1
00:00:00,120 --> 00:00:02,500
The following content is
provided under a Creative

2
00:00:02,500 --> 00:00:03,910
Commons license.

3
00:00:03,910 --> 00:00:06,950
Your support will help MIT
OpenCourseWare continue to

4
00:00:06,950 --> 00:00:10,600
offer high quality educational
resources for free.

5
00:00:10,600 --> 00:00:13,500
To make a donation or view
additional materials from

6
00:00:13,500 --> 00:00:17,430
hundreds of MIT courses, visit
MIT OpenCourseWare at

7
00:00:17,430 --> 00:00:18,680
ocw.mit.edu.

8
00:00:21,720 --> 00:00:24,580
SAMAN AMARASINGHE: Welcome
to the final competition.

9
00:00:24,580 --> 00:00:27,060
So you guys survived this far.

10
00:00:27,060 --> 00:00:28,160
And I don't know how.

11
00:00:28,160 --> 00:00:30,760
I think, probably, with
very little sleep.

12
00:00:30,760 --> 00:00:34,690
And knowing all the emails that
I have seen and all the

13
00:00:34,690 --> 00:00:36,060
check-ins.--

14
00:00:36,060 --> 00:00:37,845
I don't know how many
people had--

15
00:00:37,845 --> 00:00:39,960
Who had any sleep within
last 24 hours?

16
00:00:39,960 --> 00:00:40,960
AUDIENCE: Haha.

17
00:00:40,960 --> 00:00:42,560
SAMAN AMARASINGHE: Oh, OK.

18
00:00:42,560 --> 00:00:43,946
[LAUGHTER]

19
00:00:43,946 --> 00:00:44,650
OK.

20
00:00:44,650 --> 00:00:49,000
And also what we looked at was
the quality of the results

21
00:00:49,000 --> 00:00:53,300
keep improving because he was
running tests to check

22
00:00:53,300 --> 00:00:54,270
what's going on.

23
00:00:54,270 --> 00:00:55,550
And when we moved to
a machine, the

24
00:00:55,550 --> 00:00:56,490
things ran very fast.

25
00:00:56,490 --> 00:00:57,660
We're like, wait a minute.

26
00:00:57,660 --> 00:00:58,690
Why is this running that fast?

27
00:00:58,690 --> 00:01:02,010
Because people had checked in
more and more, faster and

28
00:01:02,010 --> 00:01:02,750
faster versions.

29
00:01:02,750 --> 00:01:05,120
So things improved, not because
we checked in faster

30
00:01:05,120 --> 00:01:07,470
machines, but you guys actually
did a lot of work.

31
00:01:07,470 --> 00:01:11,300
So first before we
get started, I

32
00:01:11,300 --> 00:01:12,630
want to thank Akamai.

33
00:01:12,630 --> 00:01:16,900
So we and Akamai got
this nice place.

34
00:01:16,900 --> 00:01:18,110
And then we are going
to give a demo.

35
00:01:18,110 --> 00:01:20,630
And we're going to have a party
sponsored by Akamai,

36
00:01:20,630 --> 00:01:21,300
which is really cool.

37
00:01:21,300 --> 00:01:22,890
So you guys actually
see something--

38
00:01:22,890 --> 00:01:25,810
what the real-life people who
care about performance do.

39
00:01:25,810 --> 00:01:26,850
So it will be really fun.

40
00:01:26,850 --> 00:01:31,640
And also, I also want to thank
most of our MIT POSSE that

41
00:01:31,640 --> 00:01:34,850
help you guys a lot in meeting
with you individually.

42
00:01:34,850 --> 00:01:37,580
So we have people from Akamai,
[UNINTELLIGIBLE]

43
00:01:37,580 --> 00:01:42,580
Digital, Google, Lockheed
Martin, MathWorks, Wordica,

44
00:01:42,580 --> 00:01:45,560
VMWare and couple of independent
consultants who

45
00:01:45,560 --> 00:01:47,360
basically put lot of their
sweat and edge

46
00:01:47,360 --> 00:01:49,230
in helping you guys.

47
00:01:49,230 --> 00:01:51,185
CHARLES LEISERSON:
And Microsoft.

48
00:01:51,185 --> 00:01:53,400
SAMAN AMARASINGHE:
And Microsoft.

49
00:01:53,400 --> 00:01:54,750
CHARLES LEISERSON: So let's
give a hand for

50
00:01:54,750 --> 00:01:56,226
the posse, MIT POSSE.

51
00:01:56,226 --> 00:01:56,719
[LAUGHTER]

52
00:01:56,719 --> 00:02:00,170
CHARLES LEISERSON: And a
couple of the sponsors.

53
00:02:00,170 --> 00:02:03,050
SAMAN AMARASINGHE: So here
is the process we

54
00:02:03,050 --> 00:02:04,090
are going to do today.

55
00:02:04,090 --> 00:02:05,620
Today's about the competition.

56
00:02:05,620 --> 00:02:08,919
So we've set up a very strict
set of rules that we are going

57
00:02:08,919 --> 00:02:11,410
to go through, which is not
going to be the way we are

58
00:02:11,410 --> 00:02:13,690
going to grade you guys.

59
00:02:13,690 --> 00:02:15,260
Because we are going to be
little bit more lenient.

60
00:02:15,260 --> 00:02:17,820
So this is the set of rules
we are going to do.

61
00:02:17,820 --> 00:02:20,960
So we had a lot of these images
that you get submitted.

62
00:02:20,960 --> 00:02:24,830
And some images were
not exactly same.

63
00:02:24,830 --> 00:02:25,970
People did little bit things.

64
00:02:25,970 --> 00:02:27,330
They made them look little
bit different.

65
00:02:27,330 --> 00:02:30,355
So what they said was instead
of we're deciding.

66
00:02:30,355 --> 00:02:35,790
We will let you guys work on
images which can participate.

67
00:02:35,790 --> 00:02:36,340
OK?

68
00:02:36,340 --> 00:02:39,330
So don't be harsh
to your friends.

69
00:02:39,330 --> 00:02:43,100
So what we'll do, we'll first
go through the images.

70
00:02:43,100 --> 00:02:48,630
And then well you guys can say
which images should be part of

71
00:02:48,630 --> 00:02:49,610
the competition.

72
00:02:49,610 --> 00:02:51,900
This is for the competition.

73
00:02:51,900 --> 00:02:56,650
Then at the end, after you get
the Akamai sponsored prize--

74
00:02:56,650 --> 00:02:57,770
not for the grading.

75
00:02:57,770 --> 00:03:00,440
So the fact that the images
start to count, that doesn't

76
00:03:00,440 --> 00:03:02,080
mean you get zero
in the grade.

77
00:03:02,080 --> 00:03:04,070
But you won't get to
participate in the

78
00:03:04,070 --> 00:03:04,820
competition.

79
00:03:04,820 --> 00:03:07,670
So after that, what
we have done is we

80
00:03:07,670 --> 00:03:08,770
looked at all your projects.

81
00:03:08,770 --> 00:03:11,830
Ran them on our 12
core machine.

82
00:03:11,830 --> 00:03:14,320
And we will show you
and seed them.

83
00:03:14,320 --> 00:03:16,980
We'll say here is the order
of everything happening.

84
00:03:16,980 --> 00:03:18,310
And there are--

85
00:03:18,310 --> 00:03:21,747
AUDIENCE: [LAUGHTER]

86
00:03:21,747 --> 00:03:24,310
SAMAN AMARASINGHE:
There are some--

87
00:03:24,310 --> 00:03:28,850
Basically, what we'll do is
we'll pick out the top set of

88
00:03:28,850 --> 00:03:30,410
seeds in them because
we don't have time

89
00:03:30,410 --> 00:03:31,850
to run through everybody.

90
00:03:31,850 --> 00:03:35,780
And basically, a lot of these
things live are not for the 48

91
00:03:35,780 --> 00:03:36,980
core machine.

92
00:03:36,980 --> 00:03:38,750
And then the--

93
00:03:38,750 --> 00:03:42,680
accepted image that ran faster,
got generated faster

94
00:03:42,680 --> 00:03:45,680
in the 48 core machine,
wins the prize.

95
00:03:45,680 --> 00:03:48,660
So that doesn't mean that when
you start grading, we will

96
00:03:48,660 --> 00:03:51,060
look at all the images and we
will do some averaging.

97
00:03:51,060 --> 00:03:52,870
We will do a lot more fair.

98
00:03:52,870 --> 00:03:53,900
But hey, this is
a competition.

99
00:03:53,900 --> 00:03:57,850
We just need to find the top
guy so we have a top group.

100
00:03:57,850 --> 00:04:00,380
So we have a much more
faster thing.

101
00:04:00,380 --> 00:04:03,940
SAMAN AMARASINGHE: So
to get started.

102
00:04:03,940 --> 00:04:06,100
So here are some of the images
people generated.

103
00:04:06,100 --> 00:04:10,530
So what we want to do is show
the reference image.

104
00:04:10,530 --> 00:04:11,830
That's the reference image.

105
00:04:11,830 --> 00:04:14,715
And what we'll do is we'll first
go through everybody's

106
00:04:14,715 --> 00:04:17,430
images generated, so you
get a feel for all the

107
00:04:17,430 --> 00:04:18,800
variations out there.

108
00:04:18,800 --> 00:04:22,820
And then after that, we'll
ask you to go out.

109
00:04:22,820 --> 00:04:25,010
We'll ask you if you don't
want that image

110
00:04:25,010 --> 00:04:27,170
to be part of this.

111
00:04:27,170 --> 00:04:28,390
[LAUGHTER]

112
00:04:28,390 --> 00:04:29,400
SAMAN AMARASINGHE: You'll see.

113
00:04:29,400 --> 00:04:29,870
OK.

114
00:04:29,870 --> 00:04:31,420
So this is the reference
image.

115
00:04:31,420 --> 00:04:33,840
Why don't you go through--

116
00:04:33,840 --> 00:04:34,210
Charles.

117
00:04:34,210 --> 00:04:34,580
Mike.

118
00:04:34,580 --> 00:04:35,290
AUDIENCE: Yep?

119
00:04:35,290 --> 00:04:36,820
SAMAN AMARASINGHE: Why
don't you go through

120
00:04:36,820 --> 00:04:37,870
the rest of the images.

121
00:04:37,870 --> 00:04:39,140
AUDIENCE: OK.

122
00:04:39,140 --> 00:04:45,120
So before we we get into the
images that we actually

123
00:04:45,120 --> 00:04:48,880
rendered in the last few hours,
I just wanted to like

124
00:04:48,880 --> 00:04:49,990
honorable mention for this.

125
00:04:49,990 --> 00:04:51,170
I think this is the coolest.

126
00:04:51,170 --> 00:04:54,570
I'm sure you guys have been
looking at a whole variety of

127
00:04:54,570 --> 00:04:59,930
broken ray tracer images But I
think this is pretty much the

128
00:04:59,930 --> 00:05:00,840
coolest thing I've seen.

129
00:05:00,840 --> 00:05:05,040
So honorable mention for that.

130
00:05:05,040 --> 00:05:07,390
And so now let's look
at things that

131
00:05:07,390 --> 00:05:08,645
people actually submitted.

132
00:05:11,400 --> 00:05:12,910
First we're going to go through
all the images and

133
00:05:12,910 --> 00:05:13,910
sort of look at the range.

134
00:05:13,910 --> 00:05:16,060
And then we'll go back
and vote on them.

135
00:05:16,060 --> 00:05:17,760
And so I'm just going to flip
through the ones on the right

136
00:05:17,760 --> 00:05:20,866
and give a couple
seconds to each.

137
00:05:20,866 --> 00:05:21,990
SAMAN AMARASINGHE: So
see if they can

138
00:05:21,990 --> 00:05:25,522
identify the artifacts.

139
00:05:25,522 --> 00:05:28,826
AUDIENCE: Which one?

140
00:05:28,826 --> 00:05:31,120
CHARLES LEISERSON: You got to
say when you're going because

141
00:05:31,120 --> 00:05:32,750
otherwise people aren't going
to realize that you've been

142
00:05:32,750 --> 00:05:33,900
through three of them already.

143
00:05:33,900 --> 00:05:35,986
Here's image one.

144
00:05:35,986 --> 00:05:37,153
SAMAN AMARASINGHE: Image--

145
00:05:37,153 --> 00:05:38,800
AUDIENCE: Here's the next one.

146
00:05:38,800 --> 00:05:41,161
SAMAN AMARASINGHE: Image two.

147
00:05:41,161 --> 00:05:42,590
CHARLES LEISERSON:
That's five.

148
00:05:42,590 --> 00:05:43,430
Four, five.

149
00:05:43,430 --> 00:05:45,732
SAMAN AMARASINGHE:
Image three.

150
00:05:45,732 --> 00:05:59,142
image four, image five, image
six, seven, eight, nine, ten,

151
00:05:59,142 --> 00:06:09,410
11, SAMAN AMARASINGHE: 12,
13, 14, 15, OK, 15.

152
00:06:09,410 --> 00:06:12,535
There are some groups that we
could not get the rendering.

153
00:06:12,535 --> 00:06:15,450
Either it crashed, or after
ten minutes it was still

154
00:06:15,450 --> 00:06:16,420
hadn't produced an image.

155
00:06:16,420 --> 00:06:18,240
So they're not a part
of the competition.

156
00:06:18,240 --> 00:06:20,230
And of course we run them
for the finals.

157
00:06:20,230 --> 00:06:22,120
OK, now the vote time.

158
00:06:22,120 --> 00:06:24,830
OK, so this image, in or out?

159
00:06:24,830 --> 00:06:27,800
Who thinks out?

160
00:06:27,800 --> 00:06:29,830
Whoa, you guys are
a harsh crowd.

161
00:06:29,830 --> 00:06:30,710
OK.

162
00:06:30,710 --> 00:06:31,160
[LAUGHTER]

163
00:06:31,160 --> 00:06:32,080
CHARLES LEISERSON:
Who thinks in?

164
00:06:32,080 --> 00:06:32,832
Who thinks in?

165
00:06:32,832 --> 00:06:34,160
SAMAN AMARASINGHE:
Who thinks in?

166
00:06:34,160 --> 00:06:34,700
CHARLES LEISERSON: We've
got to get you both.

167
00:06:34,700 --> 00:06:35,100
Who thinks in?

168
00:06:35,100 --> 00:06:36,460
SAMAN AMARASINGHE:
So we know--

169
00:06:36,460 --> 00:06:37,460
AUDIENCE: Do we get
screwed for this?

170
00:06:37,460 --> 00:06:38,280
[LAUGHTER]

171
00:06:38,280 --> 00:06:41,070
SAMAN AMARASINGHE:
Imagine that.

172
00:06:41,070 --> 00:06:41,862
OK.

173
00:06:41,862 --> 00:06:43,250
That image is out.

174
00:06:43,250 --> 00:06:44,950
You can take notes, OK.

175
00:06:44,950 --> 00:06:46,200
[INTERPOSING VOICES]

176
00:06:50,860 --> 00:06:51,580
SAMAN AMARASINGHE:
OK, next image.

177
00:06:51,580 --> 00:06:52,590
AUDIENCE: Number one's out.

178
00:06:52,590 --> 00:06:54,405
SAMAN AMARASINGHE: OK,
this one, in or out?

179
00:06:54,405 --> 00:06:55,160
Out.

180
00:06:55,160 --> 00:06:55,565
OK.

181
00:06:55,565 --> 00:06:56,420
AUDIENCE: Probably the same.

182
00:06:56,420 --> 00:06:57,770
SAMAN AMARASINGHE:
OK, it's out.

183
00:06:57,770 --> 00:07:00,390
AUDIENCE: It's got the funny
ring at the bottom, which I

184
00:07:00,390 --> 00:07:02,330
guess that's what most of
you are picking up on.

185
00:07:02,330 --> 00:07:07,006
I think that's kind of
disqualifying for the derby.

186
00:07:07,006 --> 00:07:08,590
SAMAN AMARASINGHE: OK,
how about this?

187
00:07:08,590 --> 00:07:08,670
Out?

188
00:07:08,670 --> 00:07:08,960
In or out?

189
00:07:08,960 --> 00:07:09,750
AUDIENCE: Out.

190
00:07:09,750 --> 00:07:10,510
SAMAN AMARASINGHE: Out.

191
00:07:10,510 --> 00:07:11,430
Whoa.

192
00:07:11,430 --> 00:07:13,170
CHARLES LEISERSON: Wow,
you guys are vicious.

193
00:07:13,170 --> 00:07:14,446
SAMAN AMARASINGHE: Exactly.

194
00:07:14,446 --> 00:07:17,150
OK, how about this?

195
00:07:17,150 --> 00:07:21,050
How many people think
they should be out?

196
00:07:21,050 --> 00:07:23,320
OK, I guess that'll
get in here.

197
00:07:23,320 --> 00:07:23,980
I guess--

198
00:07:23,980 --> 00:07:25,380
OK, that image is in.

199
00:07:25,380 --> 00:07:25,930
OK.

200
00:07:25,930 --> 00:07:28,380
Make sure it's in.

201
00:07:28,380 --> 00:07:30,022
How about this one?

202
00:07:30,022 --> 00:07:32,758
AUDIENCE: It's okay.

203
00:07:32,758 --> 00:07:34,688
SAMAN AMARASINGHE: Who
says it's out?

204
00:07:34,688 --> 00:07:35,938
AUDIENCE: [INTERPOSING VOICES]

205
00:07:38,592 --> 00:07:40,460
SAMAN AMARASINGHE: OK,
Who says it's in?

206
00:07:44,860 --> 00:07:46,000
I think it's in.

207
00:07:46,000 --> 00:07:47,620
CHARLES LEISERSON: You should
look at the water.

208
00:07:47,620 --> 00:07:48,100
SAMAN AMARASINGHE: Water.

209
00:07:48,100 --> 00:07:49,000
AUDIENCE: The water has holes.

210
00:07:49,000 --> 00:07:50,720
CHARLES LEISERSON: The
water has holes.

211
00:07:50,720 --> 00:07:51,600
That's what's going on.

212
00:07:51,600 --> 00:07:52,470
The water has holes.

213
00:07:52,470 --> 00:07:54,831
AUDIENCE: [INTERPOSING VOICES]

214
00:07:54,831 --> 00:07:56,142
It's the only one we're going
to be able to reference.

215
00:07:56,142 --> 00:07:58,110
SAMAN AMARASINGHE: OK, so what's
the this is in charge?

216
00:07:58,110 --> 00:07:59,132
CHARLES LEISERSON: Let's
do it again.

217
00:07:59,132 --> 00:08:00,000
SAMAN AMARASINGHE: OK.

218
00:08:00,000 --> 00:08:01,560
CHARLES LEISERSON: OK, you
have to point out--

219
00:08:01,560 --> 00:08:02,310
SAMAN AMARASINGHE: What's
the problem?

220
00:08:02,310 --> 00:08:03,500
CHARLES LEISERSON: --what's
going on in the image.

221
00:08:03,500 --> 00:08:04,900
AUDIENCE: Yeah.

222
00:08:04,900 --> 00:08:05,640
I think we should--

223
00:08:05,640 --> 00:08:07,510
I mean, do we want to get people
to like say why they

224
00:08:07,510 --> 00:08:09,083
think it's in or out?

225
00:08:09,083 --> 00:08:11,410
CHARLES LEISERSON: It's helpful
to say what's in the

226
00:08:11,410 --> 00:08:11,510
image, though.

227
00:08:11,510 --> 00:08:12,075
AUDIENCE: Yeah.

228
00:08:12,075 --> 00:08:13,005
SAMAN AMARASINGHE: OK.

229
00:08:13,005 --> 00:08:13,935
Let's take a vote again.

230
00:08:13,935 --> 00:08:16,516
How many people think
it should be out?

231
00:08:16,516 --> 00:08:18,040
AUDIENCE: OK, that's out.

232
00:08:18,040 --> 00:08:18,920
OK, yeah.

233
00:08:18,920 --> 00:08:19,590
SAMAN AMARASINGHE:
OK, next image.

234
00:08:19,590 --> 00:08:21,368
AUDIENCE: Number 5,
on the slide.

235
00:08:21,368 --> 00:08:23,370
SAMAN AMARASINGHE:
OK, in or out?

236
00:08:23,370 --> 00:08:25,658
How many people think
this should be out?

237
00:08:25,658 --> 00:08:27,495
AUDIENCE: Is there
any problem?

238
00:08:27,495 --> 00:08:28,900
I mean, what's the--

239
00:08:28,900 --> 00:08:30,308
[INTERPOSING VOICES]

240
00:08:30,308 --> 00:08:32,010
[LAUGHTER]

241
00:08:32,010 --> 00:08:34,850
SAMAN AMARASINGHE: I don't
think there's a problem.

242
00:08:34,850 --> 00:08:35,720
OK, good.

243
00:08:35,720 --> 00:08:37,039
Next one.

244
00:08:37,039 --> 00:08:38,700
How many think this
should be--

245
00:08:38,700 --> 00:08:39,210
OK.

246
00:08:39,210 --> 00:08:41,510
That's unanimous, that's out.

247
00:08:41,510 --> 00:08:46,060
How many think this
should be out?

248
00:08:46,060 --> 00:08:50,361
OK, I guess that's in.

249
00:08:50,361 --> 00:08:51,210
This another one?

250
00:08:51,210 --> 00:08:54,220
AUDIENCE: Yeah, I think it's
interesting to flip back and

251
00:08:54,220 --> 00:08:55,955
forth and see the difference
between these two.

252
00:08:55,955 --> 00:08:57,189
Like, these are very
close, right?

253
00:08:57,189 --> 00:08:58,147
AUDIENCE: Yeah.

254
00:08:58,147 --> 00:09:00,063
It's obvious they're
using this one.

255
00:09:00,063 --> 00:09:05,332
[LAUGHTER]

256
00:09:05,332 --> 00:09:07,370
SAMAN AMARASINGHE: OK,
how many think they

257
00:09:07,370 --> 00:09:08,770
should be be out?

258
00:09:11,770 --> 00:09:13,470
That group once is going
to kill everybody.

259
00:09:16,470 --> 00:09:17,635
[INTERPOSING VOICES]

260
00:09:17,635 --> 00:09:18,960
SAMAN AMARASINGHE: How
about this one?

261
00:09:18,960 --> 00:09:20,390
How many people think
this should be out?

262
00:09:20,390 --> 00:09:21,960
AUDIENCE: There's not
enough caustics.

263
00:09:21,960 --> 00:09:24,160
You don't see the floor.

264
00:09:24,160 --> 00:09:28,016
AUDIENCE: Yeah, that's
an interesting point.

265
00:09:28,016 --> 00:09:29,266
[INTERPOSING VOICES]

266
00:09:36,650 --> 00:09:38,680
AUDIENCE: Say what?

267
00:09:38,680 --> 00:09:41,462
AUDIENCE: We went past ones
that look the same.

268
00:09:41,462 --> 00:09:42,712
[INTERPOSING VOICES]

269
00:09:45,770 --> 00:09:47,350
AUDIENCE: It's really
not fair.

270
00:09:47,350 --> 00:09:49,448
When you have an image and
then put that image right

271
00:09:49,448 --> 00:09:50,732
after that it's not fair.

272
00:09:50,732 --> 00:09:52,310
SAMAN AMARASINGHE: We will keep
this in because there's

273
00:09:52,310 --> 00:09:54,630
few people want it out, but
it looks pretty good.

274
00:09:54,630 --> 00:09:56,760
OK, next image.

275
00:09:56,760 --> 00:09:58,840
OK, how many people think
this should be--

276
00:09:58,840 --> 00:10:02,500
OK, that's a clear out.

277
00:10:02,500 --> 00:10:03,951
How about this image?

278
00:10:03,951 --> 00:10:05,915
AUDIENCE: It looks OK.

279
00:10:05,915 --> 00:10:06,900
It's fine.

280
00:10:06,900 --> 00:10:07,410
SAMAN AMARASINGHE: It's fine.

281
00:10:07,410 --> 00:10:09,190
OK, anybody want it out?

282
00:10:09,190 --> 00:10:09,790
No?

283
00:10:09,790 --> 00:10:10,230
OK, good.

284
00:10:10,230 --> 00:10:11,280
That's in.

285
00:10:11,280 --> 00:10:12,125
Next image.

286
00:10:12,125 --> 00:10:14,192
How about this image?

287
00:10:14,192 --> 00:10:17,138
[INTERPOSING VOICES]

288
00:10:17,138 --> 00:10:18,611
[LAUGHTER]

289
00:10:18,611 --> 00:10:19,861
[INTERPOSING VOICES]

290
00:10:23,521 --> 00:10:25,475
CHARLES LEISERSON: We really
should start the reference--

291
00:10:25,475 --> 00:10:26,950
SAMAN AMARASINGHE: We need
another slide to see how many

292
00:10:26,950 --> 00:10:27,660
people want that out.

293
00:10:27,660 --> 00:10:29,580
CHARLES LEISERSON: --identical
images and

294
00:10:29,580 --> 00:10:31,500
people could be voting.

295
00:10:31,500 --> 00:10:33,750
[LAUGHTER]

296
00:10:33,750 --> 00:10:35,522
AUDIENCE: Is there an
explanation why all of the

297
00:10:35,522 --> 00:10:37,200
yellows have been so
far different from

298
00:10:37,200 --> 00:10:38,286
the reference image?

299
00:10:38,286 --> 00:10:40,642
[INTERPOSING VOICES]

300
00:10:40,642 --> 00:10:43,122
[LAUGHTER]

301
00:10:43,122 --> 00:10:45,590
SAMAN AMARASINGHE: So I guess
this image is in.

302
00:10:45,590 --> 00:10:48,150
OK, next one.

303
00:10:48,150 --> 00:10:48,990
How about this image?

304
00:10:48,990 --> 00:10:49,696
AUDIENCE: It's fine.

305
00:10:49,696 --> 00:10:50,480
SAMAN AMARASINGHE: Fine.

306
00:10:50,480 --> 00:10:51,710
OK, anybody want it out?

307
00:10:51,710 --> 00:10:52,220
No?

308
00:10:52,220 --> 00:10:52,510
OK.

309
00:10:52,510 --> 00:10:54,622
How about next one?

310
00:10:54,622 --> 00:10:55,872
[INTERPOSING VOICES]

311
00:11:00,390 --> 00:11:01,860
SAMAN AMARASINGHE: It
looks pretty good.

312
00:11:01,860 --> 00:11:03,555
AUDIENCE: It's pretty good.

313
00:11:03,555 --> 00:11:06,450
AUDIENCE: It's messed up on
the left and the right.

314
00:11:06,450 --> 00:11:08,050
SAMAN AMARASINGHE:
How many people--

315
00:11:08,050 --> 00:11:09,380
We have to count this one.

316
00:11:09,380 --> 00:11:10,070
This is too close.

317
00:11:10,070 --> 00:11:15,322
One, two, three, four, five,
six, seven, eight, nine, 10,

318
00:11:15,322 --> 00:11:18,540
11, 12, 13, 14, 15,
16, 17, 18, 19.

319
00:11:18,540 --> 00:11:19,480
19 out.

320
00:11:19,480 --> 00:11:21,792
Who says it should be in?

321
00:11:21,792 --> 00:11:23,756
[INTERPOSING VOICES]

322
00:11:23,756 --> 00:11:26,360
SAMAN AMARASINGHE: Oh,
I guess it's out.

323
00:11:26,360 --> 00:11:27,530
AUDIENCE: Oh, the number of
hands is more for out.

324
00:11:27,530 --> 00:11:28,305
AUDIENCE: OK, and that's it.

325
00:11:28,305 --> 00:11:29,800
SAMAN AMARASINGHE: That's it.

326
00:11:29,800 --> 00:11:33,670
So now, next thing we are going
to do is, we are going

327
00:11:33,670 --> 00:11:35,860
to show you a rank
in our seating.

328
00:11:35,860 --> 00:11:37,740
AUDIENCE: Yeah, so we
seeded you guys

329
00:11:37,740 --> 00:11:39,630
on the 12 core machines.

330
00:11:42,320 --> 00:11:45,630
So we seeded you guys based on
the time that it took for

331
00:11:45,630 --> 00:11:48,370
scene three on the
12 core machines.

332
00:11:48,370 --> 00:11:52,770
So here is the ranking so far.

333
00:11:52,770 --> 00:11:55,700
Top one is number one team.

334
00:11:55,700 --> 00:12:00,100
And we ranked about
the top 10 or so.

335
00:12:00,100 --> 00:12:02,060
SAMAN AMARASINGHE: Show
the real run time.

336
00:12:02,060 --> 00:12:05,995
AUDIENCE: I don't have the run
time, actually, for the seat.

337
00:12:05,995 --> 00:12:06,950
AUDIENCE: For scene one?

338
00:12:06,950 --> 00:12:08,190
AUDIENCE: Yeah.

339
00:12:08,190 --> 00:12:11,200
Actually, I can get
it real quick.

340
00:12:11,200 --> 00:12:12,660
SAMAN AMARASINGHE: You had
the graph, didn't you?

341
00:12:12,660 --> 00:12:13,910
AUDIENCE: I don't
have a graph.

342
00:12:16,724 --> 00:12:20,260
SAMAN AMARASINGHE: So here are
the rankings of the top ten.

343
00:12:20,260 --> 00:12:21,360
Do we have that graph?

344
00:12:21,360 --> 00:12:22,200
AUDIENCE: What?

345
00:12:22,200 --> 00:12:22,696
I don't have a graph.

346
00:12:22,696 --> 00:12:24,020
AUDIENCE: He has numbers.

347
00:12:24,020 --> 00:12:25,160
AUDIENCE: I just have
the numbers.

348
00:12:25,160 --> 00:12:25,480
[INTERPOSING VOICES]

349
00:12:25,480 --> 00:12:26,960
AUDIENCE: There's all kinds
of other numbers, too.

350
00:12:26,960 --> 00:12:27,452
[INTERPOSING VOICES]

351
00:12:27,452 --> 00:12:28,930
AUDIENCE: We'll say
unplug this.

352
00:12:28,930 --> 00:12:31,135
SAMAN AMARASINGHE: No,
nobody can see it.

353
00:12:31,135 --> 00:12:32,856
CHARLES LEISERSON: It's okay.

354
00:12:32,856 --> 00:12:35,520
AUDIENCE: Do we want to
get the rankings?

355
00:12:35,520 --> 00:12:35,820
SAMAN AMARASINGHE: No.

356
00:12:35,820 --> 00:12:38,350
Now, what we're going to do is
we're going to start from the

357
00:12:38,350 --> 00:12:44,220
bottom and start running it
on our 48 core machine.

358
00:12:44,220 --> 00:12:46,560
And while it's running, we want
that team to say a little

359
00:12:46,560 --> 00:12:49,260
bit about what they did to
get their performance.

360
00:12:49,260 --> 00:12:49,535
So--

361
00:12:49,535 --> 00:12:50,390
CHARLES LEISERSON: We need
to knock out the one--

362
00:12:50,390 --> 00:12:51,590
[INTERPOSING VOICES]

363
00:12:51,590 --> 00:12:53,720
SAMAN AMARASINGHE: Either come
prepared when you have formed

364
00:12:53,720 --> 00:12:57,100
your presentation, or make
your image run very slow.

365
00:12:57,100 --> 00:12:58,230
So you don't have to.

366
00:12:58,230 --> 00:12:58,750
So I guess--

367
00:12:58,750 --> 00:13:00,810
CHARLES LEISERSON: First of all
we have to knock out the

368
00:13:00,810 --> 00:13:02,840
ones whose images
did not render.

369
00:13:02,840 --> 00:13:03,470
SAMAN AMARASINGHE:
Oh, that's true.

370
00:13:03,470 --> 00:13:03,770
That's true.

371
00:13:03,770 --> 00:13:04,180
That's true.

372
00:13:04,180 --> 00:13:06,110
CHARLES LEISERSON: So maybe
they all get knocked out.

373
00:13:06,110 --> 00:13:08,865
And then everybody goes
home with an iPod.

374
00:13:08,865 --> 00:13:10,040
SAMAN AMARASINGHE: Or nobody.

375
00:13:10,040 --> 00:13:11,000
[LAUGHTER]

376
00:13:11,000 --> 00:13:12,250
AUDIENCE: Can we just
put the next one up?

377
00:13:14,890 --> 00:13:15,300
Here we go.

378
00:13:15,300 --> 00:13:16,461
Now we have time
for everybody.

379
00:13:16,461 --> 00:13:18,230
SAMAN AMARASINGHE:
OK, there we go.

380
00:13:18,230 --> 00:13:20,850
So this is the time for
all the images.

381
00:13:20,850 --> 00:13:22,830
So is there any of these images
today getting knocked

382
00:13:22,830 --> 00:13:23,810
out, can you check that?

383
00:13:23,810 --> 00:13:25,570
AUDIENCE: We're pretty close.

384
00:13:25,570 --> 00:13:26,910
SAMAN AMARASINGHE: Which
images get knocked out?

385
00:13:26,910 --> 00:13:28,735
AUDIENCE: So I'm looking at
them from bottom to top.

386
00:13:28,735 --> 00:13:30,646
AUDIENCE: This is the--

387
00:13:30,646 --> 00:13:31,670
I think this one's
knocked out.

388
00:13:31,670 --> 00:13:33,850
AUDIENCE: Number one,
two, and three.

389
00:13:33,850 --> 00:13:35,100
[INTERPOSING VOICES]

390
00:13:40,984 --> 00:13:44,630
AUDIENCE: I think David
Lamb and Victor

391
00:13:44,630 --> 00:13:45,410
Wang got knocked out.

392
00:13:45,410 --> 00:13:47,132
Because this is their
image here.

393
00:13:47,132 --> 00:13:48,670
SAMAN AMARASINGHE: That
got knocked out.

394
00:13:48,670 --> 00:13:50,734
The last image got
knocked out.

395
00:13:50,734 --> 00:13:52,602
[INTERPOSING VOICES]

396
00:13:52,602 --> 00:13:54,040
SAMAN AMARASINGHE: What's
the next group?

397
00:13:54,040 --> 00:13:59,458
AUDIENCE: In between--

398
00:13:59,458 --> 00:14:01,434
the A group--

399
00:14:01,434 --> 00:14:03,410
[INAUDIBLE]

400
00:14:03,410 --> 00:14:04,660
[INTERPOSING VOICES]

401
00:14:06,810 --> 00:14:08,830
AUDIENCE: This one
was the water.

402
00:14:08,830 --> 00:14:09,312
19.

403
00:14:09,312 --> 00:14:10,758
AUDIENCE: Yeah, that was.

404
00:14:10,758 --> 00:14:12,008
[INTERPOSING VOICES]

405
00:14:15,460 --> 00:14:18,434
SAMAN AMARASINGHE: Uh-oh,
they're getting knocked out.

406
00:14:18,434 --> 00:14:19,684
[INTERPOSING VOICES]

407
00:14:25,485 --> 00:14:27,916
SAMAN AMARASINGHE: OK, that
got knocked out, too.

408
00:14:27,916 --> 00:14:29,166
[INTERPOSING VOICES]

409
00:14:32,746 --> 00:14:34,950
SAMAN AMARASINGHE: Oh, man.

410
00:14:34,950 --> 00:14:37,940
Yeah, nobody competing
at this point.

411
00:14:37,940 --> 00:14:39,832
[LAUGHTER]

412
00:14:39,832 --> 00:14:41,082
[INTERPOSING VOICES]

413
00:14:44,090 --> 00:14:45,295
SAMAN AMARASINGHE:
That survives.

414
00:14:45,295 --> 00:14:46,390
That survives.

415
00:14:46,390 --> 00:14:46,755
Push!

416
00:14:46,755 --> 00:14:49,550
We have a survivor.

417
00:14:49,550 --> 00:14:53,434
So that might be a knockout
by default.

418
00:14:53,434 --> 00:14:55,904
[INTERPOSING VOICES]

419
00:14:55,904 --> 00:14:59,290
SAMAN AMARASINGHE:
Oh, that is out.

420
00:14:59,290 --> 00:15:00,110
Oh, man.

421
00:15:00,110 --> 00:15:01,986
AUDIENCE: Ours is
so close though.

422
00:15:01,986 --> 00:15:04,331
[LAUGHTER]

423
00:15:04,331 --> 00:15:06,750
[INTERPOSING VOICES]

424
00:15:06,750 --> 00:15:09,930
SAMAN AMARASINGHE:
That's all right.

425
00:15:09,930 --> 00:15:12,768
OK, I at least we have
two people right now.

426
00:15:12,768 --> 00:15:15,050
AUDIENCE: Wasn't that the one
that was really dark?

427
00:15:15,050 --> 00:15:17,770
AUDIENCE: This is the one that
was sort of darker--

428
00:15:17,770 --> 00:15:18,170
[INTERPOSING VOICES]

429
00:15:18,170 --> 00:15:18,946
AUDIENCE: Like they
don't have enough.

430
00:15:18,946 --> 00:15:20,196
[INTERPOSING VOICES]

431
00:15:22,348 --> 00:15:25,040
AUDIENCE: So here's the
two reference images.

432
00:15:25,040 --> 00:15:25,600
SAMAN AMARASINGHE:
We already voted.

433
00:15:25,600 --> 00:15:26,680
We can't vote again.

434
00:15:26,680 --> 00:15:27,105
[INTERPOSING VOICES]

435
00:15:27,105 --> 00:15:27,970
SAMAN AMARASINGHE: No--

436
00:15:27,970 --> 00:15:28,610
leave them knocked out.

437
00:15:28,610 --> 00:15:30,470
AUDIENCE: Some people voted
against it, but--

438
00:15:30,470 --> 00:15:30,660
[INTERPOSING VOICES]

439
00:15:30,660 --> 00:15:31,920
SAMAN AMARASINGHE: Yeah, but
it didn't get the quorum.

440
00:15:31,920 --> 00:15:33,540
It didn't get a quorum
to get voted out.

441
00:15:33,540 --> 00:15:36,202
[INTERPOSING VOICES]

442
00:15:36,202 --> 00:15:38,475
SAMAN AMARASINGHE: How
about other group?

443
00:15:38,475 --> 00:15:39,700
First one, let's look at that.

444
00:15:39,700 --> 00:15:41,094
AUDIENCE: Who's this?

445
00:15:41,094 --> 00:15:42,344
[INTERPOSING VOICES]

446
00:15:48,036 --> 00:15:49,910
SAMAN AMARASINGHE: OK,
that survives.

447
00:15:49,910 --> 00:15:51,160
That's good.

448
00:15:54,620 --> 00:15:57,325
Mangpo.

449
00:15:57,325 --> 00:15:59,440
OK.

450
00:15:59,440 --> 00:16:00,200
That survived, too.

451
00:16:00,200 --> 00:16:01,530
Two.

452
00:16:01,530 --> 00:16:04,622
That's all right.

453
00:16:04,622 --> 00:16:05,872
Li Chen.

454
00:16:08,302 --> 00:16:10,592
[INTERPOSING VOICES]

455
00:16:10,592 --> 00:16:11,842
SAMAN AMARASINGHE: Ooooo.

456
00:16:15,900 --> 00:16:16,250
[LAUGHTER]

457
00:16:16,250 --> 00:16:18,280
SAMAN AMARASINGHE: Oh, man.

458
00:16:18,280 --> 00:16:19,530
That's a...

459
00:16:24,470 --> 00:16:29,520
So that is a knockout
blow there.

460
00:16:29,520 --> 00:16:34,070
Now what we're going to do is,
we're going to run it on from

461
00:16:34,070 --> 00:16:35,900
48 core machine from
the bottom.

462
00:16:35,900 --> 00:16:39,230
So we will let other teams also,
if we have time, to come

463
00:16:39,230 --> 00:16:40,930
and explain what they did.

464
00:16:40,930 --> 00:16:43,050
We are running--

465
00:16:43,050 --> 00:16:43,680
who's now?

466
00:16:43,680 --> 00:16:45,920
AUDIENCE: This is a Dobs.

467
00:16:45,920 --> 00:16:46,890
SAMAN AMARASINGHE: Dobs.

468
00:16:46,890 --> 00:16:49,132
OK, why don't you guys
explain what you did.

469
00:16:49,132 --> 00:16:50,608
AUDIENCE: Come up here.

470
00:16:50,608 --> 00:16:52,084
AUDIENCE: Need the mic?

471
00:16:52,084 --> 00:16:55,036
AUDIENCE: I think
the mic is off.

472
00:16:55,036 --> 00:16:56,410
CHARLES LEISERSON: The
mic is on, but it's

473
00:16:56,410 --> 00:16:56,830
only for the video.

474
00:16:56,830 --> 00:16:58,114
SAMAN AMARASINGHE: Video.

475
00:16:58,114 --> 00:17:01,790
CHARLES LEISERSON: You have to
use the mic but you can't hear

476
00:17:01,790 --> 00:17:04,495
the speaking.

477
00:17:04,495 --> 00:17:09,140
AUDIENCE: So we parallelized
the ray tracing part.

478
00:17:09,140 --> 00:17:12,859
The two for loops for going
through all the

479
00:17:12,859 --> 00:17:13,869
pixels on the screen.

480
00:17:13,869 --> 00:17:18,069
And then we also parallelized
the main loop for caustic

481
00:17:18,069 --> 00:17:18,660
elimination.

482
00:17:18,660 --> 00:17:19,829
And we changed the--

483
00:17:19,829 --> 00:17:21,170
sort of the data structure--

484
00:17:21,170 --> 00:17:22,069
how that works.

485
00:17:22,069 --> 00:17:25,000
Used some like bounding
boxes to make the

486
00:17:25,000 --> 00:17:26,250
intersection faster.

487
00:17:28,585 --> 00:17:30,730
I don't know what else we did.

488
00:17:30,730 --> 00:17:32,200
AUDIENCE: We found ways
to get rid of--

489
00:17:32,200 --> 00:17:34,636
AUDIENCE: Mic, please.

490
00:17:34,636 --> 00:17:40,200
AUDIENCE: We found ways to get
rid of a lot of trigonometry,

491
00:17:40,200 --> 00:17:41,270
which actually helped a lot.

492
00:17:41,270 --> 00:17:46,970
And we got rid of the recursion
in locate photons,

493
00:17:46,970 --> 00:17:49,550
which was a major roadblock.

494
00:17:49,550 --> 00:17:50,370
SAMAN AMARASINGHE: Question.

495
00:17:50,370 --> 00:17:52,101
AUDIENCE: What did
you do for that?

496
00:17:52,101 --> 00:17:54,850
How'd you get rid of the trig?

497
00:17:54,850 --> 00:17:57,410
AUDIENCE: Repeat the question.

498
00:17:57,410 --> 00:18:01,355
AUDIENCE: They were converting
from XYZ coordinates into

499
00:18:01,355 --> 00:18:03,820
spherical coordinates and then
back again, over and over.

500
00:18:03,820 --> 00:18:05,320
So we just didn't do that.

501
00:18:08,785 --> 00:18:10,035
[INTERPOSING VOICES]

502
00:18:17,657 --> 00:18:18,571
[LAUGHTER]

503
00:18:18,571 --> 00:18:20,399
AUDIENCE: Slow down.

504
00:18:20,399 --> 00:18:21,630
SAMAN AMARASINGHE:
It slow down.

505
00:18:21,630 --> 00:18:21,980
Uh-oh.

506
00:18:21,980 --> 00:18:23,550
AUDIENCE: No.

507
00:18:23,550 --> 00:18:25,330
The seating was done
on scene one.

508
00:18:25,330 --> 00:18:25,630
Right?

509
00:18:25,630 --> 00:18:26,720
AUDIENCE: Oh, yeah.

510
00:18:26,720 --> 00:18:27,660
Oh.

511
00:18:27,660 --> 00:18:28,790
Wait--

512
00:18:28,790 --> 00:18:29,897
Apples and oranges.

513
00:18:29,897 --> 00:18:32,640
[INTERPOSING VOICES]

514
00:18:32,640 --> 00:18:34,910
I just ran scene three.

515
00:18:34,910 --> 00:18:36,122
[INTERPOSING VOICES]

516
00:18:36,122 --> 00:18:37,006
48.

517
00:18:37,006 --> 00:18:37,448
Scene three.

518
00:18:37,448 --> 00:18:38,780
So there might have
been a slip.

519
00:18:38,780 --> 00:18:39,540
SAMAN AMARASINGHE: No.

520
00:18:39,540 --> 00:18:43,010
But this scene one is
much more simpler.

521
00:18:43,010 --> 00:18:44,150
OK, good.

522
00:18:44,150 --> 00:18:45,490
So you're on the board now?

523
00:18:45,490 --> 00:18:46,990
[INTERPOSING VOICES]

524
00:18:46,990 --> 00:18:51,220
AUDIENCE: How is the seating
being done in scene one if it

525
00:18:51,220 --> 00:18:53,160
doesn't have a surface
that's--

526
00:18:53,160 --> 00:18:53,470
[INTERPOSING VOICES]

527
00:18:53,470 --> 00:18:55,350
AUDIENCE: --caustics.

528
00:18:55,350 --> 00:18:56,890
It might be that some
people did really

529
00:18:56,890 --> 00:18:57,940
fast on that one and--

530
00:18:57,940 --> 00:18:59,834
AUDIENCE: Because we ran
everybody's scene three.

531
00:18:59,834 --> 00:19:01,810
And we guarantee that the
top end teams are on it.

532
00:19:01,810 --> 00:19:04,774
[LAUGHTER]

533
00:19:04,774 --> 00:19:06,256
AUDIENCE: The results
are rigged.

534
00:19:06,256 --> 00:19:10,208
[LAUGHTER]

535
00:19:10,208 --> 00:19:12,060
SAMAN AMARASINGHE: He
detained the tape.

536
00:19:12,060 --> 00:19:14,380
OK, good.

537
00:19:14,380 --> 00:19:16,572
First team is in.

538
00:19:16,572 --> 00:19:17,490
Then the next team.

539
00:19:17,490 --> 00:19:18,880
AUDIENCE: That's Elizabeth.

540
00:19:18,880 --> 00:19:19,600
SAMAN AMARASINGHE: Elizabeth--

541
00:19:19,600 --> 00:19:21,541
AUDIENCE: Scott, right?

542
00:19:21,541 --> 00:19:23,530
SAMAN AMARASINGHE:
There we go.

543
00:19:23,530 --> 00:19:25,055
Why their thing is being run.

544
00:19:25,055 --> 00:19:26,936
AUDIENCE: Are you running it?

545
00:19:26,936 --> 00:19:27,810
AUDIENCE: All right.

546
00:19:27,810 --> 00:19:32,020
So I guess for the display
surface, the main optimization

547
00:19:32,020 --> 00:19:35,060
there was imagining two planes
at the top bound and the

548
00:19:35,060 --> 00:19:37,900
bottom bound and seeing if a
ray passed through them and

549
00:19:37,900 --> 00:19:41,560
then only checking that
section of the grid.

550
00:19:41,560 --> 00:19:44,300
And then even within like that
square bounding box--

551
00:19:44,300 --> 00:19:45,322
AUDIENCE: Like if there's a
square, and then the ray's

552
00:19:45,322 --> 00:19:46,036
going through like
that, there's a

553
00:19:46,036 --> 00:19:46,750
bunch of grid squares--

554
00:19:46,750 --> 00:19:48,671
AUDIENCE: Could you use
the mic, please?

555
00:19:48,671 --> 00:19:50,955
AUDIENCE: Like if a ray's
intersecting a square, there's

556
00:19:50,955 --> 00:19:53,700
still a bunch of grid squares
above and below it that you

557
00:19:53,700 --> 00:19:54,490
don't have to check.

558
00:19:54,490 --> 00:19:57,310
So we kind of imagined
it as a--

559
00:19:57,310 --> 00:19:59,370
projected it onto
a 2D surface.

560
00:19:59,370 --> 00:20:05,530
So we didn't check additional
superfluous grid squares.

561
00:20:05,530 --> 00:20:08,540
We also did the thing where
we got rid of the trig by

562
00:20:08,540 --> 00:20:10,175
transforming from polar
coordinates to XYZ

563
00:20:10,175 --> 00:20:11,320
coordinates.

564
00:20:11,320 --> 00:20:13,290
Although, before that, we used
the fast sign and cosign

565
00:20:13,290 --> 00:20:16,300
approximation, which
was almost as fast.

566
00:20:16,300 --> 00:20:20,096
Using a quadratic curve,
which was kind of cool.

567
00:20:20,096 --> 00:20:22,477
AUDIENCE: I think those
are the main--

568
00:20:22,477 --> 00:20:26,910
AUDIENCE: We did an SSE matrix
multiply, which barely speeded

569
00:20:26,910 --> 00:20:28,835
it up, but whatever.

570
00:20:28,835 --> 00:20:29,230
[LAUGHTER]

571
00:20:29,230 --> 00:20:31,790
AUDIENCE: We moved
stuff into...

572
00:20:31,790 --> 00:20:33,620
SAMAN AMARASINGHE: So you are
the team that had issue going

573
00:20:33,620 --> 00:20:35,125
to Intel to AMD.

574
00:20:35,125 --> 00:20:35,965
[INTERPOSING VOICES]

575
00:20:35,965 --> 00:20:39,250
AUDIENCE: Yeah, we were
complaining about--

576
00:20:39,250 --> 00:20:40,500
[INTERPOSING VOICES]

577
00:20:42,600 --> 00:20:43,580
SAMAN AMARASINGHE: If you use
Intel, certain instruction is

578
00:20:43,580 --> 00:20:44,830
not the same in AMD.

579
00:20:48,950 --> 00:20:50,500
OK, good.

580
00:20:50,500 --> 00:20:53,140
So we are waiting for
the time to come?

581
00:20:53,140 --> 00:20:55,270
AUDIENCE: So we ran it once,
but we forgot to time it.

582
00:20:55,270 --> 00:20:56,710
[LAUGHTER]

583
00:20:56,710 --> 00:20:59,590
AUDIENCE: It was fast enough
that we could run it again--

584
00:20:59,590 --> 00:21:01,510
[INTERPOSING VOICES]

585
00:21:01,510 --> 00:21:04,370
SAMAN AMARASINGHE: So by the
way, I want to give a chance

586
00:21:04,370 --> 00:21:06,710
for the top team that
get knocked out,

587
00:21:06,710 --> 00:21:07,930
explain what they did.

588
00:21:07,930 --> 00:21:09,770
Even though they are not in
competition because they ran

589
00:21:09,770 --> 00:21:11,520
much faster in the same run.

590
00:21:11,520 --> 00:21:12,210
Who's the top team?

591
00:21:12,210 --> 00:21:14,202
AUDIENCE: [UNINTELLIGIBLE]

592
00:21:14,202 --> 00:21:16,194
SAMAN AMARASINGHE: Who's
the top team up there?

593
00:21:16,194 --> 00:21:17,444
[INTERPOSING VOICES]

594
00:21:20,766 --> 00:21:22,143
CHARLES LEISERSON: They're
getting sleep.

595
00:21:22,143 --> 00:21:23,520
[LAUGHTER]

596
00:21:23,520 --> 00:21:26,375
SAMAN AMARASINGHE: OK, 59.

597
00:21:26,375 --> 00:21:27,925
Uh-oh, now we--

598
00:21:30,880 --> 00:21:32,970
Oops.

599
00:21:32,970 --> 00:21:34,752
OK.

600
00:21:34,752 --> 00:21:35,595
AUDIENCE: Ooo...

601
00:21:35,595 --> 00:21:37,500
SAMAN AMARASINGHE: Inversion
off that.

602
00:21:37,500 --> 00:21:39,045
OK, next team.

603
00:21:49,488 --> 00:21:52,165
AUDIENCE: So our team also
did a same trick.

604
00:21:52,165 --> 00:21:54,550
That was done in the
previous two teams.

605
00:21:54,550 --> 00:21:57,800
We got rid of the
trigonometry.

606
00:21:57,800 --> 00:22:01,600
And also used a bounding box.

607
00:22:01,600 --> 00:22:04,310
Basically described in
the previous team.

608
00:22:04,310 --> 00:22:07,160
And we also tried to use SSE.

609
00:22:07,160 --> 00:22:13,430
But as soon as the news
about an AMD machine,

610
00:22:13,430 --> 00:22:14,927
we got rid of that.

611
00:22:14,927 --> 00:22:15,324
[LAUGHTER]

612
00:22:15,324 --> 00:22:18,300
SAMAN AMARASINGHE: Oh, my
look at that speed.

613
00:22:18,300 --> 00:22:19,089
AUDIENCE: Whoa.

614
00:22:19,089 --> 00:22:20,466
SAMAN AMARASINGHE: Whoa.

615
00:22:20,466 --> 00:22:20,925
Woo!

616
00:22:20,925 --> 00:22:24,140
[LAUGHTER]

617
00:22:24,140 --> 00:22:24,445
OK.

618
00:22:24,445 --> 00:22:25,535
AUDIENCE: If you want to take--
could we take more time

619
00:22:25,535 --> 00:22:25,790
to talk about it--

620
00:22:25,790 --> 00:22:26,920
SAMAN AMARASINGHE: Yeah, sure.

621
00:22:26,920 --> 00:22:27,130
Yes.

622
00:22:27,130 --> 00:22:28,945
AUDIENCE: Because I'm sure you
guys have some things that are

623
00:22:28,945 --> 00:22:30,100
different than everybody
else's.

624
00:22:30,100 --> 00:22:30,595
In fact, I know it.

625
00:22:30,595 --> 00:22:31,585
AUDIENCE: It's on.

626
00:22:31,585 --> 00:22:32,380
It's just on.

627
00:22:32,380 --> 00:22:32,920
SAMAN AMARASINGHE: It's on.

628
00:22:32,920 --> 00:22:33,180
It's on.

629
00:22:33,180 --> 00:22:34,460
It's already recording.

630
00:22:34,460 --> 00:22:37,630
So for the cache I use like
a three-dimensional grid.

631
00:22:37,630 --> 00:22:40,420
Because I realized that as the
old times, the torrents is

632
00:22:40,420 --> 00:22:42,910
like usually one, unless
there's some outlier.

633
00:22:42,910 --> 00:22:44,820
But I don't see one
in my case.

634
00:22:44,820 --> 00:22:45,950
So I use a three-dimensional
grid.

635
00:22:45,950 --> 00:22:49,240
And use a lot of locks to
prevent contention.

636
00:22:49,240 --> 00:22:51,780
So I do the same thing
for photon map.

637
00:22:51,780 --> 00:22:54,470
I basically implement my
own photon map and

638
00:22:54,470 --> 00:22:55,650
just get rid of theirs.

639
00:22:55,650 --> 00:22:57,940
And also, the same thing,
like using a grid.

640
00:22:57,940 --> 00:23:00,300
And I spent a lot of
time debugging.

641
00:23:00,300 --> 00:23:02,515
[LAUGHTER]

642
00:23:02,515 --> 00:23:06,490
AUDIENCE: And we also do some
minor optimization.

643
00:23:06,490 --> 00:23:09,870
We implement the rendering
crest as a reducer.

644
00:23:09,870 --> 00:23:12,112
So that we--

645
00:23:12,112 --> 00:23:17,820
a hyper rod object So that every
time Cilk steals some

646
00:23:17,820 --> 00:23:20,340
other's job, you will create
a local view of

647
00:23:20,340 --> 00:23:21,650
the rendered object.

648
00:23:21,650 --> 00:23:28,200
And we found there is a artifact
that-- because we are

649
00:23:28,200 --> 00:23:30,750
seeding the rendered object.

650
00:23:30,750 --> 00:23:33,090
The default that seed
is two, right?

651
00:23:33,090 --> 00:23:38,810
And it will create some artifact
in our image, so we

652
00:23:38,810 --> 00:23:42,780
use a random number to seed
that rendered object.

653
00:23:42,780 --> 00:23:43,750
That's what we do.

654
00:23:43,750 --> 00:23:48,050
And also, we do something
like we change the--

655
00:23:48,050 --> 00:23:49,363
no, we don't have
locate photon.

656
00:23:53,610 --> 00:23:54,620
Yeah OK.

657
00:23:54,620 --> 00:23:59,310
So we have done a lot of thing
to improve the original photon

658
00:23:59,310 --> 00:24:03,415
med but after that we decide to
change our implementation

659
00:24:03,415 --> 00:24:06,145
of photon med so we
don't use that.

660
00:24:09,200 --> 00:24:10,450
OK.

661
00:24:13,870 --> 00:24:14,670
PROFESSOR 1: OK, good.

662
00:24:14,670 --> 00:24:15,740
Thank you.

663
00:24:15,740 --> 00:24:18,215
OK.

664
00:24:18,215 --> 00:24:20,170
OK, the suspense is on.

665
00:24:20,170 --> 00:24:21,460
So the final team.

666
00:24:27,620 --> 00:24:28,870
OK.

667
00:24:31,190 --> 00:24:34,050
STUDENT 8: So one of the first
things that happened is that

668
00:24:34,050 --> 00:24:36,700
we looked at the code and, like,
it was all a KD tree.

669
00:24:36,700 --> 00:24:37,980
And we're like, what's that?

670
00:24:37,980 --> 00:24:40,330
So we got rid of that
immediately and we turned it

671
00:24:40,330 --> 00:24:41,280
into something--

672
00:24:41,280 --> 00:24:42,950
I think the other groups might
have talked about it, but we

673
00:24:42,950 --> 00:24:44,750
gave it a nifty marketing
name, we

674
00:24:44,750 --> 00:24:46,470
called it a photon cube.

675
00:24:46,470 --> 00:24:50,310
And so we created
lots of these.

676
00:24:50,310 --> 00:24:53,070
We partitioned the large cube
environment into many small

677
00:24:53,070 --> 00:24:56,260
photon cubes so we could just
jump in and look in the

678
00:24:56,260 --> 00:24:58,860
relative air vicinity
of photons that

679
00:24:58,860 --> 00:25:00,630
you're looking for.

680
00:25:00,630 --> 00:25:00,850
STUDENT 9: Yeah.

681
00:25:00,850 --> 00:25:06,490
So this is like, when you do
locate photon, you know the

682
00:25:06,490 --> 00:25:10,070
reference photon and you
can create a sphere.

683
00:25:10,070 --> 00:25:15,360
So we can do constant time to
find the boundary of the

684
00:25:15,360 --> 00:25:16,810
photon cube.

685
00:25:16,810 --> 00:25:22,400
Like go from x1 to x something,
y to y and z to z.

686
00:25:22,400 --> 00:25:24,790
So then we don't have to
traverse the tree.

687
00:25:24,790 --> 00:25:30,280
And we basically parallelized
two part, so the one in render

688
00:25:30,280 --> 00:25:34,750
and another one in the trace
photon that when we try to put

689
00:25:34,750 --> 00:25:38,180
the photon in the map.

690
00:25:38,180 --> 00:25:43,030
And also when we create a photon
to the cube, there is a

691
00:25:43,030 --> 00:25:46,170
reused vector to start a
photon in each cube.

692
00:25:46,170 --> 00:25:50,770
But then it's not really good,
because when you put the

693
00:25:50,770 --> 00:25:53,510
photon in the cube, you have
to, like, new photon.

694
00:25:53,510 --> 00:25:56,410
Which means that the memory,
it's like spurts throughout

695
00:25:56,410 --> 00:25:57,860
your program.

696
00:25:57,860 --> 00:26:03,190
So after that we convert that
into a block of memory.

697
00:26:03,190 --> 00:26:05,805
And we use that when we
call locate photon.

698
00:26:08,390 --> 00:26:13,485
STUDENT 10: And for the last,
about surfaces that intersect.

699
00:26:16,450 --> 00:26:18,930
There were four kinds
of optimizations.

700
00:26:18,930 --> 00:26:27,140
First one is checking if the
origin of ray is outside of

701
00:26:27,140 --> 00:26:32,980
the maximum bound and lower
bound of surface, then and

702
00:26:32,980 --> 00:26:40,930
check whether by direction is
outside or if it's toward the

703
00:26:40,930 --> 00:26:45,480
outside then we don't have to
check, just return false.

704
00:26:45,480 --> 00:26:50,280
And second thing is like
the square cube thing.

705
00:26:50,280 --> 00:26:58,220
When rays intersect we can make
a boundary by directly

706
00:26:58,220 --> 00:27:06,330
mapping to the XYZ at the
surface at the maximum

707
00:27:06,330 --> 00:27:10,530
boundary and lower boundary
of the surface.

708
00:27:10,530 --> 00:27:18,850
And also when you iterate
through the x index, we can

709
00:27:18,850 --> 00:27:27,930
cut off again because
a bitmap.

710
00:27:27,930 --> 00:27:35,680
We only have to check with the
indexes which correspond to

711
00:27:35,680 --> 00:27:41,880
ray's z index at that cube,
at that x location.

712
00:27:41,880 --> 00:27:50,510
And the last thing was for each
square we pre-computed

713
00:27:50,510 --> 00:28:02,200
the minimum value and maximum
value of y and we compared the

714
00:28:02,200 --> 00:28:08,890
rays by value within
that range.

715
00:28:08,890 --> 00:28:13,755
And we just cut off if
it doesn't intersect.

716
00:28:13,755 --> 00:28:15,413
It doesn't overlap.

717
00:28:15,413 --> 00:28:16,400
PROFESSOR 1: OK.

718
00:28:16,400 --> 00:28:17,250
Good.

719
00:28:17,250 --> 00:28:19,120
OK, now this is the
suspense part.

720
00:28:19,120 --> 00:28:19,550
OK.

721
00:28:19,550 --> 00:28:21,240
We are going to run
this and see.

722
00:28:21,240 --> 00:28:24,090
The time to beat is 18.7.

723
00:28:24,090 --> 00:28:24,900
OK, right?

724
00:28:24,900 --> 00:28:25,630
Again.

725
00:28:25,630 --> 00:28:26,460
OK.

726
00:28:26,460 --> 00:28:27,140
There we go.

727
00:28:27,140 --> 00:28:28,560
Let's see.

728
00:28:28,560 --> 00:28:32,040
How many people think it's
going to beat 18.7.

729
00:28:32,040 --> 00:28:32,660
18.6.

730
00:28:32,660 --> 00:28:33,510
OK.

731
00:28:33,510 --> 00:28:35,070
AUDIENCE: 8, 9, 10--

732
00:28:35,070 --> 00:28:35,530
PROFESSOR 1: It's
the excitement.

733
00:28:35,530 --> 00:28:36,370
Keep counting.

734
00:28:36,370 --> 00:28:36,900
OK.

735
00:28:36,900 --> 00:28:38,550
There we go.

736
00:28:38,550 --> 00:28:44,890
AUDIENCE: 13, 14, 15,
16, 17, 18, 19--

737
00:28:44,890 --> 00:28:46,020
PROFESSOR 1: Oh, oh, oh.

738
00:28:46,020 --> 00:28:52,960
AUDIENCE: 21, 22,
23, 24, 25, 26.

739
00:28:52,960 --> 00:28:54,260
PROFESSOR 1: Oh, OK.

740
00:28:54,260 --> 00:28:57,170
It came out as 26.41.

741
00:28:57,170 --> 00:28:57,790
OK.

742
00:28:57,790 --> 00:29:01,070
[APPLAUSE]

743
00:29:01,070 --> 00:29:06,150
PROFESSOR 1: So you guys
are the winners.

744
00:29:06,150 --> 00:29:07,400
Come up here.

745
00:29:11,620 --> 00:29:13,110
AUDIENCE: Danielle.

746
00:29:13,110 --> 00:29:14,222
PROFESSOR 1: Where's Danielle?

747
00:29:14,222 --> 00:29:17,614
AUDIENCE: Uh, Danielle
disappeared.

748
00:29:17,614 --> 00:29:19,854
AUDIENCE: Yes, she did.

749
00:29:19,854 --> 00:29:20,550
PROFESSOR 1: OK.

750
00:29:20,550 --> 00:29:22,250
Let's see whether she
can find her, so OK.

751
00:29:22,250 --> 00:29:23,650
Just hold on a second until--

752
00:29:26,740 --> 00:29:27,990
AUDIENCE: Yep, there we go.

753
00:29:36,540 --> 00:29:38,395
PROFESSOR 1: You can put
it down now, though.

754
00:29:41,810 --> 00:29:43,560
PRESENTER: So this one's
the reference.

755
00:29:43,560 --> 00:29:45,820
PROFESSOR 1: Put a text meter
in, that's better.

756
00:29:45,820 --> 00:29:46,680
AUDIENCE: Which one's
reference?

757
00:29:46,680 --> 00:29:47,930
PRESENTER: This one's
the reference.

758
00:29:52,780 --> 00:29:53,880
PROFESSOR 1: I think
they don't align.

759
00:29:53,880 --> 00:29:57,200
Can you align them?

760
00:29:57,200 --> 00:30:00,070
PRESENTER: There we go.

761
00:30:00,070 --> 00:30:03,180
PROFESSOR 1: OK, so we
have a winning team.

762
00:30:03,180 --> 00:30:05,070
PROFESSOR 2: We have
a winning team.

763
00:30:05,070 --> 00:30:10,870
And so it's my pleasure to
introduce Danielle Smith, who

764
00:30:10,870 --> 00:30:19,220
is the University relations for
Akamai, and she's here to

765
00:30:19,220 --> 00:30:20,980
give you guys your prize.

766
00:30:20,980 --> 00:30:23,671
So we have three winners.

767
00:30:23,671 --> 00:30:25,770
DANIELLE: I just wanted
to make an entrance

768
00:30:25,770 --> 00:30:28,640
is why I came in.

769
00:30:28,640 --> 00:30:29,160
Thank you.

770
00:30:29,160 --> 00:30:32,900
First of all, welcome to all of
you guys for coming in, and

771
00:30:32,900 --> 00:30:34,610
we wanted to give you
these prizes.

772
00:30:34,610 --> 00:30:38,810
So we have some cool iPod Nanos,
which you may or may

773
00:30:38,810 --> 00:30:42,720
not have already, but it can't
hurt to have another.

774
00:30:46,780 --> 00:30:49,690
They kind of burn out
or you lose one or

775
00:30:49,690 --> 00:30:50,946
someone steps on it.

776
00:30:50,946 --> 00:30:55,610
[APPLAUSE]

777
00:30:55,610 --> 00:30:56,790
DANIELLE: Congratulations.

778
00:30:56,790 --> 00:30:58,230
You guys have worked
really hard.

779
00:31:04,000 --> 00:31:05,120
PROFESSOR 1: Congratulations.

780
00:31:05,120 --> 00:31:06,140
Congratulations.

781
00:31:06,140 --> 00:31:08,740
Congratulations.

782
00:31:08,740 --> 00:31:10,410
OK.

783
00:31:10,410 --> 00:31:11,150
So what's next.

784
00:31:11,150 --> 00:31:11,970
That's it.

785
00:31:11,970 --> 00:31:12,790
We are done?

786
00:31:12,790 --> 00:31:13,120
DANIELLE: All righty.

787
00:31:13,120 --> 00:31:15,600
So we have plenty of food,
it's going to be

788
00:31:15,600 --> 00:31:17,270
right on the atrium.

789
00:31:17,270 --> 00:31:20,370
I know that's the most
exciting part.

790
00:31:20,370 --> 00:31:24,690
But it should be arriving in the
next 10 to 15 minutes, so

791
00:31:24,690 --> 00:31:26,260
we have a few minutes.

792
00:31:26,260 --> 00:31:28,230
I thought maybe this would
go a little bit longer.

793
00:31:28,230 --> 00:31:29,180
PROFESSOR 1: We do
have a couple of

794
00:31:29,180 --> 00:31:30,900
things we can chat about.

795
00:31:30,900 --> 00:31:31,400
DANIELLE: Yeah.

796
00:31:31,400 --> 00:31:34,510
So just a couple of housekeeping
things.

797
00:31:34,510 --> 00:31:37,810
If you need to use the ladies'
room or the men's room, it's

798
00:31:37,810 --> 00:31:40,040
actually right outside
that door.

799
00:31:40,040 --> 00:31:44,050
Out the double doors, you'll
see it right there.

800
00:31:44,050 --> 00:31:47,150
I have invited some other Akamai
employees to come and

801
00:31:47,150 --> 00:31:49,840
mingle and say hello to you
guys, so I encourage you, if

802
00:31:49,840 --> 00:31:53,120
you see any faces that aren't
familiar, to kind of go up and

803
00:31:53,120 --> 00:31:57,310
talk about your projects
and introduce yourself.

804
00:31:57,310 --> 00:31:59,480
And thanks for coming.

805
00:31:59,480 --> 00:32:01,230
I know we wanted to switch it up
a little bit because it was

806
00:32:01,230 --> 00:32:02,040
the last class.

807
00:32:02,040 --> 00:32:04,140
We wanted to get you out of the
classroom and just kind of

808
00:32:04,140 --> 00:32:04,670
shake it up.

809
00:32:04,670 --> 00:32:07,490
So I'm glad that you
came to join.

810
00:32:07,490 --> 00:32:08,050
PROFESSOR 1: And the
knock tours.

811
00:32:08,050 --> 00:32:09,230
DANIELLE: And the knock tours.

812
00:32:09,230 --> 00:32:09,660
Thank you.

813
00:32:09,660 --> 00:32:13,160
So on the back of your name
tags, there's a number, one,

814
00:32:13,160 --> 00:32:14,410
two or three.

815
00:32:18,100 --> 00:32:18,950
One, two or three.

816
00:32:18,950 --> 00:32:21,120
So that correlates with
the knock tour that

817
00:32:21,120 --> 00:32:22,310
you're going to be on.

818
00:32:22,310 --> 00:32:24,240
We have to keep it to relatively
small groups.

819
00:32:24,240 --> 00:32:28,620
So the first tour will be at 10
past 4, the second one will

820
00:32:28,620 --> 00:32:31,840
be at 4:40, and the last
one will be at 5:10.

821
00:32:31,840 --> 00:32:33,550
PROFESSOR 1: I guess if we don't
have too many people we

822
00:32:33,550 --> 00:32:35,000
might be able to
compress some--

823
00:32:35,000 --> 00:32:35,440
DANIELLE: Yes.

824
00:32:35,440 --> 00:32:36,310
PROFESSOR 1: So we'll
see whether we

825
00:32:36,310 --> 00:32:38,100
can do that, compress--

826
00:32:38,100 --> 00:32:38,360
DANIELLE: Yes.

827
00:32:38,360 --> 00:32:41,550
So I'll just call the first tour
and you guys can go down.

828
00:32:41,550 --> 00:32:45,120
And obviously it's optional.

829
00:32:45,120 --> 00:32:46,900
Don't feel-- if you're not
interested, you don't have to

830
00:32:46,900 --> 00:32:49,310
go on the knock tour.

831
00:32:49,310 --> 00:32:50,360
I think that's it.

832
00:32:50,360 --> 00:32:50,550
PROFESSOR 1: Good.

833
00:32:50,550 --> 00:32:51,615
Thank you.

834
00:32:51,615 --> 00:32:56,265
[APPLAUSE]

835
00:32:56,265 --> 00:32:58,070
PROFESSOR 1: We can talk
about a little bit.

836
00:32:58,070 --> 00:32:59,720
So one thing.

837
00:32:59,720 --> 00:33:07,060
We are trying to get HKN
feedback up to hopefully 90%,

838
00:33:07,060 --> 00:33:09,990
and we are right now
at what, 11?

839
00:33:09,990 --> 00:33:10,980
PROFESSOR 2: No, we're
a little bit higher.

840
00:33:10,980 --> 00:33:13,910
We're now at about 16 or so.

841
00:33:13,910 --> 00:33:14,550
PROFESSOR 1: 16.

842
00:33:14,550 --> 00:33:17,150
So that's a little more way
to go from 16 to 90.

843
00:33:17,150 --> 00:33:20,790
I hope you guys go back and
then fill out this survey.

844
00:33:20,790 --> 00:33:21,310
Very important.

845
00:33:21,310 --> 00:33:23,420
PROFESSOR 2: Please take a
few minutes to do that.

846
00:33:23,420 --> 00:33:26,370
It really helps us in,
even if you want to

847
00:33:26,370 --> 00:33:28,470
say bad things, OK?

848
00:33:28,470 --> 00:33:31,100
It really helps overall to
give feedback and the

849
00:33:31,100 --> 00:33:33,300
department likes to
see these things.

850
00:33:33,300 --> 00:33:34,530
Oh yeah, thanks.

851
00:33:34,530 --> 00:33:37,370
Likes to see how it
is that people are

852
00:33:37,370 --> 00:33:38,530
responding to the courses.

853
00:33:38,530 --> 00:33:41,320
So if you hated the course, if
you liked the course, in

854
00:33:41,320 --> 00:33:43,640
particular, we have some
questions on there that are

855
00:33:43,640 --> 00:33:46,650
specific to how we teach
the course next year.

856
00:33:46,650 --> 00:33:49,890
One of the ideas we have for
next year is to introduce our

857
00:33:49,890 --> 00:33:53,190
recitation sections,
which would focus

858
00:33:53,190 --> 00:33:55,300
on the use of tools.

859
00:33:55,300 --> 00:33:55,650
OK?

860
00:33:55,650 --> 00:33:58,200
So basically the conceptual
stuff, much as we have the

861
00:33:58,200 --> 00:34:01,710
lectures now, but then also
having something where it

862
00:34:01,710 --> 00:34:06,400
really takes you through each
tool, and teaches how the use

863
00:34:06,400 --> 00:34:09,040
of those tools which would
be taught once a week.

864
00:34:09,040 --> 00:34:12,280
So if you think that's a good
idea, that's going to really

865
00:34:12,280 --> 00:34:14,449
help to have that kind
of feedback.

866
00:34:14,449 --> 00:34:17,120
And it will certainly help the
students who take it next year

867
00:34:17,120 --> 00:34:19,150
if we can persuade the
department that that's worth

868
00:34:19,150 --> 00:34:22,389
staffing the sections
next year.

869
00:34:22,389 --> 00:34:27,340
Also if you have any positive
things to say about teaching

870
00:34:27,340 --> 00:34:30,030
assistants, that's a great
place to put it.

871
00:34:30,030 --> 00:34:30,320
OK?

872
00:34:30,320 --> 00:34:33,949
If they helped you get through
this term, this is a great

873
00:34:33,949 --> 00:34:36,760
place to say thank you and it
only takes a few minutes.

874
00:34:36,760 --> 00:34:37,650
OK?

875
00:34:37,650 --> 00:34:41,380
So I'm sure that some of the
TAs put in more than a few

876
00:34:41,380 --> 00:34:43,969
minutes, OK, on your thing.

877
00:34:43,969 --> 00:34:47,040
So to take a few minutes
and say thank you on

878
00:34:47,040 --> 00:34:48,570
that form would be--

879
00:34:48,570 --> 00:34:51,820
I'm sure they would very,
very much appreciate.

880
00:34:51,820 --> 00:34:53,580
PROFESSOR 1: One thing, as
Charles mentioned, what we are

881
00:34:53,580 --> 00:34:58,110
trying to do in recitation is do
a little bit of programmer

882
00:34:58,110 --> 00:34:59,360
life skills.

883
00:34:59,360 --> 00:35:02,810
So I know all of you guys are
probably very smart, can think

884
00:35:02,810 --> 00:35:05,920
about the complex algorithms
and stuff like that.

885
00:35:05,920 --> 00:35:10,040
But there are a lot of tools
that, it's not the smartness

886
00:35:10,040 --> 00:35:12,090
but you can spend hours
trying to figure out.

887
00:35:12,090 --> 00:35:14,280
But if you know it,
it will be useful.

888
00:35:14,280 --> 00:35:17,930
So what we are trying to do is
have these recitations to say

889
00:35:17,930 --> 00:35:20,010
OK, how to use a debugger.

890
00:35:20,010 --> 00:35:23,800
And even how to change
a makefile.

891
00:35:23,800 --> 00:35:25,130
Things like that.

892
00:35:25,130 --> 00:35:28,760
So one thing, until the food
arrives, I'd like to say is,

893
00:35:28,760 --> 00:35:29,560
get some feedback.

894
00:35:29,560 --> 00:35:33,160
Think of what is a tool you'd
like to know or like to learn,

895
00:35:33,160 --> 00:35:36,270
or something you think is very
useful for people coming next

896
00:35:36,270 --> 00:35:37,110
year to learn.

897
00:35:37,110 --> 00:35:42,310
That if you have spent a lot of
time worrying and you think

898
00:35:42,310 --> 00:35:44,910
you spent way too much time
trying to understand, it would

899
00:35:44,910 --> 00:35:46,220
be easier for us to
think about that.

900
00:35:46,220 --> 00:35:47,950
I mean, I would like to
get your feedback.

901
00:35:47,950 --> 00:35:50,150
And also some of you probably
will open up with some general

902
00:35:50,150 --> 00:35:52,922
feedback on class.

903
00:35:52,922 --> 00:35:56,820
AUDIENCE: I don't know if
you want to repeat that.

904
00:35:56,820 --> 00:35:59,410
A more in depth thing
on how to use git

905
00:35:59,410 --> 00:36:00,850
would be really useful.

906
00:36:00,850 --> 00:36:04,040
Because I know it has things to
make your life easier, but

907
00:36:04,040 --> 00:36:06,802
I don't know how to
use any of them.

908
00:36:06,802 --> 00:36:08,830
AUDIENCE: Yeah. women are always
breaking the branches.

909
00:36:08,830 --> 00:36:11,160
And it's like, yeah, we maybe
know how to do this and it is

910
00:36:11,160 --> 00:36:11,790
really easy.

911
00:36:11,790 --> 00:36:13,610
And then we're like, use
this git command.

912
00:36:13,610 --> 00:36:15,434
And it's like, now your
branch is broken.

913
00:36:15,434 --> 00:36:17,819
[LAUGHTER]

914
00:36:17,819 --> 00:36:18,300
PROFESSOR 2: Good.

915
00:36:18,300 --> 00:36:19,340
So git

916
00:36:19,340 --> 00:36:19,710
PROFESSOR 1: Yeah.

917
00:36:19,710 --> 00:36:22,480
I mean that's because those are
interesting tools because,

918
00:36:22,480 --> 00:36:24,370
if you know everything
right, you can do

919
00:36:24,370 --> 00:36:25,560
a lot of fun things.

920
00:36:25,560 --> 00:36:29,690
But if you do it wrong, you end
up in huge mess that all

921
00:36:29,690 --> 00:36:31,590
you do is scored all over the
place and you don't know.

922
00:36:31,590 --> 00:36:32,330
So yes.

923
00:36:32,330 --> 00:36:34,655
I mean I can relate to that.

924
00:36:34,655 --> 00:36:36,927
AUDIENCE: If you use git
correctly, if you

925
00:36:36,927 --> 00:36:39,000
know how to use git--

926
00:36:39,000 --> 00:36:43,190
If you learn how to use git you
will find yourself using

927
00:36:43,190 --> 00:36:46,250
it even when you're working
on a one person project.

928
00:36:46,250 --> 00:36:47,400
git is awesome.

929
00:36:47,400 --> 00:36:49,181
I can highly recommend
learning it.

930
00:36:51,950 --> 00:36:53,920
AUDIENCE: What would be helpful
is that telling us

931
00:36:53,920 --> 00:36:57,700
that you are running on an AMD
box and AMD doesn't support

932
00:36:57,700 --> 00:36:59,905
SSE 4.1 instructions.

933
00:36:59,905 --> 00:37:03,090
[LAUGHTER]

934
00:37:03,090 --> 00:37:06,140
PROFESSOR 1: Yeah, I mean, this
is supposed to be X86

935
00:37:06,140 --> 00:37:07,440
compatible, all those things.

936
00:37:07,440 --> 00:37:11,860
But I mean, some people learn
the hard way that, OK, the

937
00:37:11,860 --> 00:37:13,760
word is not the same
even though--

938
00:37:13,760 --> 00:37:15,830
PROFESSOR 2: But that is part
of the lesson of the course,

939
00:37:15,830 --> 00:37:19,090
is how do you make high
performance portable?

940
00:37:19,090 --> 00:37:19,450
OK?

941
00:37:19,450 --> 00:37:25,080
And so what you choose or
don't choose to use is--

942
00:37:25,080 --> 00:37:27,140
But I agree.

943
00:37:27,140 --> 00:37:29,990
PROFESSOR 1: That's why we ran,
we want to run it the

944
00:37:29,990 --> 00:37:33,510
night before because we knew
things like this would happen.

945
00:37:33,510 --> 00:37:36,300
So we can at least give some
time to say OK, look.

946
00:37:36,300 --> 00:37:36,975
Go change something.

947
00:37:36,975 --> 00:37:37,930
Were you able to fix that?

948
00:37:37,930 --> 00:37:38,500
OK.

949
00:37:38,500 --> 00:37:39,340
Good.

950
00:37:39,340 --> 00:37:40,766
So yes.

951
00:37:40,766 --> 00:37:41,520
PROFESSOR 2: Good.

952
00:37:41,520 --> 00:37:44,320
PROFESSOR 1: And sometimes the
problem is, a lot of these

953
00:37:44,320 --> 00:37:47,595
small changes you don't know
until you run into things like

954
00:37:47,595 --> 00:37:52,100
this and oh, somewhere in page
256 in the manual that there's

955
00:37:52,100 --> 00:37:53,550
a footnote say, oh,
by the way.

956
00:37:53,550 --> 00:37:55,880
And that's what happens
in there.

957
00:37:55,880 --> 00:37:57,650
I mean if you had listed
everything that would be

958
00:37:57,650 --> 00:38:00,150
probably and entire encyclopedia
of things that

959
00:38:00,150 --> 00:38:01,300
you will see.

960
00:38:01,300 --> 00:38:04,780
PROFESSOR 2: So one other thing
that people ask about

961
00:38:04,780 --> 00:38:08,050
is, what comes after
this course?

962
00:38:08,050 --> 00:38:09,010
OK?

963
00:38:09,010 --> 00:38:13,510
So next term Saman is teaching
the compiler course.

964
00:38:13,510 --> 00:38:15,005
PROFESSOR 1: It's Martin.

965
00:38:15,005 --> 00:38:16,200
PROFESSOR 2: Are you
teaching with him?

966
00:38:16,200 --> 00:38:16,495
PROFESSOR 1: No.

967
00:38:16,495 --> 00:38:17,475
PROFESSOR 2: Oh, you're
not teaching.

968
00:38:17,475 --> 00:38:17,770
OK.

969
00:38:17,770 --> 00:38:19,110
I thought you were.

970
00:38:19,110 --> 00:38:19,810
OK.

971
00:38:19,810 --> 00:38:21,690
So in any case, compiler
course is a

972
00:38:21,690 --> 00:38:22,550
good follow-on course.

973
00:38:22,550 --> 00:38:26,210
For those who haven't done
6.046, that's a good course

974
00:38:26,210 --> 00:38:29,180
and I'll be teaching that
course next term.

975
00:38:29,180 --> 00:38:32,040
In addition, how many people
here have done a UROP?

976
00:38:34,890 --> 00:38:35,710
Few of you.

977
00:38:35,710 --> 00:38:37,790
So let me encourage you
to look for UROPs.

978
00:38:37,790 --> 00:38:40,730
You now have some skills that
will make you really valuable.

979
00:38:40,730 --> 00:38:43,370
And in particular my group
is looking for--

980
00:38:43,370 --> 00:38:44,710
[LAUGHTER]

981
00:38:44,710 --> 00:38:48,070
PROFESSOR 2: Probably two more
UROPs than I have already.

982
00:38:48,070 --> 00:38:51,390
Already Boone is one of my
UROPs, started just a

983
00:38:51,390 --> 00:38:53,070
couple weeks ago.

984
00:38:53,070 --> 00:38:57,360
And so we do have some other
projects in both architecture

985
00:38:57,360 --> 00:39:02,070
and in software-related things.

986
00:39:02,070 --> 00:39:02,920
And so.

987
00:39:02,920 --> 00:39:05,540
if you're interested, you go
to my group's web page.

988
00:39:05,540 --> 00:39:09,560
It tells you five things you
have to submit if you want to

989
00:39:09,560 --> 00:39:10,660
work in my group.

990
00:39:10,660 --> 00:39:10,950
OK?

991
00:39:10,950 --> 00:39:12,880
That includes things like
samples of your code.

992
00:39:12,880 --> 00:39:15,580
Well, you ought to
have that by now.

993
00:39:15,580 --> 00:39:16,250
And so forth.

994
00:39:16,250 --> 00:39:18,090
So let me encourage
you to do that.

995
00:39:18,090 --> 00:39:23,330
And Saman also has active
participation of

996
00:39:23,330 --> 00:39:24,670
UROPs in his group.

997
00:39:24,670 --> 00:39:24,950
PROFESSOR 1: Yeah.

998
00:39:24,950 --> 00:39:28,860
So my employees already
started doing a

999
00:39:28,860 --> 00:39:29,790
project in my group.

1000
00:39:29,790 --> 00:39:32,890
And I'm also looking for a
couple more people to do some

1001
00:39:32,890 --> 00:39:36,950
compiler, runtime system, and
a bunch of high performance

1002
00:39:36,950 --> 00:39:38,980
type projects I have.

1003
00:39:38,980 --> 00:39:43,490
And also for future course,
one course I want to also

1004
00:39:43,490 --> 00:39:45,670
mention is this IAP.

1005
00:39:45,670 --> 00:39:49,320
There's a course that is done
with collaboration with VMware

1006
00:39:49,320 --> 00:39:53,550
to look at how to build scalable
large systems using

1007
00:39:53,550 --> 00:39:56,480
this the Spring and
Java platforms.

1008
00:39:56,480 --> 00:39:58,650
So if they're interested in
trying to see how things were

1009
00:39:58,650 --> 00:40:02,040
running on one machine and run
something that can scale up to

1010
00:40:02,040 --> 00:40:04,050
hundreds of machines.

1011
00:40:04,050 --> 00:40:07,490
If you have ideas of creating
an I 8 with web 2.0,

1012
00:40:07,490 --> 00:40:09,330
interesting thing.

1013
00:40:09,330 --> 00:40:10,930
Here is an opportunity
to do that.

1014
00:40:10,930 --> 00:40:15,590
So if you're looking
for interesting IAP

1015
00:40:15,590 --> 00:40:16,630
courses, this is great.

1016
00:40:16,630 --> 00:40:18,830
PROFESSOR 2: I really encourage
every single person

1017
00:40:18,830 --> 00:40:23,580
who completed this class
to do a UROP with

1018
00:40:23,580 --> 00:40:25,430
somebody before you leave.

1019
00:40:25,430 --> 00:40:28,550
It's really a way of integrating
your knowledge.

1020
00:40:28,550 --> 00:40:32,050
And it is one of the most
valuable experiences that you

1021
00:40:32,050 --> 00:40:36,390
can have at MIT is to do a UROP
for a professor and get

1022
00:40:36,390 --> 00:40:38,680
involved in the research
of a group.

1023
00:40:38,680 --> 00:40:43,210
Because in some ways it's like
this course, in that it

1024
00:40:43,210 --> 00:40:46,470
integrates knowledge from a
lot of different sources.

1025
00:40:46,470 --> 00:40:48,140
And it isn't just one thing.

1026
00:40:48,140 --> 00:40:49,890
But even more than
this course.

1027
00:40:49,890 --> 00:40:51,070
This course we still
are dealing

1028
00:40:51,070 --> 00:40:52,320
with artificial things.

1029
00:40:52,320 --> 00:40:54,040
And in the researcher
arena, you're

1030
00:40:54,040 --> 00:40:56,870
dealing with real problems.

1031
00:40:56,870 --> 00:41:01,900
So let me just once again
encourage you to spend some

1032
00:41:01,900 --> 00:41:07,160
time getting a UROP and
certainly looking

1033
00:41:07,160 --> 00:41:11,030
for AUPs and MEng.

1034
00:41:11,030 --> 00:41:13,400
How many people here are
looking at MEng as a

1035
00:41:13,400 --> 00:41:14,982
possibility?

1036
00:41:14,982 --> 00:41:16,670
So a fair number.

1037
00:41:16,670 --> 00:41:17,560
Yeah, that's good.

1038
00:41:17,560 --> 00:41:19,530
That's great.

1039
00:41:19,530 --> 00:41:20,760
PROFESSOR 1: I want
to add to that.

1040
00:41:20,760 --> 00:41:24,730
So these days there has been a
big philosophical discussion

1041
00:41:24,730 --> 00:41:27,080
saying why do you have
to have universities?

1042
00:41:27,080 --> 00:41:32,000
You go to OCW and sit home in
your bedroom and you can

1043
00:41:32,000 --> 00:41:33,930
follow all these courses,
and you don't have to

1044
00:41:33,930 --> 00:41:35,040
pay a mighty tuition.

1045
00:41:35,040 --> 00:41:38,460
And wouldn't that
be good enough?

1046
00:41:38,460 --> 00:41:40,390
One part that's not good
enough is this class.

1047
00:41:40,390 --> 00:41:42,230
So I mean think about it.

1048
00:41:42,230 --> 00:41:45,960
I mean you can't go get the OCW
catalog and have this kind

1049
00:41:45,960 --> 00:41:46,970
of a dynamic.

1050
00:41:46,970 --> 00:41:48,220
There's no seventh floor.

1051
00:41:51,390 --> 00:41:52,810
All in OCW.

1052
00:41:52,810 --> 00:41:55,430
And you have from there to
go and get motivated.

1053
00:41:55,430 --> 00:41:56,980
So that's one part of that.

1054
00:41:56,980 --> 00:42:03,080
PROFESSOR 2: There's no John
or Reed or Eric or Josh to

1055
00:42:03,080 --> 00:42:05,990
respond quickly about
what's going on.

1056
00:42:05,990 --> 00:42:06,640
PROFESSOR 1: That's one.

1057
00:42:06,640 --> 00:42:10,280
But I think one other thing that
a lot of people miss when

1058
00:42:10,280 --> 00:42:13,740
they look at all of the
curriculum and stuff like that

1059
00:42:13,740 --> 00:42:16,090
is the UROP ability.

1060
00:42:16,090 --> 00:42:19,890
So a lot of these people, that
is something, it's very unique

1061
00:42:19,890 --> 00:42:23,080
to MIT and a lot of people
benefit hugely with UROPs.

1062
00:42:23,080 --> 00:42:26,560
I have had UROPs who started
with me as seniors.

1063
00:42:26,560 --> 00:42:30,040
By the time they are graduating
with their MEng

1064
00:42:30,040 --> 00:42:32,480
they have two first out
papers in the best

1065
00:42:32,480 --> 00:42:33,700
conference out there.

1066
00:42:33,700 --> 00:42:36,030
And everybody thinks, when they
say they are looking for

1067
00:42:36,030 --> 00:42:38,000
a job, they are finishing their
Ph.D and looking for

1068
00:42:38,000 --> 00:42:38,560
academic jobs.

1069
00:42:38,560 --> 00:42:38,730
No.

1070
00:42:38,730 --> 00:42:40,700
They are just finishing
up their masters.

1071
00:42:40,700 --> 00:42:43,270
And I have a couple of students
that managed to

1072
00:42:43,270 --> 00:42:46,870
become really expert in an
area by being a UROP.

1073
00:42:46,870 --> 00:42:50,120
So they are finishing up at MIT
as a generalist, saying

1074
00:42:50,120 --> 00:42:50,990
give me a job.

1075
00:42:50,990 --> 00:42:53,770
But you're finishing up
and have some deep

1076
00:42:53,770 --> 00:42:55,750
knowledge in some area.

1077
00:42:55,750 --> 00:42:59,880
Sitting in and at the same level
as most Ph.Ds or most

1078
00:42:59,880 --> 00:43:02,800
really deep practicing engineers
in that area.

1079
00:43:02,800 --> 00:43:06,980
So that level, even if you don't
do that, just seeing

1080
00:43:06,980 --> 00:43:08,900
what happens in a research
group, being

1081
00:43:08,900 --> 00:43:11,540
participating on that.

1082
00:43:11,540 --> 00:43:13,980
Pushing, seeing something
to the maximum limit.

1083
00:43:13,980 --> 00:43:16,950
I mean in the class,
that's a limit.

1084
00:43:16,950 --> 00:43:19,670
I mean even though this class is
somewhat open, says make it

1085
00:43:19,670 --> 00:43:21,200
run as fast as possible.

1086
00:43:21,200 --> 00:43:23,830
But most of the time in the
class, it's like, we give you

1087
00:43:23,830 --> 00:43:26,650
something, if you made that,
OK, you're done with it.

1088
00:43:26,650 --> 00:43:27,530
In research, there's no limit.

1089
00:43:27,530 --> 00:43:30,175
You really, really push
your boundaries.

1090
00:43:30,175 --> 00:43:31,130
What's that?

1091
00:43:31,130 --> 00:43:32,600
PROFESSOR 2: So what do
we have here, John?

1092
00:43:32,600 --> 00:43:36,620
JOHN: I counted in my mailbox
how many messages came from

1093
00:43:36,620 --> 00:43:40,580
6.172 staff, and it's
about 3,500 over the

1094
00:43:40,580 --> 00:43:42,960
course of this term.

1095
00:43:42,960 --> 00:43:44,692
PROFESSOR 2: Wow.

1096
00:43:44,692 --> 00:43:49,460
[APPLAUSE]

1097
00:43:49,460 --> 00:43:50,555
PROFESSOR 1: How much time--

1098
00:43:50,555 --> 00:43:53,350
PROFESSOR 2: So, this is also,
as Saman mentions, a

1099
00:43:53,350 --> 00:43:55,100
programmer life skills course.

1100
00:43:55,100 --> 00:43:58,810
And one of the things that you
should get used to in life is

1101
00:43:58,810 --> 00:44:00,760
thanking people who help you.

1102
00:44:00,760 --> 00:44:01,110
OK?

1103
00:44:01,110 --> 00:44:06,770
I think that all of the MIT
POSSE members, all the people

1104
00:44:06,770 --> 00:44:11,530
who devoted, volunteered their
time to help you, could really

1105
00:44:11,530 --> 00:44:17,350
appreciate an email that says
thank you and what exactly you

1106
00:44:17,350 --> 00:44:18,960
feel you learned from them.

1107
00:44:18,960 --> 00:44:19,500
OK?

1108
00:44:19,500 --> 00:44:21,250
Because it's one thing
to just say

1109
00:44:21,250 --> 00:44:23,500
thanks, so long, et cetera.

1110
00:44:23,500 --> 00:44:28,270
In addition, these people are
contacts, part of your network

1111
00:44:28,270 --> 00:44:33,125
of understanding what the world
of software developers

1112
00:44:33,125 --> 00:44:38,490
is all about, and can serve as
useful contacts to keep.

1113
00:44:38,490 --> 00:44:40,620
How many people use a service
like LinkedIn

1114
00:44:40,620 --> 00:44:41,580
or something to--

1115
00:44:41,580 --> 00:44:43,280
You probably all you mostly
use Facebook.

1116
00:44:43,280 --> 00:44:45,410
Anybody use LinkedIn
at this point?

1117
00:44:45,410 --> 00:44:47,180
So many professionals use--

1118
00:44:47,180 --> 00:44:50,380
I don't want to pitch for
LinkedIn, but it is one of the

1119
00:44:50,380 --> 00:44:53,470
services that lots
of people are on.

1120
00:44:53,470 --> 00:44:56,990
So what are some of the other
services that people use for

1121
00:44:56,990 --> 00:44:58,210
professional contacts?

1122
00:44:58,210 --> 00:44:59,524
AUDIENCE: The only thing
people try to

1123
00:44:59,524 --> 00:45:00,840
contact me on is LinkedIn.

1124
00:45:00,840 --> 00:45:01,590
PROFESSOR 2: Yeah.

1125
00:45:01,590 --> 00:45:04,870
So, good idea to start
connecting and building a

1126
00:45:04,870 --> 00:45:08,770
network of people who you
know and who can answer

1127
00:45:08,770 --> 00:45:10,280
questions for you.

1128
00:45:10,280 --> 00:45:10,660
OK?

1129
00:45:10,660 --> 00:45:13,410
Such as, gee, I'm considering
these two kinds of job offers.

1130
00:45:13,410 --> 00:45:14,470
What should I be looking for?

1131
00:45:14,470 --> 00:45:16,660
It might be good to get a
professional software

1132
00:45:16,660 --> 00:45:18,350
engineer's input to that.

1133
00:45:18,350 --> 00:45:18,580
OK?

1134
00:45:18,580 --> 00:45:19,460
Because you know these folks.

1135
00:45:19,460 --> 00:45:21,150
You've had a few meetings
with them.

1136
00:45:21,150 --> 00:45:21,480
OK?

1137
00:45:21,480 --> 00:45:22,380
And they know you.

1138
00:45:22,380 --> 00:45:26,150
And they may be able to say
here's my take on that, as

1139
00:45:26,150 --> 00:45:28,920
well as getting input from your
family and your friends

1140
00:45:28,920 --> 00:45:33,710
and your professors and other
people that you know.

1141
00:45:33,710 --> 00:45:42,170
But I do think that courtesy
goes a long way in life.

1142
00:45:42,170 --> 00:45:44,930
And they really have, these are
all people, as I say, who

1143
00:45:44,930 --> 00:45:46,620
spent time looking
over your code.

1144
00:45:46,620 --> 00:45:48,530
And you should have heard some
of the comments we got from

1145
00:45:48,530 --> 00:45:50,790
them after project one.

1146
00:45:50,790 --> 00:45:53,830
Oh my God, they have a single
letter variable names for the

1147
00:45:53,830 --> 00:45:54,590
whole thing.

1148
00:45:54,590 --> 00:45:57,510
Or, what are you teaching
these students?

1149
00:45:57,510 --> 00:46:01,450
And I said no, this is exactly
why we want you there.

1150
00:46:01,450 --> 00:46:02,380
OK?

1151
00:46:02,380 --> 00:46:06,350
So I think that if you feel in
particular, whether or not you

1152
00:46:06,350 --> 00:46:08,110
feel you got anything
out of it, OK?

1153
00:46:08,110 --> 00:46:10,550
But in particular, if you did
get something out of it, it

1154
00:46:10,550 --> 00:46:14,100
would be really good idea to not
forget to send some email

1155
00:46:14,100 --> 00:46:16,330
off to thank them.

1156
00:46:19,394 --> 00:46:22,030
PROFESSOR 1: OK, we want to get
some feedback from them.

1157
00:46:22,030 --> 00:46:22,470
PROFESSOR 2: Sure.

1158
00:46:22,470 --> 00:46:23,130
PROFESSOR 1: Have open mic?

1159
00:46:23,130 --> 00:46:23,660
PROFESSOR 2: Open mic?

1160
00:46:23,660 --> 00:46:24,970
PROFESSOR 1: Yeah.

1161
00:46:24,970 --> 00:46:26,860
PROFESSOR 2: Actually, open mic,
let's do it like this.

1162
00:46:26,860 --> 00:46:29,400
Because I think it'll be easier
and better for the

1163
00:46:29,400 --> 00:46:33,190
video if we just have
people come up and--

1164
00:46:33,190 --> 00:46:35,435
PROFESSOR 1: I don't know that
they're going to be too

1165
00:46:35,435 --> 00:46:35,590
intimidating.

1166
00:46:35,590 --> 00:46:38,290
PROFESSOR 2: That's OK.

1167
00:46:38,290 --> 00:46:39,070
They can handle it.

1168
00:46:39,070 --> 00:46:40,370
These are seasoned--

1169
00:46:44,130 --> 00:46:46,790
So, comments about the class.

1170
00:46:46,790 --> 00:46:48,315
PROFESSOR 1: How
we can improve?

1171
00:46:48,315 --> 00:46:49,460
We'd really like to hear that.

1172
00:46:49,460 --> 00:46:52,720
I mean things you liked
and didn't like.

1173
00:46:52,720 --> 00:46:57,620
For example, these absolute
gradings we did for projects

1174
00:46:57,620 --> 00:47:02,300
had passionate support and
control let's say before.

1175
00:47:02,300 --> 00:47:05,370
I mean, did you guys feel giving
the top marke for the

1176
00:47:05,370 --> 00:47:08,400
fastest and having at least a
log scale, no, we didn't do a

1177
00:47:08,400 --> 00:47:11,790
linear scale going down.

1178
00:47:11,790 --> 00:47:12,540
I mean, hey.

1179
00:47:12,540 --> 00:47:14,270
Was it motivating
or demotivating?

1180
00:47:16,900 --> 00:47:18,750
PROFESSOR 2: So what do people
think was the most valuable

1181
00:47:18,750 --> 00:47:20,510
thing out of the course?

1182
00:47:20,510 --> 00:47:23,610
I've had dinner with a few
of you, and you were very

1183
00:47:23,610 --> 00:47:26,470
opinionated over dinner.

1184
00:47:26,470 --> 00:47:30,480
So I don't mind hearing those
comments again, but making

1185
00:47:30,480 --> 00:47:32,360
them for other people.

1186
00:47:32,360 --> 00:47:33,680
So I know it's intimidating,
but what I want

1187
00:47:33,680 --> 00:47:34,520
to do is get a line.

1188
00:47:34,520 --> 00:47:36,360
As one person's talking
I want other

1189
00:47:36,360 --> 00:47:37,400
people to line up behind.

1190
00:47:37,400 --> 00:47:40,420
And just give us
some feedback.

1191
00:47:40,420 --> 00:47:41,260
PROFESSOR 1: Come on.

1192
00:47:41,260 --> 00:47:42,740
PROFESSOR 2: OK, come on, come
on, come on, come on.

1193
00:47:42,740 --> 00:47:43,270
Who's brave?

1194
00:47:43,270 --> 00:47:44,270
Yeah, yeah, yeah yeah.

1195
00:47:44,270 --> 00:47:47,130
You got halfway of your seat.

1196
00:47:47,130 --> 00:47:47,560
Yeah.

1197
00:47:47,560 --> 00:47:48,330
PROFESSOR 1: Yeah, come on.

1198
00:47:48,330 --> 00:47:49,350
PROFESSOR 2: I'm sure
you have an opinion.

1199
00:47:49,350 --> 00:47:51,235
Who here does not
have an opinion?

1200
00:47:51,235 --> 00:47:51,700
PROFESSOR 1: Sure.

1201
00:47:51,700 --> 00:47:52,630
Yeah.

1202
00:47:52,630 --> 00:47:54,030
PROFESSOR 2: Yeah, yeah, good.

1203
00:47:54,030 --> 00:47:56,820
And then you can
line up right--

1204
00:47:56,820 --> 00:47:58,590
there you go.

1205
00:47:58,590 --> 00:47:59,150
OK?

1206
00:47:59,150 --> 00:48:00,350
There you go.

1207
00:48:00,350 --> 00:48:04,580
And then others can come up
behind as they're motivated.

1208
00:48:04,580 --> 00:48:05,550
AUDIENCE: Good afternoon.

1209
00:48:05,550 --> 00:48:08,240
First of all, I'd like to thank
the professors and the

1210
00:48:08,240 --> 00:48:09,320
TAs for this opportunity.

1211
00:48:09,320 --> 00:48:11,310
This is absolutely wonderful.

1212
00:48:11,310 --> 00:48:13,460
And all the students
I worked with.

1213
00:48:13,460 --> 00:48:14,710
You were all great.

1214
00:48:17,040 --> 00:48:19,780
I just wanted to say when you
were talking about what the

1215
00:48:19,780 --> 00:48:23,030
most valuable thing in the class
people thought was, I

1216
00:48:23,030 --> 00:48:25,290
don't even know how to pick.

1217
00:48:25,290 --> 00:48:30,530
Of the projects that we worked
on, or that they worked on, I

1218
00:48:30,530 --> 00:48:34,440
looked at every one and just
thought, what of the content

1219
00:48:34,440 --> 00:48:38,700
of this project have I used
just in the past year?

1220
00:48:38,700 --> 00:48:41,230
And the amount of stuff that
I've made significant use of

1221
00:48:41,230 --> 00:48:47,500
that you guys covered has been
at least 75%, maybe even more,

1222
00:48:47,500 --> 00:48:49,100
of the material of the class.

1223
00:48:49,100 --> 00:48:51,210
And that means that if you look
over multiple engineers,

1224
00:48:51,210 --> 00:48:55,580
you probably pretty quickly get
coverage of 100% of it.

1225
00:48:55,580 --> 00:48:59,750
All these projects, profiling,
parallelism, algorithms, bit

1226
00:48:59,750 --> 00:49:04,640
hacks, these are all things
that I use on.

1227
00:49:04,640 --> 00:49:07,033
if not a daily basis,
then a weekly basis.

1228
00:49:07,033 --> 00:49:07,920
PROFESSOR 2: OK.

1229
00:49:07,920 --> 00:49:09,330
Good.

1230
00:49:09,330 --> 00:49:09,827
Thanks.

1231
00:49:09,827 --> 00:49:10,821
Thank you.

1232
00:49:10,821 --> 00:49:13,810
[APPLAUSE]

1233
00:49:13,810 --> 00:49:16,370
AUDIENCE: So for me, like, I
really had no experience with

1234
00:49:16,370 --> 00:49:20,130
C or C++ or really distributed
version

1235
00:49:20,130 --> 00:49:21,770
control before this class.

1236
00:49:21,770 --> 00:49:24,440
So a lot of my time I think was
spent kind of learning the

1237
00:49:24,440 --> 00:49:26,210
basics of that.

1238
00:49:26,210 --> 00:49:27,800
I spent a lot of time initially
learning the basics

1239
00:49:27,800 --> 00:49:30,280
like pointer arithmetic and
that kind of stuff.

1240
00:49:30,280 --> 00:49:32,330
And then a lot of time learning
the profiling tools.

1241
00:49:32,330 --> 00:49:35,860
So I wish, I really do wish more
time was spent kind of--

1242
00:49:35,860 --> 00:49:38,170
well, your recitation, I think,
would be a good way to

1243
00:49:38,170 --> 00:49:39,810
really get that out
of the way.

1244
00:49:39,810 --> 00:49:42,300
Because the algorithmic stuff
was more easy for me, which I

1245
00:49:42,300 --> 00:49:43,245
thought was kind of ridiculous.

1246
00:49:43,245 --> 00:49:44,190
You know?

1247
00:49:44,190 --> 00:49:45,660
[LAUGHTER]

1248
00:49:45,660 --> 00:49:48,250
AUDIENCE: The fact that these
tools are what's holding back,

1249
00:49:48,250 --> 00:49:51,050
that I have to spend a
lot of time learning.

1250
00:49:51,050 --> 00:49:53,000
PROFESSOR 1: This is actually
something very interesting

1251
00:49:53,000 --> 00:49:55,200
people have observed.

1252
00:49:55,200 --> 00:49:58,720
So if you do physics problems
or math problems, if you run

1253
00:49:58,720 --> 00:50:02,550
into an issue, after a certain
amount of time you will just

1254
00:50:02,550 --> 00:50:04,090
go ask somebody.

1255
00:50:04,090 --> 00:50:08,500
But in programming, people seem
to go on and on and on

1256
00:50:08,500 --> 00:50:10,760
trying to debug this
simple thing.

1257
00:50:10,760 --> 00:50:15,830
And part of that, part of what
we try to get is also not just

1258
00:50:15,830 --> 00:50:19,930
all the cool algorithms
and stuff like that.

1259
00:50:19,930 --> 00:50:21,490
To give you a process.

1260
00:50:21,490 --> 00:50:23,480
Because there are times when you
start saying this program

1261
00:50:23,480 --> 00:50:24,940
is slow, what you do?

1262
00:50:24,940 --> 00:50:27,130
And I think that's a missing
part sometimes about the

1263
00:50:27,130 --> 00:50:28,690
process of debugging.

1264
00:50:28,690 --> 00:50:30,980
If the program doesn't
work, what do you do?

1265
00:50:30,980 --> 00:50:31,870
Where do you start?

1266
00:50:31,870 --> 00:50:34,680
I think that's why we are trying
to do a little bit of,

1267
00:50:34,680 --> 00:50:36,670
basically, these life skills.

1268
00:50:36,670 --> 00:50:38,760
AUDIENCE: It was great to
learn but it was more

1269
00:50:38,760 --> 00:50:41,150
stressful to learn.

1270
00:50:41,150 --> 00:50:43,630
I'm not as willing to ask as a
lot of people, so it's hard

1271
00:50:43,630 --> 00:50:44,440
for me to--

1272
00:50:44,440 --> 00:50:45,750
PROFESSOR 1: So one thing we're

1273
00:50:45,750 --> 00:50:46,650
trying to do is probably--

1274
00:50:46,650 --> 00:50:47,530
PROFESSOR 2: That's
great feedback.

1275
00:50:47,530 --> 00:50:51,130
PROFESSOR 2: We will try to
formalize next year the

1276
00:50:51,130 --> 00:50:53,160
seventh floor lounge, because
a lot of people

1277
00:50:53,160 --> 00:50:54,520
who came there benefited.

1278
00:50:54,520 --> 00:50:55,940
But have it a little bit more.

1279
00:50:55,940 --> 00:50:59,610
And also get a little bit
more life skills type

1280
00:50:59,610 --> 00:51:00,470
things early on.

1281
00:51:00,470 --> 00:51:02,330
Like if you find a problem,
what will happen?

1282
00:51:02,330 --> 00:51:03,770
AUDIENCE: It could be a
lot longer than that.

1283
00:51:03,770 --> 00:51:04,250
PROFESSOR 1: Yeah.

1284
00:51:04,250 --> 00:51:06,220
PROFESSOR 2: Yes, so that's
our goal, is to make it so

1285
00:51:06,220 --> 00:51:08,880
that people come up to
speed more quickly.

1286
00:51:08,880 --> 00:51:13,260
But in some sense, another goal
but I have here, next

1287
00:51:13,260 --> 00:51:16,760
year, is that the C students
will stay in the class.

1288
00:51:16,760 --> 00:51:17,050
OK?

1289
00:51:17,050 --> 00:51:21,810
So what happened is most of the
C students dropped out.

1290
00:51:21,810 --> 00:51:23,900
So in case you want to know,
most people are going to get

1291
00:51:23,900 --> 00:51:25,210
A's and B's.

1292
00:51:25,210 --> 00:51:26,130
OK?

1293
00:51:26,130 --> 00:51:27,810
So most of the C students
dropped out.

1294
00:51:27,810 --> 00:51:30,270
And I would like to have people
complete the course,

1295
00:51:30,270 --> 00:51:33,680
even the C students, feeling
that it's useful to complete

1296
00:51:33,680 --> 00:51:34,930
the course.

1297
00:51:36,760 --> 00:51:40,320
AUDIENCE: So I think again,
as the person before me, I

1298
00:51:40,320 --> 00:51:40,980
learned a lot.

1299
00:51:40,980 --> 00:51:43,990
I hadn't had almost any
experience in C or

1300
00:51:43,990 --> 00:51:45,620
C++ before the class.

1301
00:51:45,620 --> 00:51:46,700
I had some experience
with git.

1302
00:51:46,700 --> 00:51:49,060
But I think that the class
has a lot to offer.

1303
00:51:49,060 --> 00:51:50,720
There's a lot of content
covered.

1304
00:51:50,720 --> 00:51:55,220
And I wish that, for example,
there are things that I wish

1305
00:51:55,220 --> 00:51:58,160
that I could have learned more
about, how to implement in a

1306
00:51:58,160 --> 00:51:59,260
formal way.

1307
00:51:59,260 --> 00:52:00,990
Things like SSE instructions.

1308
00:52:00,990 --> 00:52:05,265
So I feel that like there was a
clear cut off of people who

1309
00:52:05,265 --> 00:52:07,880
had a lot of experience and they
could be like, OK, I know

1310
00:52:07,880 --> 00:52:10,420
how to do assembly, I know how
to do other things, and if I

1311
00:52:10,420 --> 00:52:12,280
want to optimize this function
I can, whatever.

1312
00:52:12,280 --> 00:52:13,190
Write it down.

1313
00:52:13,190 --> 00:52:17,180
And then there's these other
people that don't have that

1314
00:52:17,180 --> 00:52:21,066
much experience that in a class
it mentions this is how

1315
00:52:21,066 --> 00:52:23,850
you use SSE or compiler
optimizations.

1316
00:52:23,850 --> 00:52:27,870
It's kind of mentioned but
not really more in depth.

1317
00:52:27,870 --> 00:52:31,190
This is a case of how do you
actually do this in real life?

1318
00:52:31,190 --> 00:52:33,950
I mean, I think maybe the
examples given in class were

1319
00:52:33,950 --> 00:52:35,410
useful but very superfluous.

1320
00:52:35,410 --> 00:52:37,340
Like you don't really know.

1321
00:52:37,340 --> 00:52:40,470
I know that I was looking
through other projects that

1322
00:52:40,470 --> 00:52:42,750
were similar to ours
for the final.

1323
00:52:42,750 --> 00:52:45,810
And I think it's a class
that you taught in IAP.

1324
00:52:45,810 --> 00:52:47,960
There was a team of people that
did a ray tracer in the

1325
00:52:47,960 --> 00:52:49,010
multi-programming.

1326
00:52:49,010 --> 00:52:52,630
And they used SSE and
CMD everywhere.

1327
00:52:52,630 --> 00:52:55,090
They packaged everything.

1328
00:52:55,090 --> 00:52:57,670
But that's something that I
said, it'll be really awesome

1329
00:52:57,670 --> 00:53:00,510
to do, but I have no idea
how to start doing it.

1330
00:53:00,510 --> 00:53:03,940
And I think that there's still
a lot of room to actually

1331
00:53:03,940 --> 00:53:05,660
apply the concepts that are
trying to be taught.

1332
00:53:05,660 --> 00:53:07,710
PROFESSOR 2: I think it's our
intention actually to increase

1333
00:53:07,710 --> 00:53:10,235
that part of the course,
the data

1334
00:53:10,235 --> 00:53:12,030
parallel part of the course.

1335
00:53:12,030 --> 00:53:13,690
Which I think we'll
be able to do.

1336
00:53:13,690 --> 00:53:17,000
One of our concerns about adding
material has been, OK,

1337
00:53:17,000 --> 00:53:18,500
that means you guys have
to learn it too.

1338
00:53:22,340 --> 00:53:26,720
But I think that if we're able
to reduce the workload of the

1339
00:53:26,720 --> 00:53:30,430
learning of the tools by having
recitations, we then

1340
00:53:30,430 --> 00:53:36,420
can look at adding other
material, and expect that the

1341
00:53:36,420 --> 00:53:39,270
load on students won't go up.

1342
00:53:39,270 --> 00:53:39,480
AUDIENCE: Right.

1343
00:53:39,480 --> 00:53:40,870
I mean even what
I want to say.

1344
00:53:40,870 --> 00:53:43,510
It's not even adding material,
what I think.

1345
00:53:43,510 --> 00:53:46,080
I feel that I was going to
class, learning all of these

1346
00:53:46,080 --> 00:53:46,670
cool things.

1347
00:53:46,670 --> 00:53:48,950
And then I was looking at my
project and it's like well,

1348
00:53:48,950 --> 00:53:50,610
these are the things that
I know how to do.

1349
00:53:50,610 --> 00:53:51,610
And I know that they
mentioned all these

1350
00:53:51,610 --> 00:53:52,580
other things in class.

1351
00:53:52,580 --> 00:53:56,980
I don't know how to make
them in my project.

1352
00:53:56,980 --> 00:53:57,720
PROFESSOR 2: Great comment.

1353
00:53:57,720 --> 00:54:01,130
PROFESSOR 1: So I think one
thing both good and bad about

1354
00:54:01,130 --> 00:54:04,340
this class is what
happened in--

1355
00:54:04,340 --> 00:54:05,550
I think it applied
to all of you.

1356
00:54:05,550 --> 00:54:09,810
When you try to teach computer
science, you get people who

1357
00:54:09,810 --> 00:54:12,450
are for the first time they are
writing any piece of code

1358
00:54:12,450 --> 00:54:14,120
more than 20 lines.

1359
00:54:14,120 --> 00:54:18,170
And you get people who are
checking in code for Apache.

1360
00:54:18,170 --> 00:54:21,660
And you have that entire range
of people in the class.

1361
00:54:21,660 --> 00:54:23,440
So how do you teach everybody?

1362
00:54:23,440 --> 00:54:26,500
And a lot of times when you give
a fixed project, either

1363
00:54:26,500 --> 00:54:29,290
somebody's completely struggling
to finish that, or

1364
00:54:29,290 --> 00:54:30,360
they are like in
a day is done.

1365
00:54:30,360 --> 00:54:31,510
They are like, so what's this?

1366
00:54:31,510 --> 00:54:32,150
What's new?

1367
00:54:32,150 --> 00:54:35,150
The thing about performance
is, asked lot of people,

1368
00:54:35,150 --> 00:54:36,830
figured out there's
no end to it.

1369
00:54:36,830 --> 00:54:38,820
And you can keep doing
it, you can do more,

1370
00:54:38,820 --> 00:54:40,120
you can do more deeper.

1371
00:54:40,120 --> 00:54:41,440
So the nice thing about--

1372
00:54:41,440 --> 00:54:43,580
one thing about this class is
that there's something for

1373
00:54:43,580 --> 00:54:45,080
everybody in that.

1374
00:54:45,080 --> 00:54:47,715
So if you are struggling still,
if you're learning, you

1375
00:54:47,715 --> 00:54:49,540
can still get to someplace,
learn a lot.

1376
00:54:49,540 --> 00:54:53,420
But if you have done that when
you are in the fourth grade,

1377
00:54:53,420 --> 00:54:55,540
you can still go and do
something more, push that.

1378
00:54:55,540 --> 00:54:58,140
So we try to keep that
balance in there.

1379
00:54:58,140 --> 00:54:59,450
So we don't want to
bore anybody.

1380
00:54:59,450 --> 00:55:01,860
And also we don't want a lot of
people dropping out of the

1381
00:55:01,860 --> 00:55:04,470
class either.

1382
00:55:04,470 --> 00:55:06,650
AUDIENCE: So continuing on with
what other people said, I

1383
00:55:06,650 --> 00:55:08,750
also ran into the same
problems with that.

1384
00:55:08,750 --> 00:55:10,440
Just the syntax,
just the tools.

1385
00:55:10,440 --> 00:55:14,140
So I guess I'll start off
thanking John and Reed,

1386
00:55:14,140 --> 00:55:17,330
they're really fast and good
at what they're doing.

1387
00:55:17,330 --> 00:55:20,070
And I feel like I've asked a
lot of dumb questions, and

1388
00:55:20,070 --> 00:55:23,488
they're good at answering
those.

1389
00:55:23,488 --> 00:55:26,356
[APPLAUSE]

1390
00:55:26,356 --> 00:55:30,560
PROFESSOR 2: So John and
Reed and Josh all took

1391
00:55:30,560 --> 00:55:32,500
the class last year.

1392
00:55:32,500 --> 00:55:32,910
OK?

1393
00:55:32,910 --> 00:55:36,040
And it's amazing, if you
ask John how he did

1394
00:55:36,040 --> 00:55:38,619
on the final project--

1395
00:55:38,619 --> 00:55:40,890
[LAUGHTER]

1396
00:55:40,890 --> 00:55:42,380
PROFESSOR 2: He didn't
do so well.

1397
00:55:42,380 --> 00:55:43,580
OK?

1398
00:55:43,580 --> 00:55:44,780
But it's amazing.

1399
00:55:44,780 --> 00:55:47,260
You'll discover, it's amazing
what you guys are going to

1400
00:55:47,260 --> 00:55:50,440
learn next semester after
being out of the class.

1401
00:55:50,440 --> 00:55:52,400
Because you're on a path to
learning this stuff for

1402
00:55:52,400 --> 00:55:53,800
yourself now.

1403
00:55:53,800 --> 00:55:54,080
OK?

1404
00:55:54,080 --> 00:55:56,280
And you'll find that in all
kinds of things you do that

1405
00:55:56,280 --> 00:55:58,730
all these skills, you can
keep building them.

1406
00:55:58,730 --> 00:56:02,530
We will definitely be looking
for TAs next fall.

1407
00:56:02,530 --> 00:56:05,910
And so people who are
interested should--

1408
00:56:05,910 --> 00:56:09,710
who are going to be
MENGs should--

1409
00:56:09,710 --> 00:56:12,840
and even if you're
undergraduate, you're still

1410
00:56:12,840 --> 00:56:15,070
allowed to work hourly
for the class.

1411
00:56:15,070 --> 00:56:18,530
In fact, that's what
John is doing.

1412
00:56:18,530 --> 00:56:21,990
So if you're interested,
please let us know.

1413
00:56:21,990 --> 00:56:23,010
Please let us know.

1414
00:56:23,010 --> 00:56:29,230
PROFESSOR 1: So you can have
this nice experience as a TA.

1415
00:56:29,230 --> 00:56:32,010
AUDIENCE: Yeah, so, and then I
want to say something that I

1416
00:56:32,010 --> 00:56:34,770
really love about this course.

1417
00:56:34,770 --> 00:56:37,410
I feel like it's such an
integrated experience.

1418
00:56:37,410 --> 00:56:42,340
It's, maybe except project one,
everything after that is,

1419
00:56:42,340 --> 00:56:45,450
you get a piece of code, as you
would normally, and then

1420
00:56:45,450 --> 00:56:48,410
everything is within
your bounds of--

1421
00:56:48,410 --> 00:56:49,980
you can do anything
that you want.

1422
00:56:49,980 --> 00:56:53,270
So I really like how it's not
limited to like, oh, we're on

1423
00:56:53,270 --> 00:56:57,040
this topic so we're going to
write a little algorithm that

1424
00:56:57,040 --> 00:57:00,090
implements this particular
algorithm, and then move on to

1425
00:57:00,090 --> 00:57:02,560
next unit, and then nothing
ever comes together.

1426
00:57:02,560 --> 00:57:05,780
So that's something I really
love about this course.

1427
00:57:05,780 --> 00:57:12,210
And the next thing I want to
say is the beta grading.

1428
00:57:12,210 --> 00:57:14,900
For project three, I was working
with Angela, I feel

1429
00:57:14,900 --> 00:57:21,510
like from beta to final, because
I don't have much

1430
00:57:21,510 --> 00:57:27,670
experience with C, so it takes
me so long to just get the

1431
00:57:27,670 --> 00:57:30,830
syntax right, to get
my own going.

1432
00:57:30,830 --> 00:57:32,580
So for beta, we didn't have
anything working.

1433
00:57:32,580 --> 00:57:33,780
It doesn't even build.

1434
00:57:33,780 --> 00:57:38,240
But with two or three more
hours, we got it working.

1435
00:57:38,240 --> 00:57:40,190
It's not super fast, but
we kept working on it.

1436
00:57:40,190 --> 00:57:44,290
And by the final we have
something I felt pretty proud.

1437
00:57:44,290 --> 00:57:47,160
But the beta grading doesn't
take that into account.

1438
00:57:47,160 --> 00:57:50,590
Like if you are not very
experienced with C, it takes

1439
00:57:50,590 --> 00:57:53,890
you a long time, longer than
other people have experienced

1440
00:57:53,890 --> 00:57:56,700
to get something going
in the first place.

1441
00:57:56,700 --> 00:58:00,290
But beta kind of penalizes you
for just not having enough

1442
00:58:00,290 --> 00:58:00,710
time to do things you want.

1443
00:58:00,710 --> 00:58:04,060
PROFESSOR 1: I think it's
a double-edged sword.

1444
00:58:04,060 --> 00:58:08,440
Because if you don't do beta,
then you have very harsh.

1445
00:58:08,440 --> 00:58:11,240
You have one time code,
it's in and out.

1446
00:58:11,240 --> 00:58:14,680
If you just really don't do,
give that much grading for

1447
00:58:14,680 --> 00:58:16,790
beta and make it optional
and stuff like that,

1448
00:58:16,790 --> 00:58:17,550
people just drop beta.

1449
00:58:17,550 --> 00:58:19,100
They don't just do
anything in it.

1450
00:58:19,100 --> 00:58:21,570
So we try to do something, give
everybody a true chance

1451
00:58:21,570 --> 00:58:24,562
and get everybody a real
motivation to actually work

1452
00:58:24,562 --> 00:58:25,250
for both sides.

1453
00:58:25,250 --> 00:58:27,840
Because for a lot of you
guys, beta defined and

1454
00:58:27,840 --> 00:58:28,910
you learned a lot.

1455
00:58:28,910 --> 00:58:30,560
Because you got something
in beta.

1456
00:58:30,560 --> 00:58:34,150
And for most of the classes I
have, we had things called

1457
00:58:34,150 --> 00:58:35,450
checkpoints.

1458
00:58:35,450 --> 00:58:38,810
For example, compiler class, you
had to do a checkpoint and

1459
00:58:38,810 --> 00:58:42,800
checkpoint will be used more
like subjectivity to grade.

1460
00:58:42,800 --> 00:58:44,510
We don't say how many grade
percent checkpoint.

1461
00:58:44,510 --> 00:58:47,630
But if checkpoint has nothing,
then we'll be very harsh

1462
00:58:47,630 --> 00:58:48,495
grading later.

1463
00:58:48,495 --> 00:58:49,490
We always provide that.

1464
00:58:49,490 --> 00:58:51,960
But still, people don't take
checkpoint that seriously.

1465
00:58:51,960 --> 00:58:55,140
But here we have to try to make
something to make people

1466
00:58:55,140 --> 00:58:58,210
get beta seriously, do something
there, and then give

1467
00:58:58,210 --> 00:59:00,350
them the opportunity
to improve.

1468
00:59:00,350 --> 00:59:01,490
I think there's a balance
in that somewhere.

1469
00:59:01,490 --> 00:59:03,513
PROFESSOR 2: And I wouldn't
worry about your grade.

1470
00:59:03,513 --> 00:59:05,080
[LAUGHTER]

1471
00:59:05,080 --> 00:59:07,850
AUDIENCE: Maybe like one idea
I was thinking is maybe

1472
00:59:07,850 --> 00:59:09,660
there's some component
of improvement

1473
00:59:09,660 --> 00:59:11,490
from beta to final.

1474
00:59:11,490 --> 00:59:13,680
Like the proportional
improvement that you made from

1475
00:59:13,680 --> 00:59:13,950
beta to final.

1476
00:59:13,950 --> 00:59:15,020
PROFESSOR 2: Oh, that's
a good idea, actually.

1477
00:59:15,020 --> 00:59:15,835
That's a good idea.

1478
00:59:15,835 --> 00:59:17,920
Yeah, to say, if you made a--

1479
00:59:17,920 --> 00:59:19,650
Although you want to be
careful about people

1480
00:59:19,650 --> 00:59:20,270
sandbagging.

1481
00:59:20,270 --> 00:59:21,566
[LAUGHTER]

1482
00:59:21,566 --> 00:59:22,430
PROFESSOR 1: Exactly.

1483
00:59:22,430 --> 00:59:23,840
AUDIENCE: And then one
last small thing--

1484
00:59:23,840 --> 00:59:25,080
PROFESSOR 2: I want to make
sure we get to everybody.

1485
00:59:25,080 --> 00:59:25,350
AUDIENCE: OK, sorry.

1486
00:59:25,350 --> 00:59:26,765
PROFESSOR 2: So if you
want to go around to

1487
00:59:26,765 --> 00:59:28,390
the end of the line.

1488
00:59:28,390 --> 00:59:30,410
AUDIENCE: That's actually
mostly all of it.

1489
00:59:30,410 --> 00:59:30,770
PROFESSOR 2: OK.

1490
00:59:30,770 --> 00:59:32,340
Thank you.

1491
00:59:32,340 --> 00:59:35,130
AUDIENCE: So I've said some of
this on my evaluation, which I

1492
00:59:35,130 --> 00:59:35,690
already turned in.

1493
00:59:35,690 --> 00:59:38,560
But I'll repeat it for
everyone else.

1494
00:59:38,560 --> 00:59:39,990
[APPLAUSE]

1495
00:59:39,990 --> 00:59:40,590
AUDIENCE: A couple things.

1496
00:59:40,590 --> 00:59:45,170
One relating to the grading
which you mentioned.

1497
00:59:45,170 --> 00:59:47,790
Honestly, I found it incredibly
stressful knowing

1498
00:59:47,790 --> 00:59:49,170
that it was competitive
grading.

1499
00:59:49,170 --> 00:59:51,490
Because we had no idea where
the benchmark was.

1500
00:59:51,490 --> 00:59:53,290
So it's like, well we've
improved it.

1501
00:59:53,290 --> 00:59:54,400
But did we improve it enough?

1502
00:59:54,400 --> 00:59:56,670
Did everyone else not improve
it any more than we did?

1503
00:59:56,670 --> 01:00:00,140
And it was just like having it
be a complete unknown was

1504
01:00:00,140 --> 01:00:00,840
pretty stressful.

1505
01:00:00,840 --> 01:00:04,300
So I don't know if there's some
way you could do like a--

1506
01:00:04,300 --> 01:00:07,450
like maybe the TAs got some
speed up, and that's the

1507
01:00:07,450 --> 01:00:10,260
benchmark to start with
or something.

1508
01:00:10,260 --> 01:00:13,390
PROFESSOR 1: I guess what we did
this time is basically we

1509
01:00:13,390 --> 01:00:16,050
gave the best grade as a
benchmark for the final.

1510
01:00:16,050 --> 01:00:19,960
And if you go above that, it's
just bragging rights.

1511
01:00:19,960 --> 01:00:21,699
AUDIENCE: I mean, when you're
working on the beta, you have

1512
01:00:21,699 --> 01:00:22,990
no idea what's going on.

1513
01:00:22,990 --> 01:00:24,590
But I did like that.

1514
01:00:24,590 --> 01:00:27,338
Once you know where the
benchmark is, then it's nice

1515
01:00:27,338 --> 01:00:28,000
to know what you're
aiming for.

1516
01:00:28,000 --> 01:00:29,530
PROFESSOR 1: So the thing about
that is, the flip side

1517
01:00:29,530 --> 01:00:31,940
of that is, we don't want
somebody that a lot of courses

1518
01:00:31,940 --> 01:00:33,770
they say OK, I'm done.

1519
01:00:33,770 --> 01:00:36,000
When you go there something, Oh
yeah, I met the thing, OK,

1520
01:00:36,000 --> 01:00:38,010
I'm getting a good grade,
OK, I move on.

1521
01:00:38,010 --> 01:00:40,990
I mean, we want people kind of
to keep pushing in time.

1522
01:00:40,990 --> 01:00:43,260
Because sometimes
the thing is--

1523
01:00:43,260 --> 01:00:45,620
If I have a minute, I'll
tell the story.

1524
01:00:45,620 --> 01:00:49,570
There's this guy named Tony
Danzig who basically invented

1525
01:00:49,570 --> 01:00:51,690
the simplex method.

1526
01:00:51,690 --> 01:00:54,210
His thing was, he was always
late to class.

1527
01:00:54,210 --> 01:00:56,180
He went to class one
day and there was--

1528
01:00:56,180 --> 01:00:58,420
normally the professor writes
the problems on the board.

1529
01:00:58,420 --> 01:01:01,010
So he copied the problems
down and went home.

1530
01:01:01,010 --> 01:01:02,360
And they were really
hard problems.

1531
01:01:02,360 --> 01:01:03,200
And so he was like, damn.

1532
01:01:03,200 --> 01:01:04,080
I don't know how
to solve this.

1533
01:01:04,080 --> 01:01:06,260
And he worked very hard
and solved it.

1534
01:01:06,260 --> 01:01:08,310
And wrote a note to the
professor saying I'm sorry

1535
01:01:08,310 --> 01:01:10,440
this is late, and put
it on his desk.

1536
01:01:10,440 --> 01:01:13,645
And a couple of hours later he
gets this excited professor

1537
01:01:13,645 --> 01:01:15,090
running around trying
to find him.

1538
01:01:15,090 --> 01:01:17,940
Because what the professor had
done is written two unsolved

1539
01:01:17,940 --> 01:01:19,190
problems on the board.

1540
01:01:21,450 --> 01:01:23,890
He came late to the class, he
didn't know that, he had two

1541
01:01:23,890 --> 01:01:25,010
problems, he copied them down.

1542
01:01:25,010 --> 01:01:26,904
He worked very hard
and solved it.

1543
01:01:26,904 --> 01:01:27,670
[LAUGHTER]

1544
01:01:27,670 --> 01:01:29,730
PROFESSOR 1: And so the
professor said divide this up,

1545
01:01:29,730 --> 01:01:30,640
you get your pay at the end.

1546
01:01:30,640 --> 01:01:31,680
Then there's lots of it.

1547
01:01:31,680 --> 01:01:32,850
So this is something
like that.

1548
01:01:32,850 --> 01:01:36,160
If you give the thing, if you
know where we are going, for

1549
01:01:36,160 --> 01:01:38,100
some people it will
be a hard climb.

1550
01:01:38,100 --> 01:01:40,102
But other people, they will get
there faster, they'll say

1551
01:01:40,102 --> 01:01:40,820
OK, I'm done.

1552
01:01:40,820 --> 01:01:43,530
And what we want to do is give
them the ability to kind of

1553
01:01:43,530 --> 01:01:46,130
really push to their limits,
not some artificial limit.

1554
01:01:46,130 --> 01:01:48,270
But not to penalize
too much for that.

1555
01:01:48,270 --> 01:01:49,630
We try to balance this out.

1556
01:01:49,630 --> 01:01:52,990
So I want everybody to go
to their best limits.

1557
01:01:52,990 --> 01:01:54,790
AUDIENCE: I had one other
thing to say.

1558
01:01:54,790 --> 01:01:57,490
Have you considered or would
it be possible to make the

1559
01:01:57,490 --> 01:01:59,540
class more than a 12
credit course?

1560
01:01:59,540 --> 01:02:01,030
[LAUGHTER]

1561
01:02:01,030 --> 01:02:01,610
AUDIENCE: No, seriously.

1562
01:02:01,610 --> 01:02:04,280
Since there is so much material
and we keep talking

1563
01:02:04,280 --> 01:02:06,390
about wouldn't it be nice to
add this and add that.

1564
01:02:06,390 --> 01:02:08,550
But I don't know about you, but
I spend a lot of time in

1565
01:02:08,550 --> 01:02:09,600
this course already.

1566
01:02:09,600 --> 01:02:12,320
And I mean, I think it's the
kind of class where you could

1567
01:02:12,320 --> 01:02:15,190
pack 16 or 18 credits worth
of material into it.

1568
01:02:15,190 --> 01:02:18,360
PROFESSOR 2: More realistic
would be to have a follow-on.

1569
01:02:18,360 --> 01:02:19,720
PROFESSOR 1: Try to make
it as two classes.

1570
01:02:19,720 --> 01:02:23,980
PROFESSOR 2: To make a
two-semester sequence.

1571
01:02:23,980 --> 01:02:30,710
But I think Saman and I still
feel that this, it's like, OK,

1572
01:02:30,710 --> 01:02:32,810
that's a new development
effort.

1573
01:02:32,810 --> 01:02:33,780
That this class still--

1574
01:02:33,780 --> 01:02:36,550
I think we feel that there's
more than incremental

1575
01:02:36,550 --> 01:02:38,770
improvement we can do to make
the class teachable.

1576
01:02:38,770 --> 01:02:41,240
For example, by introducing
recitations.

1577
01:02:41,240 --> 01:02:44,290
And so I think once--

1578
01:02:44,290 --> 01:02:50,160
maybe after next semester,
next fall we can we can--

1579
01:02:50,160 --> 01:02:52,280
But thank you, because I think
that that would be good.

1580
01:02:52,280 --> 01:02:55,010
And the whole idea of the
sections is to try

1581
01:02:55,010 --> 01:02:57,540
to reduce your load.

1582
01:02:57,540 --> 01:03:00,550
There used to be a thing that
was on all the evaluations

1583
01:03:00,550 --> 01:03:02,080
that they no longer do.

1584
01:03:02,080 --> 01:03:03,930
And this is actually before
I got to MIT.

1585
01:03:03,930 --> 01:03:08,880
They had a measure called
cums per tooling hour.

1586
01:03:08,880 --> 01:03:09,220
OK?

1587
01:03:09,220 --> 01:03:15,040
So a cum is how many units of
knowledge you're using in

1588
01:03:15,040 --> 01:03:19,650
whatever measure, and tooling
hours, like how hard you work.

1589
01:03:19,650 --> 01:03:23,870
And they got rid of it because
they said it was not

1590
01:03:23,870 --> 01:03:24,370
quantitative.

1591
01:03:24,370 --> 01:03:26,580
After all, what's a cum?

1592
01:03:26,580 --> 01:03:26,870
All right.

1593
01:03:26,870 --> 01:03:30,520
Well, regardless of the fact
they got rid of that, because

1594
01:03:30,520 --> 01:03:34,140
I think that it actually was a
very useful measure, even if

1595
01:03:34,140 --> 01:03:35,910
people are subjective
about it.

1596
01:03:35,910 --> 01:03:37,820
That's our goal in the course.

1597
01:03:37,820 --> 01:03:41,120
It's not to make people spend
a lot of time and then have

1598
01:03:41,120 --> 01:03:42,920
learned epsilon.

1599
01:03:42,920 --> 01:03:46,540
It's to have them, for the time
they put in, learn as

1600
01:03:46,540 --> 01:03:47,940
much as possible.

1601
01:03:47,940 --> 01:03:50,480
And so I think that we recognize
it's a lot more work

1602
01:03:50,480 --> 01:03:53,070
right now that you folks
are going through.

1603
01:03:53,070 --> 01:03:54,480
We hope that it's
rewarding work.

1604
01:03:54,480 --> 01:03:57,400
But we hope we can do even
better in terms of giving

1605
01:03:57,400 --> 01:03:59,520
people the opportunity to
learn these things.

1606
01:03:59,520 --> 01:04:00,740
PROFESSOR 1: So we are thinking
about cutting things.

1607
01:04:00,740 --> 01:04:02,530
We all are thinking about
what to eliminate

1608
01:04:02,530 --> 01:04:03,790
even if we add something.

1609
01:04:03,790 --> 01:04:05,680
It's very hard because we
think a lot things are

1610
01:04:05,680 --> 01:04:07,250
useful-- important.

1611
01:04:07,250 --> 01:04:08,510
But something has to give.

1612
01:04:08,510 --> 01:04:09,580
So that's the hard decision.

1613
01:04:09,580 --> 01:04:10,950
PROFESSOR 2: Yeah, we're
thinking of cutting caching

1614
01:04:10,950 --> 01:04:12,546
and parallelism.

1615
01:04:12,546 --> 01:04:16,830
[LAUGHTER]

1616
01:04:16,830 --> 01:04:18,360
PROFESSOR 1: OK.

1617
01:04:18,360 --> 01:04:18,630
PROFESSOR 2: OK.

1618
01:04:18,630 --> 01:04:21,300
We have to hurry because we
gotta be out of the room by 4.

1619
01:04:21,300 --> 01:04:22,865
AUDIENCE: I just wanted
to say that I agree.

1620
01:04:22,865 --> 01:04:24,970
That I didn't really like the
way the course was graded.

1621
01:04:24,970 --> 01:04:26,540
But I did like the material
in the course a lot.

1622
01:04:26,540 --> 01:04:27,950
So that was good.

1623
01:04:27,950 --> 01:04:30,560
But for one thing, the thing
about the competitive grading

1624
01:04:30,560 --> 01:04:33,860
was that it was way more
stressful than necessary.

1625
01:04:33,860 --> 01:04:35,660
My luck was good because
we had the kind

1626
01:04:35,660 --> 01:04:37,350
of benchmark thing.

1627
01:04:37,350 --> 01:04:41,330
It was a reasonable high
performance to attain.

1628
01:04:41,330 --> 01:04:45,650
And also I didn't like
the way some of

1629
01:04:45,650 --> 01:04:46,780
the stuff was weighted.

1630
01:04:46,780 --> 01:04:48,900
For example, it didn't really
correspond to how much time I

1631
01:04:48,900 --> 01:04:49,960
spent on it.

1632
01:04:49,960 --> 01:04:53,360
The pset was worth like 3% or
4% of the grade and it took

1633
01:04:53,360 --> 01:04:54,970
like hours, I think.

1634
01:04:54,970 --> 01:04:56,220
Many, many hours.

1635
01:04:58,800 --> 01:04:59,630
That's all.

1636
01:04:59,630 --> 01:05:00,744
PROFESSOR 1: OK, thank you.

1637
01:05:00,744 --> 01:05:02,512
PROFESSOR 2: OK.

1638
01:05:02,512 --> 01:05:06,340
AUDIENCE: There is one thing
that's pretty general about

1639
01:05:06,340 --> 01:05:07,790
course six, I guess,
is that they're

1640
01:05:07,790 --> 01:05:09,190
really lenient on prereqs.

1641
01:05:09,190 --> 01:05:12,555
Since I've only been at MIT
for like three months.

1642
01:05:12,555 --> 01:05:12,820
[LAUGHTER]

1643
01:05:12,820 --> 01:05:15,270
AUDIENCE: So yes,
I'm a freshman.

1644
01:05:15,270 --> 01:05:16,460
PROFESSOR 1: One freshman
in the class.

1645
01:05:16,460 --> 01:05:17,205
There you go.

1646
01:05:17,205 --> 01:05:20,960
AUDIENCE: I'm really thankful
that Professor Leiserson and

1647
01:05:20,960 --> 01:05:22,670
Professor Amarasinghe
let me into the

1648
01:05:22,670 --> 01:05:25,540
course without any prereqs.

1649
01:05:25,540 --> 01:05:28,610
And I like the breadth
of the topics.

1650
01:05:28,610 --> 01:05:32,600
But I think with office hours
we could have gotten a bit

1651
01:05:32,600 --> 01:05:34,090
deeper into them.

1652
01:05:34,090 --> 01:05:38,050
And one suggestion I'd like to
make is that get the nightly

1653
01:05:38,050 --> 01:05:39,300
builds working.

1654
01:05:41,400 --> 01:05:46,161
It was there for one night for
project one, and that was it.

1655
01:05:46,161 --> 01:05:49,570
[LAUGHTER]

1656
01:05:49,570 --> 01:05:51,840
AUDIENCE: So I decided
this is my last--

1657
01:05:51,840 --> 01:05:53,080
PROFESSOR 2: Can you
talk in the mic?

1658
01:05:53,080 --> 01:05:55,550
AUDIENCE: So this is my
last term at MIT.

1659
01:05:55,550 --> 01:05:58,600
And I decided to take things
easy by not taking that many

1660
01:05:58,600 --> 01:06:01,729
courses and TAing this course.

1661
01:06:01,729 --> 01:06:07,600
[LAUGHTER]

1662
01:06:07,600 --> 01:06:08,230
AUDIENCE: So yeah.

1663
01:06:08,230 --> 01:06:11,183
It would be nice to have
48 unit TA shifts too.

1664
01:06:15,040 --> 01:06:16,290
PROFESSOR 2: Thank you.

1665
01:06:18,610 --> 01:06:19,850
AUDIENCE: I just want to say
that this was like--

1666
01:06:19,850 --> 01:06:21,010
should I take it?

1667
01:06:21,010 --> 01:06:21,730
PROFESSOR 1: Sure.

1668
01:06:21,730 --> 01:06:21,950
AUDIENCE: OK.

1669
01:06:21,950 --> 01:06:24,330
That this was like
an amazing class.

1670
01:06:24,330 --> 01:06:27,100
I think everyone can agree that
we learned so much that

1671
01:06:27,100 --> 01:06:29,060
is not found in textbooks.

1672
01:06:29,060 --> 01:06:31,750
And what I thought was really
weird is that every little

1673
01:06:31,750 --> 01:06:35,480
random thing I've ever read
on the internet related to

1674
01:06:35,480 --> 01:06:38,440
coding, I guess, has come up in
this class in like, useful,

1675
01:06:38,440 --> 01:06:39,610
come up like useful
in this class.

1676
01:06:39,610 --> 01:06:42,550
Like every little bit of outside
knowledge that you

1677
01:06:42,550 --> 01:06:46,240
take yourself has appeared
useful in this class later on.

1678
01:06:46,240 --> 01:06:47,480
And I thought was really cool.

1679
01:06:47,480 --> 01:06:50,850
And related to the hardness that
people are talking about.

1680
01:06:50,850 --> 01:06:54,970
I think everyone came into the
class bracing for the absolute

1681
01:06:54,970 --> 01:06:57,610
worst, because we've heard
about you guys.

1682
01:06:57,610 --> 01:07:00,610
[LAUGHTER]

1683
01:07:00,610 --> 01:07:02,340
AUDIENCE: But, yeah.

1684
01:07:02,340 --> 01:07:05,190
I was actually really pleasantly
surprised that

1685
01:07:05,190 --> 01:07:11,400
there were helpful TAs waiting
in the G7 lounge for you, even

1686
01:07:11,400 --> 01:07:13,020
before office hours start.

1687
01:07:13,020 --> 01:07:16,580
And so I was actually really
pleasantly surprised by--

1688
01:07:16,580 --> 01:07:17,630
just like everyone, the class.

1689
01:07:17,630 --> 01:07:20,960
I feel like everyone who has
survived until this point has

1690
01:07:20,960 --> 01:07:22,690
some sort of magic skill.

1691
01:07:22,690 --> 01:07:25,560
That you really learn a lot
by working with them

1692
01:07:25,560 --> 01:07:27,110
on the group projects.

1693
01:07:27,110 --> 01:07:29,890
And so I was just
really inspired.

1694
01:07:29,890 --> 01:07:31,610
And I thought it was
a great class.

1695
01:07:31,610 --> 01:07:32,030
PROFESSOR 1: Thank you.

1696
01:07:32,030 --> 01:07:32,886
PROFESSOR 2: Thank you.

1697
01:07:32,886 --> 01:07:34,600
[APPLAUSE]

1698
01:07:34,600 --> 01:07:35,850
AUDIENCE: Hi.

1699
01:07:37,480 --> 01:07:38,280
I love the class.

1700
01:07:38,280 --> 01:07:41,600
I appreciate you Professor
Leiserson, and everything that

1701
01:07:41,600 --> 01:07:42,800
everybody else before me said.

1702
01:07:42,800 --> 01:07:43,960
I just wanted to mention
two things I

1703
01:07:43,960 --> 01:07:45,730
don't think were mentioned.

1704
01:07:45,730 --> 01:07:51,680
Beta programming, to be honest,
was amazing because

1705
01:07:51,680 --> 01:07:54,230
except the final project,
everything I did was beta

1706
01:07:54,230 --> 01:07:56,920
programming in the projects.

1707
01:07:56,920 --> 01:08:00,470
And the amount of bugs that we
avoided because of that is

1708
01:08:00,470 --> 01:08:00,990
ridiculous.

1709
01:08:00,990 --> 01:08:03,990
Especially when we're doing
pointers, and if it doesn't

1710
01:08:03,990 --> 01:08:07,400
work, make it a pointer, and all
those philosophies that we

1711
01:08:07,400 --> 01:08:09,500
came up with.

1712
01:08:09,500 --> 01:08:11,110
That's the first thing.

1713
01:08:11,110 --> 01:08:14,140
Second thing is, and I've
already mentioned this.

1714
01:08:14,140 --> 01:08:17,220
A book or readings for
this class or links.

1715
01:08:17,220 --> 01:08:20,859
Because as amazing as the
lecture slides are, a lot of

1716
01:08:20,859 --> 01:08:22,979
work done there, so
thanks a lot.

1717
01:08:22,979 --> 01:08:25,370
They're hard to connect
together, especially when

1718
01:08:25,370 --> 01:08:26,656
you're trying to figure out
what our cache-oblivious

1719
01:08:26,656 --> 01:08:32,270
algorithm is trying
to do with those.

1720
01:08:32,270 --> 01:08:34,278
So that's pretty much it.

1721
01:08:37,149 --> 01:08:38,760
Are you telling me the
stack overflow is as

1722
01:08:38,760 --> 01:08:41,600
much as this course?

1723
01:08:41,600 --> 01:08:43,479
Well, thanks a lot again
for the TAs.

1724
01:08:43,479 --> 01:08:46,184
Most of my learning came from
them, to be honest.

1725
01:08:46,184 --> 01:08:48,112
[LAUGHTER]

1726
01:08:48,112 --> 01:08:51,979
[APPLAUSE]

1727
01:08:51,979 --> 01:08:53,060
AUDIENCE: Hands on learning.

1728
01:08:53,060 --> 01:08:55,220
Hands on.

1729
01:08:55,220 --> 01:08:58,140
Thank you.

1730
01:08:58,140 --> 01:08:58,399
[LAUGHTER]

1731
01:08:58,399 --> 01:08:59,020
GUEST SPEAKER: Hi.

1732
01:08:59,020 --> 01:09:01,130
my name is Harold Pokob,
I'm the head of

1733
01:09:01,130 --> 01:09:03,930
engineering at Akamai.

1734
01:09:03,930 --> 01:09:06,700
So I wanted to use the
opportunity to thank all of

1735
01:09:06,700 --> 01:09:10,029
you for taking the class.

1736
01:09:10,029 --> 01:09:14,250
Charles was my adviser when I
was at MIT and he told me

1737
01:09:14,250 --> 01:09:15,649
about his plans for the class.

1738
01:09:15,649 --> 01:09:17,229
I thought it was a great idea.

1739
01:09:17,229 --> 01:09:22,910
Getting a lot more hands-on
experience going.

1740
01:09:22,910 --> 01:09:26,290
Back when I spent time at MIT,
it was a lot more theory, and

1741
01:09:26,290 --> 01:09:30,870
Scheme was one of the main
programming languages.

1742
01:09:30,870 --> 01:09:36,029
So in that context I really
like this type of class.

1743
01:09:36,029 --> 01:09:40,359
And I'm glad MIT is investing
in that and

1744
01:09:40,359 --> 01:09:41,800
supported from Akamai.

1745
01:09:41,800 --> 01:09:43,660
So I'm honored that we
could sponsor it.

1746
01:09:43,660 --> 01:09:45,630
I'm very happy you're here.

1747
01:09:45,630 --> 01:09:48,010
There is, I think, refreshments
downstairs in a

1748
01:09:48,010 --> 01:09:48,474
few minutes.

1749
01:09:48,474 --> 01:09:48,760
PROFESSOR 2: Yeah.

1750
01:09:48,760 --> 01:09:54,560
I'd also like to ask Harold to
say what the topic of his

1751
01:09:54,560 --> 01:09:56,282
master's thesis was.

1752
01:09:56,282 --> 01:09:57,545
[LAUGHTER]

1753
01:09:57,545 --> 01:10:00,340
GUEST SPEAKER: My master's
thesis was cache oblivious

1754
01:10:00,340 --> 01:10:01,600
algorithms.

1755
01:10:01,600 --> 01:10:03,360
[LAUGHTER]

1756
01:10:03,360 --> 01:10:07,280
PROFESSOR 2: So you have Harold
to thank for one of the

1757
01:10:07,280 --> 01:10:10,770
papers that you read
during the term.

1758
01:10:10,770 --> 01:10:13,480
GUEST SPEAKER: It's nice that
something that I did in

1759
01:10:13,480 --> 01:10:15,740
academia is still used.

1760
01:10:15,740 --> 01:10:19,950
Not all now practical business
stuff that I do these days.

1761
01:10:22,770 --> 01:10:25,650
Great to have you here, and
I hope to see some of you

1762
01:10:25,650 --> 01:10:26,320
downstairs.

1763
01:10:26,320 --> 01:10:28,950
And I've got to actually go
into a meeting at 4:00 for

1764
01:10:28,950 --> 01:10:31,570
half an hour, but I'll be
back down afterwards.

1765
01:10:31,570 --> 01:10:32,250
Thank you.

1766
01:10:32,250 --> 01:10:34,180
PROFESSOR 1: Thank you very much
for sponsoring this event

1767
01:10:34,180 --> 01:10:36,840
and getting all the food.

1768
01:10:36,840 --> 01:10:38,310
[APPLAUSE]

1769
01:10:38,310 --> 01:10:39,223
PROFESSOR 1: OK.

1770
01:10:39,223 --> 01:10:41,230
Now we have food.

1771
01:10:41,230 --> 01:10:43,330
AUDIENCE: We also have
quizzes here.

1772
01:10:43,330 --> 01:10:44,880
[LAUGHTER]