YT YouTube2Text

YouTube2Text → Transcripts → [CS61C FA20] Weekly Lecture 14.LIVE - WSC

[CS61C FA20] Weekly Lecture 14.LIVE - WSC — Transcript

by CS 61C Departmental · 10,458 words · 1,475 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:01computer and start webinar here we go
0:05and and get back to the first slide
0:06quick get back to the first
0:09slide click get back to the first
0:16slide okay
0:31welcome
0:32students hello hello hi
0:38everybody all
0:40right what a special day we have what a
0:43special guest we have today I'm so
0:44excited about this let's get started all
0:46right I think we have the numbers are 40
0:49people which is about right we usually
0:50have 30 to 40 about 50 wonderful welcome
0:53everybody to cs61c week 14 Warehouse
0:56scale computers wo great to have you all
1:00here great to have you all here uh let's
1:03share let's Jump Right In and talk about
1:05the
1:06agenda so uh our amazing guest is
1:09Professor David Patterson I believe this
1:11is our first touring Award winner Bora
1:13on our on our as our guest panel we've
1:14had some we have the DARPA lead we've
1:17had CTO of big companies it's been an
1:20amazing semester um but this is the
1:23first time we have a Tor W winner which
1:24is the highest it's Nobel Prize in
1:26Computing if you don't know uh so we're
1:27going to jump into David about 10
1:29seconds but the agenda for today is
1:30after talking today we're have a
1:31Computing the news as we all always have
1:33then midterm score update we have some
1:36midterm scores to release to you and we
1:38have some some news so that's exciting
1:40double do score a schedule and then ask
1:42us anything
1:44so boy it gives both Bor and myself and
1:47everybody on the staff the greatest of
1:48pleasure to invite Dave Patterson our
1:51colleague and friend who helped build
1:53this class by the way this isn't just a
1:54random turning Award winner this is a
1:55turning Award winner who has contributed
1:57significantly to DAV I think you're also
1:59won the Carl award I think you won like
2:01every award ACM has to give may be one
2:05of the most honored people in the entire
2:07department arguably the whole campus um
2:11Dave retired recently Professor Meritus
2:12now and he's now working as a
2:14distinguished engineer with Google some
2:15exciting things maybe he'll share some
2:17of the fun things he's working on with
2:18there he's the vice chair of the risk
2:20five uh Foundation um and he's known for
2:24partially co-invention inventing risk
2:27raid 61c the text books that are the
2:30most popular textbooks in the entire
2:32planet by the way in computer
2:33architecture there's no textbook that's
2:35more popular than Dave and and
2:38Hennessy's uh book in architecture so
2:40what a gift of his time uh to and by the
2:43way Dave was the first person I taught
2:45with in 61c he was the one who taught me
2:46what 621c was like so I owe him
2:49everything I am in 611c I owe to Dave so
2:51Dave thank you so much for the time and
2:52for joining us today what a delight oh
2:55thanks it's
2:56always great to try and catch Dan's
3:00enthusiasm because which is
3:02contagious uh I thought I'd start off
3:04about talk about five minutes tell you
3:06about RIS five and then we're just gonna
3:08open up for questions sure oh I can see
3:10the slides cool yeah oh perfect perfect
3:12and by the way you have more than that
3:13we have a lot of time we have I think
3:15I'm I'm looking B about 30 minutes so
3:17don't feel like have to rush through
3:18this don't rush through it I I'd be
3:21happy to do questions and answers either
3:23from you guys are from wonderful so uh
3:28the story goes back to 20
3:30and there's a research project right
3:32there in soda Hall where it was time for
3:36us to do computer architecture research
3:38in fact Bora was involved with this at
3:41that time and this was led by our
3:44colleague Christa sovich and he wanted
3:46to do computer architecture research and
3:49as you know you know the vocabulary of
3:51computers is the instruction set
3:53architecture well the two obvious
3:55choices were the x86 and R but it was
3:59imp possible for two reasons first they
4:01were very complicated commercial
4:04instruction sets thousands of
4:05instructions and even if we wanted to uh
4:10they protect those instruction sets
4:11because they own those instruction sets
4:13and they don't want to let anybody have
4:15it even us even though intel was one of
4:17our sponsors they wouldn't let us do
4:19that so we had to develop our own
4:20instruction set so christovich and two
4:24star grat students and your water
4:25against Lee decided okay we're just
4:27going to slap something together three
4:30months well it took them four years
4:31although along the way they had to they
4:34built some Hardware so it's it's a
4:37instruction set that's informed by
4:38actually trying to build it and it
4:40actually got better over time it
4:42actually got simpler over time
4:43surprisingly and the original idea was
4:46well as researchers we want to use it
4:48and lots of other researchers will have
4:50our problems so we'll make the
4:51instruction set which as you understand
4:53is just a specification right the
4:55instruction set is tells the hardware
4:58how it's supposed to work but then the
4:59hardware itself is a separate thing that
5:01implements that instruction set like
5:03ethernet and ether ethernet standard
5:05ethernet
5:06controllers but uh there wasn't a lot of
5:09uptake with other academic units the
5:11surprise was four years later industry
5:13started showing interest they really
5:15liked the idea of an open instruction
5:17set and once we started talking to them
5:19to understand why this industry
5:21interested in our kind of research
5:23instruction set then thought well that's
5:24a great idea to try and have an open
5:27instruction set and as an analogy is
5:29just like an open source operating
5:32system Linux is probably the most
5:35popular operating system
5:36today could we do something similar to
5:39like that for an instruction set
5:41architecture so Krista and I wrote a a
5:45technical report uh the instruction sets
5:48should be free a case for RIS five and
5:49I've learned recently uh because we just
5:52celebrated the 10th anniversary and
5:53talked to a lot of people the time
5:54actually a bunch of people read that
5:56paper and that got people decide okay
5:58we're going to try we're going to use
5:59use risk five because of the arguments
6:01in that paper okay next
6:03slide so let's go over risk five uh uh
6:08you've been using this in in the class
6:10right Dan y y y yeah yeah okay so what
6:12do I think is really important first is
6:15free and open anyone can use it I think
6:17that means since everybody many people
6:20are using the same instruction set
6:21around the world uh it's more
6:23competition and so I think it'll be more
6:26Innovative particularly for startup
6:28companies or small
6:30companies for proprietary instruction
6:33sets the first thing you have to do is
6:34sign a contract they say well okay I'll
6:36pay you this much money up front and
6:38then we'll give you a fraction of every
6:40chip that we every sell so you want to
6:42negotiate that contract carefully to to
6:45you want to be profitable so that takes
6:47a long time and it t besides money so
6:49what's nice about risk five if you're a
6:51small company or a startup company you
6:54can just say I'm going to use those five
6:55and I'll figure out who I'm going to get
6:57the processor with later because anyone
6:59can use it so that's a kind of a big
7:01business advantage independ of its
7:02technical qualities secondly I think of
7:06it as it's really elegant because of
7:08classes like 61c exactly 61c when you
7:12try and teach Assembly Language
7:14programming you don't want to show
7:16people the whole instruction set because
7:18it's you know can be hundreds or
7:20thousands of instructions and there's no
7:21real pedagogic value to that so what you
7:23do is you subset it you say here is the
7:26the essence of that instruction set risk
7:28five because it came came out of
7:29Academia for teaching as well as
7:31research that subset is official so
7:34those 4 instructions in RV is the whole
7:38is the whole simplest instruction set so
7:40all risk five software will execute in
7:43just those 47 instructions if you want
7:45so that makes it it's a new idea for
7:47instruction set architecture usually
7:49what happens is they start small and
7:51over time they get bigger and bigger and
7:53bigger kind of as you'll see as as you
7:56get into middle age your bodies will
7:58probably get bigger
8:00but the instruction set risk five is
8:03going to stay skinny for its whole life
8:05how does it accommodate other
8:07applications it has optional modular
8:09instructions like you've seen the
8:11floating Point instructions or the
8:13atomic instructions or the compact
8:15instructions those are optional if it
8:17matches your situation but you don't
8:19have to use them so that's a new idea
8:20for instructions that'ss in the past
8:23every time you added a set of
8:25instructions every implementation of
8:28that process for the next several
8:29decades had to also implement it for
8:31risk five it's optional so those are two
8:34a simple core and modular options are
8:36two uh big important changes because
8:39that can make it much more targeted for
8:42the
8:43application the third big idea which is
8:46kind of easy to do in 2010 when we were
8:48doing it is we knew it was going to be
8:51important to add accelerators to
8:53instruction sets we knew mors law was
8:55slowing down we knew that this change
8:57was happening in the technology so
8:59you're going to need accelerators so
9:01basically uh we just left lots of O code
9:04space in it now it's that was easy to do
9:07now but 20 years before that people
9:09didn't think that was important so they
9:10kind of filled up the op code space so
9:12it's hard to add more instructions it's
9:14easy and
9:16risf not surprisingly uh in 2010 we
9:19thought the future was at the edge and
9:22in the cloud so we want it to work in
9:24well in the cloud and at the edge so it
9:26has both a 32-bit and a 64-bit address
9:29shockingly because the biggest mistake
9:31you can make in computer architecture is
9:33run out address bits that's not going to
9:34happen r five because we also have a
9:37128bit address space version we're not
9:40probably going to need that for a few
9:41years but it's nice if because if
9:44instruction sets are successful they can
9:46last for 30 40 50 years we've already
9:49when it's time to go to 128bit addresses
9:52we'll be there and finally because it
9:55came out of academic roots and once The
9:58Graduate students graduate who worked on
10:00it and faculty get distracted then it
10:02could kind of fall apart because the
10:05faculty aren't paying attention anymore
10:06to that project like Linux we started a
10:10foundation it's called RIS five
10:11International and it owns that
10:13instruction set and is evolving that
10:15instruction set so that's why it's can
10:17last for decades even though academics
10:19can't you know keep uh that much
10:22attention and then finally in this kind
10:24of new world people are worried what's
10:27in their computer hardware they're
10:29worried that there's some trap door in
10:31there doing something nefarious they
10:34don't know about so by being an open
10:36instruction set several countries all
10:38over the world have had those concerns
10:40they're worried about getting technology
10:43from a competing comp uh country that
10:45might have you know be an adversary so
10:48you can design your own course and also
10:50because the open architecture allows
10:52open source implementations like written
10:55in verilog or system verog or something
10:57like that you can expect it you can know
10:59what's inside your computer so there
11:00aren't any secrets so these seven
11:03bullets are what distinguish RIS five
11:07and while there's a lot of enthusiasm
11:08for it okay next
11:10slide uh this is RIS five International
11:14has more than 600 members in 50
11:17countries it's a really worldwide
11:18phenomenon depending who you talk to
11:21they think oh this is our continent you
11:22talk to the Europeans they say oh no
11:24it's this five is primarily European or
11:26in North America well it's primary North
11:29America Asia oh no it's we're the we're
11:31the center right so they argue about it
11:33and there's 2,000 people participating
11:35in the various uh projects to expand the
11:38architecture in different modular ways
11:41so to try and convey this enthusiasm the
11:43next slide we'll see if that that is a
11:46video uh it's the 10th anniversary we
11:49were going to have a party that
11:50quarantine canceled that so we put a
11:52video together instead if you go to this
11:54website you can there's a variety of
11:57videos that we've done kind of to
11:59document history this is the three
12:01minute version that tries to capture the
12:03enthusiasm so if you guys could play
12:05that that'd be great y right for is
12:07gonna take over and do that here see if
12:11it comes through all
12:14right song
12:18beautifulity It louder what
12:24iion
12:26emach that's as loud as it goes okay
12:29that is so energizing that for those who
12:32haven't joined the community yet come on
12:34in it's a really really powerful place
12:36to be I'm actually quite impressed the
12:39kind of energy that is
12:43going so many
12:49parties the very lowest level have a
12:52common linga franka if there is a ling
12:55franka then we can start building
12:58standard computers computers are
13:03getting oops sound uh oh uh oh uh oh I I
13:07clicked the mute by accident I'm
13:10sorry
13:11me sorry my
13:14bad you can't build a society on
13:17proprietory formations the key
13:19differentiator forine from other
13:22processes is the
13:25openness the fact that you don't need to
13:28sign any agreement whatsoever to just
13:30start using it with this you can just
13:32download it you don't have to ask
13:34anyone's permission and that is a
13:37fantastic difference now that I'm doing
13:39architect use risk five and I get so
13:42many things for free I can focus on
13:44adding value on a few instructions and
13:46leverage tens of millions hundreds of
13:48millions of dollars
13:49invested over the last 10 years you go
13:52that much further by being able to
13:53leverage that infrastructure we had a
13:55workshop where we announced that we are
13:58going to RK in every GPU every chip that
14:01we have starting this year 95% of
14:05resources are put in Risk development
14:09when Western Digital announced that we
14:11were going to transition a billion cores
14:13I essentially tried to replicate that
14:16momentum that IBM's billion dollar
14:19investment did to Linux with risk five
14:21today risk five is a movement that can
14:25no longer be stopped the conversation
14:27started to change to to why not use risk
14:31five as opposed to the early years where
14:33it was like why the hell would I use
14:35something other than what I've always
14:37been using risk five is the Linux of the
14:40processor architecture world it's the
14:42transformation to an open source
14:44platform that allows many many more
14:46innovators and inventors to contribute
14:48to the advance of the technology I think
14:50of arm as a company that has paved the
14:53way for risk fire it was always destined
14:55to act as a place filler in the history
14:57of computing as a PR to this five there
15:00is an immense opportunity for RIS five
15:03to be in five years the leading Isa for
15:05computing I mean it's seems to be
15:08incredible but doesn't look so
15:11impossible
15:13now okay all right d do you want to take
15:17over the SL yep yep i' got I got here I
15:19got one last slide and then we perfect
15:21perfect here we go oops yoink there yeah
15:26so uh our this this idea that came up in
15:30the video that some of the people
15:32involved I really like the idea of a
15:34ling franka basically we showed you r
15:37five in this course and basically all
15:39the instruction sets are about the same
15:41it's not like one is gigantically better
15:44than the other so as was pointed out
15:46when we talked to the Pioneers was five
15:48if we just got one of them right we
15:50could just all Embrace that and we'd
15:52have a real linga
15:53franka for instruction sets and as a
15:56modu of uh from India pointed out if
15:58it's going to have a linga franka it
16:00needs to be open so we think that's our
16:03goal right and you know you by the time
16:05you guys if if you're freshman by the
16:07time you graduate you could see how
16:09close we are or maybe if you go to
16:11graduate school by the time you graduate
16:12by then it could be the most popular
16:14instruction set it's certainly capable
16:17of going from the smallest to largest
16:19things there's no technical thing invol
16:21Neal challenge but you know there's a
16:22lots of business issues to PR happening
16:24but this would be very exciting if uh uh
16:27instructions set the came out of
16:29Berkeley in part to help teach 61c
16:32became the instruction set that
16:34everybody in the world
16:36used okay that I'm ready for questions
16:41and we have a couple queued up already
16:42Bor you want to
16:43grab you know Dave I mean half of our
16:47labs are now risk five and projects in
16:4961 year are risk five but you know take
16:52a look in about a year year and a half I
16:54think all four will be when we have more
16:56risk five Hardware available and
16:57affordable I think we can uh we can
16:59switch the whole course and when the
17:02hive I mean I mean when the hive is all
17:03risk five I I can't wait till that day
17:06uh that would be wonderful well there's
17:08a project that I'm heading up that's
17:11called the Rios lab it stands for RIS
17:14five Inter National open source lab so
17:16there's a joint effort between chingua
17:18University and Berkeley that's located
17:21in shenen that's been around for This
17:23research institute's been there for six
17:25or seven years and we're starting a new
17:27project there called the lab and the
17:29first thing we're going to do is build
17:31what we call Pico Rio so this Rios is
17:35Spanish for rivers and so we think of
17:38the Rios laab as collecting resources
17:41from all over and creating this powerful
17:42force like a river and Pico Rio or
17:45Little River or tiny river is going to
17:48be a raspberry pight board that's uh
17:51based on rist five so it'll be trying to
17:53be open source as possible but instead
17:55of using the arm processor it'll do RIS
17:57five and our goal goal is to make that a
17:59lot less than $100 uh so we're going to
18:01do that for the first half of 2021 we
18:04hope and so that we hope to supply a a
18:07low something that students could use
18:09that has runs risk five and do software
18:12development so okay all right we got a
18:16few questions from the students ready so
18:18the first one this is an interesting one
18:20this was the same question was asked um
18:23when Andrew Waterman
18:25and were guests over here right
18:29answer to this one what aspects of the
18:31currently Frozen risk 5 Isa if any do
18:35you disagree with or would change if you
18:38could
18:41me so I I I mean Christa asanovic is
18:45just you know a real you know he's kind
18:48of a mile wide and a mile deep he knows
18:50a lot of things and he knows them really
18:52depth and he really likes to micro
18:54optimize so that explains the uh the
18:58funny swirled address bits it actually
19:01it actually does save some Hardware
19:04answer right right there's a bit and
19:08then there's a bit and then it saves
19:10some Hardware in the very lowest end uh
19:14designs but for me in because of you
19:17know the cost is having to explain it I
19:19think I would I would have just done
19:20straight address bits uh and which would
19:23have hurt the lowend designs a little
19:25bit but it would have been a lot easier
19:27to explain I think think that
19:30is that is the one that stands out to
19:34me yeah I think both Andrew and you are
19:36blaming it on K no I think that's
19:39brilliant I think I think Andrew I mean
19:42it's kind of like I've known Chris since
19:44he was a you know he took classes I'm
19:47old enough that he was a grad student
19:49here I was also Dave's grad student yeah
19:52and uh but he he just really he kind of
19:55enjoys puzzle solving so for him to kind
19:59of play with the address sh with the op
20:01code bits and I could align these four
20:02instructions with those four
20:03instructions and that might save a
20:05couple of gates here and there so it's
20:07like a puzzle solving like people enjoy
20:09do he likes doing stuff like that so for
20:11him it was kind of fun to play with muck
20:14with the op codes to get that little
20:16optimization so it was part of his
20:18personality so yeah Christa did
20:23that no nobody else would have taken the
20:26time to do that that I know I like doing
20:28that too I have to admit it okay well
20:30Bora
20:31Bora so it's like carom Maps or
20:34something like that right right we
20:37haven't we haven't covered carom maps in
20:39this class we leave it for later okay
20:41here is another question this one is uh
20:44kind of you know we we answer that but
20:46maybe Dave can say why do computer
20:48manufacturers not adopt risk 5 now given
20:51its advantages over other isas such as
20:54x86 yeah so um you know there's a first
20:58of all Hardware is not like software
21:00where you know you do a startup company
21:02and three months later you got a product
21:04right you have to design it then you
21:06have to manufacture it and it takes a
21:08couple of years I've been work you know
21:10I've known industry some but I've been
21:12at Google for four years so two or three
21:14years to start a project before a
21:16product comes out and actually deploy it
21:18when when it's a real product it's got
21:19to work that's typical so partly it's
21:23happening and we can't see it the other
21:25thing is that you guys don't know this
21:27but Intel had this problem how do people
21:29know whether this x6s and things so they
21:31did an advertising campaign on
21:33television called Intel Inside just to
21:36let you know right here's this thing
21:38it's got It's called an IBM PC but
21:41actually it's got an Intel par is there
21:44and they did advertising for that so
21:46we're not gonna be able to do that with
21:47risk five so there's GNA be a lot of
21:49stuff that you use that has risk five
21:52inside so for example Nvidia gpus have
21:55RIS five inside Western Digital drives
21:59risk five inside Samsung cell phones
22:01risk five inside so there's a whole
22:03bunch of them that we know have risk
22:05five in them but it's going to be hard
22:07for for people to notice it and we
22:10talked and and and there's no budget
22:12right there's not a there's not I mean
22:13maybe the foundation might decide to do
22:15a similar kind of risk five inside thing
22:17but it doesn't Pockets ask for funds
22:20from companies spend advertising right
22:23no not happening yeah not not happening
22:25yeah don't wait for it exactly so it
22:27takes a couple years to get out out
22:28there and that's why I wanted to show
22:30you that video what Stefan said it's
22:32gone from why should I take any risk
22:35with this instructions to like wait a
22:37minute this is working right why
22:40shouldn't I use it so that that's going
22:42on the places where it'll be the LA most
22:45difficult is ones now that you guys
22:47understand this that require binary
22:49compatibility either with arm or x86 so
22:52where does that matter IB PCS the IBN PC
22:56binary compatibility now on the other
22:58hand apple right abandon the x86
23:01instruction set like a month ago right
23:03yeah change instruction set that's hard
23:05for most companies to do uh the other
23:08one would be
23:10um uh would be the uh uh the smartphones
23:15where you Distributing applications in
23:17binary in the arm instruction set so
23:19those two will be tough but going
23:21forward if you're going to do something
23:23new and exciting that's better than a
23:25cell phone uh then you know uh you can
23:28pick what instruction set you want be
23:30because kind of binary compatibility is
23:33not as important today as it was in the
23:35past and so that lets us innovate in
23:38construction C architecture and then
23:39Here Comes this new open one that is uh
23:43that's has all kinds of potential
23:45benefits and so I think we're going to
23:46see like I said in the next five to 10
23:48years I I think it's going to get very
23:50popular and you know you'll be able to
23:52see check
23:54that all right here's another one
23:56somewhat related to that uh how how does
23:58RIS 5 plan to supplement uh legac
24:01extremely Legacy systems like Banking
24:04and medical computers well that's a
24:06really hard one like the I IBM system
24:09360 which dates back to
24:121964 IBM bet the company on this idea
24:15that you could have binary compatibility
24:17across the whole family of machines and
24:19they won that bet those so-called
24:21mainframes this was before seven years
24:24before the first
24:25microprocessor that instruction set is
24:27still alive well and doing 10 billion
24:29dollar a year business wow so you know
24:33it kind of you know it's interesting Can
24:35you capture the existing applications
24:38but it's more exciting from my
24:39perspective to get the new applications
24:42the the things that haven't been built
24:44yet if those are going to be in Risk
24:45five and if you know those older
24:47instruction sets have Legacy code that
24:48can last for a long time that's fine the
24:51other possibility that people talk about
24:53is what's called binary translation
24:55where you actually treat the instruction
24:58set
24:58as if it's a programming language and
25:01compile it into another instruction set
25:03that's a topic that people have worked
25:05on over the years you know that might
25:07work but for me it's like you know if
25:10we're going to do 20 billion Internet of
25:14Things devices or 100 billion internet
25:16things devices it'd be more exciting to
25:18see these new apps using risk 5 than it
25:20is to you know kill the IBM Mainframe
25:23Market uh that that's L for me right
25:28all right here's another question this
25:30is a good one we haven't thought of that
25:33one yet for the modular extension
25:36instruction
25:37sets is there modular Hardware to match
25:41I'm wondering how that General
25:42connection works well I think that's you
25:45know one of the exciting things about
25:48this transition the open instruction set
25:50architecture allowing open source
25:52instruction sets and then this slowing
25:54down of Mo's law meaning that we you
25:56need to start building accelerators to
25:58improve performance and performance per
26:01watt is I would say the module the the
26:04the the hardware equivalent of
26:06modularity will be these so-called IP
26:09blocks so what I've been doing at Google
26:11is building accelerators I've been
26:14helping build accelerators for machine
26:16learning so what we need is like a
26:18standard processor that runs like user
26:21interfaces and uh you know compilers and
26:24operating and system stuff like that
26:25talk over networking and these acceler
26:28ators that make machine learning really
26:30shine and that's the accelerators I've
26:32been building so I would say that's the
26:34modular Hardware equivalent of it it'll
26:37probably be blocks written in system
26:39verog or something like that that you
26:41can buy from somebody else and integrate
26:43on your chip uh would be the modularity
26:47there there's another idea that's
26:49catching on right now with the endings
26:50of mois law is Mor's law basically when
26:54he made his prediction by the way I'm
26:56sure you said Gordon moris alum Berkeley
26:58right said that okay yeah yeah yeah
27:00that's to a
27:02slide is uh is that he was betting that
27:07every year or in later revises every two
27:09years twice as many transitions per chip
27:11that basically was going to happen more
27:13transitions per area and the chips would
27:15get bigger and they just got bigger and
27:16bigger and bigger over the
27:19decades a new idea now that Mor's law
27:22sty ritten out is what called multi-
27:24chip modules instead of having one big
27:26chip it's Lots little chips that are
27:29assembled on some kind of substrate so
27:31another version of a modular Hardware
27:33idea would be this multi-chip version
27:36where you might have instead of Ip
27:38blocks you integrated a one chip like
27:40you know typically what we've done in
27:42the past maybe you'll have actually
27:44little chips that you're integrating
27:46together in some substrate that's really
27:48modular so you could have a floating
27:50Point piece a chip and you put it in the
27:53multi- chip things or you leave out the
27:54floating Point chip so that's that's a
27:57possibility of what could happen but
27:59basically it's a very exciting time in
28:01computer design because all the the way
28:03things have been designed for decades
28:05doesn't work anymore and so people are
28:07trying a lot of innovative ideas um and
28:11in instruction sets the way we build
28:12Hardware all kinds of stuff by the way
28:15this latest idea that U they they've
28:18shared about trying to build modular
28:20stuff out of chiplets is um you know
28:22based on risk 5 is Andreas solon's idea
28:25he just said he has a new startup uh he
28:28was
28:30video he must have got it for
28:34me all right here's another question
28:37this is a good one are there Vector
28:39instructions for risk five or are there
28:42plans to implement Vector instructions
28:43like AVX in Intel just for a background
28:46our current project is using Vector
28:48instructions and intrinsics uh uh to to
28:52to do the Paralis yeah so one of the
28:54jokes is about risk five is that uh
28:58kind of what caught on is simd
29:00architectures and there's a bunch of us
29:02at Berkeley who think those are really
29:03stupid designs in fact uh Andrew WAM and
29:07I wrote a a paper called simd
29:10instructions considered
29:12harmful but the contrast is these Vector
29:15architectures which are very elegant so
29:17a kind of one of the jokes is it's
29:19called risk V we called it RIS five it's
29:22called V because we wanted to have a
29:23vector instruction set so christovich
29:26and Andrew waterb have work very hard
29:29with a bunch of people in the gis five
29:30Foundation to finalize a vector
29:32instruction set so it's you know it's
29:35almost done people are starting to use
29:37it to build it so yeah there's a one of
29:39the modular extensions is a real Vector
29:42instruction set uh which many of us are
29:44excited about looking forward to how how
29:46why are they proposing it's up to you
29:49you can it's you know it's built in
29:51that's the Elegance of vector
29:53architectures is you get to adjust that
29:56but I think the people are talking about
29:58the some of the first implementations
30:00are 256 bits per clock cycle or 512 bits
30:03per clock cycle I think people are in
30:05that range right now but you know that
30:08has long legs it could be whatever
30:11people want to build that's exciting
30:13yeah it's it's F you know it'll be
30:15another year or two until we get
30:17hardware and then you know Hive will be
30:20replaced and
30:23um um here's another good one lots of
30:27this ation from projects here how
30:29similar is the way people design CPUs to
30:32our logisim project and they remember
30:35logisim it got better yeah I
30:38remember how similar
30:41um well people do design it in a harbor
30:44description language they they
30:45absolutely write it out that's logic
30:48slim is a is a harbor description
30:49language right yeah I mean it is a
30:52graphical
30:54interpretational of of uh I maybe I can
30:57answer this one why don't you answer it
30:59actually this year we built a new back
31:02end for logisim logisim now exports
31:05verog which is a standard language that
31:07people use for building
31:09CPUs and you guys don't know but we when
31:12we do when did these project
31:14competitions we actually went and
31:16synthesized your CPUs we didn't tell you
31:18that we did that but you actually there
31:20does exist alive you know you submitted
31:23a loges in file but in the back end we
31:25actually had the hardware description
31:27real Hardware description of your CPU
31:29and so we were able to comp compare them
31:32uh this is so the the embarrassing thing
31:35is verog is a kind of a terrible
31:38Hardware description language but it is
31:40popular uh it's inelegant so that's um
31:44there's a lot of reasons not to write in
31:45ver log uh there's a variety of you know
31:48even in Industry higher levels to
31:51generate it's verog is kind of like
31:52almost writing Assembly Language uh way
31:55of Designing it and so a lot of tools
31:57use it but if you wanted to get an idea
31:59of the low you kind of the worst case
32:01scenario of what Hardware design
32:02descriptions look like you look at the
32:04output of of the
32:06logisim to see it but I think if you've
32:09done logisim and then you start you
32:11could uh what if you take a follow on
32:13course I'm sure you could pick up
32:15various Hardware description languages
32:17and understand it yeah these questions
32:20keep flowing um at some point great no
32:23no we're doing we're doing fine we have
32:25we're fine okay uh and they're good I
32:29mean you know they're they're
32:31demonstrating yeah right that's
32:36right crickets crickets on their chat on
32:39their chat yeah they keep going um here
32:42is a
32:44um here's a you know I'll leave the the
32:46last one a second part of this one for
32:48for for the end um first from Riata
32:52while technical question would you or
32:55the risk five Foundation consider making
32:57instruction for Quantum Computing as it
33:00is getting developed so uh because I
33:03give this talk about you know uh things
33:06are slowing down and we need Innovation
33:09I get to ask a lot about Quantum
33:10Computing so Quantum Computing is a is a
33:13futuristic thing there's a couple of
33:15national reports that talk about that
33:18maybe it'll uh have applications that'll
33:21work in 2030 and keep in mind that
33:23Quantum Computing is is Big Computing
33:26it's big science right near zero degrees
33:29Carolin trucks going by make it not work
33:32so it it will be like in a data center
33:35if it does work so it's a really
33:37interesting topic you know when I
33:39started graduate school there were a
33:40bunch of people interested in nuclear
33:42fusion they thought I'm I'm going to
33:44spend my career on nuclear fusion
33:46nuclear fusion still doesn't work if it
33:48ever works it's going to be really
33:49fantastic uh we know it works we look up
33:52the sun yep that that's nuclear fusion
33:54but we haven't figured out how to put
33:56less energy in the get more energy out
33:58than we put in for nuclear fusion so
34:00Quantum confusion is very exciting I
34:02don't think it's going to be practical
34:04this decade so it's a 2030 thing but you
34:07know if you're excited about it you know
34:09that's really great so but I don't think
34:11it it kind of doesn't make sense to have
34:13a Quantum Computing
34:14instruction that's not the way Quantum
34:17Computing works but there are people
34:19here you could talk to if you're excited
34:20about Quantum Computing I'm sure there's
34:22many projects uh they you can learn more
34:24about it y here's another one and and a
34:28few related ones I mean this is you know
34:31there is a train of uh you know you you
34:33mentioned this so uh here are the
34:35questions related to that Apple's new M1
34:38chip is Arm based do you see the
34:41possibility of something analogous based
34:43to risk five and how why see something
34:46what say again analogous based on risk
34:49five no I would say um this happened
34:53some in the risk 5 Community where
34:55somebody would had a processor design
34:57that
34:58I said kind of these risk architectures
34:59are all pretty similar they had an
35:01existing design and they learned about
35:03risk 5 and they like the idea of this
35:05open this and this community around it
35:07and so they just changed the you know
35:09the first stage of the pipeline the
35:11instruction well after they you fetch
35:13the instruction you decode it the second
35:14stage of the pipeline that decoding they
35:16changed the decoder and they could
35:17execute RIS five so if they can that M1
35:21design that's based on the arm
35:23instruction set they could pretty easily
35:25switch it over to RIS five so I think
35:28whenever arm is successful in building
35:30something that's faster and lower power
35:33to me that's a great sign because I'm
35:35sure you could also do it in in Risk
35:37five there's not there's not like I said
35:39the instruction sets are more all more
35:41or less the same x86 is very different
35:43but the risk architectures loads in
35:46stores you know 32
35:48registers you uh similar encodings so
35:51they're kind of all the same and you
35:54know that's partly I think for Hennessy
35:56and I we got our touring award in part
35:59for our contributions to risk and in
36:00part for that textbook um so we're kind
36:03of surprised that 35 years later the
36:06best ideas are still risk you know 35
36:09years of mors law you would who know
36:11what would have happened right we're
36:12kind of surprised but since that seems
36:14to be our best ideas there uh any time
36:17something Goods happens arm we think
36:19that's good for risk five and assuming
36:21that people some PE a significant number
36:24of people will like the open model as
36:26opposed to the proprietary model
36:29and so here is another you know this is
36:32another good question and you know there
36:34is another person who I'm thinking of
36:35who could be uh who could answer this
36:38one of your former students can binary
36:40translation be done in
36:42Hardware yeah so yes as as I think you
36:45know from the class Intel to its credit
36:48once the risk ideas went out didn't give
36:50up they put a 500 person engineering
36:53team together to build the penny and pro
36:55as called that came out in I think 97
36:59and what they did amazing enough is
37:00translate from x86 into risk
37:03instructions on the fly in Hardware so
37:05kind of a hardware compiler now at that
37:07time Intel's process technology was a
37:10couple of years ahead of everybody else
37:12so they had extra stuff on the chip but
37:14their technology was years ahead so they
37:16could they could hide that cost so
37:19there's an example of somebody doing
37:21Hardware translation on the fly now
37:25there may be some in one of these
37:26discussions by talked about it there
37:29somebody claimed there were some legal
37:30issues about
37:32translating uh binary instructions and
37:34executing them and so I don't that like
37:38patents whe whether you'd be violating
37:39some patent if you did that I don't
37:42actually understand that but I have
37:44heard that claim so before you start
37:46your company you should talk to some
37:48lawyer and see if there's some truth to
37:50that uh rumor that there may be um you
37:54know you may be violating some kind of
37:55patent if you were to do a hardware
37:58translation yep um I I actually know of
38:01a patent I think that that claims that
38:05um yeah I we are getting so uh two two
38:09questions here um one you know I'm going
38:13to rephrase some of these because we
38:14have still quite a few and we are
38:16running long time and I'm going to save
38:18Riot as a second part of our question
38:19for last the one so what are the plans
38:22for these for your books uh I mean I I
38:25actually you know my age well I'll
38:27reveal my age when I tell that I
38:30actually own a first edition of the
38:33book that's how I learn you're you're
38:35really
38:36old you're really old if you if you
38:40bought a
38:44book yeah I actually own the first the
38:46second edition then like I skipped the
38:48third maybe yeah that it wasn't very
38:51good
38:52okay what's my what's your question how
38:54many total copies do do you keep track
38:56of how many total we asked once and I
38:59think uh the undergraduate's more
39:01popular I think the graduate book was I
39:04think it was a quarter million to a half
39:06a million something like that and the
39:08undergraduate one is they certainly sell
39:10more there's more undergraduates than
39:11there are grad students over this is
39:13over you know 30 years right so it's a
39:16number like
39:18that so yeah hundreds of thousands yeah
39:22wow um and um are there plans for a new
39:28oh absolutely uh this we've we uh the
39:33undergraduate book uh I just turned in
39:36the G for the second edition for the RIS
39:39five version I talked to christovich
39:42about it asked him based on teaching
39:4461c what what did he recommend he
39:47recommended going back to 32- bit
39:49addressing because you know it's it's
39:51when you're first learning your
39:52instruction set there's that confusion
39:54about 64 and 32 and there's a few
39:56instructions there
39:57so we did that uh I also don't tell them
40:00about the swirled address bits until
40:03chapter four I just I don't explain that
40:06in chapter four we explain what
40:07multiplexers are and I have a little
40:09table there that shows exactly how many
40:12three to one Ms became two to one M
40:14because of the SR address bits and I
40:17there may there may be a sarcastic tone
40:19to that
40:20section
40:22but uh and we emphasized more about uh
40:27about fundamentally the bits bits can
40:29mean anything you know that that it's
40:31kind of a strange concept that data can
40:33be anything so I try and do more kind of
40:36hit that one harder because people are
40:38surprised about that so these are a lot
40:40of suggestions from uh from Christa
40:43second edition so I just turned in the
40:44galley so it'll be for sale maybe in in
40:48March or the next
40:51Edition all right next fall definitely
40:54yeah and here is one final question uh
40:57what are a few words of advice you will
41:00give to folks who are just beginning a
41:01college CS career and maybe we can uh
41:05expand that or narrow it down a bit to
41:07people who are just wrapping x601
41:10C yes uh so I I as Dan and Bora knows I
41:16love giving advice so you you're gonna
41:18have to cut me off probably I would
41:21say you know I'm sorry that you're going
41:24to college in this in this quarantine
41:26time uh I've been around a long time
41:29this has been the strangest year in my
41:31lifetime uh so uh you guys uh survive
41:34that I I I think you picked a great
41:37major if you're you're in this to do
41:40computer science or ECS you know I this
41:42is going to be a really exciting Century
41:45just like the last century was really
41:47exciting when I there was no
41:49undergraduate degree in computer science
41:51when I went to school uh that happened a
41:53couple years later and I kind of did it
41:54myself I stumbled on this but Bo you
41:57know my uh when I did it my relatives
41:59said well okay good luck in Computing
42:02and then over the decades wow that was
42:04really smart decision you made I did it
42:07because I loved it not because I thought
42:08it had a tremendous future but it it
42:10worked out really well and I don't see
42:12anything slowing down in this decade um
42:15you know machines starting to get more
42:17intelligence more applications it just
42:20seems like this is going to be a very
42:21exciting field I the part of Google I
42:25workout is called Google brain and
42:26that's one of the big proponents of the
42:29machine learning the Deep networking I
42:31would definitely take statistics courses
42:34while you're there that's you know
42:36computer science and statistics is the
42:38foundation of this uh machine learning
42:41it's hard for us to know how how big a
42:43revolution this is how far it's going to
42:45go it's already gone pretty far uh it
42:49could change how we program instead of
42:51writing millions of lines of code
42:54already there's some examples where we
42:56in ER those lines of code basically from
43:00algorithms that look at the data so I
43:02think it's going to be a pretty farful
43:04Force you should think about going to
43:06grad school right uh life expectancies
43:10are going up uh I I retied and one of
43:13the best things about retiring was going
43:15to a retirement party and she said the
43:17number one
43:18correlation with longevity is education
43:22and so people in this crowd are going
43:23live a long time and I had been thinking
43:26based looking back WS that boy I'd be
43:28lucky if I get into my 80s and talking
43:30to I hear I added a decade to my
43:32expectations I expect to get to my 90s
43:34mid 90s you guys are a lot of you will
43:38live to a hundred being get to a 100
43:40with advances in medicine and taking
43:42care of yourself so that's a long time
43:44so thinking when you're 21 boy that's it
43:47I can't afford to go to school anymore
43:49you're gonna you're gonna have really
43:51long careers I would seriously consider
43:53graduate school the last degree you can
43:55buy no one's got PhD is it there's not
43:58like PhD plus right and spending five or
44:03six years getting acquainted to all
44:06kinds of the technologies that are going
44:07on people still willing to teach that to
44:10you and you can it's a ripe environment
44:12to learn and you get trained how to do
44:14research on your own give talks on your
44:16own part of the reason I ended up at
44:18Google is I wanted to do a sabatical
44:20there because I didn't understand why
44:21they're hiring all the phds and more or
44:23less they said look at the training you
44:25get with a PhD you can tackle Problems
44:27by yourself you know how to solve them
44:29you learn how to communicate those are
44:31valuable skills so they hire a lot of
44:33phds even though they don't have a big
44:34research brers I think going forward
44:36given life
44:38expectancies uh phds more there's going
44:40to be more opportunities for people with
44:42PhD so you should seriously think about
44:44grad school at least a master's degree
44:47you'll have plenty of time to make money
44:49after grad
44:51school don't worry about
44:53that all right great advice wonderful
44:56all right thank you Dave thank you so
44:59much Dave it's great to see Patterson
45:03everybody hey I can see you still doing
45:05you're still doing competing in the news
45:07I think I started that you did
45:09start thank you for that right wonderful
45:13all right take care Dave beautiful all
45:14right let me share my screen and get it
45:18here that was so what a delight Bora
45:20what a delight my goodness all right so
45:24I think it's this window see if that's
45:26going to work
45:28okay let me try it again okay here we go
45:30yoink we got some news I'm to share this
45:32the midterm scores so excited about that
45:34one so first start first we'll start
45:36with the computer in the new yeah yeah
45:37exactly this is a good one yeah 75 so
45:39here we go Tesla Model X gets hacked
45:41through a new relay attack Tesa said is
45:43pushing a patch through software updates
45:45by the way this is one of the benefits
45:47of having a software you know an over
45:48theair software update uh normally you
45:51have if you have a there's a software
45:52failure you have to like queue up a time
45:54well I'm Subaru on BMW got a que up a
45:56time to get back to the to the
45:58dealership this is wonderful you can
45:59update this there but I I also
46:01understand I was reading the article um
46:03that talks about there a two-step hack
46:05one of this is the there's a key fob did
46:08you read it chance to read this I was
46:09reading this while you were talking to
46:10Dave um there's a key fob that's used in
46:12the in the Tesla and they were able to
46:14basically spoof the key fob to open the
46:16door and then the catches you can't just
46:19drive away you can certainly with the
46:20key fob I think roll the windows down
46:22open the trunk and open the door but you
46:24can't drive away they say however what
46:26they want there inside the vehicle
46:28there's like a little Port you know many
46:29cars have the little Port where you can
46:31control you know plug a laptop in with a
46:33Serial port and talk to it within like a
46:36couple of minutes he plugs into the port
46:37and he overrides he says yep okay you
46:39can start driving drives away and
46:41there's a whole video on this on this
46:42link so uh certainly Tesla cares about
46:45security a great deal and they're going
46:47to be trying to push out a fix very very
46:48soon fix this this does show a few
46:51really important things that uh again
46:53somehow we got this back into risk five
46:55court right uh you know the the openness
46:58and understanding where you know
47:00building insecurity in the very root of
47:02our systems is going to be really really
47:04important going forward because you know
47:07yeah you know there is a great comment
47:10over here I liked it when the most
47:11advanced thing in my car was the
47:15radio cars cars Bor I used to
47:19own oh I didn't even have a
47:22radio
47:24car but the point here is
47:27yeah this is
47:29uh the big thing here is we like the
47:31convenience and we will love to talk to
47:33our cars you know coming you know in a
47:35few years we'll be talking a lot to a
47:37lot more to our cars than we than we are
47:40presently and there will be a computer
47:42and that computer will need or that set
47:44of computers will need to be very secure
47:46um that is there is a big debate over
47:49here how open this thing has to be and
47:51risk five is making a really big push
47:54right that uh um
47:58that uh things should be completely
48:00opened down to the last single bit and
48:03then it's a lot harder um to to to you
48:07know if there is an attack then it's
48:09very easy to trace it down and
48:11understand where it came from and I mean
48:12the topic about here is cars obviously
48:14we're just trusting that Tesla does the
48:15right thing right we putting our lives
48:16on the line to make sure Tesla does the
48:18right thing in terms of security and so
48:19car doesn't get stolen or hacked or you
48:21know somebody puts the brake on or
48:22accelerates it during a traffic jam on
48:24the other side we're talking about uh
48:26voting machines right we just had a big
48:28election a couple of weeks ago and so
48:30many voting machines are still closed
48:32systems I think the future of voting
48:34machines and David Wagner our own
48:35colleague is working on that uh is
48:37thinking about could you have a
48:39requirement from the federal point of
48:41view that says all voting machines need
48:43to be open open in software open in
48:45Hardware so that anybody can look
48:46obviously paper trail we talked about
48:48that always critical you Bo you get
48:50paper trail then you hand the paper in
48:52as your vote and maybe the voting
48:53machine is just you know you user
48:55interface to make voting easier for
48:56somebody
48:57but I love the idea of of mandating from
48:59the federal point of view an open
49:01hardware and Software System to all
49:02voting machines I think that'd be I mean
49:04it would certainly put a put a wrinkle
49:05in many companies who already have
49:07Investments there but I think that's the
49:08right way to go moving forward it's an
49:10opportunity for the next set of
49:11companies amen exactly exactly all right
49:15so I wanted to share a midterm score
49:17update uh Bor and I have written a
49:20midterm and the midterm we wrote we
49:23wanted to do better uh we were really
49:25hoping for a a better better number than
49:27this histogram but this is what we got
49:29uh as you see it's a very bell-shaped
49:31curve uh you know unfortunately we've
49:33got a median of about 50% and mean about
49:3650% and this is again this is Bora and
49:38my midterm this is what we tried and we
49:40tried to write this and wait Bora did
49:42you did you play the slides here
49:44something looks wrong with the slide
49:45here you know yeah this the age maybe
49:48yeah this is weird because because yeah
49:50the Year this is fall of
49:532018 the last time Bora and I gave it
49:57midterm
49:58oh this was the midterm okay are you
50:02ready for this this is what you get
50:05ladies and gentlemen the new school Bor
50:10Dan CeCe Justin Jerry Stephen everybody
50:15W
50:1783% clobber baby clobber all the way to
50:22the bank I love it was that good did you
50:25like it did I that was they're
50:27traumatized is that okay did you guys
50:29crush it or did you crush it oh my
50:33goodness okay this was not a lot easier
50:37fair this was a fair a lot easier
50:39midterm than what we had before this was
50:41a very fair midterm I think that it
50:43wasn't like we just gave it to you you
50:45know this is you guys did a great job
50:46you helped each other you guys are
50:48amazing I want to tell you all please do
50:51share this with your you know link to
50:53this part of the video when you share
50:54this with your friends make sure you
50:55don't just read the answer we're going
50:57to post this in about actually just
50:58posted now we just post the grades now
51:00and we're not doing the silly joke about
51:02the old school the new school we just
51:03post the normal grades it's posting
51:05right now it releases right now you guys
51:06crushed it crushed it crushed it crushed
51:09it we are so proud of you all great job
51:11everybody studying so hard working
51:13together through the you know but
51:14through the covid through the pandemic
51:16you guys Crush 61c you think you're
51:18awesome great job to everybody really
51:21really wonder this is really a testimony
51:23you know we streamlined a few things and
51:25they are going to stay forever with this
51:27class and this just shows that I think
51:31that the material went through I mean
51:34yeah and and we were seeing this on the
51:35project Scores Bar I mean this is not
51:37inconsistent right the consistency that
51:38we're saying is you guys are doing an
51:40amazing job this semester despite
51:42everything right despite all the co and
51:44all the this and all the everything
51:46we're doing we're trying to make this a
51:47better class and I think clearly from
51:49your midterm your guys are doing amazing
51:51it supposed to release now I'm seeing
51:52some people saying they didn't get it uh
51:55I I'm really proud of everybody here
51:56proud of our team I'm proud of our staff
51:58I'm proud of everybody involved on the
52:00staff part of view and proud of the
52:01students now that said some students are
52:04still struggling right we can't always
52:05celebrate unless you all got a 45 we're
52:07not we're not done yet okay we still
52:09have work to do please come talk to our
52:11staff if you didn't do so well people
52:13are on the lower end oh by the way this
52:14is very important we didn't tell you
52:15this this this number includes all the
52:19people that had to do a retake for
52:20whatever Parts they have we put them we
52:22had to put them the way this the grades
52:24were we didn't the software didn't know
52:26that they did have to do a retake so
52:27they have zeros so a lot of people have
52:29zeros or low scores they're going to be
52:31totally replaced with the retake score
52:33and that is still in process we're
52:35hopefully getting done in the next
52:36couple days so you guys even if you got
52:38a low number don't freak because it'll
52:40say by the way I think we were able to
52:41put in in the grade scope it'll say to
52:44be replaced by the midterm retake so
52:46don't panic if you got a zero because
52:48you knew because the retake isn't done
52:49yet this is before the retake so these
52:51numbers are going to go up so 83 is the
52:53lower bound we're going to be crushing
52:55this this histogram once we get the
52:57retake scores embedded into here so this
52:59is the beginning this is the beginning
53:01of a conversation before the retake
53:03comes in it's going to be amazing so
53:04great job everybody uh in terms of this
53:06really really well done let's do a
53:08quickie schedule I know we're we're
53:09running late uh we're at week 14 there's
53:13no lecture on Wednesday Happy
53:15Thanksgiving everybody no lecture on
53:17Friday uh spend some time with your
53:19family if you can if not if it's on Zoom
53:21or if it's remote please do spend time
53:22and be thankful for all the things we
53:24have uh in our society we don't have
53:26office hours this week correct no office
53:29hours this week all the Wednesday we we
53:31basically shut down after Tuesday we
53:32shut down after Tuesday and we revisit
53:34and we try to make it so that you can
53:35really take the time to you know enjoy
53:37your time that's why we moved P4 to 122
53:39so nothing really has to happen
53:41hopefully hope you started on Project
53:42four already and you're making progress
53:44but you can really take some time to
53:45recharge and reg get some energy um next
53:48week Stephen want to remind me this um
53:51next week we've got the normal uh you
53:52know recorded lecture Bora did a
53:54wonderful job on dependability parab
53:55parody ECC and raid that's it's already
53:59posted beautiful already posted that's
54:02that's his last lecture I did my last
54:03lecture on cloud computing which is
54:04today's lecture then this is important
54:07we're gonna have a wonderful guest from
54:09Apple uh I want to get the name until it
54:11actually gets finalized and and we're
54:13we're double double triple check that
54:15but we're hopefully to have a guest from
54:16Apple talk about a GPU and that lecture
54:19and then he's going to be giving another
54:21lecture on the 39th live so we think or
54:24we think it's going to be live so we're
54:25going to probably end up um using this
54:28exact time slot which is the five to
54:30four to five time slot for a live
54:32lecture on GPU we're very excited about
54:34that we're g hopefully announce it
54:36formally on Monday and then please do
54:38come live on 124 we're going to have the
54:42last session live we're gonna again
54:43cannibalize this webinar to have the
54:46last section live we'll have all the
54:48awards uh for all the prizes all that's
54:50going to be done live so please do next
54:52week is recorded lecture Monday live we
54:55think it's going to be live on Wednesday
54:57for GPU it might be pre-recorded if so
54:58we'll let you know and live on
55:01definitely live on Friday 12:4 so please
55:03do come for that uh and please note that
55:06the project 4 deadline is 122 don't
55:07forget that in in the in the last
55:11lecture we're going to have project
55:12three prizes we already have them uh
55:15project four we have to wait for all I
55:16mean there are people who have various
55:19kinds of extensions yeah we have at
55:21least six slip days we have to wait till
55:22so project four we're not going to be
55:23able to announce by exactly few extra
55:25probably yeah well and plus we'll need
55:27time like once your once your slip days
55:28are done then we have to run our project
55:30competition that's on our side so we
55:31need to be able to make sure that's
55:32checking out okay by the way if you're
55:35missing us I mean we are we are still
55:37[Laughter]
55:39around yeah and Bor do you have any room
55:42for undergraduates in your research
55:43group because I S we're going to
55:44advertise that last week yeah we'll
55:46we'll advertise asked my colleagues we
55:48have a ton I mean worries about that
55:50good job all right and I'm hearing that
55:52scores are just released so you guys are
55:53all good I'm so proud of you hope that's
55:55a great gift for you to have a nice
55:57Thanksgiving we will see you next week
55:59on Monday for our live session looking
56:00forward to that and don't forget come to
56:03we think it's going to be Wednesday live
56:04session and definitely Friday so Circle
56:07that in your calendars so we'll see you
56:08there all right guys this is a rare
56:11thing Apple speaker you know public yeah
56:15how to design the GPU we're excited
56:17about it we're excited about it good
56:18stuff all right thank you all right
56:21thank you all have a great time enjoy
56:23time with your families on video or live
56:25and it take some rest and come on back
56:26rest and recharged for our final exam
56:28we'll have our uh final exam uh details
56:32released uh on Monday 11:30 okay thank
56:35you take care folks Happy Thanksgiving
56:37everybody good to see you all all right
56:39take
56:40care all right all right

About this transcript

This page contains the full transcript of [CS61C FA20] Weekly Lecture 14.LIVE - WSC by CS 61C Departmental, generated from the public captions YouTube serves with the video. The transcript has 10,458 words across 1,475 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.