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
All right, ladies and gentlemen, welcome to CS61C Once a Week Lectures live session on Parallelism. Woo! Great to have you here. Good stuff.
Hope you, boy, were as wrapped as I was last week with CNN. I don't think I've watched more hours of CNN in my life than I did last week.
Were you all locked into that? I had it on lockdown over here. I had my second monitor always — BORIVOJE NIKOLIC: I mean, I don't watch television, but yes, I spent too many hours.
And there was an impact that I posted today's videos for the lecture last night at like 2:00 AM, because I had to make up my time, right.
DAN GARCIA
Right. Right. You were on last week. That's true.
When you're on, it's hard to have something taken. I was off so I had a little chance. That's right. Wonderful.
So let's go to what we're talking about today. Today's agenda, again, welcome to week 12 everybody. We have some wonderful guests from Apple, Jared, Dave and Anand, and we'll introduce them.
Then we're going to talk about computing in the news. I've got a delightfully humorous computing in the news to share with you. A schedule, there's no new policy updates.
Just FYI, that bulletin's usually there. We're missing that. No policy updates. And then we'll have an ask me anything to talk about anything else you have, things that are on students' minds.
So I'll do — BORIVOJE NIKOLIC: But most of the time is going to go to our chat with Apple Talks.
DAN GARCIA
Right.
BORIVOJE NIKOLIC
We have Dave Conroy, Jared Zerbe, and Anand with us.
We'd like to ask them know how they ended up over in this part of their careers, but after that, we're going to ask them for advice on internships, because we have heard earlier in the semester that Apple is interested in interns
and full-time employees. So we are going to try to spend most of the time on that. But let's start with the introductions.
And so maybe we can start with Dave, because I think people know Jared already.
Dave, how did you end or, did you decide when you started kindergarten, one day, I'm going to work for Apple?
DAN GARCIA
And tell us also what your role for Apple is now. People may not have heard that.
DAVE CONROY
Sure. I'm Dave Conroy. I'm a Fellow at Apple.
BORIVOJE NIKOLIC
Uh-oh.
DAVE CONROY
A fellow is a great role, and you get to have all the benefits of being a vice president without having to do all the management stuff of being a vice president.
So it's great for a technical person like me. I've been at Apple for 17 years, and before that, I did a short stint at a consumer electronics startup that got acquired by Microsoft.
And before that, I did 20 years at Digital Equipment. Many of you don't know who they are, but they were the people who invented minicomputers on the East Coast.
So they were like a force of their era.
DAN GARCIA
Were you an East Coast person at the time, Dave?
DAVE CONROY
Yeah. I was at Maynard.
DAN GARCIA
Okay.
DAVE CONROY
I was in Maynard and Hudson, and then I moved out to Palo Alto and worked in Bob Taylor's lab at the end. So I know all those Xerox PARC guys pretty well.
DAN GARCIA
Sure, sure.
DAVE CONROY
But my path to Apple is when I was at Digital, Apple was building this thing called Newton that some of you will remember.
And I was building a microprocessor called StrongARM which most people didn't know went into a second-generation Newton that never shipped.
And so you know, we were building this thing, and then that Newton got canceled and it all kind of disappeared.
And then years later, I got this call from this guy named Mike Colbert out of the blue who was at Apple, who knew me from back when we worked together like a decade before on this Newton and said, you should come to Apple.
It'll be fun.
And of course, this is 17 years ago, which means if you're going to come to Apple, you have to decide how comfortable you are going to a place that's going to go bankrupt in the next six months, and you're going to have to find another job.
But I like the people. And so I said, what the hell. I could find a job in six months if all goes to crap. And it turned out, it worked out kind of okay.
I've worked on a whole bunch of things at Apple.
I started off working on the Intel transition, which turned out was the reason that Mike wanted me to come, but I've worked on Silicon for the phones.
I collaborated with Intel on the chips that went into the portables. I've worked on I/O standards. I've worked on all sorts of things. So that's me.
BORIVOJE NIKOLIC
One thing, clarification here. How many Apple Fellows are there? I mean, is it like 5,000 or like less?
DAVE CONROY
Two or three if you count Phil Schiller.
JARED ZERBE
Three. I think, yeah, Phil Schiller is.
DAVE CONROY
Three if you count our marketing fellow.
BORIVOJE NIKOLIC
All right, not that many.
JARED ZERBE
We call that rare air. Well done, Dave. Well done.
BORIVOJE NIKOLIC
All right, nice.
Jared, how did you end up in that lab that's behind you?
JARED ZERBE
Oh, it's really great because I got two vice presidents behind me just in case I need backup. So I can yell back to them at any time.
No I mean, so my path to Apple was, when I started college, I thought I was going to be an aeronautical engineer.
And then I took my first aeronautical engineering class and I said, this is ridiculous. These people have no idea what's going on. I mean, you know, it's just fudge factor after fudge factor.
There's no science here at all. It's just a complete slap dash thing.
And then I started — I always loved tinkering around with computers, so I started looking at a EE, and this was back in the early to mid '80s.
So I took a class across the bay from where you guys are, and got to design and build my own chip.
And I just thought that was the best thing I had ever seen, you know, that fact that I wasn't going to be doing some problem set that some professor had dreamed up at 2:00 or 3:00 in the morning, and you know, and I could actually build something of my own and have it be something different.
I just found the ability to express ideas in Silicon to be just fascinating, and so that was 33 years ago.
After that, when I graduated — and I would like impress on people the impact of when you are interviewing and you bring to the interview a die photo, a plot and a schematic of the chip that you made, it has a different kind of impression
on people rather than like digging around in your transcript for a grade or showing someone a problem set, because for somebody in industry to see a student that's actually built something and understands how to bring it up, and where the bugs are and stuff like that, it's different.
It's a different kind of interview. Let me just say that.
DAVE CONROY
It usually goes well for the student, we try to say.
JARED ZERBE
It goes very well.
DAVE CONROY
It goes very well for the students.
JARED ZERBE
It goes very well. It opens a lot of doors. There's a class, 194, that we're helping sponsor at Berkeley for people who want to take out their own chips.
It's a lot of work, but it's a lot of fun.
BORIVOJE NIKOLIC
Do Logisim diagrams count?
JARED ZERBE
That's good. That's good. It helps, but it's not the same as actually building something. So — BORIVOJE NIKOLIC: Don't tell them that.
I mean [LAUGHING] JARED ZERBE: So after graduation, I went to a company called VLSI technology, then I went to MIPS Computer Systems. I was designing floating point units.
Then I went to Rambus for 21 years. So when I look at Dave, like, Dave's a DEC guy, you know?
So I think people still look at me and think I'm a Rambus guy, because that's where you put in a lot of time there. I've been at Apple for seven years.
I'm fortunate enough to work for the other Apple fellow that's technical, not Phil Schiller. And I work in a team called exploratory design.
Which is really a lot of fun, because we get to do a lot of crazy things across lots of different spans.
And before we get into the Q&A, I just want to make two points. Number one, no, we can't talk about anything about what's going to be announced tomorrow.
So you can just take that off of your questions list, because we're not going to get fired in the process of this meeting.
And then number two, yes, Apple has internships. Yes, Apple has internships for undergrads in hardware.
In fact, the largest organization in hardware technology, which has the most internships in the hundreds for this coming summer, has a charter to, and they are successful at, hiring 50% undergrads for internships.
So just recalibrate yourself on that.
BORIVOJE NIKOLIC
All right. And Anand? Maybe you can introduce yourself. Was your path to Apple the most straightforward?
These guys took a little bit of a winding path.
ANAND SHIMPI
Yeah, I had — DAN GARCIA: And what's your what's your role in Apple as well. Yeah, share that.
ANAND SHIMPI
Sure. So I work in a group at Apple called platform architecture. I manage a few teams that are focused on, at a high level, you can think about it as power, performance, and prototyping.
So we study how fast things are, how much power they draw, and try and kind of create new things that we might turn into products or technologies in the future.
My path to Apple was a little unusual. So I've only had two jobs for my whole life. My first job was a company I started when I was 14.
It was a deep technology publication called AnandTech, so my first name plus tech at the end. You can kind of guess what that was about.
So I started that back in April of '97, kind of showed up just before the boom or as the boom was happening, and so I rode that wave up.
I never took any VC or anything like that. This was entirely homegrown. So I rode that wave up.
I lived through the dot-com bust, I lived through the kind of rebuilding web 2.0, and then lived through the Great Recession managing a company as well. And then, again, rebuilt and continued to grow.
And as I was kind of entering into my 30s, I'd been kind of doing the same thing for a very long time.
My goal always was to at some point kind of take a step back.
I always said that every big transition my company had we always grew a lot, I ended up on a plane a lot more, and I ended up with less hair.
And so we were in kind of pivot number 4 over this kind of 18-year timespan, and I said, look, it's time for me to slow down a little bit.
And the plan was to always kind of exit and go stay with the parent company for a while and help them build other verticals similar to what I had built. I always believed that kind of my core innovation there wasn't necessarily how we approached analyzing computer and mobile hardware but our approach to journalism.
And so I wanted to go and kind of duplicate that across a bunch of different verticals.
Around the time I was having these kind of thoughts and getting the company ready to take that next step, Apple approached with an interest in hiring someone who worked for me.
Which was not unusual. We were a good pipeline for people who were students.
They would get a lot of industry experience working for us, and then oftentimes they either stuck with us for a long time or they would go work in industry.
And it kind of gave you a leg up over the competition. So they approached me trying to hire one of my folks. Again, not unusual.
It was a connection that we enabled and helped make. And in the conversation, they said, hey, it'd be kind of interesting if you joined.
And I had never thought about working for any of the companies that we covered.
I think most of them, by the time product got to us — we started as a PC hardware and technology review site, eventually expanded into consumer electronics, servers, chips, all of that stuff I always felt that I could see the politics just from the outside.
That meant that I didn't want to work at most of these companies. The only one that kind of seemed okay to me was Apple. And so I entertained the thought.
I had a conversation with the person who became my boss. And after that initial conversation, I was like, yeah, this makes no sense.
But after a bunch of these conversations, it kind of opened me up to the ability to influence and make an impact on honestly one of the greatest companies in the world.
And then I always figured, hey, the bar was low enough in journalism that a 14-year-old kid that knew nothing could kind of do well in it, so if I ever wanted to go back I could I can always do that.
And so I've been at Apple for a bit over six years now. And yeah, that's my story.
BORIVOJE NIKOLIC
Thank you. This is really funny. Do you know how many times Dan and I, when we go for these computing in the news grab stuff from AnandTech?
DAN GARCIA
We pull it every week. It's great.
BORIVOJE NIKOLIC
Here is live, in flesh, through Zoom, you're Anand on AnandTech. But this is funny, I mean, most people did not know that there was a 14-year-old kid behind that website for awhile.
ANAND SHIMPI
Yeah. I actually kept under wraps for three months when I first started.
I remember the day someone posted it in a comment that, hey, this kid goes to high school, and I thought my life over at that point.
I was like, this — DAN GARCIA: You were outed. You were outed. Look at that. Fascinating.
ANAND SHIMPI
Yeah. I kept a secret for three months. And then everyone thought my dad wrote everything. And I was like, that's hilarious.
But yeah, that's another really.
DAN GARCIA
That's brilliant. That's good.
BORIVOJE NIKOLIC
All right. And there are a couple of questions that came in. Guys, feel free asking questions. It says, did any of you get to meet Steve Wozniak or Steve Jobs?
DAVE CONROY
Well — ANAND SHIMPI: Dave, you probably had the best seat.
DAVE CONROY
By the time I showed up, Steve Wozniak was long gone. I would see him occasionally at lunches, because he always got to sit in the front row, because he always got an invitation.
I met Steve a couple of times.
Most of the stories about Steve is true, and that means that lots of people spent a lot of time trying to arrange that they perhaps didn't have to be in a meeting with Steve.
And this Mike Culbert guy that I talked about before, he and Steve were very, very trusted.
And so most of the time I could get anything I wanted to Steve simply by telling it to Mike, and then Mike would tell it to Steve.
And if Steve went, rawr, he went rawr at Mike, rather than at me, which was a win.
BORIVOJE NIKOLIC
All right.
There is another question over here. I don't know. Should we tell them, or should we keep the secret for another day? We are not done grading and checking everything.
DAN GARCIA
I feel like keeping it a secret.
BORIVOJE NIKOLIC
It says, what surprise — DAN GARCIA: I feel like we keep it a secret. I don't know.
BORIVOJE NIKOLIC
What's the prizes for project three? We'll find out in a few days.
DAN GARCIA
You'll find out. It'll be good.
JARED ZERBE
I don't know. But I heard that it's arrived, right? [LAUGHTER] DAN GARCIA: Right.
BORIVOJE NIKOLIC
It's shipped. We have — i don't know if it was right. They said it was shipped to, it's protocol somewhere in shipping and receiving.
We're going to have a ton of fun getting to that.
JARED ZERBE
Just keep the — make sure you keep feeding it and cleaning it, because otherwise, you know. [LAUGHTER] DAN GARCIA: We'll dribble some water in every couple of days, right?
JARED ZERBE
Yeah, exactly.
BORIVOJE NIKOLIC
Yes.
DAN GARCIA
Don't shake the box. People are asking, is it their own CPU? (LAUGHING) is it a custom Apple CPU?
BORIVOJE NIKOLIC
Yeah, they're going. They're going — DAN GARCIA: It's the A15 It's the A15.
BORIVOJE NIKOLIC
Yeah.
DAN GARCIA
That's funny.
BORIVOJE NIKOLIC
Yeah. All right. So one of the key things that we wanted to talk about here are to elaborate a bit on what Jared said about these internships.
I mean, lots of our students find a place in internships in software as undergrads, and they enjoy them.
And a good number goes to Apple and does a software internship. But often there is this idea that you can't get an internship in hardware.
Now, you're telling us that everybody is wrong about that. And can you tell us a little bit more? What's the secret there? What do you need to know?
Can you do it with 61C type of knowledge? I mean, do you need to pass by Dave interview, the other fellow, acting as bully?
Do they need to grill you for a little bit and see how you do?
JARED ZERBE
I sure hope not, no. I mean — (LAUGHING) no. Most people that are coming for internships will not pass the Bill and Dave feelings.
I think — DAVE CONROY: I don't think I would pass the bill. [LAUGHTER] JARED ZERBE: That's true. Yeah, these are the kinds of things that we don't want to ask too many questions about.
But I mean 61C is absolutely getting into understanding the way that machines work at a very deep level, on a register level, is absolutely very important.
There are a lot of openings in — verification is really applicable, because you really deeply understand where the machine is going.
There are a fair amount of openings in embedded software that are immediately applicable with a 61C background.
But the more that you can branch out and the more that you can do and show in terms of your projects and project background interests, in my view those are the things that stand out on a resume are projects.
How are you doing in labs? How are you doing projects? We're talking more about what it is that you've made. That makes a big difference.
I don't know, David and Anand, what you guys think about students and software and hardware, and how you see them coming in and what you see as being successful.
DAVE CONROY
I mean, as you know, I'm a super believer in internships.
When you're learning this trade, for lack of a better word, a really, really important part of it is to have a really, really solid foundation on the theory and the next click above that, and academia does a spectacular job of doing that.
But a part of learning a trade is just learning how to get your hands dirty and the millions of little jobs that have to be done.
And as someone who spent their whole life sort of building computers and shipping them to people, it's amazing how much of the work is just work as opposed to unbelievably, fancy, remarkable, earth-shattering work.
There's just a lot of stuff that has to be done, and it's important stuff. It's not — you got to do it. You have to kind of slog your way through it.
And the good news for that is it means that for someone who's sort of at the beginning of their career, like an undergraduate in EECS, that you can come to a place like Apple and do some stuff that you're perfectly capable of doing with a little bit of the help
from the people who are already there, and it's important.
This stuff that you're doing, and you're kind of learning how to set up the thing for this CAD tool, and learning how awful this CAD tool is so that you can be part of the people who know how awful this CAD tool is, that if this doesn't happen, there is no phone next year.
It's really good. And you come out of it.
You get to make a contribution to stuff that you get to see in the world and you get to learn just a little bit of the hands dirty part of the trade.
From my background, I'm a Waterloo guy. And Waterloo is a place that believes in cooperative education. You can't get an engineering degree without doing it.
And I had the good fortune to, at the beginning, get tied in with a bunch of guys in this little engineering company who remembered building computers out of vacuum tubes, and I learned a billion things from them.
It was just a really, really great.
BORIVOJE NIKOLIC
That's awesome. I can tell you — yeah, go ahead.
ANAND SHIMPI
For me, and I think this is true whether it's an internship or someone that we're just trying to hire who's out of college, I think technical depth is really important.
Be a master of some domain. Have a lot of skill that you can demonstrate in some area. But the thing that I often find is missing is passion, right?
We will interview for a position for a year and find one person out of all the interviews that year that really cares about the stuff. They really have a passion for it.
They're really enthusiastic about it. I think that's — if you have the right intersection of technical depth in something, whatever your specialty is, some technical depth and passion, you immediately stand out to me.
And it's such a hard thing to find, both amongst, I think, new college grads, people who are in the industry, or even interns.
But if we find someone like that, those people are the people that get routed to me. And they're like, you need to go talk to this person.
And so I think that I can't overstate enough the value that I think that brings.
BORIVOJE NIKOLIC
So what do these guys need to put on the resume to meet that bar? I mean, there is a lot of them that we have.
JARED ZERBE
To that point, to add on to what Anand and Dave said, I think a lot of it comes down to the person.
It really comes down to the individual — having the tenacity and the follow through. And Dave said, a lot of it is just work.
That means that what you do to differentiate the work that you do — are you doing quality work?
Are you bringing it on time? Are you making it happen? Are you supporting the people around you? Are you being a good team player?
Are you thinking not just about your block but are you thinking about other people's blocks? Are you studying the code around you?
Are you are you learning beyond what it is that you just need to do?
There's a great book about Bell Labs, and I think one of the leaders of the Bell Labs said, what you do in the eight hours per day that justifies your salary and your medical.
And what you do in the other 16 hours a day and the weekends, that's your bonus right there, right? So it's thinking about — and it's not just like pouring the coals on it.
It's really thinking deeply about these things, and spending some time, and kind of getting hooked on what it is — the joy of it, right?
And if you're not loving it, then it's not going to come out right.
So do the thing that you — it's a little bit of a trope, but do the thing that you love, and it will show through.
And one of my favorite stories about when I first started at Apple was my wife asked me, after the first week, she said, well, how's it going?
And I said, it's good, but I'm hungry all the time. I don't know why I'm hungry all the time.
And it didn't take me long to figure out it was because I would get down and start looking at a problem, and then start looking at what different people had done.
And then before I looked up, it would be like three or four hours would have passed, you know? And it'd be like, lunch flew by.
And I was just like, oh, yeah, this is cool. This is really cool. This is really, really cool.
So there are a lot of people who just love to nerd out on what it is that they do at Apple, and I think that definitely is a key characteristic.
BORIVOJE NIKOLIC
So I'll ask some questions over here on the behalf of the class, and then after that we get a few Q&A questions that are coming in.
One thing from my interaction with the students in the class, what they've figured out so far, that they seem to be every day more and more comfortable with performance programming, embedded programming, how to fit things into tight spaces and get maximum performance out of that.
And they get a good feel for hardware but are one step away, in 61C, from actually building real hardware past the logic system.
What's your advice for them? First, do you hire people who need to pack extremely tight matrix multiplication loop and tight neural net into a little embedded space that is out there?
And once, when they master that, how do they go and build their own neural net hardware or something else exciting, or maybe verify it and so on.
ANAND SHIMPI
I think I can at least speak from my orgs.
There are people I employ who even though we exist in a hardware org, their strength is being in that intersection between really understanding how code runs on bare metal and understanding the architecture.
And so they never need to ever build a chip or verify a chip or do anything like that.
But understanding and being able to exist in that kind of gap between the two, it's very rare and very important.
And so people in that space with a lot of expertise on how to make code run very quickly on the underlying architecture, that's a very important skill to have.
BORIVOJE NIKOLIC
Sounds good. And how do you make the next step?
ANAND SHIMPI
What do you mean by the next?
BORIVOJE NIKOLIC
Maybe others can say. How do you actually go towards building your own hardware? I mean, how do you get comfortable interviewing for a hardware position for example?
Maybe perhaps a question for Jared. Jared will be better equipped for that.
JARED ZERBE
Getting comfortable interviewing for a hardware position — I think there are two parts of it.
One is that you need to come into it with some confidence.
And the confidence comes kind of by doing your homework, right?
If you go into an interview thinking, you are crazy if you don't hire me, you are insane.
I am the perfect person for this job, because I've done this and this and this and this, and I really think that I can make a big impact here.
That is much more powerful than, well, I don't know if I'm going to fit here, and I'm not sure.
So I think do the homework that you feel like you need to do to be prepared, and be overprepared. When you're overprepared, then it shows. It shows all the way through.
And if there's one thing that I noticed about people that succeed at Apple, it's a certain attitude of they're there to solve the problems that mother nature has given us to solve in order to deliver the next feature, the next level of performance.
They're not there for to wait around for someone to tell them what to do. So a lot of curiosity I think is really important too. There's no magic formula.
I mean, I think there are a lot of venues for this class. There's the special pipeline. There's the FlashChat signups. There's the applying throughout.
But there's lots of this mechanics. Don't get too — go through all the mechanics, dot all the i's, cross all the t's, but do your homework.
Know your stuff and get passionate about the technology, and the roles will come to you or they'll be there when you pursue them.
DAVE CONROY
And the truth is that for a lot of complicated hardware, a really, really useful thing for a designer to have is, for lack of a better word, a sense of style, that you look at some pieces of hardware
and you say, programming this is just ugly.
And other ones you look at it and say, the person who specced this hardware understood how the software was going to feed it, because it was clear they had written a lot of software to drive low level hardware.
So don't underestimate the value of being a skilled embedded programmer.
A really skilled embedded programmer who's grounded in sort of the fundamentals of the hardware underneath it is actually a pretty attractive candidate for some hardware hardware tasks, because you don't have to teach them style.
BORIVOJE NIKOLIC
That's a good one. Here is a question — are you looking for operative EE courses?
JARED ZERBE
I was just typing an answer to that one.
BORIVOJE NIKOLIC
That's okay. We can answer live.
JARED ZERBE
Yeah, I mean, absolutely 140, 151, 152, those are great classes. But like I was saying before, there's no formula.
It's not like you're going to go tick, tick, tick, tick, and then your transcript has these things, and then all of a sudden, the UFO lands in your backyard, and it comes down and shuttles you away to Apple.
It doesn't work like that.
I mean, a big part of what happens in those classes is you get to people like Bori better, you get to know people like Dan better, or like Kirsten better, or like Rikky better.
Go learn the craft in a deep way and start working with the people that are at Berkeley that are so fantastic.
You've got Rikky Muller teaching an analog circuits class with 15 people on it. That's criminal. I would love to take that class with her.
If I was your age, whoever you are out there in Zoomland, I would go take that class right now and whatever prereqs. You know, prereqs are guidance.
I would go to professors and say, what do I need to — [LAUGHTER] See. Now, Bori is never going to invite us back.
I would go to professors and say, I am so excited about this stuff, what do I need to do to talk you into letting me take your class?
Because if you're excited about it, it comes through it every single way, and it comes through in your style too. Just like Dave was saying, it comes through in your style.
I can look at someone's schematic and tell if they were bored doing it or whether they cared about their craft. There's a key to this, and it requires some degree of passion.
And I'm going to shut up.
BORIVOJE NIKOLIC
All right. Good. Thank you. Here is another one.
What do you think have been the biggest successes and failures for Apple's hardware group, and what do you see as the main challenges going forward?
DAN GARCIA
And feel free to also share about maybe the market wasn't ready. The Newton was amazing, but we weren't ready yet — the market wasn't ready for it, even though it was a great product.
Feel free to share either marketing or engineering-wise in terms of the successes or failures.
DAVE CONROY
Well, I mean, in recent times we've had a pretty good string of successes for sure, some of them beyond our wildest dreams.
We've had things that internally have not worked out quite as we would have expected, and we don't talk about them very much.
Not because we're ashamed of them, because sometimes they end up turning into something that actually works out really, really well a little bit later.
I stopped counting the number of times we tried to build the iPad.
The iPad that you know as the iPad is not the first one, it's not the second one, it's not the third one. We built countless internal prototypes.
And either it was too big, or it was too heavy, or the display didn't work, or using a pad with a stylus is a stupid idea.
And you'd build it, and you'd say, well, that's interesting. We learned some stuff. But we don't declare it a failure, we just put it on the shelf until something happens.
And then someone says, well, that's interesting, let's take it off the shelf. And frequently those failure internal failures turn into turn into successes.
DAN GARCIA
Sometimes you can learn as much from failures as you can from successes.
DAVE CONROY
You actually usually — JARED ZERBE: Or more. DAVE CONROY: --learn more for failures than you learn from successes.
DAN GARCIA
What didn't work. What didn't work and why.
DAVE CONROY
That's actually why when I do get to interview somebody, I always like to talk about stuff that didn't work.
The discussion is usually far more interesting, and you'll learn a lot more about the person.
DAN GARCIA
That's great. That's great.
BORIVOJE NIKOLIC
And there was a Berkeley project that went from 1992 to 1995 that was called Infopad.
And it was a great success, although not the commercial — there was no immediate commercial transition.
There are a few people — the guy that kind of was the driving student behind it is now a Dean of Engineering at MIT. So you can call that a success.
Anantha Chandrakasan was the senior student.
DAVE CONROY
When I was at Digital in the research labs, HP announced that tiny little disk drive. I think it was HP.
JARED ZERBE
Yeah, yea, yeah.
DAVE CONROY
The hummingbird, right?
JARED ZERBE
Hummingbird. Yeah.
DAVE CONROY
And we looked at that and said, I know what you can do with this, you can make a music player.
And we built this thing that had the disk, and a single processor, and a bunch of mp3 decoding code that we licensed from Fraunhofer, and a bunch of software for it, and built this thing that it was about eight inches by 4 inches by an inch and a half.
And it could store like 500 songs, and it had shuffles, and equalizers, and all of those things, and it went nowhere.
Actually, that's not quite true. Compaq didn't know what to make of it, and this Korean company came to them and said, we would like to license that. And they said, sure, right?
And the Korean company licensed it and built 100 of them or something like this and tried to sell them. And nobody bought them, and it all seemed like a total failure.
Jared's laughing because he knows where this ends, right? And then we built the iPod, and the iPod was a great commercial success.
And when the iPod was a great commercial success, all of these people would come out of the woodwork and try to explain why they had this patent that read on the iPad.
And the lawyers would point to this little Korean company and say, that concept was previously available in congress, and another patent would die.
BORIVOJE NIKOLIC
All right.
DAVE CONROY
It was really funny. So I contributed to the iPod before I worked for Apple.
BORIVOJE NIKOLIC
All right. A couple of very quick ones. I think this is for Dave. Can you describe a little bit better what is good style?
DAVE CONROY
The short answer is no.
DAN GARCIA
You know it when you see it, right? You know it when you see it.
DAVE CONROY
You know it when you see it.
You sort of look at it, and you try to write some software to drive it, and you realize that this device does not have some stupid read modify write bit right in the middle of a field that you have to change every time you start I/O or something like that.
It's really hard to describe, but you sort of tell it when you see it. You see, this is really, really easy to talk to.
And the other thing is everything is clumsy, or the way the interrupt works is really crazy, and you have to read these four registers, and dismissing it's really hard.
As opposed to, it just does the right thing, right?
ANAND SHIMPI
Good style or not good style probably rears its head way, way down the line when you run into scalability issues, or you run into maintenance issues, or all of a sudden you have to grow the team 10x in order to support the thing.
Good style, good architecture, those tend to go hand in hand.
DAN GARCIA
And by the way, there's a software equivalent, which is just a good API.
I'm thinking about you're going to write some function, some method, some class, what's the API that you create for it, and style comes into that as well. Certainly.
JARED ZERBE
Style is all over the place and software. I mean, are you kidding me? When do you start documenting?
Is that 20 minutes before you turn in the assignment you start your documentation, or are you doing it all the way along?
I mean, I think the thing about style is that it shows — and I saw the question here, do you try and make the hardware as neat and pretty as possible — the funny thing about that is it's symptomatic of the mindset of the person that created it.
How important is it for things to be done in a clean way?
That means that I used to have an instructor that said, I am so lazy, I will try 1,000 times to make a proof that is one line shorter.
But the idea there is to make a really clean implementation and to understand the value that there is not just in superfluous elegance, but in terms of the quality of the work that's done.
So when you look at a wire wrap board or you look on even a proto board and the wires aren't all helter-skelter, and they're not like hanging on by one thread or something like that, it means that there's attention put to detail.
Now, if someone had it all helter-skelter and then they shoved it all in to make it look pretty because they knew the Professor was running over, sure you can cheat on stuff like that.
That'll get you a little bit of a way.
But really, the more important thing is to get into the mindset of spending the amount of time to be careful, and cognizant, and mindful of the different choices that you're making.
Because especially in programming, oh, my God, the data structure is the most important thing. You should spend 90% of your time on the data structure and then 10% of your time on your code.
Sit on your hands. Either that or you can write it five times. You can write it the first time, figure out that's wrong, throw it all out.
Write it a second time, figure out that's a little bit better but it's still wrong, throw that all out. Write it a third time — you know.
And then, eventually, you're getting the hang of it, right? You're going, okay, well then that's what it was. You don't get to become an expert without 10,000 hours, right?
And you don't spend all the 10,000 hours making perfect things. You make a lot of pretty crappy things along the way.
DAVE CONROY
There was a guy — BORIVOJE NIKOLIC: When we were — yeah.
DAVE CONROY
There was a guy a DEC, I forget who it was, but he used to always say that when he worked away on something for what seemed like an insufferable long time, and when it was all done people would come in and look at the result and say, how did that possibly take you that long to do this?
This is really easy. He knew that he had actually done a good job.
That might have been Dave Rogers who ran PDP 11 Engineering for many years.
BORIVOJE NIKOLIC
Wow. We mentioned some of these machines, by the way, in the class. Well, thank you very much. Thank you for coming, stopping by.
This was very informative. Thank you, Dave, guys. We had Dave Conroy, we had Jared Zerbe, and we had Anand of AnandTech himself in flesh.
So thank you again, and we'll see more of you. And we're going to be distributing these prizes starting for project three right next week.
And I believe there are going to be paper certificates, and then after we're done with project four we are going to distribute those as well. And internships.
You guys are ready. These guys are telling you, you are ready.
DAN GARCIA
And they want you. So make sure to apply. Make sure to apply that one, and have good things to say.
BORIVOJE NIKOLIC
All right.
DAN GARCIA
Thank you, folks.
BORIVOJE NIKOLIC
Thank you.
DAN GARCIA
That was wonderful. Nice to see you all.
JARED ZERBE
Thank you.
DAN GARCIA
Thanks for your time, everybody.
JARED ZERBE
A lot of fun. All right.
DAN GARCIA
Great stuff. Bye. Take care, folks.
JARED ZERBE
Thanks. Bye, bye.
DAN GARCIA
All right, good stuff. We're up to the schedule. I wanted to actually do — the schedule, I think I moved to here. Here we go.
Boom, there. I want to do computing in the news.
So I have this delightful computing in the news, which was just with all the insanity of the election it's nice to have something just fun and frivolous for the computing in the news.
Not like, oh, somebody hacked a voting system. And oh God, what does that mean for everybody.
Thankfully, the only news that I could find was this really delightful, whimsical story, which was — so they have now a camera — by the way, if you have a goal to be a career and as a camera operator, I might encourage you to rethink your goals, because AI is taking that job.
AI is also taking the job of radiologists. Radiologists' job is to read a black and white picture and find out where the fracture is. That's something that can be trained very easily.
So watch out — BORIVOJE NIKOLIC: My cousin is a radiologist, and he claims he is not going away.
DAN GARCIA
Well, I mean — BORIVOJE NIKOLIC: He is using AI. He's us using AI.
DAN GARCIA
No, no. I appreciate it. But I think the first pass may be being done by computers pretty soon. So in this case, in the camera operation job, it's pretty easy.
All you do is there's movement up and down, just like driving, right? I'm just doing this. That's certainly something that a computer could do. So it's moving left and right.
It's zooming. And so they figured out how to write some software to control a camera, and they had this. And it was doing well for recording soccer matches.
And it was trying to do was put the soccer ball in the center of the screen. Well, it turns out — let me show you the picture. This is a picture of what happened.
And you see, this is a referee, and the camera kept wanting to put the referee's head in the center of the screen.
The referee had a bald head, and where he was standing on the sidelines, his head is actually in the field from the projection from the camera to there.
And so it thinks, well, the ball is there. And so it kept, the entire match, going back from the game to the referee.
And then I'm like, why is the camera keep showing the referee? What's the deal? And it turns out — BORIVOJE NIKOLIC: And the lighting is just that way.
DAN GARCIA
It's exactly — a reflection exactly — the sun's on the far side. But anyway, I wanted to share that very humorous thing. They confirmed it.
They confirmed that it was a trouble that the Zoom thought it was there and it kept zooming in. And they figured the flare was shortly after the game ended, blah, blah, blah.
But it was just a lightweight, funny thing that — The larger point of this is that machine learning systems and computer-trained systems can be very brittle, and this was an example of a brittle system.
We've seen this with facial recognition, other things like this. And so this is a case of recognition just being a little brittle. So just having some fun with that.
All right, schedule. This is what our schedule looks like. I don't know if you can see. This let me Zoom in a little bit so you can see here.
We're at week 12. I've shown week 11 behind us and week 13 ahead of us. We've got no lecture on Wednesday. So take a break on Wednesday, folks, for the whole campus.
Rest and recharge if you can. And I think in two weeks after that, we're going to have Thanksgiving. So enjoy your Veteran's Day, and Happy Veterans Day to everybody.
Next week — oh, thank you — Monday, today's lecture is on SIMD and Flynn taxonomy. On Friday we start a whole new module.
So I want to encourage everybody who's behind and wants to kind of restart to jump in again. Start on Monday really, because Monday is really the first part of this section.
This whole section on parallelism, this is the topic. If you see the bottom part of my slide here, it say parallelism. This is the topic of parallelism.
So please do jump in if you were behind, and you can catch up here with 32, 33, 34, 35.
And then the fun thing is we're going to be adding 36 and 37, which is map reduce, spark, and warehouse scale computing — so how can you, literally, with a couple of lines of software, a couple of lines of Python, control a million machines.
Really amazing stuff that these abstractions are just remarkable. So we'll share that.
BORIVOJE NIKOLIC
Important comments about that. Project four depends — basically, you can start project four with your C knowledge. You can get a good head start with — DAN GARCIA: A good start, yep.
BORIVOJE NIKOLIC: --with project four. Yeah. There was a good guidance through the navigation through the project four that was done by Gia. I think it's recorded.
And please go ahead and take a look at that. In order to excel in it, you basically need to start with Monday's lecture.
And then that is a little bit longer, but I really wanted to squeeze in these things that are important for the project.
And then you have, I think about two lectures from then, Fridays and Mondays. We'll get you everything you need to do for the project — maybe Wednesdays as well.
DAN GARCIA
Yeah, maybe a little bit Wednesday as well. Good stuff. All right, great. So now let's take a look at where we are with our schedule again.
So this is another pass on that. So we're in parallelism, holiday, parallelism, TLP 1, 2, and 3, and then map reduce and spark in lecture.
In terms of homework, none of this is new information. This is all just copying and pasting from the website into this, just so you know. But 7 and 8, we moved 7 and 8.
We pushed them down and slid it to today. So they're due today. Hopefully, that isn't too bad to be able to put them together.
I've been told that it doesn't take that long to finish both of those, and hopefully you've been doing that.
We've removed 9 and 11, although they still show on the website, because they were going to be for practice only. We're going to remove the dates. We're going to make sure it's there.
But this is not new information. Just make sure you know that 9 and 11 are just for practice only — so get some work there, play with that.
You won't need them for the exams. That was a question people were asking. You will not need the four exams.
All the material — all the things you learn there we will also be teaching in lecture, and the exam will be tested on material that you could have gotten from lecture.
So you don't need homework and 9, 11 for the exam. It's not like it's really a secret thing that's there. That's different from labs. Again, this is just a clarification of a previous policy.
Lab, the material is due on the exams. But by not having you have to do a check-in, we're basically giving you flexibility to be able to do those labs anytime you want to.
And ask our TAs whenever you have any questions, for you to ask them, but it's out there. It's like reading the book. So our labs become reading the book.
The book is also on the exam, so make sure you do that. We also took a look at the project four deadline, and we moved that deadline back.
Maybe we'll pace this again into Piazza. People don't watch these videos. But project four, we moved them back so you can enjoy Thanksgiving.
It was going to be due the Monday after Thanksgiving, and no one's going to — I mean, that's going to be a painful thing for everybody in terms of Thanksgiving.
So we slid it back to be due 12/2, and you have a little bit of time at Thanksgiving. I think it's due Wednesday or Thursday. I think 12/2 is either Wednesday or Thursday.
Maybe we can — BORIVOJE NIKOLIC: Wednesday. It is Wednesday. We don't have a ton of slippage after that, because there are some people who have extra time and we need to wrap things up.
DAN GARCIA
Six slip days. We have six flip days. So yeah, we couldn't go into 2021 because of all those slip days, so we did that.
And this is also an existing information, just saying it again, the midterm retake was pushed to 11/15. 11/15 is a Sunday.
You get 24 hours, and we're going to be letting you know very early this week, CC is working on the file, and we're going to make a pass on this, and declare very early this week who needs to take it.
And then we'll let you have a chance to say, well, I wasn't on the list, but I want to be on the list, or I was on the list but I don't think I should be on the list, because you can't find my video.
So we'll try to have another pass where we'll declare a list of students we think should take it. Then will give you a chance to either add you or remove yourself from list.
We'll make a call on that and then give you the final list, and that's who we'll expect to take the exam. But that's it.
And again, we're only going to take the exam for pieces that you missed, not for the whole exam.
Unless everything is messed up in terms of your video, in which case we might ask you for the whole thing done. But that's it. All right.
BORIVOJE NIKOLIC
And that's it.
DAN GARCIA
That's it. Any questions? I'm looking at the Q&A. I see nothing. I'm looking at here.
Yeah, I think we're pretty clean. There's no information. So it's not like, well, we made some new policy and there's questions about it. So I think we're pretty clean there.
BORIVOJE NIKOLIC
And I have a collection of some internship postings. I'm going to start posting them on Piazza. And you guys are ready for many of those. I mean, you're in good shape.
Go for it.
DAN GARCIA
Good stuff. And thank you, by the way, for attending. I see about 50, 55 or 56 people were here today. So thanks, folks, for coming to this.
Let's see.
BORIVOJE NIKOLIC
We have a lot of slip days we were not done grading. So no paper certificates today yet. We are not done checking everything out from project 3B.
DAN GARCIA
Yep. Good stuff. All right. All right, folks.
BORIVOJE NIKOLIC
All right.
DAN GARCIA
Doing pretty well. Thanks, everybody. Good to see you all.
BORIVOJE NIKOLIC
Thank you. Thanks.
DAN GARCIA
We'll see you next week. All right, take care, folks. Bye, bye.
BORIVOJE NIKOLIC
Thanks. Bye.
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.