Summer Reading Reading and writing the
All year long you’ve been waiting for the chance to spend just a little time at the beach. You’re going to read, think long thoughts, and between chapters, do nothing more than watch the waves come and go in front of you. So, what are you going to read? Well, here are a couple of lists to help you decide. I would suggest you choose one title from List No. 1 and one from List No. 2. The first list has 10 books. The second list only has one. Once you’ve decided, take your two titles - one from the first list and one from the second list - and head out. Find the beach, your spot on the beach, set up your umbrella, the towel and the lunch, lather up, settle down, and dig your heels into the sand. Then follow these instructions: * Read one chapter of the book you picked from the first list. By the end of the afternoon you’ll have three things: a) A desire to retreat to a shady place to sink more deeply into the book from the first list; b) An even bigger desire to retreat to the kitchen table to sink more deeply into the (lone) book from the second list; and c) A healthy sun burn.
The Winds of War Pompeii Arundel The Snows of Kilimanjaro and Other Stories The No. 1 Ladies Detective Agency The Tears of the Giraffe Morality for Beautiful Girls The Kalahari Typing School for Men Intruder in the Dust Slaughterhouse 5
Engineering the Complex SOC --
The Great American Novel Chris Rowen is President and CEO of Tensilica Inc. He’s been in that role for quite some time now. Obviously he’s mastered the job, or he wouldn’t have found time to write a book this past year, while also serving as CEO. Of course, there's always the possibility that being a CEO isn’t that time consuming and writing a book is a clever way to wile away those extra hours that come as a perk of the job. In any case, what follows is a somewhat eclectic conversation with Chris about his book, his background, and what inspired him to attempt The Great American Novel in the midst of his busy life. Q: Give me a quick thumbnail sketch of your background. Rowen: I grew up on both coasts. My family moved just about every 5 years or so - L.A., Boston, Washington D.C., and so forth. I attended college at Harvard and studied physics. As physics major, I took this one electrical engineering class for physicists. It was on the strength of that class that I got a summer job at a little company called Intel. At Intel, I worked on DRAMs and after I graduated in 1978, I went to work full time for Intel in Oregon. It was during that time that I realized I just wasn’t going to be able to accomplish what I wanted to accomplish without an engineering degree. So I came to Stanford to get my Ph.D. and it was through my studies of semiconductor technology that I met John Hennessy, which is how I was inspired to study microprocessors. I got in on the ground floor on the MIPS effort and was one of the architects of the RISC architecture. John Hennessy and I are still in close contact. John wrote the forward for my book. I worked on all sorts of efforts at MIPS. I worked on the tools, on system design, I ran the tools department at the company. After 1992, I went to Europe as the European CTO and was the field development guy for Silicon Graphics after they bought the company. I was back in Silicon Valley by 1996, where I worked at Synopsys as General Manager of the Design Reuse Group at Synopsys. There, I recognized a fundamental transformation in the potential role of microprocessors. But the idea of pushing Synopsys in that direction was really premature. It was [inappropriate] to ask Synopsys to become a microprocessor company. So in 1997, we decided to found Tensilica. Q: Why did you decide to write this book? Rowen: The book had its beginning last year as a concept. I was clearly seeing that Tensilica was doing a pretty effective job of talking about its products, talking about opportunities for improvement that were available to our customers by using our products. But more importantly, it was clear that what we were talking about is at the center of a fundamental change that’s taking place today. We were talking about the economics of building big chips and the knowledge of how to make big chips, but we weren’t just talking about substituting Product A with Product B. I recognized we needed a different way to help people wrestle with the big questions of economics and technology. A book like this one seemed to offer an effective way to do this. Q: How did you decide what to write? Rowen: At the beginning [of the project], we had a relatively modest goal. We were targeting on a high level - the "marketing-esque" of how to communicate broad ideas - and so we started the process of writing the first several introductory chapters. However, as I got into the process of writing the book, I recognized there was an opportunity here to deal with the deeper questions, to analyze more completely the business context [within which the technology is utilized], and to take a systematic look at the technical hurdles standing in the way of adoption. The book became more substantial in scope and I decided to deal with issues that were both deep and subtle - the issues that are pervasive in the design flow. We were particularly interested in how people were going to use programmability in functions that had never been perceived as being programmable before, and how to put it all together. One of the things that make any large system fascinating - large banking systems, large communication networks, a space shuttle, for instance - is that people typically see modularization as a way to solve each sub-problem [within these large systems] in isolation. But, many would argue if you’re making something that must run for 100 hours on a AA battery, or must sell at Fry’s for $49 - you’re working on a different scale than a communications network or a space shuttle. But all of these designs require a highly efficient way to handle the complexity [inherent in the design]. That makes high-volume electronic system design, in fact, one of the grand problems of the modern world. There are both hardware developers and software developers, and both have strong roles to play in these big electronic systems. They have different perspectives and different legacy needs. But at the same time, the needs of the system architect who’s working on the end application must take those requirements down into the component parts. This combination of bottom-up and top-down design strategies is a function of the complexity of very big systems and very big chips. If you only had to deal with the cost of these systems, or only with the complexity of these systems, the problems would be much easier. But only the most powerful of tools allow you to work on the cost and the complexity at the same time. My book, at its heart, deals with the things that are highly efficient from a technological and economic perspective - including the hardware and software in our favorite gadgets. As I dug down into the problems to make it easier for both engineers and managers to understand, I found there was a hole in my own thinking and therefore in the thinking of the company. The process of sorting [out the gap in that thinking] became the vehicle to take quantum steps in our understanding at Tensilica of the key aspects of the technology, and the differences between how software people and how hardware people look at an electronic design problem. This book is organized around breakthroughs in our own thinking - in the process of writing this book, we developed our own product strategy, which we now own. Q: Does this mean every CEO should sit down and write a book? Rowen: [Laughing] I’m certainly not saying that every CEO should sit down and do this. Everybody’s got different skills, but it is a good discipline to articulate systematically what you’re trying to accomplish. I felt I was a reasonably good writer when I started the project. But by the end, I could tell that my productivity was improving. At the beginning, sitting down and writing 1000 words per evening was a very painful [process]. But with time, it became relatively straightforward to turn ideas into well-structured prose arguments. The process became easier and easier. Q: How long did it take to write the book? Rowen: The writing process lasted from April to December 2003, but the first three months were actually used to work out what we were trying to accomplish, our target audience, our ‘voice’ if you will. The bulk of the first half of this year was involved with the mechanics of going through endless technical reviews of the text, the prose reviews, the production process, sorting out the illustrations. Q: Who did the illustrations? Rowen: We subcontracted out the illustrations, although I had already done all the illustrations myself with the MS Word editor. It was the job of the illustrator to convert those sketches into polished images. We worked closely with the illustrator to make the book look decent, but it’s not the job of the illustrator in this situation to create bold new insights. If an idea wasn’t in my original illustrations, it wasn’t in the final book. Q: How did you balance family and writing a book and being a CEO all at the same time? Rowen: [Laughing] It required a lot of patience on the part of the family. There was obviously a level of occasional resentment whenever I disappeared after dinner to write, but typically that was after the kids were already in bed. Or I would work early in the morning. I would walk around the corner to a local coffeehouse and type madly before coming to work. Q: What are the 3 things I’ll know after I’ve read your book? Rowen: Let’s see ... Number one, you’ll understand why and where the use of configurable processors changes the fundamental ROI and engineering challenges in SoC design Number two, you will understand how to take big technical problems, complex, multi faceted end applications, and break them down into the appropriate set of processors, memories, interconnects, software threads, and other components, that form an efficient flexible implementation of that system Number three, whether you’re a hardware guy or a software guy who understands specifically how to use configurable processors, you'll see the wide range of problems associated each specific task at hand - network processing problem, DSPs, etc. Q: How’s the book doing? Rowen: Prentice Hall is very happy so far. In fact, they’re getting ready to do a second printing. And, it’s risen up in the rankings on Amazon quite nicely. Also for us, it’s turning out to be a useful tool and very good for prospective customers. I’ve been pleasantly surprised by the number of academics who have expressed interest. It’s obvious that there are some pretty big holes out there in terms of teaching materials about modern multi-processor SoC technology. Instructors know that their students need to be ready for the world as it exists, but there’s not really anything out there like this book to teach from. The book has a bias, of course. It doesn’t attempt to say, here are all of the different ways to design chips. Clearly, it’s saying that accepting the fact that the automatic generation of processors is a reality today, what does the existence of this kind of technology do to the nature of big SoC designs? The book talks both about the techniques and, more importantly, about the implications across all of the other dimensions of the problem - the systems architecture and the organization needed to tackle the project. I believe that once you get under the surface of the problem and you start to understand what the nature of the problem is, the book deals in a fairly reasonable and balanced way with the different issues involved. It wouldn't have made sense to have 1000 pages on the subject, but at 450 pages I think it’s a pretty thorough treatment. Q: There seems to be teaching materials for students and teaching materials for practitioners? How do we assemble materials that help people make the leap from being a student to being a practitioner? Rowen: There's a need to help people try to bridge between the two, to understand the difference between being a student and being a working engineer. The difference, of course, is that the working engineer is plunged into the economic reality of building products to make money, not just building products to learn something. The working engineer isn’t just trying to figure out which answer is the fastest or the lowest-power solution, but which product is the most appropriate to help the company make money. The other difference for working engineers is that they’re part of a team. You know, no one can release a product without doing it as a team. Today, some engineering education does a relatively good job of simulating teamwork - how do you break down a problem and communicate a goal, or goals, within a group - but I think the biggest difference between academics and practice is the complexity of the environment. The practitioner works on a team within the context of making money, within the context of the technology. This book is about the marriage of technical ideas and business ideas - how you proliferate something in the marketplace and make it useful to real customers. It’s a means to an end, writing the book - the proliferation of ideas that are useful for real companies building real products. This it not a textbook. It’s not intended for students. The primary audience is the practitioner and how to manage the design efforts. Of course, the material does have some spill over into academics, but clearly it’s focused on the guy whose job it is to design things to make money. So what this book pulls together is the threads of risk and markets and technology. That’s what companies need to examine all of the issue involved in SoC product development and manufacturing. The book makes the transition between the relatively sanitary world of academics and the real world of sausage making. [Another big laugh] Q: Would you like to be in academics? Rowen: I’m not really an academic. I don’t have the patience for the academic world. I believe very strongly that the way for me to maximize my impact on the world is not by dreaming up new technical ideas. My mission is to take these ideas and make them practical and useful for large numbers of designers. I’m far more effective as a business guy, albeit with a pretty strong technical bias at my core. Q: Is this book just the first stepping stone to writing the Great American Novel? Rowen: [Laughing] Clearly, there’s a world of difference between writing something like this and writing fiction. This material is something I understand fully. The challenge is in finding the right way to get a set of well-understood ideas across in a way that’s accessible. Writing a great novel is a completely different kind of creative effort. Q: Nonetheless, who has first dibs on the movie rights for the book? Which studios are you in conversation with? Rowen: [Really big laugh] It may be a best seller within its own world, but I’m not seeing that it’s going to be quite that popular anytime in the near future.
Editor’s Note: As mentioned, Stanford's John Hennessy wrote one of the forwards to Chris Rowen’s book; Harvard’s Clayton Christensen wrote the other one. So get your book of fiction and get Rowen’s book on SoC technology. Then head out to the beach. Don’t forget the sun screen. Editor’s Other Note: I'm fully aware that not every book on the list is an American Novel, nor is every book on the list a Great Novel. The reference to the Great American Novel was merely rhetorical. ****************************
Peggy Aycinena owns and operates EDA Confidential. She can be reached at peggy@aycinena.com
|