Analysis about Introduction

The first article I select is "A Study of Branch Prediction Strategies" written by James E.Smith. In the introduction part of this article, all three moves are included, and it is clear to find out that all three moves are orgnized in a traditional order. 

The first move, highlighted in red, introduces the main topic("it is well known that... ") and provides a brief introduction about the common ways to reduce the delay in branch technology("one can attempt to ...").

Then, in the second move, shown in blue, the author mentions the problems and limitations using common approaches("Unfortunately, a wrong prediction may lead to...").

Finally, in the third move, the author provides an introduction about the focus of this paper and a brief outline. ("This paper discusses...").

• It is well known that in a highly parallel computer system, branch instructions can break the smooth flow of instruction fetching and execution. This results in delay, because a branch that is taken changes the location of instruction fetches and because the issuing of instructions must often wait until conditional branch decisions are made.
  
• To reduce delay, one can attempt to predict the direction that a branch instruction will take and begin fetching, decoding, or even issuing instructions before the branch decision is made. Unfortunately, a wrong prediction may lead to more delay if, for example, instructions on the correct branch path need to be fetched or partially executed instructions on the wrong path need to be purged. The disparity between the delay for a correctly predicted branch and an incorrectly predicted branch points to the need for accurate branch prediction strategies. 
  
• This paper discusses branch prediction strategies with the goal of maximising the likelihood of correctly predicting the outcome of a branch. First, previously suggested branch prediction techniques are discussed. Owing to the large number of variations and configurations, only a few representative strategies have been singled out for detailed study, although several are mentioned. Then, new techniques are proposed that provide more accuracy, less cost, and more flexibility than methods used currently.
  
• Because of the wide variation in branching behaviour between different applications, different programming languages, and even individual programs, there is no good analytic model for studying branch prediction. For this reason, we used instruction trace data to measure experimentally the accuracy of branch prediction strategies. 

The second article is "Computer Interconnection Structures: Taxonomy,Characteristics, and Examples" written by George A. Anderson and E. Douglas Jensen. In the introduction part of this article, all three moves are included and orgnized in a traditional order. 

The first move, highlighted in red, introduces the main topic("one of the most active areas in computer science architecture is...") and provides a brief introduction of the popular concepts in this field, such as "distributed processor" and "distributed-function computers".

Then, in the second move, shown in blue, the author mentions that there is not enough comparison being made to evaluate the differences between each designs and approaches("A discouraging aspect of this activity, however, is the almost total lack of...").

Finally, in the third move, the author provides an introduction about the focus of this paper and a brief outline. ("Our paper is an attempt to begin filling this need.")

• Currently, one of the most active areas in computer architecture is the interconnection of computers to form systems which are called "distributed processors," "distributed-function computers," "computer networks," and similar names. These systems range in organization from two processors sharing a memory to large numbers of relatively independent computers connected over geographically long distances. A discouraging aspect of this activity, however, is the almost total lack of published information describing the rationale for various designs, or comparing the results achieved by various approaches. In part, the authors believe this condition exists because there has been no common context in which such discussion could take place, no set of design issues, no list of system characteristics to be traded off, and, in fact, not even a common nomenclature for system identification. Our paper is an attempt to begin filling this need. In it we present a naming scheme, or taxonomy, for identifying various systems of interconnected computers, and we discuss design decisions and system characteristics which we believe are germane to these architectures.
  
• The authors know of only one other general taxonomy for interconnected computers and that is a brief one (having different dimensions) with few system characteristics and no nomenclature [$1~.w74]. Some interconnection topology issues are also considered in [CHEN74] and [THtTR72], al- though these are primarily concerned with the next lower level of the interconnection design--control and communication. One level beneath these are a number of papers dealing with the design of "explicit" switches, such as crossbars [PIPe75] and permutation/sorting networks [THVR 74]. In addition, there is a wide variety of digital...

The third article is "Implementation of Precise Interrupts in Pipelines Processors" written by James E. Smith and Andrew R. Pleszkun. In the introduction part of this article, all three moves are included and orgnized in a traditional order. 

The first move, starting with "Most current computer architectures are based on a sequential model of program execution in which an architectural program counter sequences through instructions one-by-one...", defines the key term, interrupts in Pipelined Processors, and provides a basic procedure about how it works.

Then, in the second move, "If the saved process state is inconsistent with the sequential architectural model and does not satisfy the above conditions, then the interrupt is imprecise.", the author introduces a limitation of current approach in certain situation. (language signal: "If... is inconsistent, then ... is imprecise").

Finally, in the third move, the author provides an introduction about the focus of this paper and a brief outline. ("This paper describes(language signal) and compares ways of implementing precise interrupts in pipelined processors...").

Conclusion: Although some of them have subtle differences in move one, all three papers I have found are strictly following CARS model.