Assignment 1A: Review of two research papers Alessio Bucaioni Global Software Engineering European Master (GSSEM) abi11001@student.mdh.se PAPER 1 Worst-Case Execution Time Analysis for Dynamic Branch Predictors, Iain Bate and Ralf Reutemann. Is the paper well organized? In my opinion the paper is well organized and conforms to the standards. It consists of the following sections: Title, authors and affiliations Abstract Introduction Main sections Conclusions References Actually some sections are missing (e.g. contents, list of tables, list of figures..) but it still remains clear. Comment to the following sections. Title In my opinion the title is well composed. It is shorter than ten words and it does not contain any symbols or cryptic words. I think it is enough informative and exhaustive even if does not say nothing about the analysis model for bimodal branch predictors. Abstract The abstract is very concise - only one hundred and twenty-three words - but it does not have a lack of clarity. In the first eight lines the authors introduce the problem; then, they move to the explanation of the work done and in the last three lines they show a possible usage of it. No acronyms and references have been found in abstract. Introduction The introduction actually consists of three sections: Introduction, Branch prediction techniques and Related work. In my understanding, in the first section the authors briefly go through to the background and the definition of the problem – first paragraph - , their contributions to the problem, – second paragraph – the related works – third and fourth paragraph – and the structure of the paper – last paragraph. They also use some acronyms and references for the first time in the paper. Each reference is well marked: their works are clearly separated from the referred work. In the others section, Branch prediction techniques and Related work, the authors go in deep into the explanation of the background of the problem and they discuss a series of other approaches highlighting the bad points of these. Main sections In the paper we can find three main sections: Static analysis of bimodal predictors, Globalhistory predictors and Evaluation of the approach. The first section is again divided in other four subsections. I think this division is needed because they use an incremental approach to explain their method: they first start with an overview of the approach; then they fix the terminology, explain some assumptions, used later in the paper, and they explain two analyses for two basics constructs. In these subsections we can find three tables, a figure and a lot of formulas. I think each of them is shown with the appropriate notation. In the third subsection we can also find a theorem with a proof. It is not a structured proof but still clear and complete. In the other Main sections we have further analysis and an evaluation of the approaches presented. The evaluation is supported by a series of very well commented graph. Conclusions The authors take the stock of their work. In paragraphs one and two they summarize the works done while in paragraph three they point out some benefits of their approach in contrast to existing approaches. They also show limits of their work in paragraph four while in the last paragraph they propose some feasible future works. References This section contains both references from book and from other paper. It is very clear and well written. The only thing missing is the ISBN for the referred books. It could be useful to search book in a fastest way. Comment on the language used in the paper Although the language is not so formal it is used in a very formal way. The composition of the sentences is very clear and makes the reading easier. The authors, in the writing, use the first plural person “we” and try to do not use frequently the same words. Obviously the paper contains some technical words and acronyms related to the field of the work computer science of course - but each of them, especially the acronyms, are always explained before being used. General comments to the paper As I said this is a good and interesting paper. All the things used conform to the standard directives. The paragraphs are well structured. Figures and tables are always presented with caption and comments. The formulas and the proof are not confused or unreadable and well-marked. PAPER 2 Improving Direct-Mapped Cache Performance by the Addition of a Small Fully-Associative Cache and Prefetch Buffers, Norman P. Jouppi. Is the paper well organized? In my understanding the paper could be considered well organized. It contains the following sections: Title, authors and affiliations Abstract Introduction Main sections Conclusions Acknowledgments References Of course, some sections are missing but I do not think this affects the completeness of the paper. Comment to the following sections. Title Although the title is too long - fifteen words - and it contains a lot of technical words, it gives a good idea of the topic which will be treated. However in my opinion, it has a lack of appeal. Abstract Also the abstract is too much long. It’s quite boring because it looks like a list of presumably – feasible hardware techniques to improve cache performances and, even if there are not acronyms or references, sometimes the sentences are hard to understand immediately. Introduction Introduction is made up of only one section. It does not contain sections about related works, problem definition and it is not exhaustive about the work reported in the paper. Among others the only sentence which speaks about this it is quite similar to the sentence used in the Abstract section: (abstract sentence)This paper presents hardware techniques to improve the performance of caches. (introduction sentence) This paper investigates new hardware techniques for increasing the performance of the memory hierarchy. The introduction contains only two paragraphs. The first one is quite huge and tries to explain the background of the problem. The paragraph two is about the paper structure. In this section we find the first figure/table of the paper. Despite of the overall section, the table is well inserted, with its caption and it is referred in the same page. Main sections Main sections are three. Each of them, except section two - the first section - have some subsections. The first section specifies all the configurations interesting in that research. The section three speaks about miss caching and victim caching while the last section is about compulsory misses and stream buffers. All the acronyms are inserted without a preceding explanation. In the section two and also forward in the paper, the author does not always place the figures and the tables in the same page in which they are referred. However they are always captioned and well formatted even if the font used is quite tiny. Conclusions In paragraphs one, two and three the author speaks about the results of his work. He point out some benefits, especially for the stream buffers. In paragraph four he begins to speak about some limitations of these techniques. In the paragraph six he shows a graph which describes the improvement, in term of system performances, had with his techniques. In the ending paragraph the author describes some possible future works. Acknowledgments This is the shortest section. The author just cites all the people which, in some way, helped him. References In my opinion this section is enough clear and well don - except the reference five -. However the references in the text – above - are inserted without any further information, so sometimes it is difficult to understand what topic in the text it is related with the references. Comment on the language used in the paper The language used in this paper, in my understanding, it is very simple and, sometimes, not appropriate to the circumstance. The sentences are very long and in some cases this results in a hard understanding. I do not agree with the structure of the paragraph. As I said before I think the author uses too much technical words, acronyms and numbers. General comments to the paper The overall structure of the paper is good, but there are some mistakes in the sections. I think the author just does not care about some rules so, as I said before, the work, sometimes, looks quite fuzzy. For example the figures and the tables are indexed with the same notation and this could lead the readers to get confused. One appreciable thing is that the author always uses the italic for the core concepts.