WEBVTT

1
00:20:20.818 --> 00:20:28.169
Okay, Hello everyone I don't TV. I on Q talk just ended. I don't know if anyone had a chance to.

2
00:20:28.169 --> 00:20:33.449
Attended it was an excellent talk. Excellent introduction to quantum computing.

3
00:20:33.449 --> 00:20:40.648
So, a chance to introduce to that part of paranoia computing.

4
00:20:40.648 --> 00:20:46.828
So, can you hear me.

5
00:20:47.878 --> 00:20:51.269
That's here.

6
00:20:51.269 --> 00:20:57.328
And good, thank you. Okay, so, because well, that was.

7
00:20:57.328 --> 00:21:02.699
An optional part of the class. This is going to be really short right now. So, let me.

8
00:21:02.699 --> 00:21:06.239
Think we've got 2 talks left over from.

9
00:21:06.239 --> 00:21:12.088
From Thursday from Monday rather is basically that.

10
00:21:12.088 --> 00:21:18.239
And so let's see, who is.

11
00:21:18.239 --> 00:21:21.269
Who is on then.

12
00:21:22.348 --> 00:21:28.528
The 2nd.

13
00:21:28.528 --> 00:21:37.469
In parallel here.

14
00:21:39.479 --> 00:21:43.769
Whoops, this getting the parallel thing on.

15
00:21:47.548 --> 00:21:50.608
Okay.

16
00:21:53.128 --> 00:21:56.848
Silence.

17
00:21:56.848 --> 00:22:01.048
Me a 2nd, here.

18
00:22:01.048 --> 00:22:04.798
Silence.

19
00:22:06.209 --> 00:22:09.449
Okay, and they.

20
00:22:10.828 --> 00:22:15.929
Just share this.

21
00:22:24.743 --> 00:22:39.294
Okay, so is it sharing that that was this? The slides may become available also and now 1.1St. Okay the other thing is Ben in corner. If you're online, can talk to us. This 1st, 1, comment about computing is.

22
00:22:41.398 --> 00:22:45.989
We've got all these different computer systems, and they don't talk to each other.

23
00:22:47.753 --> 00:22:48.263
So,

24
00:22:48.263 --> 00:22:49.374
if you joined late,

25
00:22:49.403 --> 00:22:52.284
or if you're not in everything well,

26
00:22:52.314 --> 00:22:53.574
if you don't have,

27
00:22:53.874 --> 00:22:54.203
you know,

28
00:22:54.203 --> 00:22:56.963
if you're not watching this lecture was and you're not seeing this,

29
00:22:57.203 --> 00:22:59.544
unless you're reading the blog but in any case,

30
00:22:59.544 --> 00:23:03.564
if you're not in great scope or not getting this invitation then,

31
00:23:03.564 --> 00:23:07.344
let me know LMS you get added automatically for these,

32
00:23:07.344 --> 00:23:09.324
that just stupid final grades and.

33
00:23:11.308 --> 00:23:19.078
Use webx teams for messages, which I'm not checking during the class and so on. Okay. So any questions about.

34
00:23:19.078 --> 00:23:26.818
That so we have Ben and Connor if you want to talk, let me just stop this and.

35
00:23:26.818 --> 00:23:32.669
Yeah, Ben, I see you're online if you would like.

36
00:23:32.669 --> 00:23:37.888
Tell us about to talk in 6 computers.

37
00:23:37.888 --> 00:23:49.229
Sure, yeah, right. I don't have like a PowerPoint, but I have a little visual while. I'm talking about the 2 computers. Let me see if I can get that. Oh, that's fine. Started.

38
00:23:51.868 --> 00:24:00.838
All right, so this is just 1st, 1, this is invidious. Spleen. Let me see if I can all tab and.

39
00:24:00.838 --> 00:24:03.929
Go back to my notes. Can you still see the the visual.

40
00:24:06.479 --> 00:24:10.679
Yes. Okay. Great. So.

41
00:24:10.679 --> 00:24:16.318
1st, computer supercomputer I'm going to be talking about is saline, which is number 5.

42
00:24:16.318 --> 00:24:25.798
Has this super computer on the top 500 list that me know about it was created by invidia. It's currently located in.

43
00:24:25.798 --> 00:24:28.949
The videos headquarters.

44
00:24:28.949 --> 00:24:33.479
It's the fastest supercomputer used for commercial applications.

45
00:24:33.479 --> 00:24:37.828
And it's used by a variety of companies for its computing power.

46
00:24:37.828 --> 00:24:47.189
Customers of saline NVIDIA for usage include continental walkie, Mark and Microsoft.

47
00:24:47.189 --> 00:24:52.048
It's also used in house spine video for workloads.

48
00:24:52.048 --> 00:25:02.999
And it's a pretty big deal for in video commercially, because it uses their D system, just kind of a platform for setting up.

49
00:25:02.999 --> 00:25:07.288
Supercomputers and there are other.

50
00:25:07.288 --> 00:25:13.709
You know, research institutes and other companies that are interested in using its platform.

51
00:25:13.709 --> 00:25:22.439
University Florida was 1 of them setting up soon. It's very easy to set up compared to previous solutions.

52
00:25:22.439 --> 00:25:28.558
This was really important during the pandemic when you couldn't have a ton of people.

53
00:25:28.558 --> 00:25:33.808
Coming in and building stuff health.

54
00:25:33.808 --> 00:25:38.818
Concerns, um, so it was set up.

55
00:25:38.818 --> 00:25:42.598
We set up by 2 engineers or the course of.

56
00:25:42.598 --> 00:25:46.769
3 weeks, which is very fast.

57
00:25:46.769 --> 00:25:54.929
Achieve 63.4 flops floating point operations per 2nd.

58
00:25:54.929 --> 00:26:00.929
Fun fact about saline is it as a robotic trip that tends to it?

59
00:26:00.929 --> 00:26:09.689
It has 555520 computing cores and 1.12 gigabytes.

60
00:26:09.689 --> 00:26:15.028
1.12M gigabytes memory or 1.12 petabytes.

61
00:26:15.028 --> 00:26:20.249
It uses the epic processors for its use.

62
00:26:20.249 --> 00:26:24.179
It's the 2nd, most energy efficient supercomputer on the top of a bunch of list.

63
00:26:24.179 --> 00:26:28.709
It runs to 20.04 that 1 else. Yes.

64
00:26:28.709 --> 00:26:36.179
It's relatively new, compared to all of the other ones on the topic country list. So there wasn't a ton of.

65
00:26:36.179 --> 00:26:39.838
A lot of the information was that I could find, and it was.

66
00:26:39.838 --> 00:26:44.459
You know, and press releases, but still still interesting.

67
00:26:44.459 --> 00:26:53.969
Some interesting stuff next computer number 6, and we talking about to a.

68
00:26:53.969 --> 00:26:59.759
This was created by the Chinese, national University of defense technology.

69
00:26:59.759 --> 00:27:02.848
It's located in the National supercomputer center.

70
00:27:02.848 --> 00:27:08.159
In guangzho, China, it is 6, fastest, super computer world and I said.

71
00:27:08.159 --> 00:27:12.538
And utilizes Xenon E5 processors.

72
00:27:12.538 --> 00:27:18.689
And for a while, it was the fastest supercomputer in the world. The top 500 list had it.

73
00:27:18.689 --> 00:27:26.999
As its fastest supercomputer, or it's June 2013, November, 2013, June, 2014 and November 2014 lists.

74
00:27:26.999 --> 00:27:31.078
Also, the June, 2015 and November 201500.

75
00:27:31.078 --> 00:27:36.118
It achieved 61.4.

76
00:27:36.118 --> 00:27:41.068
Flops and it planned to double its computing power.

77
00:27:41.068 --> 00:27:44.878
Which would have made it the current 3rd, most powerful computer.

78
00:27:44.878 --> 00:27:51.239
But the U s government rejected intel's application export processors for the computer.

79
00:27:51.239 --> 00:28:01.048
They cited by nuclear research, so until was unable to export.

80
00:28:01.048 --> 00:28:04.318
Um, processors.

81
00:28:04.318 --> 00:28:10.409
For the computers, so in response to that.

82
00:28:10.409 --> 00:28:14.398
The National University of defense technology created the subway.

83
00:28:14.398 --> 00:28:20.368
Tie who light supercomputer using only domestic.

84
00:28:20.368 --> 00:28:23.848
Domestically developed processors and.

85
00:28:23.848 --> 00:28:27.298
Which I believe is the 4th.

86
00:28:27.298 --> 00:28:31.348
Computer on the top of the list, um.

87
00:28:31.348 --> 00:28:34.588
But the 2 a runs.

88
00:28:34.588 --> 00:28:38.308
It has 4.9M computing corps.

89
00:28:38.308 --> 00:28:42.719
And 2.27, petabytes of memory.

90
00:28:42.719 --> 00:28:50.159
It was reportedly used for simulation analysis and government security applications, but since it's a.

91
00:28:50.159 --> 00:28:53.578
It's used by government.

92
00:28:53.578 --> 00:28:57.179
In the National University fast technology is not exactly no.

93
00:28:57.179 --> 00:29:03.058
More of a military application, so I don't know exactly what it was all used for.

94
00:29:03.058 --> 00:29:07.078
Something interesting about it is it was accordingly difficult to use.

95
00:29:07.078 --> 00:29:12.269
And running code for it was challenging with those and experience with the machine and.

96
00:29:12.269 --> 00:29:16.439
So, it was very specific to the, to a.

97
00:29:16.439 --> 00:29:21.628
Yeah, I think that's all I have.

98
00:29:21.628 --> 00:29:27.929
Anyone has any questions yes. Thank you. Very much. Interesting.

99
00:29:27.929 --> 00:29:33.298
So, I'll start with a question, so this looks to be a bigger part of invidious business plan is.

100
00:29:33.298 --> 00:29:37.979
Yeah, like this yeah, so they've been, um.

101
00:29:37.979 --> 00:29:42.479
Working with their companies to use the computing.

102
00:29:42.479 --> 00:29:49.229
The saline for remotely using their computing power and.

103
00:29:49.229 --> 00:29:55.919
Using their framework for building up super computers and other places besides just and videos.

104
00:29:55.919 --> 00:30:00.719
Headquarters it's kind of interesting. The contrast of these 2 computers ones are very.

105
00:30:00.719 --> 00:30:04.888
Commercial 1, with a lot of, you know.

106
00:30:04.888 --> 00:30:08.098
Press releases, and the other 1, it's more of a.

107
00:30:08.098 --> 00:30:13.439
Research your military product with very different stories.

108
00:30:13.439 --> 00:30:17.128
But that is an interesting contrast between the 2 we're researching them.

109
00:30:18.239 --> 00:30:28.439
Yeah, thank you. I have a problem with some of the invidious press releases, so finally polished. I have trouble extracting the actual information. Yeah.

110
00:30:28.439 --> 00:30:32.578
Yeah, a lot of it was look, we have a robot that tends to it and not a lot of like.

111
00:30:32.578 --> 00:30:42.719
Hard information, right I probably even finding which of their are faster than the others. Sometimes, you know, they're all.

112
00:30:42.719 --> 00:30:50.459
Holidays adjectives. Okay. Which is a newer 1, the higher performing 1 anyone else have any questions.

113
00:30:51.628 --> 00:30:57.449
Okay, thank you. I was curious. I don't know for sure about it.

114
00:30:57.449 --> 00:31:02.038
The simulations they were running was they actually like.

115
00:31:02.038 --> 00:31:07.439
Per event that they were doing, like, nuclear stuff and scared, or wasn't like, just skepticism.

116
00:31:07.439 --> 00:31:13.528
Well, we don't really know from what I read the.

117
00:31:13.528 --> 00:31:19.648
Of the National University defense technology said they were using it for.

118
00:31:19.648 --> 00:31:23.909
Simulation analysis and government security applications, which is kind of a vague.

119
00:31:23.909 --> 00:31:36.058
But and I think the US Department of defense said that they were used for potentially nuclear, but I guess there's no way for us individually to know for sure. It could've just been, you know, skepticism.

120
00:31:36.058 --> 00:31:49.979
Of the of the yes becoming the most powerful computer again and 1 thing to maintain you was super computer technology priority. So who knows, you know.

121
00:31:49.979 --> 00:31:58.318
Okay.

122
00:31:58.318 --> 00:32:03.088
So, great Connor are you online?

123
00:32:05.009 --> 00:32:13.558
Yes, would you like to talk.

124
00:32:17.009 --> 00:32:23.759
Connor you logged in are you there?

125
00:32:28.048 --> 00:32:32.608
Not there at the moment.

126
00:32:32.608 --> 00:32:35.878
Silence.

127
00:32:35.878 --> 00:32:41.969
Okay, well, if not then.

128
00:32:41.969 --> 00:32:48.388
You can talk, you can talk Monday so.

129
00:32:48.388 --> 00:32:58.499
If this is since we were making the iron Q talk and the slides will be available online somewhere and I'll talk about more detail and I get to quantum computing later on.

130
00:32:58.499 --> 00:33:05.699
Then that's unless anyone has anything you would like to have another question here.

131
00:33:06.929 --> 00:33:10.378
No, I don't see you. I'm just checking.

132
00:33:10.378 --> 00:33:17.368
Unless anyone would like to talk about anything that's class for today and also.

133
00:33:17.368 --> 00:33:30.088
See, you on Monday, what we'll do Monday, I think, is finish off open and perhaps show lots of little programming examples and then move on to another parallel topic.

134
00:33:30.088 --> 00:33:35.308
I've also got to yes. Sorry. Did you want me to share my.

135
00:33:35.308 --> 00:33:41.038
My quote, unquote presentation my my laptop shut down so that's why I left.

136
00:33:41.038 --> 00:33:44.098
Um, momentarily so I apologize for that.

137
00:33:44.098 --> 00:33:49.858
Well, you are on you on the schedule for today I think that I screw up the schedule.

138
00:33:49.858 --> 00:33:57.449
No, I am on the schedule. I mean, if you've got hardware problems, we can. Do you Monday?

139
00:33:57.449 --> 00:34:02.759
No, I'm booked to reboot. I just joined on my phone, so I can.

140
00:34:02.759 --> 00:34:12.748
Doc, I just wanted to apologize for. Yeah, but if you have any video at all, then maybe wait until you can show the video you got any.

141
00:34:12.748 --> 00:34:20.429
Or if you want to just talk yeah, I can just talk. Okay. No, not just for last Monday.

142
00:34:21.568 --> 00:34:28.349
So, my topic was the applications of parallel computing and machine learning.

143
00:34:28.349 --> 00:34:43.139
So, obviously, the topic of applying parallel computing to machine learning processes, and algorithms is a topic with a lot of ongoing research as.

144
00:34:43.139 --> 00:34:53.429
The applications of machine learning as well as the size of the data sets that the algorithms employ has grown.

145
00:34:53.429 --> 00:35:02.518
Um, applications were becoming increasingly complex and time consuming, and this is led researchers and developers to want to.

146
00:35:02.518 --> 00:35:05.998
You know, more quickly and efficiently perform the necessary.

147
00:35:05.998 --> 00:35:09.748
Computations and this is especially true and.

148
00:35:09.748 --> 00:35:17.128
Areas of the commercial and development space where real time processing and speed are a customer focus.

149
00:35:17.128 --> 00:35:22.469
Think of things like autonomous driving, smart assistance, things like that.

150
00:35:22.469 --> 00:35:33.509
Up to this point machine learning, because it is a newer technology has often taken a single or serial approach to computing and processing.

151
00:35:33.509 --> 00:35:43.168
And while, sometimes this is necessary, due to certain computations and steps needing to be performed before others and then a specific order.

152
00:35:43.168 --> 00:35:49.168
There are also areas where parallel computing could be easily.

153
00:35:49.168 --> 00:35:53.909
Applied and there are some instances inconveniences in.

154
00:35:53.909 --> 00:35:59.128
That lead to the sequential processing model to be incredibly time consuming.

155
00:35:59.128 --> 00:36:09.298
And out, you know, 3 of these areas are 1, you don't actually know in advance which algorithm is going to give you the best result or the best fit to your data.

156
00:36:09.298 --> 00:36:15.809
So you have to try a lot of them and if you're performing that sequentially, obviously it will take a lot of time.

157
00:36:15.809 --> 00:36:26.188
Another 1 is that the learning process of a model is often controlled through things called hyper parameters, which is something that the user has to take.

158
00:36:26.188 --> 00:36:40.679
And there's no formula to really calculate that. So these are fine tuned over the course of iterations and often require hundreds of iterations for the model to find the hyper parameters that sit the, the design of the problem.

159
00:36:40.679 --> 00:36:49.409
And then ML models by designer super flexible, which puts them at risk of overfitting the data.

160
00:36:49.409 --> 00:36:56.159
And this is why, in your model not only links the data itself, but it also ends the noise of the data.

161
00:36:56.159 --> 00:37:00.778
This can be prevented by using a message called cross validation.

162
00:37:00.778 --> 00:37:04.318
Which is actually an area.

163
00:37:04.318 --> 00:37:09.958
That lends itself very well 2 parallel computing, but cross validation is essentially.

164
00:37:09.958 --> 00:37:15.179
Just training and evaluating the models on different subsets of the data. Multiple times.

165
00:37:15.179 --> 00:37:23.699
So, because of the iterative nature, and how often they need to be repeated this test, open up a whole world of.

166
00:37:23.699 --> 00:37:26.728
Opportunity for parallel.

167
00:37:26.728 --> 00:37:27.239
So,

168
00:37:27.233 --> 00:37:27.594
while,

169
00:37:27.594 --> 00:37:28.193
unfortunately,

170
00:37:28.193 --> 00:37:42.623
I wasn't able to find a whole lot of high level information on how machine learning and parallel computing intersect when we think of as simply processing large amounts of data and performing numerous computations often.

171
00:37:42.929 --> 00:37:49.889
Many many times we can derive a few areas that might benefit from this parallelization.

172
00:37:49.889 --> 00:38:04.469
So, for 1 to relate it to what we've kind of talked about, a lot of tasks fall into the or the single program, multi data area taxonomy. So, if we look into.

173
00:38:04.469 --> 00:38:12.599
Some of the computations, we can see a few processes that would largely benefit from parallelization technique.

174
00:38:12.599 --> 00:38:20.130
1 of these is due to the high potential for performance gain. This is a huge focus and.

175
00:38:22.315 --> 00:38:27.894
Is paralyzing the common tasks that can be shared among a lot of numerous algorithms,

176
00:38:27.925 --> 00:38:32.695
such as matrix multiplication and matrix multiplication as,

177
00:38:32.784 --> 00:38:33.085
you know,

178
00:38:33.085 --> 00:38:37.074
it's well used for implementing things like linear regression algorithms,

179
00:38:37.074 --> 00:38:43.074
which are incredibly popular and we can also use parallelization for things like distance calculation.

180
00:38:43.380 --> 00:38:48.960
Where, and distance is a common metric and machine learning.

181
00:38:48.960 --> 00:39:03.869
Used in machine learning that requires iterative calculations across numerous algorithms to determine and because these calculations are entirely independent about other calculations. The same iteration.

182
00:39:03.869 --> 00:39:07.110
You know, we can clearly use parallelization pretty well.

183
00:39:07.110 --> 00:39:11.039
Like, I had mentioned before 1 of the areas that.

184
00:39:11.039 --> 00:39:20.519
Can really benefit from parallelization is key fold cross validation. That's the data validation. Technique used.

185
00:39:20.519 --> 00:39:30.449
To avoid overfitting the data it's pretty much a common evaluation technique that's regularly employed and an model.

186
00:39:30.449 --> 00:39:34.590
To validate it, which involves just an intense processing of.

187
00:39:34.590 --> 00:39:45.780
Processing dataset segments to determine and algorithms error rate and it's a model. It's a.

188
00:39:45.780 --> 00:39:54.780
While I won't get into the technical details of how it works, or, you know, what happens in it for the sake of this discussion. We can pretty much just say that it's.

189
00:39:54.780 --> 00:39:58.829
A highly iterative and repetitive series of competition.

190
00:39:58.829 --> 00:40:07.530
Results of all of these computations then being combined and averaged into a single integrated model or predicted error rate.

191
00:40:07.530 --> 00:40:12.840
So, when performs sequentially, this can be relatively time consuming, especially when.

192
00:40:12.840 --> 00:40:20.670
Each bold or segment is paired with a copy tech funnel a complex algorithm, such as, you know.

193
00:40:20.670 --> 00:40:24.570
Linear regression. Matrix multiplication.

194
00:40:24.570 --> 00:40:30.989
So, performing these computations in parallel could drastically speed up error approximation, which is a big.

195
00:40:30.989 --> 00:40:34.619
A big component of fitting data to an model.

196
00:40:34.619 --> 00:40:43.619
And this is a standard method, so if we could figure out a way and paralyze these as well, it would speed up processing.

197
00:40:43.619 --> 00:40:57.059
Quite a bit, but, like, I had mentioned, not every process can be effectively paralyzed. Some of them do have to be sequentially performed. So there is an extent how we can parallelize these.

198
00:40:57.059 --> 00:41:07.889
But for actual applications are 2 of the most likely tools for applying core parallelization.

199
00:41:07.889 --> 00:41:12.989
In the space, and we did talk about food a little bit and I'm not sure everyone.

200
00:41:12.989 --> 00:41:18.570
5 thoughts, but from a capability standpoint, there's really the big 2.

201
00:41:18.570 --> 00:41:24.750
That's pretty much all I have. Okay. Thank you. Connor.

202
00:41:24.750 --> 00:41:28.380
Any questions.

203
00:41:28.380 --> 00:41:33.780
Questions if I have a comment to see.

204
00:41:33.780 --> 00:41:42.389
I guess some parallel application that if you go to the and video website, and then if I can say, pull up a.

205
00:41:42.389 --> 00:41:47.969
Page here, I'll just share.

206
00:41:47.969 --> 00:41:54.000
Great so this is just a page I pulled up from just a 2nd.

207
00:41:55.769 --> 00:42:10.739
Yeah, you may be able to see it. Yeah, so they again news the previous talk, they're big part of their business. Now. In fact, they're adding stuff to their hard. We mentioned matrix multiplication. They are adding.

208
00:42:10.739 --> 00:42:17.280
New hardware features, they've been doing it for several years now to the to do machine learning faster.

209
00:42:17.280 --> 00:42:31.500
What are the types of hardware are lower? Precision, 16 bit floating point and we're single precision is 32 bits, including the exponent. So, 16 best he has about 7 beds plus the exponent because that's adequate for the.

210
00:42:31.500 --> 00:42:42.539
For the matrix qualifications you mentioned, and it can be done faster since this last data and data speeds limitations. That's just 1 thing about tying it in. And you see, you can.

211
00:42:42.539 --> 00:42:46.500
Browse around here and.

212
00:42:46.500 --> 00:42:52.110
Hi, before and that sort of thing. Okay.

213
00:42:53.280 --> 00:43:00.630
The other questions, and so on, if not well, have a good weekend and see you Monday.

214
00:43:01.739 --> 00:43:08.309
Actually, uh, 1, quick question. Okay. Uh, I'm just briefly looking at your blog and I saw this at homework. 3.

215
00:43:08.309 --> 00:43:11.730
Uh, we're using parallel dot.

216
00:43:11.730 --> 00:43:16.530
You just quickly go over again, uh, how to use parallel.

217
00:43:16.530 --> 00:43:27.809
Sure, yeah, I went through it went through it Monday budgets. You connect to it. Let me show you how.

218
00:43:27.809 --> 00:43:33.869
I might do it then just to give me a minute here.

219
00:43:33.869 --> 00:43:41.670
Let's see and.

220
00:43:41.670 --> 00:43:45.449
Okay, let me.

221
00:43:45.449 --> 00:43:49.650
See, if I can share a screen.

222
00:43:58.230 --> 00:44:05.579
Okay, and let me just get this around here.

223
00:44:05.579 --> 00:44:10.079
Okay, so you're seeing a window on my laptop and.

224
00:44:10.079 --> 00:44:16.559
H, and now I'm off campus size, the VPN running.

225
00:44:16.559 --> 00:44:20.340
Silence.

226
00:44:21.894 --> 00:44:29.965
Now, I have set up, I've set up a key pair with you don't have a key pair set up.

227
00:44:29.965 --> 00:44:40.344
So, you, when you did this, it would ask for your password and your password is Erin 6, 6 0T something for 9 digits and you'd want to go and.

228
00:44:41.369 --> 00:44:52.559
You'd want to go and change it and to something else. And if we look at slash the route here, I've got a link called class 2021.

229
00:44:52.559 --> 00:45:03.869
And users are your home directory, which you don't see each other's files files oh, this thing is running.

230
00:45:03.869 --> 00:45:09.119
It's riding, it's running Linux. Of course.

231
00:45:09.119 --> 00:45:22.014
As to all the supercomputers, and just to show you what the machine is age top it's got 250 to it's got a quarter terabyte of RAM of D RAM and it's dual 14 corps.

232
00:45:24.269 --> 00:45:31.679
See on, so that's 56 hyper threads there a fair number of threads and.

233
00:45:32.699 --> 00:45:37.769
What it's what it's God is.

234
00:45:37.769 --> 00:45:45.119
What 2 GP is interesting is the 8000.

235
00:45:45.119 --> 00:45:52.050
Um, which has 48 gigabytes of memory on the GPU and 4600 CUDA cores.

236
00:45:52.050 --> 00:45:59.519
And so you're here and open an invoice talks about the machine open MP.

237
00:45:59.519 --> 00:46:06.420
And these are some programs, the Pi, the various information, the Pi.

238
00:46:06.420 --> 00:46:14.460
Here, and now you cannot write into this directory, I hope, but you can copy the.

239
00:46:14.460 --> 00:46:25.559
Copy files to somewhere else and then compile them. They all use comments. So you can copy. I'll use the directory DS let's say, and copy.

240
00:46:25.559 --> 00:46:29.070
And, um.

241
00:46:29.070 --> 00:46:34.349
Yes, and copy to make file also.

242
00:46:34.349 --> 00:46:39.539
And then I can go to DS and there are my files here. I could say something like make hello.

243
00:46:39.539 --> 00:46:47.309
And it's just using by default g plus boss later on, I'll switch to different compiler and then we can run it.

244
00:46:47.309 --> 00:46:55.769
And every time we run it, it's different, because all the threads are stomping on each other's feet. And this is the 1st lesson.

245
00:46:55.769 --> 00:47:05.909
So, that's what I have in this directory of stuff that I showed on Monday and the last year is.

246
00:47:05.909 --> 00:47:09.179
Show stuff I last year, and I'm caught.

247
00:47:09.179 --> 00:47:15.059
Think is I have to fix last year. You can go in and look at old stuff and I'll show more of that next Monday.

248
00:47:16.739 --> 00:47:25.590
Is that a start? Yes, thank you. Very much. Sure. Any more questions or other questions.

249
00:47:25.590 --> 00:47:29.969
Now, if you're not from a Linux background, you're going to have to learn a little annex, but.

250
00:47:29.969 --> 00:47:35.099
I'm not apologizing because all the big computers use Harry and sub Linux. So.

251
00:47:35.099 --> 00:47:41.250
He was amazed I haven't found any that are using.

252
00:47:41.250 --> 00:47:49.440
Microsoft at 1.180 s machine on the list of using Microsoft, and then the new list, they were all Linux.

253
00:47:49.440 --> 00:48:00.840
Other questions yes, just really quick. I had just logged into the server. I was able to get in fine, but it's asking me about creating a.

254
00:48:00.840 --> 00:48:12.449
Dot file. Oh, that's just for the shell. You can say no, should the shell gives you lots of options? I'm using Z shell so you can change it to another shell if you want. So.

255
00:48:12.449 --> 00:48:16.590
Okay, that's the 1st time you use had shell? Yeah.

256
00:48:19.320 --> 00:48:29.369
I'm introducing you to various tools that don't make your life easier, but you don't have to use them and Monday's blog we'll have more information on and so on. You can look ahead to that.

257
00:48:30.869 --> 00:48:34.739
Other.

258
00:48:34.739 --> 00:48:42.480
Okay, have a good weekend. I may be out skiing or something so I'll see you Monday.

259
00:48:42.480 --> 00:48:46.469
Silence.