YouTube2Text

[CS61C FA20] Lecture 21.1 - Pipelining I: Pipelining — Transcript

by CS 61C Departmental · 1,628 words · 297 segments · language en · Watch on YouTube

Full transcript

  1. 0:00[Music]
  2. 0:10hello welcome back to 61c
  3. 0:12we are starting a new module this module
  4. 0:16will deal with pipelining
  5. 0:17so far we have designed the functional
  6. 0:19unit and evaluated its performance
  7. 0:21but we really didn't get a sense is that
  8. 0:23performance
  9. 0:24good or bad and
  10. 0:27is that the the cycle time the only
  11. 0:30measure of performance
  12. 0:32what we'll do in the this unit will
  13. 0:33actually start measuring the performance
  14. 0:35or different ways how we measure the
  15. 0:37performance and then
  16. 0:39figuring out the way how to improve it
  17. 0:42so let's get going
  18. 0:45in our new school of machine structures
  19. 0:48we have looked at different
  20. 0:50layers of hierarchy and how can they
  21. 0:53support
  22. 0:53parallelism now we are going to take a
  23. 0:56look
  24. 0:57at parallelism at the instructional
  25. 1:00level
  26. 1:01and what do we need to do in the
  27. 1:02hardware in the execution units
  28. 1:04to support simultaneous
  29. 1:08execution of instructions
  30. 1:12we are also introducing uh
  31. 1:15one of our six great ideas in computer
  32. 1:18architecture
  33. 1:19which is performance measurement and
  34. 1:21improvement
  35. 1:23we simply what we mean by that is
  36. 1:26we need to know how to measure the
  37. 1:28performance in order to be able to
  38. 1:30improve it so let's get going
  39. 1:36we had one measure of performance which
  40. 1:38was the minimum cycle time that we
  41. 1:40needed to execute an instruction
  42. 1:42in our single cycle risk five cpu
  43. 1:46and we determine the time by trying to
  44. 1:48measure the time that it takes to
  45. 1:50execute
  46. 1:51each of the five phases of execution so
  47. 1:54we looked at the instruction fetch
  48. 1:55and found out the instruction fetch
  49. 1:57takes 200
  50. 2:00picoseconds that instruction decode
  51. 2:03takes about 100 picoseconds
  52. 2:05the alu operation that
  53. 2:10is our execute phase would take another
  54. 2:12200 picoseconds
  55. 2:13memory access will take yet another 200
  56. 2:16picoseconds and finally right back would
  57. 2:18be done we completed in 100 picoseconds
  58. 2:22when you sum all of that up the minimum
  59. 2:24cycle time was
  60. 2:25800 picoseconds now
  61. 2:28not every single instruction goes
  62. 2:31through all five phases of execution
  63. 2:33and we found out that it was just only
  64. 2:35one of them load word that
  65. 2:37went through all five phases of
  66. 2:40execution
  67. 2:41um but we can have a processor that
  68. 2:44can you know we could possibly think of
  69. 2:48a
  70. 2:48trying to clock this processor faster
  71. 2:50but then it will not be able to do
  72. 2:53a load word instruction although it will
  73. 2:55be able to do
  74. 2:56most of the other instructions so
  75. 2:59always our critical path is set by the
  76. 3:02longest
  77. 3:03instruction here that the longest time
  78. 3:06that an instruction takes
  79. 3:08in this case that was 800 picoseconds
  80. 3:10and
  81. 3:12we determined that the maximum clock
  82. 3:14frequency is equal to 1 over that
  83. 3:16those 800 picoseconds and that was 1.25
  84. 3:21gigahertz
  85. 3:24and that was one measure of a
  86. 3:25performance so our single
  87. 3:27cycle cpu can execute instructions at
  88. 3:321.25 gigahertz that means that it can
  89. 3:35start an instruction
  90. 3:36and completed it um within
  91. 3:39800 picoseconds and then can start
  92. 3:41another one that would take also 800
  93. 3:43picoseconds and so on
  94. 3:46it would finish 1.25 billion
  95. 3:51instructions per second
  96. 3:54so the question here is can we do better
  97. 3:56than that
  98. 3:58and is that the only measure of
  99. 4:01performance
  100. 4:02so generally when we take a look at the
  101. 4:05spec sheet
  102. 4:06for any kind of a product there are
  103. 4:08usually different kinds of performance
  104. 4:10measures
  105. 4:12so in this case let's try to see what do
  106. 4:14we mean by a person
  107. 4:15performance of a processor
  108. 4:20in some cases it may mean that it may
  109. 4:23have a quicker response time so
  110. 4:25if we give it one job it will will it
  111. 4:27finish it
  112. 4:28sooner so the time to execute one job
  113. 4:31may be
  114. 4:32the the the
  115. 4:35a performance measure or how many jobs
  116. 4:39can you do
  117. 4:39in a unit time so you know how many
  118. 4:42web pages can it serve or how many
  119. 4:45spoken words can it recognize or
  120. 4:47how many chess games can it do
  121. 4:50can it play in an hour
  122. 4:54or um how much energy does it take
  123. 4:57meaning how many uh how
  124. 5:01long will our battery last when running
  125. 5:04this processor running tasks on this
  126. 5:07processor like
  127. 5:08watching movies how many movies can we
  128. 5:10watch on a battery charge
  129. 5:13what is useful to think of
  130. 5:17while we are getting a sense for a
  131. 5:18process of performance to think of
  132. 5:20something that is that is
  133. 5:21much more familiar like transportational
  134. 5:23knowledge and in transportation we can
  135. 5:25take a look
  136. 5:26at the performance metrics of a sports
  137. 5:28car
  138. 5:29versus performance metrics of a bus and
  139. 5:31usually people think
  140. 5:33of a sports car as a performance car
  141. 5:38but bus is also a performance vehicle
  142. 5:41for certain under certain metrics
  143. 5:45so let's take a look at uh on the left
  144. 5:48hand side uh specs for a sports car
  145. 5:52and specs for a bus so the sports car
  146. 5:54can carry two passengers the bus can
  147. 5:56carry 50 passengers
  148. 5:58um a sports car can go at 200 miles per
  149. 6:01hour and the bus can only go at 50 miles
  150. 6:03per hour
  151. 6:05and gas mileage for the sports car is
  152. 6:07five miles per gallon
  153. 6:09um so it can go more miles per gallon
  154. 6:12than the bus that can only go
  155. 6:13two miles per gallon so let's see how
  156. 6:16they perform on a task
  157. 6:19that involves transporting a hundred
  158. 6:21passengers
  159. 6:22over a 50 mile trip and in this case
  160. 6:25we're just going to assume that
  161. 6:27we are measuring only one-way
  162. 6:30travel and
  163. 6:33both the car and the bus um
  164. 6:36immediately somehow come back to the
  165. 6:40starting point you know reload the
  166. 6:43passengers
  167. 6:43and go back so we we're just measuring
  168. 6:46the time that
  169. 6:47it is taking to transport the passengers
  170. 6:51so a sports car um will travel
  171. 6:5550 miles in 15 minutes the bus will take
  172. 6:58an hour
  173. 6:59but in every trip the
  174. 7:02the pal the sports car will only deliver
  175. 7:04two passengers and
  176. 7:05go somehow transport itself empty and
  177. 7:08take another two passengers
  178. 7:10so in total to trans transport a hundred
  179. 7:12people will take 750 minutes on the
  180. 7:15other hand the bus will take only two
  181. 7:17trips
  182. 7:17so it'll be done in 120 minutes
  183. 7:22the energy efficiency of a bus is a lot
  184. 7:24better than the energy efficiency of a
  185. 7:26car
  186. 7:27the car um uh takes
  187. 7:30five gallons per passenger to deliver
  188. 7:33them
  189. 7:34uh 50 miles to the destination if it's
  190. 7:3850 miles away
  191. 7:39and a bus takes only
  192. 7:430.5 gallons per passenger
  193. 7:46so if we had so we see a difference here
  194. 7:51and
  195. 7:51what does it mean something having a
  196. 7:53higher performance
  197. 7:55yeah a sports car is better if we have
  198. 7:58something that is of a very high value
  199. 8:00that needs to be
  200. 8:00transported 50 miles away
  201. 8:04an injured person for example
  202. 8:09but bus is much better in delivering
  203. 8:12moving a lot of people over a distance
  204. 8:15so
  205. 8:15it will do it in a much shorter time and
  206. 8:18save a lot of energy
  207. 8:19let's see how does this relate to the
  208. 8:21compute performance
  209. 8:23so in the transportation what we had was
  210. 8:27trip time in computer world that
  211. 8:30metric of performance will be program
  212. 8:32execution time
  213. 8:34how long does it take to execute a
  214. 8:37a program or for example time to update
  215. 8:40the
  216. 8:40display with a new picture
  217. 8:46the second metric that we had was the
  218. 8:49time
  219. 8:49to for in in transportation world time
  220. 8:52to transport 100 passengers
  221. 8:55in the compute world it is a metric
  222. 8:59that measures how many tasks can
  223. 9:03can we do in a given time for example
  224. 9:06in this case it would be how many web
  225. 9:09pages can be served in a given time um
  226. 9:12or how many
  227. 9:12server requests can be performed in an
  228. 9:14hour
  229. 9:16and then third one is the energy
  230. 9:18efficiency the way how we measure the
  231. 9:19energy efficiency in the cars is
  232. 9:21you know gallons per passenger in a
  233. 9:24computer
  234. 9:25we measure them we also measure energy
  235. 9:28efficiency
  236. 9:29by the amount of energy used electrical
  237. 9:32energy used
  238. 9:33so in this case we will measure the
  239. 9:36energy per task or
  240. 9:37how much energy does playing a movie
  241. 9:40take
  242. 9:41or how many movies can we watch from a
  243. 9:42battery charge
  244. 9:44keep in mind that the energy is
  245. 9:46important for
  246. 9:47both mobile devices where we kind of
  247. 9:50have a better sense for this
  248. 9:51how much energy is stored in in a
  249. 9:54battery
  250. 9:56energy is also very important for high
  251. 9:58performance servers that live in
  252. 10:00warehouse computers
  253. 10:01in data centers over there the cost of
  254. 10:04running a data center is essentially
  255. 10:07the cost of energy the hardware
  256. 10:10um you know the energy cost
  257. 10:14overshadows the hardware cost after
  258. 10:16about a
  259. 10:17year or you know year and a half after
  260. 10:19deployment so that's a that's a big deal
  261. 10:22in every
  262. 10:24[Music]
  263. 10:25application domain or in every domain of
  264. 10:27computing
  265. 10:29finally one thing that is important to
  266. 10:31notice that power is generally not a
  267. 10:33good measure
  268. 10:35of efficiency energy is much better
  269. 10:38power is a rate of exchange of energy so
  270. 10:41they're kind of related
  271. 10:43but um the the way how we can think of
  272. 10:48it
  273. 10:48is we can have um
  274. 10:51a really low performance cpu that is
  275. 10:54very low power
  276. 10:55very very low power but it would take so
  277. 10:59long to complete
  278. 11:01the task um that
  279. 11:04it would basically drain the energy by
  280. 11:06the time it's done on the other hand
  281. 11:07some better optimized cpu can probably
  282. 11:10do
  283. 11:11multiple tasks long tasks with
  284. 11:15one battery charge we'll
  285. 11:18see a bit more about that later but for
  286. 11:20now what we want to
  287. 11:22remember is we really care about the
  288. 11:24energy
  289. 11:25energy is what is stored in the battery
  290. 11:28and that's what
  291. 11:29the utility charges charges us
  292. 11:34in units for delivering us the energy
  293. 11:37not the power so that's it for the quick
  294. 11:42introduction
  295. 11:43so we are going to try to take a
  296. 11:46deeper look into the processor
  297. 11:48performance after a quick break

About this transcript

This page contains the full transcript of [CS61C FA20] Lecture 21.1 - Pipelining I: Pipelining by CS 61C Departmental, generated from the public captions YouTube serves with the video. The transcript has 1,628 words across 297 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.