| 00:03 | the teeth. Mhm, mm mm hmm. All right. Oh |
|
| 00:18 | . Uh, last time started to about open A CC Great. And |
|
| 00:27 | , so at the end of last , I will uh huh. So |
|
| 00:33 | an example, it's kind of a versions on it that illustrate different |
|
| 00:40 | So, and depending upon how long takes either from the demo of more |
|
| 00:49 | less the same thing and a few things or able to the first time |
|
| 00:53 | step depending on the first thing is . All right. So it |
|
| 01:05 | yeah. Alright. So I'm kind like, yeah, outlines such most |
|
| 01:11 | from talking about this jacoby to the situation example. And then we'll significant |
|
| 01:19 | that. And again keep um reiterate Some 10 times. Uh remember you |
|
| 01:29 | things Are put together which is important two your experience when you're trying to |
|
| 01:36 | GPS as always. Uh I said a, prepare the beginning of the |
|
| 01:44 | I think in several of the early um to get performance. It was |
|
| 01:49 | at first about data path and moving capabilities and secondly I have to use |
|
| 02:00 | . Um, so here's what the , that's a cop example. Last |
|
| 02:07 | to the and that's the solving this in the square and he's very simple |
|
| 02:19 | of the second order derivatives, I'm , non tight end of the last |
|
| 02:25 | , Yes, but there's something and will hopefully. So it's a |
|
| 02:33 | There is an approximate types of discrete . A central the difference seems. |
|
| 02:43 | , yeah, the so called sensitive , Southwestern, yeah, approximately the |
|
| 02:53 | . Well, uh neighboring points. difference between every once in the accident |
|
| 03:00 | direction and uh they were given the Between the two points and the extraction |
|
| 03:13 | . Yeah, there is a particular yeah. Yeah, yeah. Uh |
|
| 03:20 | addition Well plus -1.1% point. The wind direction. So then thing |
|
| 03:28 | there's five values involved that's broken up neighbors point Oh, that's the sense |
|
| 03:40 | for five points member. So and the competitions. Are you comin in |
|
| 03:47 | revolution position convolution kernels and that's exactly this is so different for it disciplines |
|
| 03:55 | it is college and I saw so what and and this home is matter |
|
| 04:05 | keep doing the right do some design on that and audio our bodies full |
|
| 04:15 | in the report. Yeah, is this direction and then I keep doing |
|
| 04:22 | son, it's the primary banks to conditions and control that is a converted |
|
| 04:31 | . On the other hand, it's very efficient. Use more very, |
|
| 04:37 | small examples or do things like you just illustrating some computational matters. |
|
| 04:44 | if you're serious about back to this in the past the situation, just |
|
| 04:48 | sophisticated. That's and because of the , there's nothing wrong with the method |
|
| 04:53 | is mhm All right. So and this and finally do something this |
|
| 05:06 | So you can use uh there's the as well. As long as the |
|
| 05:13 | the difference between. Uh huh the who now didn't have walls son is |
|
| 05:21 | for ability card. The absolute difference and some error. Then the drama |
|
| 05:31 | openly convert and as all those losses order for themselves around the holes but |
|
| 05:39 | the maximum number of the direction vector of. Um and then so the |
|
| 05:48 | and just go over the entire Someone directions, updates. Uh |
|
| 05:57 | Yes. They're friends argue very And the blue button uh the |
|
| 06:06 | you know, do right british So in this to value separate methods |
|
| 06:19 | reduce the value of something an Uh Oh everything using throughout. Uh |
|
| 06:31 | it's all said and done. Uh design ready and they cancel the new |
|
| 06:40 | Uh for the next 2, 3 there's a singularity. This is an |
|
| 06:46 | era. It is a point and this is an announcement. So as |
|
| 06:54 | go over the next point through the . Um and all the errors for |
|
| 07:01 | form. So that's thank you. four candles, they're getting their |
|
| 07:11 | Found out about this. Yeah. yeah. Step forward. So now |
|
| 07:19 | gonna try to figure out how to open and see on this particular |
|
| 07:27 | Yeah. One more time. Uh the structure is problem starts. Can |
|
| 07:34 | remember it? And in basketball a bond. He's a collector. Students |
|
| 07:42 | out how you and not when he they are the school time where it |
|
| 07:51 | and it just so all the new uh move over to you every I'm |
|
| 08:00 | . Uh huh. Do some Oh you don't need to move |
|
| 08:08 | Of course the code doesn't make but themselves it's available today. Do not |
|
| 08:15 | you want to do with it. . Thank you. Mhm. Sorry |
|
| 08:23 | harping on this. But it's So and there are these in terms |
|
| 08:31 | providing jobs and they were trying to there was earth damn right. A |
|
| 08:41 | of friends, everyone or Yeah. terms of it convenient factory he used |
|
| 08:50 | parallel concepts like the open empty parallel . Uh It's okay. Uh |
|
| 09:00 | Want to see who was working there constructive for for financing conflicts. And |
|
| 09:11 | if you see production costs basically that's that's the parliament of the eventual area |
|
| 09:20 | all over the country. That's our . It's over there for richer or |
|
| 09:30 | error values reduction calls and says so components of the problem it is a |
|
| 09:38 | . So Oh, something that response on top of secretly. I think |
|
| 09:53 | since the crown. Oh from it it remains. I think that event |
|
| 10:07 | . The price of the because religion HCC as the same point of the |
|
| 10:15 | does not. But they're caramels so . There's a few create the gun |
|
| 10:25 | if there's something I forget to Uh So now the first exercise. |
|
| 10:36 | . Uh Since the organization provider generated also for the host processor. |
|
| 10:46 | This is a difficult only. So , virus the army did some of |
|
| 10:54 | core of the host processor. five generations deep. You counsel. |
|
| 11:01 | that's kind of a good reference space usually the reason behind the use the |
|
| 11:07 | is that the one things to run host otherwise because it is for me |
|
| 11:15 | than just from right. The call the help in general open. S |
|
| 11:22 | is trying to make it more Open mp on the host. The |
|
| 11:28 | of the best there are. stuff. Maybe it needs to be |
|
| 11:35 | game, but since you need the Colin Edwards. So that's why I |
|
| 11:40 | it's a good thing to start. think that I like to call this |
|
| 11:44 | the host. Yeah, that's it used. Yeah, I'm sure the |
|
| 11:54 | compare car and well tonight. And the media. The other thing |
|
| 12:03 | in the use of 60 years and don't make that's because it is not |
|
| 12:11 | it's not necessarily the case of their code. Oh, so like this |
|
| 12:20 | video provider. Obviously there will be decent job. Other words they want |
|
| 12:25 | do business, but not necessarily optimize . So, but so this is |
|
| 12:31 | I work while trying or reduce one the performances on the holes as the |
|
| 12:38 | station. Let's see how much better it get under uh, things to |
|
| 12:44 | . That's it. So that's why first through one thing is 24:00 and |
|
| 12:49 | know, so and that the last , but at least yeah, pointed |
|
| 12:56 | , you know, the people who that the computer will tell you what |
|
| 13:00 | of things that students of the coast it's time to send their tireless oh |
|
| 13:09 | highlight here right now on the do it to the goal. So |
|
| 13:21 | All right. Oh, I was for in the box here and I'm |
|
| 13:25 | to take the first two. Um the first new the figured out that |
|
| 13:33 | the farmers, yeah, our work so on that basis. Uh |
|
| 13:42 | They're both just hours at the gang sorry from the here and the past |
|
| 13:52 | that generates to see people and this not for the host probably. Um |
|
| 14:01 | then same thing for the second room both. That is that they don't |
|
| 14:07 | the specialization, the characterization is for interview. Uh Huh. Right. |
|
| 14:20 | huh. That's then there's kind of government for the second certainly. Mr |
|
| 14:30 | . All right practice. So this gang level family. And then I |
|
| 14:37 | uh stop it. But yeah, it's not looking foreign operations, it |
|
| 14:46 | generate just simply generates a slightly optimized the coffee, Right? But what |
|
| 14:54 | are right market, right? So don't have the live numbers are a |
|
| 15:00 | bit a map of the source But essentially it's similar. So both |
|
| 15:09 | parents back. Right. Uh A combination. All right. |
|
| 15:18 | the news and the first battle construct that promise for the second month is |
|
| 15:29 | . Uh huh. So in this it's something where you know where they |
|
| 15:35 | you both I guess. Well, fun. It's fun. So maybe |
|
| 15:41 | other looking like this except close that thoughts of trying to back trace do |
|
| 15:50 | well. So, and then questions answers decipher the compartment important. The |
|
| 16:02 | harder. Yeah. It was not many of the later slides of P |
|
| 16:06 | C C is the video? Oh , requirements. Everyone must. |
|
| 16:16 | that's right at this point. The sense that it can be |
|
| 16:28 | He was yes. Dtcc texas company and this is C as well. |
|
| 16:38 | , okay. That's it. but you have this right? Who |
|
| 16:48 | someday and I go away. The just work and B B C C |
|
| 16:52 | there. Okay. The past is think similar or uh, organization that's |
|
| 17:03 | okay. Yeah. Maybe interested do they say. They will get to |
|
| 17:10 | that to see what happens in terms what does the cold, the different |
|
| 17:17 | organization because of this this case. gives you some information, what is |
|
| 17:22 | ? And they might just, we discover things. You know, I |
|
| 17:27 | there are verses three or 2 that completely different from other kids. So |
|
| 17:33 | all due is not better as far I remember you should. Yeah, |
|
| 17:37 | it's as a kind of a clear of open and see that you get |
|
| 17:43 | see what actually. Right, so just get a clear text and uh |
|
| 17:49 | at this end and figure out what instructions or look at the instructions about |
|
| 17:53 | this instruction needs. Cool, so it's helpful even though it's a |
|
| 17:59 | bit the first line up. Oh , it is for me, the |
|
| 18:05 | it works. I asked to just something that does something I guess we |
|
| 18:11 | try to figure out what it Um what? So so then there |
|
| 18:20 | this other than about kernels and also pronounces adults, I'm sure the referrals |
|
| 18:30 | there because of how it works. , conveniences, sopranos, directive construct |
|
| 18:41 | you can still generate call for the and also see the one extent the |
|
| 18:48 | something building on the difference. It one director vs the other. |
|
| 18:54 | For the same time it's the bulls course processor and in this paper I |
|
| 19:00 | think I was pretty much that's This that was crazy. So, |
|
| 19:16 | yes, there could be getting to Donaldson Parco who contrast major. All |
|
| 19:22 | , remember again, so that the at the open ecstasy, it started |
|
| 19:29 | kernels and it was supposed to be against the freedom to the combined er |
|
| 19:34 | the barrel directed. So and the is also a very good compiler writes |
|
| 19:41 | . So the idea was to use cardinals. That's a good chance that |
|
| 19:47 | more efficient for many. Very And the examples this recorder examples as |
|
| 19:55 | was coming today and wants to go one turn out to be the |
|
| 20:02 | Mhm. So now so here's what about and running things both uh the |
|
| 20:11 | and the right is concerts and other . So uh indeed it's pretty much |
|
| 20:23 | talking. Yeah. Not escaped. they use one or the other. |
|
| 20:35 | right now. Um so the next is not trying to do what they're |
|
| 20:46 | to do. So God generating called the use I see you know that |
|
| 20:51 | out how much again that is the is and that these timings or they |
|
| 21:01 | rather than timing didn't say exactly how it took and the stones so basically |
|
| 21:07 | to one cardinals whatever it is. these are complicated. So it's like |
|
| 21:15 | door number is not a better Like the worst because it is rape |
|
| 21:21 | to time and then the sun Yeah. Institutes Korea response rate three |
|
| 21:34 | 100 and the use of the reason use. So I stayed to |
|
| 21:41 | Well that's that's scary. Yeah, right. Yeah. Promise just as |
|
| 21:47 | sp data science. So it's using you uh from the border problems debates |
|
| 22:00 | long uh Thank you sir. 2 years old problem and Oh I think |
|
| 22:10 | your customers not just that's over The responses. Yeah, Of course. |
|
| 22:22 | respect to the artist maybe one Yeah. Mhm. We consider this |
|
| 22:28 | just for our largest problem, Mhm sorted. And it's not just |
|
| 22:38 | program with many number of publications. . Okay. A question. But |
|
| 22:53 | so now and to doing something for the units. So now I'm gonna |
|
| 23:00 | a confinement for you that thank you say I just think bash the gay |
|
| 23:13 | . Look the suspect. That's Uh huh. You Uh huh. |
|
| 23:20 | this is one of the options two uh That's a column that's seven ft |
|
| 23:33 | they introduced some of the featured coverage the you see and there is a |
|
| 23:44 | that the workforce compute capabilities nice different even after the same cold not combined |
|
| 23:56 | the G P. U. So uh and uh some clips of |
|
| 24:05 | cycle and then you go and it , thank you. I really he |
|
| 24:12 | up Cindy or vector bombs and media and the vector length in fact. |
|
| 24:22 | . And that's one of things in . Waiting for the president activation that |
|
| 24:30 | talk about something this point just uh . Both of us wanted to keep |
|
| 24:38 | a useful. Yeah. Okay. production. Yeah. Also characterized certainly |
|
| 24:53 | distance. Uh huh. I started music. You start Yeah, that's |
|
| 25:03 | second set of mr problems. Right. Right. This supporting |
|
| 25:15 | Right. So there's basically gang and vandalism for parents because kind of extra |
|
| 25:23 | sec you specified for the inner loop well. For the previous example, |
|
| 25:28 | don't think we have got my system the that's why don't we did the |
|
| 25:32 | thing that Right. It was so was the Yes, I was not |
|
| 25:47 | yes. Yes, Yes. So was a different section and the clothes |
|
| 25:56 | that's kind of food. I don't it could have been much sense because |
|
| 26:05 | could have uh generated simply instructions for group in the previous, but if |
|
| 26:11 | are not the things we don't Yeah. So maybe uh, this |
|
| 26:19 | the african, the one. What performance. So All right. I |
|
| 26:27 | got the view code that generates them use gang privately. Space for the |
|
| 26:32 | level of armaments factories things. But the game that's pretty good. |
|
| 26:39 | So let's see what we get. time not time. It's ready. |
|
| 26:50 | , okay. got down to five the execution. Why? Okay. |
|
| 27:02 | call it began paralyzed things to generate coke you barrels. Thanks Inspector |
|
| 27:12 | And it's not. And if you at suspects for the the 100 the |
|
| 27:20 | is the secret uh, Yes. want to talk to you. Isn't |
|
| 27:28 | character of brains for both single and may be exposed to protect their |
|
| 27:35 | Single uh where is this the to the peak performance is probably everything I |
|
| 27:50 | . Uh So wow, don't you can look at there's a force Content |
|
| 28:02 | as a flat rate for the so and then I'll the difference of |
|
| 28:12 | Eu I would say it's probably about least five times. Oh yeah. |
|
| 28:19 | CPU So there's a potential obviously get performance and otherwise ordered. The |
|
| 28:27 | Yes, processors that were weaker than Sun all the time. Not our |
|
| 28:36 | . Nevertheless, it's kind of his Only. Well in that case is |
|
| 28:43 | not what about 20, 20 times instead of getting a lot of problems |
|
| 28:52 | and for all. Mm huh of that are not for governments for family |
|
| 29:12 | in spite of a lot of some have worked about. So it's |
|
| 29:29 | factor of five. At least higher of the deep. You nevertheless take |
|
| 29:36 | it reduced the performance by a back a point. So, so at |
|
| 29:46 | time. Yes. Yeah. Thank . That is the point. So |
|
| 29:52 | happened? Yes. So this is gun, remember? It's a big |
|
| 30:01 | . So it so happens that there be a lot of transport that's our |
|
| 30:10 | . It works on the whole all , just go and look at he |
|
| 30:20 | a little bit more detail. Think the car function is here is the |
|
| 30:29 | in order to see what they're made . Start with the whole boots on |
|
| 30:36 | uh they move back and forth Uh so the word fix almost, |
|
| 30:48 | that on mm The concept I'm I'm back baby after this. |
|
| 30:58 | at the end of this in order produce the data. Yeah. You |
|
| 31:04 | Oh not right now. We're talking you the next apparently. And that |
|
| 31:15 | in every race. Every dimension, scam probably there's a lot of |
|
| 31:24 | Uh huh. But so it's not compiler error is this kind of thing |
|
| 31:33 | Yeah. Part of my role. I didn't think too much about what |
|
| 31:39 | to trump the weather data. It order to preserve correct foodstuff. That |
|
| 31:45 | topics that up in finder. Combined with the basic stuff. Four |
|
| 31:55 | in the book. Uh, it's from here on out there. This |
|
| 32:05 | in fact the host. Certainly You're a star move back and forth |
|
| 32:23 | . Funky. And then try to to do something on one's own. |
|
| 32:28 | here's what I talked about then. this property functions that on friday |
|
| 32:35 | Um, there was gold. There on the company. Right, |
|
| 32:42 | But it doesn't stop talking. So sense of the race allocator, it's |
|
| 32:49 | but uh, initialized by copying from boat for the senses, you know |
|
| 32:58 | . Yeah. Right. So I try to then says, well, |
|
| 33:06 | don't trust the compiler. So um , uh huh. Designer period But |
|
| 33:17 | use raid programming languages. Father, you enforcement. The final shape. |
|
| 33:26 | light inside of each one of the of dimensions of the great the future |
|
| 33:32 | Princeton some sometimes yeah. Take it of the extracted by single member there |
|
| 33:40 | there anyway, so now here is the program I'm not doing. Thank |
|
| 33:49 | combine a difficult job. So now is sources say, okay, what |
|
| 33:56 | to happen so nice in and Um, and based on information. |
|
| 34:06 | . Hey, you need to observe out to the old society. |
|
| 34:14 | The paramount rules about a what peace , initial here is what we are |
|
| 34:30 | plunger about thinking. Thanks. so now we can see it didn't |
|
| 34:40 | . Not really. So yes. , uh uh huh. Probably not |
|
| 34:52 | . And there's excited. Absolutely. difference. It shouldn't be the compartment |
|
| 35:06 | . I didn't know what to Give them what they was told to |
|
| 35:13 | . Thanks. So yes. So is just the same. It's looking |
|
| 35:19 | the code with it. Program program inserted. Data transfer taxes. |
|
| 35:27 | it's the same problem. It's the amount of data motion. So that |
|
| 35:32 | performance is pretty much is the Mm hmm. Go as far as |
|
| 35:42 | device inside. All right. So problem that's it. And in |
|
| 35:52 | so people obviously knew that that's an and the issues that the lifetime of |
|
| 36:07 | uh, defined by the farrelly So if they have a live beyond |
|
| 36:15 | . So like the same thing as opening to be different. Um, |
|
| 36:22 | in order to avoid, it's just copy. You want to buy some |
|
| 36:26 | the data to uh the longer than country in construction, they don't do |
|
| 36:36 | for instance and the states on the um for several and for this case |
|
| 36:42 | other tyrol region so it wouldn't move and forth between the two weeks. |
|
| 36:48 | then there is a construct for Yeah, so that is the data |
|
| 36:56 | that tells uh it's a way of the lifetime of data forever advocated. |
|
| 37:06 | right, so um so now I think this uh data statement for |
|
| 37:17 | this is awful. Yeah, the at the right time they raised a |
|
| 37:28 | a new is for the whole situation convergence, so no. Right. |
|
| 37:43 | will not see any couple of statements the store therefore so low cut uh |
|
| 37:53 | all of this under now some of now the coffees, so the data |
|
| 38:02 | every so so so it's not already . So that the data statement. |
|
| 38:14 | president Hello? So you don't see company company asked that once they only |
|
| 38:30 | once before initialization first, then that at all. Um Yeah. And |
|
| 38:45 | what's comfortable and a absolutely do you with you, is that really Of |
|
| 38:55 | four allocation? Sign something I haven't before. Science Yeah. Base for |
|
| 39:09 | somebody had a bigger ones going everything on the line. Oh the |
|
| 39:23 | . Oh I think there's a meaningful , you can certainly use it but |
|
| 39:31 | doesn't change anything. Was thinking to in our it is working like his |
|
| 39:46 | . Oh but this uh its residents the para beetle soul because private then |
|
| 39:59 | the terrible region absolutely have the same . So the data is the thing |
|
| 40:07 | since the lifetime the dog para will private system private just for the dinner |
|
| 40:23 | copies. It's bad. But in case because the industry is split up |
|
| 40:30 | work train concept help to gangs or Sure very but the audience separate parts |
|
| 40:40 | it or they to be very Mhm and the bottom express the characteristics |
|
| 40:51 | the fight to avoid cokie I Cokie moved there over well braces itself. |
|
| 41:00 | your monster? So yes, so just pointing these things out. I |
|
| 41:29 | think certainly in a way of controlling . Mhm I found itself redefining the |
|
| 41:37 | we were speaking. If you want be able to be a it |
|
| 41:44 | you can become measures are very Yes, but in the Crimea region |
|
| 41:55 | there is the same point um at end of the terrorists. So the |
|
| 42:07 | and no way was for the work concepts now. So when the regions |
|
| 42:23 | have listed groups or groups uh huh Oh we will proceed and I have |
|
| 42:38 | friend but it doesn't change the lifetime the I guess the other figures nothing |
|
| 42:53 | and Brexit will not survive. I imagine that there should be uh in |
|
| 43:04 | Hello. Have a process right from heart. You have these typically some |
|
| 43:15 | goal. Just want that. Then have a higher region construct. It |
|
| 43:21 | hurts a bunch of kids for Part of the rules talked about and |
|
| 43:28 | whatever number of courses available and not court processing and then the private regions |
|
| 43:37 | and your back support jobs defined by probably And at the end of our |
|
| 43:52 | , whatever it was created entirely. has forgotten On the 60th action. |
|
| 44:07 | the data set contains the one what ? The computer the kind of see |
|
| 44:15 | being allocated beyond the under the And it doesn't matter how many threads |
|
| 44:24 | of them any is then trying to the data simply. Data management. |
|
| 44:30 | be another thing. Mm Yeah. . Well, you can uh get |
|
| 44:47 | the other case statements or allegation. have the three statements and there was |
|
| 44:53 | the other people duty. So you yourself. Mm hmm. But that |
|
| 45:03 | a question where they It was a . That's fair. That that will |
|
| 45:10 | see a lot of the region should the head of it. Outside the |
|
| 45:17 | . Hopefully we will survive. And exactly want the data statement does. |
|
| 45:26 | now the point was now use the , the independence it off. Mhm |
|
| 45:36 | something that I think of the excess . Hello? Yeah. So, |
|
| 45:49 | numbers but the point is being on back. I'm not kidding. Present |
|
| 45:58 | farmer uh huh depends on the Yeah. At one time decides how |
|
| 46:10 | capitalism is forgiven size of dimension. uh, but there is a environment |
|
| 46:21 | . I'm sure that resemble in Iran you can use to get some information |
|
| 46:26 | the personal launches tells you the size effective plan that each person large. |
|
| 46:38 | No, I don't think there is option you can specifically asked for. |
|
| 46:43 | , it's a good size of Uh you know, obviously uh, |
|
| 46:48 | options. Uh, I yeah. I know some of the work. |
|
| 47:10 | sorry, it's a very good So and I would also say that |
|
| 47:21 | , just nice enough every time. condition is fast as we? Re |
|
| 47:29 | this example inclinations of it. I this in one connection of the course |
|
| 47:37 | the next you have been combined. So the behavior, it's sort of |
|
| 47:46 | all it's very different from amazon. love to hear So and that means |
|
| 47:56 | of it does. There may be organization but they also introduced more features |
|
| 48:02 | . So it's like a year ago was or code examples in different ways |
|
| 48:09 | instructive compared to what to do What kind of to be mostly well |
|
| 48:14 | fairly significant difference whether you use cardinals parallel as the directive to dealing the |
|
| 48:21 | is this time around that. So , the BBC Next one. I |
|
| 48:29 | know about what? Right. Okay. So in this case now |
|
| 48:51 | mm careful. Well, uh trying avoid the excessive things and being conscientious |
|
| 49:01 | I started out saying, be aware the weakest part is this this year |
|
| 49:07 | both the attachment china between the CPU gpu. So you don't pay |
|
| 49:14 | The point is you're likely to lose . That's supposed to gain performance. |
|
| 49:21 | Now um and so that was all the use of camels. So |
|
| 49:28 | open sec find the guy is a of advertising and try to promote the |
|
| 49:33 | directed that. So there's a thing the colonel's directive um and generating you |
|
| 49:44 | I don't think it's always different. . Thank you ourselves. But the |
|
| 49:56 | is I guess uh this is uh provider pick this up. What do |
|
| 50:03 | think? There is no explicit I think. I think so in |
|
| 50:10 | case uh the final I just needed help of a later statement in order |
|
| 50:14 | keep things become partners to get off area of a new race along using |
|
| 50:19 | next construct that is set for the are reporting. Mhm. If big |
|
| 50:27 | to do the job. Well, that countries the initiative and the previous |
|
| 50:33 | , the effective length was chosen as Here. For whatever reason the combined |
|
| 50:39 | to choose four and 2 but it's . Yeah. And the second |
|
| 50:49 | Well that's one point up there do . So the village yeah. Yes |
|
| 51:04 | better responsible analysis of sorry. Uh . Okay. From the previous taking |
|
| 51:19 | birthdays. Well let's see what the . Yeah. Yes so yeah it's |
|
| 51:29 | different graph has a little bit more . But are we talking about this |
|
| 51:38 | uh do you think? Uh All right you go. Yes. |
|
| 51:47 | . Mhm. Oh what what I'm that was unmanaged the day. All |
|
| 52:05 | so on this right because this is worst picks up. So that's what |
|
| 52:11 | said right now because with the data uh so the colonel's you see speed |
|
| 52:23 | so telling the time. Yeah. . Oh sorry about that. I |
|
| 52:31 | fix it and well uh showing a like what you have to read some |
|
| 52:42 | coming up and no one ever. huh And then being messed up in |
|
| 52:49 | basement but it shows a little bit So this is referring to for moving |
|
| 52:59 | data cost of deserve things. Oh . Uh huh. Yourself. So |
|
| 53:10 | that case the United Time Remember that travel food performance. So here goes |
|
| 53:22 | this one for the one that's so it took a move on please have |
|
| 53:29 | more relatively time. There was a for new one word party bill. |
|
| 53:40 | huh. Yeah but obviously you can see that generated it being a market |
|
| 53:50 | much. All right well let's construct quality of home and we got some |
|
| 53:59 | up or little bit more for ex . Everybody came home. I don't |
|
| 54:06 | what they say. Yeah stop there . Oh what? Okay first it's |
|
| 54:20 | now it's so all right it's a bit of deviation and I'll talk about |
|
| 54:31 | some more quickly because I want to a good example. Uh And it's |
|
| 54:37 | bit more to England. Let's So so so far it's all exercises |
|
| 54:48 | 80 Data Science Career buster. So or 3 sides from basically a tutorial |
|
| 54:59 | all this and the purpose what? it may end up may be used |
|
| 55:06 | your assignment but my ability classic mustard . Um And you said The 100 |
|
| 55:20 | . You're afraid to do that when get the assignment. The other one |
|
| 55:26 | two months to make it on. . Uh Yeah. Yeah it's uh |
|
| 55:37 | use on china that's very busy. our company adopts its order so yeah |
|
| 55:46 | see the one time student? 7 or is it fun? Thank |
|
| 56:00 | We were put thanks a lot you you know So there is some pretty |
|
| 56:07 | in the next one. What is this year turning our hearts? So |
|
| 56:14 | does the time go? Oh So they are after. Oh but |
|
| 56:24 | absolutely use that. Uh huh. of the strategy. Well this is |
|
| 56:30 | that's that's sort of the whatever all that race You know seven things up |
|
| 56:38 | the platform cold all the time. . Um Well here it is almost |
|
| 56:47 | over with you the girl. So interaction on in four steps. This |
|
| 56:58 | uh Uh huh. Both is it the time parents. You can also |
|
| 57:07 | that the computer for very long and it's all the animals are over and |
|
| 57:17 | . The next couple of slides I'm it's just we're not going to talk |
|
| 57:23 | it. But you can take a and see and try to so what |
|
| 57:31 | . So in this case uh you see you know what I want for |
|
| 57:39 | in terms of all the bars and the time went. So you can |
|
| 57:46 | the compiler information from with the uh petition bars in terms of what things |
|
| 57:58 | . So that was a little bit variation or back he's doing there's a |
|
| 58:04 | for example Korea or any right? this cigar box. So now the |
|
| 58:21 | that it's important to avoiding all I'm willing that society. But also |
|
| 58:29 | the 51 of the composers I think is about one up and so obviously |
|
| 58:36 | something that early on in the media they were around. It's difficult for |
|
| 58:43 | programmers. Uh Thanks club. Both what to do. Don't forget uh |
|
| 58:51 | forward. But there are people in to our you're right to call don't |
|
| 58:58 | it off with it. The answer that is that and I learned |
|
| 59:09 | No I'm not. And the smell uniforms. Uh so uh with they |
|
| 59:25 | version of open. Mhm. They the there was no unified member and |
|
| 59:32 | 80 in stock. They have changed produced a call management system, uniform |
|
| 59:41 | . The one We talked about, also a concept where to get the |
|
| 59:50 | an illusion of one at this. best for so right at home you |
|
| 59:58 | necessarily need the attention to the fact memories critically separated with some weak links |
|
| 60:05 | between. Uh if you're lucky the system, it's clever enough to know |
|
| 60:13 | to do in order not to excessively to have back and forth. So |
|
| 60:17 | leaving best to the runtime system to care of managing data as supposed to |
|
| 60:27 | compiler attractive insert the proper data movement or you explicitly trying to manage |
|
| 60:37 | So here is kind of the picture . And the other thing is uh |
|
| 60:46 | doesn't change everything uh you want at space uh huh. It is. |
|
| 60:57 | you cannot deal with and try to things for Okay, so here is |
|
| 61:02 | on what it says, that's just gives you one of your space and |
|
| 61:06 | simple. The point is trying to the programming without losing efficiency by having |
|
| 61:15 | functionality in the runtime system too, how data is being used and move |
|
| 61:22 | between posts and for you as necessary not more than necessary. So And |
|
| 61:33 | order to book this behavior there is managed to get someone to tend to |
|
| 61:41 | . Yeah. So here is kind what it looks so and the other |
|
| 61:49 | advantage. All right. So Uh . So the provider that is the |
|
| 61:59 | system and get information from the target kind of figure these things out. |
|
| 62:05 | uh if you know what they are validated in a way a lot of |
|
| 62:12 | . No, it's about to take of this. So now if you |
|
| 62:20 | this feature uh Yeah. So All . Don't have this. Yeah. |
|
| 62:36 | the problem they were working on by the data on the contrary. Active |
|
| 62:46 | on that. And that's something we'll back before. Uh huh. That's |
|
| 62:53 | right. Get more often. Yes. So that's not this |
|
| 63:00 | The underlying drug that's just. Uh . That which is based on buddha |
|
| 63:08 | measure of drivers and direct. So began to it also has a quarter |
|
| 63:13 | it's just not to figure out what be arrested at and when is it |
|
| 63:19 | will get but accessible successful. And in fact by just using the manage |
|
| 63:30 | there's been sitting don't worry about it think. But nevertheless, I wanted |
|
| 63:34 | to understand don't want to race. all in your producer the second |
|
| 63:42 | Yes. For instance, you may practice start the manager and see if |
|
| 63:47 | get the performance is probably done if doesn't work to try to figure out |
|
| 63:54 | can't do that. But again, So that we want to and then |
|
| 64:04 | was also the kernel version rights, can do the same thing uh Donald |
|
| 64:10 | much less. It has to have cardinal corrected on that. It's going |
|
| 64:16 | look at the other than that people had sorted out what to do with |
|
| 64:21 | and then the managed context. Um cold is further similar as you look |
|
| 64:27 | the information from the compiler and now can see what in this case at |
|
| 64:33 | we're some other job arses designers, there are consistent uh you versions uh |
|
| 64:50 | the first uh like this big hundreds of people Oh, was carl |
|
| 65:01 | a little bit better than the I want to be presentable. |
|
| 65:14 | So um and they said yes, there some of the inverted ways. |
|
| 65:23 | great, you know that right? , yeah, they are. |
|
| 65:35 | there is also so mhm sort of single right on this. Well, |
|
| 65:46 | there uh you know, since He's 24 yeah or so, but |
|
| 66:04 | this case so happy. The advantage performance. Yeah. Oh yeah. |
|
| 66:19 | from the question for us, they the peak performance of the office. |
|
| 66:24 | , and I think you get this the same as I call it |
|
| 66:28 | The same practice speed. We don't to decide the rest of the |
|
| 66:35 | but the same ratio. seriously the are, you know, on the |
|
| 66:44 | is about the sufficient respect for these seat of the holes. So |
|
| 67:01 | Uh huh is comparing these cause I'm going to so through by assembled it |
|
| 67:08 | you can sit down and try to what happens added the cardinals versus parallel |
|
| 67:14 | but there are many, many issues that they are still correct. Um |
|
| 67:24 | this was against, the first thing , I don't know. Oh |
|
| 67:33 | what's different between these phones around it now anyway, so then I have |
|
| 67:40 | best of text slides here. Uh there's not much difference. Mm So |
|
| 67:54 | may be more relevant to what it a year ago. I forgot to |
|
| 67:58 | . Yeah, but there's something here I will make sure they are, |
|
| 68:05 | . I'm saying it I don't Excellent bar graphs and it's the first |
|
| 68:16 | last questions. Well come it's got dr demos it was something that thanks |
|
| 68:29 | thing program when we have one questions a lot of it has served. |
|
| 68:39 | about the data management. Yeah, a different whether it's or minimize state |
|
| 68:49 | that handmade emotional. Is it is always this product? As for |
|
| 69:04 | Uh huh, wow. Yeah. . It's very politician. That's |
|
| 69:12 | I wouldn't say that for this Yeah and the manage yes you and |
|
| 69:22 | a few percent is better I think some of the other cases it was |
|
| 69:26 | your percentage know where they're parallel so seems to be at this point and |
|
| 69:33 | just buy this one example. I worry too much about the performance. |
|
| 69:40 | fact that it's a cry overseas girl as was shown in the in the |
|
| 69:47 | it's our fault that there was you directed to the code with the colonel's |
|
| 69:57 | the current code there was more or . Uh other rooms were supposed to |
|
| 70:05 | comment was very difficult. Mhm Well compared and complicated last year at that |
|
| 70:18 | the above the loop holster football much better. That's true. That's one |
|
| 70:27 | . The other thing I would say only one example it would be different |
|
| 70:34 | are depending on output that combine risks at the house. That number. |
|
| 70:42 | yeah, I'm sure um this one you can say comments as competitive place |
|
| 70:51 | as well but the application is on we does it to the house now |
|
| 70:58 | a firearm. Yeah that's what I'm to have, how what the the |
|
| 71:07 | understands data. Any provocation. Uh . Yes. I mean the serpent |
|
| 71:17 | the combined data is not whether they're . It is failing to revenge of |
|
| 71:24 | that the company wants to be competitive present also just co construct different |
|
| 71:35 | Uh They do cut from some range application that they used for benchmarking and |
|
| 71:42 | try to figure out but uh thanks optimize so it covers this larger |
|
| 71:50 | one organisation may be good for some not so great for others actually potentially |
|
| 71:58 | performance for some institute trade off that can find the right time to figure |
|
| 72:04 | of the balance. The training also optimistic some stuff for the market targeting |
|
| 72:14 | they're already customers when we have So that's nice. Yeah. |
|
| 72:26 | Part of the sauce. Right. and sometimes it's just a simple or |
|
| 72:34 | like that. But the race sizes matter. Development performance. Uh we're |
|
| 72:42 | only performance as such, but what trying to optimize so. Yeah. |
|
| 72:48 | . Some applications. Thank you. in transport. Uh huh. That's |
|
| 72:58 | very violent corner or library type code some disciplines they used tons of calls |
|
| 73:08 | very small instances started doing that. doing statistics six or something like that |
|
| 73:17 | this point. State and management and . It's totally different compared to if |
|
| 73:23 | do the unfortunate so would and harder gender appropriate difference will it's determined reputations |
|
| 73:32 | be very competitive where you know, of instances of operations and relatively small |
|
| 73:40 | sets four. They do something different if you want to optimize something for |
|
| 73:47 | on very large states. So I'm giving examples of trade offs to make |
|
| 73:54 | we're that's the same thing that file and things like this. I don't |
|
| 74:10 | anyone. Yeah, bioinformatics people that to hands up in the Brazilian small |
|
| 74:21 | and uh also the health systems have been optimized for that. So they're |
|
| 74:26 | sluggish. I think we're very Uh there's a we're after 50 were |
|
| 74:34 | larger files. Oh, um that's . About time. So next time |
|
| 74:46 | start with them on top of Talk a little bit more about positional |
|
| 74:53 | of all I see stuff to see . Mhm. Mhm. Comes to |
|
| 75:08 | . They mentioned NPR is for There has they're going up to |
|
| 75:16 | Families. Course evidence. Yeah. they have to be laws and I |
|
| 75:21 | they have a bunch of others back the city that they are so for |
|
| 75:30 | you know, this very commonly used . It's not always easy to, |
|
| 75:40 | that's why many vendors, they have were trying to hide the complexity of |
|
| 75:47 | good performance out of their hard work the and users so they can call |
|
| 75:50 | library instead. So the media is different. So they have yeah, |
|
| 75:58 | 50s in it. And there are things for are seeing warning in the |
|
| 76:04 | . So because a video or Right. Yes. And the |
|
| 76:13 | V. So I guess I am sole school libraries may not be as |
|
| 76:23 | because they haven't had as much money their best and software something recently recently |
|
| 76:29 | do very well. And now they found more to some for himself, |
|
| 76:34 | of the numbers. Just keep get to buy their artist but certainly it's |
|
| 76:45 | to understand the underlying things begin and things. And that's part of our |
|
| 76:53 | exercise. Should you know, I'm gonna I'm gonna that if um |
|
| 77:00 | were asked to compare yeah with K. L. In terms of |
|
| 77:07 | |
|
