- Email: [email protected]
Two books in one on GaAs design and new RISC architectures
Book reviews Useful RISC text from ARM development team leader Furber, S B VLSlRISCarchitecture and organization Marcel Dekker, NY, USA (1989) pp 371 The book is a welcome addition to the expanding literature on RISC systems. It is particularly valuable because the author belongs to an industrial team which created one of the existing RISC systems--the Acorn RISC Machine (ARM). Many details involved in the design of the ARM are included; about a quarter of the book (chapter 4) is dedicated to this topic. It constitutes one of the most important assets of the book. Another remarkable asset of the book is the author's success in squeezing in a discussion on a great variety of computer architecture issues (processor design, technology, memory, VLSI, cache, software) within a relatively small number of pages and a single chapter (chapter 1). The author's insights and analysis at the end of the book are interesting, instructive and stimulating. According to the author's preface, the book is written for 'final-year undergraduate and postgraduate students of computer science, and
for engineers who study or design computer architectures.' This reviewer fully agrees with this statement, but the following caveat should be added: 'the reader should have some basic knowledge about the concept of RISC'. A reader who has no previous knowledge of RISC will find the first chapter of the book (pp 1-69) somewhat difficult to follow. Granted that there does not exist a formal definition of RISC,there do exist some definite concepts and practices, implemented in many systems. Nowhere in the book does the author offer a concise, clear and concentrated presentation of what are the ideas, principles and practices of RISC and what are its advantages and disadvantages (this information should have been placed within the first 20 pages). This information is in the book, but it is presented in separate small pieces scattered over chapter 1. Throughout chapter I the concept of RISC is mentioned a number of times without the benefit of sufficient and timely elucidation. The uninitiated reader is recommended to read chapters I and 2 and then reread chapter1 again for a better application of the author's very
helpful, informative and important statements. The lists of references in the book leave much to be desired. There exist several books dedicated to RISC systems and architecture which appeared prior to this book, and there are several textbooks (such as J P Hayes, McGraw Hill, 1988) on computer architecture and organization, which contain chapters or sections dedicated to RISC. The presentation of the various instruction sets tables for various systems (IBM ROMP, Acorn, SPARC, MIPS R2000, AT&T CRISP, Am29000, HP Precision Architecture, M88000, Intel 80960, Transputer) is vague and incomplete. The information would have been more useful had the author just reproduced the tables from the manufacturers' manuals. The book is priced too high (over $100) for class adoption as a main text, but it should be included in the reading list in undergraduate and graduate courses on computer architecture and organization.
D Tabak George Mason University VA, USA
'Two books in one' on GaAs design and new RISC architectures Milutinovic, V (Ed.) Microprocessor design for GaAs technology Prentice Hall, Englewood Cliffs, NJ, USA (1990) £43.85 pp 330 There's a compelling argument that if A is good and B is good, then A plus B is even better. Sometimes this is valid and sometimes it isn't: ice-cream is good and so is pizza, but nobody ever got rich selling ice-creamtopped pizza! Milutinovic has combined two hot topics in modem microprocessor systems design: RISC architectures and gallium arsenide technology. Whether the sum of these two topics is, for the reader, greater than the whole is unclear. The first third of the book is largely devoted to topics in the field of GaAs
Vol 14 No 6 July/August 1990
technology. Gallium arsenide devices are very much faster than their silicon equivalents, but they are not electrically compatible with conventional logic elements. GaAs devices have very small voltage swings (i.e. as low as 0.6V) and therefore correspondingly small noise margins. Moreover, it is much harder to work with GaAs than with silicon, which means that device yields are lower. Several chapters (each contributed by an authority on GaAs technology) look at the present situation and discuss developments in the production of GaAs VLSI devices. Much of this section is aimed at the student of semiconductor technology (i.e. the electronic engineer or the
physicist). Topics covered range from device physics to GaAs logic elements to device layout and chip fabrication. This section provides the student with a good and quite readable account of the current state of GaAs technology. It would be suitable for a student taking a course in VLSI design who wanted to look at the GaAs world. The remaining two thirds of this book are really devoted to microprocessor architecture with a strong RISC flavour. Although the various contributions on architecture are peppered with GaAs terminology, ultimately it does not really matter whether these architectures are implemented by GaAs, plain silicon, or even steam valves.
421
B o o k r e v i e w s .... The first chapter in the architecture section is entitled 'Design concepts for microprocessor-based GaAs systems'. Do not be fooled by this title - - it is really an introduction to RISC architectures. This emphasis on RISC architectures is entirely sensible since RISC processors use less silicon (or Gag,s) real-estate to achieve the same power as a more traditional complex instruction set architecture of the 68020 or 80386 variety. If it is harder to fabricate large GaAs chips than silicon chips, it follows that it is reasonable to explore architectures that do not waste real-estate by implementing instructions that are seldom executed. The second paper in the architecture section covers GaAs versions of conventional bit-slice components (e.g. the 2900 series), together with components ranging from SSI devices to DRAMs to 8-bit microprocessors. In addition to this overview, more advanced architectures such as
systolic arrays and even waferscale GaAs are introduced. Two chapters cover 32-bit GaAs processors; one from CDC and one from McDonnel Douglas. Both chapters provide in-depth coverage of the respective architectures (unlike many overview papers). I would certainly recommend these two chapters to any student of computer architecture, not least because they cover so many aspects of processor architecture and implementation. The final paper in this book is from General Electric and describes an experimental 32-bit GaAs processor (much of the research was performed in the mid 1980s). This paper describes a RISC-like processor and provides a lot of interesting numeric (statistical) data on the performance of the machine. Returning to my opening comments, this book is really two books: a GaAs technology primer and a description of several RISC-like
architectures. It is difficult to say whether the sum of its parts is greater than the whole. A student specializing in device physics might be interested only in the GaAs technology part, while a student of computer science might be interested only in the architecture part. However, both parts of this book are interesting and authoritative. The editor has resisted the temptation to include too many shallow papers and has selected a smaller number of good in-depth papers. Since there is a dearth of good texts covering new architectures (as opposed to the philosophy of computer architecture), I would recommend this book for its coverage of architectures alone. The level is suitable for final year undergraduates or for students beginning postgraduate research into computer architecture.
Alan Clements Teesside Polytechnic, UK
o ,.Inmn~ qliD. ,.uumma~ ~ml~m.. -,1Inn., O ,.1mira..imm~n..,
O O O ,.,ginuum
O O ~ O
0 0
~--~~~~-~_--
~ O
,qMINm', o .,aiUlU~
O o
_~ ~ ~ - - -
General Editor: Professor Keith Baker, University of Reading, UK Image and Vision Computing is the international, interdisciplinary journal on computer-based imaging and vision processing methods, theory and application. Coverage • • • • •
computer-based imaging image analysis image understanding image enhancement and restoration vision system architectures
• • • • •
computer vision application.~, machine vision medical imaging optical flow remote sensing
Published quarterly For further details and a FREE sample copy, please write to: Donna McMahon, Butterworth Scientific Ltd., Westbury House, Bury Street, Guildford, Surrey GU2 5BH, UK Telephone: (0483) 300966 Fax: (0483) 301563 Telex: 859556 SCITEC G
Butterworth Scientific 422
Microprocessors and Microsystems
for engineers who study or design computer architectures.' This reviewer fully agrees with this statement, but the following caveat should be added: 'the reader should have some basic knowledge about the concept of RISC'. A reader who has no previous knowledge of RISC will find the first chapter of the book (pp 1-69) somewhat difficult to follow. Granted that there does not exist a formal definition of RISC,there do exist some definite concepts and practices, implemented in many systems. Nowhere in the book does the author offer a concise, clear and concentrated presentation of what are the ideas, principles and practices of RISC and what are its advantages and disadvantages (this information should have been placed within the first 20 pages). This information is in the book, but it is presented in separate small pieces scattered over chapter 1. Throughout chapter I the concept of RISC is mentioned a number of times without the benefit of sufficient and timely elucidation. The uninitiated reader is recommended to read chapters I and 2 and then reread chapter1 again for a better application of the author's very
helpful, informative and important statements. The lists of references in the book leave much to be desired. There exist several books dedicated to RISC systems and architecture which appeared prior to this book, and there are several textbooks (such as J P Hayes, McGraw Hill, 1988) on computer architecture and organization, which contain chapters or sections dedicated to RISC. The presentation of the various instruction sets tables for various systems (IBM ROMP, Acorn, SPARC, MIPS R2000, AT&T CRISP, Am29000, HP Precision Architecture, M88000, Intel 80960, Transputer) is vague and incomplete. The information would have been more useful had the author just reproduced the tables from the manufacturers' manuals. The book is priced too high (over $100) for class adoption as a main text, but it should be included in the reading list in undergraduate and graduate courses on computer architecture and organization.
D Tabak George Mason University VA, USA
'Two books in one' on GaAs design and new RISC architectures Milutinovic, V (Ed.) Microprocessor design for GaAs technology Prentice Hall, Englewood Cliffs, NJ, USA (1990) £43.85 pp 330 There's a compelling argument that if A is good and B is good, then A plus B is even better. Sometimes this is valid and sometimes it isn't: ice-cream is good and so is pizza, but nobody ever got rich selling ice-creamtopped pizza! Milutinovic has combined two hot topics in modem microprocessor systems design: RISC architectures and gallium arsenide technology. Whether the sum of these two topics is, for the reader, greater than the whole is unclear. The first third of the book is largely devoted to topics in the field of GaAs
Vol 14 No 6 July/August 1990
technology. Gallium arsenide devices are very much faster than their silicon equivalents, but they are not electrically compatible with conventional logic elements. GaAs devices have very small voltage swings (i.e. as low as 0.6V) and therefore correspondingly small noise margins. Moreover, it is much harder to work with GaAs than with silicon, which means that device yields are lower. Several chapters (each contributed by an authority on GaAs technology) look at the present situation and discuss developments in the production of GaAs VLSI devices. Much of this section is aimed at the student of semiconductor technology (i.e. the electronic engineer or the
physicist). Topics covered range from device physics to GaAs logic elements to device layout and chip fabrication. This section provides the student with a good and quite readable account of the current state of GaAs technology. It would be suitable for a student taking a course in VLSI design who wanted to look at the GaAs world. The remaining two thirds of this book are really devoted to microprocessor architecture with a strong RISC flavour. Although the various contributions on architecture are peppered with GaAs terminology, ultimately it does not really matter whether these architectures are implemented by GaAs, plain silicon, or even steam valves.
421
B o o k r e v i e w s .... The first chapter in the architecture section is entitled 'Design concepts for microprocessor-based GaAs systems'. Do not be fooled by this title - - it is really an introduction to RISC architectures. This emphasis on RISC architectures is entirely sensible since RISC processors use less silicon (or Gag,s) real-estate to achieve the same power as a more traditional complex instruction set architecture of the 68020 or 80386 variety. If it is harder to fabricate large GaAs chips than silicon chips, it follows that it is reasonable to explore architectures that do not waste real-estate by implementing instructions that are seldom executed. The second paper in the architecture section covers GaAs versions of conventional bit-slice components (e.g. the 2900 series), together with components ranging from SSI devices to DRAMs to 8-bit microprocessors. In addition to this overview, more advanced architectures such as
systolic arrays and even waferscale GaAs are introduced. Two chapters cover 32-bit GaAs processors; one from CDC and one from McDonnel Douglas. Both chapters provide in-depth coverage of the respective architectures (unlike many overview papers). I would certainly recommend these two chapters to any student of computer architecture, not least because they cover so many aspects of processor architecture and implementation. The final paper in this book is from General Electric and describes an experimental 32-bit GaAs processor (much of the research was performed in the mid 1980s). This paper describes a RISC-like processor and provides a lot of interesting numeric (statistical) data on the performance of the machine. Returning to my opening comments, this book is really two books: a GaAs technology primer and a description of several RISC-like
architectures. It is difficult to say whether the sum of its parts is greater than the whole. A student specializing in device physics might be interested only in the GaAs technology part, while a student of computer science might be interested only in the architecture part. However, both parts of this book are interesting and authoritative. The editor has resisted the temptation to include too many shallow papers and has selected a smaller number of good in-depth papers. Since there is a dearth of good texts covering new architectures (as opposed to the philosophy of computer architecture), I would recommend this book for its coverage of architectures alone. The level is suitable for final year undergraduates or for students beginning postgraduate research into computer architecture.
Alan Clements Teesside Polytechnic, UK
o ,.Inmn~ qliD. ,.uumma~ ~ml~m.. -,1Inn., O ,.1mira..imm~n..,
O O O ,.,ginuum
O O ~ O
0 0
~--~~~~-~_--
~ O
,qMINm', o .,aiUlU~
O o
_~ ~ ~ - - -
General Editor: Professor Keith Baker, University of Reading, UK Image and Vision Computing is the international, interdisciplinary journal on computer-based imaging and vision processing methods, theory and application. Coverage • • • • •
computer-based imaging image analysis image understanding image enhancement and restoration vision system architectures
• • • • •
computer vision application.~, machine vision medical imaging optical flow remote sensing
Published quarterly For further details and a FREE sample copy, please write to: Donna McMahon, Butterworth Scientific Ltd., Westbury House, Bury Street, Guildford, Surrey GU2 5BH, UK Telephone: (0483) 300966 Fax: (0483) 301563 Telex: 859556 SCITEC G
Butterworth Scientific 422
Microprocessors and Microsystems