Tác giả giữ toàn bộ bản quyền (All Rights Reserved) — không phải giấy phép mở. Bạn vẫn xem và học miễn phí trên OpenER; muốn tái sử dụng cần xin phép tác giả.
DAN GARCIA
Can I start webinar? Say yes.
BORA NIKOLIC
Hang on. Hang on.
DAN GARCIA
Uh-oh, I started. It's live. Ah, I push the button. Sorry.
[LAUGHS] BORA NIKOLIC: So look among our participants if we have our guest to come out.
DAN GARCIA
Okay, perfect.
BORA NIKOLIC
All right.
DAN GARCIA
Welcome, everybody. How many? We got 64. We should have had a guess on — we should have had our little gamble on how many people we'll see.
I hope more people come today. Today's the day of big announcements. We're excited to share all these with you live. Welcome, everybody.
Bora, I'm going to highlight you if that's okay. Spotlight you. Bora, is that all right? Okay, let me figure out how to do this.
I can't seem to do it. Okay, click Spotlight for Everyone. There we go. Okay, you are spot lit.
And we're past 4:10. We should start. Welcome, everybody, to CS 61C. Wooo!
BORA NIKOLIC
Hello.
DAN GARCIA
All right. You're up, Bora.
BORA NIKOLIC
I'm up?
DAN GARCIA
Yeah.
BORA NIKOLIC
All right. Hi, everyone. Well, I am trying to still spot our — DAN GARCIA: I can change slides for — I can change the slides.
BORA NIKOLIC
No, you're good.
DAN GARCIA
Oh, you're looking for — BORA NIKOLIC: I'm looking for our — DAN GARCIA: Oh, in our participants list. Right, right.
BORA NIKOLIC
Right, in the participants list because I sent out both the panelist invitation and the regular invitation. Or I sent a regular invitation to her admin.
And I — doing these kind of things is a non-trivial thing if you are in a government office. All right. Hi, everybody.
I am hoping everybody is doing better now with the project behind us.
And we are going to try to help really everybody get through the next few weeks in a better way.
So our agenda for today is we want to go over the week four. We are looking at a very self-contained module.
That is RISC-V. We have covered almost all of it. So if you watch this week's lectures, you are ready to go to cover homework four and projects 2a and b.
So it's just the project stack. So things should go relatively smoothly. And there should be enough time. There are few more important announcements, but I'm going to save that for a few minutes later.
Now — DAN GARCIA: Bora, can I just add a quickie — BORA NIKOLIC: Yes. DAN GARCIA: --here while we're here? We're kind of doing our little two microphones microphone thing.
Most people have seen a C-like language before. That was the requirement for this class. But most people have not seen any kind — I'll just add this — any kind of assembly code.
I mean, it would be rare if somebody came into 61C having touched assembly before. So basically, this is new for everybody. And we understand that.
And we're hoping that through our courses of the lab and the homework and the project, we're kind of ramping up — assuming you've done none of this before.
Remember, the C was you've mostly done algorithms before, done software engineering before, and a C-like language. This is just C's particulars.
And we try to — hopefully, you enjoyed project 1A and 1B in terms of making it fun with some graphical output and this life and all that stuff.
But two is new. And we're trying to channel in some of the excitement of the machine learning part.
You want to talk a little bit about that, of what 2a's going to do for them? But I'm sure they've seen it.
But it's exciting to try to bring this really fun idea of number recognition, but do it at the level of the assembler.
So this is new, and I hope you enjoy project 2a as well. At the end, we'll always take a survey of what projects you've liked.
And in the past, people have liked one and they've also liked two. So hopefully you'll enjoy two as well.
BORA NIKOLIC
Yeah, and there is a natural downshift in speed that we did going into week five module.
The first thing, a slower speaker than Dan as we have already established.
And I really want — I started recording essentially at the speed that — the pace that we would deliver this in live lectures. And then I realize that I'm just not giving enough details.
And I'm never — even with myself, I'm never really stopping, never letting this information sink in. So I watched my own videos and said, oh, I need to slow down a bit.
And I need to double up on explanations. So what you find out, that the videos are a little bit longer now. But I'm repetitive.
So I mean, if you feel like — that it is too much, skip, fast forward, get 10-second forward button going.
But I try to always explain the same thing from approximately two angles. Every time there is a new concept, I try to give two explanations. And hopefully I'm succeeding.
We know that people have not seen machine language and people have not seen assembly before. But it is intuitive. It is very simple. There are no deep concepts.
But it is different way of thinking. So spend some time, try to figure that out. From what am I seeing from Piazza, questions that have come through so far, people seem to get it.
But few — if their concept that they're missing or they're not clear, keep calling these questions.
DAN GARCIA
And I want to add, make sure the people — if they missed this last time, we're the number one department in the world, arguably, in computer science and computer engineering.
I can't think of anybody better on our entire faculty, the top in the world, than Bora to deliver these lectures on RISC-V. So I'm just so proud to be able to learn from Bora.
I want to say, sit back and just — I'm in awe of all that you know about RISC-V. You know the history, you know the back story, and you know where it's going.
I learned so much from listening to you. So I just want to add that at the end because I really appreciate it.
BORA NIKOLIC
Thank you. I don't deserve that, but that's okay. [LAUGHTER] One thing that I did notice is that some of our — somebody just spotted a three-year-old bug that we have in our coding.
And one thing that — why these bugs exist — students don't this. RISC-V has — since it's so new, it's changed some of the encoding in some instructions.
And — DAN GARCIA: It's a living, breathing beast. You got to keep up with it. Yeah, exactly.
BORA NIKOLIC
No, but it's now frozen since last year. Since fall last year, it is frozen. So the RISC-V spec is ratified. Although, it has not been reflected in every single one of our slides.
So we need to do some effort in cleaning.
DAN GARCIA
And it's almost like we need to draw a line in the sand and say, you're learning RISC-V, version bloop. Is it version 2020 October?
What version are we learning as we're kind of working with something? So maybe — are there big lines in the sand where they say, this is new, you know, C99, C18.
Are there lines in the sand where they say, this is a new change and we're giving it a new name? It's always important to know what those are.
Exactly.
BORA NIKOLIC
Unfortunately, they just give them numbers. So whatever you want, 0.7, 0.6 or something like that to be ratified.
DAN GARCIA
Right.
BORA NIKOLIC
But that's the one that's went through the vote. And that is the one that is frozen.
DAN GARCIA
Yeah.
BORA NIKOLIC
There are a few interesting things that happened there. They dropped two instructions from the base. Maybe more — five, six instructions from the base.
So actually, what we teach in 61C, which you'll find out when you watch this, is literally the base instruction set for RV 32R.
So any off-the-self compilers should work with your code if we get all these encodings right.
So a commercial compiler should be able to compile your code, not just Venus, which is quite exciting. And you can do things with it later on.
And when there is really this RISC-V PC, you can run it on RISC-V, whoever gets their hands on that. I mean, that's quite interesting, quite amazing, I think.
But I want to move on because we have other important stuff.
We were supposed to have a very important guest, but I think the guest is — we might have overestimated our abilities or her abilities. But let me run through a few pieces of news.
And let me see if I can share some slides.
Some people might have seen some of these news before.
What you'll find out, a lot of research in the US and leading universities funded by the agency that is called DARPA, which stands for Defense Advanced Research Projects Agency, and some people think that they do some very creepy stuff.
No, their idea is to move for the technological agenda. If you read the mission statement, is to prevent technological surprises.
DARPA was founded in 1959 after US was surprised by the launch of Sputnik.
And at that time, President Kennedy said, this is not going to happen again, started DARPA to prevent technological surprise. And then I think that has been amended to prevent technological surprises by creating them.
And there are quite a few things that have happened that are very relevant to this class and the things that we do.
In this slide, what you will see, there have been projects with quite a bit of impact. They started this thing, which was called ARPANET, that turned into the internet.
A minor accomplishment. I mean, that was a tiny project.
DAN GARCIA
[LAUGH] That's true. That's exactly right.
BORA NIKOLIC
It was a couple of terminals back then. And now that's it.
There was a big program back in the '70s and '80s to develop very large scale integration, meaning better and better chips to ride the Moore's law.
We talked about Moore's law, but every few years, there was somebody that came up with a reason why will Moore's law stop, why will it end.
So initially, Moore projected it was going to go until 1975. And there are people that are having the end date written — 1978, 1980, 1983.
I mean, every few years, there was a reason why there should be no more Moore's law. And some folks in DARPA took that as a mission.
Well, we are not going to let Moore's law stop because it's a good thing for everything. So RISC I and II were DARPA projects. Some people did not know that.
I mean — DAN GARCIA: I didn't know that. I didn't know that, yeah.
BORA NIKOLIC
But they were funded by DARPA. Similarly — RISC-V was not invented by DARPA, but our first realizations of RISC-V were part of DARPA projects.
It was kind of a funny thing then. DARPA called for building specialist. That's really difficult to do with somebody else's instructions said that you can't touch. But we had our own instruction set.
They could do whatever we wanted. They could build a specialized computer. We clearly aced that part.
So well, DARPA has a new set of things that they're after now. It is a big piece of news.
So these people that we have elected to represent us in Congress and Senate do tend to make us upset. But occasionally, they do get together and make these big bipartisan moves.
And the big one that has been boiling or cooking up this summer is about the so-called chips bill that is supposed to be $22 billion total bill to advance semiconductors in the US.
That involves some manufacturing. Some people have seen about the lab coming up in Arizona.
But also a lot of funding for innovation, to innovate things, to extend whatever is the Moore's law and whatever comes after the Moore's law that will help develop this technology.
So in that sense, DARPA got a new director. And she was supposed to join us. She might join us. We don't know.
Well, still — DAN GARCIA: I'm looking. I don't see her on the attendees list.
I'm hoping that she didn't just — BORA NIKOLIC: You know, today was her first day at the job, first day in the office taking the job.
She's, I think, only the second or third female to lead the DARPA. And her previous appointment was at Berkeley, which is very interesting.
So we tried to get her on the first day, but I'm guessing she is busy. She might join us.
DAN GARCIA
Do you know the size of the budget she controls, Bora? Do you know? I mean, DARPA is a very deep pocketed organization, much deeper than NSF.
BORA NIKOLIC
It's up to a billion, I think. It's an annual budget.
It probably goes from — I think about 300 million out of this bill are supposed to be managed by DARPA. So this chips bill is enormous.
It's too big for DARPA.
DAN GARCIA
Wow.
BORA NIKOLIC
But they're supposed to do the research.
And I think about $200 to $300 million are going to be managed by DARPA to try to advance things, because DARPA has the ability to start the projects, manage them, review them, kill them as they don't turn out to be working outright, and keep going.
Yeah.
DAN GARCIA
While you were doing this, I was just looking this up. I know we're trying to stall here. But I was looking this up, and it says that DARPA's annual budget is $3.56 billion.
NSF's annual budget is $8.3 billion, so it's bigger than DARPA's. But the CISE budget, CISE, which is the computing and information technology, is a smaller part of that.
So comparing CISE to that, DARPA's a lot — I mean, DARPA's everything, but the computing part of NSF is smaller than that.
BORA NIKOLIC
Yeah, DARPA is focused primarily on technology, right. NSF funds all kinds of things.
And we want somebody progressive later, assuming that she can — there is a lot of science that is outside of technology. DARPA is straight engineering.
Now — DAN GARCIA: Can I bring up another — BORA NIKOLIC: Yeah.
DAN GARCIA
Can I just bring up another computing in the news that we don't have a slide for? But I just want to talk about them. I'll even just — I'll highlight.
Let me just grab my thing and highlight myself really fast. Boop. Spotlight. Remove spotlight.
Okay, so I wanted to add that there is a — blink.
There is a — they would have — had some lawyers not come into to play to order a stay order, they would have quit TikTok and WeChat.
And there were a lot — boy, oh boy, there are a lot of concern about many people who are using this for their own livelihoods.
A lot of people are using — more people using WeChat for their livelihoods than TikTok. But there are some artists who are using TikTok for their livelihoods.
And it looks like both of those are going to be resolved now. It looks like — that TikTok is going to be bought by Walmart and Oracle. Oracle's here in the Bay Area.
And it looks like WeChat was just going to be canceled altogether. And there was a — a judge ordered a stay of that order.
So right now, those people would have — I mean, you could have your whole business based on WeChat and the community you've built around that. It's mostly a Chinese app.
It's amazing how you can put your credit card in there, do so many things through that. But it sounds like they're — right now, they're just — they're in a holding pattern right now.
A lot of people were trying to download all the pictures and all the contacts they could as much as possible.
But even those are very hard to do because you only know them through WeChat.
So this is — for a big community who uses WeChat as their lifeblood, this was a really, really big deal. And it looks like they're just on hold now.
So that's an issue. So we'll see with that. But it was scary. It was really hard.
I was feeling for all the people who were just using it, passively using it. You know, not harming — good, working American companies using WeChat to connect with their customers.
And all of that would have been gone. Imagine taking your phones away and you don't have any other way to find the customers and all those contacts you had.
So crazy. So I think they're both resolved. I think the TikTok is going to be fine.
What I'm hearing is 20% is going to be owned by Walmart and Oracle, and 80% will still be owned by China.
But there'll be enough seats on the board to be able to make the people in Washington happy. I don't know what the future of WeChat is. I don't know.
But I would take your data off of it if you can.
If there's some way to grab your data off of it and grab the context off of it, try to do that because who knows what will happen when the stay gets lifted.
But that's it. I just wanted to add that. That was relevant for a lot of people in CS10. I wanted to share that as well here.
BORA NIKOLIC
So in the meantime, while Dan was talking, I found a spare — just for today, thank you spare guest. It's Jared Zerbe. Let's allow him to talk.
DAN GARCIA
Oh, wonderful! Hi, Jared. Good to see you.
BORA NIKOLIC
I don't know if you guys remember — JARED ZERBE: Hi. BORA NIKOLIC: --is he allowed to — yeah, he's allowed to talk.
[LAUGHTER] I guess — yeah, I've even got my BWC anniversary jacket on here if I can get my video going. How are you guys doing?
This seems like a pretty fun crowd.
DAN GARCIA
Good to see you.
JARED ZERBE
You all made it through the first assignment, so that's all good. Assembly language programming. I want to give a shout-out to that. That was one of my first joys in life was doing that.
And I think the students should all really appreciate the efficiency and the speed that you get by bypassing something from a compiler.
So by making life easier for programmers, you make things run less efficiently.
And there's a lot to be gained when you really put the time in to do some of the programming, so I hope you guys get an appreciation of that.
BORA NIKOLIC
If you guys do not realize, Jared is the — works at Apple. And he gave that big introduction to their chips program.
JARED ZERBE
[LAUGHS] BORA NIKOLIC: --about a month ago.
JARED ZERBE
It's our little chips program at Berkeley is what it is.
I encourage people to look more deeply into that and to — I want to encourage people to apply to also to the Master's and PhD scholarships and fellowships.
So think about that if they're interested in VLSI.
And if they're interested in hardware, I think this is a great class to really kind of get back and forth and play in both sides of the world, really dig into the hardware aspects of software.
DAN GARCIA
And to whet their appetite.
I mean, the goal of this is to get people excited about it, to be able to say, wow, I want to take that upper division course and that upper division course.
And then boom, maybe there's a career path for me. So this is the beginning for a lot of people of a whole career path. So excited to get people excited about this material.
BORA NIKOLIC
We'll see what happens with this chips bill. I mean, that may change quite a bit of a landscape on — I mean, nothing is set, especially not these days.
JARED ZERBE
[LAUGHS] Yeah.
BORA NIKOLIC
But it may change completely the landscape of jobs around here and the perspective of where do things go and where do things go.
Of course, I mean, when somebody dangles $22 billion, there's quite a few hands that go up in the air, ready to do something with that.
JARED ZERBE
Well, I think when you're — especially when you're a student and you're starting to — you're trying to figure out your career, when you see moves occurring like this, try and think at a meta level.
Try and think about not just that — you know, there is $22 billion. I mean, it sounds like a big number. It is a big number. But try and think about why.
You know, why there are organizations doing the things that they're doing. And even for our kickoff the other week, right, you know, why is Apple spending all this time and money and resources?
I got fancy slides with animations and all this. Why is this person here talking to me about this?
And the reason is because it's important to them in some meaningful and impactful way, right?
And certainly, at Apple over the last 13 years, I guess, the amount of silicon that's — and the impact of silicon has become massive, right?
So the impact of being able to control our own destinies is enormous. When you see what can be done when you do custom silicon. It's quite fantastic.
DAN GARCIA
And I think Apple's unique in that way to control both the hardware and the software. But the hardware, they didn't fully control all the way down to the silicon.
So now it's like from the pixel at the top level of the GUI all the way down through the software to the lower layers of the OS, that's all controlled by them to the hardware.
And they're building the hardware all the way down to the lowest level. So it's very exciting to kind of remove all of the OEMs.
All the other OEMs that are out of the picture, and make it Apple all the way down.
The picture we've shown in 61C from the top to the bottom, imagine one company controlling all of that. I know of no other company that does that, by the way.
Microsoft, Oracle, nobody else goes top to the bottom like you do. I mean, even Intel doesn't, right? They don't write software at the top level.
So I don't know of anybody else who has that control to be able to deliver that user experience. I mean, I'm a big Apple fan. I mean, maybe you can tell.
But my point is I don't know of anybody else who does that ever, ever in the history of — JARED ZERBE: Well, we just pointed out, you have vertical integration.
I mean, there was IBM, right? So IBM was the classic vertically integrated company. And you'd buy a big 360.
And they had their fabs at one end, and they have their software at the other end.
And the danger when you have that is like how do you avoid collapsing on yourself or having these kind of inter-group fights, right?
There's those famous sociological experiments where anybody — they could take a group of people and then divide them in two any way.
Like, the people that are taller or shorter, men or women, or people born on odd days or even days.
Any time they did a division, they would eventually start competing and warring against each other.
DAN GARCIA
Interesting.
JARED ZERBE
So part of the — I think the amazing thing for me — I've been at Apple seven years.
The amazing thing for me is that people do have — you know, oh, , what about this? And what about that? You know, a conflict on this.
And then there's the classic lots and lots of heavy arguments. But then people will line up behind the product. So you just go [WHOOSH].
And so when there's that defining force and there's the overriding force, it's almost like you imagine that you're in — if you were in the army in 1944 or something like that.
How can you possibly organize a group like that? Well, at least they're uniting behind or fighting a common enemy, you know? And for us, it's uniting behind delivering an amazing product experience, right?
So when you have that uniting force, it makes a big difference. And I think that helps the companies from collapsing when they become this vertical integration challenge.
But what an opportunity. You're right. I mean, it's just an amazing thing to be able to talk to software guys.
And just today, we're talking with the algorithms guys and we're talking about — I was just talking with process people yesterday. And it's just — it's amazing. It's quite an opportunity.
DAN GARCIA
It's also no excuses, right? Before, well, we would do this except that we can't control the hardware. And we only wish that they would add a feature to the instruction set.
No, go fix it! Go make a new instruction set.
JARED ZERBE
Exactly. Go — DAN GARCIA: Go make a new — go make a feature. Go make a new command and do that. No more excuses.
It's a fascinating thing, you know?
BORA NIKOLIC
They don't own their own instruction set, however.
DAN GARCIA
No, that's true.
JARED ZERBE
But we have our architectural license, so you'd be amazed at the kinds of things that are in our instruction set that you might just not like very much.
BORA NIKOLIC
There is a magic word — architectural license for — JARED ZERBE: That's right, architectural license gives you — BORA NIKOLIC: That costs a lot.
DAN GARCIA
Right.
BORA NIKOLIC
Not very many people can afford that.
JARED ZERBE
That's true. That's true.
DAN GARCIA
There's a question. Can I share this question in the chat for you, by the way?
JARED ZERBE
Yeah, yeah.
DAN GARCIA
What does this mean if you own the entire top to bottom system, vertical integration, about antitrust? Isn't there something about competition or something? You want to respond to what that issue might be?
JARED ZERBE
Oh, we have lots of competition. I mean, I don't think there's any — there's no worries about competition. They have — DAN GARCIA: Right.
JARED ZERBE
If you look at Android ecosystem or you look at all different kinds of — I mean, even Macs, PCs, and the volumes out there, and on phones, on anything, there's plenty and plenty and plenty of competition.
The difference — I think what you're thinking about is the vertical integration versus horizontal — DAN GARCIA: Right, exactly. That's what it is. JARED ZERBE: --kind of spread, right?
DAN GARCIA
Yeah, exactly.
JARED ZERBE
If somebody dominates the entire disk drive industry, then they own the whole industry.
And then all of a sudden, it isn't very long from that moment where there's some clever sales guy who goes, huh, I wonder what sets the price of all these drives.
Oh, it's me.
DAN GARCIA
Right. Yeah.
JARED ZERBE
That's called a monopoly. And so that's different. Vertical integration is owning a small piece of lots of different things that you need to bring your product to market.
DAN GARCIA
Right.
JARED ZERBE
And even in that respect, we don't own our fabs, right? We're a fabless semiconductor company. So we know what we're good at. And we know what we're not as good at.
And running fabs is not our forte. But it's an amazing kind of technological exploration you can do with fabs when you are a company that has the size and impact.
DAN GARCIA
So the analogy is what if Tesla decided to make their tires and make the — obviously, they buy leather from other people and they buy tires from other people and buy other things.
What if they made everything and they make a factory? They still have many competitors for electric cars. They certainly are — there's no anti-trust questions about Tesla.
There's a ton of competition in that electric car market.
So you could still do what they — they could model Apple, top to bottom, but still there's no antitrust issues because there's a ton of people at that level.
At the level of selling cars, there's a ton of people in that market just as there's a lot of people in the iPhone, back, et cetera, watches market.
Every market, it's clogged. So that's not an issue there.
JARED ZERBE
Yeah, absolutely. Tesla's a great example.
BORA NIKOLIC
IBM sells still vertically integrated stuff. I mean, you can buy Z-series chips from IBM. I mean, you only can buy a Z-series machine.
And that's what you buy if you are Bank of America or airline or something like that. I mean, those things — they run only one application, it's like DB2, you know, database system.
And that's it. They may change the user interface here and there, but it is — every new generation of Z-series is there to support the giant databases.
JARED ZERBE
Yeah, yeah.
Well, I mean, I think one of the things that you have to realize also is that one of the challenges these days in SOCs is — we just had an announcement earlier in the week that included this latest SOC.
There's an 814 SOC. It's got over $11 billion transistors on it.
And that's just — the amount of complexity involved in something like that is mind numbing, not just in terms of I'll be plunking down every transistor, but just managing all the different blocks, managing the communication between the blocks, managing the team, managing who does what and how they all communicate to each other
and how you verify something that has the kind of unbridled complexity of something that has that many layers of it. It's like a small city, right?
You have to — So in many aspects, are a lot of the things that you learn when you're doing software at a high level and kind of managing the programs with these modules that are coming in and out are very similar.
There's lots of similarities in terms of SOC integration as well.
BORA NIKOLIC
Yeah. By the way, I mean, what we believe — we have been talking also in some of our office hours. We've been talking about the end of Moore's law and the need for specialization.
Well, we have to do it. I think Apple is showing the way, well, one, that seems to be the right way of doing specialization. And you will have to do it.
I mean, you can't just go buy chips in a grocery store and work or Best Buy and specialize those. And that's why somebody was asking, what are other companies like?
Well, everybody's trying to do that. But it's a lot harder to capture a market like that, like cellphones at the moment.
The big story, the big question there is, what will Apple do next? And we're going to find that out.
JARED ZERBE
[LAUGHS] Well, yeah, I mean, there's — the funny thing is that — and if I were to kind of put myself outside the company again and then look in, I think — looking at Apple from the outside, I think that people don't really understand how many different engineering programs we have
running in parallel and how lean in operation we really run. I mean, every program is a lean program. It's not like — I mean, and there are lots and lots and lots of them.
And there are programs where you just hammer on something for a long time, and then go it's missing this fundamental ingredient. We'll put it on the shelf.
And then five years later, this fundamental ingredient comes along. And you go, ah, let me take this off the shelf and restart it. That happens a lot.
I mean, if you two — say, just you, Dan and Bora, were to guess, how long do you — you know, you heard this announcement about Macs going to Apple silicon, right?
So how long do you think that's been in discussion?
BORA NIKOLIC
[LAUGHS] 10 years.
DAN GARCIA
I'll throw a number out. I'll say a year just so you can blow me away with how long it takes to — JARED ZERBE: Bora was much — BORA NIKOLIC: 10 years.
10 years.
JARED ZERBE
Bora much closer, yeah. 10 years. 10 years is more — DAN GARCIA: Wow. JARED ZERBE: --of the proper order of magnitude.
And when we're in discussion of something, that doesn't mean that someone goes to a whiteboard and draws a picture and goes, oh, wouldn't this be nice?
And then, techs stare around and say, oh, yeah. No, no. Discussion at Apple means that you go build stuff, you test it, you try it, you poke it, your prod it.
You know, all this kind of stuff, right? So I mean, that means that — you can imagine how much investigation goes on kind of in parallel.
It's really an engineering — just a powerhouse of lots of different parallel engineering.
DAN GARCIA
I mean, this is maybe too much inside baseball. Do you ever grow two competing groups to see who would win? Like, it's the idea of we don't know what the future is.
JARED ZERBE
You know, I've read that.
DAN GARCIA
Let's have a couple of these different groups and only one's going to succeed at the end of the day. But you're all kind of in a race against each other.
And — JARED ZERBE: I've read that. I read that in the — DAN GARCIA: I don't know. JARED ZERBE: — Steve Jobs thing, that like, oh, he didn't tell this group about it.
And this — DAN GARCIA: Oh.
JARED ZERBE
You know, and I — DAN GARCIA: Yeah, yeah, yeah, yeah. JARED ZERBE: --read about that in the past. I'm not aware of anything like that going on right now.
Everybody's too lean. And if you find out something that somebody else is doing that's similar to what you're doing, you go, hey, hey, hey, what are you doing over there?
And you try and pool some effort, right? You try and collaborate and figure out what's going on. But there's a lot of little investigations too.
BORA NIKOLIC
Two super quick questions.
JARED ZERBE
Yeah.
BORA NIKOLIC
I mean one because we have a few other important announcements.
JARED ZERBE
Great.
BORA NIKOLIC
So the first one, I mean, Apple went public about a year ago that Apple will make a modem, a 5G modem. That is a good thing.
So I guess they have been thinking about it for about a couple of years before that. So I think you can start your stopwatches and see when is this going to hit the market.
I think they'll license Qualcomm for another five years. So you can think about that. That's about seven years' time to own that kind of a thing. But the majority don't have to come.
There is a good question here. I don't know if you can answer this one. This is from Benjamin. It says, Jared mentioned that we should appreciate assembly because of its speed and efficiency.
But at companies like Apple, is the assembly almost entirely abstracted beneath compilers? Or is there still development done entirely in assembly?
DAN GARCIA
For even higher orders of optimization.
BORA NIKOLIC
Yeah.
JARED ZERBE
Oh yeah. So — DAN GARCIA: Picking below the layer just to optimize your final code, yeah.
JARED ZERBE
Yeah, so I just heard a story related by somebody who Bora knows very well. And the story basically goes that software developers work — you know, they'll try at the highest level they can.
And then it doesn't perform the way they need to. And then they'll try and get rid of some of the efficiencies and it still doesn't work. Then they'll scratch their heads and parallelize.
And it doesn't work. So the answer is that — is there firmware, assembly-level programming work done at all levels of — absolutely.
Absolutely. And my advice to students is find the thing — find the layer of abstraction that you like playing. And the layers — most people like playing in multiple layers.
Find the layers that you like playing in and just get really good at it. Just really just jump in with both feet and just really immerse yourself in that.
And then also get really good — so that's what we call the T model. So you own your area. Going to own your number one is what we call it.
Own your number one area of specialization. But then also understand and get some curiosity and breadth as well.
So you want to own some adjacencies too because you want to say, okay, well, I can bring my expertise here, over here, or over here. And then in — who knows what?
In 20 years, you may actually decide, oh, I want to — I've gotten really good at this, and I've kind of drilled this to the point where I'm more interested in this over here.
So it's a great way of appreciating not just your one singular area, but also other areas. But yeah, I guess to answer the question succinctly, absolutely.
All levels of programming, from the highest level to the very, very bottom of the stack.
BORA NIKOLIC
So thank you, Jeremy, for stepping in. I mean, I'm assuming that Victoria doesn't have an easy time calling in from Washington.
on her first day of the job and probably not easy to communicate out of that building. Zoom may not be — let you connect — DAN GARCIA: That's true.
Zoom might not be blessed. That's true. Never thought about that.
JARED ZERBE
She's trying to ping into the firewall.
DAN GARCIA
Yeah, exactly.
JARED ZERBE
All right. I'll bus out then. Thanks a lot.
DAN GARCIA
Thank you for dropping in. That's wonderful. That's great.
BORA NIKOLIC
Jared will be back, I think, with a few friends at some later date.
DAN GARCIA
Hey, that's going to be fun. Look forward to that.
BORA NIKOLIC
All right.
DAN GARCIA
Great to see you, Jared.
BORA NIKOLIC
Thank you. We have a few other important announcements. Let's move it over to Dan.
DAN GARCIA
I'm happy to take it over. Yeah, happy to take it over. Spotlight me if we can. And then I'll jump in.
BORA NIKOLIC
I think you are spotlighted.
DAN GARCIA
I am spotlighted. And then I will go to my slides, which are right here.
BORA NIKOLIC
Boop. Spotlight.
DAN GARCIA
Okay, spotlight me. Can you see me? Jared's still there. We've got to drop him off of the spotlight, I think.
Can we figure out how to do that? You can't see my slides or me because he's still got it. How do we do this?
BORA NIKOLIC
I see you.
JARED ZERBE
I can see you and your slides.
DAN GARCIA
Oh, really? Funny. Really weird.
BORA NIKOLIC
I spotlighted you. I spotlighted you.
DAN GARCIA
Okay. All I see is Jared in this. Okay. Well, wonderful.
I'll assume that you can see me because I can't see me in my Zoom window. It's a little weird. All right, schedule. Here's what the schedule looks like.
Last Friday — actually two Fridays ago, 9/11 was the RISC intro, RISC-V intro. Last week was load/store, decisions, and procedures. That basically taught you how to code.
By the end of those four lectures, I could say take this C code and translate it to RISC-V, and you should be able to make that happen. So that's pretty cool.
This week is about lifting the hood even more. It's an exciting week, where we say, you know what?
Let's actually see if we can move down to the machine code layer and show you how to encode these as ones and zeros.
You're going to need that when you get to my lecture, which I just finished on Friday, which is CALL, compiling, assembling, linking, and loading.
And to understand how the whole thing works, we're going to take a C program and go all the way down to the actual a.out.
And then even the last L of CALL is what does a loader do to actually run that program? So that's pretty exciting.
This is a fun series of lectures, really lifting the onion, revealing the layers of the onion to show you how this stuff works.
If you have ever wondered how this stuff works, this is the week where you're kind of making the connection between C to the hardware. It's pretty exciting.
After this week — we're not showing what the week after this is. It's all hardware. It's almost like taking a bite of an In-N-Out burger. You come from the top.
Your teeth bite down until you get to the meat. The meat's the ISA layer.
And then the bottom teeth come in and bite from the bottom up as you're biting through bread and then lettuce and tomato.
So we're going to start from the bottom then of the hardware layer and come all the way up. And we're going to meet in the middle.
So it's like your teeth — I'll show a video. Your teeth actually meet in the middle, and that's a special, magical day.
The reason we bring up the timings of this is that all of this material, RISC-V and the lower levels of instructional coding and CALL, are needed for your homework four, which, again, is going to be due 9/30 and your project 2a, which is due 9/28 and 10/5, okay?
So pretty cool stuff. Let's go to the next slide. We got some big announcements.
BORA NIKOLIC
I just wanted — DAN GARCIA: I need some — go back. Go back.
BORA NIKOLIC
No, you're good. You're good.
DAN GARCIA
Oh, yeah.
BORA NIKOLIC
A few important things about this. It is just seven lectures with quite a bit of a payload. So it's six hours of videos with a big payload.
Clicker questions, we have increased slightly, just tiny bit difficulty of Clicker questions. So they make a little bit better bridge towards the homework.
Homework, we added a little bit more of difficulty to the homework such that they start looking more like midterm problems. But they're not quite at the midterm problem level.
So we'll need to practice a little bit. But if you do the project — if you do the homework, that gets you to the project.
If you do the project, you're ready to practice midterms. And we will try to prevent surprises. We know that we surprise you. We will try to be absolutely clear.
We surprised you in the quest. We will be absolutely clear what to expect in the midterm. So midterm covers this material. This is a great block.
There's an advantage of being — there are not many advantages of being online, but this is one advantage that you have — nicely packaged seven videos that cover a complete module, everything that you need to know from assembly to actually instruction and coding.
And majority of the things that we actually need to do projects 2a and b are in the first RISC-V assembly language sets of videos and in the CALL — a little bit of that in the CALL.
And that's it.
DAN GARCIA
Yeah, good stuff. Good stuff. All right, here we go. Big announcements.
We have the next set of slides. So first big announcement. This is some time. But don't show.
Don't show. You removed the — ah.
BORA NIKOLIC
Back, back.
DAN GARCIA
Ah!
BORA NIKOLIC
Why don't I have control? No?
DAN GARCIA
Yeah! [LAUGHS] This one. I got to control it. I've got my PowerPoint.
You gave it all away. Ah! Oh, no. Don't read it.
Oh! Here we go. Okay.
BORA NIKOLIC
No, but it's you still.
DAN GARCIA
No, it was me. But then you clicked on a new slide. And then, it — BORA NIKOLIC: Oh, no. I'm sorry.
DAN GARCIA
No, no. It's all right. No, it's OK. Because where both co-piloting, you could override anything.
Anyway, longer stuff. So first one is that we're listening. Thank you so much for the feedback. We want to continue to listen to what's happening.
And we want to respond to everything that are going wrong. We want to try to reduce the pain points. So the first is quest grading concerns.
I want to add we've just seen the quest numbers of the grades. They've come out. We were looking forward, out of 30 points, to having the quest average be 25, 22, 23.
I mean, high. The quest is 15. So we were unhappy. So we didn't do a good job there.
We didn't write an exam that our students could totally crush and gain confidence, right? Easy, get in, get out in an hour, and gain confidence. It did neither of those.
It lost your confidence. And you got angry, which is reasonable. And it took more time than we wanted to. So both of those things are broken.
We learned from that. [WHISTLES] Okay, lessons learned. But there were other grading concerns. And we said, it's low.
We've actually — by the way, we've taught this class a lot here. We baked in the points of this. This is an absolute graded course. It's not a curved course.
So we baked it in basically assuming, okay, the quest should be around the average of 25. Now, let's do the other stuff. When the quest is below, we didn't do a good job.
And it means that it throws off what we think the distributions will be at the end. We're not going to curve it. But we're worried about that.
So we need your grades to go up. So we need to have a question or two on the midterm that's very solid and can replace. Now, I want to say this.
In the past, how we have done a quest clobber is what's called a z-score. Okay, here's what happens. What you do is you look at how you sit related to the mean.
Remember, this is kind of a curve, right? Whatever the average is, you look to where you sit. Are you above the mean by one sigma?
Are you below the mean by one sigma on the quest? Then you take the midterm.
And we're talking about — we're talking about the quest clobber, is the questions on the midterm that are exactly clobberable. Like it's the C and number representation stuff, right?
Nothing else. Just for that part of it.
BORA NIKOLIC
They will be labeled. They will be labeled.
DAN GARCIA
And they'll be labeled. Exactly. Huge label. This is the quest clobber question, okay?
Typically, what we had done — and this is just historically. This is what we would have done until now, until my announcement.
We would have said, whatever those numbers are, whatever your relation to the mean is — let's say you're one sigma above, meaning you get above your neighbors. So it's kind of a curve.
You get above your neighbors, one sigma above, but you get — let's say you were one sigma below on the quest.
We go back and map it back and figure out what would one sigma above been on the quest. And that's your new grade.
So what that means is it's only if you do better than you did before. So it only affects half of the students.
Meaning of all the students who are sitting here with a bad grade, only half of them can benefit because they have to do better than they did before against other people.
We are changing that. For the first time ever, ever in 61C, we are changing that so that the clobber's going to be based on either the curve or the raw score map.
Basically, raw score, let's say out of 30. Let's make the question on the midterm around the quest material out of 30.
Whatever score that is out of 30, compare with your actual quest score out of 30. If the midterm is higher, it's a new score, okay? And if that's better, well, that's your new grade.
Or if the curve is better, then that's your new grade. So whichever's better for you, we'll do that. That's a big deal. So we're giving that to you.
Never done before in 61C history. It's always been a z-score. It's always been a standard deviation based thing. Now it's the raw score as well so that everyone could benefit.
If literally everybody — Here's an example. If everybody aced the curve — let's say everybody gets either 28, 29, or 30.
Well, the problem is only people — let's say, so the average is 29. Well, the 28s are below and they couldn't get a benefit in the old way.
Now they can all get a benefit. You get it? Now you can all get 28, 29, 30 on the quest. It's a big deal.
We're listening to you. We're fixing it, okay? So that's new. And we're making sure you all know that.
Number one. Project office hours were — BORA NIKOLIC: Oh, yes. DAN GARCIA: --overloaded. Oh yeah, go, go, go.
Go back.
BORA NIKOLIC
Somebody gets a point if they can repeat what just Dan said. [LAUGHS] DAN GARCIA: Play this video on 2x on YouTube and see if you can understand a single word of it.
Let's see what we got here. Okay. Next.
BORA NIKOLIC
But the point this is good for you.
DAN GARCIA
This is really good for you, yeah. We're listening. And we're listening. By the way, we want to see the numbers go up.
Literally, we would be happy, high fiving unlike everybody else. They're like, oh, I can't give all A's. We can give all A's! This is an absolute graded course.
If you all rock the quest and get all A's, we want that. We're not trying to separate you and the weed from the chaff. No, we want you all to do really well.
All right. Project office hours, overloaded, okay? We noticed last week that project office hours — people were not getting their help in projects. And I think it's a factor of two things.
I think it's because people had to do project one alone, so you had no one to talk to.
And two, we were using Zoom and Zoom doesn't allow people — in a traditional project office hours held in the Woz, there's a whole line of people at the door.
And they're all helping each other while they're waiting to see a TA or see a staff member. So we switched over to Discord to try to resolve it.
We did that last week. We did it dynamically. In the middle of a week, we did that, okay? So that's the first thing.
When you're in Discord, have conversations. Teach each other to debug. Work each other, okay? Support each other.
This is an absolute graded course. There's no reason you need to compete with each other. Rock star, helping each other. Love that, love that, love that.
Two, workload concerns. Okay, I'm glad to hear that some of the folks are appreciating Discord. We know that this is a hard week. We know.
We actually circled this week.
Bora and I and Steven and Cici and all the other head TAs, we circled weeks like three, four, five, and six as being — it's like the hardest part of the class.
You've got to learn your C. You've got to get in your RISC-V. You've got a back-to-back project. It's the hardest time of the class. If we can get past that, it's more relaxed.
I want to make sure you guys know this. There is no lab for week six. We're in week five. There's no lab for a week six.
Week eight has a really light lab. Week nine has no lab. You guys hear this? So it gets easier.
This is the hard — if you can bear with it — by the way, I would prefer that it's hard early so that you know what you're getting into come drop date.
The worst thing we can do is do a bait and switch, where it's easy, easy, easy, easy, and then all of a sudden, drop day hits, and it gets harder.
That's unfair. So we made it hard — not that we made it hard. It just happens to be this is the hardest part of the semester.
Not that we intentionally — this is just where it all packed in. You got to get your C and your RISC, so you can get on to really good projects.
So it's hard. So again, no lab for week six. Eight has a really easy lab. Nine has no lab.
And week seven is going to teach you the tool that you need for project three. So it's all awesome, okay? So this is the hardest part.
If you can survive this, we're going to have a lot — it's an easy road. You climb this steep road, it's coasting after this. We are giving you three more project slip days!
Wooo! So this allows you to relax, lets your stuff — if you need to get a good night's sleep, get a good night's sleep! We want you to consider your health.
This is why we have our lab check-ins. One of the questions we should ask you is, how's your health? We need you to focus on your health, make sure you're doing okay.
And maybe one other class is busy just when yours is, so fine. Use some slip days to react to that. Again, you now have six slip days, okay?
Big deal.
BORA NIKOLIC
Important thing — we are giving control to you. I mean, we had so many of these requests in the last minute. Can I please — DAN GARCIA: Definitely.
BORA NIKOLIC
You know, something happened.
DAN GARCIA
Yes, exactly.
BORA NIKOLIC
Of course things are going to happen this semester. Try to pace yourself. We give this to you to try to pace yourself.
DAN GARCIA
Exactly.
BORA NIKOLIC
And if you need extra time, well, you have six days now. But manage yourself. You know, front load whatever you need to.
DAN GARCIA
Yeah.
BORA NIKOLIC
Take care of yourself. Take care of other classes.
DAN GARCIA
Yeah. And this is the second biggest one. We are moving project two to a partner project. We thought — and giving 2a three more days!
The reason for this is we think that the reason people were having such long lines is that they're all solo. You can't ask your friends. And we said, there's a trade-off of this.
When you go to a partner project, there's a chance that one partner may not learn all the things because they divide it in half and one person half, the other person learns the other half.
We want all of you to learn all of it. So help each other learn. Teach each other. Form a learning community.
And then maybe you'll each answer each other's question. We've seen the numbers go down significantly once these are partner projects.
But we still remain concerned, that we want you to learn all of the material. So both of you teach yourself all the material. How you divide it up is up to you.
We'll maybe give some best practices on Piazza. But we want you to know that we are thinking about your workload concerns. Project two is now a partner project.
Go find your best buddies. Go pair up. Find somebody. Work together.
And you also have three more days for project 2a to support that. So these are big deal changes that will hopefully fundamentally change your workload in the next couple of weeks.
Also, people are asking about quest retakes. We made a deep decision — this is a very deep decision — that we're only going to ask for quest retakes in the most extreme cases.
So your camera dropped out for two seconds as it fell off your desk. You pulled it up. We're going to say fine to that. So only in the most extreme cases.
We're going to let you know what they are. We've had to write a full new exam.
This is hard to do, by the way, especially in — it's not just write it on paper and then print it in a PDF and go. You have to write all the auto-graders.
It is hard. And Cici and I and Bora and Steven and the staff are working on it. It's hard. But we are still — we're making sure it's right.
We're making sure it's beta tested really well. So give us some time for this — probably next week — until we let you know. You now have a regrades.
You've seen your — we're going to release the first exam. You'll see — or at least the cracking. You're going to see your scores and the question you were asked on the original exam.
But because we're releasing that, we can't reuse that. So now we have to write a whole new exam. We didn't plan on that. So give us some more time for this.
I know this is hard, but give us some more time for this and we'll let you know in the next week or so if you're going to do this.
But fewer people — far fewer people are going to do the retake than before. We're mostly going to trust you using the Berkeley honor code on that kind of thing.
So if, you know, my camera dropped down for a second, I covered my clock, we'll probably waive all that. We'll let you know, okay? So we'll let what the line is.
Don't worry about what those questions are. Am I on the list? We'll look at the list. We'll look at what you commented about what it is.
We'll look at your videos. And we'll make that — that takes a lot of time. Give us the time to do that. We'll do it right.
And we'll make sure that the people who do have to retake have a good, very solid exam. We're going to ask a survey.
We're going to make sure — we're going to do check in on how things are going, how you responded to these new policy changes, where there's — I'm sorry for going over a couple of minutes.
There's a survey coming. And this one last comment I want to add, which is our staff has said, Dan, can you remind the students that we need to set some boundaries?
People are — know these other — they know our staff from other classes, and they're DMing them on Facebook. Can you help me with this?
Can you help me — outside of their hours where they're being paid, they can't do that. The union says that students can only work for a certain number of hours.
We've allocated them for the support hours, for public office hours. And if you're asking there, they have to say — we're telling them they have to say no.
This is not fair for you to ask all this other time. There are other students also. Not a single person on our staff aside from Bora and I are not other students.
Everybody else is a student too, manage their own classes. So we need to have some boundaries and work with the students within the space of the office hours, et cetera, we give you.
But please no DMing on Facebook and no DMing on all those other things, okay? I forced them to say no to those because it's too much work for them.
And that's it, okay? Very reasonable. That's it. We're over time.
Thank you very much, folks. Bora, you want to say the last couple of words?
BORA NIKOLIC
Oh, just one thing. Guys, no matter how do you feel about the quest, you can pass pretty much any C interview question with the knowledge that you have.
15 is good enough to pass an interview question. And so feel good about that. We are going to make sure that you feel — there are less assembly level questions.
Although, there are in interviews. And you'll feel like that in two weeks, in just two to three weeks. And then we'll move on.
I mean, I feel good about what we have accomplished so far. So — DAN GARCIA: Yeah, good stuff. Good stuff. Folks, keep up the good spirits.
I know the spirits were low after the quest, after the project one and two workload. But we're trying to do these things to help you guys. We really feel for you.
And we hope you can get through these couple of weeks. We know that. Here's more office — I'm sorry — more slip days, partnerships. Find your good partners, folks.
Find some partners for project two. Getting a great partner can make a really big deal, okay? We care about you. We're thinking about you.
Tell us what else we can do. Give us — we'll do the survey and we'll have a chance for you to tell us any more we can do.
We want to help you guys make this class the best class on campus. Thank you so much, folks. We'll see you next week. Wooo!
All right, take care, folks.
BORA NIKOLIC
Thank you.
DAN GARCIA
Thank you. Thank you.
BORA NIKOLIC
Bye.
DAN GARCIA
All right.
Đăng nhập để đánh giá học liệu này.
Đăng nhập để bình luận — chưa ai bình luận về học liệu này.
Không có khóa học liên quan.