Information Technology Reference
In-Depth Information
colloquial, poorly punctuated, and there were spelling errors (not reproduced here),
but it is direct and frank.
To improve the chance of a cache hit almost a complete recodewas necessary to
the data structure routines but no run with the new code showed any improve-
ment. The cache may have been too small but more likely the problemwas the
operating system and instruction prefetch getting the cache dirty. Also after
recoding a couple of extra machine instructions were needed for each access
so the saving of having a few more hits was lost.
For researchers educated in an English-speaking country, it is easy to forget that
English is not the first language of a great many readers. Simple writing allows these
readers to easily understand your work. Also, popular writing often uses shared cul-
tural elements as references. Slang (“home run”), values (“cool”), analogies (“like
turning left from the right lane”), and events (“the Northeastern power outage”) may
well be meaningless to readers living in other countries, or in other times. Even
dates can be confusing: in the United States, dates are often written month-day-
year, but elsewhere this notation almost invariably means day-month-year. Write
for everybody.
Examples
Use an example whenever it adds clarification. A small example often means the
difference between communication and confusion, particularly if the concept being
illustrated is fundamental to understanding the paper. People learn by generalizing
from concrete instances, and examples can give substance to abstract concepts.
In a semi-static model, each symbol has an associated probability representing
its likelihood of occurrence. For example, if the symbols are characters in text,
then a common character such as “e” might have an associated probability
of 12%.
Each example should be an illustration of one concept; if you don't know what an
example is illustrating, change it.
Examples can be blocks of text with a heading such as “Example 3.5” or detailed
discussions of specific instances where a technique can be used, but often an infor-
mative example is just a few words.
Large document collections, such as a repository of newspaper articles, can
be managed with the same techniques.
Special cases, such as the empty set, need to be handled separately.
Algorithms that involve bit manipulation cannot be efficiently implemented
in these languages. For example, Huffman coding is impractical because it
involves processing a stream one bit at a time.
