count, I: 10,000,000 instructions Average CPI for the program: 2.5 cycles/instruction. CPU clock rate: 200 MHz. Thus: C = 1/(200x10 )= 5x10 seconds Using the same program with these changes: A new compiler used: New executed instruction count, I: 9,500,000 New CPI: 3.0 Faster CPU implementation: New clock rate = 300 MHz Thus: C = 1/(300x10 )= 3.33x10 seconds What is the speedup with the changes?
6 -9 6 -9

Speedup Speedup

= Old Execution Time = Iold xx CPIold = Old Execution Time = Iold CPIold New Execution Time Inew xx CPInew New Execution Time Inew CPInew

xx Clock cycleold Clock cycleold xx Clock Cyclenew Clock Cycle

new

Speedup = (10,000,000 x 2.5 x 5x10-9) / (9,500,000 x 3 x 3.33x10-9 ) = .125 / .095 = 1.32 or 32 % faster after changes.
Clock Cycle = C = 1/ Clock Rate

T = I x CPI x C

EECC550 - Shaaban
#10 Lec # 3 Winter 2009 12-8-2009

Instruction Types & CPI

Given a program with n types or classes of instructions executed on given CPU with the following characteristics: a

Ci = Count of instructions of typei executed CPIi = Cycles per instruction for typei

i = 1, 2, . n

Then:
i.e average or effective CPI

Depends on CPU Design

CPI = CPU Clock Cycles / Instruction Count I

Executed

Where:

CPU clock cycles =

i =1

( CPI C )
i i

Executed Instruction Count I = Ci

T = I x CPI x C

EECC550 - Shaaban
#11 Lec # 3 Winter 2009 12-8-2009

Instruction Types & CPI: An Example

An instruction set has three instruction classes:
Instruction class A B C CPI 1 2 3
For a specific CPU design

Two code sequences have the following instruction counts:

Code Sequence 1 2 Instruction counts for instruction class A B C 2 1 2 4 1 1

CPU cycles for sequence 1 = 2 x 1 + 1 x 2 + 2 x 3 = 10 cycles CPI for sequence 1 = clock cycles / instruction count i.e average or effective CPI = 10 /5 = 2 CPU cycles for sequence 2 = 4 x 1 + 1 x 2 + 1 x 3 = 9 cycles CPI for sequence 2 = 9 / 6 = 1.5
CPU clock cycles =
i =1 n

( CPI C )
i i

CPI = CPU Cycles / I

EECC550 - Shaaban
#12 Lec # 3 Winter 2009 12-8-2009

Instruction Frequency & CPI

Given a program with n types or classes of instructions with the following characteristics:
Ci = Count of instructions of typei executed CPIi = Average cycles per instruction of typei Fi = Frequency or fraction of instruction typei executed = Ci/ total executed instruction count Executed/Instruction Count I = = Ci I Where:
Then:
Ci

i = 1, 2, . n

CPI = (CPI i F i )
n i =1

i.e average or effective CPI

Fraction of total execution time for instructions of type i = T = I x CPI x C

CPIi x Fi CPI

EECC550 - Shaaban
#13 Lec # 3 Winter 2009 12-8-2009

Instruction Type Frequency & CPI: A RISC Example

Program Profile or Executed Instructions Mix

CPIi x Fi
Depends on CPU Design

Base Machine (Reg / Reg) Op Freq, Fi CPIi

Given

CPI % Time 23% = 45% = 14% = 18% = .5/2.2 1/2.2 .3/2.2 .4/2.2

CPIi x Fi .5 1.0 .3 .4

ALU Load Store Branch

50% 20% 10% 20%

Typical Mix n

1 5 3 2

Sum = 2.2
i

i.e average or effective CPI

CPI = C ( PI
i= 1

F i )

CPI = .5 x 1 + .2 x 5 + .1 x 3 + .2 x 2 = 2.2 = .5 + 1 + .3 + .4
T = I x CPI x C

EECC550 - Shaaban
#14 Lec # 3 Winter 2009 12-8-2009

Metrics of Computer Performance

(Measures)
Application Programming Language Compiler ISA Datapath Control Function Units Transistors Wires Pins Cycles per second (clock rate). (millions) of Instructions per second MIPS (millions) of (F.P.) operations per second MFLOP/s Execution time: Target workload, SPEC, etc.

Megabytes per second.

Each metric has a purpose, and each can be misused.

EECC550 - Shaaban
#15 Lec # 3 Winter 2009 12-8-2009

Choosing Programs To Evaluate Performance

Levels of programs or benchmarks that could be used to evaluate performance: Actual Target Workload: Full applications that run on the target machine. Real Full Program-based Benchmarks:
Select a specific mix or suite of programs that are typical of targeted applications or workload (e.g SPEC95, SPEC CPU2000).

Small Kernel Benchmarks:

Also called synthetic benchmarks

Key computationally-intensive pieces extracted from real programs. Examples: Matrix factorization, FFT, tree search, etc. Best used to test specific aspects of the machine.

Microbenchmarks:
Small, specially written programs to isolate a specific aspect of performance characteristics: Processing: integer, floating point, local memory, input/output, etc.

EECC550 - Shaaban
#16 Lec # 3 Winter 2009 12-8-2009

Pros

Types of Benchmarks
Actual Target Workload

Cons

Representative

Very specific. Non-portable. Complex: Difficult to run, or measure. Less representative than actual workload. Easy to fool by designing hardware to run them well. Peak performance results may be a long way from real application performance

Portable. Widely used. Measurements useful in reality.

Full Application Benchmarks

Easy to run, early in the design cycle. Identify peak performance and potential bottlenecks.

Small Kernel Benchmarks

Microbenchmarks

EECC550 - Shaaban
#17 Lec # 3 Winter 2009 12-8-2009

The most popular and industry-standard set of CPU benchmarks. Programs application domain: Engineering and scientific computation SPECmarks, 1989:
10 programs yielding a single number (SPECmarks).

SPEC: System Performance Evaluation Corporation

SPEC92, 1992:
SPECInt92 (6 integer programs) and SPECfp92 (14 floating point programs).

SPEC95, 1995:
SPECint95 (8 integer programs):
go, m88ksim, gcc, compress, li, ijpeg, perl, vortex

SPECfp95 (10 floating-point intensive programs):

tomcatv, swim, su2cor, hydro2d, mgrid, applu, turb3d, apsi, fppp, wave5

Performance relative to a Sun SuperSpark I (50 MHz) which is given a score of SPECint95 = SPECfp95 = 1 CINT2000 (11 integer programs). CFP2000 (14 floating-point intensive programs) Performance relative to a Sun Ultra5_10 (300 MHz) which is given a score of SPECint2000 = SPECfp2000 = 100 CINT2006 (12 integer programs). CFP2006 (17 floating-point intensive programs) Performance relative to a Sun Ultra Enterprise 2 workstation with a 296-MHz UltraSPARC II processor which is given a score of SPECint2006 = SPECfp2006 = 1

SPEC CPU2000, 1999:

SPEC CPU2006, 2006:

All based on execution time and give speedup over a reference CPU

EECC550 - Shaaban
#18 Lec # 3 Winter 2009 12-8-2009

SPEC95 Programs
Programs application domain: Engineering and scientific computation
B enchm ark go m 88ksim gcc com press li ijpeg perl vortex tom catv sw im su2cor hydro2d m grid applu trub3d apsi fpppp w ave5 D escription Artificial intelligence; plays the gam e of G o M otorola 88k chip sim ulator; runs test program The G nu C com piler generating SPAR C code C om presses and decom presses file in m em ory Lisp interpreter G raphic com pression and decom pression M anipulates strings and prim e num bers in the special-purpose program m ing language A database program A m esh generation program Shallow w ater m odel w ith 513 x 513 grid quantum physics; M onte C arlo sim ulation Astrophysics; H ydrodynam ic N aiver Stokes equations M ultigrid solver in 3-D potential field Parabolic/elliptic partial differential equations Sim ulates isotropic, hom ogeneous turbulence in a cube Solves problem s regarding tem perature, w ind velocity, and distribution of pollutant Q uantum chem istry Plasm a physics; electrom agnetic particle sim ulation

Integer

Floating Point

Resulting Performance relative to a Sun SuperSpark I (50 MHz) which is given a score of SPECint95 = SPECfp95 = 1

EECC550 - Shaaban
#19 Lec # 3 Winter 2009 12-8-2009

Sample SPECint95 (Integer) Results

Source URL: http://www.macinfo.de/bench/specmark.html

Sun SuperSpark I (50 MHz) score = 1

T = I x CPI x C

EECC550 - Shaaban
#20 Lec # 3 Winter 2009 12-8-2009

Sample SPECfp95 (Floating Point) Results

Source URL: http://www.macinfo.de/bench/specmark.html

Sun SuperSpark I (50 MHz) score = 1

T = I x CPI x C

EECC550 - Shaaban
#21 Lec # 3 Winter 2009 12-8-2009

Benchmark
164.gzip 175.vpr 176.gcc 181.mcf 186.crafty 197.parser 252.eon 253.perlbmk 254.gap 255.vortex 256.bzip2 300.twolf 168.wupwise 171.swim 172.mgrid 173.applu 177.mesa 178.galgel 179.art 183.equake 187.facerec 188.ammp 189.lucas 191.fma3d 200.sixtrack 301.apsi

SPEC CPU2000 Programs

Language
C C C C C C C++ C C C C C Fortran 77 Fortran 77 Fortran 77 Fortran 77 C Fortran 90 C C Fortran 90 C Fortran 90 Fortran 90 Fortran 77 Fortran 77

Descriptions

CINT2000 (Integer)
11 programs

Compression FPGA Circuit Placement and Routing C Programming Language Compiler Combinatorial Optimization Game Playing: Chess Word Processing Computer Visualization PERL Programming Language Group Theory, Interpreter Object-oriented Database Compression Place and Route Simulator Physics / Quantum Chromodynamics Shallow Water Modeling Multi-grid Solver: 3D Potential Field Parabolic / Elliptic Partial Differential Equations 3-D Graphics Library Computational Fluid Dynamics Image Recognition / Neural Networks Seismic Wave Propagation Simulation Image Processing: Face Recognition Computational Chemistry Number Theory / Primality Testing Finite-element Crash Simulation High Energy Nuclear Physics Accelerator Design Meteorology: Pollutant Distribution

CFP2000 (Floating Point)

14 programs

Programs application domain: Engineering and scientific computation

Source: http://www.spec.org/cpu2000/

EECC550 - Shaaban
#22 Lec # 3 Winter 2009 12-8-2009

Integer SPEC CPU2000 Microprocessor Performance 1978-2006

Performance relative to VAX 11/780 (given a score = 1)

EECC550 - Shaaban
#23 Lec # 3 Winter 2009 12-8-2009

Top 20 SPEC CPU2000 Results (As of March 2002)

Top 20 SPECint2000 Top 20 SPECfp2000
int base 790 790 788 697 621 648 537 587 568 497 4 495 1533 511 483 449 438 417 409 431 358 328 MHz Processor 1300 POWER4 1000 Alpha 21264C 1050 UltraSPARC-III Cu 2200 Pentium 4 Xeon 2200 Pentium 4 833 Alpha 21264B 800 Itanium 833 Alpha 21264A 1667 Athlon XP 750 PA-RISC 8700 Athlon MP 547 600 MIPS R14000 675 SPARC64 GP 900 UltraSPARC-III 1400 Athlon 1400 Pentium III 500 PA-RISC 8600 450 POWER3-II 500 Alpha 21264 400 MIPS R12000 fp peak 1169 960 827 802 801 784 701 644 642 581 504 529 509 482 458 456 440 433 422 407 fp base 1098 776 701 779 779 643 701 571 596 526 499 371 427 426 437 397 426 383 382 # 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 MHz 1300 2200 2200 1667 1000 1400 1050 1533 750 833 1400 833 600 675 900 552 750 700 800 400 Processor POWER4 Pentium 4 Pentium 4 Xeon Athlon XP Alpha 21264C Pentium III UltraSPARC-III Cu Athlon MP PA-RISC 8700 Alpha 21264B Athlon Alpha 21264A MIPS R14000 SPARC64 GP UltraSPARC-III PA-RISC 8600 POWER RS64-IV Pentium III Xeon Itanium MIPS R12000 int peak 814 811 810 724 679 664 610 609 604 571 554 533 500 478 467 441 439 438 365 353

Performance relative to a Sun Ultra5_10 (300 MHz) which is given a score of SPECint2000 = SPECfp2000 = 100

Source: http://www.aceshardware.com/SPECmine/top.jsp

EECC550 - Shaaban
#24 Lec # 3 Winter 2009 12-8-2009

Top 20 SPECint2000 Top 20 SPECfp2000 # MHz Processorint peakint base MHz Processor fp peak fp base 1 2933 Core 2 Duo EE 3119 3108 2300 2 3000 Xeon 51xx 3102 3089 1600 3 2666 Core 2 Duo 2848 2844 3000 4 2660 Xeon 30xx 2835 2826 2933 5 3000 Opteron 2119 1942 2660 6 2800 Athlon 64 FX 2061 1923 1600 7 2800 Opteron AM2 1960 1749 2667 8 2300 POWER5+ 1900 1820 1900 9 3733 Pentium 4 E 1872 1870 3000 103800 Pentium 4 Xeon1856 1854 2800 112260 Pentium M 1839 1812 3733 123600 Pentium D 1814 1810 2800 132167 Core Duo 1804 1796 2700 143600 Pentium 4 1774 1772 2160 153466 Pentium 4 EE 1772 1701 3730 162700 PowerPC 970MP1706 1623 3600 172600 Athlon 64 1706 1612 3600 182000 Pentium 4 Xeon LV 1668 1663 1700 192160 SPARC64 V 1620 1501 1700 Performance relative to a Sun Ultra5_10 (300 MHz) which is given a score of SPECint2000 = SPECfp2000 = 100 201600 Itanium 2 1590 1590 3466
Source: http://www.aceshardware.com/SPECmine/top.jsp

Top 20 SPEC CPU2000 Results (As of October 2006)

POWER5+ 3642 3369 DC Itanium 2 3098 3098 Xeon 51xx 3056 2811 Core 2 Duo EE3050 3048 Xeon 30xx 3044 2763 Itanium 2 3017 3017 Core 2 Duo 2850 2847 POWER5 2796 2585 Opteron 2497 2260 Opteron AM2 2462 2230 Pentium 4 E 2283 2280 Athlon 64 FX 2261 2086 PowerPC 970MP22592060 SPARC64 V 2236 2094 Pentium 4 Xeon21502063 Pentium D 2077 2073 Pentium 4 2015 2009 2600 Athlon 641829 POWER4+ 1776 1642 EECC550 - Shaaban Pentium 4 EE 1724 1719
#25 Lec # 3 Winter 2009 12-8-2009

Benchmark
400.perlbench 401.bzip2 403.gcc 429.mcf 445.gobmk 456.hmmer 458.sjeng 462.libquantum 464.h264ref 471.omnetpp 473.astar 483.Xalancbmk 410.bwaves 416.gamess 433.milc 434.zeusmp 435.gromacs 436.cactusADM 437.leslie3d 444.namd 447.dealII 450.soplex 453.povray 454.calculix 459.GemsFDTD 465.tonto 470.lbm 481.wrf 482.sphinx3

SPEC CPU2006 Programs

Language
C C C C C C C C C C++ C++ C++ Fortran Fortran C Fortran C/Fortran C/Fortran Fortran C++ C++ C++ C++ C/Fortran Fortran Fortran C C/Fortran C

Descriptions

CINT2006 (Integer)
12 programs

CFP2006 (Floating Point)

17 programs

PERL Programming Language Compression C Compiler Combinatorial Optimization Artificial Intelligence: go Search Gene Sequence Artificial Intelligence: chess Physics: Quantum Computing Video Compression Discrete Event Simulation Path-finding Algorithms XML Processing Fluid Dynamics Quantum Chemistry Physics: Quantum Chromodynamics Physics/CFD Biochemistry/Molecular Dynamics Physics/General Relativity Fluid Dynamics Biology/Molecular Dynamics Finite Element Analysis Linear Programming, Optimization Image Ray-tracing Structural Mechanics Computational Electromagnetics Quantum Chemistry Fluid Dynamics Weather Prediction Speech recognition

Programs application domain: Engineering and scientific computation

Source: http://www.spec.org/cpu2006/

EECC550 - Shaaban
#26 Lec # 3 Winter 2009 12-8-2009

Example Integer SPEC CPU2006 Performance Results

For 2.5 GHz AMD Opteron X4 model 2356 (Barcelona)

EECC550 - Shaaban
#27 Lec # 3 Winter 2009 12-8-2009

For a specific program running on a specific CPU the MIPS rating is a measure of how many millions of instructions are executed per second:
MIPS Rating = Instruction count / (Execution Time x 106) = Instruction count / (CPU clocks x Cycle time x 106) = (Instruction count x Clock rate) / (Instruction count x CPI x 106) = Clock rate / (CPI x 106)

Computer Performance Measures : MIPS (Million Instructions Per Second) Rating

Major problem with MIPS rating: As shown above the MIPS rating does not account for the count of instructions executed (I).

A higher MIPS rating in many cases may not mean higher performance or better execution time. i.e. due to compiler design variations.

In addition the MIPS rating:

Does not account for the instruction set architecture (ISA) used.
Thus it cannot be used to compare computers/CPUs with different instruction sets.

Easy to abuse: Program used to get the MIPS rating is often omitted.
Often the Peak MIPS rating is provided for a given CPU which is obtained using a program comprised entirely of instructions with the lowest CPI for the given CPU design which does not represent real programs.

T = I x CPI x C

EECC550 - Shaaban
#28 Lec # 3 Winter 2009 12-8-2009

Computer Performance Measures : MIPS (Million Instructions Per Second) Rating

Under what conditions can the MIPS rating be used to compare performance of different CPUs?
The MIPS rating is only valid to compare the performance of different CPUs provided that the following conditions are satisfied: 1 The same program is used
(actually this applies to all performance metrics)

2 The same ISA is used 3 The same compiler is used (Thus the resulting programs used to run on the CPUs and obtain the MIPS rating are identical at the machine code (binary) level including the same instruction count)

EECC550 - Shaaban
#29 Lec # 3 Winter 2009 12-8-2009

Compiler Variations, MIPS & Performance:

For a machine (CPU)An Example classes: with instruction
Instruction class A B C CPI 1 2 3

For a given high-level language program, two compilers produced the following executed instruction counts:
Code from: Compiler 1 Compiler 2 Instruction counts (in millions) for each instruction class A B C 5 1 1 10 1 1

The machine is assumed to run at a clock rate of 100 MHz.

EECC550 - Shaaban
#30 Lec # 3 Winter 2009 12-8-2009

Compiler Variations, MIPS & Performance:

MIPS = Clock rate / (CPI x 106) = 100 MHz / (CPI x 106)

An Example (Continued)

CPI = CPU execution cycles / Instructions count n

CPU clock cycles =
i =1

( CPI C )
i i

CPU time = Instruction count x CPI / Clock rate

For compiler 1: CPI1 = (5 x 1 + 1 x 2 + 1 x 3) / (5 + 1 + 1) = 10 / 7 = 1.43 MIPS Rating1 = 100 / (1.428 x 106) = 70.0 MIPS CPU time1 = ((5 + 1 + 1) x 106 x 1.43) / (100 x 106) = 0.10 seconds For compiler 2: CPI2 = (10 x 1 + 1 x 2 + 1 x 3) / (10 + 1 + 1) = 15 / 12 = 1.25 MIPS Rating2 = 100 / (1.25 x 106) = 80.0 MIPS
MIPS rating indicates that compiler 2 is better while in reality the code produced by compiler 1 is faster

CPU time2 = ((10 + 1 + 1) x 106 x 1.25) / (100 x 106) = 0.15 seconds

EECC550 - Shaaban
#31 Lec # 3 Winter 2009 12-8-2009

MIPS (The ISA not the metric) Loop Performance Example

High Memory $6 points here

For the loop: for (i=0; i<1000; i=i+1){ x[i] = x[i] + s; } MIPS assembly code is given by:
loop: lw addi lw add sw addi bne $3, $6, $4, $5, $5, $2, $6, 0($1) $2, 4000 0($2) $4, $3 0($2) $2, 4 $2, loop ; ; ; ; ; ; ;
$2 initially points here

X[999] X[998]

Last element to compute

. . . .
X[0] Low Memory First element to compute

load s in $3 $6 = address of last element + 4 load x[i] in $4 $5 has x[i] + s store computed x[i] increment $2 to point to next x[ ] element last loop iteration reached?

The MIPS code is executed on a specific CPU that runs at 500 MHz (clock cycle = 2ns = 2x10-9 seconds) with following instruction type CPIs :

For this MIPS code running on this CPU find:

Instruction type CPI ALU 4 Load 5 Store 7 Branch 3

1- Fraction of total instructions executed for each instruction type 2- Total number of CPU cycles 3- Average CPI 4- Fraction of total execution time for each instructions type 5- Execution time 6- MIPS rating , peak MIPS rating for this CPU

X[ ] array of words in memory, base address in $2 , s a constant word value in memory, address in $1

EECC550 - Shaaban
#32 Lec # 3 Winter 2009 12-8-2009

MIPS (The ISA) Loop Performance Example (continued)

The code has 2 instructions before the loop and 5 instructions in the body of the loop which iterates 1000 times, Thus: Total instructions executed, I = 5x1000 + 2 = 5002 instructions

Number of instructions executed/fraction Fi for each instruction type:

ALU instructions = 1 + 2x1000 = 2001 Load instructions = 1 + 1x1000 = 1001 Store instructions = 1000 Branch instructions = 1000
n

CPIALU = 4 CPILoad = 5 CPIStore = 7

FractionALU = FALU = 2001/5002 = 0.4 = 40% FractionLoad = FLoad = 1001/5002= 0.2 = 20% FractionStore = FStore = 1000/5002 = 0.2 = 20%

2 3 4

CPU clock cycles = CPI

i= 1

CPIBranch = 3 FractionBranch= FBranch = 1000/5002= 0.2 = 20%

i

Ci

= 2001x4 + 1001x5 + 1000x7 + 1000x3 = 23009 cycles Average CPI = CPU clock cycles / I = 23009/5002 = 4.6 Instruction type Fraction of execution time for each instruction type: ALU
Fraction of time for ALU instructions = CPIALU x FALU / CPI= 4x0.4/4.6 = 0.348 = 34.8% Fraction of time for load instructions = CPIload x Fload / CPI= 5x0.2/4.6 = 0.217 = 21.7% Fraction of time for store instructions = CPIstore x Fstore / CPI= 7x0.2/4.6 = 0.304 = 30.4% Fraction of time for branch instructions = CPIbranch x Fbranch / CPI= 3x0.2/4.6 = 0.13 = 13%

CPI 4 5 7 3

Load Store Branch

5 6

Execution time = I x CPI x C = CPU cycles x C = 23009 x 2x10-9 = = 4.6x 10-5 seconds = 0.046 msec = 46 usec MIPS rating = Clock rate / (CPI x 106) = 500 / 4.6 = 108.7 MIPS
The CPU achieves its peak MIPS rating when executing a program that only has instructions of the type with the lowest CPI. In this case branches with CPIBranch = 3 Peak MIPS rating = Clock rate / (CPIBranch x 106) = 500/3 = 166.67 MIPS

EECC550 - Shaaban
#33 Lec # 3 Winter 2009 12-8-2009

Computer Performance Measures : MFLOPS (Million FLOating-Point Operations Per Second)

A floating-point operation is an addition, subtraction, multiplication, or division operation applied to numbers represented by a single or a double precision floating-point representation. MFLOPS, for a specific program running on a specific computer, is a measure of millions of floating point-operation (megaflops) per second:

MFLOPS = Number of floating-point operations / (Execution time x 106 )

MFLOPS rating is a better comparison measure between different machines (applies even if ISAs are different) than the MIPS rating. Applicable even if ISAs are different Program-dependent: Different programs have different percentages of floating-point operations present. i.e compilers have no floating- point operations and yield a MFLOPS rating of zero. Dependent on the type of floating-point operations present in the program. Peak MFLOPS rating for a CPU: Obtained using a program comprised entirely of the simplest floating point instructions (with the lowest CPI) for the given CPU design which does not represent real floating point programs.

EECC550 - Shaaban
#34 Lec # 3 Winter 2009 12-8-2009

Quantitative Principles of Computer Design

Amdahls Law:
The performance gain from improving some portion of a computer is calculated by: i.e using some enhancement
Speedup = Performance for entire task using the enhancement Performance for the entire task without using the enhancement Execution time without the enhancement Execution time for entire task using the enhancement
Recall: Performance = 1 /Execution Time
3rd Edition: Chapter 4.5

or Speedup =

Here: Task = Program

4th Edition: Chapter 1.8

EECC550 - Shaaban
#35 Lec # 3 Winter 2009 12-8-2009

Performance Enhancement Calculations: Amdahl's Law

The performance enhancement possible due to a given design improvement is limited by the amount that the improved feature is used

Amdahls Law: Performance improvement or speedup due to enhancement E:

Execution Time without E Speedup(E) = -------------------------------------Execution Time with E
original

Performance with E = --------------------------------Performance without E

Suppose that enhancement E accelerates a fraction F of the execution time by a factor S and the remainder of the time is unaffected then:

Execution Time with E = ((1-F) + F/S) X Execution Time without E Hence speedup is given by: Execution Time without E 1 Speedup(E) = --------------------------------------------------------- = -------------------((1 - F) + F/S) X Execution Time without E (1 - F) + F/S F (Fraction of execution time enhanced) refers
to original execution time before the enhancement is applied
#36 Lec # 3 Winter 2009 12-8-2009

EECC550 - Shaaban

Pictorial Depiction of Amdahls Law

Enhancement E accelerates fraction F of original execution time by a factor of S

Before: Execution Time without enhancement E: (Before enhancement is applied) shown normalized to 1 = (1-F) + F =1

Unaffected fraction: (1- F)

Affected fraction: F

Unchanged

Unaffected fraction: (1- F)

After: Execution Time with enhancement E:

F/S
What if the fraction given is after the enhancement has been applied? How would you solve the problem? (i.e find expression for speedup)

Execution Time without enhancement E 1 Speedup(E) = ------------------------------------------------------ = -----------------Execution Time with enhancement E (1 - F) + F/S

EECC550 - Shaaban
#37 Lec # 3 Winter 2009 12-8-2009

Performance Enhancement Example

For the RISC machine with the following instruction mix given earlier:
Op ALU Load Store Freq 50% 20% 10% Cycles 1 5 3 CPI(i) .5 1.0 .3 % Time 23% 45% 14%

CPI = 2.2

Branch 20% 2 .4 18% If a CPU design enhancement improves the CPI of load instructions from 5 to 2, what is the resulting performance improvement from this enhancement:

Fraction enhanced = F = 45% or .45 Unaffected fraction = 1- F = 100% - 45% = 55% or .55 Factor of enhancement = S = 5/2 = 2.5 Using Amdahls Law:
1 1 Speedup(E) = ------------------ = --------------------- = (1 - F) + F/S .55 + .45/2.5 1.37

EECC550 - Shaaban
#38 Lec # 3 Winter 2009 12-8-2009

An Alternative Solution Using CPU Equation

Op ALU Load Store Freq 50% 20% 10% Cycles 1 5 3 CPI(i) .5 1.0 .3 % Time 23% 45% 14%

CPI = 2.2

Branch 20% 2 .4 18% If a CPU design enhancement improves the CPI of load instructions from 5 to 2, what is the resulting performance improvement from this enhancement:
New CPI of load is now 2 instead of 5

Old CPI = 2.2 New CPI = .5 x 1 + .2 x 2 + .1 x 3 + .2 x 2 = 1.6

Original Execution Time Speedup(E) = ----------------------------------New Execution Time Instruction count x old CPI x clock cycle = ---------------------------------------------------------------Instruction count x new CPI x clock cycle old CPI = ------------ = new CPI 2.2 --------1.6

= 1.37

Which is the same speedup obtained from Amdahls Law in the first solution.

T = I x CPI x C

EECC550 - Shaaban
#39 Lec # 3 Winter 2009 12-8-2009

Performance Enhancement Example

100 ----------------------------------------------------Execution Time with enhancement

A program runs in 100 seconds on a machine with multiply operations responsible for 80 seconds of this time. By how much must the speed of multiplication be improved to make the program four times faster?

Desired speedup = 4 =

Execution time with enhancement = 100/4 = 25 seconds 25 seconds = (100 - 80 seconds) + 80 seconds / S 25 seconds = 20 seconds + 80 seconds / S 5 = 80 seconds / S S = 80/5 = 16

Alternatively, it can also be solved by finding enhanced fraction of execution time: F = 80/100 = .8 1 1 1 and then solving Amdahls speedup equation for desired enhancement factor S

Speedup(E) = ------------------ = 4 = (1 - F) + F/S

----------------- = --------------(1 - .8) + .8/S .2 + .8/s

Solving for S gives S= 16

Hence multiplication should be 16 times Machine = CPUto get an overall speedup of 4. faster

EECC550 - Shaaban
#40 Lec # 3 Winter 2009 12-8-2009

Performance Enhancement Example

For the previous example with a program running in 100 seconds on a machine with multiply operations responsible for 80 seconds of this time. By how much must the speed of multiplication be improved to make the program five times faster?
100 ----------------------------------------------------Execution Time with enhancement

Desired speedup = 5 =

Execution time with enhancement = 100/5 = 20 seconds 20 seconds = (100 - 80 seconds) + 80 seconds / s 20 seconds = 20 seconds + 80 seconds / s

0 = 80 seconds / s

No amount of multiplication speed improvement can achieve this.

EECC550 - Shaaban
#41 Lec # 3 Winter 2009 12-8-2009

Extending Amdahl's Law To Multiple Enhancements

n enhancements each affecting a different portion of execution time

Suppose that enhancement Ei accelerates a fraction Fi of the original execution time by a factor Si and the remainder of the i = 1, 2, . n time is unaffected then:

Speedup =

Original Execution Time

((1 F ) + F ) XOriginal Execution Time

i i i

Unaffected fraction

S
i

Speedup =

((1 F ) + F )
i i

What if the fractions given are after the enhancements were applied? How would you solve the problem? (i.e find expression for speedup)

Note: All fractions Fi refer to original execution time before the enhancements are applied.

EECC550 - Shaaban
#42 Lec # 3 Winter 2009 12-8-2009

Amdahl's Law With Multiple Enhancements: Example

Three CPU performance enhancements are proposed with the following speedups and percentage of the code execution time affected: Speedup1 = S1 = 10 Speedup2 = S2 = 15 Speedup3 = S3 = 30 Percentage1 = F1 = 20% Percentage1 = F2 = 15% Percentage1 = F3 = 10%

While all three enhancements are in place in the new design, each enhancement affects a different portion of the code and only one enhancement can be used at a time. What is the resulting overall speedup?
S peedu p = 1

((1 F ) + F )
i i i

Speedup = 1 / [(1 - .2 - .15 - .1) + .2/10 + .15/15 + .1/30)] = 1/ [ .55 + .0333 ] = 1 / .5833 = 1.71

EECC550 - Shaaban
#43 Lec # 3 Winter 2009 12-8-2009

Pictorial Depiction of Example

Before: Execution Time with no enhancements: 1
i.e normalized to 1

S1 = 10

S2 = 15

S3 = 30

Unaffected, fraction: .55

F1 = .2

F2 = .15

F3 = .1 / 30

/ 10
Unchanged

/ 15

Unaffected, fraction: .55

After: Execution Time with enhancements: .55 + .02 + .01 + .00333 = .5833 Speedup = 1 / .5833 = 1.71 Note: All fractions Fi refer to original execution time.
What if the fractions given are after the enhancements were applied? How would you solve the problem?

EECC550 - Shaaban
#44 Lec # 3 Winter 2009 12-8-2009

Reverse Multiple Enhancements Amdahl's Law

Multiple Enhancements Amdahl's Law assumes that the fractions given refer to original execution time. If for each enhancement Si the fraction Fi it affects is given as a fraction of the resulting execution time after the enhancements were applied then:
i i

((1 F ) + F S ) XResulting ExecutionTime Speedup=

i i i

Unaffected fraction

Resulting Execution Time

Speedup=

(1 i F i) + i F i S i
1

= (1 i F i ) + i F i S i

For the previous example assuming fractions given refer to resulting execution time after the enhancements were applied (not the original execution time), then:
Speedup = (1 - .2 - .15 - .1) + .2 x10 + .15 x15 + .1x30 = .55 + 2 + 2.25 + 3 = 7.8

i.e as if resulting execution time is normalized to 1

EECC550 - Shaaban
#45 Lec # 3 Winter 2009 12-8-2009