During the first week of this course, we built a suite of 15 elementary logic gates. And in the next week of the course, we're going to take this basic functionality and use it in order to build some more interesting and powerful chips like adders and arithmetic logic unit. Now whenever we teach this course every week of instruction, there's a whole set of issues and questions that come up about different aspects of building hardware and software systems. So we've decided to devote the last unit of each week to answer such questions. And we are calling this unit this last unit every week a perspective unit. So here are the questions that typically come up at the end of this unit in which we deal with elementary logic gates. So the first question of the day is is it possible to build a computer starting with a logic gate other than NAND? Well I'll take this question and the answer is indeed yes, it's possible. For example, you can use another elementary gate called NOR, which stands for not or, and base your entire computer on this atomic building block. Likewise, it is quite natural to start with the suite consisting of and, or and not gates and use them to build a system. And there are a number of other possibilities. And in fact it's quite similar to the way geometry can be founded on different sets of axioms. And each one of them can be yet another point of departure to build, you know, all the power and theorems of geometry. This is not such a bad analogy. However, it turns out that NAND gates are very popular in physical implementations of hardware systems. Because there are many integrated circuit technologies in which it is quite cheap or relatively cheap to build NAND gates. As you well know during the course during the first week, we treated these NAND gates as black box abstractions. So, the next question, which is directly related to the first one, is well, if you actually had to build a, a NAND gate, how would you go about it? Norm, maybe you can answer this question. >> Okay so, in this course we sort of made a point of not going into this question. Because we feel that is already physics or electronics, rather than computer science. But still, let us see an example. There are of course many different technologies to implement and navigate and as in any other gate. But let me show you, let me show you the simplest example, if you wish, of such an implementation. So here is an NMOS implementation of a NAND gate. [NOISE] We have a plus voltage, which is going to be one, our logical one. And the minus voltage, which is our logical zero, and we're going to connect them as follows. A little resistor here which is a wi connection, and then two nice transistors, NMOS transistors, which we connect this way. This is going to be our a input, and this is going to be our b input. And this is going to be our output. So these are the two transistors that we have. So how does this implementation work? The basic functionality of one of these endmost transistors, is that whatever that input gate has, gets high voltage, then it connects the two other terminals. If it gets a negative voltage inside, it disconnects the two terminals. So let's see what happens. If both a and b are one, then we get the high voltage here, high voltage here. These two terminals are connected by this transistor. These two terminals are connected by this transistor. So we get the connection all the way from the output, to negative. And since this is a weak connection and this is a strong connection, our output is going to be negative or zero. As we want in a NAND gate, because when a NAND gets one and one as inputs, it gets zero as output. In any other case, one of these two inputs is going to be negative, which means a low voltage. That when these two connected to the low variable of voltage input, it's going to disconnect the two terminals. In which case, there is not going to be a connection from the minus sign to the output. And so, this weak connection to the plus is the one that's going to rule, and we're going to get the plus output, which is exactly what we want in a NAND gate, which is two. So, this is a NMOS implementation of a NAND gate. There are many other technologies. This one is not used so much anymore, since the 70s or 80s. And the important thing in this course is, we do not care about the technology. We do not care neither about this implementation or if there are any other implementation. We want to abstract that away, away inside of gates that get true and false and output true and false and we don't care how exactly. >> So welcome back from the world of resistors and transistors. And once again, this is a level of detail that we are not going to get into in this course. So here's the next question that we have. And it's about hardware description languages, and, the question is. How does the HDL that we used in this course, and in this week, how does it compare to real HDL languages that hardware engineers normally use? Well first of all, our HDL is a very real language, because you can use it to design computers and simulate computers, and that's what HDLs are all about. At the same time obviously, industrial strength HDL languages like Verilog and VHDL are far more complex and powerful than our HDL. Typically they have a syntax which is some mix of the HDL that we use and something that looks like the C programming language. And they feature all sorts of high level programming constructs, like For and While, which eliminate the need to write a lot of repetitive HDL code. So it's very nice to use these capabilities. And they also feature like our language does, they feature the ability to model and simulate the notion of time and clocks. Without which, you wouldn't be able to build sequential logic. You know, logic that ends up with which, with which you end up building things like memories and and counters, and so on. So these languages are very nice. At the same time, they are quite complicated and would, it would take you at least a month or so to master the these languages in order to begin writing some code for this course. So as an alternative, Norm and I decide to, decided to design a very simple dialect or very simple version of HDL and offer it to you. It has all of the capabilities that we need in order to build the computer that we will build in this course, and you can learn it in one hour. The next question, which is for Norm, is as follows the chips that we built so far were quite simple. How does one go about designing complex chips containing hundreds of parts and connections? >> Well, the truth is [COUGH] that there is no simple general way of designing complex circuits. It's really a complicated design challenge, and you need human ingenuity to do it well. Now there are many techniques that you learn in in digital circuit digital circuit construction courses. For example, there are techniques for card no maps that allows you to optimize a gate with a small number of inputs. Sometimes you can use various tools. For example, there are so-called silicon compilers, where you specify what the functionality is that you want. And the silicon compiler already has inside it a lot of logic and a lot of algorithms that can optimize gates for you. But again, this is, it is not perfect algorithms. These are all heuristics because the general problem of so-called NP complete, you cannot find a computer program that does it perfectly. The real answer is, is that after all these tools and all these techniques and so on, at the end of the day, you use the usual tools of computer science modularity and abstraction. You break a complex problem into simpler parts, and the simpler parts are simpler to optimize and to construct. And at the end of the day, after you use all your tools, after you use all your techniques, you go back and use this modularity at the idea. >> So these were the questions that we chose to focus on in this first perspective unit of week one. As you can see, this particular format of the perspective unit is open-ended. There are numerous questions that can come up. Once again, we don't want to deal with the level of transistors and resistors. You know, this is something that belongs to electrical engineering. We want to focus on computer science. And when it comes to hardware technologies and so on, there are many questions that Norm and I cannot answer either. So we are we welcome any questions that can come up. You can post these questions on the course Q&A forum. And if there are other students who are who have some knowledge about these areas you are perfectly welcome to go to the Q & A forum of the website of this course and answer these questions on your own.
