CPE 626:Writing Scholarly Papers

Aleksandar MilenkovicE-mail: milenka@ece.uah.edu

Web: http://www.ece.uah.edu/~milenka

A. Milenkovic 2

Why Are We Bothering?

Useful for writing your final report.

Useful to know the style for doing future research.

This is how academics communicate You should know the technique Proven over time

A. Milenkovic 3

Sources

Stephen A. Edwards – lecture slides http://www.cs.columbia.edu/~sedwards/

Veljko Milutinovic V. Milutinovic,

“The Best Method for Presentation of Research Results,”IEEE TCCA Newsletter, Sep. 1996.http://tab.computer.org/tcca/NEWS/sept96/bestmeth.pdf

A. Milenkovic 4

Research Papers

Clear statement of the problem being addressed What has been done before and what is new Proposed solution Results achieved

Literature survey should have included the first two Final report should include all four

A. Milenkovic 5

Typical Outline

Abstract Introduction Related Work Description of problem solution Experimental results Conclusions and future work Bibliography

A. Milenkovic 6

The Abstract

Short 2-3 paragraphs 100-150 words

Introduce problem in first paragraph Describe your approach in second Brief conclusions and impact in third

Abstract must stand alone as pure English No bibliographic citations No mathematics

A. Milenkovic 7

A Sample Abstract

Improvement of the branch predictors has been one of focal points in the computer architecture research during the last decade, from two-level predictors to the use of genetic algorithms. The most of the research efforts try to use the real, already implemented, branch predictor sizes and organization for comparison and evaluation. Yet, a little is known about exact predictor implementation in Intel processors, apart from a few size hints in the Intel manuals and a valuable but unverified hacker efforts. On the other hand, Intel processors include performance monitoring counters that can count the events related to branches, and Intel provides a powerful VTune Performance Analyzer enabling the easy access to performance counters.

In this paper, we propose a series of simple experiments that explore the organization and size of a branch predictor, and use it to investigate Pentium III and Pentium 4 predictor implementations.

Experiments showed that BTB has 4 ways in both implementations. PIII predictor has just the local history component, while P4 has both global and local history components. Such knowledge could be used in further predictor research, as well as in the design of new, architecture-aware compilers.

A. Milenkovic 8

The Introduction

Describe both the area you’re working on and what you’ve found Cut to the chase early

“My field is interesting and here’s what I’ve done”

Don’t repeat the abstract

Orient the readers about what they should expect Some references are appropriate here,

but they need not be exhaustive

A. Milenkovic 9

Related Work

Describe the relevant work of others You won’t be exactly duplicating their work,

so contrast your results with theirs

Be respectful to authors:

Smith [1] describes a system for real-time scheduling…

Don’t use citations as nouns:

In [1], a system for real-time scheduling is described

A. Milenkovic 10

Goals of the Related Work Section

Orient the reader Part of your job is to figure out the state of the field and

communicate that clearly They need to know how your work is a huge step forward

from what you’ve done

Convince the reader you are knowledgeable Academics work by consensus There’s an accepted body of knowledge Yours won’t be added to it unless you understand and

acknowledge it, too

A. Milenkovic 11

Bibliography

Many different styles that differ only slightly All have the same goal of providing enough information

so that a reader can find what you read

Information in most citations: Author(s) of the work Title of the work Where it appeared (if a conference or journal publication) Date (at least year) at which it appeared

A. Milenkovic 12

Example Bibliographic Entries

A paper in a conference

[1] A. Milenkovic, V. Milutinovic, "Cache Injection: A Novel Technique for Tolerating Memory Latency in Bus-Based SMPs," Lecture Notes in Computer Science, vol. 1900, Proceedings of the 6th International Euro-Par Conference, Munich, Germany, August/September 2000, pp. 558-566.

Authors’ names, “Title,” Conference title, Conference Location, Month 20xx, pp. xx-yy.

A. Milenkovic 13

Example Bibliographic Entries

A book

[2] M. J. Smith, Application-Specific Integrated Circuits, Addison-Wesley, New York, 1997.

Author, Title, Publisher, 19xx.

A. Milenkovic 14

Example Bibliographic Entries

A journal paper

[1] A. Milenkovic,"Achieving High Performance in Bus-Based Shared Memory Multiprocessors," IEEE Concurrency, Vol. 8, No. 3, July-September 2000, pp. 36-44.

Author(s), “Title,” Journal, Vol. x, No. y, month, 20xx, pp. xx-yy.

A. Milenkovic 15

Example Bibliographic Entries

A technical manual

[1] M. D. Smith, Energy-efficient cache memories, Phd Thesis, University of Alabama in Huntsville, 2001.

Author(s). Title, Organization, 20xx.

A. Milenkovic 16

Citing Web Pages

URLs are not generally considered reasonable scholarly citations Not peer-reviewed Very fragile Can easily be changed

Best used for pointing people to projects or companies

A. Milenkovic 17

The Body

Describe what you’ve done with reasonable detail

Make sure to describe any unexpected or particularly difficult problems Goal is to simplify life for those in the future

Describe what you did and how it turned out

No requirement to include everything “Thanks, but that’s a little more information

than I needed to know.”

A. Milenkovic 18

Clear Writing

Be concise This is the main goal Think of it as an engineering problem:

o How can I communicate using the fewest words?

Be emphatic Avoid passive constructions

A. Milenkovic 19

Clear Writing

Avoid wordy idiomsInstead of Prefer

make assumption assume

is a function of depends on

is an illustration of illustrates

Avoid inactive verbs Avoid writing “to be” Not “In Smith, a clear-cut distinction was made …” Instead “Smith made a clear-cut distinction…”

A. Milenkovic 20

Clear Writing

Start each paragraph with a topic sentence:

You will be amazed by how much this helps. People will sing your praises. Professors will grade you higher. Your peers will call you “The Exalted One.” Little children will bow in your presence.

Technical writing is not a murder mystery: Explain whodunit immediately Suspense is not the point

A. Milenkovic 21

That vs. Which

Restrictive clauses use “that” Added to constraint the number of different things

being considered Don’t need commas

“Buildings that have white walls are common here.”

Nonrestrictive clauses use “which” Added to give additional information without changing

collection being considered Need commas

“Stucco buildings, which have white walls, are common here.”

A. Milenkovic 22

Experimental Results

Include important results that actually show something Avoid including endless tables or graphs

if they don’t further your point

Clearly label and explain how to interpret your graph What are you measuring? What are you controlling? What is your experimental setup?

A. Milenkovic 23

Experimental Results

Don’t do this:

0

10

20

30

40

50

60

70

80

90

1st Qtr 2nd Qtr 3rd Qtr 4th Qtr

East

West

North

A. Milenkovic 24

Experimental Results

0.5

K

1K

2K

4K

8K

16

K

32

K

sha

adpcm-decsusan

qsortpatricia

dijkstramad

cjpegblowfish

fftgs

djpeglame

rijndael

0

20

40

60

80

100

120

140

160

180

200M

iss

es

pe

r 1

00

0 in

str

uc

tio

ns

Cache Size [Bytes]

Benchmarks

Data Cache Misses

A. Milenkovic 25

Experimental Results

Graphs often illustrate trends and behaviors much more clearly than tables

Consider using graphics where you can Often much more succinct

A. Milenkovic 26

Conclusion Section

Here, summarize the results, say what worked and what didn’t, and explain what remains to be done

Be careful not to save the best for last I’ve read many papers where the contents of the conclusions

section should have been written in the introduction This is not a murder mystery

A. Milenkovic 27

Paper Format

Most conferences and workshops require a common format

Two columns 10 point

Usually Times Roman Single-spaced

Paper is precious: don’t waste it