YT YouTube2Text

YouTube2Text → Transcripts → [CS61C FA20] Weekly Lecture 07.LIVE - RISC-V Datapath

[CS61C FA20] Weekly Lecture 07.LIVE - RISC-V Datapath — Transcript

by CS 61C Departmental · 9,875 words · 1,616 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00session okay and we'll start it okay
0:03we're on we're starting perfect we're
0:14live
0:20and people are here here welcome
0:28everybody okay
0:36all right
0:42good to see everybody here all right
0:44let's get started ladies and gentlemen
0:46welcome to
0:47week seven once a week lecture 61c
0:50logic risk 5 datapath
0:54great to see you we are almost at the
0:56halfway point
0:58that's amazing it's like it's like doing
1:00a podcast with you poor it's really
1:01great
1:02it's like a video podcast we go back and
1:04forth
1:06radio show radio show it's like a radio
1:08show i'd love this this is great anyway
1:09it's great to see everybody as always uh
1:12we've got a full agenda we should just
1:13jump right in
1:15yes exactly so
1:19so here is our first here's today's
1:22agenda
1:23we're welcome again welcome to week
1:24seven and as always we do kind of the
1:25same
1:26script with before uh we're competing
1:28the news we're going to mention in fact
1:30in a moment we have our wonderful guest
1:32victoria coleman is going to join us and
1:33borah will lead that conversation
1:35uh we do the week's plan and then we
1:36actually didn't add this to this but but
1:38stay tuned because we're going to
1:40review uh the workload and wellness
1:43survey so we didn't add this to the
1:44agenda but that is on the agenda today
1:46to have a deeper review and then a
1:47conversation of what do we take out of
1:49that so we're going to have a nice
1:49conversation at the end of it
1:51after we bid victoria ado after her her
1:54introduction so
1:55stay tuned this is going to be an
1:56exciting time uh not no big
1:58announcements but a nice conversation
1:59hopefully around
2:00uh the output of that workload and
2:01wellness survey
2:04big tech news board take it away yeah we
2:07mentioned this already uh i mean there
2:09is a huge bill that is going through the
2:11congress and the senate about uh
2:13semiconductor industry
2:14and generally design in the u.s um
2:18it is uh depending who you are asking on
2:21which house you ask somebody
2:23it's up to 22 billion dollars over the
2:25next 10 years that's going to change
2:27a lot of things and it is going to
2:29change the
2:31perhaps trajectory of your careers
2:34you who are going to be entering this
2:36field in about two years
2:38um there is a lot of stuff that is going
2:41to happen
2:42there um so we do
2:45have our guests today that
2:48we tried to get a few weeks ago
2:53and that is dr victoria coleman we
2:55actually tried to get her here on her
2:57first day when she
2:58became uh when she's born when she was
3:01sworn in as a
3:02director of darpa which is a defense
3:05advanced
3:06research project said in agency but that
3:09was a bit too tough
3:10so she's here two weeks into her new job
3:13and this is uh
3:14you know you guys should need to
3:16appreciate how big of a deal is to have
3:18our director show up in a class
3:21and we are very very very appreciative
3:24about this
3:27so we're going to ask victoria a few
3:29questions and hopefully her
3:31uh her audio is doing better now
3:34hi victoria thank you for uh joining us
3:38you're welcome can you hear me at all
3:40perfect
3:41much better thank you thank you
3:51wonderful technology is amazing but it
3:53works
3:55thank you thank you dan for inviting me
3:58along
3:59um i and you know i have to start with
4:01an apology i did try
4:03well i did try and fail to join you
4:06literally the first day on the job
4:08um it was time got away from me but
4:11i'm really pleased to uh to be here
4:13today um
4:14as uh as as some of you probably know i
4:18have a
4:18um a
4:22kind of weird little appointment that at
4:24berkeley
4:25that i'm very proud of um and i was i
4:28was really pleased actually to uh
4:30to be able to keep my appointment while
4:32i'm at the agency so that i'm
4:34i'm really happy um so when when this
4:37story is over
4:38you know i'll be back on campus and we
4:41will have many more war stories to
4:43exchange
4:44when i when i get there um but for now
4:48i'm out here on east coast missing
4:50california
4:53but really pleased to be here and you
4:55know if i can offer any
4:58um you know any thoughts any insights
5:01into
5:01the kind of things that you have in mind
5:03i'm happy to uh bona you talked a little
5:05bit about
5:06uh the legislation um
5:09is this let me ask you a few talk about
5:11first of all
5:13a few few questions first um it's uh we
5:16usually
5:17so this is perhaps the largest computer
5:19architecture class in the
5:21in the us and likely in the world at the
5:23moment
5:24and maybe in history maybe in history
5:26also
5:28that's great
5:32many of us are still learning over here
5:34how to
5:35to handle that um we
5:38have uh we decided to to keep students
5:41um animated and entertained by bringing
5:44interesting speakers every week
5:45this is something that we couldn't do i
5:47mean we you know nobody would fly from
5:48east coast or from
5:50from somewhere else uh to come to a
5:52class
5:53um we usually ask our speakers to
5:56introduce themselves how
5:57did you how did you you we i i know that
6:00you had a very
6:00long and and productive career in in
6:04industry and you know
6:05we know that you're sponsoring
6:06researcher at berkeley
6:08from different positions
6:11that you were in um and
6:14um but what we would like to you know
6:17briefly for our students
6:18tell us how you know what what do you do
6:21and how
6:22did you get there and then i'll ask you
6:24follow up after that how does one become
6:26a darker director
6:29all right so uh a a long and windy road
6:34uh and frankly you know a a
6:37a sequence of unlikely events
6:41um i uh you know as you can tell by the
6:44accent i wasn't born here i was born in
6:46greece
6:47um so age 18 you know i packed my bags i
6:50went to the uk to go to university there
6:52i'm a computer scientist
6:54um so i'm a lot of sympathy for a class
6:57like this i mean i'm really excited to
6:59hear this is the largest
7:00uh computer architecture class like i'm
7:02glad that somebody is still working on
7:04architecture it's like
7:06wonderful to hear um
7:09i think uh honestly it's
7:13well two things one is it's about uh
7:15what you want to do
7:17and the other one is about the
7:19environment in which you work
7:21giving you the opportunities to to do
7:23these things
7:24so i um when i first came to the us in
7:271998 i worked at sri um
7:30which you know back in the day used to
7:32be the stanford research institute
7:34so we had a lot of work going on
7:37with darpa in fact all the work that i
7:38did was with darpa and i kind of fell in
7:40love with the agency
7:42um i mean i i had to become a computer
7:44scientist because i watched
7:47i'm old enough to watch the landing on
7:49the moon
7:50um it was the first thing that you
7:51watched on on the tv i remember my dad
7:53went out and bought this
7:55great black and white tv set so you know
7:57that kind of changed my life and it was
7:59kind of interesting
8:00when i realized that had it not been for
8:02darpa we would have never gotten
8:04to the moon because like you know 75 was
8:07a key component of the apollo missions
8:11so you know it's um i
8:15i just fell in love with it um i i left
8:18sri
8:18in 2004 because you know as an academic
8:22you know some of you will appreciate
8:23this as an academic you aspire to get
8:26a real job it's so called a real job so
8:29getting closer and closer to product
8:30so i spent um a dozen years
8:33working in cooperating companies like
8:35intel and samsung and hp
8:37but i always wanted that
8:40i i knew that going to the agency or
8:44working
8:44in snt for national security was the
8:47place that i was
8:48most excited about so i did a lot of
8:53kind of volunteer advisory work as i was
8:55doing
8:56um as i was pursuing my career in
8:58corporate um
8:59and you know it came to the point
9:01frankly for me personally that
9:04you know i started telling people that i
9:05was going to work to support
9:08my dod habit i mean i ended up
9:11you know working 130 days a year
9:14on top of my day job doing the dod work
9:16and then it dawned on me like
9:18if i enjoy this so much maybe i should
9:20do it for a living
9:21so that was a big light bulb um
9:24and i decided to uh to to to join the
9:28agency now
9:29it's kind of interesting you ask how
9:31does one get to be darpa director
9:33um you know being darpa director is
9:37an exceptionally amazing job i think for
9:39me
9:40in my book it's the best job on the
9:42planet but there are
9:43many other very very good jobs at arpa
9:47so the program managers in darpa have a
9:49ton of fun
9:50um actually i think they have they have
9:52better their job is a better job than
9:54the director job
9:55um my appointment is political but many
9:59people don't know this but the director
10:00of darpa is the
10:02it's a political appointment it's the
10:03only political appointment in the whole
10:05of the agency
10:06so you become the developer director by
10:10virtue of the
10:12administration and of course you know
10:15every administration has different
10:17ways of picking its people
10:22in this case i think it was
10:26a matter of respect for the work that
10:28i've done in the past and also
10:30hopefully some trust about the work that
10:33i will do in the future in the agency
10:37[Music]
10:38it was uh it was really quite
10:42emotional i'll tell you i was born in
10:45the first week i was here
10:47and i was sworn in by michael kracios
10:49who is
10:50the cto of the united states and he also
10:53is the acting
10:54under secretary of defense for research
10:56and engineering
10:57and he is also
11:01a of greek descent um so i was thinking
11:04what are the chances of this
11:06you know having to to kind of two greeks
11:09to sue the greeks in the chain of
11:10command
11:11darpa and the department i don't think
11:14there are many places in the world where
11:16that can happen
11:17honestly um so i yeah i feel
11:21fortunate i feel humble i feel like um
11:25actually week two i'm a little less
11:26worried weak one was like yeah
11:28let me not break anything um you know
11:30this is an amazing
11:32amazing agency and uh
11:35you know i mean of course there's
11:36nothing that there's no so perfect that
11:38we cannot make it better so that's what
11:40we're gonna try and do but um
11:42first of all we're not gonna break
11:43anything
11:45uh bro did this get close to what you
11:47were trying to uh to elicit
11:49yes yes thank you very much uh that
11:52that's great yeah i mean students
11:53probably don't know you don't have very
11:55many
11:55layers of reporting above you um
12:01this is just a bit of a joke um students
12:04may not know what siri
12:05is and uh probably the best widely known
12:08product of c
12:09of of sri is siri sorry students don't
12:12know what sri is but siri is
12:16the best known product of sri
12:19um one and we don't mention what does it
12:22stand for it has a word that we don't
12:24mention over here
12:28it's a stanford research institute it
12:30has nothing to do with the stanford
12:32campus
12:34we don't do that
12:38quick uh um another uh quick question so
12:42this uh chip's bill has been going on
12:45for
12:46a while and you know i every time i try
12:49try to
12:50read something about it we it changes
12:52shape
12:53which is how things happen they're
12:55propagating from uh
12:57through the government it has
13:00one of the things that we have recently
13:02seen through the announcements that it
13:03has a
13:04big push for the workforce
13:08you know tell us something about the the
13:10the
13:12chips built and this uh
13:16part of it that has to deal with the
13:18workforce because we have 1200 people in
13:20a class and this
13:21may actually affect what uh uh what
13:24they're going to be doing for the next
13:2630 years or so
13:28yeah yeah it's kind of interesting i
13:29mean when you know when we talk about
13:31you know semiconductor you know the
13:33first thing that people think about is
13:35manufacturing it's you know it's uh
13:38moore's law blah blah but
13:40at the end of the day as you say it's
13:41all about the people um and it's kind of
13:43interesting you know we did a um
13:45back in 2018 the defense science board
13:48did a study
13:49um on uh on microelectronics
13:53um kind of leadership in the u.s and
13:56there were a number of recommendations
13:58that we made one of the recommendations
14:01was to
14:02to uh vastly increase the number of um
14:05sponsored phd positions that the
14:08department
14:09um has with particular emphasis on
14:13microelectronics and the idea there was
14:15that
14:16you know the department would pay for
14:19people's phds and then people would give
14:21back by committing to work
14:23the same number of years that they were
14:26sponsored by the department into either
14:29a government
14:30lab or a us-based company
14:34so you know i think that that is kind of
14:37part of the
14:38part of the picture but of course you
14:39know you don't get to do a phd in
14:40semiconductors or microelectronics
14:42if you haven't started you know probably
14:45at age
14:46you know 14 or 15 you know directing
14:48yourself
14:49you know in that um in that career so it
14:51starts much
14:52earlier than that um and
14:56you know it turns out that actually the
14:58department has a program it's a small
15:00program right now
15:01that uh you know starts by incentivizing
15:04high schoolers
15:06to um to consider microelectronics as a
15:09career and that kind of helps them
15:10through undergraduate years
15:12and certainly through to the graduate
15:14program
15:15so i think these are some of the tools
15:16that were at our disposal but i think
15:18that is you know behind the question
15:19there's a bigger
15:20kind of strategy right and
15:24you know for for a long time in the
15:26business um
15:27there was this perception that it didn't
15:31matter
15:31if if someone conducted manufacturing
15:34went overseas because
15:36we kept like the high value part of the
15:40the process here in the united states
15:42namely design
15:46and architecture
15:49and eda and so on
15:53so for a long time you know people
15:56weren't paying attention to the talent
15:57equation of this
15:59but you know one of the things that
16:01slowly becomes
16:02true if you don't pay attention is that
16:05you wake up one day
16:07and you don't have the skilled workforce
16:10domestically to even do kind of the
16:13innovative work on the design and the
16:15eda
16:16because think about that if you are a
16:19you know if you're an aspiring phd
16:21student
16:21you're gonna go and spend you know five
16:23six years of your life
16:25say at a school like berkeley at a
16:27school like mit
16:29you will only do that if there are good
16:32great
16:33jobs to be had here in
16:36uh in the united states and if if there
16:39aren't
16:39then you're gonna take your skills and
16:41your talents and go
16:42to the places that do have those
16:45opportunities
16:46so i think gradually what you you know
16:48what you're seeing is that this killing
16:50of the workforce here domestically if he
16:54let
16:54manufacturing go overseas kind of
16:56completely
16:57um i um i had the great fortune of
17:01working
17:01a little bit with andy grove when i was
17:04at intel
17:05and you know he really he really
17:08believed that
17:09unless you were able to scale
17:11manufacturing somewhat like a
17:12manufacturing in the us
17:14uh you will eventually lose the plot and
17:16you know and and here's somebody
17:18that really ought to know right i mean
17:20he i think
17:21him and involved more single-handedly
17:23kind of crafted
17:25the um the semiconductor business that
17:28we know today and of course
17:30another another a glorious alum of
17:33berkeley so i i always i always felt
17:36that his words kind of resonated
17:39um you know building
17:42enough kind of scaling capacity here at
17:44home to attract the best and the
17:46brightest and keep them in the fall so
17:48that
17:48we could indeed have them be part of the
17:51innovation kind of army if you like
17:53right that creates the amazing ideas
17:55but i don't think i don't think this
17:57industry is all about
17:59ideas it's also about making things and
18:01we need to have both here domestically
18:04which kind of circles back uh borrowed
18:06to the to the bill
18:07so the chips bill and before that the um
18:10the american foundries act
18:12they're all kind of trying to get it
18:15with this problem i mean everybody is
18:16watching
18:17uh the rate at which um manufacturing of
18:20semiconductors has shifted overseas so i
18:22think today
18:24uh we'll have 12 percent of global
18:25production uh steady state
18:28i believe in a decade we will be at nine
18:29percent um
18:31so that's a very uncomfortable kind of
18:34place to be in
18:35so there are you know lots of people on
18:37the hill um
18:38in the you know in companies in the
18:40department uh
18:42in in the associations trying to you
18:44know to figure out how can you know
18:46diagnose first of all why this is
18:48happening and then to try and find some
18:50ways of
18:51um kind of problem solving part of it
18:55clearly is
18:56um for the lack of virtual world
19:00incentives i mean one of the reasons why
19:03um you know companies
19:06like tsmc are very successful
19:09as well as you know having great
19:10technology is that they get a lot of
19:12support
19:13domestically from the government um
19:16here in the united states at the federal
19:17level um
19:19there are really no incentives for
19:21semiconductor companies
19:22so you know you may agree or disagree
19:24with that um
19:26but what it does do um is it changes the
19:29economics
19:30of uh manufacturing here versus in a
19:33place where these incentives exist
19:35so many folks are trying to get kind of
19:37their head around
19:38um what would happen if we were to
19:41inject
19:42us you know some incentives in in the
19:44business would that change that
19:46would it change the economics enough and
19:48would it change the economics
19:50long term for us to be able to uh
19:52redress the balance if you like
19:55um so that's part of the kind of
19:57conversation on the hill
19:58the other part of the conversation is
20:00what does the dod need
20:02because you know we certainly need um
20:05need to have
20:07access to um you know high-end parts and
20:10if for example
20:12all the production that we rely on is
20:14overseas you know perhaps
20:16the supply chain might dry up for a
20:17variety of reasons so the dod
20:20has very specific needs so this builds
20:22as well as looking
20:24overall at industry health they're also
20:27looking
20:28out for the needs of the department and
20:30also
20:31everybody comprehends that the way that
20:34you know
20:34we lead is by investing in r d so if
20:38if you like to abstract away from you
20:41know from the detail
20:42there are these three big pieces there's
20:44what do you do to
20:46to redress the financial balance the
20:48economics what do you do for the
20:49department of defense
20:51and then what do you do with r d and of
20:53course the r
20:54d piece is where the um
20:58the um kind of workforce piece uh
21:01falls in kind of squarely
21:04that's where one would look at in fact i
21:06think the latest version of the bell
21:08uh calls out specifically workforce
21:10development is one of the four kind of
21:12key tenants
21:13of uh what they call the nstc
21:15international
21:16semiconductor technology center so i
21:19think there is a recognition
21:20um but of course we need to keep
21:22reminding ourselves
21:24that it's you know without the workforce
21:25we have nothing and it's
21:27complicated because it's not just about
21:29you know the people that were in the
21:31bunny suits and i
21:32in the in the fab it's also about
21:35building a pipeline of talent that will
21:39you know that will supply the thousands
21:42that we need
21:42in all the different um kind of steps
21:46of the life cycle of semiconductors all
21:49the way from you know inception
21:51fabrics to you know packaging and
21:53delivery
21:55thank you we have a couple of questions
21:58from students we need to keep it short
22:00because we have
22:01other things to do with this class but a
22:04few things so most of
22:05the students in this class will be in
22:06the design pipeline
22:08meaning though you know there is very
22:09little use for the fabs if there is no
22:11if there are no chips
22:13to be manufactured yet to be made
22:17um and i in about three weeks we'll have
22:21extra
22:21you know another 1200 people who have
22:24designed the microprocessors
22:26from the you know from the gates up and
22:28that's quite exciting
22:30and that enables a path if they want to
22:32pursue uh
22:33to go that way one thing that we have
22:35noticed this class
22:36is actually sponsored by apple because
22:39they've
22:40perceived the same issue that there is
22:42not enough designers
22:45around here and they they prefer to hire
22:48locally
22:49we got two quick questions from the
22:50students one is
22:52what is approximately darpa's annual
22:55budget and where does it go
22:56to companies or universities okay so the
22:59budget the darpa
23:01um you know various from year to year
23:03but roughly it's about three and a half
23:04billion dollars annually
23:06um i could not tell you
23:09exactly what the the breakdown is
23:12but the darkwood typically sponsors
23:14research at universities
23:16at companies uh large or small and small
23:21darpa does none of the work inside the
23:24agency
23:24the way that we work is by bringing on
23:27board
23:28program managers who are experts in
23:29their field and they come
23:32with you know a strong desire to solve a
23:34problem
23:35in their domain um they pitch a program
23:38if they're successful they
23:39they they get money and you know if if
23:42you would like to
23:43you know to um to kind of
23:46squint and look at these programs you'd
23:48say the average size of a problem would
23:49be maybe
23:5040 million dollars over say three years
23:53this is roughly the level uh the level
23:56of effort
23:57that they um that they typically kind of
23:59walk away with
24:01um so you know
24:04the other thing that you should know
24:05about darpa is it is an agency which
24:08means it has
24:09a great deal of building independence in
24:11picking the elements of work that we do
24:14because we do very high risk work in
24:18other words work
24:18that honestly should not be relied upon
24:21as a source of roadmap elements
24:24because high risk work sometimes doesn't
24:26doesn't work out
24:27but on the other hand when it pays off
24:29it really pays off you know
24:31big time so so darpa and the legislation
24:34around it
24:35is crafted in such a way as to give the
24:37agency the freedom that it needs in
24:39order to you know to really innovate out
24:41the box and do amazing things and also
24:43it has the funds that it needs in order
24:45to be able to build out
24:47uh this ideas and because it has
24:50because we were part of the defense
24:51ecosystem we also have pull
24:53from our customers within uh the
24:56department having said that as you know
24:58oftentimes many of the technologies that
25:00we sponsor uh end up being dual use
25:02technologies
25:03so they end up in many other places
25:06as well as the um services
25:10our students may not know that risk one
25:12and race two were
25:14darpa projects back in the eighties
25:17with dave patterson because darpa was
25:19looking for a disruption
25:20risk five was not developed by darpa but
25:23the first prototypes of s5 processors
25:25were
25:26used in darpa projects um one uh
25:29final quick thing uh the question from a
25:32student
25:32is now that you are at darpa
25:37are there any specific domains or
25:39technologies that you'll be trying to
25:42promote microelectronics
25:49that's great thank you
25:52i think uh you know we are already
25:55thinking
25:55you know what we should be doing beyond
25:57eri so many of you will be familiar
26:00um i think with the electronics
26:02resurgence initiative
26:03which is a program that darpa it's it's
26:06actually it's a it's a collection of
26:07problems that darpa have put together 32
26:09years ago
26:11the idea that really being to reverse
26:14this trend
26:15of um you know leadership in in
26:17microelectronics kind of you know
26:19sliding away from
26:20from our hands here in the united states
26:22um
26:23the program was very successful in
26:25engaging non-traditional
26:27players in the research that we do in
26:30particular
26:30companies that typically would not be
26:33defense contractors
26:35um you know the angles of the world for
26:36example
26:38we are already looking to see what that
26:41program should look like
26:43for the next five years and we're also
26:45looking
26:46to really problem solve around some
26:49structural aspects
26:51of um of manufacturing
26:54again it's all about the economics the
26:56reason why you know
26:57manufacturing is going abroad is about
26:59economics so oftentimes darpa is looking
27:01for
27:02you know uh for work that has this these
27:05levers
27:06that um you know you apply a little
27:08force but you have a big effect
27:10um so we are you know
27:15putting ideas together uh internally um
27:18in
27:18certain days yet um so don't hold me um
27:21go home and to answer on this but uh you
27:24can expect to see
27:26um you know new ideas a new um
27:29you know a new interest at our part for
27:32building out um what we do in
27:34microelectronics
27:35but also how we connect innovations in
27:37like electronics with what happens in
27:39the world of computation
27:41right because it's not just about a
27:42hardware it's not just about the
27:43software
27:44we also will be looking at ways of
27:45connecting the two
27:47so that any innovation that we that
27:50would put forward on the converse side
27:52actually manifests itself and provides
27:55value to the end users who very
27:57very unlikely to be just hardwood users
28:00right that's not how um
28:02how typically people use
28:04microelectronics
28:05i like i say i i probably should have
28:08said you know check back in a month
28:09or two but um these are my thoughts you
28:12know two weeks
28:13two weeks into the agency and the way
28:15that the agency works of course you know
28:17there's a
28:18there are lots of very very smart people
28:19that have great ideas one of the things
28:21that i'm working on
28:22is bringing them all together and you
28:24know um and having
28:26a good old kind of brainstorming session
28:28and deciding how
28:30um how we should prioritize the kinds of
28:32things that we all want to do and what
28:33should we do more often less of
28:35so stay tuned um in the next you know
28:38um next few weeks months these things
28:40should become clear
28:42and um you know i'm i'm from the bay
28:44area
28:46our home is in the bay area i'm going to
28:48be out there a lot so
28:50i hope to uh to be able to come by and
28:52uh perhaps meet you all in person
28:54uh when coffee 19 permits the opening of
28:57the campus
28:59and by the way thank you very much just
29:01put that plug into that
29:02uh darpa is doing a ton of work on kobe
29:0419. so
29:05stay tuned on that as well thank you
29:09very much
29:09these folks in the in the student in the
29:11class are under undergraduate students
29:14almost all of them and i think they will
29:16only in about a couple of years realize
29:18oh
29:18we were in a zoom room with a darpa
29:19director that's exactly right
29:23they don't appreciate it at all they
29:24don't appreciate it
29:31[Laughter]
29:34thank you very much thank you for
29:36joining us so much for taking the time
29:48no i mean dan was in nbc news just
29:51saying you know what
29:57thank you thank you goodbye good to see
29:59you good to see victoria wonderful thank
30:01you that was so great
30:02thank you wonderful that was great
30:06back to work it doesn't get much bigger
30:08than that people don't realize what
30:10we're getting we're getting ceo
30:12darpa director do you realize who we're
30:13getting this this semester what an
30:15amazing year we have yeah
30:16there are only two levels of reporting
30:17above her just so you know guys i mean
30:20it's pretty pretty thin pretty
30:23remarkable pretty remarkable
30:25yeah um okay
30:29we're gonna jump right into the the
30:30schedule yeah so
30:33let's take a look at that uh we're in
30:36the week where we're looking at
30:37finishing up logic blocks that's my last
30:39lecture this week and then we jump into
30:40data path and control
30:42these are among the most exciting
30:44lectures uh
30:45that that we have all year in fact i
30:48want to celebrate
30:49borah we'll have a special special
30:52something next week when we finally
30:54make the connection with control we're
30:55finally going to be able to make the
30:57hardware and software link and we'll
30:58have some celebration live with
31:00all the students that's really going to
31:01be fun we're finally coming together you
31:04saw this this whole
31:05material starting monday 928 we're
31:07biting that biting that burger
31:10a veggie burger if you'd like all the
31:11way from the top down at the software
31:12angle
31:13you now have machine code and now we're
31:14biting it from the bottom up so sds is
31:16the lowest level then state is higher
31:18than that logic and logic blocks are
31:19above that
31:20then you have these bigger blocks big
31:22logical blocks that you can now
31:23wire together and part of what the
31:25wiring together is what's called the
31:26data path and then you have to then
31:28control them you have to control the
31:30the muxes if you've watched the lecture
31:31on on logic you'll see
31:33and actually today's lecture of logic
31:34blocks we're talking about how what a
31:35mux is
31:36how do you wire those boxes boxes like
31:38two to one guys so or maybe four to one
31:40guys or eight to one guys so how do you
31:42who how do you control these guys who's
31:44the marionette that does that that's the
31:45control
31:46so the kind of the brawn is the data
31:48path and the brain is the control
31:51how do you even set that up so all of
31:53that is what borah is going to be
31:54sharing through his lectures wednesday
31:56friday
31:56and then next monday and then we jump in
32:00how do we make that faster
32:01so how do we even jump in and by the way
32:03all this material that you see there on
32:05this slide is on the midterm
32:07the part that's not though friday you
32:08know wednesday uh 10 21 and friday 10 23
32:12is not so he's not even listing it here
32:14this is why we're doing that so we're
32:15listing all the material that's there
32:17pipelining's about how do you make that
32:18faster if you
32:19if you had and we'll make analogies to
32:21laundry and all those things so that's
32:22gonna be a great set of lectures for
32:24that
32:24and then we talk about caches but you
32:25didn't hear me say that but that's what
32:27we're seeing after that and that's all
32:28post midterm material but that caches is
32:29right after that
32:31so looking at the deadlines for project
32:32i'll get myself out of the way of this
32:34slide here you can see
32:35so looking at the the deadlines here
32:37project 2b is is
32:38due today 10-5 lab five is this week and
32:42that's
32:42we i think we had a break before for the
32:44lab and this is to talk about logisim
32:46to get you set up for project three
32:47that's critical you gotta learn how to
32:49lose
32:49logisim to be able to work with project
32:50three um and you can see here
32:53and there's a connection between the
32:54colors go ahead
32:56one one quick uh comment uh lab is
33:00pretty light um i i i've done it uh
33:03the previous version of that it's pretty
33:05light it shouldn't be very time
33:06consuming
33:07but it is to introduce you to logisim
33:10which will enable you to do the project
33:12and i think the next lab
33:14as well um it is you know logisim is a
33:17bit
33:17finicky tool you know you need to be
33:19disciplined and
33:21pay attention to that lab um you know
33:23there are many over there gotchas that
33:25that
33:26you could get into um
33:30into um if you are not very disciplined
33:33so pay attention to that
33:34go through it it's not um conceptually
33:38challenging it is just learning the tool
33:41and then when you learn the tool you can
33:43build anything including a response
33:44pipeline
33:45and it's remarkable that i mean when
33:47bore and i went to school they didn't
33:48have these tools we were all doing
33:50this on paper the fact that you have
33:51these simulators are just remarkable so
33:53the idea of venus and logisim
33:55changes the entire education experience
33:57so we're really appreciative of the
33:58folks at logisim
33:59who built those tools that we use
34:02speaking of that i think
34:03we have run run i think one ryan is
34:06going to give us like
34:07five minute preview of the project uh oh
34:10very nicely
34:11ties into the into the lecture so if you
34:15watch
34:15uh this wednesdays and fridays lectures
34:18wednesdays one is up and fridays should
34:20should be i i put all the videos where
34:22they're supposed to be
34:24they just need to be linked together um
34:27so if you as soon as you watch that you
34:29can actually we will launch the project
34:31tomorrow and you can
34:32take a look at it and get going with it
34:34it basically follows the project
34:36but you know doing it for real in
34:38logistics so
34:39uh ryan take it away you've got about
34:42five minutes
34:43whoo okay welcome welcome right thank
34:46you so much
34:46yep i'm going to pin you here here we go
34:48add your spotlight it's all you my
34:49friend
34:50all right so hey i'm uh ryan i'm a
34:54master's student boris actually my
34:55advisor for my master's degree and i'm
34:58the project lead for project three
35:00and then the other ryan brian thornton
35:01is uh the co-project lead
35:03um so project three as you've kind of
35:05heard is basically taking this
35:07simulation tool logisim which is
35:09uh you basically draw out logical
35:11circuits so
35:12wires and gates or gates more complex
35:15things like adders
35:17and you're going to create the risk 5
35:20data path and eventually the control
35:21logic create an overall core
35:23like you've been learning about in
35:24lecture
35:26so sort of the key component of this
35:30project is just having a good
35:32fundamental understanding of
35:34the risk five instruction set and how
35:37each instruction actually works
35:39and so that i think the lectures from
35:41this week which go over the data path in
35:43particular and the lab which was for
35:44logistics together will kind of
35:46give you that foundation from which you
35:47can build all of this
35:49um so some quick like
35:53updates uh tips and tricks and things to
35:55be aware of
35:58the first is you're going to get this
35:59logisim out of the
36:01largest binary jar file
36:05out of the project starter please only
36:08use that one it's a custom one that i've
36:10added some modifications for
36:12uh that help with this class and if you
36:15use the wrong version
36:16uh it can lead to some complications
36:18down the line um
36:20the second one is try doing that i i did
36:22i did that already
36:24in my attempt today um it causes um
36:29you know you don't want that
36:32yeah yeah i thought they had an
36:33installation that looked pretty good
36:36um uh do you need anything to share
36:39are you able to share us do you have
36:41anything to share and see a little demo
36:42we're gonna see a demo here
36:44um i'm gonna release yeah i'm so i'm
36:47gonna release some uh
36:48demo videos that i'll record probably
36:50later tonight and they'll be in the spec
36:52when it gets released
36:53of sort of me walking through how to do
36:55one of the components
36:57a little toy example so that everyone
36:58can kind of see the way
37:00that i as a ta think about this and how
37:02i'm
37:03building components and how i'm testing
37:05my work so hopefully that will help
37:07you all as you're working on your
37:08project understand um
37:10sort of better the flow that is
37:14designed for you um so just another tip
37:17or trick is to be very careful with the
37:18spec first all the way through
37:21and then start um spec reading is a very
37:24important part of the design process
37:26and just make sure you you understand
37:28everything feel free to ask questions on
37:29piazza if you need to clarify anything
37:32it's better to understand that first and
37:35do it a second rather than do it first
37:37and ask questions later
37:38um and then the third thing is that we
37:41have a design contest this year thanks
37:43to
37:44uh being sponsored by apple
37:47um so at the end of this contest we're
37:49going to take all the submissions
37:51that pass 100 of the tests uh hidden
37:54invisible
37:55and we're going to convert them uh into
37:57gate equivalent models basically so
37:59we're going to say you had
38:01this adder which is equal to 200 nand
38:04gates
38:05and then we sum together all the
38:06components of your circuit and we get
38:07this one composite score which is
38:09basically how
38:09big is this thing you've made um and
38:12then the lowest
38:13size will basically win points for this
38:16section of the contest so another thing
38:18to think about as you're designing is
38:19just
38:20you know what's the simplest way i can
38:22do this you can do things a very
38:23complicated way using very large blocks
38:25so you can kind of abstract it down to
38:26the
38:27simplest logical forms and use small
38:29blocks and small wires
38:30and if you are willing to go through
38:32that effort you will be rewarded for
38:34that
38:34in the contest um so yeah the spec will
38:37be posted
38:39tomorrow sometime um there'll be some
38:41walk through videos some more guides
38:43if you have any questions ryan and i
38:45will be in the q and a in chat for the
38:47rest of lecture
38:48and uh feel free to come to our office
38:50hours uh mine are monday
38:53morning uh monday morning and then
38:55tuesday morning i have some special
38:57project office hours
38:59brian can chime in with his if he wants
39:01and yeah otherwise post on piazza
39:03send us emails just please ask a lot of
39:05questions and we're here to help
39:08do you wanna add anything ryan bryan t
39:10uh yeah so
39:12just a little bit about lab5 uh make
39:15sure to do that as early as possible
39:18um because logism labs and projects take
39:20a little bit more time than coding
39:22things
39:23um i have lots of lab office hours
39:26throughout the week on tuesday to
39:28thursday
39:30um i think my ta office hours on
39:33wednesdays
39:36perfect well thanks for both thanks to
39:38those to you that's just wonderful
39:40so one thing to for everybody should
39:42appreciate is the
39:43the the project and
39:46the project itself has been reshaped a
39:48little bit as
39:49through the lens of folks who have gone
39:52through upper division classes recently
39:54that
39:55go through there is five processor
39:56design and
39:58uh really reflect that um for masses now
40:02better you know these are the guys that
40:04went through the through the ranks
40:06of the 61c to 151 152
40:10you know graduate classes and understand
40:12the process uh
40:14you know what are very fresh to the
40:15process and that is all reflected in the
40:17spec
40:18in in the in the design and you know
40:20when you come out of this
40:22you have a processor in your hands and
40:24don't don't do it
40:25just for the pair of airpods uh we'll
40:28try to get a few of
40:30a few extra airpods over there but do it
40:33so you'll have a rs5 core in your hands
40:38and as i said for me i don't know i
40:39don't i don't want to suddenly the vote
40:41but for many people they find
40:42that that the processor design is the
40:43most fun project in all the 621c we have
40:45some fun projects and they find like
40:47this is so cool to actually build this
40:48finally
40:49so some people really get into it that's
40:50wonderful thank you ryan thank you very
40:52much
40:53yeah let's jump in thanks so much folks
40:56good stuff you guys can stay or you can
40:58relax thanks all right let's keep on
41:01let's go
41:04i know got a lot of stuff to do all
41:05right let me get in here all right here
41:06we go
41:07so uh
41:10next so this is the pipe project three
41:13intro we did some of that
41:14um born and i took a little
41:18took a look uh we continue to look at
41:19the responses there's a ton of them to
41:21read through we want to make sure to
41:22read everybody's comments
41:24761 of you replied thank you so much for
41:26that that's a little bit more
41:27more than half maybe 60 of you to
41:29respond um
41:31and the the the i would say overall i'm
41:33gonna borrow your first
41:34take on this but my first reaction is
41:36these numbers were sobering um these
41:38numbers were were not numbers that i was
41:40um expecting what's harder is we don't
41:42have a baseline we never did this during
41:43a norm
41:44a non-covid semester so it's harder to
41:46know how people
41:47you know we're doing we used to do hey
41:49how's how's the midterms there but it's
41:50less about your health
41:51that's about you know mental health and
41:52workload um so but we're hearing
41:55you can see these numbers we're sharing
41:56with you now are just remarkable in
41:58terms of how overworked you are how
42:00overtaxed you are how mentally exhausted
42:02you are and we're not even in week
42:04seven yet we're just starting week seven
42:06um so
42:07we're just sharing all this borrower
42:08feel free to chime in with other
42:10thoughts you had about when we see these
42:11numbers but they i thought they were
42:12very very sobering
42:13um so how are you doing emotionally you
42:16know we were hoping for fives and sixes
42:17and sevens and some of you were in that
42:19area
42:19but most of you are averaging around
42:21three so you're a little bit below four
42:22is kind of
42:23neutral most of you are kind of feeling
42:25feeling a little bit on the low side and
42:27seven is great and one is awful okay
42:30um exhaustion was even stronger in terms
42:32of a signal for us to try to make some
42:34changes um you know again we're looking
42:36at fives and six and sevens and people
42:38feeling okay we do know as we mentioned
42:40that the earlier part of 61c is harder
42:42and it does
42:43ease up at the end so we're now you know
42:45the foot is we're kind of
42:47taking the foot off the pedal now uh
42:48with project two finishing up and then
42:50it's you know only two projects are
42:51still left for
42:52for for i think just eight weeks left so
42:54so it's much less of that and it's
42:56thinking about
42:56kind of like it all sink in but but the
42:58material comes fast and furious you know
43:00every
43:00three days is a new topic that's that's
43:03positives and minuses the minuses are
43:05it's a new kind of idea to wrap your
43:06mind around the positives are if you got
43:08behind
43:08you can rejoin and jump on the bus again
43:10and be part of it so that's the part
43:11that's it
43:12that's harder but nevertheless this this
43:14is again sobering
43:16in terms of how exhausted you are seven
43:17is doing great and one is just totally
43:20zero batteries are dead at one and you
43:22know we got
43:23one twos and threes yeah my take on this
43:27is the
43:27the first graph was not that bad i was
43:29hoping for a four
43:31on the average we got the three um yeah
43:33yeah
43:34but this one is uh this is not a 100
43:37meter dash right i mean this course is
43:39not under meter dash this is more like a
43:425k race um so we need to make to make
43:46sure that you guys can
43:47pace yourself uh uh i mean we have have
43:49to to not
43:50over over bark you in the first yeah
43:54in the first mile you have you have two
43:56more miles and and that was actually a
43:58comment or two by people in the overall
44:00general thing was this does the first
44:01beginning of six weeks he have to be
44:02this hard
44:03does it have to kill us in the first
44:05couple of weeks could it just you know
44:06many semes many courses 162 and
44:08you know project courses they get hard
44:10in the last five weeks and the first
44:12five are set up and then it's a big
44:13project that just overwhelms you
44:14i mean what's the what's 150 more right
44:17150 is this like
44:19unit class where that you know people
44:20just drowning for the last
44:22eight weeks or seven weeks or something
44:23or six weeks for that project and living
44:25in the lab
44:26there's a little difference experiencing
44:27we front load a little bit of that which
44:29it's a lighter at the end so it's a
44:30little different and people said dude
44:31does it have to be that way could you
44:33just switch it back to there
44:34um so we need to be we need to have a
44:36rethought we actually did
44:37none of you know this but we had all
44:39this the lead staff had teased
44:40us went through and did an estimate of
44:42how many hours per week with this and
44:44these were just very low
44:45i think our estimates were much lower
44:46than they really are
44:48in terms of that and also we made a few
44:50mistakes we made
44:51a few mistakes as we were walking yeah
44:53and uh we are very conscious of that
44:57and and on top of that this this also
44:58reflects your your
45:00exhaustion for other classes too this
45:02isn't just 61c but since you would see
45:03it might be for many of you that think
45:04that that really is is hard
45:06um i do want to set the survey so let me
45:08get to the next one um again how are you
45:10juggling that's very similar
45:12you're it's hard to manage a little bit
45:13less less than average here
45:15um prior knowledge we were pleased we
45:17were thinking well maybe if that's the
45:19case that people came in without
45:20strength a strong 61 and b
45:22experience and maybe you know that that
45:24then that that that's maybe the argument
45:26for maybe oh we need to
45:28hold the line stronger if half of you
45:29hadn't taken six to one a or b
45:31or something and somehow snuck into the
45:32class and the reason you're struggling
45:34is because of that
45:34but it's not most of you have you know
45:36completed six to one a and b well and
45:37done well in that
45:38um and most of you don't i actually see
45:40programming either today
45:42i did some i tried to do a correlation
45:44and there is clearly a correlation uh
45:47between how you you feel uh about your
45:50uh 61ab and uh how are you
45:53doing in this class right uh or in
45:55general and
45:56there is correlation but it's not that
45:58strong so
45:59there are people who did extremely well
46:01in in amd but
46:04or think uh they're very confident about
46:06their a and b but they're
46:07still not doing great now so things
46:09happen we and we
46:10very much um
46:13appreciate having being able to know
46:15that
46:17um this is the time on the quest we
46:20already knew this we already knew this
46:21from the
46:21piazza responses but this is kind of
46:23more of a you know more data points here
46:25and the numbers we're seeing here if you
46:27look at the average it's around four to
46:28five hours
46:29for an exam that we was hoping to have
46:31be an hour two max and people were
46:32mostly doing it in
46:33four or five hours and a lot of people
46:34taking a lot longer than that so
46:36this leads me to the next question i
46:38want to get to but let me let me
46:39find we have we have a quick poll i want
46:41to get some sense of
46:42but i want to just show you the rest of
46:44those numbers before we do this project
46:45one timing
46:46uh by the way everybody did amazingly
46:48well on project one and one b the
46:49average i think the median is 100 for
46:51both of those parts so
46:52people are doing very well but they're
46:54putting the hours in as well and i think
46:55the we need to look at these numbers and
46:57then
46:57we had an estimate for how long project
46:58one should take uh we need to look to
47:00see if if it's at all in online but a
47:02lot
47:02of high numbers so yeah the estimate was
47:04around the mean the the the
47:06the mean but the problem is we have a
47:09very long um
47:10the long tail i mean even it's a long
47:11tail even if you throw out some of those
47:14uh that may not be completely honest
47:17answers
47:18um this is a very long tail and this did
47:20hit um
47:21some people and not necessarily because
47:23of the of
47:2561a or b it was just thanks um
47:28that had to do with the providing tests
47:30and and and things like those
47:32and and part and part of our perspective
47:34is we thought that people learn how to
47:35do tests in 61b that was the point of
47:37that but
47:37we were hearing loud and clear people
47:38were really not ready to do tests yet so
47:40this
47:40there might be a breakdown in kind of a
47:42and b of preparing students4c
47:44but we're thinking that they're ready to
47:45do their own tests and in fact they're
47:47not
47:47possibly so that that also can reduce
47:49the work time at workload if we give you
47:50a set more tests
47:52um overall you're supposed to be
47:54spending six hours per week
47:56uh it's 12 12 hours per week for a four
47:57year class and the numbers are
47:59much higher than that that we're seeing
48:00but there is a there are a fair number
48:02of folks who are in the lower numbers
48:03but
48:03you know a big chunk of 15 at big chunk
48:05of 20. on an even big number blips at 20
48:07and 28. and that's just that's
48:08inappropriate in terms of
48:09our point of view we need to fix this we
48:10need to be better and we need to not be
48:12overloading as well so again this is all
48:14things for us to fix
48:15how many of these we can fix they may
48:16have to fix in the spring but we're
48:18going to see we can do to try to
48:19compensate for this here
48:20so to this end any midterm coding
48:23problems
48:25there's going to be a chance to do some
48:26coding and see a smaller problem
48:28to replace your quest we have to do that
48:30and there's some coding in mips
48:31um we are going to have all test cases
48:34given to you
48:35um we're going to reserve some of those
48:36to prevent you from hard coding if we
48:38give you every single test you can just
48:40basically look at our tests and just get
48:41it working for those we're going to try
48:42to
48:43support you there but we're going to
48:44have a couple of you know you know if
48:46example split you're supposed to do the
48:47splitting thing for your thing
48:48if we have a similar thing we're going
48:49to obviously have a big long one and
48:51that you know some points for handling
48:53are hidden one hidden test but mostly
48:54we're giving the test to you
48:56so that's very fair we have to obviously
48:57have a couple of them so you don't just
48:58hard cut to that
48:59we will announce how long it took all
49:00rtas we will survey our tas and announce
49:03that in advance as well as announce
49:04every piece that's on every exam you're
49:06going to have six questions they'll be
49:07on this topic you already know what the
49:08topics are they're everything up to up
49:10and through pipelining but we'll make
49:11sure that
49:12you know and you are going to be coding
49:13so be ready for that but we will give
49:14you your tests
49:15or a fair number of those tests we're
49:17going to lie for 48 calendar hours take
49:19the exam that's another piece of it
49:20people said
49:20we don't you know i was busy on sunday
49:22sunday even though we announced it very
49:24well
49:24that was a bad day for me either because
49:26religious reasons or other scheduling
49:27things
49:28give me more flexible time so we're
49:29gonna have one weekend day and one
49:30weekday
49:31and i have a poll right now that we're
49:35launching
49:36that i'd like you all to fill in two
49:37questions we're asking you number one
49:39what day would you what days would you
49:41prefer to take it um
49:43we're going to probably give a 48-hour
49:44time limit for what we believe is a
49:46two-hour exam
49:47that may be there so that's what we're
49:49aiming for
49:50the poll is open please do click on the
49:52poll if you can i see nobody's voted yet
49:53please jump into the polls so we can see
49:54those numbers
49:55and we can see when people are taking it
49:57because we have to staff piazza to
49:58answer these questions
50:00the second question is we are of two
50:03minds
50:04of whether we give you all 48 hours and
50:06say
50:07it really shouldn't take you more than
50:08three you know two three or four hours
50:10max we believe but do we give you all of
50:14this some people said if you give me 48
50:15hours i will spend 48 hours you have to
50:17stop me from my best interest
50:19so help me limit that other the other
50:22counter to that is
50:23some people take longer when i was in
50:25college i took longer to do exams i did
50:27well i got to berkeley
50:28but i took longer to do exams than my
50:30neighbors i would just be more
50:31refined and quieter and i would i would
50:33just make sure i was more perfect with
50:34my answers
50:35so i want to you know react to that the
50:38students who need more time i think it's
50:39more
50:40fair to give them the more time but some
50:41people are saying no don't have an
50:42unlimited time thing
50:44so that that that that that that's at
50:46issue so we're we wanted to get a sense
50:47of that and
50:48keep on voting we have about uh 92
50:50percent of you voting
50:51and you are it was funny boy if you see
50:54you can't see it you can't see this only
50:55i can see these
50:56these are these are very funny i say
50:57that i see them oh you can see
50:59look look at the second one let me share
51:02this with you
51:03this is no signal whatsoever let me
51:06close this 96
51:07voted heading poll i'm going to share
51:09the results with you
51:10um and what we're seeing is
51:13uh you're all across the board friday
51:15saturday sunday and monday
51:16most of you are saturday and sunday will
51:18work for when you're going to do it
51:20but some of you need a friday or
51:21saturday but we need to limit we can't
51:23just have it open for
51:24for four days when our staff needs to be
51:26able to do other things if you have a
51:28four days maybe it'll be a longer time
51:29for you to get an answer
51:30so that's harder if we're not staffing
51:32all of it all the time for that
51:33the second one i've seen here are just
51:36evenly split whether we should not have
51:38an upper limit
51:40um you know i'm seeing 45 saying no
51:42upper limit and
51:43a little bit less than 40 percent maybe
51:45exactly 40-ish 41
51:47uh on yes upper limit so it's it's
51:52i don't know if you have a situation
51:54here an average of data here it's the
51:55same situation
51:57with about the the these uh little
51:59clicker questions you know we we are
52:01going to make
52:02exactly a half of the the class i'm
52:03happy unhappy exactly
52:05exactly and whatever way we make unhappy
52:07i mean if we if we had to lean in one
52:09way i would lean toward
52:10no upper limit just because you can
52:12satisfy more but we would ask those who
52:14need an upper limit
52:15to set one internally that's what we'd
52:17ask if we'd end up if we end up leaning
52:19that way
52:19we would ask him please just set a timer
52:21and stop working on it after two hours
52:23or an hour or whatever
52:24we will by the way break us up into
52:25pieces like we did last time and we're
52:26the same proctoring as last time and
52:27hopefully people have
52:28practiced their proctoring stuff so that
52:30they and it'll be smaller times with
52:31smaller pieces it won't be a long
52:33two hour video we have shorter pieces so
52:34i think we'll have probably i think we
52:35said four or five questions or so
52:37and it's five small pieces for that and
52:38you'll stop and start your video for
52:39that yeah
52:40all right i'm all done
52:44yeah yeah thank you so much folks for
52:45that feedback this is a fascinating poll
52:49with no data but again to see you good
52:51luck with this continue with this
52:52um we're week seven and we'll be back
52:54next week any last
52:56thoughts labs and projects uh on from
52:59tomorrow all right
53:00good stuff all right folks take care
53:02thanks so much for joining me let's live
53:03session folks
53:04wonderful bye

About this transcript

This page contains the full transcript of [CS61C FA20] Weekly Lecture 07.LIVE - RISC-V Datapath by CS 61C Departmental, generated from the public captions YouTube serves with the video. The transcript has 9,875 words across 1,616 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.