CS250, UC Berkeley Fall ‘10Lecture 01, Introduction 1CS250VLSI Systems DesignFall 2010Krste Asanovic’, John WawrzynekwithJohn LazzaroandYunsup Lee (TA)CS250, UC Berkeley Fall ‘10Lecture 01, Introduction 1Why the heck is it CS250 and not EE250?2‣We answer that with a course history (with a few embedded lessons). Warning: What follows is principally from memory. I’ve done my best to be accurate, but some errors or misinterpretations might exist.Starts in 1958 with the invention of the Integrated Circuit independently by Robert Noyce (co-founder of Fairchild Semiconductor Corporation) and Jack Kilby (engineer at Texas Instruments).CS250, UC Berkeley Fall ‘10Lecture 01, IntroIC Design in the 70’s and early 80’sThe Intel 4004 microprocessor, which was introduced in 1971. The 4004 contained 2300 transistors and performed 60,000 calculations per second. Courtesy: Intel.Introduced to help sell memory chips!Federico Faggin,Ted Hoff,Stan Mazor‣Circuit design, layout, and processing tightly linked. ‣Logic design and layout was “random”‣Chip design was the domain of industry (Fairchild, Intel, Texas Instruments, …). These were IC processing companies. Those who controlled the physics controlled the creative agenda!3CS250, UC Berkeley Fall ‘10Lecture 01, Introduction 1"Listen to the silicon; find out what it's telling you."Meanwhile at Caltech…‣Carver Mead was designing and building prototype ICs (with help from his friends at Intel)‣His background was in physical electronics (invented several semiconductor devices such as the GaAs MESFET) but was deeply interested in the interaction of physical implementation and the higher level design of electronic systems:4CS250, UC Berkeley Fall ‘10Lecture 01, Introduction 1CS At Caltech‣Ivan Sutherland became founding head of the computer science division at CIT in 1974 (after leaving E&S)‣He and Mead teamed up to get the division off the ground making IC design (Integrated Systems) a key component of the research and teaching.‣My take: ‣These two believed that IC design was at the heart of computer science because CS was largely about inventing and building computing devices.‣The future of computing was integrated circuits:‣Very flexible, “boundless” growth potential (was on an exponential grow curve with no end in sight!) ‣Close to “pure thought” with few constraints and “nasty realities”‣The potential of “LSI” was not going to be reached with the status quo in industry.‣Worked together over the next 6 years to establish the faculty, industrial ties, curriculum, research projects with silicon structures as a key component.‣They set off to build their own machines (OM1, OM2).5CS250, UC Berkeley Fall ‘10Lecture 01, Introduction 1Pushing forward (1)‣The reality of integrated circuits:‣Wires are expensive (area, delay, power), transistors are cheap.‣Pre-ICs, the opposite was true.‣Therefore, plan the communication and the layout‣Exploit locality, think about the “geometry” of the problem from the beginning. Choose algorithms/designs accordingly.‣Algorithms/designs represented as communication graphs in a large number of dimensions, not a good idea.6CS250, UC Berkeley Fall ‘10Lecture 01, Introduction 1Pushing Forward (2)‣Put IC design expertise into the hands of those best qualified to take advantage of its potential:‣Those with intimate knowledge of computation and algorithms: computer scientists!‣Traditionally, IC design had been stratified:7Algorithm / architecture Micro-architecture Circuit design Layout ‣Emergence of the “tall thin designer”. Spans all levels of the design and implementation stack.‣Would lead to more successful innovation and highly optimized designs.CS250, UC Berkeley Fall ‘10Lecture 01, Introduction 1Pushing Forward (3)‣How to enable system architects: ‣Managing the complexity was the key challenge. Manipulating multiple levels of design complexity was difficult and projected to get much worse looking forward (remember Moore’s Law).‣Providing universal access to IC fabrication.‣Solutions: 1. Ideas from software2. New design representations3. Computer aided design tools4. Silicon “foundries”5. Education8All linkedCS250, UC Berkeley Fall ‘10Lecture 01, Introduction 1Ideas from Programming(help manage complexity)‣“Structured Programming” was getting popular (Dijstra, el. al.)‣No goto statements‣Block organization.‣Use of hierarchy, abstraction (sub-routines).‣“Structured Design” for ICs:‣Exploit regularity and symmetry ‣Use and reuse common sub-blocks (flip-flops, gates, arithmetic, etc.)‣Represent designs hierarchically9CS250, UC Berkeley Fall ‘10Lecture 01, Introduction 1Design Representations (1)‣Previously, to generate the mask information for fabrication, the designed needed intimate knowledge of the manufacturing process. Even once this knowledge was distilled to a set of “Geometric Design Rules”, this set of rules was voluminous with many special cases.‣Mead and associates come up with a much simplified set of design rules (single page description). A sort of “API” or abstraction of the process (back end processing could automatically convert this information into masks).10‣Sufficiently small set that designers could memorize.‣Sufficiently abstract to allow process engineers to shrink the process and preserve existing layouts. ‣Process resolution becomes a “parameter”, λ.CS250, UC Berkeley Fall ‘10Lecture 01, Introduction 1Scalable CMOS Design Rules‣Created with the transition from nMOS to CMOS (a much nicer technology), around 1985.‣Little changed over the years.11CS250, UC Berkeley Fall ‘10Lecture 01, Introduction 1Design Representations (2)‣Caltech Intermediate Form (CIF)‣Capture layout information, needed to generate masks and process.‣ASCII text file with geometric primitives and hierarchical definitions.‣Simple and human readable.‣Easy to generate and parse.‣Common sub-blocks could be reused from one design to the next (output pad drivers, etc.)12A sample CIF "wire" statement. The statement is: W25 100 200 100 100 200 200 300 200;CS250, UC Berkeley Fall ‘10Lecture 01, Introduction 1Design Representations (3)‣Previously, designed were represented by hand drawings. Then masks where made by transferring drawings to rubylith.‣Base layer of heavy transparent dimensionally stable Mylar. A thin film of deep red