A TEACHING SCHEMA FOR MULTI-CORE PROGRAMMING

Yun Gao

and Xuejie Zhang

1,2

High Performance Computing Center, Yunnan University, No.2 North Cuihu Road, Kunming, China

School of Information Science and Engineering, Yunnan University, No.2 North Cuihu Road, Kunming, China

Keywords: Multi-core programming, Teaching schema, Positivity, Instance.

Abstract: As multi-core processors become more and more widespread, multi-core programming attracts attention of

the entire society. On this background, the multi-core programming curricula have been opened to the senior

undergraduate students in many universities. At present, the teaching schemas about multi-core

programming are scarcely mentioned. In this paper, we present a teaching schema on the specific

characteristics of multi-core programming curriculum, combining our teaching practice. The schema pays

attention to arousing students’ enthusiasm and positivity, showing instance demos and improving students’

consciousnesses of optimizing multi-core programs. During our practical teaching, the schema has obtained

a better teaching effect.

1 INTRODUCTION

With the rapid development of computer hardware

technologies and the increase of modern society

requirement to the information processing speed,

multi-core processor emerges as the times requires.

It is named multi-core processor that two or many

processing cores are integrated on a processor chip,

abbreviated as CMP. But the improved hardware

performance can not be transformed into high

software execution performance by the traditional

programming (Shameem and Jason, 2006) (LIU and

LIANG, 2007) (HE and WANG, 2007). The multi-

core programming on CMP has aroused general

concern of the entire society. On the background, the

correlative curricula have already been opened to the

senior undergraduate students in many universities.

However, which teaching schema should be taken to

effectively improve teaching efficiency and impel

students to understand the kernel of multi-core

programming?

Multi-core programming should be in the

programming category. In the recent years, many

teaching methods of traditional programming based

on one or more single-core processors have been

proposed, such as task driving method (REN and LI,

2008), practice teaching method (SUN, 2008), and

so on (Christopher Haynes, 2009) (Christoph W.

Kessler, 2006). These methods can be used by multi-

core teaching for reference. Multi-core programming

on CMP is different from traditional programming in

nature, so we should take special teaching schema.

At present, the teaching schemas about multi-core

programming are scarcely mentioned.

In the following sections, by analyzing the

specific characteristics of multi-core programming

curriculum, we present a teaching schema on its

specific characteristics and our teaching practice,

aiming at the senior undergrduate students, and then

show its practical effects. The schema pays attention

to arousing students’ enthusiasm and positivity,

showing instance demos and improving students’

consciousnesses of optimizing multi-core programs.

2 SPECIFIC CHARACTERISTICS

It is useful and necessary to analyze the curricular

specific characteristics for selecting teaching schema.

The multi-core programming curriculum has mainly

several following characteristics distinguished from

other programming.

2.1 Stronger Previous Foundations

It is well known that traditional programming in

C/C++ or Java hardly needs any curricular basis.

But, for designing a reasonable multi-core multi-

thread parallel program, it is helpful and necessary

to know the hardware frame of CMP, the scheduling

mechanism of operating system (OS), the basic

195

Gao Y. and Zhang X. (2010).

A TEACHING SCHEMA FOR MULTI-CORE PROGRAMMING.

In Proceedings of the 2nd International Conference on Computer Supported Education, pages 195-198

DOI: 10.5220/0002857401950198

 SciTePress

theory concerning design and analysis of algorithms

and to master a programming language.

Consequently, there are several necessary previous

curricular foundations for multi-core programming

curriculum, including Computer Composition

Principle, Operating System, Compiler Principle,

Design and Analysis of Algorithms, C language, and

so on. Multi-core programming curriculum is

usually offered to senior undergraduate or

postgraduate.

2.2 Non-figurative Theory

The theories of multi-core programming are more.

Multi-core program is mainly realized by a multi-

thread way on CMP. When realizing multi-core

programs, there are four questions or challenges

faced by parallel programming designers, and they

are multi-thread synchronization, multi-thread

communication, load balance and programming

scalability, which are also the kernel of multi-core

programming. (Shameem and Jason, 2006)

Nevertheless, students often mention that it is bald

and abstract to understand its theories and that it is

especially difficult to apply them during factual

programming. By this phenomenon, we must arouse

their enthusiasm and positivity in an example style,

which will be explained in the next section.

2.3 Needing Programming Practices

The theory must be applied finally in the practice,

and the theory is improved only by the practice.

According to this principle, for understanding the

nonfigurative theory of multi-core programming

concretely, it is necessary to adopt a practice way.

During the actual teaching, explanation in an

instance way can promote students’ intellect, and

programming practice by students themselves can

enhance students’ operating ability. Ultimately,

students can master the kernel of multi-core

programming in a high effective style.

2.4 Support of Hardware and Software

Multi-core programming must be developed based

on multi-core hardware and corresponding software

environment. Therefore, the support of experimental

environments is necessary to multi-core

programming curriculum. But multi-core computer

is not popularized to each senior undergraduate

student. So, many universities have built multi-core

labs for further promoting the laboratorial conditions

of multi-core teaching at present.

3 A TEACHING SCHEMA

In this section, we present a teaching schema of

multi-core programming curriculum on its specific

characteristics and our teaching practice, aiming at

the senior undergraduate students having learned

previous curricula. Our expectant teaching purpose

is that students can be interested in multi-core

programming, master its kernel, and establish multi-

core programming foundation for their future work

or study. Our schema detailed in the following pays

attention to arousing students’ enthusiasm and

positivity, showing instance demos and improving

students’ consciousnesses of optimizing multi-core

programs.

3.1 Arousing Students’ Enthusiasm

and Positivity

To whatever curriculum, including multi-core

programming, students’ study enthusiasm and

positivity is closely relative to the final study

efficiency. For multi-core programming curriculum,

we must firstly make students know its necessity and

importance before they study the concrete content.

In our schema, it is reached by answering two

following questions.

 Is it to study multi-core programming?

 Can the efficiency of application be improved

by multi-core programming?

3.1.1 Studying Multi-core Programming is

Necessary

Many questions can certainly be preferably solved

by single-thread programs. But some questions do

exist and must be solved by multi-thread parallel

programs. Students can cognize that multi-thread

parallel programming is necessary under some

circumstances by correlative examples. The

following is an example playing two pictures in a

mosaic style synchronously in a dialog window.

When realizing it with a single-thread program,

there are two visible disadvantages. One is that user

messages can not be responded in good time, and the

other is that two pictures can not be played

synchronously (see Figure 1). That is to say, the

requirement can not be reached by single-thread

program. If realizing it with a multi-thread program,

two foregoing disadvantages can be avoided and the

application requirement can be achieved perfectly.

Two picture can be played in a mosaic style in left

and right synchronously (see Figure 2).

CSEDU 2010 - 2nd International Conference on Computer Supported Education

196

Figure 1: When left played, right must wait until left is

finished.

Figure 2: Picture is played synchronously in left and right.

3.1.2 Efficiency of Application can be

Improved

Firstly, the executing time of multi-core program is

less than that of single-thread for a same multi-core

platform and a same question. Secondly, on multi-

core platform, genuine parallel can be achieved. So

the executable efficiency of application can be

improved in a genuine sense. The following is an

example calculating the pi value in an integral style

(Michael J. Quinn, 2003). The hardware

configuration of platform is dual-core CPU, Genuine

Intel(R) CPU T2080 1.73GHz.

The single-thread code and its executing result

are as follows.

num_steps = 1000000;

step = 1.0/(double)num_steps;

start = clock();

for (i=0; i<num_steps; i++) {

x = (i + 0.5)*step;

sum += 4.0/(1.0+ x*x);

}

pi = sum*step;

stop = clock();

The pi value is 3.141592653590.

The executing time is 0.109 seconds.

The dual-thread code and its executing result are

as follows with OpenMP.

num_steps = 1000000;

step = 1.0/(double)num_steps;

start = clock();

omp_set_num_threads(NUM_THREADS);

#pragma omp parallel for

reduction(+:sum) private(x)

for (i=0; i<num_steps; i++) {

x = (i + 0.5)*step;

sum += 4.0/(1.0+ x*x);

}

pi = sum*step;

stop = clock();

The pi value is 3.141592653590.

The executing time is 0.063 seconds.

From the results, multi-thread program takes less

time and obtains a higher efficiency than single-

thread program, and speedup ratio is approximately

1.73, that is to say, the performance of this program

is increased by 73%.

3.2 Showing Instance Demos

Multi-core programming can be realized on

multiform OS and in multiform programming

languages. In the limited curriculum hour, it is

almost impossible to explain all involved grammars

or functions. It is also useless all the same if these

grammars or functions can not be used in factual

programming. Therefore, in our schema, we only

explain some representative grammars or functions

to make students master the kernel of multi-core

programming. For the grammars or functions not

taught, students can completely master them in a

self-study way.

As an example, barrier is one of the effective

means to realize multi-thread synchronization, and a

thread meeting barrier can not continue until all

threads in the same parallel area reach the same

barrier. (Shameem and Jason, 2006) If this is

explained only by language, it is difficult for

students to use barrier. On the contrary, instance

demos can help students not only to master its usage

but also to sense the synchronization mechanism.

The following is an example calculating variable y

and z synchronously.

#pragma omp parallel shared(x,y,z)

num_threads(2){

int tid = omp_get_thread_num();

if (tid == 0) {

y = fn70(tid);

}

else {

z = fn80(tid);

}

#pragma omp barrier

#pragma omp for

for ( k=0; k<100; k++) {

x[k] = y+z+fn10(k)+fn20(k);

}

A TEACHING SCHEMA FOR MULTI-CORE PROGRAMMING

197

3.3 Improving Students’ Consciousness

of Optimizing Multi-core Programs

Multi-core program is an effective way to exhibit the

multi-core hardware performance. But it must be

mentioned that the efficiency of multi-core program

on CMP can be worse than serial program if multi-

core program is not designed in a better thinking.

Therefore, in our schema, we think much of

cultivating students’ consciousness to optimize

realized multi-core programs, a very important habit

to a software designer. In our factual teaching, we

can compare several different realizations for a same

problem.

As an example, for there is no relativity between

the data of the first “for” circle block and that of the

second “section” code block, one thread can

continue to run directly witout waiting. Under the

circumstance, we can import a “nowait” clause to

remove the hidden barrier (Michael J. Quinn, 2003),

and so the waiting time for synchronization is saved.

#pragma omp parallel {

#pragma omp for nowait

for (int k=0;k<m;k++){

fn1(k);

fn2(k);

}

#pragma omp sections private(y,z)

{

#pragma omp section {

y = sectionA(x);

fn7(y);

}

#pragma omp section {

z = sectionB(x);

fn8(z);

}

4 CONCLUSIONS

In this paper, we present a teaching schema of multi-

core programming curriculum on its specific

characteristics and our teaching practice, aiming at

the senior undergraduate students. The schema pays

attention to arousing students’ enthusiasm and

positivity, showing instance demos and improving

students’ consciousnesses of optimizing multi-core

programs. During our practical teaching with the

schema, it is proved that students’ positivity can be

aroused, students’ practical ability can be enhanced

and students’ consciousness to optimize programs

can be improved.

As multi-core programming is a newly arisen

technique, both its theory and its teaching schema

need further to be perfected all the same. In our

future work, we will further explore more effective

teaching schemas about multi-core programming by

our teaching practices.

ACKNOWLEDGEMENTS

The authors thank Intel-Yunnan University Multi-

core Technology Lab and High Performance

Computing Center of Yunnan University for

experimental support. This work was also supported

by the Project of Building up Characteristic

Disciplines under Grant No. TS11135 from Ministry

of Education of China and by Innovation Group

Project from Yunnan University of China.

REFERENCES

Shameem Akhter and Jason Roberts, 2006. Multi-core

Programming Increasing Performance through

Software Multi-threading, Intel Press.

Liu Jin-guang and Liang Man-gui, 2007. The Development

and the Software System Architecture of Multi-core

Multi-threading Processor. Microprocessors.

He Jun and Wang Biao, 2007. Research on Architecture

Design of Multi-core Processor. Computer

Engineering.

Ren Jing-ying, Li Ying, 2008. Application of Duty

Actuation Teaching Method in C++ Language

Programming, Combining Curricula with Infomation

Technologies.

Sun Geng, 2008. Project Teaching Method in C++

Programming, Computer Education.

Michael J. Quinn, 2003. Parallel Programming in C with

MPI and OpenMP, McGraw Hill Higher Education.

Christopher Haynes, 2009. Experience with an Analytic

Approach to Teaching Programming Languages,

Proceedings of the 29th SIGCSE Technical

Symposium on Computer Science Education.

Christoph W. Kessler, 2006. Teaching Parallel

Programming Early, Proceedings of Workshop on

Developing Computer Science Education.

CSEDU 2010 - 2nd International Conference on Computer Supported Education

198