[CS61C FA20] Lecture 40.LIVE - Summary — Transcript

Full transcript

1. 0:02DAN GARCIA: Okay, perfect.
2. 0:05Let me look at participants.
3. 0:06Welcome, everybody, to the last class, 61C.
4. 0:10So happy to see you all here.
5. 0:1245-- see if we can get to it.
6. 0:13Boy, it would be great if we could actually
7. 0:15get to a hundred.
8. 0:16Boy, that would be wild-- one out
9. 0:18of every 10 students showing up for the last live class.
10. 0:20[LAUGHS] What a stretch that would be.
11. 0:23By the way, before I start, let me share that it's important
12. 0:27for you to sign up and do the course evaluations,
13. 0:30course-evaluations.berkeley.edu.
14. 0:33We need to get up to 90%.
15. 0:35You get a free point.
16. 0:36So please.
17. 0:37This is just wonderful, wonderful to sign that up.
18. 0:40So do that while you're waiting.
19. 0:42And we will start recording.
20. 0:44Actually, recording now.
21. 0:46And we will start in a second.
22. 0:51All right.
23. 0:51Bora, you feeling okay?
24. 0:52You ready?
25. 0:53BORA NIKOLIC: I am good.
26. 0:54DAN GARCIA: All right, good stuff.
27. 0:55BORA NIKOLIC: Are you driving the slides?
28. 0:56DAN GARCIA: I'm driving the slides, yep.
29. 0:57You can all see that.
30. 0:58Hopefully, you can all see.
31. 0:59We're not doing mm-hmm today because of audiovisual reasons.
32. 1:02You'll see the reasons why.
33. 1:03But hopefully, you can see us on this week.
34. 1:04Probably should pin ourselves too.
35. 1:06I just realized we should probably--
36. 1:07is there a way to-- ahh.
37. 1:09When you're driving screen, you lose all the windows.
38. 1:11Let me see if I can pin myself.
39. 1:12BORA NIKOLIC: That's right.
40. 1:13I can pin myself.
41. 1:14DAN GARCIA: Spotlight here.
42. 1:15Spotlight myself and spotlight you as well.
43. 1:18How about that, Bora?
44. 1:19BORA NIKOLIC: I got it perfect.
45. 1:19DAN GARCIA: And that way our TAs are not being spotlighted.
46. 1:21They don't have to worry about being online.
47. 1:23Perfect, okay.
48. 1:24So let's get rolling.
49. 1:27Welcome.
50. 1:27We got to 77.
51. 1:28BORA NIKOLIC: All right.
52. 1:29DAN GARCIA: Ladies and gentlemen,
53. 1:30welcome to the last class in 61C, lecture, I think, 45.
54. 1:36It's amazing.
55. 1:36Friday, week 15-- woo!
56. 1:41BORA NIKOLIC: I mean, can you imagine that we're almost done?
57. 1:44DAN GARCIA: Can you believe it?
58. 1:45I can't believe it.
59. 1:46This is incredible.
60. 1:47I mean, what a road this has been.
61. 1:49Holy moly.
62. 1:50BORA NIKOLIC: Yes.
63. 1:51DAN GARCIA: If you had told me the bumps in the road,
64. 1:52I was, ah, come on.
65. 1:53There won't be bumps in the road.
66. 1:54Bumps in the road?
67. 1:55They're mountains.
68. 1:56You're all here together.
69. 1:59BORA NIKOLIC: I can tell you this was the craziest semester
70. 2:02that I've had in my 21 years at Berkeley, period.
71. 2:05DAN GARCIA: Me too.
72. 2:05Me too.
73. 2:05BORA NIKOLIC: I mean, and the craziest one.
74. 2:07And then-- yeah.
75. 2:09And then comes the spring 2020.
76. 2:12DAN GARCIA: Well, there's like three things that
77. 2:14are distinguished this year.
78. 2:15One, it's the craziest semester.
79. 2:17The second is that these students
80. 2:19are walking into the final with the highest
81. 2:21grade I have ever seen in 61C.
82. 2:22And I've taught this, I think, 10 times.
83. 2:24I've never seen this histogram, this total course points,
84. 2:27going to the final.
85. 2:28So these students are as primed as they are.
86. 2:31All that work you guys put into project 1, 2, 3, and 4
87. 2:34and the studying for the exams, wonderful.
88. 2:36It has really paid off in terms of your performance.
89. 2:38BORA NIKOLIC: But not only that, we
90. 2:44get some feedback about the performance in interviews.
91. 2:47Our friends at Apple told us that all the firmware positions
92. 2:53that needed basically understanding of C
93. 2:56and how does it work with hardware
94. 2:57were sold to Berkeley students, Berkeley undergrads,
95. 3:02five days after posting.
96. 3:03That's it.
97. 3:04DAN GARCIA: Wow, wow.
98. 3:04BORA NIKOLIC: I mean, nobody failed.
99. 3:06Apparently, everybody passing through.
100. 3:07DAN GARCIA: Wow.
101. 3:08Wow.
102. 3:09BORA NIKOLIC: I feel really good about that.
103. 3:10DAN GARCIA: Congratulations.
104. 3:11Congratulations, everybody.
105. 3:12BORA NIKOLIC: I mean, I really feel good about that.
106. 3:14It all feels like it was worth it.
107. 3:16DAN GARCIA: And you made Berkeley proud.
108. 3:18I mean, this is what we are.
109. 3:19Our job is to set you up.
110. 3:20We're the coach.
111. 3:21Like, we're the ones on the corners in the boxing ring.
112. 3:25You can do it.
113. 3:26You're a lean, mean 61C machine.
114. 3:28And we send you off into the round.
115. 3:30And you've all done very, very well.
116. 3:32So great job, everybody.
117. 3:33BORA NIKOLIC: And you also create
118. 3:35opportunity for the next class and the class after that.
119. 3:39And it feels good.
120. 3:42It feels like it's worth doing all of this.
121. 3:44DAN GARCIA: I love it.
122. 3:45I love it.
123. 3:46BORA NIKOLIC: So what we have done in this class--
124. 3:48let's try to recap it a bit.
125. 3:50DAN GARCIA: Sure.
126. 3:52BORA NIKOLIC: So we tried to convey six really great ideas
127. 3:57in computer architecture.
128. 3:59So we drilled on the abstraction.
129. 4:01So we have been talking about building
130. 4:03really complex systems.
131. 4:05But we can't build these complex systems
132. 4:07by having everybody always touch every gate at the very bottom
133. 4:11when writing a very complex system.
134. 4:13DAN GARCIA: Too much detail.
135. 4:14Too much detail.
136. 4:15You'll just drown.
137. 4:16You'll drown in detail, exactly.
138. 4:18BORA NIKOLIC: And too many things going on.
139. 4:20So we build these layers of abstraction.
140. 4:22And these layers have served us really well.
141. 4:25We have been able to tackle them layer by layer,
142. 4:28going from top to the bottom and back from bottom to the top.
143. 4:32And I'm hoping that everybody feels really comfortable
144. 4:36now understanding how does a computing system work.
145. 4:42That's the first big, big point that was drilled down
146. 4:49throughout the class.
147. 4:50The second one was this thing of Moore's law
148. 4:53that got us where we are now.
149. 4:55And there is the scare that Moore's law may be going away,
150. 5:02that the technology is not going to continue scaling
151. 5:06at the same pace as it did before.
152. 5:10And that's true.
153. 5:11But it is still getting better.
154. 5:13And new things are going to happen.
155. 5:15And we just need to understand that we
156. 5:18are going to have to deal with power limits.
157. 5:20We need to figure out how to go towards designing systems
158. 5:25with slightly better technology and a little bit more energy
159. 5:28efficient.
160. 5:29So we'll have to deal with design under power constraints.
161. 5:34And that's going to be big.
162. 5:35I mean, that's the task for your generation.
163. 5:37You're going to be dealing with that.
164. 5:39There is no miracle technology that
165. 5:41is going to come and suddenly start scaling
166. 5:44into subatomic levels.
167. 5:46You guys will have to figure this out.
168. 5:48That's the mission for your entire generation.
169. 5:51And you know what to do.
170. 5:52DAN GARCIA: And quantum may be coming.
171. 5:53That's going to be exciting.
172. 5:54When quantum is here, that changes everything.
173. 5:56I can't wait to teach the new 61C with quantum
174. 5:59as one of the things we'll use.
175. 6:00I just can't wait for that.
176. 6:02BORA NIKOLIC: Yeah.
177. 6:02I mean, but don't hold your breath right away.
178. 6:05I mean, we've got to make do--
179. 6:06DAN GARCIA: (LAUGHING) I agree.
180. 6:07I agree.
181. 6:07BORA NIKOLIC: --with CMOS for another 10, 15 years or so.
182. 6:09DAN GARCIA: I agree.
183. 6:11BORA NIKOLIC: Exactly.
184. 6:13Exactly.
185. 6:13I mean, we talked about that.
186. 6:15We are seeing glimpses of this using specialization.
187. 6:18And yeah, we have to change 61C as we go along, the big thing.
188. 6:23I'll get back to that.
189. 6:24So the third thing that we did is this principle of locality
190. 6:28and memory hierarchy.
191. 6:30And that is huge.
192. 6:32It looks like we are working with this infinitely
193. 6:35fast and infinitely large memory system, although it is not.
194. 6:40It is limited.
195. 6:41I mean, we most of the time are working
196. 6:44with those few registers that we got in L1 cache.
197. 6:48And everything else is there just
198. 6:49to show up right in time for us to grab it and do computation.
199. 6:54DAN GARCIA: It's a beautiful abstraction
200. 6:55to have the abstraction to make you
201. 6:57think you get this benefit of infinite size and infinite
202. 6:59speed.
203. 7:00L1 speed at size of disk--
204. 7:02that is just a great lie that we tell you.
205. 7:05And it just works.
206. 7:06I mean, it is wonderful.
207. 7:07A lot of, obviously, engineering behind it,
208. 7:09but it's just pretty cool that you can think that way.
209. 7:11BORA NIKOLIC: The other thing, keep
210. 7:12in mind one of the challenges you'll
211. 7:14hear a lot of people talking about,
212. 7:16oh, there is some other technology that
213. 7:19is going to come and rescue us.
214. 7:22We are going to do non computing and so on.
215. 7:25And typically, when you drill down,
216. 7:29they forget about this principle of memory hierarchy.
217. 7:35And the life-- working without being
218. 7:39able to build these luxury applications that we have
219. 7:43running on top of the operating system
220. 7:45with multithreading and many processes and all of that
221. 7:50working simultaneously and not impacting each other, that is
222. 7:54going to be really tough like.
223. 7:56I don't think we're going to see much of that.
224. 7:58So keep in mind this is a very basic principle.
225. 8:01Don't violate it.
226. 8:02We do need infinitely big and infinitely fast
227. 8:04memory built out of cheap components.
228. 8:09The fourth thing that came initially
229. 8:12to help out with the issues with power limitations and slowdowns
230. 8:17of Moore's law is the parallelism.
231. 8:20That is a huge thing that we drill through.
232. 8:22So if you took this class 10 years ago,
233. 8:24there was almost no parallelism.
234. 8:25DAN GARCIA: That's true.
235. 8:26That's true.
236. 8:27Gate level-- there was some gate level, I'm sure.
237. 8:30BORA NIKOLIC: But that's not how people taught it.
238. 8:33Dan, you taught the class back then.
239. 8:35DAN GARCIA: I did.
240. 8:36I did.
241. 8:37BORA NIKOLIC: So things now, parallelism
242. 8:38is in the forefront.
243. 8:39And what we are going to see going forward,
244. 8:43we are going to see specialization,
245. 8:45a different kind of parallelism.
246. 8:46We are going to have these units that
247. 8:48are really good at their stuff.
248. 8:51It's not just CPUs, and GPUs, and everything else.
249. 8:56There is going to be a whole bunch of different units
250. 8:58that are really good for their particular stuff,
251. 9:01for particular things.
252. 9:02We need to figure out how to run them all together concurrently
253. 9:06and efficiently.
254. 9:08And we need to program them really efficiently.
255. 9:10And that's a pretty big thing, pretty big deal,
256. 9:14how to do that.
257. 9:17The fifth big principle is about the performance measurement
258. 9:22and improvement.
259. 9:24And this is one of the first--
260. 9:26we didn't tell you that early on.
261. 9:28But that was one of the big Patterson's contributions
262. 9:31from late '70s.
263. 9:33Before that, building computers was like a black magic.
264. 9:37And the first edition of Patterson, Hennessey's book
265. 9:42actually set up the right way of thinking about computers.
266. 9:45First, let's try to figure out how do we evaluate performance.
267. 9:50And then let me know what we are dealing with.
268. 9:53We know what to improve on.
269. 9:55And there, what we talked about in the class
270. 9:59is just IPCs and the iron law of computing
271. 10:06but also different ways of how we measure performance.
272. 10:10Because taking just one angle of measuring
273. 10:13the performance often gives us a wrong idea
274. 10:18how the actual system is going to perform
275. 10:21from user's point of view.
276. 10:23And we mentioned this multiple times.
277. 10:25AMD is known for beating all the benchmarks.
278. 10:30And users don't like it in the end.
279. 10:32They've gotten a lot better over time.
280. 10:34But this is something that we are going to see,
281. 10:37especially going forward in the era of specialization.
282. 10:41We have the Apple M1.
283. 10:48It's a different system.
284. 10:49And it is going to be doing some things better.
285. 10:51But some other things may not be.
286. 10:53Some applications may not be running quite well.
287. 10:57And then the sixth big idea we just touched on
288. 11:00is dependability by redundancy.
289. 11:03It is important.
290. 11:04It is what makes our systems run.
291. 11:07And we can rely on them for doing really,
292. 11:10really important things, like health care, and life support,
293. 11:14and finances, and everything else.
294. 11:16DAN GARCIA: And self-driving cars.
295. 11:17And self-driving cars.
296. 11:18When there's a mud splotch on my-- on sensor on the right,
297. 11:21does my car continue to drive?
298. 11:22It does.
299. 11:23How does it do that?
300. 11:24All of that.
301. 11:26BORA NIKOLIC: Now, that also is entangled with everything else.
302. 11:30I mean, we need to--
303. 11:32there is this really, really important thing,
304. 11:35where we need to figure out how do we measure
305. 11:37how dependable our system is.
306. 11:40We have put some numbers over there.
307. 11:41But that's something-- if you look
308. 11:43at all these ASIL standards, people
309. 11:45are trying to matricise that.
310. 11:46How often are we allowed to have our self-driving engine fail?
311. 11:53We are going to see that--
312. 11:56we are talking about quantum.
313. 11:59Those are going to fail a lot more often, at least the way
314. 12:02how we see them now.
315. 12:03They're going to be less reliable than what we have now.
316. 12:06We'll have to live with that.
317. 12:08They're going to be wonderful.
318. 12:09But they're not going to work all the time.
319. 12:12This is going to get to the forefront.
320. 12:15That's it.
321. 12:16I'm going to switch over to Dan.
322. 12:18DAN GARCIA: Sure.
323. 12:19And what's interesting is I remember
324. 12:21Dave has this wonderful line from point 5, which
325. 12:23was great metrics yield great performance.
326. 12:27Great metrics yield great improvement.
327. 12:29If you don't have a metric--
328. 12:30my user interface is better than your user interface.
329. 12:32Really?
330. 12:33On what metric?
331. 12:34There is no metric.
332. 12:34It's a soft, squishy one.
333. 12:35And that's why it's hard to measure
334. 12:37these things that are squishy.
335. 12:38You make a really clean metric.
336. 12:39You can optimize.
337. 12:40And by the way, you know what they're optimizing for now?
338. 12:42This is something probably, Bora, in the last five years,
339. 12:44I think.
340. 12:45Used to be, what's your performance?
341. 12:46What's your spec rating?
342. 12:47Now it's spec rating per watt.
343. 12:49It's, how power-efficient are you?
344. 12:51That's a whole new way of thinking.
345. 12:53How efficient are you per watt?
346. 12:54Everything's mobile.
347. 12:55Everything's on the laptop.
348. 12:57Certainly, the Mac Pro plugged in,
349. 12:58who cares how many watts you're using?
350. 13:01But if you're doing anything else-- laptop, mobile, watch--
351. 13:04you need to have performance per watt.
352. 13:06So that's a big new idea way of thinking that wasn't there,
353. 13:08I think, five years or maybe even three years ago.
354. 13:11It was just raw performance.
355. 13:12BORA NIKOLIC: Yeah, and that's a huge thing as well, I mean.
356. 13:15And it also needs to be shaped up over time.
357. 13:18And it's not just instructions per watt now.
358. 13:21It is often some particular task.
359. 13:24How many facial recognitions can we do per watt and per second.
360. 13:31And that's a pretty big deal.
361. 13:33When we are looking at a phone, that facial recognition engine
362. 13:39has 20 milliseconds to recognize you.
363. 13:42Otherwise, you are not going to like your phone anymore.
364. 13:45You're going to give up on that.
365. 13:4720 milliseconds under a limited power budget
366. 13:49is a pretty challenging compute task, a lot more challenging
367. 13:52than project 4.
368. 13:54[CHUCKLES] But project 4 does lay a good foundation
369. 13:59for building something like that.
370. 14:01DAN GARCIA: Sure.
371. 14:02Let's keep on moving.
372. 14:03Good stuff.
373. 14:04BORA NIKOLIC: Yeah, you go.
374. 14:04DAN GARCIA: So this is--
375. 14:05I mean, Bora talked a lot about this.
376. 14:07This was the picture I used to teach when I taught this
377. 14:09in 2001 with Dave Patterson.
378. 14:11This was our course--
379. 14:13machine structures, the old school model, just thinking
380. 14:15about a single machine--
381. 14:16a single core, even--
382. 14:17and trying to look at all the layers of abstraction.
383. 14:20We talked about great idea number one.
384. 14:21With the ISA being the most important,
385. 14:23we spent a lot of time talking about the design
386. 14:26of this new, wonderful, open ISA that we have,
387. 14:28that we've learned about, that our team, Berkeley's team,
388. 14:31has contributed to.
389. 14:32Just love that.
390. 14:33So we spent a lot of time thinking about this slide.
391. 14:36We also talked about this slide, which is--
392. 14:39it's a lot more complicated.
393. 14:41You've got not one machine but multiple machines.
394. 14:44And every machine has multiple cores.
395. 14:46How do they talk?
396. 14:46How do you have cache coherency?
397. 14:48What are the issues?
398. 14:49What's the fourth C?
399. 14:50Parallelism came in at many layers.
400. 14:52We've got parallelism with requests, high level.
401. 14:55We talked about thread level parallelism,
402. 14:57parallelism instructions through pipelining, parallel data.
403. 15:00Let's do some SIMD and do things all at once.
404. 15:02And hardware descriptions all operating at the same time.
405. 15:04So we talked about parallelism, the many, many layers.
406. 15:06And I think all of you are fluent in all five layers.
407. 15:09You can imagine a future project 4 might involve somebody
408. 15:12touching every single one of them,
409. 15:13some project that actually involves if you really
410. 15:15optimize it, you touch on every single one of those.
411. 15:18And that would be really exciting.
412. 15:19Maybe future TAs could think about a project that actually
413. 15:21makes them-- to get that number, to get the top of the score,
414. 15:24you actually have to do something explicitly
415. 15:26about that in a clever way.
416. 15:27BORA NIKOLIC: May make you wish to take 61C again.
417. 15:30DAN GARCIA: Yeah, right?
418. 15:31BORA NIKOLIC: By the way, we have a hundred people now.
419. 15:35DAN GARCIA: Hey!
420. 15:36We got [INAUDIBLE]---- yeah, that's great.
421. 15:37Wonderful, wonderful.
422. 15:38BORA NIKOLIC: Ah, it was a hundred a minute ago.
423. 15:40DAN GARCIA: Yeah, yeah, yeah.
424. 15:41It came out just before-- just as you said it.
425. 15:44And this is also a big idea here.
426. 15:47We have connected-- okay, I'll tell the story.
427. 15:53This was the first time.
428. 15:54My own version of this class I took a hundred years ago,
429. 15:561980-- boy, man, 1986.
430. 15:58I'm dating myself.
431. 15:59Bora, you're probably similar.
432. 16:01You and I are the same age,
433. 16:02I remember learning this, and sitting in class,
434. 16:05and realizing-- it was called 6.004 at MIT.
435. 16:08It's basically 61C.
436. 16:09They are almost identical classes.
437. 16:11And I remember saying this is the first time I understand
438. 16:14it all.
439. 16:14This is the first time I understand
440. 16:16the connection between how software works at the highest
441. 16:19level.
442. 16:20Here's C. And above that, here's Python and things above that.
443. 16:23Here's something.
444. 16:24Here's Snap! above that.
445. 16:26All the way down to--
446. 16:27BORA NIKOLIC: Dan, there was no Python back then.
447. 16:28DAN GARCIA: I know, right?
448. 16:29Right.
449. 16:30Python was only on the TV show, Monty Python.
450. 16:34It was a gleam in Guido's eye.
451. 16:36So it came down.
452. 16:37I assembled down to machine code.
453. 16:40Now I have an executable program, just 1's and 0's.
454. 16:42How do you run that?
455. 16:43At the point of me walking into that class--
456. 16:44and hopefully, you walked into this class,
457. 16:46you didn't know what that was.
458. 16:46Those 1's and 0's meant nothing to you.
459. 16:48Now they do.
460. 16:49You could literally-- if I give you
461. 16:51on the final exam 1's and 0's from a final executable,
462. 16:54a simple thing, you could actually
463. 16:56go all the way back up to C and understand what
464. 16:58the heck that means.
465. 16:59And you could also build a machine that does that.
466. 17:01And you did that in project 3.
467. 17:03So I think of this where it's almost
468. 17:06like the two connections between hardware and software
469. 17:09are connected.
470. 17:10And I want people--
471. 17:11I want the 101 of you to actually-- if you can,
472. 17:14get ready here.
473. 17:15I want you to clear out.
474. 17:16Step back from your computer.
475. 17:17Nobody else is on the screen except for Bora and myself.
476. 17:20I'm going to try to do this if I can.
477. 17:21I got a background.
478. 17:23I got to do this thing.
479. 17:24I want you to stand up.
480. 17:25I want everybody to stand up.
481. 17:27Here, let's do it.
482. 17:27Everybody stands up.
483. 17:28This is a moment.
484. 17:30This is a moment for celebration.
485. 17:31BORA NIKOLIC: I'll zoom out.
486. 17:31I'll zoom out.
487. 17:32DAN GARCIA: I can't zoom out.
488. 17:33BORA NIKOLIC: I'm zoomed out.
489. 17:34DAN GARCIA: Bora has more--
490. 17:34Bora has higher and higher.
491. 17:35I'm going to show you my chest.
492. 17:37That's all I got.
493. 17:38I'm standing up.
494. 17:39I'm standing up, okay?
495. 17:40BORA NIKOLIC: I can wave.
496. 17:40DAN GARCIA: Everybody standing up.
497. 17:42Here we go.
498. 17:43This is the moment where you made hardware and software
499. 17:46connection, where the connection between hardware and software
500. 17:50comes together.
501. 17:51[IMITATES ANGELIC SINGING]
502. 17:53The first time ever.
503. 17:54I want you to call home.
504. 17:56Get your phone out.
505. 17:57Call home.
506. 17:58We've done it!
507. 17:59Celebration!
508. 18:00[MUSIC - LUDWIG VAN BEETHOVEN, "SYMPHONY NO.
509. 18:029 IN D MINOR, OP.
510. 18:03125"]
511. 18:04Woo!
512. 18:05Do a dance.
513. 18:06Do a dance.
514. 18:07[CHOIR SINGING GERMAN]
515. 18:18We did it.
516. 18:23Woo.
517. 18:32Woo.
518. 18:33Thank you so much.
519. 18:35BORA NIKOLIC: And zoom back in.
520. 18:36DAN GARCIA: Zoom back in.
521. 18:37Now, I will say this is all weird.
522. 18:40And it's all virtual.
523. 18:41So you don't get the equivalent.
524. 18:43But when we would do this, Bora, we'd do this live,
525. 18:46we asked the entire class to stand up,
526. 18:49and walk around, and dance with us around the class.
527. 18:52And we would have everybody--
528. 18:53I just want you to remember that moment,
529. 18:55that there's never been a time before where you've really
530. 18:58been able to connect that.
531. 18:59And the moment that you actually finish, the last tie when
532. 19:03you connect the two railroads, the last tie that was connected
533. 19:05was when you figured out the control logic
534. 19:08to make the RISC-V machine work.
535. 19:10That was the last-- the moment that hit,
536. 19:12that's when that connection made, where you could actually
537. 19:14teach it from the top down and the bottom up, and build it
538. 19:17there, and tell your friends.
539. 19:18So I encourage you to try to teach your friends how
540. 19:20this works, to share some of the wonders of it.
541. 19:23All right.
542. 19:24BORA NIKOLIC: And there is a benefit
543. 19:25of being remote this semester.
544. 19:26We didn't make you call home right away at that point.
545. 19:29[LAUGHTER]
546. 19:30Otherwise, we would have.
547. 19:31DAN GARCIA: Right.
548. 19:32That was a big deal.
549. 19:32It's hard to do to this asynchronously
550. 19:34when we make a video and ship it off to you.
551. 19:35We don't know when these things happen.
552. 19:36But nice to have a shared experience.
553. 19:38These shared experiences are definitely missed sometimes.
554. 19:41All right.
555. 19:43We've got a pretty wide open calendar here, folks.
556. 19:46You've got two weeks.
557. 19:47This is a Friday.
558. 19:48This is the summary lecture.
559. 19:50You've got all the rest time to put things together,
560. 19:53to get yourself gathered together.
561. 19:55You have a unique exam situation.
562. 19:58The exam situation is going to be the same
563. 20:00as it was in the midterm.
564. 20:02We give you 24 hours.
565. 20:02We've already announced.
566. 20:03This is nothing new.
567. 20:04It's going to be three sections-- a Quest,
568. 20:06a midterm, and a final.
569. 20:08Roughly a 1, 3, 2 kind of a model in terms of the number
570. 20:12of half hour segments, where the first question will be about
571. 20:15half an hour-ish, the midterm probably 90 minutes-ish--
572. 20:18we definitely leave room for flex--
573. 20:20and the final section for about an hour-ish, totaling up
574. 20:24to three hours.
575. 20:25And you can take it any time over 24 hours.
576. 20:28And the same recording rules--
577. 20:29I'm seeing a question.
578. 20:30Same recording rules is there.
579. 20:31Nothing changed.
580. 20:32Do everything the same.
581. 20:33But only three sections you have to worry about, and not two.
582. 20:36BORA NIKOLIC: And we are going to have-- we are here to help.
583. 20:39DAN GARCIA: And points definitely-- points
584. 20:40distributed similar to the time limits.
585. 20:41That's exactly right, yes.
586. 20:43BORA NIKOLIC: Yeah, we are here to help during the RRR week.
587. 20:45There will be extra office hours.
588. 20:49We will hold them.
589. 20:51There will be a dense offering of office hours.
590. 20:56So come armed with your questions.
591. 20:59I'll definitely hold an office hour.
592. 21:01I can even try to hold it in person,
593. 21:03but it's limited to 12 people.
594. 21:06I don't know how is that going to work.
595. 21:08But I can try.
596. 21:09DAN GARCIA: I may be with my nose
597. 21:11to the grindstone working with CeCe on the team
598. 21:13to get the PrairieLearn.
599. 21:14So I may not offer that many office hours.
600. 21:16But I definitely do know that we care about you.
601. 21:18And we are putting our hours into where
602. 21:20they need to be spent to really make
603. 21:21sure we give it a very nice beta-tested exam for you guys,
604. 21:24definitely.
605. 21:25And you have to complete it in three hours.
606. 21:27A question from Josh-- yes, you have
607. 21:29to complete it in three hours, but not in a three-hour window.
608. 21:31Three hours overall time-- like, this plus that plus that
609. 21:34is three hours.
610. 21:35And how you divide those three sections up is there.
611. 21:38Good stuff.
612. 21:38Good question.
613. 21:40All right.
614. 21:42So we said this before.
615. 21:44And by the way, I ought to make this clarification.
616. 21:47Somebody got stumbled one time.
617. 21:49I was really embarrassed when they said, oh yeah,
618. 21:51I aced the midterms.
619. 21:52That means I don't need to take the final, right?
620. 21:54The midterm part of the final, I didn't take it.
621. 21:55I just left it blank because I aced it, right?
622. 21:57No!
623. 21:57No.
624. 21:58No, these are harder questions.
625. 21:59You've got to take that whole final.
626. 22:01So everyone takes the whole final.
627. 22:04Even if you aced Quest or midterm,
628. 22:05you've got to take the whole final.
629. 22:06Don't make that mistake.
630. 22:08That would be embarrassing.
631. 22:10And if you did well in any class--
632. 22:14and this is for 10-- this second part of the slide,
633. 22:16I talk about every class I teach.
634. 22:18If you did well, please give back.
635. 22:20The usual path is academic intern, tutor, or reader,
636. 22:22and then TA, and then head TA, and then summer instructor.
637. 22:26The highest pantheon of elevation
638. 22:28of a student instructor is the summer instructor.
639. 22:30So some of our students have done that this semester.
640. 22:33And it's been great to work with them.
641. 22:34So--
642. 22:35BORA NIKOLIC: And this is--
643. 22:36DAN GARCIA: Go ahead.
644. 22:36BORA NIKOLIC: --really, really important.
645. 22:37We really rely on you guys.
646. 22:39I mean, if you have done well, if you like this class,
647. 22:42join the teaching crew.
648. 22:43DAN GARCIA: Yeah, definitely.
649. 22:45BORA NIKOLIC: This semester, we went with a lot of new TAs
650. 22:48because the previous TAs, a set of lots of TAs
651. 22:51that were around before, graduated.
652. 22:54And guess what?
653. 22:54In May, another round of former TAs are going to graduate.
654. 22:59We need new ones.
655. 23:00DAN GARCIA: Right, right.
656. 23:00BORA NIKOLIC: And that's really important.
657. 23:02Also, it's a great experience.
658. 23:03I mean, one thing is to take a class.
659. 23:09A completely different level of understanding
660. 23:11is to actually teach someone else how to do things.
661. 23:16And you [INAUDIBLE].
662. 23:17DAN GARCIA: You might think you know something
663. 23:17until you try to teach it.
664. 23:18[LAUGHS]
665. 23:21BORA NIKOLIC: Yep, exactly.
666. 23:25DAN GARCIA: So there's two questions that are coming up.
667. 23:27I'll just answer live.
668. 23:30Do you have to start at a 61C AI and then jump to this?
669. 23:33But if you have AI experience-- if you
670. 23:35have lots of experience for another class,
671. 23:37sometimes we can jump into that.
672. 23:38You can be jumped into a paid position of a tutor reader
673. 23:42without having to go through 61C AI.
674. 23:43If you have no experience, you have to go through
675. 23:45at least an AI experience.
676. 23:47But if you do have experience, it's
677. 23:48a good question, anonymous attendee, for that one.
678. 23:51And exam strictly timed this time.
679. 23:53That means, again-- actually, the exam is same
680. 23:56timed as it was in the midterm.
681. 23:57It's just that we're not going to write an exam that's
682. 23:59much longer than hopefully an hour and a half
683. 24:01for the TAs and the three hours.
684. 24:03So basically, it's even the same exam style
685. 24:05as the midterm, same idea, but basically a smaller scope.
686. 24:08You don't have, I think, whatever, six hours
687. 24:09or something for the midterm.
688. 24:11Now it's definitely three max, three max
689. 24:12divided into three pieces.
690. 24:14All right.
691. 24:16I love this.
692. 24:17I'll do my again--
693. 24:19(IMITATING GODFATHER) the Godfather
694. 24:20answers all of life's questions.
695. 24:22You come to us.
696. 24:22The reason the Godfather answers all life's questions
697. 24:25is a wonderful meta point here.
698. 24:29But what we mean by this is you need somebody to vouch for you.
699. 24:32You need somebody to tell--
700. 24:33if you're going into research, somebody's got to say,
701. 24:36this person is good.
702. 24:37You go into a job.
703. 24:38Sometimes on a job you don't need
704. 24:39to vouch as much because you can be in an interview.
705. 24:41And you self-vouch by showing off
706. 24:43that you know the stuff in an interview.
707. 24:45But most of the time, if you go to the next level of academics,
708. 24:48you need to have somebody tell you that they believe.
709. 24:51I'm from a position of authority.
710. 24:52And this person is a rock star.
711. 24:54Somebody's got to tell you that.
712. 24:55And so the way you do that is get involved in research.
713. 24:58And that's one of the reasons that Berkeley
714. 25:00is the top university in the world, or at least one of them.
715. 25:02So--
716. 25:03BORA NIKOLIC: And you--
717. 25:04DAN GARCIA: Go ahead, Bora.
718. 25:04BORA NIKOLIC: --heard the pretty strong pitch
719. 25:06from Dave Patterson.
720. 25:10Go to grad school.
721. 25:11I mean, if you can.
722. 25:12DAN GARCIA: Yeah, definitely.
723. 25:13BORA NIKOLIC: If you can, by any means.
724. 25:15Don't stop with a bachelor's degree.
725. 25:17I mean, you can.
726. 25:18Nobody will fault you at that.
727. 25:20But go for it.
728. 25:23The thing there is it becomes a lot more personal.
729. 25:26Instead of having a thousand people,
730. 25:28it's your advisor and a few people.
731. 25:31And it's a big commitment for that advisor to take you on.
732. 25:36So somebody has to vouch for you, like Dan and I.
733. 25:41DAN GARCIA: And they know you great.
734. 25:43I mean, they know you for years.
735. 25:44I mean, I just love writing letters
736. 25:46for people I've worked for for several years,
737. 25:48either in GamesCrafters or other things that I've done,
738. 25:51because I can say--
739. 25:51I can speak-- it's not just, well, they've taught for me,
740. 25:53but they've done this research.
741. 25:54They've done these things.
742. 25:55There's a lot of elements to a person
743. 25:56that you can really boast about.
744. 25:57This person has dedication, and passion, and blah blah.
745. 26:00And I can just continue to talk about that.
746. 26:02If they're a student--
747. 26:03in fact, if you ask me for a letter,
748. 26:04I won't give you a letter if you're my student.
749. 26:06I don't know you well enough.
750. 26:07Only when you go through my teaching,
751. 26:09research, or service activities do
752. 26:10I get to know you enough to write a letter that means you.
753. 26:12Otherwise, it's a meaningless letter.
754. 26:13It's called actually a weak letter.
755. 26:14And weak letters hurt your case more than no letter at all.
756. 26:17So I always say no to any student.
757. 26:18But I say yes to every former person
758. 26:21who worked with me in a teaching, research,
759. 26:22or service role as well.
760. 26:26All right.
761. 26:27So do you want to take this one over, Bora?
762. 26:31BORA NIKOLIC: Yeah, of course.
763. 26:33I mean, I'll paraphrase this a bit and butcher it, possibly.
764. 26:39Dave would say this always better than I would.
765. 26:45There are many opportunities in this field.
766. 26:50So we'll talk a little bit more about that.
767. 26:56For huge majority of students in the class,
768. 26:59this was the entry into the systems area
769. 27:01of computer science or what's commonly
770. 27:04known as computer engineering.
771. 27:06How do we make systems work?
772. 27:09There is a bunch of upper division classes
773. 27:11that will help you understand that.
774. 27:13And there are many career opportunities
775. 27:17that are out there.
776. 27:22Dave Patterson and John Hennessey won the Turing Award
777. 27:27three years ago.
778. 27:28And they made a point that this is a golden time,
779. 27:32golden opportunity, golden era of computer architecture.
780. 27:37These golden eras happen from time to time
781. 27:40when we have a big transition to a new paradigm.
782. 27:43And we are now transitioning to this domain of specialization.
783. 27:49Somebody has to do it.
784. 27:51We'll have 1,200 candidates in this class that
785. 27:56can participate in that.
786. 27:58But it probably needs many more people than that.
787. 28:01There are two directions that Dave mentions.
788. 28:03One thing, there is a huge gap now
789. 28:09between what we call productivity programming
790. 28:12and performance programming.
791. 28:14So far, before this class, you have
792. 28:16learned about productivity.
793. 28:18But in this class, you have encountered
794. 28:21what performance means.
795. 28:23And by the way, a side benefit of that, whenever
796. 28:25you see a job listing or internship
797. 28:30listing for performance programming,
798. 28:33you're most likely already qualified for that.
799. 28:36I mean, you'll get more qualified
800. 28:37as you take undergrad-- more upper div classes.
801. 28:41But you're already good enough to hit those internships.
802. 28:46One of the big things is that this gap
803. 28:48between productivity and performance
804. 28:50is growing every year.
805. 28:52And it's the gap between high level applications,
806. 28:57frameworks-- even Python down to C is huge.
807. 29:00It's 1,000x in [INAUDIBLE].
808. 29:02And we have seen that.
809. 29:04One of the big challenges is make productivity frameworks
810. 29:10run as fast as performance.
811. 29:11I think that Dave has been quoted, whoever does
812. 29:14that will win a Turing Award.
813. 29:17It's not an easy thing to do.
814. 29:19And then the other thing is about this domain
815. 29:23specialization and domain-specific applications.
816. 29:27Building hardware, building efficient compilers
817. 29:30for that and efficient frameworks,
818. 29:32it is something that is pretty hard.
819. 29:36And we have seen our glimpse of that.
820. 29:40You have seen a little bit how that is done.
821. 29:43The first successful coprocessor was a floating point processor.
822. 29:48That's the first accelerator.
823. 29:50Then we are seeing vectors.
824. 29:52And your vector accelerators, you're using them in the class
825. 29:57now.
826. 29:57And I'm hoping you're having fun.
827. 30:00But there, in project 4, you're actually
828. 30:03using accelerator using relatively
829. 30:06low-level programming.
830. 30:09These high-level frameworks do help.
831. 30:12But they don't always find the right thing.
832. 30:15And they don't compile always quite the right way.
833. 30:19There is a big challenge of doing that.
834. 30:21And what we are going to see more and more of these classes
835. 30:24of accelerators showing up.
836. 30:26And frameworks that are going to try to hit them
837. 30:28are going to be really important.
838. 30:31DAN GARCIA: It's tough.
839. 30:32Nice summary.
840. 30:32Nice summary.
841. 30:33Dave would be proud.
842. 30:34That was great.
843. 30:35Do you want to talk about this one too?
844. 30:36Yeah.
845. 30:37BORA NIKOLIC: Yeah, we are seeing
846. 30:38a glimpse of this showing up.
847. 30:42We are seeing these systems on a chip that have not just--
848. 30:45CPUs are not going anywhere.
849. 30:47Whoever is telling you that the computing is going
850. 30:49to completely change and we are not
851. 30:51going to see this general purpose course,
852. 30:54I would love them--
853. 30:56ask them to show you how are we going
854. 30:59to program bare metal, that kind of stuff, and never live with--
855. 31:03never use an operating system again.
856. 31:05Good luck with that.
857. 31:06Good luck with user acceptance of that kind of a thing.
858. 31:10But we are going to see these systems that
859. 31:13have incredible performance when they are ran properly,
860. 31:17when there are going to be multiple processor cores, GPUs,
861. 31:20accelerators, interfaces, memory systems of different kinds.
862. 31:24And we use the picture here of Apple A13 Bionic.
863. 31:27But M1 chip is an example of that.
864. 31:30And we are going to see more of those.
865. 31:32And by the way, one funny thing about this class--
866. 31:36we use this Anandtech often as a source of our pictures
867. 31:40and so on.
868. 31:40And we brought you Anand live to--
869. 31:44DAN GARCIA: No other class that I've ever seen.
870. 31:46No other class in the history of Berkeley--
871. 31:47BORA NIKOLIC: Yeah, never happened before.
872. 31:48DAN GARCIA: --as far as I know.
873. 31:49It's true.
874. 31:50BORA NIKOLIC: Yeah, and he's not a 14-year-old kid anymore.
875. 31:52[LAUGHTER]
876. 31:54He's hiring.
877. 31:56And I think he's the one that had a few of our undergrads.
878. 31:59DAN GARCIA: That's great.
879. 32:00That's great.
880. 32:01BORA NIKOLIC: Things are going to go in this.
881. 32:04And what we are seeing now--
882. 32:05I mean, you have been following these news.
883. 32:08Amazon has now their own systems on a chip.
884. 32:11They have been announced to be in the cloud.
885. 32:14Apparently, they have the client one.
886. 32:16So Alexa is going to start speaking, is
887. 32:19going to get custom hardware.
888. 32:21(IN A STAGE WHISPER) And there are rumors
889. 32:22that it may be this time.
890. 32:23We don't know.
891. 32:24DAN GARCIA: Woo.
892. 32:25Woo.
893. 32:26That would be nice coup.
894. 32:27That'd be a great coup.
895. 32:28BORA NIKOLIC: And this is going to be all over the place.
896. 32:31And they're all going to be different.
897. 32:32Somebody has to figure out how to work with all of that.
898. 32:36DAN GARCIA: Outstanding.
899. 32:37Outstanding.
900. 32:38Good stuff.
901. 32:39So a little taste of getting involved
902. 32:42with some of the projects I'm working on--
903. 32:43as you know, I'm passionate about CS for all.
904. 32:46I hope you know that.
905. 32:47I care about bringing computer science
906. 32:49to people who are really just passionate about learning,
907. 32:53learning CS who may not have had the resources to get there.
908. 32:55So I do have some projects to improve CS10
909. 32:59and to improve Snap! the language.
910. 33:00And if you're interested, I have a project manager
911. 33:03of BJC, who is Lauren Mock.
912. 33:04She's our former head TA.
913. 33:05Just shoot her an email saying I want
914. 33:07to join one of Dan's groups.
915. 33:08What do you have?
916. 33:08And then she'll cue up your names.
917. 33:09Then we'll have a big meeting at the beginning of next semester
918. 33:11and do that.
919. 33:12I also have had this project I've
920. 33:13been working on since I was--
921. 33:15gosh, I was a sophomore.
922. 33:16I was a sophomore in college.
923. 33:18And I created this game theory tool called GAMESMAN.
924. 33:21And I've continued to keep it rolling.
925. 33:22And it was my master's project.
926. 33:23And I have now a group that I've been running for now 20 years.
927. 33:27Started it in 2001.
928. 33:28So next year is going to be the 20th anniversary
929. 33:30of GamesCrafters.
930. 33:31So please do that if you care about game theory.
931. 33:33We call it computational game theory,
932. 33:34using computer horsepower, i.e. parallelism-- nudge, nudge,
933. 33:37wink, wink, say no more, say no more--
934. 33:39to try to attack game theory trees.
935. 33:43Bora.
936. 33:44BORA NIKOLIC: And we do also have to do development first.
937. 33:47I mean, 61C is--
938. 33:50it's never twice the same class that is being taught.
939. 33:54I would love to change some of the infrastructure that we run.
940. 33:58As the RISC-V harder becomes more available,
941. 34:03we may go for changing our compute infrastructure to be
942. 34:07RISC-V--
943. 34:07DAN GARCIA: Oh, that'd be so good.
944. 34:09That'd be so good.
945. 34:10BORA NIKOLIC: Yeah.
946. 34:11Right, you can imagine building your own core.
947. 34:13And then fast forward, you're running software on that core--
948. 34:16DAN GARCIA: Oh, wow.
949. 34:17Wow.
950. 34:18BORA NIKOLIC: --and the accelerators and so on.
951. 34:19It'll be a few years.
952. 34:20But we will need a big effort to put that together
953. 34:25in the meantime.
954. 34:27I did mention this before.
955. 34:29But now this is an open call.
956. 34:31We do have a bunch of research projects.
957. 34:35And we do involve undergrads.
958. 34:37And why am I--
959. 34:38and the next three slides are pretty important.
960. 34:40My graduate students and senior undergrads
961. 34:44are telling me nobody is knocking on their lab doors.
962. 34:47Well, perhaps there is nobody in Soda,
963. 34:50and the lab doors are closed, and there is nobody inside.
964. 34:53It's a lot harder to make a connection.
965. 34:55I'm trying to make a connection.
966. 34:57And this is open call.
967. 34:58If you think you're doing really well in this class
968. 35:01and then you have it, you can be involved in research.
969. 35:06Please join us.
970. 35:08And here is stuff that we do.
971. 35:10My group is a little different than the others.
972. 35:12But we build a ton of chips, a lot of chips.
973. 35:15And guess what?
974. 35:16Pretty much every single one of them has a RISC-V--
975. 35:19or every single one in this picture
976. 35:21has a RISC-V home core in there.
977. 35:22DAN GARCIA: Wow.
978. 35:23BORA NIKOLIC: Look where they date, to May 20,
979. 35:26and that was a Raven-1 chip.
980. 35:28Had the first RISC-V core.
981. 35:30Remember, RISC-V ISA was formed by--
982. 35:35the first draft was in August of 2010.
983. 35:39Six months later, we taped out the chip based on that.
984. 35:43DAN GARCIA: Wow.
985. 35:43BORA NIKOLIC: The chip didn't work.
986. 35:45Its internal codename was a train wreck
987. 35:48because it was just--
988. 35:49DAN GARCIA: [LAUGHS] Oh, that's brilliant.
989. 35:50You're learning.
990. 35:51BORA NIKOLIC: But that is what the reason--
991. 35:51that's a beauty.
992. 35:52I mean, it basically was like a 61C project.
993. 35:56I mean, it was built--
994. 35:59DAN GARCIA: That's great.
995. 36:00BORA NIKOLIC: --pretty much with an understanding
996. 36:02that what you guys have now.
997. 36:04But we get better over time.
998. 36:06So the thing that you have in the corner over there, EAGLEX,
999. 36:10is a 20-core RISC-V pretty capable--
1000. 36:12DAN GARCIA: Wow.
1001. 36:13Wow.
1002. 36:14BORA NIKOLIC: --system that booted Linux and has all kinds
1003. 36:17of other things on it.
1004. 36:19And you can play with it.
1005. 36:20And we need-- some of our current TAs are helping us out,
1006. 36:23write the applications for it.
1007. 36:25But you can run--
1008. 36:27write applications through its lab applications
1009. 36:30for that particular processor.
1010. 36:32Next slide.
1011. 36:33DAN GARCIA: That's great.
1012. 36:34Boop.
1013. 36:36BORA NIKOLIC: A few big things--
1014. 36:37I'm associate affiliated with a few research labs.
1015. 36:40And we do traditionally engage quite a few undergrads.
1016. 36:46At that lab, has this big infrastructural project
1017. 36:49that is called Chipyard.
1018. 36:51We have a whole bunch of--
1019. 36:54a good collection of RISC-V cores and tools
1020. 36:58that go along that.
1021. 36:59And there is an automated software framework that
1022. 37:03lets you build your own chips.
1023. 37:04You can try those chips.
1024. 37:05You can build them on an FPGA.
1025. 37:07Or you can build them in an ASIC.
1026. 37:09We typically have undergrads.
1027. 37:11And several of your TAs have been working on that as well.
1028. 37:16Join us, please.
1029. 37:17Next slide.
1030. 37:20Here is the-- what is in there, there is a generator now.
1031. 37:24We make you write these cores from scratch when we start,
1032. 37:29when we build this Chipyard thing.
1033. 37:30We give you cores.
1034. 37:31We have pre-configured cores written in a more abstract way.
1035. 37:36But you can make a single core RISC-V. Or you
1036. 37:39can make a few cores RISC-V. And then
1037. 37:41we have custom accelerators.
1038. 37:42And we are practicing that future
1039. 37:44that we are imagining, where the domain specialization is
1040. 37:47going to be the thing.
1041. 37:49So we make all kinds of things.
1042. 37:50We don't always know what is that going to be.
1043. 37:54Is it going to work well?
1044. 37:56And there is plenty of surprises over there.
1045. 37:58Oh, we thought this was going to work.
1046. 37:59Not really well.
1047. 38:02But that's what the research is.
1048. 38:04So try it out.
1049. 38:05And it is going to help you see the future a little bit
1050. 38:09earlier.
1051. 38:10And that's it.
1052. 38:11Back to you, Dan.
1053. 38:12DAN GARCIA: Beautiful.
1054. 38:13Sure.
1055. 38:13Let's talk about opportunities next semester.
1056. 38:15So if you're thinking about what courses--
1057. 38:17I think that was one question.
1058. 38:18What courses should we take after 61C?
1059. 38:21Here's a little taste of it.
1060. 38:23152 is an obvious course if you like CPU design, memory,
1061. 38:25and architecture in general.
1062. 38:27Please think about that.
1063. 38:28162 is the OS side of that.
1064. 38:30So if you think-- if you appreciate the VMOS and just
1065. 38:32want it on more, and more, and more.
1066. 38:34And I think 162 will really teach
1067. 38:35you to be a better programmer as well,
1068. 38:37even if you don't end up studying systems or OS.
1069. 38:40You'll be a better C programmer.
1070. 38:41BORA NIKOLIC: It's an intense class.
1071. 38:42DAN GARCIA: Yeah, it's really hard.
1072. 38:43But it's great.
1073. 38:44BORA NIKOLIC: These two classes really build up
1074. 38:46on top of what we have done.
1075. 38:47I mean, what we have done follows.
1076. 38:47DAN GARCIA: They build your muscles in that way.
1077. 38:50I offer three DeCals.
1078. 38:51I'll just give you a little taste of them.
1079. 38:53One, I mentioned GamesCrafters before.
1080. 38:54Please do join that.
1081. 38:55This time has changed.
1082. 38:57It's not Monday, Wednesday-- this old slide, 10:00 to 11:00.
1083. 38:59I think it's 4:00 to 5:00.
1084. 39:01Check my website.
1085. 39:01I think it's 4:00 to 5:00 PM next semester.
1086. 39:04So sorry about that.
1087. 39:04There is an iOS DeCal, if you want
1088. 39:06to learn how to program iOS devices, and UCBUGG DeCal,
1089. 39:09if you want to learn how to make Pixar short films.
1090. 39:12That's an amazing course.
1091. 39:13That's also a lot of work.
1092. 39:14That's three units.
1093. 39:15You don't need anything other than 3D interest to do that.
1094. 39:17So that's just great.
1095. 39:19You want to cover this one, Bora?
1096. 39:20BORA NIKOLIC: More on the EE side, many of you
1097. 39:24probably have taken 16A.
1098. 39:26And that's good enough to get you into a lot of entry
1099. 39:30into the EE side of things.
1100. 39:33If you want to get into 120 series, which
1101. 39:37are signals, and systems, and communications,
1102. 39:38you probably should take 16B to get going with that.
1103. 39:43But if you are interested more into circuits
1104. 39:47and physical devices, you can go straight from 16A into 105
1105. 39:52and then dive into 140 series.
1106. 39:54There is also devices that is 130 series.
1107. 39:57The next one I'm teaching--
1108. 39:59I'm guessing I'm going to see many of you in 151.
1109. 40:03So we just got our teaching assignments.
1110. 40:05I'm teaching EECS151, which is Intro to Digital Systems
1111. 40:08and Circuits.
1112. 40:09And that class just basically follows
1113. 40:12exactly where we stopped.
1114. 40:13We are going to repeat some of that--
1115. 40:16there'll be other students who might have not
1116. 40:18seen RISC-V in that class.
1117. 40:20We are going to repeat RISC-V and then just get going, learn
1118. 40:24a new language, Verilog, and build a bigger RISC-V system
1119. 40:27and get it to work.
1120. 40:29It is one thing--
1121. 40:31I often get questions.
1122. 40:32It has two labs.
1123. 40:33And they're separate.
1124. 40:34You can sign up for LA and LB.
1125. 40:37Generally, not a good idea to sign up for both.
1126. 40:40That's seven units, and seven Berkeley units
1127. 40:42is a lot of units.
1128. 40:43I mean, there are students who do that, but do
1129. 40:46think about it before signing up for seven units of heavy lab
1130. 40:50project class.
1131. 40:51Usually, people take five units, so the lectures
1132. 40:54plus one of the labs.
1133. 40:55And then in the next semester, they take the other lab.
1134. 40:59DAN GARCIA: Wonderful.
1135. 41:00I just looked up on the side.
1136. 41:01GamesCrafters is 3:00 to 4:00 PM next spring.
1137. 41:04So think about that if you want to do that.
1138. 41:06This is my slide I created a couple of years ago.
1139. 41:09And I still think it's true.
1140. 41:10With all the whiz bang, oh, there's cloning,
1141. 41:13and TVs on your wristwatch, and all the thing,
1142. 41:15I think the biggest invention that we're
1143. 41:16going to see in our lifetimes is going
1144. 41:18to be the ability to directly connect
1145. 41:19your brain to a machine.
1146. 41:20And I think we've already seen Elon Musk--
1147. 41:22I think one of our Computing in the News,
1148. 41:23if you remember that, Bora, is we brought--
1149. 41:24Elon Musk is trying to do that and play with that exploit.
1150. 41:27That is a very exciting space.
1151. 41:29BORA NIKOLIC: A bunch of our alumni.
1152. 41:31And I think today posted they're hiring again.
1153. 41:33DAN GARCIA: Wow.
1154. 41:34Wow.
1155. 41:34I mean, this is, I think, the hottest, most interesting space
1156. 41:37in terms of future, like blue-sky research.
1157. 41:40Boy, the Matrix is coming, folks, just the ability
1158. 41:43to be able to have input and output,
1159. 41:46to be able to download a dream and upload a dream.
1160. 41:49Blows my mind in terms of science fiction.
1161. 41:50But I think it's going to be here soon.
1162. 41:52So very excited about that.
1163. 41:54And may be able to download all your thoughts of life.
1164. 41:56You have some Albert Einstein level intelligence.
1165. 41:59Download it so you have it kind of a thing, surviving it.
1166. 42:03But wonderful stuff.
1167. 42:03So Jose Carmena is one of our colleagues
1168. 42:06who was working that very, very, very hot research area--
1169. 42:09brain-machine interfaces.
1170. 42:10So please connect with him if that's interesting to you.
1171. 42:13Bora, project 3 awards.
1172. 42:16Ladies and gentlemen, the awards!
1173. 42:17BORA NIKOLIC: We finally--
1174. 42:18we got our first 10 winners.
1175. 42:19Congratulations.
1176. 42:20DAN GARCIA: Woo!
1177. 42:22Congratulations, everybody, for your hard work in project 3.
1178. 42:26Man, oh, man, that was a lot of work to put that together.
1179. 42:29BORA NIKOLIC: So what you won is our--
1180. 42:32each one of you won a pair of Apple AirPods Pro.
1181. 42:36DAN GARCIA: Wow.
1182. 42:37BORA NIKOLIC: We will need your shipping addresses.
1183. 42:40We are going to contact you to--
1184. 42:42if you would like to receive the award.
1185. 42:44We can give it to you in person.
1186. 42:46I mean, maybe.
1187. 42:47We might be able to arrange for that.
1188. 42:49But more likely, we are going to ship it to you.
1189. 42:52And we'll need your address.
1190. 42:53DAN GARCIA: Congratulations.
1191. 42:55BORA NIKOLIC: Congratulations.
1192. 42:55DAN GARCIA: Congratulations to all that group.
1193. 42:56Wonderful, wonderful.
1194. 42:57BORA NIKOLIC: This is spectacular.
1195. 42:58I mean, sometimes we have small prizes, $30 prizes.
1196. 43:03These are a--
1197. 43:04DAN GARCIA: This is a big deal.
1198. 43:05BORA NIKOLIC: We have a couple of dozen of these.
1199. 43:07DAN GARCIA: Again, we've never done this before.
1200. 43:09Usually, we give you a handshake and a certificate.
1201. 43:12We've never given a hundred dollar gift for this.
1202. 43:15Congratulations.
1203. 43:15BORA NIKOLIC: Well, this is $259, $259 each.
1204. 43:18DAN GARCIA: Wow.
1205. 43:19Wow.
1206. 43:21Thank you, Apple.
1207. 43:22And thank you to the students who put their hard work in,
1208. 43:23folks.
1209. 43:24Great job, everybody.
1210. 43:25Great job, folks.
1211. 43:26BORA NIKOLIC: Yep.
1212. 43:27DAN GARCIA: And there are more awards coming I understand,
1213. 43:30Bora.
1214. 43:30BORA NIKOLIC: Yeah.
1215. 43:31So we have more awards for project 4,
1216. 43:34top performance in project 4, and also top performance
1217. 43:36in the class.
1218. 43:37And our top TAs are going to get awards.
1219. 43:41DAN GARCIA: Wow.
1220. 43:41Wow.
1221. 43:42Great job, folks.
1222. 43:43Again, wonderful.
1223. 43:44BORA NIKOLIC: Speaking of--
1224. 43:44DAN GARCIA: All right.
1225. 43:45Speaking of TAs--
1226. 43:46BORA NIKOLIC: Speaking of TAs--
1227. 43:47DAN GARCIA: --let's give thanks to all the hard work
1228. 43:50done by our TAs.
1229. 43:53This is all a set of just thank yous at this point.
1230. 43:55This is like the end of an Oscar speech.
1231. 43:57First, we want to thank Stephan and CeCe.
1232. 43:59They were outstanding head TAs up and down,
1233. 44:02handling all the situations we did.
1234. 44:04Thank you so much for your hard work and dedication
1235. 44:06to this class.
1236. 44:07And this is just pages after pages.
1237. 44:09We have a huge staff.
1238. 44:10So thank you Jonathan, Nikhil, Sunay, Daniel,
1239. 44:13Cynthia, and Kevin, again, for all of this.
1240. 44:15Caroline, Jie, Kimberly, Max, Anson, Taejong, Jerry, Alice,
1241. 44:20Justin, Seah, Ryan, Emily, Ryan, Ivy, Robin, Luke, Avi, and Lu--
1242. 44:27almost done--
1243. 44:28Ryan, Dayeol, and Anthony, thank you so much.
1244. 44:31Just give a hand for our TAs and our staff!
1245. 44:33Amazing group of TAs who were there.
1246. 44:36By the way, our TAs are overworked at a level--
1247. 44:39I don't even want to know.
1248. 44:40I mean, I want to know because I want to fix it.
1249. 44:42But our TAs put in so many extra hours
1250. 44:43that you guys don't even realize.
1251. 44:45So please-- thank you so--
1252. 44:46let's give our thanks to the TAs.
1253. 44:48They really--
1254. 44:48BORA NIKOLIC: It's like Oscar ceremony--
1255. 44:49we need to thank our families, friends, producers.
1256. 44:52DAN GARCIA: Yeah, yeah.
1257. 44:52[LAUGHTER]
1258. 44:53That's right.
1259. 44:53Let me thank my family for letting
1260. 44:55me steal every weekend for the entire semester.
1261. 44:57That's definitely where the-- prize goes to my wife.
1262. 44:58BORA NIKOLIC: Exactly, and staying up till 2:00, 3:00 AM.
1263. 45:01DAN GARCIA: The prize goes to my wife
1264. 45:03for not having a weekend with her for the whole--
1265. 45:05all right.
1266. 45:06I also want to thank-- there's another slide that I think
1267. 45:09might have gotten lost here.
1268. 45:10I want to thank also the tutors and the AIs for our family.
1269. 45:14This is a lot of work by a lot of people.
1270. 45:17This is where the wheels meet the road, all that contact, all
1271. 45:19that face-to-face time you had with your tutors and your AIs.
1272. 45:23Thanks to those folks.
1273. 45:24Great job, everybody.
1274. 45:26Shoo, here we go.
1275. 45:29The future for future Cal alumni--
1276. 45:32what is the future?
1277. 45:34Look at this.
1278. 45:35This new millennium is amazing.
1279. 45:37Ubiquitous computing-- anywhere you
1280. 45:38go, you've got a computer on your phone, in your hand
1281. 45:42as fast as a computer from two generations ago.
1282. 45:45What can you do with it?
1283. 45:46Nanotechnology-- making things smaller and smaller and small.
1284. 45:48The internet of things means that everything
1285. 45:50is an IP address now.
1286. 45:51I have light bulbs in my house that are IP addresses, Bora.
1287. 45:53This is crazy.
1288. 45:54If you told me, Dan, you're going to have light bulbs,
1289. 45:55I was like, come on.
1290. 45:56A light bulb lamp, yes, I have--
1291. 45:57BORA NIKOLIC: Dan, they're not secure.
1292. 45:58They're not secure.
1293. 45:58DAN GARCIA: I know they're not.
1294. 46:00Wait-- why did it go red?
1295. 46:01Thank you so much, hacker.
1296. 46:02So I've got this going on.
1297. 46:04We've got AI.
1298. 46:04The parallel revolution is hitting.
1299. 46:06Domain specialization, as Bora mentioned, exactly that.
1300. 46:10Again, rapid changes in technology.
1301. 46:11Post-PC era-- we talked about the eras of computing.
1302. 46:14Remember the lecture.
1303. 46:15I talked about that, where some of the post
1304. 46:17PC-era, the power of the cloud is in your hand
1305. 46:20and on your watch.
1306. 46:21It's incredible.
1307. 46:22You've got the world's best education at Cal.
1308. 46:24Never give up.
1309. 46:26"The best way to predict the future-- is to invent it!"
1310. 46:28Alan Kay.
1311. 46:29So the future is up to you.
1312. 46:31That's it.
1313. 46:32Folks, I want to thank Bora for the teamwork.
1314. 46:36Bora, I really have enjoyed this year with you,
1315. 46:38my friend, this semester with you.
1316. 46:40It's been a challenge because of all the other things
1317. 46:42that came out.
1318. 46:42But I couldn't imagine somebody better
1319. 46:44to ride the train with you.
1320. 46:46So thank you so much for all the things you've done.
1321. 46:47BORA NIKOLIC: Dan, and thank you as well and to your family
1322. 46:50for letting us borrow you.
1323. 46:51DAN GARCIA: [LAUGHS] Yeah.
1324. 46:53Yours as well.
1325. 46:54BORA NIKOLIC: Lending us to you.
1326. 46:55DAN GARCIA: And yours as well.
1327. 46:56Oh my goodness.
1328. 46:57Let me get out of this.
1329. 46:59Get out here.
1330. 46:59Don't leave yet.
1331. 47:00Don't leave.
1332. 47:00We've got one more--
1333. 47:01let me see.
1334. 47:02I might have--
1335. 47:03Bora, do you have one more thing?
1336. 47:05I might have one more--
1337. 47:06BORA NIKOLIC: Yes, one more thing.
1338. 47:07DAN GARCIA: I might just have one more thing.
1339. 47:10I don't know if many of you folks are fans--
1340. 47:12BORA NIKOLIC: We've never done this before.
1341. 47:13DAN GARCIA: Yeah, we've never done this before.
1342. 47:14This is brand new.
1343. 47:15I don't know if we have fans of the Hamilton soundtrack,
1344. 47:19Disney+ and everything.
1345. 47:20But I might need some help with this from folks who might need
1346. 47:25to-- so TAs, if you can unmute, I might need-- all the TAs--
1347. 47:28Anson, Ryan, Lu, Bora, I need some help here.
1348. 47:30BORA NIKOLIC: Dan, I'm here.
1349. 47:31I'm here.
1350. 47:32DAN GARCIA: But you have to go to--
1351. 47:33BORA NIKOLIC: I'll give it a try.
1352. 47:34I'll give it a try.
1353. 47:34I'll give it a try.
1354. 47:35DAN GARCIA: So I need everybody to give me
1355. 47:37a little bit of a backup here going on.
1356. 47:40So we'll see-- shoo.
1357. 47:42We'll see if this works.
1358. 47:43And there is some Zoom lag.
1359. 47:45So I want to apologize already for--
1360. 47:47BORA NIKOLIC: We have never done this.
1361. 47:48DAN GARCIA: The Zoom lag is going
1362. 47:50to be just very unfortunate.
1363. 47:52And this is much better live in front of the class.
1364. 47:54But we'll try.
1365. 47:55So folks--
1366. 47:56BORA NIKOLIC: Let's give it a try.
1367. 47:56DAN GARCIA: --give this a try.
1368. 47:57TAs, Ryan, Anson, Lu, you've got to unmute.
1369. 47:59You got to say with us.
1370. 48:00They don't know.
1371. 48:00They're like, what do you mean?
1372. 48:01You didn't tell us this.
1373. 48:02I didn't tell anybody.
1374. 48:03Nobody knows what they're supposed to be doing.
1375. 48:05But we're going to give this a shot anyway.
1376. 48:07Bora, at least you've got my back here.
1377. 48:08Here we go.
1378. 48:08All right?
1379. 48:09BORA NIKOLIC: Exactly.
1380. 48:09DAN GARCIA: Ladies and gentlemen--
1381. 48:10BORA NIKOLIC: Let's do it.
1382. 48:11DAN GARCIA: All right, on 3.
1383. 48:12Let me turn the volume down just a little bit so
1384. 48:13I'm not overwhelming myself.
1385. 48:14Okay.
1386. 48:14On 3.
1387. 48:15Here we go.
1388. 48:17Woo.
1389. 48:17[MUSIC - LIN MANUEL MIRANDA, "TEN DUEL COMMANDMENTS"]
1390. 48:20(RAPPING) 3, 4, 5, 6--
1391. 48:21BORA NIKOLIC: (RAPPING) 1, 2, 3, 4, 5, 6, 61C.
1392. 48:24It's the six big ideas.
1393. 48:27Computer architecture.
1394. 48:31DAN GARCIA: Number 1--
1395. 48:32BORA NIKOLIC: Number 1--
1396. 48:32DAN GARCIA: --it is all about the layers of abstraction.
1397. 48:34If you understand, no need for further action.
1398. 48:37BORA NIKOLIC: Number 2--
1399. 48:38DAN GARCIA: If you don't, grab a friend, peer instruction.
1400. 48:40They'll tell you that's Moore's law
1401. 48:42was an exponential function.
1402. 48:43BORA NIKOLIC: Number 3--
1403. 48:44DAN GARCIA: Have a second cache just
1404. 48:46in case the first one is the miss.
1405. 48:48Locality of time and space, this hierarchy,
1406. 48:51with smaller costly faster, always a local copy
1407. 48:54of the bigger, cheaper master.
1408. 48:56BORA NIKOLIC: Number 4--
1409. 48:57DAN GARCIA: If not fast enough, parallel!
1410. 48:59Though pruning nondeterminism can be hell.
1411. 49:02You widen up your registers, accelerate with SIMD,
1412. 49:05or you go multi-core and then you speed it up with MIMD.
1413. 49:08BORA NIKOLIC: Number 5--
1414. 49:09DAN GARCIA: Measure performance for improvement.
1415. 49:11The use of solid metrics, yo, is a movement.
1416. 49:15To get--
1417. 49:15BORA NIKOLIC: Number 6--
1418. 49:16DAN GARCIA: --dependability, add one, two, or three
1419. 49:19backups for redundancy.
1420. 49:211, 2, 3, 4, 5, 6--
1421. 49:22BORA NIKOLIC: 2, 3, 4, 5 6--
1422. 49:24DAN GARCIA: 61C!
1423. 49:24BORA NIKOLIC: 61C!
1424. 49:25DAN GARCIA: Good luck on your exam!
1425. 49:28Woo!
1426. 49:30Folks, what a pleasure it's been teaching this class with you.
1427. 49:34We are so excited for you to go rock star that final exam,
1428. 49:37crush that, get the highest A grade in the history
1429. 49:41of this class, no curve.
1430. 49:42You guys are going to be great.
1431. 49:44You guys are ready.
1432. 49:44We love you guys.
1433. 49:45All right, good stuff.
1434. 49:47Good stuff.
1435. 49:48Thank you again for coming to this last class, everybody.
1436. 49:52BORA NIKOLIC: And we've got to vacate this.
1437. 49:53We have a class after us.
1438. 49:55DAN GARCIA: Yeah, that's true.
1439. 49:55We've got a class right now, actually, Chem 1A.
1440. 49:57So we've got to get out.
1441. 49:58Take care, folks.
1442. 50:00Thanks, everybody.
1443. 50:01BORA NIKOLIC: All right.
1444. 50:01We'll answer the questions in--
1445. 50:03hit me up with the questions next week.
1446. 50:05DAN GARCIA: Perfect.
1447. 50:06All right.
1448. 50:06BORA NIKOLIC: All right, bye.
1449. 50:06DAN GARCIA: Take care, folks.
1450. 50:08I'm writing the exam.
1451. 50:09All right.
1452. 50:09Thanks, folks.
1453. 50:10Bye.
1454. 50:10BORA NIKOLIC: Bye.
1455. 50:11DAN GARCIA: Good stuff.