[CS61C FA20] Lecture 31.1 - I/O: I/O Devices — Transcript

Full transcript

1. 0:00[Music]
2. 0:10hello
3. 0:11welcome back to our module that deals
4. 0:13with
5. 0:14virtual memory and operating system
6. 0:16support but now we have figured out
7. 0:18pretty much how does
8. 0:19virtual memory work so what you would
9. 0:21like to do
10. 0:22in order to complete our compute system
11. 0:26is to add io devices so let's get into
12. 0:29that
13. 0:31we made tremendous progress so far in
14. 0:35this class we
15. 0:36build almost a computer so we started
16. 0:39with the high level c program and
17. 0:41practices that
18. 0:42through the first project then we
19. 0:45learned the assembly language
20. 0:47for this five isa and practiced that and
21. 0:50really cemented it
22. 0:52through project two then we built a data
23. 0:54pod that pretty much can
24. 0:55execute our program from project two
25. 0:58minus the equals but that's not very
26. 1:01difficult although
27. 1:03not too educational to implement over
28. 1:06there
29. 1:06you know they've just added a burden so
30. 1:08we can assume that we can build
31. 1:10we can run pretty much everything we
32. 1:12have built
33. 1:13in uh project two and we can also
34. 1:16compile our project one and run it over
35. 1:19on our core
36. 1:20we added cache memory and then we
37. 1:23connected the virtual memory so all
38. 1:25what we need in order to have a computer
39. 1:28is to add
40. 1:29io devices such that we can hook up a
41. 1:31keyboard and a mouse and display our
42. 1:33results
43. 1:34on the screen and perhaps connect it to
44. 1:38a network
45. 1:40the key thing here is you would like to
46. 1:42understand the principles how these
47. 1:44io devices are connected such that we
48. 1:47don't have to do
49. 1:48one at a time using some special
50. 1:51procedure so let's get into that so how
51. 1:55do we interact with the devices
52. 1:57so we have a program running on a cpu
53. 2:00and then
54. 2:01it would like to print something on the
55. 2:02printer or receive
56. 2:06input from a keystroke so how do we do
57. 2:08that
58. 2:09so we have our processor and the memory
59. 2:11and there is some code that is running
60. 2:12on that
61. 2:13so we need to have an i o interface for
62. 2:16keyboards network mice displays and so
63. 2:19on
64. 2:20and you would like to do that in a
65. 2:23uniform way so that we don't have to do
66. 2:25it
67. 2:26separately for each one of them
68. 2:28generally
69. 2:29there will be a plethora of devices and
70. 2:31they're going to have very
71. 2:32different demands or the ways how they
72. 2:36present their data or how do they take
73. 2:38their data
74. 2:38but you would like to try to treat them
75. 2:40in a uniform way
76. 2:42they would be generally connected
77. 2:43through some hierarchy of buses and the
78. 2:46way how we can
79. 2:48view these buses are like you know
80. 2:51highways um where a lot of data is
81. 2:54moving
82. 2:55um at a relatively high rate and then
83. 2:58there are things
84. 2:59connected off ramps to that highway
85. 3:04we're only going to get into the depths
86. 3:07and details
87. 3:08of designing buses we can just view them
88. 3:12as something abstract that is connected
89. 3:15to the processor
90. 3:16memory system
91. 3:19then the way how these devices
92. 3:23talk to the rest of the world is
93. 3:26generally through standardized interface
94. 3:29that consists of command
95. 3:32and status registers and data registers
96. 3:36and generally is the operating system
97. 3:39that
98. 3:41checks the status where the weather
99. 3:42devices are ready to talk to us
100. 3:45and then orchestrates how
101. 3:48do the processes underneath access these
102. 3:51devices
103. 3:53so standardized interface they generally
104. 3:57have commands
105. 3:57let's say print something and then the
106. 4:00data register contains
107. 4:02what to actually print all right
108. 4:07how do we support this kind of
109. 4:10i o interfaces through our isa
110. 4:14um so it's relatively basic
111. 4:17at the very bottom and the leaf level of
112. 4:19what does the
113. 4:20processor need to do it needs to be able
114. 4:22to read the sequence of bytes
115. 4:24and write the sequence of bytes whether
116. 4:25there are commands
117. 4:27or the data so how do we do that
118. 4:30well we have two options one design
119. 4:32special input output instructions
120. 4:34and the matching hardware that would be
121. 4:36talking to each type of device this used
122. 4:38to be done in the past there used to be
123. 4:40io instructions
124. 4:41in processors but that is that is turned
125. 4:44out not to be a good idea because
126. 4:46these devices change we get new
127. 4:49standards new interfaces every few years
128. 4:52and the old ones get old and
129. 4:55abandoned the obsolete and our isa is
130. 4:57stuck with supporting
131. 4:59stuff that may be 10 or 20 or 40 years
132. 5:02old
133. 5:03so most of the things nowadays are
134. 5:05actually
135. 5:07done through something that is called
136. 5:08memory mapped i o
137. 5:11portion of generally address space some
138. 5:14low addresses in our
139. 5:16memory space in our memory locations
140. 5:20are dedicated to i o that's
141. 5:24where our i o command and
142. 5:28data registers are placed so we don't
143. 5:31use that part of a memory it is not
144. 5:33visible to our program
145. 5:35it is managed by the os and
146. 5:38our processes may be allowed
147. 5:41to to to to read or write from there
148. 5:46so if they're allowed they can use
149. 5:48normal load and store
150. 5:50instructions to access them or they
151. 5:52would be accessible through system calls
152. 5:55and this is the way how risk 5 does it
153. 5:58so here is an example
154. 6:02there are some addresses in this example
155. 6:04all addresses below
156. 6:057 ffff
157. 6:08are memory are reserved for memory
158. 6:10mapped io that's where all our
159. 6:12peripheral devices are going to copy
160. 6:14their control and status registers and
161. 6:16they're going to
162. 6:17read and write to them that's going to
163. 6:19be our way of communication so our
164. 6:21program and data memory is going to
165. 6:23start from the address eight
166. 6:25zero zero zero zero zero zero and go up
167. 6:29to the top of the memory so every single
168. 6:32io device is going to have a copy of its
169. 6:36registers
170. 6:37in the memory mapped i o region
171. 6:42one thing to keep in mind there is a
172. 6:44variety of devices out there and
173. 6:46they may be running at very different
174. 6:48speeds so if our microprocessor
175. 6:51is uh just a single quarter it's five
176. 6:53processor says five stage pipeline
177. 6:55operating at one
178. 6:56uh gigahertz it can do four
179. 6:59gb per second worth of loads and stores
180. 7:02it can write in every cycle
181. 7:04assuming that a mat is somewhere around
182. 7:08one it can in every cycle do one
183. 7:11four byte load or four byte store
184. 7:16but there is a whole bunch of devices
185. 7:18that
186. 7:19may not be able to use that or generally
187. 7:21will not be
188. 7:22able to use them unless they're dram
189. 7:25so let's take a look at what is out
190. 7:28there
191. 7:29so if you're working with a keyboard
192. 7:31keyboard really the rate that we talk to
193. 7:33a keyboard is
194. 7:34something of the order of tensor bytes
195. 7:36per second
196. 7:38right i mean that's the speed the
197. 7:40maximum speed that we can type
198. 7:42um you know humans and human fingers
199. 7:46can produce few tensor bytes at the max
200. 7:48per second if it is me it's you know
201. 7:50maybe a few bytes per second
202. 7:52um and then if you're dealing with audio
203. 7:55signals which are typically
204. 7:57associated you know bluetooth is now
205. 7:59used to carry uh
206. 8:01general audio signals it is few
207. 8:04hundreds of cubes per second it can go
208. 8:06up to three
209. 8:08mega bytes per second and then when we
210. 8:11talk about
211. 8:12wi-fi and and ethernet
212. 8:15and um you know hard disk drives we can
213. 8:18go up to higher speeds
214. 8:19but none of them get anywhere close
215. 8:21they're typically
216. 8:23at least an order of magnitude below the
217. 8:25speed of what the processor can do
218. 8:27until we get to like thunderbolt and
219. 8:31newer pressure dram interfaces
220. 8:35now those are generally those that
221. 8:37exceed the bandwidth of what the core
222. 8:40can
223. 8:40read and write are designed to support
224. 8:43higher
225. 8:43end processors and those that have
226. 8:45multiple cores because each of these
227. 8:47cores
228. 8:48may have a desire to load or store the
229. 8:51data to the memory
230. 8:54so generally
231. 8:58common io devices neither deliver nor
232. 9:00accept data matching the processor speed
233. 9:03we have to have some way
234. 9:04to adjust to that and we're going to
235. 9:06take a look at a few
236. 9:09possible options how to do that after a
237. 9:11quick break see you then