The HLF Portraits: David A. Patterson
This is a modal window.
Das Video konnte nicht geladen werden, da entweder ein Server- oder Netzwerkfehler auftrat oder das Format nicht unterstützt wird.
Formale Metadaten
| Titel |
| |
| Serientitel | ||
| Anzahl der Teile | 66 | |
| Autor | ||
| Lizenz | Keine Open-Access-Lizenz: Es gilt deutsches Urheberrecht. Der Film darf zum eigenen Gebrauch kostenfrei genutzt, aber nicht im Internet bereitgestellt oder an Außenstehende weitergegeben werden. | |
| Identifikatoren | 10.5446/44108 (DOI) | |
| Herausgeber | ||
| Erscheinungsjahr | ||
| Sprache |
Inhaltliche Metadaten
| Fachgebiet | ||
| Genre | ||
| Abstract |
|
The HLF Portraits51 / 66
2
9
14
16
17
18
20
23
25
26
27
29
30
34
35
36
38
40
43
46
50
53
57
58
00:00
Elektronisches ForumBitFamilie <Mathematik>MathematikStereometrieBildschirmfensterProgrammierumgebungKalkülPunktrechnungResultanteTabelleTermZentralisatorGüte der AnpassungTeilbarkeitExogene VariableAutomatische HandlungsplanungVerzweigendes ProgrammGewicht <Ausgleichsrechnung>VererbungshierarchieKlasse <Mathematik>VersicherungsmathematikerNeuroinformatikMinimalgradMultiplikationsoperatorRechter WinkelEinsComputeranimationBesprechung/Interview
07:36
ComputerspielMathematikSchaltnetzProgrammierungMAPSoftwaretestEntscheidungstheorieAggregatzustandAlgebraisches ModellKalkülMereologiePhysikalisches SystemRechenwerkZahlenbereichQuick-SortFlächeninhaltGüte der AnpassungNichtlineares GleichungssystemOnline-KatalogFaktor <Algebra>Metropolitan area networkVererbungshierarchieKlasse <Mathematik>GradientQuelle <Physik>t-TestPlastikkarteNeuroinformatikMinimalgradMultiplikationsoperatorSchlussregelRechter WinkelAssoziativgesetzBesprechung/Interview
14:52
ProgrammierungNeuroinformatikComputerspielHardwareInformatikMathematikComputerarchitekturSoftwareGebäude <Mathematik>MatrizenrechnungStörungstheorieSoftwaretestGrenzschichtablösungEinfach zusammenhängender RaumFakultät <Mathematik>GruppenoperationMereologieProjektive EbeneRechenwerkStatistische HypotheseStellenringTechnische InformatikFamilie <Mathematik>Faktor <Algebra>Prozess <Informatik>Notepad-ComputerPunktVererbungshierarchieKlasse <Mathematik>Gradientt-TestVollständiger VerbandWeg <Topologie>Schreib-Lese-KopfMinimalgradMultiplikationsoperatorStandardabweichungTuring-TestRechter WinkelOffice-PaketOrtsoperatorGamecontrollerEinsBesprechung/Interview
24:30
ComputerspielFormale SpracheHardwareInformatikMustererkennungOrdnung <Mathematik>ProgrammierspracheComputerarchitekturSoftwareSpieltheorieProgrammierungExpertensystemMittelwertProgrammverifikationEntscheidungstheorieGesetz <Physik>AggregatzustandAuswahlaxiomBitFakultät <Mathematik>MereologiePhysikalisches SystemProjektive EbeneTermZahlenbereichFlächeninhaltTeilbarkeitFamilie <Mathematik>AbstandCASE <Informatik>Prozess <Informatik>DatenfeldRoutingPunktVererbungshierarchieKartesische KoordinatenUmsetzung <Informatik>t-TestSchreib-Lese-KopfNeuroinformatikMinimalgradMultiplikationsoperatorSchlussregelSchreiben <Datenverarbeitung>Rechter WinkelBell and HowellEinsBesprechung/Interview
33:52
Formale SpracheHardwareMikroprozessorMustererkennungNatürliche ZahlComputerarchitekturRückkopplungSoftwareProgrammierungEichtheorieFormale SemantikDigitalisierungGebäude <Mathematik>ProgrammverifikationGesetz <Physik>BildschirmmaskeBitComputersimulationDivergente ReiheGrundraumHöhere ProgrammierspracheMereologieMultiplikationPhysikalische TheoriePhysikalisches SystemProjektive EbeneSpeicherabzugStellenringFlächeninhaltReelle ZahlParametersystemCoprozessorHypermediaCASE <Informatik>ProgrammfehlerDatenfeldSampler <Musikinstrument>PunktArithmetische FolgeWort <Informatik>Quelle <Physik>Hilfesystemt-TestDigital Equipment CorporationWeg <Topologie>DifferenteNeuroinformatikElement <Gruppentheorie>MultiplikationsoperatorTuring-TestNetzbetriebssystemAssoziativgesetzEinsBesprechung/Interview
43:15
HardwareMikroprozessorComputerarchitekturHeimcomputerFrequenzGebäude <Mathematik>Generator <Informatik>SoftwaretestBildschirmmaskeGruppenoperationLeistung <Physik>MaßerweiterungProjektive EbeneTermVirtuelle MaschineZählenZahlenbereichFestplatteFlächeninhaltTechnische InformatikGüte der AnpassungFamilie <Mathematik>PrototypingInternetworkingParametersystemCoprozessorCASE <Informatik>Perfekte GruppeCoxeter-GruppeDatenfeldSampler <Musikinstrument>Notepad-ComputerPunktProgrammierparadigmaDisk-ArraySchnittmengeArithmetische FolgeWort <Informatik>t-TestPunktwolkeDifferenteMini-DiscNeuroinformatikAlgorithmische Lerntheoriesinc-FunktionKontextbezogenes SystemMultiplikationsoperatorKollaboration <Informatik>Rechter WinkelMooresches GesetzEinsBesprechung/Interview
52:37
Elektronisches ForumVorlesung/KonferenzComputeranimation
Transkript: Englisch(automatisch erzeugt)
00:18
Dave I'd like to start at the beginning actually before you I'd like to know a little bit about your family your parents, okay
00:25
so Long term my ancestors on my father's side came from Scotland and they were relocated in Ireland In Northern Ireland to help suppress the Catholics. Yes, Scott Irish. Yes, Scott's Irish We're our my family religion was Presbyterian, which is actually the Church of Scotland. So that goes back hundreds of years and
00:46
As a result for our weddings we wear kilts And my mother is largely Swedish and My parents met in Illinois. It was right
01:02
The Graduating from high school. My grandfather had served in World War two and lived in San Diego and moved away And my father was going to join him, but he couldn't leave his high school sweetheart So they got married Right after he graduated from high school and then moved to California
01:22
So you were born in in Illinois, but they were quickly. Yeah, I was I've been here since before I was four years old. Oh, so essentially California is the environment in which you grew and it's California boy and it's around LA Yes in the beaches of Redonda of Los Angeles Redonda Beach and Torrance
01:43
Which have gotten a lot more fashionable than in our days And so I was just here this weekend and they're pretty expensive places to live in amazing We're not in our day. I say our day because I grew up not far from you. Okay, so This is a family. I don't mean educated but professional
02:02
So no one had graduated from college in either branch of the family My dad started going to college in Illinois and then They couldn't afford housing in the Illinois at the time. So at the time in 1951
02:20
Housing was cheaper in California than Illinois hard to imagine. Yeah, so they Got in the car and drove to California and he had finished. I think his junior year at Northwestern was partway through his new year and his plan was You know coming out and raise a family and go part-time to finish his bachelor's degrees at USC
02:42
So we became a family of four. I was the oldest Yes, and after you know more than ten years going part-time and he hadn't yet finished His joke was he wanted to graduate from college before I did but he lost that bet I graduated from UCLA and then he graduated from
03:03
University of Southern California a couple years after I did so I was the first of my college to credit and Your mother never did no, she unfortunately very smart she Got bad advice about going to college as a woman Graduating from high school in 1946 and kind of missed that window of opportunity and then had me and then
03:24
two Sisters and a brother and kind of got tied up. I didn't get that chance fair enough. Um, a household full of books Parents with ideas of what you should do, you know, what's no they Really encouraged, you know school was important
03:42
the thing that my parents tell me now is what they did was they'd sit around the kitchen table and as kids got older They would help with the homework and what they noticed was That I would answer a lot of the questions when my kids Younger siblings would ask the question. I try and I just taken the material so I try and answer it
04:03
So eventually my parents didn't show up at the table anymore and it was just my the four of us and I would help them out and my really Sister who was the next? Oldest one also would help and both of us ended up in education She was a high school teacher and I went to you know, taught at Berkeley So I pretty sure that growing up experience explaining things to my younger siblings, you know influenced my love of teaching
04:28
Excellent. So I'm assuming you went to a public school not a private school, right when we grew up in Southern, California You know The only private schools were for the Catholic kids then private schools and if not you went to the public school
04:41
I remember that actually private schools were for the kids who couldn't make it a public school Sometimes if there were private schools. Yeah, they were I mean not Catholic schools, but not the other ones Yeah, if that yeah, they were for kids who couldn't make it in the public schools. Okay, so Solid elementary education good teachers
05:01
Yeah in high school Elementary school. I don't think there was anything special. So good. It was a good suburban school, right? High school. I was fortunate to have a really great math teacher and this they really didn't teach calculus in high school where I went but he was a he was a broad guy who could do many things and I
05:24
So he was he was able we got up I was put in the advanced math class and and Was able to get to calculus there and when people asked me when I wanted to do when I went to college I had no idea but my high school math teacher always wanted to be an actuary and so
05:43
When people ask me, I would say an actuary. I didn't really know what an actuary was Called recently that an actuary is like a bookkeeper without the central planning personality That's what he wanted to do and that was your answer yeah, that was your answer are no computers at all
06:02
You know, I had never thought about computers. I knew they existed but didn't really cross my mind. There's another factor I want to know about the young Dave Because I've talked to your future colleague and he said he was lazy You knew was pretty smart in Elementary school and high school, but he said he really needed and it didn't happen till he went to Catholic school
06:25
He needed someone to say, you know, don't be lazy, you know start. No, what were you were you hard-working? I was pretty hard-working I I think one of the things that really shaped me as I think about in high school is I was small I was both one of the youngest people. I was my birthday was just before the age cut off
06:43
So I was almost always the youngest person and I think my family kind of matured late So I was a small person and when I was a freshman I probably weighed a hundred pounds and really was five to freshman in high school freshman in high school And then that but for some reason my parents thought I'd be good at wrestling. I have no idea why but I
07:04
wrestled all the way through high school and then in college and so I think wrestling was a big experience to shape me and then Wrestling uses up a lot of time and it takes a lot of discipline Not only is it tremendously physically demanding you have there's weight classes So you have to diet so that you get your up and weight class. So I think there was a lot of
07:24
Discipline around being able to participate in in sports and you didn't you used up a fair amount of time so I think the parents example, of course, everybody's gonna do their homework and then Being in the family of four or having your responsibilities and the wrestling and so I think I I work pretty hard
07:42
Well, okay, at least people kids work really hard today for that era. I was a hard worker Got it. Got it and for California, okay, so You've already been noticed by your math teacher in the sense of having somebody with ability I want to tell you one please please like that my mom So I got sick when I was in eighth grade and I missed a couple of weeks of classes
08:04
So I tried I had to make up a bunch of tests and when I was making up the tests I thought I saw the answers imprinted, you know I could see the dents in the paper the numbers and I was way behind so I just started copying the numbers down To try and catch up with the class because I was so far behind
08:22
I thought that was shortcut. So my eighth grade teacher just thought I was an idiot and he said he came and Interviewed my mom for the parent-teacher conference and she said You know this not everybody's cut out for college and I think you know, I don't your son absolutely cannot take algebra there. So
08:42
She my mom who has a sense of temper like me just was umbrage But I was forced between my eighth grade and freshman year to take a general Math class in summer school to shoot proved that I could take algebra It's the opposite of a mentor It's a doubter yes, and I think my mom to this day wish she could go
09:04
Visit his grave and let him know How wrong? Yeah, this guy who he didn't think was could go to college actually did okay did okay. Okay, so still I'm sorry. I'm gonna make you a senior. Okay, you're gonna have to face the ideas by then Somebody's figured out you're ready for college. Yes
09:21
How does the decision go to where you're gonna go? So I have Important that's part and part of my story is I have a high school sweetheart But we started dating and my junior year name. Her name was Linda Crandell at the time and so I applied to UCLA and get into UCLA as senior But now I'm it's kind of like my dad. I have to make this decision
09:42
I realized at the time that if I go off to college, you know The chances of Linda and I drifting apart are probably pretty high right and I asked my mother recently about this decision She said the other thing that influenced my decision is do I want to go to UCLA or do I want to go to the local? Community college besides my girlfriend was the wrestling team. So the wrestling team in El Camino College was great
10:05
In fact, it was even though it's only a two-year college. It was a better wrestling team than UCLA So she said the wrestling team influenced my decisions So some combination my girlfriend and the wrestling team got me to go to El Camino Community College
10:21
You went to El Camino first. I went to El Camino and rather than go went to UCLA Wow Okay. Well, I'm gonna have to Help you deal with that decision So you're there for two years two years. I'm on the wrestling team both years you know, we win the state championship my first year and you know, I'm and I'm not that good. But the second year I become Southern California champion
10:44
So for the wrestling that was the best decision. Yeah, and for the love that was yes, and then we Maintain, yeah The more important thing is my girlfriend who became my fiance and then we got married right after my junior year Now part of the reason for that from the Patterson history is my dad
11:02
Got his graduated from a community college and then went to Northwestern, but he never finished his bachelor's degree So my parents were first worried that we would follow in their footsteps Is that I wouldn't have any degree of any sort and that would limit my career opportunities They kind of for at the time I needed the permission my parents get married
11:22
If a man is under 21, woman could be a teen man 21 need that permission. So they you know They told us get your associate of arts degree before you get married just because you know, you never know what's gonna happen Right. So that's that's that Allowed us to get married after my sophomore year
11:41
I'm gonna presume to say something since we both came from the same area and I actually took some courses at El Camino But when I was in high school, I know it I'm gonna Suggest that two years were intellectually wasted if not morally and so forth Yeah, I did pretty well in the math class In fact, I remember my wrestling coach
12:02
you know a lot of the wrestlers weren't great scholars and He had this rule is you can't take Don't take you can't take math or you can't take advanced math and he didn't know I was taking calculus He found out later that I was doing it, but I remember being in the classroom and you know being able to work out some of the
12:21
Some of the equations look for the instructor on the blackboard. So it probably Probably, you know, the math didn't challenge me But you know the other classes were really good the philosophy and other classes were okay And the wrestling was a really great experience, you know As I say when I give talks is you know
12:41
I was blessed with both a high school coach and a wrestling coach who believed that we would be more successful if we Formed a team that if we can blend the personalities together and bond as a team, even though It's an individual sport. There's no tag team. Yeah, yeah that we would be more successful And so I think part of my skills is I'm good at leadership and forming teams and stuff like that
13:04
I absolutely got that from wrestling and plays well with others And I will talk about that at the fundamental stage of your career too because it's not true of everyone But UCLA decides you might have a talent enough to go to UCLA after finishing. Yeah Well, they let me back in after
13:21
They didn't they didn't send me back. They didn't send you back. Okay, so now tell me the stage of Education in computers or math or whatever it is you decide to specialize in by going to UCLA Well, yeah, I would say I thought of myself as a math major at El Camino I don't know if he actually had to pick a major
13:41
And then that's what I did at UCLA And I'm in my dream, you know, I'm married where I'm working at my dad's factory to support and my wife's working to Support ourselves as a college students and I'm a math major and then it Matt UCLA's was on the quarter system Still is in the quarter system. So I'm two quarters into my junior year and then for my spring quarter
14:07
The math class I want to take was canceled. And so I had to find a replacement So I just kind of looked in the catalog and oh There's a computer course. I Guess I'll take that and in fact as I look back it wasn't even a full
14:21
Semester course. It was only a half. It was two units and where most courses are for you So as I have courses up, I guess I'll take that so I take it course in that era this is 1966 right it's punch cards and Fortran, right but as amazingly as it's hard to explain
14:40
It's kind of like old video games back in the day They were very exciting to find how primitive they look same thing was true with programming here This was I could get these ideas to come alive kind of my brain would come alive and even today people talk about When they're talking about the program the anthropomorphism Then I'm then I test this thing and I do that and I these ideas come alive inside the computer and they work
15:04
Yes, you know, it's what Brooks. Koch thought stuff. It's great. So I was hooked. That was it I had found the thing that I wanted to do the math was getting more abstract But the computing was exciting and concrete. So this is before UCLA had an undergraduate degree. Yes
15:20
Yes, so but my senior year I took every computer class I could find the business school took taught some Engineering taught some I took every single course I could Then what happens to me? In my senior year, it's actually I think at the end of the winter quarter so I've got one quarter left of my undergraduate education and
15:45
At the end of a course that I did well in I just after the course was over I just casually mentioned to the person teaching it That boy, I'd sure rather do computer stuff to pay for college than to work in my dad's factory in downtown LA and I it was an innocent comment
16:02
Yes, but this young faculty member took it upon himself to go find me a job as undergraduate Let's get several several projects He found me a position to work at and one of the research projects And so I started working there and everybody else was a grad student and I really liked it And so I talked to my wife. Well, what do you think about me getting a master's degree?
16:21
Well, and I said it's you know master's degree is pretty efficient. It's only one year. She said sure Why don't you go once you go do that? And so I applied pretty late in the process to get get into grad school the master's degree Which you know because those worked as undergrads was able to get in but up until You know it had that professor not got me the job. I certainly wouldn't gone to grad
16:43
I think we need his name Remember what do you remember his name Jean-loubert? Okay there I have thanked him profusely over the years But he yeah, he it was right the kindness of his heart that he made that happen and it completely changed my life You might have sensed you had some talent. He was possible. It's possible. He thought I would be
17:04
Useful as a as undergraduate and would sponsor me as a grad student So there was such a thing as a master's degree in computer science at that point Yeah, kind of the graduate programs in computer science started popping up in the in the mid 60s, right? Just ready for you
17:21
Yeah The you know the first PhD in computer science is maybe Purdue or Stanford in 63 years for and so they started Quickly coming, you know all the probably use many UC campuses had PhD and master's programs But the undergraduate thing was gonna come later Now I'm assuming you had to do a master's thesis. Maybe yes, not comparable to a PhD, but it's a serious project
17:42
What was that project? I can't The Dissertation Yeah, I'm gonna have to look that one up Clearly you're in a process because at some point you're gonna go beyond a masters
18:01
Yes, so I should tell the rest of the story Yeah Is so I get so I go to graduate school and in my office is of four people the other three all getting PhDs So I go back and tell my wife after probably pretty early. So well, what do you think about me getting a PhD? And we did have nobody in my family go to the graduate school as far as I knew a PhD that's a doctorate
18:21
So it's like a medical doctor. It takes you four years So that's what I knew and she said well if you're smart enough get a PhD you should go and get it Now what had happened at home is you know as the were the old we were the first The oldest ones that are family and the first one to get married. So they told us to wait a while We had kids so we waited a whole year before we had kids which seemed like a long time for us
18:43
But by the time about the same time I graduated from colleges I became a father so I was going to graduate school with a wife in a infant and She was staying home taking care of the infant and so where was the money coming from? Well, we lived on a research assistant salary of 20 hours a week
19:03
which was Which was I was not much money, right? Very small. In fact, we qualified for Food stamps we qualified as if we were officially poor by by standards given how much money we had and and
19:20
wearing with school and then while I was a grad student my Second son was born about 20 months later. So we then became you know living on our race 20 Our week our a salary with a family of four But you must have known this was the right track Something in you is telling you this is not just an indulgence
19:42
You know, you know, I'm naturally optimistic We were living in UCLA's married student housing. We we were enjoying ourselves. There are lots of other young families there We couldn't afford to have a car We've made ends meet there was this Entrepreneurial person in UCLA's married student housing who had contacts in Hollywood. He said well, how would you like?
20:06
To have babysitters when you go out at night who are graduates Parents who are graduate students at UCLA be your babysitter. What would you think about that? So that was a pretty popular notion. So my wife and I would take turns kind of
20:20
tricking the Hollywood celebrities by riding our bicycles to their house because we didn't have a car and where they'd give us a whole dollar For fuel well, we would just pocket that dollar because we wrote our bikes there and we would babysit You know Hollywood celebrities, you know and that you know helped make Ends meet and there were places we went weren't that had
20:41
Oscars on the mantle there, right? We would ride our bikes and babysit and head back So we helped we would supplement our incomes that way then my parents lived near lived 30 miles away That's right. And they would you know, give you a hot meal everyone. Yes They would pick us up and take us into Westwood at UCLA
21:01
Take us to a movie and take us to the Mars maybe even babysit everyone's and maybe someone's don't yeah But we didn't have they can give us financial support But they give this kind of a motion sport and and one of my wife was pregnant They loaned us a card so that we can get her to the hospital No, you couldn't remember your master's thesis. I'm just guessing you can remember I can
21:23
Okay, so get me to the thesis that you remember. Well now the next part of the story is maybe three years into this RA ship the grant ends and There's not a follow-on grant for me to get on so I need to find a job My advisor
21:41
Helps, are you I guess he makes the connection with a local aerospace company Hughes Aircraft, which is located in Culver City, which is Six or seven miles from the UCLA campus, right? So I go there and get interviewed and I'd taken some hardware classes, but I thought of myself as a software person
22:00
Well, this job was to build Computers for airplanes so airborne computers So I got tested on my ability to understand hardware computer architecture and I passed that test But that became my practical education in computer design. It was building Computers for Hughes aircraft and that excites you or were you feeling?
22:22
this was a kind of byway until you got to what you really wanted to do I Would say I was just thrilled to have a job Because I wanted to finish, you know, I'd started the PhD I wanted to be able to finish They were willing to let me work part-time well, I finished my PhD and
22:41
And so that and we made a lot more money and working 20 hours a week at Hughes than being an RA so I think we're very happy to get the job and then Maybe it's kind of my natural optimistic nature or and there were great people to work with In there are other people who were working on their PhD in this in this group
23:01
So it's great people work with and we got to design computers, which is a very exciting thing There's that excitement of software if you got these ideas in your head and you see it come alive in the computer Another very exciting thing is a design hardware and see it built and work. That's on there I got an excitement from that as well. So To me it was well, that's how I would pay the bills
23:22
Well, I got my PhD and it was interesting, but I didn't actually say from now on I'm only building hardware I felt I could do both now. Oh, I Still want to find out your PhD thesis. So at that time the way computers were built was
23:40
They well the the hardest part of building a computer isn't designing arithmetic units and things like that It's the control the brains of the The data path is the brawn the brain is the control. So that's the heart thing control So it turns out a Turing Award winner Maurice Wilkes
24:00
It knew that he built kind of the first operational computer and he came up with this idea that he called micro programming is we could specify can We can specify control as a matrix and you could think of putting in the zeros and ones of control It's like programming and he called it micro programming. So in the
24:23
1970s most computers control was done as micro programming so in Including the computers were doing it Hughes aircraft And so the idea was well, why don't I make software to make it easier to write micro programs? Okay, so that was what my dissertation was about and how interested
24:43
Was your PhD advisor in this how interested was the community with did this seem an exciting idea My advisor my long my advisor who had stuck with me all this time was willing to supervise this area You know, I was supporting myself at Hughes It wasn't something that he was an expert in he was more kind of in programming languages and correctness
25:04
so my actual dissertation was building a language to write programs and a Verification system to prove that micro programs were correct, right which he was much more interested in verification side of things But he was he it was more incidental to his research rather than mainline to it. I
25:21
assume the Destination was well received and yeah, you you graduate with some broader recognition I eventually yeah after I know it's the greatest thing about being the field long time at the time It took me seven and I had a wife and two kids Seven and a half years to graduate right back then that was a long time right these days. It's kind of average right now
25:46
Comes to next. Okay, probably the next question is how did I get a job? How'd you get a job? Yeah So what happens is my wife was born up here in Northern California and lived her parents went to Berkeley High School And she lived here since she was 11 years old, but she got a great education
26:01
Berkeley High was amazing. Well, my my my parents-in-law did that right and so my wife grew was here She was 11 then her parents moved to Southern California where but she loved You know Berkeley area and always wanted to move back. So I'm now applying for jobs Like most people I'm applying for both industry and academic jobs and she keeps asking me
26:24
You know, what about Berkeley and I said, you know, I tried to she didn't understand the academics. Well, I said look Berkeley's up here UCLA's are here. It's really something graduating from here doesn't go to there That's just you just don't understand academics. But you know, I started getting job offers like from Bell Labs
26:41
yes, and nice places like that and I said And she said what about Berkeley and I said, oh, okay and call them up Ah, so she forced me as a graduate student a PhD student at UCLA to call the head of the department At UCLA of head of computer science, so I asked my application so she forced me to do it
27:05
So I still remember putting my finger in my ear put the phone up and hi, this is Dave You know Berkeley's the one place I hadn't heard of that if you were interested, you know I take it seriously, but you know, I'm getting into other offers I just so he was speaking to Elwin Berlekamp and of a world-famous
27:24
Person in games But kind of a mumbler so it's kind of hard to understand he says well Dave I've got your application here and you're in the top ten, but not the top five and I remember Oh, thanks very much professor Berlekamp and I was completely relieved
27:42
That was just I thought it was gonna be so much worse. My wife made me do it So I hang up I heard the story this side if he hangs up he finds out. He said that to anybody who called Really top ten about the top five is just a nice thing to say but he takes my application out and looks at it and There was a colleague that they wanted to build up the hardware piece of it and he was heading to Southern, California
28:05
Anyway, so he thought he dropped by and talked to me. He did we had a nice conversation I got invited up here to have an interview that went well a job But in terms of my kind of career preparation as I visited I said well I could either do software thing I could do software things or hardware things
28:21
But most people were just in software, but Berkeley was trying to build up their hardware architecture thing. That's simply what I could do that and So that was that led to the job offer for Berkeley so I'm gonna roll back just a minute not so much in your life as to an issue
28:42
Which is that decision point between academia and industry what goes into it? I mean you wanted a good job you wanted to support your family and so forth But were the implications profound as to whether you wound up in industry or academia next part of the story So we're we were the first ones in our family get married. But while I'm this longtime graduate student
29:05
Which was not welcome by my in-laws They referred to staying in the they referred to staying in school for a long time as pulling a dave So I thought you know and then even later they probably Dave so I thought you know my career turned out pretty well
29:21
but even 20 years later when my sister-in-law's Daughter marries. She marries a guy who got I think an MBA and then a law degree in she Referred to him while I was there as pulling a dave. That was a part of the family. Okay, but going back So our younger sisters
29:41
Get married They get cars and houses and we're still living in there and soon housing seems like there'll be a rule against that So it's clearly my wife and my son's had sacrificed a lot for me to be a graduate student And so it's time to make decision well, you know, I That was up to my wife, you know, she had sacrificed a lot
30:01
and so I said look you've got this job offer at Berkeley and we got the actual offer letter and the original letter was $13,500 and even if we account for inflation, that's like $50,000 a year. I Think actually what it was was that I got the offer in
30:20
May and I think what they do is give you a They won't you'll get a bonus once you actually finish your PhD So just in case that wasn't you needed more So they're gonna lowball you because they want to make sure you finish that PhD And I actually got maybe 10% more but the time is 13,500 and that's all I knew so I said look Linda
30:42
It's your decision. You've sacrificed a lot. So she asked me the questions well Dave if we Go to industry and get more money. Can we later go to Berkeley? No And I said, okay if we go to Berkeley and we want to change our minds. Can we later go to the industry?
31:00
Yes, so she thinks that ever says, okay, we'll be poor but proud That but that was the decision had she picked the other route I would have been disappointed but you know What else could I but what would the implications of that have been? Of course? That's the road not taken but you know, I've taken I go who went to industry, you know, the I actually
31:23
Followed up with the guys. I was at Hughes aircraft with kind of 20-ish years later 20. Let's see 20 25 years later they were and You know, they were Some of them were retired and I thought wow, they're much younger than me trying. They must done really well No, what was happening was the aerospace industry was collapsing and
31:44
Companies were acquiring each other and they were starting to lay people off. So I mean that would have been a Pretty if had I just stayed at Hughes aircraft all those years That would have been a not very exciting thing And in fact when I met them like 20 years later They were still talking about some of the same airplanes that I remember them talking about back then
32:04
That's what I'm wondering about. What whether because in the end you could shift quite a bit in the field Whether it would have been constrained by those circumstances. I mean, we don't know, you know Yeah, I have pursued the ideas in any case But yeah, I think the question would have been you know We were close to my family and all everybody was in Southern California what I've stated Hughes
32:24
To be near my family or would we have recognized the opportunity of Linda's desire to Move Northern California and gone work for Intel or something, right? Which would have been which would have been as intellectually Not just financially as intellectually exciting. Yeah, because because the aerospace industry
32:41
You know is a trailing technology right? Well for some things like missiles or something I'm sure they're the best in the world at but they they tend to be way behind everything else They didn't need to be at the right of the art. So Berkeley was the right choice shirt for lots of reasons. Yeah
33:00
Broadly, what's the that's a big question that I'm sure can I hope it can be answered in a brief way But the state of thinking about the computer at that point. I mean when you enter Berkeley as a junior faculty member the faculty member and The state of the field or at least the state of the field at Berkeley
33:21
Oh There's two things there Berkeley is In computer science is not one of the top departments of the top three departments are Stanford MIT and Carnegie Mellon or maybe the order is even Stanford Carnegie Mellon MIT at that time. They're heavily funded by DARPA and they're the big three
33:44
Berkeley is probably number four but a distance number four. Berkeley doesn't have any the DARPA funding and that's the big Distinguishing factor they had had some DARPA funding a few years before but that project kind of crashed and burned Literally apparently that is the person who was leading it, you know the computers they were building caught fire
34:06
There was there was earlier ones that had some all-star people in it, but they had Gone and started what they called Berkeley Computer Corporation a startup company that left Berkeley and all these Fantastic people that didn't do very well, but they became kind of the core of Xerox PARC
34:24
So this is Butler Lamson Chuck Thacker, two Turing Award winners. We're in that collection that used to be at Berkeley Did the startup company and then were the people who made Xerox PARC happen. Is that before your time?
34:40
Okay, so you're right so Berkeley's kind of lost that systems building piece then right and they want to build it back up and that's You know, it's you are part of that's part. I'm one of the people they wanted to build back up that that strength They used to have okay, so Get me to your insight But which begins the insight
35:02
Which which you did together with Hennessy were actually okay pointed away Well before that the guy who helped recruit me Or who interviewed me. Yes, and then you know recommended me for the interview was Al to Spain So when I came back, I liked the idea of having him a more senior mentor. He was an associate professor
35:20
Also, Carlos they can who had been at Bell Labs had just been visiting and had just joined so that we were two hires That the Spain was recruited from University of Utah just a couple of years before to try and build up this area and so Carlo and I now started off working together. Mm-hmm So he you know, he was kind of a kind of philosophical and we had these great ideas of what how?
35:46
Computers should be designed and did this project that he called x-tree. That was extraordinarily ambitious We were in design our own microprocessor our own operating system You know build a multi processor all these big ideas Sadly, we had absolutely no
36:02
resources No, no research funding or anything like that. So in surprisingly kind of three years into my career at Berkeley I'm contacted by Some people at digital equipment who was one of the leading
36:21
Computer manufacturers at that time. Yes in 1979 and asked me if I Person who used to be a professor at Carnegie Mellon Sam Fuller Who's now at Digital Equipment Corporation says I remember your dissertation on micro programming and verification. We're building these Media computers that's huge amounts of microcode and all kinds of bugs
36:41
How'd you like to come spend a sabbatical and we'll use your ideas in your dissertation in the real world? Very cool. So we do a sabbatical in Boston in the fall of 79 But it had this other consequence as I tried to describe the project We were doing on at Berkeley and nobody had heard of it and and and it gave me some time to think about well
37:06
You know, we're this is extraordinarily ambitious and we don't have any resources. Can you really do this at university? And basically I I Took that pause in my career to come back and think about what you can do in the universities What you can't do in universities
37:21
And came back and right after that I came back in January of 1980 That's when we started the risk project. So I ended up doing that on my own The ideas have been in the air a little bit I had done a graduate project and what was called high-level language computer architecture where the theory was What you should do is build really complicated
37:42
Computers that were much closer to the semantics of the languages and that was going to make better computers and make software easier And we had studied that material in the graduate courses and thought it didn't make a lot of sense We should you shouldn't be doing things in hardware issues compilers more and that and kind of the micro programming stuff was kind of a natural
38:04
Segway into reduced instruction set computers and then plus kind of my recognition that You know one of the opportunities you could have at a university is you could tie your research to courses Because courses had deadlines right and academia doesn't have deadlines, but courses have deadlines
38:21
So if I could put the research ideas associated with courses we could make Pretty rapid progress with deadlines and get students to volunteer by taking these courses to work on it. So That was you know, the sabbatical gave me the opportunity to think and have the insight Fortunately in my fourth year I started working on the risk stuff and did a series of four courses where the first one was investigate the ideas and then
38:47
To learn about how to design chips And then two courses where we actually built the first risk microprocessor so in a series of four Courses we built the first risk microprocessor now
39:00
It's not that everybody is leaping up and saying you're on the right track at this point. Nope So what happens is this so the spring I think it's the spring in spring semester So we're starting to investigate these ideas and starting to modify compilers and
39:21
Simulators and stuff and starting to get an idea that this might this is this might be an interesting idea So I write a paper that's called the case for the reduced instruction set computer and I write that during the summer and send it to a
39:44
Feedback and one of my former students Dave Ditzel really liked it and made a lot of comments I asked him if he wouldn't be a co-author but and Also as part of my nature for some reason I get feedback from other people about it I also send it to my friends at DEC and they were building, you know
40:00
This really and the reduced instruction set computer ideas if we think of Computing computers when software talks to hardware hardware has a vocabulary You have to speak you have to use the words of the computers vocabulary And as I said the philosophy at the time was that a very rich vocabulary Which this micro programming stuff made easier to do and we were arguing the reduced instruction set computer
40:26
That's the wrong way to go. Uh-huh with Morris law and things rapidly changing You should have a small vocabulary of simple words, which would make this micro programming even than necessary so totally against Common sense. Yes in the field. The conventional wisdom was kind of software is hard to build because computer
40:47
Vocabularies are too primitive and this is the the bane of all of our problems. And so When I wrote this paper, yes, my friends at DEC wrote Contrarian paper and so these two papers the case for the reduced
41:03
Instruction set computer and comments on the case for used instruction showed up in people's mailboxes at the same time and they were you know, they were Like this, right? And so this was really interesting controversy. Yeah different ideas and and VLSI
41:21
microprocessors microchips, where should things go and so this got tremendous attention and where the reduced instruction set computer side was You know a small minority so fortunately John Hennessy started working in this area. We were both funded by DARPA
41:41
And he started doing the design as well and fortunately for me despite Berkeley and Stanford being rivals Yes, as local competitors we Recognize that here's somebody We're on the same side of this argument I need as much help as I can get and so, you know
42:03
I welcome John's help on this area and he and I and people at IBM who were also in this area we would Form we were on the we were on the same team and we would participate in debates At comp at these conferences arguing the case for reduced instruction computers
42:20
And there's plenty of people on the other side argue other side But over I'd say the course of like 18 months there were three or four of these debates that John and I participated in and we kind of Switched the the intellectual argument around people could see okay. I understand Technically why this might be true then the next question is but is this commercially relevant, right? But it's like technically, okay
42:44
You guys were heretics before maybe maybe what you're saying makes some sense But you know, it's just an academic pipe dream. This will never have impact It's like changing like railroads have a certain gauge. You guys are proposing to change the gauge of the railroad. It'll never happen. So
43:03
The temperamental elements and both of you is stubbornness or knowing that you're right when everybody's to a lot of people are telling you You're wrong. What's why do you keep going on that? Yeah, I would say John first of all, John and I are a lot of we grew up in the opposite coasts, you know He's Catholic. We both are in large families
43:22
Went to you know, big public schools, but we're very much alike in our worldview As it was that's why we are such good co-authors collaborators together I think I'm up. I think we're both optimistic. I think we just believed our arguments are right, right? We were logical people. Here's the set of arguments if semiconductor technology is changing fast
43:45
it makes more sense to have a simple instruction set that can track the fast-changing technology and Compilers are the right thing to close that gap not not hardware because it'll take you longer to build the hardware It's really complicated. So we just thought these are the right arguments that matches
44:02
Microprocessor technology, so we were convinced of the correctness of our thoughts And we're willing to be convinced. Otherwise, which we participate in the debates, but nobody nobody Nobody Nobody changed our minds. Nobody came up the arguments except the fact that this will never fly commercially, right?
44:22
So John started a company MIPS, which is named after a Stanford project and I started consulting for Sun Microsystems Where they decided one of our former graduate students at Berkeley was one of the founders of MIPS. This was Bill Joy and even despite
44:41
Sun Microsystems being a pretty young company in the back of his mind He thought the risk ideas were the right things to do. He couldn't buy them. So he decided or convinced his the leadership that they should start their own risk project to build their own Microprocessors by being a small company because these are better ideas than buying microprocessors from Intel or Motorola
45:01
I'm very interested in talked about it with John as well about this Relationship now in a very different question from industry versus academia the extent to which industry can in your field Be a good testing ground that it carries through an idea in another form and
45:20
Helps you do the thinking about it or maybe just the convincing about it. Yeah, I think I think a lot of Certainly my reaction and I think a lot of students when they go to college and they take these really provocative Thoughtful courses is some of the humanities and social sciences field. It's like You'll never be able to figure this out. Right? That's okay. Here's two opposing ideas
45:44
How are you gonna decide who's right? And what's kind of nice about engineering and sciences is you can have these diametrically opposed ideas But there's experiments you can try and do just see which one wins so you can make progress for us in computer architecture that Test field is beside you building your prototypes is basically industry you you're
46:05
Your success is transferring your ideas into industry They're able to make better computers because that's how everybody gets to have a computer you buy them, right? It's not not gonna make them yourself. So our ultimate testing of our ideas is the marketplace, you know, which has
46:22
It's not a perfect testing point because they can have other reasons for success other than your technical ideas But if you're successful The technical ideas play an important role in art and can be the enabling role But that's kind of where we settle the architecture debates It often involves spending hundreds of millions or billions of dollars
46:42
But companies will make these bets build computers in this new way. And if they're right, it can be Successful for decades and if they're wrong, they could lose a lot of money There's nobody on the other side now on this point, right today for risk stuff Yeah, not today in terms of number of chips sold it's it's overwhelming
47:03
99% You know Intel kept pushing their style of instruction set which had heavily microcode It's still for sale today still dominates the cloud, but the cloud is a very small piece of the overall marketplace Is the rest of your career building on these?
47:20
Insights or do you take on other? Issues and challenges, so I think we ended up doing Kind of two generations of the risk processor through those courses and we did two more research projects that breast processors So I kind of as we look back retrospectively we call, you know, we can think of them as risk one to risk four at Berkeley
47:42
after that We just I think one of the things I didn't say is when I came here as a assistant professor And I got advice of what to do. They said look Berkeley doesn't count the number of papers you publish Berkeley wants impact and I love that. I just love that challenge impact
48:00
We want you to change the world and I don't know if that was actually whether they really back that up But that's what they told me. And so I love that we want to impact change the world So after the risk stuff which started having pickup my colleague Randy Katz Had a Macintosh and the first Macintosh of time as you remember did not have a hard disk they were floppy
48:23
Yes Which are terrible and then he got a it's the first personal computer a disk sitting next to his Macintosh He said, huh? That's interesting. I wonder what we could do with that. So Randy and I led an effort of Replacing disks in the big computers, which were the size of washing machines and we said why not replace them
48:42
With 40 of these little PC disks and then we kind of calculated since disks are rotating You could have tremendously more performance and better latency from 40 small disks one big one and then we Kind of wrote those ideas out and sent it to a colleague in industry and he said pretty interesting ideas in terms of performance
49:02
but by the way The chances of 40 things failing is basically 40 times higher that one of 40 would fail than one thing failing So, you know, that's a bad idea. So we said okay, we got to work on reliability So the idea was to take the 40 small ones and add some more so we'd have redundant ones
49:21
So in case one failed and if you do it, right actually, you know 44 of them can be more reliable than the one big one. And so the name of that project was redundant array of Inexpensive disks were the inexpensive big PC disks. That's called raid. Great. So that project was also very popular And you know really big impact lots of people everybody uses it today. We're
49:46
we're heading to the end of a too short conversation, but Everybody's gonna be angry at me if I don't talk to somebody who's been so invested in computer architecture About the present. I mean you have already said to in certain context that this is actually an exciting period
50:04
You guys didn't do it all all the questions have not been answered What's exciting about this period in computer architecture? Well my whole career we've had Moore's law since let's see since I graduated from high school There's been Moore's law and it's over and Moore's law was doubling number of transistors originally every year
50:23
He predicted than every two years and it's shocking that it's over and people even today even in my department Don't want to believe it's over but you know The data is clear it's over and that has driven in computer design for a long time people still want faster Lower power computers despite the end of Moore's law. So that involves what?
50:43
People like Hennessy and I do we have to come up with new designs of computers Even though the transistors aren't any better. So it puts more of the emphasis on computer architects. Ah That's why we think this is going to be a golden age of computer architecture which are article That's the title of our Turing word lecture, so again, you don't have to be a prophet if you don't want to but
51:07
Number of years from now 510. What what will be possible? Yeah. Well, we definitely I think To me, it's pretty clear that it's the computing is exciting at the two ends in the cloud and at the edge
51:23
That's where things are going to happen There's this new I think we're in the middle of another Transforming paradigm so in my career the microprocessor changed the world the Internet changed the world. I think machine learning is one of those and what's exciting about machine learning is
51:44
They need tremendously more compute power. Basically. The first word is machine for it to learn you need a lot of machines So here's a group that needs as much computing power as we can build right when Moore's law is ending, right? So what? Who architects believe is that it's hard to build faster general-purpose microprocessors
52:03
That's really hard to do maybe you know instead of doubling every 18 months like they did during the risk era Maybe they're going to double every 20 years But building something just for one narrow relatively narrow area like machine learning. There's huge opportunities there So there's tremendous excitement right now both at big companies and at startups in building
52:25
Dedicated computer architectures to make machine learning work and it's a very heady time Thank you very much. Okay, always nice to end it heady time
Empfehlungen
Serie mit 66 Medien