WEBVTT

1
00:01:34.050 --> 00:01:40.590
William Cheng: Hello. Welcome to the live Lecture eight. On June 12 2020

2
00:01:42.060 --> 00:01:46.800
William Cheng: Anybody has any question please feel free to unmute the microphone.

3
00:01:52.560 --> 00:01:52.860
William Cheng: Morning.

4
00:01:53.790 --> 00:01:59.520
Hardik Mahipal Surana: I had a question not in this respect to the lecture but LOOKS LIKE TO THE ASSIGNMENT, THE LAW. Sure.

5
00:02:00.060 --> 00:02:02.910
Hardik Mahipal Surana: So I'm actually entering into a case where

6
00:02:04.110 --> 00:02:12.720
Hardik Mahipal Surana: Once I press Control see my monitoring that does all of the tasks that were mentioned in the discussion slides and I tried

7
00:02:13.260 --> 00:02:29.880
Hardik Mahipal Surana: In the end I broadcast all of the server threads and in the summertime. Once I broadcast my my thread comes out of the condition variable queue and then it checks for the global flag which was said. And at that point, when

8
00:02:31.980 --> 00:02:39.990
Hardik Mahipal Surana: The controls. He was not pressed it starts, you know, removing a packet from Q2. And then when controls I press Control see before.

9
00:02:40.530 --> 00:02:50.100
Hardik Mahipal Surana: I run the score up my my global flag SAYS THAT SHANNON has not been initiated and after I do the check. Then controls. He gets pressed. So I'm in the middle of

10
00:02:50.700 --> 00:02:58.740
Hardik Mahipal Surana: Trying to remove a packet from Q2, but at this stage, since controls. He has been pressed, I shouldn't be removing the package from you today. So

11
00:02:58.920 --> 00:03:06.420
William Cheng: So that's why in the control C catching threat as soon as you return from from Sig. Wait, you should lock them he attacks.

12
00:03:07.230 --> 00:03:08.040
Hardik Mahipal Surana: I am doing

13
00:03:08.910 --> 00:03:18.000
William Cheng: Well, so in that case, if the server has the new tax law, then the control see catching threat they will block or mutates me tatlock right

14
00:03:19.230 --> 00:03:19.740
Hardik Mahipal Surana: Years.

15
00:03:20.940 --> 00:03:26.910
Hardik Mahipal Surana: So I'm actually getting some segmentation fault issues because I'm trying to access a packet from Q2.

16
00:03:30.390 --> 00:03:32.640
William Cheng: You only do that if you have the new tax law.

17
00:03:32.790 --> 00:03:33.870
Hardik Mahipal Surana: Right. Yes.

18
00:03:34.440 --> 00:03:36.210
William Cheng: So in that case, there should be no conflict.

19
00:03:37.680 --> 00:03:54.840
Hardik Mahipal Surana: I do. Ideally, yes, but I don't understand why I'm facing that issue. I'm getting a few segmentation falls and I tried using GDP to understand where that is happening and it shows me points where I'm trying to reference a pointer from the packet

20
00:03:56.400 --> 00:04:03.120
William Cheng: Right, so, so, so either you unlock the new attacks when you're not supposed to, or you have a memory corruption bug somewhere.

21
00:04:03.900 --> 00:04:12.900
William Cheng: That cause something were to happen, right. Because once you have memory corruption bugs and all kinds of weird stuff can happen. So you got to make sure that you're not doing something that will corrupt memory.

22
00:04:13.920 --> 00:04:23.010
William Cheng: But, but then in that case, I mean, you know, what's important is to understand how this works. Right. You know, if you have the new time slot. That means that no other thick and get the new tax.

23
00:04:24.120 --> 00:04:33.570
William Cheng: Right. So if you have the new time slot and another another threat apparently looks like it has the new tags. Well, then that means that something is wrong with your address space.

24
00:04:35.070 --> 00:04:39.090
William Cheng: Okay, so, so if that actually can happen. That means that there's a memory corruption bug.

25
00:04:40.470 --> 00:04:50.370
William Cheng: Okay at the other possibility is that that's not happening. But it looks like it's happening, you know, then then again you have to debug your code to find out how. How can something like that happen.

26
00:04:53.970 --> 00:05:02.160
William Cheng: Right, I mean Debugging is not, there's really no fixed procedure, but this is how you debug it and you'll find all your bugs on if there's a procedure like that there will be no bugs.

27
00:05:03.660 --> 00:05:09.690
William Cheng: Okay, so what you need to do is that you need to, you know, he says, the first of all you need to find out exactly where your program crashed.

28
00:05:10.470 --> 00:05:21.360
William Cheng: Okay, so, so if you, you know, if you're saying that you feel if your server threat wakes up from the pizza, pizza, a new tags. Sorry. Peter condition. Wait.

29
00:05:21.960 --> 00:05:28.020
William Cheng: Right, so, so what's what's important to understand is that before you call pizza condition where you have to be that slot.

30
00:05:28.710 --> 00:05:37.830
William Cheng: Right. So, what you call Peter condition way now temporarily, you're going to have the pieces, you're going to have to meet as a lot. But as soon as the Peter condition where return. You have to be tax law.

31
00:05:39.510 --> 00:05:47.640
William Cheng: Okay, so if you made a programming mistake and you know, then you, you then you don't call Peter a condition where you have the new tax law. Well, then that will be a problem.

32
00:05:48.540 --> 00:05:53.520
William Cheng: Okay, but if you, you know, it says there's remember that the rest of people pleaser condition way is

33
00:05:54.030 --> 00:06:04.650
William Cheng: Piece of effects not why not guard piece recognition way so you have to record like that. So in this case, where you copies are conditional way. You know, you must have the music's loud. So therefore, when Peter condition were returned. You still have the meantime.

34
00:06:05.880 --> 00:06:11.190
William Cheng: If you have the music's loud and no other thread can, you know, can, can, can only view tax credit.

35
00:06:11.610 --> 00:06:22.920
Hardik Mahipal Surana: So ideally, once we wake up from the condition variable and we have the new tax law again at that stage, since we have the mutations. We just need to check the global flag once and then

36
00:06:24.600 --> 00:06:27.300
Hardik Mahipal Surana: Check whether the packet. There is a packet being serviced or

37
00:06:28.260 --> 00:06:44.490
William Cheng: No, no, no. The cold look like this right piece it. Okay, so, so let me do this right piece of me. That's okay so so so piece I mean tatlock right, wow, while nod guard right P thread condition. Wait, right.

38
00:06:46.140 --> 00:06:51.000
William Cheng: So, so what is this God right then this guard in the, I think in the warm up to

39
00:06:52.020 --> 00:06:57.540
William Cheng: In the warm up to FAQ, it shows you that, you know, Q2 is

40
00:06:59.820 --> 00:07:03.090
William Cheng: A future is empty, and it's not time to quit right

41
00:07:04.680 --> 00:07:15.780
William Cheng: Okay, so, so the the the scenarios that you know the the scenario that you have is that Q2 is not empty right so Q2 is not empty, then that they even jump out of the infinite loop right

42
00:07:17.010 --> 00:07:23.190
William Cheng: Yes, right. So if you jump out of the infinite loop. The first thing you still have to do is to check if Control C has been pressed

43
00:07:25.800 --> 00:07:30.750
William Cheng: Yes, right. Because if Control C has been pressed, you must not, you know, secure package so so it

44
00:07:32.220 --> 00:07:41.640
William Cheng: Says if right if you are, you know, if you have to quit. Why, in that case, you have to, you have to, you know, basically a long commute acts and just get out right

45
00:07:44.010 --> 00:07:49.140
William Cheng: Okay but but otherwise. Otherwise, if you reach your it must mean that your QQ is not empty, right.

46
00:07:50.490 --> 00:08:04.050
William Cheng: Yes wife cuter is not empty, and then when he said a DQ the pack of on Q2. Because at this point, you have the new tax law, if you will, if you put a defect, apart from Q2 and you get a segmentation ball. Well, that means that you, you have a blog somewhere else.

47
00:08:05.280 --> 00:08:11.910
William Cheng: Because how could it be that Q2 is not is not empty, and when you DQ a packet, you're gonna get a segmentation ball that's not possible.

48
00:08:13.290 --> 00:08:15.150
William Cheng: Unless you have a bug. Great.

49
00:08:15.240 --> 00:08:18.900
Hardik Mahipal Surana: I am emptying both q1 and q2. Before I broadcast the so

50
00:08:20.700 --> 00:08:21.450
William Cheng: Why would you do that.

51
00:08:23.400 --> 00:08:24.000
Hardik Mahipal Surana: Because

52
00:08:24.030 --> 00:08:28.020
William Cheng: We need you. I will. Are you talking about the coat we press Control. See, yes.

53
00:08:28.440 --> 00:08:30.690
William Cheng: Yeah so. So again, my recommendation.

54
00:08:31.800 --> 00:08:37.410
William Cheng: My recommendation. Okay. You don't need to follow that. But my recommendation is that what you've press Control see right

55
00:08:38.280 --> 00:08:51.900
William Cheng: You know, you, you. What will you return from Subway you locked me tags you can sell your rivals where you could cancel the packets of that and then you said you said a global variable to say it's time to quit or you broadcast the CV and you unlock the new tax yourself terminate.

56
00:08:53.670 --> 00:09:09.150
William Cheng: Okay, leave all the packets in Q1, Q2. I keep posting this empty in the class, we will go, I said that, you know, my recommendation is that the main threat will be the one that remove all those packets. Okay, well, what, why is that a better, better, better way

57
00:09:12.210 --> 00:09:12.750
William Cheng: I mean,

58
00:09:14.190 --> 00:09:18.450
Hardik Mahipal Surana: I think because once we cancel the package to concerts and so what

59
00:09:19.320 --> 00:09:35.310
Hardik Mahipal Surana: If that currently servicing a packet. Then once the model of the control see catching thread exists, then we know for a fact that we have initiated the cancer for the packet and to concerts and then the remaining package already in the kids need to be remote to the main thread can

60
00:09:36.270 --> 00:09:48.240
William Cheng: My logic is actually simpler right you know the die. This is that your main thread create those fourth way john with all those four threads. OK. So now when you try to remove Haggis on q1 and q2. There's no other thread running right

61
00:09:49.020 --> 00:10:01.500
William Cheng: Yes. So, therefore, you don't have to worry about. You don't even have to log the new tax if you don't want to write because because all the other thread, they're not, you know, I know you know up because no other side is adding packet or removing packet

62
00:10:04.050 --> 00:10:10.980
William Cheng: So, so you know again inside, Control C catching threat if you want to remove all the packet, that's fine, because you have the new tax law. You can do it, do all that.

63
00:10:11.370 --> 00:10:22.170
William Cheng: Okay, then in that case, when you remove the packet. Okay. Is that a contrasting that with the new tax law, then it's not possible for the server to be removing a packet. At the same time, right.

64
00:10:23.910 --> 00:10:30.390
William Cheng: Because you have the music loud. Yes. So you must have a bug somewhere, you just need to see if you just need to find the bug. Okay. Okay.

65
00:10:30.870 --> 00:10:37.560
William Cheng: Right, so, so sometimes talking about can be very, very tricky and it can be very time consuming and sometimes one bug will take one or two days to debug.

66
00:10:37.950 --> 00:10:49.020
William Cheng: I mean, that's another reason you know I want people to, you know, to try to get through to submit the assignment early. So this way, just in case you have a nasty bug, they will take you one or two days to debug and then that can still happen.

67
00:10:51.240 --> 00:10:53.760
Hardik Mahipal Surana: Sure, I can go to position.

68
00:10:54.060 --> 00:11:04.260
William Cheng: Yeah, so, so, yeah. So when you try to debug it again. So how do you develop the program it but we, you know, for you to give up the program. First you have to make predictions, right. Yep. Say, I think this is what's going to happen.

69
00:11:05.130 --> 00:11:11.820
William Cheng: Okay, and then you see if that happens or not. If it turns out that didn't happen either your assumption is wrong or your code is wrong.

70
00:11:13.950 --> 00:11:22.560
William Cheng: Right, so, so in that case you need to say why is my assumption round is my code is wrong. And then if your assumptions is wrong. What do you need to get a new assumption. If you're somebody is correct. Well, then it must be a code is wrong.

71
00:11:23.550 --> 00:11:27.690
William Cheng: Right. So in that case, you, you, you, you try to find out exactly where your code made a mistake.

72
00:11:30.030 --> 00:11:42.720
William Cheng: Okay. Well, of course, in the worst case scenario is the assumption is wrong and the code is wrong. So in that case, you have to sort of, you know, start, you know, really, we think about what assumption or what what what what kind of code, you should be writing. Yes.

73
00:11:44.190 --> 00:11:52.290
William Cheng: Okay, so I hope he does. Oh, yeah, yeah. So, so what what what I'm saying. Over here is that if you know exactly where all your critical section code is

74
00:11:52.740 --> 00:12:05.220
William Cheng: You know, then, then the condition that you just mentioned just cannot happen. Right, so therefore it must be, you know, one of the one of the assumption there is wrong because you'll be because you're you're you're you. You have a bug in your code.

75
00:12:07.380 --> 00:12:08.640
William Cheng: Okay, okay.

76
00:12:11.760 --> 00:12:24.420
William Cheng: Yeah, because if you said, if, if the cure packet and now the you know the the the the Q1, or Q2, you know, the degree they get a segmentation fall I mean that that doesn't sell you know if you have no bugs and that's not possible.

77
00:12:26.820 --> 00:12:34.710
William Cheng: Okay, so one thing that you need to sort of convince yourself is that if I have no bug, then this condition is not possible because of all these reasons, because all the things that we learned in class.

78
00:12:35.700 --> 00:12:39.990
William Cheng: Right. And if that's true, then, then it must be a bug in your code and then you just need to find a bug.

79
00:12:44.730 --> 00:12:49.650
William Cheng: Yeah so so multi threading. That's one of the reason multithreading is sort of a harder thing to debug.

80
00:12:49.920 --> 00:12:57.120
William Cheng: Because, you know, these kind of things are happening any water right. The, the, the, the control see press a thread is you can trust you can be present anytime

81
00:12:57.450 --> 00:13:05.220
William Cheng: And the you know the servers wreck and run it anytime. So, therefore, you know, all kinds of timing can actually happen. So that makes it harder to debug.

82
00:13:08.730 --> 00:13:18.630
Hardik Mahipal Surana: Another question I had was, what happens or what should happen when we press Control Z. As soon as we start the program. I mean, before you and the emulation.

83
00:13:21.150 --> 00:13:29.670
William Cheng: Well, then, then you progress you get killed right away. Right. Yeah, but why don't you stop so so so Rob, what I would do is that you know as soon as the main program starts

84
00:13:30.180 --> 00:13:39.900
William Cheng: I will block until see right away when you can also you know parts command line argument first and then and then block, Control C. But once you start printing, you know, emulation parameters.

85
00:13:40.290 --> 00:13:48.930
William Cheng: Right. So the first thing that we print the print emulation parameter you should print up before you create any threats. Okay. So, therefore, before you print those things out. You should block control see

86
00:13:50.340 --> 00:13:54.960
William Cheng: Raza once you start, you know, printing those things out. What, then, then, and then in that case, you're in control. Right.

87
00:13:56.430 --> 00:14:08.670
William Cheng: So again, the timing is very important. And then you. You're the you're the programmer. You get to decide where you're going, where you want to do this and you know making you know making a bad decision can affect the the running of your program.

88
00:14:13.290 --> 00:14:24.930
William Cheng: Okay, so, yeah. So, so I just want to mention that meant this the spec doesn't say exactly when you should do this right, so part of the exercise is to sort of go through this kind of thinking and try to sort of figure out what is reasonable and what is not.

89
00:14:26.070 --> 00:14:26.310
Okay.

90
00:14:28.740 --> 00:14:30.060
William Cheng: Thank you. You're welcome.

91
00:14:31.860 --> 00:14:36.450
William Cheng: Okay. So, by the way, you know, the weekend is coming. And if you continue to have trouble.

92
00:14:37.320 --> 00:14:43.440
William Cheng: You know, just just just send me email, right, and tell me what you're you know what you're seeing. And then I you know I can actually give you

93
00:14:44.040 --> 00:14:54.240
William Cheng: Additional pointer of, you know, what will you know what does that break point and I'm a again, if you look at this cut over here. There are quite a few places you can set break point right so you can either set a breakpoint everywhere.

94
00:14:55.020 --> 00:15:10.560
William Cheng: So it's so it's so and and what you're trying to do is to find the first time at the bug happen, right. You want to catch as early as early as possible because the later you catch it, that the worst things that the worst things will get ok ok ok

95
00:15:15.090 --> 00:15:17.190
Lalit Gupta: OK, I have a question regarding my luck.

96
00:15:18.570 --> 00:15:22.080
Lalit Gupta: In the lecture it I think you've talked about the can we do Malik.

97
00:15:23.250 --> 00:15:29.460
Lalit Gupta: We get a page of memory, which is I think you said 4K of memory for k of memory.

98
00:15:30.660 --> 00:15:42.870
William Cheng: So that's how the colonel allocate memory. So we called Malik and three, you know, so. So since we use the first memory Alka alligator. Then in that case you don't get multiple four kilobytes.

99
00:15:45.900 --> 00:15:49.920
William Cheng: So the multiple four kilobytes. That's only done when you're using the buddy system.

100
00:15:51.060 --> 00:16:00.930
Lalit Gupta: So, so in our computer in the inevitable prediction that we are doing in our laptop. So what is the thing that is happening isn't doing like our

101
00:16:02.220 --> 00:16:03.750
Lalit Gupta: Page event vendor Malik.

102
00:16:04.140 --> 00:16:11.610
William Cheng: On so so so there are three memory alligator right one is in the user space in user space we're doing the first fit memory alligator.

103
00:16:12.120 --> 00:16:19.170
William Cheng: OK. And then inside the Colonel, they use two different kinds of memory alligator. One is the buddy system and the other one is a slap alligator.

104
00:16:21.270 --> 00:16:29.700
Lalit Gupta: Okay, so you. So you're saying that if you do in the program. We do Malak, let's say, a size of

105
00:16:30.840 --> 00:16:35.160
Lalit Gupta: A size of and then it will give only four byte of memory, right.

106
00:16:35.340 --> 00:16:36.210
William Cheng: Exactly, yes.

107
00:16:36.780 --> 00:16:38.280
Lalit Gupta: Or if I wanted to give more than that.

108
00:16:39.030 --> 00:16:44.400
William Cheng: It will give the exact well so so again because of the implementation their boundary tags. Right.

109
00:16:45.660 --> 00:16:55.470
William Cheng: Okay. So we asked for four bytes over here. Here's four bytes of memory. We're going to have boundary tack on top and boundary tag on the bottom right in the lecture, I said, You know, I

110
00:16:56.010 --> 00:17:02.220
William Cheng: Sort of mentioned that we're going to be really, really wastefully memory. So we're going to have eight bytes of boundary tag at the top and a bouncer budget budget, I get

111
00:17:03.300 --> 00:17:10.980
William Cheng: Some of the memory allocated. There are more memory efficient, maybe the only have four bytes of memory at the top of four bytes of boundary tag at the top and for by the boundary tagging the bottom

112
00:17:11.190 --> 00:17:15.690
William Cheng: So in that case, we tried to call Malik for you're going to end up getting a getting tall bites.

113
00:17:17.010 --> 00:17:22.200
Lalit Gupta: Yes, yes, yes. Yeah, but we have we have the access to. We have is only for the Fulbright

114
00:17:23.130 --> 00:17:31.050
William Cheng: While the you can access anything. Is that your is that is how you address space. You're just not supposed to touch anything outside these four bytes.

115
00:17:32.160 --> 00:17:41.280
Lalit Gupta: Yes. So, so when so so why does the colonel on to ask for for pay your memory.

116
00:17:41.850 --> 00:17:43.800
William Cheng: Because first America is too slow.

117
00:17:45.570 --> 00:17:51.330
William Cheng: Right. I mean, so, so over here, there's a slide that says you know that that is that the problem with the first remember alligator.

118
00:17:52.500 --> 00:17:54.240
William Cheng: Where is that, let me take a look

119
00:17:55.860 --> 00:18:00.330
William Cheng: Right, so first memory alligator. It's order n. In the worst case,

120
00:18:01.410 --> 00:18:10.050
William Cheng: So it's not suitable to use inside the colonel instead of Colonel, we need to have predictable worst case performance. So therefore, we use the buddy system and Assad alligator.

121
00:18:12.360 --> 00:18:18.690
Lalit Gupta: Okay. And why is it like okay to you ok to use the sausage algorithm in the user space.

122
00:18:20.460 --> 00:18:21.150
William Cheng: Um,

123
00:18:22.650 --> 00:18:23.340
William Cheng: Well, I guess.

124
00:18:25.170 --> 00:18:40.890
William Cheng: That's sort of a design choice. So typically, what's done is that you know that people found that this is acceptable. I heard numbers, something like, you know, whenever you write a program your programs spent about 20% of its time in Malaga free

125
00:18:42.060 --> 00:18:53.520
William Cheng: Okay, so your program is losing performance by 20% because they're using Malika free. Okay, so is that acceptable. Some people will say yes, some people say no. I guess I'll get different people at different opinion.

126
00:18:55.020 --> 00:19:03.660
Lalit Gupta: Okay, so you when you are saying 20% of the time. This means 20% of the time the user space is calling or the the coroner spaces calling this

127
00:19:05.250 --> 00:19:10.650
William Cheng: Doesn't use Malcolm free so so only the user space program use Malcolm free. So, for example, your warm up to

128
00:19:11.430 --> 00:19:21.750
William Cheng: Okay, the one if you measure the number of CPU second that's using. Okay. Some people will say that it's going to spend 20% of the time inside Malcolm free

129
00:19:24.720 --> 00:19:25.290
Lalit Gupta: Okay.

130
00:19:25.980 --> 00:19:32.310
William Cheng: So that sounds really big, whereas I sounds like a lot of time is wasted on that volume free, but now some people say, that's fine.

131
00:19:34.440 --> 00:19:36.150
Lalit Gupta: Okay, so. Okay.

132
00:19:37.620 --> 00:19:45.240
William Cheng: So some people say that if you want to use a buddy system in the user space program you will be wasting so much memory that it become not acceptable.

133
00:19:46.650 --> 00:19:48.900
Lalit Gupta: Okay, basically, to kind of a trade off like

134
00:19:48.900 --> 00:19:57.000
William Cheng: Yeah. Everything is a trade off. There's, there's, you know, very often, you know, I mean, for, for example, if you compare first fit and and best fit.

135
00:19:57.870 --> 00:20:09.060
William Cheng: You know, people study that to death and eventually they say well you know you should always use first fit because most of the time, it works better than best fit. Okay, but it's still there possible. You know, it is possible that best fit can still do better.

136
00:20:10.110 --> 00:20:16.260
William Cheng: But, you know, sort of, people will eventually come up with a conclusion that first fit is always better than best fit on the average

137
00:20:17.070 --> 00:20:26.220
William Cheng: Okay, but a lot of time is just trade off. Right, very, very rare that you have something to say something is always better than the other one. And typically, you know, if we can prove it mathematically, just as better than B and that's great.

138
00:20:27.030 --> 00:20:31.050
William Cheng: But you need to somehow sort of demonstrate is greater than B and but most of the time is just trade offs.

139
00:20:32.280 --> 00:20:36.810
William Cheng: I mean grad school. We're going to see all kinds of trade offs. Very, very rarely something is better than me.

140
00:20:38.070 --> 00:20:38.640
Lalit Gupta: Yeah.

141
00:20:39.930 --> 00:20:42.120
Lalit Gupta: So in the kernel space.

142
00:20:43.980 --> 00:20:45.330
Lalit Gupta: By default, it uses this

143
00:20:46.500 --> 00:20:50.040
Lalit Gupta: Body system. And then if we want to allocate some kind of

144
00:20:52.140 --> 00:20:59.190
Lalit Gupta: Some kind of space, such as PCB, etc. We kind of thing. Then it, then it will use this slab elevator. Right.

145
00:20:59.580 --> 00:21:13.680
William Cheng: Right, so the slap allocate it is also think about the slap alligator is building on top of the buddy system, right, because in order for you to get a slap you need to, you know, ask the buses and say, Hey, you know, give me 64 kilobytes. Give me 16 kilobytes or something like that.

146
00:21:14.670 --> 00:21:19.560
Lalit Gupta: Okay, and then we will structure this thing for tangible structure just for globe. Right.

147
00:21:19.800 --> 00:21:23.850
William Cheng: Yeah, we'll put them into this. So once we get a big slap right like a 64 kilobytes.

148
00:21:24.030 --> 00:21:30.870
William Cheng: We're gonna, you know, put them put all the process control blah in them into array and then we're going to allocate them one at a time, right, then it's going to be real fast.

149
00:21:31.770 --> 00:21:35.400
Lalit Gupta: Okay. Okay, this one is not clear. Thank you.

150
00:21:36.180 --> 00:21:36.450
Okay.

151
00:21:40.320 --> 00:21:45.900
William Cheng: Yeah, so our kernel source code come with both the buddy system and the you know the

152
00:21:46.440 --> 00:21:54.390
William Cheng: The snap alligator. So, you know, if you have, if you have time after semester is over, you can try to study it and it's extremely complicated.

153
00:21:56.850 --> 00:21:57.240
William Cheng: Actually

154
00:21:58.440 --> 00:22:04.200
William Cheng: You know, I'm also not saying that first memory alpha is not complicated when you're doing your kernel three

155
00:22:04.680 --> 00:22:10.050
William Cheng: Some of the bugs, you're going to encounter is because it because it's due to the first memory alligator.

156
00:22:10.710 --> 00:22:19.530
William Cheng: So, so you're going to get a little bit of familiar. You gotta get a little bit of a little bit familiar with the first MMA alligator and and then you got to see that's also pretty complicated.

157
00:22:20.460 --> 00:22:27.900
William Cheng: Okay, so all these memory alligator because they have to work all the time and they have to be fairly efficient Sunday. In the end, they're all very, very complicated.

158
00:22:30.270 --> 00:22:31.470
William Cheng: Yeah, it's crazy.

159
00:22:41.670 --> 00:22:54.450
William Cheng: So today we don't have, like, you know, two students in the room, any, any more questions. If there's no questions, then I'm just going to you know get stop recording let you guys talk a little bit and then we'll close it down. Yeah.

160
00:22:54.540 --> 00:22:55.740
Lalit Gupta: I have no question. Sure.

161
00:22:55.770 --> 00:22:56.130
William Cheng: Go ahead.

162
00:22:59.340 --> 00:22:59.490
Oh,

163
00:23:00.540 --> 00:23:01.260
Lalit Gupta: Yeah, no question.

164
00:23:01.740 --> 00:23:02.040
Okay.

165
00:23:03.990 --> 00:23:09.180
William Cheng: Alright, so, so in this case I'm gonna stop recording. Okay, let's see.