[CS61C FA20] Lecture 31.2 - I/O: I/O Polling — Transcript

Full transcript

1. 0:00[Music]
2. 0:09hello
3. 0:10and welcome back to our discussion about
4. 0:12input
5. 0:13output devices we've seen that we often
6. 0:16deal with a variety of devices
7. 0:18and they usually work with very
8. 0:20different data rates some of them are
9. 0:22very slow
10. 0:23and some of them can be producing a lot
11. 0:25of data that
12. 0:26processor will need to keep up with
13. 0:30we also have seen that generally the way
14. 0:34how we work with these devices is
15. 0:37by talking to its memory mapped
16. 0:40registers their memory map registers
17. 0:42so they have two types of registers
18. 0:47or at the minimum one of each type
19. 0:50there would be a control register and a
20. 0:52data register
21. 0:56the these control registers
22. 0:59would basically tell us whether it's
23. 1:00okay to read
24. 1:02or write the data to the device um
25. 1:06we can use an analogy of a
26. 1:10flagman on the road that tells us
27. 1:12whether it's okay or not
28. 1:13to go through the draw part of the road
29. 1:16if they have a
30. 1:18raised flag that means that we should
31. 1:21generally stop
32. 1:23and then if they uh drop that flag then
33. 1:26it's okay to to go
34. 1:31similarly that's how the control
35. 1:34register tells us whether
36. 1:35the device is ready to receive the data
37. 1:38if we're writing to it
38. 1:39or it has the data ready for the
39. 1:41processor to read from
40. 1:43so where is the data that should be read
41. 1:46from that's in the data register
42. 1:48it's also the register that we are going
43. 1:50to write to
44. 1:52so in a procedure called polling
45. 1:56a processor repeatedly or and
46. 1:59periodically
47. 2:00checks the control register
48. 2:04and there is usually one bit over there
49. 2:08that is a ready bit ready bit means that
50. 2:11either
51. 2:11the device is ready to receive the data
52. 2:15or it has some data that is ready to be
53. 2:17sent to the processor
54. 2:20at ready signal is typically a single
55. 2:22bit
56. 2:23and it is controlled by the device
57. 2:26the device sets it by changing its value
58. 2:29from zero to one
59. 2:34processor then will load from
60. 2:37the device or right to the device
61. 2:40whether it's an input or output device
62. 2:43and the io device
63. 2:46resets the control signal from one to
64. 2:49zero
65. 2:50as we said this is a procedure called
66. 2:52polling
67. 2:54so let's see uh an example of a polling
68. 2:57code
69. 2:58this is by the way the simplest
70. 3:00procedure car you know
71. 3:02simplest method that we will work that
72. 3:04we can use
73. 3:05to work with input output devices
74. 3:08so in this example we have a memory map
75. 3:11that contains
76. 3:13one input device and one output device
77. 3:15and each of these devices just as one
78. 3:17control and one data register and as
79. 3:21so input devices one input control
80. 3:24register
81. 3:24and one data register and output devices
82. 3:27one control register one data register
83. 3:29so here are two loops one for
84. 3:33working with input device the other one
85. 3:35that conversely works with the output
86. 3:36device
87. 3:39conveniently the first
88. 3:42entry into the memory map is set to be a
89. 3:45value
90. 3:46seven ffff000 last three digits being
91. 3:49zeros
92. 3:51make it convenient to use a louie to
93. 3:54load that address into the temporary
94. 3:56register and we don't have to follow it
95. 3:58with
96. 3:58an ad immediate so then we enter
97. 4:03this loop the loop essentially loads the
98. 4:06value
99. 4:07from the control register input control
100. 4:10register
101. 4:11and checks whether the ready bit is has
102. 4:14been asserted
103. 4:15whether it's equal to one it is going to
104. 4:18stay in this loop
105. 4:19as long as that is equal to zero this
106. 4:22end immediate is going to
107. 4:23to to keep you know is is going to
108. 4:26produce a
109. 4:28value of 0 unless
110. 4:31the ready bit which is the least
111. 4:33significant bit in this register
112. 4:35is asserted if it is it will exit the
113. 4:38loop
114. 4:39and load the data i'll do whatever
115. 4:42it needs to do with the data and
116. 4:44continue polling
117. 4:45at some point
118. 4:50the output polling works exactly the
119. 4:53same way
120. 4:54we would like to for example write to
121. 4:57a display from a1 while
122. 5:00we are again going to luay we are going
123. 5:03to
124. 5:04enter a weight loop now we are going to
125. 5:06check the appropriate registers
126. 5:07and we are going to write to it
127. 5:11when the device is ready
128. 5:17so this is very simple and convenient
129. 5:20but is not always the most efficient way
130. 5:23of dealing with io devices so let's
131. 5:25assume
132. 5:25our standard processor that we have
133. 5:27dealing we have been dealing with that
134. 5:29runs at one gigahertz
135. 5:30clock rate so let's say that it takes
136. 5:33400 clock cycles for a polling operation
137. 5:36so we call a polling routine you know
138. 5:39check the device whether it's a keyboard
139. 5:40or wi-fi um you know whether it's uh
140. 5:44ready to to send us the data or receive
141. 5:47it
142. 5:47and then return so how much of a time
143. 5:51what percentage of its clock cycles does
144. 5:53the processor need to spend on polling
145. 5:56so let's take a look at a simple example
146. 5:58usually mouse needs to be pulled
147. 6:00something like 30 times per second in
148. 6:02order to avoid
149. 6:04jittery motions on the screen um
150. 6:08so let's see what's our cost of polling
151. 6:11the
152. 6:12the polling amounts to see if it has
153. 6:15changed this data if it has changed its
154. 6:17location
155. 6:19so when we're pulling a mouse let's say
156. 6:22we said that we need 30 poles per second
157. 6:24each pole takes
158. 6:26400 clocks to complete
159. 6:29so that's 12 000 clocks per second
160. 6:34so if processor is pulling a mouse then
161. 6:37that's a really tiny fraction of the
162. 6:40total clock clock cycles that are being
163. 6:44they're available in each second so we
164. 6:46divide this number of 12
165. 6:48k with a billion clock cycles though
166. 6:51that is 0.001
167. 6:55so that's fine
168. 6:58um that that's not a big load for a
169. 7:00processor and processor can pull that
170. 7:02kind of an
171. 7:03input device however it's kind of not
172. 7:05practical if we are pulling all of the
173. 7:07devices and we don't know when to stop
174. 7:09pulling for a polling mouse you know if
175. 7:12we know that somebody
176. 7:14if there is no not going to be mouse
177. 7:16activity
178. 7:17but it's fine for now um
179. 7:21let's take a look at a different device
180. 7:22the device that has a lot more data for
181. 7:24example
182. 7:25um if you would like to pull a disk and
183. 7:27the disk
184. 7:28let's say is a moderately fast one that
185. 7:32is capable of producing 16 megabytes per
186. 7:34second
187. 7:35and it we need to
188. 7:39do 16 bytes per pole so
189. 7:42we need 1 million pulse per second
190. 7:45so how um how do we
191. 7:49um you know how how many cycles do we
192. 7:51need to spend on that what's the
193. 7:53percentage of cycles
194. 7:54so we need to do one million poles per
195. 7:56second it's going to cost us again
196. 7:58400 cycles to do each poll so that's a
197. 8:02400 million
198. 8:05clocks per second that we would spend on
199. 8:08polling the disk
200. 8:09well we only have a billion cycles
201. 8:13per second that basically means that
202. 8:16this processor
203. 8:17will be doing nothing else than polling
204. 8:2040
205. 8:21of clock cycles basically would uh
206. 8:24render that processor don't be able to
207. 8:27do
208. 8:28much other things and by the way also
209. 8:30polling disks that
210. 8:32produce a lot of data in larger chunks
211. 8:34pulling them to receive small chunks
212. 8:36doesn't really make that much sense
213. 8:40so trying to get 16 bytes per pole
214. 8:43may not be the best way of doing things
215. 8:46so we need to come up with a better way
216. 8:49of working with these devices that can
217. 8:50produce
218. 8:51a lot of data and what we'll find out
219. 8:53when we discover that
220. 8:54we'll most likely use that better
221. 8:56mechanism for everything else
222. 8:59we'll get to that after a quick break
223. 9:01see you then