Scribd
Upload
48 views

Micro Processor and Assembly Language

This document summarizes key topics related to evaluating microprocessor performance, including clock speed, CPI, MIPS rate, benchmarks like SPEC, and Amdahl's Law. It discusses techniques for assessing instruction execution rate and issues with using MIPS as the sole measure. The document is divided into sections covering these performance assessment topics from a textbook on microprocessors and assembly language.

Uploaded by

Jaleed Masood
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
48 views

Micro Processor and Assembly Language

This document summarizes key topics related to evaluating microprocessor performance, including clock speed, CPI, MIPS rate, benchmarks like SPEC, and Amdahl's Law. It discusses techniques for assessing instruction execution rate and issues with using MIPS as the sole measure. The document is divided into sections covering these performance assessment topics from a textbook on microprocessors and assembly language.

Uploaded by

Jaleed Masood
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 16

“Evolution and Performance” in

[ Microprocessor and
Assembly Language ]

Lecture-05

M. M. Yasin
[email protected]
[ Topics Covered ]
• Performance Assessment Techniques
1. Clock Speed/Rate

2. CPI

3. MIPS Rate

4. SPEC Benchmark

5. Amdahl’s Law

Note: Topics covered are from chapter 3 of William Stallings book.

Fall 2019 - M. M. Yasin 1.2


[ Performance Assessment ]

Key Parameters in Evaluation Criteria


Size
Cost
Security
Reliability
Performance
Energy Saving and Power Consumption

Note: Evaluation Criteria may vary among systems that are designed for
different applications.

Fall 2019 - M. M. Yasin 1.3


1. Clock Speed/Rate

“Rate of pulses”
e.g., 1GHz processor, how many pulses per second???

Fall 2019 - M. M. Yasin 1.4


Clock Speed & Instructions per Second
Execution of an
instruction involves: Fetching,
Decoding,
Loading and Storing data
Performing arithmetic and logical
operations

Deduction: Simple clock speed is not an accurate/true measure of


different Processors’ Performance, hence requiring
complex assessment mechanism. (Why???)

Fall 2019 - M. M. Yasin 1.5


2. Instruction Execution Rate

Avg. Cycle Per Instruction:

CPIi : Number of cycles required for instruction type i


Ii : Number of executed instructions of type i for a given program
Ic : Instruction count for a program as the number of machine
instructions executed for that program

Processor Time to execute a given program:

𝜏 is constant cycle time and 𝜏 = 1/f, where f is the processor clock frequency.

Fall 2019 - M. M. Yasin 1.6


Since Instruction Execution involves Processor Cycle Time
as well as Memory Cycle Time
and these aren't equal,
therefore need to refine the above relation.

p: Number of processor cycles needed to decode/execute an instruction


m: Number of memory references needed
k: Ratio b/w memory and processor cycle time

Fall 2019 - M. M. Yasin 1.7


Above five performance factors are influenced/affected by
below four system attributes:
1. Instruction Set Design
2. How effective the Compiler is in producing an efficient
machine language program
3. Processor Implementation
4. Cache and Memory Hierarchy

Fall 2019 - M. M. Yasin 1.8


3. Processor Performance Measures

Millions of Instructions Per Second:

Millions of FLoating-point Operations Per Second:

Fall 2019 - M. M. Yasin 1.9


3. Processor Performance Measures
MIPs and MFLOPs are not true measures to evaluate system
performance.
Consider, A=B+C ;Assume all are in main memory
with CISC, add mem(B), mem(C), mem(A)

while on a typical RISC Machine,


load mem(B), reg(1);
load mem(C), reg(2);
add reg(1), reg(2), reg(3);
store reg(3), mem(A);

Execution times equal, but CISC = 1 MIPS and RISC = 4 MIPS

Fall 2019 - M. M. Yasin 1.10


4. Benchmarking
• Using a set of Benchmark Programs, called Benchmark Suite

• SPEC Benchmark (System Performance Evaluation Corporation)

I. SPEC CPU2006: For Processor-intensive applications rather than I/O

II. SPECjvm98: For Combined HW and SW aspects of the Java Virtual

Machine (JVM) client platform

III. SPECjbb2000 (Java Business Benchmark): for Evaluating server-side

Java-based electronic commerce applications

IV. SPECweb99: For WWW Servers

V. SPECmail2001: For systems acting as Mail Server


Fall 2019 - M. M. Yasin 1.11
4. Benchmarking

• Averaging a number of Benchmark Programs

To obtain a reliable comparison of the performance of various

computers, a number of benchmark programs are run on each

machine, then average the results.

These results could be averaged by,

Arithmetic Mean, Harmonic Mean or Geometric Mean.

Fall 2019 - M. M. Yasin 1.12


5. Amdahl’s Law

Consider a program running on a single processor such that

( 1-f ): fraction of the execution time involves code that is inherently

serial.

f: fraction involves code that is infinitely parallelizable.

T: Total execution time of the program using a single processor.

Fall 2019 - M. M. Yasin 1.13


5. Amdahl’s Law

Then the Speedup using a parallel processor with N processors that

fully exploits the parallel portion of the program is as follow:

Fall 2019 - M. M. Yasin 1.14


5. Amdahl’s Law
Amdahl’s law generalization to evaluate any Design and Technology

Improvement in a computer system:

Fall 2019 - M. M. Yasin 1.15


5. Amdahl’s Law
Suppose that a feature of the system is used (during execution) a

fraction of the time f, before enhancement, and that the speedup of

that feature after enhancement is SUf .

Then the overall Speedup of the system is:

Fall 2019 - M. M. Yasin 1.16

You might also like